Risk Reclassification Beyond BMI by Waist Circumference and Waist-to-Hip Ratio Across Nine Cardiovascular Outcomes: Results from the Cross-Cohort Collaboration

Abstract

Background:

Body mass index (BMI) is commonly used to assess total adiposity yet does not provide insight into overall body composition, potentially obscuring meaningful heterogeneity in cardiovascular disease (CVD) risk. Waist circumference (WC) and waist-to-hip ratio (WHR) are surrogate measures of central adiposity that are commonly discordant with BMI and may provide additional prognostic information.

Objectives:

To quantify the reclassification and misclassification of traditional normal weight/overweight/obesity categories using central adiposity, and to evaluate the prognostic implications for CVD risk using the Cross-Cohort Collaboration (CCC).

Methods:

We included 260,959 participants from 15 cohorts of the Cross Cohort Collaboration (CCC) with harmonized data on either WC or WHR data and at least one of nine outcomes: time to first fatal and non-fatal myocardial infarction (MI), fatal and non-fatal stroke, heart failure (HF), atrial fibrillation (AF), total coronary heart disease (CHD), total cardiovascular disease (CVD), CHD mortality, CVD mortality, and all-cause mortality. Multivariable Cox proportional hazards models were used to estimate the hazard ratio (HR) of higher WC and WHR for each outcome across BMI categories.

Results:

Our sample included 260,861 individuals with WC data and 220,370 with WHR data, with a median follow-up of 20.0 years (25th percentile 12.7; 75th percentile 23.5). Among individuals with normal weight, 5% had high WC and 18% had high WHR; among those with overweight, 39% had high WC and 40% had high WHR. Among those with obesity, 9% had low WC and 45% had low WHR.

Among individuals with normal weight or overweight, clinically-defined high WC or WHR was associated with 15–50% greater risk for most outcomes. Among those with obesity, low WC was not associated with a significantly different risk compared with normal weight and low WC, except for all-cause mortality, for which risk was significantly lower. In contrast to males, females with obesity but low WHR retained significantly higher risk for all outcomes compared to normal weight and low WHR; however, the HRs were smaller compared with those with obesity and high WHR.

Among individuals with obesity, the population attributable risk associated with elevated WC or WHR ranged from 13% to 49%, with the highest estimates observed for elevated WC in HF and AF.

Conclusion:

WC and WHR identifies misclassification of conventional BMI categories and reclassifies CVD risk across normal weight, overweight, and obesity, adding diagnostic and prognostic value.

Central Illustration:

Central Adiposity Assessment Beyond BMI: Results from the Cross Cohort Collaboration (CCC)

Description: Reference category for Cox Models is Normal Weight/Low WC and Normal Weight/Low WHR. Models adjusted for age, sex, race and ethnicity, education, alcohol use, smoking status, diabetes hypertension, antihypertension medication use, dyslipidemia, and lipid lowering medication.

Numbers in bold are statistically significant.

High WC is defined as WC≥88cm and High WHR as WHR≥ 0.85.

BMI categories defined as the following: Normal weight: BMI 18.5 – <25.0 kg/m2, Overweight: BMI 25.0 – <30.0 kg/m2, Obesity: ≥30 kg/m2.

Models include a shared frailty term for cohort and site to account for within-site correlation and unobserved heterogeneity.

Abbreviations: WC=waist circumference, WHR=waist-to-hip ratio, BMI=body mass index, HF=heart failure, AF=atrial fibrillation, CVD= cardiovascular disease

Introduction:

Body mass index (BMI), calculated as weight in kilograms divided by height in meters squared, is endorsed by the World Health Organization (WHO) as a clinical tool to diagnose obesity. However, BMI lacks precision in assessing total body fat distribution, which is a crucial factor in the etiology of most obesity-related diseases.¹ Visceral fat is more metabolically active than subcutaneous fat, leading to increased secretion of pro-inflammatory cytokines into the bloodstream and surrounding tissues.² These cytokines have been shown to increase insulin resistance, activate the renin-angiotensin-aldosterone system (RAAS) and the sympathetic nervous system (SNS), enhance the production of very low-density lipoprotein (VLDL), and contribute to overall metabolic dysregulation.³ Through these mechanisms, central adiposity has been shown to be an independent risk factor for type 2 diabetes, hypertension, dyslipidemia, the formation of atherosclerotic plaques, as well as the development and progression of metabolic-associated steatotic liver disease (MASLD).^2–4

Despite growing evidence linking central adiposity to adverse cardiometabolic outcomes, BMI remains the most used clinical metric to define obesity and stratify cardiovascular risk. However, prior studies demonstrate that discordant phenotypes, particularly individuals with normal BMI but elevated central adiposity, have higher risk of several cardiovascular outcomes, whereas some individuals with obesity and lower central adiposity exhibit comparatively lower risk^5–10. Furthermore, the observation that patients with established cardiovascular disease and moderately elevated BMI may have improved prognosis, often termed the obesity paradox, may in part reflect underlying heterogeneity in fat distribution rather than a protective effect of excess body mass¹¹.

