March 10th, 2010

Is the Patient-Selection Process for Cardiac Cath Broken?

We welcome Pamela S. Douglas, MD, to answer our questions about her research team’s NEJM study on the diagnostic yield of coronary angiography. We encourage you to ask yours.
Background: In this observational study of the CathPCI database of the National Cardiovascular Data Registry (NCDR), 38% of about 400,000 patients without known coronary artery disease who underwent elective cardiac catheterization were ultimately found to have obstructive CAD.
Did the patients who had undergone noninvasive testing (84% of your cohort) include those who had a screening catheterization for new-onset cardiomyopathy? If so, could that have biased the results, given that typically about half of such patients have nonobstructive CAD?
We excluded patients who underwent diagnostic catheterization for cardiac indications other than “to rule out CAD,” as the NCDR describes it. That exclusion comprises an array of cath indications, including cardiomyopathy, that patients may have. Although the NCDR is an observational database and the indications for cath were selected by each site, we suspect that few patients with new-onset cardiomyopathy were in our cohort.
Noninvasive testing added little predictive value to that of clinical risk factors and symptoms in your study. Would such testing have had a better predictive value for obstructive CAD if the analysis had included only patients with stress tests (exercise, nuclear, or echo) and not patients with ECGs, resting echos, or CTs?
One would certainly hope so, but we simply cannot know for sure. Notably, Bayesian principles tell us that no test performs well in populations whose pretest risk levels are at the extremes. In our study, the low prevalence of obstructive CAD at cath suggests that the noninvasive tests for ischemia were done in a low-risk population. In the unlikely event that all patients who underwent noninvasive tests also went on to catheterization, there would have been, at best, an intermediate probability of disease. In either case, it is impossible to assess the performance of these tests, regardless of the type, using our data.
Should we as practitioners conclude that we’re doing too many useless noninvasive tests and cardiac catheterizations?
This has been posed as an either/or choice, but the data make clear that a large number of patients in our study were at low clinical risk (30% were asymptomatic, 29% had low Framingham risk). Given that decisions about whether to proceed to cath begin with a clinical evaluation, it’s at least one of the potential source points for the problem. We do not have adequate data to assess whether the actual ordering of tests and caths also contributes to the problem. In addition, we must remember that a finding of no obstructive CAD can be an important result and is not proof that testing was unnecessary. Plaque rupture generally occurs in patients with nonobstructive CAD, and even a finding that a patient has no lesions at all can be reassuring and may alter care.
What policies or practices should change as a result of this study?

We have merely identified a problem; we have not found its cause. More research is needed to identify the optimal practice at each of the decision points that lead up to a diagnostic cath: clinical evaluation and risk assessment, noninvasive test selection and performance, and the decision to proceed to cath. The National Heart, Lung, and Blood Institute has just funded the PROMISE trial, in which clinical outcomes following a diagnostic strategy of functional testing will be compared with outcomes after CT angiography in patients with stable chest pain.
For now, benchmark rates for finding no obstructive disease in a carefully defined elective-cath population could be added to the regular NCDR reports that sites already receive. The sites would be able to use this information to assess and improve their practices. Decision-support tools can also be employed, as David J. Brenner recommends in his editorial. The American College of Cardiology has developed appropriate-use criteria for noninvasive testing and revascularization. Such criteria for diagnostic cath might be very helpful in quality-improvement efforts.

4 Responses to “Is the Patient-Selection Process for Cardiac Cath Broken?”

  1. Overlooking a major point

    This is a very good and useful study however it misses a major aspect of heart attack prevention and implicitly supports old and erroneous paradigms.

    Most heart attacks result from rupture of unstable, non-obstructive plaque. Any test looking for obstruction, either cath or stress imaging will miss the majority of subjects who are destined to suffer an MI. Furthermore, the utility of screening for obstruction is questionable since repairing obstruction is not associated with improved outcomes over medical management.

    I would suggest that Bayesian principles do not apply to a test that is very sensitive for the presence as well as the absence of disease. With coronary calcium imaging, a 0 score is highly predictive of non- events and positive scores are incrementally predictive of events, therefore, screening low risk populations is successful. As it has also been shown that statistically the majority of heart attacks occur in Framingham “low risk” subjects; ignoring the risk in this population is a mistake.

    Competing Interests:
    1. I hate heart attacks and strokes
    2. I have an ownership interest in a facility that can do coronary calcium imaging and ultrasound.

  2. Low Yield Tests

    Calcium scoring is no better at identifying “vulnerable” plaques and plaque rupture than coronary angiography. I fail to see it’s utility in “low risk” individuals.

  3. Coronary calcium imaging in “low risk” subjects

    Keel an et al. (Circulation. 2001;104:412-417.) demonstrated that EBT coronary calcium imaging was more predictive of MI than coronary angiography. However for the purpose of this discussion, I can accept that they are equivalent. The initial elevated calcium score identifies the subject who has had unstable plaque in the past. In most circumstances, that vulnerable subject will still have unstable plaque. The follow up calcium score when stable identifies the person whose physiology has changed and is no longer producing unstable plaque. The follow up calcium score when increasing by more than 15% annually, identifies the individual who is continuing to lay down unstable plaque. This observation is clearly shown in a study by Raggi et al, in Arteriosclerosis, Thrombosis, and Vascular Biology 2004;24:1272. The observation that stable plaque is associated with very few events and progression of plaque is associated with significantly more events is further discussed in ACCF/AHA 2007 Expert Consensus Document Circulation Jan 23, 2007. Finally, there is a compelling reason to use coronary calcium imaging in the “low risk” population. Although it is certain that “high risk” individuals are at greater individual risk for MI than low risk subjects, it however is lost on the guideline writers that the majority of heart attacks actually occur in “low risk” subjects. Khot, et al. JAMA. 2003 demonstrated that 81% of men with heart attacks had at least one risk factor. This study is often quoted to support the use of risk factors alone, without the benefit of atherosclerosis imaging, to screen for heart attack risk. Looking at this DATA differently, what percentage of subjects would have qualified for medical intervention of risk based on risk factors alone. The answer is sobering. 62% of these men had 0 or 1 risk factor and would almost always be considered “low risk” and would qualify for lipid management only if their LDL cholesterol was >160. This is no different than the general adult population where 63% of subjects have 0 or 1 risk factor as documented in AHA Heart Disease and Stroke Statistics 2007, Circulation 1-2007. To make matters worse, a study by Sachdeva A et al. Am Heart J 2009; 157:111-117, demonstrated that 50% of heart attacks occur in subjects with LDL cholesterol <100 and 77% had LDL <130. Without belaboring the subject, the guidelines fail miserably by not identifying the vast majority of subjects at risk for MI.

  4. In summary

    As CAC imaging is remarkably good at predicting heart attack risk ( dramatically better than all risk factors combined and 10X more predictive than HS-CRP (N Engl J Med 2008;358:1336-45) while conventional risk factors fail, and the majority of heart attacks occur in “low risk” individuals, it is time we start doing routine atherosclerosis imaging. EBT CAC imaging is a great non-invasive, inexpensive, very low radiation measure of MI risk and carotid US is the best non-invasive measure of stroke risk.

    Competing Interests: Through the use of EBT calcium imaging and carotid ultrasound screening, I have been able to dramatically reduce the incidence of heart attack and stroke in my Internal Medicine practice. In order to provide this technology to my community, it was necessary that I make a financial investment in an EBT scanner after my community hospital refused to purchase one.