Determining the pretest probability of angiographically significant (≥50% stenosis) coronary artery disease (CAD) in symptomatic patients relies on the Diamond and Forrester (DF) classification, which was derived from a cohort referred for invasive coronary angiography. The accuracy of this approach in patients referred for noninvasive coronary angiography is not fully known. Consecutive patients without known CAD referred for coronary computed tomographic angiography (CCTA) were evaluated. Chest pain was prospectively categorized as nonanginal, atypical angina, typical angina, or asymptomatic. The pretest likelihood of angiographically significant CAD was estimated using DF classification and compared with observed rates of angiographically significant CAD on CCTA. Among 1,027 patients (41% women; mean age 50 ± 12 years), 38 (4%) had nonanginal symptoms, 643 (63%) had atypical angina, 72 (7%) had typical angina, and 274 (26%) were asymptomatic. The prevalence of angiographically significant CAD in patients with nonanginal chest pain, atypical angina, typical angina, and no symptoms was 1 (3%), 55 (9%), 14 (19%), and 25 (9%), respectively (p <0.001). DF classification significantly overestimated angiographically significant CAD prevalence across all symptom classifications, genders, and ages despite adjustment for risk factors (p <0.001 for all comparisons). DF classification had an area under the receiver-operating characteristic curve of 0.72 (95% confidence interval 0.66 to 0.78), which was not significantly different from age alone (0.69) or age, symptoms, and risk factors (0.68). In conclusion, in a low- to intermediate-risk cohort referred for CCTA, DF classification significantly overestimated angiographically significant CAD prevalence across all age, gender, and symptom strata. The DF classification may overestimate the pretest probability of angiographically significant CAD in contemporary patients referred for CCTA.
Current guidelines for the evaluation of patients with chest pain dictate further testing on the basis of the pretest estimation of the probability of angiographically significant (≥50% stenosis) coronary artery disease (CAD) using the method reported by Diamond and Forrester (DF). This commonly used and easily applied system, first described in 1979, incorporates age, gender, and symptom description and was derived from a cohort referred for invasive coronary angiography. Recent studies have shown that the DF criteria may overestimate the prevalence of angiographically significant CAD in contemporary patient populations, including those referred for noninvasive coronary computed tomographic angiography (CCTA), but additional studies in this area are required to corroborate these important findings. We sought to compare the expected prevalence of angiographically significant CAD predicted by DF classification with the observed prevalence of angiographically significant CAD in patients clinically referred for 64-slice CCTA.
Methods
We evaluated 1,027 consecutive patients without known CAD who were referred for CCTA at Walter Reed Army Medical Center (Washington, District of Columbia) from July 2006 to December 2010. Patient symptoms of chest pain were prospectively classified according to the DF criteria as nonanginal, atypical angina, or typical angina. Typical angina was defined as possessing the following 3 characteristics: (1) substernal location, (2) occurs with exertion or emotional stress, and (3) is consistently relieved with rest or nitroglycerin. Atypical angina was defined by having 2 of the aforementioned criteria, and chest pain possessing <2 of the criteria was defined as nonanginal. Patients presenting with dyspnea without chest pain or without chest pain or dyspnea (asymptomatic) were analyzed separately. We determined the pretest probability of angiographically significant CAD using the age, gender, and symptom table from the American College of Cardiology and American Heart Association guidelines for the management of patients with chronic stable angina, which are based primarily on the DF criteria.
Each coronary computed tomographic angiographic examination was performed on the same 64-slice scanner (LightSpeed VCT; GE Medical Systems, Waukesha, Wisconsin) in accordance with Society of Cardiovascular Computed Tomography guidelines. All scans were jointly interpreted by a trained (American College of Cardiology competency level II or III), credentialed cardiologist and radiologist who reached consensus. Maximal epicardial vessel luminal stenosis was visually estimated, with patients categorized as having (1) normal coronary arteries, (2) nonobstructive CAD (<50% stenosis), or (3) ≥50% visual luminal stenosis in ≥1 epicardial coronary artery segment ≥1.5 mm in diameter (angiographically significant CAD).
Continuous variables were compared using Student’s t tests or Mann Whitney U tests, as appropriate, and are expressed as mean ± SD. Categorical variables were compared using chi-square tests and are presented as frequencies with percentages. Receiver-operating characteristic curves were constructed to predict the presence of angiographically significant CAD on CCTA, comparing age versus symptoms (DF criteria) versus the Morise score, which incorporates age, risk factors, and DF criteria symptoms. Two-tailed p values <0.05 were considered statistically significant.
Results
Baseline characteristics of the study cohort are listed in Table 1 . Atypical angina (63%) was the most common symptom classification prompting referral for CCTA. The cohort was generally at low to intermediate pretest risk for angiographically significant CAD, with an overall expected prevalence of angiographically significant CAD (stenosis ≥50%) according to the DF method of 32%. The expected prevalence of angiographically significant CAD when stratified by gender was 38% for men and 23% for women. The overall observed prevalence of angiographically significant CAD compared with the predicted prevalence of angiographically significant CAD using the DF classification is listed in Table 2 . Regardless of stratification by typicality of the presenting symptoms, DF classification significantly overestimated the prevalence of angiographically significant CAD (p <0.001). Typical angina predicted the highest prevalence of angiographically significant CAD (19%) but was significantly lower than the expected rate of angiographically significant CAD (81%) in patients with this type of chest pain (p <0.001).
| Variable | Value |
|---|---|
| Men | 606 (59%) |
| Age (yrs) | 50 ± 12 |
| Diabetes mellitus | 112 (10%) |
| Hyperlipidemia (patient identified or treated) | 562 (51%) |
| Smokers | 135 (12%) |
| Family history (premature coronary heart disease) | 290 (26%) |
| Hypertension | 562 (51%) |
| Body mass index (kg/m 2 ) | 29 ± 5 |
| Total cholesterol (mg/dl) | 190 ± 38 |
| Low-density lipoprotein cholesterol (mg/dl) | 116 ± 33 |
| High-density lipoprotein cholesterol (mg/dl) | 53 ± 21 |
| Category | Observed | Predicted |
|---|---|---|
| Nonanginal chest pain | 1 (3%) | 6 (15%) ∗ |
| Atypical angina | 55 (9%) | 264 (41%) ∗ |
| Typical angina | 14 (19%) | 58 (81%) ∗ |
When patients were stratified into their appropriate age, gender, and symptom categories, DF classification significantly overestimated the prevalence of angiographically significant CAD across all categories (p <0.001; Figure 1 ). Patients with typical angina had a higher prevalence of angiographically significant CAD among all age and gender subgroups. When adjusted for the major cardiovascular risk factors hypertension, hyperlipidemia, and smoking, DF classification significantly overestimated the prevalence of obstructive CAD (p <0.001). For the prediction of any angiographically significant CAD, DF classification had an area under the curve of 0.72 (95% confidence interval 0.66 to 0.78, p <0.001) on receiver-operating characteristic curve analysis ( Figure 2 ). Incorporating standard cardiovascular risk factors using the Morise score for the prediction of angiographically significant CAD, the area under the curve was 0.68 (95% confidence interval 0.63 to 0.74), whereas age alone had an area under the curve of 0.69 (95% confidence interval 0.63 to 0.75) ( Figure 2 ). These 3 areas under the curve were not statistically different.
