Evaluating a new obesity indicator for stroke risk prediction: comparative cohort analysis in rural settings of two nations Authors: Feifei Yao, Jing Cui, Yuncheng Shen, Yuting Jiang, Yuanyuan Li, Xiaona Liu, Hongqi Feng, Zhe Jiao, Chang Liu, Fulan Hu, Wei Zhang, Dianjun Sun Published: 2024 ABSTRACT AbstractBackgroundWhile the TyG index has been studied in relation to stroke risk, there is a lack of research integrating fat distribution indicators like Body Roundness Index (BRI) and Fat Mass Index (FMI). Additionally, comparative studies across multiple regions are scarce. This study investigates the association between obesity-related parameters and stroke incidence, examining the mediation effects of multimorbidity, using data from rural areas in China and the United Kingdom.MethodsThis cohort study included 60,685 participants (6,980 from China and 53,705 from UK). The obesity-related parameters were calculated using established formulas. The TyG index was determined as ln [TG (mg/dL) × GLU (mg/dL) / 2]. Additionally, composite indices were created by multiplying the TyG index by BMI, WC, FMI, and RBI to assess obesity-related risks. Cox regression analyses were employed on the relationship between Triglyceride Glucose index related parameters and stroke risk. Multiple mediation analysis was applied to assess the contributions of multimorbidity to obesity indicators in stroke occurrence.ResultsAfter excluding those who developed stroke within two years of enrollment, the Chinese cohort (6,638 subjects, median follow-up 4.33 years) had 237 ischemic and 21 hemorrhagic strokes. The UK cohort (53,631 subjects, median follow-up 13.85 years) had 742 ischemic and 316 hemorrhagic strokes. Chinese residents had lower BMI but higher visceral obesity (BRI), higher prevalence of multimorbidity, and higher stroke incidence compared to UK residents.Cox analyses demonstrated significant associations between BMI/TyG indices and ischemic stroke in both Chinese and UK populations, which diminished after adjusting for multimorbidity. In the Chinese rural cohort, only TyG-BRI (HR:1.13, 95%CI:0.99–1.30) approached statistical significance after full adjustment for mediators. In contrast, in the UK cohort, significant associations persisted for most TyG Index indicators when full adjustment for mediators, including BMI (HR: 1.17, 95% CI: 1.09–1.26), TyG-BMI (HR: 1.16, 95% CI: 1.07–1.26), TyG-WC (HR: 1.13, 95% CI: 1.03–1.25), TyG-FMI (HR: 1.17, 95% CI: 1.07–1.28), and TyG-RBI (HR: 1.15, 95% CI: 1.06–1.24). TyG-BRI also showed the best predictive performance for ischemic stroke in Chinese rural residents (AUC > 0.7) and exhibited an almost linear relationship with ischemic stroke occurrence. Additionally, TyG-BRI presented a U-shaped relationship with the risk of hemorrhagic stroke incidence in the UK (p overall = 0.041, p non-linear = 0.017). Multimorbidity mediated the relationship between TyG indices, and ischemic stroke incidence in both cohorts. The mediation percentage for multimorbidity was higher than the sum of individual chronic diseases, with a higher mediation percentage in the Chinese cohort (up to 51%) compared to the UK cohort (up to 27.2%).ConclusionsChinese rural residents exhibit higher levels of visceral obesity compared to residents in UK, leading to greater stroke susceptibility mediated by multimorbidity. These findings underscore the importance of comprehensive management of multimorbidity for stroke prevention. The TyG-BRI may serve as a promising predictor of ischemic stroke incidence among rural community residents.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12889-024-20631-5.Keywords: TyG index, Obesity, Stroke, Rural residents, Multimorbidity Introduction Stroke accounted for the highest age-standardized Disability-Adjusted Life Years (DALYs) among nervous system in 2021, are the leading cause of DALYs globally, impacting the health significantly [1]. Notably, the mortality associated with stroke was higher in Asia compared to Europe and North America [2]. While high-income countries have witnessed a decline in stroke incidence, the trend in China and other middle- to low-income nations is on an upward trajectory, and China at the top in terms of stroke prevalence [3]. In China, the northeastern region has the highest stroke incidence (365.2 per 100,000 person-years) [4], underscoring the urgent need for enhanced primary prevention efforts in the region. Body mass index (BMI) is commonly employed to estimate obesity, findings from Mendelian randomization studies based on data from European cohorts indicate that a higher BMI does not necessarily correlate with an increased stroke risk [5]. This discovery aligns with the observed "obesity paradox," where obesity, as defined by BMI, is simultaneously recognized as an independent stroke risk factor in numerous studies, while other research reported no significant link between obesity and stroke risk, lower mortality rates among individuals with higher BMI [6], and even lower incidence of stroke [7]. Approximately 10–30% of obese individuals are classified as metabolically healthy, contributing to the complexity of the "obesity paradox" [8, 9]. This paradox may account for the varied body composition among individuals with similar BMI levels [10], implying that individuals with the same BMI could have markedly different stroke risks due to the discrepancy in fat distribution and metabolism. In recent years, the Triglyceride-Glucose (TyG) index and its derivatives (including the TyG-BMI and the TyG-wasit circumference,TyG-WC), have emerged as superior to traditional obesity metrics. These measures more precisely unravel visceral fat levels and offer a more accurate stroke risk prediction [11]. However, the predictive validity of these new metrics for stroke in community-dwelling populations, especially in rural areas, remains unexplored, despite the rapid increase in stroke prevalence in rural China over the past decade, which surpassed urban rates in 2018 [12]. The body round index (BRI) combines height and waist circumference has been proven to better reflect visceral fat proportions compared to traditional indices like BMI, WC, and hip circumference [13]. Abdominal visceral adipose tissue was more strongly associated with adverse cardiometabolic risk factors than subcutaneous adipose tissue [14]. Nonetheless, no studies have yet explored the relationship between the TyG index and BRI, or the fat mass index(FMI) [15], which measures visceral obesity and fat quantity, on the incidence of stroke. TyG and other obesity-related indicators are established risk factors for diabetes, lipid health, heart diseases, and hypertension [16, 17], all of which are crucial stroke risk factors [18]. Among stroke participants, hypertension, dyslipidemia, heart disease and diabetes were highly prevalent, yet most of these conditions remained uncontrolled in rural northeast of China [19]. These chronic diseases may mediate stroke development, but research on their mediating roles is extremely limited, especially in resource-poor rural settings. Previous studies have shown that blood pressure, diabetes, and lipid levels mediate the relationship between BMI, testosterone, environmental noise exposure, and stroke incidence [20–23], while these studies only considered the mediating effects of individual chronic disease. Globally, about one-third of adults including a substantial proportion in low and middle income countries [24], and more than half of adults with any chronic condition, experience multimorbidity [25]. The impact of multiple simultaneous conditions on individuals is not merely additive but often interacts in complex ways [26]. However, the extent to which multimorbidity (diabetes, lipid disorders, heart diseases, and hypertension) mediates the development of stroke remains unclear. Collectively, the present study aims to compare the association of BMI and TyG indicators with stroke incidence across subtypes among rural residents in China and the UK, by combining TyG with measures of obesity (BMI), fat distribution (BRI and WC), and body composition (FMI), and to investigate the mediating role of multimorbidity in this relationship and assess the predictive performance of these indicators. Methods Study design This study utilized data from two cohorts: the Environment and Chronic Disease in Rural Areas of Heilongjiang, China (ECDRAHC) [27] study and the UK Biobank [28]. The ECDRAHC study, was carried out from November 2018 to September 2019 and included over 9,000 community participants aged 35 to 74 from nine regions within Mingshui County, Heilongjiang Province. The UK Biobank is a population-based cohort study that enrolled over 500,000 participants aged 39 to 74 years from 22 assessment centers throughout the UK. Recruitment for the UK Biobank took place between 2006 and 2022. Both studies obtained written informed consent from all participants and adhered to the principles of the Declaration of Helsinki [29]. