Triglyceride-glucose related indices, cardiometabolic multimorbidity, and risk of incident osteoarthritis: A prospective cohort study in UK Biobank Authors: Kai-Bo Sun, Yuan-Gang Wu, Yu-Huan Xiang, Meng-Ying Li, Yang Ding, Ling-Sheng Zhang, Yi Zeng, Hui-Qi Xie, Bin Shen Published: 2026 ABSTRACT BackgroundMetabolic dysfunction may contribute to osteoarthritis (OA), but routine markers for identifying future OA risk remain limited.ObjectiveTo examine associations of triglyceride-glucose (TyG) and adiposity-integrated TyG indices with hospital-recorded incident OA, and whether these associations differed by cardiometabolic disease (CMD) burden.DesignProspective cohort study using UK Biobank data.MethodsWe included 255,232 participants without baseline OA and with complete exposure, covariate, and selected metabolomic data. Baseline TyG, TyG-body mass index (BMI), TyG-waist circumference (WC), and TyG-waist-to-height ratio (WHtR) were analyzed in quartiles. CMD burden was categorized as no CMD, one CMD, or cardiometabolic multimorbidity (CMM). Hospital-recorded incident OA was identified from linked hospital inpatient records using ICD-10 codes M15-M19. Cox proportional hazards models estimated hazard ratios (HRs) and 95% confidence intervals (CIs), adjusting for demographic, socioeconomic, and lifestyle factors. Restricted cubic spline, stratified, and joint exposure analyses were performed.ResultsDuring follow-up, 43,554 participants developed OA. Compared with the lowest quartile, the highest quartile was associated with a higher risk of hospital-recorded incident OA for TyG-BMI (HR 2.18, 95% CI 2.12–2.24), TyG-WC (HR 2.14, 95% CI 2.08–2.21), and TyG-WHtR (HR 2.00, 95% CI 1.94–2.06), whereas TyG alone showed only a weak association (HR 1.03, 95% CI 1.00–1.06). CMD burden was independently associated with OA risk (one CMD: HR 1.37, 95% CI 1.29-1.46; CMM: HR 1.44, 95% CI 1.31-1.58). Across CMD strata, the strongest associations were observed among participants with CMM, with HRs of 2.05 (95% CI 1.61–2.61), 1.99 (95% CI 1.61–2.46), and 1.97 (95% CI 1.56–2.49) for TyG-BMI, TyG-WC, and TyG-WHtR, respectively. No statistically significant interactions were detected.ConclusionAdiposity-integrated TyG indices and greater CMD burden were associated with hospital-recorded incident OA, whereas TyG alone showed only a weak association. Joint-category analyses showed higher observed risk estimates among participants with both characteristics, without evidence of multiplicative interaction. Introduction Osteoarthritis (OA) is a leading cause of pain and disability worldwide, and growing evidence suggests that systemic metabolic and inflammatory factors contribute to its development and progression.1–3 Obesity, type 2 diabetes, and cardiovascular disease are frequently associated with OA, indicating that metabolic dysfunction may influence OA risk beyond local joint loading alone.4–6 Insulin resistance is a shared feature of these metabolic disorders and may link systemic metabolic dysfunction to joint degeneration through low-grade inflammation, oxidative stress, mitochondrial dysfunction, and altered chondrocyte metabolism.7,8 Because direct measurement of insulin resistance is impractical in large population-based studies and routine clinical settings, simple surrogate markers based on routinely measured metabolic traits are needed. 9 The triglyceride-glucose (TyG) index, calculated from triglyceride and glucose levels, has emerged as a reliable and accessible surrogate marker of insulin resistance.10,11 Although TyG has been extensively studied in cardiometabolic and metabolic diseases, its relevance to OA remains less well established.12–15 However, OA risk may not be fully captured by insulin resistance alone, because adiposity contributes to OA through both mechanical loading and adipose tissue-derived inflammatory pathways.16,17 Composite TyG-related indices that combine TyG with body mass index (BMI), waist circumference (WC), or waist-to-height ratio (WHtR) may therefore better reflect the combined burden of insulin resistance and adiposity. In addition to subclinical metabolic phenotypes captured by TyG-related indices, established cardiometabolic diseases (CMDs) may represent clinically recognized manifestations of systemic metabolic dysfunction.18,19 In this study, cardiometabolic multimorbidity (CMM) was defined as the coexistence of two or more CMDs, including hypertension, type 2 diabetes, ischemic heart disease, stroke, and dyslipidemia. Greater CMD burden may contribute to OA development through inflammatory and metabolic pathways that overlap with key processes in OA pathogenesis, including cartilage catabolism, synovial inflammation, and subchondral bone remodeling.20–22 Accumulating evidence suggests that CMD burden may contribute to musculoskeletal degeneration, including accelerated cartilage loss and increased OA risk.23,24 Nevertheless, most prior studies have examined individual cardiometabolic conditions, leaving the role of cumulative CMD burden in OA risk unclear. For population-level OA risk stratification and prevention, it is important to determine whether routinely available metabolic indices and established CMD burden provide complementary information. It remains unclear whether adiposity-integrated TyG indices are more informative than TyG alone, whether cumulative CMD burden independently contributes to hospital-recorded incident OA, and whether their combination identifies individuals at particularly high risk. Using prospective data from the UK Biobank, we examined the associations of TyG and adiposity-integrated TyG indices with hospital-recorded incident OA, evaluated the association between CMD burden and OA risk, and assessed their joint and stratified associations with OA development. Methods Data source and study population This study was based on the UK Biobank, a large prospective population-based cohort comprising adults recruited across the United Kingdom. 25 This study was conducted under the generic ethical approval obtained by the UK Biobank from the North West–Haydock Research Ethics Committee (REC reference: 11/NW/0382) on June 17, 2011. All participants provided written informed consent prior to enrolment in the study. This study was conducted under UK Biobank application no. 548384. This research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki. The study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for cohort studies. 