Types of studies

We included randomised controlled trials (RCTs). We applied no restrictions on language or publication status.

Types of participants

We included people with haematological malignancies receiving treatment with intensive chemotherapy or stem cell transplantation (or both). We included participants of all ages, and included both inpatients and outpatients.

When trials consisted of mixed populations of participants (e.g. people with diagnoses of solid tumours), we used only data from the haematological subgroups. If subgroup data for haematological participants were not provided (after contacting the authors of the trial), we excluded trials if less than 80% of participants had a haematological disorder. We excluded any participants that were not treated with intensive chemotherapy or a stem cell transplant as these participants are the focus of another review (Desborough 2016). We included participants with non‐malignant haematological disorders (e.g. aplastic anaemia, congenital bone marrow failure syndromes) that were being treated with an allogeneic stem cell transplant. We also excluded people treated with non‐intensive chemotherapy (such as azacitidine, decitabine and lenalidomide) because the degree of thrombocytopenia is much less profound with a lower risk of bleeding. Trials in people with chronic bone marrow failure using these agents are reported in another review (Desborough 2016).

Types of interventions

We included the two following comparisons:

• alternative agent versus prophylactic platelet transfusion;

• alternative agent and prophylactic platelet transfusion versus placebo or no treatment and prophylactic platelet transfusion.

We considered the following interventions:

• experimental intervention: alternative agents:

  • artificial platelet substitutes;

  • PPP;

  • rFVIIa;

  • fibrinogen;

  • TPO mimetics;

  • DDAVP.

We placed no restrictions on the dose of alternative agents used.

• Comparator intervention:

  • comparison one: alternative agent versus prophylactic platelet transfusion. The comparator was prophylactic platelet transfusions. Transfusion of platelet concentrates, prepared either from individual units of whole blood or by apheresis, and given prophylactically to prevent bleeding. Prophylactic platelet transfusions are typically given when blood platelet counts fall below a given trigger level. There was no restriction on the dose or frequency of platelet transfusions, neither was there a restriction on the transfusion trigger level, although we took this information into account in the analysis where available;

  • comparison two: alternative agent and prophylactic platelet transfusion versus placebo or no treatment and prophylactic platelet transfusion. The comparator was prophylactic platelet transfusions and placebo or no treatment. There was no restriction on the dose or frequency of platelet transfusions used in addition to the alternative agents, but the dose of prophylactic platelet transfusions received and the platelet transfusion threshold at which they were given was the same in both arms of the trial.

Types of outcome measures

Electronic searches

Searching other resources

Selection of studies

We updated the selection of studies from that performed for the previous version of this review (Estcourt 2012b).

Two review authors (MD, LE) independently performed an initial screen of all electronically derived citations and abstracts of papers identified by the review search strategy for relevance. We excluded clearly irrelevant trials at this stage.

Two review authors (MD, LE) independently assessed the full texts of all potentially relevant trials formally for eligibility against the criteria. We resolved all disagreements by discussion with a third review author (SS). We sought further information from trial authors if the article contained insufficient data to make a decision about eligibility. We designed a trial eligibility form for trials of platelet transfusion to help in the assessment of relevance, which included ascertaining whether the participants had haematological malignancies, and whether the two groups could be defined in the trial on the basis of use of an alternative agent to prophylactic platelet transfusions. We recorded the reasons why potentially relevant trials did not meet the eligibility criteria.

Data extraction and management

The data extraction was updated from that performed for the previous version of this review (Estcourt 2012b). This included data extraction for all trials that were included since the previous review and also for all review outcomes that were not part of the previous review (e.g. platelet transfusion interval, quality of life).

Two review authors (MD, LE) independently conducted data extraction according to the guidelines proposed in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). We resolved disagreements between the review authors by consensus without the need for a third review author. The review authors were not blinded to names of authors, institutions, journals or outcomes of the trials. The data extraction forms were piloted in the previous version of this review (Estcourt 2012b). Due to minor changes in the format, the forms were piloted on a further trial, thereafter the two review authors (MD, LE) extracted data independently for all the trials as follows.

Assessment of risk of bias in included studies

The 'Risk of bias' assessment was updated from that performed for the previous version of this review (Estcourt 2012b).

Two review authors (MD, LE) assessed all newly included trials for possible risk of bias (as described in the Cochrane Handbook for Systematic Reviews of Interventions, (Higgins 2011b). The assessment included information about the design, conduct and analysis of the trial. Each criterion was evaluated on a three‐point scale: low risk of bias, high risk of bias, or unclear risk of bias. To assess risk of bias, we included the following questions in the 'Risk of bias' table for each included trial.

• Was the allocation sequence adequately generated?

• Was allocation adequately concealed?

• Was knowledge of the allocated intervention adequately prevented during the trial (including an assessment of blinding of participants, personnel and outcome assessors)?

• Were incomplete outcome data adequately addressed (for every outcome separately)?

• Were reports of the trial free of selective outcome reporting?

• Was the trial apparently free of other problems that could put it at risk of bias?

Measures of treatment effect

For dichotomous outcomes, we recorded the number of outcomes in the treatment and control groups and estimated the treatment effect measures across individual trials as the relative effect measures (odds ratio (OR) with 95% confidence interval (CI)).

