The association between cardiometabolic index and hypertension-diabetes comorbidity among adults aged 45 and older: evidence from two national databases.

NHANES is a cross-sectional study conducted in the United States that collects health and nutrition information from a nationally representative sample [ 23 ]. The National Center for Health Statistics (NCHS) Data Use Policy allows public access to anonymized health survey data for unrestricted analysis [ 24 ]. The China Health and Retirement Longitudinal Study (CHARLS) aims to collect high-quality microdata on families and people aged 45 and above in China. These data can enhance the examination of population aging in China and promote multidisciplinary research on aging-related matters. All respondents provided written informed consent before participation. Interviewers received intensive standardized training and subsequently carried out face-to-face interviews using structured questionnaires. Individuals interviewed in 2011–2012 comprised the baseline cohort and were followed up in 2013, 2015, 2018 and 2020. the present study used data from the 2011 baseline wave and conducted a cross-sectional analysis [ 25 ]. NHANES administers standardized questionnaires during in-home interviews and at Mobile Examination Centers (MECs), alongside physical measurements and laboratory testing. CHARLS conducts face-to-face, computer-assisted personal interviews (CAPI) at sampled households with periodic physical or biomarker modules. Because participants were aged 45 years and older, the findings may not be generalizable to younger populations. For the NHANES study, we analyzed data from 1999 to 2020. Initially, the study included 107,622 participants. We excluded 28,470 individuals lacking hypertension and diabetes data; 47,903 individuals lacking CMI-related data; 17,606 individuals aged < 45 years; and 674 individuals with missing covariate data. A total of 12,996 individuals were ultimately included. In CHARLS, we analyzed data from 2011. Of the 17,705 participants, we excluded 3,092 individuals with missing hypertension and diabetes data, 4,959 individuals with missing CMI-related data, 186 individuals aged < 45 years, and 108 individuals with missing covariate data. The final sample comprised 9,360 individuals (Fig.  1 ). The specific number of missing variables is presented in Table S13 . Fig. 1 Flow chart of study population Flow chart of study population The CMI measurement for each participant was calculated using the formula: CMI = [waist circumference (cm)/height (cm)] x [triglycerides (mg/dL)/high-density lipoprotein cholesterol (mg/dL)] [ 18 , 19 ]. Waist-to-height ratio (WHtR) = waist circumference (cm)/height (cm); TG/HDL-C = triglycerides (mg/dL)/high-density lipoprotein cholesterol (mg/dL). Hypertension was defined as any of the following: (1) a systolic blood pressure of no less than 140 mmHg or a diastolic blood pressure of no less than 90 mmHg; (2) a self-reported physician diagnosis of hypertension; or (3) current use of antihypertensive medication. In the primary pooled analysis, to maximize data comparability, both datasets used physician diagnosis reported by participants as the criterion for defining diabetes. Participants who answered “yes” to the question, “Has a doctor or healthcare professional ever told you that you have diabetes?” were identified as having the condition. In the NHANES data, we additionally conducted a sensitivity analysis using glycated hemoglobin and fasting plasma glucose as diagnostic indicators where data were available (see Supplementary Table S7 ). Hypertension–diabetes comorbidity was defined as having both hypertension and diabetes. This study included covariates from NHANES, including age, sex, education level, smoking status, alcohol consumption, kidney disease, and cardiovascular disease. Smoking status was determined by the following question: “Have you ever smoked at least 100 cigarettes in your lifetime?” Alcohol consumption frequency was evaluated using the following question over the past 12 months: “On average, how often do you drink any type of alcoholic beverage per week, month, or year?” Participants were also asked: “Including alcoholic beverages such as whiskey or gin, beer, wine, spirits, and any other type of alcoholic drink, have you consumed at least 12 drinks of any type of alcoholic beverage in any given year?” Cardiovascular disease was determined by whether participants reported a diagnosis of any of the following: (1) coronary heart disease, (2) heart attack, (3) stroke, (4) angina, or (5) congestive