26 Participants were recruited between March 2006 and October 2010 and were followed from baseline until the first diagnosis of osteoarthritis, death, loss to follow-up, or October 31, 2022, whichever occurred first. Participants eligible for the present analysis were required to have baseline assessment data, linked hospital inpatient outcome records, and no recorded diagnosis of OA at or before baseline. Participants were excluded if they had prevalent OA or missing data required to calculate the TyG-related indices, including serum triglycerides, serum glucose, BMI, waist circumference, and height; missing data for any prespecified covariates; or missing measurements for the selected metabolites. After these exclusions, 255,232 participants were included in the final analytic cohort shown in Figure 1. Participants were followed from baseline assessment until the first diagnosis of OA, death, loss to follow-up, or the end of follow-up, whichever came first.Figure 1.Flow chart of participant selection and baseline data collection. Flow chart of the study population selection from the UK Biobank cohort. Participants were sequentially excluded for missing data on TyG-related indices, anthropometric measures, prespecified covariates, or targeted metabolites. The final analytic cohort included 255,232 participants. The figure also summarizes the baseline data collected, including sociodemographic factors, anthropometric measurements, behavioral and lifestyle factors, selected metabolites, and cardiometabolic conditions.Abbreviations: BMI, body mass index; HDL-C, high-density lipoprotein cholesterol; IDL-C, intermediate-density lipoprotein cholesterol; OA, osteoarthritis; TyG, triglyceride-glucose. Flow chart of participant selection and baseline data collection. Flow chart of the study population selection from the UK Biobank cohort. Participants were sequentially excluded for missing data on TyG-related indices, anthropometric measures, prespecified covariates, or targeted metabolites. The final analytic cohort included 255,232 participants. The figure also summarizes the baseline data collected, including sociodemographic factors, anthropometric measurements, behavioral and lifestyle factors, selected metabolites, and cardiometabolic conditions. Abbreviations: BMI, body mass index; HDL-C, high-density lipoprotein cholesterol; IDL-C, intermediate-density lipoprotein cholesterol; OA, osteoarthritis; TyG, triglyceride-glucose. Covariates and outcome ascertainment Baseline covariates included age, sex, ethnic group, educational attainment, BMI category, smoking status, drinking status, and Townsend deprivation index (TDI). These variables were selected a priori based on their established associations with cardiometabolic risk and OA. In models for TyG-BMI, TyG-WC, and TyG-WHtR, BMI category was not included as an adjustment covariate when appropriate to avoid overadjustment for body-size components already incorporated into the exposure. The primary outcome was hospital-recorded incident OA during follow-up. Hospital-recorded incident OA was defined as the first recorded hospital inpatient diagnosis of OA (ICD-10 codes M15–M19) after baseline. Therefore, the outcome represents clinically recorded OA rather than radiographic or symptom-defined OA. Participants with prevalent OA at or before baseline were excluded from the incident analysis. Follow-up time was calculated from the baseline assessment date to the first occurrence of OA or censoring. TyG-related indices and CMD burden We evaluated four TyG-related indices: TyG, TyG-BMI, TyG-WC, and TyG-WHtR. UK Biobank reported triglyceride and glucose concentrations in mmol/L. Before calculation of the TyG index, triglyceride concentrations were converted to mg/dL by multiplying by 88.57, and glucose concentrations were converted to mg/dL by multiplying by 18.0. The TyG index was calculated as Ln [triglycerides (mg/dL) × serum glucose (mg/dL)/2]. Because blood samples were not uniformly collected under fasting conditions, the term “serum glucose” was used instead of “fasting glucose”. TyG-BMI, TyG-WC, and TyG-WHtR were derived by multiplying TyG by body mass index (BMI), waist circumference (WC), and waist-to-height ratio (WHtR), respectively. 27 Each TyG-related index was analyzed both as a continuous variable and by quartiles (Q1–Q4). Baseline CMD burden was defined using five prespecified cardiometabolic conditions: hypertension (I10-15), type 2 diabetes (E11), ischemic heart disease (I21-25), stroke (I60-64), and dyslipidemia (E78). For each participant, the number of baseline CMDs was summed and categorized as no CMD, one CMD, or CMM(19). For joint analyses, each TyG-related index was further dichotomized into low (Q1–Q2) and high (Q3–Q4) levels. A six-level joint exposure variable was then created by combining TyG-related index level (low/high) with CMD burden category, with the low-index/no-cardiometabolic-disease group used as the reference. Metabolomic biomarkers We additionally evaluated a set of selected baseline metabolic biomarkers, including low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), total triglycerides, apolipoprotein B(ApoB)/apolipoprotein A1 (ApoA1) ratio, total fatty acids, omega-3 fatty acids, monounsaturated fatty acids (MUFA), glucose measured by nuclear magnetic resonance (NMR), lactate measured by NMR, and glycoprotein acetyls. For descriptive visualization, raw metabolite distributions were displayed with violin plots stratified by hospital-recorded incident OA status, whereas standardized z-scores were used to construct heatmaps of metabolite differences. Statistical analysis Baseline characteristics were summarized as median (interquartile range) for skewed continuous variables, and number (percentage) for categorical variables. Group differences were compared using one-way ANOVA or the Kruskal–Wallis test for continuous variables and the chi-squared test for categorical variables. Associations of TyG-related indices and CMD burden with hospital-recorded incident OA were assessed using Cox proportional hazards regression. Univariable and multivariable models were fitted, with results reported as hazard ratios (HRs) and 95% confidence intervals (CIs). The multivariable models adjusted for age, sex, ethnic group, educational attainment, smoking status, drinking status, and TDI; BMI category was additionally adjusted for in models of TyG, where appropriate. Cumulative incidence patterns across TyG quartiles and CMD burden groups were visualized using Kaplan–Meier–based cumulative hazard curves, and between-group differences were evaluated with the log-rank test. Potential nonlinear associations between TyG-related indices and hospital-recorded incident OA were assessed using restricted cubic spline (RCS) Cox regression models with four knots. Nonlinearity was evaluated using the spline terms in the fitted models. To examine combined risk patterns, we cross-classified low/high TyG-related indices with CMD burden groups and estimated the corresponding OA risk using cumulative hazard curves and multivariable Cox models. To formally assess effect modification, we fitted interaction terms between dichotomized TyG-related indices and CMD burden categories, and evaluated P for interaction using likelihood ratio testing. All analyses were performed using R software (version 4.4.1). The main R packages included dplyr and related tidyverse tools for data management, survival for Cox regression, survminer for survival visualization, rms for restricted cubic spline analyses, ggplot2 for statistical graphics, and broom for model output tidying. All statistical tests were two-sided, and P < 0.05 was considered statistically significant. Results