To date, there is a lack of large, comprehensive studies examining the implications when measures of central and general adiposity are discordant. To address this gap, we harmonized waist circumference (WC), waist-to-hip ratio (WHR), two anthropomorphic measures commonly used as surrogates for the extent of central fat in the body, with BMI data from 15 National Institutes of Health (NIH) and National Heart, Lung, and Blood Institute (NHLBI) prospective cohorts as part of the Cross Cohort Collaboration (CCC) study. In this analysis of participants with no reported history of CHD, we aimed to quantify reclassification when adding WC or WHR to BMI, and to demonstrate that measures of central adiposity identify meaningful risk heterogeneity within BMI categories.

Methods:

Data Availability Statement

The data and methods used in this study will be made available exclusively to researchers certified by CCC-Obesity and individual cohorts’ steering committees, only for the purpose of reproducing the results or replicating the procedures.

Ethical Approval and Informed Consent

This study was approved by the Johns Hopkins School of Medicine Institutional Review Board (Approval number: IRB00182463; Date: 24 June 2020). Parent cohorts all had participants sign informed consent and were covered by local IRBs.

Study Population:

The Cross-Cohort Collaboration (CCC) is a harmonized dataset of 24 prospective cohort studies, with 322,782 participants predominantly in the U.S. Rationale and details of the CCC study have been described elsewhere.¹² Our study, on behalf of the CCC-Obesity Working Group, included a subset of the CCC that had information on WC or WHR and followed participants for at least one of our outcomes of interest (N=274,844). To reduce the likelihood of confounding due to undiagnosed cancer or other pathological conditions that result in weight loss, we excluded participants with BMI <18.5 kg/m² (N=2,617). We also excluded participants with reported history of CHD at baseline (N=11,268), resulting in a final sample of 260,959. Our final sample size included N=260,861 with information on WC, N=220,370 with available information on WHR, and N=220,306 with both measures. Individuals with a reported history of stroke (N=4,851), HF (N=1,951), and AF (N=8,994) at baseline were excluded from their respective analyses. Inclusions and exclusions of our study sample are displayed in Supplemental Figure S1.

The cohorts that are included in this study are the following: 1) Atherosclerosis Risk in Communities (ARIC) Study, 2) Coronary Artery Risk Development in Young Adults (CARDIA) Study, 3) Cardiovascular Health Study (CHS), 4) Chronic Renal Insufficiency Cohort (CRIC)¹³, 5) Dallas Heart Study (DHS), 6) the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil), 7) Framingham Heart Study (FHS) third generation, 8) Jackson Heart Study (JHS), 9) Multi-Ethnic Study of Atherosclerosis (MESA), 10) the REasons for Geographic and Racial Differences in Stroke Study (REGARDS) 11) Rancho Bernardo Study (RBS), 12) the Strong Heart Study (SHS), 13) Study of Osteoporotic Fractures (SOF), 14) the Study of Women’s Health Across the Nation (SWAN), and 15) Women’s Health Initiative (WHI).

Definition of Harmonized Exposures

The following harmonized measures in the CCC were used: WC (in centimeters), WHR (WC divided by hip circumference, both in centimeters), and BMI, calculated as weight (in kilograms) divided by the square of height (in meters). Recommended NHLBI and WHO binary clinical cut points of both WC and WHR representing high/low WC (> 88cm for females, >102 for males)¹⁴ and high/low WHR (> 0.85 for females, >0.90 for males)¹⁵ were also assessed. BMI (kg/m²) was categorized based on the WHO classification as follows: normal weight (BMI 18.5 to <25 kg/m²), overweight (BMI 25.0 to <30 kg/m²), and obesity (BMI ≥30.0 kg/m²).¹⁶ Cohort specific protocols on measurement of WC and hip circumference is provided in Supplemental Table S1. Distribution of WC, hip circumference, and WHR are presented by cohort in Supplemental Table S2.

Cardiovascular and Mortality Outcomes:

A total of nine outcomes relevant to cardiovascular health were collected and harmonized in CCC: fatal and non-fatal myocardial infarction (MI), fatal and non-fatal stroke, heart failure (HF), diagnosed atrial fibrillation (AF), total CHD, total CVD, CHD mortality, CVD mortality, and all-cause mortality. Total CHD events were defined as a composite of MI, coronary revascularization, or coronary death. Total CVD events were defined as a composite of all atherosclerotic CVD events including CHD, stroke, or cardiovascular death (coronary death, stroke death, other atherosclerotic death, or other CVD death). Not all cohorts collected all nine outcomes; therefore, each final harmonized outcome reflects the maximum amount of data available across the included cohorts. Sample size and follow-up time for each outcome are presented overall and by cohort in Supplemental Table S3–4.

Harmonization of covariates:

The definitions of all harmonized sociodemographic and traditional risk factors have been previously described.¹² Briefly, information on age, sex, race and ethnicity (White, Black or African American, Asian, Hispanic, American Indian or Alaskan, other), education (less than high school, completed high school, college or advanced degree), smoking status (never, former, current), alcohol use (yes/no) were harmonized across the cohorts. Cardiometabolic parameters and medication use were harmonized and used to define the following cardiovascular risk factors: hypertension; defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or use of antihypertensive medications. Diabetes; defined as a fasting blood glucose level ≥126 mg/dL, non-fasting blood glucose ≥200 mg/dL, previous diagnosis of diabetes (treated or untreated), or use of anti-diabetes medications. Hyperlipidemia; defined as either: 1) total cholesterol >240 mg/dL; 2) triglycerides >200 mg/dL; or 3) low-density lipoprotein cholesterol >160 mg/dL. Data harmonization in the CCC followed published best practices in the field and has been coded in a master file to enable replication. This study was conducted and reported in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.

