Abstract
INTRODUCTION: Cardiovascular disease (CVD) remains the leading cause of mortality in Brazil. Accurate risk stratification is essential for guiding primary prevention, yet the concordance between global tools remains uncertain in specific populations. This study aimed to evaluate the agreement between the Framingham Global Risk Score (FGRS) and the 2019 World Health Organization (WHO) cardiovascular risk charts in a Brazilian population, assessing both calibration bias and the utility of non-laboratory models. METHODS: An observational, cross-sectional study was conducted at Hospital de Clínicas de Uberlândia (HC-UFU), a tertiary care center in Uberlândia, Brazil. A convenience sample of 140 adults (aged 40-74 years) was evaluated. Risk of 10-year cardiovascular events was estimated using the FGRS and WHO charts (Tropical Latin America region), applying both laboratory-based and non-laboratory (body mass index (BMI)-based) algorithms. Agreement was assessed using weighted kappa (kappa) and Bland-Altman analysis. We tested two threshold strategies: a standardized cut-off (≥20% for both) and tool-specific thresholds (FGRS ≥20% vs. WHO ≥10%) to assess clinical equivalence. RESULTS: The WHO charts systematically underestimated CVD risk compared to the FGRS. Using the standard ≥20% threshold, the FGRS identified 59.3% of men as high-risk, whereas WHO identified only 11.1%, revealing a substantial "prevention gap." Agreement between the FGRS and WHO was generally fair (kappa < 0.40) but improved significantly when a lower threshold (≥ 10%) was applied to the WHO charts. Conversely, internal consistency between laboratory and non-laboratory models was robust for both tools (kappa = 0.81), validating the use of BMI-based scores. CONCLUSION: In this high-risk Brazilian population, WHO charts yielded significantly lower risk estimates than Framingham, potentially excluding eligible patients from statin therapy if standard thresholds are used. Adopting a lower treatment threshold (≥10%) for WHO charts may help achieve clinical equivalence with Framingham. Non-laboratory models demonstrated high reliability and offer a viable alternative for risk screening in resource-constrained settings.