Predictive performance of the Framingham risk score in obstructive sleep apnea cohorts: insights from the SAVE, RICCADSA, and ISAACC trials

The Framingham Risk Score (FRS) is a widely used clinical tool for estimating 10-year cardiovascular (CV) risk based on traditional risk factors. In patients with obstructive sleep apnea (OSA) and established cardiovascular disease, where disease-specific CV risk tools are lacking, clinicians often use general primary-prevention risk tools such as the FRS or the ACC/AHA ASCVD Risk Score, even though these tools are not formally validated for secondary-prevention risk estimation. The purpose of this study is to evaluate the predictive performance of the FRS and ACC/AHA ASCVD Risk Score for CV events and mortality in patients with OSA enrolled in secondary-prevention CPAP trials, and to determine whether the FRS can guide personalized treatment strategies in OSA. We conducted an individual participant data meta-analysis of three randomized controlled trials (SAVE, ISAACC, RICCADSA) including 3,676 adults with OSA and established CV disease. Associations between risk scores and outcomes were assessed using logistic regression and Cox proportional hazards models with mixed effects to account for between-study variability. Predictive performance was evaluated using area under the receiver operating characteristic curve (AUC). Interaction analyses were performed to determine whether baseline CV risk modified CPAP treatment effects. Higher FRS was significantly associated with increased risk of CV events across all trials (meta-analysis OR = 1.03 per point, p < 0.001). However, predictive discrimination was modest (AUC range: 0.56–0.62). The ASCVD score showed similar patterns. The apnea-hypopnea index (AHI) showed no consistent association with outcomes and did not improve predictive performance when added to the risk scores. Addition of the oxygen desaturation index (ODI) failed to produce consistent gains in CV prediction. The FRS did not identify subgroups with differential benefit from CPAP (interaction p = 0.56 for CV events and p = 0.15 for CV death). General CV risk scores demonstrate limited predictive accuracy in patients with OSA and do not inform CPAP treatment decisions. These findings underscore the need for OSA-specific, mechanism-based risk stratification approaches to enable precise prediction of CV risk and individualized treatment strategies. Not applicable. This study is a secondary analysis of previously registered trials: SAVE (NCT00738179), ISAACC (NCT01335087), and RICCADSA (NCT00519597).