GLP-1 CVOT Evidence Synthesis Engine

Abstract

Extracting meta-analysis data from CG.gov represents a shift from static data visualization to active clinical decision support. Within our advanced  engine, data is processed through rigorous statistical guards. We use SUCRA scores to identify "Category Leaders,". The τ² thresholds trigger automated heterogeneity and by integrating Bayesian posterior probabilities with rule-based logic we classify molecules like Tirzepatide by their actual probability of superiority. This framework emulates target trials, by adjusting for population shifts identified in PCA geometry. This rules-based synthesis prevents the misinterpretation of noisy registry data with an "Executive Intelligence" text output that mirrors high-level health technology assessments. We take a pool of divergent trials converting them into a coherent evidence-based hierarchy for modern therapeutic guidelines. 3-point MACE pool (added for transparency): A homogeneous pool of the GLP-1 receptor agonist cardiovascular outcome trials reporting 3-point MACE—SUSTAIN-6, HARMONY, and SELECT—gives a pooled odds ratio of 0.78 with no heterogeneity (I²=0%); SOUL is consistent with and fits this pool (OR 0.80, I²=0%). ELIXA is reported separately and is not included in this pool because its only available endpoint is a 4-point MACE, which is not poolable with the 3-point MACE trials and is the sole driver of heterogeneity if forced into it.