Coronary Microvascular Dysfunction Is Associated With Myocardial Ischemia and Abnormal Coronary Perfusion During Exercise

Supplemental Digital Content is available in the text.


METHODS
The data that support the findings of this study are available from the corresponding author on reasonable request.

Study Population
Patients undergoing elective diagnostic angiography for investigation of exertional chest pain were enrolled in the study ( Figure 1). Inclusion criteria were preserved LV systolic function (ejection fraction >50%) and unobstructed coronary arteries (<30% diameter stenosis and/or fractional flow reserve >0.80). Exclusion criteria were intolerance to adenosine, chronic kidney disease (estimated glomerular filtration rate <30 mL/min/m 2 ), concomitant valve disease (greater than mild on echocardiography), recent acute coronary syndrome, cardiomyopathy, or any neuromuscular comorbidity that may affect their ability to perform bicycle exercise. Antianginal medications were stopped, and patients abstained from caffeine 24 hours before all study visits. Subjects gave written informed consent in accordance with the protocol approved by the UK National Research Ethics Service (17/LO/0203). The study was registered with the National Institute for Health Research UK Clinical Research Network portfolio database (Central Portfolio Management System identifier: 33170).

Catheterization Protocol
Catheterization was performed through the right radial artery using standard coronary catheters. All patients received 1 mg intravenous midazolam and intraarterial unfractionated heparin (70 U/kg) before intracoronary physiological measurements. A dual-pressure and Doppler sensor-tipped 0.014-inch intracoronary wire (Combowire, Philips Volcano, California) was used to measure distal coronary pressure and average peak flow velocity in the left anterior descending artery, as previously described. 10 Hemodynamic measurements were recorded under resting conditions, during intravenous adenosine-mediated hyperemia (140 mcg/kg / min) and continuously during bicycle exercise, using a specially adapted supine ergometer (Ergosana, Bitz, Germany) attached to the catheter laboratory table (Figure 2). Exercise began at a workload of 30 W and increased every 2 min by 20 W. Where muscle weakness restricted increasing workloads, resistance was fixed at the maximum tolerated level and exercise continued until exhaustion. After full recovery from exercise or hyperemia, a second set of resting hemodynamic data was acquired before the final condition.

Analysis of Coronary Physiological Data
Signals were sampled at 200 Hz, with data exported into a custom-made study manager program (Academic Medical

Clinical Perspective
What Is New?
• In patients with angina without obstructive coronary artery disease, diminished coronary flow reserve in response to pharmacological vasodilatation identifies those with a maladaptive physiological response to exercise and global myocardial ischemia. • There are 2 distinct endotypes of microvascular dysfunction: functional and structural, with differing degrees of systemic disease involvement and distinct mechanisms of ischemia.
What Are the Clinical Implications?
• Patients with angina and nonobstructive coronary artery disease can have exercise pathophysiology and global myocardial ischemia; the measurement of coronary flow reserve will help to characterize this population. • Not all microvascular dysfunction is mechanistically identical; distinct endotypes may have differing prognosis and warrant individualized therapies. • Targeting specific mechanistic abnormalities may improve the poorer prognosis observed in this population and would need to be studied in therapeutic trials.
Centre, University of Amsterdam, Netherlands) and analyzed on custom-made software, Cardiac Waves (Kings College London, UK). Microvascular resistance (MR) was calculated as distal coronary pressure divided by flow velocity for each condition. Hyperemic MR was dichotomously classified as normal (< 2.5 mmHg/cm/s) or elevated ≥2.5 mmHg/cm/s. 12 Wave intensity was calculated as the product of the derivatives of distal coronary pressure and flow velocity, each with respect to time (d(distal coronary pressure)/dt × d(flow velocity)/dt), and wave separation performed as previously described. 13 For each patient, 4 dominant waves were identified and included in our analysis: (1) backward compression wave (BCW), causing flow deceleration during isovolumetric contraction in early systole; (2) forward compression wave, causing flow acceleration, associated with peak aortic pressure; (3) forward expansion wave, causing flow deceleration associated with the fall in aortic pressure in late systole; and (4) backward expansion wave (BEW), causing flow acceleration during isovolumetric relaxation in early diastole. Perfusion efficiency was calculated as the percentage of accelerating wave intensity in relation total wave intensity, using areas under the respective curves. Patients were classified offline as controls or more correctly as having normal CFR (CFR ≥2.5) and MVD (CFR <2.5) with researchers blinded to this classification throughout the study protocol.

