A risk-management approach based on the Framingham risk score (FRS), although useful in preventing future coronary artery disease (CAD) events, is unable to identify a considerable portion of patients with CAD who need aggressive medical management. Coronary artery calcium (CAC), an anatomic marker of atherosclerosis, correlates well with presence and extent of CAD. This study investigated mortality risk associated with CAC score and FRS in subjects classified as “low risk” versus “high risk” based on FRS. In total 730 veterans without known CAD (61 ± 10 years old, 12.8% women) underwent measurement of their FRS and CAC. Subjects were classified as “discordant low risk” (DLR) if their FRS was <10% and CAC score was ≥100 (n = 108, 14.8%) or “discordant high risk” (DHR) if their FRS was ≥20% and CAC score was 0 (n = 104, 14.2%). Survival analysis was used to compare mortality rates associated with FRS and CAC in DLR versus DHR subjects. Mortality rate during the mean 48-month follow-up was 7.3% (n = 53) including 18.5% (n = 20) in the DLR group and 7.7% (n = 8) in the DHR group, respectively. Adjusted relative risks of mortality were 5.46 (95% confidence interval [CI] 2.44 to 12.20, p = 0.0001) in subjects with CAC score ≥100 compared to CAC score 0 and 1.35 (95% CI 1.01 to 4.32, p = 0.04) in subjects with FRS ≥20% compared to FRS <10%. Adjusted relative risk of mortality was 3.6 (95% CI 1.57 to 8.34, p = 0.003) for DLR compared to DHR. Areas under the receiver operator curve to predict mortality were 0.72 for FRS, 0.82 for CAC score, and 0.92 for the combination. In conclusion, the prognostic value of CAC to predict future mortality is superior to the FRS. Addition of CAC score to FRS significantly improves the identification and prognostication of patients without known CAD.
Recent studies have shown that a significant proportion of patients categorized as low to intermediate risk by the Framingham risk score (FRS) have significant coronary artery calcium (CAC). Furthermore, recent cumulative evidence has pointed to the prognostic value of detecting atherosclerosis measured by extent of CAC rather than risk factor assessment by FRS to identify subjects at high risk for coronary artery disease (CAD). We therefore sought to compare in patients without known CAD the predictive value of a CAC score and FRS for future mortality, mortality risk associated with CAC and FRS in subjects classified as “low risk” versus “high risk” based on the FRS, and whether addition of a CAC score to FRS can further improve its predictive value for mortality.
Methods
The study population consisted of 730 consecutive veterans with suspected CAD but without previous known CAD who underwent CAC scanning. Subjects with established cardiovascular disease, stroke, diabetic retinopathy, end-stage renal disease, Raynaud syndrome, infection, cancer, immunosuppression, systemic inflammation status, or end-stage liver disease were excluded.
Conventional cardiovascular risk factors were assessed according to the FRS. Medical information was obtained from the Veterans Administration Computerized Patient Record System. Subjects receiving antihypertensive drugs or with systolic or diastolic blood pressures ≥140 or >90 mm Hg, respectively, were considered hypertensive. Subjects were considered dyslipidemic according to the National Cholesterol Education Program/Adult Treatment Panel III guidelines for lipid-lowering therapy. Subjects were considered diabetic if they were being treated for diabetes. Risk factors were determined and the FRS was calculated to assess risk of developing coronary events (myocardial infarction or coronary heart disease death) over the next 10 years. Framingham scoring categorizes subjects with multiple risk factors into those with 10-year risk for coronary heart disease of high risk (>20%), moderate risk (10% to 20%), and low risk (<10%). Mortality verified by the Social Security Death Index was obtained from electronic medical records.
CAC was detected using a Siemens Definition dual-source 64-slice scanner (Siemens Medical Solutions, Forchheim, Germany). The dual-source computed tomographic protocol has been described previously and validated by comparison to electron-beam computed tomography. Coronary arteries were imaged in 30 to 40 contiguous 3-mm slices during mid-diastole using electrocardiographic triggering during a 15-second breath-hold. CAC was considered present in a coronary artery when a density >130 HU was detected in ≥3 contiguous pixels (>1 mm 2 ) overlying that coronary artery and quantified.