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. Study population The inclusion criteria for the study population were specified as individuals who have complete records for height, weight, WC, body fat mass, triglycerides (TG), and glucose (GLU). Exclusion criteria included subjects under-weight (BMI < 18.5), and those with a prior diagnosis of stroke, cancer, transient cerebral ischemia (TIA), individuals with implausible or extreme anthropometric measurements (including WC, weight, height, and body fat mass, for example, cases where height is less than 50 cm, weight exceeds 200 kg, or body fat mass exceeds total weight), and individuals who were pregnant at baseline (Supplementary Fig. 1). Measurement and definition of BMI/TyG-obesity indicators FMI is calculated as the ratio of body fat mass to the square of height (m2) [15]. BMI was calculated as the ratio between weight (kg) and the height (m2). BRI is calculated as follows [13]: 364·2–365·5* (height/2)2-(WC/π/2)2height/2. The TyG index is calculated as ln [ TG (mg/dL) × GLU (mg/dL) / 2] [30]. The TyG index parameters were defined as previously described [31]. TyG-BMI index = TyG index × BMI, TyG-WC index = TyG index × WC, TyG-FMI index = TyG index × FMI; TyG-RBI index = TyG index* RBI. Obesity defined according to BMI standards, for Chinese people, the categories were: underweight (< 18.5), healthy weight (18.5–24), overweight [24–28], and obesity (≥ 28) [32]: For European people, the categories were: underweight (< 18.5), healthy weight (18.5–25), overweight [25–30], and obesity (≥ 30) [33]. Main outcome measures Electronic medical records from two hospitals in the Mingshui community, covering nearly all stroke cases in Mingshui County, were utilized. Records from 2016 to 2018, along with participants' baseline stroke status, confirmed stroke histories. Follow-up continued until a stroke was identified or until November 2023, whichever came first. For rural UK participants, data on fatal and nonfatal stroke incidents were obtained through national electronic health databases, including the Hospital Episode Statistics (HES) and mortality data from the Office for National Statistics. Follow-up started at the study's inception and continued until a stroke was identified or until December 2022, whichever came first. Stroke events were classified according to the International Classification of Diseases, 10th Revision (ICD-10) and 9th Revision (ICD-9) codes. The classifications included subarachnoid hemorrhages (ICD-10 I60; ICD-9 430), intracerebral hemorrhages (ICD-10 I61; ICD-9 431) and ischemic stroke (ICD-10 I63, ICD-9 434). Covariates and mediators Several potential confounding variables were collected through questionnaires, which included age, sex, smoking status (classified into current smoker, former smoker, and non-smoker), drinking status (divided into current drinker, former drinker, and non-drinker), the frequency of leisure exercise (categorized as had regular exercise and never have regular exercise), family history of CVD and family history of stroke. Hypertension was defined as having a diastolic BP of ≥ 90 mmHg and/or a systolic BP of ≥ 140 mmHg, or the self-reported use of antihypertensive medications. Type 2 diabetes was identified by a fasting plasma glucose level exceeding 126 mg/dL and/or the current use of anti-diabetic medications in ECDRAHC. Participants in UK who had self-reported or hospital diagnoses of type 2 diabetes were identified as diagnosed patients, and who had random glucose level ≥ 11.1 mmol/L or glycated hemoglobin (HbA1c) level ≥ 48 mmol/mol [34]. A family history of stroke was noted if a parent, sibling, or child had experienced a stroke. TC/HDL-C was calculated based on conventional lipids in clinical practice [35], which verified superior than traditional lipid profiles when predicting incidence of stroke [36]. In the ECDRAHC cohort, heart diseases were defined as participants with self-reported coronary heart disease (CHD) or rheumatic heart disease at baseline. In the UK Biobank cohort, heart conditions included self-reported heart attack, myocardial infarction, atrial fibrillation (AF), or heart failure, along with angina or heart attack diagnosed by a physician. According to the definition of multimorbidity [26], the participants were grouped based on the number of chronic conditions they had: hypertension, diabetes, TC/HDL ratio, and heart disease. Due to the lack of an established health threshold for the TC/HDL ratio, it was divided into two categories using the median values (ECDRAHC: 3.65, UK Biobank: 3.965). Participants were then categorized into three groups: healthy (no chronic conditions), one chronic condition, and multimorbidity (two or more chronic conditions). Statistical analysis Multiple imputation method (mice package) was employed to address the missing values, and diagnostic plots subsequently utilized to evaluate the imputation quality. Details on the missing data are provided in Supplementary Table 1. For non-normally distributed continuous covariates, medians and interquartile ranges were used. Categorical covariates were described by frequencies and percentages. Baseline characteristics were categorized by TyG index quartiles and analyzed using the Kruskal–Wallis rank sum test for continuous variables and the χ2 test for categorical variables. Associations between BMI and TyG index parameters were evaluated with Spearman's correlation coefficients. Only baseline variables were included in the analysis. To investigate the association between BMI, TyG index-related parameters, and stroke incidence, Cox proportional hazards regression models were used. For the ECDRAHC cohort, Cox proportional hazards regression models were used, and the proportionality of hazards was verified for all models. For the UK Biobank cohort, we employed a competing risks Cox regression model using the Fine and Gray method, to account for the possibility of non-stroke-related deaths as competing events. Variables potentially impacting stroke incidence or TyG index-related parameters, as identified from existing literature on causal stroke risk factors, were included [18], and incorporating these into directed acyclic graph (DAG) [37, 38] (Supplementary Fig. 2). We constructed a crude univariable model and multivariable models for further analysis. The basic multivariate models were adjusted for potential confounders, which included demographic factors (sex and age treated as a continuous variable), family history of CVD and stroke, and behavioral factors (smoking status, alcohol consumption, and frequency of physical activity). The UK Biobank models were further adjusted for glycated hemoglobin (HbA1c) levels. We sequentially adjusted for potential mediators identified through DAGs, namely heart diseases, hypertension, diabetes and TC/HDL to identify independent, direct associations. In the Cox statistical analysis, BMI and TyG index-related parameters were considered in terms of quartiles, with comparisons made between each quartile and the first quartile, and the impact of each one standard deviation (SD) increase on the risk of stroke. To evaluate the model's performance, receiver operating characteristic (ROC) curves were generated for the cox regression model of stroke, considering an area under the curve (AUC) greater than 0.6 as indicative of higher accuracy [39]. We used restricted cubic splines (RCS) curves with 4 knots to establish the non-linear relationships of per-1-SD increases of TyG indices with incidence of stroke. Kaplan–Meier analysis was used to plot cumulative incidence rate of stroke. Moreover, a random-effects model was used to pool the data for all outcomes from the 2 cohorts. A multiple mediation was analyzed by using mma package in R [40] to explore how TyG index-related parameters impact stroke incidence, it allows for the simultaneous