heart failure. Kidney disease was determined by the question: “Have you ever been told by a doctor or other health professional that you had weak or failing kidneys?” In CHARLS, alcohol consumption was determined based on the question: “Did you drink any alcoholic beverages, such as beer, wine, or liquor in the past year? How often?” Respondents who answered “(1) Drink more than once a month” or “(2) Drink but less than once a month” were classified as drinkers, while those who answered “no” were classified as non-drinkers. Smoking was determined by the question: “Have you ever chewed tobacco, smoked a pipe, smoked self-rolled cigarettes, or smoked cigarettes or cigars?” Responding “Yes” classified the individual as a smoker; otherwise, the individual was classified as a non-smoker. Cardiovascular disease was assessed based on the question: “Have you been diagnosed with a heart attack, coronary heart disease, angina, congestive heart failure, or other heart problems or stroke by a doctor?”. Kidney disease was determined by the question: “Have you been diagnosed with kidney disease (excluding tumors or cancer) by a doctor?”. Covariates were preselected based on prior literature linking obesity and dyslipidemia to hypertension or diabetes, while considering variables available and harmonizable across both NHANES and CHARLS. Specifically, we adjusted for demographic factors (age, sex, education level) and behavioral or clinical factors (smoking, alcohol consumption, kidney disease, history of cardiovascular disease) to reduce potential confounding of the association between CMI and hypertension–diabetes comorbidity. The specific variable screening methods are shown in Table S8. As CMI showed a skewed distribution, we applied a ln transformation (LnCMI = ln(CMI)) to approximate normality. For normally distributed variables, we employed t-tests and reported means ± standard deviations; for skewed variables, we used the Kruskal–Wallis rank-sum test and reported medians (interquartile ranges), categorizing results by LnCMI tertiles. NHANES data were analyzed using multi-cycle survey weights. The analysis was conducted in the R statistical environment, relying on the survey package, which provides the svyglm function for fitting generalized linear models to complex survey data. We examined the association between LnCMI and the odds of hypertension–diabetes comorbidity. Three models were constructed in each database: Model I was unadjusted; Model II was adjusted for age, sex, and education; and Model III was adjusted for age, sex, education level, smoking status, alcohol consumption, kidney disease, and cardiovascular disease. Furthermore, we stratified by sex (male/female), education level (below high school/high school/above high school), smoking status (yes/no), alcohol consumption (yes/no), kidney disease (yes/no), and cardiovascular disease (yes/no), and assessed for interactions. In CHARLS, Model I was unadjusted; Model II incorporated age, sex, and education; and Model III included age, sex, education, kidney disease, cardiovascular disease, smoking, and alcohol consumption. To investigate nonlinearity, we performed restricted cubic spline (RCS) and threshold effect analyses. To evaluate discriminative ability, receiver operating characteristic (ROC) analysis was used. We compared a base multivariable model (covariates only) versus an extended model adding CMI, and computed AUCs based on predicted probabilities. Sensitivity analyses: (1) We conducted separate sensitivity analyses excluding CVD for each population to validate the robustness of the results. (2) Since diabetes mellitus (DM) was defined solely by self-report and is prone to misclassification, we added HbA1c and fasting blood glucose measurements to the diabetes diagnosis in NHANES for sensitivity analysis. (3) We employed decision curve analysis and bootstrap internal validation to enhance the robustness of the results. All statistical analyses were performed using R software (version 4.5.0). A two-tailed P value < 0.05 was considered statistically significant.

