Highlights
- External validation of multiple cardiovascular disease (CVD) risk scores in a large Mexican cohort showed high discriminatory power but poor calibration.
- All tested models, including Framingham, Globorisk, and WHO, consistently overpredicted the 10-year risk of fatal CVD, particularly among women.
- Recalibration of the Globorisk-fatal model significantly improved its predictive accuracy, aligning estimated risks with observed clinical outcomes in the Mexico City population.
- The study underscores the necessity of using population-specific or recalibrated tools to avoid over-medicalization and ensure efficient healthcare resource allocation.
The Burden of Cardiovascular Disease in Latin America
Cardiovascular disease (CVD) remains the leading cause of mortality across Latin America, presenting a formidable challenge to public health systems already strained by varying socioeconomic conditions and an epidemiological transition. Despite the high prevalence of risk factors such as obesity, diabetes, and hypertension in the region, clinical practice has long relied on risk prediction equations developed in high-income countries, primarily in North America and Europe. These tools, such as the Framingham Risk Score or the SCORE2 equation, are fundamental for identifying high-risk individuals who require intensive preventive interventions, including statin therapy and aggressive blood pressure management.
However, the generalizability of these models to diverse populations remains a point of contention. Genetic backgrounds, lifestyle factors, and healthcare access differ significantly between the populations where these models were derived and the populations where they are applied. In Mexico, where the burden of metabolic disease is unique, the potential for these global models to miscalculate risk—either through underestimation or overestimation—carries significant clinical and economic implications. Overestimation can lead to unnecessary pharmacological treatment, while underestimation might result in missed opportunities for life-saving preventive care.
Study Design: The Mexico City Prospective Study
To address this gap, researchers conducted a comprehensive external validation and recalibration study using data from the Mexico City Prospective Study. This analysis involved 112,262 adults aged 40 years and older who were free of cardiovascular disease at baseline. The cohort was predominantly female (67%), with a mean age of 54.6 years. This prospective, observational, population-based analysis provided a robust dataset to evaluate how well established risk scores perform in a real-world Mexican setting.
The primary endpoint was the 10-year risk of fatal CVD, which included fatal myocardial infarction (MI) and stroke. Researchers evaluated several prominent risk prediction models, including the laboratory-based and office-based versions of the Framingham score, Globorisk, Globorisk-LAC (specifically designed for Latin America and the Caribbean), the World Health Organization (WHO) scores, and the SCORE2 equations. The evaluation focused on two key performance metrics: discrimination (the ability of the model to differentiate between those who will and will not have an event) and calibration (the agreement between the predicted probability and the observed frequency of events).
Key Findings: The Overestimation Gap
Over a median follow-up period of 10 years, 2,429 fatal CVD events were recorded, consisting of 1,667 myocardial infarctions and 762 strokes. The analysis of these events yielded several critical insights for clinicians and policy experts.
Discrimination Performance
In terms of discrimination, all models performed remarkably well. Harrell’s c-statistics, which measure the model’s ability to rank patients correctly by risk, ranged from 0.761 to 0.805 in males and from 0.797 to 0.831 in females. Notably, the Globorisk-fatal model demonstrated the highest discrimination in women (0.831, 95% CI 0.821-0.841), while the laboratory-based WHO-MI model performed best in men (0.805, 95% CI 0.783-0.827). These high c-statistics suggest that the variables included in these models—such as age, systolic blood pressure, smoking status, and cholesterol levels—are indeed strong predictors of cardiovascular mortality in the Mexican population.
The Calibration Challenge
While the models were excellent at ranking patients, they failed significantly in calibration. Every model tested consistently overestimated the actual risk of fatal CVD. This trend was particularly pronounced in women and at higher risk deciles. For instance, the predicted risk curves often sat significantly above the observed risk line, indicating that the ‘global’ equations were predicting many more deaths than actually occurred in the Mexico City cohort. This systematic overprediction suggests that the baseline risk constants in these models do not align with the current epidemiological reality of Mexico City.
Recalibration: A Path Toward Precision
Recognizing the poor calibration of the global models, the researchers performed a sex-specific recalibration of the Globorisk-fatal model. Recalibration involves adjusting the model’s intercept or baseline hazard to better reflect the average risk in the target population while maintaining the relative weights of the individual risk factors.
The results of this recalibration were striking. The recalibrated Globorisk-fatal model showed a much tighter agreement between predicted and observed risks across all risk categories. By adjusting the model to account for the specific mortality rates observed in the Mexican cohort, the researchers were able to create a tool that is significantly more reliable for clinical decision-making in this region. This recalibrated tool provides a more accurate foundation for determining which patients truly meet the threshold for intervention based on their absolute 10-year risk.
Clinical Implications and Expert Commentary
The findings of this study have immediate implications for clinical practice in Mexico and potentially other Latin American nations. The consistent overestimation of risk by standard scores suggests that many Mexican patients may be classified as ‘high risk’ and prescribed preventive medications like statins when their actual absolute risk is relatively low. This not only increases healthcare costs but also exposes patients to potential side effects without a proportional benefit in risk reduction.
Experts in the field note that the high discrimination of these scores is encouraging, as it confirms that the traditional risk factors remain the primary drivers of CVD. However, the calibration failure highlights the ‘calibration gap’ often seen in middle-income countries undergoing rapid health transitions. Factors such as improvements in acute care for MI and stroke, or regional differences in dietary patterns and physical activity, may explain why observed mortality is lower than what models based on older or different cohorts would predict.
Furthermore, the study highlights a common issue in medical research: the underrepresentation of diverse populations in the development of clinical tools. As we move toward an era of personalized and precision medicine, the use of ‘one-size-fits-all’ risk equations is increasingly viewed as inadequate. The success of the recalibrated Globorisk-fatal model serves as a call to action for local health authorities to support the development and validation of population-specific risk assessment tools.
Conclusion
The external validation of CVD risk scores in Mexico City provides a clear message: while global models are useful for identifying which patients are at higher risk relative to others, they are currently inaccurate in predicting the absolute risk of fatal events in the Mexican population. The recalibrated Globorisk-fatal model offers a superior alternative for clinicians, providing a more accurate assessment that can guide better-informed treatment decisions. Future research should focus on validating these recalibrated tools in other regions of Mexico and exploring whether the addition of novel biomarkers or social determinants of health could further refine these predictions.
Funding and Clinical Data
This research was supported by the Instituto Nacional de Geriatría in Mexico. The study utilized data from the Mexico City Prospective Study, a significant longitudinal effort designed to understand the health trajectories of adults in one of the world’s largest urban environments.
References
1. Perezalonso-Espinosa J, et al. External validation and recalibration of cardiovascular risk scores for prediction of 10-year risk of fatal cardiovascular disease: a prospective, observational, population-based cohort analysis of adults in Mexico City. Lancet Reg Health Am. 2026 Feb 16;56:101403. doi: 10.1016/j.lana.2026.101403.
2. Global Burden of Disease 2019 Case Study. Cardiovascular Diseases in Latin America. The Lancet, 2020.
3. Lloyd-Jones DM. Cardiovascular risk prediction: basic concepts, current status, and future directions. Circulation. 2010;121(15):1768-1777.
