Quimioterapia de dosis alta y rescate con células madre hematopoyéticas autólogas para niños con neuroblastoma de alto riesgo
Resumen
Antecedentes
A pesar del desarrollo de nuevas opciones de tratamiento, el pronóstico de los pacientes con neuroblastoma de alto riesgo sigue siendo malo; más de la mitad de los pacientes experimentan una recurrencia de la enfermedad. La quimioterapia de dosis alta y el rescate con células madre hematopoyéticas (es decir, tratamiento mielosupresor) podrían mejorar la supervivencia. Esta revisión es la segunda actualización de una revisión Cochrane publicada anteriormente.
Objetivos
Objetivo principal
Comparar la eficacia, es decir, la supervivencia general y libre de eventos, de la quimioterapia de alta dosis y el rescate de médula ósea autóloga o de células madre con la terapia convencional en niños con neuroblastoma de alto riesgo.
Objetivos secundarios
Determinar los efectos adversos (p.ej., enfermedad veno‐oclusiva del hígado) y los efectos tardíos (p.ej., trastornos endocrinos o malignidades secundarias) relacionados con el procedimiento y los posibles efectos de estos procedimientos en la calidad de vida.
Métodos de búsqueda
Se realizaron búsquedas en las bases de datos electrónicas del Registro Cochrane Central de Ensayos Controlados (Cochrane Central Register of Controlled Trials, CENTRAL) (The Cochrane Library 2014, número 11), MEDLINE/PubMed (1966 hasta diciembre de 2014) y EMBASE/Ovid (1980 hasta diciembre de 2014). Además, se realizaron búsquedas en las listas de referencias de los artículos pertinentes y las actas de congresos de la International Society for Paediatric Oncology (SIOP) (de 2002 hasta 2014), American Society for Pediatric Hematology and Oncology (ASPHO) (de 2002 hasta 2014), Advances in Neuroblastoma Research (ANR) (de 2002 hasta 2014) y la American Society for Clinical Oncology (ASCO) (de 2008 hasta 2014). Se buscaron los ensayos en curso en el registro del ISRCTN (www.isrct.com) y en el registro del Instituto Nacional de Salud (www.clinicaltrials.gov). Ambos registros fueron examinados en abril de 2015.
Criterios de selección
Ensayos controlados aleatorizados (ECA) que compararan la eficacia del tratamiento mielosupresor con el tratamiento convencional en pacientes con neuroblastoma de alto riesgo.
Obtención y análisis de los datos
Dos autores seleccionaron los estudios, extrajeron los datos y evaluaron el riesgo de sesgo de forma independiente. Cuando fue apropiado, se realizó el agrupamiento de los estudios. Se calculó el riesgo relativo (RR) y el intervalo de confianza (IC) del 95% para los resultados dicotómicos. Para la evaluación de los datos de supervivencia, se calculó el cociente de riesgos instantáneos (CRI) y el IC del 95%. Si los cocientes de riesgos instantáneos no se presentaban en el estudio, se utilizó el método de Parmar. Se utilizó un modelo de efectos aleatorios.
Resultados principales
Se identificaron tres ECA que incluían a 739 niños. Todos utilizaron una edad de un año como el punto de corte para la estratificación del riesgo previo al tratamiento. En la primera búsqueda actualizada se identificó un manuscrito en el que se informaba de datos de seguimiento adicionales para uno de estos ECA, mientras que en la segunda actualización se identificó una fe de erratas de este estudio. Hubo una diferencia estadística significativa en la supervivencia sin eventos a favor del tratamiento mielosupresor sobre la quimioterapia convencional o la ausencia de tratamiento adicional (tres estudios, 739 pacientes; CRI 0,78, IC del 95%: 0,67 a 0,90). Hubo una diferencia estadística significativa en la supervivencia general a favor del tratamiento mielosupresor sobre la quimioterapia convencional o ningún tratamiento adicional (dos estudios, 360 pacientes; CRI 0,74, IC del 95%: 0,57 a 0,98). Sin embargo, cuando se incluyeron datos de seguimiento adicionales en los análisis, la diferencia en la supervivencia libre de eventos siguió siendo estadísticamente significativa (tres estudios, 739 pacientes; CRI 0,79, IC del 95%: 0,70 a 0,90), pero la diferencia en la supervivencia general ya no fue estadísticamente significativa (dos estudios, 360 pacientes; CRI 0,86, IC del 95%: 0,73 a 1,01). El metaanálisis de la enfermedad maligna secundaria y la muerte relacionada con el tratamiento no mostró diferencias estadísticas significativas entre los grupos de tratamiento. Los datos de un estudio (379 pacientes) mostraron una incidencia significativamente mayor de efectos renales, neumonitis intersticial y enfermedades veno‐oclusivas en el grupo mielosupresor en comparación con la quimioterapia convencional, mientras que para las infecciones graves y la septicemia, no se identificaron diferencias significativas entre los grupos de tratamiento. No se presentó información sobre la calidad de vida. En los estudios individuales, se evaluaron diferentes subgrupos, aunque los resultados no fueron claros en todos los estudios. Todos los estudios tenían limitaciones metodológicas.
Conclusiones de los autores
Sobre la base de la evidencia disponible actualmente, el tratamiento mielosupresor parece funcionar en cuanto a la supervivencia sin eventos. Para la supervivencia general, actualmente no hay evidencia del efecto cuando se incluyen datos de seguimiento adicionales. No es posible establecer conclusiones definitivas con respecto a los efectos adversos y la calidad de vida, aunque debe tenerse presente la posibilidad de niveles más altos de efectos adversos. No es posible establecer una conclusión definitiva con respecto al efecto del tratamiento mielosupresor en diferentes subgrupos. Esta revisión sistemática sólo permite llegar a una conclusión sobre el concepto de tratamiento mielosupresor; no se puede llegar a ninguna conclusión sobre la mejor estrategia de tratamiento. Los ensayos futuros sobre la administración del tratamiento mielosupresor para el neuroblastoma de alto riesgo deben centrarse en la identificación del mejor régimen mielosupresor o de inducción. El diseño de estudio óptimo para responder estas preguntas es un ECA. Dichos ECA deben realizarse en poblaciones de estudio homogéneas (p.ej., estadio de la enfermedad y edad del paciente) y deben incluir un seguimiento a largo plazo. Se deben tener en cuenta diferentes grupos de riesgo, mediante el uso de las definiciones más recientes.
Se debe tener presente que recientemente, el punto de corte de la edad para las enfermedades de alto riesgo cambió de un año a 18 meses. Como resultado, es posible que los pacientes que presentaban enfermedades que ahora se clasifican como de riesgo intermedio hayan estado incluidos en los grupos de alto riesgo. En consecuencia, la relevancia de los resultados de estos estudios respecto a la práctica actual puede ser cuestionada. Pueden sobrestimarse las tasas de supervivencia debido a la inclusión de pacientes con enfermedades de riesgo intermedio.