Table  1 delineates the attributes of the study population in the NHANES investigation. A total of 12,996 participants aged 45 and above were enrolled, with a mean age of 62.58 ± 10.97 years. Males constituted 49.9% and females 50.1%. The tertile cut-points for LnCMI were: −2.402 to − 0.053 ( ≤ − 0.053), − 0.053 to 0.639 (≤ 0.639), and 0.639 to 4.519 (≤ 4.519). Overall, the prevalence of HTN–DM comorbidity was 14.7%, increasing across LnCMI tertiles (low: 8.3%, middle: 14.9%, high: 20.7%). Significant differences across tertiles were observed in age, weight, height, waist circumference, TG, HDL-C, LnCMI, sex, education level, smoking status, drinking, CVD and kidney disease (all P  < 0.05). Compared with the lowest LnCMI group, individuals in the highest LnCMI group exhibited increased height, weight, waist circumference, TG, CVD prevalence, and kidney disease prevalence; decreased HDL-C and drinking prevalence; and increased smoking prevalence (all P  < 0.05). The prevalence of hypertension–diabetes comorbidity increased with rising LnCMI tertiles ( P  < 0.05). Table 1 Survey-weighted baseline characteristics of NHANES participants (1999–2020) aged ≥ 45 years by LnCMI tertiles LnCMI Level Overall Tertiles Group P low middle high n 12,996 4332 4332 4332 Age (mean (SD)) 62.58 (10.97) 62.26 (11.16) 63.33 (11.06) 62.16 (10.64) < 0.001 Height(cm) (mean (SD)) 166.60 (10.08) 166.42 (9.81) 166.32 (10.09) 167.06 (10.32) 0.0011 Weight (kg) (median [IQR]) 78.80 [67.60, 92.40] 71.10 [61.40, 82.80] 79.50 [69.00, 92.30] 86.55 [74.40, 100.20] < 0.001 Waist (cm) (mean (SD)) 101.45 (15.03) 93.51 (13.54) 102.32 (13.39) 108.53 (14.18) < 0.001 TG (mg/dL) (median [IQR]) 113.00 [80.00, 165.00] 70.00 [57.00, 85.25] 113.00 [96.00, 133.00] 193.00 [156.00, 248.00] < 0.001 HDL-C (mg/dL) (median [IQR]) 52.00 [43.00, 63.00] 66.00 [57.00, 78.00] 52.00 [45.00, 59.00] 41.00 [36.00, 47.00] < 0.001 LnCMI (mean (SD)) 0.30 (0.80) -0.56 (0.40) 0.29 (0.20) 1.17 (0.47) < 0.001 Gender (%) Female 6507 (50.1) 2449 (56.5) 2187 (50.5) 1871 (43.2) < 0.001 Male 6489 (49.9) 1883 (43.5) 2145 (49.5) 2461 (56.8) Education (%) above high school l 6211 (47.8) 2367 (54.6) 2063 (47.6) 1781 (41.1) < 0.001 below high school l 3739 (28.8) 972 (22.4) 1268 (29.3) 1499 (34.6) high school 3046 (23.4) 993 (22.9) 1001 (23.1) 1052 (24.3) Smoking (%) no 6294 (48.4) 2269 (52.4) 2125 (49.1) 1900 (43.9) < 0.001 yes 6702 (51.6) 2063 (47.6) 2207 (50.9) 2432 (56.1) Drinking (%) no 3688 (28.4) 1133 (26.2) 1266 (29.2) 1289 (29.8) < 0.001 yes 9308 (71.6) 3199 (73.8) 3066 (70.8) 3043 (70.2) Hypertension (%) no 5120 (39.4) 1997 (46.1) 1640 (37.9) 1483 (34.2) < 0.001 yes 7876 (60.6) 2335 (53.9) 2692 (62.1) 2849 (65.8) DM (%) no 10,564 (81.3) 3856 (89.0) 3520 (81.3) 3188 (73.6) < 0.001 yes 2432 (18.7) 476 (11.0) 812 (18.7) 1144 (26.4) CVD (%) no 10,717 (82.5) 3735 (86.2) 3590 (82.9) 3392 (78.3) < 0.001 yes 2279 (17.5) 597 (13.8) 742 (17.1) 940 (21.7) HTN–DM comorbidity (%) no 11,091 (85.3) 3971 (91.7) 3685 (85.1) 3435 (79.3) < 0.001 yes 1905 (14.7) 361 (8.3) 647 (14.9) 897 (20.7) Kidney disease (%) no 12,437 (95.7) 4186 (96.6) 4158 (96.0) 4093 (94.5) < 0.001 yes 559 (4.3) 146 (3.4) 174 (4.0) 239 (5.5) Continuous variables: P values were determined using linear regression models; categorical variables: P values were assessed using chi-square tests Abbreviations: CMI, cardiometabolic index; TG, triglycerides; HDL-C, high-density lipoprotein cholesterol; CVD, cardiovascular disease; HTN–DM comorbidity, presence of both hypertension and diabetes; IQR, interquartile range; SD, standard deviation Survey-weighted baseline characteristics of NHANES participants (1999–2020) aged ≥ 45 years by LnCMI tertiles Continuous variables: P values were determined using linear regression models; categorical variables: P values were assessed using chi-square tests Abbreviations: CMI, cardiometabolic index; TG, triglycerides; HDL-C, high-density lipoprotein cholesterol; CVD, cardiovascular disease; HTN–DM comorbidity, presence of both hypertension and diabetes; IQR, interquartile range; SD, standard deviation Table  2 delineates the attributes of CHARLS. A total of 9,360 people aged 45 and older were enrolled, with a median age of 59.00 [52.00–66.00] years. Males accounted for 46.5% of the study population, while females constituted 53.5%. The tertile cut-points for LnCMI were: −3.475 to − 0.214 ( ≤ − 0.214), − 0.214 to 0.485 (≤ 0.485), and 0.485 to 6.070 (≤ 6.070). The prevalence of hypertension–diabetes comorbidity was 3.6%, increasing across LnCMI tertiles (low: 1.6%, middle: 2.6%, high: 6.4%). Compared with the lowest tertile, participants in the highest tertile had higher TG and waist circumference and lower HDL-C, and differed in smoking/drinking status and CVD prevalence (all P  < 0.05). Table 2 Baseline characteristics of CHARLS participants (2011) aged ≥ 45 years by LnCMI tertiles LnCMI Level Overall Tertiles Group P low middle high n 9360 3120 3120 3120 Age (median [IQR]) 59.00 [52.00, 66.00] 59.00 [52.00, 66.00] 59.00 [52.00, 66.00] 58.00 [52.00, 65.00] 0.0023 Height(cm) (median [IQR]) 157.55 [151.90, 164.00] 157.90 [152.10, 164.10] 157.40 [151.40, 164.00] 157.60 [151.90, 164.02] 0.0372 Weight(kg) (median [IQR]) 57.80 [50.90, 65.80] 53.90 [47.80, 60.02] 57.60 [50.70, 65.00] 63.05 [55.50, 70.90] < 0.001 Waist(cm) (median [IQR]) 84.60 [77.88, 92.00] 78.80 [73.20, 84.40] 85.00 [78.80, 91.20] 91.00 [84.30, 97.20] < 0.001 TG(mg/dL) (median [IQR]) 106.20 [75.22, 154.88] 66.38 [54.87, 79.65] 105.32 [89.39, 123.90] 185.41 [146.91, 248.91] < 0.001 HDL-C(mg/dL) (median [IQR]) 49.10 [40.21, 59.92] 62.24 [54.12, 71.91] 49.48 [43.69, 56.44] 38.27 [32.86, 44.46] < 0.001 LnCMI (median [IQR]) 0.12 [-0.40, 0.69] -0.61 [-0.88, -0.40] 0.12 [-0.04, 0.30] 0.96 [0.69, 1.38] < 0.001 Gender (%) Female 5009 (53.5) 1468 (47.1) 1722 (55.2) 1819 (58.3) < 0.001 Male 4351 (46.5) 1652 (52.9) 1398 (44.8) 1301 (41.7) Education (%) above high school 294 (3.1) 84 (2.7) 107 (3.4) 103 (3.3) 0.3407 below high school 8445 (90.2) 2838 (91.0) 2809 (90.0) 2798 (89.7) high school 621 (6.6) 198 (6.3) 204 (6.5) 219 (7.0) Smoking (%) no 5687 (60.8) 1751 (56.1) 1929 (61.8) 2007 (64.3) < 0.001 yes 3673 (39.2) 1369 (43.9) 1191 (38.2) 1113 (35.7) Drinking (%) no 6295 (67.3) 1885 (60.4) 2173 (69.6) 2237 (71.7) < 0.001 yes 3065 (32.7) 1235 (39.6) 947 (30.4) 883 (28.3) Hypertension (%) no 5464 (58.4) 2117 (67.9) 1846 (59.2) 1501 (48.1) < 0.001 yes 3896 (41.6) 1003 (32.1) 1274 (40.8) 1619 (51.9) DM (%) no 8814 (94.2) 3016 (96.7) 2972 (95.3) 2826 (90.6) < 0.001 yes 546 (5.8) 104 (3.3) 148 (4.7) 294 (9.4) CVD (%) no 8093 (86.5) 2776 (89.0) 2729 (87.5) 2588 (82.9) < 0.001 yes 1267 (13.5) 344 (11.0) 391 (12.5) 532 (17.1) HTN–DM comorbidity (%) no 9026 (96.4) 3069 (98.4) 3038 (97.4) 2919 (93.6) < 0.001 yes 334 (3.6) 51 (1.6) 82 (2.6) 201 (6.4) Kidney disease (%) no 8732 (93.3) 2888 (92.6) 2913 (93.4) 2931 (93.9) 0.0918 yes 628 (6.7) 232 (7.4) 207 (6.6) 189 (6.1) Continuous variables: P values were determined using linear regression models; categorical variables: P values were assessed using chi-square tests Abbreviations: CMI, cardiometabolic index; TG, triglycerides; HDL-C, high-density lipoprotein cholesterol; CVD, cardiovascular disease; HTN–DM comorbidity, presence of both hypertension and diabetes; IQR, interquartile range; SD, standard deviation Baseline characteristics of CHARLS participants (2011) aged ≥ 45 years by LnCMI tertiles Continuous variables: P values were determined using linear regression models; categorical variables: P values were assessed using chi-square tests Abbreviations: CMI, cardiometabolic index; TG, triglycerides; HDL-C, high-density lipoprotein cholesterol; CVD, cardiovascular disease; HTN–DM comorbidity, presence of both hypertension and diabetes; IQR, interquartile range; SD, standard deviation Table  3 of the NHANES study indicates that LnCMI exhibited a positive association with hypertension–diabetes comorbidity across all three models. In Model I, each unit increase in LnCMI was associated with higher odds of hypertension–diabetes comorbidity (OR = 1.994, 95% CI: 1.818–2.186). In Model II, each unit increase in LnCMI was associated with an OR of 2.068 (95% CI: 1.864–2.295). In Model III, each unit increase in LnCMI was associated with an OR of 1.995 (95% CI: 1.794–2.218). In the LnCMI tertile analysis, this association remained statistically significant: in the fully adjusted model, the odds of hypertension–diabetes comorbidity were higher in the highest tertile than in the lowest tertile (OR = 3.696, 95% CI: 3.057–4.468; P for trend < 0.001). In addition, restricted cubic spline (RCS) analyses were used to examine potential nonlinearity between LnCMI and hypertension–diabetes comorbidity. Threshold effect analyses suggested an inflection point in this association. In NHANES, in the fully adjusted model, the inflection point was 0.286. In a two-piecewise model, when LnCMI < 0.286, each unit increase in LnCMI was associated with an OR of 2.068 (95% CI: 1.758–2.444; P  < 0.001). When LnCMI ≥ 0.286, each unit increase in LnCMI was associated with an OR of 1.557 (95% CI: 1.405–1.725; P  < 0.001). The likelihood ratio test supported the two-piecewise model ( P  < 0.05), as depicted in Fig.  2 and Table S1. Table 3 Weighted multivariable logistic regression of hypertension–diabetes comorbidity on LnCMI, NHANES LnCMI Model I OR (95%CI P value) Model II OR (95%CI P value) Model III OR (95%CI P value) Continue 1.994(1.818, 2.186), P  < 0.001 2.068(1.864, 2.295), P  < 0.001 1.995(1.794, 2.218), P  < 0.001 LnCMI Tertile group low 1(Ref) 1(Ref) 1(Ref) middle 2.286(1.947, 2.684), P  < 0.001 2.195(1.867, 2.579), P  < 0.001 2.172(1.842, 