Continuous data are presented as mean ± SD, and categorical data are reported as a percentage or absolute number. T , chi-square, and Mann–Whitney tests were used to assess differences between groups. Subjects were classified as “discordant low risk” (DLR) if the FRS was <10% and CAC score was ≥100 (n = 108, 14.8%) or “discordant high risk” (DHR) if the FRS was ≥20% and CAC score was 0 (n = 104, 14.2%). Kaplan–Meier survival curves were constructed for the DHR and DLR groups and compared by log-rank test. Effect of cardiovascular risk factors on cardiovascular mortality in the DHR and DLR cohorts was determined using stepwise Cox regression survival analysis. Multivariable Cox regression survival analyses were employed to assess mortality risk associated with CAC score and FRS overall and in subjects classified as low risk versus high risk based on the FRS with and without adjustment for cardiovascular risk factors. Predictive performance of the model and extent of overfitting were assessed through internal validation with bootstrapping.
Time-dependent receiver operator characteristic curves were constructed for models I (FRS), II (CAC), and III (CAC + FRS). Area under the curve was calculated to predict the ability of each model to detect cardiovascular-related mortality, with an area under the curve of 0.50 indicating no accuracy and a curve of 1.00 indicating maximal accuracy. All statistical analyses were performed with SAS 9.2 (SAS Institute, Cary, North Carolina; available at: http://www.sas.com ) and PASW 18 (SPSS, Inc., Chicago, Illinois; available at: http://www.spss.com ). This study was approved by the institutional review board committee of the Greater Los Angeles Veterans Administration Health Care System, Los Angeles, California.
Results
The study population consisted of 730 consecutive subjects; 86.7% were men and the mean age was 61 ± 10 years. Demographic and conventional risk factors are presented in Table 1 . In total 21.8% had intermediate likelihood of coronary heart disease but no symptoms, 31.5% had atypical chest pain, 32.1% had typical chest pain, 12.4% had abnormal treadmill test result, and 2.2% had syncope. There was no significant association between presence of symptoms and clinical outcome in the DLR and DHR cohorts and the entire study cohort (p >0.05).
| Variable | |
|---|---|
| Age (years) | 61 ± 10 |
| Men | 636 (87.1%) |
| Smoker | 223 (26%) |
| Hypertension ⁎ | 397 (55.8%) |
| High cholesterol † | 352 (49.4%) |
| Diabetes mellitus ‡ | 186 (26.1%) |
| Family history of coronary heart disease § | 372 (51%) |
| Body mass index (kg/m 2 ) | 29.7 ± 7.1 |
| Creatinine (mg/dl) | 1.04 ± 0.22 |
| Coronary artery calcium score | 396 ± 339 |
| 0 | 289 (39.6%) |
| 1–100 | 152 (20.8%) |
| ≥100 | 289 (39.6%) |
| Framingham risk score (%) | 12.4 ± 7.8 |
| 0–9 | 286 (39.2%) |
| 10–20 | 220 (30.1%) |
| ≥20 | 224 (30.7%) |
| Total cholesterol (mg/dl) | 167 ± 52 |
| Low-density lipoprotein (mg/dl) | 97 ± 43 |
| High-density lipoprotein (mg/dl) | 43 ± 18 |
| Triglyceride (mg/dl) | 126 ± 85 |
| Systolic blood pressure (mm Hg) | 125 ± 23 |
| Diastolic blood pressure (mm Hg) | 78 ± 15 |
| Follow-up (months) | 48 ± 14 |
| Cardiovascular event | 53 (7.3%) |
⁎ Self-reported diagnosis of hypertension, prescribed medication for hypertension, or current blood pressure >140 mm Hg systolic or >90 mm Hg diastolic (>130/80 mm Hg if diabetic).
† Self-reported diagnosis of high cholesterol, prescribed medication for high cholesterol, or current total cholesterol >200 mg/dl.
‡ Self-reported diagnosis of diabetes (type 1 or 2) or prescribed medication for diabetes.
§ First-degree relative; <65 years of age in women, <55 years of age in men.
Mortality rate was 7.3% (n = 53 deaths) during the mean 48-month follow-up period. Survival rates was 91.9% in patients with FRS <10% and 81.9% in patients with FRS ≥20% (p = 0.0001). Similarly, survival rate decreased from 97.9% in patients with CAC score of 0 to 85.2% in patients with CAC score ≥100 (p = 0.0001). After adjustment for traditional cardiovascular risk factors, relative risks of death were 5.46 (95% confidence interval [CI] 2.44 to 12.20, p = 0.0001) in patients with a CAC score ≥100 versus 0 and 1.35 (95% CI 1.01 to 4.30, p = 0.04) in patients with a FRS ≥20% compared to <10%. As shown in Figure 1 , mortality receiver operator characteristic areas under the curve were 0.72 (95% CI 0.68 to 0.75, p = 0.0001) for FRS and 0.82 (95% CI 0.79 to 0.85, p = 0.0001) for CAC. Addition of CAC score to the FRS increased the area under the curve from 0.72 to 0.92 (p = 0.001).