Statistical Analysis

Baseline characteristics are presented across sex-specific high/low categories of WC and WHR within each BMI category. Normally distributed continuous variables are presented as mean ± standard deviation, while non-normally distributed continuous variables are presented as median (25^th and 75^th percentile). Categorical variables are presented as numbers and the total column percentage (%).

For each of the nine outcomes, the total sample size, number of events, prevalence (%), and incidence rate (per 1,000 person-years) are reported overall and within each category of BMI across WC and WHR sex-specific quartiles. Using Poisson regression models, we calculated the adjusted incidence rates for each of our nine outcomes across high/low WC and WHR across each BMI category. Adjusted incidence rates were calculated at mean levels of age, sex, race and ethnicity, smoking status, alcohol use, and education level within the study sample.

To explore the implications of discordant BMI and WC or WHR, we created two 4×3 categorical variables. One set represents the combination of WC quartiles and BMI categories, while the other represents WHR quartiles and BMI categories. The distribution of each is presented by sex in Supplemental Figure S2. Using multivariable Cox regression models, we estimated the HR, and 95% CI associated with higher WC and WHR across BMI categories for each outcome. Normal weight individuals in the lowest quartile of WC or WHR (quartile 1) served as reference categories.

Similarly, we also created two 2×3 categorical variables. One set represents the combination of high/low WC and BMI categories, while the other represents high/low WHR and BMI categories. The distribution of each is presented by sex in Supplemental Figure S3. The associated HR and 95% CI were also plotted and presented for all nine outcomes by sex using normal weight /low WC and normal weight/low WHR as reference categories.

To address potential confounding by sociodemographic and traditional CVD risk factors, the following model definition was used; Model 1 is adjusted for age, sex, race and ethnicity, education, alcohol use, smoking status, diabetes, hypertension, antihypertension medication use, hyperlipidemia, and lipid lowering medication use.

We also evaluated outcomes across BMI categories using two complementary models: the risk classification model, which estimated the cardiovascular risk by BMI category without continuous BMI adjustment, and the BMI-conditional model, additionally adjusting for continuous BMI to estimate associations fully independent of overall body mass. All estimates from the forest plots in the analysis was presented using full Model 1 adjustments without continuous BMI (the risk classification model).

The population attributable fraction (PAF) of high/low WC and WHR was calculated within each BMI category using multivariable Cox proportional hazards models adjusted for age, sex, race and ethnicity, education, smoking status, and alcohol use. PAF represents the proportion of events attributable to high WC or WHR within each BMI category. PAF and their corresponding 95% CIs was calculated using the STATA command punafcc¹⁷.

All models include a shared frailty term for cohort and site to account for within-site correlation and unobserved heterogeneity. In cases where risk factor prevalence was missing in less than 10% of participants, multiple imputation using logistic regression was performed within each cohort, utilizing non-missing risk factor data.¹⁸ We applied Bonferroni correction for multiple comparisons, interpreting significance at a threshold of 0.0056 (α=0.05/9) to account for the nine tests conducted. All analyses were performed with Stata/SE 17.0 (StataCorp LP, College Station, TX).

Results:

Our study sample included 260,861 individuals with WC data and 220,370 with WHR data (Table 1). Among individuals with normal weight, 4,182(5%) had high WC and 13,964 (18%) had high WHR; among those with overweight, 35,885 (39%) had high WC and 31,743 (40%) had high WHR. Among those with obesity, 6,952 (9%) had low WC and 28,258 (45%) had low WHR (Central Illustration).

Table 1.

Baseline Characteristics by High and Low Waist Circumference and Waist-to-Hip Ratio Across BMI Categories