PICO
Resumen en términos sencillos
Quimioterapia de dosis alta y trasplante de células madre en comparación con tratamiento convencional en niños con neuroblastoma de alto riesgo
A pesar del desarrollo de nuevas opciones de tratamiento, el pronóstico de los pacientes con neuroblastoma de alto riesgo sigue siendo deficiente; en más de la mitad de los pacientes la enfermedad reaparece. El rescate de células madre sustituye a las células madre formadoras de sangre que fueron destruidas por la quimioterapia de alta dosis para recuperar la médula ósea. También se conoce como tratamiento mielosupresor y podría mejorar la supervivencia de estos pacientes. Una decisión bien fundada sobre la administración del tratamiento mielosupresor en el tratamiento de los niños con neuroblastoma de alto riesgo debe basarse en evidencia de alta calidad sobre la efectividad en cuanto al tratamiento de los tumores y los efectos secundarios.
Esta revisión sistemática se centró en los estudios aleatorizados que comparan la efectividad del tratamiento mielosupresor con el tratamiento convencional en niños con neuroblastoma de alto riesgo. Los autores encontraron tres estudios que incluían a 739 pacientes. Dichos estudios aportan evidencia de que el tratamiento mielosupresor mejora la supervivencia sin eventos (o sea, el tiempo hasta que ocurre un determinado evento como por ejemplo, la progresión del tumor, el desarrollo de un segundo tumor o la muerte por cualquier causa). En cuanto a la supervivencia general (o sea, el tiempo hasta que el paciente muere por cualquier causa, no sólo a causa del tumor o del tratamiento, sino también, por ejemplo, debido a un accidente automovilístico), no existe evidencia de un mejor resultado en los pacientes que reciben tratamiento mielosupresor. Los efectos secundarios como los efectos renales (riñón), la neumonitis intersticial (un tipo de enfermedad pulmonar) y la enfermedad veno‐oclusiva (un trastorno en el que se observa la obstrucción de algunas de las venas pequeñas del hígado) fueron más frecuentes en los pacientes que recibieron tratamiento mielosupresor en comparación con la quimioterapia convencional. Debe señalarse que esta revisión sistemática sólo permite establecer una conclusión sobre el concepto del tratamiento mielosupresor; no se pudo llegar a ninguna conclusión sobre la mejor estrategia de tratamiento con respecto, por ejemplo, a los tipos de agentes quimioterapéuticos y el uso de la radioterapia. Se necesitan más estudios de investigación de alta calidad.
Se debe señalar que recientemente, el punto de corte para la edad para las enfermedades de alto riesgo cambió de un año a 18 meses. Como resultado, es posible que los pacientes que presentaban enfermedades que ahora se clasifican como de riesgo intermedio estuviesen incluidos en los grupos de alto riesgo. En consecuencia, la relevancia de los resultados de estos estudios respecto a la práctica actual puede ser cuestionada. Pueden sobrestimarse las tasas de supervivencia debido a la inclusión de pacientes con enfermedades de riesgo intermedio.
Authors' conclusions
Background
Description of the condition
Neuroblastoma is the most common extracranial solid tumour in children comprising 8% to 10% of all childhood cancers. The incidence is nearly 10 per 1,000,000 children under the age of 15 years and 90% of cases are diagnosed in the first 10 years of life. Children with neuroblastoma mostly present with abdominal disease and more than half of the cases have advanced disease (defined as stage III or IV disease) (Aydin 2009; Brodeur 2006; Goldsby 2004).
Neuroblastoma is one of the most challenging and enigmatic neoplasms of childhood because of its biological heterogeneity and contrasting patterns of clinical behaviour. Some young infants with favourable disease may experience complete spontaneous regression while older children with metastatic disease mostly relapse despite initial response to chemotherapy.
The prognosis and management of children with neuroblastoma is highly dependent on clinical, histopathological and biological characteristics, and they are stratified into risk groups based on prognostic factors (Goldsby 2004; Maris 2005; Maris 2007; Weinstein 2003). For pre‐treatment risk stratification traditionally an age of one year was used as a cut‐off point. The low‐risk disease group included cases younger than one year who had stage I, II or IV‐S disease with favourable histopathology and no MYCN oncogene amplification (Brodeur 2006; Goldsby 2004; Maris 2005; Maris 2007; Weinstein 2003).
High‐risk neuroblastoma cases were traditionally characterised by an age older than one year, disseminated disease, MYCN oncogene amplification and unfavourable histopathologic findings (Brodeur 2006; Goldsby 2004; Maris 2005; Weinstein 2003). However, in recent years a new neuroblastoma risk classification system has been developed by the International Neuroblastoma Risk Group (INRG) Task Force resulting in standardised approaches for the initial evaluation and treatment stratification of neuroblastoma patients (Cohn 2009). In this INRG classification system an age cut‐off of 18 months is used for pre‐treatment risk stratification, as opposed to the traditional cut‐off of one year. In many of the currently published studies patients with what is now understood to be intermediate‐risk disease were classified as high‐risk patients. Consequently the relevance of the results of these studies to current practice can be questioned.
Description of the intervention
Substantial improvement has been achieved in the cure of patients with low‐risk neuroblastoma resulting in survival rates up to 90% (Brodeur 2006; Goldsby 2004). In high‐risk cases, despite intensified combination chemotherapies, surgery, radiotherapy and the use of differentiation agents, prognosis improved only modestly with long‐term survival in less than one‐third of patients (Brodeur 2006; De Bernardi 2003; Goldsby 2004; Weinstein 2003; Peinemann 2015a). In the last two decades higher remission rates have been achieved with intensive induction chemotherapy regimens combined with surgical resection, external irradiation or both (Brodeur 2006; Castel 1995; Goldsby 2004; Kaneko 2002; Kushner 2004; Laprie 2004; Sawaguchi 1990). The effects of retinoic acid post consolidation therapy for high‐risk neuroblastoma patients treated with autologous hematopoietic stem cell transplantation are currently not clear (Peinemann 2015b).
The challenge is to maintain remission since more than half of patients with high‐risk disease develop systemic disease recurrence with or without a relapse at the primary tumour site (Brodeur 2006; Goldsby 2004; Matthay 1993; Weinstein 2003). AntiGD2 antibody‐based immunotherapy has been shown to improve event‐free and overall survival in high‐risk neuroblastoma patients in remission after multimodality treatment (Yu 2010). Therapy failures are mostly attributed to development of resistance to chemotherapy and minimal residual disease is considered an important cause of recurrence (Burchill 2004; Keshelava 1998; Reynolds 2001; Reynolds 2004).
