Provider and center effect in multicenter randomized controlled trials of surgical specialties: An analysis on patient-level data

Objective: To estimate the importance of center and provider effect and its implication on the estimation of treatment effect in surgical randomized controlled trials. Summary Background Data: Provider and center effect may play an important role in the estimated treatment effect of multicenter surgical randomized controlled trials (RCTs). However, such effects are rarely accounted for in surgical RCTs. Methods: Analysis of patient-level data from 3 large surgical randomized controlled trials. One trial in ophthalmology comparing retinal detachment rate after retinal reattachment in 225 patients operated on by 32 providers across 10 centers; one trial in orthopaedics comparing Harris Hip Score after total hip replacement in 496 patients operated on by 22 providers across 18 centers; one trial in general surgery comparing recurrence rate of hernia repair in 200 patients operated on by 88 providers across I I centers. A provider effect and a center effect were searched for by comparing nonadjusted and adjusted models. An analysis of volume (ie, number of procedures performed during the course of the trial) effect and, when relevant, a treatment-by-volume interaction was also sought. Results: A significant provider effect was found in ophthalmology (P < 0.0001); center effect was not significant. In orthopaedics, significant provider (P = 0.0037) and volume effect (P = 0.019) were found; a correlation was found between provider effect and volume (r = 0.5, P = 0.018); moreover, a treatment-by-volume interaction was found (P = 0.033); treatment effect became significant when adjusting for volume and provider. In general surgery, center effect was more prominent than provider effect, although none were found significant. Conclusions: Provider and center effects play a significant role in the estimation of treatment effect of large randomized controlled surgical trials. Not accounting for such effects may lead to biased estimates and misleading conclusions. These effects should be accounted for in the design and analysis of such trials.