Annals of the American Thoracic Society · Mar 26, 2026
Performance of cardiovascular disease risk prediction tools in chronic obstructive pulmonary disease
By Salimian S, Hawkins NM, Amegadzie JE et al.
RATIONALE: Emerging evidence suggests that a widely used risk scoring tool, QRISK3, substantially underestimates cardiovascular disease risk in patients with chronic obstructive pulmonary disease (COPD), raising concerns about the validity of comparable risk assessment tools used in North America. OBJECTIVES: We examined the performance of 3 risk equations-simplified Predicting Risk of Cardiovascular Disease EVENTs (PREVENT), Pooled Cohort Equations, and the 2008 global Framingham Risk Score to estimate 10-year total cardiovascular disease risk in individuals with COPD. METHODS: Individuals ≥ 40 years of age with COPD were identified from 5 longitudinal, community-based epidemiologic North American cohort studies. The risk was derived from each model using model-specific definitions with no major differences in the setting, time horizon, outcome, or predictors with those used in the original model development studies, except for the COPD eligibility criteria. Discrimination (using time-dependent area under the receiver operating characteristic curve), calibration (using observed to the average estimated risk ratio [O/E] and calibration plot), and clinical utility (using decision curve) were assessed. RESULTS: PREVENT demonstrated the highest discrimination: 0.76 (95% CI, 0.74, 0.77) followed by Pooled 0.66 (0.64, 0.69) and Framingham 0.57 (0.54, 0.59). The PREVENT underestimated risk (O/E: 1.25 [95% CI, 1.2, 1.3]), whereas the Pooled and Framingham overestimated risk, by approximately 30% (0.73 [0.67, 0.80] and 0.67 [0.62, 0.72]). These discrepancies varied by age and sex, with a more pronounced underestimation in younger adults with PREVENT and overestimation in older adults with Pooled. All 3 models demonstrated clinical utility across a range of risk thresholds. CONCLUSIONS: The models exhibit variable levels of miscalibration but retain clinical utility. With further calibration, their accuracy and predictive power may be improved.
doi.org/10.1093/annalsats/aaoag073Open accessPubMed 41883123Biomarkers