Table 4_Pain as a predictor of incident frailty in middle-aged and older adults with metabolic dysfunction-associated steatotic liver disease: a prospective cohort study using machine learning and external validation.docx

Modified Poisson regression (prespecified four-model framework including baseline frailty index) estimated relative risks (RR); g-computation estimated population attributable fractions (PAF); and mediation by depressive symptoms was assessed. Net reclassification improvement (NRI) and integrated discrimination improvement (IDI) quantified the incremental value of pain.</p>Results<p>The analysis included 3,622 CHARLS (median 56 years, 70% female) and 2,059 ELSA participants (median 64 years, 53% female).

Frailty incidence was 24.49% (CHARLS) and 21.18% (ELSA). Logistic regression performed best (internal test AUC 0.740, 95% CI 0.706–0.774; external validation AUC 0.753, 0.728–0.779). Pain was the strongest predictor (mean absolute SHAP value 0.184).

After full adjustment including baseline frailty index, pain remained associated with incident frailty (CHARLS: RR 1.219, 95% CI 1.085–1.368; ELSA: 1.310, 1.114–1.539). Under stated assumptions, the PAF was 8.01% (3.14%–12.66%) in CHARLS and 11.98% (4.76%–19.27%) in ELSA. Adding pain improved reclassification and discrimination (continuous NRI 0.397 and 0.409; IDI 0.024 and 0.023).

Depressive symptoms mediated 74.0% (CHARLS) and 46.1% (ELSA) of the total effect; associations were consistent across most subgroups and strongest at baseline FI <0.10 (CHARLS: RR 1.794, 1.289–2.496; ELSA: 1.583, 1.125–2.227).</p>Conclusion<p>Pain is a strong predictor of incident frailty in MASLD, with depressive symptoms as the principal mediator. Whether pain reduction modifies frailty risk requires interventional evidence.

Routine pain screening could facilitate frailty risk assessment; an online calculator is provided for individualized risk prediction (baozhadekele.shinyapps.io/MASLD_Frailty/).</p>