Discontinuation and nonpublication of interventional clinical trials conducted in ophthalmology

Discontinuation of interventional clinical trials and nonpublication of completed trials represent a waste of already scarce resources. We sought to identify the prevalence of discontinuation and nonpublication of interventional clinical trials conducted in ophthalmology patients.

A retrospective, cross-sectional study of ophthalmology-based interventional clinical trials in ClinicalTrials.gov dating back to 1972 was conducted. χ² tests were used to determine any potential associations between trial characteristics and trial completion. The Mann-Whitney U test was used to compare median time to publication between academic and industry-sponsored trials.

Of 2926 included trials, 413 (14%) were discontinued early with only 19 of these studies being published. A total of 2027 (81%) of the 2513 completed trials were not published. A total of 277 419 participants were enrolled in unpublished, completed trials, whereas an estimated 20 843 participants were enrolled in trials that were never completed and never to led to publication. The odds of study discontinuation among industry-sponsored trials were 27% greater than government/academia-sponsored (National Institutes of Health, National Eye Institute, National Institutes of Health Clinical Center, U.S. and foreign university-based teaching hospitals) trials (odds ratio [OR] = 1.27, 95% confidence interval [CI] 1.03-1.57, p = 0.03). Furthermore, the odds of nonpublication among industry-sponsored trials were 55% more than those funded by government/academia (OR = 1.55, 95% CI 1.27-1.89, p < 0.0001).

The nonpublication of many completed trials and of preliminary results of trials that are discontinued early dilutes the quality and decreases the comprehensive nature of the medical literature. This occurs in both industry and academia. Greater transparency through the publication of the entire spectrum of clinical trials' results as well as those that are terminated early could minimize publication bias and thus lead to a more robust medical literature.