	Waist Circumference (N= 260,861)						Waist-to-Hip Ratio (N=220,370)
	Normal Weight		Overweight		Obesity		Normal Weight		Overweight		Obesity
	Low WC	High WC	Low WC	High WC	Low WC	High WC	Low WHR	High WHR	Low WHR	High WHR	Low WHR	High WHR
N	82,459	4,182	57,146	35,885	6,952	72,744	63,112	13,964	46,858	31,743	28,258	35,049
Age, years	61.0± 11.1	63.2±10.3	60.8±10.5	62.8± 9.5	59.5±10.6	60.6±9.4	60.9±10.6	62.3±10.4	61.5±9.7	61.7±9.9	60.9±9	60.4±9.1
Sex, % Female	70856, 86.64%	4029, 97.13%	43078, 75.97%	31280, 87.87%	5325, 77.25%	62479, 86.52%	59408,94.7%	10017,72.08%	44838,96.37%	22822,72.24%	27739,98.91%	29926,85.85%
Race and Ethnicity
White	67240, 82.35%	3170, 76.61%	42879, 75.74%	26757, 75.27%	4463, 64.86%	47230, 65.48%	53889, 85.98%	10249, 73.93%	38733, 83.29%	22895, 72.61%	21258, 75.83%	23853, 68.5%
Black	9020, 11.05%	619, 14.96%	10187, 17.99%	6379, 17.94%	1992, 28.95%	19921, 27.62%	4759, 7.59%	1977, 14.26%	5433, 11.68%	5234, 16.6%	5267, 18.79%	7372, 21.17%
Hispanic	470, 0.58%	70, 1.69%	671, 1.19%	578, 1.63%	96, 1.4%	1221, 1.69%	225, 0.36%	245, 1.77%	242, 0.52%	794, 2.52%	183, 0.65%	847, 2.43%
Asian	3184, 3.9%	98, 2.37%	1205, 2.13%	470, 1.32%	69, 1%	409, 0.57%	2455, 3.92%	822, 5.93%	780, 1.68%	894, 2.84%	178, 0.63%	300, 0.86%
American or Alaskan Indian	1724, 2.11%	179, 4.33%	1659, 2.93%	1360, 3.83%	261, 3.79%	3337, 4.63%	1340, 2.14%	563, 4.06%	1312, 2.82%	1701, 5.39%	1147, 4.09%	2442, 7.01%
Education
Less than High School	5388, 6.59%	599, 14.44%	4825, 8.51%	4267, 11.99%	632, 9.17%	9225, 12.78%	2964, 4.72%	2136, 15.37%	2835, 6.09%	4604, 14.57%	2266, 8.08%	5231, 15.01%
Completed High School	16393, 20.05%	1121, 27.03%	12242, 21.59%	8396, 23.59%	1615, 23.43%	17431, 24.14%	11847, 18.88%	3708, 26.68%	9302, 19.99%	8144, 25.78%	6149, 21.93%	8896, 25.52%
More than High School	59997, 73.37%	2428, 58.53%	39634, 69.9%	22935, 64.43%	4646, 67.4%	45555, 63.09%	47922, 76.39%	8054, 57.95%	34391, 73.91%	18843, 59.65%	19629, 69.99%	20732, 59.47%
Body Mass Index, kg/m2	22.6±1.6	23.7±1.2	26.9±1.3	28±1.3	32.6±5.2	35.4±4.9	22.5±1.6	23.2±1.4	27.2±1.4	27.5±1.4	34.9±5.1	35±4.6
Systolic Blood Pressure, mm Hg	123±18.8	126.8±20.4	126.4±17.9	128.9±18.2	127.6±16.7	130.7±17.3	122.5±18.6	127.3±20.4	126.4±17.9	129.4±18.8	129.5±17	131.6±17.7
HDL-Cholesterol, mg/dL	59.7±17.2	57.8±16.5	51.6±15	52.3±15.1	49.5±14.9	49±13.7	62.9±16.5	55±16.3	57.7±15.2	49.8±14	54.1±13.6	47.9±13
LDL-Cholesterol, mg/dL	119.6±35.4	127.3±40.1	125.1±35.9	127.4±38.3	121.7±35.5	122.6±36.8	119.6±35	129.8±37.2	127.5±35.2	132.5±37.7	126.2±35.6	128.8±37.1
Triglycerides, mg/dL	89 (65, 125)	111 (82, 155)	107 (77, 155)	121 (87, 173)	109 (77, 163)	124 (89, 176)	81 (60, 112)	110 (81, 154)	95 (68, 136)	126 (91, 179)	107 (76, 150)	134 (97, 190)
Total Cholesterol, mg/dL	200.4±40.1	211.6±43.5	202.4±41.1	208.5±43.2	198.2±40.8	201.2±42.4	201.3±39.9	210.7±41.7	207.7±40.7	212.1±42.4	204.9±40.7	208.9±43.1
Hypertension	21583, 26.4%	1664, 40.13%	19287, 34.02%	15247, 42.85%	2661, 38.62%	36616, 50.72%	14386, 22.94%	5086, 36.62%	13671, 29.39%	13159, 41.67%	10705, 38.18%	17096, 49.06%
Hypertension Medication	9747, 12.06%	1005, 24.72%	10138, 18.15%	9000, 25.69%	1500, 22.11%	23310, 32.83%	4952, 7.96%	2679, 19.68%	5279, 11.43%	7197, 23.26%	4722, 16.94%	9756, 28.57%
Hyperlipidemia	10551, 14.01%	1031, 26.5%	10200, 19.57%	8423, 25.64%	1185, 18.68%	14986, 22.46%	6524, 11.4%	3549, 26.98%	6428, 15.32%	8983, 30.2%	3851, 15.1%	8438, 25.71%
Lipid-Lowering Medication	5695, 7.34%	479, 12.11%	6358, 11.77%	5062, 14.87%	850, 12.8%	10836, 15.5%	3045, 5.14%	1270, 9.66%	3647, 8.28%	3596, 11.99%	2336, 8.63%	3914, 11.7%
Diabetes	3061, 3.76%	411, 10%	3442, 6.11%	4342, 12.28%	539, 7.87%	13571, 18.94%	1332, 2.13%	1315, 9.5%	1372, 2.95%	3977, 12.63%	1949, 6.97%	6832, 19.67%
Diabetes Medication	1602, 2.2%	292, 8%	2081, 4.26%	3033, 9.55%	333, 6%	9780, 15.47%	526, 0.88%	770, 5.81%	688, 1.53%	2548, 8.48%	1235, 4.54%	4776, 14.46%
Current Alcohol Use	42447, 51.94%	2090, 50.45%	26923, 47.51%	15581, 43.81%	2649, 38.44%	24356, 33.75%	32158, 51.29%	7565, 54.5%	20391, 43.84%	15352, 48.65%	8551, 30.49%	12041, 34.58%
Smoking Status
Current Smoker	9983, 12.33%	742, 18.09%	5255, 9.36%	4406, 12.52%	513, 7.51%	6609, 9.28%	6182, 9.95%	2816, 20.4%	3341, 7.26%	4436, 14.16%	1637, 5.9%	3680, 10.67%
Former Smoker	29133, 35.97%	1424, 34.72%	21899, 39.01%	13955, 39.64%	2424, 35.5%	28431, 39.92%	22842, 36.78%	4794, 34.73%	18066, 39.24%	12438, 39.7%	10860, 39.16%	14173, 41.09%
Never Smoker	41874, 51.7%	1935, 47.18%	28976, 51.62%	16840, 47.84%	3891, 56.99%	36186, 50.8%	33084, 53.27%	6193, 44.87%	24637, 53.51%	14454, 46.14%	15232, 54.93%	16638, 48.24%

