Profilaxis del tromboembolia durante el tratamiento con asparaginasa en adultos con leucemia linfoblástica aguda
Resumen
Antecedentes
El riesgo de tromboembolia venosa aumenta en los adultos y se potencia con la quimioterapia con asparaginasa, y la tromboembolia venosa introduce un riesgo secundario de retraso del tratamiento y de interrupción prematura de los principales agentes antileucémicos, lo que puede comprometer la supervivencia. Sin embargo, no se conocen a ciencia cierta los efectos beneficiosos ni perjudiciales de la tromboprofilaxis primaria en adultos con leucemia linfoblástica aguda (LLA) tratados con asparaginasa.
Objetivos
Los objetivos primarios fueron evaluar los efectos beneficiosos y perjudiciales de la tromboprofilaxis primaria en el primer episodio de tromboembolia venosa sintomática en adultos con LLA que recibían tratamiento con asparaginasa, en comparación con placebo o ninguna tromboprofilaxis.
Los objetivos secundarios consistían en comparar los efectos beneficiosos y perjudiciales de los diferentes grupos de tromboprofilaxis sistémica primaria mediante la estratificación de los resultados principales por tipo de fármaco (heparinas, antagonistas de la vitamina K, pentasacáridos sintéticos, inhibidores directos de la trombina por vía parenteral, anticoagulantes orales de acción directa y hemoderivados para la sustitución de la antitrombina).
Métodos de búsqueda
El 2 de junio de 2020 se realizó una búsqueda exhaustiva de bibliografía sin restricciones de idioma, que incluyó (1) búsquedas electrónicas en Pubmed/MEDLINE; Embase/Ovid; Scopus/Elsevier; Web of Science Core Collection/Clarivate Analytics; y el Registro Central Cochrane de Ensayos Controlados (Cochrane Central Register of Controlled Trials, CENTRAL) y (2) búsquedas manuales en (i) las listas de referencias de los estudios identificados y las revisiones relacionadas; (ii) los registros de ensayos clínicos (ClinicalTrials.gov; el registro International Standard Randomized Controlled Trial Number (ISRCTN); la Plataforma de registros internacionales de ensayos clínicos (ICTRP) de la Organización Mundial de la Salud; y los fabricantes farmacéuticos de asparaginasa, incluidos Servier, Takeda, Jazz Pharmaceuticals, Ohara Pharmaceuticals y Kyowa Pharmaceuticals), y iii) los resúmenes de congresos (de las reuniones anuales de la American Society of Hematology (ASH); la European Haematology Association (EHA); la American Society of Clinical Oncology (ASCO); y la International Society on Thrombosis and Haemostasis (ISTH). Todas las búsquedas se realizaron a partir de 1970 (el momento de la introducción de la asparaginasa en el tratamiento de la LLA). Se estableció contacto con los autores de los estudios pertinentes para identificar cualquier material no publicado, datos faltantes o información relacionada con los estudios en curso.
Criterios de selección
Ensayos controlados aleatorizados (ECA); incluidos los diseños de ensayos clínicos controlados, cuasialeatorizados, cruzados y aleatorizados por grupos) que compararon cualquier anticoagulante preventivo parenteral/oral o intervención mecánica con placebo o ninguna tromboprofilaxis, o que compararon dos intervenciones anticoagulantes preventivas diferentes en adultos de al menos 18 años de edad con LLA, tratados con ciclos de quimioterapia con asparaginasa. Para la descripción de los efectos perjudiciales se eligieron para inclusión los estudios observacionales no aleatorizados con un grupo control.
Obtención y análisis de los datos
Con el uso de un formulario estandarizado de obtención de datos, dos autores de la revisión realizaron de forma independiente la revisión y la selección de los estudios, extrajeron los datos, evaluaron el riesgo de sesgo para cada desenlace mediante herramientas estandarizadas (herramienta RoB 2.0 para los ECA y herramienta ROBINS‐I para los estudios no aleatorizados) y la certeza de la evidencia de cada desenlace mediante el enfoque GRADE. Los desenlaces principales incluyeron primer episodio de tromboembolia venosa sintomática, mortalidad por todas las causas y hemorragia grave. Los desenlaces secundarios incluyeron tromboembolia venosa asintomática, mortalidad relacionada con la tromboembolia venosa, episodios adversos (es decir, hemorragia no grave clínicamente relevante y trombocitopenia inducida por la heparina en los ensayos que utilizaron heparinas) y calidad de vida. Los análisis se realizaron según las guías del Manual Cochrane de revisiones sistemáticas de intervenciones (Cochrane Handbook for Systematic Reviews of Interventions). En el caso de los estudios no aleatorizados, todos los estudios (incluidos los estudios considerados con importante riesgo de sesgo en al menos uno de los dominios ROBINS‐I) se evaluaron en un análisis de sensibilidad que exploró los factores de confusión.
Resultados principales
Se identificaron 23 estudios no aleatorizados que cumplieron con los criterios de inclusión de esta revisión, de los cuales diez estudios no proporcionaron datos de desenlaces en adultos con LLA. En la evaluación del "Riesgo de sesgo" se incluyeron los 13 estudios restantes y se identificó una definición no válida del grupo control en dos estudios, además se consideró que los desenlaces de nueve estudios tenían un importante riesgo de sesgo en al menos uno de los dominios ROBINS‐I y que los desenlaces de dos estudios tenían un alto riesgo de sesgo.
No se evaluaron los efectos beneficiosos de la tromboprofilaxis, ya que no se incluyeron ECA. En el análisis descriptivo principal de los efectos perjudiciales se incluyeron dos estudios retrospectivos no aleatorizados con desenlaces que se consideraron con alto riesgo de sesgo. Un estudio evaluó concentrados de antitrombina en comparación con ningún concentrado de antitrombina. No se sabe con certeza si los concentrados de antitrombina tienen un efecto sobre la mortalidad por todas las causas (riesgo relativo [RR] 0,55; intervalo de confianza [IC] del 95%: 0,26 a 1,19 [análisis por intención de tratar]; un estudio, 40 participantes; evidencia de certeza muy baja). No se sabe con certeza si los concentrados de antitrombina tienen un efecto sobre la mortalidad relacionada con la tromboembolia venosa (RR 0,10; IC del 95%: 0,01 a 1,94 [análisis por intención de tratar]; un estudio, 40 participantes; evidencia de certeza muy baja). No se sabe si los concentrados de antitrombina tienen un efecto sobre la hemorragia grave, la hemorragia no grave clínicamente relevante o la calidad de vida de los adultos con LLA tratados con quimioterapia con asparaginasa, ya que los datos no fueron suficientes. El estudio restante (224 participantes) evaluó la profilaxis con heparina de bajo peso molecular versus ninguna profilaxis. Sin embargo, en este estudio se notificó que no se disponía de datos suficientes sobre los efectos perjudiciales, incluida la mortalidad por todas las causas, la hemorragia grave, la mortalidad relacionada con la tromboembolia venosa, la hemorragia no grave clínicamente relevante, la trombocitopenia inducida por la heparina y la calidad de vida.
En el análisis de sensibilidad de los efectos perjudiciales, que explora el efecto de los factores de confusión, también se incluyeron nueve estudios no aleatorizados con desenlaces considerados con importante riesgo de sesgo debido principalmente a los factores de confusión no controlados. Tres estudios (179 participantes) evaluaron el efecto de los concentrados de antitrombina y seis estudios (1224 participantes) evaluaron el efecto de la profilaxis con diferentes tipos de heparinas. Al analizar la mortalidad por todas las causas, la mortalidad relacionada con la tromboembolia venosa y la hemorragia grave (estudios de heparina solamente), con la inclusión de todos los estudios con desenlaces extraíbles para cada comparación (antitrombina y heparina de bajo peso molecular), se observó que el tamaño de los estudios fue pequeño; hubo pocos episodios; IC amplios que cruzan la línea de ningún efecto; y una heterogeneidad significativa a partir de la inspección visual de los diagramas de bosque (forest plots). Aunque la heterogeneidad observada podría surgir por la inclusión de un escaso número de estudios con diferencias en cuanto a los participantes, las intervenciones y las evaluaciones de los desenlaces, es inevitable la probabilidad de que el sesgo debido a la falta de control de los factores de confusión sea la causa de la heterogeneidad. No fue posible realizar análisis de subgrupos debido a la escasez de datos.
Conclusiones de los autores
No se sabe, a partir de la evidencia actualmente disponible, si la tromboprofilaxis utilizada en adultos con LLA tratados con asparaginasa se asocia con efectos beneficiosos clínicamente significativos y efectos perjudiciales aceptables. La investigación existente sobre esta pregunta es únicamente de diseño no aleatorizado, con factores de confusión de importantes a críticos, y con escasa potencia estadística con una imprecisión significativa. Cualquier estimación de los efectos sobre la base de la evidencia insuficiente existente es muy incierta y es probable que cambie con las investigaciones futuras.
PICO
Resumen en términos sencillos
Prevención de los coágulos sanguíneos en adultos diagnosticados de leucemia linfoblástica aguda y tratados con quimioterapia con asparaginasa
¿Cuál es el objetivo de esta revisión?
El objetivo de esta revisión Cochrane fue determinar si los medicamentos actualmente disponibles para prevenir coágulos sanguíneos presentan un equilibrio considerable entre los efectos beneficiosos y perjudiciales en adultos con un subtipo de cáncer sanguíneo llamado leucemia linfoblástica aguda (LLA) tratados con quimioterapia que contiene asparaginasa.
Mensajes clave
No está claro si los productos derivados de la sangre (hemoderivados) como los concentrados de antitrombina para prevenir coágulos sanguíneos en adultos con LLA tratados con quimioterapia con asparaginasa se asocian con efectos perjudiciales inaceptables como la muerte. La confianza en los estudios identificados es limitada debido a su diseño no aleatorizado con pocos participantes y sin control de los factores subyacentes que podrían influir en el desenlace.
No fue posible evaluar los efectos perjudiciales relacionados con otro tipo de prevención de coágulos sanguíneos, como la heparina de bajo peso molecular, porque no se notificaron muertes ni hemorragias. No se consideraron efectos beneficiosos como la reducción del riesgo de coágulos, debido a la calidad baja de la evidencia identificada (no hay estudios aleatorizados). Se necesitan estudios aleatorizados de calidad alta, ya que su realización reduce el riesgo de que los factores subyacentes influyan en los desenlaces.
¿Qué se estudió en esta revisión?
La LLA es un subtipo de cáncer de la sangre que se produce por la transformación maligna de glóbulos blancos inmaduros de origen linfoide. En los últimos 20 años, la supervivencia históricamente deficiente de los adultos con esta enfermedad ha mejorado de forma notable, principalmente debido a la introducción de un tratamiento antileucémico más intensivo inspirado en el utilizado en niños con LLA. Una de las características de este tratamiento es el uso intensivo del agente antileucémico llamado asparaginasa. Sin embargo, esto tiene un precio. La edad avanzada y el tratamiento con asparaginasa aumentan el riesgo de que se formen coágulos sanguíneos, que pueden bloquear los vasos sanguíneos (predominantemente en los llamados venas) y cambiar el flujo sanguíneo normal o hacer que fracciones del coágulo pasen a otras venas situadas en órganos importantes, lo que da lugar a graves problemas de salud. Además, los médicos son propensos a interrumpir el tratamiento con asparaginasa si se producen coágulos, lo que podría comprometer la supervivencia de estas personas.
No se sabe si los tratamientos existentes de prevención de coágulos sanguíneos protegen contra los coágulos y no causan hemorragias ni muerte en los adultos con esta enfermedad. Los medicamentos incluyen heparinas, antagonistas de la vitamina K, pentasacáridos sintéticos, inhibidores directos de la trombina, anticoagulantes orales de acción directa o hemoderivados para la sustitución de la antitrombina (disminución de la producción en el cuerpo de esta proteína anticoagulante durante el uso de la asparaginasa) y la prevención mecánica como las medias elásticas graduadas.
Por lo tanto, se analizaron todos los estudios de investigación disponibles actualmente en adultos con LLA tratados con quimioterapia con asparaginasa que recibieron tratamiento preventivo para los coágulos en comparación con placebo o ninguna prevención. Se examinaron los desenlaces siguientes: primer episodio de coágulo venoso sintomático, muerte por cualquier causa, hemorragia grave, muerte relacionada con coágulos sanguíneos, coágulo venoso asintomático, hemorragia no grave clínicamente relevante, disminución del recuento de plaquetas sanguíneas inducida por la heparina y calidad de vida.
¿Cuáles son los resultados de esta revisión?
Se encontraron 23 estudios, de los cuales dos se incluyeron en los análisis principales para ayudar a responder la pregunta y nueve en un análisis adicional para ayudar a describir las limitaciones de la evidencia. Los 12 estudios restantes no se pudieron incluir debido a que faltaban desenlaces o al importante riesgo de sesgo, ya que las personas del grupo control no eran controles "reales".
Los dos estudios del análisis principal se realizaron en el Reino Unido/Canadá y compararon el tratamiento con concentrados de antitrombina con ningún tratamiento y la heparina de bajo peso molecular con ninguna heparina de bajo peso molecular, respectivamente, en personas con LLA. No se sabe si los concentrados de antitrombina en adultos con LLA tratados con quimioterapia con asparaginasa mejoran/reducen la muerte por todas las causas/coágulos sanguíneos, porque la confianza en la evidencia fue muy baja (un estudio, 40 adultos). No fue posible evaluar los efectos perjudiciales relacionados con el uso de la heparina de bajo peso molecular porque no se notificaron muertes por todas las causas/coágulos sanguíneos, hemorragias graves/no graves clínicamente relevantes ni disminución de las plaquetas inducida por la heparina. Ninguno de los estudios analizó la calidad de vida.
No se encontraron estudios que evaluaran otros tratamientos preventivos. No se examinaron efectos beneficiosos como la reducción del riesgo de coágulos sanguíneos, debido a la escasa certeza de la evidencia.
¿Qué grado de actualización tiene esta revisión?
Los autores de la revisión buscaron estudios publicados hasta el 2 de junio de 2020.
Authors' conclusions
Summary of findings
| Antithrombin concentrates compared to no antithrombin concentrates for thromboprophylaxis in adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy | ||||||
| Patient or population: thromboprophylaxis in adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy Setting: United Kingdom (inpatient setting) Intervention: antithrombin concentrates Comparison: no antithrombin concentrates | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with no antithrombin concentrates | Risk with antithrombin concentrates | |||||
| All‐cause mortality ‐ Intention‐to‐treat analysis | 538 per 1.000 | 296 per 1.000 | RR 0.55 | 40 | ⊕⊝⊝⊝ | The evidence is very uncertain about the effect of antithrombin concentrates on all‐cause mortality. f,g |
| First‐time symptomatic venous thromboembolism | Benefits not assessed for non‐randomised studies in this review (outcome reported in study) | 40 | ‐ | |||
| Major bleeding ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Venous thromboembolism‐related mortality ‐ Intention‐to‐treat analysis | 154 per 1.000 | 15 per 1.000 | RR 0.10 | 40 | ⊕⊝⊝⊝ | The evidence is very uncertain about the effect of antithrombin concentrates on mortality secondary to thromboembolism. f,g |
| Quality of life ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Asymptomatic venous thromboembolism | Benefits not assessed for non‐randomised studies in this review (outcome not reported in study) | (0 studies) | ‐ | |||
| Clinically relevant non‐major bleeding ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_415691309867741365. | ||||||
| a. Risk of bias: downgraded one level for serious risk of bias due to confounding. | ||||||
| Enoxaparin compared to no enoxaparin for thromboprophylaxis in adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy | ||||||
| Patient or population: adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy Setting: Canada (inpatient setting) Intervention: enoxaparin Comparison: no enoxaparin | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with no enoxaparin | Risk with enoxaparin | |||||
| All‐cause mortality ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| First‐time symptomatic venous thromboembolism | Benefits not assessed for non‐randomised studies in this review (outcome reported in study) | 224 | ‐ | |||
| Major bleeding | No major bleeding events in the enoxaparin group. Number of events not reported in the control group. 1 | 224 | ‐ | |||
| Thrombosis‐related mortality ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Quality of life ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Asymptomatic venous thromboembolism ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Adverse events: clinically relevant non‐major bleeding / heparin‐induced thrombocytopenia ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_415827266990160046. | ||||||
| 1. Schulman S, Kearon C.. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non‐surgical patients.. Journal of Thrombosis and Haemostasis.; 2005. | ||||||
Background
Acute lymphoblastic leukaemia (ALL) is a serious haematological malignancy with historically poor survival in the adult population (Stock 2013; Toft 2012). However, the five‐year event‐free survival rates (events: death, leukaemic relapse, and second malignancy) are improving and approaching 75% with the introduction of paediatric‐inspired treatment protocols for younger adults decades ago, especially in the Philadelphia chromosome‐negative subgroup (DeAngelo 2015; Ribera 2008; Stock 2019; Toft 2018). Now cure rates are to a higher degree challenged by emerging treatment‐related toxicities, including venous thromboembolism, with an incidence that increases with age (Grace 2011; Stock 2011; Toft 2016; Truelove 2013).
Description of the condition
Venous thromboembolism has been reported in as many as 42% of adults with ALL during chemotherapy (Grace 2017), compared with 1.4% to 2.2% at the time of ALL diagnosis (De Stefano 2005; Ziegler 2005). Thus, venous thromboembolism rarely occurs at initial presentation of ALL but rather during the initial phases of ALL treatment (Caruso 2007), which most likely indicates a therapeutic relation, including a state of tumour lysis‐induced procoagulation.
Compared with conventional adult ALL protocols, the backbone of the paediatric‐inspired ALL regimens is based on higher cumulative doses of non‐myelosuppressive agents (glucocorticoids, vincristine, and asparaginase), earlier and more frequent central nervous system prophylaxis, and prolonged maintenance therapy (Boissel 2003; De Bont 2004; Ramanujachar 2007; Stock 2008; Toft 2016). The successful outcomes in children with ALL have been ascribed to the intensive use of asparaginase, being a key component of the therapy (Raetz 2010). Asparaginase is an enzyme that takes advantage of the fact that the cancer cells (malignant lymphoblasts) have limited ability to synthesise the amino acid asparagine and depend on external sources—in contrast to normal cells that synthesise their own asparagine. Thus, it deprives the malignant lymphoblasts of asparagine by catalysing its hydrolysis to aspartic acid and ammonia, inhibiting the protein synthesis and ultimately leading to cell death (Van den Berg 2011). However, as a result of asparagine depletion—hence global plasma protein synthesis impairment, its use has been associated with complex disruption of the balance between haemostatic and fibrinolytic pathways, predominantly inducing a procoagulant state. This is thought to be mediated by decreased synthesis of the anticoagulant proteins (protein S, protein C, and antithrombin) (Mitchell 1994; Truelove 2013). Additionally, venous thromboembolism has been observed more frequently to coincide with concomitant administration of asparaginase and corticosteroids in children with ALL (Athale 2003; Nowak‐Gottl 2001). The currently Food and Drug Administration approved formulations of asparaginase include Escherichiacoli‐derived L‐asparaginase (native or polyethylene glycol‐conjugated (pegylated; with extended half‐life) and Erwinia chrysanthemi‐derived asparaginase. Reported venous thromboembolic rates seem higher in those receiving pegylated asparaginase (Rank 2018; Stock 2019) compared with native Escherichiacoli‐derived asparaginase (Gugliotta 1992); this has not been investigated in a randomised setting. The impact of these agents, other concomitantly given drugs, and physiologic factors in the pathogenesis of venous thromboembolism remain unknown. Nevertheless, the development of venous thromboembolism during treatment may ultimately result in premature discontinuation of asparaginase, which has been associated with inferior survival (Aldoss 2013; Silverman 2001).
The majority of thromboembolic events are reported to be of venous origin (Athale 2003; Caruso 2007; De Stefano 2005). Apart from age and the treatment regimen, several other risk factors of venous thromboembolism have been reported such as the malignancy itself, prior venous thromboembolism, infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, smoking, and obesity (Dobromirski 2012; Kahn 2012; Streiff 2011; Truelove 2013). Whether Philadelphia chromosome status has an impact on the risk of venous thromboembolism is unknown; however, a high incidence of venous thromboembolism has been observed in people with Philadelphia chromosome‐positive ALL, even in the absence of asparaginase therapy (Chen 2019).
Despite the awareness of the increased risk of venous thromboembolism in adults with ALL, few coherent recommendations on the use of thromboprophylaxis in adults with ALL are published (Stock 2011; Zwicker 2020), as no high‐quality evidence exists. The subgroup of adults at particularly high risk of venous thromboembolism, who indeed could benefit from pre‐emptive anticoagulant treatment, has not been clearly defined. Yet, several different risk‐score models of venous thromboembolism have been proposed (Mitchell 2010; Rank 2018; Roininen 2017).
Description of the intervention
This review encompasses the benefits and harms of primary systemic thromboprophylaxis and mechanical thromboprophylaxis in adults with ALL during asparaginase‐containing chemotherapy compared with placebo or no thromboprophylaxis or comparing different groups of primary thromboprophylaxis.
Currently available drugs for the prevention of venous thromboembolism include the parenteral anticoagulants unfractionated heparin and low molecular weight heparin; the oral vitamin K antagonists; the synthetic pentasaccharides (fondaparinux, idaparinux, and idrabiotaparinux); the parenteral direct thrombin inhibitors (argatroban, desirudin, bivalirudin, and lepirudin); the direct oral anticoagulants (the direct thrombin inhibitor, dabigatran, and the factors Xa inhibitors rivaroxaban, apixaban, edoxaban, and betrixaban); and blood‐derived products for antithrombin replacement (intravascular antithrombin concentrates, cryoprecipitate, or fresh frozen plasma; all containing antithrombin (Mintz 1979)).
Low molecular weight heparin is usually preferred to vitamin K antagonists in individuals receiving concurrent myelosuppressive therapy due to efficacy (Piatek 2012) and as multiple drug interactions; food interactions/malnutrition; and liver dysfunction lead to fluctuations in the international normalised ratio (INR), which is used for therapeutic drug monitoring of vitamin K antagonists. Not to mention the slow onset of action as well as long half‐life of vitamin K antagonists, rendering them unmanageable with the frequent temporary cessation of anticoagulation during ALL therapy required for example, intrathecal chemotherapy or chemotherapy‐induced thrombocytopenia (Bauer 2006). Moreover, the production of the natural vitamin K‐dependent anticoagulants, protein C and S, decreases following initial administration of vitamin K antagonists secondary to the vitamin K depletion by vitamin K antagonists (Esmon 1987). Hypothetically, this may make clotting worse in presence of concurrent asparaginase‐induced protein C and S deficiency. Furthermore, unfractionated heparin is preferable in persons with renal failure or at a high risk of haemorrhage warranting rapid anticoagulation reversal, as the risk of bleeding events increases due to chemotherapy‐induced thrombocytopenia. On the other hand, heparins introduce the risk of heparin‐induced thrombocytopenia, and long‐term use of low molecular weight heparin for up 24 months has been associated with decreased bone mineral density in adults (Gajic‐Veljanoski 2016). Moreover, heparins and synthetic pentasaccharides directly rely on antithrombin for their mechanism of action (Kuhle 2006), which is usually depleted during asparaginase treatment. The flexibility of the reactive site loop of antithrombin increases as a result of a heparin‐/pentasaccharide‐induced conformational change, which enhances the anticoagulant activity of antithrombin against factor Xa, IIa (thrombin), and various others by more than 1000‐fold (heparin) (Björk 1982; Chuang 2001), or specifically against factor Xa by 300‐fold (pentasaccharide) (Bauer 2003; Bauer 2006). In this case, anti‐Xa as well as antithrombin activity monitoring with activity‐adapted antithrombin replacement may be necessary to maintain adequate anticoagulation during treatment with asparaginase. Currently available antithrombin concentrates are either derived from plasma or purified from transgenic goat milk as humanised recombinant protein. Only one adequately powered randomised trial to date exists supporting the use of antithrombin replacement as primary prevention of venous thromboembolism in children with ALL (Greiner 2019). Potentially, direct oral anticoagulants are better alternatives to low molecular weight heparin, as they act independently of antithrombin and can be administered orally (Sibson 2018). However, the bleeding risk needs to be balanced against the sparse evidence regarding the few available specific antidotes for direct oral anticoagulants (idarucizumab for the reversal of the direct thrombin inhibitor dabigatran; andexanet alfa and ciraparantag for factor Xa inhibitors), although direct oral anticoagulants have a short half‐life when compared with vitamin K antagonists (Godier 2019). Moreover, drug‐interactions with azole‐antimycotics used for fungal prophylaxis in ALL therapy complicate administration of direct oral anticoagulants. Generally, consensus guidelines suggest that full‐dose thromboprophylaxis should be administered and is without excessive bleeding risk, when the platelet count is above 50 x 109/L (Watson 2015).
Currently available mechanical interventions for the prevention of venous thromboembolism include intermittent pneumatic compression and graduated elastic stockings.
How the intervention might work
In a systematic review and meta‐analysis including 19,958 participants from nine randomised trials, thromboprophylaxis (unfractionated heparin/low molecular weight heparin/fondaparinux) was associated with a reduction in symptomatic non‐fatal and fatal pulmonary embolism and a non‐significant increase in major bleeding events in at‐risk hospitalised medically ill people (people with cancer included in one of nine trials, but unclear if any had an acute leukaemia diagnosis) (Dentali 2007). In a Cochrane systematic rreview and meta‐analysis including 12,352 participants from 26 trials (one of 26 trials enrolled children with ALL), primary thromboprophylaxis with low molecular weight heparin significantly reduced the incidence of symptomatic venous thromboembolism in people with solid/haematological cancer undergoing chemotherapy in the ambulatory setting, albeit data on the risk of major bleeding were inconclusive and suggested caution (Di Nisio 2016).
Two placebo‐controlled double‐blind randomised controlled trials (RCTs) contributed to this evidence by showing a significant reduction of venous thromboembolic events and a non‐significant increase in major bleeding events during prophylaxis with low molecular weight heparin (nadroparin/semuloparin) in people with cancer (people with leukaemia not eligible) treated with chemotherapy in the ambulatory setting (Agnelli 2009; Agnelli 2012). Yet, a randomised trial of thromboprophylaxis with dalteparin versus observation found a significantly higher incidence of clinically relevant bleeding but similar major bleeding in each arm and no significant reduction in venous thromboembolism in people with cancer (people with leukaemia not eligible) at high risk of venous thromboembolism (Khorana score ⋝3) in an outpatient setting (Khorana 2017). This study was underpowered. Of note, the Khorana score (range zero to six) is a validated tool to identify individuals with solid cancer in the outpatient setting at high risk of chemotherapy‐associated venous thromboembolism including cancer site, haemoglobin level or use of red cell growth factors, platelet count, leukocyte count, and body mass index (Khorana 2008). Importantly, this score has not been validated specifically for acute leukaemia, in which factors such as cancer site, platelet and leucocyte count have different underlying basis, and may therefore not be generalisable to an ALL setting.
Even more conflicting results have been reported for thromboprophylaxis with factor Xa inhibitors in people with cancer (people with acute leukaemia not eligible) at intermediate‐to‐high risk of venous thromboembolism (Khorana score ⋝2) receiving chemotherapy in an outpatient setting. In two placebo‐controlled double‐blind RCTs, Carrier and colleagues demonstrated a significantly lower rate of venous thromboembolism but a higher rate of major bleeding events for apixaban (Carrier 2019), whereas Khorana and colleagues did not find any difference in the incidence of venous thromboembolism or major bleeding with rivaroxaban (Khorana 2019). Furthermore, randomisation of additional intermittent pneumatic compression for critically ill persons treated with systemic thromboprophylaxis versus only systemic thromboprophylaxis in an intensive care unit setting, did not show any difference in incidence of venous thromboembolism (unclear if any individuals with acute leukaemia were included) (Arabi 2019). This controversy regarding the benefits and harms of direct oral anticoagulants was also seen in two recently published systematic reviews and meta‐analyses. Neumann and colleagues found no difference in the risk of venous thromboembolism but a slightly increased risk of major bleeding for direct oral anticoagulants when compared to low molecular weight heparin in hospitalised medically ill people (unclear if any individuals with acute leukaemia were included) (Neumann 2020), whereas direct oral anticoagulants compared with placebo significantly reduced the incidence of venous thromboembolism without any increased risk of major bleeding in people with cancer in the ambulatory setting (none of the included trials enrolled people with acute leukaemia) (Wang 2020). Worth noticing is the different populations of interest for the two reviews that might not translate into comparable findings.
Nevertheless, various different working groups have published clinical practice guidelines for thromboprophylaxis and generally recommend the use of primary thromboprophylaxis for people with cancer undergoing major surgery, being hospitalised, and being at high risk of venous thromboembolism (Khorana score ⋝2) in the outpatient setting, if the bleeding risk is acceptable (the European Society for Medical Oncology (ESMO) Mandala 2011; the American College of CHEST Physicians, Kahn 2012; the Scientific and Standardisation Committee on haemostasis and malignancy of the International Society on Thrombosis and Haemostasis (SSC of ISTH), Di Nisio 2014 and Wang 2019; Norwegian group, Kristiansen 2014; Canadian group, Easaw 2015; the National Comprehensive Cancer Network (NCCN) Streiff 2011, Streiff 2015; the American Society of Hematology (ASH) Schünemann 2018; the American Society of Clinical Oncology (ASCO), Key 2019; and the international Initiative on Thrombosis And Cancer (ITAC), Farge 2019). Additionally, the CHEST group recommends against pharmacologic thromboprophylaxis in hospitalised medically ill people at low risk of venous thromboembolism or at high risk of bleeding. Further, the group recommends thromboprophylaxis with heparins over no prophylaxis in ambulatory persons with solid tumours who have additional risk factors of venous thromboembolism (history of prior venous thromboembolism, immobilisation, hormonal therapy, treatment with angiogenesis inhibitors, thalidomide, and lenalidomide) and a low risk of bleeding (Kahn 2012; Kristiansen 2014). Yet, the American Society of Hematology guidelines strongly recommend systemic venous thromboembolic prophylaxis during hospitalisation in acutely or critically ill people who have an acceptable bleeding risk—and recommend use of mechanical prophylaxis in case of unacceptable bleeding risk (Schünemann 2018). The American Society of Clinical Oncology and the Scientific and Standardisation Committee on haemostasis and malignancy of the International Society on Thrombosis and Haemostasis guidelines suggest apixaban, rivaroxaban, and low molecular weight heparin are the preferred agents of choice (Key 2019; Wang 2019); direct oral anticoagulants being the first option in ambulatory people with cancer at high risk of venous thromboembolism and not at high risk of bleeding (Wang 2019). Specifically, the international Initiative on Thrombosis And Cancer group recommend for thromboprophylaxis use of heparins in individuals with cancer who have undergone surgery; use of heparins or fondaparinux in persons with cancer who are hospitalised; use of rivaroxaban/apixaban in persons with cancer at intermediate‐to‐high risk of venous thromboembolism (Khorana score ⋝2) who are ambulatory and receiving systemic anticancer treatment—not actively bleeding or at high risk of bleeding; use of vitamin K antagonists/low molecular weight heparin/low‐dose aspirin for individuals treated with immunomodulatory drugs combined with steroids or other systemic anticancer treatment (Farge 2019).
