Nuclear cardiology for detection of myocardial ischemia has a long and proud history that spans many decades (1). As currently used, stress myocardial perfusion imaging (MPI) using single-photon emission computed tomography has evolved over the years in its use of hardware, software, radionuclides, acquisition protocols, processing algorithms, interpretation methods, and stress modalities. The use has also felt the impact of practice guidelines; appropriateness criteria; radiation risk concerns; laboratory accreditation; user certification; pre-authorization; declining reimbursement; and the increasing use of positron emission tomography and emergence of other imaging and nonimaging modalities, such as computed tomography, stress echocardiography, stress magnetic resonance imaging; and fractional flow reserve measurements. Those of us who have been fortunate enough to witness the evolution remain amazed by the value that MPI continue to provide to patient care, albeit when used wisely.

Lessons learned from the application of MPI to clinical care have been many. Of these, 3 have stood out as significant over the years. The first involves the impact of post-test referral bias on specificity of the test. Thus, selective referral of patients with abnormal MPI to coronary angiography artificially lowers specificity of MPI (2). Second, coronary angiography should not be used as a “gold standard” for MPI, because anatomic coronary artery stenosis is “neither a sufficient nor a necessary cause” for the presence of myocardial ischemia (3). Third, MPI provides prognostic information independent and complementary to that obtained from anatomic modalities and/or risk factors and has proven itself in numerous studies to be useful for risk stratification in diverse populations (4). An important consideration that is often overlooked links the last 2 points; the variability in ischemic burden as it relates to coronary anatomy is intimately associated with the independent prognostic data provided by MPI.

In this issue of the Journal, Rozanski et al. (5) provide a fourth important lesson: There has been a dramatic decline in the frequency of myocardial ischemia detected by stress MPI over the last 2 decades (Figure 09_gr1). This observation is based on data derived from stress MPI performed on 39,515 patients at a single institution (after excluding 12,174 patients with prior myocardial infarction [MI], coronary revascularization, cardiomyopathy, or valvular heart disease). The sheer number of studies performed at 1 center is in itself remarkable and a testimony to the widespread adoption of this imaging modality by the community. The importance of this observation is further emphasized by the expertise of this group. Therefore, it is worthwhile to examine the reasons behind this decline and future implications.

There has been a progressive decline in cardiovascular deaths in the United States since the 1970s. From 1998 to 2008, the rate of death attributable to cardiovascular disease declined 30.6% (6). Reflecting these changes, the annual cardiac event rates in contemporary trials in which participants were on optimal medical therapy such as the DIAD (Detection of Ischemia in Asymptomatic Diabetics) study (0.6% cardiac death/MI) (7), the WOMEN (What Is the Optimal Method for Ischemia Evaluation in Women) study (8) (1.1% coronary death/hospitalization for acute coronary syndrome or heart failure), COURAGE (Clinical Outcomes Utilizing Revascularization and Aggressive Drug Evaluation) trial (4.1% death/MI) (9), and BARI 2D (Bypass Angioplasty Revascularization Investigation 2 Diabetes) study (4.7% death/MI/stroke) (10) were low. A large proportion of this welcome decline in event rates has been attributed to modern medical therapy and control of risk factors (11). Thus, the observed decrease in ischemia detected by Rozanski et al. (5) could also be a reflection of this trend. In COURAGE (12), medical therapy and coronary revascularization were associated with significant decrease in ischemia detection by imaging. Reduction in ischemia has been reported with common antianginal medications, ranolazine and statins (13). This salutatory effect observed even after short-term therapy is independent of changes in large-vessel coronary artery stenosis severity and is yet another reason why coronary angiography is a not a reliable reflection of myocardial perfusion pattern. It is provocative to suggest that more aggressive medical therapy results in less ischemia, which in turn is associated with better outcomes.

Another aspect of this decline has to do with the change in MPI referral patterns. With the introduction of numerous guidelines and especially appropriateness criteria (first in 2005 and then in 2009) (14) that limit indications for MPI referral, and growing concerns over radiation exposure with MPI, one would expect detection of more and not less ischemia. This is reflected in the decreasing proportion of MPI performed in asymptomatic individuals in the current study. Similarly, the increase in proportion of patients with diabetes and obesity would again have predicted a higher incidence of ischemia. Other changes in patient characteristics such as decreasing age, smoking, and lower blood pressure are consistent with the observed decline in ischemia. In propensity-matched analysis that controlled for these changes, the temporal decline in ischemia persisted.

