841-P: Phenomapping-Derived Tool to Personalize Chiglitazar Treatment in Type 2 Diabetes—A Post Hoc Analysis of Randomised Clinical Trials

Abstract

Chiglitazar is a peroxisome proliferator-activated receptor (PPAR) pan-agonist showing comparable glucose-lowering efficacy with sitagliptin in patients with type 2 diabetes. To identify the subgroup with higher benefits with chiglitazar thus personalizes drug therapy, a post-hoc analysis using a machine learning algorithm was applied in ChiglitAzar Monotherapy with Placebo (CAMP) and ChiglitAzar Monotherapy with Sitagliptin (CAMS) study. We constructed a phenomap based on 13 baseline variables from 1069 patients with diabetes assigned to chiglitazar (32mg or 48mg, n=822) and sitagliptin (n=247). Personalized HbA1c decline at week 24 was estimated using least squares regression weighted for phenotypic distances between each participant. High-benefit group (HBG) was defined as patients with estimated HbA1c decline higher in chiglitazar than sitagliptin, and low-benefit group (LBG) was defined as the opposite.677 (63.3%) patients were allocated to the HBG and 392 (36.7%) were allocated to the LBG. Chiglitazar showed higher HbA1c decline than sitagliptin In the HBG and lower HbA1c decline in the LBG (relative decline 0.67%, 95%CI [0.50, 0.83] in HBG and -0.93% [-1.18, -0.68] in LBG, p for interaction<0.001). There was no significant inter-group heterogeneity in changes in HOMA-IR, HDL, triglyceride and body weight. Dosage analysis suggested the majority (n=635, 77.3%) would gain more glucose-lowering benefits from chiglitazar 48mg than 32mg. To facilitate efficient identification of HBG in the clinic, we developed ML-PANPPAR, a machine learning model based on extreme gradient boosting algorithm (XGBoost). The model showed robust performance (AUC=0.933 in the internal testing subset) to recognise HBG with five simple variables including sex, BMI, HbA1c, HDL and fasting insulin.Our phenomapping-derived tool provides support to select specific patients to receive chiglitazar for better glucose-lowering efficacy. Disclosure Q.Huang: None. X.Zou: None. W.Jia: None. L.Ji: Other Relationship; Eli Lilly and Company, Novo Nordisk, Merck & Co., Inc., Bayer Inc., Sanofi-Aventis U.S., Roche Pharmaceuticals, MSD Life Science Foundation, AstraZeneca, Boehringer Ingelheim Inc., Abbott, Metronics. Funding Beijing Nova Cross Program (Z211100002121169); Beijing Nova Program of Science and Technology (Z191100001119026 to X.Z.)