Prediction of cardiac and noncardiac mortality after percutaneous coronary intervention

Background-Current risk models for predicting long-term mortality after percutaneous coronary intervention are restricted to all-cause mortality. We sought to develop novel risk models for the prediction of cardiac and noncardiac mortality after percutaneous coronary intervention. Methods and Results-We retrospectively evaluated patients who underwent index percutaneous coronary intervention at Mayo Clinic from 2003 to 2008. Long-term deaths were ascertained through scheduled prospective surveillance. Cause of death was determined via telephone interviews, medical records, and autopsy reports. Fine and Gray extension of Cox proportional hazards models was used to model cause-specific cumulative incidence. Candidate variables and interactions were chosen a priori, without variable selection methods. Resulting models were mapped to an integer-based risk score. The study comprised 6636 patients followed up over a median of 62 months (25th, 75th percentiles: 45, 77 months). There were 1488 deaths, 518 (35%) cardiac, 938 (63%) noncardiac, and 32 (2%) unknown. The 5-year predicted cardiac mortality ranged from 0.6% to 97%, with a corrected c-statistic of 0.82. Risk factors for cardiac death included age, body mass index, ejection fraction, and history of congestive heart failure. The integer score for noncardiac death included age, medicine index, body mass index, current smoker, noncardiac Charlson index and cardiac Charlson index, and accommodated significant age-based interactions for smoking and the 2 Charlson indices. Predicted noncardiac mortality at 5 years ranged from 0.2% to 81%, with a corrected c-statistic of 0.77. Conclusions-We report novel risk models to predict cardiac and noncardiac long-term mortality after percutaneous coronary intervention.