2.562), P  < 0.001 high 3.8594(3.2404,4.5967) p  < 0.001 3.939(3.274, 4.738), P  < 0.001 3.696(3.057, 4.468), P  < 0.001 P for trend < 0.001 < 0.001 < 0.001 Model I was unadjusted Model II was adjusted for age, sex and education Model III was adjusted for age, sex, education, smoking status, alcohol consumption, kidney disease and cardiovascular disease Weighted multivariable logistic regression of hypertension–diabetes comorbidity on LnCMI, NHANES Model I was unadjusted Model II was adjusted for age, sex and education Model III was adjusted for age, sex, education, smoking status, alcohol consumption, kidney disease and cardiovascular disease Fig. 2 Restricted cubic spline (RCS) analysis of the association between LnCMI and hypertension–diabetes comorbidity in NHANES. Survey-weighted logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) across three models (Model I–III, as defined in Table 3 ). Solid lines represent adjusted ORs, shaded areas denote 95% CIs, and the dashed line indicates OR = 1 Restricted cubic spline (RCS) analysis of the association between LnCMI and hypertension–diabetes comorbidity in NHANES. Survey-weighted logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) across three models (Model I–III, as defined in Table 3 ). Solid lines represent adjusted ORs, shaded areas denote 95% CIs, and the dashed line indicates OR = 1 In the CHARLS investigation, Table  4 indicates that LnCMI exhibited a positive association with hypertension–diabetes comorbidity across all three models. In Model I, each unit increase in LnCMI was associated with an OR of 1.790 (95% CI: 1.610–1.988). In Model II, each unit increase in LnCMI was associated with an OR of 1.808 (95% CI: 1.623–2.012). In Model III, each unit increase in LnCMI was associated with an OR of 1.760 (95% CI: 1.574–1.964). This association remained statistically significant in tertile analyses: in the fully adjusted model, the odds of hypertension–diabetes comorbidity were higher in the highest tertile than in the lowest tertile (OR = 3.788, 95% CI: 2.782–5.249; P for trend < 0.001). We also examined potential nonlinearity using restricted cubic spline (RCS) (Fig.  3 and Table S2) and performed threshold effect analysis. In the fully adjusted model, the LnCMI inflection point was 0.121. In a two-piecewise model, when LnCMI < 0.121, each unit increase in LnCMI was associated with an OR of 2.683 (95% CI: 1.757–4.253; P  < 0.001). When LnCMI ≥ 0.121, each unit increase in LnCMI was associated with an OR of 1.599 (95% CI: 1.374–1.848; P  < 0.001). The likelihood ratio test supported the two-piecewise model ( P  < 0.05). Table 4 Multivariable logistic regression of hypertension–diabetes comorbidity on LnCMI, CHARLS LnCMI Model I OR(95%CI P value) Model II OR(95%CI P value) Model III OR(95%CI P value) Continue 1.790(1.610, 1.988), P  < 0.001 1.808(1.623, 2.012), P  < 0.001 1.760(1.574, 1.964), P  < 0.001 LnCMI Tertile group low 1(Ref) 1(Ref) 1(Ref) middle 1.624(1.145, 2.325), P  = 0.007 1.573(1.108, 2.255), P  = 0.012 1.549(1.089, 2.225), P  = 0.016 high 4.144(3.061, 5.714), P  < 0.001 4.098(3.020, 5.661), P  < 0.001 3.788(2.782, 5.249), P  < 0.001 P for trend < 0.001 < 0.001 < 0.001 Model I was unadjusted Model II was adjusted for age, sex and education Model III was adjusted for age, sex, education, smoking status, alcohol consumption, kidney disease and cardiovascular disease Multivariable logistic regression of hypertension–diabetes comorbidity on LnCMI, CHARLS Model I was unadjusted Model II was adjusted for age, sex and education Model III was adjusted for age, sex, education, smoking status, alcohol consumption, kidney disease and cardiovascular disease Fig. 3 Restricted cubic spline (RCS) analysis of the association between LnCMI and hypertension–diabetes comorbidity in CHARLS. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated across three models. Solid lines represent adjusted ORs, shaded areas denote 95% CIs, and the dashed line indicates OR = 1 Restricted cubic spline (RCS) analysis of the association between LnCMI and hypertension–diabetes comorbidity in CHARLS. