Are multiple primary outcomes analysed appropriately in randomised controlled trials? A review

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Medical Research Integrity

Background:

• Randomized controlled trials (RCTs) frequently employ multiple primary outcomes to assess various treatment effects.
• The statistical implications of analyzing multiple outcomes without appropriate adjustments for multiplicity are often overlooked.
• Neurology and psychiatry are disease areas where the use of multiple primary outcomes in RCTs is prevalent.

Purpose of the Study:

• To review the current practices in analyzing randomized controlled trials (RCTs) with multiple primary outcomes.
• To identify the methods used to safeguard inferences when multiple primary outcomes are present.
• To raise awareness regarding potential issues arising from the analysis of multiple outcomes in clinical trials.

Main Methods:

• A systematic review of RCTs published between July 2011 and June 2014 in high-impact medical journals.
• Focus on neurology and psychiatry RCTs due to their frequent use of multiple outcomes.
• Data extraction included the number of primary outcomes, multiplicity adjustment methods, and sample size calculation details.

Main Results:

• Out of 209 reviewed RCTs, 60 (29%) utilized multiple primary outcomes.
• A significant majority (75%) of these trials did not adjust for multiplicity in their statistical analyses.
• Bonferroni's correction was the most common method (used in 25% of trials that did account for multiplicity).

Conclusions:

• The use of multiple primary outcomes in RCTs is common in clinical research.
• Inappropriate statistical analysis, specifically the lack of multiplicity adjustment, is widespread.
• Clear articulation of primary outcomes and justification of analytical methods are crucial for authors conducting and reporting RCTs.