The Diamond-Forrester (DF) algorithm overestimates the likelihood of significant coronary artery disease (≥50% stenosis, CAD50). The aim of the present study was to evaluate whether the addition of a diagonal earlobe crease (DELC) enhances the predictive ability of DF to detect CAD50 by coronary computed tomographic angiography (CTA). We evaluated 430 patients referred for CTA for symptoms, cardiovascular risk factors, and CAD50 likelihood using DF. Observers blinded to CTA findings evaluated the presence of DELC. The diagnostic accuracy and relation of DF, DELC, and DF + DELC for predicting CAD50 in patients with chest pain were evaluated using receiver operating characteristics curve (area under curve) analyses and multivariate logistic regression analyses. In 199 patients with chest pain, the sensitivity and specificity for CAD50 were 96% and 20% for DF (AUC 0.59, p = 0.59), 91% and 32% for DELC (AUC 0.62, p = 0.03), and 91% and 41% for DF + DELC (AUC 0.66, p = 0.004). On multivariate analyses DELC was the only independent predictor of CAD50 (odds ratio 3.6, 95% confidence interval 1 to 12.9, p = 0.048). DF + DELC increased the predictive ability to detect CAD50 above cardiovascular risk factors (odds ratio 5.6, 95% confidence interval 1.6 to 19.8, p = 0.007). In patients with chest pain, the presence of DELC is related to CAD50 beyond DF. A combined variable of DF + DELC provides superior discriminatory ability for detecting CAD50 than either method alone.
Recent studies have demonstrated overestimation of stenotic coronary artery disease (≥50% stenosis, CAD50) prevalence when using standard algorithms for estimating its pretest probability. One of these algorithms is the Diamond-Forrester (DF) algorithm, which predicts the likelihood of CAD50, based on typicality of chest pain, age, and gender. These data suggest that it may be valuable to refine estimates of CAD50 by incorporating additional components of the clinical presentation and physical examination. One simple clinical sign that may be a potential candidate is the presence of a diagonal earlobe crease (DELC; Figure 1 ). Although this was first associated with coronary artery disease (CAD) almost 4 decades ago, its popularity has since waned, owing to controversy regarding its relation to CAD. Despite this, there is now emerging evidence that this simple clinical sign may be related to the presence, extent, and severity of CAD. The aim of the present study was therefore to evaluate whether the addition of a DELC finding to the standard DF algorithm (DF + DELC) enhances its predictive ability in detecting CAD50 as determined by coronary computed tomography angiography (CTA).
Methods
We prospectively studied 459 consecutive patients who underwent coronary CTA at our hospital over a 9-month period. All patients underwent a detailed assessment of symptoms, cardiovascular risk factors, CAD50 risk using the DF classification system, and the presence of DELC before the coronary CTA. Patients were subsequently excluded from the study if they had a history of CAD (myocardial infarction, coronary stenting, and previous bypass surgery) and if an expert reader did not consider the coronary CTA image quality to be good or excellent. Written informed consent was obtained for all patients and the study was approved by our institutional review board.
After a clinical history, patients were dichotomously divided into those having chest pain or not. For those with chest pain, typical angina pectoris was rigidly defined as: (1) substernal, jaw, or arm pressure-like pain, (2) induced by exertion, and (3) resolved with rest or use of nitroglycerin. Atypical angina pectoris was defined as 2 of the previously mentioned features. Nonanginal chest pain was defined as 1 or none of the previously mentioned features.
Cardiovascular risk factors were determined by preset criteria. Hypertension was defined as a systolic blood pressure of >140 mm Hg, a diastolic blood pressure of >90 mm Hg, or antihypertensive drug use. Smoking status was defined as a current smoker or past heavy smoker (>20 pack-years). Diabetes mellitus was defined as a previously established diagnosis, insulin or oral hypoglycemic therapy, fasting glucose of >126 mg/dl, or nonfasting glucose of >200 mg/dl. Family history of CAD was defined as myocardial infarction, coronary revascularization, or sudden cardiac death in a first-degree male relative <55 years of age or female relative <65 years of age.
The pretest probability of CAD50 was calculated using the original DF table of probabilities (generating a “DF probability”) and treated as a categorical variable. Patients >69 years of age were assigned the expected pretest probability of the oldest age group in the algorithm. Those who were asymptomatic were also included in this study, based on their probability of having CAD50 according to extended tables published by the American Heart Association in its guidelines. Patients with “intermediate” or “high” DF probability were considered suspected of having CAD50.