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated across three models. Solid lines represent adjusted ORs, shaded areas denote 95% CIs, and the dashed line indicates OR = 1 In the NHANES study, we conducted subgroup analyses with interaction testing to evaluate whether the association between LnCMI and hypertension–diabetes comorbidity differed across subgroups. Figure  4 illustrates that with every 1-unit increment in LnCMI, the odds of hypertension–diabetes comorbidity increased by 75.3% in men (OR = 1.753, 95% CI: 1.572–1.954) and by 136.4% in women (OR = 2.136, 95% CI: 2.009–2.782). Among smokers, the association was OR = 1.868 (95% CI: 1.650–2.114), and among non-smokers, the association was OR = 2.191 (95% CI: 1.879–2.555). Interaction testing suggested that sex and smoking status modified the association between LnCMI and hypertension–diabetes comorbidity. No statistically significant interactions were observed for age, education level, alcohol consumption, kidney disease, or CVD (all P for interaction > 0.05). Fig. 4 Subgroup analysis of the association between LnCMI and hypertension–diabetes comorbidity in NHANES. ORs and 95% CIs were estimated using survey-weighted logistic regression (Model III). Interaction P values were obtained by including interaction terms between LnCMI and subgroup indicators. The vertical dashed line indicates OR = 1 Subgroup analysis of the association between LnCMI and hypertension–diabetes comorbidity in NHANES. ORs and 95% CIs were estimated using survey-weighted logistic regression (Model III). Interaction P values were obtained by including interaction terms between LnCMI and subgroup indicators. The vertical dashed line indicates OR = 1 In the CHARLS investigation, we examined the association between LnCMI and hypertension–diabetes comorbidity across subgroups using interaction testing. Figure  5 illustrates that among men, the association between LnCMI and hypertension–diabetes comorbidity was OR = 1.668 (95% CI: 1.402–1.974), and among women, the association was OR = 1.807 (95% CI: 1.557–2.096). Interaction tests indicated no substantial effect modification by sex on the relationship between LnCMI and hypertension–diabetes comorbidity. The positive association between LnCMI and hypertension–diabetes comorbidity was also consistent across CVD, education level, smoking, kidney disease, age, and alcohol intake subgroups (all interaction P values > 0.05). Fig. 5 Subgroup analysis of the association between LnCMI and hypertension–diabetes comorbidity in CHARLS. ORs and 95% CIs were estimated using logistic regression (Model III). Interaction P values were obtained by including product terms between LnCMI and subgroup indicators. The vertical dashed line indicates OR = 1 Subgroup analysis of the association between LnCMI and hypertension–diabetes comorbidity in CHARLS. ORs and 95% CIs were estimated using logistic regression (Model III). Interaction P values were obtained by including product terms between LnCMI and subgroup indicators. The vertical dashed line indicates OR = 1 To assess the discriminative ability of various indicators (TG/HDL-C, BMI, WHtR, and CMI) for hypertension–diabetes comorbidity, we performed receiver operating characteristic (ROC) curve analyses in both databases. As shown in Tables S3–S4 and Fig.  6 , CMI demonstrated a higher AUC than the other indicators in CHARLS, while WHtR slightly exceeded CMI in NHANES. These discrepancies may reflect differences in population composition, measurement modules, and cardiometabolic characteristics between the two surveys. Fig. 6 Receiver operating characteristic (ROC) curves comparing the discriminatory ability of CMI, TG/HDL-C, WHtR, and BMI for hypertension–diabetes comorbidity in NHANES and CHARLS. Calculations in NHANES were performed using survey weights. Adjusted variables included age, sex, education, smoking status, alcohol consumption, kidney disease, and cardiovascular disease Receiver operating characteristic (ROC) curves comparing the discriminatory ability of CMI, TG/HDL-C, WHtR, and BMI for hypertension–diabetes comorbidity in NHANES and CHARLS. Calculations in NHANES were performed using survey weights. Adjusted variables included age, sex, education, smoking status, alcohol consumption, kidney disease, and cardiovascular disease The association between LnCMI and hypertension–diabetes comorbidity remained robust in sensitivity analyses using the fully adjusted model (Model III) after excluding CVD (Tables S5–S6). In the NHANES sensitivity analysis, diabetes was defined as HbA1c ≥ 6.5%, fasting plasma glucose ≥ 126 mg/dL, or self-reported diabetes; results showed a consistent association between LnCMI and hypertension–diabetes comorbidity (OR = 2.218, 95% CI: 1.940–2.333; P  < 0.001) (Table S7). Decision curve analysis suggested a higher net benefit of the CMI-extended model than the base model across clinically relevant threshold probabilities (Figure S2; Tables S9 – S10 ). Bootstrap internal validation showed minimal optimism and stable calibration (Table S11 ).

