A continuous composite endpoint methodology improves statistical efficiency: proof of concept in celiac disease

Clinical trials in celiac disease (CeD) use histologic, immunologic, and serologic biomarkers that capture complementary aspects of disease activity but are generally analyzed independently. Integrating these measurements may improve clinical trial endpoint performance. We developed a continuous composite endpoint (CCE) methodology that standardizes complementary biomarkers onto a common scale to provide a single quantitative endpoint. CCE performance was evaluated in two independent CeD datasets: a 14-day gluten challenge study assessing histologic deterioration and immune activation, and a 52-week study of newly diagnosed patients assessing mucosal recovery, serologic normalization, and gluten-specific immune responses. Performance was compared with individual measures using statistical significance, effect size, variance, responder classification, and biological correlations. Leave-one-out and missing-component sensitivity analyses evaluated robustness, while simulations assessed type I error, statistical power, sample-size requirements, component correlation, and heterogeneity of biomarker response. In the gluten challenge study, the histologic composite (VCIEL) increased Cohen’s d from 1.92 for villous height-to-crypt depth ratio (Vh) and 1.85 for intraepithelial lymphocytes (IEL) to 2.30. In the recovery study, combining serology and tetramer measurements increased Cohen’s d from 1.81 and 1.39, respectively, to 2.47. Simulations maintained type I error near 0.05 and, for four complementary biomarkers with a common effect size of d = 0.30, reduced estimated sample size for 80% power from 90 for an individual biomarker to 45 for the CCE at an inter-component correlation of 0.3. Benefits diminished with increasing correlation and were not observed when the response was confined primarily to a single biomarker. CCE methodology can improve clinical trial endpoint performance by integrating complementary continuous biomarkers while preserving their quantitative information. These findings provide proof of concept in CeD and support prospective evaluation in CeD and other diseases characterized by complementary continuous biomarkers.