Alternative agents to prophylactic platelet transfusion for preventing bleeding in people with thrombocytopenia due to chronic bone marrow failure: a meta‐analysis and systematic review

Michael JR Desborough; Andreas V Hadjinicolaou; Anna Chaimani; Marialena Trivella; Paresh Vyas; Carolyn Doree; Sally Hopewell; Simon J Stanworth; Lise J Estcourt

doi:10.1002/14651858.CD012055.pub2

Alternative agents to prophylactic platelet transfusion for preventing bleeding in people with thrombocytopenia due to chronic bone marrow failure: a meta‐analysis and systematic review

Authors' declarations of interest

Version published: 31 October 2016 Version history

https://doi.org/10.1002/14651858.CD012055.pub2

Collapse all Expand all

Abstract

available in

Background

People with thrombocytopenia due to bone marrow failure are vulnerable to bleeding. Platelet transfusions have limited efficacy in this setting and alternative agents that could replace, or reduce platelet transfusion, and are effective at reducing bleeding are needed.

Objectives

To compare the relative efficacy of different interventions for patients with thrombocytopenia due to chronic bone marrow failure and to derive a hierarchy of potential alternative treatments to platelet transfusions.

Search methods

We searched for randomised controlled trials (RCTs) in the Cochrane Central Register of Controlled Trials (the Cochrane Library 2016, Issue 3), MEDLINE (from 1946), Embase (from 1974), CINAHL (from 1937), the Transfusion Evidence Library (from 1980) and ongoing trial databases to 27 April 2016.

Selection criteria

We included randomised controlled trials in people with thrombocytopenia due to chronic bone marrow failure who were allocated to either an alternative to platelet transfusion (artificial platelet substitutes, platelet‐poor plasma, fibrinogen concentrate, recombinant activated factor VII (rFVIIa), desmopressin (DDAVP), recombinant factor XIII (rFXIII), recombinant interleukin (rIL)6 or rIL11, or thrombopoietin (TPO) mimetics) or a comparator (placebo, standard of care or platelet transfusion). We excluded people undergoing intensive chemotherapy or stem cell transfusion.

Data collection and analysis

Two review authors independently screened search results, extracted data and assessed trial quality. We estimated summary risk ratios (RR) for dichotomous outcomes. We planned to use summary mean differences (MD) for continuous outcomes. All summary measures are presented with 95% confidence intervals (CI).

We could not perform a network meta‐analysis because the included studies had important differences in the baseline severity of disease for the participants and in the number of participants undergoing chemotherapy. This raised important concerns about the plausibility of the transitivity assumption in the final dataset and we could not evaluate transitivity statistically because of the small number of trials per comparison. Therefore, we could only perform direct pairwise meta‐analyses of included interventions.

We employed a random‐effects model for all analyses. We assessed statistical heterogeneity using the I² statistic and its 95% CI. The risk of bias of each study included was assessed using the Cochrane 'Risk of bias' tool. The quality of the evidence was assessed using GRADE methods.

Main results

We identified seven completed trials (472 participants), and four ongoing trials (recruiting 837 participants) which are due to be completed by December 2020. Of the seven completed trials, five trials (456 participants) compared a TPO mimetic versus placebo (four romiplostim trials, and one eltrombopag trial), one trial (eight participants) compared DDAVP with placebo and one trial (eight participants) compared tranexamic acid with placebo. In the DDAVP trial, the only outcome reported was the bleeding time. In the tranexamic acid trial there were methodological flaws and bleeding definitions were subject to significant bias. Consequently, these trials could not be incorporated into the quantitative synthesis. No randomised trial of artificial platelet substitutes, platelet‐poor plasma, fibrinogen concentrate, rFVIIa, rFXIII, rIL6 or rIL11 was identified.

We assessed all five trials of TPO mimetics included in this review to be at high risk of bias because the trials were funded by the manufacturers of the TPO mimetics and the authors had financial stakes in the sponsoring companies.

The GRADE quality of the evidence was very low to moderate across the different outcomes.

There was insufficient evidence to detect a difference in the number of participants with at least one bleeding episode between TPO mimetics and placebo (RR 0.86, 95% CI 0.56 to 1.31, four trials, 206 participants, low‐quality evidence).

There was insufficient evidence to detect a difference in the risk of all‐cause mortality between those treated with a TPO mimetic and placebo (RR 0.74, 95%CI 0.52 to 1.05, five trials, 456 participants, very low‐quality evidence).

There was no evidence for a difference in the incidence of transfusion reactions between those treated with TPO mimetics and placebo (pOR 0.06, 95% CI 0.00 to 3.44, one trial, 98 participants, very low‐quality evidence).

There was no evidence for a difference in thromboembolic events between TPO mimetics and placebo (RR 1.41, 95%CI 0.39 to 5.01, five trials, 456 participants, very‐low quality evidence).

There was no evidence for a difference in drug reactions between TPO mimetics and placebo (RR 1.12, 95% CI 0.83 to 1.51, five trials, 455 participants, low‐quality evidence).

No trial reported the number of days of bleeding per participant, platelet transfusion episodes, mean red cell transfusions per participant, red cell transfusion episodes, transfusion‐transmitted infections, formation of antiplatelet antibodies or platelet refractoriness.

In order to demonstrate a reduction in bleeding events from 26 in 100 to 16 in 100 participants, a study would need to recruit 514 participants (80% power, 5% significance).

Authors' conclusions

There is insufficient evidence at present for thrombopoietin (TPO) mimetics for the prevention of bleeding for people with thrombocytopenia due to chronic bone marrow failure. There is no randomised controlled trial evidence for artificial platelet substitutes, platelet‐poor plasma, fibrinogen concentrate, rFVIIa, rFXIII or rIL6 or rIL11, antifibrinolytics or DDAVP in this setting.

PICOs

Population

Intervention

Comparison

Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Plain language summary

available in

Alternative agents instead of platelet transfusions to prevent bleeding for people who have bone marrow disorders and low platelet counts

Review question

We evaluated the evidence about whether giving agents that can replace, or reduce platelet transfusion (artificial platelets, platelet‐poor plasma, fibrinogen concentrate, recombinant activated factor VII (rFVIIa), recombinant factor XIII (rFXIII), recombinant interleukin (rIL)6 or rIL11, desmopressin (DDAVP), thrombopoietin (TPO) mimetics or antifibrinolytic drugs), to people with a low platelet count prevents bleeding and whether these alternative agents are associated with side effects. Our target population was people with bone marrow disorders which prevent them from producing enough platelets. We excluded people undergoing intensive chemotherapy or stem cell transplantation.

Background

People with low platelet counts due to bone marrow disorders are vulnerable to bleeding which may be severe or life‐threatening. In order to treat, or prevent bleeding, they are often given platelet transfusions. However, platelet transfusions are associated with risks such as infection and transfusion reactions. Consequently, there is interest in whether it is possible to use alternative treatments to prevent bleeding. These treatments include: man‐made platelets (artificial platelets); stimulating the person's body to produce more platelets (recombinant interleukin (rIL)6, rIL11, TPO mimetics); increasing the levels of proteins in the blood that help the body to form a clot (platelet‐poor plasma, fibrinogen concentrate, recombinant activated factor VII (rFVIIa), recombinant factor XIII (rFXIII)); and preventing a blood clot from breaking down (antifibrinolytics) There may be risks associated with agents that prevent bleeding; the most important being an increased risk of forming unwanted blood clots, which could be potentially life‐threatening.

Study characteristics

The evidence is current to April 2016. We identified 11 randomised controlled trials, of which seven had been completed. Of the seven completed trials, five trials (456 participants) assessed TPO mimetics, one trial (eight participants) assessed tranexamic acid and one trial (eight participants) assessed DDAVP. The trial of DDAVP only assessed the bleeding time: the time taken for bleeding to stop after a small cut is made in the participant's forearm. It did not assess any of the outcomes of interest to this review. The trial of tranexamic acid had significant methodological flaws in the way bleeding was reported. No randomised trial of artificial platelet substitutes, platelet‐poor plasma, fibrinogen concentrate, rFVIIa, rFXIII, rIL6 or rIL11 was identified. Consequently, quantitative analysis was only performed on the five trials assessing TPO mimetics. Four of these trials included adults with myelodysplastic syndrome (MDS) and one trial assessed adults with MDS or acute myeloid leukaemia (AML). We assessed all five trials of TPO mimetics included in this review to be at high risk as the manufacturers if the TPO mimetics were directly involved in the design and publication of the trials.

Differences in severity of disease and number of participants undergoing chemotherapy between trials meant that network meta‐analysis could not be performed. A requirement of network meta‐analysis is that participants in each trial should meet the eligibility criteria for each trial that is included.

The four ongoing trials are all comparing TPO mimetics versus placebo; they are expected to recruit 837 participants in total and are due to be completed by December 2020.

Key results

TPO mimetics may make little or no difference to the number of participants with any bleeding or severe/life‐threatening bleeding. We are very uncertain whether TPO mimetics reduce the risk of mortality. TPO mimetics probably reduce the number of participants who need a platelet transfusion. We are very uncertain whether TPO mimetics reduce the risk of transfusion reactions or risk of thromboembolism. TPO mimetics may have little or no effect on the risk of drug reactions.

No trial reported the number of days bleeding per participant, platelet transfusion episodes, mean red cell transfusions per participant, red cell transfusion episodes, transfusion‐transmitted infections, formation of antiplatelet antibodies or platelet refractoriness.

Quality of the evidence

The quality of the evidence was low or very low for all outcomes except the number of participants receiving a platelet transfusion which was moderate‐quality evidence.

Authors' conclusions

Implications for practice

There is inadequate evidence to recommend the use of thrombopoietin (TPO) mimetics for bone marrow failure, although the results of four ongoing trials may change this. There is moderate quality evidence that TPO mimetics reduce the number of participants requiring a platelet transfusion compared to placebo. However further data will be necessary in order to assess other clinical outcomes. One randomised trial of tranexamic acid versus placebo, and one randomised controlled trial of desmopressin (DDAVP) versus placebo were identified but neither trial was reported in a way that allowed the extraction of any clinical data. There were no randomised controlled trials assessing artificial platelet substitutes, platelet‐poor plasma, rFVIIa, rFXIII, interleukin 6, interleukin 11 or fibrinogen concentrate for people with bone marrow failure.

Implications for research

Our search strategy has identified four further trials of TPO mimetics (eltrombopag) with 837 participants, which are presently underway for people with bone marrow failure. In order to demonstrate a fall in bleeding events from 26 in 100 to 16 in 100 participants (as seen in the eltrombopag data), a study would need to recruit 514 participants (80% power, 5% significance) and it is likely that the publication of additional data from ongoing trials will answer this question. There are no adequate randomised controlled trials assessing artificial platelet substitutes, platelet‐poor plasma, rFVIIa, rFXIII, interleukin 6, interleukin 11, fibrinogen concentrate, DDAVP or antifibrinolytics for people with bone marrow failure and this remains a potential area for future research.