The idea that further increasing dose intensity may overcome chemotherapy resistance has provided a rationale for aggressive high‐dose chemotherapy consolidation protocols (Cheung 1991; Pritchard 1995). Such myeloablative chemotherapy regimens utilise effective high‐dose drug combinations which can be safely escalated to levels above those causing bone marrow ablation. Rapid bone marrow reconstitution can be achieved by autologous stem cell rescue. Autologous peripheral blood stem cells (PBSC) are the preferred source for rescue as this has several advantages over autologous bone marrow grafts. For example, the procedure of PBSC collection is easier, the incidence of tumour cell contamination is lower, and the yield of stem cells is higher (Brodeur 2006; Cohn 1997; Ladenstein 1994). A possible limitation of using autologous products is the risk of tumour cell contamination in the graft, which has been shown to contribute to relapse. Considerable efforts have been made to detect and remove tumour (negative purging) or select progenitor cells (positive purging) before reinfusion (Ladenstein 2004). Disease status prior to stem cell rescue has a crucial influence on final outcome. Patients in complete, very good partial or partial remission have a better prognosis, while those with stable disease or no response have a poor outcome (Ladenstein 2004).
Patients undergoing stem cell rescue may experience toxicities related to conditioning regimens like veno‐occlusive disease of the liver or haemorrhagic cystitis (Bollard 2006). Growth failure, endocrine disorders such as gonadal or thyroid dysfunction, hearing impairment, renal impairment, orthopedic complications as well as the occurrence of secondary malignancies are among the late complications following high‐dose chemotherapy and stem cell rescue (Bollard 2006; Trahair 2007).
Why it is important to do this review
Many retrospective and prospective non‐randomised studies support the use of a myeloablative consolidation in neuroblastoma. Retrospective studies mostly suggest that intensification of consolidation therapy with autologous stem cell rescue following high‐dose chemotherapy improves survival (Castel 1995; Di Caro 1994; Matthay 1995; Philip 1997; Stram 1996; Verdeguer 2004). The results of non‐randomised pilot studies by the Children's Cancer Group also suggest a modest prolongation of event‐free survival for children with high‐risk neuroblastoma (Matthay 1995).
This is the second update of the first systematic review evaluating the current state of evidence on the efficacy of high‐dose chemotherapy and autologous haematopoietic stem cell rescue compared with conventional therapy in patients diagnosed with high‐risk neuroblastoma (Yalçin 2010; Yalçin 2013).
Objectives
Primary objective
To compare the efficacy, that is event‐free and overall survival, of high‐dose chemotherapy and autologous bone marrow or stem cell rescue with conventional therapy in children with high‐risk neuroblastoma.
Secondary objectives
To determine adverse effects (e.g. veno‐occlusive disease of the liver) and late effects (e.g. endocrine disorders or secondary malignancies) related to the procedure and possible effects of these procedures on quality of life.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) that compared the efficacy of high‐dose chemotherapy and autologous bone marrow or stem cell rescue with conventional therapy were considered for inclusion.
Types of participants
Participants included children (aged < 21 years) with high‐risk neuroblastoma. High‐risk neuroblastoma is defined as a combination of the following characteristics: unfavourable age (as defined by the authors of the included studies), advanced stage disease (as defined by the authors of the included studies), presence of MYCN oncogene amplification, or unfavourable histopathologic findings (as defined by the authors of the included studies) or both. Recently the age cut‐off for high‐risk disease was changed from one year to 18 months. As a result it is possible that patients with what is now classified as intermediate‐risk disease were included in the high‐risk group.
Types of interventions
Interventions included high‐dose chemotherapy with autologous bone marrow or stem cell rescue versus conventional therapy (i.e. either conventional chemotherapy or no further treatment). High‐dose chemotherapy is defined as chemotherapy sufficient to require stem cell rescue. Conventional chemotherapy is defined as chemotherapy at a lower dose than the high‐dose chemotherapy without the need for stem cell rescue.
Types of outcome measures
Primary outcomes
-
Event‐free survival (usually defined as the time to recurrence or progression of disease or death from any cause or varying definitions, as used by the authors).
-
Overall survival (defined as the time to death from any cause).
Secondary outcomes
-
Adverse events (e.g. stomatitis, diarrhoea, fatigue, anaemia, leukopenia, thrombocytopenia or veno‐occlusive disease of the liver) and late effects (e.g. endocrine disorders or secondary malignancies).
-
Quality of life.
Search methods for identification of studies
We searched the following electronic databases: the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2014, issue 11), MEDLINE/PubMed (from 1966 to December 2014) and EMBASE/Ovid (from 1980 to December 2014). The search strategies for the different electronic databases (using a combination of controlled vocabulary and text word terms) are shown in Appendix 1, Appendix 2 and Appendix 3.
We located information about trials not registered in CENTRAL, MEDLINE or EMBASE, either published or unpublished, by searching the reference lists of relevant articles and review articles. We also scanned the conference proceedings of the International Society for Paediatric Oncology (SIOP) (from 2002 to 2014), the American Society for Pediatric Hematology and Oncology (ASPHO) (from 2002 to 2014), Advances in Neuroblastoma Research (ANR) (from 2002 to 2014) and the American Society for Clinical Oncology (ASCO) (from 2008 to 2014), if available electronically and otherwise by handsearching. We searched for ongoing trials by scanning the ISRCTN register (www.isrct.com) and the National Institute of Health Register (www.clinicaltrials.gov). Both registers were screened in April 2015. We imposed no language restriction.
Data collection and analysis
Selection of studies
Two authors independently screened the search results to identify studies meeting the inclusion criteria. We obtained full manuscripts for any study that appeared to meet the inclusion criteria based on title, or abstract or both for closer inspection. We clearly reported reasons for exclusion for any study considered for the review. Discrepancies between authors were resolved by consensus. No third party arbitration was needed.
Data extraction and management
Two authors independently extracted data using standardised forms. We extracted the following data:
-
the characteristics of the participants at the time of randomisation (such as age, sex, stage of disease, histology, MYCN oncogene amplification, received induction treatment, received external irradiation, remission status at myeloablative treatment and stem cell rescue);
-
the characteristics of interventions (details of myeloablative therapy: drugs used, routes of delivery, dose, timing, use of total body irradiation; details of the stem cell rescue: timing and methods of cell harvest, timing of stem cell rescue, number of cells infused, contamination with tumour cells);
-
details of supportive care (such as the use of prophylactic antibiotics, growth factors, granulocyte infusions);
-
details of outcome measures as described above; and
-
length of follow up.
In case of disagreement between authors, we re‐examined the abstracts and articles and discussed these until consensus was achieved. No third party arbitration was needed.