In summary, current guidelines for thromboprophylaxis in people with cancer based on high‐quality RCTs primarily concern cancer types such as advanced solid tumours and metastatic cancer as well as lymphoma and multiple myeloma. People diagnosed with acute leukaemia are frequently excluded from RCTs investigating prevention of venous thromboembolism due to their high risk of bleeding from prolonged periods of severe chemotherapy‐induced thrombocytopenia, disease‐ or treatment‐related distortion of haemostatic/fibrinolytic pathways, and more complex clinical setting. Yet, guidelines tailored to people with solid cancer are often applied to the clinical setting of acute leukaemia, as no such guidelines based on RCTs exist for people with acute leukaemia clearly outlining the benefits and harms associated with use of thromboprophylaxis (Annibali 2018).
Why it is important to do this review
The improved event‐free survival rate in people with ALL during the last decades may be jeopardised by the incidence of venous thromboembolism, as it is a frequently encountered treatment‐related toxicity in adults with ALL. Venous thromboembolism introduces a risk of treatment delay, omission, or premature discontinuation of the key anti‐leukaemic agent asparaginase and represents a non‐negligible risk of morbidity and highly likely compromised survival outcomes. Evidence‐based guidelines for pre‐emptive antithrombotic therapy in adults with ALL treated with asparaginase‐containing regimens are lacking, and the benefits of thromboprophylaxis need to be balanced against the harms such as (major) bleeding events.
Objectives
The primary objectives of this review were to assess the benefits and harms of primary thromboprophylaxis for first‐time symptomatic venous thromboembolism in adults with ALL receiving therapy with asparaginase compared with placebo or no thromboprophylaxis.
The secondary objectives were to compare the benefits and harms of different groups of primary systemic thromboprophylaxis in adults with ALL receiving therapy with asparaginase by stratifying the main results per type of drug (heparins, vitamin K antagonists, synthetic pentasaccharides, parenteral direct thrombin inhibitors, direct oral anticoagulants, and antithrombin substitutions).
Methods
Criteria for considering studies for this review
Types of studies
We planned to include:
-
randomised controlled trials (RCTs) including quasi‐randomised trials/controlled clinical trials (CCTs), cross‐over trials, and cluster‐randomised trials; however, none were included.
-
cohort studies and case‐control studies; we included only studies with control groups. Further, these studies were included only to describe the harms of thromboprophylaxis and were not included in meta‐analyses.
Types of participants
Adults aged at least 18 years diagnosed with ALL and treated with chemotherapy including asparaginase.
We excluded participants aged below 18 years, since another Cochrane Review already addressed the prevention of venous thromboembolism in children with cancer and tunnelled central venous catheters (Schoot 2013).
Types of interventions
We planned to conduct:
-
network meta‐analysis 1 including all trials of thromboprophylaxis; thus, comparing primary thromboprophylactic treatments (unfractionated heparin, low molecular weight heparin, vitamin K antagonists, synthetic pentasaccharides, parenteral direct thrombin inhibitors, direct oral anticoagulants, and antithrombin substitutions), mechanical thromboprophylaxis (intermittent pneumatic compression or graduated elastic stockings), or combined thromboprophylaxis with placebo or no thromboprophylaxis, or comparing systemic with mechanical thromboprophylaxis.
-
network meta‐analysis 2 including only drug trials; thus, comparing primary systemic prophylactic treatments for the prevention of venous thromboembolism (unfractionated heparin, low molecular weight heparin, vitamin K antagonists, synthetic pentasaccharides, parenteral direct thrombin inhibitors, direct oral anticoagulants, and antithrombin substitutions) directly (if available) or indirectly by using placebo‐controlled studies as data source.
If studies reported concomitant thromboprophylactic interventions (not part of direct comparisons) with pharmacologic drugs (heparins/vitamin K antagonists/synthetic pentasaccharides/parenteral direct thrombin inhibitors/direct oral anticoagulants) and blood‐derived products (antithrombin/fresh frozen plasma/cryoprecipitate) or mechanical interventions, we evaluated the following interventions as primary, listed in decreasing order: pharmacologic drugs, blood‐derived products, and mechanical interventions—if data were available.
No RCTs were identified as eligible. If in future updates RCTs become available for inclusion, RCTs and non‐RCTs will be analysed separately; only RCTs will be included in the meta‐analyses.
We did not impose any restrictions on frequency, duration, dosage, intensity, or route of administration.
Types of outcome measures
We included the following outcomes in both the primary and secondary objectives.
Of note, outcomes of benefits were not evaluated, as no RCTs were included in the main analyses. Given the inclusion of only non‐RCTs, outcomes of harms were analysed in a main descriptive analysis.
Primary outcomes
-
First‐time symptomatic venous thromboembolism during ALL treatment with asparaginase until four weeks after the last asparaginase dose (the cut‐off time point of measurable pegylated asparaginase activity (Henriksen 2017)) (benefit; dichotomous).
-
All‐cause mortality (harm; dichotomous).
-
Major bleeding events during treatment with primary systemic prophylactic treatment for the prevention of venous thromboembolism (harm; dichotomous), defined as non‐traumatic fatal bleeding and/or symptomatic haemorrhage in a critical area or organ (intracranial, intraspinal, intraocular, retroperitoneal, intra‐articular, pericardial, or intramuscular with compartment syndrome) and/or bleeding causing a decrease in haemoglobin level of at least 20 g/L (1.24 mmol/L) within one day, or leading to transfusion of at least 2 units of whole blood or red blood cells (Schulman 2005)).
Secondary outcomes
-
Mortality secondary to a venous thromboembolism event (harm; dichotomous).
-
Asymptomatic venous thromboembolism during ALL treatment with asparaginase until four weeks after the last asparaginase dose (benefit; dichotomous).
-
Adverse events during treatment with systemic primary prophylactic treatment (i.e. clinicall‐ relevant non‐major bleeding events defined as any non‐traumatic, non‐skin bleeding not fulfilling criteria for major bleeding (Schulman 2005) and heparin‐induced thrombocytopenia for trials using heparins) (harms; dichotomous).
-
Quality of life assessment during treatment with systemic primary prophylactic treatment using validated tools (harm; ordinal).
We grouped multiple variants of each of the outcome measures into appropriate groups to simplify comparisons, if necessary.
Search methods for identification of studies
We did not use any filters or restrictions.
Electronic searches
We searched the following electronic databases.
-
PubMed/MEDLINE (from 1970—the time of the introduction of asparaginase in adult ALL treatment—to 2020) (search strategy in Appendix 1)
-
Embase/Ovid (from 1970 to 2020) (search strategy in Appendix 2)
-
Scopus/Elsevier (from 1970 to 2020) (search strategy in Appendix 3)
-
Web of Science Core Collection/Clarivate Analytics (from 1970 to 2020) (search strategy in Appendix 4)
-
The Cochrane Central Register of Controlled Trials (CENTRAL) (the Cochrane Library, 2020, latest issue) (search strategy in Appendix 5)
The search strategies used for the above mentioned electronic databases are available in the appendices.
-
We used the "Related data—similar articles" option in PubMed/MEDLINE.
-
We used the "Find similar" option of included papers from the search string in Embase/Ovid.
-
We used the "Highly Cited in Field" feature in Web of Science Core Collection.
Searching other resources
Handsearches
Reference lists
-
The reference lists of identified studies and related reviews (with awareness of the risk of citation bias)
Trials
-
ClinicalTrials.gov registry (www.clinicaltrials.gov)
-
The International Standard Randomised Controlled Trial Number (ISRCTN) registry (www.isrctn.com/)
-
World Health Organization's International Clinical Trials Registry Platform (ICTRP) (http://apps.who.int/trialsearch/)
-
Pharmaceutical manufacturers of asparaginase (Servier and Takeda (pegylated asparaginase), Jazz Pharmaceuticals and Ohara Pharmaceuticals (Erwiniachrysanthemi‐derived asparaginase), and Kyowa Pharmaceuticals (native Escherichia coli‐derived L‐asparaginase))
Conference proceedings
-
The American Society of Hematology (ASH) (from 1970 to 2019)
-
The European Haematology Association (EHA) (from 1994—the year of the first EHA congress—to 2019)
-
The American Society of Clinical Oncology (ASCO) (from 1970 to 2019)
-
The International Society on Thrombosis and Haemostasis (ISTH) (from 1970 to 2019).
We included the above mentioned literature from handsearches, if we were able to obtain adequate information from either the abstract or personal communication with the authors of the relevant studies. Additionally, we contacted the authors of relevant studies to identify any unpublished material, missing data, or information regarding ongoing studies.
The most recent set of searches was conducted on 02 June 2020, from which all records have been screened for relevance. Any studies meeting the eligibility criteria have been fully incorporated into the review.
Data collection and analysis
We summarised data in accordance with standard Cochrane methodologies. We did not allow any restrictions on language. We planned to analyse data from different study designs separately (RCTs and non‐RCTs); however, we did not include any RCTs.
Selection of studies
We uploaded the results of the electronic searches and handsearches to the Covidence systematic review software for screening and study selection (www.covidence.org), removing duplicate records of the same study. Two review authors (C.U.R. and L.S.L.) independently screened the titles and abstracts of identified articles for eligibility. The two review authors were not blinded to the journal, authors, institution, or results. We retrieved all full text of articles judged as potentially eligible by at least one of the review authors. The same two review authors independently screened the full‐text articles for eligibility using a standardised form with inclusion and exclusion criteria and resolved disagreements by discussion or by consulting a third review author (B.A.N.). The two review authors compared reports by author names, location, setting, sample sizes, intervention details, and date to detect any duplicate publications. We contacted study authors to resolve uncertainties or if information was inadequate. We produced a flow chart of the selection process, created a list of excluded studies (covering all studies that on the surface appeared to meet the eligibility criteria but on further inspection did not), and described reasons for exclusion of any study considered for this review.
Data extraction and management
Two review authors (C.U.R. and L.S.L.) independently performed data extraction using a standardised data collection form including the following.
General information
-
Study identification (ID) (citation ID)
-
Review author ID
-
Review author ID checking extracted data
-
Study author contact details
-
Publication type
-
Language
Characteristics of included studies
Methods
-
Aim of study/primary and secondary outcomes
-
Study design
-
Unit of allocation (RCTs) / method of assigning the intervention (non‐RCTs)
-
Start and end dates
-
Total study duration
-
Follow‐up start and end dates
-
Follow‐up duration
-
Stopping rules
-
Source of funding
-
Possible conflicts of interest
Participants
-
Population description
-
Setting
-
ALL characteristics (e.g. Philadelphia chromosome status, immunophenotype, central nervous system involvement, risk group (including definition))
-
Treatment characteristics (e.g. ALL treatment protocol (paediatric versus adult); treatment phase; intrathecal chemotherapy; corticosteroids type and dosing; asparaginase type, dose (unit), frequency, route of administration, and duration (including the time of the last dose))
-
Venous thromboembolism characteristics (e.g. co‐morbidities, risk factors of venous thromboembolism)
-
Study inclusion and exclusion criteria
-
Method of recruitment of participants
-
Total number screened for inclusion
-
Total number randomised/recruited
-
Clusters
-
Number randomised/allocated per group
-
Number of allocated participants, who received planned treatment in each treatment arm
-
Source of control group (non‐RCTs)
-
Number missing including reasons
-
Number of participants moved from one group to another including reasons
-
Baseline imbalances
-
Age
-
Sex
-
Race/ethnicity
-
Other relevant socio demographics
-
Subgroups measured
-
Subgroups reported
Intervention groups
-
Group name
-
Description/type
-
Duration of treatment period
-
Timing/frequency
-
Antithrombin replacement threshold
-
Dose (unit)
-
Delivery/route of administration
-
Providers
-
Co‐interventions
-
Economic information
-
Resource requirements
-
Integrity of delivery
-
Compliance
-
Truncation/omission/delay/dose reduction including reason (e.g. thrombocytopenia)
Data and analysis
Dichotomous outcomes (all‐cause mortality, first‐time symptomatic venous thromboembolism, major bleeding events, venous thromboembolism‐related mortality, asymptomatic venous thromboembolism, and adverse events (clinically relevant non‐major bleeding events and heparin‐induced thrombocytopenia).
Ordinal data (quality of life assessment)
-
Definitions, diagnostic criteria, and diagnostic methods
-
Causes for mortality
-
Venous thromboembolism‐related mortality: how and who defined the causal relation between venous thromboembolism and death
-
Timing
-
Person reporting/measuring
-
Unit of measurement (bleeding events, quality of life)
-
Scales: upper and lower limits (quality of life)
-
Is the outcome tool validated? (quality of life)
-
Imputation of missing data
-
Power
-
Results (e.g. number with event and number measured in each group at time points or by subgroup, if relevant)
-
Time points measured but not reported
-
Any other results reported
-
Unit of analysis
-
Statistical methods used and appropriateness of these
-
Re‐analysis required? Possible? Results?
Other
-
Key conclusions of the study authors
-
References to other relevant studies
-
Correspondence required for further information
We extracted all data from studies reported in more than one publication directly into a single data collection form. The two review authors resolved disagreements by discussion or by consulting a third review author (B.A.N.). We contacted authors from identified studies for additional information when necessary.
If we are able to extract data only from figures in future updates, we will use Plot Digitizer software (http://plotdigitizer.sourceforge.net/).
Assessment of risk of bias in included studies
We assessed the effect of assignment to intervention (the 'intention‐to‐treat' effect).
Non‐RCTs
We assessed the risk of bias for non‐RCTs in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Sterne 2019b), using the ROBINS‐I tool (Sterne 2016) with signalling questions for the following domains.
-
Bias due to confounding
-
Bias in selection of participants
-
Bias in classification of interventions
-
Bias due to deviations from intended interventions
-
Bias due to missing data
-
Bias in measurement of outcomes
-
Bias in selection of the reported result
We used the following response options to the signalling questions: 'yes'; 'probably yes'; 'no'; 'probably no'; and 'no information'. We rated the possible risk of bias in each of the seven domains as ‘low risk’, ‘moderate risk’, ‘serious risk’, ‘critical risk’, or ‘no information’ for each available outcome of each included study. We assessed an overall risk of bias by the following method.
-
'Low risk of bias': the study outcome is judged to be at low risk of bias in all of the tool's seven domains.
-
'Moderate risk of bias': the study outcome is judged to be at low to moderate risk of bias in all of the tool's seven domains.
-
'Serious risk of bias': the study outcome is judged to be at serious risk of bias in at least one of the tool's seven domains.
-
'Critical risk of bias': the study outcome is judged to be at critical risk of bias in at least one of the tool's seven domains.
-
'No information on bias': when information in one or more key domains is lacking.
We excluded studies with outcomes that we judged to be critical in at least one of the ROBINS‐I domains from the main descriptive analysis and included all studies no matter their risk of bias in a sensitivity analysis.
We prespecified the following main confounding factors and co‐interventions.
-
Age (variability in the age of persons included; younger adults 18 to 45 years versus older adults > 45 years)
-
Sex
-
Paediatric(‐inspired) treatment protocols versus conventional adult treatment protocols
-
Asparaginase type (native Escherichia coli‐derived, Erwiniachrysanthemi‐derived, and pegylated asparaginase) and cumulative dose
-
Type and dosing of intrathecal chemotherapy and corticosteroids
-
Risk factors of venous thromboembolism (e.g. prior venous thromboembolism, infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, smoking, and obesity)
-
Co‐morbidity (e.g. liver disease or renal insufficiency)
-
Co‐interventions (antithrombin concentrates, cryoprecipitate, fresh frozen plasma, fibrinogen concentrates, and platelet infusions)
RCTs
We did not identify any RCTs as eligible. However, if we include RCTs in future updates, we will evaluate the risk of bias for RCTs using the Cochrane 'Risk of bias' tool (RoB 2.0 tool) (Sterne 2019a) using signalling questions for the following domains
-
Bias arising from the randomisation process (selection bias)
-
Bias due to deviations from intended interventions (performance bias)
-
Bias due to missing outcome data (attrition bias)
-
Bias in measurement of the outcome (detection bias)
-
Bias due to selective reporting (reporting bias)
We will use the following response options to the signalling questions: 'yes'; 'probably yes'; 'no'; 'probably no'; and 'no information', ultimately rating the possible risk of bias in each of the domains as ‘low risk’, ‘high risk’, or 'some concerns'. We will judge outcomes/studies with at least one domain of high risk to be at high risk of bias overall.
Overall
We compared outcomes between trial registrations/protocols and published reports. Where trial registrations/protocols were not available, we compared outcomes reported in the methods and results sections.
Two review authors (C.U.R. and L.S.L.) independently assessed the overall risk of bias for each included outcome of each included study and at study level using the above mentioned tools. If a discrepancy between the review authors occurred and no agreement could be reached, a third review author was involved (B.A.N.). In the case of insufficient reporting, we contacted study authors for additional information. We took the evaluated risk of bias in included studies into account when interpreting the results of the review.
We have presented the results of the 'Risk of bias' assessments as additional tables for each included study.
We addressed the risk of bias at the systematic review level using ROBIS tool (Whiting 2016), which included the following three phases with associated signalling questions provided for each phase.
-
Assess relevance.
-
Identify concerns with the review process including the domains: study eligibility criteria, identification and selection of studies, data collection and study appraisal, and synthesis and findings.
-
Judge the overall risk of bias in the interpretation of the review findings (results and conclusions) and whether this interpretation considered limitations identified in any of the phase 2 domains.
Measures of treatment effect
Non‐RCTs
For dichotomous outcomes, we recorded the number of events and the total number of participants in both the treatment and control groups and calculated the risk ratio (RR) with corresponding 95% confidence interval (CI) for each study. In future updates, we will extract and report adjusted effect measures such as the RR with a 95% CI from statistical analyses adjusting for baseline differences (such as Poisson regressions or logistic regressions) or the ratio of RRs (i.e. the post‐intervention RR/pre‐intervention RR), if available—in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Reeves 2019).
In future updates, ordinal outcomes will be recorded either as dichotomous data or continuous data depending on the way that the study authors present their data. Thus data extraction will involve calculation of numbers of participants above and below a defined threshold and/or mean values and standard deviations (SDs). If available, we will extract and report the absolute change from a statistical analysis adjusting for baseline differences (such as regression models, mixed models, or hierarchical models), or the relative change adjusted for baseline differences in the outcome measures (i.e. the absolute post‐intervention difference between the intervention and control groups as well as the absolute pre‐intervention difference between the intervention and control groups/the post‐intervention level in the control group). Nonetheless, we will extract data in all given forms, as we will not known the most common form used until all future studies have been reviewed.
RCTs
We did not identify any RCTs eligible for inclusion.
In future updates regarding dichotomous outcomes from RCTs, we will record the number of events and the total number of participants in both the treatment and control groups. Moreover, we will report the pooled RR with corresponding 95% CI, in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2019a; Deeks 2019b). Ordinal outcomes will be meta‐analysed as dichotomous data or continuous data depending on the way that the study authors performed the original analyses. Thus data extraction will involve calculation of numbers of participants above and below a defined threshold and/or mean values and standard deviations (SDs). We will extract data in all given forms, as we will not know the most common form used until all future studies have been reviewed. For continuous outcomes originating from the same scale, we will perform analyses using the mean difference (MD) with 95% CIs. If continuous outcomes originate from different scales, we plan to use standardised mean difference (SMD). We will report the number needed to treat to for an additional beneficial outcome (NNTB) and the number needed to treat for an additional harmful outcome (NNTH) with 95% CIs from the summary statistic (RR) of the meta‐analyses according to the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2019b). We will carry out the statistical analyses using Review Manager Web (RevMan Web 2019), and the statistical computing software R (R 2019).
Unit of analysis issues
In future updates, we will conduct the analysis at the same level as the allocation for cluster‐randomised trials, if relevant. However, we will seek statistical advise, if the clusters markedly vary in size potentially leading to unnecessary reduction of the precision of the effect estimate. If available and relevant, we plan to extract a direct estimate of the required effect measure (e.g. a RR with its CI) from an analysis that properly accounted for the cluster design (Higgins 2019c).
For cross‐over trials in future updates, if relevant, we will include and analyse only the first treatment given, if the wash‐out period does not eliminate any carry‐over effect of thromboprophylactic drugs regarding elimination half‐life. We will contact the authors of trials reporting only paired analyses (i.e. not reporting data for the first period separately) to avoid omission of these and, thus, potential introduction of bias at the meta‐analysis level (Higgins 2019c).
If the study authors have not adjusted for correlation, we plan to adjust the results in accordance with the advice given in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019c).
For studies with multiple treatment groups, two review authors (C.U.R. and L.S.L.) excluded subgroups that were considered irrelevant to the analysis. We tabulated all subgroups in the 'Characteristics of included studies' table. When appropriate, we combined groups to create a single pairwise comparison. If this was not possible, we selected the most appropriate pair of interventions and excluded the others (Higgins 2019a). Multiple observations for the same outcome were unlikely to be included in this review. In future updates, if participants are randomised more than once, we will contact the authors of the study to provide us with data associated with the initial randomisation.
Dealing with missing data
If data of relevance to the analyses were missing or unclear, we contacted the authors of the concerned study to retrieve the missing data. If unsuccessful and possible in future updates, our analyses will be based on the number reaching follow‐up, and we will perform analyses for worse‐ and best‐case scenarios. We will also record the number of persons lost to follow‐up for each study, when available.
We analysed data by the effect of assignment to intervention (the intention‐to‐treat effect). If we encountered insufficient data, we carried out the analyses for each study twice including missing data as success and as failure, respectively. In addition, we performed an as‐treated/per protocol analysis (Higgins 2019b).
Assessment of heterogeneity
In future updates, we will only consider combining the data and performing network meta‐analyses, if the clinical/participant and methodological/study characteristics of individually included studies are sufficiently homogeneous in terms of participants, interventions, and outcomes. Before conducting the network meta‐analyses, we will make sure that all treatments are 'jointly randomisable' for the persons and study settings and carefully evaluate the following.
-
Homogeneity within each comparison, e.g. is there a similar participant population or treatment dose?
-
Similarity across all comparisons, i.e. are the studies combinable?
-
Consistency between the results from direct and indirect comparisons.
Thus, we plan to assess heterogeneity of the treatment effects between trials within each comparison in the network, evaluating how much variation of treatment effect there is amongst the studies in each comparison by both visual inspection of the forest plots and by using the Chi2 test with a significance level of P < 0.1. We will use the I2 heterogeneity statistic to help quantify the degree of potential heterogeneity—classifying the results as low (I2 = 0% to 40%); moderate (I2 = 30% to 60%); substantial (I2 = 50% to 90%); and considerable (I2 = 75% to 100%) heterogeneity across trials (Deeks 2019b). If statistical heterogeneity is considerable, we will not report the overall summary statistic. We plan to investigate possible reasons for heterogeneity by sensitivity and subgroup analyses (Deeks 2019b), if substantial heterogeneity (I2 ≥ 50%) is revealed.
Assessment of reporting biases
We searched multiple electronic databases, conference proceedings, and trial registries for both unpublished and published studies dealing with publication bias, location bias, and time lag bias. We had no limits or language restrictions in the search strategies, thus avoiding language bias. Duplicate reports of the same study were identified and excluded using the Covidence systematic review software (www.covidence.org) and by comparing authors, location, setting, and sample size, thus avoiding duplicate publication bias. We addressed selective outcome reporting bias by searching for a trial registration, a protocol, or by scrutinising the methods section for a list of outcomes for comparison with the reported published outcomes of each included study.
In future updates, we will evaluate publication bias and other biases related to small‐study size by constructing a funnel plot of effect estimates and using a linear regression test for ordinal/continuous outcomes, if we have a minimum of 10 studies included in the meta‐analysis—otherwise the power of the test will be too low to distinguish chance from real asymmetry (Page 2019; Sterne 2011). We will consider P < 0.1 as significant (Page 2019).
Data synthesis
Non‐RCTs
We showed the results of included studies in forest plots but without the pooled estimate—as non‐RCTs were not included in the meta‐analyses. We have produced a narrative report and presented the data in tables, as we included only observational studies.
RCTs
Only RCTs could have been included in the meta‐analyses; however, none were identified as eligible. In future updates, if studies are sufficiently homogenous in their study design, we will conduct two main network meta‐analyses according to the recommendations of Cochrane (Cipriani 2013; Chaimani 2019; Salanti 2008; Salanti 2011; Salanti 2012), aiming for a ranking of thromboprophylactic treatments using summary outputs from the network meta‐analysis. We will use the random‐effects model for all analyses, as we anticipate that true effects were related but were not the same for included studies.
We will conduct the following two main network meta‐analyses:
-
network meta‐analysis 1 including trials of all types of thromboprophylaxis; thus comparing primary thromboprophylactic treatments (unfractionated heparin, low molecular weight heparin, vitamin K antagonists, synthetic pentasaccharides, parenteral direct thrombin inhibitors, direct oral anticoagulants, and antithrombin substitutions), mechanical thromboprophylaxis (intermittent pneumatic compression or graduated elastic stockings), or combined thromboprophylaxis with placebo or no thromboprophylaxis, or comparing systemic with mechanical thromboprophylaxis;
-
network meta‐analysis 2 including drug trials only; thus comparing primary systemic prophylactic treatments for the prevention of venous thromboembolism (unfractionated heparin, low molecular weight heparin, vitamin K antagonists, synthetic pentasaccharides, parenteral direct thrombin inhibitors, direct oral anticoagulants, and antithrombin substitutions) directly (if available) or indirectly by employing placebo‐controlled studies as data source.
Different thresholds within the comparisons will only be grouped together, if they are considered to be clinically similar. If we found considerable heterogeneity across studies and we identify a cause for the heterogeneity, we will explore this with subgroup analyses. If we cannot find a cause for the heterogeneity, then we will not perform a meta‐analysis but comment on the results narratively and present the results from all studies in tables.
We will report the NNTB and NNTH with 95% CIs from the summary statistic (RR) of the meta‐analyses according to the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2019b). We will carry out data analyses using Review Manager Web (RevMan Web 2019) and the statistical computing software R (R 2019). We plan to use R package 'netmeta' for the network meta‐analyses.
Subgroup analysis and investigation of heterogeneity
If adequate data were available, we performed subgroup analyses for each of the following outcomes in order to assess the effect on heterogeneity.
-
Adults treated according to paediatric‐inspired treatment protocols versus adults treated according to adult treatment protocols
-
Philadelphia chromosome‐positive versus Philadelphia‐chromosome‐negative participants
-
Participants with versus without inherited thrombophilia traits
-
Asparaginase type (Escherichia coli‐derived versus pegylated asparaginase versus Erwiniachrysanthemi‐derived)
Sensitivity analysis
We included all non‐randomised studies no matter their risk of bias in a sensitivity analysis.
In future updates, we will assess the robustness of the results by performing the following sensitivity analyses, when possible.
-
Fixed‐effect meta‐analysis—as large differences in estimated effects indicate small‐study effects.
-
Including studies with a ‘low risk of bias’ in a new sensitivity meta‐analysis.
-
Including RCTs/quasi‐randomised trials in a new sensitivity meta‐analysis (excluding cluster‐randomised and cross‐over trials).
-
Including studies with less than a 20% drop out in a new sensitivity meta‐analysis.
Summary of findings and assessment of the certainty of the evidence
We used the GRADE approach to evaluate the certainty of evidence (Atkins 2004; Guyatt 2008; Schünemann 2019a; Schünemann 2019b), including risk of bias, inconsistency, imprecision, indirectness, publication bias—and additional considerations for non‐RCTs (e.g. large or very large effects, presence of a dose‐response gradient, and a result that opposes any plausible residual confounding). We rated the certainty of the evidence as 'very low', 'low', 'moderate', or 'high' using the GRADE considerations. Outcome data from non‐randomised studies start at 'low' certainty. In future updates, we will also use the CINeMA software (CINeMA 2017) to evaluate the confidence in the findings from the meta‐analyses.
Two review authors (C.U.R. and L.S.L.) independently assessed the certainty of the body of evidence of each outcome using the above mentioned tools. If a discrepancy between the review authors occurred and no agreement could be reached, a third review author was involved (B.A.N.). In the case of insufficient reporting, we contacted the study authors for additional information.
We presented a ’Summary of findings’ table according to the guidelines of the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2019a) including only the outcomes listed below in order of most relevant endpoints for participants.
-
All‐cause mortality
-
First‐time symptomatic venous thromboembolism
-
Major bleeding events
-
Venous thromboembolism‐related mortality
-
Quality of life
-
Asymptomatic venous thromboembolism
-
Adverse events (i.e. clinically relevant non‐major bleeding and heparin‐induced thrombocytopenia for trials using heparins)
We generated the 'Summary of findings' table using the GRADEpro GDT software (GRADEpro GDT 2015). In future updates, the 'Summary of findings' table will be created using the MAGICapp software (MAGICapp 2018), if network meta‐analyses are included.
Results
Description of studies
See Characteristics of included studies; Characteristics of excluded studies; Characteristics of studies awaiting classification.
Results of the search
See flow diagram in Figure 1.
Flow diagram
As of 02 June 2020, we retrieved a total of 2706 references from database searching and 27 references from handsearching, which were gathered in an EndNote X8.1 library and directly imported into Covidence as one XML file. A total of 963 duplicates were automatically removed in Covidence. Following title and abstract screening, we excluded 1694 of 1770 reports. We considered 76 reports eligible for full‐text screening, of which 41 reports (23 studies) met the inclusion criteria of this review and seven reports (six studies; four full‐text articles, one study protocol, and two conference abstracts) await classification; we excluded 28 reports (22 studies).
We have contacted the corresponding and/or first author of the six studies awaiting classification of in‐ or exclusion, respectively, as we do not know whether the participants were diagnosed with ALL in half of the studies, and whether participants with ALL received asparaginase treatment. See Characteristics of studies awaiting classification for details.
We did not identify any prospectively registered ongoing trials that met the inclusion criteria.
Noteworthy is that we identified 14 studies by handsearching reference lists; the majority being studies including people with a cancer or haematological malignancy diagnosis, which was not caught by our search strategy given the lack of ALL indexing. None were included after full‐text screening; three studies are awaiting classification.