Summary of findings

Open in table viewer

Summary of findings 1. Thrombopoietin mimetic versus placebo

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments

Patient or population: People with chronic bone marrow failure Intervention: Thrombopoietin mimetics Comparison: Placebo
Outcomes	Risk with placebo	Risk with Thrombopoietin mimetic	Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments
Number of participants with at least one bleeding episode	Study population		RR 0.86 (0.56 to 1.31)	206 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}
Number of participants with at least one bleeding episode	315 per 1000	271 per 1000 (176 to 413)	RR 0.86 (0.56 to 1.31)	206 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}
Number of participants with at least one severe or life‐threatening bleeding episode	Study population		RR 0.31 (0.04 to 2.26)	40 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
	154 per 1000	48 per 1000 (6 to 348)	RR 0.31 (0.04 to 2.26)	40 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
All‐cause mortality	Study population		RR 0.74 (0.52 to 1.05)	456 (5 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3}
All‐cause mortality	237 per 1000	176 per 1000 (123 to 249)	RR 0.74 (0.52 to 1.05)	456 (5 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3}
Proportion of participants receiving a platelet transfusion	Study population		RR 0.76 (0.61 to 0.95)	206 (4 RCTs)	⊕⊕⊕⊝ MODERATE ¹
Proportion of participants receiving a platelet transfusion	658 per 1000	500 per 1000 (401 to 625)	RR 0.76 (0.61 to 0.95)	206 (4 RCTs)	⊕⊕⊕⊝ MODERATE ¹
Transfusion reactions	Study population		pOR 0.06 (0.00 to 3.44)	98 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 3 4}
Transfusion reactions	29 per 1000	2 per 1000 (0 to 94)	pOR 0.06 (0.00 to 3.44)	98 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 3 4}
Thromboembolism	Study population		RR 1.41 (0.39 to 5.01)	456 (5 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3}
Thromboembolism	19 per 1000	27 per 1000 (8 to 97)	RR 1.41 (0.39 to 5.01)	456 (5 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3}
Drug reactions	Study population		RR 1.12 (0.83 to 1.51)	455 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}
Drug reactions	271 per 1000	303 per 1000 (225 to 409)	RR 1.12 (0.83 to 1.51)	455 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}
Number of red cell transfusions per participant	Meta‐analysis not possible		Not estimable	98 (1 RCT)	‐	Participants treated with eltrombopag received mean 4.8 units red blood cells whereas those treated with placebo received mean 6.3 units over 6 months
*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). CI: Confidence interval; RR: Risk ratio; pOR: Peto Odds ratio;
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
¹ Downgraded one point due to risk of bias ² Downgraded one point due to imprecision ³ Downgraded two points due to low event rate ⁴ Downgraded one point due to indirectness

Background

Description of the condition

The bone marrow is the production site for red blood cells, white blood cells and platelets from stem cells, during the processes termed collectively as haematopoiesis. Bone marrow failure disorders encompass a wide range of diseases that cause quantitative (reduced numbers) or qualitative (reduced function) defects of red cells, white cells and platelets.

Clinical symptoms of people with bone marrow failure disorders are related to cytopenia, that is, the failure to produce adequate numbers of normal red cells, white cells, or platelets. People can present with fatigue and shortness of breath due to anaemia, recurrent infections due to neutropenia (reduced numbers of white cells ‐ neutrophils), and bleeding or bruising due to thrombocytopenia (reduced numbers of platelets). Bleeding is a result of a failure to produce adequate numbers of platelets because of insufficient numbers of bone marrow megakaryocytes (cells in the bone marrow that produce platelets) or megakaryocyte dysfunction. Bone marrow failure disorders can also be associated with an increased risk of progression to acute leukaemia.

Bone marrow failure disorders can be classified according to the underlying pathophysiology, into four broad categories: myelodysplastic syndromes (MDS), primary myelofibrosis, acquired aplastic anaemia, and inherited bone marrow failure disorders.

MDS encompasses a diverse group of disorders that are characterised by dysplasia in one or more cell lines (blood cells have an abnormal shape or size), ineffective haematopoiesis, and an increased risk of developing acute myeloid leukaemia (AML). Overall, the incidence of MDS is estimated at between 2.3 to 4.5 per 100,000 per year; however, incidence increases markedly with age, peaking in those aged over 80 years (> 30 per 100,000 per year) (Dinmohamed 2014; Ma 2007; Ma 2012; Neukirchen 2011). Several cohort studies have evaluated the incidence of thrombocytopenia at diagnosis (platelet count < 100 x 10⁹/L), which affects 23% to 93% of people with newly diagnosed MDS, depending on the cohort (Kantarjian 2007). Cohort studies report that haemorrhage is the cause of death in 14% to 24% cases of MDS (Foucar 1985; Gupta 1999; Kantarjian 2007; Konstantopoulos 1989; Lidbeck 1980).

Primary myelofibrosis is a clonal myeloproliferative disease whereby the normal bone marrow is replaced by fibrosis, resulting in bone marrow failure. It has an incidence of 2.2 to 9.9 per million per year (Titmarsh 2014). People may develop a number of symptoms including fatigue, sweats, fevers, weight loss and an enlarged spleen, as well as symptoms of bone marrow failure (Tefferi 2013).

Acquired aplastic anaemia is a disease that results in a hypocellular bone marrow with quantitative defects of all three cell lines. The incidence in Europe and North America is two per million population per year and has a biphasic age distribution with increasing numbers of cases in those aged 10 to 25 years and those over 60 years (Heimpel 2000; Issaragrisil 2006; Montané 2008). The incidence in Asia is higher, with estimates ranging from 3.9 to 7.4 per million per year (Young 2008). The underlying cause of aplastic anaemia is unknown in most cases, but different reports have associated it with certain industrial chemicals (Young 2008), agricultural pesticides (Issaragrisil 2006; Muir 2003), drugs (Issaragrisil 2006; Young 2008), and hepatitis viruses (Rauff 2011).

Inherited bone marrow failure disorders that result in thrombocytopenia include those associated with a global haematopoietic defect such as Fanconi anaemia, Dyskeratosis congenita, or Swachman‐Diamond syndrome, as well as disorders associated with isolated thrombocytopenia, such as thrombocytopenia with absent radii (TAR) and amegakaryocytic thrombocytopenia (Alter 2007). The most common inherited bone marrow disorder is Fanconi anaemia, which has a reported incidence of approximately 1 in 360,000 live births, with a carrier frequency of 1 in 300 (Swift 1971).

Treatment is tailored to the needs of individual people but may include intensive treatment with allogeneic stem cell transplantation (Dokal 2008). Other people are managed symptomatically, with low‐dose chemotherapy, or in the case of aplastic anaemia, with immunosuppressive agents, with a focus on maintaining quality of life, prolonging life and delaying transformation to acute leukaemia.

Description of the intervention

Platelet transfusions are of some benefit in managing active bleeding for people with bone marrow failure and severe thrombocytopenia. The standard practice in most haematology units across the developed world is to use prophylactic transfusions in line with guidelines (BCSH 2003; BCSH 2004; NBA 2012; Schiffer 2001; Slichter 2007; Tinmouth 2007). For chronic bone marrow failure, prophylactic platelet transfusions are standard for people with a platelet count less than 10 x 10⁹/L and a haemorrhagic phenotype, but for people without a haemorrhagic phenotype, platelet transfusions are not given. It is still uncertain how best to use platelet transfusions to prevent severe and life‐threatening bleeding (Estcourt 2011). Alternative agents which could replace or reduce platelet transfusions may be more effective than platelet transfusions at controlling bleeding and will have a different side‐effect profile. Alternatives include artificial platelet substitutes, platelet‐poor plasma, recombinant factor VIIa (rFVIIa), fibrinogen, recombinant factor XIII (rFXIII), thrombopoietin (TPO) mimetics and antifibrinolytic drugs.

How the intervention might work

In normal haemostasis (formation of a blood clot), platelets form a primary haemostatic plug that is consolidated by the deposition of cross‐linked fibrin. Platelet adhesion depends on normal platelet function, the presence of von Willebrand factor, and extracellular matrix components such as collagen and fibronectin (Ruggeri 2007). When the platelet count is low, the standard treatment has been to transfuse platelets, although this procedure can be associated with hazards such as infection, transfusion reactions and formation of anti‐platelet antibodies. Additionally, 18% to 23% people with bone marrow failure due to MDS have a haemorrhagic phenotype regardless of their platelet count (Kantarjian 2007). Alternatives to platelet transfusion aim to either simulate the effects of platelets (artificial platelet substitutes), stimulate additional fibrin formation (platelet‐poor plasma, recombinant factor VIIa and fibrinogen), promote von Willebrand factor release and platelet function (desmopressin), or increase platelet production (TPO mimetics). These agents aim to promote haemostasis without the side effects associated with platelet transfusions. The important adverse event for any pro‐haemostatic intervention is thrombosis, and any of these interventions have the potential to cause it. Other specific adverse events are listed below with the description of the intervention.

Artificial platelet substitutes

Artificial platelet substitutes such as microspheres of human albumin coated with fibrinogen. lyophilised platelets, infusible plasma membranes; and liposomes with inserted platelet receptors aim to reproduce the active components of platelets without associated adverse events (Desborough 2016a). Artificial platelets are not yet in routine clinical use, so their costs and adverse events are unclear at present.

Platelet‐poor plasma (PPP)

Platelet‐poor plasma (PPP) is a source of clotting factors and fibrinogen and can be administered intravenously. PPP is a blood component and is associated with a small risk of transfusion reactions and transfusion‐transmitted infections (Desborough 2012).

Recombinant factor VIIa (rFVIIa)

Recombinant factor VIIa (rFVIIa) is an intravenous drug licensed for people with haemophilia and inhibitory allo‐antibodies, and for prophylaxis and treatment of people with congenital factor VII deficiency. It is used off‐license in a number of other settings, including operations where blood loss cannot be controlled by other means. However, the effectiveness of its use in people without haemophilia is unproven. This is an expensive agent compared to platelet transfusion and repeated doses every two to three hours are often necessary (Joint Formulary Committee 2016). It has an advantage of not being a biological agent (Simpson 2012).

Fibrinogen concentrate

The final step of the coagulation cascade is the formation of a fibrin clot. The substrate for fibrin is fibrinogen, which is converted into fibrin by the action of thrombin. Fibrinogen concentrate is administered intravenously and may result in some reduction in surgical bleeding when administered pre‐operatively, although the overall quality of evidence for this is low (Wikkelsø 2013). Fibrinogen concentrate is a blood component and is associated with a theoretical risk of viral infection. However, viral inactivation is involved in its manufacture and is likely to make this risk very low (Franchini 2012).

Recombinant factor XIII (rFXIII)

In a normal clot, when single strands of fibrin have been formed, they are cross‐linked by factor XIII, giving the clot strength. Trials of rFXIII have taken place in people undergoing cardiac surgery to assess whether this reduced postsurgical bleeding (Karkouti 2013).

Desmopressin (DDAVP)

Desmopressin (DDAVP) is a vasopressin analogue that increases the plasma levels of factor VIII (FVIII) and von Willebrand factor (vWF) two‐ to three‐fold. It is used to treat people with mild haemophilia A or von Willebrand disease and has also been used to treat people with uraemia, liver cirrhosis, congenital platelet function disorders and drug‐induced platelet dysfunction (Svensson 2014). It can be administered intravenously, subcutaneously or intranasally. These different routes of administration result in different levels of vWF and factor VIII response (Mannucci 1987). If we include trials comparing more than one route of administration of DDAVP, then we will perform sensitivity analyses to determine if they can be combined as a single node. DDAVP is a well‐tolerated medication, but it is associated with facial flushing and can potentially cause hyponatraemic seizures in people who are not fluid‐restricted (Svensson 2014). It has a short duration of action and is more likely to be used for prophylaxis prior to procedures than for long‐term prophylaxis. It is not a biological product and is less expensive than platelet transfusion (Joint Formulary Committee 2016).

Thrombopoietin (TPO) mimetics

The liver synthesises thrombopoietin (TPO), which is the key regulator of bone marrow platelet production. TPO mimetics have been used in several disease states to promote both megakaryopoiesis and thrombopoiesis (Kuter 2014). The two main TPO mimetics in current use are romiplostim (weekly injection) and eltrombopag (daily oral tablet), both of which are recommended by the National Institute for Health and Care Excellence (NICE) for use in adults with immune thrombocytopenia (ITP) who have severe disease and a high risk of bleeding (NICE 2011; NICE 2013). While a systematic review found that these agents improve platelet counts, there was no evidence that TPO receptor agonists reduced the risk of significant bleeding for people with ITP (Zeng 2011). TPO mimetics are more expensive than platelet transfusions (Joint Formulary Committee 2016). Interleukin 6 and interleukin 11 may also act as stimulants of thrombopoiesis (Gordon 1995; Kurzrock 2001; Tsimberidou 2005). Recombinant interleukin 6 and 11 are not in routine clinical use, so their costs are unclear at present.

