Evaluation of non-invasive imaging parameters in coronary microvascular disease: a systematic review

Background Coronary microvascular dysfunction (CMD) is an important underlying cause of angina pectoris. Currently, no diagnostic tool is available to directly visualize the coronary microvasculature. Invasive microvascular reactivity testing is the diagnostic standard for CMD, but several non-invasive imaging techniques are being evaluated. However, evidence on reported non-invasive parameters and cut-off values is limited. Thus, we aimed to provide an overview of reported non-invasive parameters and corresponding cut-off values for CMD. Methods Pubmed and EMBASE databases were systematically searched for studies enrolling patients with angina pectoris without obstructed coronary arteries, investigating at least one non-invasive imaging technique to quantify CMD. Methodological quality assessment of included studies was performed using QUADAS-2. Results Thirty-seven studies were included. Ten cardiac magnetic resonance studies reported MPRI and nine positron emission tomography (PET) and transthoracic echocardiography (TTE) studies reported CFR. Mean MPRI ranged from 1.47 ± 0.36 to 2.01 ± 0.41 in patients and from 1.50 ± 0.47 to 2.68 ± 0.49 in controls without CMD. Reported mean CFR in PET and TTE ranged from 1.39 ± 0.31 to 2.85 ± 1.35 and 1.69 ± 0.40 to 2.40 ± 0.40 for patients, and 2.68 ± 0.83 to 4.32 ± 1.78 and 2.65 ± 0.65 to 3.31 ± 1.10 for controls, respectively. Conclusions This systematic review summarized current evidence on reported parameters and cut-off values to diagnose CMD for various non-invasive imaging modalities. In current clinical practice, CMD is generally diagnosed with a CFR less than 2.0. However, due to heterogeneity in methodology and reporting of outcome measures, outcomes could not be compared and no definite reference values could be provided.

The Coronary Vasomotion Disorders International Study Group (COVADIS) determined the following criteria to diagnose CMD: presence of symptoms and objective documentation of myocardial ischemia, absence of obstructive CAD (< 50% stenosis and/or fractional flow reserve < 0.8) and confirmed reduced coronary flow reserve (CFR) (and/or inducible microvascular spasm). However, assessment of CMD remains challenging, as no tools are available to directly visualize the coronary microvasculature. In fact, the current golden standard to diagnose CMD is invasive measurement of CFR in epicardial arteries without functionally relevant stenosis [12]. The CFR depicts the increase in coronary blood flow in response to vasoactive agents [2,4,9] and provides indirect quantification of coronary microvascular blood flow [5,13].
The invasive nature and high costs of coronary reactivity testing (CRT) initiated the search for a non-invasive alternative to diagnose CMD, including myocardial perfusion reserve index (MPRI) measured using cardiac magnetic resonance imaging (CMR), and CFR using positron emission tomography (PET) and transthoracic echocardiography (TTE) [13][14][15][16]. Yet, the cut-off value for CFR to diagnose CMD differs between these modalities, is not well validated and, even though sex-differences in coronary physiology are known, the need for a sex-specific cut-off value remains under debate [17]. To date, a CFR below 2.0 to 2.5 is deemed diagnostic for CMD [9,16,[18][19][20].
The (dis)advantages of these non-invasive imaging techniques in the diagnosis of CMD have been discussed extensively before [21]. However, it is unclear which outcome parameters and corresponding cut-off values should be used to diagnose CMD. As such, this systematic review aims to provide an overview of currently reported reference and cut-off values for diagnosing CMD in a non-invasive manner.

Search strategy
On October 15, 2018 the PubMed and EMBASE databases were systematically searched for non-invasive imaging studies on CMD. The search was updated on November 1, 2020. Studies were considered for eligibility without date restriction. The search terms and synonyms of 'coronary microvascular dysfunction' , 'nonobstructive coronary disease' and 'imaging' , including the imaging modalities CMR, PET and TTE were used. A broad search strategy was performed as studies on CMD are limited and nomenclature of CMD is not standardized. Therefore, search terms were searched for in ' All Fields' .
The detailed search strategy is provided in Additional file 1: Search Strategy.