High-Resolution Cardiac Magnetic Resonance Imaging Protocol
All scans were performed on a dedicated 3-Tesla cardia magnetic resonance (CMR) scanner (Achieva, Philips Healthcare, Netherlands). Contiguous short-axis slices were acquired from the base to the apex to calculate LV function and mass (CVI42, v5.1.1, Circle Cardiovascular Imaging, Calgary, Ontario, Canada). After 3 minutes of intravenous adenosine (140 mcg/kg / min), stress perfusion data were acquired in 3 short-axis slices using a saturation-recovery k-t sensitivity encoding accelerated gradient-echo method. 14 A dualbolus gadobutrol (Gadovist, Bayer, Berlin, Germany) contrast agent scheme was used to correct for signal saturation of the arterial input function as previously described. 15 Resting perfusion imaging was performed 15 minutes after stress, before acquisition of late gadolinium enhancement imaging (total contrast agent dose 0.2 mmol/kg). 16 A proton density acquisition was performed before stress and rest acquisitions to correct for spatial inhomogeneities of surface coils. 17 Before quantitative analysis, the perfusion images were motion corrected according to published methods. 18 Quantitative analysis was performed as previously described by Fermi-constrained deconvolution. 19 Myocardial blood flow estimates (MBF) were quantified in ml/min / g during rest and hyperemic stress; myocardial perfusion reserve was defined as the ratio between stress and rest perfusion. Endocardialto-epicardial-perfusion ratios were calculated during hyperemic stress and during rest, by dividing stress MBF estimates within the inner and outer layers of myocardium. Inducible ischemia was defined as endocardial-to-epicardial perfusion ratio <1.0 during hyperemia (subendocardial relative hypoperfusion during pharmacological vasodilation). 11 NT-proBNP (N-terminal pro-B-type natriuretic peptide) levels were measured before angiography using conventional clinical assays. Diastolic function was assessed using standard transthoracic echocardiography.

Statistical Analyses
Sample size was estimated on the basis of the coprimary end points, exercise perfusion efficiency and frequency of ischemia. Assuming an allocation ratio of 1:1, 66 patients would be needed to detect a minimum difference in exercise perfusion efficiency of 12% (predicted SD, 17%) and 58 patients to detect a 50% relative decrease in the frequency of ischemia (predicted rate in MVD, 70%) at 80% power and 5% significance. To allow for potentially unequal allocation, data censoring as a result of quality issues, and incomplete datasets, we aimed to enroll 85 patients. Statistical analysis was performed using SPSS version 24 (IBM Corp., Armonk, New York). Normality of data was visually assessed (using histograms and the normal Q-Q plot) and using the Shapiro-Wilk test. Continuous normal data are expressed as mean±SD and compared using paired or unpaired Student t tests as appropriate. Nonnormal data are expressed as median±interquartile range and compared using Mann-Whitney U test, and categorical variables were compared with chi-square tests. Twosided P-values <0.05 were considered nominally significant, with no correction for multiplicity of testing. Baseline variables found to correlate with exercise perfusion efficiency or inducible ischemia on univariate analysis (P<0.05) were assessed by a multiple linear regression model.