^*Data are presented as frequency and percentage n (%), mean ± standard deviation, or median (25^th and 75^th percentile). High WC is defined as > 88cm for females, >102 cm for males and high WHR as > 0.85 for females, >0.90 for males.

BMI categories defined as the following: Normal weight: BMI 18.5 - <25.0 kg/m², Overweight: BMI 25.0 - <30.0 kg/m², Obesity: ≥30 kg/m². LDL-C, low density lipoprotein cholesterol; HDL-C, high density lipoprotein cholesterol; WC, Waist Circumference; WHR, Waist-to-Hip Ratio.

Within each BMI category, higher WC (vs lower WC) and higher WHR (vs lower WHR) were each associated with a greater prevalence of less than a high school education, hypertension, hyperlipidemia, diabetes, and current smoking (Table 1). Furthermore, across each BMI category, individuals with higher WC (vs lower WC) were more likely to be female, less likely to report current alcohol use, and have a higher BMI. In contrast, individuals with higher WHR were less likely to be female and more likely to report current alcohol use compared to their lower WHR counterparts. Within BMI categories, mean BMI were similar across high versus low WHR groups.

Adjusted incidence rates are presented by high and low WC and WHR within each BMI category in Figure 1. Within each BMI category and for each outcome, higher WC (vs low WC) and higher WHR (vs low WHR) were associated with higher adjusted incidence rates per 1000 person-years. Lower rates of all-cause mortality were observed in overweight compared to normal-weight individuals; however, the differences were more pronounced across WC categories. In individuals with obesity, those with high WHR had higher adjusted incidence rates overall compared to WC categories. The total number of events and unadjusted incidence rates for each outcome by BMI and sex-specific quartiles of WC and WHR is presented in Supplementary Table S5–S6.

Figure 1:

Adjusted Incidence Rates for Nine CVD-related Outcomes across Low/High Waist Circumference and Waist-to-Hip Ratio by BMI Categories.

Description: *Incidence rates were calculated at mean levels of age, sex, race and ethnicity, smoking status, alcohol use, and education.

BMI categories were defined as the following: Normal weight: BMI 18.5 to <25.0 kg/m2, Overweight: BMI 25.0 to <30.0 kg/m2, Obesity: BMI ≥30.0 kg/m2.

High WC is defined as > 88cm for females, >102 for males and high WHR as > 0.85 for females, >0.90 for males BMI: body mass index; MI: myocardial infarction, HF: heart failure; AF: atrial fibrillation; CHD: coronary heart disease; CVD: cardiovascular disease, PYs: Person-years; WC, Waist Circumference; WHR, Waist-to-Hip Ratio.

Among individuals with normal weight, PAFs for high WC and high WHR ranged across outcomes from 1.2–2.4% and 4.8–10.5%, respectively; among those with overweight, 9.9–15.3% and 5.4–23.7%; and among those with obesity, 28.8–48.9% and 12.8–27% (Table 2). The highest PAF was observed for high WC among individuals with obesity, particularly for AF (48.9% [41.2%, 55.6%]) and HF (46.2% [30.1%, 58.6%]). Among those with overweight, the highest PAFs were observed for high WHR, specifically for HF (23.7% [18.7%, 28.3%]) and MI (22.2% [19.0%, 25.2%]).

Table 2:

Population Attributable Fraction (PAF) of High Waist Circumference and High Waist-to-Hip Ratio within Body Mass Index Categories.

