Discovery of a predictive model for anlotinib-induced hand-foot syndrome based on metabolomics

The purpose of this study was to investigate the exposure-toxicity relationship between anlotinib (ANL) and the hand-foot syndrome (HFS) and to establish an early warning model of ANL-induced HFS based on exposure level and metabolomics. The cancer patients treated with ANL were enrolled in a prospective cohort study. Plasma samples were collected at baseline and steady-state trough concentration points. Exposure concentrations of ANL were examined and the exposure-toxicity relationship was analyzed. The untargeted metabolomics was carried out and the early warning model was constructed using binary logistic regression. Totally 65 lung cancer patients were enrolled and the steady-state trough concentration of ANL was 13.8 ± 7.8 (mean±SD). The occurrence of HFS wasn't associated with ANL exposure concentration. Untargeted metabolomics revealed 6 differential metabolites in baseline samples from patients with or without HFS, and 28 differential metabolites were identified in baseline samples from patients with grade ≥ 2 or grade < 2 HFS, allowing to distinguish between patients with or without grade ≥ 2 HFS. A logistic regression model Y= 0.107*Tromethamine-1.447*(6,9-Dioxo-Decanoic acid) + 5.697*(9-O-Acetylneuraminic acid) + 161.586 (AUC=0.958, 95 % CI: 0.878-1.000, P < 0.0001) for early warning of grade ≥ 2 HFS was constructed. HFS didn't correlated to ANL steady-state trough concentration. Disturbed baseline metabolic profiles showed close association with ANL-induced HFS, resulting an early warning model which may improve the clinical application of ANL after verification with large sample size.