Author + information
- Received February 27, 2014
- Revision received May 20, 2014
- Accepted May 20, 2014
- Published online September 2, 2014.
- ∗Department of Diagnostic Radiology, Yale University School of Medicine, New Haven, Connecticut
- †Atlantic Medical Imaging, Galloway, New Jersey
- ↵∗Reprint requests and Correspondence:
Dr. Kevin M. Johnson, Yale University School of Medicine, Department of Diagnostic Radiology, 55 Lock Street, New Haven, Connecticut 06520.
Background Accurate assignment of statin therapy is a major public health issue.
Objectives The American Heart Association and the American College of Cardiology released a new guideline on the assessment of cardiovascular risk (GACR) to replace the 2001 National Cholesterol Education Program (NCEP) Adult Treatment Panel III recommendations. The aim of this study was to determine which method more accurately assigns statins to patients with features of coronary imaging known to have predictive value for cardiovascular events and whether more patients would be assigned to statins under the new method.
Methods The burden of coronary atherosclerosis on computed tomography angiography was measured in several ways on the basis of a 16-segment model. Whether to assign a given patient to statin therapy was compared between the NCEP and GACR guidelines.
Results A total of 3,076 subjects were studied (65.3% men, mean age 55.4 ± 10.3 years, mean age of women 58.9 ± 10.3 years). The probability of prescribing statins rose sharply with increasing plaque burden under the GACR compared with the NCEP guideline. Under the NCEP guideline, 59% of patients with ≥50% stenosis of the left main coronary artery and 40% of patients with ≥50% stenosis of other branches would not have been treated. The comparable results for the GACR were 19% and 10%. The use of low-density lipoprotein targets seriously degraded the accuracy of the NCEP guideline for statin assignment. The proportion of patients assigned to statin therapy was 15% higher under the GACR.
Conclusions The new American Heart Association/American College of Cardiology guideline matches statin assignment to total plaque burden better than the older guidelines, with only a modest increase in the number of patients who were assigned statins.
- cholesterol guidelines
- computed tomography
- coronary artery disease
- low-density lipoprotein
- risk estimates
In November 2013, the American Heart Association (AHA) and the American College of Cardiology (ACC) jointly published the guideline on the assessment of cardiovascular risk (GACR) (1,2) to replace the National Cholesterol Education Program (NCEP) Adult Treatment Panel (ATP) III recommendations, released in 2001 (3). Some have suggested that the new guideline overestimates risk and will result in too many patients being put on statins for life.
The new guideline, like the old, is based on epidemiologic analysis, validated in varied populations. The accuracy of an epidemiologic model to pinpoint the risks in an individual is limited. However, coronary computed tomographic angiography (CTA) can directly detect the presence and severity of plaque burden and stenoses in each patient.
The important clinical decision is whether to start a particular patient on statins for life. Thus, we structured this analysis around these specific questions:
1. Which method is more accurate in assigning treatment to patients with features of coronary imaging known to have predictive value for cardiovascular events?
2. What accounts for the difference in performance of the methods, if any?
3. Does the proportion of patients on statins differ between the methods?
Our premise is that more plaque, especially more proximal and stenotic plaque, means more risk. In particular, we were interested in the extremes: How frequently are patients with very heavy atherosclerotic burden not assigned statins? Are many patients with no plaque assigned statins?
The computed tomographic angiograms of consecutive patients obtained at a private outpatient radiology practice were reviewed. Patients had presented for various reasons, including stable atypical chest pain, indeterminate stress test results, multiple risk factors, and a strong family history. Risk factors and patient histories were retrospectively collated from patient intake forms and medical records. Patients were included if they were 21 to 90 years of age and had readily available lipid data. Patients younger than 40 years or older than 75 years, without known atherosclerotic cardiovascular disease (ASCVD) or low-density lipoprotein (LDL) cholesterol ≥190 mg/dl, were not included in further analyses, because the GACR makes no recommendations for such patients. The study complied with Health Insurance Portability and Accountability Act guidelines and received a waiver from the local institutional review board.
Baseline cholesterol estimation
The NCEP and GACR regression equations for risk estimation were developed using data from a time when few patients were on statins, and they do not specifically account for statin use. In the present study, the baseline total cholesterol and LDL cholesterol values of patients on statins at the time of imaging were estimated from the measured values. In one large study, the average decrease in LDL cholesterol with statin therapy was 34% and that of total cholesterol 25% (4). For patients on statins, this corresponds to increasing the measured values of LDL cholesterol by 52% and total cholesterol by 33%. Analyses were repeated in the subgroup of patients not on statins at the time of imaging to confirm conclusions based on all patients. In addition, a sensitivity analysis was performed by varying the LDL decrease between 10% and 70% with a proportional change in total cholesterol level.