Antifibrinolytic drugs

Fibrinolysis is the process by which blood clots are broken down after they have been formed. Anti‐fibrinolytic drugs block this process, resulting in greater clot strength. The three most commonly used antifibrinolytic drugs are tranexamic acid, aprotinin and epsilon‐aminocaproic acid. A previous Cochrane systematic review assessed these agents (Estcourt 2016), which are included in this review for comparison with other potential interventions as part of our planned network meta‐analysis. Antifibrinolytics are cheaper than platelet transfusions (Joint Formulary Committee 2016).

Why it is important to do this review

This review is focused on whether alternative agents to prophylactic platelet transfusions are effective for the prevention and control of life‐threatening thrombocytopenic bleeding. Platelet transfusions are expensive and may lead to adverse events such as infections and platelet refractoriness, particularly in groups of people who receive multiple transfusions, such as those with chronic bone marrow failure. Some people with bone marrow failure bleed despite apparently adequate platelet numbers, and alternative methods for managing bleeding will be necessary. This review is also important for the developing world, where access to safe blood components is much more limited (Verma 2009).

Objectives

To compare the relative efficacy of different treatments for thrombocytopenia (artificial platelet substitutes, platelet‐poor plasma, fibrinogen, rFVIIa, rFXIII, thrombopoietin mimetics, antifibrinolytic drugs or platelet transfusions) in people with chronic bone marrow failure and to derive a hierarchy of potential alternate treatments to platelet transfusions.

Methods

Criteria for considering studies for this review

Types of studies

Only randomised controlled trials (RCTs) were included.

Types of participants

We included inpatients and outpatients of all ages with thrombocytopenia due to chronic bone marrow failure. Only data from the bone marrow failure subgroups were used for trials consisting of mixed populations of participants (e.g. those with diagnoses of immune thrombocytopenic purpura). If subgroup data for participants with bone marrow failure were not available (even after contacting the authors of the trial), we excluded the trial if fewer than 80% of participants had bone marrow failure. We excluded any participants who did not have thrombocytopenia due to bone marrow failure, as well as participants undergoing intensive chemotherapy or stem cell transplantation, as this is the focus of another Cochrane review (Desborough 2016b). We included participants with bone marrow failure syndromes (e.g. aplastic anaemia, congenital bone marrow failure syndromes, MDS and myelofibrosis) who were not being treated with intensive chemotherapy or an allogeneic stem cell transplant.

Types of interventions

We considered the following interventions (alternative agents that could replace or reduce platelet transfusion) without restrictions on the dose compared to each other or to placebo.

Artificial platelet substitutes.
Platelet‐poor plasma.
Recombinant factor VIIa (rFVIIa).
Fibrinogen.
Recombinant factor XIII (rFXIII).
TPO mimetics (we analysed the most commonly used TPO mimetics, eltrombopag and romiplostim, separately and in combination).
Interleukin 6 or interleukin 11.
Desmopressin.
Anti‐fibrinolytics (such as tranexamic acid).

We included randomised controlled trials (RCTs) that evaluated one or more of the interventions listed above. We report the findings for all interventions in the results and conclusions of the review.

Types of outcome measures

We categorised all outcomes according to short‐, medium‐, and long‐term outcomes. Studies that met the other inclusion criteria were included in this review regardless of whether they included these outcomes. We reported the exact definition of these time frames over time periods that were common to as many studies as possible (e.g. up to 30 days, one to six months, and greater than six months from day of randomisation). We planned to use the primary outcomes and adverse events to develop a hierarchy of treatments.

Primary outcomes

Number of participants with at least one bleeding episode
Number of participants with at least one severe or life‐threatening bleeding episode
Number of days bleeding occurred per participant

Secondary outcomes

Mortality
- Overall mortality
- Mortality due to bleeding
- Mortality due to infection
Platelet transfusions
- Proportion of participants requiring a platelet transfusion
- Number of units of platelets transfused per participant
- Mean number of platelet transfusion episodes per participant
Red cell transfusions
- Proportion of participants requiring a red cell transfusion
- Number of units of red cells transfused per participant
- Number of red cell transfusion episodes per participant
Adverse events (e.g. transfusion reactions, transfusion‐transmitted infections, thromboembolism, development of platelet antibodies, development of platelet refractoriness, drug reactions)

Search methods for identification of studies

The Systematic Review Initiative (SRI) Information Specialist (CD) formulated the search strategies in collaboration with the Cochrane Haematological Malignancies Group. The search included all possible comparisons formed by the interventions of interest.

Electronic searches

Bibliographic databases

We searched for randomised controlled trials in the following databases.

CENTRAL, DARE, HTA & NHSEED (the Cochrane Library 2016, Issue 3) (Appendix 1)
MEDLINE (1946 to 27 April 2016) (Appendix 2)
Embase (1974 to 27 April 2016) (Appendix 3)
CINAHL (1937 to 27 April 2016) (Appendix 4)
PUBMED (epublications only) (Appendix 5)
TRANSFUSION EVIDENCE LIBRARY (1980 to 27 April 2016) (Appendix 6)
LILACS (1982 to 227 April 2016) (Appendix 7)
IndMed (1986 to 27 April 2016) (Appendix 8)
KoreaMed (1997 to 27 April 2016) (Appendix 9)
Web of Science (Conference Proceedings Citation Index‐ Science (CPCI‐S) ‐ 1990 to 27 April 2016) (Appendix 10)

We combined searches in MEDLINE, Embase and CINAHL with adaptations of the Cochrane RCT search filters, as described in Chapter 6 of the Cochrane Handbook for Systematic Reviews of Interventions (Lefebvre 2011). We combined searches in CINAHL with the relevant SIGN RCT studies filter (www.sign.ac.uk/methodology/filters.html). We presented all search strategies in the appendices as indicated. There were no restrictions on language or publication status.

Ongoing studies:

We identified ongoing trials with searches of ClinicalTrials.gov (http://clinicaltrials.gov/ct2/search) (Appendix 11), the WHO International Clinical Trials Registry Platform (ICTRP) (http://apps.who.int/trialsearch/) (Appendix 12) and the Hong Kong Clinical Trials Registry (http://www.hkclinicaltrials.com/) (Appendix 13) to 27 April 2016.

Searching other resources

We handsearched references of all included trials, relevant review articles, and current treatment guidelines for further literature, limiting these searches to the 'first generation' reference lists.

Data collection and analysis

Selection of studies

We selected studies according to the methods described in Chapter 7 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). Two review authors (MD, AH) working independently initially screened all electronically derived citations and abstracts of papers identified by the review search strategy for relevance. Clearly irrelevant studies were excluded at this stage. The same two review authors then independently assessed the full texts of all potentially relevant trials for eligibility against the criteria outlined above. We designed a study eligibility form to assess the relevance of trials on platelet transfusion, which helped ascertain whether the participants had thrombocytopenia due to bone marrow failure and whether trial arms differed according to their use of an alternative agent to prophylactic platelet transfusions. We recorded the reasons why potentially relevant studies failed to meet the eligibility criteria and displayed the results of the search in a PRISMA flow chart (Hutton 2015) (Figure 1).

Figure 1

Study flow diagram.

Data extraction and management

Two review authors (MD, AH) conducted data extraction according to the guidelines proposed by Cochrane (Higgins 2011a). Disagreements between the review authors were resolved by consensus without the need for a third review author. The review authors were not blinded to names of authors, institutions, journals, or the outcomes of the trials. A related review team had previously piloted the data extraction forms (Desborough 2016b). The two authors (MD, AH) independently extracted the following data for all the studies.

General information: review author’s name, date of data extraction, study ID, reference manager number, first author of study, author’s contact address (if available), citation of paper and objectives of the trial.
Trial details: trial design, location, setting, sample size, power calculation, treatment allocation, randomisation, blinding, inclusion and exclusion criteria, reasons for exclusion, comparability of groups, length of follow‐up, stratification, stopping rules described, statistical analysis, results, conclusion and funding.
Characteristics of participants: age, sex, ethnicity, total number recruited, total number randomised, total number analysed, types of bone marrow failure, severity of disease, baseline platelet count, numbers lost to follow‐up, dropouts (percentage in each arm) with reasons, protocol violations, previous treatments, current treatment and prognostic factors. We used the type of bone marrow failure, severity of disease and baseline platelet count for the evaluation of the transitivity assumption (Jansen 2013; Salanti 2012).
Characteristics of interventions: number of study arms, description of experimental arm(s), description of control arm, type of platelet transfusion given, timing of intervention, dosage of platelet given, compliance to interventions, additional interventions given especially in relation to red cell transfusions and any other differences between interventions.
Outcomes: number and severity of bleeding episodes, mortality (all causes), mortality due to bleeding, mortality due to infection, mean number of platelet and red cell transfusions, proportion of participants requiring each type of transfusion and adverse events (e.g. transfusion reactions, transfusion‐transmitted infections, thromboembolism, development of platelet antibodies, development of platelet refractoriness, drug reactions). We used both full‐text versions and abstracts to retrieve data. We extracted arm‐level data rather than study‐level data. One review author (MD) entered data into software, and another (AH) checked the data entry for accuracy.
Data on potential effect modifiers: For each individual study, we extracted data on the following study, intervention and population characteristics that may have acted as effect modifiers.

- Cause of bone marrow failure.
- Severity of disease.
- Baseline platelet count.
- Concurrent medications.

Assessment of risk of bias in included studies

We assessed the quality of all RCTs using the Cochrane 'Risk of bias' criteria, as described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b).Two review authors (MD, AH) worked independently to assess each element of potential bias listed below as carrying a high, low or unclear risk. We describe the judgement statements upon which review authors have assessed potential bias in the Characteristics of included studies table. We reached consensus on the degree of risk of bias by comparing the review authors' statements. We used the 'Risk of bias' assessment to explore statistical heterogeneity in each included study. We used Cochrane's tool for assessing risk of bias (low, high or unclear risk) in the following areas.

Selection bias (random sequence generation and allocation concealment).
Performance bias (blinding of participants and personnel).
Detection bias (blinding of outcome assessment).
Attrition bias (incomplete outcome data).
Reporting bias (selective reporting).
Other bias.

We assessed risk of bias separately for each key outcome of the review.

Measures of treatment effect

We recorded the number of events and the total number of participants in both the treatment and control groups for dichotomous outcomes (number of participants with at least one bleeding episode, number of participants with at least one severe or life‐threatening bleeding episode, overall mortality, mortality due to bleeding, mortality due to infection, proportion of participants requiring a platelet transfusion, proportion of participants requiring a red cell transfusion, adverse events),

If data were available, we planned to record the mean, standard deviation and total number of participants in both the treatment and control groups for continuous outcomes (number of days bleeding occurred per participant, number of units of platelets transfused per participant, mean number of platelet transfusion episodes per participant, number of units of red cells transfused per participant, number of red cell transfusion episodes per participant). For studies providing only study‐level data, we would have extracted the reported effect size with the corresponding standard error.

We planned to analyse continuous outcomes measured using the same scale, using the mean difference (MD) with 95% confidence intervals (CIs).

Relative treatment effects

We reported risk ratios (RRs) with a 95% CI for dichotomous outcomes. When we could not report the available data in any of the formats described above, we provided a descriptive summary of the available information.

We estimated the pairwise relative treatment effects of the competing interventions using the proportion of participants with significant bleeding, the proportion of participants with an adverse events and the proportion of participants requiring a platelet transfusion. We then analysed these dichotomous outcomes by calculating a RR.

Relative treatment ranking

We also considered the use of the surface area under the cumulative ranking curve (SUCRA) to obtain a hierarchy of the competing interventions for the primary outcomes and the adverse events (Salanti 2011).