Included studies
See Characteristics of included studies.
We assessed a total of 23 studies (1 RCT and 22 non‐RCTs) merged from 41 records as eligible for this review, of which 14 were full‐text articles, two were correspondences/letters, and seven were conference abstracts.
An overview of the included studies has been provided in Table 1. We contacted the corresponding author (if not available, the first author/co‐author) of each included study to retrieve additional data needed for analysis; approximately half of the contacted study authors responded to our request (overview of author correspondence and additional data in Table 2).
| Study | Risk of bias assessment | Inclusion in descriptive analyses | Annotations |
|---|---|---|---|
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in main descriptive analysis | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in main descriptive analysis | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | no | Not included in descriptive analysis. Critical risk of bias due to control group inconsistency. Participants who did not receive antithrombin supplementation had antithrombin activity levels above the target of supplementation (<70%) cannot be classified as controls. Of note, the aim of the study was to assess the safety profile of Erwinia chrysanthemi‐derived asparaginase in adults with ALL. | |
| yes | no | Not included in descriptive analysis. Critical risk of bias due to control group inconsistency. Unclear if antithrombin activity was monitored in all included participants, and if only participants who had antithrombin activity levels ≥ 70% were allocated to the control group. | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data (and control group inconsistency) | |
| no | no | Complete lack of outcome data (and control group inconsistency) |
| Response from study author | Study | Format | Provided data | Annotations |
|---|---|---|---|---|
| YES | letter |
| Invalid control group definition. | |
| full report |
| |||
| full report |
| |||
| conference abstract |
| |||
| correspondence |
| |||
| full report |
| |||
| full report | ‐ | Study database closed in 2005; data not available. | ||
| full report |
| Most of the requested additional data (outcome data, patient/disease characteristics) could not be provided due competing interests | ||
| full report |
| |||
| full report |
| |||
| conference abstract |
| |||
| NO | full report | ‐ | Complete lack of data of interest. | |
| conference abstract | ‐ | Incomplete outcome data. | ||
| full report | ‐ | Complete lack of data of interest. | ||
| full report | ‐ | Complete lack of data of interest. | ||
| full report | ‐ | Incomplete outcome data. Suspected invalid control group definition. | ||
| conference abstract | ‐ | Incomplete outcome data. | ||
| full report | ‐ | Complete lack of data of interest. | ||
| conference abstract | ‐ | Complete lack of data of interest. | ||
| full report | ‐ | Complete lack of data of interest. | ||
| conference abstract | ‐ | Complete lack of data of interest. Control group inconsistency. | ||
| full report | ‐ | Incomplete outcome data. | ||
| conference abstract | ‐ | Complete lack of data of interest. Control group inconsistency. |
We identified 10 of the 23 included studies (1 RCT and 9 non‐RCTs comprising seven full‐text articles Abdelkefi 2004; Belmonte 1991; Couturier 2015; Gugliotta 1990; Hunault‐Berger 2008; Lauw 2013; Pogliani 1995 and three conference abstracts McCloskey 2017; Renteria 2018; Underwood 2019) that all met the inclusion criteria but provided no outcome data for adults aged at least 18 years with an ALL diagnosis. Accordingly, we did not include these 10 studies in the 'Risk of bias' assessment and the main descriptive analysis. These studies are presented in a separate overview in Table 3.
| Study | Participants | Interventions | Outcomes of harms |
|---|---|---|---|
| 108 participants (4 to 58 years) | Intervention (4/55 ALL) Control (5/53 ALL) Co‐interventions: | Outcome data for participants aged 4‐58 years withhaemato‐oncologicaldiseases. Major bleeding All‐cause mortality, venous thromboembolism‐related mortality, clinically relevant non‐major bleeding, and quality of life were not reported. | |
| 30 participants (16‐58 years) with ALL (n = 23) or lymphoblastic lymphoma treated with Escherichia coli‐derived L‐asparaginase‐containing consolidation chemotherapy. | Intervention (17/17 ALL) Antithrombin substitution schedule A: Antithrombin substitution schedule B:
Control (6/13 ALL) Co‐interventions: | Outcome data for participants aged 16‐58 years with ALL or lymphoblastic lymphoma. All‐cause mortality, major bleeding, clinically relevant non‐major bleeding, venous thromboembolism‐related mortality, and quality of life were not reported. | |
| 706 adults (18 to 59 years) with Philadelphia chromosome‐negative ALL or lymphoblastic lymphoma treated with paediatric‐inspired Escherichia coli‐derived L‐asparaginase‐containing induction chemotherapy. | Intervention (n=179) Control (n=89) Co‐interventions: | Thromboprophylaxisdata for 268/706 of the included adults with ALL or lymphoblastic lymphoma. Venous thromboembolism‐related mortality All‐cause mortality, major bleeding, clinically‐relevant non‐major bleeding, heparin‐induced thrombocytopenia, and quality of life were not reported. | |
| 15 participants (16 to 58 years) with ALL or lymphoblastic lymphoma treated with Escherichia coli‐derived L‐asparaginase‐containing consolidation. | Intervention (n = 7) Control (n = 8) Co‐interventions: | Outcome data for participants aged 16‐58 years with ALL or lymphoblastic lymphoma. Bleeding events All‐cause mortality, venous thromboembolism‐related mortality, and quality of life were not reported. | |
| 214 participants (15 to 59 years) diagnosed with either non‐Burkitt type ALL (n = 191) or T‐cell lymphoblastic lymphoma receiving Escherichia coli‐derived asparaginase‐containing induction therapy. | Intervention (n = 85) Control (n = 127) Co‐interventions: | Thromboprophylaxisdata for 212/214 of the included participants aged 15‐59 years with ALL orlymphoblasticlymphoma. Bleeding events Venous thromboembolism‐related mortality All‐cause mortality, heparin‐induced thrombocytopenia, and quality of life were not reported. | |
| 236 participants (16 to 59 years) with newly diagnosed ALL treated according to the HOVON‐37 ALL remission induction including Escherichia coli L‐asparaginase. | Intervention (n = 193) Control (n = 43) Co‐interventions: | Outcome data for participants aged 16‐59 years with ALL. | |
| 27 adults (19 to 49 years) with newly diagnosed Philadelphia chromosome‐negative ALL (non‐Burkitt type) treated according to the paediatric‐inspired AHCVAD (augmented hyper‐CVAD) protocol including pegylated asparaginase. | Intervention Control Co‐interventions: | Outcome data for adults with ALL. | |
| 20 participants (16 to 58 years) with ALL treated according to ALL regimens including L‐asparaginase during induction treatment. | Intervention (n = 8) Control (n = 12) Co‐interventions: | Outcome data for participants aged 16‐58 years with ALL. | |
| 12 adults (45 to 76 years) with Philadelphia chromosome‐negative ALL treated according to an age‐based, dose‐adjusted pegylated asparaginase‐based regimen with a CALGB 10403 backbone. | Intervention (n = 8) Control (n = 4) Co‐interventions:Cryoprecipitate (if fibrinogen <120mg/dL). | Outcome data for adults with ALL. Treatment‐related mortality (all‐cause mortality?) Major bleeding, venous thromboembolism‐related mortality, clinically relevant non‐major bleeding, and quality of life were not reported. | |
| 46 adults (18 to 38 years) with ALL treated with an pegylated asparaginase‐based regimen. | Intervention Control Co‐interventions: | Outcome data for adults with ALL. All‐cause mortality, major bleeding, venous thromboembolism‐related mortality, clinically relevant non‐major bleeding, and quality of life were not reported. |
In summary, we found 13 relevant studies (all non‐RCTs), of which 11 studies did not report important baseline characteristics and/or control for important baseline confounders (including invalid control group definitions in two of 11 studies). The remaining two studies were conducted in UK/Canada and compared antithrombin with no antithrombin concentrates and low molecular weight heparin with no low molecular weight heparin, respectively, for people with ALL.
We characterised the 13 studies to be of non‐randomised follow‐up (observational) design including a total of 1734 participants; 12 of 13 studies were retrospective with prospective registration of venous thromboembolism in two studies (Orvain 2020; Rank 2018), and one study had a combined retrospective and prospective design (Grace 2018). The studies were from Canada (one study), Canada/USA (one study), Belgium/France/Switzerland (one study), Italy (one study), Nordic/Baltic (Estonia and Lithuania) countries (one study), Portugal (one study), USA (six studies), and UK (one study). Four studies did not include evaluation of the preemptive antithrombotic intervention of interest for this review as their primary study aim (Bigliardi 2015; Elliott 2004; Freixo 2017; Rank 2018). Five studies included only Philadelphia chromosome‐negative ALL (Bigliardi 2015; Grace 2018; Orvain 2020; Rank 2018; Sibai 2020), two studies included both Philadelphia chromosome‐negative and ‐positive ALL (Elliott 2004; Farrell 2016), and the remaining studies did not include any information on Philadelphia chromosome status. Thromboprophylaxis was administered during induction treatment in six studies (Al Rabadi 2017; Bigliardi 2015; Elliott 2004; Farrell 2016; Orvain 2020; Umakanthan 2016), during age‐modified intensification phase in one study including only participants in first complete remission (Sibai 2020), from induction phase throughout asparaginase treatment in two studies (Grace 2018; Rank 2018), during any asparaginase‐containing phase of ALL treatment (i.e. induction, consolidation, maintenance, salvage) in one study (Chen 2019), and unspecified/during asparaginase‐based chemotherapy in the remaining three conference abstracts (Freixo 2017; George 2020; Grose 2018). The thromboprophylactic treatments reported were low molecular weight heparin (Al Rabadi 2017; Freixo 2017; Rank 2018; Sibai 2020; Umakanthan 2016), low molecular weight heparin/unfractionated heparin (Grace 2018; Orvain 2020), and antithrombin concentrates (Bigliardi 2015; Chen 2019; Elliott 2004; Farrell 2016; George 2020; Grose 2018). None of the participants in the comparator groups received the thromboprophylactic intervention of interest. None of the included studies used primary pharmacological thromboprophylaxis with vitamin K antagonists, synthetic pentasaccharides, parenteral direct thrombin inhibitors, direct oral anticoagulants (oral direct thrombin inhibitor or factors Xa inhibitors), or primary non‐pharmacological thromboprophylaxis such as intermittent pneumatic compression and graduated elastic stockings.
Two non‐randomised follow‐up studies evaluated heparin prophylaxis (unspecified; low molecular weight heparin/unfractionated heparin) versus no heparin prophylaxis.
-
Grace 2018 included 85 adults with newly diagnosed Philadelphia chromosome‐negative ALL enrolled on a paediatric‐inspired Dana Farber Cancer Institute (DFCI) treatment protocol. Included adults received either no heparin prophylaxis or varying type and dosage of low molecular weight heparin/unfractionated heparin prophylaxis once daily subcutaneously (enoxaparin 30 mg to 40 mg (induction) and 30 mg to 100 mg (consolidation), dalteparin 5000 international units (IU) (induction and consolidation), or unfractionated heparin 5000 U (consolidation)), recommended from ALL diagnosis or at least on day seven (the day of the first asparaginase dose) through induction (four weeks), consolidation I (nine weeks), central nervous system therapy (three weeks), and consolidation II (24 to 27 weeks). The median follow‐up was 52 months, range was not reported.
-
Orvain 2020 evaluated 784 adults (prophylaxis data available in 662/784) with newly diagnosed Philadelphia chromosome‐negative ALL treated according to the Group for Research of Adult Acute Lymphoblastic Leukaemia (GRAALL)‐2005 protocol. Participants received either no heparin prophylaxis or prophylaxis with low molecular weight heparin (dose not reported, subcutaneously)/unfractionated heparin (100 UI/kg, intravenous continuous infusion) once daily, recommended during induction. Follow‐up time was not reported.
Five non‐randomised follow‐up studies evaluated prophylaxis with low molecular weight heparin prophylaxis versus no prophylaxis with low molecular weight heparin.
-
Al Rabadi 2017 recruited 38 adults diagnosed with ALL and treated with asparaginase‐containing induction treatment. Participants either received no prophylaxis with low molecular weight heparin or prophylaxis with low molecular weight heparin (enoxaparin, 1 mg/kg subcutaneously) once daily from the first dose of asparaginase until the day of discharge (30 days of induction treatment). Follow‐up time was not reported.
-
Freixo 2017 included 26 adults with ALL treated according to the Hemato‐Oncologie voor Volwassenen Nederland (HOVON) 100/China Lymphoma Collaborative Group (CLCG) treatment protocols. Participants received either no prophylaxis with low molecular weight heparin or prophylaxis with low molecular weight heparin (enoxaparin, 40 mg to 60mg subcutaneously) once daily during intensive chemotherapy. Follow‐up time was not reported.
-
Rank 2018 enrolled 274 adults with newly diagnosed Philadelphia chromosome‐negative ALL treated according to the Nordic society Of Paediatric Haematology and Oncology (NOPHO) ALL2008 protocol. Participants received either no prophylaxis with low molecular weight heparin or prophylaxis with low molecular weight heparin (dosage details not reported) from the first dose of pegylated asparaginase (day 30); duration was at the discretion of the treating physician. The median follow‐up was 4.3 years (interquartile range 2.5 to 6.4 years).
-
Sibai 2020 included 224 adults (of note, ≥17 years of age) with Philadelphia chromosome‐negative ALL in complete remission treated according to a modified version the Dana Farber Cancer Institute (DFCI) 91‐01 protocol. Participants received either no prophylaxis with low molecular weight heparin or weight‐adjusted prophylaxis with low molecular weight heparin (enoxaparin subcutaneously once daily; mean dose administered 0.79 mg/kg, range 0.39 to 1.2) from day 1 in cycle 1 (post‐induction) of intensification chemotherapy until the completion of the intensification phase (< 60 years: 30 weeks and ≥ 60 years: 21 weeks). Follow‐up time was not reported.
-
Umakanthan 2016 evaluated 17 adults with ALL treated with an asparaginase‐containing chemotherapy regimen. Participants either received no prophylaxis with low molecular weight heparin or prophylaxis with low molecular weight heparin (enoxaparin 40 mg subcutaneously) once daily up to three weeks following the pegylated asparaginase administration during induction treatment; duration of enoxaparin prophylaxis was 21 days in 66.7% and seven days in 33.3%. Follow‐up time was not reported.
Six non‐randomised follow‐up studies assessed prophylactic antithrombin concentrates versus no prophylactic antithrombin concentrates.
-
Bigliardi 2015 included 13 adults with newly diagnosed Philadelphia chromosome‐negative ALL treated with induction chemotherapy according to three different regimens including either Erwiniachrysanthemi‐derived (n = 11) or Escherichiacoli‐derived (n = 2) asparaginase. Participants received prophylactic antithrombin concentrates (50 IU/kg intravenously, if antithrombin activity level <70%); the frequency of antithrombin activity monitoring was at the discretion of the treating physician. Participants who did not receive any antithrombin concentrates had antithrombin activity levels ≥70%. The median follow‐up was 31 months (range 6 to 65 months).
-
Chen 2019 included 75 adults with ALL treated with different asparaginase‐containing ALL regimens receiving at least one dose of pegylated asparaginase during any phase of treatment (i.e. induction, consolidation, maintenance, salvage). Prophylactic antithrombin concentrates (Thrombate III when antithrombin activity levels < 60 %; dose IV (units) = [desired‐baseline antithrombin levels x body weight (kg)] / 1.4) were given during pegylated asparaginase therapy. The frequency of antithrombin activity monitoring was at the discretion of the treating physician. Median antithrombin dose was 3564 units (range 1755 to 7423), the median number of antithrombin doses received was two (range zer to 12). Follow‐up time was not reported.
-
Elliott 2004 included 54 adults (of note, ≥ 17 years of age) with newly diagnosed Philadelphia chromosome‐negative and ‐positive ALL treated with native Escherichia coli‐derived asparaginase‐containing induction chemotherapy according to two different regimens. Prophylactic antithrombin concentrates (Thrombate III intravenously (dose not reported), if antithrombin activity level < 70%); the frequency of antithrombin activity monitoring was at the discretion of the treating physician. It is unclear whether participants who did not receive any antithrombin concentrates had antithrombin activity levels ≥ 70%. Follow‐up time was not reported.
-
Farrell 2016 evaluated 40 adults diagnosed with Philadelphia chromosome‐negative and ‐positive ALL and treated according to different asparaginase‐containing regimens (UK Acute Lymphoblastic Leukaemia (UKALL) XII, UKALL2003, German Multicentre Acute Lymphoblastic Leukaemia (GMALL), or UKALL14). Participants received either no antithrombin supplementation or antithrombin supplementation, if antithrombin levels < 70 IU/dL (Kybernin P, IV dose (IU) = [(120‐pre‐level) x patient weight kg] / 1.4) during phase I induction. Antithrombin levels were monitored three times weekly during and after the period of asparaginase therapy. The total antithrombin doses per treated participant during induction ranged from 3000 to 54,000 IU (median 75,000 IU). Participants received a median of 2.5 antithrombin doses (range 1 to 11). Follow‐up time was not reported.
-
George 2020 included 61 adults with ALL treated according different adult/paediatric asparaginase‐containing regimens. Participants received either no antithrombin activity monitoring or antithrombin activity monitoring thrice weekly and supplementation, if antithrombin levels < 50%, according to weight‐based dosing (3000 units (< 70 kg), 4000 units (70 kg to 100 kg), and 5000 units (>100 kg) with a target antithrombin activity level of 120%). Participants received a median number of two antithrombin doses (range 0‐5). Follow‐up time was not reported.
-
Grose 2018 enrolled 46 adults with either ALL (n = 43), lymphoblastic lymphoma or NK‐/T‐cell lymphoma treated according to various asparaginase‐based regimens (Berlin‐Frankfurt‐Münster (BFM)/hyperCVAD/Linker/Larson/SMILE) including either Escherichia coli‐derived L‐asparaginase, pegaspargase, or Erwinia chrysanthemi‐derived L‐asparaginase. Controls received no antithrombin monitoring or supplementation. Participants in the intervention group received antithrombin monitoring and supplementation, if antithrombin levels < 60%; antithrombin levels were monitored post‐asparaginase administration. Follow‐up time was not reported.
Excluded studies
See Characteristics of excluded studies.
We excluded a total of 22 studies merged from 28 reports for the following reasons: no ALL diagnosis (Bern 1990; Karthaus 2005; Lavau‐Denes 2013; Mismetti 2003; Monreal 1996; Ratcliffe 1999; Young 2009), or no information available regarding inclusion of participants with ALL (Couban 2005 / author correspondence), paediatric population (Cohen 2010), no asparaginase treatment during intervention (Chojnowski 2002; Verso 2005), no intervention (Baile 2017; Bakalov 2020; Kroll 2013; Napolitano 2014; Peterson 2013), no control group (Boban 2019; Barreto 2017; Freyer 2020; Magagnoli 2005; Mazzucconi 1994), and study protocol (Klaassen 2017; study ended, no future report will be published / author correspondence).
Risk of bias in included studies
We assessed the risk of bias for 13 of 23 included studies (for details see Included studies) using ROBINS‐I (version 19, September 2016; Sterne 2016), given that all 13 studies were non‐RCTs; the assessments are summarised at outcome level for each study in Table 4; Table 5; Table 6; Table 7; Table 8; Table 9; Table 10; Table 11; Table 12; Table 13; Table 14; Table 15; Table 16, and summarised for only available primary outcomes of harms in Figure 2; Figure 3, as well as available secondary outcomes of harms in Figure 4, respectively, according to the listed method in Assessment of risk of bias in included studies. Of note, none of the included studies provided outcome data for clinically relevant non‐major bleeding, heparin‐induced thrombocytopenia (studies of heparins), or quality of life.
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for. Conference abstract—no information reported about baseline characteristics or presence of confounders such as age, sex, type of ALL regimen including intrathecal chemotherapy and steroid use, risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, oral contraceptives, inherited thrombophilia traits, smoking, and obesity), co‐morbidities, and co‐interventions (antithrombin, fresh frozen plasma, cryoprecipitate, fibrinogen, platelets). The distribution of central venous catheters between the two groups was not reported. Quote: "All thromboses in the prophylaxis group were associated with central venous catheters.." No method for dealing with potential confounders. Sparse information regarding the control group, quote: "Fifteen patients receiving similar induction protocols who did not receive any prophylactic anticoagulation were used as the control group." No information reported of the time period; unknown if a substantial change in supportive care or awareness/detection of venous thromboembolism could have been present. However, the study authors report that all participants received identical asparaginase treatment (type and dose). |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design; only a conference abstract was available. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that start of follow‐up and start of intervention coincided for all participants. Quote: "We implemented A fixed dose (1 mg/kg/day) of LMWH prophylaxis for adults with ALL undergoing induction therapy with L‐asparaginaseat our institution in 2012. In this study we report our outcomes with this protocol." Selection of the intervention group, quote: "Retrospective review of twenty‐three patients who received prophylactic anticoagulation with the LMWHenoxaparin(1mg/kg/day) while undergoing induction therapy with L‐asparaginasefor ALL." Limited information on selection of the control group, quote: "Fifteen patients receiving similar induction protocols who did not receive any prophylactic anticoagulation were used as the control group." The historical cohort may confound the results. |
| Bias in classification of interventions | All outcomes | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended interventions | All outcomes | Low risk | It appears that all participants in the intervention group received the intended thromboprophylaxis; however, only a conference abstract was available. |
| Bias due to missing data | All outcomes | No information | All data appeared to be reported; however, only a conference abstract was available. |
| Bias in measurement of outcomes | Venous thromboembolism‐related mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
| Major bleeding | Moderate risk | Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). | |
| Bias in selection of the reported result | All outcomes | Moderate risk | Only a conference abstract was available. No pre‐registered protocol or statistical analysis plan were available; however, the reported outcomes as described in the methods section were consistent with an a priori plan. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for. Conference abstract—no method for dealing with potential confounders presented. Quote: "There were no differences between the 2 groups (before and after protocol) concerning age, gender, ALL phenotype, risk,hyperleukocytosis, LDH level, SNC involvement, coagulation at diagnosis and treatment response." Statistical method for comparison and raw data not presented. Participants were treated according to either HOVON 100 (n = 22) or CLCG (n = 4) chemotherapy protocols including L‐asparaginase (type and dosing not presented); intrathecal chemotherapy and steroid use not mentioned. No information reported about presence of risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, smoking, and obesity) or co‐morbidities. Unclear if participants receiving prophylaxis with low molecular weight heparin were at higher risk of venous thromboembolism compared with controls. Participants received antithrombin, fibrinogen, and plasma concentrates; number of participants who received co‐interventions and distribution between the low molecular weight heparin prophylaxis and no low molecular weight heparin prophylaxis group not presented. Further, the study authors showed no attempt trying to disentangle the effect of co‐interventions from that of low molecular weight heparin prophylaxis, as this was not the primary study aim. No information reported on follow‐up that potentially included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design; only a conference abstract was available. Evaluation of low molecular weight heparin prophylaxis was not a primary aim of the study, thus we assume that participants were selected into the study regardless of low molecular weight heparin prophylaxis data. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appears that start of follow‐up and start of intervention coincide for all participants. It appears that all potentially eligible participants for the specified study period were included. Quote: "In 2016 our group adapted the protocol described byBiddeciet al. for prophylactic replacement therapy with Fibrinogen (FI) Concentrate and Antithrombin (AT) Concentrate." Quote: "From 2014, a total of 26 patients (15 males, 11 females; median age 35 years, 22Bprecursor‐ALL, 4 T‐ALL) were treated with intensive chemotherapy (...) 10 of the 26 patients were enrolled after the protocol implementation." Quote: "All patients but 1 started prophylactic LMWH during L‐Asp therapy in the protocol group but only 7 (44%) were given LMWH before 2016." |
| Bias in classification of interventions | All‐cause mortality | Moderate risk | Intervention was not defined in the methods section and unclearly defined in the results section, as evaluation of thromboprophylaxis with low molecular weight heparin was not the primary aim of the study. Thus, assignments of intervention status were determined retrospectively. |
|
| Serious risk | ||
| Bias due to deviations from intended intervention | All outcomes | No information | No information reported on deviations from the prophylaxis with low molecular weight heparin, as this was not the primary study aim. |
| Bias due to missing data | All outcomes | No information | All data of interest were not reported, as evaluation of thromboprophylaxis with low molecular weight heparin was not the primary study outcome, and only a conference abstract was available. Selected relevant outcome data have been obtained by author correspondence. |
| Bias in measurement of outcomes |
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, none of the outcomes were subjective. |
|
| Moderate risk | Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). Imaging confirmation not described. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). | |
| Bias in selection of the reported result | All outcomes | No information | There is too little information to make judgement, as only an abstract was available. No pre‐registered protocol or statistical analysis plan were available. Of note, evaluation of prophylaxis with low molecular weight heparin was not the primary study aim. Outcome data have been obtained by study author correspondence. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for, and no method for dealing with potential confounders was presented. Baseline characteristics presented and were similar between groups regarding age, sex, white blood cell count at ALL diagnosis, immunophenotype, mediastinal mass, performance status 0‐2 (a proxy for immobilisation; unknown in n = 5), and body mass index. Twenty‐two per cent (11/49) were obese (body mass index ≥ 30) in the control group, and 27% (10/36) were obese in the intervention group. Eighteen percent had performance status 2 (9/49) in the control group and 6% (2/36) in the intervention group. Only a higher median platelet count at day 32 in the intervention group (320 x 103/L) differed significantly from that in the control group (224 x 103/L). Additional confounders such as prior venous thromboembolism, infections, central venous catheters, oral contraceptives, inherited thrombophilia traits, or smoking) and co‐morbidities were not assessed for the two groups. All participants were treated according to a paediatric‐inspired ALL protocol; however, the type, route of administration and dose of asparaginase was changed in August 2011 as part of an amendment due to a high number of safety events. Intrathecal chemotherapy and steroid use were not mentioned. Participants treated prior to the amendment (control group) received pegylated apsaraginase (2500 IU/m2 intravenously) during induction and 2500 IU/m2 every two weeks during consolidation, whereas participants treated after the amendment (intervention group) received one dose of native Escherichia coli‐derived asparaginase (25,000 IU/m2 intramuscularly) during induction followed by pegylated asparaginase (2000 IU/m2 intravenously) every three weeks during consolidation. Only participants treated after the amendment received heparin prophylaxis. Co‐interventions reported only for the intervention group and included antithrombin replacement (if antithrombin activity levels <30%; n=1), cryoprecipitate (if fibrinogen < 50mg/dL; n = 4), and platelet transfusions (if <50x109/L ; 24/33 during induction and 11/23 during consolidation). Quote: "Although a direct comparison of VTE rates before and after instituting the use of prophylactic anticoagulation is limited by the change in the schedule, dosing, and type ofasparaginaseadministered during induction, the use ofcryoprecipitate, FFP and antithrombin replacement was minimal, suggesting that the observed decrease in VTE incidence is due to the use of prophylactic anticoagulation and not coagulation factor supplementation." Quote: "Comparison of the rate of thrombosis before and after the implementation of guidelines for prophylactic anticoagulation is also limited by the modifications that were made simultaneously to improve the safety of the study." The type, dose, and timing of heparin prophylaxis varied. Quote: "Although 9 thrombotic events occurred while patients were on prophylactic anticoagulation, there was a significant variability in dose, timing, and likely compliance with daily administration of prophylaxis in the outpatient setting making it hard to determine if prophylaxis failed to prevent the development of VTE." The follow‐up time included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Mixed prospective/retrospective design. Quote: "Between September 2007 and June 2013, 110 eligible adult patients with newly diagnosed ALL were enrolled on a high riskpediatricinspired treatment regimen through the DFCI ALL Consortium.." Quote: "The data safety and monitoring committee put the study on hold between September 2010 and August 2011 due to a high number of safety events related to IV PEG‐asparaginase, including VTE. (...) and mandated that patients with a Philadelphia‐chromosome positive (Ph+)leukemiabe removed from this study." Quote: "Of the 110 patients, 22 patients with a Ph+leukemiaand 3 patients not prescribed prophylactic anticoagulation (2 induction failures and 1 who was transplanted directly after induction) were excluded from analysis; 49 patients treated prior to the amendment and 36 treated after the amendment wereanalyzed." Selection into the study was based on participants' characteristics observed after the start of the study (Philadelphia chromosome status); Philadelphia chromosome status may be related to the risk of venous thromboembolism. It appears that start of follow‐up and start of intervention coincide for all participants. |
| Bias in classification of interventions | All outcomes | Moderate risk | The recommended intervention was clearly defined (thromboprophylaxis was recommended for all participants post‐amendment); however Quote: "While these guidelines offered management suggestions for the use of prophylactic anticoagulation, use was not mandated; failure to use prophylactic anticoagulation or to assess coagulation factor levels was not a protocol violation." Quote: "The decision to use prophylactic anticoagulation was left to the treating physician." Some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | All outcomes | Low risk | The reported deviations from intended intervention are what would be expected in usual practice of ALL treatment. 3/36 participants in the intervention group did not receive prophylactic treatment during induction due to ALL treatment‐related toxicities including severe thrombocytopenia, neutropenia, and infection. 12/36 patients only received venous thromboembolism prophylaxis during induction and not in consolidation due to development of venous thromboembolism, haematopoietic stem cell transplantation in first complete remission immediately after induction, induction failure, relapse, or other patient‐specific factors including thrombocytopenia. Thromboprophylaxis was held for procedures (n=24 during induction; n = 19 during consolidation), platelet count < 30x109/L (n = 12 during induction; n = 2 during consolidation), and bleeding symptoms (n = 1 during induction; n = 1 during consolidation), and at a patient's request (n = 1 during consolidation). |
| Bias due to missing data | All outcomes | Low risk | Data appear to be reasonably complete. Clarifications of selected relevant outcome data have been obtained by author correspondence. |
| Bias in measurement of outcomes |
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, none of the outcomes were subjective. |
|