Hypertension is one of the common risk factors for ischemic heart disease, stroke, other cardiovascular diseases (CVD), chronic kidney disease, and other conditions [ 1 – 4 ]. The latest 2023 NHANES data released in the United States indicates that the prevalence of hypertension among adults aged 18 and older reached 47.7% [ 5 ]. Hypertension imposes a heavy burden on healthcare and socioeconomic systems, with related annual expenditures in the United States estimated at approximately $219 billion in 2019 [ 6 ]. According to the latest national report from the International Diabetes Federation (IDF), the number of people with diabetes aged 20–79 in the United States reached approximately 38.5 million in 2024, with an age-standardized prevalence rate of 13.7%. In terms of economic burden, diabetes-related healthcare expenditures in the United States reached approximately $404.5 billion in 2024. The number of patients is projected to reach approximately 43 million by 2050 [ 7 ]. The American Diabetes Association (ADA) assessment of socioeconomic burden similarly indicates that the total economic cost of diabetes in 2022 was approximately $412.9 billion [ 8 ]. Additionally, recent nationwide data analysis in China shows that the prevalence of hypertension among adults aged ≥ 18 years was approximately 31.6% in 2021–2022 [ 9 ]. Concurrently, the WHO (2023) Country Profile Report indicates that approximately 256.7 million individuals aged 30–79 in China have hypertension, highlighting its immense scale [ 10 ]. Regarding economic burden, a recent study in rural southwest China found that the per capita economic burden of hypertension rose from $1,369 to $2,244 (2021 data) between 2011 and 2021, indicating sustained growth in disease burden and resource consumption [ 11 ]. Given that hypertension and diabetes frequently coexist and exert synergistic effects on cardiovascular and renal complications, establishing a simplified indicator for stratifying the prevalence of hypertension–diabetes comorbidity holds significant public health implications. Obesity is a major risk factor for cardiovascular diseases, hypertension, stroke, diabetes, and other conditions [ 12 – 14 ]. The body mass index, albeit the most prevalent metric for evaluating human obesity, fails to adequately represent body composition or the distribution of visceral fat [ 15 ]. The Cardiometabolic Index (CMI) is a novel metric that reflects adiposity and lipid profiles [ 16 ]. The computation utilizes the ratio of triglyceride-to-high-density lipoprotein cholesterol and the ratio of waist to height [ 17 , 18 ]. Currently, CMI is being used in research on various diseases, such as cardiovascular disease [ 19 ], diabetes [ 20 ], and endometriosis [ 21 ]. The triglycerides, height, waist circumference, and high-density lipoprotein cholesterol that constitute the CMI are clinically simple and readily obtainable, enhancing the CMI’s universality [ 22 ]. The CMI may be used for screening and prevalence stratification, aiding in the earlier identification of high-prevalence populations. This may facilitate the implementation of preventive strategies, potentially reducing subsequent cardiovascular and renal complications and the associated healthcare burden. Prior research has investigated the association between the cardiometabolic index (CMI) and hypertension or diabetes. Nonetheless, no studies have yet established the association between CMI and hypertension–diabetes comorbidity among individuals with both conditions. In addition, prior research within the American demographic has predominantly concentrated on adults, lacking specific investigations into the association between CMI and hypertension or diabetes among middle-aged and older populations. Moreover, previous studies have not performed direct comparisons between Chinese and American populations. This study investigates the association between CMI and hypertension–diabetes comorbidity among individuals aged 45 years and older in China and the United States.

Both datasets showed an association between LnCMI and the prevalence of hypertension–diabetes comorbidity in persons aged 45 years and older. In both populations, the prevalence of hypertension–diabetes comorbidity increased with increasing LnCMI, and this pattern remained statistically significant whether LnCMI was modeled as a continuous or categorical variable. In NHANES, the inflection point for LnCMI was 0.286, while in CHARLS it was 0.121. Subgroup analyses further supported the consistency of this association across several demographic and clinical categories. In NHANES, interaction testing suggested effect modification by sex and smoking status, indicating that the association between LnCMI and hypertension–diabetes comorbidity may differ by these factors. ROC analyses indicated that WHtR had the highest discriminative ability in NHANES (AUC = 0.739), while CMI showed the highest discrimination in CHARLS (AUC = 0.739). Sensitivity analyses supported the robustness of the association between LnCMI and hypertension–diabetes comorbidity. Compared with WHtR and TG/HDL-C alone, LnCMI incorporates both central adiposity (WHtR) and lipid traits (TG/HDL-C), which may capture complementary cardiometabolic information. In our subgroup analysis, the association between LnCMI and hypertension–diabetes comorbidity was stronger in women and non-smokers in NHANES (P for interaction = 0.003), suggesting potential subgroup-specific utility. CMI, as an index reflecting visceral fat and metabolic dysfunction, has demonstrated predictive value in conditions such as cardiovascular diseases [ 19 ], diabetes [ 20 ], endometriosis [ 21 ], kidney stones [ 26 ] and osteoporosis [ 27 ]. CMI provides a useful proxy for visceral adiposity and shows strong correlations with metabolic conditions, closely linked to obesity [ 28 ]. First, excessive visceral fat distribution is associated with changes in inflammatory mediators and endothelial function [ 29 ]. Second, the renin-angiotensin-aldosterone system can be activated, leading to impaired sodium excretion and increased sodium reabsorption in the renal tubules [ 30 ]. Additionally, visceral fat tissue participates in blood pressure regulation through the secretion of adipokines that regulate arterial tone, such as leptin and adiponectin [ 31 ]. Finally, excessive visceral fat tissue increases insulin