The presence of a DELC was determined by consensus by 2 trained observers before coronary CTA. A DELC was defined as a wrinkle-like line extending diagonally from the tragus across the lobule to the rear edge of the auricle of the ear, not related to sleeping position or wearing earrings. The observers were blinded to the clinical history or any other previous cardiac imaging results. The overall interobserver agreement for the presence of a DELC was 97.4%.
Coronary CTA was performed on all patients using a SOMATOM Definition dual-source scanner (Siemens Medical Systems, Forchheim, Germany). Imaging protocol has been previously described in detail. Beta blockade with oral and/or intravenous metoprolol was used whenever heart rate at rest exceeded 70 beats/min. All patients received 0.4 mg of sublingual nitroglycerin, 3 to 5 minutes before scanning. Patients were scanned after injection of 80 ml of iodinated contrast at 5 to 6 ml/second, triggered by >100 Hounsfield units in the ascending aorta, in a single breath hold. Scanning parameters included heart rate–dependent pitch (0.2 to 0.45 detector width per rotation), 330 ms gantry rotation time, 100 or 120 kVp tube voltage depending on patient body mass index, and 330 to 350 mA reference tube current. The acquired coronary CTA data were reconstructed in mid-diastole and at end-systole using 0.6 mm slice-thickness (0.75 mm if body mass index was >35 kg/m 2 ), 0.3 mm slice increment, 250 mm field-of-view, 512 × 512 matrix, and B26f “medium-smooth” kernel. If reconstruction from standard phases of the cardiac cycle resulted in uninterpretable segments, additional phases were reconstructed and analyzed. Reconstructed data were transferred to a Siemens workstation (LEONARDO, Siemens Medical solutions, Forchheim, Germany) for interpretation.
For the purposes of this research, coronary CTA interpretation was performed by 2 American Heart Association level-3 expert readers who were blinded to the presence or absence of DELC, using the modified American Heart Association 15 segment coronary artery tree model. Presence of CAD50 was recorded for each coronary artery segment, using technique previously described by our group. Consensus interpretation was obtained to resolve discrepancies between readers and the Cohen’s Kappa coefficient for the interobserver agreement was >0.9. Researchers assessing for the presence of a DELC were blinded to coronary CTA findings.
All continuous variables included in the analysis are presented as mean ± SD. Variables with non-normal distributions are presented as median with range. Univariate analyses were performed on continuous variables using the 2-sample t test for normally distributed variables and the Mann-Whitney U test for non-normally distributed data. Receiver operating characteristic curves were used to derive the sensitivity, specificity, positive predictive value, negative value, and overall accuracy for DF, DELC, and DF + DELC in predicting the presence of CAD50. Multivariate logistic regression was used to determine the independent predictors of CAD50 in the whole population and in patients with chest pain after adjusting for DF, DELC, and conventional cardiovascular risk factors (smoking, hypertension, hypercholesterolemia, diabetes mellitus, and a family history of CAD). A second model was also constructed to determine whether DF + DELC was an independent predictor of CAD50 after adjusting for the same cardiovascular risk factors. Statistical significance for all analyses was set at the 5% level. All data were collected and analyzed using SPSS for MAC, version 21 (IBM, Somers, New York).
Results
From the preliminary cohort of 459 patients, 29 patients were excluded owing to a history of CAD or less than good image quality. This left a final study cohort of 430 patients. The primary indication for the coronary CTA was chest pain in 199 (46%), a previous equivocal functional test in 146 (34%), as an assessment pre–noncardiac surgery in 46 (11%), and multiple cardiovascular risk factors in 39 patients (9%). The demographic characteristics for both cohorts are listed in Table 1 .
| Variable | Entire cohort (n = 430) | Chest pain (n = 199) |
|---|---|---|
| DELC | 307 (71%) | 143 (72%) |
| Age (years) | 61±13 | 61±14 |
| Men | 264 (61%) | 105 (53%) |
| Hypertension | 244 (57%) | 114 (57%) |
| Hyperlipidemia | 271 (63%) | 127 (64%) |
| Smokers | 162 (38%) | 74 (37%) |
| Diabetes mellitus | 65 (15%) | 38 (19%) |
| CAD family history | 131 (30%) | 60 (30%) |
| >50% stenosis | 71 (17%) | 34 (17%) |
| Total cholesterol (mg/dL) | 166±40 | 168±40 |
| LDL cholesterol (mg/dL) | 96±33 | 96±32 |
| HDL cholesterol (mg/dL) | 48±18 | 48±19 |
| Triglycerides (mg/dL) | 114±83 | 122±82 |
| Glucose (mg/dL) | 94±29 | 95±30 |
Of the 430 patients who underwent coronary CTA, 71 patients (17%) had CAD50. The DF classification correctly identified 34 of these patients (48%), but incorrectly attributed an intermediate-high likelihood of CAD50 in 132 of 359 patients (37%). In the 199 patients with chest pain, 34 (17%) had CAD50. The DF classification correctly identified 33 of these patients (97%), but incorrectly attributed an intermediate-high likelihood of CAD50 in 132 (80%) of 165 patients. The overall diagnostic accuracy of the DF classification for predicting CAD50 in the entire cohort and in patients with chest pain is listed in Table 2 .