Assessment of risk of bias in included studies
Two authors independently performed assessment of risk of bias in the included studies. For this second update we used the most recent recommendations of the Childhood Cancer Group (that is selection bias, performance bias, detection bias (for each outcome separately), attrition bias (for each outcome separately), reporting bias and other bias). We used the 'risk of bias' items and definitions of low risk, unclear risk and high risk as described in the module of the Cochrane Childhood Cancer Group (Kremer 2014), which is based on the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). All RCTs were scored using the new 'risk of bias' items. Discrepancies between authors were resolved by consensus. No third party arbitration was needed. The risk of bias in the included studies was taken into account in the interpretation of the review's results.
Measures of treatment effect
The risk ratio (RR) was calculated for dichotomous outcomes along with the corresponding 95% confidence interval (CI). For the assessment of survival, we used the generic inverse variance function of RevMan to combine logs of the hazard ratios (HR) and estimate corresponding 95% CI. We used Parmar's method if hazard ratios were not reported in the study (Parmar 1998).
Assessment of heterogeneity
We assessed heterogeneity both by visual inspection of the forest plots and by a formal statistical test for heterogeneity, i.e. the I² statistic (Higgins 2003). If there was evidence of substantial heterogeneity (I² > 50%), this was reported.
Assessment of reporting biases
In addition to the evaluation of reporting bias, as described in the Assessment of risk of bias in included studies section, we planned to construct a funnel plot to graphically ascertain the existence of publication bias. However, as a rule of thumb, tests for funnel plot asymmetry should be used only when there are at least 10 studies included in the meta‐analysis, because when there are fewer studies the power of the tests is too low to distinguish chance from real asymmetry (Sterne 2011). Since only three trials could be included in the review, we did not construct funnel plots.
Data synthesis
We entered data into Review Manager (RevMan 2014) and analysed data according to the guidelines of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). We pooled studies for meta‐analysis when appropriate. We extracted data by allocation intervention, irrespective of compliance with the allocated intervention, in order to allow for an 'intention‐to‐treat' (ITT) analysis. We used a random‐effects model for the estimation of treatment effects throughout the review. Where possible, we presented data separately for different prognostic factors such as disease status at the time of stem cell rescue, MYCN amplification status and serum lactate dehydrogenase levels. For outcomes where only one study was available, we were unable to calculate a RR if one of the treatment groups experienced no events and the Fischer’s exact test was used instead; this option is not available in RevMan and therefore we used http://graphpad.com/quickcalcs/contingency2/.
Sensitivity analysis
For all outcomes for which pooling was possible we performed sensitivity analyses for all quality criteria separately. We excluded low quality studies and studies for which the quality was unclear, and compared the results of the good quality studies with the results of all available studies.
Results
Description of studies
Results of the search
We ran searches of the electronic databases CENTRAL, MEDLINE/PubMed and EMBASE/Ovid in January 2009 for the original version of this review. This search yielded a total of 1463 references. Initial screening excluded 1427 references which clearly did not meet all the criteria for considering studies for this review. We obtained 36 references in full for closer inspection. Three studies fulfilled all the criteria for considering studies for this review and were included in the review (Berthold 2005; Matthay 1999; Pritchard 2005). We excluded the remaining 33 articles for reasons described in the Characteristics of excluded studies table.
For the first update we ran searches of CENTRAL, MEDLINE/PubMed and EMBASE/Ovid in June 2012 yielding a total of 266 references which were added to the search results from January 2009. Initial screening excluded 265 references which clearly did not meet the inclusion criteria. We obtained one full‐text study for closer inspection. It fulfilled the inclusion criteria and was thus included. This study provides additional follow‐up data for the Matthay 1999 study.
For the second update we ran searches of CENTRAL, MEDLINE/PubMed and EMBASE/Ovid in December 2014 yielding a total of 313 references. Initial screening excluded all 313 references; no studies needed to be seen in full text.
We scanned the reference lists of relevant articles and reviews and conference proceedings (i.e. SIOP, ASPHO and ASCO) and did not identify any additional eligible studies. One trial was added to the Characteristics of excluded studies table. By scanning the ANR conference proceedings we identified one study that has not yet been published in full and is awaiting further assessment (see Characteristics of studies awaiting classification table). This study provides additional follow‐up for the Berthold 2005 study. We identified an erratum on one of the included studies (Matthay 1999). No eligible ongoing trials were identified.
In summary, the total number of included RCTs was three (described in four publications and one erratum). Characteristics of included studies are summarised below (for further details see the Characteristics of included studies table). We also identified one study awaiting assessment.
Included studies
All three eligible RCTs were multi‐centre studies, two of which were based in Europe (Berthold 2005; Pritchard 2005) and one in North America (Matthay 1999). Patients were recruited between January 1982 (Pritchard 2005) and November 2002 (Berthold 2005). The total number of patients included in the three studies was 739; 370 children received high‐dose chemotherapy and stem cell rescue, whereas 369 children received either conventional chemotherapy (Berthold 2005; Matthay 1999) or no further treatment (Pritchard 2005). All trials used different induction regimens (i.e. different chemotherapeutic agents and different (cumulative) dosages) and different myeloablative treatments. In Berthold 2005 and Matthay 1999 there were also differences in treatment patients received in the study itself (for example, external radiotherapy, therapeutic MIBG‐I131, immunotherapy and retinoic acid treatment). See the Characteristics of included studies table for further details. All patients were diagnosed with high‐risk neuroblastoma. The definitions of high‐risk neuroblastoma differed between studies (See Characteristics of included studies table). Matthay 1999 included newly diagnosed patients. It was unclear if the patients in the other studies were newly diagnosed. None of the studies reported the exact age of the participants; only the number of cases above and below one year of age was reported.
Risk of bias in included studies
See the risk of bias section of the Characteristics of included studies table and Figure 1 for the exact scores per included study.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study (it should be noted that in the study describing additional follow‐up the associated risk of detection bias was thus unclear for all outcomes except overall survival)
Allocation
For evaluating selection bias we have assessed the random sequence generation and the allocation concealment. The risk of selection bias was low in two studies (Berthold 2005; Pritchard 2005), while in one study it was unclear (Matthay 1999).
Blinding
For evaluating performance bias we have assessed the blinding of participants and personnel. In two studies there was a high risk of bias (Berthold 2005; Matthay 1999), while in one study it was unclear (Pritchard 2005). However, it should be noted that due to the nature of the interventions blinding of care providers and patients was virtually impossible.
For evaluating detection bias we have evaluated the blinding of outcome assessors for all separate outcomes, with the exception of overall survival since for that outcome blinding is not relevant and the risk of detection bias was thus automatically judged as low for all three studies evaluating this outcome. All three studies evaluated event‐free survival; in two studies the risk of detection bias was unclear (Berthold 2005; Pritchard 2005), while in the other study it was low (Matthay 1999). However, for the additional follow‐up data of this study it became unclear. Two studies evaluated adverse effects; in one study the risk of detection bias was unclear (Berthold 2005), while in the other study it was low (Matthay 1999). However, again, for the additional follow‐up data of this study it was unclear if the outcome assessor evaluating adverse events was blinded.
