Department of Cardiology, Lillebælt Hospital Vejle, Kabbeltoft 25, DK-7100 Vejle, Denmark.
Atherosclerosis. 2012 Feb;220(2):557-62. doi: 10.1016/j.atherosclerosis.2011.11.027. Epub 2011 Nov 25.
To compare the performance of five risk models (Diamond-Forrester, the updated Diamond-Forrester, Morise, Duke, and a new model designated COronary Risk SCORE (CORSCORE) in predicting significant coronary artery disease (CAD) in patients with chest pain suggestive of stable angina pectoris.
Retrospective cohort for creation of CORSCORE by means of logistic regression analysis. Prospective cohort for validation of the five risk models using receiver operating characteristics (ROC) curve analysis, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Significant CAD was defined as lumen area diameter reduction ≥50% at coronary angiography. All risk models include information on age, sex, and symptoms. In addition the Duke, Morise, and CORSCORE models include information on tobacco use and hypercholesterolemia. Duke and Morise also include information on diabetes. History of myocardial infarction is used by the Duke and CORSCORE models whereas hypertension is included in the Morise and CORSCORE models. The Duke model includes information on electrocardiogram (ECG) changes and the Morise model includes information on family history, body mass index, obesity, and oestrogen status.
4781 retrospective and 633 prospective consecutive patients referred for coronary angiography were included. The area under the ROC for the updated Diamond-Forrester, Duke, and CORSCORE was significantly larger than for the Diamond-Forrester (p≤0.001). The IDI was significantly higher for the Duke as compared to all other models (p≤0.006).
The Duke, updated Diamond-Forrester, and CORSCORE risk models are most efficient in predicting CAD in a contemporary cohort of patients with symptoms suggestive of angina. The updated Diamond-Forrester may most operational in daily clinical practice since it is calculated from the lowest number of clinical variables.
比较 5 种风险模型（Diamond-Forrester、更新后的 Diamond-Forrester、Morise、Duke 和新的 Coronary Risk SCORE（CORSCORE）模型）在预测有稳定型心绞痛症状的胸痛患者中存在严重冠状动脉疾病（CAD）的表现。
采用逻辑回归分析方法进行 CORSCORE 模型的回顾性队列创建。采用受试者工作特征（ROC）曲线分析、净重新分类改善（NRI）和综合判别改善（IDI）对 5 种风险模型进行前瞻性队列验证。显著 CAD 定义为冠状动脉造影时管腔面积直径减少≥50%。所有风险模型均包含年龄、性别和症状的信息。此外，Duke、Morise 和 CORSCORE 模型还包含吸烟和高胆固醇血症的信息。Duke 和 Morise 模型还包含糖尿病的信息。心肌梗死史用于 Duke 和 CORSCORE 模型，而高血压包含在 Morise 和 CORSCORE 模型中。Duke 模型包含心电图（ECG）变化的信息，而 Morise 模型包含家族史、体重指数、肥胖和雌激素状况的信息。
纳入 4781 例回顾性和 633 例前瞻性连续冠状动脉造影患者。更新后的 Diamond-Forrester、Duke 和 CORSCORE 的 ROC 曲线下面积明显大于 Diamond-Forrester（p≤0.001）。与其他所有模型相比，Duke 的 IDI 显著更高（p≤0.006）。
在有胸痛症状的当代患者队列中，Duke、更新后的 Diamond-Forrester 和 CORSCORE 风险模型在预测 CAD 方面最为有效。更新后的 Diamond-Forrester 可能在日常临床实践中最具操作性，因为它是根据最少的临床变量计算出来的。