Unit of analysis issues

We dealt with unit of analysis issues according to the recommendations in Chapter 16 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011c).

Cross‐over trials

Our search strategy identified two cross‐over trials, but we were unable to use any data from these trials. If relevant cross‐over trials are identified in future updates of this review, we will not assess the long‐term outcomes mortality and proportion of participants in complete remission. We will assess other outcomes if the timing of the outcome measure occurred before the cross‐over and if outcomes after the cross‐over are not biased by the treatment before the cross‐over. We will examine each trial individually to determine this eventuality.

Cluster‐randomised trials

We did not find any relevant cluster randomised trials, but for future updates of this review we plan to analyse cluster‐randomised trials at the individual participant level, accounting for the cluster design and seek statistical advice.

Studies with multiple treatment groups

We treated studies with multiple treatment groups as different independent two‐arm studies. Where appropriate, the control group was split between the two intervention groups according to the guidelines in Chapter 16 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011c). In the network meta‐analysis the correlation included in the relative effects from multi‐arm studies can be modelled properly.

Dealing with missing data

We dealt with missing data according to the recommendations in Chapter 16 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011c). We contacted authors in order to obtain information that was missing or unclear in the published report. In trials that included thrombocytopenic participants with bone marrow failure as well as participants with other causes of thrombocytopenia, we extracted data for the bone marrow failure subgroup from the general trial data. We recorded the number of participants lost to follow‐up for each trial. We analysed data according to the intention‐to‐treat (ITT) principle, but if insufficient data were available, we planned to present per protocol (PP) analyses (Higgins 2011c). We also considered to perform a sensitivity analysis to evaluate the robustness of results when we move away from the available‐case analysis using the informative missingness parameter framework (Mavridis 2014; White 2008).

Assessment of heterogeneity

Assessment of clinical and methodological heterogeneity within treatment comparisons

When we considered the clinical and methodological characteristics of individual studies to be sufficiently homogenous, we combined the data to perform a meta‐analysis (Deeks 2011). We did not report the overall summary statistic if excessive heterogeneity was present.

Assumptions when estimating the heterogeneity

We estimated different heterogeneity variances for each pairwise comparison. For the network meta‐analysis, we would have considered a common heterogeneity parameter for each outcome.

Measures and tests for heterogeneity

We assessed statistical heterogeneity within each pairwise comparison using the I² statistic and its 95% CI, which measured the percentage of variability that could not be attributed to random error (I² > 50% moderate heterogeneity, I² > 80% considerable heterogeneity). For the network meta‐analysis, the assessment of heterogeneity would be based on the magnitude of the on the magnitude of the heterogeneity variance parameter (Tau²) and its comparison with previously suggested empirical distributions (Rhodes 2015; Turner 2012). We would also estimate a total I² value for heterogeneity in the network (Jackson 2014).

Assessment of transitivity across treatment groups

We assessed whether the transitivity assumption is likely to hold by comparing epidemiologically and statistically, when possible, the clinical and methodological characteristics of the studies grouped by comparison (Jansen 2013; Salanti 2012). We considered that transitivity would be violated by differences across comparisons in the severity of disease, baseline platelet count and co‐interventions such as chemotherapy.

Assessment of reporting biases

We planned to explore the presence of small‐study effects in direct meta‐analyses by generating a funnel plot and statistically using a linear regression test if sufficient studies had been available. We considered a P value of less than 0.10 significant for this test (Sterne 2011). We also had planned to use contour‐enhanced funnel plots to assess whether publication bias is likely to operate (Peters 2008), as well as comparison‐adjusted funnel plots and network meta‐regression models to assess the presence of small‐study effects in the entire network (Chaimani 2012; Chaimani 2013).

Data synthesis

Direct meta‐analysis

We performed direct meta‐analyses according to the recommendations in Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions, using aggregated data for analyses (Deeks 2011). Where there were sufficient data with enough similarities (in participants, interventions, settings and outcome measurement) between the data, we undertook meta‐analyses using the Review Manager 5 software (RevMan 2014). One review author (MD) entered the data into the software, and a second (AH) checked it for accuracy. We used the random‐effects model to pool the data when meta‐analysis was feasible. We used the Mantel‐Haenszel method for dichotomous outcomes and planned to use the inverse variance method for continuous outcomes. We planned to use the random‐effect model for sensitivity analysis.

Network meta‐analyses

We could not perform a network meta‐analysis because the included studies had important differences in the baseline severity of disease for the participants and in the number of participants undergoing chemotherapy. This raised important concerns about the plausibility of the transitivity assumption in the final dataset and we could not evaluate transitivity statistically because of the small number of trials per comparison. Therefore, we could only perform direct pairwise meta‐analyses of included interventions. For future updates of this review, we will perform network meta‐analysis in Stata (StataCorp 2011) using the method of multivariate meta‐analysis that treats the different comparisons in studies as different outcomes (White 2012).

Subgroup analysis and investigation of heterogeneity

We had planned to perform subgroup analyses and network meta‐regression for each of the following variables in order to explain heterogeneity and/or inconsistency if sufficient studies had been available.

Type of bone marrow failure disorder (MDS, aplastic anaemia, myelofibrosis or congenital bone marrow failure disorder)
Severity of disease
Baseline platelet count
Study precision

Sensitivity analysis

We assessed the robustness of the overall results by performing the following sensitivity analyses where appropriate with respect to those trials deemed to be at high risk of bias. For dichotomous data, we assessed the influence of participant dropout, analysing separately RCTs with less than 20% dropout, RCTs with 20% to 50% dropout and RCTs with greater than 50% dropout. We used the random‐effects model for sensitivity analyses as part of the exploration of heterogeneity.

'Summary of findings' table

We used an approach that extends the GRADE system into network meta‐analysis to build a 'Summary of findings' table, as suggested in Chapter 11 of the Cochrane Handbook for Systematic Reviews of Interventions (Salanti 2014; Schünemann 2011). We included the following outcomes for each type of comparison listed below.

Number of participants with at least one bleeding episode.
Number of participants with life‐threatening or fatal bleeding.
Number of platelet transfusions per participant.
Number of red cell transfusions per participant.
Adverse events – thromboembolism.
Adverse events – transfusion or drug reactions.

Results

Description of studies

See Characteristics of included studies; Characteristics of excluded studies; and Characteristics of ongoing studies.

Results of the search

The database searches identified 4608 references which were reduced to 4104 after duplicates were removed. These references were screened by two review authors (MD, AH) according to the criteria defined above, and we excluded 4012 references as either not an RCT or clearly outside the scope of this review (See PRISMA diagram Figure 1). The full text of the remaining 92 references were obtained. Sixty‐seven were excluded (two review articles, 26 not RCTs, three incorrect intervention, 14 wrong participant group, four ongoing trials and 18 secondary citations of excluded studies). Seven trials that were excluded for being in the wrong participant group (Archimbaud 1999; Geissler 2003; Han 2015; Higby 1974; Miao 2012; Moskowitz 2007; Schiffer 2000) were included in a separate review assessing alternatives, and adjuncts, to prophylactic platelet transfusion for people with haematological malignancies undergoing intensive chemotherapy or stem cell transplantation (Desborough 2016b). In total, seven studies in 25 references were deemed eligible for inclusion (Fricke 1991; Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Mannucci 1986; Platzbecker 2015; Wang 2012). The four ongoing trials are expected to be reported (EudraCT 2010‐022890‐33; EudraCT 2014‐000174‐19; NCT02099747; NCT02158936).

Included studies

Seven completed trials reported in 25 papers were included in the analysis (see Characteristics of included studies for full details of each study).

Design

Seven trials were published as full‐text articles (published in 25 papers) between 1986 and 2015 (Fricke 1991; Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Mannucci 1986; Platzbecker 2015; Wang 2012). All seven were published in English. Three trials were parallel group two‐arm trials (Giagounidis 2014; Greenberg 2013; Platzbecker 2015), two were parallel group three‐arm trials (Kantarjian 2010; Wang 2012) and two were cross‐over trials (Fricke 1991; Mannucci 1986).

Sample sizes

The trials included 472 participants with numbers ranging from eight (Fricke 1991; Mannucci 1986) to 250 (Giagounidis 2014).

Setting

Four trials were conducted in the USA (Fricke 1991; Greenberg 2013; Kantarjian 2010; Wang 2012), one was conducted in Italy and Spain (Mannucci 1986), one was conducted in Australia, Canada, France, Germany, Italy, Poland, UK and the USA (Giagounidis 2014), and one was conducted in Brazil, Denmark, France, Germany, Hong Kong, Italy, South Korea, Taiwan, UK and the USA (Platzbecker 2015).

Participants

Four trials included only participants with MDS (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012), one included participants with MDS and AML if no intensive treatment was planned (Platzbecker 2015), one included aplastic anaemia and MDS (Fricke 1991) and one included aplastic anaemia and familial thrombocytopenia (Mannucci 1986). In two trials, participants were not receiving chemotherapy (Giagounidis 2014; Platzbecker 2015), in three trials participants were treated with low‐dose chemotherapy: azacitidine (Kantarjian 2010), decitabine (Greenberg 2013) and lenalidomide (Wang 2012). In two trials, it was unclear if participants were receiving any other treatment (Fricke 1991; Mannucci 1986).

Interventions

All the interventions included in the review reduce platelet transfusion rather than replace it directly. Four trials compared romiplostim with placebo (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012), and one trial compared eltrombopag with placebo (Platzbecker 2015).Two trials used weekly subcutaneous romiplostim 750 μg (Giagounidis 2014; Greenberg 2013), two trials used weekly subcutaneous romiplostim 500 μg and romiplostim 750 μg (Kantarjian 2010; Wang 2012), and one trial used daily oral eltrombopag 50 mg once daily, which was increased every two weeks based on the patients' platelet and peripheral bone marrow blast counts (doses of 100 mg, 200 mg and 300 mg, or 100 mg and 150 mg for patients of East Asian heritage) (Platzbecker 2015). In one trial, treatment continued for 26 weeks, followed by a four‐week washout, then an optional continuation of 24 weeks (using the same treatment as at the initial randomisation) followed by another four‐week washout (Giagounidis 2014). One trial continued treatment for 26 weeks followed by an optional additional 26 weeks (Platzbecker 2015). One trial continued treatment for the duration of four cycles of decitabine: approximately 16 to 24 weeks (Greenberg 2013), one trial continued treatment for the duration of four cycles of azacitidine: approximately 16 weeks, and one trial continued treatment for the duration of four cycles of lenalidomide: approximately 16 weeks (Wang 2012).

One trial compared tranexamic acid versus placebo (Fricke 1991), and one compared desmopressin (DDAVP) with placebo (Mannucci 1986).

No trials assessed artificial platelet substitutes, platelet‐poor plasma, rFVIIa, rFXIII, interleukin 6, interleukin 11, fibrinogen concentrate.

Outcomes

No trial reported all the outcomes of interest. Four trials reported data for our primary outcome of number and severity of bleeding episodes (Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). Five trials also reported overall mortality, death from bleeding, platelet transfusions and adverse events (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). Four trials reported risk of death from infection (Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012) and one reported red cell transfusion requirements (Platzbecker 2015). One trial did not report any outcomes of interest (Mannucci 1986), and one trial did not report any outcomes in a way that could be interpreted due to methodological flaws in the trial design (Fricke 1991).

Incidence of bleeding and platelet transfusion

The incidence of bleeding in the control arms of the five trials of thrombopoietin (TPO) mimetics ranged from 8% (Wang 2012) to 54% (Kantarjian 2010). The proportion of participants receiving a platelet transfusion in the control arms of the five trials of TPO mimetics ranged from 33% (Wang 2012) to 79% (Platzbecker 2015). The trials of DDAVP versus placebo (Mannucci 1986) and tranexamic acid versus placebo (Fricke 1991) did not report sufficient detail of the burden of bleeding or platelet transfusion in these groups.