| Moderate risk | Venous thromboembolism‐related symptoms are likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). Additionally, as a result of a multi‐centre study, differences in care and awareness may have influenced diagnostic suspicion. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available; however, the outcome measurements as described in the methods section and the analyses were consistent and in agreement with an a priori plan. Of note, the cumulative incidence of pulmonary embolism, a subgroup of serious venous thromboembolism, between the two groups was highlighted in the abstract and results section, although not being prespecified in the methods section. However, this addition did not change the overall direction of results. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | In a multivariate logistic regression model, the authors explored risk factors of venous thromboembolism at ALL diagnosis including older age, female sex, high body mass index, high platelet count. Baseline characteristics not presented between heparin prophylaxis and no‐heparin prophylaxis groups. Baseline confounders not controlled for in the intervention groups of interest apart from selected risk factors of venous thromboembolism, quote: "Although patients who received heparin prophylaxis also had a higher platelet count at diagnosis of ALL in comparison to those who did not (79 versus 59 x 109/L, p=0.05), the association between heparin prophylaxis and an increased risk of VTE persisted after adjustment for platelet counts at diagnosis (Supplementary Table 1). Especially, there was no difference between patients who received heparin prophylaxis or not regarding personal and familial thromboembolic events." Poor‐risk prognostic factors (ALL immunophenotype, high white blood cell count (>30 x109/L), central nervous system involvement, and poor prednisone response), prior venous thromboembolism, oral contraceptives before diagnosis, and a history of smoking as confounders were described for the total cohort but not analysed for the heparin versus no heparin prophylaxis groups. Infections, immobilisation, and co‐morbidities were not assessed for the two intervention groups of interest. Quote: "Hereditary thrombotic risk factors (protein S deficiency, protein C deficiency, activated protein C,prothrombin G20210A and factor V Leiden polymorphisms) and lupus anticoagulant wereseldomlyevaluated (42/110 patients, 38%, with thrombosis and 191/677, 28%, without thrombosis were tested) and their impact on thrombosis could not be evaluated." The authors highlight the presence of central venous lines as a potential confounder, quote: "This last parameter could not be evaluated in our study since all patients have central venous lines. Increased use of CVL, which is recommended for all patients in our protocol, might have caused the higher thrombotic rate in our study as we observed 50 DVT of the upper limb, for which CVL is arecognizedrisk factor." All participants were treated according to the same paediatric‐inspired ALL protocol including same dosing of intrathecal chemotherapy, steroids, and pegylated asparaginase (risk group distribution not reported for the heparin versus no heparin prophylaxis groups). Participants received antithrombin concentrates (if antithrombin activity levels≤60%; 87% of participants), fibrinogen or fresh frozen plasma (if fibrinogen <0.5g/L; 9% received fibrinogen concentrates), and platelets (if platelet count <20x109/L); the distribution between the heparin prophylaxis and no‐heparin prophylaxis groups not presented. Further, the study authors showed no attempt trying to disentangle the effect of co‐interventions from that of prophylaxis with unfractionated heparin/low molecular weight heparin. Quote: "Interestingly, patients who received heparin prophylaxis with no antithrombin prophylaxis (38 patients, 6%) were more likely to experience thrombosis (24%) in comparison to other patients (603 patients, 94%) who experienced 9.6% of VTE (p=0.02)." The authors highlight the type of heparin prophylaxis used as a potential confounder along with the potential important difference between participants who received heparin prophylaxis and participants who did not. The two groups might not be directly comparable. Quote: "Unfractionatedheparin (UFH) at 100 IU/kg/day in continuous infusion was recommended during induction; low molecular weight heparin (LMWH) at prophylactic doses in subcutaneous injection could be used as replacement at the physician’s discretion." Quote: "In our study, even though all patients were supposed to receive heparin prophylaxis, all did not receive such prophylaxis. We observe that it was more frequently administered to patients with less intensive thrombocytopenia. After adjustment for platelet counts at diagnosis, heparin prophylaxis was surprisingly associated with an increased risk of VTE. We have no clear explanation for this observation and the lack of details regarding the modality of heparin prophylaxis in our study (type, dose, therapeutic level monitoring) limits the generalization of this result. Patients deemed to be at increased bleeding risk, and thus at decreased thrombotic risk, might have received less heparin prophylaxis. The extensive use ofunfractionated heparin in our study might also explain the lack of efficacy of heparin prophylaxis that we observed. Given the severe antithrombin depletion induced byasparaginase,unfractionatedheparin is probably not the most effective anticoagulant in this setting, while some data advocate the use of low molecular weight heparin (LMWH) (3.5% of VTE versus 8% in those who receivedunfractionated heparin)." |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design (prospective registration of venous thromboembolic events). Of 813 people screened, 29 were excluded due to non‐eligibility criteria (n=11), loss to follow‐up (n=12), lack of toxicity data (n=3), and consent withdrawal (n = 3). We assumed that participants were selected into the study regardless of low molecular weight heparin prophylaxis data. Missing data on thromboprophylaxis in 122/784 participants; however, none were excluded on that account. Selection of participants into the study did not appear to be related to intervention, outcome, or prognostic factors. It appeared that start of follow‐up and start of prophylaxis with low molecular weight heparin coincided for all participants. |
| Bias in classification of interventions | All outcomes | Serious risk | Intervention was not clearly defined in the methods section. Quote: "Unfractionatedheparin (UFH) at 100 IU/kg/day in continuous infusion was recommended during induction; low molecular weight heparin (LMWH) at prophylactic doses in subcutaneous injection could be used as replacement at the physician’s discretion. Heparin prophylaxis was recommended from day 1 of induction until end of induction." Assignments of intervention status were determined retrospectively. |
| Bias due to deviations from intended intervention | All outcomes | Low risk | Potential deviations from the intended intervention, as described in the methods section, reflected what would be expected in usual practice: Quote: "When platelets were below 20 G/l, heparin prophylaxis was interrupted, and patients received a platelet transfusion. Heparin prophylaxis was resumed after platelet transfusion." |
| Bias due to missing data | All outcomes | Serious risk | Participants were excluded due to lack of outcome data. Quote: "Twenty‐nine patients were excluded due to non‐eligibility criteria (11 patients), lost to follow‐up (12 patients), lack of data regarding toxicity (3 patients), and consent withdrawal (3 patients)." Additional outcome data have been obtained by author correspondence. Missing intervention data in 122/784 participants. |
| Bias in measurement of outcomes | All outcomes | Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). Additionally, as a result of a multi‐centre study, differences in care and awareness may have influenced diagnostic suspicion. |
| Bias in selection of the reported result | All outcomes | Serious risk | No pre‐registered protocol or statistical analysis plan were available. Given the retrospective and exploratory design, several variables (risk factors of venous thromboembolism) and associations between intervention, co‐interventions, and outcomes were not prespecified in the methods section and described only in the results section. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | The study authors adjusted for age (adults aged 18 to 45 years compared with children), sex, body mass index, white blood cell count, immunophenotype, central nervous system status (leukaemic infiltration), presence of mediastinal mass, enlarged lymph nodes, palpable splenomegaly, induction corticosteroids, minimal residual disease at day 29, treatment group at day 29, and prophylaxis with low molecular weight heparin. Quote: "In a multiple Cox regression analysis with delayed entry at day 29 (N 5 1594; 114 TE events), the adjusted hazard ratio (HRa) of TE was significantly increased in ages 10.0 to 17.9 years (HRa, 4.9; 95% CI, 3.1‐7.8; P , .0001) and ages 18.0 to 45.9 years (HRa, 6.06; 95% CI, 3.65‐10.1; P , .0001) compared with children younger than 10.0 years, and for mediastinal mass at ALL diagnosis (HRa, 2.1; 95% CI, 1.0‐4.3; P 5.04; Table 3)." Noteworthy is that the analysis included participants aged 1‐45 years and compared people with versus without development of thromboembolism. Baseline characteristics and confounders such as age, sex, risk factors of venous thromboembolism (central venous catheters, infections, inherited thrombophilia traits, and obesity) were evaluated for participants with and without development of thromboembolism (these were not reported for the heparin prophylaxis and no‐heparin prophylaxis groups, as this was not part of the primary study aim). Unpublished data obtained by author correspondence: median age of 26.5 years (range 18.6‐43.0) in the low molecular weight heparin prophylaxis group and 26.0 years (range 18.0 to 45.8) in the no prophylaxis group; 41% male sex in the low molecular weight heparin prophylaxis group and 67% in the no prophylaxis group; 37% T‐cell immunophenotype in the low molecular weight heparin prophylaxis group and 30% in the no prophylaxis group. No information reported regarding immobilisation, oral contraceptives, smoking, co‐morbidities, or co‐interventions. None of the included participants had a history of thromboembolism. All participants were treated according to the same paediatric ALL protocol including the same dosing of pegylated asparaginase, steroids, and intrathecal chemotherapy. Given the multi‐centre design and lack of protocol guidelines for thromboprophylaxis, prophylaxis with low molecular weight heparin was only used at some centres; 41/233 adults received prophylaxis. The follow‐up time included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All‐cause mortality | Moderate risk | Retrospective design (prospective registration of venous thromboembolic events). Given that evaluation of prophylaxis with low molecular weight heparin was not a primary study aim, participants were selected into the study regardless of low molecular weight heparin prophylaxis data. Consecutive participants with newly diagnosed ALL treated according to the NOPHO ALL2008 protocol were included. Excluded participants accounted for in detail (participants with registered thromboembolism in the NOPHO registry were excluded because of missing imaging confirmation of TE (n=1) and missing data (n = 8)). Quote: "Of 1861 patients with ALL, the following were excluded: 21 patients with acuteleukemiaof ambiguous lineage, 54 with ALL predisposition syndromes (eg, Down syndrome or ataxiatelangiectasia), and 1 not treated according to the ALL2008 protocol. One hundred fifty patients with ALL developed TE, of whom 13 were excluded because of missing imaging confirmation of TE (N = 1), superficial thrombophlebitis (N = 1), septic emboli (N = 1), central venous line (CVL) dysfunction registered as asymptomatic TE (N = 2), and missing data (N = 8). Thus, a total of 1772 patients with Ph‐ ALL, among whom were 137 registered TE cases, were included in the study." Start of follow‐up and start of prophylaxis with low molecular weight heparin coincided for all participants. |
|
| Critical risk | ||
| Bias in classification of interventions | All‐cause mortality | Moderate risk | Intervention was not clearly defined in the methods or results section, as evaluation of thromboprophylaxis with low molecular weight heparin was not the primary study aim. Methods section, quote: "No common recommendations for routineantithromboticprophylaxis exist in the ALL2008 protocol." Assignments of intervention status were determined retrospectively. |
|
| Serious risk | ||
| Bias due to deviations from intended intervention | All outcomes | No information | No information reported on deviations from the prophylaxis with low molecular weight heparin, as this was not the primary study aim. |
| Bias due to missing data | All outcomes | Serious risk | All data of interest are not reported, as evaluation of prophylaxis with low molecular weight heparin was not the primary study aim. Selected relevant outcome data have been obtained by author correspondence. |
| Bias in measurement of outcomes | All‐cause mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
| First‐time symptomatic venous thromboembolism | Moderate risk | Venous thromboembolism‐related symptoms were both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. As a result of a multi‐centre study, differences in care and awareness may have influenced diagnostic suspicion. | |
| Asymptomatic venous thromboembolism | Serious risk | Additionally, as a result of a multi‐centre study, differences in care and awareness may have influenced the incidence of asymptomatic venous thromboembolic events; no screening was done. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available. However, the outcome measurements as described in the methods section and the analyses were consistent and in agreement with an a priori plan. Of note, evaluation of prophylaxis with low molecular weight heparin was not the primary study aim. |
| Overall bias | All‐cause mortality | Critical risk | Based on the critical risk of bias due to confounding. |
|
| Critical risk | Based on the critical risk of bias due to confounding and in selection of participants into the study. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | First‐time symptomatic venous thromboembolism | Serious risk | Baseline confounders not controlled for, and no method dealing with potential confounders presented. Baseline characteristics between the prophylaxis with low molecular weight heparin and control groups were similar in regard to median age and weight as well as sex distribution. All participants were treated according to a paediatric‐inspired ALL regimen. However, 123/125 participants in the low molecular weight heparin prophylaxis group received native Escherichia coli‐derived asparaginase and 3/125 received pegylated asparaginase. Type of asparaginase formulation was not reported for the historical control group. Additionally, participants aged <60 years received asparaginase 12,500 IU/m2 intramuscularly once weekly for 30 weeks of intensification treatment, whereas participants aged ≥60 years received a shortened intensification (21 weeks) with asparaginase dose reduction (6000 IU/m2 intramuscularly once weekly). Accordingly, dexamethasone was also dose‐reduced from 9 mg/m2 to 6 mg, twice daily from day 1‐5. Intrathecal chemotherapy was identical. No information reported about potential confounders such as presence of risk factors of venous thromboembolism (infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, obesity (body mass index), and smoking) or co‐interventions. Participants with prior venous thromboembolism were excluded. The study authors paid attention to the challenge of heparin dosing in relation to co‐morbidities and weight using weight‐adjusted doses and excluding participants with renal insufficiency. Quote: "Patients with baseline renal impairment (creatinine clearance <30 ml/min) were excluded in this study." The study was conducted from 2001‐2018; including a historical control group and an intervention group after implementation of thromboprophylaxis as the standard of care (in 2009 according to the associated former publications). However, the reporting of time periods in relation to cases and controls was inconsistent throughout the paper: Quote: "Patients from the historical cohort, treated between 2001 and 2010, did not receive any anticoagulation prophylaxis. In contrast, patients treated between 2011 and 2017 received anticoagulation prophylaxis..." Quote: "...and (ii) increased monitoring and awareness of VTE in the prophylaxis group (2011–2018) than in the control group (2001–2009)." Nonetheless, quality in supportive care and awareness/detection of venous thromboembolism were likely to have changed over this period of time. |
| Bias in selection of participants into the study | First‐time symptomatic venous thromboembolism | Moderate risk | Retrospective design. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that start of follow‐up and start of intervention coincide for all participants. Limited information on selection of the controls, quote: "Patients from the historical cohort, treated between 2001 Quote: "Patients from the historical cohort, treated between 2001 and 2010, did not receive any anticoagulation prophylaxis. In contrast, patients treated between 2011 and 2017 received anticoagulation prophylaxis..." versus Quote: "...and (ii) increased monitoring and awareness of VTE in the prophylaxis group (2011–2018) than in the control group (2001–2009)." Exclusion of participants were accounted for. Quote: "Patients were excluded from this study for the following reasons: (i) received full‐dose LMWH anticoagulation therapy at the beginning of intensification due to prior thrombosis, (ii) underwent allogeneic stem cell transplant during intensification, (iii) relapsed before the seventh cycle of intensification, (iv) did not complete ≥7 cycles of intensification due to other patient factors such as liver toxicity, non‐haemorrhagic death, or loss to follow‐up." Additionally, people with Philadelphia chromosome‐positive ALL were excluded, as they did not receive asparaginase treatment. |
| Bias in classification of interventions | First‐time symptomatic venous thromboembolism | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | First‐time symptomatic venous thromboembolism | Low risk | Deviations from the intended intervention was in accordance with usual practice. Quote: "Among those who did not develop VTE in the prophylaxis group, seven patients had PLT <50 x 109/l and five had PLT <30 x 109/l. Comparatively, among the 17 patients who experienced VTE, four had PLT <50 x 109/l and one patient had PLT <30 x 109/l. Prophylaxis was temporarily held if PLT |
| Bias due to missing data | First‐time symptomatic venous thromboembolism | Low risk | Data appeared to be reasonably complete. |
| Bias in measurement of outcomes | First‐time symptomatic venous thromboembolism | Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. Quote: "The mechanism of this difference is unclear, but may be explained by: (i) a potential increase in antithrombin levels as intensification treatment progresses, and (ii) increased monitoring and awareness of VTE in the prophylaxis group (2011–2018) than in the control group (2001–2009)." |
| Bias in selection of the reported result | First‐time symptomatic venous thromboembolism | Serious risk | No pre‐registered protocol or statistical analysis plan was available. The main outcome measurements were described in the methods section and reported in the results section in agreement with an a priori plan. However, potential predictors, hence risk factors, of venous thromboembolism including ALL immunophenotype and body weight ≤ vs > 80 kg were mentioned only in the results section and abstract; the former yielding statistical significance. |
| Overall bias | First‐time symptomatic venous thromboembolism | Serious risk | Based on the serious risk of bias due to confounding and in selection of the reported result. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | First‐time symptomatic venous thromboembolism | Critical risk | Baseline confounders not controlled for. Conference abstract—no information reported about baseline characteristics between the enoxaparin prophylaxis and the no‐prophylaxis groups or presence of confounders such as age, sex, risk factors of venous thromboembolism (infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, smoking, and obesity) or co‐morbidities. No method for dealing with potential confounders. No participant had prior venous thromboembolism. Participants (41%) received cryoprecipitate (if fibrinogen <100mg/dL); unknown distribution between the two groups—disentangling this effect from that of enoxaparin prohylaxis was not considered. Additionally, different ALL protocols including different type and dosing of asparaginase were used, quote: "88% percent receivedpediatric‐typechemoregimen andintramuscularpeg‐asparaginase; the most common doses ofpeg‐asparaginase included2000 mg/m2 (41%) and 2500 mg/m2 (35%)." Intrathecal chemotherapy and steroid use highly likely varied too (not reported). Sparse information regarding the control group, quote: "This is asingle‐centerstudy of 17 consecutive adult patients (>18 years old) treated for ALL withpeg‐asparaginasecontaininginduction regimen before and after the establishment of institutional policy touse enoxaparin prophylaxis." No information reported on the time period; however, unknown if a substantial change in supportive care or awareness/detection of venous thromboembolism could have been present even though only 17 consecutive people were included—indicating a rather short period of time. |
| Bias in selection of participants into the study | First‐time symptomatic venous thromboembolism | Moderate risk | Retrospective design; only a conference abstract was available. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that start of follow‐up and start of intervention coincided for all participants. Quote: "This is a single‐centerstudy of 17 consecutive adult patients (>18 years old) treated for ALL with peg‐asparaginase containinginduction regimen before and after the establishment of institutional policy touse enoxaparin prophylaxis. Patients were identified from hospital research database." It appeared that all potentially eligible participants for the specified study years have been included. |
| Bias in classification of interventions | First‐time symptomatic venous thromboembolism | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | First‐time symptomatic venous thromboembolism | Low risk | It appeared that all participants in the intervention group received the intended thromboprophylaxis; however, only a conference abstract was available. |
| Bias due to missing data | First‐time symptomatic venous thromboembolism | No information | All data appear to be reported; however, only a conference abstract was available. Clarifications of selected data have been obtained from study author correspondence. |
| Bias in measurement of outcomes | First‐time symptomatic venous thromboembolism | Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). Imaging confirmation not described. Quote: "Medical records were reviewed for the occurrence of any clinically relevant bleeding and VTE within 3 months of receiving peg‐asparaginase." |
| Bias in selection of the reported result | First‐time symptomatic venous thromboembolism | Moderate risk | Only a conference abstract was available. No pre‐registered protocol or statistical analysis plan were available; however, the reported outcomes as described in the methods section were consistent with an a priori plan. |
| Overall bias | First‐time symptomatic venous thromboembolism | Critical risk | Based on the critical risk of bias due to confounding. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors's judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding |
| Critical risk | Given that evaluation of prophylactic antithrombin concentrates for prevention of venous thromboembolism was not a study aim, baseline confounders not controlled for, and no method for dealing with potential confounders was presented. Baseline characteristics presented and were similar between groups regarding age (median age of 32.5 years (range 20 to 58) in the antithrombin group and 32 years (range 21 to 57) in the no‐antithrombin group) but with unequal sex distribution (m/f: 2/2 in the antithrombin group and 8/1 in the no‐antithrombin group). No apparent difference in ALL immunophenotype between the groups was present (B‐/T‐cell ALL: 2/2 in the antithrombin group and 4/5 in the no‐antithrombin group), and all included participants had Philadelphia chromosome‐negative ALL. The distribution of peripherally central venous catheter was uneven between the groups (2/4 in the antithrombin group and 6/9 in the no‐antithrombin group). Additional confounders such as prior venous thromboembolism, infections, oral contraceptives, inherited thrombophilia traits, or smoking) and co‐morbidities were not reported for the two groups. The participants received asparaginase‐based induction chemotherapy according to three different ALL regimens including either Erwinia chrysanthemi‐derived (n = 11) or native Escherichia coli‐derived (n=2) L‐asparaginase as well as different steroid dosing. Two of the four participants in the antithrombin group received Escherichia coli‐derived asparaginase. No information on intrathecal chemotherapy. The study authors note the highly heterogeneous cohort limiting a comparison‐‐quote: "Even if we cannot make any direct comparison between Erwinia and E. coliasparginasesin terms of efficacy and toxicity profiles from our series, we could suggest that remission induction chemotherapy regimens with Erwinia asparaginase in first line may be feasible in adult patients." Importantly, the allocation to prophylactic treatment with antithrombin concentrates was dependent on the antithrombin activity levels, as an antithrombin level < 70% qualified for treatment with antithrombin concentrates. Hence, the no‐antithrombin group solely comprises participants with antithrombin levels > 70% and cannot be defined as a valid comparator group. Furthermore, the co‐interventions including fibrinogen and fresh frozen plasma were unequally administrated between the groups based on fibrinogen levels and hepatic insufficiency, respectively. Quote: "The antithrombin (AT) level was < 70% (median AT level 100%, range 62 – 131%) in two out of 11 patients (18.2%) receiving Erwinia asparaginase and in two of two cases (100%) receiving E. coli asparaginase, whereas the fibrinogen level was < 100 mg/dL (median fibrinogen level 217 mg/dL, range 70 – 887 mg/dL) in eight out of 11 cases (72.7%) receiving Erwinia asparaginase, and was invariably > 100 mg/dL in the two patients treated with E. coli asparaginase." Quote: "Prophylactic AT(50 IU/kg) and fibrinogen (20 mg/kg) concentrate supplementations were administered for an AT level < 70% and fibrinogen level < 80 mg/dL, respectively." Quote: "Two of the 11 patients (18.2%) treated with Erwinia asparaginase, namely patients 3 and 10, received prophylactic AT and fibrinogen concentrates, whereas in the two patients receiving E. coli asparaginase, only AT concentrate supplementation was provided." Quote: "Of note, fresh frozen plasma (15 mL/kg) was also infused to patient 10 on day + 26 due to hepatic insufficiency." The study authors showed no attempt trying to disentangle the effect of co‐interventions from that of antithrombin concentrates, as this was not the primary study aim. The study retrospectively evaluated people treated between 2006‐2013. During a time period of seven years a substantial change in supportive care or venous thromboembolism awareness/detection could have been present. The follow‐up time included the time with administration of therapeutic anticoagulation for participants with development of venous thromboembolism, potentially influencing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Given that evaluation of prophylaxis with antithrombin concentrates was not a primary study aim, we assume that participants were selected into the study regardless of prophylaxis data. It appeared that all potentially eligible participants for the specified study period were included. Quote: "We retrospectivelyanalyzed48 adultsaffectedwith ALL, consecutively observed at our institution between January 2006 and December 2013." Selection into the study was based on retrospective collection of information regarding participants' characteristics (Philadelphia chromosome status); Philadelphia chromosome status may be related to the risk of venous thromboembolism. Yet, the study authors argue that asparaginase often is excluded from the Philadelphia chromosome‐positive ALL treatment. Quote: "The patients with Philadelphia‐positive ALL received remission induction regimens based on tyrosine kinase inhibitors combined Excluded participants accounted for in detail. Quote: "Thirty‐three patients did not receive asparaginase because of either Philadelphia‐positivity (21 patients) or advanced age (12 patients aged > 60 years), while 13 patients of median age 32 years (range 20 – 58 years) received either Erwinia asparaginase (11 patients) or E. coli asparaginase (two patients) during remission induction treatment, according to three different regimens (Table I)." It appeared that start of follow‐up and start of intervention coincided for all participants. |
| Bias in classification of interventions
|
| Moderate risk | Intervention was not defined in the methods paragraph (format of study reference: letter), as evaluation of thromboprophylaxis with antithrombin concentrates was not the primary aim of the study. However, preset criteria for antithrombin substitution as part of clinical practice seemed to be present at the start of intervention: Quote: "Prophylactic AT(50 IU/kg) and fibrinogen (20 mg/kg) concentrate supplementations were administered for an AT level < 70% and fibrinogen level < 80 mg/dL, respectively." Assignments of intervention status were determined retrospectively. |
|
| Serious risk | ||
| Bias due to deviations from intended intervention | All outcomes | No information | No information reported on deviations from the prophylaxis with antithrombin concentrates, as this was not the primary study aim. |
| Bias due to missing data | All outcomes | Low risk | Data appear to be reasonably complete. Clarifications of selected relevant outcome data have been obtained by author correspondence. |
| Bias in measurement of outcomes
|
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
|
| Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). Systematic imaging confirmation not described. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available (study reference format: letter); however, the reported outcomes seemed in agreement with the scope (safety of asparaginase therapy) as outlined in the methods paragraph consistent with an a priori plan. Of note, evaluation of prophylactic antithrombin concentrates was not the primary study aim. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding, as the control group is not a valid comparator group. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for, and no method for dealing with potential confounders presented. Baseline characteristics shown and were similar between antithrombin and no‐antithrombin groups in regard to age; 10% more male sex in the no‐antithrombin group (reported as non‐significant). Potential confounders such as risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, central venous catheters, inherited thrombophilia traits, smoking, and obesity) and co‐morbidities not reported. A similar proportion of people between antithrombin and no‐antithrombin groups had hormonal therapy use. Participants received concomitant enoxaparin prophylaxis, fresh frozen plasma, and cryoprecipitate supplementation. Quote: "Cryoprecipitateuse was significantly higher in the AT group (55%) than the control group (11%) (p < .001). Fresh frozen plasma use was not significantly different between the two groups." A total of 26% (12/47) participants in the antithrombin group and 7%(2/28) participants in the no‐antithrombin group received enoxaparin prophylaxis. No adjustments for these co‐interventions were done. Quote: "Univariate analysis showed that the use of prophylactic anticoagulation,cryoprecipitate, or fresh frozen plasma did not have a significant effect on the incidence of VTE." Quote: "Because there is nostandardizedpractice for AT or fibrinogen replacement in these patients, our data implies that physicians who monitor AT levels for replacement were also more likely to order fibrinogen levels and prescribecryoprecipitatefor low fibrinogen levels (generally <100 mg/dL)." Study authors recognised cryoprecipitate as a potential confounder, quote: "This study did not find a benefit in bleeding risk withcryoprecipitateuse, and the higher frequency ofcryoprecipitateuse in the AT group may have attenuated any beneficial effects from AT supplementation. This may explain why a larger proportion of thrombotic events were observed in the AT group although this difference was not significant." Participants in the antithrombin and the no‐antithrombin groups were treated according various ALL regimens, all containing pegylated asparaginase at different doses. Quote: "Patients in the AT group received a median of two doses of PEG‐Asp compared to three doses in the control group (p = .04). The dose of PEG‐Asp in the AT group was also marginally lower, although this difference was not significant." Intrathecal chemotherapy and steroid use highly likely varied too (not reported); the latter acknowledged by the study authors, quote: "Another limitation is that the specific corticosteroid (i.e. dexamethasone vs. prednisone) used by patients in each group was not accounted for." |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that start of follow‐up and start of intervention coincided for all participants. Quote: "...electronic medical records of 116 patients who received PEG‐Asp at City of Hope MedicalCenterbetween 27 April 2014 and 31 October 2017 were reviewed. Inclusion criteria were of age 18 years, a diagnosis of ALL or T‐celllymphoblasticlymphoma warranting treatment with a PEG‐Asp‐containing regimen, and receipt of at least one dose of PEG‐Asp during any phase of their ALL regimen (i.e. induction, consolidation, maintenance, salvage). Patients were assigned to the AT group if a physician monitored AT levels with the intention to replace if levels were low (i.e. <60%). All other patients were assigned to the control group." It appeared that all potentially eligible participants for the specified study years were included. Selection into groups was based on the intention‐to‐treat principle. |
| Bias in classification of interventions | All outcomes | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | All outcomes | Moderate risk | The study authors report deviations from the intended intervention, for which they have no explanation. Quote: "Of the remaining 10 patients in the AT group who did not receive AT, none of them experienced a VTE event, and 6 of them maintained an AT level above 60%. For unknown reasons, three of these patients did not receive AT when levels were below 60%, and one patient did not have AT levels drawn." |
| Bias due to missing data | All outcomes | Low risk | Data appeared to be reasonably complete. Clarifications of outcome data obtained from study author correspondence. |
| Bias in measurement of outcomes | Venous thromboembolism‐related mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
| First‐time symptomatic venous thromboembolism | Moderate risk | Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available; however, the outcome measurements as described in detail in the methods section and the analyses were consistent and in agreement with an a priori plan. Analysis was presented by intention‐to‐treat in the results and abstract sections (10 participants in the intervention group were intended to receive AT supplementation but did not; none of them developed venous thromboembolism). No results (raw numbers) were reported in regard to the univariate analysis evaluating the impact of pre‐selected variables on the incidence of venous thromboembolism, quote: "Univariate analysis showed that the use of prophylactic anticoagulation,cryoprecipitate, or fresh frozen plasma did not have a significant effect on the incidence of VTE." |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding |
| Critical risk | Given that evaluation of prophylactic antithrombin concentrates for prevention of venous thromboembolism was not the primary study aim, baseline characteristics were not presented for the participants receiving antithrombin concentrates and those who did not. Baseline characteristics presented for participants with and without development of venous thromboembolism, respectively. Baseline confounders not controlled for, and no method for dealing with potential confounders was presented. Most participants had central venous catheters, quote: "Surgically placed Hickman catheters provided central access for 98%." Additional confounders such as prior venous thromboembolism, infections, oral contraceptives, inherited thrombophilia traits, or smoking) and co‐morbidities were not reported for the antithrombin and no‐antithrombin groups. The participants with newly diagnosed Philadelphia chromosome‐positive or ‐negative ALL received native Escherichia coli‐derived asparaginase‐based induction chemotherapy according to two different ALL regimens including different asparaginase, steroid, and intrathecal chemotherapy dosages. Importantly, it is unclear whether antithrombin activity was monitored for all participants and whether all participants who had antithrombin activity levels <70%/≥ 70% did/did not receive antithrombin concentrates, respectively (and thus comprise the comparison group). Quote: "Monitoring of coagulation parameters (PT, APTT, Fb), AT activity (Chromogenix,Coamatic, Antithrombin Kit) and actor replacement was at the discretion of the treating physician." Quote: "Median baseline (prior to ASP) AT activity (normal: 80 – 120%) was 94% (77 – 151%). Median nadir AT during ASP therapy was 47% (30 – 90%), significantly different from baseline (P<0.0001)." Quote: "The practice of ‘‘prophylactic’’ AT replacement varied amongst physicians and AT was administered to 17 patients when AT activity fell to < 70%." Furthermore, the co‐interventions included fresh frozen plasma or cryoprecipitate (if fibrinogen < 100 mg/dL) and platelet transfusions (if <10 x109/L in the absence of fever (< 20 x109/L), bleeding or anticoagulation therapy (<30 x109/L)). The study authors showed no attempt trying to disentangle the effect of co‐interventions from that of antithrombin concentrates (but rather in regard to thrombosis development), as this was not the primary study aim. Quote: "The incidence of TE was significantly lower among patients who received AT (0/17) than those who did not (10/37, P=0.021). FFP or CPT (given for Fb 5100 mg/dl) had no significant effect on incidence of TE and there was no significant difference in factor replacement administered to either group." Quote: "For appropriate comparison of the platelet count during ASP therapy, in those with and those without TE, the platelet count on the day of maximal ASP effect (time of maximal nadir of AT and/ or Fb) was taken as a time point for comparison between the 2 groups. The median platelet counts (range) at this time in those with and those without TE were 136 x 109/l (28 – 223) and 64 x 109/l (9 – 274), respectively. As can be seen from these platelet counts, platelet transfusions were generally not indicated in this population at the time of ASP therapy. There was no observed effect of platelet transfusions on subsequent TE events. The impact of the higher platelet count observed in those with TE is unknown." The study retrospectively evaluated people treated between 1994‐2003. During a time period of nine years a substantial change in supportive care or venous thromboembolism awareness/detection could have been present. No information reported on follow‐up that potentially included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Given that evaluation of prophylaxis with antithrombin concentrates was not a primary study aim, we assume that participants were selected into the study regardless of prophylaxis data. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that all potentially eligible participants for the specified study period were included. Quote: "This study only included adults with newly diagnosed ALL receiving ASP (a component of 2 contemporary regimens used as standard care at our institution) during induction. Patients with relapsed ALL or lymphoidblastictransformation of chronicmyelogenousleukemia were excluded." Quote: "Ten of 54 (18.5%) consecutive adults developed symptomatic, objectively confirmed TE.." It appeared that start of follow‐up and start of intervention coincided for all participants. |
