Coronary artery disease (CAD) is a major cause of systolic heart failure (HF). Identifying CAD as a cause of systolic HF has prognostic and treatment implications. Whether all patients with systolic HF of unclear etiology should undergo coronary angiography has been controversial. We sought to derive and validate a clinical prediction rule to exclude CAD as a cause of systolic HF. A derivation cohort was formed of consecutive patients who had undergone coronary angiography with a primary diagnosis of systolic HF of unclear etiology (ejection fraction <50%). Using multivariate logistic regression analysis, we derived a prediction rule for severe CAD (≥50% diameter stenosis in the left main, 3-vessel CAD, and 2-vessel CAD involving the proximal left anterior descending artery). The diagnostic performance of the defined prediction rule was prospectively validated in a separate cohort recruited from 2 institutions. Of the 124 patients in the derivation cohort, 27% had CAD, including 15% with severe CAD. The independent predictors of severe CAD included diabetes (odds ratio 5.1, p = 0.005), electrocardiographic Q waves or left bundle branch block (odds ratio 3.8, p = 0.02), and ≥2 nondiabetes risk factors: age (men ≥55 or women ≥65 years), dyslipidemia, hypertension, and tobacco use (odds ratio 4.8, p = 0.02). A prediction rule of having ≥1 independent predictor identified 97% of the patients with CAD and 100% of the patients with severe CAD. In the prospective validation cohort of 143 patients, the prediction rule had 98% sensitivity and 18% specificity for CAD but 100% sensitivity for severe CAD. In conclusion, a simple clinical prediction rule can accurately identify patients with CAD and eliminate the need for angiography in a substantial proportion of patients with systolic HF, with potentially significant cost savings and risk avoidance.
In many patients with congestive heart failure (HF) due to left ventricular systolic dysfunction HF, the etiology is apparent, such as previous myocardial infarction or valvular disease. However, a substantial number of patients have systolic HF of uncertain etiology. It is common practice to determine whether systolic HF is caused by coronary artery disease (CAD), because coronary artery bypass grafting (CABG) could offer symptomatic and survival benefits beyond medical therapy, particularly for severe disease. However, selecting patients with systolic HF of unclear etiology for coronary angiography has been controversial. The American College of Cardiology Foundation/American Heart Association guidelines consider coronary angiography to be appropriate for patients with angina pectoris or a history of myocardial infarction, because the likelihood of CAD in these patients is thought to be high and the survival benefit of CABG, based on decades-old data, appears to be limited to those who experience angina. However, many patients with significant CAD might not report angina nor have a clear history of previous infarction, and many patients with systolic HF due to nonischemic cardiomyopathy experience angina. Some cardiologists, therefore, have recommended performing angiography in all patients who are potential candidates for revascularization; however, this would be costly and has associated risks. This approach has been considered a class IIa indication in the American College of Cardiology Foundation/American Heart Association guidelines with level of evidence C (expert opinion), highlighting the lack of clinical data. Although selection of patients with systolic HF to undergo angiography and appropriate revascularization continues to be an important issue, no study has evaluated this. In the present investigation, we sought to derive and validate a simple clinical tool to help clinicians predict the absence of severe CAD and eliminate unnecessary angiographic procedures.
Methods
From a derivation cohort, we identified the clinical predictors of CAD in patients with systolic HF of unknown etiology and derived a simple clinical prediction rule that was prospectively validated in a separate validation cohort.
The cardiac catheterization database at John H. Stroger, Jr. Hospital of Cook County (Chicago, Illinois) was queried for all consecutive patients with a primary diagnosis of congestive HF who had undergone coronary angiography during a 5-year period (January 1, 1995 to December 31, 1999). The health records were reviewed to ensure that congestive HF had been the primary indication for coronary angiography. The exclusion criteria were (1) normal systolic function, defined as a left ventricular ejection fraction (LVEF) ≥50%; (2) known CAD (defined as a history of myocardial infarction, previous revascularization procedure, or previous angiographically documented CAD); (3) an etiology presumed to be responsible for congestive HF (e.g., severe valvular disease, Adriamycin-induced cardiomyopathy, or peripartum cardiomyopathy); and (4) another indication for performing coronary angiography (e.g., malignant arrhythmia or aortic root aneurysm). Patients with chest pain were included only if their primary diagnosis and dominant clinical symptoms were congestive HF.
