Development and temporal validation of a prognostic prediction model for incident painful diabetic peripheral neuropathy in patients with diabetes aged over 45 years in China: a multicentre cohort study

Résumé fourni par la source

Painful diabetic peripheral neuropathy (PDPN) substantially impairs quality of life in people with diabetes mellitus (DM). This study aimed to develop a preliminary prognostic model for incident PDPN among Chinese DM patients free of baseline PDPN. This multicentre, prospective observational cohort consecutively recruited DM patients aged ≥ 45 years from eight hospitals in China. Patients recruited between January 1, 2019 and March 31, 2023 constituted the derivation cohort, whereas those recruited between April 1, 2023 and December 1, 2024 constituted the temporal validation cohort. A Firth-penalized Cox proportional hazards regression was applied to develop the prediction model, with variable selection guided by the Akaike information criterion for Firth’s penalized partial likelihood. Model performance was evaluated in both the derivation and validation cohorts with respect to discrimination (Harrell’s C-index and time-dependent AUC), calibration (calibration plots, intercepts, and slopes), overall predictive accuracy (Brier scores), and clinical utility (decision curve analysis). To assess and correct for overfitting, the entire model-building procedure was repeated in 1000 bootstrap samples from the derivation cohort, and optimism-corrected performance estimates were obtained. Four independent predictors were identified: female sex (HR=2.12, 95% CI 1.05–4.40), DM duration ≥1 year (HR=2.40, 95% CI 1.15–5.33), TCSS > 5 (HR=4.14, 95% CI 1.30–10.45), and elevated LDL-C (HR=4.50, 95% CI 1.42–11.24). Performance was consistently superior for 2-year versus 1-year risk prediction. In the derivation cohort, the bootstrap-corrected 2-year AUC was 0.718. In the temporal validation cohort, the model maintained discrimination (AUC=0.701 for 2-year risk) with well-calibrated predictions (2-year calibration slope=0.896, intercept=− 0.120). Decision curve analysis confirmed positive net clinical benefit for 2-year risk prediction across threshold probabilities of 0.03–0.30. This study developed and temporally validated a preliminary prognostic model for PDPN risk using readily accessible clinical parameters. The model may help identify Chinese DM patients aged over 45 years at increased risk of PDPN, although independent external validation remains required before clinical implementation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.