A clinical risk score for prediction of stent thrombosis

Abstract

The aim was to develop a clinically useful patient risk score predictive for stent thrombosis (ST). Using readily available baseline clinical and angiographic characteristics, a Cox proportional hazards multivariate model was used to identify significant (p <0.10) predictors of ST through 1 year in 2,487 patients receiving a TAXUS Express (Boston Scientific Corp., Natick, Massachusetts) drug-eluting stent (DES) in the ARRIVE 1 registry. Hazard ratios of significant predictors were rounded to an integer value ranging from 2 to 5. These values were summed for a maximum possible score of 24. The model was validated using 1-year data from a similar DES data set (ARRIVE 2, n = 4,820 patients). The 8 significant predictors found were thienopyridine therapy discontinuation before 6 months, insulin-requiring diabetes, smoker at baseline, left main stent placement, multiple stent placement, lesion length >28 mm, moderate to severe lesion calcification, and reference vessel diameter <3 mm. Model discrimination was high, indicated by an area under the receiver-operator characteristic curve of 0.819. Stratification of patients into low-, medium-, and high-risk groups showed that ST developed in 0.8% of patients with a score <6, 3.6% of patients with a score of 7 to 13, and 12.6% of patients with a score >or=14. In conclusion, using 8 readily available clinical and angiographic characteristics, we defined an ST risk score for patients receiving a DES during the first year. Analysis of patients from ARRIVE 1 and 2 showed that most (73%) were in the lowest risk category, with 25% in the moderate risk category. Less than 2% were at highest risk of developing ST.