The data collected included demographic information, coronary risk factors, the presence and character of chest pain, LVEF, serum creatinine, and coronary angiography (diameter stenosis percentage and the coronary segments involved). Significant CAD was defined ≥50% diameter stenosis by visual inspection in any major epicardial coronary vessel. Severe CAD was defined as significant CAD of the left main coronary artery, 3-vessel CAD, or 2-vessel CAD involving the proximal left anterior descending artery. The definition of severe CAD was determined from subsets of coronary anatomy that have been shown to have long-term survival benefit from CABG. Electrocardiograms taken immediately before angiography were evaluated for Q waves consistent with previous myocardial infarction or left bundle branch block (LBBB) precluding the assessment of Q waves by a cardiologist who was unaware of the patient’s clinical and angiographic findings.
Consecutive patients with a primary diagnosis of congestive HF who had undergone coronary angiography at John H. Stroger, Jr. Hospital of Cook County (formerly Cook County Hospital) and Rush University Medical Center (Chicago, Illinois) during a 2-year period (January 14, 2008 to March 15, 2010) were prospectively enrolled in a validation cohort. The inclusion and exclusion criteria were identical to those in the derivation cohort. The same data points described in the derivation cohort were tabulated prospectively but with the addition of intravascular ultrasound and quantitative coronary angiography software findings, which were available for the assessment of borderline lesions.
From the derivation cohort, a stepwise backward conditional elimination multivariate logistic regression analysis was used to identify the independent predictors of severe CAD. The elimination criterion at each step was p >0.10 of the greatest value. The identified independent predictors of CAD were used to derive a prediction rule with optimal sensitivity, which was reported in an abstract well before the validation phase of the study. The prediction rule was then prospectively validated in the validation cohort by plotting the predicted cases of CAD (test positive) and the observed cases (disease positive) in 2 × 2 tables to calculate the prediction rule’s sensitivity and specificity for significant and severe CAD.
The 2-tailed Student t test was used to compare normally distributed continuous variables, and the Mann-Whitney U test was used to compare continuous variables that did not adhere to a normal distribution. The chi-square test was used to compare categorical variables. The Mantel-Haenszel extension of the chi-square test for trend was used to compare graded responses. The continuous variables are expressed as the mean ± SD, and the dichotomous variables are expressed as frequency and percentage. Relative risk is expressed as the odds ratio, with 95% confidence intervals. p Values <0.05 were considered statistically significant. PASW, version 18, software (SPSS, Chicago, Illinois) was used for the statistical analyses. The institutional review boards of the participating institutions approved the study.
Results
A database query of the cardiac catheterization laboratory of John H. Stroger, Jr. Hospital of Cook County yielded 124 consecutive patients who had undergone coronary angiography for the evaluation of systolic HF of unknown etiology and had met the inclusion and exclusion criteria. The baseline characteristics of the derivation cohort are summarized in Table 1 . The clinical characteristics of the patients with and without CAD are summarized in Table 2 . The patients with CAD were older, had a greater prevalence of diabetes, and were more likely to have multiple CAD risk factors other than diabetes. Furthermore, the patients with severe CAD tended to have a greater prevalence of Q waves or LBBB. The mean LVEF and the prevalence of chest pain or angina were not different between those with and without CAD ( Table 2 ). A family history of CAD was not different in patients with and without CAD (odds ratio 1). Thus, we did not include this risk factor in the nondiabetes CAD risk factors in the subsequent analyses.
| Variable | Derivation Cohort (n = 124) | Validation Cohort (n = 143) | p Value |
|---|---|---|---|
| Age (yrs) | 55 ± 11 | 56 ± 10 | 0.37 |
| Male gender | 72 (58) | 88 (62) | 0.56 |
| Male age ≥55 yrs, female age ≥65 yrs | 50 (40) | 54 (38) | 0.67 |
| Left ventricular ejection fraction (%) | 32 ± 10 | 22 ± 9 | <0.001 |
| Creatinine (mg/dl) | 1.3 ± 1.0 | 1.2 ± 0.7 | 0.28 |
| Hypertension ∗ | 89 (72) | 108 (76) | 0.49 |
| Diabetes mellitus | 33 (27) | 56 (39) | 0.03 |
| Smoker | 43 (35) | 60 (42) | 0.22 |
| Dyslipidemia † | 21 (17) | 99 (69) | <0.001 |
| Family history of premature CAD | 7 (6) | 14 (10) | 0.21 |
| No chest pain | 65 (52) | 93 (65) | |
| Nonanginal chest pain | 23 (19) | 29 (20) | |
| Anginal chest pain | 24 (19) | 21 (15) | |
| Unknown | 12 (10) | 0 (0) | |
| Electrocardiographic findings | |||
| Q waves | 23 (19) | 18 (13) | 0.18 |
| LBBB | 15 (12) | 23 (16) | 0.35 |
| Coronary angiographic findings | |||
| No significant CAD ‡ | 91 (73) | 91 (64) | 0.09 |
| Nonobstructive CAD § | 20 (16) | 16 (11) | 0.24 |
| No. of coronary arteries narrowed ≥50% | 33 (27) | 52 (36) | 0.09 |
| 1 | 9 (7) | 16 (11) | 0.27 |
| 2 | 9 (7) | 14 (10) | 0.46 |
| 2 (including proximal LAD) | 3 (2) | 2 (1) | 0.54 |
| 3 | 13 (10) | 17 (12) | 0.72 |
| Left main | 2 (2) | 5 (3) | 0.34 |
| Severe CAD ¶ | 18 (15) | 24 (17) | 0.61 |
∗ Documented 2 readings of blood pressure ≥140/90 mm Hg or receiving antihypertensive medical therapy.
