Scientific Reports · Nov 17, 2026
Higher insulin sensitivity estimated by SPISE is associated with lower odds of incident knee osteoarthritis in Chinese adults
By Cui X, Li M, Wu J et al.
To investigate the association between insulin sensitivity estimated using the Single Point Insulin Sensitivity Estimator (SPISE) and knee osteoarthritis (KOA) among Chinese middle-aged and older adults. We conducted a longitudinal analysis using data from the China Health and Retirement Longitudinal Study (CHARLS). KOA was the outcome and SPISE was the exposure. Feature selection was performed using Boruta, and selected variables were included in multivariable logistic regression models. Dose–response patterns were assessed using restricted cubic splines, and prespecified subgroup analyses were conducted to examine consistency across strata.To contextualize SPISE’s clinical utility, we compared its predictive performance against the triglyceride-to-HDL cholesterol ratio (TG/HDL) and body mass index (BMI) using area under the receiver operating characteristic curve (AUC, with between-marker comparisons performed using DeLong’s test and Holm correction for multiple comparisons), net reclassification improvement (NRI), and integrated discrimination improvement (IDI). To complement the observational findings, we additionally conducted a rat experiment (12 male Sprague-Dawley rats, blank vs. model group) to explore the metabolic phenotype associated with altered SPISE, acknowledging that this experiment does not directly validate the SPISE-KOA association. A total of 2,709 participants were included, of whom 252 had KOA (9.30%). In the fully adjusted model, each 1-unit increase in SPISE was associated with lower odds of KOA (odds ratio 0.7939, 95% confidence interval 0.6826 to 0.9203; P = 0.0025). Compared with SPISE Q1, SPISE Q4 was associated with lower odds of KOA (0.5973, 0.3944 to 0.8982; P = 0.0139); however, Q2 and Q3 did not differ significantly from Q1, and the significant trend test (P for trend = 0.0093) was driven primarily by this extreme-quartile contrast, so a graded dose-response relationship could not be confirmed. Restricted cubic splines supported an overall association (P = 0.011) with no evidence of nonlinearity (P = 0.84). Subgroup analyses showed no significant interactions by age, sex, hypertension, or smoking.In head-to-head comparisons, the AUC of SPISE for incident KOA was 0.5606 (95% CI 0.5241 to 0.5972): discrimination was significantly higher than that of TG/HDL (0.5065, 95% CI 0.4699 to 0.5431; AUC difference 0.0541, 95% CI 0.0279 to 0.0804; DeLong P < 0.001) but not superior to BMI (0.5805, 95% CI 0.5435 to 0.6175; AUC difference − 0.0199, 95% CI -0.0399 to 0.0001; P = 0.0512), with conclusions unchanged after Holm correction; the incremental gain was small and non-significant (NRI 0.04, 95% CI -0.02 to 0.11; IDI 0.008, P = 0.12), indicating that SPISE did not significantly improve discrimination or reclassification beyond conventional metabolic markers and that its standalone discriminative ability is limited.In the rat experiment, the model group showed significantly lower SPISE than the blank group (28.63 ± 1.35 vs. 31.40 ± 1.68, P = 0.010), consistent with the human data. The negative correlation between SPISE and BMI in rats (r = -0.744, P = 0.006) is mathematically expected because BMI is a denominator term in the SPISE formula and was not interpreted as an independent biological finding. Higher SPISE was associated with lower odds of KOA in a nationally representative Chinese cohort. The association appeared approximately linear and was broadly consistent across key subgroups. However, the association was primarily driven by the extreme quartile contrast (Q4 vs. Q1), and SPISE showed no statistically significant incremental predictive value over BMI and TG/HDL. The exploratory rat experiment supported metabolic phenotype consistency but did not validate the SPISE-KOA association directly, owing to the lack of histological KOA confirmation and the absence of a lean surgical control group. These findings suggest that SPISE is an insulin-free metabolic correlate of KOA; because it provided no significant incremental predictive value over BMI or TG/HDL, its practical advantage lies in its physiological specificity and freedom from insulin assays rather than in superior predictive performance, and its clinical translation requires further validation against imaging-confirmed outcomes.