T. J. Mouw1, S. W. Hong1, S. Sarwar1, A. E. Fondaw2, A. Walling3, M. Al-Kasspooles1, P. J. DiPasco1 1Unversity Of Kansas Medical Center,General Surgery,Kansas City, KS, USA 2University Of Kansas School of Medicine – Kansas City, Kansas City, KS, USA 3Unversity Of Kansas School of Medicine – Wichita, Family and Community Medicine, Wichita, KS, USA
Introduction:
Trial early discontinuation is a complex issue with both financial and ethical implications. It has been previously reported that over 20% of surgical trials will be discontinued prematurely and many of those which reach completion will not publish. Previous studies have been limited in scope owing to the need for manual review of selected trials. To date there has been no broad analysis comparing surgical and non-surgical registered clinical trials.
Methods:
The US National Institutes of Health registry at clinicaltrials.gov was accessed 7/7/17 and all US trials from 2005-2017 were downloaded by status (Completed, ongoing, and discontinued). An algorithm was developed to automatically assign trials as “surgical” or “non-surgical” based on trial type and inclusion of surgical keywords generated from a list of 10,000 trial titles and descriptions. The algorithm was validated by testing a subset of trials against a team of blinded residents and medical students. A primary analysis was conducted of all US trials based on the assigned designation of surgical and non-surgical per the trial status. Significance was established via two-tailed z-test. The reasons for discontinuation between surgical and non-surgical trials were examined and tabulated. A univariate multiple logistic regression using SPSS version 20.0 was performed to assess the impacts of trial design, characteristics, and funding sources on trial discontinuation and completion.
Results:
The database search yielded 82,719 non-surgical and 5779 surgical trials after automatic assignment. The algorithm for assignments had an overall accuracy of 87.99% (95%CI 86.85-89.13%) and was associated with a +LR of 6.09 and -LR of 0.093.
Significant differences were observed in trial status (Non-surg vs surg: Completed: 55.51% vs 39.49%, Ongoing: 33.42% vs 44.54%, and Discontinued: 11.07% vs 15.97%, p <0.001 each). Industry was more likely to fund non-surgical trials (44.00% vs 32.50%, p <0.001). Surgical trials were more likely to discontinue due to poor recruitment (44.65% vs 34.74% p<0.001). Industry funding was associated with increased discontinuation (OR 1.63 p<0.001). This remained true for the surgical subset of trials funded by industry (OR 1.25 p=0.041). Reaching enrollment and/or phase 1, reporting results, and NIH funding were all protective against discontinuation while randomization had no effect.
Conclusion:
Surgical trials are less likely to reach completion compared to non-surgical trials. This study establishes industry funding as a contributory factor to trial discontinuation. However, it is not clear if this is due to forces in play after trial initiation or if it is due to relative exclusivity of selection criteria among the different trial sponsors. Poor study recruitment is an major cause for early trial discontinuation and surgical trials are more susceptible to this than non-surgical trials.