NCEP risk estimation
Using the NCEP ATP III panel recommendations, the percent risk for coronary heart disease (CHD) at 10 years was calculated under its modified Framingham model (3). Then, risk factors were assessed: smoking (any in the past month), hypertension (history of hypertension or systolic pressure ≥140 mm Hg), high-density lipoprotein (counted as 1 if <40 mg/dl and as −1 if >60 mg/dl), age (men, ≥45 years; women, ≥55 years), and family history of premature CHD in a first-degree relative (men, <55 years; women, <65 years). Patients were categorized as low risk (0 or 1 risk factor), intermediate risk (≥2 risk factors and <10% risk for CHD at 10 years), moderately high risk (≥2 risk factors and 10% to 20% risk for CHD in 10 years), or high risk (≥2 risk factors and >20% risk for CHD in 10 years). Patients with known CHD, diabetes, or peripheral atherosclerosis were classified as high risk regardless of their risk factor count or Framingham estimate.
The decision to start statins was based on an LDL cholesterol target keyed to the risk category: low risk, <160 mg/dl; intermediate risk, <130 mg/dl; moderately high risk, <100 mg/dl; and high risk, <70 mg/dl.
GACR risk estimation
Under the GACR, 4 groups benefit from statins (2):
1. Known ASCVD with no New York Heart Association functional class II to IV heart failure or dialysis
2. LDL cholesterol ≥190 mg/dl
3. Age 40 to 75 years with diabetes and LDL >70 and <190 mg/dl without ASCVD
4. Age 40 to 75 without ASCVD or diabetes, LDL >70 and <190 mg/dl, and estimated 10-year risk of 7.5% or higher using new pooled cohort equations
Statins were assigned if a patient met any of these criteria. The heart failure and dialysis exclusions were not applied, because these data were unavailable.
Patients not already on beta-blockers received 100 mg of metoprolol by mouth about 1 h before CTA. If the heart rate remained higher than 72 beats/min, an additional 50 to 100 mg was administered. Heart rate, electrocardiogram, and blood pressure were monitored. Sublingual nitroglycerin spray 1/150 g was given 1 min before image acquisition. A 64-slice computed tomographic scanner was used (VCT 64, GE Healthcare, Waukesha, Wisconsin). An initial bolus-timing single-slice scan using 20 ml of contrast (iodine content 320 mg/ml) (Visipaque, GE Healthcare) was followed by a 20-ml saline bolus. Contrast injection of 80 ml at a flow rate of 5 ml/s was followed by a saline chaser, 50 ml at a flow rate of 5 ml/s. Collimation was 0.625 mm, rotation time 0.35 ms, scan pitch 0.16 to 0.24 depending on heart rate, field of view 25 cm, reconstruction matrix 512 × 512, table feed 20 mm/rotation, tube voltage 120 kVp, and tube current 450 to 800 mA. Electrocardiographic dose modulation was used when possible. Electrocardiographically gated datasets were reconstructed from 70%, 75%, and 80% of the R-R cycle length; additional reconstruction windows were used if motion artifacts were present. Cases after June 2006 mostly used prospective gating, while the rest were retrospectively gated. Overall, approximately two-thirds of the cases in this study were done with prospective gating.
Plaque burden and stenoses scoring
Methods of scoring plaque burden have been discussed in detail previously (5). The coronary tree was analyzed using the AHA’s 1975 16-segment definitions (6). Cases with image quality designated as “limited” for any reason were excluded. Heavy calcification was not an exclusion criterion.
Each coronary examination was scored using the following 3 methods:
1. Segmental plaque burden score (SPS): For each segment, plaque amount was scored as none or trace (0), mild (1), moderate (2), or heavy (3). These refer to the volume of plaque within a segment, whether calcified or not. In the case of multiple lesions, the score reflected the segment as a whole. The SPS was calculated as the sum of the individual segment scores.
2. Segmental stenosis score (SSS): Same as above, but using an estimate of worst diameter stenosis per segment rather than plaque volume, scored as very mild, <30% (0); mild, 30% to 49% (1); moderate, 50% to 69% (2); or severe, ≥70% (3). The SSS was calculated as the sum of the individual segment scores.
3. Segmental involvement score (SIS): For each segment, plaque amount was scored as absent or trace (0) or present (1), whether calcified or not. The SIS was calculated as the sum of the individual segment scores.
