The global rise in obesity has prompted interest in anthropometric indices that go beyond body mass index to better capture cardiometabolic risk. This study compared three commonly used indices — BMI, waist circumference (WC), and body roundness index (BRI) — for their prognostic associations with major clinical outcomes in a large, nationwide Korean cohort. The primary intent was to determine whether any single index consistently outperforms others across outcomes such as atherosclerotic cardiovascular disease (ASCVD), all-cause mortality, and atrial fibrillation (AF).
This retrospective cohort analysis used linked records from the Korean National Health Insurance Service (NHIS) health screening database. Standardized measurements and laboratory tests were obtained from routine national health screenings, which are governed by uniform quality control protocols. Data access for this analysis occurred between 30 January and 30 March 2026.
Participants were adults aged ≥55 years who underwent health screening between 2009 and 2012. From an initial sample of ~1.5 million, exclusions applied for prior MI, stroke, heart failure, or AF at baseline; extreme anthropometric values; missing key variables; or invalid mortality records. The final analytic cohort comprised 1,199,633 individuals, followed for up to 10 years.
Height, weight, and waist circumference were measured during standardized examinations. BMI was calculated as weight (kg) divided by height (m) squared. WC was measured at the midpoint between the lower rib margin and the iliac crest. BRI was computed using height and WC according to a previously validated formula cited in the source.
Primary outcomes were:
Secondary outcomes included the individual ASCVD components and cancer-related mortality.
Prognostic discrimination was evaluated using receiver operating characteristic (ROC) curves with area under the curve (AUC) calculations for each index and outcome. To address potential non-linear associations, restricted cubic spline (RCS) models were constructed. Multivariable Cox proportional hazards models estimated hazard ratios across quintiles of each anthropometric index, adjusted for age, sex, diabetes, hypertension, dyslipidaemia, smoking status, and Charlson comorbidity index. The third quintile (Q3) served as the reference due to observed nadirs within middle ranges. Landmark analyses excluded early events (first 1–2 years) to explore reverse causation.
Mean age was 63.5 ± 7.3 years; 47.2% of participants were male. Prevalence of hypertension was 49.5%, diabetes 13.2%, dyslipidaemia 48.8%, and current smoking 16.2%. Mean BMI, WC, and BRI were 24.04 ± 3.02 kg/m2, 82.4 ± 8.3 cm, and 3.75 ± 1.04, respectively.
During follow-up, ASCVD occurred in 9.6% of participants, all-cause mortality in 10.9%, cancer-related mortality in 4.2%, and AF in 4.7%. ROC comparisons showed outcome-specific differences in discriminatory performance:
Overall, discriminatory ability of any single index was modest, with AUC values not exceeding 0.60.
RCS models revealed predominantly non-linear relations between anthropometric indices and outcomes. Patterns varied by index and outcome:
These non-linear, often bimodal associations attenuate ROC discrimination when a single continuous marker must capture increased risk at both extremes.
After adjustment for established risk factors, many associations were attenuated. Key adjusted findings included:
Landmark analyses excluding events occurring within the first 1–2 years demonstrated similar AUCs and non-linear risk patterns, suggesting that early reverse causation (pre-existing illness causing low weight) did not fully account for the elevated risks observed at lower index values.
In this large nationwide cohort, no single anthropometric index uniformly outperformed others across major clinical outcomes. Instead, index performance was outcome-specific: BRI performed best for ASCVD, BMI for all-cause mortality (driven by underweight-associated risk), and WC for AF. The modest AUCs highlight the limited standalone prognostic value of anthropometric measures relative to traditional cardiometabolic risk factors. Non-linear associations (J- and U-shaped) further complicate discrimination by a single continuous index.
The study was approved by the institutional review board of Seoul National University Bundang Hospital. NHIS data used are not publicly available because of confidentiality rules but may be accessed by qualified researchers following NHIS governance procedures. Funding was provided by the National Research Foundation Bio and Medical Technology Development Program (Korean Government MSIT), and the authors declared no competing interests.
Each anthropometric index captures different biological aspects of adiposity and shows outcome-specific prognostic associations. Clinicians and researchers should consider the specific clinical endpoint when selecting anthropometric measures for risk assessment rather than relying on a single universal index.