Study selection
To assess eligibility, the results from the literature search were initially screened by title and abstract and subsequently for full text. Article selection and data extraction were performed independently by two reviewers (RGMK and FG). Observational studies and randomized controlled trials providing baseline outcome measurements were considered for inclusion.
Studies were included if they enrolled participants with AP (i.e. effort angina or anginal equivalents) and CAG or coronary computed tomography angiography confirmed absent or nonobstructive CAD (based on the definition described in the study protocol of the included studies), or healthy participants without prior history of cardiovascular disease or AP as a control group and reported the results of a non-invasive imaging method with use of pharmacological stress (i.e. flow parameters measured with either CMR, PET or TTE) to diagnose CMD.
Studies written in languages other than English or Dutch, exclusively consisting of participants with comorbidities, i.e. CAD, diabetes mellitus, aortic stenosis or cardiomyopathies, were excluded. Studies were excluded if outcomes were not reported as flow parameters, if patients were stratified according to the outcome of interest or if patient or control groups contained fewer than 10 participants.

Quality assessment
A methodological quality assessment was performed with the QUADAS-2 (Tool for the Quality Assessment of Diagnostic Accuracy Studies [22]). Studies were assessed for concerns of applicability ('low' , 'high' or 'unclear') and for risk of bias ('low' , 'high' or 'unclear') on four key domains (patient selection, index test, reference standard and flow and timing). The assessments per domain were combined into an overall risk of bias and concern of applicability.

Data extraction and analysis
The variables of interest were extracted using a standardized data collection form. Post-hoc evaluations within one clinical trial assessing the same imaging modality were considered as one study. Due to heterogeneity of the included studies, a meta-analysis of the results was not possible.

Search results
The search yielded a total of 6976 results, 2568 studies in Pubmed and 4408 studies in Embase. Removal of duplicates resulted in 5238 unique entries. After title and abstract screening, 443 possibly relevant studies were obtained. The full texts of these studies were screened to select those that met the inclusion criteria as provided in the methods section. One relevant study was obtained through cross-reference checking. Thirty-seven studies met the inclusion criteria and were included in the final analysis. The search and inclusion and exclusion of relevant studies are summarized in Fig. 1. The characteristics of the included studies are summarized in Table 1. Quality assessment of included studies showed a clear description of the reference standard was not part of the study protocol in most of the included studies. The full assessment is provided in Additional file 2: Table S1 [22].

Demographic information
The number of patients included in each study was generally small, with a median study population of 22 patients (range 11 to 963, 89% women) and median of 18 controls (range 10 to 268, 33% women). The mean age in patient groups ranged from 50.0 ± 7.0 to 66.0 ± 10.0 years of age and 35.3 ± 3.9 to 62.6 ± 9.1 years of age in control groups. The specific demographic information per study is summarized in Table 1.

Flow parameters
Different flow parameters were reported (Table 2). In CMR studies, the myocardial perfusion reserve index (MPRI) was most often reported. Other parameters reported were myocardial perfusion reserve (MPR) and CFR. MPR is defined as the ratio between the relative upslope of myocardial signal intensity (obtained with the use of gadolinium as contrast agent) during stress and rest. In contrast to MPR, the MPRI is corrected for left ventricular contrast signal intensity, allowing for a reduction in signal differences within the image and intra-individual level differences in signal intensity due to heart rate and blood pressure [6,23,24]. As such, MPRI is often the preferred outcome measure as it seems to be more accurate in quantifying coronary microvascular blood flow. In one study, CMR-derived CFR results were presented [25]. CFR was calculated and measured in the exact same way as the MPR and can therefore be considered as a synonym of MPR.
In PET studies microvascular function was usually quantified with CFR. Other outcome parameters were MPR or myocardial flow reserve (MFR). MPR and MFR were calculated based on the same methods and measurements as CFR and could therefore be used interchangeably. CFR was defined as the ratio between hyperemic and resting myocardial blood flow (MBF) [26,27] which was expressed in ml/min/g [28]. CFR was often corrected for rate pressure product (RPP), defined as heart rate multiplied by systolic blood pressure and represents cardiac metabolic demand. This correction is recommended as it reduces variability in outcomes due to person-level differences in systolic blood pressure and heart rate [2,27].
In TTE studies CFR was used. Similar to PET and CMR, a variety of equivalent terms were reported, namely CFR, coronary blood flow (CBF) and coronary flow velocity reserve (CFVR). CFR, CBF and CFVR were all defined as the ratio of peak stress and rest coronary blood flow velocities (CBFV), usually obtained by spectral Doppler measurements.