Incomplete outcome data
For evaluating attrition bias we have assessed incomplete outcome data for all separate outcomes. Three studies evaluated both event‐free and overall survival; in all studies the risk of attrition bias was unclear. Two studies evaluated adverse effects other than treatment‐related death; in both studies the risk of attrition bias was unclear (Berthold 2005; Matthay 1999). Two studies evaluated treatment‐related death; in one study the risk of attrition bias was low (Berthold 2005), while in the other study it was high (Matthay 1999).
Selective reporting
For evaluating reporting bias we have assessed selective reporting. We defined 'all expected outcomes' as reporting on event‐free survival, overall survival and early and late adverse effects (irrespective of the number of evaluated adverse effects; for studies presenting additional follow‐up only late adverse effects were considered appropriate). In one study we judged the risk of reporting bias to be low (Matthay 1999), while in two studies it was judged to be high (Berthold 2005; Pritchard 2005).
Other potential sources of bias
For evaluating other potential sources of bias we have assessed the difference in length of follow‐up between treatment arms and inappropriate influence of funders. In all three studies there was an unclear risk of other bias. For a more detailed description of the different items see the risk of bias section of the Characteristics of included studies table.
Effects of interventions
Event‐free survival
Data on event‐free survival were extracted from three trials (Berthold 2005; Matthay 1999; Pritchard 2005) with a total of 739 patients (see Figure 2). Two of these trials included secondary malignant disease as an event (Berthold 2005; Matthay 1999), whereas the other did not (Pritchard 2005; see the Characteristics of included studies table for the exact definitions). However, since only a very small percentage of patients developed secondary malignant disease (see the description of adverse effects later on), we were able to pool the results of these three trials. There was a significant statistical difference in event‐free survival in favour of myeloablative therapy over conventional chemotherapy or no further treatment (HR 0.78, 95% CI 0.67 to 0.90, P = 0.0006). No heterogeneity was detected for this comparison (I2 = 0%).
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.1 Event‐free survival without additional follow‐up data.
The meta‐analysis of the two trials including secondary malignant disease as an event also showed a significant statistical difference in event‐free survival in favour of myeloablative therapy over conventional chemotherapy (HR 0.79; 95% CI 0.67 to 0.92, P = 0.002, 674 patients). No heterogeneity was detected for this comparison (I2 = 0%). The study not including secondary malignant disease as an event showed no significant statistical difference between myeloablative therapy and no further treatment (HR 0.73, 95% CI 0.49 to 1.07, P = 0.11, 65 patients).
Additional follow‐up data from the Matthay 1999 study have been published. The length of follow‐up was not reported, but the median follow‐up of patients alive without an event was 7.7 years (range 130 days to 12.8 years). Relapse was included as an event in the follow‐up study, as opposed to the original publication. The results of this meta‐analysis of three trials including additional follow‐up data were in line with the earlier data. There was a significant statistical difference in event‐free survival in favour of the myeloablative therapy group over conventional chemotherapy or no further treatment (HR 0.79, 95% CI 0.70 to 0.90, P = 0.0003, 739 patients). No heterogeneity was detected for this comparison (I2 = 0%). When additional follow‐up data were included the meta‐analysis of the two trials including secondary malignant disease as an event again showed a significant statistical difference in event‐free survival in favour of the myeloablative therapy group over conventional chemotherapy (HR 0.80, 95% CI 0.70 to 0.92, P = 0.001, 674 patients). No heterogeneity was detected for this comparison (I2 = 0%). The results of the analysis of the study not including secondary malignant disease as an event did not change, since no new data were available (i.e. no significant difference between the treatment groups). See Figure 3.
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.2 Event‐free survival including additional follow‐up data of Matthay 1999.
The above mentioned HRs were calculated using the complete follow‐up period of the trials. The individual studies also reported three or five‐year event‐free survival rates. Berthold 2005 and Matthay 1999 reported significantly better three‐year event‐free survival rates in the myeloablative therapy group compared to conventional chemotherapy (P = 0.0221 and P = 0.034 respectively). Pritchard 2005 found no significant statistical difference in five‐year event‐free survival between the treatment groups (P = 0.08). Additional follow‐up data from Matthay 1999 showed a significant statistical difference in five‐year event‐free survival in favour of the myeloablative therapy group (P = 0.0434).
It should be noted that the individual studies all used different starting points for event‐free survival, either years since randomisation or years from diagnosis. In the Matthay 1999 study randomisation was performed a median of 60 days after diagnosis. Pritchard 2005 did not report the time of randomisation but the decision to randomise was made five to nine months after diagnosis. In the Berthold 2005 study randomisation was performed a median of 39 days after diagnosis.
Subgroups
Subgroups are summarised descriptively (i.e. as reported in the individual studies).
Berthold 2005 performed subgroup analyses based on commonly accepted risk factors. Three‐year event‐free survival was significantly better in the myeloablative therapy group compared to conventional chemotherapy: (1) in patients who had a complete remission or very good partial remission to induction chemotherapy/before randomisation (P = 0.0083); (2) in patients with raised serum concentration of lactate dehydrogenase (P = 0.0357; cut‐off value not reported); and (3) in patients receiving antibody treatment (P = 0.0061). By contrast, no significant statistical difference in three‐year event‐free survival was identified between the treatment groups: (1) in patients who had partial remission, mixed remission or stable disease before randomisation (P = 0.1809); (2) in patients treated with retinoic acid (P = 0.9732); (3) in patients with MYCN amplification with stage I, II, III, or IV‐S disease or with stage IV disease aged younger than one year (P = 0.2183); (4) in patients with MYCN amplification (P = 0.0669); (5) in patients without MYCN amplification (P = 0.0643); (6) in patients with normal concentrations of serum lactate dehydrogenase (P = 0.3132; cut‐off value not mentioned); and (7) in patients with stage IV disease and age more than one year (P = 0.0549).
Matthay 1999 performed subgroup analyses based on stage IV disease, MYCN amplification, unfavourable histopathological findings, a serum ferritin level of at least 143 ng per millilitre, a partial response to initial chemotherapy (as compared to complete response or nearly complete response), age at diagnosis more than two years and, among stage IV patients, bone metastases at diagnosis, and the presence at diagnosis of more than 100 tumour cells per 105 normal nucleated bone marrow cells. Matthay 1999 reported that event‐free survival among patients assigned to myeloablative therapy was longer in each subgroup (univariate analysis) than that among patients assigned to conventional chemotherapy (no P values reported). This difference in event‐free survival was most pronounced among the subgroups of patients who were older than two years at diagnosis (P = 0.01) and among patients with MYCN amplification (P = 0.03). Additional follow‐up data showed no significant statistical difference between treatment groups in event‐free survival in patients with stage IV disease.
