August 22nd, 2011
Panel: Coronary Calcium vs. CRP for Predicting Cardiovascular Events
CardioExchange Editors, Staff
Drs. Paul Ridker and Sanjay Kaul offer their perspectives on a new coronary artery calcium/C-reactive protein study, published in the Lancet, and three of the study authors respond. The study showed that people with low LDL levels and high CRP levels may benefit from CAC scans to identify the folks who are most likely to benefit from statin therapy. Data came from 950 people enrolled in the Multi-Ethnic Study of Atherosclerosis (MESA) who had met the entry criteria for the JUPITER study. We invite you to read the discussion—and then join it.
Paul Ridker: Participants in MESA were not randomized to statin therapy, and there are no statin trials based on CAC scores (in contrast to high-sensitivity CRP); nor do statins lower CAC. A major lesson learned by the cardiology community is that without direct trial data, we just don’t know whether interventions work or not in a given population. So I congratulate these authors on providing evidence to undertake just such a trial, as has been suggested to the imaging community many times before. But in the absence of formal testing, it is imprudent to use a technology we know is associated with radiation exposure, expense, and a considerable “incidentaloma” rate without knowing that it actually identifies individuals who preferentially benefit from any specific therapy.
One other thing worth considering: the CORONA, GISSI-HF, 4D, and AURORA trials all failed to show any benefit of statin therapy. What do these trials all have in common? Severe underlying atherosclerosis with presumably high coronary calcium scores. So we just can’t make the assumption that people with high CAC scores benefit from statin therapy—we have to do real trials. Remember that the imaging community also assumed for years that statins would reduce coronary calcium, but when trials were finally done, that turned out to be a completely wrong assumption.
Sanjay Kaul: The finding that, compared with CRP, CAC is more strongly associated with CHD or CVD extends previous observations. However, I would like to have seen the comparative effects on a more clinically relevant risk-prediction metric, such as reclassification, that can help guide treatment decisions.
The NNT results are based on exploratory analysis using assumptions that are not verifiable in the MESA cohort—for example, 46% risk reduction in outcomes with rosuvastatin observed in JUPITER, which is likely to be an overestimate.
The observation that 75% of the events were clustered in the 25% of patients with a CAC score of >100 does not necessarily mean that selective therapeutic targeting of such patients (test all, treat few) would be more cost-effective than unconditional treatment (test none, treat all). It is a defensible hypothesis in dire need of validation.
The Study Authors Respond
Michael J. Blaha, Roger S. Blumenthal, and Khurram Nasir (study authors): We thank Drs. Ridker and Kaul for their insightful comments and agree that our paper is best considered a provocative exploratory analysis — one demanding follow-up investigation. Most agree that this should include a clinical trial with random allocation of statins after CAC testing. The NHLBI is currently considering such an RCT. Unfortunately, the results would not be available for many years, and the design of such as trial is immensely challenging (see Blumenthal and Hasan, J Am Coll Cardiol 2011; 57:1601). The complex trial design continues to be the main barrier to definitive evidence of the benefit of subclinical atherosclerosis testing.
We would like to highlight the two distinct aims of our paper. First, and most important, we sought to determine whether CAC testing could further risk-stratify the “JUPITER-eligible” population. Such patients are known to benefit from rosuvastatin, as per the JUPITER trial results. Within this group, CAC discriminates risk up to 20-fold, identifying JUPITER-eligible patients both at very low risk and very high risk. Expressed as a predicted 5-year number needed to treat (NNT-5), we see that JUPITER-eligible individuals with CAC=0 are unlikely to gain a net benefit from statin treatment, whereas those with measurable CAC, especially CAC >100, have a highly favorable NNT. We believe this has important potential public health implications for statin allocation in primary prevention. Regarding Dr. Kaul’s concern about the optimistic 46% event reduction from JUPITER, we agree, and our paper includes sensitivity analyses with more “real-world” expectations about statin benefit.
The second aim directly compares CAC vs. hsCRP for risk prediction among all patients with “normal” cholesterol (LDL<130). CAC is clearly superior here in the MESA cohort. Dr. Ridker brings up the critical point that we don’t know whether CAC “identifies individuals who preferentially benefit from any specific therapy.” This is true, but it is probably also true for hsCRP. Without a low-hsCRP arm, JUPITER could be considered a lipid-lowering trial, not just a biomarker trial. Studies from the Heart Protection Study and some secondary analyses from JUPITER show that higher hsCRP does not necessarily identify preferential benefit with statin (similar relative risk reduction across hsCRP levels). As Dr. Ridker has previously noted, the value of hsCRP is in identifying higher absolute cardiovascular risk among people not eligible for statins according to current guidelines. JUPITER was indeed a landmark achievement and has greatly expanded the population eligible for statin therapy.
Our data from the MESA cohort indicated that CAC is likely a better predictor of absolute risk than hsCRP, but CAC testing necessitates the need for low levels of ionizing radiation and the challenges of dealing with incidental findings that necessitate a follow-up CT scan (5%–10% of subjects in some prior studies.) We agree wholeheartedly that demonstration of improved risk prediction does not necessarily translate to cost-effectiveness of CAC testing. Incidental findings remain a critical issue.