| Bias in classification of interventions | All outcomes | Serious risk | The intervention was unclearly described in the methods as well as results sections. Of note, evaluation of thromboprophylaxis with antithrombin concentrates was not the primary aim of the study. Quote (methods section): "Monitoring of coagulation parameters (PT, APTT, Fb), AT activity (Chromogenix,Coamatic, Antithrombin Kit) and factor replacement was at the discretion of the treating physician. AT replacement consisted of Thrombate III1 Bayer AG,Leverkusen, Germany)." Quote (results section): "The practice of ‘‘prophylactic’’ AT replacement varied amongst physicians and AT was administered to 17 patients when AT activity fell to <70%." Quote: "Although not all patients had a baseline value to enable statistical comparison, the majority had levels of AT and fibrinogen assessed during the course of ASP therapy to allow comparison of these parameters in those with and without TE events." Unclear whether lack of antithrombin activity monitoring affected the classification of intervention. Assignments of intervention status were determined retrospectively. |
| Bias due to deviations from intended intervention | All outcomes | No information | No information reported on deviations from the prophylaxis with antithrombin concentrates, as this was not the primary study aim. |
| Bias due to missing data | All outcomes | Serious risk | Unclear what data were missing in which participants. Quote: "Although not all patients had a baseline value to enable statistical comparison, the majority had levels of AT and fibrinogen assessed during the course of ASP therapy to allow comparison of these parameters in those with and without TE events." |
| Bias in measurement of outcomes
|
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
|
| Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All thrombotic events were confirmed by imaging. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available. However, the outcome measurements as described in the methods section and the analyses were consistent and in agreement with an a priori plan. Of note, evaluation of prophylactic antithrombin concentrates was not the primary study aim. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding, as the control group is likely not a valid comparator group. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Serious risk | Baseline confounders controlled for, quote: "Cox regressionmodelingwas performed and included whether patient was managed using the AT replacement strategy, age, sex, presence of a CVC, WBC at diagnosis, steroid used, ALL trial protocol, type ofasparaginase, and whether or not FFP was given. In this model, only management using the AT replacement strategy was statistically significantly associated with incidence of thrombosis." Data not presented. Baseline characteristics shown and were similar between antithrombin and no‐antithrombin groups in regard to age and sex. Unequal distribution of central venous catheters (7/30 in the antithrombin group vs 2/15 in the no‐antithrombin group, P = 0.7) and fresh frozen plasma use (4/30 in the antithrombin group vs 11/15 in the no‐antithrombin group, P<0.001). Other potential confounders such as risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, oral contraceptives, inherited thrombophilia traits, smoking, and obesity), and co‐morbidities not reported and adjusted for. Participants received treatment according to different ALL regimens. Consequently, differences in asparaginase formulation and dosing were reported (pegylated asparaginase treatment in 19/30 in the antithrombin group vs 3/15 in the no‐antithrombin group. Quote: "More doses were required in patients who had received PEGylatedasparaginase(median of three doses; range 0–11 doses) as compared with those who received nativeasparaginase(median one dose, range 0–9). Four of 11 patients (36%) who received nativeasparaginasedid not require any AT replacement, as compared with two of 19 (10.5%) in those receiving PEGylatedasparaginase." Also a difference in corticosteroid use were observed (dexamethasone use in 20% (3/15) in the no‐antithrombin group vs 73% (22/30) in the antithrombin group, P<0.01; prednisolone use in 80% (12/15) in the no‐antithrombin group vs 30% (9/30) in the antithrombin group, P<0.01). No report of intrathecal chemotherapy. The study authors acknowledged the potential confounders, quote: "It is unlikely that transitions to different ALL trial protocols (between the observation and replacement cohorts) have influenced this dramatic change in VTE incidence, as other than the corticosteroid formulation used with induction, the drugs used in remission induction have not changed substantially. As a result of transition between various trials over the study period, the proportion of patients receiving standard compared to PEGylatedasparaginasediffered between the two groups, however as noted above a published study showed no apparent difference in thrombotic risk between these formulations." The study was conducted from 2005‐2013; including a historical control group (2005‐2009) and an intervention group (2009‐2013) after implementation of the antithrombin replacement protocol in 2009. Quality in supportive care and awareness/detection of venous thromboembolism were likely to have changed over this period of time. Follow‐up time included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Selection into the study did not appear to be related to intervention, outcome or any prognostic factor. It appeared that start of follow‐up and start of intervention coincided for all participants. Quote: "Forty‐five consecutive patients with a diagnosis of ALL treated withasparaginase‐containing phase I induction protocols were included in this retrospective cohort observational study." Quote: "The first 15 historical patients (2005–2009) received standard therapy with no AT replacement but thrombotic events were recorded. (...) Subsequently, a protocol of AT replacement was instituted (described below) and delivered to the subsequent 30 patients (2009–2013)." It appeared that all potentially eligible participants for the specified study period have been included. |
| Bias in classification of interventions | All outcomes | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | All outcomes | Low risk | The reported deviations from the intended intervention reflect usual practice. Quote: "In seven patients, AT levels did not fall below 70 iu/dl and they therefore did not receive AT replacement." |
| Bias due to missing data | All outcomes | Low risk | Data were reasonably complete; type of ALL protocol received was missing in one participant. Clarifications of outcome data obtained from study author correspondence. |
| Bias in measurement of outcomes |
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, none of the outcomes were subjective. |
| First‐time symptomatic venous thromboembolism | Moderate risk | Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available; however, the venous thromboembolism outcome measurements as described in the methods section and the analyses were consistent and in agreement with an a priori plan. The mortality outcomes were not mentioned in the methods section but reported in the results section. Analysis was presented by intention‐to‐treat in the results and abstract sections (7 participants in the intervention group were intended to receive antithrombin supplementation but did not; none of them developed venous thromboembolism). |
| Overall bias | All outcomes | Serious risk | Based on the serious risk of bias due to confounding. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | Venous thromboembolism‐related mortality | Critical risk | Baseline confounders not controlled for. Conference abstract—no method dealing with potential confounders presented. Baseline characteristics presented and were similar regarding sex. The mean age was 39 years (range 21 to 73) in the antithrombin group and 32 years (range 19‐63) in the no‐antithrombin group. Unclear reporting of body mass index (2.01 in the no‐antithrombin group and 1.84 in the antithrombin group, unit?). No information reported about presence of risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, and smoking), co‐morbidities, or co‐interventions. Participants in the antithrombin and the no‐antithrombin groups were treated according to various different ALL regimens including different type and dosing asparaginase (total total doses of asparaginase was 66 in the no‐antithrombin group and 46 in the antithrombin group; 91% received pegylated asparaginase in the no‐antithrombin group and 100% in the antithrombin group). No reporting of the type/dosing of corticosteroids and intrathecal chemotherapy. The study was conducted from 2011‐2018; including a historical control group (2011‐2014) and an intervention group (2014‐2018) after implementation of the antithrombin replacement protocol in 2014 (our interpretation). Quality in supportive care and awareness/detection of venous thromboembolism were likely to have changed over this period of time. |
| Bias in selection of participants into the study | Venous thromboembolism‐related mortality | Moderate risk | Retrospective design; only a conference abstract was available. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that all potentially eligible participants for the specified study period were included. Limited information on selection of participants, quote: "This retrospective study evaluated patients 18 years of age or older who received L‐asparaginase,pegaspargase, orasparaginaseErwiniachrysanthemibetween January 2011 and June 2018 for ALL,lymphoblasticlymphoma, or NK/T‐cell lymphoma." Quote: "6 patients did not receive AT monitoring while 20 patients were monitored and received AT repletion." It appeared that start of follow‐up and start of intervention coincided for all participants. |
| Bias in classification of interventions | Venous thromboembolism‐related mortality | Moderate risk | Intervention was not clearly defined in the methods (or results) section; only a conference abstract was available. Some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | Venous thromboembolism‐related mortality | Low risk | It appeared that all participants in the intervention group received the intended thromboprophylaxis; however, only a conference abstract was available. |
| Bias due to missing data | Venous thromboembolism‐related mortality | Low risk | All data appeared to be reported; a conference abstract with supplemental tables were available. |
| Bias in measurement of outcomes | Venous thromboembolism‐related mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the reported outcome of interest was objective. |
| Bias in selection of the reported result | Venous thromboembolism‐related mortality | Moderate risk | Only a conference abstract was available. No pre‐registered protocol or statistical analysis plan were available; however, the outcome was described in the methods section consistent with an a priori plan. |
| Overall bias | Venous thromboembolism‐related mortality | Critical risk | Based on the critical risk of bias due to confounding. |
[ROBINS‐I tool version 19, September 2016]
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for. Correspondence—no method dealing with potential confounders presented. Baseline characteristics not similar between groups with median age of 33 years (range 19 to 77) in the antithrombin group and 38 years (range 20 to 75) in the no‐antithrombin group; 63% versus 72% male sex in the antithrombin and no‐antithrombin group, respectively. Slightly more B‐cell immunophenotype in the antithrombin group and T‐cell immunophenotype in the no‐antithrombin group. No information reported about potential confounders such as presence of risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, obesity, and smoking) or co‐morbidities. Only participants in the antithrombin group underwent monitoring of fibrinogen levels and received cryoprecipitate (if fibrinogen levels <100 mg/dl or <150 mg/dl with suspected bleeding); in total, 31 of 43 participants in the antithrombin group received fibrinogen supplementation. Participants in the antithrombin and the no‐antithrombin groups were treated according to different ALL regimens containing pegylated asparaginase, not further described. Thus, type and dosing of intrathecal chemotherapy, corticosteroids, and asparaginase potentially vary (not reported). Quote: "Patients were treated with variousasparaginasecontaining regimens including apediatricbased regimen." The study was conducted from 2009‐2019. Participants before and after implementation of an antithrombin activity monitoring and supplementation practice in 2012, respectively, were enrolled. Quality in supportive care and awareness/detection of venous thromboembolism likely could have changed over this period of time. No information reported on follow‐up that potentially included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality (outcome data obtained from author correspondence). |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that all potentially eligible participants for the specified study period were included. Limited information on selection of participants, quote: "Patients were identified using our institution’s hematological malignancy registry." Quote: "Patients in the control group were derived from a historical cohort prior to implementation of the AT supplementation protocol." It appeared that start of follow‐up and start of intervention coincided for all participants. |
| Bias in classification of interventions | All outcomes | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | All outcomes | Moderate risk | Limited information. 27/43 participants received the intended intervention in the antithrombin group, likely due to antithrombin activity levels above 50% (but not written directly in text). |
| Bias due to missing data | All outcomes | Low risk | All data appeared to be reported. |
| Bias in measurement of outcomes | All‐cause mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
| First‐time symptomatic venous thromboembolism | Moderate risk | Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). Imaging confirmation not described. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available; however, the venous thromboembolism outcome was described in the methods section consistent with an a priori plan. Of note, all‐cause mortality was not part of the described or reported outcomes, and outcome data were obtained by study author correspondence. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
[ROBINS‐I tool version 19, September 2016]
Risk of bias summary ‐ primary outcomes of harms
Risk of bias summary ‐ primary outcomes of harms
Risk of bias summary ‐ secondary outcomes of harms
Of the 13 included non‐randomised studies, we classified one or more available outcomes of 10 studies as being at critical risk of bias due to confounding (including two studies with invalid control group definition) and outcomes of one study at critical risk of bias due to confounding and in selection of participants into the study, as summarised in Table 17. Ten of these 11 studies were retrospective follow‐up studies, and one had a mixed retrospective/prospective design. All of these studies failed to control for baseline confounding regarding the intervention groups of interest; the majority did not describe important baseline characteristics for intervention and control groups. We judged the outcomes of remaining two studies to be at serious risk of bias due to confounding (Farrell 2016; Sibai 2020) and in selection of the reported result (Sibai 2020).
| Study | Outcomes | Domain at critical risk of bias |
|---|---|---|
|
| Bias due to baseline confounding | |
|
| Bias due to confounding and invalid control group definition | |
|
| Bias due to baseline confounding | |
|
| Bias due to confounding and invalid control group definition | |
|
| Bias due to baseline confounding | |
|
| Bias due to baseline confounding | |
|
| Bias due to baseline confounding | |
|
| Bias due to baseline confounding | |
|
| Bias due to baseline confounding | |
|
|
| Bias due to baseline confounding |
|
| Bias due to confounding and in selection of participants into the study | |
|
| Bias due to baseline confounding |
Of note, four of 13 included non‐randomised studies were conference abstracts published between 2016 to 2018. We contacted the first and/or the corresponding authors in regard to all included studies for relevant data. If we could not find any contact information, we contacted one of the co‐authors of the study. Authors from nine of 13 studies responded.
Bias due to confounding
We classified 11 of 13 non‐randomised studies as being at critical risk of bias for all assessed outcomes regarding baseline confounding; the main reasons hereof were unclear definition of the intervention group/thromboprophylaxis indication (Freixo 2017/Table 5; Orvain 2020/Table 7); inclusion of various ALL treatment regimens and/or lack of baseline information, as only brief correspondence or conference abstracts were available (Al Rabadi 2017/Table 4; George 2020/Table 16; Grose 2018/Table 15; Umakanthan 2016/Table 10); and substantial differences between intervention and control groups (Chen 2019/Table 12; Grace 2018/Table 6; Rank 2018/Table 8). Importantly, fundamental issues regarding the control group definition were identified in two studies (Bigliardi 2015/Table 11 ; Elliott 2004/Table 13), in which the allocation to prophylactic treatment with antithrombin concentrates was dependent on the antithrombin activity levels (an antithrombin level < 70% qualified for prophylaxis). Hence, it was clear in one study (Bigliardi 2015) and suspected in the other study (Elliott 2004) that the group without antithrombin substitution solely comprised participants with antithrombin levels ≥ 70%; thus representing invalid comparator groups.
In Farrell 2016 (Table 14), baseline confounders such as age, sex, presence of central venous catheter, white blood cell count at diagnosis, steroid used, ALL trial protocol, type of asparaginase, fresh frozen plasma co‐intervention, and antithrombin concentrates were controlled for in a multiple Cox regression model of the time to venous thromboembolism; albeit raw data were not presented. Despite similar age and sex, the study authors reported substantial differences between the antithrombin and no‐antithrombin groups in relation to fresh frozen plasma infusions and ALL regimens with various types/dosing of asparaginase and corticosteroids. No information was provided on risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, oral contraceptives, inherited thrombophilia traits, smoking, and obesity) and co‐morbidities. Finally, the study was conducted from 2005 to 2013, including a historical control group (2005 to 2009) and an intervention group (2009 to 2013)—after implementation of the antithrombin replacement protocol in 2009. Quality in supportive care and awareness/detection of venous thromboembolism may have changed over this period of time.
In Sibai 2020 (Table 9), baseline confounders were not controlled for. However, baseline characteristics between the low molecular weight heparin and no‐low molecular weight heparin groups were similar in regard to median age and weight as well as sex distribution. All participants were treated according to a paediatric‐inspired ALL regimen. However, 123/125 participants in the low molecular weight heparin prophylaxis group received native Escherichia coli‐derived asparaginase and 3/125 received pegylated asparaginase. Type of asparaginase formulation was not reported for the historical control group. Additionally, participants aged < 60 years received asparaginase 12,500 IU/m2 intramuscularly once weekly for 30 weeks of intensification treatment, whereas participants aged ≥ 60 years received a shortened intensification (21 weeks) with asparaginase dose reduction (6000 IU/m2 intramuscularly once weekly). Accordingly, dexamethasone was also dose‐reduced from 9 mg/m2 to 6 mg, twice daily from days one to five. Intrathecal chemotherapy was identical. No information reported about potential confounders such as presence of risk factors of venous thromboembolism (infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, obesity (body mass index), and smoking) or co‐interventions. Participants with prior venous thromboembolism were excluded. The study authors paid attention to the challenge of heparin dosing in relation to co‐morbidities and weight, excluding participants with baseline creatinine clearance < 30 mL/minute and adjusting enoxaparin dose according to weight. Finally, the study was conducted from 2001 to 2018; including a historical control group and an intervention group after implementation of thromboprophylaxis as the standard of care (in 2009 according to the associated former publications). Quality in supportive care and awareness/detection of venous thromboembolism were likely to have changed over this period of time.
Based on the above mentioned descriptions, we classified these two remaining non‐randomised studies as being at serious risk of bias for the domain of confounding for all assessed outcomes.
Bias in selection of participants into the study
We classified one of 13 non‐randomised studies as being at critical risk of bias regarding selection bias for the outcomes of first‐time symptomatic venous thromboembolism and asymptomatic venous thromboembolism (Rank 2018/Table 8), as nine participants in this study with a diagnosis of thromboembolism in the central NOPHO registry had missing data; thus the diagnosis of thromboembolism could not be confirmed in these people. We judged the all‐cause mortality outcome as being at moderate risk of bias in this domain.
We classified the remaining 12 studies as being at moderate risk of bias for all assessed outcomes, mainly due to limited information on selection of controls and the retrospective design. In Sibai 2020 (Table 9) and Farrell 2016 (Table 14), selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that all potentially eligible participants for the specified study period had been included. Of note, in Sibai 2020, people who did not receive asparaginase therapy or received less than seven cycles of chemotherapy (early protocol discontinuation not because of venous thromboembolism or bleeding event) due to early leukaemic relapse, haematopoietic stem cell transplantation, non‐haemorrhagic complications, loss to follow‐up, or death were excluded. Participants were not excluded due to missing data. Follow‐up and start of intervention appeared to coincide for all participants.
Bias in classification of interventions
We classified five of 13 non‐randomised studies as being at serious risk of bias regarding bias in classification of interventions for venous thromboembolism‐related and bleeding outcomes (Bigliardi 2015/Table 11; Elliott 2004/Table 13; Freixo 2017/Table 5; Orvain 2020/Table 7; Rank 2018/Table 8). The intervention with heparin prophylaxis (Freixo 2017; Orvain 2020; Rank 2018) or antithrombin concentrates (Bigliardi 2015; Elliott 2004) was not clearly defined and determined retrospectively. Details related to type, dosing and duration of thromboprophylaxis for all participants were unknown. We judged the all‐cause mortality outcome, assessed in Freixo 2017, Bigliardi 2015 and Rank 2018, to be at moderate risk of bias, as it is unlikely that knowledge of or risk of this outcome affected the classification of intervention.
The remaining eight studies were classified as moderate risk of bias for all outcomes, mainly due to the retrospective aspect of assignment into groups. In Farrell 2016 (Table 14) and Sibai 2020 (Table 9), intervention was clearly defined in the methods section. However, some aspects of the assignments of intervention status were determined retrospectively, but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome.
Bias due to deviations from intended interventions
We assessed the effect of assignment to thromboprophylaxis in the 13 included non‐randomised studies, of which we classified two as being at moderate risk of bias due to deviations from intended interventions for all assessed outcomes, as the study authors could not explain 4/10 deviations (Chen 2019/Table 12) or provided no explanation for 16 deviations (George 2020/Table 16). We could not assess the risk of bias in four studies (Bigliardi 2015/Table 11; Elliott 2004/Table 13; ; Freixo 2017/Table 5; Rank 2018/Table 8), as evaluation of thromboprophylaxis was not not a primary study aim; thus, no information on deviations was reported.
The remaining seven studies were classified as low risk of bias for all outcomes, as the described deviations from the intended intervention was in accordance with what would be expected in usual practice. In Farrell 2016 (Table 14), seven participants did not receive the intended antithrombin concentrates, as their measured antithrombin activity levels were above the defined target level of substitution. In Sibai 2020 (Table 9), all included participants received the intended prophylaxis with low molecular weight heparin, which was temporarily withheld due to low platelet counts < 30 x 109/L.
Bias due to missing data
We classified three of 13 non‐randomised studies as being at serious risk of bias due to missing data for all assessed outcomes, as data on intervention were missing in 122/784 participants (Orvain 2020/Table 7) and 9/1,772 (274 adults) participants registered with thromboembolism were excluded due to missing data (Rank 2018/Table 8). The involved number of people with missing baseline values and data on antithrombin activity monitoring were unclear in the remaining study (Elliott 2004/Table 13). We could not assess the risk of bias in three studies (Al Rabadi 2017/Table 4; Freixo 2017/Table 5; Umakanthan 2016/Table 10), as only conference abstracts were available.
The remaining seven studies including one conference abstracts (Grose 2018/Table 15) were classified as low risk of bias for all outcomes, as data appeared to be reasonably complete; supplemental tables accompanied the abstract. In Farrell 2016 (Table 14) and Sibai 2020 (Table 9), data were reasonably complete.
Bias in measurement of outcomes
Given the retrospective design in all included non‐randomised studies, blinding of outcome assessors was not done. We classified one of 13 non‐randomised studies as being at serious risk of bias in measurements of outcomes for the outcome of asymptomatic venous thromboembolism (Rank 2018/Table 8), because no screening was done and as a result of the multi‐centre study design differences in care and awareness may have influenced the reported incidence. We judged the outcome of first‐time symptomatic venous thromboembolism to be at moderate risk of bias in this study, as symptoms/clinical signs that raise the clinical suspicion and lead to confirmation by imaging have a subjective element. The mortality outcome was classified as being at low risk of bias, given the objective measurement.
We classified 11 studies as being at moderate risk of bias for the first‐time symptomatic venous thromboembolism and/or major bleeding outcome (Al Rabadi 2017/Table 4; Bigliardi 2015/Table 11; Chen 2019/Table 12; Elliott 2004/Table 13; ; Farrell 2016/Table 14; Freixo 2017/Table 5; George 2020/Table 16; Grace 2018/Table 6; Orvain 2020/Table 7; Sibai 2020/Table 9; Umakanthan 2016/Table 10), given the potential subjective element in both the diagnosis of symptomatic venous thromboembolism and major bleeding. In addition, as a result of the multi‐centre study design, differences in care and awareness may have influenced diagnostic suspicion in two of these studies (Grace 2018; Orvain 2020). Of note, no information on imaging confirmation of venous thromboembolism was reported in four of these studies (Bigliardi 2015; Freixo 2017; George 2020; Umakanthan 2016); however, we assume that the venous thromboembolic events were assessed radiologically—although not described. We classified the mortality outcomes as being at low risk of bias due to the objectivity in measurement. In both Farrell 2016 (Table 14) and Sibai 2020 (Table 9), the clinical suspicion of symptomatic venous thromboembolism relied on symptoms—although objectively confirmed by imaging in both studies.
We judged the remaining study to be at low risk of bias for all outcomes (Grose 2018/Table 15), as the study authors only presented relevant data on objectively measured mortality outcomes.
Bias in selection of the reported result
We did not identify any pre‐registered protocols or statistical analysis plan for any of the included 13 non‐randomised studies, being exploratory by design. We classified the risk of bias in selection of the reported result as being at serious risk for all assessed outcomes in two of 13 studies (Orvain 2020/Table 7; Sibai 2020/Table 9). In Orvain 2020, several variables (risk factors of venous thromboembolism) and associations between intervention, co‐interventions, and outcomes were not prespecified in the methods section and described only in the results section. In Sibai 2020, potential predictors, hence risk factors, of venous thromboembolism including ALL immunophenotype and body weight ≤ versus > 80 kg were mentioned only in the results and abstract sections; the former yielding statistical significance. Thus, one of the main results presented was likely to have been selected on the basis of results from multiple subgroup analyses.
We judged 10 studies to be at moderate risk of bias for all outcomes (Al Rabadi 2017/Table 4; Bigliardi 2015/Table 11; Chen 2019/Table 12; Elliott 2004/Table 13; Farrell 2016/Table 14; Grace 2018/Table 6; George 2020/Table 16; Grose 2018/Table 15; Rank 2018/Table 8; Umakanthan 2016/Table 10), as the outcome measurements as described in the methods section and the analyses were consistent and in agreement with an a priori plan. Of note, in Chen 2019 and Farrell 2016 the presented main analysis was by intention‐to‐treat, reporting the number of participants who actually received antithrombin concentrates in the intervention group in close proximity. Moreover, in the study by Farrell 2016, the mortality outcomes were presented only in the results section.
We could not assess the risk of bias in this domain for the remaining study, as only a conference abstract was available including scarce information on the outcomes of interest (Freixo 2017/Table 5).
Allocation
‐
Blinding
‐
Incomplete outcome data
‐
Selective reporting
‐
Other potential sources of bias
‐
Effects of interventions
See: Summary of findings 1 Summary of Findings Table ‐ Antithrombin concentrates compared to no antithrombin concentrates for thromboprophylaxis in adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy; Summary of findings 2 Summary of Findings Table ‐ Enoxaparin compared to no enoxaparin for thromboprophylaxis in adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy
Of the 13 studies assessed for risk of bias, we included two studies with outcomes classified as being at serious risk of bias in our main descriptive analysis of harms. Of the remaining 11 studies with outcomes classified as being at critical risk of bias, nine studies were included along with the two studies from the main analysis in a sensitivity analysis of harms to explore the effect of confounding. Harms included all‐cause mortality (primary outcome), major bleeding (primary outcome), venous thromboembolism‐related mortality (secondary outcome), clinically‐relevant non‐major bleeding/heparin‐induced thrombocytopenia (secondary outcomes), and quality of life (secondary outcome).
The remaining two studies (Bigliardi 2015; Elliott 2004) were not considered in the descriptive analyses due to fundamental issues involving invalid comparator group, as identified in the previous 'Risk of bias' assessment.
Given the lack of included RCTs, we did not analyse benefits (first‐time symptomatic venous thromboembolism (primary) and asymptomatic venous thromboembolism (secondary)) and could not conduct effects analyses using network meta‐analyses—as stated in the methods section.
Main descriptive analysis
See summary of findings Table 1 of harms for prophylactic antithrombin concentrates versus no prophylactic antithrombin concentrates and summary of findings Table 2 of harms for low molecular weight heparin prophylaxis (enoxaparin) versus no low molecular weight heparin (enoxaparin) prophylaxis.
For the main descriptive analysis, we included two non‐randomised studies, in which available outcomes were not at critical risk of bias; one study (Farrell 2016) of 40 adults receiving prophylactic antithrombin concentrates (Kybernin P) or no antithrombin concentrates and one study (Sibai 2020) of 224 adults receiving prophylaxis with low molecular weight heparin (enoxaparin) or no prophylaxis with low molecular weight heparin.
Prophylactic antithrombin concentrates versus no prophylactic antithrombin concentrates
All‐cause mortality. The one included non‐randomised study (Farrell 2016) of 40 adults found no association between reduced all‐cause mortality in those who received prophylactic antithrombin concentrates compared with those who did not receive antithrombin concentrates (risk ratio (RR) 0.55, 95% confidence interval (CI) 0.26 to 1.19; Analysis 1.1.1, Figure 5). The effect estimates are based on the intention‐to‐treat principle; seven participants in the antithrombin group did not receive antithrombin concentrates due to antithrombin activity levels above the target for substitution. We performed an 'as‐treated' analysis excluding these seven participants, of whom none died during the study period (Analysis 1.1.2, Figure 5). Overall, the direction of the results did not markedly change (RR 0.74, 95% CI 0.36 to 1.55; 33 participants). We presented only the effect estimate based on the intention‐to‐treat principle in the summary of findings table. All‐cause mortality was downgraded from low to very low quality due to serious risk of bias and imprecision.
Forest plot of prophylactic antithrombin concentrates versus no antithrombin concentrates, outcome: all‐cause mortality
Venous thromboembolism‐related mortality. Neither was an association found between reduced venous thromboembolism‐related mortality in those who received prophylactic antithrombin concentrates compared with those who did not receive antithrombin concentrates, based on the intention‐to‐treat principle (RR 0.10, 95% CI 0.01 to 1.94; Analysis 1.2.1, Figure 6). We did not see any difference in effect estimate of venous thromboembolism‐related mortality in the 'as‐treated' analysis (Analysis 1.2.2, Figure 6). We presented only the effect estimate based on the intention‐to‐treat principle in the summary of findings table. Venous thromboembolism‐related mortality was downgraded from low to very low quality due to serious risk of bias and imprecision.
Forest plot of prophylactic antithrombin concentrates versus no antithrombin concentrates, outcome: venous thromboembolism‐related mortality
The study did not report on bleeding events or quality of life. We could not explore potential heterogeneity by subgroup analyses of type of ALL regimen, Philadelphia chromosome status, presence of inherited thrombophilia traits, and asparaginase formulation, as relevant data were not available. Given the small study size, few events, and wide CIs containing both clinically relevant benefit and harm, we cannot draw a conclusion regarding the effect of prophylactic antithrombin concentrates on overall mortality and venous thromboembolism‐related mortality based on the available evidence (very low certainty of evidence). Additionally, the concerns of the internal validity of the study further decrease our confidence in the effect estimates.