Excluded studies

We excluded 67 studies from the review (See Characteristics of excluded studies for further details).

Two studies were review articles (NIHR 2014; Reynolds 2000).
Twenty‐six studies were not randomised controlled trials (ACTRN12610000641099; Antun 2013; Castamann 1997; Cattan 1963; Desmond 2014; Dickinson 2014; EudraCT 2012‐004886‐42; Fenaux 2013; Gerrits 2015; Gordon 1995; Kantarjian 2007; Kantarjian 2012; Kurzrock 2001; Mittelman 2012; Montero 2006; NCT01286038; NCT01481220; Olnes 2012; Pecci 2010; Perez Ruixo 2012; Ramadan 2015; Schrezenmeier 1995; Sekeres 2009; Svensson 2014; Will 2009; Young 1997).
Three studies were incorrect interventions (Khan 2015; NCT01893372; NCT02094417).
Fourteen were trials on the wrong participant group (Archimbaud 1999; Bussel 2007; Chen 2010; Geissler 2003; Giles 2005; Han 2015; Higby 1974; ISRCTN73545489; Miao 2012; Moskowitz 2007; NCT02094248; NCT02578901; Schiffer 2000; Usuki 2007).
Four ongoing studies (EudraCT 2010‐022890‐33; EudraCT 2014‐000174‐19; NCT02099747; NCT02158936).
Eighteen studies were secondary citations for excluded studies.

Ongoing studies

We identified four ongoing studies (see Characteristics of ongoing studies) (EudraCT 2010‐022890‐33; EudraCT 2014‐000174‐19; NCT02099747; NCT02158936). We will monitor the progress of these trials and on publication (assuming eligibility), we will include them in future updates of this review. Two trials are due to be completed by December 2020 (NCT02099747; NCT02158936). Two trials have not reported an expected completion date (EudraCT 2010‐022890‐33; EudraCT 2014‐000174‐19). All four of the ongoing studies are comparing eltrombopag versus placebo in the following settings: low/intermediate risk MDS (EudraCT 2010‐022890‐33), intermediate/high‐risk MDS in combination with azacitidine (NCT02158936), moderate aplastic anaemia (EudraCT 2014‐000174‐19) and severe/very severe aplastic anaemia (NCT02099747). These trials are planning to include 837 participants in total.

Risk of bias in included studies

See the ’Risk of bias’ tables within Characteristics of included studies for details of our assessment for each study and Figure 2 for a tabular summary.

Figure 2

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Sequence generation

Thrombopoietin mimetics

Romiplostim

Four trials were assessed at unclear risk of bias because they did not report details of the randomisation sequence (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012).

Eltrombopag

The trial of eltrombopag was at low risk of bias because it used a permuted block randomisation schedule (Platzbecker 2015).

Tranexamic acid

The trial of tranexamic acid versus placebo was at unclear risk of bias as it did not report sufficient information on sequence generation for a judgement to be made (Fricke 1991).

DDAVP

The trial of DDAVP versus placebo was considered to be at low risk of bias as it used a computer‐generated randomisation sequence (Mannucci 1986).

Concealment of treatment allocation

Thrombopoietin mimetics

Romiplostim

Two trials were at low risk of bias because they used an interactive voice response system (Giagounidis 2014; Wang 2012). Two trials were considered at unclear risk of bias because they did not report sufficient details about concealment of treatment allocation (Greenberg 2013; Kantarjian 2010).

Eltrombopag

The trial of eltrombopag versus placebo was considered to be at low risk of bias as it used an interactive voice response system (Platzbecker 2015).

Tranexamic acid

The trial of tranexamic acid versus placebo was considered to be at unclear risk of bias as insufficient information was reported for concealment of treatment allocation to be assessed (Fricke 1991).

DDAVP

The trial of DDAVP versus placebo was considered at low risk of bias, as it used sealed envelopes for concealment of treatment allocation (Mannucci 1986)

Performance Bias

Participants

Thrombopoietin mimetics

Romiplostim

Two trials were at low risk of bias as the methodology for blinding of participant was adequate (Greenberg 2013; Wang 2012). Two trials were at unclear risk of bias as they stated they were double‐blind, placebo‐controlled trials but did not provide details of the methodology (Giagounidis 2014; Kantarjian 2010).

Eltrombopag

The trial of eltrombopag versus placebo was considered to be at low risk of bias as the methodology for blinding of participants was adequate (Platzbecker 2015).

Tranexamic acid

The trial of tranexamic acid versus placebo was considered at unclear risk of bias. It was double‐blinded and tranexamic acid and placebo were identical in appearance. However, tranexamic acid levels were taken weekly and before transfusion. Overall success of tranexamic acid was defined as either five failures of placebo and none of tranexamic acid or seven failures of placebo and one of tranexamic acid. Overall failure of tranexamic acid was defined as two failed courses of tranexamic acid. Sequential courses continued until overall success or failure of tranexamic acid could be determined. It is unclear how this assessment was performed without unblinding the analysis (Fricke 1991).

DDAVP

The trial of DDAVP versus placebo was considered to be at low risk of bias as it was a double‐blind, placebo‐controlled trial with matching intervention and placebo (Mannucci 1986).

Study Personnel

Thrombopoietin mimetics

Romiplostim

Two trials were at low risk of bias as methodology for blinding of participant was adequate (Greenberg 2013; Wang 2012). Two trials were at unclear risk of bias as they stated they were double‐blind, placebo‐controlled trials but did not provide details of the methodology (Giagounidis 2014; Kantarjian 2010).

Eltrombopag

The trial of eltrombopag versus placebo was considered at low risk of bias as the methodology for blinding of study personnel was adequate (Platzbecker 2015).

Tranexamic acid

The trial of tranexamic acid versus placebo was considered to be at unclear risk of bias. It was double‐blinded and tranexamic acid and placebo were identical in appearance. However, tranexamic acid levels were taken weekly and before transfusion. Overall success of tranexamic acid was defined as either five failures of placebo and none of tranexamic acid or seven failures of placebo and one of tranexamic acid. Overall failure of tranexamic acid was defined as two failed courses of tranexamic acid. Sequential courses continued until overall success or failure of tranexamic acid could be determined. It is unclear how this assessment was performed without unblinding the analysis (Fricke 1991).

DDAVP

The trial of DDAVP versus placebo was considered at low risk of bias as it was a double‐blind, placebo‐controlled trial with matching intervention and placebo (Mannucci 1986).

Blinding of study analysts

Thrombopoietin mimetics

Romiplostim

Two trials were at low risk of bias as methodology for blinding of study analysts was adequate (Greenberg 2013; Wang 2012). Two trials were at unclear risk of bias as they stated they were double‐blind placebo‐controlled trials but did not provide details of the methodology (Giagounidis 2014; Kantarjian 2010).

Eltrombopag

The trial of eltrombopag versus placebo was considered to be at low risk of bias as methodology for blinding of study analysts was adequate (Platzbecker 2015).

Tranexamic acid

DDAVP

The trial of DDAVP versus placebo was considered to be at low risk of bias as it was a double‐blind placebo‐controlled trial with matching intervention and placebo (Mannucci 1986).

Incomplete outcome data

Thrombopoietin mimetics

Romiplostim

Three trials were considered to be at low risk of bias as they analysed data on an intention‐to‐treat basis and all participants were accounted for in the final analysis (Greenberg 2013; Kantarjian 2010; Wang 2012). One trial was considered at unclear risk of bias because it was stopped early (Giagounidis 2014).

Eltrombopag

The trial of eltrombopag versus placebo was considered at low risk of bias as it analysed data on an intention‐to‐treat basis and all participants were accounted for in the final analysis (Platzbecker 2015).

Tranexamic acid

In the trial of tranexamic acid versus placebo was considered at high risk of bias, as only three out of eight participants completed the study (Fricke 1991).

DDAVP

The trial of DDAVP versus placebo was considered to be at low risk of bias, as all participants were accounted for in the final analysis (Mannucci 1986).

Selective reporting

Thrombopoietin mimetics

Romiplostim

Four trials were considered to be at low risk of bias, as all pre‐specified outcomes from their protocols were included in the final manuscript (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012).

Eltrombopag

The trial of eltrombopag versus placebo was considered to be at high risk of bias from selective outcome reporting, as it stated that quality of life would be assessed but this was not included in the final paper (Platzbecker 2015).

Tranexamic acid

The trial of tranexamic acid versus placebo was at high risk of bias as it stated that severity of bleeding would be assessed but this was not included in the final paper (Fricke 1991).

DDAVP

The trial of DDAVP versus placebo was considered to be at unclear risk of bias as the protocol was not available (Mannucci 1986).

Other potential sources of bias

Thrombopoietin mimetics

Romiplostim

Four trials were considered to be at high risk of bias because: in two trials at least one author had served on an advisory board and received honoraria from the drug company sponsor (Giagounidis 2014; Wang 2012); in one trial, one of the authors received payment from the sponsor for writing the manuscript (Greenberg 2013); and in four trials, each of the following was applicable to at least one author: received research funding, worked as a consultant, was an employee, or stockholder in the sponsoring company (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012). Systematic review evidence demonstrates that when clinical trials are sponsored by the manufacturing company, the results are favourable more commonly than when trials have other sources of funding. This potential bias can not be explained by standard 'Risk of bias' assessments (Lundh 2012).

Eltrombopag

The trial of eltrombopag versus placebo was considered to be at high risk of bias because each of the following was applicable to at least one author: served on an advisory board, received honoraria, received research funding, worked as a consultant, was an employee or stockholder in the sponsoring company (Platzbecker 2015).

Tranexamic acid

The trial of tranexamic acid versus placebo was considered to be at high risk of bias as one patient was kept in the study even though they received HLA‐matched platelets (this was pre‐defined as a reason for treatment failure), whereas two other patients were withdrawn after commencing HLA‐matched platelet transfusions. There was significant heterogeneity in the number of courses of treatment each patient received (zero to more than 20) (Fricke 1991).

DDAVP

The trial assessing DDAVP versus placebo was considered to be at unclear risk of other sources of bias. Von Willebrand factor (vWF) levels were two to three times the expected level at baseline which may have reduced the effect of DDAVP, as DDAVP acts by increasing plasma vWF levels (Mannucci 1986).

Effects of interventions

See: Summary of findings 1 Thrombopoietin mimetic versus placebo

No trials assessed artificial platelet substitutes, platelet‐poor plasma, rFVIIa, rFXIII, interleukin 6, interleukin 11 or fibrinogen concentrate.

There were five trials with 456 participants comparing TPO mimetics versus placebo (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). The type of TPO mimetic, dose and duration of administration varied between trials. One trial with eight participants assessed tranexamic acid compared with placebo (Fricke 1991) and one trial with eight participants compared DDAVP with placebo (Mannucci 1986).

In the DDAVP trial, the only outcome reported was the bleeding time (Mannucci 1986). This is a test used to estimate bleeding tendency and is performed by making a small cut in a participant's forearm and timing how long it takes for the bleeding to stop. The bleeding time is no longer used as a clinical test, as it is not considered to be a reliable measure of bleeding risk (Lehman 2001). No clinical outcomes were reported in this trial, so it could not be included in the quantitative synthesis. We did not include the tranexamic acid trial in the quantitative synthesis due to significant methodological flaws: bleeding was not defined equally between the tranexamic acid and control group; only three out of eight participants completed the trial; there was high risk of reporting bias; and 1/8 participants completed the trial despite meeting a pre‐specified reason for exclusion (Fricke 1991).