Patient Characteristics
Eighty-five patients were recruited into the study (45 were classified as having MVD and 40 had normal CFR). Groups were well-matched for cardiovascular risk factors and preprocedural medications, whereas there were more women in the MVD group (87% vs 68%; P=0.03) ( Table 1). All patients had resting and hyperemic coronary physiology measurements, 68 patients performed the full bicycle exercise protocol, and 65 underwent the 3-Tesla perfusion CMR protocol ( Figure 1). The MVD and normal CFR groups exercised for the same duration (394±109 vs 408±146 s; P=0.67) and workload (63±27 vs 71±27 W; P=0.25).  By wave intensity analysis, both groups had similar profiles at rest. During exercise and hyperemia, the magnitude of all waves increased from rest, although the relative changes in accelerating and decelerating waves was significantly different in the 2 groups. Typical coronary pressure and flow waveforms, with corresponding wave intensity analysis profiles during peak exercise, are shown in Figure 3. In those with normal CFR, perfusion efficiency was enhanced during exercise (from 59±11% to 65±14%; P=0.02) and remained unchanged during hyperemia (57±18%; P=0.14) (Figure 4a). In contrast, in patients with MVD, perfusion efficiency decreased from rest (61±12%), during peak exercise (44±10%) and hyperemia (42±11%; P<0.001 for both conditions). The difference between normal CFR and MVD groups was mainly in the microcirculation-derived backward waves rather than in the aorta-derived forward waves; at peak exercise, patients with MVD had proportionately larger decelerating BCW (36±11 vs 24±12%; P<0.001) and smaller accelerating BEW (30±6 vs 46±12%; P<0.001) than did those with normal CFR (Figure 4b).

DISCUSSION
This is the first study to directly assess coronary blood flow during exercise in patients with microvascular dysfunction and compare these changes with high-resolution perfusion imaging. Among patients with angina and unobstructed coronary arteries, those with diminished coronary flow reserve have a higher prevalence of inducible myocardial ischemia and reduced global perfusion reserve in addition to reduced coronary perfusion efficiency during physical exercise compared with controls with preserved coronary flow reserve. We distinguished 2 microvascular dysfunction endotypes, those with low resting microvascular resistance (functional) and those with high minimal microvascular resistance (structural), both displaying similar myocardial perfusion characteristics and distinct exercise pathophysiology. The relationship with diminished coronary vasodilator reserve and these pathophysiological changes may underlie the poorer prognosis exhibited by patients with microvascular dysfunction and whether each endotype warrants individualized therapies will need to be studied further.

Diminished Flow Reserve
The historical perception of MVD is that high vascular resistance during periods of increased demand or pharmacological vasodilatation is the key mechanistic abnormality. This structural MVD endotype may represent architectural changes such as capillary rarefaction and downstream myocardial hypertrophy and fibrosis. Indeed, there are several physiological parallels with heart failure with preserved ejection fraction, where coronary blood flow is elevated during rest and similarly diminished during vasodilator hyperemia. 20 However, this state of raised minimal MR may be surmountable over time with appropriate treatment and so the term structural should not be considered to imply an irreversible disease process. Many authors have suggested that high minimal microvascular resistance is a more reliable tool for identification of patients with MVD and that this should replace measurements such as CFR, where resting conditions are integrated particularly in a population where there is mixed epicardial artery disease. 21 However, in ORIGINAL RESEARCH ARTICLE our NOCAD population, 62% of patients with MVD had preserved minimal MR (or functional MVD) despite a diminished CFR, yet they displayed a similar pathological phenotype as the traditional structural MVD group. The finding of low resting MR suggests that this is a functional disorder rather than being primarily the result of structural vascular changes leading to elevated minimal MR. This is consistent with recent observations by other groups where elevated myocardial and coronary blood flow was the likely mechanism of diminished flow reserve in MVD. 22,23 In this context, elevated resting flow could be a response to increased myocardial oxygen demand or represent disordered autoregulation. We did not find a statistically significant difference in mechanical cardiac work at rest, as estimated by ratepressure product, but it should be noted that myocardial oxygen demand is also dependent on basal metabolic activity and calcium cycling. 24 High baseline MBF has been proposed to be due to increased myocardial oxygen consumption secondary to myocardial metabolic derangements, in studies of patients fitting the historical Cardiac Syndrome X definition. [25][26][27] Diminished flow reserve in patients with diabetes mellitus has been shown to be due to elevated resting flow in the early stages of the disease. 22 This may be a response to increased basal oxygen consumption as the metabolic shift in diabetics, from glucose to fatty acid oxidation, results in fewer ATP molecules produced per molecule of oxygen consumed; this theory would need to be explored further in MVD. While increased resting flow can lead to a reduction in vasodilator capacity, it may also play a role in disease progression. Chronically raised coronary blood flow may precipitate structural vascular changes, a process commonly found in other organs, such as the renal or pulmonary vascular beds. 28,29 This may also explain the bimodal distribution of microvascular resistance found in patients with diabetes mellitus, with an early functional disorder progressing to structural changes. 22 Whether functional MVD represents a precursor to structural MVD and early intervention may prevent disease progression is currently unknown. At present, no diseasemodifying therapies exist specifically for MVD.

