Making the Most of Your Study Budget: Optimizing the Number of Sites in a Multicenter Clinical Trial to Reduce Costs and Time to Study Completion

Multicenter clinical trials are crucial to advance the science behind clinical care but are also complex and costly. Trialists are called upon to propose gold-standard studies which address the most critical questions while government funding has leveled if not reduced. Currently there are no evidence-based approaches for the financial management of multicenter trials in an orthopaedic trauma population. One key cost driver in multicenter trials is the number of sites. This project proposes a model for determining the optimal number of sites in a prospective multicenter trial. Our hypothesis is that the optimal number of sites can be determined based on study characteristics, known costs, and predictable site enrollment contributions. This is a retrospective marginal analysis of 12 completed studies conducted as part of an orthopaedic research consortium. The studies represented a wide range of research priorities and varied by injury volume with the highest enrolling 1,054 patients and the lowest enrolling just over 30. They also varied by design complexity with the least complex being an observational study with light data collection and the most complex a placebo controlled, double-blinded RCT. Using enrollment and financial data, the primary analysis was to determine the optimal number of sites for each study by modeling the marginal cost of each site added to the study. Secondary analyses were performed to determine the model’s sensitivity to variation in overhead costs. For every study the optimal number of sites was lower than the actual number of sites that participated. Excess sites ranged from 2 to 39 sites and excess costs ranged from $17K to $330K across the 12 studies, with a median excess cost of $96K. These costs represented, on average, 7% [95% CI: 6%, 9%] of the study budget. Secondary analyses demonstrated that as overhead costs increase, so do the optimal number of sites as it becomes more advantageous to complete the study as quickly as possible. The results indicate that previous trials were not optimized in terms of the number of sites. Hopefully this model can be used by future clinical researchers to answer critical clinical questions in a more cost-effective manner