Low molecular weight heparin prophylaxisversus no low molecular weight heparin prophylaxis
The one included non‐randomised study (Sibai 2020) of 224 adults reported zero major bleeding events in the enoxaparin group during preemptive enoxaparin treatment; however the number of events were not reported for the control group. Neither all‐cause mortality, venous thromboembolism‐related mortality, quality of life, heparin‐induced thrombocytopenia, nor clinically relevant non‐major bleeding events were reported.
Of note, included participants were aged ≥ 17 years; however, we do not know, how many participants were aged <18 years.
Sensitivity analysis
In addition to the two studies from the main analysis, we included nine studies (1,403 adults with ALL) with outcomes classified as being at critical risk of bias at study level in our sensitivity analysis of harms to explore the effect of bias due to confounding. Three studies (179 people with ALL) explored antithrombin concentrates versus no antithrombin concentrates (Chen 2019; George 2020; Grose 2018), and six studies (1224 adults with ALL) reported on heparin prophylaxis versus no heparin prophylaxis (Al Rabadi 2017; Freixo 2017; Grace 2018; Orvain 2020; Rank 2018; Umakanthan 2016).
Prophylactic antithrombin concentrates versus no prophylactic antithrombin concentrates
All‐cause mortality. In the sensitivity analysis of harms, one of three studies evaluating antithrombin concentrates provided data on all‐cause mortality (George 2020); we have presented this study in a forest plot alongside the one study from the main descriptive analysis at serious risk of bias (Analysis 2.1, intention‐to‐treat analysis, Figure 7; 101 adults with ALL). We observed small study sizes, few events, and wide CIs crossing the line of no effect. The effect estimates seem diverging across studies, likely being attributable to the small number of evaluated studies. With only two studies included, it is important to acknowledge the large potential impact, if the average effect of one study differs in size or direction. Moreover, the critical failure to control for confounding was associated with inconsistency between the studies.
Sensitivity analysis. Forest plot of prophylactic antithrombin concentrates versus no antithrombin concentrates, outcome: all‐cause mortality
Venous thromboembolism‐related mortality. Two of four studies evaluating antithrombin concentrates reported on venous thromboembolism‐related mortality (Chen 2019; Grose 2018); we have presented these in a forest plot alongside the one study from the main descriptive analysis at serious risk of bias (Analysis 2.2, intention‐to‐treat analysis, Figure 8; 158 adults with ALL). In Grose 2018, no venous thromboembolism‐related deaths occurred in 43 participants with ALL, and RRs were therefore not calculable. Of note, the analysis of Chen 2019 is based on the intention‐to‐treat principle; 10/47 participants in the antithrombin group did not receive antithrombin concentrates due to antithrombin activity levels > 60 %, no antithrombin monitoring or unknown reason. Using an 'as‐treated' approach, we did not see any substantial impact on the effect estimate (results not shown). Overall, we observed small study sizes, few events, and wide CIs crossing the line of no effect. The effect estimates seem in agreement, albeit the critical failure to control for confounding was considerable.
Sensitivity analysis. Forest plot of prophylactic antithrombin concentrates versus no antithrombin concentrates, outcome: venous thromboembolism‐related mortality (intention‐to‐treat)
None of the studies reported on bleeding events or quality of life. We were not able to explore heterogeneity by subgroup analyses for any of the above mentioned outcomes, as relevant data regarding type of ALL regimen, Philadelphia chromosome status, presence of inherited thrombophilia traits, and asparaginase formulation were not available.
Heparin prophylaxis versus no heparin prophylaxis
All‐cause mortality. In the sensitivity analysis of harms, three of six studies evaluating low molecular weight heparin/unfractionated heparin prophylaxis reported on all‐cause mortality (Freixo 2017; Grace 2018; Rank 2018); we have presented these studies in a forest plot (Analysis 3.1, Figure 9; 385 adults with ALL). We could not include the study from the main descriptive analysis at serious risk of bias due to lack of outcome data. We observed small study sizes, few events, and wide CIs crossing the line of no effect. The effect estimates seem diverging—likely being attributable to the small number of evaluated studies. Of note, different types of heparin were used. With only three studies included, it is important to acknowledge the large potential impact, if the average effect of one study differs in size or direction. Moreover, the critical failure to control for confounding was associated with inconsistency between the studies.
Sensitivity analysis. Forest plot of heparin prophylaxis versus no heparin prophylaxis, outcome: all‐cause mortality
Major bleeding. Four of six studies evaluating low molecular weight heparin/unfractionated heparin prophylaxis reported on major bleeding events (Al Rabadi 2017; Freixo 2017; Grace 2018; Orvain 2020); we have presented these studies in a forest plot (Analysis 3.2, Figure 10; 933 adults with ALL). We could not include the study from the main descriptive analysis at serious risk of bias due to lack of outcome data. We observed wide CIs crossing the line of no effect. The effect estimates seem diverging—likely being attributable to the small number of evaluated studies. With only four studies included, it is important to acknowledge the large potential impact, if the average effect of one study differs in size or direction. Moreover, the critical failure to control for confounding was associated with inconsistency between the studies.
Sensitivity analysis. Forest plot of heparin prophylaxis versus no heparin prophylaxis, outcome: major bleeding
Venous thromboembolism‐related mortality. Three of six studies evaluating low molecular weight heparin/unfractionated heparin prophylaxis reported on venous thromboembolism‐related mortality (Al Rabadi 2017; Freixo 2017; Grace 2018); we have presented these studies in a forest plot (Analysis 3.3, Figure 11; 149 adults with ALL). No venous thromboembolism‐related deaths occurred in two of the studies; thus, RRs were not calculable. We could not include the study from the main descriptive analysis at serious risk of bias due to lack of outcome data. We observed small study sizes, few events, and wide CIs crossing the line of no effect.
Sensitivity analysis. Forest plot of heparin prophylaxis versus no heparin prophylaxis, outcome: venous thromboembolism‐related mortality
None of the studies reported on clinically relevant non‐major bleeding, heparin‐induced thrombocytopenia, or quality of life. We were not able to explore heterogeneity by subgroup analyses for any of the above mentioned outcomes, as relevant data regarding type of ALL regimen, Philadelphia chromosome status, presence of inherited thrombophilia traits, and asparaginase formulation were not available. Although the observed heterogeneity could arise through differences in participants, interventions and outcome assessments, the possibility that bias was the cause of heterogeneity is inevitable.
Discussion
Summary of main results
Two retrospective non‐randomised studies met the inclusion criteria for our main descriptive analysis of harms. We judged the outcomes of two studies to be at serious risk of bias using ROBINS‐I. One of these studies involving 40 adults was included in our main comparison of antithrombin concentrates versus no antithrombin concentrates for thromboprophylaxis. The remaining study of 224 adults reported on enoxaparin versus no enoxaparin thromboprophylaxis but did not provide any extractable outcome data of harms. We judged outcomes of an additional number of nine non‐randomised studies to be at critical risk of bias and included these along with the two studies at serious risk of bias in our sensitivity analysis of harms exploring confounding. Of these nine studies, three studies (179 adults) explored antithrombin concentrates versus no antithrombin concentrates for thromboprophylaxis and six studies (1224 adults) evaluated heparin thromboprophylaxis versus no heparin thromboprophylaxis. None of these studies controlled for confounding.
We are uncertain whether antithrombin concentrates have an effect on all‐cause mortality and venous thromboembolism‐related mortality in adults with ALL treated with an asparaginase‐based chemotherapy regimen (very low certainty of evidence). The sensitivity analysis emphasised our concern about confounding. We do not know the impact of antithrombin concentrates on major bleeding, clinically relevant non‐major bleeding, and quality of life due to insufficient or lack of data for these outcomes.
We do not know the impact of enoxaparin prophylaxis on all‐cause mortality, major bleeding, venous thromboembolism‐related mortality, clinically relevant non‐major bleeding, heparin‐induced thrombocytopenia, and quality of life due to insufficient or lack of data for these outcomes. In our sensitivity analyses including studies with outcomes at a critical risk of bias, presence of small study sizes; few events; and CIs crossing the line of no effect confirmed the unreliable evidence.
Given the lack of included RCTs, hence pooled effect estimates from meta‐analyses of benefits, we know the impact of neither antithrombin concentrates nor enoxaparin prophylaxis on first‐time symptomatic venous thromboembolism and asymptomatic venous thromboembolism.
None of the included studies evaluated vitamin K antagonists, synthetic pentasaccharides, parenteral direct thrombin inhibitors, direct oral anticoagulants, or mechanical interventions.
Of note, we did not succeed in retrieving additional data from study authors in 12/23 included studies. In addition, 10/23 studies included different haematological malignancies and/or wide age span, from which outcome‐specific data for adults with ALL treated with an asparaginase‐based regimen could not de directly extracted. This could potentially introduce bias. Finally, two studies were excluded from the descriptive analyses due to invalid comparator group definition identified in the'Rrisk of bias' assessment.
Overall completeness and applicability of evidence
The studies from the descriptive analyses covered the continents of Europe and North America and included participants with different pheno‐ and genotypes of ALL at different stages of treatment, which should increase the applicability of the results.
Yet, there is insufficient or lack of reliable evidence to be confident of the effects of antithrombin concentrates and enoxaparin prophylaxis on all‐cause mortality, first‐time symptomatic venous thromboembolism, major bleeding, venous thromboembolism‐related mortality, quality of life, clinically relevant non‐major bleeding, heparin‐induced thrombocytopenia (for heparins), and asymptomatic venous thromboembolism. In particular, the paucity of data to evaluate the impact of interventions on quality of life is important, given the potential burden of both oral, subcutaneous, and intravenous route of thromboprophylaxis administration. Additionally, there is insufficient high‐quality evidence to draw conclusions about the likelihood of a reduction in venous thromboembolic events and mortality when administering antithrombin concentrates or enoxaparin prophylaxis to adults with ALL undergoing asparaginase‐based chemotherapy. Antithrombin concentrates and heparins are frequently used for thromboprophylaxis in people with ALL, despite inadequate documentation. A Cochrane meta‐analysis of people with cancer treated with chemotherapy in an ambulatory setting found that prophylaxis with low molecular weight heparin significantly reduced the incidence of symptomatic venous thromboembolism; however, the risk of major bleeding suggested caution (Di Nisio 2016). Given the risk of chemotherapy‐induced thrombocytopenia and disrupted haemostatic/fibrinolytic balance in people with ALL during asparaginase‐based chemotherapy, we do not think that this evidence is sufficiently generalisable to ALL.
We considered non‐randomised studies as part of the inclusion criteria of this review, as they potentially could be sufficient to show clinically appreciable benefits and/or harms for primary thromboprophylaxis in cases of very large effects, presence of a dose‐response gradient, or results that opposed any plausible residual confounding. However, we judged 11 of 13 included non‐randomised studies to be at critical risk of bias regarding outcomes; hence, the studies were too problematic to provide any useful evidence on the effects of intervention. In the remaining two studies, the basis for evidence was not of sufficient robustness to draw any conclusions about important benefits or harms, given the serious risk of bias due to confounding, concerns of internal validity, and imprecision of effect estimates.
Quality of the evidence
The overall certainty of evidence in relation to the effect of antithrombin concentrates on all‐cause mortality and venous thromboembolism‐related mortality was very low. Any estimate of effect was highly uncertain and is likely to change with further research. The design of most research conducted in this field involving adults with ALL is non‐randomised, and studies are compromised by uncontrolled confounding and small sample sizes. To this adds the insufficient reporting of harms associated with pre‐emptive anticoagulation. Because of this, we were unable to draw any conclusions on the effects of primary thromboprophylaxis (antithrombin concentrates and enoxaparin) in adults with ALL undergoing asparaginase‐containing chemotherapy.
In the assessment of effect estimates of all‐cause mortality and venous thromboembolism‐related mortality for the comparison of antithrombin concentrates versus no antithrombin concentrates, we downgraded the evidence to very low certainty by one level for serious risk of bias primarily due to confounding and by two levels for imprecision of effect estimates involving small study sizes, few events, and broad CIs that included both beneficial and detrimental effects. Of note, outcome data from non‐randomised studies started at low quality and were not upgraded. Bleeding events and quality of life could not be assessed, as these outcomes were not reported. See summary of findings Table 1.
We could not assess the certainty of evidence of effect estimates of harms for the comparison of enoxaparin prophylaxis versus no enoxaparin prophylaxis, as these outcomes were insufficiently reported. See summary of findings Table 2.
Of note, benefits associated with thromboprophylactic treatment such as reduction in incidences of first‐time symptomatic venous thromboembolism and asymptomatic venous thromboembolism were not evaluated, although reported in studies, as only non‐randomised studies were included.
Potential biases in the review process
We evaluated the potential biases in the review process using ROBIS tool (Whiting 2016).
Study eligibility criteria
We adhered to the pre‐defined objectives and eligibility criteria, which were unambiguous and appropriate for the review question. We made appropriate restrictions in eligibility criteria based on study characteristics enrolling studies conducted from 1970 (the time of introduction of asparaginase in adult ALL treatment) and case control or cohort studies only with a control group. No restrictions were based on information sources (i.e. publication status/format, language, or availability of data). In summary, we had low concerns regarding the risk of bias in eligibility criteria.
Identification and selection of studies
We conducted a comprehensive and unrestricted search; a detailed list of all sources searched can be found in Search methods for identification of studies. Additionally, we contacted the authors of relevant studies to identify any unpublished material, missing data, or information regarding ongoing studies. Two review authors (L.S.L. and C.U.R.) independently and systematically searched, screened, and selected the studies. Of note, a third review author (B.A.N.) independently evaluated the selection of one eligible study (Rank 2018) due to conflicts of interest by one of the review authors involved (C.U.R.).
We based our search on an appropriately designed search strategy, albeit reports not specifically indexed by ALL (including all synonymous terms) were not retrieved. Noteworthy is that we identified several reports including haematological diseases by handsearching reference lists, in which the unspecific inclusion of acute leukaemia (not indexed) was described. Moreover, we introduced a risk of citation bias by handsearching reference lists. Further, existing databases and smaller congresses from e.g. African or Asian countries not commonly used or attended by European and American researchers were not searched. Finally, the two review authors who undertook the search were neither blinded to the journal, authors, institution, or results when assessing studies for eligibility. In theory, this could influence the decision to exclude a study.
We believe that we minimised the risk of citation bias from the handsearch of reference lists and the likelihood of publication bias by contacting authors in this field and pharmaceutical companies to retrieve unpublished data. Moreover, the independent assessment of study eligibility by at least two review authors substantially reduced the risk of bias related to the lack of blinding of the review authors. Therefore, we had low concerns regarding the risk of bias in identification and selection of studies.
Data collection and study appraisal
We identified 29 potentially eligible studies after full‐text screening, of which six studies are awaiting classification due to lack of data regarding ALL diagnosis and/or asparaginase therapy. We have reached out to the study authors to obtain the information needed for the decision on eligibility.
Of the included 23 studies, we contacted the corresponding authors (if not available, the first author/co‐author) of all studies to retrieve all relevant data needed for analysis; 11/23 contacted study authors responded to our request (Table 2).
Two review authors (L.S.L. and C.U.R.) independently and systematically extracted data from all included studies using a predefined, standardised data collection form, thus minimising data extraction errors. Additionally, these two authors assessed the risk of bias using the standardised ROBINS‐I tool and the quality of evidence using the GRADE approach. We acknowledge that these assessments leaves room for different interpretations, in particular when quality of reporting is poor. We added quotes and provided arguments for our judgements, thus allowing the reader to reach different judgements and conclusions. In addition, a third review author (B.A.N.) independently extracted data and assessed the risk of bias regarding one of the included studies (Rank 2018) due to conflicts of interest by one of the review authors involved (C.U.R.).
Despite our systematic and standardised approach, we had high concerns regarding the risk of bias in data collection and study appraisal due to the lack of response from contacted authors and, thus, data retrieval, in approximately half of the included studies; 42% of these (5/12) being conference proceedings.
Synthesis and findings
Of the 23 studies, which met the eligibility criteria of this review, we did not include 12 studies in our analyses due to invalid control group definition in two studies and lack of outcome data in 10 studies.
We judged nine of the 11 studies included in our analyses to be at critical risk of bias regarding outcomes, and we decided to exclude these from the main descriptive analysis and conduct a sensitivity analysis including all studies no matter their risk of bias. This decision was not originally described in our review protocol but was properly explained and reasoned.
We believe that our synthesis was appropriate given the nature and similarity in the research questions, study designs, and outcomes across the included studies. However, we were only able to address the observed between‐study heterogeneity by visual inspection of the forest plots without pooled effect estimates and not by using the Chi2 test and the I2 heterogeneity statistic to help quantify the degree of potential heterogeneity as no meta‐analyses were performed. For this reason, we could not perform any of the sensitivity analyses, as described in the methods section. Additionally, subgroup analyses could not be conducted for the assessed primary outcomes exploring the impact of characteristics of participants (presence of Philadelphia chromosome; inherited thrombophilia traits) and characteristics of treatment (paediatric‐inspired versus conventional adult ALL regimens; type of asparaginase formulation including Escherichia coli‐derived, pegylated asparaginase, and Erwiniachrysanthemi‐derived), as relevant data were not available. We included only two non‐randomised studies, judged to be at serious risk of bias at outcome‐level, in our main descriptive analysis accompanied by nine studies at critical risk of bias at outcome‐level in a sensitivity analysis (< 10 studies in each comparison); hence no funnel plot could be generated. Consequently, our findings were not robust and highly uncertain. All biases in primary studies were addressed.
We had high concerns regarding the risk of bias in synthesis and findings due to the lack of outcome data in 8/23 included studies.
Overall risk of bias
In summary, we judged this review to be at low risk of bias, as we explained and accounted for all of the above mentioned concerns and major limitations including the incomplete correspondence with study authors (lack of additional data retrieval) in the interpretation of the findings. We considered the relevance of identified studies to the research question in the selection process. In addition, given the involvement of 6/9 review authors in one of the included studies, two review authors without any involvement/conflicts of interest independently selected; extracted data; and assessed risk of bias for that particular study, minimising the introduction of bias. Moreover, we did not emphasise results or base our interpretation of findings and conclusions on the statistical significance.
We had departures from our review protocol, which mainly consisted of wording corrections; clarifications; and updates regarding references, tools, and software programmes (see Differences between protocol and review). However, we made the following additions to the methods section before conducting our first search: (1) addition of intrathecal chemotherapy, corticosteroids and important co‐interventions to the listed confounding factors; (2) exclusion of studies with outcomes at critical risk of bias from the main descriptive analysis (included in a sensitivity analysis); and (3) revision of the listed predefined subgroup analyses. We made one addition after the first search involving a hierarchy for selection of interventions, if more than one intervention in a study, as listed in decreasing order: pharmacologic drugs, blood‐derived products, and mechanical interventions. Nonetheless, we do not believe that these departures introduced a potential source of bias in this review.
Agreements and disagreements with other studies or reviews
To our knowledge, no systematic review exists evaluating the benefits and harms associated with the use of thromboprophylaxis in adults with acute lymphoblastic leukaemia treated with asparaginase‐based regimens. However, this review agrees with existing expert panel guidelines (Stock 2011; Zwicker 2020) and reviews (Annibali 2018; Fulcher 2020; Sorigue 2020) that all acknowledge the low‐quality evidence in this area of research.
A recently published systematic review and network meta‐analysis of thromboprohylaxis in children with cancer (1318 children; 97.5% ALL) compared antithrombin concentrates, low molecular weight heparin, vitamin K antagonists, and standard of care (no prophylaxis/low molecular weight heparin for central venous catheter patency) (Pelland‐Marcotte 2019). Low molecular weight heparin was the only agent significantly associated with lower odds of venous thromboembolism when compared with standard of care. No significant differences in odds of major bleeding were shown for any agent. Noteworthy is that all included six studies were judged to be at moderate or high risk of bias, and a high degree of heterogeneity was observed. In children with ALL, only two frequently cited open‐label RCTs on thromboprohylaxis to date have been conducted. The most recent THROMBOTECT trial enrolled 949 children with ALL and found a significant reduction in incidence of venous thromboembolism and no difference in major bleeding during induction therapy for the antithrombin concentrates (if antithrombin activity levels < 80%) and low molecular weight heparin (enoxaparin) arms when compared with low dose unfractionated heparin (Greiner 2019). Concerns regarding study limitations, as stated by Male and colleagues, include the open‐label, not masked design, the fact that 33% declined treatment of the subcutaneously injected enoxaparin, and allowance of cross‐over between the treatment arms—introducing substantial risk of bias. Yet, the results were confirmed in an as‐treated analysis (Male 2019). The underpowered open‐label PARKAA trial randomising 109 children with ALL to either antithrombin concentrates or no antithrombin concentrates during induction treatment found no difference in incidence of venous thromboembolism between the two arms (Mitchell 2003). A third RCT, the ongoing open‐label PREVAPIX‐ALL study, is randomising children with ALL or lymphoblastic lymphoma to primary thromboprohylaxis with apixaban (direct oral anticoagulant) versus no prophylaxis during induction treatment (clinicaltrials.gov identifier: NCT02369653).
We did not evaluate the benefits, i.e. reduction in venous thromboembolic events, of thromboprophylaxis, as no RCTs evaluating thromboprophylaxis to date have been conducted in adults with ALL. To this adds that we judged the majority of outcomes of included non‐randomised studies in this field to be at critical risk of bias mainly due to uncontrolled confounding. We judged the certainty of evidence for the two studies included in the main descriptive analysis to be very low due to serious risk of bias and imprecision, thus limiting our confidence in the findings of harms. Most recently, the low‐quality data have also been acknowledged in a short review (Fulcher 2020).
The clinical practice guidelines for thromboprophylaxis in people with cancer, as outlined in the Background/How the intervention might work section, may not be directly applicable for people with ALL, primarily because of the leukaemia‐ and treatment‐induced interference with blood counts and the haemostatic/fibrinolytic balance. The current USA expert panel guidelines and the Scientific and Standardisation Committee on haemostasis and malignancy of the International Society on Thrombosis and Haemostasis highlight the limited evidence in the field of ALL but suggest antithrombin supplementation for antithrombin activity levels < 50% to 60% either alone or in combination with low molecular weight heparin thromboprophylaxis in adults with ALL during asparaginase‐containing induction (Stock 2011; Zwicker 2020). Antithrombin activity monitoring is suggested to be weekly during asparaginase treatment by the Scientific and Standardisation Committee on haemostasis and malignancy of the International Society on Thrombosis and Haemostasis . Particularly, this committee suggests thromboprophylaxis with low molecular weight heparin for those at high risk of venous thromboembolism (e.g. obesity or prior venous thromboembolism) in the outpatient setting during induction and intensification phases (Zwicker 2020).
Further, the USA guidelines recommend cryoprecipitate concentrates for bleeding prophylaxis, replacing fibrinogen, as cryoprecipitate concentrates contain high levels of factor VIII, which is highly thrombogenic (Stock 2011). Specifically in regard to the plasma‐derived products for antithrombin supplementation, a general concern in relation to the use of these products has involved asparagine repletion; thus counteracting the anti‐leukaemic action of asparaginase. Currently available guidelines suggest against fresh frozen plasma for thromboprophylaxis in people with ALL for this reason (Steiner 2008; Stock 2011; Zwicker 2020). Worryingly, participants receiving antithrombin replacement had an increased relapse rate in the paediatric THROMBOTECT trial (Greiner 2019), which may be a chance finding but certainly raises concern.