The Mechanism of Ischemia
Within the structural MVD group, exercise-induced hypertension leads to increased myocardial oxygen

ORIGINAL RESEARCH ARTICLE
demand during exercise. Attenuated reduction in afterload with exercise would interrupt the usual synergistic response of the coronary and peripheral circulations and predispose to ischemia; in theory, large vessel vasodilators may enhance the normal synergistic adaptation to exercise preferentially among this disease endotype. 30 The structural MVD group have more established cardiovascular risk factors and poorly controlled hypertension, a process similarly associated with diminished maximal flow. 31 In functional MVD, minimal microvascular resistance during stress is preserved compared with those with preserved CFR; therefore, this measurement cannot account for the ischemic changes identified during perfusion imaging. Wave intensity analysis provides unique insight into cardiac-coronary coupling among both MVD endotypes. Whereas coronary perfusion efficiency increases or is maintained with exercise and hyperemia in the healthy heart, in MVD perfusion efficiency decreases with both forms of stress. 10 Reduced perfusion efficiency in MVD demonstrates that greater energy needs to be expended to achieve the same degree of coronary blood flow augmentation, providing an ischemic substrate in both functional and structural MVD groups. The decreased perfusion efficiency with exercise in MVD is primarily determined by attenuated augmentation of the BEW, which is usually the main driver of coronary perfusion in health. The BEW is generated during ventricular lusitropy, thus flow acceleration will be diminished when there is diastolic dysfunction. 32 Conversely, subendocardial hypoperfusion results in diastolic dysfunction, while transmural hypoperfusion results in overt systolic and diastolic dysfunction. 33 While establishing causality remains challenging because of the intricate interplay of these parameters, exercise induced diastolic dysfunction and ischemia will amplify each other in a potentially deleterious cycle. In our study, both patient groups had similar resting diastolic function in terms of resting echocardiography while a trend toward increasing NT-proBNP values from patients with normal CFR, to functional MVD and structural MVD may support the notion of temporal progression of MVD disease states.
During systole, compression of the subendocardium causes retrograde filling of the subepicardial vessels, therefore antegrade subendocardial filling occurs exclusively in diastole. Myocardial oxygen demand is increased in the subendocardium, increased vasculature within this layer ensures perfusion is maintained throughout the cardiac cycle and indeed in health, pharmacological vasodilatation maintains hyperperfusion within the subendocardial layer. 34 We have demonstrated a reduction in coronary perfusion efficiency during vasodilator hyperemia in MVD, accentuation of the BCW and attenuation of the BEW will disproportionately affect subendocardial blood flow. Loss of privileged perfusion of this layer in MVD may be the driving force behind subendocardial ischemia leading to diastolic dysfunction and inefficient coronary perfusion during exercise. Myocardial contraction in the

ORIGINAL RESEARCH ARTICLE
face of higher vascular resistance within the subendocardial layer may also account for the exaggerated BCW during isovolumetric contraction, while conversely a larger BCW can diminish systolic perfusion. Indeed, the mechanics of image acquisition during perfusion CMR whereby the basal-, mid-, and apical-LV slices are acquired during mid-diastole, mid-systole, and late-diastole, respectively, demonstrates that subendocardial perfusion in MVD is impaired throughout the cardiac cycle ( Table 2). The ability of the coronary microvasculature to dilate during periods of stress seems pivotal to preventing the maladaptive changes to exercise and myocardial perfusion observed within the MVD cohort. With therapies shown to improve outcome in patients with MVD as defined by low CFR, further work will need to be carried out to determine whether the distinct functional and structural MVD endotypes should be managed differently. 35,36

