Curr Cardiol Rep (2014) 16:472 DOI 10.1007/s11886-014-0472-9

NUCLEAR CARDIOLOGY (V DILSIZIAN, SECTION EDITOR)

Assessing the Prognostic Implications of Myocardial Perfusion Studies: Identification of Patients at Risk vs Patients who May Benefit from Intervention? Paul Cremer & Rory Hachamovitch

# Springer Science+Business Media New York 2014

Abstract Stress myocardial perfusion imaging (MPI) has a well-established role in improving risk stratification. Recent analyses, compared with older data, suggest that the yield of stress MPI has decreased. In part, this trend relates to testing patients with heterogeneous, but improved, risk factor modification. In this setting, positron emission tomography with myocardial flow reserve enhances risk stratification as it reflects the end result of atherosclerosis. Recent studies have also emphasized the clinical impact of incremental risk stratification by assessing net reclassification improvement (NRI). Previous retrospective studies have favored an ischemic threshold to select patients that benefit from revascularization, but this finding has not been corroborated in randomized trials. However, no large randomized trial has directly tested a strategy of revascularization for patients with at least a moderate amount of ischemia at risk. Unfortunately, even when faced with a significantly abnormal MPI result, subsequent action is too often absent.

Keywords Coronary artery disease . Diabetes mellitus . Early revascularization . Myocardial perfusion imaging . Single photon emission computed tomography . Positron emission tomography . Net reclassification improvement . Myocardial flow reserve

Introduction For years, stress myocardial perfusion imaging (MPI) has aided in the accurate diagnosis of obstructive coronary artery disease (CAD) and refined the risk for future cardiac events. In the present healthcare environment, however, the utility of an imaging test is no longer defined in terms of incremental prognostication. Instead, a test is only useful to the extent that it also selects patients that will benefit from specific treatments. In this context, we review single photon emission computed tomography (SPECT) and positron emission tomography (PET) MPI with 3 questions in mind: (1) Is risk stratification achieved? (2) Can MPI select patients that benefit from revascularization? (3) Are MPI results used in clinical practice? With regard to the first question, the importance of baseline clinical risk is emphasized, and recent studies incorporating net reclassification improvement (NRI) as well as advances in PET myocardial flow reserve (MFR) are highlighted. The decision to revascularize patients with chronic CAD with or without left ventricular systolic dysfunction is framed within the discord between observational studies and randomized trials. Finally, the issue of downstream clinical decisions is discussed with new data from the SPARC registry.

Risk Stratification This article is part of the Topical Collection on Nuclear Cardiology P. Cremer : R. Hachamovitch (*) Section of Cardiovascular Imaging, Department of Cardiovascular Medicine, Heart and Vascular Institute, Cleveland Clinic, Desk J1-5 9500 Euclid Avenue, Cleveland, OH 44915, USA e-mail: [email protected]

Importance of Pre-Test Risk Assessment and Limitation of Binary Variables The incremental prognostic value of MPI abnormalities when added to clinical and exercise testing variables has been well

472, Page 2 of 8

Curr Cardiol Rep (2014) 16:472

established. Risk associated with specific MPI results is clearly a function of pre-imaging risk factors [1–3]. However, until recently, no study had assessed on a large scale the temporal trends in risk factors and SPECT results. In a single center study of 39,515 patients without known CAD, the rates of abnormal tests decreased from 40.9 % in 1991 to 8.7 % in 2009 [4••]. However, during the same period, the calculated pre-test likelihood of CAD actually increased from 40.1 % to 49.2 %. Over this time, there was increased use of aspirin, lipid lowering agents, and anti-hypertensives. Consequently, contemporary SPECT stress testing may be performed in patients in whom risk factor modification has been applied.