consideration of multiple mediators of various types. These mediators including heart disease, hypertension, diabetes and TC/HDL, as identified through DAGs and Cox regression analyses. To minimize reverse causality and selection bias, participants with less than 2 years of follow-up were excluded from all Cox models. Sensitivity analyses were conducted using data without imputation for missing values, and subgroup analyses were stratified by gender and number of chronic diseases. All p-values are two-tailed, with significance set at p < 0.05. Analyses were performed using R version 4.4.0. For more detailed methods, refer to the supplementary materials. Results Descriptive statistics This study includes 60,685 participants: 6980 (11.50%) from the ECDRAHC cohort and 53,705 (88.50%) from the UK Biobank cohort. For the ECDRAHC cohort, during a median follow-up of 4.33 years (IQR: 4.22–4.97), total of 29 hemorrhagic strokes and 481 ischemic strokes were observed. After excluding individuals diagnosed within the first two years, the final cohort included 6638 subjects, with 237 ischemic and 21 hemorrhagic stroke events) (Table 1). For the UK Biobank cohort, during a median follow-up of 13.85 years (IQR: 13.29–14.71), we initially observed 322 hemorrhagic and 805 ischemic strokes. After excluding individuals diagnosed within the first two years, the final cohort included 53,631 subjects, with 742 ischemic and 316 hemorrhagic strokes (Table 2). Table 1General characteristics of study subjects (ECDRAHC) grouped by quartile of TyG indexCharacteristicOverall, N = 6,890Participants, No. (%) quartile of baseline TyG-indexp-value1Q1[6.72,8.3], N = 1,685Q2(8.3,8.67], N = 1,700Q3(8.67,9.06], N = 1,747Q4(9.06,11.7], N = 1,758Age, Median (IQR)57.00 (51.00, 64.00)56.00 (50.00, 64.00)57.00 (51.00, 64.00)58.00 (52.00, 64.00)58.00 (52.00, 64.00) < 0.001Gender, n (%) < 0.001 Male2,660.0 (38.6%)784.0 (46.5%)626.0 (36.8%)594.0 (34.0%)656.0 (37.3%) Female4,230.0 (61.4%)901.0 (53.5%)1,074.0 (63.2%)1,153.0 (66.0%)1,102.0 (62.7%)Exercise frequency, n (%)0.011 No5,071.0 (73.6%)1,285.0 (76.3%)1,258.0 (74.0%)1,272.0 (72.8%)1,256.0 (71.4%) Yes1,819.0 (26.4%)400.0 (23.7%)442.0 (26.0%)475.0 (27.2%)502.0 (28.6%)Drinking status, n (%) < 0.001 Never or almost never4,830.0 (70.1%)1,123.0 (66.6%)1,233.0 (72.5%)1,254.0 (71.8%)1,220.0 (69.4%) Only occasionally1,007.0 (14.6%)304.0 (18.0%)217.0 (12.8%)238.0 (13.6%)248.0 (14.1%) Almost at least once a week1,053.0 (15.3%)258.0 (15.3%)250.0 (14.7%)255.0 (14.6%)290.0 (16.5%)Smoking status, n (%)0.756 Ever smoker318.0 (4.6%)76.0 (4.5%)74.0 (4.4%)79.0 (4.5%)89.0 (5.1%) Never smoker3,875.0 (56.2%)974.0 (57.8%)946.0 (55.6%)979.0 (56.0%)976.0 (55.5%) Smoker2,697.0 (39.1%)635.0 (37.7%)680.0 (40.0%)689.0 (39.4%)693.0 (39.4%)Family history of stroke, n (%)0.633 No5,543.0 (80.4%)1,340.0 (79.5%)1,382.0 (81.3%)1,408.0 (80.6%)1,413.0 (80.4%) Yes1,347.0 (19.6%)345.0 (20.5%)318.0 (18.7%)339.0 (19.4%)345.0 (19.6%)Family history of CVD, n (%)2,111.0 (30.6%)469.0 (27.8%)519.0 (30.5%)544.0 (31.1%)579.0 (32.9%)0.013BMI, Median (IQR)24.59 (22.35, 27.02)23.16 (21.23, 25.33)24.21 (22.07, 26.53)25.11 (22.99, 27.52)25.96 (23.79, 28.28) < 0.001FMI, Median (IQR)7.25 (5.67, 8.86)6.09 (4.76, 7.58)6.98 (5.53, 8.60)7.72 (6.15, 9.16)8.15 (6.71, 9.62) < 0.001BRI, Median (IQR)4.43 (3.66, 5.28)3.87 (3.24, 4.60)4.33 (3.60, 5.08)4.61 (3.86, 5.43)4.91 (4.18, 5.80) < 0.001WC (cm), Median (IQR)88.00 (82.00, 94.00)84.00 (79.00, 89.00)87.00 (81.00, 93.00)89.00 (83.00, 95.00)91.00 (86.00, 97.00) < 0.001Obesity, n (%) < 0.001 Normal weight2,175.0 (31.6%)810.0 (48.1%)608.0 (35.8%)438.0 (25.1%)319.0 (18.1%) Obesity1,449.0 (21.0%)152.0 (9.0%)276.0 (16.2%)440.0 (25.2%)581.0 (33.0%) Over weight3,266.0 (47.4%)723.0 (42.9%)816.0 (48.0%)869.0 (49.7%)858.0 (48.8%)Hypertension, n (%) < 0.001 No3,185.0 (46.2%)977.0 (58.0%)833.0 (49.0%)760.0 (43.5%)615.0 (35.0%) Yes3,705.0 (53.8%)708.0 (42.0%)867.0 (51.0%)987.0 (56.5%)1,143.0 (65.0%)Diabetes, n (%) < 0.001 No6,494.0 (94.3%)1,660.0 (98.5%)1,666.0 (98.0%)1,681.0 (96.2%)1,487.0 (84.6%) Yes396.0 (5.7%)25.0 (1.5%)34.0 (2.0%)66.0 (3.8%)271.0 (15.4%)Heart diseases, n (%)0.046 No5,965.0 (86.6%)1,488.0 (88.3%)1,477.0 (86.9%)1,503.0 (86.0%)1,497.0 (85.2%) Yes925.0 (13.4%)197.0 (11.7%)223.0 (13.1%)244.0 (14.0%)261.0 (14.8%)Multimorbidity, n (%) < 0.001 Healthy1,569.0 (22.8%)721.0 (42.8%)460.0 (27.1%)290.0 (16.6%)98.0 (5.6%) Multimorbidity2,625.0 (38.1%)231.0 (13.7%)485.0 (28.5%)744.0 (42.6%)1,165.0 (66.3%) One chronic disease2,696.0 (39.1%)733.0 (43.5%)755.0 (44.4%)713.0 (40.8%)495.0 (28.2%)GLU(mmol/L), Median (IQR)4.60 (4.10, 5.30)4.20 (3.80, 4.80)4.50 (4.10, 5.00)4.70 (4.20, 5.30)5.30 (4.60, 6.50) < 0.001TG(mmol/L), Median (IQR)1.60 (1.10, 2.20)0.90 (0.80, 1.00)1.40 (1.20, 1.50)1.90 (1.60, 2.10)2.80 (2.30, 3.50) < 0.001HDL-C(mmol/L), Median (IQR)1.50 (1.30, 1.80)1.60 (1.40, 1.90)1.60 (1.40, 1.80)1.50 (1.30, 1.70)1.40 (1.20, 1.70) < 0.001TC(mmol/L), Median (IQR)5.50 (4.80, 6.30)5.10 (4.50, 5.70)5.50 (4.80, 6.20)5.70 (5.05, 6.50)6.00 (5.10, 6.90) < 0.001TyG-index, Median (IQR)8.68 (8.31, 9.08)8.06 (7.86, 8.19)8.50 (8.40, 8.59)8.86 (8.76, 8.96)9.38 (9.20, 9.70) < 0.001TyG BMI index, Median (IQR)214.21 (189.75, 241.13)184.63 (168.58, 203.19)205.65 (187.49, 225.95)221.86 (203.21, 244.29)246.47 (224.58, 269.67) < 0.001TyG WC index, Median (IQR)762.65 (692.27, 841.34)672.49 (626.83, 718.08)738.65 (685.25, 789.93)787.75 (735.76, 844.39)869.44 (807.82, 930.96) < 0.001TyG FMI index, Median (IQR)63.41 (48.55, 78.82)48.70 (37.80, 60.93)59.26 (46.90, 73.11)68.50 (54.48, 81.26)77.22 (63.66, 92.42) < 0.001TyG MMI index, Median (IQR)0.55 (0.50, 0.59)0.51 (0.47, 0.55)0.54 (0.50, 0.58)0.56 (0.52, 0.60)0.59 (0.54, 0.63) < 0.001TyG BRI index, Median (IQR)38.65 (31.14, 47.06)30.92 (25.75, 36.78)36.70 (30.45, 43.29)40.69 (34.26, 48.11)46.85 (39.56, 55.26) < 0.0011Kruskal-Wallis rank sum test; Pearson’s Chi-squared test General characteristics of study subjects (ECDRAHC) grouped by quartile of TyG index 1Kruskal-Wallis rank sum test; Pearson’s Chi-squared test Table 2General characteristics of study subjects (UK Biobank) grouped by quartile of TyG indexCharacteristicOverall, N = 53,705UKB Participants, No. (%) quartile of baseline TyG-indexp-value1Q1, N = 13,427Q2, N = 13,425Q3, N = 13,426Q4, N = 13,427Gender, n (%) < 0.001 Female28,947.0 (53.9%)8,337.0 (62.1%)7,761.0 (57.8%)6,990.0 (52.1%)5,859.0 (43.6%) Male24,758.0 (46.1%)5,090.0 (37.9%)5,664.0 (42.2%)6,436.0 (47.9%)7,568.0 (56.4%)Age, Median (IQR)58.00 (51.00, 63.00)55.00 (48.00, 62.00)59.00 (52.00, 64.00)60.00 (53.00, 64.00)59.00 (53.00, 63.00) < 0.001Ethnic, n (%)0.019 Other514.0 (1.0%)137.0 (1.0%)148.0 (1.1%)129.0 (1.0%)100.0 (0.7%) White53,191.0 (99.0%)13,290.0 (99.0%)13,277.0 (98.9%)13,297.0 (99.0%)13,327.0 (99.3%)WC (cm), Median (IQR)89.00 (80.00, 98.00)82.00 (74.00, 92.00)87.00 (78.00, 96.00)91.00 (82.00, 99.00)95.00 (87.00, 102.00) < 0.001Obesity, n (%) < 0.001 Normal weight18,645.0 (34.7%)6,991.0 (52.1%)5,312.0 (39.6%)3,857.0 (28.7%)2,485.0 (18.5%) Obesity11,314.0 (21.1%)1,581.0 (11.8%)2,345.0 (17.5%)3,198.0 (23.8%)4,190.0 (31.2%) Over weight23,746.0 (44.2%)4,855.0 (36.2%)5,768.0 (43.0%)6,371.0 (47.5%)6,752.0 (50.3%)Drinking status, n (%) < 0.001 Almost at least once a week40,820.0 (76.0%)10,407.0 (77.5%)10,206.0 (76.0%)10,154.0 (75.6%)10,053.0 (74.9%) Never or almost never2,798.0 (5.2%)622.0 (4.6%)698.0 (5.2%)747.0 (5.6%)731.0 (5.4%) Only occasionally10,087.0 (18.8%)2,398.0 (17.9%)2,521.0 (18.8%)2,525.0 (18.8%)2,643.0 (19.7%)BMI, Median (IQR)26.48 (24.05, 29.38)24.84 (22.69, 27.44)25.96 (23.66, 28.78)26.95 (24.67, 29.83)28.05 (25.77, 30.82) < 0.001RBI, Median (IQR)3.81 (2.96, 4.78)3.13 (2.43, 4.04)3.60 (2.82, 4.54)4.00 (3.20, 4.93)4.39 (3.65, 5.30) < 0.001FMI, Median (IQR)7.98 (6.21, 10.25)7.16 (5.54, 9.11)7.85 (6.09, 10.05)8.35 (6.49, 10.74)8.65 (6.85, 11.03) < 0.001Smoking status, n (%) < 0.001 Yes22,147.0 (41.2%)5,822.0 (43.4%)5,627.0 (41.9%)5,533.0 (41.2%)5,165.0 (38.5%) No31,558.0 (58.8%)7,605.0 (56.6%)7,798.0 (58.1%)7,893.0 (58.8%)8,262.0 (61.5%)Exercise frequency, n (%) < 0.001 Low9,570.0 (17.8%)2,066.0 (15.4%)2,253.0 (16.8%)2,458.0 (18.3%)2,793.0 (20.8%) Moderate21,364.0 (39.8%)5,219.0 (38.9%)5,276.0 (39.3%)5,478.0 (40.8%)5,391.0 (40.2%) High22,771.0 (42.4%)6,142.0 (45.7%)5,896.0 (43.9%)5,490.0 (40.9%)5,243.0 (39.0%)Family history of stroke, n (%) < 0.001 No38,805.0 (72.3%)9,927.0 (73.9%)9,691.0 (72.2%)9,622.0 (71.7%)9,565.0 (71.2%) Yes14,900.0 (27.7%)3,500.0 (26.1%)3,734.0 (27.8%)3,804.0 (28.3%)3,862.0 (28.8%)Family history of CVD, n (%) < 0.001 No29,592.0 (55.1%)7,707.0 (57.4%)7,323.0 (54.5%)7,316.0 (54.5%)7,246.0 (54.0%) Yes24,113.0 (44.9%)5,720.0 (42.6%)6,102.0 (45.5%)6,110.0 (45.5%)6,181.0 (46.0%)Diabetes, n (%) < 0.001 No51,630.0 (96.1%)12,830.0 (95.6%)12,938.0 (96.4%)12,917.0 (96.2%)12,945.0 (96.4%) Yes2,075.0 (3.9%)597.0 (4.4%)487.0 (3.6%)509.0 (3.8%)482.0 (3.6%)Hypertension, n (%) < 0.001 No25,453.0 (47.4%)7,980.0 (59.4%)6,578.0 (49.0%)5,806.0 (43.2%)5,089.0 (37.9%) Yes28,252.0 (52.6%)5,447.0 (40.6%)6,847.0 (51.0%)7,620.0 (56.8%)8,338.0 (62.1%)Heart diseases, n (%) < 0.001 No51,746.0 (96.4%)12,754.0 (95.0%)12,922.0 (96.3%)12,969.0 (96.6%)13,101.0 (97.6%) Yes1,959.0 (3.6%)673.0 (5.0%)503.0 (3.7%)457.0 (3.4%)326.0 (2.4%)Multimorbidity, n (%) < 0.001 Healthy13,284.0 (24.7%)6,625.0 (49.3%)4,114.0 (30.6%)2,050.0 (15.3%)495.0 (3.7%) Multimorbidity17,351.0 (32.3%)1,525.0 (11.4%)2,986.0 (22.2%)5,200.0 (38.7%)7,640.0 (56.9%) One chronic disease23,070.0 (43.0%)5,277.0 (39.3%)6,325.0 (47.1%)6,176.0 (46.0%)5,292.0 (39.4%)GLU(mmol/L), Median (IQR)5.74 (4.98, 6.50)4.96 (4.31, 5.55)5.59 (4.96, 6.21)5.97 (5.31, 6.61)6.53 (5.81, 7.25) < 0.001HDL-C(mmol/L), Median (IQR)1.43 (1.20, 1.71)1.57 (1.33, 1.86)1.51 (1.27, 1.80)1.40 (1.19, 1.65)1.27 (1.09, 1.48) < 0.001TC(mmol/L), Median (IQR)5.77 (5.03, 6.52)5.03 (4.44, 5.62)5.62 (5.00, 6.23)5.98 (5.34, 6.62)6.53 (5.82, 7.25) < 0.001TG(mmol/L), Median (IQR)1.48 (1.04, 2.13)0.85 (0.72, 0.99)1.24 (1.09, 1.42)1.72 (1.51, 1.97)2.73 (2.28, 3.41) < 0.001TC/HDL, Median (IQR)3.97 (3.30, 4.79)3.16 (2.77, 3.61)3.68 (3.23, 4.20)4.22 (3.71, 4.81)5.12 (4.46, 5.83) < 0.001TyG-index, Median (IQR)8.81 (8.40, 9.22)8.14 (7.94, 8.28)8.61 (8.51, 8.71)9.00 (8.90, 9.11)9.52 (9.35, 9.75) < 0.001TyG BMI index, Median (IQR)234.17 (206.65, 266.15)200.98 (182.47, 222.43)223.66 (203.53, 248.23)242.60 (221.85, 269.03)269.51 (245.37, 297.29) < 0.001TyG FMI index, Median (IQR)70.46 (54.03, 91.64)57.64 (44.53, 73.90)67.63 (52.39, 86.53)75.18 (58.45, 96.58)82.93 (65.65, 105.97) < 0.001TyG MMI index, Median (IQR)0.48 (0.43, 0.53)0.46 (0.41, 0.51)0.48 (0.43, 0.53)0.49 (0.44, 0.54)0.49 (0.45, 0.55) < 0.001TyG WC index, Median (IQR)785.20 (685.51, 885.49)663.93 (598.27, 744.96)747.95 (672.42, 826.60)816.50 (741.37, 891.71)906.99 (830.06, 985.37) < 0.001TyG BRI index, Median (IQR)33.71 (25.48, 43.07)25.25 (19.55, 32.77)31.05 (24.25, 39.13)35.98 (28.91, 44.31)42.21 (34.87, 51.04) < 0.0011Pearson’s Chi-squared test; Kruskal–Wallis rank sum test General characteristics of study subjects (UK Biobank) grouped by quartile of TyG index 1Pearson’s Chi-squared test; Kruskal–Wallis rank sum test The two cohorts had similar median ages (57.00 vs. 58.00 years), WC (88.00 cm vs. 89.00 cm) and both had a higher proportion of females (ECDRAHC: 61.40% vs 38.60%.; UKB: 53.90% vs 46.10%.). While, the UK Biobank cohort had a higher prevalence of stroke family history (19.60% vs.27.70%), higher BMI (24.59 vs.26.48), higher prevalence of CVD family history (30.60% vs. 44.90%), higher TyG index levels (8.68 vs.8.81), and higher FMI (7.25 vs.7.98). Conversely, the UK Biobank cohort had a lower prevalence of diabetes (5.70% vs.3.90%), lower rates of hypertension (53.80% vs.52.60%), lower prevalence of heart disease (13.40% vs.3.60%), lower BRI (4.43 vs. 3.81), lower prevalence of multimorbidity (38.10% vs. 32.30%) and lower age-adjusted incidence rate of stroke (401.89/100,000 person-years vs. 41.91/100,000 person-years). In both cohorts, when stratified by TyG index quartiles, the highest TyG index quartile exhibited older age, a higher proportion of males, and increased rates of obesity, hypertension, diabetes, TC and lower HDL level compared to the lowest quartile. The detailed information is presented in Table 1 and Table 2. Correlation analysis between BMI and TyG series indices In the ECDRAHC cohort, correlation coefficients (0.28–0.92) are generally lower than those in the UK Biobank cohort (0.28–0.94). All correlations are statistically significant (p < 0.05). In the ECDRAHC cohort, BMI shows the strongest correlation with TyG-BMI (0.92), followed by TyG-FMI (0.83), and the weakest with the TyG index (0.28) (Supplementary Fig. 3a). In the UK Biobank cohort, BMI also has the strongest correlation with TyG-BMI (0.94), followed by TyG-BRI (0.87), and the weakest with the TyG index (0.28) (Supplementary Fig. 3b). Associations of BMI and TyG-related indices with incidence of stroke risk When basic covariates were included, BMI was significantly associated with ischemic stroke risk in the UKB cohort but not in the ECDRAHC cohort. TyG-related parameters were associated with the occurrence of ischemic stroke in both cohorts, except for the TyG index in the UKB cohort. Most indicators were not significantly associated with hemorrhagic stroke incidence in either cohort, except for TyG-FMI and TyG-BRI, which were negatively associated with hemorrhagic stroke incidence in the UK (Fig. 1). Association of BMI and TyG indices with stroke risk. ECDRAHC(a1-a3) adjusted for demographic factors (sex and age treated as a continuous variable), family history of CVD and stroke, and behavioral factors (smoking status, alcohol consumption, and frequency of physical activity). Additionally, the UK Biobank database(b1-b3) adjustments incorporated glycated hemoglobin (HbA1c) levels Potential mediators were included in Cox regression models, in both cohorts, the positive correlations between TyG index parameters and ischemic stroke incidence gradually decreased (Fig. 2), a trend not seen for hemorrhagic stroke (Supplementary Fig. 4). In the ECDRAHC cohort, the RCS curves indicated nearly linear trends for TyG indices (P overall > 0.05, P non-linear > 0.05). In the UK Biobank cohort, there was a nearly linear increase in ischemic stroke incidence with higher BMI and TyG indices (except the TyG Index) (P overall < 0.05, P non-linear > 0.05) Supplementary Fig. 5. Sequential adjustment for potential mediating effects for ischemic stroke in two cohorts and combined meta-analysis. ECDRAHC: analyses adjusted for age, gender, physical exercise frequency, smoking status, drinking status, family history of stroke. family history of CVD, TC/HDL, diagnosed with heart disease, diabetes, and hypertension, in UKB cohort additionally adjusted for HbA1c. All exposures are presented as per 1-SD increase. The standard deviation values for each exposure in UKB are as follows: BMI: 4.38, TyG-BMI: 45.95, TyG-WC: 140.94, TyG-FMI: 31.24, TyG-BRI: 13.87. The standard deviation values for each exposure in ECDRAHC are as follows: BMI: 3.62, TyG-BMI: 38.87, TyG-WC: 110.24, TyG-FMI: 23.26, TyG-BRI: 12.11 In fully adjusted models, the TyG-BRI index in the ECDRAHC cohort showed marginal significance for ischemic stroke incidence (HR 1.13, 95% CI: 0.99–1.30, p = 0.069). In the UK Biobank cohort, BMI and TyG-obesity parameters remained significant for ischemic stroke, with HRs from 1.13 to 1.15. (Fig. 2). For hemorrhagic stroke, the TyG index in the ECDRAHC cohort showed an increased risk in Q4 vs. Q1 (HR: 4.62, 95% CI: 1.04–20.53) and per 1-SD increase (HR: 1.55, 95% CI: 1.01–2.40), though pooled effects were not significant (Supplementary Table 2). In the UK Biobank cohort, TyG-FMI and TyG-BRI indices had a significant protective effect against hemorrhagic stroke, with consistent pooled results. Dose–response analysis revealed a significant non-linear relationship between the TyG-BRI index and hemorrhagic stroke (p overall = 0.041, p non-linear = 0.017), with lower risk between 2.4-SD and 4.2-SD (approximately 29.12–66.59) (Fig. 3). Non-linear analysis of TyG-BRI and TyG-FMI in relation to hemorrhagic stroke risk in UKB cohort. adjusted for age, gender, physical exercise frequency, smoking status, drinking status, family history of stroke. family history of CVD, TC/HDL, diagnosed with heart disease, diabetes, and hypertension and HbA1c Mediating effects of multimorbidity on the relationship between TyG index parameters and ischemic stroke incidence In the ECDRAHC cohort, hypertension mediated 8.4% and the combined effects of multiple health conditions mediated 24.7% of the relationship between TyG-BRI and ischemic stroke incidence. In the UK Biobank cohort, the mediation effects were: hypertension 4.7%, TC/HDL 17.0%, heart disease 3.0%, and combined effects 25.9%. Generally, the combined mediation effects of multimorbidity provided the highest contribution to