In an earlier paper (5), we established reference values for the plaque-scoring methods using data from patients with known coronary artery disease. For the SPS, the category cutoff values were 0 (zero), 1 to 3 (mild), 4 to 7 (moderate), and ≥8 (heavy). A similar procedure was used to establish the values for the SSS method (0, 1 to 3, 4 to 8, and ≥9) and the SIS method (0, 1 or 2, 3 or 4, and >4).
SPS is an overall plaque burden score. However, SSS and SIS are better established in the published research. In particular, SSS is closely allied to the Duke prognostic score. Other investigators have shown outcome data for SIS and SSS (7,8) All 3 have been included.
Certain findings may correlate with higher risk, in particular patterns of more proximal disease (7). Two prognostic assessments of this type were used:
1. Modified Duke prognostic index: This index, derived from conventional angiographic data, correlates with cardiac morbidity and mortality (9). For each segment, stenosis is graded visually as very mild (<30% diameter), mild (30% to 49%), moderate (50% to 69%), or severe (≥70%). Then permutations of lesions carrying increasing risk for a coronary event are defined. The original scheme used 16 combinations, but we adopt the following modification. The categories are 0, no or trace plaque; 1, very mild or mild stenoses; 2, ≥2 mild stenoses with 1 proximal or 1 moderate stenosis; 3, 2 moderate stenoses or 1 severe stenosis; 4, 3 moderate stenoses or 2 severe stenoses or proximal left anterior descending coronary artery (LAD) severe stenosis; 5, 3 severe stenoses or 2 severe stenoses with proximal LAD involved; and 6, moderate or severe left main stenosis.
2. Criteria emphasized in a meta-analysis of 11 prognostic studies by Bamberg et al. (8): Left main stenosis ≥50%, any stenosis ≥50%, and any plaque. The number of proximal segments with stenosis ≥50% was also assessed, defined as proximal right coronary artery, middle right coronary artery, left main coronary artery, proximal LAD, middle LAD, proximal circumflex coronary artery, and obtuse marginal branch, following Hadamitzky et al. (7).
Discrete variables are reported as proportions and continuous variables as mean ± SD, if approximately normally distributed, or as median (interquartile range) if not. Statistical significance was defined at the ≤0.05 level for all analyses (2-tailed). Spearman’s rho was used for correlation because several important variables were substantially skewed; 95% confidence intervals (CI) for rho were calculated using Fisher’s z transformation. Logistic regression was used, with the decision whether to start a statin as the dichotomous variable and various measures of disease severity as the continuous variable. Unit odds ratios with 95% CIs were calculated. McNemar’s test was used to estimate the significance of the difference in proportions of patients assigned to statin therapy by the 2 risk methods. JMP Pro version 10.0.0 (SAS Institute Inc., Cary, North Carolina) was used to perform the analysis.
Between February 2004 and November 2009, 3,346 patients met the inclusion criteria. Of the cases, 270 (8.1 %) were judged as technically limited and were excluded. The remaining 3,076 subjects were studied (65.3% men, mean age 55.4 ± 10.3 years, mean age of women 58.9 ± 10.3 years, >90% Caucasian) (Table 1). The median NCEP Framingham 10-year risk was 10% for men and 3% for women. The most common reasons for undergoing CTA were hypercholesterolemia, family history of CHD, hypertension, smoking, and various kinds of nonacute chest pain (Table 2).
At the time of imaging, 1,362 patients (44.3%) were not on statins or other lipid-lowering therapy; the remaining 1,714 patients (55.7%) were on statins. Patients not on statins had less disease than patients on statins: the 10th, 25th, 50th, 75th, 90th, and 97.5th percentile scores for SPS were 0, 0, 1, 3, 6.7, and 12 and 0, 1, 3, 8, 12, and 18, respectively.
Correlation of 10-year risk estimates with plaque
The newer risk estimation method (GACR) was correlated marginally better with plaque and stenosis burden than the older method (NCEP). For all patients, the correlation of SPS with the NCEP 10-year risk estimate was 0.51 (95% CI: 0.48 to 0.54), whereas with the GACR estimate, it was 0.56 (95% CI: 0.53 to 0.59) (Table 3).
Probability of assigning statins
The probability of prescribing statins as a function of plaque burden under each method is shown by the logistic regression results in Figure 1. The probability curve rose sharply with increasing plaque burden. For the corresponding odds ratios in Table 4, there is no overlap of the 95% CIs; in all cases, the GACR performed better than the NCEP guideline.