Baseline characteristics After exclusion of participants with missing baseline data, metabolite measurements, covariates, or outcome information, 255,232 individuals were included in the final analytic cohort. Among them, 137,957 (54.0%) were women and 117,275 (46.0%) were men, with a median age of 58 years (IQR, 50–64 years). Compared with participants who did not develop OA during follow-up, those who developed OA were older, had a higher BMI, and had lower educational attainment. They were also more likely to be women and to have type 2 diabetes or ischemic heart disease at baseline. Detailed baseline characteristics were presented in Table 1. In addition, participants who developed OA had higher circulating levels of triglycerides, total fatty acids, omega-3 fatty acids, MUFA, and glucose, shown in Figure 2.Table 1.Baseline characteristics according to incident OA status.CharacteristicOverall N = 255,232 1 No incident OA N = 211,678 1 Incident OA N = 43,554 1 p-value 2 Age, years 58.00 (50.00, 64.00)57.00 (49.00, 63.00)62.00 (56.00, 66.00)<0.001 Sex, n (%) <0.001 Female137,957 (54%)112,042 (53%)25,915 (60%)​ Male117,275 (46%)99,636 (47%)17,639 (40%)​ Ethnicity, n (%) <0.001 Asian5,082 (2.0%)4,433 (2.1%)649 (1.5%)​ Black3,508 (1.4%)3,054 (1.4%)454 (1.0%)​ Mixed1,038 (0.4%)895 (0.4%)143 (0.3%)​ Other2,097 (0.8%)1,825 (0.9%)272 (0.6%)​ White243,507 (95%)201,471 (95%)42,036 (97%)​ BMI category, kg/m 2 , n (%) <0.001 <18.51,311 (0.5%)1,184 (0.6%)127 (0.3%)​ ≥3060,424 (24%)45,278 (21%)15,146 (35%)​ 18.5–24.984,414 (33%)74,341 (35%)10,073 (23%)​ 25–29.9109,083 (43%)90,875 (43%)18,208 (42%)​ Educational attainment, n (%) <0.001 Higher education83,006 (33%)72,613 (34%)10,393 (24%)​ No education43,598 (17%)32,565 (15%)11,033 (25%)​ Secondary education84,015 (33%)70,163 (33%)13,852 (32%)​ Vocational education44,613 (17%)36,337 (17%)8,276 (19%)​ Townsend deprivation index -2.23 (-3.69, 0.37)-2.25 (-3.70, 0.34)-2.18 (-3.66, 0.55)<0.001 Smoking status, n (%) <0.001 Current26,891 (11%)22,547 (11%)4,344 (10%)​ Never140,088 (55%)118,225 (56%)21,863 (50%)​ Previous88,253 (35%)70,906 (33%)17,347 (40%)​ Drinking status, n (%) <0.001 Current235,797 (92%)196,197 (93%)39,600 (91%)​ Never10,632 (4.2%)8,548 (4.0%)2,084 (4.8%)​ Previous8,803 (3.4%)6,933 (3.3%)1,870 (4.3%)​ Diabetes, n (%) 624 (0.2%)383 (0.2%)241 (0.6%)<0.001 Ischemic heart disease, n (%) 752 (0.3%)511 (0.2%)241 (0.6%)<0.001 Stroke, n (%) 229 (<0.1%)172 (<0.1%)57 (0.1%)0.002 Hypertension, n (%) 3,540 (1.4%)2,359 (1.1%)1,181 (2.7%)<0.001 Dyslipidemia, n (%) 1,207 (0.5%)798 (0.4%)409 (1.0%)<0.001 CMD burden, n (%) <0.001 No CMD250,554 (98.2%)208,523 (98.5%)42,031 (96.5%)​ One CMD3,421 (1.3%)2,354 (1.1%)1,067 (2.5%)​ CMM1,257 (0.5%)801 (0.4%)456 (1.0%)​ TyG 8.12 (7.76, 8.48)8.11 (7.75, 8.47)8.19 (7.84, 8.55)<0.001 TyG-BMI 217.26 (191.08, 248.10)214.68 (189.09, 244.60)231.01 (202.73, 264.50)<0.001 TyG-WC 727.54 (635.20, 821.42)720.83 (629.07, 814.07)760.42 (667.60, 855.13)<0.001 TyG-WHtR 4.31 (3.81, 4.83)4.26 (3.77, 4.78)4.53 (4.02, 5.07)<0.001 LDL-C 1.73 (1.45, 2.03)1.73 (1.45, 2.03)1.73 (1.43, 2.03)0.2 HDL-C 1.28 (1.09, 1.52)1.28 (1.09, 1.53)1.28 (1.09, 1.51)<0.001 Triglycerides 1.20 (0.88, 1.64)1.19 (0.87, 1.62)1.27 (0.93, 1.71)<0.001 ApoB/ApoA1 ratio 0.58 (0.47, 0.70)0.58 (0.47, 0.70)0.58 (0.47, 0.69)<0.001 Total fatty acids 11.79 (10.37, 13.44)11.73 (10.33, 13.38)12.06 (10.62, 13.75)<0.001 Omega-3 (%) 4.17 (3.34, 5.16)4.15 (3.33, 5.15)4.26 (3.41, 5.24)<0.001 MUFA (%) 23.44 (21.81, 25.38)23.36 (21.75, 25.29)23.82 (22.15, 25.77)<0.001 Glucose (NMR) 3.55 (3.11, 4.03)3.54 (3.10, 4.01)3.62 (3.17, 4.10)<0.001 Lactate (NMR) 3.83 (3.17, 4.58)3.83 (3.17, 4.59)3.83 (3.16, 4.56)0.11 Glycoprotein acetyls 0.80 (0.73, 0.88)0.80 (0.72, 0.88)0.82 (0.75, 0.90)<0.0011IQR, median (Q1, Q3); n (%).2Wilcoxon rank sum test; Pearson’s Chi-squared test.Ethnicity was self-reported and classified according to UK Biobank categories as White, Black, Asian, Chinese, Mixed, and Other. Educational attainment was self-reported at baseline and categorized as no formal qualifications, vocational education, secondary education, or higher education. Vocational education included NVQ, HNC, HND, and other professional qualifications; secondary education included GCSE, CSE, O-level, and A-level qualifications; and higher education was defined as a college or university degree. Townsend deprivation index (TDI) was derived at baseline from participants’ residential postcode using census-based indicators of unemployment, car ownership, home ownership, and household overcrowding, with higher values indicating greater socioeconomic deprivation.Abbreviations: OA, osteoarthritis; BMI, body mass index; IQR, interquartile range; CMD, cardiometabolic disease; CMM, cardiometabolic multimorbidity; TYG, triglyceride-glucose; WC, waist circumference; WHtR, waist-to-height ratio; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; ApoB, apolipoprotein B; ApoA1, apolipoprotein A1; MUFA, monounsaturated fatty acids; NMR, glucose measured by nuclear magnetic resonance.Figure 2.Baseline metabolite distributions according to incident osteoarthritis status. (A) Violin plots showing the baseline distributions of selected metabolites in participants with and without incident osteoarthritis during follow-up. (B) Heatmap showing standardized mean differences (z-scores) in selected metabolites between participants with no incident osteoarthritis and those who developed incident osteoarthritis. Positive z-scores indicate relatively higher metabolite levels, whereas negative z-scores indicate relatively lower metabolite levels.Abbreviations: ApoB, apolipoprotein B; ApoA1, apolipoprotein A1; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; MUFA, monounsaturated fatty acids; NMR, nuclear magnetic resonance; OA, osteoarthritis. Table 1.Baseline characteristics according to incident OA status.CharacteristicOverall N = 255,232 1 No incident OA N = 211,678 1 Incident OA N = 43,554 1 p-value 2 Age, years 58.00 (50.00, 64.00)57.00 (49.00, 63.00)62.00 (56.00, 66.00)<0.001 Sex, n (%) <0.001 Female137,957 (54%)112,042 (53%)25,915 (60%)​ Male117,275 (46%)99,636 (47%)17,639 (40%)​ Ethnicity, n (%) <0.001 Asian5,082 (2.0%)4,433 (2.1%)649 (1.5%)​ Black3,508 (1.4%)3,054 (1.4%)454 (1.0%)​ Mixed1,038 (0.4%)895 (0.4%)143 (0.3%)​ Other2,097 (0.8%)1,825 (0.9%)272 (0.6%)​ White243,507 (95%)201,471 (95%)42,036 (97%)​ BMI category, kg/m 2 , n (%) <0.001 <18.51,311 (0.5%)1,184 (0.6%)127 (0.3%)​ ≥3060,424 (24%)45,278 (21%)15,146 (35%)​ 18.5–24.984,414 (33%)74,341 (35%)10,073 (23%)​ 25–29.9109,083 (43%)90,875 (43%)18,208 (42%)​ Educational attainment, n (%) <0.001 Higher education83,006 (33%)72,613 (34%)10,393 (24%)​ No education43,598 (17%)32,565 (15%)11,033 (25%)​ Secondary education84,015 (33%)70,163 (33%)13,852 (32%)​ Vocational education44,613 (17%)36,337 (17%)8,276 (19%)​ Townsend deprivation index -2.23 (-3.69, 0.37)-2.25 (-3.70, 0.34)-2.18 (-3.66, 0.55)<0.001 Smoking status, n (%) <0.001 Current26,891 (11%)22,547 (11%)4,344 (10%)​ Never140,088 (55%)118,225 (56%)21,863 (50%)​ Previous88,253 (35%)70,906 (33%)17,347 (40%)​ Drinking status, n (%) <0.001 Current235,797 (92%)196,197 (93%)39,600 (91%)​ Never10,632 (4.2%)8,548 (4.0%)2,084 (4.8%)​ Previous8,803 (3.4%)6,933 (3.3%)1,870 (4.3%)​ Diabetes, n (%) 624 (0.2%)383 (0.2%)241 (0.6%)<0.001 Ischemic heart disease, n (%) 752 (0.3%)511 (0.2%)241 (0.6%)<0.001 Stroke, n (%) 229 (<0.1%)172 (<0.1%)57 (0.1%)0.002 Hypertension, n (%) 3,540 (1.4%)2,359 (1.1%)1,181 (2.7%)<0.001 Dyslipidemia, n (%) 1,207 (0.5%)798 (0.4%)409 (1.0%)<0.001 CMD burden, n (%) <0.001 No CMD250,554 (98.2%)208,523 (98.5%)42,031 (96.5%)​ One CMD3,421 (1.3%)2,354 (1.1%)1,067 (2.5%)​ CMM1,257 (0.5%)801 (0.4%)456 (1.0%)​ TyG 8.12 (7.76, 8.48)8.11 (7.75, 8.47)8.19 (7.84, 8.55)<0.001 TyG-BMI 217.26 (191.08, 248.10)214.68 (189.09, 244.60)231.01 (202.73, 264.50)<0.001 TyG-WC 727.54 (635.20, 821.42)720.83 (629.07, 814.07)760.42 (667.60, 855.13)<0.001 TyG-WHtR 4.31 (3.81, 4.83)4.26 (3.77, 4.78)4.53 (4.02, 5.07)<0.001 LDL-C 1.73 (1.45, 2.03)1.73 (1.45, 2.03)1.73 (1.43, 2.03)0.2 HDL-C 1.28 (1.09, 1.52)1.28 (1.09, 1.53)1.28 (1.09, 1.51)<0.001 Triglycerides 1.20 (0.88, 1.64)1.19 (0.87, 1.62)1.27 (0.93, 1.71)<0.001 ApoB/ApoA1 ratio 0.58 (0.47, 0.70)0.58 (0.47, 0.70)0.58 (0.47, 0.69)<0.001 Total fatty acids 11.79 (10.37, 13.44)11.73 (10.33, 13.38)12.06 (10.62, 13.75)<0.001 Omega-3 (%) 4.17 (3.34, 5.16)4.15 (3.33, 5.15)4.26 (3.41, 5.24)<0.001 MUFA (%) 23.44 (21.81, 25.38)23.36 (21.75, 25.29)23.82 (22.15, 25.77)<0.001 Glucose (NMR) 3.55 (3.11, 4.03)3.54 (3.10, 4.01)3.62 (3.17, 4.10)<0.001 Lactate (NMR) 3.83 (3.17, 4.58)3.83 (3.17, 4.59)3.83 (3.16, 4.56)0.11 Glycoprotein acetyls 0.80 (0.73, 0.88)0.80 (0.72, 0.88)0.82 (0.75, 0.90)<0.0011IQR, median (Q1, Q3); n (%).2Wilcoxon rank sum test; Pearson’s Chi-squared test.Ethnicity was self-reported and classified according to UK Biobank categories as White, Black, Asian, Chinese, Mixed, and Other. Educational attainment was self-reported at baseline and categorized as no formal qualifications, vocational education, secondary education, or higher education. Vocational education included NVQ, HNC, HND, and other professional qualifications; secondary education included GCSE, CSE, O-level, and A-level qualifications; and higher education was defined as a college or university degree. Townsend deprivation index (TDI) was derived at baseline from participants’ residential postcode using census-based indicators of unemployment, car ownership, home ownership, and household overcrowding, with higher values indicating greater socioeconomic deprivation.Abbreviations: OA, osteoarthritis; BMI, body mass index; IQR, interquartile range; CMD, cardiometabolic disease; CMM, cardiometabolic multimorbidity; TYG, triglyceride-glucose; WC, waist circumference; WHtR, waist-to-height ratio; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; ApoB, apolipoprotein B; ApoA1, apolipoprotein A1; MUFA, monounsaturated fatty acids; NMR, glucose measured by nuclear magnetic resonance. Baseline characteristics according to incident OA status. 1IQR, median (Q1, Q3); n (%). 2Wilcoxon rank sum test; Pearson’s Chi-squared test. Ethnicity was self-reported and classified according to UK Biobank categories as White, Black, Asian, Chinese, Mixed, and Other. Educational attainment was self-reported at baseline and categorized as no formal qualifications, vocational education, secondary education, or higher education. Vocational education included NVQ, HNC, HND, and other professional qualifications; secondary education included GCSE, CSE, O-level, and A-level qualifications; and higher education was defined as a college or university degree. Townsend deprivation index (TDI) was derived at baseline from participants’ residential postcode using census-based indicators of unemployment, car ownership, home ownership, and household overcrowding, with higher values indicating greater socioeconomic deprivation. Abbreviations: OA, osteoarthritis; BMI, body mass index; IQR, interquartile range; CMD, cardiometabolic disease; CMM, cardiometabolic multimorbidity; TYG, triglyceride-glucose; WC, waist circumference; WHtR, waist-to-height ratio; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; ApoB, apolipoprotein B; ApoA1, apolipoprotein A1; MUFA, monounsaturated fatty acids; NMR, glucose measured by nuclear magnetic resonance. Baseline metabolite distributions according to incident osteoarthritis status. (A) Violin plots showing the baseline distributions of selected metabolites in participants with and without incident osteoarthritis during follow-up. (B) Heatmap showing standardized mean differences (z-scores) in selected metabolites between participants with no incident osteoarthritis and those who developed incident osteoarthritis. Positive z-scores indicate relatively higher metabolite levels, whereas negative z-scores indicate relatively lower metabolite levels. Abbreviations: ApoB, apolipoprotein B; ApoA1, apolipoprotein A1; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; MUFA, monounsaturated fatty acids; NMR, nuclear magnetic resonance; OA, osteoarthritis. When the study population was stratified by quartiles of the triglyceride-glucose (TyG) index, participants in the highest quartile (Q4) had a higher mean BMI and TDI than those in the lower quartiles (Table S1). Baseline characteristics according to CMD burden are summarized in Table S2. Participants with CMM were older than those without CMD and had a higher BMI, higher TDI, higher TyG-related index values, and lower educational attainment. Metabolically, the Q4 group and CMM group showed higher levels of triglycerides, the ApoB/ApoA1 ratio, MUFA, glucose, lactate, and glycoprotein acetyls. Associations of TyG-related indices with OA incidence During a median follow-up of 12.55 years, 43,554 participants (17.06%) developed hospital-recorded OA over approximately 3,203,162 person-years, corresponding to an incidence rate of 13.6 cases per 1,000 person-years. Kaplan–Meier curves showed a higher cumulative incidence of OA across increasing quartiles of each TyG-related index, with the highest incidence consistently observed in Q4 (all log-rank P < 0.05). In univariable Cox models using Q1 as the reference, the Q4 quartiles of TyG, TyG-BMI, TyG-WC, and TyG-WHtR were associated with 48%, 142%, 99%, and 132% higher risks of hospital-recorded incident OA, respectively. After adjustment for age, sex, ethnic group, educational group, smoking status, drinking status, and TDI; BMI category was additionally included only in the TyG model. Compared with Q1, the adjusted for Q4 were 1.03 (1.00–1.06) for TyG, 2.18 (2.12–2.24) for TyG-BMI, 2.14 (2.08–2.21) for TyG-WC, and 2.00 (1.94–2.06) for TyG-WHtR (all P < 0.05) in Figure 3.Figure 3.Associations of TyG-related index quartiles with incident osteoarthritis. Kaplan–Meier–based cumulative hazard curves and multivariable Cox regression analyses showing the associations between quartiles of TyG-related indices and incident osteoarthritis. Participants were categorized into quartiles (Q1–Q4) according to baseline levels of TyG-related indices, with Q1 serving as the reference group. A graded increase