Usefulness of the logistic clinical SYNTAX score for predicting 1-year mortality in patients undergoing percutaneous coronary intervention of the left main coronary artery.

OBJECTIVES

To externally validate the logistic clinical SYNTAX in patients undergoing percutaneous coronary intervention (PCI) of the left main coronary artery (LMCA).

BACKGROUND

A novel version of the clinical SYNTAX score (CSS) has been recently introduced to overcome the limitations of categorical risk scores for PCI by the incorporation of clinical variables and the SYNTAX score (SXscore) into a logistic formula for predicting 1-year mortality.

METHODS

Core and extended models of the logistic CSS (logCSS(core) and logCSS(ext)) were applied to 400 patients undergoing LMCA PCI, and their performance was compared with those of the standard CSS, SXscore, and age, creatinine, and ejection fraction (ACEF) score by different measures of discrimination and calibration.

RESULTS

A significant gradient in the distribution of 1-year mortality was noted with all the models (P < 0.001), with the SYNTAX and ACEF scores showing the lowest (0.64) and the highest (0.75) c-statistic, respectively. Based on Somers' D(xy) rank correlation, discrimination of both the logCSS(core) and the logCSS(ext) was numerically lower than that of the ACEF score. The logCSS(core) was miscalibrated toward underpredicting all-cause mortality in low-predicted probabilities, while the logCSS(ext) tended to underpredict in low-predicted probabilities and overpredict in high-predicted probabilities. Slope and intercept values reflected a better calibration ability of the logCSS(core) compared with the logCSS(ext). The global accuracy of the logCSS(core) was superior to that of the logCSS(ext) (Brier score 0.087 vs. 0.095).

CONCLUSIONS

A personalized approach to risk stratification of LMCA PCI with the logistic CSS is feasible and of potential clinical utility. However, in this study, the logistic CSS did not achieve superior discrimination compared with other categorical models, and suffered from mild to moderate miscalibration.