| DF | DELC | DF+DELC | |
|---|---|---|---|
| Entire cohort (n = 430) | |||
| Sensitivity | 48% | 90% | 45% |
| Specificity | 63% | 32% | 72% |
| Positive predictive value | 20% | 21% | 25% |
| Negative predictive value | 86% | 94% | 87% |
| Positive likelihood ratio | 1.29 | 1.32 | 1.6 |
| Negative likelihood ratio | 0.83 | 0.31 | 0.76 |
| Accuracy | 61% | 42% | 68% |
| AUC | 0.56 | 0.61 | 0.64 |
| p | 0.14 | 0.003 ∗ | <0.001 ∗ |
| Patients with chest pain (n = 199) | |||
| Sensitivity | 97% | 91% | 91% |
| Specificity | 20% | 32% | 41% |
| Positive predictive value | 20% | 22% | 24% |
| Negative predictive value | 97% | 95% | 96% |
| Positive likelihood ratio | 1.21 | 1.34 | 1.54 |
| Negative likelihood ratio | 0.15 | 0.28 | 0.22 |
| Accuracy | 33% | 42% | 49% |
| AUC | 0.59 | 0.62 | 0.66 |
| p | 0.12 | 0.03 ∗ | 0.004 ∗ |
The presence of a DELC correctly identified CAD50 in 64 of the 71 patients (90%) from the entire 430 patient cohort, but incorrectly attributed an intermediate-high likelihood of CAD50 in 243 patients (68%). In the 199 patients with chest pain DELC correctly identified 31 of the 34 patients (91%) with CAD50, but incorrectly attributed an intermediate-high likelihood of CAD50 in 112 patients (68%). The overall diagnostic accuracy of the DF classification for predicting CAD50 in the entire cohort and in patients with chest pain is listed in Table 2 .
The relation between DELC and CAD50 was consistent irrespective of gender. From the initial cohort of 430 patients there were 166 women. Of the 23 women with CAD50, 22 had a DELC (p = 0.004). Similarly there were 264 men. Of the 48 men with CAD50, 42 had a DELC (p = 0.009).
On multivariate logistic regression analyses with CAD50 as the dependent variable, only a DELC (odds ratio [OR] 3.7, 95% confidence interval [CI] 1.6 to 8.4, p = 0.002), hyperlipidemia (OR 2.1, 95% CI 1.1 to 4.0, p = 0.02), and a smoking history (OR 2.0, 95% CI 1.2 to 3.4, p = 0.01) were independent predictors of CAD50. In patients with symptoms of chest pain, on multivariate analyses, only a DELC remained as an independent predictor of having CAD50 (OR 3.6, 95% CI 1.0 to 12.9, p = 0.048; Table 3 ).
| Variable | Odds Ratio | 95% Confidence Intervals | p |
|---|---|---|---|
| Entire Cohort (n = 430) | |||
| DELC | 3.7 | 1.6 – 8.4 | 0.002 ∗ |
| DF category | 1.2 | 0.7 – 2.1 | 0.49 |
| Hypertension | 1.3 | 0.7 – 2.3 | 0.38 |
| Hypercholesterolemia | 2.1 | 1.1 – 4.0 | 0.02 ∗ |
| Smokers | 2.0 | 1.2 – 3.4 | 0.01 ∗ |
| Diabetes mellitus | 1.9 | 1.0 – 3.6 | 0.07 |
| Family History | 0.9 | 0.5 – 1.6 | 0.69 |
| Chest pain (n = 199) | |||
| DELC | 3.6 | 1.0 – 12.9 | 0.048 ∗ |
| DF category | 1.7 | 0.7 – 3.8 | 0.22 |
| Hypertension | 1.3 | 0.6 – 3.1 | 0.53 |
| Hypercholesterolemia | 1.8 | 0.7 – 4.9 | 0.22 |
| Smokers | 1.7 | 0.8 – 3.8 | 0.18 |
| Diabetes mellitus | 2.0 | 0.8 – 4.8 | 0.14 |
| Family History | 0.9 | 0.4 – 2.1 | 0.78 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