Reclassifying Risk with SPECT and PET MPI Over the past 3 decades, adequate risk stratification with MPI has been based upon the foundation that a normal scan equates low risk with risk increasing as results become increasingly abnormal. With this background, enhanced risk stratification has been an iterative process with a recent focus on assessing the net reclassification of risk when MPI abnormalities are added to traditional risk factors. NRI, compared with conventional metrics that also inform prognosis, highlights the clinical importance of risk assessment. Specifically, NRI expresses how a test changes assigned risk category with the implicit assumption that this change impacts subsequent management. Recent studies evaluating NRI with SPECT and PET are shown in Table 1. In an analysis of 4575 patients from the Myoview Prognosis Registry, ischemia on a stress SPECT had incremental prognostic value in predicting annual cardiac death or myocardial infarction when assessed in conjunction with Duke Treadmill Score (DTS) and pre-test probability of CAD [5•]. With the addition of MPI ischemia, NRI was robust. 35.8 % of patients changed risk category, which included 5 %. Newly identified events were more likely to have had pharmacologic stress testing (82.1 %), intermediate to high pre-test likelihood of CAD (75.5 %), or an intermediate DTS. Conversely, newly

identified non-events more frequently had exercise testing (41.6 %), 0 or 1 cardiac risk factor (56.1 %), and less frequently had angina (89.8 %). Similar refinement of risk assessment has been studied in PET MPI. A recent multicenter registry of 6037 patients demonstrated the value of PET abnormalities to predict cardiac death [6•]. When percent scarred or ischemic myocardium was added to a model that included numerous clinical variables, the NRI for cardiac death was 11.6 %. Risk categories for annual cardiac death were defined as low (3 %). This threshold for risk assignment is well accepted and similar to the study previously referenced [5•, 7]. However, in the context of the practicing physician, NRI is most useful if a change in risk category affects downstream clinical decisions that then improve patient outcomes. The ability of these risk categories to provide that differentiation is unclear. NRI has additional limitations that warrant discussion. First, all reclassifications are treated equally. However, clinically, a false negative test that precludes prevention of a cardiac death or a myocardial infarction (MI) is much more devastating than a false positive test that then leads to negative confirmatory testing. Moreover, the risk categories and test thresholds chosen have enormous impact on the number of reclassifications. NRI will be greatest when assessed among numerous risk categories with small absolute variations for a test with a low threshold for abnormal. In such a scenario, even though incremental prognostication will be impressive, clinical impact will be limited.

Improved Risk Stratification with Quantitative Myocardial Blood Flow An advantage of PET is quantitative analysis of myocardial blood flow and calculation of MFR upon pharmacologic stress. MPI with SPECT, or PET without quantitative measurement, depends on flow heterogeneity. Given this limitation, research has focused on whether MFR can provide incremental risk

Table 1 Recent SPECT and PET MPI studies emphasizing NRI for risk stratification

Patients Model Test characteristic NRI Risk categories

Shaw et al. [5•]

Dorbala et al. [6•]

Murthy et al. [9•]

4575 patients 32 % with prior CAD SPECT ischemia added to Duke treadmill score and pre-test CAD likelihood >5 % ischemia 35.8 % 4

6037 patients 27 % with prior PCI or CABG PET ischemia or scar added to clinical variables >10 % ischemia or scar 11.6 % 3

2783 patients 42 % with prior CAD PET MFR added to clinical variables, rest and stress LVEF, ischemia and scar MFR 10 % ischemia. In 29.8 % of patients with events, stress MPI showed normal perfusion and lacked any other high-risk finding including reduced left ventricular ejection fraction (LVEF), significant wall motion abnormalities, increased end-systolic volume, or transient ischemic dilation. Given this concern for misclassification of high-risk patients with SPECT MPI, risk stratification using PET MFR has been addressed in a study of 2783 diabetics [18••]. A significant increase in annual cardiac mortality was noted in patients without relative perfusion defects if MFR was reduced (0.4 % vs 3.5 %). Furthermore, when diabetic patients without known CAD but with impaired MFR were compared with nondiabetics with known CAD, rates of cardiac death were similar (2.8 %/year and 2.0 %/year, respectively). Among diabetic patients without CAD and with preserved MFR, rates of death were similar to patients without CAD or diabetes and with preserved MFR (0.3 %/year and 0.5 %/year, respectively; Figure 1). Since cardiovascular disease is the