in OA risk was observed across increasing quartiles of TyG-related indices. (A, D, G, J) Kaplan–Meier curves showing cumulative hazard of hospital-recorded incident OA across quartiles of TyG (A), TyG-BMI (D), TyG-WC (G), and TyG-WHtR (I). Shaded areas represent 95% confidence intervals. Log-rank test P values are presented. (B, E, H, K) Univariable Cox regression analyses showing hazard ratios (HRs) and 95% confidence intervals (CIs) for hospital-recorded incident OA across quartiles. (C, F, I, L) Multivariable Cox regression analyses showing adjusted HRs and 95% CIs for hospital-recorded incident OA across quartiles. Models were adjusted for age, sex, ethnicity, educational level, smoking status, drinking status, and Townsend deprivation index. Additional adjustment for BMI was applied where appropriate.Abbreviations: CI, confidence interval; HR, hazard ratio; OA, osteoarthritis; Q, quartile; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose–body mass index; TyG-WC, triglyceride-glucose–waist circumference; TyG-WHtR, triglyceride-glucose–waist-to-height ratio. Associations of TyG-related index quartiles with incident osteoarthritis. Kaplan–Meier–based cumulative hazard curves and multivariable Cox regression analyses showing the associations between quartiles of TyG-related indices and incident osteoarthritis. Participants were categorized into quartiles (Q1–Q4) according to baseline levels of TyG-related indices, with Q1 serving as the reference group. A graded increase in OA risk was observed across increasing quartiles of TyG-related indices. (A, D, G, J) Kaplan–Meier curves showing cumulative hazard of hospital-recorded incident OA across quartiles of TyG (A), TyG-BMI (D), TyG-WC (G), and TyG-WHtR (I). Shaded areas represent 95% confidence intervals. Log-rank test P values are presented. (B, E, H, K) Univariable Cox regression analyses showing hazard ratios (HRs) and 95% confidence intervals (CIs) for hospital-recorded incident OA across quartiles. (C, F, I, L) Multivariable Cox regression analyses showing adjusted HRs and 95% CIs for hospital-recorded incident OA across quartiles. Models were adjusted for age, sex, ethnicity, educational level, smoking status, drinking status, and Townsend deprivation index. Additional adjustment for BMI was applied where appropriate. Abbreviations: CI, confidence interval; HR, hazard ratio; OA, osteoarthritis; Q, quartile; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose–body mass index; TyG-WC, triglyceride-glucose–waist circumference; TyG-WHtR, triglyceride-glucose–waist-to-height ratio. To examine the contribution of the adiposity components, we additionally compared BMI, WC, and WHtR with their corresponding TyG-derived composite indices (Table 2). BMI, WC, and WHtR were each positively associated with hospital-recorded incident OA. In the multivariable models, the HRs (95% CIs) comparing Q4 with Q1 were 1.76 (1.45–2.14) for BMI, 1.96 (1.85–2.08) for WC, and 1.75 (1.58–1.94) for WHtR. The corresponding estimates were 2.18 (2.12–2.24) for TyG-BMI, 2.14 (2.08–2.21) for TyG-WC, and 2.00 (1.94–2.06) for TyG-WHtR. Thus, the composite indices showed numerically stronger associations than their corresponding adiposity measures, although the adiposity measures themselves remained significantly associated with OA.Table 2.Comparative associations of adiposity measures and their corresponding TyG-derived composite indices with incident OA.VariableCategoryUnivariate HR (95% CI)P valueMultivariate HR (95% CI)P valueBMI quartilesQ1ReferenceNANANABMI quartilesQ21.25 (1.18, 1.32)<0.0011.12 (1.01, 1.24)0.032BMI quartilesQ31.42 (1.28, 1.58)<0.0011.33 (1.11, 1.59)0.002BMI quartilesQ42.18 (1.94, 2.45)<0.0011.76 (1.45, 2.14)<0.001TyG-BMI quartilesQ1NANANANATyG-BMI quartilesQ21.37 (1.33, 1.41)<0.0011.27 (1.23, 1.31)<0.001TyG-BMI quartilesQ31.71 (1.66, 1.76)<0.0011.56 (1.52, 1.61)<0.001TyG-BMI quartilesQ42.42 (2.35, 2.48)<0.0012.18 (2.12, 2.24)<0.001WC quartilesQ1ReferenceNANANAWC quartilesQ21.21 (1.13, 1.30)<0.0011.23 (1.08, 1.40)0.002WC quartilesQ31.36 (1.23, 1.50)<0.0011.42 (1.21, 1.72)0.005WC quartilesQ41.76 (1.56, 1.99)<0.0011.96 (1.85, 2.08)<0.001TyG-WC quartilesQ1NANANANATyG-WC quartilesQ21.33 (1.29, 1.37)<0.0011.31 (1.27, 1.35)<0.001TyG-WC quartilesQ31.54 (1.50, 1.59)<0.0011.60 (1.55, 1.65)<0.001TyG-WC quartilesQ41.99 (1.94, 2.05)<0.0012.14 (2.08, 2.21)<0.001WHtR quartilesQ1ReferenceNANANAWHtR quartilesQ21.12 (1.03, 1.22)0.0081.17 (1.03, 1.33)0.016WHtR quartilesQ31.44 (1.28, 1.62)<0.0011.42 (1.19, 1.69)<0.001WHtR quartilesQ42.05 (1.86, 2.26)<0.0011.75 (1.58, 1.94)<0.001TyG-WHtR quartilesQ1NANANANATyG-WHtR quartilesQ21.34 (1.30, 1.38)<0.0011.24 (1.21, 1.28)<0.001TyG-WHtR quartilesQ31.67 (1.62, 1.72)<0.0011.52 (1.47, 1.56)<0.001TyG-WHtR quartilesQ42.32 (2.26, 2.39)<0.0012.00 (1.94, 2.06)<0.001Abbreviations: OA, osteoarthritis; TyG, triglyceride-glucose; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio. Table 2.Comparative associations of adiposity measures and their corresponding TyG-derived composite indices with incident OA.VariableCategoryUnivariate HR (95% CI)P valueMultivariate HR (95% CI)P valueBMI quartilesQ1ReferenceNANANABMI quartilesQ21.25 (1.18, 1.32)<0.0011.12 (1.01, 1.24)0.032BMI quartilesQ31.42 (1.28, 1.58)<0.0011.33 (1.11, 1.59)0.002BMI quartilesQ42.18 (1.94, 2.45)<0.0011.76 (1.45, 2.14)<0.001TyG-BMI quartilesQ1NANANANATyG-BMI quartilesQ21.37 (1.33, 1.41)<0.0011.27 (1.23, 1.31)<0.001TyG-BMI quartilesQ31.71 (1.66, 1.76)<0.0011.56 (1.52, 1.61)<0.001TyG-BMI quartilesQ42.42 (2.35, 2.48)<0.0012.18 (2.12, 2.24)<0.001WC quartilesQ1ReferenceNANANAWC quartilesQ21.21 (1.13, 1.30)<0.0011.23 (1.08, 1.40)0.002WC quartilesQ31.36 (1.23, 1.50)<0.0011.42 (1.21, 1.72)0.005WC quartilesQ41.76 (1.56, 1.99)<0.0011.96 (1.85, 2.08)<0.001TyG-WC quartilesQ1NANANANATyG-WC quartilesQ21.33 (1.29, 1.37)<0.0011.31 (1.27, 1.35)<0.001TyG-WC quartilesQ31.54 (1.50, 1.59)<0.0011.60 (1.55, 1.65)<0.001TyG-WC quartilesQ41.99 (1.94, 2.05)<0.0012.14 (2.08, 2.21)<0.001WHtR quartilesQ1ReferenceNANANAWHtR quartilesQ21.12 (1.03, 1.22)0.0081.17 (1.03, 1.33)0.016WHtR quartilesQ31.44 (1.28, 1.62)<0.0011.42 (1.19, 1.69)<0.001WHtR quartilesQ42.05 (1.86, 2.26)<0.0011.75 (1.58, 1.94)<0.001TyG-WHtR quartilesQ1NANANANATyG-WHtR quartilesQ21.34 (1.30, 1.38)<0.0011.24 (1.21, 1.28)<0.001TyG-WHtR quartilesQ31.67 (1.62, 1.72)<0.0011.52 (1.47, 1.56)<0.001TyG-WHtR quartilesQ42.32 (2.26, 2.39)<0.0012.00 (1.94, 2.06)<0.001Abbreviations: OA, osteoarthritis; TyG, triglyceride-glucose; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio. Comparative associations of adiposity measures and their corresponding TyG-derived composite indices with incident OA. Abbreviations: OA, osteoarthritis; TyG, triglyceride-glucose; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio. Multivariable-adjusted restricted cubic spline analyses showed that the association between TyG and hospital-recorded incident OA was not significantly nonlinear (P for nonlinearity = 0.158) (Figure 4). In contrast, TyG-BMI, TyG-WC, and TyG-WHtR all showed significant nonlinear associations with hospital-recorded incident