Network meta‐analysis

We could not perform a network meta‐analysis because the included studies had important differences in the baseline severity of disease for the participants and in the number of participants undergoing chemotherapy. Three trials only included participants undergoing low‐dose chemotherapy (Greenberg 2013; Kantarjian 2010; Wang 2012) and two trials did not include participants undergoing chemotherapy (Giagounidis 2014; Platzbecker 2015). One trial included participants with AML in addition to MDS (Platzbecker 2015), whereas the other trials only included participants with MDS (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012). There was variation through all trials in the international prognostic scoring system (IPSS) of the participants included in the trials. This raised important concerns about the plausibility of the transitivity assumption in the final dataset and we could not evaluate transitivity statistically because of the small number of trials per comparison. Therefore, we could only perform direct pairwise meta‐analyses of included interventions. Full details of the proposed methodology are included in Differences between protocol and review. Consequently, only direct pairwise meta‐analyses were performed and the results of these are described below.

Thrombopoietin (TPO) mimetics

Primary outcomes

The number of participants with at least one bleeding episode

Four trials (206 participants) reported the number of participants with at least one bleeding episode (Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). There was insufficient evidence to detect a difference in the risk of a bleeding episode between those treated with treated with a TPO mimetic and placebo (RR 0.86, 95% CI 0.56 to 1.31, I² = 0%, four trials, 206 participants, low‐quality evidence) (Analysis 1.1; Figure 3).

Figure 3

Thrombopoietin mimetic versus placebo. Number of participants with at least one bleeding episode

Romiplostim

There was insufficient evidence to detect a difference in the number of participants with at least one bleeding episode between romiplostim 500 μg and placebo (RR 1.39, 95% CI 0.58 to 3.28, I² = 0%, two trials, 39 participants) (Kantarjian 2010; Wang 2012), or romiplostim 750 μg and placebo (RR 0.83, 95%CI 0.45 to 1.54, I² = 23%, three trials, 69 participants) (Greenberg 2013; Kantarjian 2010; Wang 2012) (Analysis 1.1; Figure 3).

Eltrombopag

There was insufficient evidence to detect a difference in the number of participants with at least one bleeding episode between eltrombopag and placebo (RR 0.59, 95% CI 0.27 to 1.31, one trial, 98 participants) (Platzbecker 2015) (Analysis 1.1; Figure 3).

The number of participants with at least one episode of severe or life‐threatening bleeding

One trial (40 participants) reported the number of participants with at least one episode of severe or life‐threatening bleeding (Kantarjian 2010). There was insufficient evidence to detect a difference in the risk of a life‐threatening bleed between those treated with a TPO mimetic and placebo (RR 0.31, 95% CI 0.04 to 2.26, one trial, 40 participants, low‐quality evidence) (Kantarjian 2010) (Analysis 1.2).

Romiplostim

There was insufficient evidence to detect a difference in the risk of a life‐threatening bleed between those treated with romiplostim 500 μg and placebo (RR 0.46, 95% CI 0.03 to 6.20, one trial, 19 participants) (Kantarjian 2010) or romiplostim 750 μg and placebo (RR 0.18, 95% CI 0.01 to 3.88, one trial, 21 participants) (Analysis 1.2).

Eltrombopag

Outcome not reported.

Number of days of bleeding per participant

Outcome not reported in any trial.

Secondary outcomes

Mortality

All‐cause mortality

Five trials (456 participants) reported all‐cause mortality (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). There was insufficient evidence to detect a difference in the risk of all‐cause mortality between those treated with a TPO mimetic and placebo (RR 0.74, 95% CI 0.52 to 1.05, I² = 0%, five trials, 456 participants, very low‐quality evidence) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012) (Analysis 1.3; Figure 4).

Figure 4

Thrombopoietin mimetic versus placebo. All‐cause mortality.

Romiplostim

There was no evidence for a difference in all‐cause mortality between romiplostim 500 μg and placebo (RR 0.19, 95%CI 0.19, 95% CI 0.01 to 4.15, I² = 0%, two trials, 40 participants) (Kantarjian 2010; Wang 2012). There was no evidence for a difference in overall mortality between romiplostim 750 μg and placebo (RR 0.81, 95% CI 0.49 to 1.35, I² = 0%, four trials, 318 participants) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012) (Analysis 1.3; Figure 4).

Eltrombopag

There was no evidence for a difference in overall mortality between eltrombopag and placebo (RR 0.70, 95%CI 0.42 to 1.15, one trial, 98 participants) (Platzbecker 2015) (Analysis 1.3; Figure 4).

Mortality due to bleeding

Five trials (457 participants) reported mortality due to bleeding (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). There was no evidence for a difference in mortality due to bleeding between TPO mimetics and placebo (RR 0.44, 95% CI 0.07 to 2.69, I² = 38%, five trials, 457 participants) (Analysis 1.4).

Romiplostim

There were no deaths from bleeding in the intervention arms of either study assessing romiplostim 500 μg (Kantarjian 2010; Wang 2012). There was no evidence for a difference in mortality from bleeding between romiplostim 750 μg and placebo (RR 0.14, 95% CI 0.02 to 1.22 I² = 0%, four trials, 319 participants) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012) (Analysis 1.4).

Eltrombopag

There was no evidence for a difference in mortality due to bleeding between eltrombopag and placebo (RR 1.33, 95% CI 0.27 to 6.49, one trial, 98 participants) (Platzbecker 2015) (Analysis 1.4).

Mortality due to infection

Four trials (206 participants) reported mortality due to infection (Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). There was no evidence for a difference in mortality due to infection between TPO mimetics and placebo (RR 0.62, 95% CI 0.32 to 1.19, I² = 0%, four trials, 206 participants) (Analysis 1.5).

Romiplostim

There was no evidence for a difference in mortality from infection between romiplostim 500 μg and placebo (RR 0.19, 95% CI 0.01 to 4.15, I² = 0%, two trials, 40 participants) (Kantarjian 2010; Wang 2012). There was no evidence for a difference in mortality from infection between romiplostim 750 μg and placebo (RR 0.43, 95% CI 0.06 to 3.24, I² = 0%, three trials, 68 participants) (Greenberg 2013; Kantarjian 2010; Wang 2012) (Analysis 1.5).

Eltrombopag

There was no evidence for a difference in mortality from infection between eltrombopag and placebo (RR 0.69, 95% CI 0.34 to 1.41, one trial, 98 participants) (Platzbecker 2015) (Analysis 1.5).

Platelet transfusions

Proportion of participants requiring a platelet transfusion

Four trials (206 participants) reported the proportion of patients requiring a platelet transfusion (Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). There was a significant reduction in the number of platelet transfusions for those treated with TPO mimetics compared to placebo (RR 0.76, 95% CI 0.61 to 0.95, I² = 0, four trials, 206 participants, moderate‐quality evidence) (Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012) (Analysis 1.6; Figure 5).

Figure 5

Proportion of participants receiving a platelet transfusion.

Romiplostim

There was no evidence for a difference in the proportion of participants requiring a platelet transfusion between romiplostim 500 μg and placebo (RR 0.71, 95%CI 0.35 to 1.44, two trials, 39 participants) (Kantarjian 2010; Wang 2012) or romiplostim 750 μg and placebo (RR 0.70, 95% CI 0.42 to 1.15, I² = 0, three trials, 69 participants) (Greenberg 2013; Kantarjian 2010; Wang 2012) (Analysis 1.6; Figure 5).

Eltrombopag

There was no evidence for a difference in the proportion of participants requiring a platelet transfusion between eltrombopag and placebo (RR 0.79, 95% CI 0.61 to 1.02, one trial, 98 participants) (Platzbecker 2015) (Analysis 1.6; Figure 5).

Number of platelet units transfused per participant

Five trials (456 participants) reported platelet units transfused per participant but with insufficient information for combination into meta‐analysis (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). Results are summarised in Table 1.

Open in table viewer

Table 1. Platelet transfusions per participant

Trial	Inclusion criteria (platelet count)	Duration of follow‐up	Platelet transfusions per participant (mean)
Trial	Inclusion criteria (platelet count)	Duration of follow‐up	Placebo	Romiplostim 500 μg	Romiplostim 750 μg	Eltrombopag
Kantarjian 2010	Any	16 weeks	6.08	8.08	2.43	‐
Wang 2012	Any	16 weeks	3.92	0.21	6.23	‐
Giagounidis 2014	< 20 x 10⁹/L or ≥ 20 and bleeding	26 weeks	15.6	‐	11.1	‐
Greenberg 2013	Any	16‐20 weeks	5.8	‐	3.1	‐
Platzbecker 2015	< 30 x 10⁹/L or transfusion dependent	26 weeks	28.8	‐	‐	37.8

Results reported as mean platelet transfusions per participant. Standard deviations not reported in any trial.

Mean number of platelet transfusion episodes per participant

No trial reported mean number of platelet transfusion episodes as an outcome.

Red cell transfusions

Proportion of participants requiring a red cell transfusion

No trial reported proportion of participants requiring a red cell transfusion.

Number of red cell units transfused per participant

One trial (98 participants) reported proportion of patients requiring a red cell transfusion but with insufficient information for combination into meta‐analysis (Platzbecker 2015). Participants treated with eltrombopag received mean 4.8 units red blood cells whereas those treated with placebo received mean 6.3 units.

Mean number of red cell transfusion episodes per participant

No trial reported mean number of red cell transfusion episodes per participant.

Adverse events

Transfusion reactions

One trial reported transfusion reactions (Platzbecker 2015). There was no evidence for a difference in the incidence of transfusion reactions between those treated with eltrombopag and placebo (Peto odds ratio (pOR) 0.06, 95% CI 0.00 to 3.44, one trial, 98 participants, very low‐quality evidence). The event rate was low with one transfusion reaction reported in total (Analysis 1.7).

Transfusion‐transmitted infection

No trial reported transfusion‐transmitted infection as an outcome.

Thromboembolic events

Five trials (456 participants) reported thromboembolic events (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). There was no evidence for a difference in thromboembolic events between TPO mimetics and placebo (RR 1.41, 95% CI 0.39 to 5.01, I² = 0%, five trials, 456 participants, very low‐quality evidence) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012) (Analysis 1.8; Figure 6).

Figure 6

Thrombopoietin mimetic versus placebo. Thromboembolism.

Romiplostim

There were no thromboembolic events in either study assessing romiplostim 500 μg (Kantarjian 2010; Wang 2012). There was no evidence for a difference in thromboembolic events between romiplostim 750 μg and placebo (RR 1.43, 95% CI 0.23 to 8.77, I² = 0%, four trials, 318 participants) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012) (Analysis 1.8; Figure 6)

Eltrombopag

There was no evidence for a difference in thromboembolic events between eltrombopag and placebo (RR 1.06, 95% CI 0.10 to 11.30, one trial, 98 participants) (Platzbecker 2015) (Analysis 1.8; Figure 6).

Formation of anti‐HLA antibodies

No trial reported formation of anti‐HLA antibodies as an outcome.

Platelet refractoriness

No trial reported platelet refractoriness as an outcome.

Drug reactions

Five trials (455 participants) reported drug reactions (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). There was no evidence for a difference in drug reactions between TPO mimetics and placebo (RR 1.12, 95% CI 0.83 to 1.51, five trials, 455 participants, low‐quality evidence) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012) (Analysis 1.9).

Romiplostim

There was no evidence for a difference in drug reactions between romiplostim 500 μg and placebo (RR 0.62, 95% CI 0.23 to 1.70, I² = 0%, two trials, 40 participants) (Kantarjian 2010; Wang 2012) or romiplostim 750 μg and placebo (RR 1.17, 95% CI 0.80 to 1.70, I² = 0%, four trials, 317 participants) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012) (Analysis 1.9).

Eltrombopag

There was no evidence for a difference in drug reactions between eltrombopag and placebo (RR 0.80, 95% CI 0.24 to 2.63, one trial, 98 participants) (Platzbecker 2015) (Analysis 1.9).

Tranexamic acid

There was one trial of tranexamic acid versus placebo (Fricke 1991). We extracted no data from this study due to major methodological problems in the study design. In addition to the high risk of bias in terms of attrition bias, reporting bias and other bias (see text section above, Figure 2 and the Risk of bias in included studies table), there was a variable number of study cycles depending on the results of previous cycles of treatment. All these factors meant that it was impossible to fully understand the data in this trial and we took the decision not to include this trial in the assessment of 'effects of interventions'.