ischemic stroke incidence across different TyG-related indices in both populations, exceeding the sum of individual effects from separate chronic diseases. The combined mediation percentages of multimorbidity were higher in the ECDRAHC cohort compared to the UK Biobank cohort. These contributions ranged from 24.7% to 51.0% in the ECDRAHC cohort and from 25.1% to 25.9% in the UK Biobank cohort. Significant combined mediation effects of multimorbidity in the UK Biobank cohort were observed only for BMI and TyG-BRI index related to ischemic stroke incidence (Table 3). Table 3Mediating effects of hypertension, diabetes, heart diseases and TC/HDL on the relationship between TyG Indices and ischemic stroke riskOutcome/TyG index-related parametersECDRAHCUKBRelative contribution, % (95% CI)aPRelative contribution, % (95% CI)aPBMI Mediating effects of hypertension//9.0%(5.6%,13.0%) < 0.001 Mediating effects of diabetes///0.101 Mediating effects of TC/HDL///0.160 Mediating effects of heart diseases//8.3%(3.9%,11.7%)0.001 Joint mediation effects of hypertension, diabetes, 0heart diseases and TC/HDL//25.1%(13.5%,35.0%) < 0.001TyG index Mediating effects of hypertension13.3%(4.7%,49.1%)0.006// Mediating effects of diabetes/0.220// Mediating effects of TC/HDL/0.060// Mediating effects of heart diseases/0.100// Joint mediation effects of hypertension, diabetes, 0heart diseases and TC/HDL40.7%(15.6%,111.3%)0.018//TyG-BMI index Mediating effects of hypertension12.1%(5.9%,90.2%)0.0043.8%(2.1%,11.6%)0.040 Mediating effects of diabetes/0.160/0.520 Mediating effects of TC/HDL/0.146/0.080 Mediating effects of heart diseases/0.760-1.8%(-0.4.8%,-0.6%)0.020 Joint mediation effects of hypertension, diabetes, 0heart diseases and TC/HDL38.6%(12.6%,123.8%)0.00727.2%(2.6%,60.9%)0.020TyG-WC index Mediating effects of hypertension11.9%(4.5%,34.5%)0.0205.5%(0.4%,15.5%)0.020 Mediating effects of diabetes/0.360/0.541 Mediating effects of TC/HDL14.6%(1.9%,30.8%) < 0.001/0.080 Mediating effects of heart diseases/0.140/0.660 Joint mediation effects of hypertension, diabetes, 0heart diseases and TC/HDL30.1%(6.7%,57.1%) < 0.001/0.060TyG-FMI index Mediating effects of hypertension17.5%(7.6%,69.3%)0.004/0.760 Mediating effects of diabetes/0.240/0.500 Mediating effects of TC/HDL19.2%(4.2%,83.7%)0.0202.4%(0.2%,6.3%)0.020 Mediating effects of heart diseases/0.060-2.2%(-8.8%,-0.8%)0.020 Joint mediation effects of hypertension, diabetes, 0heart diseases and TC/HDL51.0%(23.9%,136.2%)0.020/0.360TyG-BRI index Mediating effects of hypertension8.4%(3.8%,16.4%) < 0.0014.7%(1.1%,11.4%)0.040 Mediating effects of diabetes/0.260/0.660 Mediating effects of TC/HDL/0.06017.0%(6.4%,33.2%)0.020 Mediating effects of heart diseases/0.1203.0%(0.7%,3.2%)0.020 Joint mediation effects of hypertension, diabetes, 0heart diseases and TC/HDL24.7%(11.9%,42.7%) < 0.00125.9%(10.6%,48.4%)0.001aPercentage of the BMI and TyG indices difference in stroke incidence risk explained by chronic diseases in the multivariable mediation model Mediating effects of hypertension, diabetes, heart diseases and TC/HDL on the relationship between TyG Indices and ischemic stroke risk aPercentage of the BMI and TyG indices difference in stroke incidence risk explained by chronic diseases in the multivariable mediation model Evaluation of predictive performance of TyG index parameters In the UK Biobank cohort, AUC values ranged from 0.732 to 0.736, with BMI (SD) showing the highest AUC (0.736), followed by indices TyG RBI index (SD),BMI (quartile) and TyG-FMI (SD). Sensitivity values ranged from 0.712 to 0.751, with TyG-BMI (SD) achieving the highest sensitivity (0.751), indicating better detection of stroke risk. Specificity was moderate, ranging from 0.611 to 0.647, with TyG-BRI (SD) showing the highest value (0.647) (Supplementary Fig. 6 b1-b2). In ECDRAHC cohort, AUC ranging between 0.708 and 0.716, with the highest observed for TyG-BRI (quartile). Sensitivity was notably higher, ranging from 0.762 to 0.814, with BMI (SD)(0.814) showing the best performance, followed by TyG-FMI (0.810), TyG-BMI (0.805) and TyG-BRI (0.801). Specificity ranging from 0.533 to 0.583, with TyG-BRI (quartile) (0.583) showing the highest value (Supplementary Fig. 6 a1-a2). Ischemic stroke incidence and HRs according to the prevalence of multimorbidity K-M curves showed a significant association between the cumulative number of chronic diseases and the incidence rate of ischemic stroke in both cohorts (log-rank test p < 0.001). After adjusting for confounders in the Cox regression analysis, a significant increased risk of ischemic stroke was observed only in the ECDRAHC cohort not in UK cohort for individuals with multimorbidity (HR: 1.83, 95% CI: 1.19–2.83) compared to those with no chronic diseases (Supplementary Fig. 7 and Supplementary Fig. 8). Discussion In this multicohort study across two countries, significant differences were found in the predictive value of BMI for ischemic stroke between the two populations. Composite indices combining the TyG index with obesity consistently linked to an increased risk of ischemic stroke in both countries. However, this significant association was mediated by chronic diseases, with multimorbidity playing a primary mediating role, exceeding the combined mediation effect of individual chronic diseases. This highlights the complex interactions between comorbidities in the occurrence of ischemic stroke. Furthermore, the mediating effect of multimorbidity on ischemic stroke incidence was more pronounced in rural China, indicating greater susceptibility among rural Chinese residents and underscoring the need for targeted interventions. Notably, the TyG-BRI index within a suitable range showed a protective effect against hemorrhagic stroke, suggesting distinct mechanisms for ischemic and hemorrhagic strokes. The “obesity paradox” in CVD remains an issue of ongoing debate [41]. Our study found that higher BMI is significantly associated with increased ischemic stroke incidence among UK rural residents, but not among Chinese rural residents. Despite similar age and gender distributions, UK residents had higher BMI and more family histories of CVD, but lower prevalence of diabetes, heart disease, hypertension, prevalence of multimorbidity and incidence of stroke. This disparity may be due to ethnic differences, as Asians have higher visceral fat levels at the same BMI compared to Western populations [42]. Visceral obesity is a primary metabolic and CVD risk predictor in obesity [43]. In a heart cohort study, abdominal visceral adipose tissue was more strongly associated with adverse cardiometabolic risk factors than subcutaneous adipose tissue [14]. BRI used to assess visceral fat levels is more effective than both BMI and Waist-to-Hip Ratio (WHR) in predicting metabolic syndromes [44], which are known risk factors for stroke. Our results further confirms that the TyG-BRI index is superior to other TyG-related indices in predicting ischemic stroke in rural Chinese. Although no direct studies have explored the relationship between TyG-BRI and stroke risk, evidence suggests that sarcopenia combined with high BRI is associated with increased stroke risk [45]. These findings indicate that TyG-BRI may serve as an alternative measure to BMI for assessing ischemic stroke risk in rural Chinese populations. In our study, despite higher BMI levels, UK residents had lower BRI levels than Chinese rural residents. In addition, tertile-based analysis of BRI further confirmed that the positive association of ischemic stroke and BMI was only found in the highest BRI tertile among UK residents. BMI does not differentiate fat distribution, which may explain the inconsistency in BMI-stroke risk across populations. The TyG index was associated with ischemic stroke incidence in rural Chinese residents but not in UK residents. This disparity may be partly due to racial differences and the non-fasting peripheral blood collection in the UK Biobank cohort, which likely influenced blood glucose levels. The positive association of TyG-obesity measures with ischemic stroke attenuated after sequential adjustment for potential mediating factors, aligning with findings from other studies on this attenuated trend [7, 46, 47], and in the fully adjusted models, the significant association between TyG-related indices and ischemic stroke disappeared in rural Chinese residents, with only TyG-BRI approaching significance and showing an