Survival prediction models for coronary intervention: strategic decision support.

Abstract

For a given patient with coronary artery disease, it is uncertain which therapy, percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG), maximizes long-term survival. Hence, we developed survival models for CABG and PCI using bare-metal stents (BMS) or drug-eluting stents (DES), programmed a decision-support tool, and identified its potential usefulness. From 1995 to 2007, 23,182 patients underwent primary isolated CABG (n=13,114) or first-time PCI with BMS (n=6,964) or DES (n=3,104). Follow-up was 6.3±3.9 years. Survival models were developed independently for each therapy, then all factors appearing in any of the three models were forced into a final model for each. These were programmed into a decision-support tool. Predicted differences in 5-year survival for the same patient among the three therapies were calculated. Unadjusted survival was 96%, 86%, and 68% at 1, 5, and 10 years after CABG, 94%, 83%, and 68% after BMS, and 95% and 84% (no 10-year estimate) after DES, respectively. Risk factors for early and mid-term mortality were identified, leading to variable-rich (25 variables) prediction models. Patients most likely to experience a 5-year survival benefit from DES were those undergoing emergency revascularization for acute infarction, and patients most likely to benefit from CABG had extensive coronary artery disease and numerous comorbidities. Detailed prediction models for prognosis after PCI and CABG are useful for developing a clinically relevant, strategic decision-support tool that reveals who may experience a long-term survival benefit from each modality.