† Total cholesterol >200 mg/dl, low-density lipoprotein >160 mg/dl, triglycerides >200 mg/dl, high-density lipoprotein <35 mg/dl in men or <45 mg/dl in women, or receiving lipid-lowering therapy.
‡ Luminal diameter stenosis ≥50% in ≥1 epicardial coronary arteries.
§ Luminal diameter stenosis 20% to 49% in ≥1 epicardial coronary arteries.
¶ Significant (≥50%) left main disease, 3-vessel disease, or 2-vessel disease that included proximal LAD.
| Variable | Significant CAD | Severe CAD | ||||||
|---|---|---|---|---|---|---|---|---|
| CAD | No CAD | OR (95% CI) | p Value | Severe CAD | No Severe CAD | OR (95% CI) | p Value | |
| Age (yrs) | 58 ± 10 | 54 ± 11 | 1.5 (1.01–2.2) ∗ | 0.04 † | 62 ± 9 | 54 ± 11 | 2.2 (1.2–3.8) ∗ | 0.006 † |
| Men | 20 (61) | 52 (57) | 0.9 (0.4–2.0) | 0.73 | 10 (56) | 62 (58) | 0.9 (0.3–2.4) | 0.82 |
| Any chest pain | 11 (33) | 36 (40) | 0.8 (0.3–1.8) | 0.53 | 7 (39) | 40 (38) | 1.1 (0.4–2.9) | 0.93 |
| Angina pectoris | 7 (21) | 17 (19) | 1.2 (0.4–3.2) | 0.75 | 6 (33) | 18 (17) | 2.44 (0.8–7.4) | 0.10 |
| LVEF (%) | 33 ± 9 | 31 ± 10 | 1.2 (0.8–1.8) ‡ | 0.45 | 32 ± 9 | 31 ± 10 | 1.1 (0.6–1.9) ‡ | 0.73 |
| Creatinine (mg/dl) | 1.4 ± 0.8 | 1.3 ± 1.3 | 1.03 (0.8–1.4) § | 0.85 | 1.5 ± 0.9 | 1.3 ± 1.2 | 1.14 (0.8–1.6) § | 0.44 |
| DM | 19 (58) | 14 (15) | 7.5 (3.1–18.3) | <0.001 † | 10 (56) | 23 (22) | 4.5 (1.6–12.7) | 0.003 † |
| Risk factors other than DM | 2.2 ± 1.0 | 1.5 ± 1.0 | 2.1 (1.3–3.3) ¶ | <0.001 † | 2.2 ± 1.1 | 1.6 ± 1.0 | 1.9 (1.1–3.2) ¶ | 0.01 † |
| Male age ≥55 yrs, female age ≥65 yrs | 19 (58) | 31 (34) | 2.6 (1.2–5.9) | 0.02 † | 12 (67) | 38 (36) | 3.7 (1.2–10.3) | 0.01 † |
| HTN | 29 (88) | 60 (66) | 3.8 (1.2–11.6) | 0.02 † | 16 (89) | 73 (69) | 3.6 (0.8–16.6) | 0.08 |
| Tobacco use | 16 (48) | 27 (30) | 2.2 (1.0–5.1) | 0.052 † | 9 (50) | 34 (32) | 2.1 (0.8–5.8) | 0.14 |
| Dyslipidemia | 7 (21) | 14 (15) | 1.5 (0.5–4.1) | 0.45 | 2 (11) | 19 (18) | 0.6 (0.1–2.7) | 0.48 |
| Family history | 2 (6) | 5 (6) | 1.1 (0.2–6.0) | 0.90 | 1 (6) | 6 (6) | 1.0 (0.1–9.0) | 0.99 |
| ≥2 Risk factors other than DM | 27 (82) | 44 (48) | 4.8 (1.8–12.8) | 0.001 † | 15 (83) | 56 (53) | 4.5 (1.2–16.3) | 0.02 † |
| ≥2 Risk factors: male age ≥55 yrs, female age ≥65 yrs, HTN, tobacco, dyslipidemia | 26 (79) | 43 (47) | 4.1 (1.6–10.5) | 0.002 † | 15 (83) | 54 (51) | 4.8 (1.3–17.6) | 0.01 † |
| Q waves or LBBB | 11 (33) | 27 (30) | 1.2 (0.5–2.8) | 0.70 | 9 (50) | 29 (27) | 2.7 (1.0–7.4) | 0.054 † |
∗ OR for 10-year increment in age.
