Stress CMR, Revascularization, and All-Cause Mortality: Do We have a Final Answer?

Stress first pass cardiac magnetic resonance imaging (CMR) perfusion imaging has had a growing clinical impact since the early feasibility and validation studies of the 1990’s.¹ First, large single center studies such as the CE-MARC trial (Clinical Evaluation of Magnetic Resonance imaging in Coronary heart disease) demonstrated its greater specificity and negative predictive value as compared to single photon emission computed tomography (SPECT).² Then, 2 large multicenter studies established the diagnostic accuracy and clinical utility of stress CMR. The MR-INFORM (The Myocardial Perfusion CMR versus Angiography and FFR to Guide the Management of Patients with Stable Coronary Artery Disease) study compared invasive and noninvasive strategies in the management of patients presenting with angina chest pain.³ Nine hundred ten patients were randomized to invasive angiography with fractional flow reserve (FFR) measures or stress perfusion CMR and followed for outcomes for 1 year. There were fewer revascularizations (36% vs. 45%, p-0.005) in the CMR arm with no difference in adverse cardiac events or freedom from angina at follow-up, suggesting that the noninvasive approach was at least equivalent, if not better. Finally, the recent GadaCad (Gadubutrol-enhanced CMR to detect Coronary Artery Disease) trials combined 2 phase III clinical trials in 764 patients from 47 centers world-wide.⁴ Stress CMR had sensitivities of 79% and 87% and specificities of 87% and 73% for the detection of single vessel and multi-vessel CAD respectively using 70% stenosis defined by angiography, leading to U.S. Food and Drug Administration of approval of gadobutrol for this indication. Previously, gadolinium chelates had been used off label for stress perfusion CMR in the U.S.

The next frontier for stress perfusion CMR has been assessing its prognostic value. It is against this backdrop that the paper by Pezel et al is presented in this issue of Circulation Cardiovascular Imaging.⁵ This paper synthesizes data from their team’s extensive database over the last several years from subgroups of the same patient population with different baseline characteristics studied in their prior publications.^6–9 The present study examines 31,752 consecutive patients who underwent stress perfusion CMR between 2008 and 2018. Of these, 98% completed the protocol and 98% of those had diagnostic quality imaging, both of which are quite remarkable. The median follow-up of 6.0 years is likewise impressive with over 200,000 patient years of follow-up. This makes this particular study the largest CMR study of its kind to date, single center or otherwise.

The authors examined the effect of the CMR imaging findings on all-cause mortality as well as the effect of coronary revascularization based on the imaging findings on the same endpoint. They demonstrated that CMR findings of inducible ischemia and late gadolinium enhancement (LGE) in an infarct pattern were independent predictors of mortality with similar hazard ratios (HR) of approximately 1.6. Revascularization was an independent predictor of survival with a HR of 0.58. In a subset of patients matched for baseline characteristics, the benefit of revascularization remained, but was evident only in patients with severe ischemia (>6 segments) but not in those with less ischemia (<6 segments).

The authors should be congratulated for carrying out this tour-de-force study in such a large population. Certainly, it would have been advantageous to have data in regards to cardiovascular-specific mortality as stress CMR would not be expected to predict death from cancer or other causes. However, in a study this large, it is quite difficult and costly to perform the ascertainment of cause of death short of querying a national death registry, which is what was done in this case. The relationship between inducible ischemia and LGE and subsequent mortality confirms findings in other registries such as the SPINS (Stress CMR Perfusion Imaging in the U.S.) registry of 2349 patients from 13 centers which demonstrated excellent risk reclassification of stress CMR for cardiac death or nonfatal myocardial infarction.¹⁰ Another large multi-center cohort study of 9000 patients with over 48,000 patient-years of follow-up showed that stress CMR was an independent predictor of mortality in the overall group, as well as several subgroups.¹¹ Similarly, a prior meta-analysis by our group of nearly 12,000 patients showed that the presence of ischemia was associated with an odds ratio of 6.5 for MI and cardiovascular death and LGE had a similar association.¹²

The more difficult part of the study is teasing out the effect of revascularization. There are multiple reasons that patients don’t go on to revascularization, including co-morbidities such as co-existing malignancy, severity and extent of CAD, etc. and it is by no means surprising that these patients are much sicker than those that get revascularized. The authors performed propensity-matching in their multivariable analysis but this can’t possibly account for all of the differences between those who do or do not get revascularized. In fact, the non-revascularized group had fewer segments with LGE and number of ischemic segments, and a trend towards less hypertension, raising questions about how well the groups were matched. Another question was why 13% of the patients with ischemia didn’t get sent for angiography and thus could not be revascularized? These patients, included as part of the non-revascularized group, likely had substantial co-morbidities and thus, it is not unexpected that they would have worse outcomes.

The findings that CMR-guided revascularization reduced mortality is inconsistent with the findings of the ISCHEMIA trial in which patients with significant ischemia were randomized to invasive or medical therapy.¹³ However, CMR was performed considerably less frequently in ISCHEMIA than were other stress imaging modalities. Regardless, medical therapy was shown to be equivalent to revascularization when ischemia was demonstrated. The findings in the present study that benefit was limited to those patients with >6 segments of ischemia is potential fodder for the design of future randomized trials of patients with profound ischemia such as these.

So, is the final answer in? Yes, in the case of stress CMR predicting all-cause mortality using the presence of ischemia and LGE. The wealth of single and multi-center registries and meta-analyses suggest that this is indeed the case. MR INFORM tells us that the absence of ischemia is certainly a reason to not perform catheterization and/or revascularization. As to whether revascularization reduces mortality, randomized studies such as ISCHEMIA suggest that medical therapy is equivalent to revascularization and observational studies, even with propensity matching, are no match for randomization. Thus, in this instance, the verdict is still out. Randomized studies with entry criteria using ischemia and the extent of ischemia as identified by stress CMR would be useful in this regard.