OA (all P for nonlinearity < 0.001).Figure 4.Dose–response associations between TyG-related indices and incident osteoarthritis. Restricted cubic spline analyses showing the multivariable-adjusted dose–response associations of (A) TyG, (B) TyG-BMI, (C) TyG-WC, and (D) TyG-WHtR with incident osteoarthritis. Solid lines represent hazard ratios, and shaded areas indicate 95% confidence intervals. The dashed horizontal line indicates a hazard ratio of 1.0. P values for nonlinearity are shown in each panel.Abbreviations: CI, confidence interval; HR, hazard ratio; OA, osteoarthritis; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose–body mass index; TyG-WC, triglyceride-glucose–waist circumference; TyG-WHtR, triglyceride-glucose–waist-to-height ratio. Dose–response associations between TyG-related indices and incident osteoarthritis. Restricted cubic spline analyses showing the multivariable-adjusted dose–response associations of (A) TyG, (B) TyG-BMI, (C) TyG-WC, and (D) TyG-WHtR with incident osteoarthritis. Solid lines represent hazard ratios, and shaded areas indicate 95% confidence intervals. The dashed horizontal line indicates a hazard ratio of 1.0. P values for nonlinearity are shown in each panel. Abbreviations: CI, confidence interval; HR, hazard ratio; OA, osteoarthritis; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose–body mass index; TyG-WC, triglyceride-glucose–waist circumference; TyG-WHtR, triglyceride-glucose–waist-to-height ratio. Association of metabolic biomarkers with OA incidence We further compared baseline metabolic biomarker profiles according to hospital-recorded incident OA status. Participants who developed OA showed higher standardized levels of glycoprotein acetyls, glucose, MUFA, omega-3 fatty acids, total fatty acids, and triglycerides than those who remained free of OA. Violin plots showed broadly consistent distributional patterns, with greater dispersion for LDL-C, triglycerides, and MUFA and an opposite trend for HDL-C among participants who developed OA. Association between CMD burden and OA incidence CMD burden was positively associated with hospital-recorded incident OA. Kaplan–Meier curves showed a stepwise increase in cumulative OA incidence with increasing numbers of CMDs (Figure 5). In univariable Cox models, compared with participants without CMD, the HRs (95% CIs) were 2.05 (1.93–2.17) for one CMD and 2.49 (2.27–2.73) for cardiometabolic multimorbidity. After multivariable adjustment, these associations remained significant, although attenuated, with adjusted HRs (95% CIs) of 1.37 (1.29–1.46) and 1.44 (1.31–1.58), respectively.Figure 5.Association between cardiometabolic disease burden and incident osteoarthritis. (A) Kaplan–Meier–based cumulative hazard curves for incident osteoarthritis according to cardiometabolic disease burden categories: no cardiometabolic disease, one cardiometabolic disease, and cardiometabolic multimorbidity. Numbers at risk are shown below the plot. Log-rank P value is presented. (B) Univariable Cox regression analysis showing hazard ratios and 95% confidence intervals for incident osteoarthritis according to cardiometabolic disease burden. (C) Multivariable Cox regression analysis showing adjusted hazard ratios and 95% confidence intervals for incident osteoarthritis according to cardiometabolic disease burden. The reference group was participants without cardiometabolic disease.Abbreviations: CI, confidence interval; CMD, cardiometabolic disease; HR, hazard ratio; OA, osteoarthritis. Association between cardiometabolic disease burden and incident osteoarthritis. (A) Kaplan–Meier–based cumulative hazard curves for incident osteoarthritis according to cardiometabolic disease burden categories: no cardiometabolic disease, one cardiometabolic disease, and cardiometabolic multimorbidity. Numbers at risk are shown below the plot. Log-rank P value is presented. (B) Univariable Cox regression analysis showing hazard ratios and 95% confidence intervals for incident osteoarthritis according to cardiometabolic disease burden. (C) Multivariable Cox regression analysis showing adjusted hazard ratios and 95% confidence intervals for incident osteoarthritis according to cardiometabolic disease burden. The reference group was participants without cardiometabolic disease. Abbreviations: CI, confidence interval; CMD, cardiometabolic disease; HR, hazard ratio; OA, osteoarthritis. In the joint analyses, participants with both high adiposity-integrated TyG indices and CMM had the highest point estimates for hospital-recorded incident OA (Figure 6). We then assessed whether CMD burden modified the associations between TyG-related indices and hospital-recorded incident OA using stratified analyses (Figure 7). TyG alone showed little association with OA risk across CMD strata, with HRs (95% CIs) for high versus low TyG of 1.01 (0.99–1.03) in participants without CMD, 1.01 (0.89–1.14) in those with one CMD, and 1.10 (0.90–1.33) in those with cardiometabolic multimorbidity; no significant interaction was observed (P for interaction = 0.671). In contrast, TyG-BMI, TyG-WC, and TyG-WHtR were consistently associated with higher OA risk in all strata. For TyG-BMI, the HRs (95% CIs) were 1.63 (1.59–1.66), 1.59 (1.39–1.82), and 2.05 (1.61–2.61), respectively; for TyG-WC, they were 1.57 (1.54–1.61), 1.56 (1.38–1.77), and 1.99 (1.61–2.46); and for TyG-WHtR, they were 1.55 (1.51–1.58), 1.48 (1.29–1.69), and 1.97 (1.56–2.49). Although these associations appeared stronger among participants with cardiometabolic multimorbidity, interaction tests were not statistically significant for TyG-BMI (P for interaction = 0.148), TyG-WC (P for interaction = 0.082), or TyG-WHtR (P for interaction = 0.096).Figure 6.Joint associations of TyG-related indices and cardiometabolic disease burden with incident osteoarthritis. Kaplan–Meier–based cumulative hazard curves and Cox regression analyses showing the joint associations of TyG-related indices and cardiometabolic disease burden with incident osteoarthritis. Participants were categorized into six groups according to TyG-related index level (low: Q1–Q2; high: Q3–Q4) and cardiometabolic disease burden (no cardiometabolic disease [CMD], one CMD, and cardiometabolic multimorbidity [CMM]). (A, D, G, J) Kaplan–Meier curves showing cumulative hazard of hospital-recorded incident OA across joint exposure groups for TyG (A), TyG-BMI (D), TyG-WC (G), and TyG-WHtR (J), respectively. Shaded areas represent 95% confidence intervals. Log-rank P values are shown. (B, E, H, K) Univariable Cox regression analyses presenting hazard ratios (HRs) and 95% confidence intervals (CIs) for hospital-recorded incident OA across joint exposure groups. (C, F, I, L) Multivariable Cox regression analyses presenting adjusted HRs and 95% CIs for hospital-recorded incident OA across joint exposure groups. Models were adjusted for age, sex, ethnicity, educational level, smoking status, drinking status, and Townsend deprivation index. Additional adjustment for BMI was applied where appropriate. In all analyses, the low TyG-related index with