Discontinuation and non-publication of clinical trials in cardiovascular medicine.

Background: Appropriate dissemination of clinical data is crucial for minimising bias. Despite this, high rates of study discontinuation and non-publication have been reported among clinical trials. Cardiovascular medicine receives a substantial proportion of academic funding; however, predictors of non-publication among cardiovascular trials are not well-established.
Methods: The National Clinical Trials database was searched for cardiovascular trials completed between January 2010 and January 2014. Associated publications were identified in Medline or Embase. Relevant variables were extracted and subject to chi-squared and logistic regression to identify predictors of discontinuation and non-publication.
Results: After reviewing 2035 trials, 431 trials were included, of which 82.1% (n=354; 119,233 participants) were completed. Among completed trials, 70.3% (n=249; 99,095 participants) were published. Industry funding was associated with increased likelihood of non-publication (odds ratio [OR] 2.84; 95% confidence interval [CI] 1.47-5.51; P=0.002), while non-randomised studies were more likely to remain unpublished than randomised counterparts. Industry-funded studies were over three times more likely to be discontinued than those sponsored by academic institutions (OR 3.89; CI 1.54-9.83; P=0.004). Trials studying heart failure and atrial fibrillation were more likely to be discontinued compared to trials studying coronary artery disease (OR 2.83; CI 1.23-6.51; and OR 3.10; CI 1.21-7.96, respectively). Of the total 135,714 participants, 25,565 were recruited into unpublished studies.
Conclusions: Discontinuation and non-publication of cardiovascular trials are common, resulting in data from thousands of participants remaining unpublished. Funding source and randomisation are strong predictors of non-publication, while sponsor type, phase and blinding status are key predictors of discontinuation.
(Copyright © 2017 Elsevier B.V. All rights reserved.)