# Comparison of laboratory-based and non-laboratory-based WHO and GLOBORISK CVD risk scores: A cross-sectional analysis of the APCAPS cohort

## TL;DR

This study compares lab and non-lab CVD risk scores in rural India, finding they agree well overall but slightly underestimate risk in some high-risk groups.

## Contribution

The study evaluates the validity of non-laboratory CVD risk scores in a rural Indian cohort and their correlation with subclinical atherosclerosis.

## Key findings

- Non-laboratory-based CVD risk scores showed high agreement with laboratory-based models (κ = 0.82 for WHO, κ = 0.72 for GLOBORISK).
- Both risk scores correlated strongly with subclinical atherosclerosis markers like CIMT, PWV, and AIx.
- Non-lab models underestimated risk in diabetics and those with high cholesterol by 1.2-4.2%.

## Abstract

Cardiovascular diseases (CVDs) represent a growing public-health challenge in India, where nearly one in four deaths is CVD-related. Accurate risk stratification underpins targeted prevention, yet laboratory-dependent tools are often impractical in resource-limited settings. The World Health Organization (WHO) and GLOBORISK initiatives both offer non-laboratory-based 10-year CVD risk algorithms alongside their laboratory-based counterparts. We aimed to compare laboratory- and non-laboratory-based WHO and GLOBORISK CVD risk scores, assess their concordance, and examine relationships with sub-clinical atherosclerosis in a rural Indian cohort.

We conducted a cross-sectional analysis of 2,465 adults (1,184 men, 1,281 women) aged 40−74 years from the third wave (2010−12) of the Andhra Pradesh Children and Parents Study (APCAPS). Participants with prior CVD were excluded. Ten-year CVD risk was calculated using sex-specific WHO (South Asia) and India-calibrated GLOBORISK models, both laboratory-based (age, sex, smoking, systolic blood pressure, diabetes, total cholesterol) and non-laboratory-based (age, sex, smoking, systolic blood pressure, BMI) algorithms. Categorical agreement was quantified via percentage agreement and quadratic weighted kappa (κ); continuous agreement by Bland-Altman analysis. We also evaluated linear associations between each risk score (categorical and continuous) and three sub-clinical atherosclerosis markers: carotid intima-media thickness (CIMT), pulse-wave velocity (PWV), and augmentation index (AIx), through sex-stratified multi-level linear regression with random intercept at the household level, adjusting for multiple testing (p < 0.01).

Median WHO-CVD-risk was 6.0% (IQR 4% − 9%) in men and 3.0% (2% − 4%) in women for both lab and non-lab models; median GLOBORISK-CVD-risk was 12.0% (9% − 16%) for lab-model vs. 15.0% (10% − 16%) for non-lab-model in men and 5.0% (3% − 9%) for lab-model vs. 5.0% (3% − 9%) for non-lab-model in women. Categorical agreement was substantial to almost perfect: WHO κ = 0.82 (overall), GLOBORISK κ = 0.72. Bland-Altman analyses demonstrated mean differences <1% between lab- and non-lab-based scores, though non-lab models underestimated risk by 4.2% in diabetics and 1.2% in participants with total cholesterol ≥200 mg/dL. Both risk scores showed positive, dose-response relationships with CIMT, PWV, and AIx (p-trend<0.001), with each SD increase in CVD-scores associated with clinically meaningful increases in all three markers of sub-clinical atheroscerosis.

Non-laboratory-based WHO and GLOBORISK CVD risk scores exhibit high overall agreement with laboratory-based models and correlate strongly with subclinical atherosclerosis in rural India. However, modest underestimation in high-risk subgroups (diabetics, hypercholesterolemia) warrants cautious interpretation. These findings support the feasibility of non-lab risk assessment in resource-constrained settings, while underscoring the need for prospective validation against hard cardiovascular outcomes prior to large-scale implementation.