Coronary Flow Reserve and Inducible Ischemia
The Coronary Vasomotion Disorders International Study Group agreed upon the following criteria for the diagnosis of microvascular angina: (1) presence of symptoms suggestive of myocardial ischemia; (2) objective documentation of myocardial ischemia; (3) absence of obstructive CAD >50% coronary diameter reduction and/or fractional flow reserve >0.80); and (4) confirmation of a reduced coronary blood flow reserve and/or inducible microvascular spasm. 37 We have demonstrated that a CFR<2.5 (in our NOCAD cohort) accurately predicted myocardial ischemia and exercise pathophysiology. Emerging data suggests that a binary CFR threshold of 2.5 more accurately identifies patients with MVD than the less sensitive value of 2.0, both currently accepted by international guideline committees and experts within the field. 37-40 Whether a CFR<2.5 alone is sufficient to diagnose MVD (when noninvasive ischemia testing has not been performed) is something the guideline committees may wish to consider, but in the meantime this study should provide clinicians with a strong recommendation to perform invasive coronary physiology in patients who undergo cardiac catheterization. Recent studies have also demonstrated the correlation between global measurements of myocardial perfusion and invasive coronary functional testing. 41 Within our study, 22% of patients with normal vasodilator reserve demonstrated inducible ischemia on quantitative perfusion imaging, and this group of patients may have other ischemic pathologies. One explanation is that CFR integrates epicardial and endocardial blood flow within a given territory; good perfusion in the former may mask ischemia affecting the latter.

Study Limitations
This was a mechanistic single-center study with relatively small numbers of patients in each group. The lack of correction for multiplicity increases the possibility for type 1 error for each of the analyses presented. Invasive CFR is currently the most accepted method for classifying MVD in patients with NOCAD. Like all biological measurements, CFR is a continuous variable and for this study a dichotomous CFR threshold of 2.5 was adopted to define MVD, acknowledging that a lower threshold may have had enhanced specificity at the cost of sensitivity. Moreover, given that our control group consisted of patients who were not healthy volunteers but who had symptoms that had led to angiography, we adopted a 2.5 threshold to ensure that their endothelium-independent microvascular function was truly normal. Indeed, patients in the control group may have occult coronary abnormalities such as endothelial dysfunction or coronary vasospasm that could be unmasked during provocation testing. The study was powered to detect differences in exercise and myocardial perfusion physiology between patients with preserved CFR and diminished CFR, not among MVD endotypes. A larger sample size may have enabled the identification of differences between the 2 endotypes. A normal hyperemic microvascular resistance value is currently undefined in a NOCAD population because of the absence of a gold standard measure to assess microvascular function in vivo. However, our study describes 2 distinct endotypes with differing degrees of systemic disease involvement based around a 2.5 mmHg/ cm/s value. We were not able to directly assess myocardial perfusion during exercise in this study; therefore, we were unable to demonstrate exercise-induced ischemia per se, although vasodilator mediated perfusion heterogeneity during CMR is widely regarded as a surrogate of ischemia in routine clinical practice. Subendocardial hypoperfusion during hyperemia represents a very early stage of the ischemic cascade and so its presence may not correlate perfectly with later stages, such as wall motion abnormalities. However, this is a widely adopted index for identifying the presence of inducible ischemia, particularly in a NOCAD population and is strongly recommended in practice guidelines. Because of the length of this protocol, not all study patients underwent CMR; however, this was factored into the study design and statistical power calculation.

CONCLUSION
Patients with MVD, defined by diminished vasodilator reserve in the cardiac catheter laboratory, have inducible ischemia and inefficient coronary perfusion during exercise. These findings are consistent among the 2 distinct MVD endotypes, defined by either abnormal resting or minimal microvascular resistances and reflect