Pritchard 2005 reported significantly better five‐year event‐free survival in patients aged more than one year with stage IV disease assigned to the myeloablative therapy group (P = 0.01).
Overall survival
Data on overall survival were reported in three trials (Berthold 2005; Matthay 1999; Pritchard 2005). The results from two of these trials could be pooled (Berthold 2005; Pritchard 2005). There was a significant statistical difference in overall survival in favour of the myeloablative therapy group over conventional chemotherapy or no further treatment (HR 0.74, 95% CI 0.57 to 0.98, P = 0.04, 360 patients). No heterogeneity was detected for this comparison (I2 = 0%). See Figure 4. The other study (Matthay 1999) only provided descriptive results: overall survival was similar for both regimens (379 patients).
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.3 Overall survival without additional follow‐up data.
As opposed to the original publication, the additional follow‐up data from the Matthay 1999 study included information on overall survival. The length of follow‐up was not mentioned, but the median follow‐up of patients alive without an event was 7.7 years (range 130 days to 12.8 years). In contrast to the earlier results, the meta‐analysis including the additional follow‐up data showed no significant statistical difference in overall survival between treatment groups (HR = 0.86, 95% CI 0.73 to 1.01, P = 0.06, 739 patients). No heterogeneity was detected (I2 = 0%) (Figure 5).
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.4 Overall survival including additional follow‐up data of Matthay 1999.
The above mentioned HR is calculated using the complete follow‐up period of the trials. The individual studies also reported three‐ or five‐year overall survival rates. Berthold 2005 and Matthay 1999 found no significant statistical difference in three‐year overall survival between the treatment groups (P = 0.0875 and P = 0.87 respectively). Pritchard 2005 found no significant statistical difference in five‐year overall survival between the treatment groups (P = 0.1). Additional follow‐up data from Matthay 1999 showed no significant statistical difference in five‐year overall survival between the treatment groups (P = 0.3917). As opposed to the data provided in the additional follow‐up publication (at five years overall survival was significantly better in the myeloablative therapy group compared to the conventional chemotherapy group), in the erratum it was stated that at five years the overall survival curves were not significantly different (P = 0.08).
It should be noted that the individual studies all used different starting points for overall survival, either years since randomisation or years from diagnosis. In the Matthay 1999 study randomisation was performed a median of 60 days after diagnosis. Pritchard 2005 did not report the time of randomisation but the decision to randomise was made five to nine months after diagnosis. In the Berthold 2005 study randomisation was performed a median of 39 days after diagnosis.
Subgroups
Subgroups are summarised descriptively (i.e. as reported in the individual studies).
Berthold 2005 performed subgroup analyses based on commonly accepted risk factors. Three‐year overall survival was significantly better in the myeloablative therapy group compared to conventional chemotherapy: (1) in patients who had a complete remission or very good partial remission before randomisation (P = 0.0436), or (2) in patients receiving antibody treatment (P = 0.0221). By contrast, no significant difference in three‐year event‐free survival was identified between the treatment groups in the following subgroups: (1) in patients who had partial remission, mixed remission or stable disease before randomisation (P = 0.1311); (2) in patients treated with retinoic acid (P = 0.8918); (3) in patients with MYCN amplification with stage I, II, III or IV‐S disease or with stage IV disease aged younger than one year (P = 0.2287); (4) in patients with MYCN amplification (P = 0.1658); (5) in patients without MYCN amplification (P = 0.1921); (6) in patients with normal concentrations of serum lactate dehydrogenase (P = 0.1985; cut‐off value not reported); (7) in patients with raised serum concentration of lactate dehydrogenase (P = 0.1247; cut‐off value not reported); and (8) in patients with stage IV disease and aged more than one year (P = 0.1820).
Matthay 1999 did not report any subgroup analyses regarding overall survival in the original manuscript. Using additional follow‐up data a subgroup analysis of patients with stage IV disease found no significant statistical difference in overall survival between the treatment groups.
Pritchard 2005 reported that five‐year overall survival in patients aged more than one year with stage IV disease was significantly better in the myeloablative therapy group than in the no further treatment group (P = 0.03).
Adverse effects
Since all patients receiving antineoplastic treatment will suffer from adverse events, we decided to analyse only severe and life‐threatening effects. We defined this as toxicity grade three or higher. Only adverse effects occurring from the start of the randomised treatment (i.e. myeloablative therapy or control) were eligible for inclusion in this review. See Figure 6 and Figure 7.
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.5 Adverse effects grade 3 or higher without additional follow‐up data.
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.6 Adverse effects grade 3 or higher including additional follow‐up of Matthay 1999.
Treatment‐related death
Data on treatment‐related death could be extracted from two trials with a total of 574 patients (Berthold 2005; Matthay 1999). There were 12 deaths among 278 patients randomised to the myeloablative therapy group and five among 296 in the control group. There was no significant statistical difference in treatment‐related death between the treatment groups (RR 2.53, 95% CI 0.17 to 37.12, P = 0.50). However, unexplained heterogeneity was detected (I2 = 78%). It should be noted that it was not possible to perform an ITT analysis using the Matthay 1999 data, therefore we performed an as‐treated analysis.
When additional follow‐up data from Matthay 1999 were included in the meta‐analysis there was still no significant statistical difference in treatment‐related death between the treatment groups (RR 1.08, 95% CI 0.65 to 1.78, P = 0.78). There were 25 deaths among 271 patients in the myeloablative group compared to 26 among 284 patients in the control group (RR 1.08, 95% CI 0.65 to 1.78, P = 0.78). No unexplained heterogeneity was detected for this comparison (I2 = 0%). Again, it was not possible to perform an ITT analysis for the study of Matthay 1999, so an as‐treated analysis was performed.
Secondary malignant disease
Data on secondary malignant disease could be extracted from two trials with a total of 674 patients (Berthold 2005; Matthay 1999). There were two cases of secondary malignant disease among 338 patients in the myeloablative group (i.e. one acute myeloid leukaemia and one unknown malignancy) compared to two cases among 336 conventional chemotherapy patients (i.e. one acute lymphoblastic leukaemia and one unknown malignancy). The meta‐analysis showed no significant statistical difference between the treatment groups (RR 0.99; 95% CI 0.14 to 7.00, P = 0.99). No heterogeneity was detected for this comparison (I2 = 0%).
When additional follow‐up data from the Matthay 1999 study were included in the meta‐analysis there was still no significant statistical difference between the treatment groups (RR 1.48, 95% CI 0.24 to 9.01, P = 0.67). There were three cases of secondary malignant disease among 338 patients in the myeloablative therapy group (i.e. two cases of acute myeloid leukaemia and one follicular carcinoma of the thyroid) compared to two among 336 patients in the conventional chemotherapy group (i.e. both acute lymphoblastic leukaemia). No unexplained heterogeneity was detected for this comparison (I2 = 0%).