The GACR assigned fewer patients with no plaque to statins and more patients with heavy plaque to statins. Of patients with heavy plaque (SPS ≥ 8), 53% would be assigned to statin therapy under the NCEP guideline and 92% under the GACR. Of patients with no or trace plaque, 41% would be assigned to statin therapy under the NCEP guideline and 36% under the GACR (Table 5). Of note, under the NCEP guideline, 59% of patients with ≥50% stenosis of the left main coronary artery and 40% of patients with ≥50% stenosis of other branches would not have been treated. The comparable results for the GACR were 18% and 10%. As shown in Table 5, the NCEP guideline was more likely to assign statins to patients with little plaque and less likely to assign statins to those with much plaque.
The sensitivity analysis for estimated baseline cholesterol values did not alter our conclusions (Figure 2).
Different performance of the methods
Patients with known atherosclerosis and diabetes are considered high risk in both schemes. Removing LDL cholesterol of 190 mg/dl or greater as a condition (criterion 2) resulted in little change (data not shown). Replacing the GACR 10-year risk equation with the NCEP equation (criterion 4) also resulted in little change (Figure 3).
Although the GACR and NCEP methods partition patients into groups of similar kinds, the NCEP method then applies LDL targets. However, the correlations of serum LDL cholesterol levels with most plaque measures are not significantly different from zero (Figure 4). This includes plaque features known to correlate with risk and includes patients with LDL of 190 mg/dl or greater.
We found that this lack of correlation largely destroyed the discrimination achieved by the risk categorization. To demonstrate this, we altered the NCEP method by not applying the LDL targets. Instead, all patients in NCEP categories 3 (moderately high risk) and 4 (high risk) were assigned to statins, and those in categories 1 and 2 were not. The performance of this altered NCEP approach was similar to the GACR (Central Illustration), representing a substantial improvement.
Proportion of patients assigned to statins
The proportion of patients assigned to statin therapy under the NCEP guideline was 53%, compared with 61% under the GACR (all patients) (Table 6).
Compared with the NCEP ATP III guideline, the GACR better discriminates between patients with little or no plaque and those with much plaque. Under the new guidelines, the probability of being assigned statin therapy better tracks the underlying plaque and stenosis severity.
In this study, plaque burden is used as a surrogate for cardiac events, which are the ultimate end point to determine accuracy. This hypothesis is partially proved in symptomatic patients but remains unproved in the primary prevention setting. Other physiologic, as opposed to epidemiologic, factors also contribute to risk, including the blood’s propensity to clot and the vulnerability of plaque rupture. CTA cannot assess the thrombogenicity of blood. In the present study, we did not evaluate the radiodensity of the plaque, which has been shown to correlate to some degree with vulnerability but with technical difficulty because of limited spatial and contrast resolution.
There was not a significant difference in the correlation of the 10-year risk estimates with plaque or stenosis burden. Nonetheless, there was a marked difference in the probability of being assigned to statin therapy as a function of the underlying plaque and prognostic measures. If accuracy is defined as assigning statins to patients with greater risk for disease, and not to those without, the GACR outperformed the NCEP guideline substantially.
The use of LDL cholesterol targets by the NCEP led to this difference. The GACR and NCEP methods partition patients into risk groups in similar ways, but the NCEP method then applies LDL targets to decide who gets statin therapy. The correlation of serum LDL cholesterol levels with various plaque measures is not significantly different from zero. This lack of correlation largely destroys the discrimination achieved by the risk categorization. The AHA/ACC committee cited the lack of clinical trial data to support its decision to no longer use targets (2). This study reinforces this methodological decision, by showing that these targets degrade the accuracy of assignment of patients to statin therapy, in the sense of correlation with underlying atherosclerosis severity.
The proportion of patients on statins is modestly larger under the GACR, about 15% more patients than under the NCEP guideline (Table 6). This is in approximate agreement with projections by others (10). However, Ridker and Cook (11) showed overestimation of risk when the GACR is applied. If the appropriate risk threshold desired is 7.5%, the nominal threshold would need to be raised to compensate for this overestimation, and the proportion of patients on statins would decrease. Ideally, such an adjustment would be based on a comprehensive cost-benefit analysis.
There are still large numbers of patients with no plaque being assigned to statin therapy under both guidelines, but the GACR performed slightly better than did the NCEP guideline. The event risk in patients with normal results on CTA is known to be very low (8,12–14). At the same time, statins are not without complications and expense (15). It does not appear that these risk estimators can distinguish patients with little or no disease from those in whom it is clearly present. The simple absence of calcification on coronary imaging likely could substantially improve this discrimination in asymptomatic patients, at least in nondiabetics (16).
