Using preliminary data and prospective power analyses for mid-stream revision of projected group and subgroup sizes...

Pragmatic clinical trials are commonly used in patient-centered outcomes research to assess heterogeneity of treatment effects. Patient-Centered Outcomes Research Institute (PCORI) methodology standards for assessing heterogeneity of treatment effects are extremely rigorous, but their implementation in real-world settings can be difficult. Predicting recruitment effectiveness and subgroup characteristics is often challenging and may require mid-stream revision of projected group and subgroup sizes. Yet, little real-world data are available to demonstrate methodologically valid approaches to address situations where such revisions are necessary. These data were used for mid-stream revision of group and subgroup sizes in the Management of Diabetes in Everyday Life (MODEL) clinical trial. The planned number of randomized participants retained over the one-year study period was reduced from 800 to 581 due to recruitment difficulties among potential participants residing in rural areas. Prospective power analyses are based on the revised target of 581 participants retained and the proportions of 167 participants with various key baseline characteristics, who had been randomized in MODEL by January 2018, as reported to the Patient Center Outcomes Research Institute (PCORI) and the MODEL Data Safety and Monitoring Committee. Power calculations are based on two-sided t-tests with type-I error rates of 0.05 and the assumption that effect sizes will range from small (standardized difference = 0.36) to medium (= 0.50). The primary outcome variables are how many days in the previous week participants 1) ate healthy meals, 2) participated in at least 30 minutes of physical activity, and 3) took medications as prescribed. The POWER procedure of SAS 9.4 was used for all analyses. These data, along with the approach, can assist statisticians as they plan future pragmatic clinical trials evaluating heterogeneity of treatment effects. These data can help inform investigators, conducting patient-centered outcomes research, as they define subgroups for either confirmatory analyses for testing heterogeneity of treatment effects or for exploratory analyses where estimation of confidence bounds may be useful for generating future hypotheses. (This work was supported through a Patient-Centered Outcomes Research Institute (PCORI) Project Program Award (SC15-1503-28336), www.ClinicalTrials.gov and Identifier: NCT02957513 [1].).