Until the results of a CAC RCT are available, use of either CAC or hsCRP as “tie-breakers” for determining statin benefit is reasonable. The 2010 ACCF/AHA Guideline for Assessment of Cardiovascular Risk in Asymptomatic Adults gives CAC scoring a IIa recommendation in the intermediate risk group and hsCRP measurement a IIa recommendation in those meeting the JUPITER entry criteria. However, we posit this provocative question: What is the future of primary prevention — treating based on direct measurement of the disease we propose to treat (coronary atherosclerosis) or treating based on a single blood test? We look forward to future research in this area.
Khurram Nasir (personal view): In my opinion, the major aim of our study was to assess a key question: whether we can identify a subset of individuals meeting JUPITER criteria using CAC testing who are less likely to derive any significant benefit.
As our study clearly demonstrates, nearly half of individuals within this nationally representative and ethnically diverse cohort meeting eligibility for statin therapy according to JUPITER criteria had no CAC and experienced an extremely low event rate during almost 6 years of follow-up, with a very high estimated number needed to treat for 5 years to prevent one cardiovascular event. This has very important implications: If all individuals meeting JUPITER eligibility criteria can be further stratified by CAC testing prior to committing them to long-term statin therapy, we can identify one out of every other individual in this group from whom we can safely withhold pharmacotherapy and focus on lifestyle modification.
I believe in the setting of such a low risk for CVD events with absence of CAC, the burden of proof that this specific subset of individuals will benefit from statin therapy in the setting of meeting any criteria for lipid-lowering pharmacotherapy lies with those who advocate for it. In the current environment of rising health care costs and shrinking resources, we cannot afford to treat a large number of individuals to prevent few events.
The next challenge for us is definitely to engage all pertinent stakeholders to urgently consider the feasibility of a strategy supplemented by CAC testing, to separate out the subset of individuals with absent, thus, a very low CVD risk, and focus on a small subset of individuals with high atherosclerotic burden of subclinical atherosclerosis in whom the majority of events occur.
The FDA required at least 1 additional risk factor besides the JUPITER entry criteria of age and elevated hsCRP level to define eligibility for treatment with rosuvastatin. This was based on a post hoc subgroup analysis in which the event rate in the placebo arm was nearly double in those who met the age and elevated hsCRP level criteria plus at least 1 additional risk factor (incidence rate of 15.5 vs 7.7 per 1000 patient-years) compared with those based on age and elevated hsCRP level criteria alone (incidence rate of 7.9 vs 7.1 per 1000 patient-years). Moreover, the treatment benefit was observed only in the former subgroup (hazard ratio [HR], 0.51; 95% confidence interval [CI], 0.41-0.64) compared with the latter subgroup, which represented approximately 25% of the JUPITER cohort (HR, 0.91; 95% CI, 0.56-1.46) (Pint = .03).
Thus, accounting for 1 additional traditional risk factor identified high-risk patients and treatment response in JUPITER-eligible patients without subjecting them to the CAC testing and it’s potential downsides (cost, radiation exposure, incidentalomas, etc.)
So, this observation begs 2 questions:
1. How would CAC perform compared with 1 additional risk factor?
2. Can CAC further discriminate those with JUPITER eligibility criteria plus 1 additional risk factor?
Jupiter failed to prove that subjects with elevated HS-CRP responded to therapy any better than subjects with normal HS-CRP might have responded. To imply otherwise is imprudent at best.
When Dr. Ridker’s writes:”it is imprudent to use a technology we know is associated with radiation exposure, expense, and a considerable “incidentaloma” rate without knowing that it actually identifies individuals who preferentially benefit from any specific therapy.” I assume he is arguing against nuclear stress imaging.
Coronary calcium imaging has been unequivocally shown to identify individuals at risk for heart attacks dramatically better than all risk factors combined. It is not a stretch to say that individuals with atherosclerosis will benefit from atherosclerosis therapy.
As no technology including Framingham risk stratification, stress imaging, ultrasound, or HS-CRP has been demonstrated to improve outcomes, is Dr. Ridker suggesting that all primary prevention is “imprudent”?
As the “imaging community” has never bothered to do outcome studies, has anyone asked why? I am aware of several proposals to look at outcome studies based on calcium imaging that have failed to be funded. If the NIH would fund such a study, we would have the DATA. Who is involved in the decisions to not fund these studies? I suspect that the irrational detractors of calcium imaging such as Dr. Ridker and Dr. Nissen just might influence the decisions to not fund CAC intervention studies.
The best outcome study we currently have is the St. Francis sub-study; Treatment of Asymptomatic Adults With Elevated Coronary Calcium Scores With Atorvastatin, Vitamin C, and Vitamin E; J Am Coll Cardiol, 2005; 46:166-172. In this study, the improvement in the combined endpoint of coronary vascular events was 30% with a p value of 0.08. For those who do not understand P values, that means that there is a 92% chance that this result is a consequence of the treatment, not just random chance.
A sub-group with calcium scores over 400 demonstrated a 42% reduction in the combined endpoint of all coronary vascular events with a p value of 0.046. As this was not a pre-specified secondary endpoint, it is not of unassailable value however there is a 95.4% chance that this reduction was due to the use of atorvastatin in this group rather than just chance.
The observation that the higher the calcium score, the greater the relative improvement with atorvastatin is a compelling observation, one that was notably absent with Jupiter.
It is time for those “thought leaders” such as Drs. Ridker to come clean with their conflicts of interest and step aside so that the truth can be told and heart attacks can be prevented in the most effective way possible, using atherosclerosis screening.
Competing interests pertaining specifically to this post, comment, or both:
I work at a facility where EBT calcium imaging is used to stratify risk for MI. Our MI rates are extraordinarily low.