Another form of change in referral pattern has to do with the source of the referral. It is possible that early on, this academic imaging center had a broad catchment area. Over the years, with MPI becoming more office-based and more available in the community, patients with high likelihoods of disease were shifted to outpatient/office centers, whereas referrals from noncardiologists who are less likely to have significant disease became over-represented in hospital-based academic centers. Changing patterns of resource utilization is another potential explanation for this trend. It is possible that individuals with high pre-likelihood ratios for coronary disease (and abnormal perfusion) are triaged to other noninvasive imaging tests or directly to angiography in more recent years. Data that address these concerns are not provided, but there is no reason to suspect such trends explain these findings, as nationwide there is still concern that far too many invasive angiograms show no significant coronary disease (15).

During the duration of this study, we have witnessed a change in stress modalities. Over these 2 decades, pharmacologic stress (mostly vasodilator) increased from less than one-third in the first epoch to more than one-half in the last. It is unlikely that this shift is behind the declining rate of abnormal perfusion. First, the increase in myocardial blood flow is more robust with vasodilators than with exercise, and vasodilator MPI should be at least as sensitive for detecting ischemia as exercise MPI is. Second, prevalence of abnormal perfusion was higher among patients undergoing pharmacologic MPI versus exercise MPI in this study. In fact, pharmacologic (vs. exercise) MPI was independently associated with an abnormal MPI in all 4 epochs with increasing odds ratio (for 2006 to 2009: 2.63, p < 0.001). Also, the decline in abnormal MPI was at least as dramatic (if not more so) in the exercise subgroup versus the pharmacologic subgroup.

Finally, as pointed out earlier, MPI has been constantly changing over time. It is certainly possible that a change in interpretation threshold for an abnormal MPI is responsible for the decline in ischemia detection due to technical advancements in hardware, software, and imaging protocols even though the interpretation was done by same readers. It is interesting that most of the decline for abnormal perfusion occurred in the first decade with little decline after 2001 ((Figure 09_gr1) in Rozanski et al. [5]). Another interesting observation is that mortality rates although declined during the same period did so at a much slower rate (Figure 09_gr1). The ratio of abnormal perfusion/annual mortality was 9.8:1 in the first period and 3.6:1 in the last (Table 1 in Rozanski et al. [5]). Because outcomes are provided for the overall population and not by perfusion pattern, it is difficult to interpret these findings. Our analysis of the data shows that if risk of death for patients with normal MPI remained constant over time, then risk of death in those with abnormal MPI has more than doubled even in the setting of decreasing overall risk ((Figure 09_gr2), scenario A). Alternatively, relative risk for abnormal/normal MPI could have remained constant over time if risk of both normal and abnormal MPI has increased (Figure 09_gr2, scenario B). The first possibility is compatible with better risk stratification using MPI in the current era, whereas the second is compatible with changing referral patterns. The more likely possibility is that these trends reflect a combination of these 2 scenarios. Obviously, in the absence of data, this analysis is purely speculative.

The implications of this study could be far-reaching if they are true nationwide. The registry planned by the American Society of Nuclear Cardiology could provide this much-needed data. What lies in the balance is defining a population for which MPI is still cost-effective if the rate of detected abnormalities is low and decreasing. For example, in this study, the frequency of ischemia in patients without typical angina referred for exercise MPI and who achieved target heart rate in the contemporary cohort was very low. It may arguably be more effective to perform exercise treadmill testing on these patients without MPI as an initial test. A recent study by Bourque et al. (16) showed that patients who have high exercise capacity with no ischemic ST-segment changes are very unlikely to have ischemia on MPI. Such an approach will be useful only in the subset of patients suitable for exercise stress, an ever-shrinking proportion of patients. We have recently introduced a novel concept that may help in better risk-stratifying patients with normal perfusion by vasodilator MPI (17). Another implication is that it may become increasingly difficult to recruit patients for clinical trials, whether they involve new imaging agents, stress agents, or evaluation of the impact of treatment (drug, device, revascularization, gene therapy, or treatment strategy) on ischemia reduction. The large National Heart, Lung, and Blood Institute–sponsored ISCHEMIA (International Study of Comparative Health Effectiveness With Medical and Invasive Approaches) trial will be a good test for this new hurdle as it will specifically target subjects with moderate-to-large areas of ischemia.

We are grateful to Rozanski et al. (5) for this contribution, as we were 3 decades ago to their other seminal contribution of declining specificity due to referral bias and their introduction of the term normalcy rate (2).

⁎Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.