In total 108 patients (14.8%) were classified as DLR, whereas 104 patients (14.2%) were classified as DHR ( Table 2 ). There were no significant differences in age, gender, smoking status, family history of premature CAD, body mass index, creatinine, systolic and diastolic blood pressures, and lipid profile between the DLR and DHR cohorts. Prevalence of hypertension, hypercholesterolemia, and diabetes mellitus was higher in the DHR cohort compared to the DLR cohort.
| Variable | DLR (FRS <10%, CAC score ≥100) (n = 108) | DHR (FRS ≥20%, CAC score 0) (n = 104) | p Value |
|---|---|---|---|
| Age (years) | 58 ± 12 | 61 ± 13 | 0.1 |
| Men | 79% | 87% | 0.1 |
| Smoker | 26% | 23% | 0.7 |
| Hypertension ⁎ | 38% | 60% | 0.002 |
| High cholesterol † | 38% | 54% | 0.01 |
| Diabetes mellitus ‡ | 0 | 34% | 0.0001 |
| Family history of coronary heart disease § | 52% | 49% | 0.8 |
| Body mass index (kg/m 2 ) | 28.1 ± 6.4 | 28.8 ± 6.8 | 0.3 |
| Creatinine (mg/dl) | 1.01 ± 0.19 | 1.01 ± 0.23 | 0.9 |
| Coronary artery calcium | 215 ± 48 | 0 | 0.0001 |
| Framingham risk score (%) | 6.8 ± 1.6 | 24.8 ± 2.1 | 0.0001 |
| Total cholesterol (mg/dl) | 159.6 ± 61.8 | 151.8 ± 63.4 | 0.2 |
| Low-density lipoprotein (mg/dl) | 92.9 ± 45.1 | 86.9 ± 51.3 | 0.4 |
| High-density lipoprotein (mg/dl) | 39.1 ± 18.6 | 41.9 ± 26.4 | 0.4 |
| Triglyceride (mg/dl) | 119.3 ± 44.1 | 103.2 ± 46.3 | 0.7 |
| Systolic blood pressure (mm Hg) | 118 ± 18 | 121 ± 20 | 0.6 |
| Diastolic blood pressure (mm Hg) | 81 ± 12 | 83 ± 10 | 0.5 |
| Antihypertensive therapy ∥ | 100% | 100% | — |
| Antidiabetic therapy ∥ | 100% | 100% | — |
| Statin therapy ∥ | 100% | 100% | — |
| Cardiovascular event | 18.5% | 7.7% | 0.0001 |
⁎ Self-reported diagnosis of hypertension, prescribed medication for hypertension, or current blood pressure >140 mm Hg systolic or >90 mm Hg diastolic (>130/80 mm Hg if diabetic).
† Self-reported diagnosis of high cholesterol, prescribed medication for high cholesterol, or current total cholesterol >200 mg/dl.
‡ Self-reported diagnosis of diabetes (type 1 or 2) or prescribed medication for diabetes.
§ First-degree relative; <65 years of age in women, <55 years of age in men.
∥ Percent antihypertensive, antidiabetic, and statin therapies in subjects with diagnosed hypertension, diabetes, and hypercholesterolemia, respectively.
Mortality rates were 18.5% (n = 20 deaths) in the DLR group and 7.7% (n = 8 deaths) in the DHR group ( Figure 2 ). After adjustment for age, gender, diabetes mellitus, hypertension, hypercholesterolemia, family history of premature coronary heart disease, smoking status, antihypertensive therapy, diabetic and statin therapies, the multivariable relative risk of death was 3.61 (95% CI 1.57 to 8.34, p = 0.003) in the DLR compared to the DHR group ( Table 3 ). Furthermore, adjustment for risk factors, Cox regression survival analysis with bootstrapping technique survival analysis showed that diabetes mellitus and low-density lipoprotein were independent predictors of a cardiovascular cause for mortality in the DHR cohort ( Table 4 ). In addition, risk factor–adjusted survival analysis showed that CAC and high-density lipoprotein were independent predictors of a cardiovascular cause for mortality in the DLR cohort ( Table 4 ).