For the patients already on statins or other lipid-lowering medications at the time of the imaging, the analysis estimated baseline cholesterol values by increasing the measured value. This added uncertainty to the results. For this reason, the analyses were repeated in the subgroup of patients not on statins at the time of imaging and in those who did not require estimation of baseline cholesterol values. The sensitivity analysis for estimated baseline cholesterol values did not alter our conclusions.
Heart failure and dialysis status were unknown and thus could not be used with the GACR criteria. Seriously ill patients did not account for many cases in this nonacute outpatient setting. The inclusion of a few patients of this type would not be expected to materially change our results.
In this single-center retrospective study, the population was mostly Caucasian, from a single suburban, smaller city area. Referral bias is a consideration; the patients were referred mostly by primary care physicians and cardiologists because of nonspecific symptoms or indeterminate stress test results or because of multiple risk factors. Nonetheless, this patient spectrum can be expected to reflect the population of patients likely to undergo CTA outside an acute chest pain setting in predominantly white communities. Inclusion was biased somewhat toward higher risk patients (Table 1). In contrast, the population sent for such a study in daily clinical settings is likely to be biased in a similar way. As shown in Table 2, few patients had symptoms, but many had risk factors. These are the patients one might be most inclined to image if the goal is to determine who in fact has a heavy plaque burden and are therefore (presumably) at high risk for events.
Risk factors were by patient report for the most part. The laboratory data were a sample of convenience in that manpower limitations allowed the retrieval of laboratory data on only a subset of all patients who had presented for CTA.
The 2013 guideline proposed by the AHA and the ACC (2) results in a better discrimination of total plaque and stenosis burden than does the 2001 NCEP ATP III guideline (3). On the basis of our findings, it is a reasonable hypothesis that the new guideline will better predict coronary events, given that it better correlates with the severity of underlying atherosclerosis. Outcome studies are needed to confirm this hypothesis. The proportion of patients assigned to statins under the new guideline is modestly larger than under the older recommendations. The GACR may overestimate risk, but after adjusting for this, the proportion of patients assigned to statins may not be much different than under the older guideline.
COMPETENCY IN MEDICAL KNOWLEDGE: A key difference between the 2013 ACC/AHA guidelines for the management of blood cholesterol and previous recommendations under the NCEP involves the selection of patients for statin therapy on the basis of an overall assessment of cardiovascular risk rather than LDL threshold criteria.
COMPETENCY IN PATIENT CARE: Following the 2013 ACC/AHA guidelines for the management of blood cholesterol will likely assign more patients with coronary disease to statin therapy and result in treating approximately 15% more patients overall.
TRANSLATIONAL OUTLOOK: Because available guidelines recommend treatment of a large number of patients with relatively mild or no vascular disease, better methods are needed to identify low-risk patients who could safely forgo statin therapy.
The authors thank Benjamin Johnson for assistance with data analysis and Wayne Martyn, RT, RN, and Hilary Johnson for data collection and entry.
Both authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- American College of Cardiology
- American Heart Association
- atherosclerotic cardiovascular disease
- Adult Treatment Panel
- coronary heart disease
- confidence interval
- computed tomographic angiography
- AHA/ACC 2013 guideline on the assessment of cardiovascular risk
- left anterior descending coronary artery
- low-density lipoprotein
- National Cholesterol Education Program
- segmental involvement score
- segmental plaque burden score
- segmental stenosis score
- Received February 27, 2014.
- Revision received May 20, 2014.
- Accepted May 20, 2014.
- American College of Cardiology Foundation
- Goff D.C.,
- Lloyd-Jones D.M.,
- Bennett G.,
- et al.
- Stone N.J.U.,
- Robinson J.,
- Lichtenstein A.H.,
- et al.
- Cleeman J.I.,
- Grundy S.M.,
- Becker D.,
- et al.
- Pedersen T.R.,
- Olsson A.G.,
- Faergeman O.,
- et al.
- Hadamitzky M.,
- Achenbach S.,
- Al-Mallah M.,
- et al.
- Bamberg F.,
- Sommer W.H.,
- Hoffmann V.,
- et al.
- Mark D.B.,
- Nelson C.L.,
- Califf R.M.,
- et al.
- Andreini D.,
- Pontone G.,
- Mushtaq S.,
- et al.
- Hadamitzky M.,
- Täubert S.,
- Deseive S.,
- et al.
- Hulten E.A.,
- Carbonaro S.,
- Petrillo S.P.,
- et al.