UCSF

Objective

To assess how inadequate reporting of cointerventions influences estimated treatment effects in recent cardiovascular trials.

Methods

Medline/Embase were systematically searched from January 1, 2011 to July 1, 2021 for trials evaluating pharmacologic interventions on clinical cardiovascular outcomes published in 5 high-impact journals. Information on adequate vs inadequate reporting of cointerventions, blinding, risk of bias due to deviations of intended interventions (low vs high/some concerns), funding (nonindustry vs industry), design (superiority vs noninferiority), and results were assessed by 2 reviewers. The association with effect sizes was assessed using meta-regression random-effect analysis, expressed as ratios of odds ratios (ROR). RORs of >1.0 indicated that trials with the methodological factor pointing to lower quality report larger treatment estimates.

Results

In total, 164 trials were included. Of the 164 trials, 124 (74%) did not adequately report cointerventions; 89 of the 164 trials (54%) provided no information regarding cointerventions, and 70 of the 164 (43%) were at risk of bias due to inadequate blinding. Moreover, 86 of the 164 (53%) were at risk of bias due to deviation of intended interventions. Of the 164 trials, 144 (88%) were funded by the industries. Trials with inadequate reporting of cointerventions had larger treatment estimates for the primary end point (ROR, 1.08; 95% CI, 1.01-1.15; I²=0%). No significant association with results for blinding (ROR, 0.97; 95% CI, 0.91-1.03; I²=66%), deviation of intended interventions (ROR, 0.98; 95% CI, 0.92-1.04; I²=0%), or funding (ROR, 1.01; 95% CI, 0.93-1.09; I²=0%) was found.

Conclusion

We conclude that trials with inadequate reporting of cointerventions showed larger treatment effect estimates, potentially indicating overestimation of therapeutic benefit.

Trial registration

Prospero Identifier: CRD42017072522.