Reporting conflicts of interest in randomised trials of patient blood management interventions in patients requiring major surgery: a systematic review and meta-analysis

Abstract

Objective This study aimed to systematically review the effects of declared and undeclared conflicts of interest on randomised controlled trials (RCTs) of patient blood management (PBM) interventions.Design We performed a secondary analysis of a recently published meta-analysis of RCTs evaluating five common PBM interventions in patients undergoing major surgery.Data sources The databases searched by the original systematic reviews were searched using subject headings and Medical Subject Headings terms according to search strategies from the final search time-points until 1 June 2019.Eligibility criteria RCTs on PBM irrespective of blinding, language, date of publication and sample size were included. Abstracts and unpublished trials were excluded. Conflicts of interest were defined as sponsorship, funding or authorship by industry, professional PBM advocacy groups or blood services.Data extraction and synthesis Three independent reviewers extracted the data and assessed the risk of bias. Pooled treatment effect estimates were reported as risk ratios (RRs) or standardised mean difference with 95% CIs. Heterogeneity was quantified using the I2 statistic.Results Three hundred and eighty-nine RCTs totalling 53 635 participants were included. Thirty-two trials (8%) were considered free from important sources of bias. There was reporting bias favouring PBM interventions on transfusion across all analyses. In trials with no declared author conflicts of interest, the treatment effect on mortality was RR 1.12 (0.86 to 1.45). In trials where author conflicts of interest were declared, the treatment effect on mortality was RR 0.84 (0.69 to 1.03), with significant reporting bias favouring PBM interventions. Trials with declared conflicts linked to professional PBM advocacy groups (five studies, n=977 patients) reported statistically significant reductions in mortality RR 0.40 (0.17 to 0.92), unlike other groups.Conclusions Low certainty of the evidence that guides PBM implementation is confounded by evidence of reporting bias, and the effects of declared and undeclared conflicts of interest, favouring PBM on important trial outcomes.