approximately linear relationship. In contrast, the association persisted in the UK Biobank cohort. The attenuating trend highlights the mediating roles of hypertension, diabetes, heart diseases, and the TC/HDL ratio in the pathogenesis of ischemic stroke, particularly among rural Chinese populations. As we know, no previous studies have reported the combined mediating effect of multimorbidity on stroke incidence. We found that the mediating percent of multimorbidity on ischemic stroke incidence is greater than the sum of individual chronic diseases. This indicates that the influence of multiple coexisting chronic conditions on stroke risk involves complex interactions rather than a simple additive effect. In practice, multimorbidity can arise from the same cause, one condition can increase the risk of another, or the side effects of one treatment can trigger a completely different disease, necessitating multidisciplinary collaboration and decision making [48]. Unfortunately, most clinical practice guidelines and healthcare organizations focus on managing single diseases [49], leading to fragmented and contradictory care for people with multimorbidity, increasing their treatment burden [50]. Research, clinical guidelines, and healthcare services should shift focus from single conditions to the needs of increasingly complex patient populations. Effective management of multimorbidity requires balancing single-disease focus with comprehensive care. Although evidence on effective management strategies for multimorbidity is still lacking, research suggests that patient-centered management, involving care coordination, self-management, and medication management, can improve health outcomes [25]. Additionally, we found a higher proportion of mediation by multimorbidity in rural Chinese residents compared to rural UK residents. In both populations, individuals with multimorbidity had a significantly higher cumulative risk of ischemic stroke compared to healthy individuals. However, Cox regression models revealed a significant increase in ischemic stroke risk for those with multimorbidity only among rural Chinese residents, not in the UK Biobank cohort. Subgroup analysis also supported this finding. These results suggest that the high incidence of stroke in rural China is likely related to the high prevalence and suboptimal control of multimorbidity. Multimorbidity is a growing global challenge with substantial effects on individuals, carers, and society. While, access to health care is more difficult for people in socioeconomically deprived areas who have higher levels of multimorbidity [51]. Urgent reconfiguration of healthcare is needed to support the management of multimorbidity in northeast rural residents. And, population-level approaches may be needed to overcome economic barriers and address early determinants of multimorbidity, including socioeconomic deprivation and lower education levels [52, 53]. For healthy individuals, lifestyle changes, including weight management, dietary improvements, increased physical activity, and not smoking, are advocated to prevent multimorbidity and mitigate the risk of ischemic stroke [54]. Previous research [55] and our study did not find a significant association between the TyG index and the occurrence of hemorrhagic stroke, and Mendelian randomization studies have demonstrated a significant causal relationship between the TyG index and ischemic stroke but not with hemorrhagic stroke [56]. However, when combining the TyG index with the visceral fat indicator BRI, a significant nonlinear relationship with hemorrhagic stroke was observed in the UK Biobank population. This relationship was not observed among rural Chinese residents, likely due to the small sample size, with only 21 hemorrhagic stroke cases recorded. Additionally, BRI has been shown to be nonlinearly associated with cardiovascular and all-cause mortality [57]. Given the non-fasting blood samples in the UK Biobank population, further studies are needed to validate this finding. Classic risk factors for stroke show differing impacts on the three main types of stroke: ischemic stroke, intracerebral hemorrhage, and subarachnoid hemorrhage [58]. This suggests distinct pathogenic mechanisms for ischemic and hemorrhagic strokes, necessitating separate discussions on their risk factors. Identifying specific predictive factors for each type of stroke is essential for effective population risk screening. So far, the underlying mechanism is not well understood. TyG index is a marker of insulin resistance and metabolic dysfunction [59], implicates lipid and glucose abnormalities that are risk factors for atherosclerosis, thereby increasing ischemic stroke risk. Differences in the regulation of fat metabolism by genes such as UCP1 and UCP12 among various ethnic groups [60, 61] may elucidate the disparities in the association between obesity and stroke risk across different populations. Adipose tissue exerts protective effects by secreting adipokines like adiponectin, which have anti-inflammatory properties and enhance endothelial function [62]. However, excessive visceral fat leads to chronic inflammation and vascular instability [62], increasing hemorrhagic stroke risk. This dual role of adipose tissue may explain the significant non-linear relationship between the TyG-BRI index and hemorrhagic stroke risk observed in our study. Future research should explore these mechanisms to confirm protective thresholds and identify therapeutic targets. In summary, this study highlights the importance of the TyG-BRI index as a significant predictor of ischemic stroke risk, particularly in rural populations, emphasizing the need for tailored public health interventions that address the complexities of multimorbidity and its impact on stroke incidence. In addition, in the clinical practice, more attention should be paid to higher TyG-BRI index individuals. The strengths of this study include data from electronic medical records to accurately distinguish between ischemic and hemorrhagic strokes, providing more reliable classification than self-reported stroke incidents. The large prospective follow-up of participants allows for stronger causal inference compared to cross-sectional studies. Additionally, this study is the first to comprehensively consider the combined mediating effects of multimorbidity on the incidence of different stroke subtypes. Comparative analyses across two countries systematically unravel differences in obesity-related factors among populations with varying economic statuses and ethnic backgrounds. However, the study has several limitations. First, the limited sample size of males, especially hemorrhagic stroke cases in the ECDRAHC cohort, resulted in wide confidence intervals for the TyG index and hemorrhagic stroke risk, preventing multiple mediation analysis in males. Second, the use of the UK Biobank database, with racial differences, may affect the generalizability of TyG index impacts on hemorrhagic stroke risk, necessitating future validation in a large rural Chinese sample. Third, lack of data on lipid-lowering medication use may introduce confounding effects. Fourth, non-fasting glucose levels in the UK Biobank cohort could affect results, though HbA1c adjustments were made. Lastly, differing heart disease definitions between populations likely led to underestimation of heart disease prevalence in Chinese rural residents compared to the UK Biobank data. Conclusion Through comparisons with rural cohorts from UK, revealed that residents in rural China, despite having lower BMI levels, exhibit higher levels of visceral obesity and a higher prevalence of chronic diseases. This mainly contributes to an increased susceptibility to stroke, mediated by multimorbidity in rural China. Multimorbidity accounted for a greater mediation percentage compared to individual chronic diseases. To maintain an optimal range of the TyG-BRI may reduce the incidence risk of hemorrhagic stroke, though this requires further validation in large populations with fasting peripheral blood samples. Supplementary Information Supplementary Material 1. Supplementary Figure 1. Flowchart of participant selection Supplementary Material 2. Supplementary Figure 2. Directed Acyclic Graph. Note: A directed acyclic graph represents associations between covariates and primary exposure and outcome. Pink circles represent ancestors of the exposure and outcome (ie, confounders), blue circles represent ancestors of the outcome (ie, causal determinants of the outcome). Green lines represent causal paths, and pink lines represent biasing paths. The minimally sufficient adjustment set represents covariates such that the adjustment for this set of variables will minimize confounding bias when estimating the association between the exposure and the outcome. The minimally sufficient adjustment set was determined using the DAGitty software3. 