Risk of bias summary ‐ primary outcomes of harms
Risk of bias summary ‐ primary outcomes of harms
Risk of bias summary ‐ secondary outcomes of harms
Forest plot of prophylactic antithrombin concentrates versus no antithrombin concentrates, outcome: all‐cause mortality
Forest plot of prophylactic antithrombin concentrates versus no antithrombin concentrates, outcome: venous thromboembolism‐related mortality
Sensitivity analysis. Forest plot of prophylactic antithrombin concentrates versus no antithrombin concentrates, outcome: all‐cause mortality
Sensitivity analysis. Forest plot of prophylactic antithrombin concentrates versus no antithrombin concentrates, outcome: venous thromboembolism‐related mortality (intention‐to‐treat)
Sensitivity analysis. Forest plot of heparin prophylaxis versus no heparin prophylaxis, outcome: all‐cause mortality
Sensitivity analysis. Forest plot of heparin prophylaxis versus no heparin prophylaxis, outcome: major bleeding
Sensitivity analysis. Forest plot of heparin prophylaxis versus no heparin prophylaxis, outcome: venous thromboembolism‐related mortality
Comparison 1: Antithrombin concentrates versus no antithrombin concentrates, Outcome 1: All‐cause mortality
Comparison 1: Antithrombin concentrates versus no antithrombin concentrates, Outcome 2: Venous thromboembolism‐related mortality
Comparison 2: Sensitivity analysis: Antithrombin concentrates versus no antithrombin concentrates, Outcome 1: All‐cause mortality (intention‐to‐treat analysis)
Comparison 2: Sensitivity analysis: Antithrombin concentrates versus no antithrombin concentrates, Outcome 2: Venous thromboembolism‐related mortality (intention‐to‐treat analysis)
Comparison 3: Sensitivity analysis: Heparin versus no heparin, Outcome 1: All‐cause mortality
Comparison 3: Sensitivity analysis: Heparin versus no heparin, Outcome 2: Major bleeding
Comparison 3: Sensitivity analysis: Heparin versus no heparin, Outcome 3: Venous thromboembolism‐related mortality
| Antithrombin concentrates compared to no antithrombin concentrates for thromboprophylaxis in adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy | ||||||
| Patient or population: thromboprophylaxis in adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy Setting: United Kingdom (inpatient setting) Intervention: antithrombin concentrates Comparison: no antithrombin concentrates | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with no antithrombin concentrates | Risk with antithrombin concentrates | |||||
| All‐cause mortality ‐ Intention‐to‐treat analysis | 538 per 1.000 | 296 per 1.000 | RR 0.55 | 40 | ⊕⊝⊝⊝ | The evidence is very uncertain about the effect of antithrombin concentrates on all‐cause mortality. f,g |
| First‐time symptomatic venous thromboembolism | Benefits not assessed for non‐randomised studies in this review (outcome reported in study) | 40 | ‐ | |||
| Major bleeding ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Venous thromboembolism‐related mortality ‐ Intention‐to‐treat analysis | 154 per 1.000 | 15 per 1.000 | RR 0.10 | 40 | ⊕⊝⊝⊝ | The evidence is very uncertain about the effect of antithrombin concentrates on mortality secondary to thromboembolism. f,g |
| Quality of life ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Asymptomatic venous thromboembolism | Benefits not assessed for non‐randomised studies in this review (outcome not reported in study) | (0 studies) | ‐ | |||
| Clinically relevant non‐major bleeding ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_415691309867741365. | ||||||
| a. Risk of bias: downgraded one level for serious risk of bias due to confounding. | ||||||
| Enoxaparin compared to no enoxaparin for thromboprophylaxis in adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy | ||||||
| Patient or population: adults with acute lymphoblastic leukaemia treated with asparaginase‐based chemotherapy Setting: Canada (inpatient setting) Intervention: enoxaparin Comparison: no enoxaparin | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with no enoxaparin | Risk with enoxaparin | |||||
| All‐cause mortality ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| First‐time symptomatic venous thromboembolism | Benefits not assessed for non‐randomised studies in this review (outcome reported in study) | 224 | ‐ | |||
| Major bleeding | No major bleeding events in the enoxaparin group. Number of events not reported in the control group. 1 | 224 | ‐ | |||
| Thrombosis‐related mortality ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Quality of life ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Asymptomatic venous thromboembolism ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| Adverse events: clinically relevant non‐major bleeding / heparin‐induced thrombocytopenia ‐ not reported | ‐ | ‐ | ‐ | ‐ | ‐ | |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_415827266990160046. | ||||||
| 1. Schulman S, Kearon C.. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non‐surgical patients.. Journal of Thrombosis and Haemostasis.; 2005. | ||||||
| Study | Risk of bias assessment | Inclusion in descriptive analyses | Annotations |
|---|---|---|---|
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in main descriptive analysis | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | yes | Included in main descriptive analysis | |
| yes | yes | Included in sensitivity analysis due to critical risk of bias | |
| yes | no | Not included in descriptive analysis. Critical risk of bias due to control group inconsistency. Participants who did not receive antithrombin supplementation had antithrombin activity levels above the target of supplementation (<70%) cannot be classified as controls. Of note, the aim of the study was to assess the safety profile of Erwinia chrysanthemi‐derived asparaginase in adults with ALL. | |
| yes | no | Not included in descriptive analysis. Critical risk of bias due to control group inconsistency. Unclear if antithrombin activity was monitored in all included participants, and if only participants who had antithrombin activity levels ≥ 70% were allocated to the control group. | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data | |
| no | no | Complete lack of outcome data (and control group inconsistency) | |
| no | no | Complete lack of outcome data (and control group inconsistency) |
| Response from study author | Study | Format | Provided data | Annotations |
|---|---|---|---|---|
| YES | letter |
| Invalid control group definition. | |
| full report |
| |||
| full report |
| |||
| conference abstract |
| |||
| correspondence |
| |||
| full report |
| |||
| full report | ‐ | Study database closed in 2005; data not available. | ||
| full report |
| Most of the requested additional data (outcome data, patient/disease characteristics) could not be provided due competing interests | ||
| full report |
| |||
| full report |
| |||
| conference abstract |
| |||
| NO | full report | ‐ | Complete lack of data of interest. | |
| conference abstract | ‐ | Incomplete outcome data. | ||
| full report | ‐ | Complete lack of data of interest. | ||
| full report | ‐ | Complete lack of data of interest. | ||
| full report | ‐ | Incomplete outcome data. Suspected invalid control group definition. | ||
| conference abstract | ‐ | Incomplete outcome data. | ||
| full report | ‐ | Complete lack of data of interest. | ||
| conference abstract | ‐ | Complete lack of data of interest. | ||
| full report | ‐ | Complete lack of data of interest. | ||
| conference abstract | ‐ | Complete lack of data of interest. Control group inconsistency. | ||
| full report | ‐ | Incomplete outcome data. | ||
| conference abstract | ‐ | Complete lack of data of interest. Control group inconsistency. |
| Study | Participants | Interventions | Outcomes of harms |
|---|---|---|---|
| 108 participants (4 to 58 years) | Intervention (4/55 ALL) Control (5/53 ALL) Co‐interventions: | Outcome data for participants aged 4‐58 years withhaemato‐oncologicaldiseases. Major bleeding All‐cause mortality, venous thromboembolism‐related mortality, clinically relevant non‐major bleeding, and quality of life were not reported. | |
| 30 participants (16‐58 years) with ALL (n = 23) or lymphoblastic lymphoma treated with Escherichia coli‐derived L‐asparaginase‐containing consolidation chemotherapy. | Intervention (17/17 ALL) Antithrombin substitution schedule A: Antithrombin substitution schedule B:
Control (6/13 ALL) Co‐interventions: | Outcome data for participants aged 16‐58 years with ALL or lymphoblastic lymphoma. All‐cause mortality, major bleeding, clinically relevant non‐major bleeding, venous thromboembolism‐related mortality, and quality of life were not reported. | |
| 706 adults (18 to 59 years) with Philadelphia chromosome‐negative ALL or lymphoblastic lymphoma treated with paediatric‐inspired Escherichia coli‐derived L‐asparaginase‐containing induction chemotherapy. | Intervention (n=179) Control (n=89) Co‐interventions: | Thromboprophylaxisdata for 268/706 of the included adults with ALL or lymphoblastic lymphoma. Venous thromboembolism‐related mortality All‐cause mortality, major bleeding, clinically‐relevant non‐major bleeding, heparin‐induced thrombocytopenia, and quality of life were not reported. | |
| 15 participants (16 to 58 years) with ALL or lymphoblastic lymphoma treated with Escherichia coli‐derived L‐asparaginase‐containing consolidation. | Intervention (n = 7) Control (n = 8) Co‐interventions: | Outcome data for participants aged 16‐58 years with ALL or lymphoblastic lymphoma. Bleeding events All‐cause mortality, venous thromboembolism‐related mortality, and quality of life were not reported. | |
| 214 participants (15 to 59 years) diagnosed with either non‐Burkitt type ALL (n = 191) or T‐cell lymphoblastic lymphoma receiving Escherichia coli‐derived asparaginase‐containing induction therapy. | Intervention (n = 85) Control (n = 127) Co‐interventions: | Thromboprophylaxisdata for 212/214 of the included participants aged 15‐59 years with ALL orlymphoblasticlymphoma. Bleeding events Venous thromboembolism‐related mortality All‐cause mortality, heparin‐induced thrombocytopenia, and quality of life were not reported. | |
| 236 participants (16 to 59 years) with newly diagnosed ALL treated according to the HOVON‐37 ALL remission induction including Escherichia coli L‐asparaginase. | Intervention (n = 193) Control (n = 43) Co‐interventions: | Outcome data for participants aged 16‐59 years with ALL. | |
| 27 adults (19 to 49 years) with newly diagnosed Philadelphia chromosome‐negative ALL (non‐Burkitt type) treated according to the paediatric‐inspired AHCVAD (augmented hyper‐CVAD) protocol including pegylated asparaginase. | Intervention Control Co‐interventions: | Outcome data for adults with ALL. | |
| 20 participants (16 to 58 years) with ALL treated according to ALL regimens including L‐asparaginase during induction treatment. | Intervention (n = 8) Control (n = 12) Co‐interventions: | Outcome data for participants aged 16‐58 years with ALL. | |
| 12 adults (45 to 76 years) with Philadelphia chromosome‐negative ALL treated according to an age‐based, dose‐adjusted pegylated asparaginase‐based regimen with a CALGB 10403 backbone. | Intervention (n = 8) Control (n = 4) Co‐interventions:Cryoprecipitate (if fibrinogen <120mg/dL). | Outcome data for adults with ALL. Treatment‐related mortality (all‐cause mortality?) Major bleeding, venous thromboembolism‐related mortality, clinically relevant non‐major bleeding, and quality of life were not reported. | |
| 46 adults (18 to 38 years) with ALL treated with an pegylated asparaginase‐based regimen. | Intervention Control Co‐interventions: | Outcome data for adults with ALL. All‐cause mortality, major bleeding, venous thromboembolism‐related mortality, clinically relevant non‐major bleeding, and quality of life were not reported. |
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for. Conference abstract—no information reported about baseline characteristics or presence of confounders such as age, sex, type of ALL regimen including intrathecal chemotherapy and steroid use, risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, oral contraceptives, inherited thrombophilia traits, smoking, and obesity), co‐morbidities, and co‐interventions (antithrombin, fresh frozen plasma, cryoprecipitate, fibrinogen, platelets). The distribution of central venous catheters between the two groups was not reported. Quote: "All thromboses in the prophylaxis group were associated with central venous catheters.." No method for dealing with potential confounders. Sparse information regarding the control group, quote: "Fifteen patients receiving similar induction protocols who did not receive any prophylactic anticoagulation were used as the control group." No information reported of the time period; unknown if a substantial change in supportive care or awareness/detection of venous thromboembolism could have been present. However, the study authors report that all participants received identical asparaginase treatment (type and dose). |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design; only a conference abstract was available. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that start of follow‐up and start of intervention coincided for all participants. Quote: "We implemented A fixed dose (1 mg/kg/day) of LMWH prophylaxis for adults with ALL undergoing induction therapy with L‐asparaginaseat our institution in 2012. In this study we report our outcomes with this protocol." Selection of the intervention group, quote: "Retrospective review of twenty‐three patients who received prophylactic anticoagulation with the LMWHenoxaparin(1mg/kg/day) while undergoing induction therapy with L‐asparaginasefor ALL." Limited information on selection of the control group, quote: "Fifteen patients receiving similar induction protocols who did not receive any prophylactic anticoagulation were used as the control group." The historical cohort may confound the results. |
| Bias in classification of interventions | All outcomes | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended interventions | All outcomes | Low risk | It appears that all participants in the intervention group received the intended thromboprophylaxis; however, only a conference abstract was available. |
| Bias due to missing data | All outcomes | No information | All data appeared to be reported; however, only a conference abstract was available. |
| Bias in measurement of outcomes | Venous thromboembolism‐related mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
| Major bleeding | Moderate risk | Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). | |
| Bias in selection of the reported result | All outcomes | Moderate risk | Only a conference abstract was available. No pre‐registered protocol or statistical analysis plan were available; however, the reported outcomes as described in the methods section were consistent with an a priori plan. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for. Conference abstract—no method for dealing with potential confounders presented. Quote: "There were no differences between the 2 groups (before and after protocol) concerning age, gender, ALL phenotype, risk,hyperleukocytosis, LDH level, SNC involvement, coagulation at diagnosis and treatment response." Statistical method for comparison and raw data not presented. Participants were treated according to either HOVON 100 (n = 22) or CLCG (n = 4) chemotherapy protocols including L‐asparaginase (type and dosing not presented); intrathecal chemotherapy and steroid use not mentioned. No information reported about presence of risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, smoking, and obesity) or co‐morbidities. Unclear if participants receiving prophylaxis with low molecular weight heparin were at higher risk of venous thromboembolism compared with controls. Participants received antithrombin, fibrinogen, and plasma concentrates; number of participants who received co‐interventions and distribution between the low molecular weight heparin prophylaxis and no low molecular weight heparin prophylaxis group not presented. Further, the study authors showed no attempt trying to disentangle the effect of co‐interventions from that of low molecular weight heparin prophylaxis, as this was not the primary study aim. No information reported on follow‐up that potentially included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design; only a conference abstract was available. Evaluation of low molecular weight heparin prophylaxis was not a primary aim of the study, thus we assume that participants were selected into the study regardless of low molecular weight heparin prophylaxis data. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appears that start of follow‐up and start of intervention coincide for all participants. It appears that all potentially eligible participants for the specified study period were included. Quote: "In 2016 our group adapted the protocol described byBiddeciet al. for prophylactic replacement therapy with Fibrinogen (FI) Concentrate and Antithrombin (AT) Concentrate." Quote: "From 2014, a total of 26 patients (15 males, 11 females; median age 35 years, 22Bprecursor‐ALL, 4 T‐ALL) were treated with intensive chemotherapy (...) 10 of the 26 patients were enrolled after the protocol implementation." Quote: "All patients but 1 started prophylactic LMWH during L‐Asp therapy in the protocol group but only 7 (44%) were given LMWH before 2016." |
| Bias in classification of interventions | All‐cause mortality | Moderate risk | Intervention was not defined in the methods section and unclearly defined in the results section, as evaluation of thromboprophylaxis with low molecular weight heparin was not the primary aim of the study. Thus, assignments of intervention status were determined retrospectively. |
|
| Serious risk | ||
| Bias due to deviations from intended intervention | All outcomes | No information | No information reported on deviations from the prophylaxis with low molecular weight heparin, as this was not the primary study aim. |
| Bias due to missing data | All outcomes | No information | All data of interest were not reported, as evaluation of thromboprophylaxis with low molecular weight heparin was not the primary study outcome, and only a conference abstract was available. Selected relevant outcome data have been obtained by author correspondence. |
| Bias in measurement of outcomes |
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, none of the outcomes were subjective. |
|
| Moderate risk | Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). Imaging confirmation not described. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). | |
| Bias in selection of the reported result | All outcomes | No information | There is too little information to make judgement, as only an abstract was available. No pre‐registered protocol or statistical analysis plan were available. Of note, evaluation of prophylaxis with low molecular weight heparin was not the primary study aim. Outcome data have been obtained by study author correspondence. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for, and no method for dealing with potential confounders was presented. Baseline characteristics presented and were similar between groups regarding age, sex, white blood cell count at ALL diagnosis, immunophenotype, mediastinal mass, performance status 0‐2 (a proxy for immobilisation; unknown in n = 5), and body mass index. Twenty‐two per cent (11/49) were obese (body mass index ≥ 30) in the control group, and 27% (10/36) were obese in the intervention group. Eighteen percent had performance status 2 (9/49) in the control group and 6% (2/36) in the intervention group. Only a higher median platelet count at day 32 in the intervention group (320 x 103/L) differed significantly from that in the control group (224 x 103/L). Additional confounders such as prior venous thromboembolism, infections, central venous catheters, oral contraceptives, inherited thrombophilia traits, or smoking) and co‐morbidities were not assessed for the two groups. All participants were treated according to a paediatric‐inspired ALL protocol; however, the type, route of administration and dose of asparaginase was changed in August 2011 as part of an amendment due to a high number of safety events. Intrathecal chemotherapy and steroid use were not mentioned. Participants treated prior to the amendment (control group) received pegylated apsaraginase (2500 IU/m2 intravenously) during induction and 2500 IU/m2 every two weeks during consolidation, whereas participants treated after the amendment (intervention group) received one dose of native Escherichia coli‐derived asparaginase (25,000 IU/m2 intramuscularly) during induction followed by pegylated asparaginase (2000 IU/m2 intravenously) every three weeks during consolidation. Only participants treated after the amendment received heparin prophylaxis. Co‐interventions reported only for the intervention group and included antithrombin replacement (if antithrombin activity levels <30%; n=1), cryoprecipitate (if fibrinogen < 50mg/dL; n = 4), and platelet transfusions (if <50x109/L ; 24/33 during induction and 11/23 during consolidation). Quote: "Although a direct comparison of VTE rates before and after instituting the use of prophylactic anticoagulation is limited by the change in the schedule, dosing, and type ofasparaginaseadministered during induction, the use ofcryoprecipitate, FFP and antithrombin replacement was minimal, suggesting that the observed decrease in VTE incidence is due to the use of prophylactic anticoagulation and not coagulation factor supplementation." Quote: "Comparison of the rate of thrombosis before and after the implementation of guidelines for prophylactic anticoagulation is also limited by the modifications that were made simultaneously to improve the safety of the study." The type, dose, and timing of heparin prophylaxis varied. Quote: "Although 9 thrombotic events occurred while patients were on prophylactic anticoagulation, there was a significant variability in dose, timing, and likely compliance with daily administration of prophylaxis in the outpatient setting making it hard to determine if prophylaxis failed to prevent the development of VTE." The follow‐up time included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Mixed prospective/retrospective design. Quote: "Between September 2007 and June 2013, 110 eligible adult patients with newly diagnosed ALL were enrolled on a high riskpediatricinspired treatment regimen through the DFCI ALL Consortium.." Quote: "The data safety and monitoring committee put the study on hold between September 2010 and August 2011 due to a high number of safety events related to IV PEG‐asparaginase, including VTE. (...) and mandated that patients with a Philadelphia‐chromosome positive (Ph+)leukemiabe removed from this study." Quote: "Of the 110 patients, 22 patients with a Ph+leukemiaand 3 patients not prescribed prophylactic anticoagulation (2 induction failures and 1 who was transplanted directly after induction) were excluded from analysis; 49 patients treated prior to the amendment and 36 treated after the amendment wereanalyzed." Selection into the study was based on participants' characteristics observed after the start of the study (Philadelphia chromosome status); Philadelphia chromosome status may be related to the risk of venous thromboembolism. It appears that start of follow‐up and start of intervention coincide for all participants. |
| Bias in classification of interventions | All outcomes | Moderate risk | The recommended intervention was clearly defined (thromboprophylaxis was recommended for all participants post‐amendment); however Quote: "While these guidelines offered management suggestions for the use of prophylactic anticoagulation, use was not mandated; failure to use prophylactic anticoagulation or to assess coagulation factor levels was not a protocol violation." Quote: "The decision to use prophylactic anticoagulation was left to the treating physician." Some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | All outcomes | Low risk | The reported deviations from intended intervention are what would be expected in usual practice of ALL treatment. 3/36 participants in the intervention group did not receive prophylactic treatment during induction due to ALL treatment‐related toxicities including severe thrombocytopenia, neutropenia, and infection. 12/36 patients only received venous thromboembolism prophylaxis during induction and not in consolidation due to development of venous thromboembolism, haematopoietic stem cell transplantation in first complete remission immediately after induction, induction failure, relapse, or other patient‐specific factors including thrombocytopenia. Thromboprophylaxis was held for procedures (n=24 during induction; n = 19 during consolidation), platelet count < 30x109/L (n = 12 during induction; n = 2 during consolidation), and bleeding symptoms (n = 1 during induction; n = 1 during consolidation), and at a patient's request (n = 1 during consolidation). |
| Bias due to missing data | All outcomes | Low risk | Data appear to be reasonably complete. Clarifications of selected relevant outcome data have been obtained by author correspondence. |
| Bias in measurement of outcomes |
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, none of the outcomes were subjective. |
|
| Moderate risk | Venous thromboembolism‐related symptoms are likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). Additionally, as a result of a multi‐centre study, differences in care and awareness may have influenced diagnostic suspicion. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available; however, the outcome measurements as described in the methods section and the analyses were consistent and in agreement with an a priori plan. Of note, the cumulative incidence of pulmonary embolism, a subgroup of serious venous thromboembolism, between the two groups was highlighted in the abstract and results section, although not being prespecified in the methods section. However, this addition did not change the overall direction of results. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | In a multivariate logistic regression model, the authors explored risk factors of venous thromboembolism at ALL diagnosis including older age, female sex, high body mass index, high platelet count. Baseline characteristics not presented between heparin prophylaxis and no‐heparin prophylaxis groups. Baseline confounders not controlled for in the intervention groups of interest apart from selected risk factors of venous thromboembolism, quote: "Although patients who received heparin prophylaxis also had a higher platelet count at diagnosis of ALL in comparison to those who did not (79 versus 59 x 109/L, p=0.05), the association between heparin prophylaxis and an increased risk of VTE persisted after adjustment for platelet counts at diagnosis (Supplementary Table 1). Especially, there was no difference between patients who received heparin prophylaxis or not regarding personal and familial thromboembolic events." Poor‐risk prognostic factors (ALL immunophenotype, high white blood cell count (>30 x109/L), central nervous system involvement, and poor prednisone response), prior venous thromboembolism, oral contraceptives before diagnosis, and a history of smoking as confounders were described for the total cohort but not analysed for the heparin versus no heparin prophylaxis groups. Infections, immobilisation, and co‐morbidities were not assessed for the two intervention groups of interest. Quote: "Hereditary thrombotic risk factors (protein S deficiency, protein C deficiency, activated protein C,prothrombin G20210A and factor V Leiden polymorphisms) and lupus anticoagulant wereseldomlyevaluated (42/110 patients, 38%, with thrombosis and 191/677, 28%, without thrombosis were tested) and their impact on thrombosis could not be evaluated." The authors highlight the presence of central venous lines as a potential confounder, quote: "This last parameter could not be evaluated in our study since all patients have central venous lines. Increased use of CVL, which is recommended for all patients in our protocol, might have caused the higher thrombotic rate in our study as we observed 50 DVT of the upper limb, for which CVL is arecognizedrisk factor." All participants were treated according to the same paediatric‐inspired ALL protocol including same dosing of intrathecal chemotherapy, steroids, and pegylated asparaginase (risk group distribution not reported for the heparin versus no heparin prophylaxis groups). Participants received antithrombin concentrates (if antithrombin activity levels≤60%; 87% of participants), fibrinogen or fresh frozen plasma (if fibrinogen <0.5g/L; 9% received fibrinogen concentrates), and platelets (if platelet count <20x109/L); the distribution between the heparin prophylaxis and no‐heparin prophylaxis groups not presented. Further, the study authors showed no attempt trying to disentangle the effect of co‐interventions from that of prophylaxis with unfractionated heparin/low molecular weight heparin. Quote: "Interestingly, patients who received heparin prophylaxis with no antithrombin prophylaxis (38 patients, 6%) were more likely to experience thrombosis (24%) in comparison to other patients (603 patients, 94%) who experienced 9.6% of VTE (p=0.02)." The authors highlight the type of heparin prophylaxis used as a potential confounder along with the potential important difference between participants who received heparin prophylaxis and participants who did not. The two groups might not be directly comparable. Quote: "Unfractionatedheparin (UFH) at 100 IU/kg/day in continuous infusion was recommended during induction; low molecular weight heparin (LMWH) at prophylactic doses in subcutaneous injection could be used as replacement at the physician’s discretion." Quote: "In our study, even though all patients were supposed to receive heparin prophylaxis, all did not receive such prophylaxis. We observe that it was more frequently administered to patients with less intensive thrombocytopenia. After adjustment for platelet counts at diagnosis, heparin prophylaxis was surprisingly associated with an increased risk of VTE. We have no clear explanation for this observation and the lack of details regarding the modality of heparin prophylaxis in our study (type, dose, therapeutic level monitoring) limits the generalization of this result. Patients deemed to be at increased bleeding risk, and thus at decreased thrombotic risk, might have received less heparin prophylaxis. The extensive use ofunfractionated heparin in our study might also explain the lack of efficacy of heparin prophylaxis that we observed. Given the severe antithrombin depletion induced byasparaginase,unfractionatedheparin is probably not the most effective anticoagulant in this setting, while some data advocate the use of low molecular weight heparin (LMWH) (3.5% of VTE versus 8% in those who receivedunfractionated heparin)." |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design (prospective registration of venous thromboembolic events). Of 813 people screened, 29 were excluded due to non‐eligibility criteria (n=11), loss to follow‐up (n=12), lack of toxicity data (n=3), and consent withdrawal (n = 3). We assumed that participants were selected into the study regardless of low molecular weight heparin prophylaxis data. Missing data on thromboprophylaxis in 122/784 participants; however, none were excluded on that account. Selection of participants into the study did not appear to be related to intervention, outcome, or prognostic factors. It appeared that start of follow‐up and start of prophylaxis with low molecular weight heparin coincided for all participants. |
| Bias in classification of interventions | All outcomes | Serious risk | Intervention was not clearly defined in the methods section. Quote: "Unfractionatedheparin (UFH) at 100 IU/kg/day in continuous infusion was recommended during induction; low molecular weight heparin (LMWH) at prophylactic doses in subcutaneous injection could be used as replacement at the physician’s discretion. Heparin prophylaxis was recommended from day 1 of induction until end of induction." Assignments of intervention status were determined retrospectively. |
| Bias due to deviations from intended intervention | All outcomes | Low risk | Potential deviations from the intended intervention, as described in the methods section, reflected what would be expected in usual practice: Quote: "When platelets were below 20 G/l, heparin prophylaxis was interrupted, and patients received a platelet transfusion. Heparin prophylaxis was resumed after platelet transfusion." |
| Bias due to missing data | All outcomes | Serious risk | Participants were excluded due to lack of outcome data. Quote: "Twenty‐nine patients were excluded due to non‐eligibility criteria (11 patients), lost to follow‐up (12 patients), lack of data regarding toxicity (3 patients), and consent withdrawal (3 patients)." Additional outcome data have been obtained by author correspondence. Missing intervention data in 122/784 participants. |
| Bias in measurement of outcomes | All outcomes | Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). Additionally, as a result of a multi‐centre study, differences in care and awareness may have influenced diagnostic suspicion. |
| Bias in selection of the reported result | All outcomes | Serious risk | No pre‐registered protocol or statistical analysis plan were available. Given the retrospective and exploratory design, several variables (risk factors of venous thromboembolism) and associations between intervention, co‐interventions, and outcomes were not prespecified in the methods section and described only in the results section. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | The study authors adjusted for age (adults aged 18 to 45 years compared with children), sex, body mass index, white blood cell count, immunophenotype, central nervous system status (leukaemic infiltration), presence of mediastinal mass, enlarged lymph nodes, palpable splenomegaly, induction corticosteroids, minimal residual disease at day 29, treatment group at day 29, and prophylaxis with low molecular weight heparin. Quote: "In a multiple Cox regression analysis with delayed entry at day 29 (N 5 1594; 114 TE events), the adjusted hazard ratio (HRa) of TE was significantly increased in ages 10.0 to 17.9 years (HRa, 4.9; 95% CI, 3.1‐7.8; P , .0001) and ages 18.0 to 45.9 years (HRa, 6.06; 95% CI, 3.65‐10.1; P , .0001) compared with children younger than 10.0 years, and for mediastinal mass at ALL diagnosis (HRa, 2.1; 95% CI, 1.0‐4.3; P 5.04; Table 3)." Noteworthy is that the analysis included participants aged 1‐45 years and compared people with versus without development of thromboembolism. Baseline characteristics and confounders such as age, sex, risk factors of venous thromboembolism (central venous catheters, infections, inherited thrombophilia traits, and obesity) were evaluated for participants with and without development of thromboembolism (these were not reported for the heparin prophylaxis and no‐heparin prophylaxis groups, as this was not part of the primary study aim). Unpublished data obtained by author correspondence: median age of 26.5 years (range 18.6‐43.0) in the low molecular weight heparin prophylaxis group and 26.0 years (range 18.0 to 45.8) in the no prophylaxis group; 41% male sex in the low molecular weight heparin prophylaxis group and 67% in the no prophylaxis group; 37% T‐cell immunophenotype in the low molecular weight heparin prophylaxis group and 30% in the no prophylaxis group. No information reported regarding immobilisation, oral contraceptives, smoking, co‐morbidities, or co‐interventions. None of the included participants had a history of thromboembolism. All participants were treated according to the same paediatric ALL protocol including the same dosing of pegylated asparaginase, steroids, and intrathecal chemotherapy. Given the multi‐centre design and lack of protocol guidelines for thromboprophylaxis, prophylaxis with low molecular weight heparin was only used at some centres; 41/233 adults received prophylaxis. The follow‐up time included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All‐cause mortality | Moderate risk | Retrospective design (prospective registration of venous thromboembolic events). Given that evaluation of prophylaxis with low molecular weight heparin was not a primary study aim, participants were selected into the study regardless of low molecular weight heparin prophylaxis data. Consecutive participants with newly diagnosed ALL treated according to the NOPHO ALL2008 protocol were included. Excluded participants accounted for in detail (participants with registered thromboembolism in the NOPHO registry were excluded because of missing imaging confirmation of TE (n=1) and missing data (n = 8)). Quote: "Of 1861 patients with ALL, the following were excluded: 21 patients with acuteleukemiaof ambiguous lineage, 54 with ALL predisposition syndromes (eg, Down syndrome or ataxiatelangiectasia), and 1 not treated according to the ALL2008 protocol. One hundred fifty patients with ALL developed TE, of whom 13 were excluded because of missing imaging confirmation of TE (N = 1), superficial thrombophlebitis (N = 1), septic emboli (N = 1), central venous line (CVL) dysfunction registered as asymptomatic TE (N = 2), and missing data (N = 8). Thus, a total of 1772 patients with Ph‐ ALL, among whom were 137 registered TE cases, were included in the study." Start of follow‐up and start of prophylaxis with low molecular weight heparin coincided for all participants. |
|
| Critical risk | ||
| Bias in classification of interventions | All‐cause mortality | Moderate risk | Intervention was not clearly defined in the methods or results section, as evaluation of thromboprophylaxis with low molecular weight heparin was not the primary study aim. Methods section, quote: "No common recommendations for routineantithromboticprophylaxis exist in the ALL2008 protocol." Assignments of intervention status were determined retrospectively. |
|
| Serious risk | ||
| Bias due to deviations from intended intervention | All outcomes | No information | No information reported on deviations from the prophylaxis with low molecular weight heparin, as this was not the primary study aim. |
| Bias due to missing data | All outcomes | Serious risk | All data of interest are not reported, as evaluation of prophylaxis with low molecular weight heparin was not the primary study aim. Selected relevant outcome data have been obtained by author correspondence. |
| Bias in measurement of outcomes | All‐cause mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
| First‐time symptomatic venous thromboembolism | Moderate risk | Venous thromboembolism‐related symptoms were both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. As a result of a multi‐centre study, differences in care and awareness may have influenced diagnostic suspicion. | |
| Asymptomatic venous thromboembolism | Serious risk | Additionally, as a result of a multi‐centre study, differences in care and awareness may have influenced the incidence of asymptomatic venous thromboembolic events; no screening was done. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available. However, the outcome measurements as described in the methods section and the analyses were consistent and in agreement with an a priori plan. Of note, evaluation of prophylaxis with low molecular weight heparin was not the primary study aim. |
| Overall bias | All‐cause mortality | Critical risk | Based on the critical risk of bias due to confounding. |
|
| Critical risk | Based on the critical risk of bias due to confounding and in selection of participants into the study. | |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | First‐time symptomatic venous thromboembolism | Serious risk | Baseline confounders not controlled for, and no method dealing with potential confounders presented. Baseline characteristics between the prophylaxis with low molecular weight heparin and control groups were similar in regard to median age and weight as well as sex distribution. All participants were treated according to a paediatric‐inspired ALL regimen. However, 123/125 participants in the low molecular weight heparin prophylaxis group received native Escherichia coli‐derived asparaginase and 3/125 received pegylated asparaginase. Type of asparaginase formulation was not reported for the historical control group. Additionally, participants aged <60 years received asparaginase 12,500 IU/m2 intramuscularly once weekly for 30 weeks of intensification treatment, whereas participants aged ≥60 years received a shortened intensification (21 weeks) with asparaginase dose reduction (6000 IU/m2 intramuscularly once weekly). Accordingly, dexamethasone was also dose‐reduced from 9 mg/m2 to 6 mg, twice daily from day 1‐5. Intrathecal chemotherapy was identical. No information reported about potential confounders such as presence of risk factors of venous thromboembolism (infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, obesity (body mass index), and smoking) or co‐interventions. Participants with prior venous thromboembolism were excluded. The study authors paid attention to the challenge of heparin dosing in relation to co‐morbidities and weight using weight‐adjusted doses and excluding participants with renal insufficiency. Quote: "Patients with baseline renal impairment (creatinine clearance <30 ml/min) were excluded in this study." The study was conducted from 2001‐2018; including a historical control group and an intervention group after implementation of thromboprophylaxis as the standard of care (in 2009 according to the associated former publications). However, the reporting of time periods in relation to cases and controls was inconsistent throughout the paper: Quote: "Patients from the historical cohort, treated between 2001 and 2010, did not receive any anticoagulation prophylaxis. In contrast, patients treated between 2011 and 2017 received anticoagulation prophylaxis..." Quote: "...and (ii) increased monitoring and awareness of VTE in the prophylaxis group (2011–2018) than in the control group (2001–2009)." Nonetheless, quality in supportive care and awareness/detection of venous thromboembolism were likely to have changed over this period of time. |