	Normal Weight		Overweight		Obesity
	High WC	High WHR	High WC	High WHR	High WC	High WHR
MI	1.9% (1%, 2.9%)	10.5% (8.1%, 12.9%)	11.4% (9.2%, 13.6%)	22.2% (19%, 25.2%)	34% (25.5%, 41.6%)	19.7% (15.7%, 23.5%)
Stroke	2.4% (1.6%, 3.3%)	9.2% (7.1%, 11.2%)	9.9% (7.7%, 12.1%)	18.2% (15.7%, 20.6%)	31.7% (23.6%, 38.9%)	22.8% (19.8%, 25.6%)
HF	1.8% (0.4%, 3.2%)	9.7% (5.7%, 13.4%)	15.3% (12.7%, 17.9%)	23.7% (18.7%, 28.3%)	48.9% (41.2%, 55.6%)	27% (22.2%, 31.4%)
AF	1.7% (−0.3%, 3.6%)	6.4% (0.6%, 11.8%)	12.7% (8.9%, 16.4%)	5.4% (−4.7%, 14.5%)	46.2% (30.1%, 58.6%)	20.8% (9.4%, 30.7%)
CHD	2% (1.3%, 2.6%)	9.1% (7.6%, 10.6%)	10.6% (8.9%, 12.2%)	17.2% (15.1%, 19.2%)	35.4% (29.4%, 40.9%)	17.9% (15.3%, 20.5%)
CVD	1.7% (1.2%, 2.1%)	7% (5.9%, 8.1%)	10.1% (8.9%, 11.4%)	14.4% (12.8%, 15.9%)	32.5% (28%, 36.7%)	15.6% (13.7%, 17.5%)
CHD Mortality	2.1% (1.4%, 2.8%)	8.9% (7.2%, 10.6%)	11.2% (9%, 13.3%)	15.4% (12.8%, 17.9%)	36.3% (28.5%, 43.3%)	18.4% (15.1%, 21.5%)
CVD Mortality	1.8% (1.3%, 2.3%)	6.4% (5.2%, 7.6%)	11% (9.4%, 12.5%)	12% (10.1%, 13.8%)	36.6% (31.1%, 41.6%)	13.8% (11.4%, 16.2%)
All-Cause Mortality	1.2% (1%, 1.5%)	4.8% (4.2%, 5.3%)	9.7% (8.9%, 10.4%)	9.6% (8.7%, 10.5%)	28.8% (25.8%, 31.7%)	12.8% (11.5%, 13.9%)

Population attributable fractions were estimated using multivariable Cox proportional hazards models adjusted for age, sex, race and ethnicity, education, smoking status, and alcohol use, with a shared frailty term for cohort and site.

BMI categories were defined as the following: Normal weight: BMI 18.5 to <25.0 kg/m², Overweight: BMI 25.0 to <30.0 kg/m², Obesity: BMI ≥30.0 kg/m².

High WC is defined as > 88cm for females, >102 cm for males and high WHR as > 0.85 for females, >0.90 for males.

BMI: body mass index; MI: myocardial infarction, HF: heart failure; AF: atrial fibrillation; CHD: coronary heart disease; CVD: cardiovascular disease, WC, Waist Circumference; WHR, Waist-to-Hip Ratio.

Among individuals with normal weight, higher WC quartiles were significantly associated with a greater risk for all outcomes (Figure 2). Furthermore, compared to individuals with normal weight and a WC in the lowest quartile, those with obesity and a WC in the 1st or 2nd quartile did not experience a significantly different hazard for any outcomes except all-cause mortality, in which risk was significantly lower in the 2^nd quartile. Model estimates for Figure 2 is presented in Supplemental Tables S7. Results of WC quartiles by BMI categories are further presented by sex in Supplemental Tables S8–S9.

Figure 2.

Hazard Ratio and 95% Confidence Interval for MI, Stroke, HF, AF, CHD, CVD, CHD Death, CVD Death, and All-Cause Mortality Associated with Waist Circumference Quartiles across BMI Categories.

Description: Reference category is Normal Weight/1st Quartile WC.

Models adjusted for age, sex, race and ethnicity, education, alcohol use, smoking status, diabetes hypertension, antihypertension medication use, dyslipidemia, and lipid lowering medication.

Sex-specific quartiles for WC in cm. are defined as follows; Females: Q1(32.5–77), Q2 (77–86), Q3(86–96.5), Q4 (96.5–334). Males: Q1(32–90), Q2 (90–97.1), Q3(97.1–106), Q4 (106–244).

BMI categories defined as the following: Normal weight: BMI 18.5 - <25.0 kg/m2, Overweight: BMI 25.0 - <30.0 kg/m2, Obesity: ≥30 kg/m2.

Models include a shared frailty term for cohort and site to account for within-site correlation and unobserved heterogeneity.

Abbreviations: WC=waist circumference, BMI=body mass index, MI=myocardial infarction, HF=heart failure, AF=atrial fibrillation, CHD=coronary heart disease, CVD= cardiovascular disease.

Similarly, among individuals with normal weight, higher quartiles of WHR were associated with a higher risk for all outcomes, except AF (Figure 3). Compared to individuals with normal weight and a WHR in the lowest quartile, individuals with obesity and a WHR in the 1st or 2nd quartile showed a higher hazard for all outcomes yet estimates trended lower that their 3^rd and 4^th quartile counterparts. Model estimates for Figure 3 are presented in Supplemental Tables S10. Results of WC and WHR quartiles by BMI categories are further presented by sex in Supplemental Tables S11–S12.

Figure 3.

Hazard Ratio and 95% Confidence Interval for MI, Stroke, HF, AF, CHD, CVD, CHD Death, CVD Death, and All-Cause Mortality Associated with Waist-to-Hip Ratio Quartiles across BMI Categories.

Reference category is Normal Weight/1st Quartile WHR. Models adjusted for age, sex, race and ethnicity, education, alcohol use, smoking status, diabetes hypertension, antihypertension medication use, dyslipidemia, and lipid lowering medication.