DDAVP

There was one trial of DDAVP versus placebo (Mannucci 1986). This study had a single outcome: bleeding time. It did not assess any of the outcomes of interest in this review. Consequently, no data could be extracted from this trial.

Discussion

Network meta‐analysis

We could not perform a network meta‐analysis because the included studies had important differences in the baseline severity of disease for the participants, and in the number of participants undergoing chemotherapy. This raised important concerns about the plausibility of the transitivity assumption in the final dataset and we could not evaluate transitivity statistically because of the small number of trials per comparison. Therefore, we could only perform direct pairwise meta‐analyses of included interventions.

Pairwise meta‐analysis

Each of the interventions identified in this review have the potential to reduce platelet transfusion, but are not direct alternatives.

Thrombopoietin (TPO) mimetics

Five trials reported in 23 papers reported the use of TPO mimetics (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012).

Efficacy

There was insufficient evidence to detect a difference in the risk of a life‐threatening bleed between those treated with a TPO mimetic and placebo (RR 0.31, 95% CI 0.04 to 2.26, one trial, 40 participants, low‐quality evidence) (Kantarjian 2010). Severe or life‐threatening bleeding events were rare and consequently, the quality of the evidence was reduced by imprecision. There was a small improvement in the RR of severe or life‐threatening bleeding as the romiplostim dose increased but the event numbers were too low to assess if this is a dose‐related effect.

There was insufficient evidence to detect a difference in the risk of all‐cause mortality between those treated with a TPO mimetic and placebo (RR 0.74, 95% CI 0.52 to 1.05, five trials, 456 participants, very low‐quality evidence) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012).

There was insufficient evidence to detect a difference in mortality from bleeding (RR 0.44, 95% CI 0.07 to 2.69, five trials, 457 participants) or mortality from infection (RR 0.62, 95% CI 0.32 to 1.19, four trials, 206 participants) (Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012) between those treated with TPO mimetics and placebo.

There was a significant reduction in the number of participants receiving any platelet transfusion between those treated with TPO mimetics and placebo (RR 0.76, 95% CI 0.61 to 0.95, four trials, 206 participants, moderate‐quality evidence) (Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). Data for units of platelets and units of red blood cells per participant were not reported adequately for meta‐analysis.

Adverse events

There was no evidence for a difference in the incidence of transfusion reactions between those treated with TPO mimetics and placebo (Peto odds ratio (pOR) 0.06, 95% CI 0.00 to 3.44, one trial, 98 participants, very low‐quality evidence) (Platzbecker 2015), but with only a single reported transfusion reaction, interpretation of this finding is limited by imprecision.

There was no evidence for a difference in thromboembolic events between TPO mimetics and placebo (RR 1.41, 95% CI 0.39 to 5.01, five trials, 456 participants, very low‐quality evidence) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012).

There was no evidence for a difference in drug reactions between TPO mimetics and placebo (RR 1.12, 95% CI 0.83 to 1.51, five trials, 455 participants, low quality evidence) (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012).

No trial reported number of days bleeding per participant, platelet transfusion episodes, mean red cell transfusions per participant, red cell transfusion episodes, transfusion‐transmitted infections, formation of antiplatelet antibodies or platelet refractoriness.

Tranexamic acid

One trial reported tranexamic acid (Fricke 1991). We did not include this trial in the quantitative synthesis due to significant methodological flaws: bleeding was not defined equally between the tranexamic acid and control group; only three out of eight participants completed the trial; there was high risk of reporting bias; and 1/8 participants completed the trial despite meeting a pre‐specified reason for exclusion.

DDAVP

One trial reported DDAVP (Mannucci 1986). The only outcome reported in this trial was the bleeding time. This is a test used to estimate bleeding tendency and is performed by making a small cut in a participant's forearm and timing how long it takes for the bleeding to stop. The bleeding time is no longer used as a clinical test, as it is not considered to be a reliable measure of bleeding risk (Lehman 2001). No clinical outcomes were reported in this trial, so it was not included in the quantitative synthesis.

Other agents

No trial assessed artificial platelets, platelet‐poor plasma, activated factor VII, fibrinogen concentrate or recombinant interleukin 6 or 11. The remainder of this discussion will focus only on TPO mimetic as there is inadequate trial data for assessing the other interventions.

Overall completeness and applicability of evidence

Four trials only included participants with MDS and one included participants with MDS and AML. The severity of MDS varied between trials. One trial planned to only include IPSS low and intermediate‐1 MDS, but recruited one participant with IPSS intermediate‐2 disease (Giagounidis 2014). Three trials aimed to recruit participants with IPSS low, intermediate‐1 and intermediate‐2 risk disease (Greenberg 2013; Kantarjian 2010; Wang 2012), although one of these trials recruited a single participant with IPSS high‐risk disease (Greenberg 2013). One trial did not use the IPSS classification but included relatively high‐risk MDS (refractory anaemia with an excess of blasts) and AML (Platzbecker 2015). No participants with aplastic anaemia or congenital bone marrow failure conditions were included. Two trials excluded participants receiving chemotherapy (Giagounidis 2014; Platzbecker 2015). The remaining three trials included participants who were receiving low‐dose chemotherapy: azacitidine (Kantarjian 2010), decitabine (Greenberg 2013) and lenalidomide (Wang 2012). All of the trials only included adults. No trials assessed artificial platelets, platelet‐poor plasma, activated factor VII, fibrinogen concentrate or recombinant interleukin 6 or 11. Trials were identified of tranexamic acid (data not extracted because of methodological flaws) (Fricke 1991) and DDAVP (data not extracted as none of the outcomes of interest in this review were reported) (Mannucci 1986).

Four trials (206 participants) reported the number of participants with any bleeding episode. There was no evidence for a difference in this outcome and in order to demonstrate a reduction in number of participants with any bleeding episode from 26 in 100 to 16 in 100 participants (as seen in the eltrombopag data), a study would need to recruit 514 participants (80% power, 5% significance). Eight hundred and thirty‐seven further participants are due to be recruited into future trials by December 2020, 451 of which are of eltrombopag in MDS, so this question may be answered when these data are available.

One trial (40 participants) reported the number of participants with a severe/life‐threatening bleeding episode. There was no evidence for a difference in the risk of severe/life‐threatening bleeding between TPO mimetics and placebo. In order to demonstrate a reduction in severe or life‐threatening bleeding events from 15 in 100 to 8 in 100 (as seen with romiplostim 500 μg), a study would need to recruit 646 participants (80% power, 5% significance). This question may be answered once the four ongoing trials are completed.

Five trials (456 participants) reported overall mortality. There was no evidence for a difference in the risk of overall mortality between those treated with TPO mimetics and placebo. In order to demonstrate a reduction in overall mortality from 24 in 100 to 19 in 100 (as seen with the pooled TPO mimetic data), a study would need to recruit 2114 participants (80% power, 5% significance) and consequently, even with the additional data provided by the four ongoing trials, this question is unlikely to be answered.

Four trials (206 participants) reported the number of participants who received a platelet transfusion. A reduction was noted in the number of participants who received a platelet transfusion between those treated with TPO mimetics and placebo. In order to demonstrate a reduction in number of participants who received a platelet transfusion from 66 in 100 to 50 in 100 (as seen with the pooled TPO mimetic data), a study would need to recruit 292 participants (80% power, 5% significance). Consequently, these studies are close to being adequately powered and it would be expected that this question will be answered when the ongoing trials have been reported. A reduction in the number of participants receiving platelet transfusion would be clinically significant.

Transfusion reactions were rare events with only a single reaction recorded in any trial. It is unlikely that a significant difference will found in transfusion reactions, even once future trials are published. Five trials (456 participants) reported thromboembolism. We found no evidence for a difference in the risk of thromboembolism between participants treated with a TPO mimetics and control. In order to detect an increase in thrombosis incidence from 2 in 100 to 4 in 100 (as seen with the pooled TPO mimetic data), a study would need to recruit 2278 participants (80% power, 5% significance). Even with the addition of data from the ongoing trials, there will be insufficient data to determine if this increase in the risk of thromboembolism is present.

Five trials (455 participants) reported drug reactions. No significant difference was found between the groups. Five trials (455 participants) reported units of platelets per participant and one trial (98 participants) reported number of participants with at least one red cell transfusion, but not in a way that could be incorporated into meta‐analysis.

No trial reported days of bleeding per participant, platelet transfusion episodes, mean red cell transfusions per participant, red cell transfusion episodes, transfusion‐transmitted infections, formation of antiplatelet antibodies or platelet refractoriness.

Quality of the evidence

We considered all trials of TPO mimetics to be at high risk of bias because in three trials at least one author had served on an advisory board and received honoraria from the drug company sponsor (Giagounidis 2014; Platzbecker 2015; Wang 2012); in one trial, one of the authors received payment from the sponsor for writing the manuscript (Greenberg 2013) and in five trials, at least one author had received research funding, worked as a consultant, was an employee, and stockholder in the sponsoring company (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Platzbecker 2015; Wang 2012). Systematic review evidence demonstrates that when clinical trials are sponsored by the manufacturing company, the results are favourable more commonly than when trials have other sources of funding. This potential bias can not be explained by standard 'Risk of bias' assessments (Lundh 2012). One trial was also considered to be at risk of reporting bias, as it did not report all the outcomes prespecified in its protocol (Platzbecker 2015). Consequently, we downgraded all outcomes one point for risk of bias (Figure 2).

We did not downgrade the number of platelet transfusions per participant for imprecision, as it was close to being adequately powered. We downgraded all other outcomes one point for imprecision as they were underpowered. We downgraded overall mortality, transfusion reactions and thromboembolism two points due to imprecision.

We downgraded transfusion reactions one point due to indirectness, as this outcome was dependent on a significant difference in the number of blood component transfusions between those receiving TPO mimetics and placebo. A full summary of quality of the evidence is included in summary of findings Table 1.

Potential biases in the review process

There were no obvious biases within the review process. We conducted a wide search and the relevance of each paper identified was carefully assessed, and no restrictions were made for the language in which the paper was originally published. Original authors and sponsors were given the opportunity to provide additional data to clarify the results of their trials but none put forward any new information. We could not formally assess publication bias, as our primary outcomes were reported in four papers, and only for trials assessing TPO mimetics. The interim results of one ongoing trial were published in 2012 but we do not believe this represents publication bias because we are not expecting the trial to be published until at least 2019 (the trial opened in France in June 2014 and follow‐up for individual participants is for five years) (EudraCT 2010‐022890‐33).

Agreements and disagreements with other studies or reviews

A systematic review and meta‐analysis of TPO mimetics for MDS published in 2014 (Prica 2014), assessed the same four trials of romiplostim that were included in this review (Giagounidis 2014; Greenberg 2013; Kantarjian 2010; Wang 2012). At the time of publication of that review no data had been published on eltrombopag. The authors noted no evidence for a difference in the risk of bleeding between those treated with romiplostim and placebo, although they did note a reduction in expose‐adjusted bleeding rate. They also noted a reduction in exposure adjusted platelet transfusion rate. With the exception of number of participants receiving any platelet transfusion, these meta‐analyses are underpowered and the results of ongoing studies are likely to lead to adequately powered meta‐analysis for bleeding risk. A further detailed systematic review of TPO mimetics in MDS is presently underway (Dodillet 2012).

Figure 1

Study flow diagram.

Navigate to figure in ReviewOpen in new tab

Figure 2

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Navigate to figure in ReviewOpen in new tab

Figure 3

Thrombopoietin mimetic versus placebo. Number of participants with at least one bleeding episode

Navigate to figure in ReviewOpen in new tab

Figure 4

Thrombopoietin mimetic versus placebo. All‐cause mortality.

Navigate to figure in ReviewOpen in new tab

Figure 5

Proportion of participants receiving a platelet transfusion.