| Bias in selection of participants into the study | First‐time symptomatic venous thromboembolism | Moderate risk | Retrospective design. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that start of follow‐up and start of intervention coincide for all participants. Limited information on selection of the controls, quote: "Patients from the historical cohort, treated between 2001 Quote: "Patients from the historical cohort, treated between 2001 and 2010, did not receive any anticoagulation prophylaxis. In contrast, patients treated between 2011 and 2017 received anticoagulation prophylaxis..." versus Quote: "...and (ii) increased monitoring and awareness of VTE in the prophylaxis group (2011–2018) than in the control group (2001–2009)." Exclusion of participants were accounted for. Quote: "Patients were excluded from this study for the following reasons: (i) received full‐dose LMWH anticoagulation therapy at the beginning of intensification due to prior thrombosis, (ii) underwent allogeneic stem cell transplant during intensification, (iii) relapsed before the seventh cycle of intensification, (iv) did not complete ≥7 cycles of intensification due to other patient factors such as liver toxicity, non‐haemorrhagic death, or loss to follow‐up." Additionally, people with Philadelphia chromosome‐positive ALL were excluded, as they did not receive asparaginase treatment. |
| Bias in classification of interventions | First‐time symptomatic venous thromboembolism | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | First‐time symptomatic venous thromboembolism | Low risk | Deviations from the intended intervention was in accordance with usual practice. Quote: "Among those who did not develop VTE in the prophylaxis group, seven patients had PLT <50 x 109/l and five had PLT <30 x 109/l. Comparatively, among the 17 patients who experienced VTE, four had PLT <50 x 109/l and one patient had PLT <30 x 109/l. Prophylaxis was temporarily held if PLT |
| Bias due to missing data | First‐time symptomatic venous thromboembolism | Low risk | Data appeared to be reasonably complete. |
| Bias in measurement of outcomes | First‐time symptomatic venous thromboembolism | Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. Quote: "The mechanism of this difference is unclear, but may be explained by: (i) a potential increase in antithrombin levels as intensification treatment progresses, and (ii) increased monitoring and awareness of VTE in the prophylaxis group (2011–2018) than in the control group (2001–2009)." |
| Bias in selection of the reported result | First‐time symptomatic venous thromboembolism | Serious risk | No pre‐registered protocol or statistical analysis plan was available. The main outcome measurements were described in the methods section and reported in the results section in agreement with an a priori plan. However, potential predictors, hence risk factors, of venous thromboembolism including ALL immunophenotype and body weight ≤ vs > 80 kg were mentioned only in the results section and abstract; the former yielding statistical significance. |
| Overall bias | First‐time symptomatic venous thromboembolism | Serious risk | Based on the serious risk of bias due to confounding and in selection of the reported result. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | First‐time symptomatic venous thromboembolism | Critical risk | Baseline confounders not controlled for. Conference abstract—no information reported about baseline characteristics between the enoxaparin prophylaxis and the no‐prophylaxis groups or presence of confounders such as age, sex, risk factors of venous thromboembolism (infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, smoking, and obesity) or co‐morbidities. No method for dealing with potential confounders. No participant had prior venous thromboembolism. Participants (41%) received cryoprecipitate (if fibrinogen <100mg/dL); unknown distribution between the two groups—disentangling this effect from that of enoxaparin prohylaxis was not considered. Additionally, different ALL protocols including different type and dosing of asparaginase were used, quote: "88% percent receivedpediatric‐typechemoregimen andintramuscularpeg‐asparaginase; the most common doses ofpeg‐asparaginase included2000 mg/m2 (41%) and 2500 mg/m2 (35%)." Intrathecal chemotherapy and steroid use highly likely varied too (not reported). Sparse information regarding the control group, quote: "This is asingle‐centerstudy of 17 consecutive adult patients (>18 years old) treated for ALL withpeg‐asparaginasecontaininginduction regimen before and after the establishment of institutional policy touse enoxaparin prophylaxis." No information reported on the time period; however, unknown if a substantial change in supportive care or awareness/detection of venous thromboembolism could have been present even though only 17 consecutive people were included—indicating a rather short period of time. |
| Bias in selection of participants into the study | First‐time symptomatic venous thromboembolism | Moderate risk | Retrospective design; only a conference abstract was available. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that start of follow‐up and start of intervention coincided for all participants. Quote: "This is a single‐centerstudy of 17 consecutive adult patients (>18 years old) treated for ALL with peg‐asparaginase containinginduction regimen before and after the establishment of institutional policy touse enoxaparin prophylaxis. Patients were identified from hospital research database." It appeared that all potentially eligible participants for the specified study years have been included. |
| Bias in classification of interventions | First‐time symptomatic venous thromboembolism | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | First‐time symptomatic venous thromboembolism | Low risk | It appeared that all participants in the intervention group received the intended thromboprophylaxis; however, only a conference abstract was available. |
| Bias due to missing data | First‐time symptomatic venous thromboembolism | No information | All data appear to be reported; however, only a conference abstract was available. Clarifications of selected data have been obtained from study author correspondence. |
| Bias in measurement of outcomes | First‐time symptomatic venous thromboembolism | Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). Imaging confirmation not described. Quote: "Medical records were reviewed for the occurrence of any clinically relevant bleeding and VTE within 3 months of receiving peg‐asparaginase." |
| Bias in selection of the reported result | First‐time symptomatic venous thromboembolism | Moderate risk | Only a conference abstract was available. No pre‐registered protocol or statistical analysis plan were available; however, the reported outcomes as described in the methods section were consistent with an a priori plan. |
| Overall bias | First‐time symptomatic venous thromboembolism | Critical risk | Based on the critical risk of bias due to confounding. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors's judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding |
| Critical risk | Given that evaluation of prophylactic antithrombin concentrates for prevention of venous thromboembolism was not a study aim, baseline confounders not controlled for, and no method for dealing with potential confounders was presented. Baseline characteristics presented and were similar between groups regarding age (median age of 32.5 years (range 20 to 58) in the antithrombin group and 32 years (range 21 to 57) in the no‐antithrombin group) but with unequal sex distribution (m/f: 2/2 in the antithrombin group and 8/1 in the no‐antithrombin group). No apparent difference in ALL immunophenotype between the groups was present (B‐/T‐cell ALL: 2/2 in the antithrombin group and 4/5 in the no‐antithrombin group), and all included participants had Philadelphia chromosome‐negative ALL. The distribution of peripherally central venous catheter was uneven between the groups (2/4 in the antithrombin group and 6/9 in the no‐antithrombin group). Additional confounders such as prior venous thromboembolism, infections, oral contraceptives, inherited thrombophilia traits, or smoking) and co‐morbidities were not reported for the two groups. The participants received asparaginase‐based induction chemotherapy according to three different ALL regimens including either Erwinia chrysanthemi‐derived (n = 11) or native Escherichia coli‐derived (n=2) L‐asparaginase as well as different steroid dosing. Two of the four participants in the antithrombin group received Escherichia coli‐derived asparaginase. No information on intrathecal chemotherapy. The study authors note the highly heterogeneous cohort limiting a comparison‐‐quote: "Even if we cannot make any direct comparison between Erwinia and E. coliasparginasesin terms of efficacy and toxicity profiles from our series, we could suggest that remission induction chemotherapy regimens with Erwinia asparaginase in first line may be feasible in adult patients." Importantly, the allocation to prophylactic treatment with antithrombin concentrates was dependent on the antithrombin activity levels, as an antithrombin level < 70% qualified for treatment with antithrombin concentrates. Hence, the no‐antithrombin group solely comprises participants with antithrombin levels > 70% and cannot be defined as a valid comparator group. Furthermore, the co‐interventions including fibrinogen and fresh frozen plasma were unequally administrated between the groups based on fibrinogen levels and hepatic insufficiency, respectively. Quote: "The antithrombin (AT) level was < 70% (median AT level 100%, range 62 – 131%) in two out of 11 patients (18.2%) receiving Erwinia asparaginase and in two of two cases (100%) receiving E. coli asparaginase, whereas the fibrinogen level was < 100 mg/dL (median fibrinogen level 217 mg/dL, range 70 – 887 mg/dL) in eight out of 11 cases (72.7%) receiving Erwinia asparaginase, and was invariably > 100 mg/dL in the two patients treated with E. coli asparaginase." Quote: "Prophylactic AT(50 IU/kg) and fibrinogen (20 mg/kg) concentrate supplementations were administered for an AT level < 70% and fibrinogen level < 80 mg/dL, respectively." Quote: "Two of the 11 patients (18.2%) treated with Erwinia asparaginase, namely patients 3 and 10, received prophylactic AT and fibrinogen concentrates, whereas in the two patients receiving E. coli asparaginase, only AT concentrate supplementation was provided." Quote: "Of note, fresh frozen plasma (15 mL/kg) was also infused to patient 10 on day + 26 due to hepatic insufficiency." The study authors showed no attempt trying to disentangle the effect of co‐interventions from that of antithrombin concentrates, as this was not the primary study aim. The study retrospectively evaluated people treated between 2006‐2013. During a time period of seven years a substantial change in supportive care or venous thromboembolism awareness/detection could have been present. The follow‐up time included the time with administration of therapeutic anticoagulation for participants with development of venous thromboembolism, potentially influencing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Given that evaluation of prophylaxis with antithrombin concentrates was not a primary study aim, we assume that participants were selected into the study regardless of prophylaxis data. It appeared that all potentially eligible participants for the specified study period were included. Quote: "We retrospectivelyanalyzed48 adultsaffectedwith ALL, consecutively observed at our institution between January 2006 and December 2013." Selection into the study was based on retrospective collection of information regarding participants' characteristics (Philadelphia chromosome status); Philadelphia chromosome status may be related to the risk of venous thromboembolism. Yet, the study authors argue that asparaginase often is excluded from the Philadelphia chromosome‐positive ALL treatment. Quote: "The patients with Philadelphia‐positive ALL received remission induction regimens based on tyrosine kinase inhibitors combined Excluded participants accounted for in detail. Quote: "Thirty‐three patients did not receive asparaginase because of either Philadelphia‐positivity (21 patients) or advanced age (12 patients aged > 60 years), while 13 patients of median age 32 years (range 20 – 58 years) received either Erwinia asparaginase (11 patients) or E. coli asparaginase (two patients) during remission induction treatment, according to three different regimens (Table I)." It appeared that start of follow‐up and start of intervention coincided for all participants. |
| Bias in classification of interventions
|
| Moderate risk | Intervention was not defined in the methods paragraph (format of study reference: letter), as evaluation of thromboprophylaxis with antithrombin concentrates was not the primary aim of the study. However, preset criteria for antithrombin substitution as part of clinical practice seemed to be present at the start of intervention: Quote: "Prophylactic AT(50 IU/kg) and fibrinogen (20 mg/kg) concentrate supplementations were administered for an AT level < 70% and fibrinogen level < 80 mg/dL, respectively." Assignments of intervention status were determined retrospectively. |
|
| Serious risk | ||
| Bias due to deviations from intended intervention | All outcomes | No information | No information reported on deviations from the prophylaxis with antithrombin concentrates, as this was not the primary study aim. |
| Bias due to missing data | All outcomes | Low risk | Data appear to be reasonably complete. Clarifications of selected relevant outcome data have been obtained by author correspondence. |
| Bias in measurement of outcomes
|
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
|
| Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). Systematic imaging confirmation not described. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available (study reference format: letter); however, the reported outcomes seemed in agreement with the scope (safety of asparaginase therapy) as outlined in the methods paragraph consistent with an a priori plan. Of note, evaluation of prophylactic antithrombin concentrates was not the primary study aim. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding, as the control group is not a valid comparator group. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for, and no method for dealing with potential confounders presented. Baseline characteristics shown and were similar between antithrombin and no‐antithrombin groups in regard to age; 10% more male sex in the no‐antithrombin group (reported as non‐significant). Potential confounders such as risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, central venous catheters, inherited thrombophilia traits, smoking, and obesity) and co‐morbidities not reported. A similar proportion of people between antithrombin and no‐antithrombin groups had hormonal therapy use. Participants received concomitant enoxaparin prophylaxis, fresh frozen plasma, and cryoprecipitate supplementation. Quote: "Cryoprecipitateuse was significantly higher in the AT group (55%) than the control group (11%) (p < .001). Fresh frozen plasma use was not significantly different between the two groups." A total of 26% (12/47) participants in the antithrombin group and 7%(2/28) participants in the no‐antithrombin group received enoxaparin prophylaxis. No adjustments for these co‐interventions were done. Quote: "Univariate analysis showed that the use of prophylactic anticoagulation,cryoprecipitate, or fresh frozen plasma did not have a significant effect on the incidence of VTE." Quote: "Because there is nostandardizedpractice for AT or fibrinogen replacement in these patients, our data implies that physicians who monitor AT levels for replacement were also more likely to order fibrinogen levels and prescribecryoprecipitatefor low fibrinogen levels (generally <100 mg/dL)." Study authors recognised cryoprecipitate as a potential confounder, quote: "This study did not find a benefit in bleeding risk withcryoprecipitateuse, and the higher frequency ofcryoprecipitateuse in the AT group may have attenuated any beneficial effects from AT supplementation. This may explain why a larger proportion of thrombotic events were observed in the AT group although this difference was not significant." Participants in the antithrombin and the no‐antithrombin groups were treated according various ALL regimens, all containing pegylated asparaginase at different doses. Quote: "Patients in the AT group received a median of two doses of PEG‐Asp compared to three doses in the control group (p = .04). The dose of PEG‐Asp in the AT group was also marginally lower, although this difference was not significant." Intrathecal chemotherapy and steroid use highly likely varied too (not reported); the latter acknowledged by the study authors, quote: "Another limitation is that the specific corticosteroid (i.e. dexamethasone vs. prednisone) used by patients in each group was not accounted for." |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that start of follow‐up and start of intervention coincided for all participants. Quote: "...electronic medical records of 116 patients who received PEG‐Asp at City of Hope MedicalCenterbetween 27 April 2014 and 31 October 2017 were reviewed. Inclusion criteria were of age 18 years, a diagnosis of ALL or T‐celllymphoblasticlymphoma warranting treatment with a PEG‐Asp‐containing regimen, and receipt of at least one dose of PEG‐Asp during any phase of their ALL regimen (i.e. induction, consolidation, maintenance, salvage). Patients were assigned to the AT group if a physician monitored AT levels with the intention to replace if levels were low (i.e. <60%). All other patients were assigned to the control group." It appeared that all potentially eligible participants for the specified study years were included. Selection into groups was based on the intention‐to‐treat principle. |
| Bias in classification of interventions | All outcomes | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | All outcomes | Moderate risk | The study authors report deviations from the intended intervention, for which they have no explanation. Quote: "Of the remaining 10 patients in the AT group who did not receive AT, none of them experienced a VTE event, and 6 of them maintained an AT level above 60%. For unknown reasons, three of these patients did not receive AT when levels were below 60%, and one patient did not have AT levels drawn." |
| Bias due to missing data | All outcomes | Low risk | Data appeared to be reasonably complete. Clarifications of outcome data obtained from study author correspondence. |
| Bias in measurement of outcomes | Venous thromboembolism‐related mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
| First‐time symptomatic venous thromboembolism | Moderate risk | Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available; however, the outcome measurements as described in detail in the methods section and the analyses were consistent and in agreement with an a priori plan. Analysis was presented by intention‐to‐treat in the results and abstract sections (10 participants in the intervention group were intended to receive AT supplementation but did not; none of them developed venous thromboembolism). No results (raw numbers) were reported in regard to the univariate analysis evaluating the impact of pre‐selected variables on the incidence of venous thromboembolism, quote: "Univariate analysis showed that the use of prophylactic anticoagulation,cryoprecipitate, or fresh frozen plasma did not have a significant effect on the incidence of VTE." |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding |
| Critical risk | Given that evaluation of prophylactic antithrombin concentrates for prevention of venous thromboembolism was not the primary study aim, baseline characteristics were not presented for the participants receiving antithrombin concentrates and those who did not. Baseline characteristics presented for participants with and without development of venous thromboembolism, respectively. Baseline confounders not controlled for, and no method for dealing with potential confounders was presented. Most participants had central venous catheters, quote: "Surgically placed Hickman catheters provided central access for 98%." Additional confounders such as prior venous thromboembolism, infections, oral contraceptives, inherited thrombophilia traits, or smoking) and co‐morbidities were not reported for the antithrombin and no‐antithrombin groups. The participants with newly diagnosed Philadelphia chromosome‐positive or ‐negative ALL received native Escherichia coli‐derived asparaginase‐based induction chemotherapy according to two different ALL regimens including different asparaginase, steroid, and intrathecal chemotherapy dosages. Importantly, it is unclear whether antithrombin activity was monitored for all participants and whether all participants who had antithrombin activity levels <70%/≥ 70% did/did not receive antithrombin concentrates, respectively (and thus comprise the comparison group). Quote: "Monitoring of coagulation parameters (PT, APTT, Fb), AT activity (Chromogenix,Coamatic, Antithrombin Kit) and actor replacement was at the discretion of the treating physician." Quote: "Median baseline (prior to ASP) AT activity (normal: 80 – 120%) was 94% (77 – 151%). Median nadir AT during ASP therapy was 47% (30 – 90%), significantly different from baseline (P<0.0001)." Quote: "The practice of ‘‘prophylactic’’ AT replacement varied amongst physicians and AT was administered to 17 patients when AT activity fell to < 70%." Furthermore, the co‐interventions included fresh frozen plasma or cryoprecipitate (if fibrinogen < 100 mg/dL) and platelet transfusions (if <10 x109/L in the absence of fever (< 20 x109/L), bleeding or anticoagulation therapy (<30 x109/L)). The study authors showed no attempt trying to disentangle the effect of co‐interventions from that of antithrombin concentrates (but rather in regard to thrombosis development), as this was not the primary study aim. Quote: "The incidence of TE was significantly lower among patients who received AT (0/17) than those who did not (10/37, P=0.021). FFP or CPT (given for Fb 5100 mg/dl) had no significant effect on incidence of TE and there was no significant difference in factor replacement administered to either group." Quote: "For appropriate comparison of the platelet count during ASP therapy, in those with and those without TE, the platelet count on the day of maximal ASP effect (time of maximal nadir of AT and/ or Fb) was taken as a time point for comparison between the 2 groups. The median platelet counts (range) at this time in those with and those without TE were 136 x 109/l (28 – 223) and 64 x 109/l (9 – 274), respectively. As can be seen from these platelet counts, platelet transfusions were generally not indicated in this population at the time of ASP therapy. There was no observed effect of platelet transfusions on subsequent TE events. The impact of the higher platelet count observed in those with TE is unknown." The study retrospectively evaluated people treated between 1994‐2003. During a time period of nine years a substantial change in supportive care or venous thromboembolism awareness/detection could have been present. No information reported on follow‐up that potentially included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Given that evaluation of prophylaxis with antithrombin concentrates was not a primary study aim, we assume that participants were selected into the study regardless of prophylaxis data. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that all potentially eligible participants for the specified study period were included. Quote: "This study only included adults with newly diagnosed ALL receiving ASP (a component of 2 contemporary regimens used as standard care at our institution) during induction. Patients with relapsed ALL or lymphoidblastictransformation of chronicmyelogenousleukemia were excluded." Quote: "Ten of 54 (18.5%) consecutive adults developed symptomatic, objectively confirmed TE.." It appeared that start of follow‐up and start of intervention coincided for all participants. |
| Bias in classification of interventions | All outcomes | Serious risk | The intervention was unclearly described in the methods as well as results sections. Of note, evaluation of thromboprophylaxis with antithrombin concentrates was not the primary aim of the study. Quote (methods section): "Monitoring of coagulation parameters (PT, APTT, Fb), AT activity (Chromogenix,Coamatic, Antithrombin Kit) and factor replacement was at the discretion of the treating physician. AT replacement consisted of Thrombate III1 Bayer AG,Leverkusen, Germany)." Quote (results section): "The practice of ‘‘prophylactic’’ AT replacement varied amongst physicians and AT was administered to 17 patients when AT activity fell to <70%." Quote: "Although not all patients had a baseline value to enable statistical comparison, the majority had levels of AT and fibrinogen assessed during the course of ASP therapy to allow comparison of these parameters in those with and without TE events." Unclear whether lack of antithrombin activity monitoring affected the classification of intervention. Assignments of intervention status were determined retrospectively. |
| Bias due to deviations from intended intervention | All outcomes | No information | No information reported on deviations from the prophylaxis with antithrombin concentrates, as this was not the primary study aim. |
| Bias due to missing data | All outcomes | Serious risk | Unclear what data were missing in which participants. Quote: "Although not all patients had a baseline value to enable statistical comparison, the majority had levels of AT and fibrinogen assessed during the course of ASP therapy to allow comparison of these parameters in those with and without TE events." |
| Bias in measurement of outcomes
|
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
|
| Moderate risk | Given the retrospective design, blinding of outcome assessors was not done. Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All thrombotic events were confirmed by imaging. Part of the assessment of major bleeding events was likely based on symptoms (reported by participants) and clinical signs (assessed by clinicians). | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available. However, the outcome measurements as described in the methods section and the analyses were consistent and in agreement with an a priori plan. Of note, evaluation of prophylactic antithrombin concentrates was not the primary study aim. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding, as the control group is likely not a valid comparator group. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Serious risk | Baseline confounders controlled for, quote: "Cox regressionmodelingwas performed and included whether patient was managed using the AT replacement strategy, age, sex, presence of a CVC, WBC at diagnosis, steroid used, ALL trial protocol, type ofasparaginase, and whether or not FFP was given. In this model, only management using the AT replacement strategy was statistically significantly associated with incidence of thrombosis." Data not presented. Baseline characteristics shown and were similar between antithrombin and no‐antithrombin groups in regard to age and sex. Unequal distribution of central venous catheters (7/30 in the antithrombin group vs 2/15 in the no‐antithrombin group, P = 0.7) and fresh frozen plasma use (4/30 in the antithrombin group vs 11/15 in the no‐antithrombin group, P<0.001). Other potential confounders such as risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, oral contraceptives, inherited thrombophilia traits, smoking, and obesity), and co‐morbidities not reported and adjusted for. Participants received treatment according to different ALL regimens. Consequently, differences in asparaginase formulation and dosing were reported (pegylated asparaginase treatment in 19/30 in the antithrombin group vs 3/15 in the no‐antithrombin group. Quote: "More doses were required in patients who had received PEGylatedasparaginase(median of three doses; range 0–11 doses) as compared with those who received nativeasparaginase(median one dose, range 0–9). Four of 11 patients (36%) who received nativeasparaginasedid not require any AT replacement, as compared with two of 19 (10.5%) in those receiving PEGylatedasparaginase." Also a difference in corticosteroid use were observed (dexamethasone use in 20% (3/15) in the no‐antithrombin group vs 73% (22/30) in the antithrombin group, P<0.01; prednisolone use in 80% (12/15) in the no‐antithrombin group vs 30% (9/30) in the antithrombin group, P<0.01). No report of intrathecal chemotherapy. The study authors acknowledged the potential confounders, quote: "It is unlikely that transitions to different ALL trial protocols (between the observation and replacement cohorts) have influenced this dramatic change in VTE incidence, as other than the corticosteroid formulation used with induction, the drugs used in remission induction have not changed substantially. As a result of transition between various trials over the study period, the proportion of patients receiving standard compared to PEGylatedasparaginasediffered between the two groups, however as noted above a published study showed no apparent difference in thrombotic risk between these formulations." The study was conducted from 2005‐2013; including a historical control group (2005‐2009) and an intervention group (2009‐2013) after implementation of the antithrombin replacement protocol in 2009. Quality in supportive care and awareness/detection of venous thromboembolism were likely to have changed over this period of time. Follow‐up time included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality. |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Selection into the study did not appear to be related to intervention, outcome or any prognostic factor. It appeared that start of follow‐up and start of intervention coincided for all participants. Quote: "Forty‐five consecutive patients with a diagnosis of ALL treated withasparaginase‐containing phase I induction protocols were included in this retrospective cohort observational study." Quote: "The first 15 historical patients (2005–2009) received standard therapy with no AT replacement but thrombotic events were recorded. (...) Subsequently, a protocol of AT replacement was instituted (described below) and delivered to the subsequent 30 patients (2009–2013)." It appeared that all potentially eligible participants for the specified study period have been included. |
| Bias in classification of interventions | All outcomes | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | All outcomes | Low risk | The reported deviations from the intended intervention reflect usual practice. Quote: "In seven patients, AT levels did not fall below 70 iu/dl and they therefore did not receive AT replacement." |
| Bias due to missing data | All outcomes | Low risk | Data were reasonably complete; type of ALL protocol received was missing in one participant. Clarifications of outcome data obtained from study author correspondence. |
| Bias in measurement of outcomes |
| Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, none of the outcomes were subjective. |
| First‐time symptomatic venous thromboembolism | Moderate risk | Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). All venous thromboembolic events were confirmed by imaging. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available; however, the venous thromboembolism outcome measurements as described in the methods section and the analyses were consistent and in agreement with an a priori plan. The mortality outcomes were not mentioned in the methods section but reported in the results section. Analysis was presented by intention‐to‐treat in the results and abstract sections (7 participants in the intervention group were intended to receive antithrombin supplementation but did not; none of them developed venous thromboembolism). |
| Overall bias | All outcomes | Serious risk | Based on the serious risk of bias due to confounding. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | Venous thromboembolism‐related mortality | Critical risk | Baseline confounders not controlled for. Conference abstract—no method dealing with potential confounders presented. Baseline characteristics presented and were similar regarding sex. The mean age was 39 years (range 21 to 73) in the antithrombin group and 32 years (range 19‐63) in the no‐antithrombin group. Unclear reporting of body mass index (2.01 in the no‐antithrombin group and 1.84 in the antithrombin group, unit?). No information reported about presence of risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, and smoking), co‐morbidities, or co‐interventions. Participants in the antithrombin and the no‐antithrombin groups were treated according to various different ALL regimens including different type and dosing asparaginase (total total doses of asparaginase was 66 in the no‐antithrombin group and 46 in the antithrombin group; 91% received pegylated asparaginase in the no‐antithrombin group and 100% in the antithrombin group). No reporting of the type/dosing of corticosteroids and intrathecal chemotherapy. The study was conducted from 2011‐2018; including a historical control group (2011‐2014) and an intervention group (2014‐2018) after implementation of the antithrombin replacement protocol in 2014 (our interpretation). Quality in supportive care and awareness/detection of venous thromboembolism were likely to have changed over this period of time. |
| Bias in selection of participants into the study | Venous thromboembolism‐related mortality | Moderate risk | Retrospective design; only a conference abstract was available. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that all potentially eligible participants for the specified study period were included. Limited information on selection of participants, quote: "This retrospective study evaluated patients 18 years of age or older who received L‐asparaginase,pegaspargase, orasparaginaseErwiniachrysanthemibetween January 2011 and June 2018 for ALL,lymphoblasticlymphoma, or NK/T‐cell lymphoma." Quote: "6 patients did not receive AT monitoring while 20 patients were monitored and received AT repletion." It appeared that start of follow‐up and start of intervention coincided for all participants. |
| Bias in classification of interventions | Venous thromboembolism‐related mortality | Moderate risk | Intervention was not clearly defined in the methods (or results) section; only a conference abstract was available. Some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | Venous thromboembolism‐related mortality | Low risk | It appeared that all participants in the intervention group received the intended thromboprophylaxis; however, only a conference abstract was available. |
| Bias due to missing data | Venous thromboembolism‐related mortality | Low risk | All data appeared to be reported; a conference abstract with supplemental tables were available. |
| Bias in measurement of outcomes | Venous thromboembolism‐related mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the reported outcome of interest was objective. |
| Bias in selection of the reported result | Venous thromboembolism‐related mortality | Moderate risk | Only a conference abstract was available. No pre‐registered protocol or statistical analysis plan were available; however, the outcome was described in the methods section consistent with an a priori plan. |
| Overall bias | Venous thromboembolism‐related mortality | Critical risk | Based on the critical risk of bias due to confounding. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Bias | Outcome | Authors' judgement | Support for judgement |
|---|---|---|---|
| Bias due to confounding | All outcomes
| Critical risk | Baseline confounders not controlled for. Correspondence—no method dealing with potential confounders presented. Baseline characteristics not similar between groups with median age of 33 years (range 19 to 77) in the antithrombin group and 38 years (range 20 to 75) in the no‐antithrombin group; 63% versus 72% male sex in the antithrombin and no‐antithrombin group, respectively. Slightly more B‐cell immunophenotype in the antithrombin group and T‐cell immunophenotype in the no‐antithrombin group. No information reported about potential confounders such as presence of risk factors of venous thromboembolism (prior venous thromboembolism, infections, immobilisation, central venous catheters, oral contraceptives, inherited thrombophilia traits, obesity, and smoking) or co‐morbidities. Only participants in the antithrombin group underwent monitoring of fibrinogen levels and received cryoprecipitate (if fibrinogen levels <100 mg/dl or <150 mg/dl with suspected bleeding); in total, 31 of 43 participants in the antithrombin group received fibrinogen supplementation. Participants in the antithrombin and the no‐antithrombin groups were treated according to different ALL regimens containing pegylated asparaginase, not further described. Thus, type and dosing of intrathecal chemotherapy, corticosteroids, and asparaginase potentially vary (not reported). Quote: "Patients were treated with variousasparaginasecontaining regimens including apediatricbased regimen." The study was conducted from 2009‐2019. Participants before and after implementation of an antithrombin activity monitoring and supplementation practice in 2012, respectively, were enrolled. Quality in supportive care and awareness/detection of venous thromboembolism likely could have changed over this period of time. No information reported on follow‐up that potentially included the time with therapeutic anticoagulation post‐venous thromboembolism for participants with development of venous thromboembolism—when assessing all‐cause mortality (outcome data obtained from author correspondence). |
| Bias in selection of participants into the study | All outcomes | Moderate risk | Retrospective design. Selection into the study did not appear to be related to intervention, outcome, or any prognostic factor. It appeared that all potentially eligible participants for the specified study period were included. Limited information on selection of participants, quote: "Patients were identified using our institution’s hematological malignancy registry." Quote: "Patients in the control group were derived from a historical cohort prior to implementation of the AT supplementation protocol." It appeared that start of follow‐up and start of intervention coincided for all participants. |
| Bias in classification of interventions | All outcomes | Moderate risk | Intervention was clearly defined in the methods section; some aspects of the assignments of intervention status were determined retrospectively—but it is unlikely that this was affected by knowledge of the outcome or risk of the outcome. |
| Bias due to deviations from intended intervention | All outcomes | Moderate risk | Limited information. 27/43 participants received the intended intervention in the antithrombin group, likely due to antithrombin activity levels above 50% (but not written directly in text). |
| Bias due to missing data | All outcomes | Low risk | All data appeared to be reported. |
| Bias in measurement of outcomes | All‐cause mortality | Low risk | Given the retrospective design, blinding of outcome assessors was not done. However, the outcome was objective. |
| First‐time symptomatic venous thromboembolism | Moderate risk | Venous thromboembolism‐related symptoms were likely both subjective (assessed by the participant) and objective (clinical signs assessed by the responsible clinician). Imaging confirmation not described. | |
| Bias in selection of the reported result | All outcomes | Moderate risk | No pre‐registered protocol or statistical analysis plan were available; however, the venous thromboembolism outcome was described in the methods section consistent with an a priori plan. Of note, all‐cause mortality was not part of the described or reported outcomes, and outcome data were obtained by study author correspondence. |
| Overall bias | All outcomes | Critical risk | Based on the critical risk of bias due to confounding. |
| [ROBINS‐I tool version 19, September 2016] | |||
| Study | Outcomes | Domain at critical risk of bias |
|---|---|---|
|
| Bias due to baseline confounding | |
|
| Bias due to confounding and invalid control group definition | |
|
| Bias due to baseline confounding | |
|
| Bias due to confounding and invalid control group definition | |
|
| Bias due to baseline confounding | |
|
| Bias due to baseline confounding | |
|
| Bias due to baseline confounding | |
|
| Bias due to baseline confounding | |
|
| Bias due to baseline confounding | |
|
|
| Bias due to baseline confounding |
|
| Bias due to confounding and in selection of participants into the study | |
|
| Bias due to baseline confounding |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1.1 All‐cause mortality Show forest plot | 1 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 1.1.1 Intention‐to‐treat analysis | 1 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 1.1.2 'As‐treated' analysis | 1 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 1.2 Venous thromboembolism‐related mortality Show forest plot | 1 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 1.2.1 Intention‐to‐treat analysis | 1 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 1.2.2 'As‐treated' analysis | 1 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 2.1 All‐cause mortality (intention‐to‐treat analysis) Show forest plot | 2 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 2.1.1 Non‐randomised studies at serious risk of bias | 1 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 2.1.2 Non‐randomised studies at critical risk of bias | 1 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 2.2 Venous thromboembolism‐related mortality (intention‐to‐treat analysis) Show forest plot | 3 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 2.2.1 Non‐randomised studies at serious risk of bias | 1 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 2.2.2 Non‐randomised studies at critical risk of bias (intention‐to‐treat analysis) | 2 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 3.1 All‐cause mortality Show forest plot | 3 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 3.1.1 Non‐randomised studies at critical risk of bias | 3 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 3.2 Major bleeding Show forest plot | 4 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 3.2.1 Non‐randomised studies at critical risk of bias | 4 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 3.3 Venous thromboembolism‐related mortality Show forest plot | 3 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
| 3.3.1 Non‐randomised studies at critical risk of bias | 3 | Risk Ratio (IV, Random, 95% CI) | Totals not selected | |