Sex differences
Only one of the included studies compared outcomes between men and women. Sestito et al. [56] determined CBF (defined as the ratio of diastolic CBF velocity at peak stress and rest) using TTE in 71 patients diagnosed with CMD (48 women, 67.6%). No significant difference in CBF was found between men and women (CBF 2.09 ± 0.60 and 2.03 ± 0.50, respectively). Furthermore, the proportion of women as compared to men in the patient groups was much larger (89%) as compared with the control population (33%).

Discussion
We provided an overview of currently used non-invasive imaging techniques and corresponding reference values for CMD in patients with AP and no or nonobstructive CAD as well as healthy subjects. We found quite some differences between reported non-invasive imaging parameters to assess CMD, which we have summarized in Fig. 2. These differences reflect the large heterogeneity between the studied population as well as the rapidly developing imaging techniques and protocols per imaging technique, which result in a variety of different study protocols. Due to the heterogeneity between the included studies we were unable to perform a formal meta-analysis and provide clear clinically applicable cut-off values to diagnose CMD. MPRI was found to correlate well with invasive measurements obtained with CRT, such as index of microcirculatory resistance and CFR [15,23]. Therefore, MPRI could potentially serve as a non-invasive alternative to CRT. Current literature proposes two different cut-off values, namely 1.40 and 1.84 [6,15], corresponding with the results found in this review. However, the results of this review suggest a grey area of MPRI values, as some overlap is seen between MPRI in patients and healthy controls. Stress MBF values can aid in differentiating CMD from normal coronary microvascular function. Liu et al. [15] have shown that a decreased stress MBF (i.e. less than 2.30 ml/min/g) is suggestive of CMD in patients with inconclusive MPRI values. Furthermore, some CMR studies now investigate the clinical applicability of quantitative myocardial tissue characterization with rest and stress T1 mapping as an alternative [29,33]. Ischemic myocardial tissue can be differentiated from healthy tissue based on distinct properties at T1 mapping during rest and stress conditions, without the use of contrast agents. However, the diagnostic value of T1 mapping in diagnosing CMD still needs extensive validation [29].
Currently, PET is the most frequently applied and validated non-invasive imaging technique in quantifying microvascular blood flow. PET is considered the golden standard of non-invasive diagnosis of CMD, although discordance between invasive fractional flow reserve (FFR) and non-invasive CFR is reported in up to 30% of cases [16,26,27,61]. CMD is generally diagnosed with a CFR less than 2.0 if corrected for RPP or less than 2.5 if uncorrected [28,35,45,47]. However, no evidence-based cut-off values for CFR in PET are available yet. Similarly, no cut-off values for CFR in TTE have yet been determined and generally a cut-off value of less than 2.0 for the diagnosis of CMD is applied [18,58,[62][63][64][65]. The study of Hildick-Smith et al. showed CFR with use of adenosine stress TTE to be well above this applied cut-off value, i.e. a mean CFR 3.7 in healthy controls and 5.9 in athletes [66]. However these study population comprised men of 27 years of age and could therefore not be directly used as reference value for the, mainly older and female, population of interest at risk for CMD. TTE assessment of CFR with Doppler echocardiography has been validated against intracoronary Doppler measurements and outcomes correlate well [2,5,60,67].

Causes of heterogeneity in measured outcome parameters Patient groups
The heterogeneity in outcomes observed in this systematic review is most likely the result of differences in inclusion criteria applied across several studies and differences in the use of imaging techniques. Although most studies investigated patients with typical AP and no or nonobstructive CAD during diagnostic CAG, the setting in which participants were recruited was not reported clearly. Furthermore, the definition of no or nonobstructive CAD and the control population was often unclear and, if documented, heterogeneous among the included studies (Additional file 6: Table S5). Therefore, a more homogeneous definition could not be applied in the search method. Hence, we suggest the use and documentation of standardized criteria as reported by COVADIS [12].