If continuous outcomes had been reported in a way that permitted analysis, we intended to record the mean and standard deviations (SD). For continuous outcomes measured using the same scale, the effect measure would have been the mean difference (MD) with 95% CI, or for outcomes measured using different scales, the effect measure would have been the standardised mean difference (SMD). For time‐to‐event outcomes, we planned to extracted the hazard ratio (HR) from published data according to Parmar 1998 and Tierney 2007. When appropriate, we planned to assess the number needed to treat for an additional beneficial outcome (NNTB) with CIs and the number needed to treat for an additional harmful outcome (NNTH) with CIs.

Unit of analysis issues

We did not prespecify in the original protocol how we would deal with any unit of analysis issues. In one trial, there were unit of analysis issues as three participants were re‐randomised; we used only data from one randomisation in the quantitative analysis (Higby 1974).

The trials by Geissler (Geissler 2003‐T1; Geissler 2003‐T2) included 47 participants who had previously been treated in the trials by Archimbaud (Archimbaud 1999‐T1; Archimbaud 1999‐T2), so we did not combine data from these trials in a meta‐analysis.

We did not prespecify in the original protocol how we would deal with multi‐arm trials. For two outcomes that included multi‐arm trials (duration of thrombocytopenia and proportion of participants in complete remission), we split participants from the control arm equally between the intervention arms.

Dealing with missing data

We dealt with missing data according to the recommendations in 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 participants with haematological malignancies as well as participants with solid tumours or non‐malignant haematological disorders, we extracted data for the malignant haematology subgroup from the general trial data. When this could not be done, we contacted the trial author.

Within an outcome, when there were missing data, the preferred analysis was intention‐to‐treat (ITT) analysis. We recorded the number of participants lost to follow‐up for each trial.

Assessment of heterogeneity

If trials were sufficiently homogenous in their trial design, we planned to conduct meta‐analysis and assess the statistical heterogeneity (Deeks 2011). However, due to problems with the way trials were reported, we performed no meta‐analyses. We planned to assess statistical heterogeneity of treatment effects between trials using a Chi² test with a significance level at P < 0.1. We planned to use the I² statistic to quantify possible heterogeneity (I² > 50% moderate heterogeneity, I² > 80% considerable heterogeneity). When necessary, we intended to explore potential causes of heterogeneity by sensitivity and subgroup analyses. This was not possible with the final data set that was obtained.

Assessment of reporting biases

We did not assess for potential publication bias (small‐trial bias) by generating a funnel plot, and statistically test using a linear regression test because the search identified an insufficient number of adequately reported trials reporting the primary outcome. We will perform such testing in future updates of this review if the search identifies at least 10 trials reporting the primary outcome. If this is the case, we will consider a P < 0.1 significant for this test (Sterne 2011).

Data synthesis

We performed analyses according to the recommendations of Cochrane (Deeks 2011). For statistical analysis, we entered data into Review Manager 5 (RevMan 2012).

• We used the random‐effects model for pooling the data, using the Mantel‐Haenszel method for dichotomous outcomes, and the inverse variance method for continuous outcomes.

• We used the random‐effects model for sensitivity analyses as part of the exploration of heterogeneity. When heterogeneity was above 80%, we did not perform a meta‐analysis and commented on the results as a narrative.

We used GRADEprofiler to create 'Summary of findings' tables as suggested in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2011). We reported 'Summary of findings' tables that included the following outcomes.

• Number of participants with at least one bleeding episode.

• Total number of days on which bleeding occurred per participant.

• Number of participants with at least one episode of severe or life‐threatening bleeding.

• Time from randomisation to first bleeding episode.

• All‐cause mortality.

• Number of platelet transfusions per participant within 30 days from the start of the trial.

• Adverse effects: thromboembolic events.

• Quality of life.

For future updates, we will produce separate 'Summary of findings' tables for each type of alternative agent if the search identifies trials of these agents.

Subgroup analysis and investigation of heterogeneity

The studies did not report data in sufficient detail to perform subgroup analysis or to investigate heterogeneity. For future reviews, we intend to use the following methodology.

Two subgroup analyses: we will use fever and participants' diagnostic and treatment subgroups. We will consider performing subgroup analyses on the following characteristics:

• presence of fever (greater than 38 °C);

• underlying disease;

• type of treatment (autologous HSCT, allogeneic HSCT, or chemotherapy alone);

• age of the participant (paediatric, adults, older adults (over 60 years)).

We did not perform meta‐regression as no subgroup contained more than 10 trials (Deeks 2011). For future updates of this review, if the search identifies sufficient trials for subgroup analysis, we will compare differences between subgroups using a random‐effects model when the two subgroups are independent following the guidance in the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). If this is not possible, then we will comment on the differences as a narrative.

We did not perform as investigation of heterogeneity between trials but for future updates this will include:

• age of the trial (as the type of platelet component has changed since the mid‐1970s);

• different platelet component doses;

• different prophylactic platelet transfusion thresholds.

Sensitivity analysis

We did not perform a sensitivity analysis due to an inadequate number of adequately reported trials. For future updates of this review, we will assess robustness of the overall results with sensitivity analysis with respect to those trials deemed to be at high risk of bias.

For dichotomous data, we will assess the influence of participant drop‐out, analysing separately RCTs with less than 20% drop‐out, RCTs with 20% to 50% drop‐out and RCTs with greater than 50% drop‐out. We will use the random‐effects model for sensitivity analyses as part of the exploration of heterogeneity.