no CMD group was used as the reference.Abbreviations: CI, confidence interval; CMD, cardiometabolic disease; CMM, cardiometabolic multimorbidity; HR, hazard ratio; OA, osteoarthritis; Q, quartile; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose–body mass index; TyG-WC, triglyceride-glucose–waist circumference; TyG-WHtR, triglyceride-glucose–waist-to-height ratio.Figure 7.Stratified associations between TyG-related indices and incident osteoarthritis according to cardiometabolic disease burden. Forest plots showing the associations of high versus low TyG-related indices with incident osteoarthritis across strata of cardiometabolic disease burden. (A) TyG, (B) TyG-BMI, (C) TyG-WC, and (D) TyG-WHtR. Hazard ratios and 95% confidence intervals were estimated from multivariable Cox regression models within each cardiometabolic disease burden stratum. P values for interaction were derived from likelihood ratio tests comparing models with and without the multiplicative interaction term.Abbreviations: CI, confidence interval; CMD, cardiometabolic disease; HR, hazard ratio; OA, osteoarthritis; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose–body mass index; TyG-WC, triglyceride-glucose–waist circumference; TyG-WHtR, triglyceride-glucose–waist-to-height ratio. Joint associations of TyG-related indices and cardiometabolic disease burden with incident osteoarthritis. Kaplan–Meier–based cumulative hazard curves and Cox regression analyses showing the joint associations of TyG-related indices and cardiometabolic disease burden with incident osteoarthritis. Participants were categorized into six groups according to TyG-related index level (low: Q1–Q2; high: Q3–Q4) and cardiometabolic disease burden (no cardiometabolic disease [CMD], one CMD, and cardiometabolic multimorbidity [CMM]). (A, D, G, J) Kaplan–Meier curves showing cumulative hazard of hospital-recorded incident OA across joint exposure groups for TyG (A), TyG-BMI (D), TyG-WC (G), and TyG-WHtR (J), respectively. Shaded areas represent 95% confidence intervals. Log-rank P values are shown. (B, E, H, K) Univariable Cox regression analyses presenting hazard ratios (HRs) and 95% confidence intervals (CIs) for hospital-recorded incident OA across joint exposure groups. (C, F, I, L) Multivariable Cox regression analyses presenting adjusted HRs and 95% CIs for hospital-recorded incident OA across joint exposure groups. Models were adjusted for age, sex, ethnicity, educational level, smoking status, drinking status, and Townsend deprivation index. Additional adjustment for BMI was applied where appropriate. In all analyses, the low TyG-related index with no CMD group was used as the reference. Abbreviations: CI, confidence interval; CMD, cardiometabolic disease; CMM, cardiometabolic multimorbidity; HR, hazard ratio; OA, osteoarthritis; Q, quartile; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose–body mass index; TyG-WC, triglyceride-glucose–waist circumference; TyG-WHtR, triglyceride-glucose–waist-to-height ratio. Stratified associations between TyG-related indices and incident osteoarthritis according to cardiometabolic disease burden. Forest plots showing the associations of high versus low TyG-related indices with incident osteoarthritis across strata of cardiometabolic disease burden. (A) TyG, (B) TyG-BMI, (C) TyG-WC, and (D) TyG-WHtR. Hazard ratios and 95% confidence intervals were estimated from multivariable Cox regression models within each cardiometabolic disease burden stratum. P values for interaction were derived from likelihood ratio tests comparing models with and without the multiplicative interaction term. Abbreviations: CI, confidence interval; CMD, cardiometabolic disease; HR, hazard ratio; OA, osteoarthritis; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose–body mass index; TyG-WC, triglyceride-glucose–waist circumference; TyG-WHtR, triglyceride-glucose–waist-to-height ratio. Discussion In this large prospective cohort study, three main findings are noteworthy. First, adiposity-integrated TyG indices, including TyG-BMI, TyG-WC, and TyG-WHtR, were consistently associated with a higher hazard of hospital-recorded incident OA, whereas TyG alone showed only a weak association. Second, greater CMD burden was independently associated with OA. Third, participants with both higher adiposity-integrated TyG indices and greater CMD burden had the largest point estimates in the joint-category analyses; however, the absence of statistically significant interaction indicates that these findings should not be interpreted as evidence of multiplicative effect modification or biological synergy. The TyG index is widely used as a surrogate marker of insulin resistance and has been associated with type 2 diabetes and cardiovascular disease.28–30 Although TyG reflects insulin resistance, OA-related metabolic risk likely extends beyond glucose-lipid dysregulation alone.31,32 Adiposity may be linked to OA through both mechanical loading and endocrine-inflammatory pathways, including adipokine release, low-grade inflammation, and synovial activation.33–36 Several mechanisms may help explain these associations. Insulin resistance and adiposity are associated with systemic low-grade inflammation, lipid metabolic disturbance, and adipose tissue-derived mediators that may affect the joint microenvironment.37–39 Within OA joint tissues, inflammatory and hypoxic stress can shift chondrocyte metabolism toward glycolysis, with increased lactate production and reduced mitochondrial oxidative metabolism. This metabolic reprogramming may promote catabolic and inflammatory responses through pathways involving hypoxia-inducible factor-1 alpha (HIF-1α), glucose transporter 1 (GLUT1), lactate dehydrogenase A (LDHA), and nuclear factor kappa B (NF-κB).40,41 In parallel, excess adiposity and hypertriglyceridemia may promote lipotoxic stress, altered fatty acid handling, and cholesterol imbalance in joint tissues. These lipid-related disturbances can activate inflammatory signaling pathways such as Toll-like receptor 4 (TLR4)/NF-κB, disrupt lipid-regulatory pathways, including peroxisome proliferator-activated receptor (PPAR), liver X receptor (LXR), and sterol regulatory element-binding protein (SREBP), and contribute to synovial inflammation and cartilage catabolism.42–44 Therefore, adiposity-integrated TyG indices may capture a more OA-relevant risk profile by combining insulin resistance with body size or central adiposity. We further observed that CMD burden was independently associated with hospital-recorded incident OA after multivariable adjustment. The largest increase in risk occurred between participants with no CMD and those with one CMD, whereas the additional increase from one CMD to cardiometabolic multimorbidity was more modest. This pattern suggested that the transition from metabolic health to clinically recognized cardiometabolic dysfunction may be particularly relevant to OA development. The coexistence of multiple CMDs likely reflects cumulative exposure to systemic inflammation, endothelial dysfunction, oxidative stress, and