Road to Hierarchical Diabetes Management at Primary Care (ROADMAP) Study in China: Statistical Analysis Plan for a Cluster Randomized Controlled Trial (Preprint)

Background: As the management of type 2 diabetes remains suboptimal in primary care, the Road to Hierarchical Diabetes Management at Primary Care (ROADMAP) study was designed and conducted in diverse primary care settings to test the effectiveness of a three-tiered diabetes management model of care in China. Objective: This paper aims to predetermine the detailed analytical methods for the ROADMAP study before the database lock to reduce potential bias and facilitate transparent analyses. Methods: The ROADMAP study adopts a community-based, cluster randomized controlled trial design that compares the effectiveness of a tiered diabetes management model on diabetes control with usual care among patients with diabetes over a 1-year study period. The primary outcome is the control rate of glycated hemoglobin (HbA(1c)) <7% at 1 year. Secondary outcomes include the control rates of ABC (HbA(1c), blood pressure, and low-density lipoprotein cholesterol [LDL-C], individual and combined) and fasting blood glucose, and the change in each outcome. The primary analysis will be the log-binomial regression with generalized estimating equation (GEE), which accounts for the clustering within communities, for binary outcomes and linear regression with GEE for continuous outcomes. For both, the baseline value of the analyzed outcome will be the covariate. The other covariate further adjusted models and the repetitive models after multiple imputation (when more than 10% of observations in HbA(1c) after 1 year are missing) will be used for sensitivity analysis. Five prespecified subgroup analyses have also been planned to explore the heterogeneity of the intervention effects by adding the subgroup variable and its interaction with the intervention to the primary model. Results: This plan has been finalized, approved, and signed off by the principle investigator, co-principle investigator, and lead statisticians as of November 22, 2019, and made public on the institutional website without any knowledge of intervention allocation. Templates for the main figure and tables are presented. Conclusions: This statistical analysis protocol was developed for the main results of the ROADMAP study by authors blinded to group allocation and with no access to study data, which will guarantee the transparency and reduce potential bias during statistical analysis.