1. The final minimally sufficient adjustment sets for estimating the direct effect of TyG index-related parameters on Stroke: Age, Diabetes, Drinking, Family history of CVD, Family history of stroke, Gender, Heart diseases, Hypertension, Leisure exercise frequency, Smoking, TC/HDL in ECDRAHC cohort. In UKB cohort additionally adjusted for HbA1c Supplementary Material 3. Supplementary Figure 3. The correlation of BMI and TyG related parameters in ECDRAHC(a) and UKB (b). Note:***p<0.001 Supplementary Material 4. Supplementary Figure 4. Sequential adjustment for potential mediating effects for hemorrhagic stroke. Note: All exposures are presented as per 1-SD increase. The standard deviation values for each exposure are as follows: BMI: 4.38, TyG-BMI: 45.95, TyG-WC: 140.94, TyG-FMI: 31.24, TyG-BRI: 13.87 Supplementary Material 5. Supplementary Figure 5. Multivariable adjusted dose-response associations between BMI, TyG related parameters and ischemic stroke. Note: ECDRAHC(a1-a5) adjusted for adjusted for age, gender, smoking status, drinking status, family history of CVD, family history of stroke, physical activity, tc/dhl, hypertension, diabetes, heart diseases; UKB(b1-b5) additionally adjusted for HbA1c Supplementary Material 6. Supplementary Figure 6. ROC curves for predicting ischemic stroke. Note: ECDRAHC cohort (a1: per 1-SD increase, a2: quartile increase) UKB cohort (b1: per 1-SD increase, b2: quartile increase) Supplementary Material 7. Supplementary Figure 7. Incidence and HRs of ischemic stroke in ECDRAHC cohort according to the number of chronic diseases. Note: Kaplan-Meier curves show cumulative incidence by number of chronic diseases. Bar chart shows HRs of incidence of ischemic stroke associated with number of chronic diseases, adjusted for adjusted for age, gender, physical activity, family history of CVD, family history of stroke, smoking status, drinking status and BMI Supplementary Material 8. Supplementary Figure 8. Incidence and HRs of ischemic stroke in UK biobank cohort according to the number of chronic diseases. Note: Kaplan-Meier curves show cumulative incidence by number of chronic diseases. Bar chart shows HRs of incidence of ischemic stroke associated with number of chronic diseases, adjusted for adjusted for age, gender, physical activity, family history of CVD, family history of stroke, smoking status, drinking status and TyG-BRI Supplementary Material 9. Supplementary Table 1. Description of missing data in ECDRAHC and UKB cohorts Supplementary Material 10. Supplementary Table 2. Association of BMI and TyG indices with Hemorrhagic Stroke Risk (Cohort Analysis and Meta-Analysis). Note: Note:ECDRAHC:Models ajusted for Age, Drinking, Family history of CVD, Family history of stroke, Gender, Leisure exercise frequency, Smoking, Diabetes, Heart diseases, Hypertension, TC/HDL UKB additionally adjusted for HbA1c Supplementary Material 11. Supplementary Table 3. Relationship between TyG indices and the risk of ischemic stroke (including missing variables) in ECDRAHC and UKB Databases. Note: Mod1: crude model. Mod2: adjusted for basic risk factors (Age, Drinking, Family history of CVD, Family history of stroke, Gender, Leisure exercise frequency, Smoking, UKB additionally adjusted for HbA1c). Mod3: additionally adjusted for mediators (Diabetes, Heart diseases, Hypertension, TC/HDL)1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 12. Supplementary Table 4. Association between BMI and ischemic stroke risk stratified by BRI tertiles in the UKB cohort. Note: adjusted for basic risk factors (Age, Drinking, Family history of CVD, Family history of stroke, Gender, Leisure exercise frequency, Smoking, and HbA1c) Supplementary Material 13. Supplementary Table 5. Association of obesity with the incidence of ischemic stroke in UKB cohorts. Note:models adjusted for age,gender,physical activity,smoking status, drinking status,family history of CVD, family history of stroke,TC/HDL,TG,HbA1c,hypertension, diabetes, heart diseases1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 14. Supplementary Table 6. Association of BMI and TyG index series with the incidence of ischemic stroke in ECDRAHC and UKB cohorts stratified by gender. Note: ECDRAHC: models adjusted for age, smoking status, drinking status, family history of CVD, family history of stroke, physical activity, tc/dhl, hypertension, diabetes, heart diseases . UKB:additionally adjustde for HbA1c based on ECDRAHC 1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 15. Supplementary Table 7. Association of BMI and TyG index parameters with the incidence of hemorrhagic stroke in UKB cohorts stratified by gender. Note:ECDRAHC: models adjusted for age, smoking status, drinking status, family history of CVD, family history of stroke, physical activity, tc/dhl, hypertension, diabetes, heart diseases UKB:additionally adjustde for HbA1c based on ECDRAHC 1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 16. Supplementary Table 8. Mediating effects of hypertension, diabetes, heart diseases and TC/HDL on the relationship between TyG Indices and ischemic stroke risk. Note:*Percentage of the BMI and TyG indices difference in stroke incidence risk explained by chronic diseases in the multivariable mediation model Supplementary Material 17. Supplementary Table 9. Association of BMI and TyG-related indices with ischemic stroke risk stratified by number of chronic diseases: cox regression analysis in UKB and ECDRAHC cohorts. Note: ECDRAHC: adjusted for age, gender, physical activity, family history of CVD, family history of stroke, smoking status, drinking status UKB:additionally adjusted for HbA1c1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 18. Supplementary Table 10. Association of BMI and TyG-related indices with ischemic stroke risk in UKB and ECDRAHC cohorts. Note: ECDRAHC: adjusted for age, smoking status, drinking status, family history of CVD, family history of stroke, physical activity, non-HDL, hypertension, diabetes, heart diseases. UKB:additionally adjusted for HbA1c1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 19. Acknowledgements We gratefully acknowledge the UK Biobank for providing the data used in this study. Special thanks to Dr. Xiangyu Tang from the Department of Gastroenterology at Shenzhen Qianhai Shekou Free Trade Zone Hospital for his invaluable clinical and technical support. Abbreviations Triglyceride-Glucose index Body Mass Index Waist Circumference Environment and Chronic Disease in Rural Areas of Heilongjiang, China United Kingdom Biobank World Health Organization Diastolic Blood Pressure Systolic Blood Pressure Glycated Hemoglobin Total Cholesterol to High-Density Lipoprotein Cholesterol Confidence Interval Hazard Ratio Subarachnoid Hemorrhage Ischemic Stroke Intracerebral Hemorrhage Hospital Episode Statistics Strengthening the Reporting of Observational Studies in Epidemiology International Classification of Diseases Rheumatic Heart Disease Coronary Heart Disease Cardiovascular Disease Directed Acyclic Graph Interquartile Range Restricted Cubic Splines Area Under the Curve Multiple Mediation Analysis Authors’ contributions Feifei Yao conceived of and designed the study, led the initial drafting of the manuscript, performed the statistical analyses, participated in the field investigation, and follow-ups for the ECDRAHC cohort. Jing Cui and Yuncheng Shen were involved in the field investigation, medical record follow-up, and data organization for the ECDRAHC cohort. Yuting Jiang, Yuanyuan Li, Xiaona Liu, Hongqi Feng, Zhe Jiao, and Chang Liu contributed to ECDRAHC field investigation, data organization, quality control, and laboratory testing. Fulan Hu performed the UKB data analyses. Fulan Hu, Wei Zhang and Dianjun Sun were responsible for the decision to submit the manuscript and had full access to the data. Wei Zhang and Dianjun Sun responsible for critical review of the manuscript, obtained funding, supervised the study. Funding National Health Commission of the People’s Republic of China; Central transfer payments endemic disease project (N/A); Special development funds of local colleges from the central government (Study on prevention and control of major diseases in Heilongjiang province, N/A); Heilongjiang Academy of Medical Sciences (Chronic Disease Cohort study in Rural areas of Northeast China, CR201801);Noncommunicable Chronic Diseases-National Science and Technology Major Project. Data availability Data availability statement: ECDRAHC data are available upon reasonable request, and UK biobank is available in the public UK Biobank Resource (www.ukbiobank.ac.uk/) Declarations Ethics approval and consent to participate UK Biobank was constructed under ethical approval obtained by the North West Multi-Centre Research Ethics Committee (REC reference: 11/NW/03820) and all participants provided written informed consent prior to participation. ECDRAHC cohort involves human participants and was approved by Harbin Medical University. All authors consent to the publication of this manuscript. Consent for publication All authors consent to the publication of this manuscript. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Fulan Hu, Email: hufu1525@163.com. Wei Zhang, Email: zwhxd@126.com. Dianjun Sun, Email: hrbmusdj@163.com. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Figure 1. Flowchart of participant selection Supplementary Material 2. Supplementary Figure 2. Directed Acyclic Graph. Note: A directed acyclic graph represents associations between covariates and primary exposure and outcome. Pink circles represent ancestors of the exposure and outcome (ie, confounders), blue circles represent ancestors of the outcome (ie, causal determinants of the outcome). Green lines represent causal paths, and pink lines represent biasing paths. The minimally sufficient adjustment set represents covariates such that the adjustment for this set of variables will minimize confounding bias when estimating the association between the exposure and the outcome. The minimally sufficient adjustment set was determined using the DAGitty software3. 1. The final minimally sufficient adjustment sets for estimating the direct effect of TyG index-related parameters on Stroke: Age, Diabetes, Drinking, Family history of CVD, Family history of stroke, Gender, Heart diseases, Hypertension, Leisure exercise frequency, Smoking, TC/HDL in ECDRAHC cohort. In UKB cohort additionally adjusted for HbA1c Supplementary Material 3. Supplementary Figure 3. The correlation of BMI and TyG related parameters in ECDRAHC(a) and UKB (b). Note:***p<0.001 Supplementary Material 4. Supplementary Figure 4. Sequential adjustment for potential mediating effects for hemorrhagic stroke. Note: All exposures are presented as per 1-SD increase. The standard deviation values for each exposure are as follows: BMI: 4.38, TyG-BMI: 45.95, TyG-WC: 140.94, TyG-FMI: 31.24, TyG-BRI: 13.87 Supplementary Material 5. Supplementary Figure 5. Multivariable adjusted dose-response associations between BMI, TyG related parameters and ischemic stroke. Note: ECDRAHC(a1-a5) adjusted for adjusted for age, gender, smoking status, drinking status, family history of CVD, family history of stroke, physical activity, tc/dhl, hypertension, diabetes, heart diseases; UKB(b1-b5) additionally adjusted for HbA1c Supplementary Material 6. Supplementary Figure 6. ROC curves for predicting ischemic stroke. Note: ECDRAHC cohort (a1: per 1-SD increase, a2: quartile increase) UKB cohort (b1: per 1-SD increase, b2: quartile increase) Supplementary Material 7. Supplementary Figure 7. Incidence and HRs of ischemic stroke in ECDRAHC cohort according to the number of chronic diseases. Note: Kaplan-Meier curves show cumulative incidence by number of chronic diseases. Bar chart shows HRs of incidence of ischemic stroke associated with number of chronic diseases, adjusted for adjusted for age, gender, physical activity, family history of CVD, family history of stroke, smoking status, drinking status and BMI Supplementary Material 8. Supplementary Figure 8. Incidence and HRs of ischemic stroke in UK biobank cohort according to the number of chronic diseases. Note: Kaplan-Meier curves show cumulative incidence by number of chronic diseases. Bar chart shows HRs of incidence of ischemic stroke associated with number of chronic diseases, adjusted for adjusted for age, gender, physical activity, family history of CVD, family history of stroke, smoking status, drinking status and TyG-BRI Supplementary Material 9. Supplementary Table 1. Description of missing data in ECDRAHC and UKB cohorts Supplementary Material 10. Supplementary Table 2. Association of BMI and TyG indices with Hemorrhagic Stroke Risk (Cohort Analysis and Meta-Analysis). Note: Note:ECDRAHC:Models ajusted for Age, Drinking, Family history of CVD, Family history of stroke, Gender, Leisure exercise frequency, Smoking, Diabetes, Heart diseases, Hypertension, TC/HDL UKB additionally adjusted for HbA1c Supplementary Material 11. Supplementary Table 3. Relationship between TyG indices and the risk of ischemic stroke (including missing variables) in ECDRAHC and UKB Databases. Note: Mod1: crude model. Mod2: adjusted for basic risk factors (Age, Drinking, Family history of CVD, Family history of stroke, Gender, Leisure exercise frequency, Smoking, UKB additionally adjusted for HbA1c). Mod3: additionally adjusted for mediators (Diabetes, Heart diseases, Hypertension, TC/HDL)1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 12. Supplementary Table 4. Association between BMI and ischemic stroke risk stratified by BRI tertiles in the UKB cohort. Note: adjusted for basic risk factors (Age, Drinking, Family history of CVD, Family history of stroke, Gender, Leisure exercise frequency, Smoking, and HbA1c) Supplementary Material 13. Supplementary Table 5. Association of obesity with the incidence of ischemic stroke in UKB cohorts. Note:models adjusted for age,gender,physical activity,smoking status, drinking status,family history of CVD, family history of stroke,TC/HDL,TG,HbA1c,hypertension, diabetes, heart diseases1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 14. Supplementary Table 6. Association of BMI and TyG index series with the incidence of ischemic stroke in ECDRAHC and UKB cohorts stratified by gender. Note: ECDRAHC: models adjusted for age, smoking status, drinking status, family history of CVD, family history of stroke, physical activity, tc/dhl, hypertension, diabetes, heart diseases . UKB:additionally adjustde for HbA1c based on ECDRAHC 1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 15. Supplementary Table 7. Association of BMI and TyG index parameters with the incidence of hemorrhagic stroke in UKB cohorts stratified by gender. Note:ECDRAHC: models adjusted for age, smoking status, drinking status, family history of CVD, family history of stroke, physical activity, tc/dhl, hypertension, diabetes, heart diseases UKB:additionally adjustde for HbA1c based on ECDRAHC 1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 16. Supplementary Table 8. Mediating effects of hypertension, diabetes, heart diseases and TC/HDL on the relationship between TyG Indices and ischemic stroke risk. Note:*Percentage of the BMI and TyG indices difference in stroke incidence risk explained by chronic diseases in the multivariable mediation model Supplementary Material 17. Supplementary Table 9. Association of BMI and TyG-related indices with ischemic stroke risk stratified by number of chronic diseases: cox regression analysis in UKB and ECDRAHC cohorts. Note: ECDRAHC: adjusted for age, gender, physical activity, family history of CVD, family history of stroke, smoking status, drinking status UKB:additionally adjusted for HbA1c1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 18. Supplementary Table 10. Association of BMI and TyG-related indices with ischemic stroke risk in UKB and ECDRAHC cohorts. Note: ECDRAHC: adjusted for age, smoking status, drinking status, family history of CVD, family history of stroke, physical activity, non-HDL, hypertension, diabetes, heart diseases. UKB:additionally adjusted for HbA1c1 HR = Hazard Ratio, CI = Confidence Interval Supplementary Material 19. Data Availability Statement Data availability statement: ECDRAHC data are available upon reasonable request, and UK biobank is available in the public UK Biobank Resource (www.ukbiobank.ac.uk/)