altered energy metabolism.45–49 Established CMDs may reflect cumulative exposure to systemic inflammation, endothelial dysfunction, oxidative stress, and altered energy metabolism, processes that have been implicated in cartilage degeneration, subchondral bone remodeling, and synovial inflammation.50–52 These findings extend previous evidence linking systemic metabolic dysregulation to OA by showing that CMD burden remains associated with hospital-recorded incident OA after adjustment for traditional risk factors, including age and BMI.8,53 Although participants with both elevated adiposity-integrated TyG indices and CMM had the highest risk of hospital-recorded incident OA, we did not observe statistically significant multiplicative interactions. This finding suggested that CMD burden did not clearly amplify the association between TyG-related indices and OA risk, but rather that these associations were broadly consistent across CMD strata. A plausible explanation is that adiposity-integrated TyG indices and established CMDs are not entirely independent risk domains. Both reflect systemic metabolic dysfunction, with TyG-related indices capturing earlier metabolic abnormalities and CMDs representing clinically recognized disease burden.54–57 Therefore, the joint findings should not be interpreted as evidence of biological synergy, but rather as showing that combining routine metabolic indices with CMD history can identify individuals with accumulated metabolic risk and particularly high OA risk. The potential clinical relevance of these findings was underscored by the accessibility of the indices examined. TyG-BMI, TyG-WC, and TyG-WHtR were derived from routinely available biochemical and anthropometric measurements and could therefore help characterize metabolic risk profiles associated with future OA. These indices were not intended to replace clinical, imaging, or symptom-based assessments but could complement existing approaches by providing additional information on metabolic risk before clinically advanced joint disease developed. The strong associations observed for adiposity-integrated TyG indices also highlight the importance of metabolic health and central adiposity in OA prevention. 58 Interventions targeting insulin resistance, visceral adiposity, and systemic inflammation, including emerging metabolic therapies such as glucagon-like peptide-1 receptor agonists (GLP-1RAs), may therefore warrant further investigation for reducing both cardiometabolic and OA-related risk.59–61 Recent evidence that semaglutide improves obesity-associated knee OA outcomes and may exert weight-independent metabolic effects in joint tissues further supports this translational direction.62,63 However, more interventional studies are needed to determine whether modifying these metabolic profiles can prevent OA progression. Strengths and limitations This study has several strengths. First, it was based on a large prospective cohort with exclusion of prevalent OA at baseline, allowing assessment of hospital-recorded incident OA and strengthening the temporal relationship between metabolic exposure and outcome. Second, by examining TyG together with multiple composite TyG-related indices, we were able to compare different dimensions of metabolic dysfunction and adiposity, providing a more nuanced evaluation of their relative associations with OA risk. Third, the simultaneous assessment of CMD burden, joint exposure categories, and stratified interaction analyses offered a more integrated view of how subclinical metabolic dysfunction and overt CMD relate to OA development. Several limitations should also be acknowledged. First, residual confounding cannot be excluded despite extensive adjustment for major demographic, lifestyle, and socioeconomic factors. Second, TyG-related indices were measured at baseline only and may not capture longitudinal changes in metabolic status over time. Third, OA ascertainment was based on linked clinical records, which may have introduced misclassification and may preferentially capture clinically recognized or more severe cases. Fourth, the complete-case approach resulted in the exclusion of a substantial proportion of the original UK Biobank cohort. Participants with complete exposure, covariate, and metabolomic data may have differed systematically from those with missing data, potentially introducing selection bias and limiting the generalizability of the findings. Multiple imputation was not performed because missingness involved several exposure-defining biochemical and anthropometric components, and imputation could have affected the derived TyG-related indices and their quartile classification. Additionally, death was treated as a censoring event in the cause-specific Cox models, and a Fine–Gray subdistribution hazard analysis was not performed. Because mortality may be higher among participants with greater cardiometabolic disease burden, the reported hazard ratios do not directly estimate the cumulative probability of OA in the presence of competing death. Because fasting status was not standardized in UK Biobank, non-fasting glucose and triglyceride measurements may have reduced the accuracy of the derived TyG index. Finally, as an observational study, causal inference cannot be established, and further mechanistic, longitudinal, and interventional studies are needed to validate these findings. Conclusion In conclusion, adiposity measures, their corresponding TyG-derived composite indices, and greater CMD burden were prospectively associated with hospital-recorded incident OA, whereas TyG alone showed only a weak association. The composite-index associations appeared to be driven substantially by their adiposity components. Although participants with both higher composite indices and greater CMD burden had the highest observed hazard estimates, no multiplicative interaction was detected. These observational findings do not establish causality and require confirmation using broader OA definitions and formal prediction analyses. Supplemental material Supplemental material - Triglyceride-glucose related indices, cardiometabolic multimorbidity, and risk of incident osteoarthritis: A prospective cohort study in UK Biobank Supplemental material for Triglyceride-glucose related indices, cardiometabolic multimorbidity, and risk of incident osteoarthritis: A prospective cohort study in UK Biobank by Kai-Bo Sun, Yuan-Gang Wu, Yu-Huan Xiang, Meng-Ying Li, Yang Ding, Ling-Sheng Zhang, Yi Zeng, Hui-Qi Xie, Bin Shen in Therapeutic Advances in Musculoskeletal Disease Supplemental material - Triglyceride-glucose related indices, cardiometabolic multimorbidity, and risk of incident osteoarthritis: A prospective cohort study in UK Biobank Supplemental material for Triglyceride-glucose related indices, cardiometabolic multimorbidity, and risk of incident osteoarthritis: A prospective cohort study in UK Biobank by Kai-Bo Sun, Yuan-Gang Wu, Yu-Huan Xiang, Meng-Ying Li, Yang Ding, Ling-Sheng Zhang, Yi Zeng, Hui-Qi Xie, Bin Shen in Therapeutic Advances in Musculoskeletal Disease