Fig. 1 Annualized cardiac mortality stratified according to CAD, diabetes mellitus (DM), and MFR. (Adapted with permission from Murthy VL, Naya M, Foster CR, et al. Association between coronary vascular dysfunction and cardiac mortality in patients with and without diabetes mellitus. Circulation. 2012;126:1858–68) [18••]

472, Page 4 of 8

major driver of mortality in diabetes, improved risk stratification with MFR could be particularly valuable. PET MFR: Reflecting the Final Common Pathway Across the Spectrum of Clinical Risks Furthermore, intuitively, the cardiovascular risk conferred by diabetes is a continuous variable related to the duration and severity of the illness. The same is true for other cardiovascular risk factors. A patient with well-controlled hyperlipidemia for 5 years is certainly at lower risk than a like patient with uncontrolled hyperlipidemia for 20. Hence, as discussed previously, risk assessment using clinical diagnoses as binary variables may be increasingly less reliable. PET MFR, to the extent that it more accurately reflects atherosclerosis and thrombosis across the coronary vasculature, may represent a common end result of many clinical risk factors. Accordingly, particularly in the current era of heterogeneous risk factor modification, PET MFR may continue to enhance risk stratification. Diabetics with CAD: Ongoing Dissonance Between Risk Conferred and Treatment Offered The management of diabetics with known CAD has been the focus of many clinical trials, and BARI-2D demonstrated no difference in death among patients assigned to revascularization vs medical management [19]. However, diabetics who underwent CABG had decreased major adverse cardiac events when compared with medical therapy, even though 5-year event rates were still high, 22.4 % vs 30.5 %. Given this particularly high-risk, there has been interest in continued risk stratification. The nuclear substudy of BARI-2D consequently assessed cardiovascular risk in 1505 patients who underwent SPECT MPI at 1 year following randomization [20•]. Importantly, all SPECT MPIs were interpreted at a core lab, and 1-year testing rates did not vary by randomization to revascularization or medical therapy. A significant association between the severity of abnormal MPI results and risk of subsequent cardiovascular events was noted. Cox proportional hazards modeling revealed the percent of fixed perfusion defect was the strongest predictor of cardiac death or MI (HR 1.16). The percent of abnormal myocardium was also significantly associated with adverse events (HR 1.11); of note, the amount of ischemic myocardium was not predictive. This finding again illustrates the frequent disconnect between the result of most prognostic significance and the reason a test is ordered. Specifically, fixed myocardial defect may better inform cardiac death, but may or may not represent modifiable substrate. Alternatively, ischemic myocardium is more readily addressable, though in certain clinical situations, the best treatment remains murky.

Curr Cardiol Rep (2014) 16:472

MPI to Guide Revascularization in Chronic CAD Early Revascularization for Moderate to Severe Ischemia on MPI The simple question of whether stress MPI can identify patients with stable CAD that benefit from early revascularization continues to plague clinicians. This uncertainty arises primarily from the disconnect between positive observational studies that can be affected by unmeasured covariates and negative randomized controlled trials where nuclear substudies have been underpowered. In a landmark observational study of 10,627 patients with suspected CAD that underwent SPECT MPI, patients with no or a mild amount of inducible ischemia had a survival advantage with medical therapy [21]. Conversely, at a value of >10 %–12.5 % ischemic myocardium, there was a survival benefit for revascularization. Moreover, revascularization seemed to neutralize the poor prognostic impact of ischemia. These results were expanded in a follow-up study including 13,555 patients followed for a median of 8.7 years [22•]. This study included 4758 patients with known CAD and had 3893 deaths. Again, patients with more than ~12.5 % ischemic myocardium had improved survival with early revascularization. This interaction occurred in patients with previous revascularization but not in those with a previous MI. However, when patients with >10 % fixed myocardial defect were excluded from the analysis, extensive ischemia did identify a survival benefit with early revascularization among patients with a prior MI. These results suggest that the magnitude of possible scar attenuates the benefit of revascularization identified by significant ischemia. Randomized Trials With Nuclear Substudies Underpowered to Test an Ischemic Threshold Despite over a decade of observational data that supports an ischemic threshold to identify patients that will benefit from early revascularization, no prospective trial has confirmed these findings. In contrast, the COURAGE trial showed no difference in death or nonfatal myocardial infarction when patients were treated with medical therapy or PCI plus medical therapy [23]. Although 70 % of patients in COURAGE had nuclear imaging, the nuclear substudy of COURAGE included only 314 patients that had pretreatment as well as 6–18-month follow-up gated SPECT MPIs [24]. Importantly, studies were assessed at a core laboratory by nuclear cardiologists blinded to site and randomization. Based upon observational reports, the primary endpoint was >5 % ischemia reduction [10, 25]. Among patients who underwent PCI, there was a statistically significant increase in ischemia reduction (33 % vs 19 %, P=0.0004). The nuclear substudy of BARI-2D yielded similar results [20•]. There were less stress perfusion abnormalities in patients assigned to

