Outcome reporting bias in randomized-controlled trials investigating antipsychotic drugs

M Lancee,J J Luykx,C M C Lemmens,C H Vinkers,R S Kahn

doi:10.1038/tp.2017.203

M Lancee, J J Luykx + Show 3 more

Open Access

https://doi.org/10.1038/tp.2017.203

Copy DOI

Abstract

Recent literature hints that outcomes of clinical trials in medicine are selectively reported. If applicable to psychotic disorders, such bias would jeopardize the reliability of randomized clinical trials (RCTs) investigating antipsychotics and thus their extrapolation to clinical practice. We therefore comprehensively examined outcome reporting bias in RCTs of antipsychotic drugs by a systematic review of prespecified outcomes on ClinicalTrials.gov records of RCTs investigating antipsychotic drugs in schizophrenia and schizoaffective disorder between 1 January 2006 and 31 December 2013. These outcomes were compared with outcomes published in scientific journals. Our primary outcome measure was concordance between prespecified and published outcomes; secondary outcome measures included outcome modifications on ClinicalTrials.gov after trial inception and the effects of funding source and directionality of results on record adherence. Of the 48 RCTs, 85% did not fully adhere to the prespecified outcomes. Discrepancies between prespecified and published outcomes were found in 23% of RCTs for primary outcomes, whereas 81% of RCTs had at least one secondary outcome non-reported, newly introduced, or changed to a primary outcome in the respective publication. In total, 14% of primary and 44% of secondary prespecified outcomes were modified after trial initiation. Neither funding source (P=0.60) nor directionality of the RCT results (P=0.10) impacted ClinicalTrials.gov record adherence. Finally, the number of published safety endpoints (N=335) exceeded the number of prespecified safety outcomes by 5.5 fold. We conclude that RCTs investigating antipsychotic drugs suffer from substantial outcome reporting bias and offer suggestions to both monitor and limit such bias in the future.

Highlights

Even though clinical trials constitute the backbone of evidencebased medicine, a substantial proportion of their outcomes have proven overstated, flawed or difficult to reproduce.[1,2,3] To ensure the transparency and reliability of clinical trial research, the National Institutes of Health (NIH) and the Food and DrugAdministration (FDA) created a national clinical trials registry and results database (ClinicalTrials.gov) in 2000.4 This registry constitutes the largest online clinical trials database.[5]
Submission of study results to ClinicalTrials.gov is mandatory for US-based clinical trials following the 2007 Food and DrugAdministration (FDA) Amendments act,[6] which aims to prevent outcome reporting bias, defined as selective publication of clinical trial outcomes.[7,8]
This study signals discrepancies between outcomes prespecified on ClinicalTrials.gov and published outcomes in 85% of randomized clinical trials (RCTs) investigating antipsychotic drugs, with 23% of RCTs showing discrepancies in primary outcomes and 81% in secondary outcomes

Summary

Introduction

Even though clinical trials constitute the backbone of evidencebased medicine, a substantial proportion of their outcomes have proven overstated, flawed or difficult to reproduce.[1,2,3] To ensure the transparency and reliability of clinical trial research, the National Institutes of Health (NIH) and the Food and Drug. Administration (FDA) created a national clinical trials registry and results database (ClinicalTrials.gov) in 2000.4 This registry constitutes the largest online clinical trials database.[5] Submission of study results to ClinicalTrials.gov is mandatory for US-based clinical trials following the 2007 FDA Amendments act,[6] which aims to prevent outcome reporting bias, defined as selective publication of clinical trial outcomes.[7,8]. Meta-analyses risk producing biased or inflated results and thereby introduce bias in the translation of published evidence into clinical practice.[8,17]

Methods

Results

Conclusion