‡ OR for 10-point increment in LVEF.
§ OR for 1-mg/dl increment in serum creatinine.
¶ OR for 1 increment in number of risk factors other than DM.
A multivariate logistic regression analysis with stepwise backward elimination determined that the only independent predictors for significant CAD were diabetes mellitus and ≥2 nondiabetes risk factors (age, men ≥55 and women ≥65 years, hypertension, tobacco use, and dyslipidemia). The other covariates entered in the model (LVEF, angina, any chest pain, Q waves or LBBB, gender, and creatinine) were sequentially eliminated ( Table 3 ). The Hosmer and Lemeshow test showed a good model fit at each step (p >0.72 for all). However, diabetes mellitus, ≥2 nondiabetes risk factors, and Q waves or LBBB were independently predictive of severe CAD. The other covariates entered in the model (i.e., LVEF, gender, creatinine, any chest pain, and angina) were sequentially eliminated ( Table 4 ). Similarly, the Hosmer and Lemeshow test showed a good model fit at each step (p >0.11 for all).
| Significant CAD | OR | 95% CI | p Value |
|---|---|---|---|
| Covariates evaluated in model at step 1 | |||
| DM | 10.5 | 3.6–31.2 | <0.001 |
| ≥2 Risk factors: male age ≥55 yrs, female age ≥65 yrs, HTN, tobacco, dyslipidemia | 4.7 | 1.6–14.1 | 0.006 |
| Q wave or LBBB | 1.5 | 0.5–4.2 | 0.47 |
| Any chest pain | 0.6 | 0.2–2.4 | 0.48 |
| Angina | 1.3 | 0.2–6.8 | 0.76 |
| Male gender | 1.8 | 0.6–5.5 | 0.27 |
| LVEF | 1.1 | 0.6–1.9 | 0.77 |
| Creatinine | 0.8 | 0.5–1.2 ∗ | 0.25 |
| Independent predictors remaining in model | |||
| DM | 7.8 | 3.0–20.2 | <0.001 |
| ≥2 Risk factors: male age ≥55 yrs, female age ≥65 yrs, HTN, tobacco, dyslipidemia | 4.4 | 1.6–12.2 | 0.004 |
| Severe CAD | OR | 95% CI | p Value |
|---|---|---|---|
| Covariates evaluated in model at step 1 | |||
| DM | 5.2 | 1.5–17.6 | 0.008 |
| ≥2 Risk factors: male age ≥55 yrs, female age ≥65 yrs, HTN, tobacco, dyslipidemia | 4.5 | 1.1–18.5 | 0.04 |
| Q wave or LBBB | 3.4 | 1.0–11.9 | 0.05 |
| Any chest pain | 0.4 | 0.05–3.8 | 0.44 |
| Angina | 4.0 | 0.4–42.5 | 0.25 |
| Male gender | 1.2 | 0.4–4.1 | 0.77 |
| LVEF | 1.05 | 0.5–2.0 | 0.89 |
| Creatinine | 0.9 | 0.6–1.4 ∗ | 0.69 |
| Independent predictors remaining in model | |||
| DM | 5.2 | 1.7–16.2 | 0.005 |
| ≥2 Risk factors: male age ≥55 yrs, female age ≥65 yrs, HTN, tobacco, dyslipidemia | 5.1 | 1.3–20.3 | 0.02 |
| Q wave or LBBB | 3.7 | 1.2–11.7 | 0.03 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