Sex-specific quartiles for WHR are defined as follows; Females: Q1(0.28–0.76), Q2 (0.76–0.81), Q3(0.81–0.87), Q4 (0.87–2.88). Males: Q1(0.33–0.90), Q2 (0.90–0.95), Q3(0.95–0.99), Q4 (0.99–2.27).

BMI categories defined as the following: Normal weight: BMI 18.5 - <25.0 kg/m2, Overweight: BMI 25.0 - <30.0 kg/m2, Obesity: ≥30 kg/m2.

Models include a shared frailty term for cohort and site to account for within-site correlation and unobserved heterogeneity.

Abbreviations: WHR=waist-to-hip ratio, BMI=body mass index, MI=myocardial infarction, HF=heart failure, AF=atrial fibrillation, CHD=coronary heart disease, CVD= cardiovascular disease

When evaluating clinically defined cut points of WC and WHR among females with normal weight, those with a high WC or high WHR had a higher hazard for most outcomes except HF and AF, where only high WHR was significantly associated with a higher risk of HF (Supplemental Figure S4). Furthermore, compared to females with normal weight and low WC, those with obesity and a low WC did not experience a significantly different hazard for any outcomes except all-cause mortality, in which risk was significantly lower. Contrary to that seen with low WC, females with obesity and a low WHR experienced a significantly higher hazard for all outcomes when compared to females with normal weight and a low WHR, however the risk remained lower than their high WHR counterpart. Model estimates for the overall sample are presented in Supplemental Tables S13. Estimates for Supplemental Figure 4 are presented in Supplemental Tables S14.

Among males with normal weight, those with a clinically defined high WC had a significantly greater hazard for the risk of CVD mortality and all-cause mortality while those with a high WHR had a significantly greater hazard for the risk of HF, CHD, CVD, CHD mortality, and all-cause mortality (Supplemental Figure 5). Furthermore, compared to males with normal weight and low WC, those with obesity and a low WC did not experience a significantly different hazard for any outcomes except all-cause mortality, in which risk was significantly lower. Compared to males with normal weight and low WHR, those with obesity and low WHR similarly did not experience a significantly different hazard for most outcomes, except AF, in which the relative hazard was significantly higher. Model estimates for Supplemental Figure 5 is presented in Supplemental Tables S15.

BMI Conditional Analysis

The addition of continuous BMI to the model adjustments minimally attenuated the associations of high WC and WHR in individuals with normal weight. Greater attenuation was observed in those with overweight or obesity; however, most estimates, particularly for individuals with obesity and low WHR, were attenuated toward the null, with the largest differences seen for HF and AF. In individuals with obesity and low WC, adjusting for continuous BMI resulted in significantly reduced risk for several outcomes.

Discussion:

In our study of participants without a reported history of CHD, we demonstrated that relying on BMI alone, without accounting for measures of central adiposity, can obscure meaningful heterogeneity in CVD risk. Indeed, it appears that WC and WHR reclassify risk defined by traditional BMI thresholds. For example, among individuals with normal weight in our sample, 5% had elevated central adiposity defined by WC, while 18% were classified as high WHR. These normal-weight phenotypes were associated with higher risk across most outcomes, although the magnitude varied depending on the central adiposity measure and the outcome assessed.

Prior studies have also reported that normal weight individuals with central obesity have greater risk for both diabetes and hypertension compared to their normal weight counterparts^19,,20, highlighting the important role central adiposity may play in the development of CVD. Several smaller studies have also shown that these individuals may experience the highest risks of both all-cause and CVD-related mortality⁵,^21–23.

In a study of 15,184 adults aged 18 to 90 years published from the NHANES III (Third National Health and Nutrition Examination Survey), Sahakyan et. al. demonstrated that men with normal BMI and central obesity (defined by WHR), had twice the mortality risk compared to participants with overweight or obesity (HR 2.24, 95% CI 1.52–3.32 and HR 2.42, 95% CI 1.30–4.53, respectively). Also, that women with normal weight and central obesity had a 32% increase in risk of mortality compared to women with obesity (HR 1.32, 95 % CI 1.15–1.51)⁵.

At the high end of the BMI spectrum, we observed that among both females and males with obesity, low WC was generally not associated with significantly different risk for most outcomes compared with individuals with normal weight and low WC, except for all-cause mortality, where risk was lower. Similar results were seen in a large Korean cohort study, in which individuals with general obesity but without abdominal obesity did not have significantly elevated risk of major adverse cardiac events (adjusted HR 1.06, 95% CI 0.98–1.16 in men; HR 1.07, 95% CI 0.88–1.30 in women)²⁴. While the precise mechanism underlying this lower risk of all-cause mortality is not fully elucidated, both cardiorespiratory fitness (CRF) and body composition (specifically muscle versus fat mass) have been shown to be important mediators of all-cause mortality risk.²⁵⁻²⁷ Although WC alone is not a direct measure of the latter, this WC range may correspond to a subgroup of individuals with a higher prevalence of better CRF or a higher muscle mass composition.

