External validation and cross-center transportability of phenotypic age acceleration and laboratory frailty for postoperative mortality prediction: a dual-center cohort study

This is the authors' abstract. We don't add key points for this paper.

Background Phenotypic Age Acceleration (PAA) and the Laboratory Frailty Index (FI-Lab) can be derived from routine biomarkers to describe biologic aging and accumulated laboratory deficits. Their prognostic behavior across different clinical settings, however, remains uncertain. We examined whether associations with postoperative mortality were consistent between two hospitals and whether either measure added predictive information beyond standard clinical variables. Methods We analyzed 1,223 adults (671 from Shandong and 552 from Jiangsu) with linked health-examination/questionnaire information, perioperative characteristics, and follow-up through 540 days. Phenotypic Age was calculated from chronological age and nine biomarkers. Primary PAA used the Shandong age–Phenotypic Age relation as a fixed reference, and FI-Lab was the proportion of 17 prespecified laboratory deficits. Center-specific and center-stratified Cox models assessed 540-day mortality. Ridge logistic models were trained in Shandong and externally validated in Jiangsu for 365-day mortality; the reverse direction was treated as a secondary reciprocal transportability sensitivity analysis. As the fixed Shandong reference induced age dependence in Jiangsu, sensitivity analyses recalculated PAA as center-specific age–Phenotypic Age residuals with center-specific standardization. Secondary analyses covered the 12-month composite of death or incident diabetes, chronic kidney disease, or cardiovascular disease. Longitudinal analyses of within-person PAA and FI-Lab changes were exploratory; exact person-specific reassessment intervals were unavailable, and follow-up chronological age for PAA was approximated as baseline age + 1 year. All laboratory measurements used to calculate PAA and FI-Lab were obtained from the preoperative panel within 24 h before the index operation, and completeness of this panel was required for inclusion. Results During 1,694.66 person-years, 74 participants died, including 57 within 365 days. In the fully adjusted center-stratified model, each 1-SD increase in PAA was associated with mortality [hazard ratio (HR), 1.34; 95% confidence interval (CI), 1.03–1.75], whereas the FI-Lab estimate was borderline (HR, 1.28; 95% CI, 1.00–1.65). Cross-center interaction estimates were imprecise (PAA interaction ratio, 1.06; 95% CI, 0.68–1.65; FI-Lab interaction ratio, 1.48; 95% CI, 0.94–2.32). Fixed-reference PAA was strongly inversely correlated with chronological age in Jiangsu ( r = − 0.74); when PAA was recalculated using center-specific age–Phenotypic Age residuals and standardized within center, the fully adjusted pooled association remained (HR, 1.39; 95% CI, 1.09–1.77). In the primary Shandong-to-Jiangsu external validation, the study clinical model had an area under the receiver operating characteristic curve (AUC) of 0.60 (95% CI, 0.45–0.73); adding PAA, FI-Lab, or both yielded AUCs of 0.60, 0.63, and 0.63, respectively, without a consistent decision-curve net-benefit gain. FI-Lab, but not PAA, was associated with the 12-month death-or-incident-disease composite (RR, 1.36; 95% CI, 1.08–1.71). Among 820 participants with repeated measurements, exploratory mean changes were + 2.90 years for PAA and + 0.02 for FI-Lab. Calibration estimates in the primary external-validation cohort were imprecise because only 23 deaths occurred by 365 days; nonsignificant calibration tests were not interpreted as evidence of good calibration. Conclusions Laboratory-derived aging and frailty measures were related to postoperative mortality across the two cohorts, and the PAA association was robust to center-specific residualization. However, absolute fixed-reference PAA values were not directly transportable across centers because the age–Phenotypic Age relations differed. Neither marker consistently improved prediction beyond the study clinical model, whose external discrimination was modest; calibration remained imprecise. Larger multicenter studies with rigorous cross-center laboratory harmonization are needed before routine perioperative use.