Paradigm of pretest risk stratification before coronary computed tomography.

BACKGROUND

The optimal method of determining the pretest risk of coronary artery disease as a patient selection tool before coronary multidetector computed tomography (MDCT) is unknown.

OBJECTIVE

We investigated the ability of 3 different clinical risk scores to predict the outcome of coronary MDCT.

METHODS

This was a retrospective study of 551 patients consecutively referred for coronary MDCT on a suspicion of coronary artery disease. Diamond-Forrester, Duke, and Morise risk models were used to predict coronary artery stenosis (>50%) as assessed by coronary MDCT. The models were compared by receiver operating characteristic analysis. The distribution of low-, intermediate-, and high-risk persons, respectively, was established and compared for each of the 3 risk models.

RESULTS

Overall, all risk prediction models performed equally well. However, the Duke risk model classified the low-risk patients more correctly than did the other models (P < 0.01). In patients without coronary artery calcification (CAC), the predictive value of the Duke risk model was superior to the other risk models (P < 0.05). Currently available risk prediction models seem to perform better in patients without CAC. Between the risk prediction models, there was a significant discrepancy in the distribution of patients at low, intermediate, or high risk (P < 0.01).

CONCLUSIONS

The 3 risk prediction models perform equally well, although the Duke risk score may have advantages in subsets of patients. The choice of risk prediction model affects the referral pattern to MDCT.