In contrast, females with obesity and low WHR retained higher hazards for most outcomes compared with those with normal weight and low WHR, although risks remained lower than in their high-WHR counterparts. This pattern is possibly largely driven by residual differences in BMI within the obesity group. Among individuals with obesity, WC appeared more strongly correlated with overall adiposity, with lower mean BMI observed in the low WC group compared to high WC, suggesting that WC may preferentially capture a lower-risk, less severe obesity phenotype. In contrast, a similar BMI gradient was not observed across WHR strata, suggesting that comparisons by WHR may retain greater residual confounding by BMI. As a result, a portion of the observed risk among individuals with obesity and low WHR likely reflects unaccounted differences in BMI severity rather than true differences in adiposity distribution. This interpretation is supported by prior analyses of the CCC dataset, which showed that higher obesity severity increased risk for multiple outcomes²⁸. Furthermore, in BMI-conditional models adjusting for continuous BMI, the excess risk associated with low WHR was largely attenuated to null for most outcomes.

The PAF estimates underscore the substantial population-level impact of central adiposity across BMI strata. While PAFs for excess central adiposity were modest among individuals with normal weight and overweight, among those with obesity nearly half of HF and AF events and more than a third of CHD and CHD mortality events occurred in participants with concurrent elevated WC. These findings suggest that central adiposity is a major driver of cardiovascular risk, with its population-level impact particularly pronounced at higher BMI. Together with the elevated risk observed among normal weight individuals with high WC or WHR, these observations support incorporating central adiposity measurement into routine cardiovascular risk assessment across the BMI spectrum.

The International Atherosclerosis Society (IAS) and International Chair on Cardiometabolic Risk (ICCR) Working Group on Visceral Obesity released a 2017 consensus stating that BMI alone is not sufficient to properly assess, evaluate, or manage the cardiometabolic risk associated with higher adiposity; these groups further recommended waist circumference to be added as a standard measurement in clinical practice alongside BMI, for classifying obesity.²⁹ More recently, the Lancet Diabetes & Endocrinology Commission recently released a consensus outlining new clinical criteria for diagnosing obesity.³⁰ These criteria include either: (1) direct body fat measurement using imaging modalities, (2) in addition to BMI, at least one anthropometric measure such as WC, WHR, or waist-to-height ratio (WHtR), or (3) at least two anthropometric measures (WC, WHR, or WHtR), with assessments of such using validated methods that are appropriate for age, sex, and ethnicity. Additionally, for individuals with BMI > 40 kg/m², excess adiposity can be assumed without further measurement.

Age may modify the association between central adiposity and CVD outcomes^31−34, potentially reflecting sex-specific changes in body composition that occur with advancing age. For instance, after menopause, women tend to accumulate more central fat around the abdomen due to lower estrogen levels, which may result in higher rates of visceral adiposity in older women.³⁵ Further, studies have shown that even after adjusting for total fat mass, men tend to accumulate visceral fat more rapidly than pre-menopausal women, suggesting that men develop visceral fat earlier in life.³⁶ Therefore future studies are needed to investigate the association of age with relationships between WC, WHR, and CVD risk among men and women to better understand the implications of central adiposity across all age ranges.

There are several strengths of our study. To date, studies assessing CVD risk in relation to BMI and central adiposity measures have been limited in providing comparable long-term risk estimates across multiple CVD subtypes, particularly when accounting for traditional CVD risk factors. By leveraging rigorous harmonization across high-quality cohorts within the CCC, our study addresses these limitations, offering a comprehensive analysis of adjudicated CVD outcomes while establishing WC and WHR as strong discriminators of CVD risk, with sufficient power to study those individuals with discordant BMI. Additionally, our large sample size and diverse cohort composition enables a detailed, sex-specific evaluation of central obesity’s association with future CVD risk, with strong generalizability to broader populations. A further strength is our predominantly female representation, which provides sufficient power to examine sex specific implications of risk, in a context where prior evidence has largely been derived from male-dominant cohorts.

This study also has limitations. Our study did not have measures of physical activity, diet, or genetic obesity-risk, each of which have been shown to play significant roles in the development of CVD^37–39. Our study also includes only a single assessment of WC and WHR, which may limit our understanding of how changes in central fat accumulation over the life course influence future CVD risk⁴⁰. Furthermore, given the recent rapid rise in obesity, the implications for future CVD risk may be underrepresented by virtue of our decades-long follow-up of the constituent cohorts. There is increasing evidence suggesting that genetic variants associated with obesity may have a more pronounced effect in individuals born more recently, particularly with regards to proposed gene-environment interaction⁴¹. In addition, given the observational nature of our study, we cannot exclude residual confounding and cannot establish causality.

Conclusion:

In conclusion, our findings emphasize the critical role of identifying elevated central adiposity, even in individuals with a normal BMI or with a BMI in the overweight range. Relying solely on BMI may result in misclassification of cardiovascular risk across a wide range of cardiovascular outcomes. We encourage clinicians to consider central adiposity distribution across the entire BMI spectrum when evaluating cardiovascular risk in primary prevention settings.

Non-standard Abbreviations and Acronyms

AF

Atrial fibrillation

CCC

Cross-Cohort Collaboration

CHD

Coronary heart disease

CVD

Cardiovascular disease

DBP

Diastolic blood pressure

HF

Heart failure

LDL-C

Low-density lipoprotein cholesterol

SBP

Systolic blood pressure

WC

Waist Circumference

WHR

Waist-to-hip Ratio