Methodological differences
Unclarity of the used reference standard, as reflected by the risk of bias assessment (Additional file 2: Table S1), may have introduced significant bias. Moreover, it was often unclear whether researchers were blinded to the reference standard when interpreting results from the index test.
Furthermore, measurement of MPRI in CMR might contribute to the inconsistent results observed in this systematic review. MPRI can be measured transmural, but also subendo-or epicardial. Several studies show subendocardial MPRI to be decreased more often than epicardial MPRI in CMD patients [34,39,64,68], which might indicate subendocardial MPRI to be more valuable in diagnosing CMD as compared with epicardial or transmural MPRI. Unfortunately, in this systematic review only transmural MPRI values were included.
Regarding PET, correction for RPP is not standard which results in decreased comparability of outcomes. Moreover, the use of different radioactive tracers (15O-water, 13N-ammonia and Rubidium-82) could result in varying outcomes due to differences in characteristics and processing of images [14,27,28]. The use of a specific radiotracer might require a specific cut-off value to diagnose CMD [14]. Similar concerns apply to the use of various vasoactive agents to achieve hyperemia in stress perfusion imaging. Adenosine and dipyridamole are most commonly administered to achieve hyperemia. However, adenosine seems to be superior to dipyridamole with regard to attaining maximal hyperemia and their effects are not identical. Therefore the use of the different types and doses of vasoactive agent could have contributed to the differences found in the outcome parameters [27,69,70].
Lastly, this systematic review highlights the discordance in nomenclature and reporting of outcomes. Standardization of outcome parameters reported could increase comparability of studies assessing reference values for the diagnosis of CMD.

Sex differences
In the present analysis women were highly represented in the patient groups (89%) compared to control groups (33%). Therefore, sex differences could contribute to discrepancies between studies resulting in decreased comparability between CFR and MPRI measurements in patient and control groups. Kobayashi et al. [17] measured coronary vascular diameter with quantitative CAG and intravascular ultrasound in patients with AP and nonobstructive CAD and found a smaller vascular diameter in women. Furthermore, they showed a significantly higher resting CBF in women. The latter is consistent with findings by Opstal et al. [71] and Chareonthaitawee et al. [72] who studied coronary blood flow in healthy subjects with 13N-ammonia PET (n = 206) and 15O-water PET (n = 169), respectively. These findings suggest sex differences in flow parameters. High resting myocardial flow volumes could decrease CFR (in PET and TTE) or MPRI (in CMR) in women compared to men as flow parameters are determined as the ratio of stress and rest perfusion. Although sex differences in resting MBF and CFR have been observed in invasive CRT [3,17,73], only one of the included studies included assessed sex differences regarding CBF and reported no significant sex differences [56]. These findings are consistent with another study comparing non-invasive CFR between men and women using PET [74]. Therefore, further research is needed to establish whether or not sexspecific cut-off values are required for the non-invasive diagnosis of CMD.

Recommendations for future research
The studies included in this review show heterogeneity in study methodology and outcome. This contributes to the discrepancies in outcomes between studies and to the lack of consensus regarding definition and cut-off values for CMD in non-invasive imaging modalities. We emphasize the need for large validation studies and suggest standardization of outcome parameters to reduce heterogeneity and increase comparability of studies. This is needed to provide clinically applicable, possibly sexspecific, reference values for the diagnosis of CMD in the future.
Furthermore, during this systematic review, we found several other imaging modalities that are studied for their potential to diagnose CMD, such as myocardial contrast echocardiography (MCE) [63,75,76], CT-perfusion [77] and absolute quantification of myocardial perfusion by CMR [78]. Even though, current evidence is still limited so the clinical significance and applicability in regular care of these modalities in CMD diagnosis remains unclear, current research shows promising results. For example, Bechsgaard et al. [77] studied CT myocardial perfusion in women with angina and no obstructive CAD, (defined as < 50% stenosis), in comparison with female controls. They showed CT-perfusion is able to identify decreased global myocardial perfusion and impaired increase of myocardial blood flow during adenosine provocation in women with angina and non obstructive CAD as compared with the control group. The use of CT-perfusion in addition to the commonly performed CCTA could in the future play an important role in the evaluation of CMD early in the evaluation of patients with angina.

Study limitations
The number of studies investigating non-invasive imaging techniques to diagnose CMD is limited. As such, the results of this systematic review are based on limited data. Hence, only an indication of reference and cut-off values could be provided. Furthermore, a formal metaanalysis could not be performed due to heterogeneity of included studies. In addition, the risk of selection bias in the included studies was high. Also, the heterogeneity in the definition of non obstructive CAD might also have impacted the results as the included studies comprise patients with both completely normal or non obstructed coronaries. Unfortunately, as the definition of nonobstructive CAD was unclear or heterogeneous, it was not possible to separately analyse outcomes for patients with normal coronaries as compared to nonobstructive CAD. These analyses would have been of additive value as in patients with nonobstructive CAD there might still be epicardial stenosis that could impact myocardial blood flow. These limitations emphasize the importance of standardization of imaging protocols and analyses, patient selection and reporting of outcome measurements to obtain reliable and clinically relevant cut-off values for CMD.