Other adverse effects
Data on serious infections, sepsis, renal effects, interstitial pneumonitis and veno‐occlusive disease could be extracted from one trial with a total of 379 patients (Matthay 1999). There was a significantly higher incidence of renal effects, interstitial pneumonitis and veno‐occlusive disease in the myeloablative group compared to conventional chemotherapy, whereas for serious infections and sepsis no significant difference between the treatment groups was identified (see Figure 6 and Figure 7 for more information; since for veno‐occlusive disease none of the 190 patients in the control group experienced an event (while 17 out of 189 patients in the myeloablative therapy group did), we used the Fischer's exact test; P = 0.0001). These toxicity data did not change with additional follow‐up.
Quality of life
None of the studies evaluated quality of life.
Sensitivity analyses
The results of the sensitivity analyses for the risk of bias criteria were consistent among the trials and did not differ from the overall analyses.
Discussion
Neuroblastoma is the most common extracranial solid tumour in children. Unfortunately, despite the development of new treatment options, the prognosis of high‐risk patients is poor; and disease recurs in more than half of patients (Brodeur 2006; Goldsby 2004; Matthay 1993; Weinstein 2003). High‐dose chemotherapy and stem cell transplantation might increase survival in these patients. This systematic review evaluated the current state of evidence on the efficacy of high‐dose chemotherapy and autologous haematopoietic stem cell rescue compared with conventional therapy in patients diagnosed with high‐risk neuroblastoma. Only randomised controlled trials (RCTs) were included since it is widely recognised that a RCT is the only study design which can be used to obtain unbiased evidence on the use of different treatment options, provided that the design and execution of the RCTs are adequate.
We identified three RCTs that compared high‐dose chemotherapy and autologous haematopoietic stem cell rescue (i.e. myeloablative therapy) with conventional chemotherapy or no further treatment (Berthold 2005; Matthay 1999; Pritchard 2005). The first updated literature search identified a manuscript reporting additional follow‐up data from the Matthay 1999 study and the second one an erratum for this study. The exact length of follow‐up was not reported, but the median follow‐up of patients alive without an event was 7.7 years (range 130 days to 12.8 years). Even though all patients were diagnosed with high‐risk neuroblastoma, definitions of high‐risk patients differed between the studies, as did the required response to induction therapy to be eligible for randomisation. The three included RCTs used an age of one year as the cut‐off point for pre‐treatment risk stratification. Recently the age cut‐off for high‐risk disease was changed from one year to 18 months. As a result it is possible that patients with what is now classified as intermediate‐risk disease were included in the high‐risk groups. Consequently the relevance of the results of these studies to the current practice can be questioned. Survival rates may be overestimated due to the inclusion of patients with intermediate‐risk disease. Randomisation was performed at different time intervals since diagnosis. All trials used different induction treatments, myeloablative treatments and control treatments. Also, in two trials the received treatment slightly differed between patients, but we assumed that as a result of randomisation these differences in treatment were evenly distributed in both treatment groups. Two studies used bone marrow cells (Matthay 1999; Pritchard 2005) and one study used peripheral blood cells (Berthold 2005) for stem cell transplant (See the Characteristics of included studies table for further information).
The individual studies reported three or five‐year event‐free survival. In the Berthold 2005 and Matthay 1999 studies (including the additional follow‐up data from Matthay 1999) a significant statistical difference in favour of myeloablative therapy was identified. In the Pritchard 2005 study this difference was not statistically significant. Our meta‐analysis using their complete follow‐up period also showed a significant statistical difference in event‐free survival in favour of the myeloablative therapy group (HR 0.78, 95% CI 0.67 to 0.90, P = 0.0006). When the two studies in which secondary malignant disease was included as an event were analysed separately (Berthold 2005; Matthay 1999), this finding was confirmed. In the study which did not include secondary malignant disease as an event, no significant statistical difference in event‐free survival between the treatment groups was identified (Pritchard 2005). However, the direction of the result of this study was the same as the overall result of all trials. The results of the meta‐analysis of event‐free survival did not change when additional follow‐up data from Matthay 1999 were included in the analysis.
The individual studies also reported three‐ or five‐year overall survival. In all three studies (including the additional follow‐up data from Matthay 1999) no significant statistical difference between the treatment groups was identified. Our original meta‐analysis using the complete follow‐up period showed a significant statistical difference in overall survival in favour of the myeloablative therapy group (HR 0.74, 95% CI 0.57 to 0.98, P = 0.04; 2 studies, 360 patients). However, when additional follow‐up data from Matthay 1999 was included in the meta‐analysis the overall result was no longer statistically significant (HR 0.86, 95% CI 0.73 to 1.01, P = 0.43; 3 studies, 739 patients). A possible explanation could be the mortality caused by different late effects (e.g. as described by Mertens 2001; Reulen 2010). Another possible explanation could be that treatment options for progressive disease or relapse in patients who received myeloablative therapy are more limited than in patients who received conventional therapy or no further treatment. However, more data are needed for a definitive conclusion.
For two evaluated adverse effects (toxicity grade three or higher) it was possible to perform a meta‐analysis. No significant statistical differences in secondary malignant disease or treatment‐related death were identified between the treatment groups (both with the original data and with the inclusion of additional follow‐up data from Matthay 1999). While unexplained heterogeneity was detected in the original analysis of treatment‐related death, this was not the case when additional follow‐up data were included. Serious infections, sepsis, renal effects, interstitial pneumonitis and veno‐occlusive disease were only evaluated in the Matthay 1999 study. There was a significantly higher incidence of renal effects, interstitial pneumonitis and veno‐occlusive disease in the myeloablative therapy group, whereas for serious infections and sepsis no significant statistical differences between the treatment groups were identified. These toxicity data did not change with additional follow‐up. None of the included studies evaluated quality of life.
In the individual studies different subgroups were evaluated, but the results were not univocal in all studies. As a result, no definitive conclusions can be made regarding the effect of myeloablative therapy or conventional treatment in the different subgroups.
In this review we performed intention‐to‐treat (ITT) analyses, since they provide the most realistic and unbiased answer to the question of clinical effectiveness (Lachin 2000; Lee 1991). However, for treatment‐related death it was not possible to perform an ITT analysis using the Matthay 1999 data and we therefore performed an as‐treated analysis.
'No evidence of effect', as identified in this review, is not the same as 'evidence of no effect'. The reason that some studies did not identify a significant statistical difference between study groups could be the fact that the number of patients included in these studies was too small to detect a difference between the treatment groups (i.e. low power). Furthermore, the length of follow‐up could have been too short to detect a significant difference between the treatment groups.