Navigate to figure in ReviewOpen in new tab

Figure 6

Thrombopoietin mimetic versus placebo. Thromboembolism.

Navigate to figure in ReviewOpen in new tab

Analysis 1.1

Comparison 1: Thrombopoietin mimetic versus placebo, Outcome 1: Number of participants with at least one bleeding episode

Navigate to figure in ReviewOpen in new tab

Analysis 1.2

Comparison 1: Thrombopoietin mimetic versus placebo, Outcome 2: Number of participants with at least one severe or life‐threatening bleeding episode

Navigate to figure in ReviewOpen in new tab

Analysis 1.3

Comparison 1: Thrombopoietin mimetic versus placebo, Outcome 3: All‐cause mortality

Navigate to figure in ReviewOpen in new tab

Analysis 1.4

Comparison 1: Thrombopoietin mimetic versus placebo, Outcome 4: Mortality due to bleeding

Navigate to figure in ReviewOpen in new tab

Analysis 1.5

Comparison 1: Thrombopoietin mimetic versus placebo, Outcome 5: Mortality due to infection

Navigate to figure in ReviewOpen in new tab

Analysis 1.6

Comparison 1: Thrombopoietin mimetic versus placebo, Outcome 6: Proportion of participants receiving a platelet transfusion

Navigate to figure in ReviewOpen in new tab

Analysis 1.7

Comparison 1: Thrombopoietin mimetic versus placebo, Outcome 7: Transfusion reactions

Navigate to figure in ReviewOpen in new tab

Analysis 1.8

Comparison 1: Thrombopoietin mimetic versus placebo, Outcome 8: Thromboembolism

Navigate to figure in ReviewOpen in new tab

Analysis 1.9

Comparison 1: Thrombopoietin mimetic versus placebo, Outcome 9: Drug reactions

Navigate to figure in ReviewOpen in new tab

Summary of findings 1. Thrombopoietin mimetic versus placebo

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments

Patient or population: People with chronic bone marrow failure Intervention: Thrombopoietin mimetics Comparison: Placebo
Outcomes	Risk with placebo	Risk with Thrombopoietin mimetic	Relative effect (95% CI)	№ of participants (studies)	Quality of the evidence (GRADE)	Comments
Number of participants with at least one bleeding episode	Study population		RR 0.86 (0.56 to 1.31)	206 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}
Number of participants with at least one bleeding episode	315 per 1000	271 per 1000 (176 to 413)	RR 0.86 (0.56 to 1.31)	206 (4 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}
Number of participants with at least one severe or life‐threatening bleeding episode	Study population		RR 0.31 (0.04 to 2.26)	40 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
	154 per 1000	48 per 1000 (6 to 348)	RR 0.31 (0.04 to 2.26)	40 (1 RCT)	⊕⊕⊝⊝ LOW ^{1 2}
All‐cause mortality	Study population		RR 0.74 (0.52 to 1.05)	456 (5 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3}
All‐cause mortality	237 per 1000	176 per 1000 (123 to 249)	RR 0.74 (0.52 to 1.05)	456 (5 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3}
Proportion of participants receiving a platelet transfusion	Study population		RR 0.76 (0.61 to 0.95)	206 (4 RCTs)	⊕⊕⊕⊝ MODERATE ¹
Proportion of participants receiving a platelet transfusion	658 per 1000	500 per 1000 (401 to 625)	RR 0.76 (0.61 to 0.95)	206 (4 RCTs)	⊕⊕⊕⊝ MODERATE ¹
Transfusion reactions	Study population		pOR 0.06 (0.00 to 3.44)	98 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 3 4}
Transfusion reactions	29 per 1000	2 per 1000 (0 to 94)	pOR 0.06 (0.00 to 3.44)	98 (1 RCT)	⊕⊝⊝⊝ VERY LOW ^{1 3 4}
Thromboembolism	Study population		RR 1.41 (0.39 to 5.01)	456 (5 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3}
Thromboembolism	19 per 1000	27 per 1000 (8 to 97)	RR 1.41 (0.39 to 5.01)	456 (5 RCTs)	⊕⊝⊝⊝ VERY LOW ^{1 3}
Drug reactions	Study population		RR 1.12 (0.83 to 1.51)	455 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}
Drug reactions	271 per 1000	303 per 1000 (225 to 409)	RR 1.12 (0.83 to 1.51)	455 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}
Number of red cell transfusions per participant	Meta‐analysis not possible		Not estimable	98 (1 RCT)	‐	Participants treated with eltrombopag received mean 4.8 units red blood cells whereas those treated with placebo received mean 6.3 units over 6 months
*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). CI: Confidence interval; RR: Risk ratio; pOR: Peto Odds ratio;
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
¹ Downgraded one point due to risk of bias ² Downgraded one point due to imprecision ³ Downgraded two points due to low event rate ⁴ Downgraded one point due to indirectness

Summary of findings 1. Thrombopoietin mimetic versus placebo

Navigate to table in Review

Table 1. Platelet transfusions per participant

Trial	Inclusion criteria (platelet count)	Duration of follow‐up	Platelet transfusions per participant (mean)
Trial	Inclusion criteria (platelet count)	Duration of follow‐up	Placebo	Romiplostim 500 μg	Romiplostim 750 μg	Eltrombopag
Kantarjian 2010	Any	16 weeks	6.08	8.08	2.43	‐
Wang 2012	Any	16 weeks	3.92	0.21	6.23	‐
Giagounidis 2014	< 20 x 10⁹/L or ≥ 20 and bleeding	26 weeks	15.6	‐	11.1	‐
Greenberg 2013	Any	16‐20 weeks	5.8	‐	3.1	‐
Platzbecker 2015	< 30 x 10⁹/L or transfusion dependent	26 weeks	28.8	‐	‐	37.8
Results reported as mean platelet transfusions per participant. Standard deviations not reported in any trial.

Table 1. Platelet transfusions per participant

Navigate to table in Review

Comparison 1. Thrombopoietin mimetic versus placebo

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Number of participants with at least one bleeding episode Show forest plot	4	206	Risk Ratio (M‐H, Random, 95% CI)	0.86 [0.56, 1.31]

1.1.1 Romiplostim 500 μg	2	39	Risk Ratio (M‐H, Random, 95% CI)	1.39 [0.58, 3.28]
1.1.2 Romiplostim 750 μg	3	69	Risk Ratio (M‐H, Random, 95% CI)	0.83 [0.45, 1.54]
1.1.3 Eltrombopag	1	98	Risk Ratio (M‐H, Random, 95% CI)	0.59 [0.27, 1.31]
1.2 Number of participants with at least one severe or life‐threatening bleeding episode Show forest plot	1	40	Risk Ratio (M‐H, Random, 95% CI)	0.31 [0.04, 2.26]

1.2.1 Romiplostim 500 μg	1	19	Risk Ratio (M‐H, Random, 95% CI)	0.46 [0.03, 6.20]
1.2.2 Romiplostim 750 μg	1	21	Risk Ratio (M‐H, Random, 95% CI)	0.18 [0.01, 3.88]
1.3 All‐cause mortality Show forest plot	5	456	Risk Ratio (M‐H, Random, 95% CI)	0.74 [0.52, 1.05]

1.3.1 Romiplostim 500 μg	2	40	Risk Ratio (M‐H, Random, 95% CI)	0.19 [0.01, 4.15]
1.3.2 Romiplostim 750 μg	4	318	Risk Ratio (M‐H, Random, 95% CI)	0.81 [0.49, 1.35]
1.3.3 Eltrombopag	1	98	Risk Ratio (M‐H, Random, 95% CI)	0.70 [0.42, 1.15]
1.4 Mortality due to bleeding Show forest plot	5	457	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.07, 2.69]

1.4.1 Romiplostim 500 μg	2	40	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
1.4.2 Romiplostim 750 μg	4	319	Risk Ratio (M‐H, Random, 95% CI)	0.14 [0.02, 1.22]
1.4.3 Eltrombopag	1	98	Risk Ratio (M‐H, Random, 95% CI)	1.33 [0.27, 6.49]
1.5 Mortality due to infection Show forest plot	4	206	Risk Ratio (M‐H, Random, 95% CI)	0.62 [0.32, 1.19]

1.5.1 Romiplostim 500 μg	2	40	Risk Ratio (M‐H, Random, 95% CI)	0.19 [0.01, 4.15]
1.5.2 Romiplostim 750 μg	3	68	Risk Ratio (M‐H, Random, 95% CI)	0.43 [0.06, 3.24]
1.5.3 Eltrombopag	1	98	Risk Ratio (M‐H, Random, 95% CI)	0.69 [0.34, 1.41]
1.6 Proportion of participants receiving a platelet transfusion Show forest plot	4	206	Risk Ratio (M‐H, Random, 95% CI)	0.76 [0.61, 0.95]

1.6.1 Romiplostim 500 μg	2	39	Risk Ratio (M‐H, Random, 95% CI)	0.71 [0.35, 1.44]
1.6.2 Romiplostim 750 μg	3	69	Risk Ratio (M‐H, Random, 95% CI)	0.70 [0.42, 1.15]
1.6.3 Eltrombopag	1	98	Risk Ratio (M‐H, Random, 95% CI)	0.79 [0.61, 1.02]
1.7 Transfusion reactions Show forest plot	1		Peto Odds Ratio (Peto, Fixed, 95% CI)	Totals not selected

1.7.1 Eltrombopag	1		Peto Odds Ratio (Peto, Fixed, 95% CI)	Totals not selected
1.8 Thromboembolism Show forest plot	5	456	Risk Ratio (M‐H, Random, 95% CI)	1.41 [0.39, 5.01]

1.8.1 Romiplostim 500 μg	2	40	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
1.8.2 Romiplostim 750 μg	4	318	Risk Ratio (M‐H, Random, 95% CI)	1.43 [0.23, 8.77]
1.8.3 Eltrombopag	1	98	Risk Ratio (M‐H, Random, 95% CI)	1.06 [0.10, 11.30]
1.9 Drug reactions Show forest plot	5	455	Risk Ratio (M‐H, Random, 95% CI)	1.12 [0.83, 1.51]

1.9.1 Romiplostim 500 μg	2	40	Risk Ratio (M‐H, Random, 95% CI)	0.62 [0.23, 1.70]
1.9.2 Romiplostim 750 μg	4	317	Risk Ratio (M‐H, Random, 95% CI)	1.17 [0.80, 1.70]
1.9.3 Eltrombopag	1	98	Risk Ratio (M‐H, Random, 95% CI)	0.80 [0.24, 2.63]

Comparison 1. Thrombopoietin mimetic versus placebo

Navigate to table in Review

Cochrane Review language

Website language

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

PICOs

PICOs

Population

Intervention

Comparison

Outcome

Plain language summary

Alternative agents instead of platelet transfusions to prevent bleeding for people who have bone marrow disorders and low platelet counts

Visual summary

Authors' conclusions

Implications for practice

Implications for research

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Artificial platelet substitutes

Platelet‐poor plasma (PPP)

Recombinant factor VIIa (rFVIIa)

Fibrinogen concentrate

Recombinant factor XIII (rFXIII)

Desmopressin (DDAVP)

Thrombopoietin (TPO) mimetics

Antifibrinolytic drugs

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Bibliographic databases

Ongoing studies:

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Relative treatment effects

Relative treatment ranking

Unit of analysis issues

Cross‐over trials

Cluster‐randomised trials

Studies with multiple treatment groups

Dealing with missing data

Assessment of heterogeneity

Assessment of clinical and methodological heterogeneity within treatment comparisons

Assumptions when estimating the heterogeneity

Measures and tests for heterogeneity

Assessment of transitivity across treatment groups

Assessment of reporting biases

Data synthesis

Direct meta‐analysis

Network meta‐analyses

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

'Summary of findings' table

Results

Description of studies

Results of the search

Included studies

Design

Sample sizes