Curr Cardiol Rep (2014) 16:472

revascularization vs medical management (3 % vs 9 % of myocardium, P=0.01). Upon exploratory analysis of the COURAGE nuclear substudy, the magnitude of residual ischemia on follow-up MPI was proportional to the risk of death, though this result was considered hypothesis-generating as the study was underpowered for this analysis. The nuclear core lab substudies were also not designed to assess an ischemic threshold where revascularization is preferred over medical therapy alone. In the COURAGE substudy, only one-third of patients had moderate to severe ischemia. Furthermore, both studies had notable rates of crossover from medical therapy to PCI that will bias the intentionto-treat analysis in favor of the null hypothesis. In the BARI2D substudy, 16 % of patients assigned to medical therapy underwent revascularization within 1 year; in the COURAGE substudy, 21 % of patients had PCI after their second SPECT MPI [20•, 24]. Discrepancies Between On-Site and Core Lab SPECT Studies A larger post-hoc substudy of COURAGE included 1381 patients who had stress SPECT MPI. Similar to the main study, there was no difference in 5-year death rates [26•]. In patients with moderate to severe ischemia, the death rate was 21 % with medical therapy alone whereas it was 24 % among patients who also had PCI. However, unlike the previously mentioned substudies, SPECT MPI results were interpreted on-site. Moreover, a 6-segment model was used that is known to be less accurate than quantitative assessment with a 17segment model [27]. Finally, contrary to many studies previously discussed, there was no graded relationship between the number of ischemic or scarred segments and clinical outcome. Accordingly, the generalizability of these on-site SPECT MPI studies in relation to studies interpreted in core labs or in previously published cohorts is questionable. A Randomized Controlled Trial to Test the Concept of an Ischemic Threshold Given these limitations, whether MPI can identify the patient that will benefit from early revascularization is still unknown. The ISCHEMIA trial has been designed for this purpose with the intention to enroll 8000 patients with moderate to severe ischemia. Eligible patients will undergo a blinded coronary CT angiogram (CCTA) and will then be randomly assigned to medical management or medical management as well as cardiac catheterization and revascularization as indicated [28••]. Of note, unlike other trials, patients assigned to medical management will not have a diagnostic cardiac catheterization. A small prospective study has demonstrated an agreement of 92 % regarding the decision to revascularize based upon SPECT and CCTA when compared with SPECT and invasive coronary angiography [29]. A feasibility assessment also

Page 5 of 8, 472

suggested that this approach was acceptable to cardiologists, though this method does represent a paradigm shift from MPI as a gatekeeper to MPI followed by CCTA as a definitive evaluation [30]. Revascularization with Impaired Left Ventricular Systolic Function Patients with severe CAD resulting in significant left ventricular dysfunction have increased morbidity and mortality as well as increased complications upon revascularization [14]. Consequently, there has been intense interest to guide invasive management based upon the identification of recoverable myocardium. Unfortunately, observational and randomized trials have again been discordant. The PARR-2 trial randomized 430 patients with an LVEF