The risk of bias in the included studies was difficult to assess due to a lack of reporting. As a result, the presence of selection bias, performance bias, detection bias, attrition bias, reporting bias and other bias could not be ruled out. However, at the moment this is the best available evidence from RCTs comparing the efficacy of high‐dose chemotherapy and autologous haematopoietic stem cell rescue with conventional therapy in patients diagnosed with high‐risk neuroblastoma. With regard to performance bias, it should be noted that due to the nature of the interventions, blinding of care providers and patients was virtually impossible.
The results of myeloablative therapy must be viewed in the context of the different treatment regimens used in the included studies. The included studies differed in terms of the preceding therapy (e.g. induction chemotherapy, surgery and radiotherapy), the myeloablative regimen with or without total body irradiation, the stem cells (e.g. whether purging was used or not, and the origin of the stem cells, i.e. bone marrow or peripheral blood) and the presence and type of consolidation chemotherapy. As a result, in this systematic review we can only provide conclusions on the concept of myeloablative treatment versus conventional or no further treatment. We cannot make conclusions with regard to specific versions of these treatments. However, the concept seems to work in terms of event‐free survival, both for all randomised patients treated with myeloablative therapy and for specific subgroups. For overall survival there is currently no evidence of effect. No definitive conclusions can be made regarding adverse effects and quality of life.
We are awaiting the full text publication of the study published as a conference abstract (Hero 2010), which presented additional follow‐up data from the Berthold 2005 study. Hero 2010 reported that myeloablative therapy proved to be effective with respect to long‐term outcome, but is complicated by acute and long‐term toxicity. However, as it is known that information provided in conference abstracts is often substantially different from information provided in subsequent full text publications (Yoon 2012), these results were not included in the analyses of this systematic review.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study (it should be noted that in the study describing additional follow‐up the associated risk of detection bias was thus unclear for all outcomes except overall survival)
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.1 Event‐free survival without additional follow‐up data.
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.2 Event‐free survival including additional follow‐up data of Matthay 1999.
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.3 Overall survival without additional follow‐up data.
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.4 Overall survival including additional follow‐up data of Matthay 1999.
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.5 Adverse effects grade 3 or higher without additional follow‐up data.
Forest plot of comparison: 1 Myeloablative therapy versus control, outcome: 1.6 Adverse effects grade 3 or higher including additional follow‐up of Matthay 1999.
Comparison 1 Myeloablative therapy versus control, Outcome 1 Event‐free survival without additional follow‐up data.
Comparison 1 Myeloablative therapy versus control, Outcome 2 Event‐free survival including additional follow‐up data of Matthay 1999.
Comparison 1 Myeloablative therapy versus control, Outcome 3 Overall survival without additional follow‐up data.
Comparison 1 Myeloablative therapy versus control, Outcome 4 Overall survival including additional follow‐up data of Matthay 1999.
Comparison 1 Myeloablative therapy versus control, Outcome 5 Adverse effects grade 3 or higher without additional follow‐up data.
Comparison 1 Myeloablative therapy versus control, Outcome 6 Adverse effects grade 3 or higher including additional follow‐up of Matthay 1999.
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Event‐free survival without additional follow‐up data Show forest plot | 3 | 739 | Hazard Ratio (Random, 95% CI) | 0.78 [0.67, 0.90] |
| 1.1 Event‐free survival including secondary malignant disease as an event without additional follow‐up data | 2 | 674 | Hazard Ratio (Random, 95% CI) | 0.79 [0.67, 0.92] |
| 1.2 Event‐free survival not including secondary malignant disease as an event without additional follow‐up data | 1 | 65 | Hazard Ratio (Random, 95% CI) | 0.73 [0.49, 1.07] |
| 2 Event‐free survival including additional follow‐up data of Matthay 1999 Show forest plot | 3 | 739 | Hazard Ratio (Random, 95% CI) | 0.79 [0.70, 0.90] |
| 2.1 Event‐free survival including additional follow‐up data and secondary malignant disease as an event | 2 | 674 | Hazard Ratio (Random, 95% CI) | 0.80 [0.70, 0.92] |
| 2.2 Event‐free survival including additional follow‐up data and not including secondary malignant disease as an event | 1 | 65 | Hazard Ratio (Random, 95% CI) | 0.73 [0.49, 1.07] |
| 3 Overall survival without additional follow‐up data Show forest plot | 2 | 360 | Hazard Ratio (Random, 95% CI) | 0.74 [0.57, 0.98] |
| 4 Overall survival including additional follow‐up data of Matthay 1999 Show forest plot | 3 | 739 | Hazard Ratio (Random, 95% CI) | 0.86 [0.73, 1.01] |
| 5 Adverse effects grade 3 or higher without additional follow‐up data Show forest plot | 2 | Risk Ratio (M‐H, Random, 95% CI) | Subtotals only | |
| 5.1 Treatment‐related death | 2 | 574 | Risk Ratio (M‐H, Random, 95% CI) | 2.53 [0.17, 37.12] |
| 5.2 Secondary malignant disease | 2 | 674 | Risk Ratio (M‐H, Random, 95% CI) | 0.99 [0.14, 7.00] |
| 5.3 Serious infections | 1 | 379 | Risk Ratio (M‐H, Random, 95% CI) | 1.02 [0.84, 1.23] |
| 5.4 Sepsis | 1 | 379 | Risk Ratio (M‐H, Random, 95% CI) | 0.93 [0.67, 1.30] |
| 5.5 Renal effects | 1 | 379 | Risk Ratio (M‐H, Random, 95% CI) | 2.28 [1.28, 4.04] |
| 5.6 Interstitial pneumonitis | 1 | 379 | Risk Ratio (M‐H, Random, 95% CI) | 9.55 [2.26, 40.43] |
| 6 Adverse effects grade 3 or higher including additional follow‐up of Matthay 1999 Show forest plot | 2 | Risk Ratio (M‐H, Random, 95% CI) | Subtotals only | |
| 6.1 Treatment‐related death | 2 | 555 | Risk Ratio (M‐H, Random, 95% CI) | 1.08 [0.65, 1.78] |
| 6.2 Secondary malignant disease | 2 | 674 | Risk Ratio (M‐H, Random, 95% CI) | 1.48 [0.24, 9.01] |
| 6.3 Serious infections | 1 | 379 | Risk Ratio (M‐H, Random, 95% CI) | 1.02 [0.84, 1.23] |
| 6.4 Sepsis | 1 | 379 | Risk Ratio (M‐H, Random, 95% CI) | 0.93 [0.67, 1.30] |
| 6.5 Renal effects | 1 | 379 | Risk Ratio (M‐H, Random, 95% CI) | 2.28 [1.28, 4.04] |
| 6.6 Interstitial pneumonitis | 1 | 379 | Risk Ratio (M‐H, Random, 95% CI) | 9.55 [2.26, 40.43] |