resistance, leading to metabolic dysfunction, oxidative stress, and vascular dysfunction [ 32 ]. Together, these pathways support the biological plausibility that higher CMI is associated with hypertension and may help prevalence stratification. To our knowledge, this study is the first to examine the association between LnCMI and hypertension–diabetes comorbidity using a direct comparison of databases from China and the United States. By specifically focusing on middle-aged and older populations, the findings are more targeted. Methodologically, data underwent ln transformation to improve model fit. Our findings indicate that higher LnCMI was associated with higher odds of hypertension–diabetes comorbidity. Furthermore, consistent with previous studies indicating sex-dependent associations between CMI and hypertension [ 33 ], our analysis suggests that the association between LnCMI and hypertension–diabetes comorbidity may differ by sex [ 33 ]. Estrogen modulates key cardiovascular processes, including vascular nitric oxide production, endothelial function, sympathetic activity, and the renin-angiotensin system. A reduction in estrogen levels can impair endothelial relaxation, increase vascular stiffness, and promote salt sensitivity. By exerting potent pro-inflammatory and pro-oxidative effects, triglyceride-rich remnants and triglyceride-displaced HDL-C disrupt key vascular protective mechanisms, including reverse cholesterol transport and native HDL-C antioxidant capacity. This HDL-C dysfunction may be particularly detrimental for women, as sex hormones and their receptors differentially regulate the endothelial–HDL-C interface, potentially heightening vascular risk during this transition [ 34 – 37 ]. Unlike previous studies, the association between LnCMI and hypertension–diabetes comorbidity may differ by smoking status because smoking drives chronic inflammation [ 38 ], oxidative stress [ 39 ], impaired endothelial function [ 40 ], sympathetic activation [ 41 ], and insulin resistance [ 42 ]. When baseline risk pathways in smokers are already elevated by tobacco exposure, the relative incremental association of higher CMI may be less pronounced compared with non-smokers. Among non-smokers, elevated CMI levels often more strongly reflect central obesity and dyslipidemia, aligning with pathways such as the renin-angiotensin-aldosterone system [ 43 ], adipokines [ 44 ], insulin resistance [ 45 ], and oxidative stress [ 46 ]. Differences in ROC performance between NHANES and CHARLS suggest that the discriminatory ability of CMI may depend on population context. Our conjecture is that in NHANES, central fat distribution captured by WHtR likely concentrated much of the discriminative information for comorbidity, whereas in CHARLS, TG/HDL-C combined with central fat distribution and an atherogenic lipid profile may provide additional differentiation. Differences in measurement patterns and baseline prevalence distributions may also contribute. Furthermore, our primary findings are based on multivariable association models and nonlinear associations observed in both populations. There was a marked difference in baseline comorbidity prevalence (NHANES 14.7% vs. CHARLS 3.6%). This may help explain the differing inflection points of LnCMI between the two cohorts (NHANES 0.286 vs. CHARLS 0.121) and the variation in discriminative rankings (WHtR was highest in NHANES, while CMI performed best in CHARLS). Furthermore, we observed more pronounced associations among women and non-smokers in NHANES, potentially attributable to environmental, geographic, and demographic factors. Additionally, the consistent positive associations across both datasets support the robustness of LnCMI as a population-level marker for prevalence stratification. A principal advantage of this study is the use of NHANES and CHARLS data, both of which have large sample sizes and national representativeness. To our knowledge, this is the first study to investigate the association between LnCMI and hypertension–diabetes comorbidity in adults aged 45 years and older using harmonized analyses across China and the United States. We validated the findings using subgroup analyses, RCS, interaction testing, threshold analyses, ROC analysis, decision curve analysis, and bootstrap internal validation. Another key strength is cross-national verification using two representative surveys. Consistent associations across populations with distinct sociocultural backgrounds, healthcare systems, and cardiometabolic profiles enhance external validity. Because CMI derives from routine anthropometric and lipid measurements, it may serve as a practical tool for screening and prevalence stratification in community and primary care settings. We recognize several limitations. First, the cross-sectional design constrains causal inference, and residual confounding from unmeasured variables may exist. Second, because participants were aged 45 years and older from China and the United States, the findings may not be generalizable to other age groups or settings. Third, diagnostic bias in hypertension and diabetes may vary across age groups and survey settings. In addition, because we relied on parent survey instruments, we did not conduct independent re-validation of the questionnaires. Potential measurement error and mode differences between NHANES (home/MEC) and CHARLS (CAPI) may remain. Therefore, we conducted sensitivity analyses using NHANES laboratory data and provided reconciliation tables to enhance transparency (Table S8). Future longitudinal studies should evaluate temporal relationships, assess incremental predictive value beyond established measures, and determine whether CMI-guided strategies improve preventive outcomes.

We conducted parallel analyses in NHANES and CHARLS and compared the consistency of associations across countries. Overall, higher LnCMI (log-transformed CMI) was associated with a higher prevalence of hypertension–diabetes comorbidity in persons aged 45 years and older in both China and the United States. We also observed evidence of nonlinearity and potential inflection points in the association between LnCMI and hypertension–diabetes comorbidity. CMI may be a practical marker for prevalence stratification, but longitudinal studies are needed to evaluate predictive utility and clinical impact.

Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1