Publication Rates and Characteristics of Registered Randomized Clinical Trials in Digital Health: A Cross Sectional Analysis.

BACKGROUND: Clinical trials are key to advancing evidence-based medical research. The medical research literature has identified the impact and risks of publication bias in clinical trials. Selective publication for positive outcomes or non-publication of negative results could lead to misdirection of subsequent research, justify further research, and result in literature reviews leaning towards positive outcomes. Digital health randomized clinical trials face specific challenges, including a high attrition rate, usability issues, and insufficient prior formative research. These challenges may become contributing factors to non-publication of trials' results. To our knowledge, no study exists that has analyzed and reported the characteristics of non-publication rates within the domain of digital health trials.

OBJECTIVE: The primary research objective was to examine the prevalence and characteristics of non-published digital health randomized clinical trials, including eHealth, mHealth, and telehealth clinical trials, registered in ClinicalTrials.gov. Our secondary research objective was to determine whether industry funding is a contributing factor to trials non-publication.

METHODS: To identify digital health trials, a list of 47 search terms and phrases was developed through an iterative process and applied to the "Title," "Interventions," and "Outcome Measures" fields of registered clinical trials with completion dates between April 1st, 2010 and April 1st, 2013. The search was based on the full dataset exported from the ClinlicalTrials.gov database, with 265,657 registered clinical trials entries downloaded on February 10th, 2018, to allow for up to nearly five years for the publication of the study after trial completion. To identify publications related to the results of the trials, we extracted the registered clinical trials content from the ClinicalTrials.gov website in XML format and identified relevant publications through a comprehensive approach that included an automated as well as a manual publication identification process.

RESULTS: In total, 6717 articles matched the a priori search terms and phrases, of which 803 trials matched our latest completion date criteria. 556 randomized trials were included in this study after screening. We found that 150 (27%) of all included trials remain unpublished five years after the completion date of each trial. In bivariate analyses, statistically significant differences in trial characteristics between published and unpublished trials were found for the intervention target condition (cancer having the largest non-publication rate at 45%, while addiction/smoking cessation trials having the lowest non-publication rate at 16%), country (U.S. at 33% and non-U.S. at 18%), trial size (small trials at 52%, larger trials at 30%), clinical trial phases, recruitment and prospective trials registration. In multivariate analyses, differences in trial characteristics between published and unpublished trials remained statistically significant for the intervention target condition, country, trial size, trial phases and recruitment, with the odds of publication for non-U.S.-based trials being significant and 3.3 (CI:1.845-5.964) times more likely to be published compared to the reference group of the U.S.-based trials. We observed a trend of 1.5 times higher non-publication rate for industry-funded trials, However, the trend was not statistically significant.

CONCLUSIONS: In the domain of digital health, non-publication of registered clinical trials results is prevalent at 27%, which is lower than non-publication rates in other fields. There are substantial differences in publication rates between trials funded by industry and non-industry sponsors. Further research is required to define further determinants and reasons for non-publication, and more importantly to articulate the impact and risk of publication bias in the field of digital health clinical trials.