Impact of Isolated Coronary Microvascular Disease Diagnosed Using Various Measurement Modalities on Prognosis: An Updated Systematic Review and Meta-Analysis

Myocardial ischemia has been traditionally considered to be attributed to atherosclerosis and obstructive atherothrombotic events in epicardial coronary arteries. However, a significant proportion of patients presenting with symptoms and signs of ischemic heart disease exhibit no significant coronary artery stenosis [1]. One potential etiology for chest pain in the absence of obstructive atherosclerosis is coronary microvascular disease (CMD) [2]. CMD is a condition that affects the structure and/or function of the coronary microcirculation, leading to impaired coronary flow reserve (CFR) and microcirculatory resistance (MR) [3]. Impaired CFR or MR indicates the presence of CMD. Invasive measurement using coronary angiography [4] as well as noninvasive methods such as echocardiography (ECHO) [5], positron emission tomography (PET) [6], and cardiac magnetic resonance (CMR) [7] can be utilized to assess coronary microvascular function.

CMD can be classified into four main types: (i) CMD in the absence of myocardial diseases and obstructive coronary artery disease (CAD), (ii) CMD in myocardial diseases, (iii) CMD in obstructive CAD, and (iv) iatrogenic CMD [8]. Patients presenting with angina in the absence of obstructive CAD are often suspected of having non-cardiogenic symptoms. Prior investigations have demonstrated association between CMD and increased risk of mortality and major adverse cardiac event (MACE) in patients without obstructive CAD [9]. The evaluation of microvascular function through appropriate modalities is crucial in identifying patients at increased risk of adverse outcomes. Furthermore, systematic review and meta-analysis have revealed elevated rates of all-cause or cardiac mortality and MACE in the isolated CMD population [10, 11]. However, these studies did not incorporate emerging technologies, such as the wire-free angiography‐based index known as the coronary angiography-derived index of microcirculatory resistance (caIMR) [12, 13]. Due to a lack of data, the prognostic utility of coronary flow indices has not been systematically quantified across a wide broad range of measurement modalities in the isolated CMD population. Hence, the purpose of this systematic review and meta-analysis was to further clarify the association between isolated CMD as diagnosed by various measurement modalities and clinical outcomes.

This review was conducted in accordance to the PRISMA statement [14] and registered in the International Prospective Register of Systematic Reviews (CRD42023412647).

A search in electronic databases was conducted including PubMed, The Cochrane library, and EMBASE databases for studies published until March 2023. The full-search strategy is provided in the online supplementary Table 1 (for all online suppl. material, see https://doi.org/10.1159/000533670).

All studies included in the meta-analysis adhered to the following inclusion criteria: subjects without obstructive CAD (stenosis ≥50%) on invasive coronary angiography or those with negative coronary computed tomography or negative stress myocardial ischemia if coronary angiography data were not available; subjects without a history of heart transplantation, cardiomyopathy, or aortic stenosis; prospectively measured coronary microvascular function using either invasive or noninvasive methods, with indices of coronary flow measurement including CFR, coronary flow velocity reserve (CFVR), myocardial blood flow reserve (MBFR), myocardial flow reserve (MFR), myocardial perfusion reserve index (MPRI), hyperemic microcirculatory resistance (HMR), index of microcirculatory resistance (IMR), caIMR, and angiography‐based index of microcirculatory resistance (angio‐IMR) reported; clinical events, including all-cause mortality and/or MACE were reported and compared between two groups of populations divided by a cut-off value for indices. Exclusion criteria included non-English language, nonhuman studies, reviews, editorials, abstracts; studies with insufficient data to estimate hazard ratios (HRs) or odds ratios (ORs) and their 95% confidence intervals (CI), follow-up duration less than 3 months. In order to strictly limit the study population to isolated CMD, we excluded studies of patients with heart failure with reduced ejection fraction. Due to the possibility of obstructive lesions observed in coronary angiography presenting with fractional flow reserve values >0.8, we excluded studies that did not confirm coronary artery stenosis less than 50% on coronary angiography in subjects with functionally insignificant epicardial coronary stenosis.

Two authors (Xing Yu-Luo and Yao Kun-Liu) screened all articles and selected those meeting the prespecified inclusion and exclusion criteria. We resolved disagreements by consensus with a third investigator (Bo Zheng). If there was overlap of the study population, the article with the greatest number of patients was used for the analysis.

The primary endpoint was the all-cause mortality, with cardiac death or cardiovascular death as a substitute if all-cause mortality was not provided. The secondary endpoint was MACE. Definitions of MACE varied across the included studies (online suppl. Table 2), including a combination of the following: death or cardiac death or cardiovascular death; nonfatal acute coronary syndrome; nonfatal stroke; development or hospitalization for heart failure; rehospitalization for angina; revascularization.

Two authors (Luo X. and Liu Y.) independently extracted the data from included studies, verified by a third author (Zheng B.). Data extraction was conducted according to the following subheadings: editorial information (lead author and publication year), study population description (number of patients for each study, age, body mass index, and percentage of female population), prevalence of risk factors (such as smoking, hypertension, hyperlipidemia, and diabetes), method of CMD determination, outcomes (event description) using number of events, and adjusted time-to-event data, presented as HRs of CMD versus non-CMD. We excluded studies reporting the HR using indices of coronary microvascular flow as a continuous variable. Where event rate data were missing from a manuscript, we searched any available online supplementary material. Where studies investigated two or more populations divided by a baseline characteristic (e.g., female and male) and provided outcomes of interest for each, they were considered separate studies for the purposes of meta-analysis.

Two investigators (Luo X. and Liu Y.) assessed the included studies for risk of bias using the Newcastle-Ottawa Quality Assessment Scale for cohort studies [15]. A quality score was calculated for three major components of cohort studies: selection of study groups (0–4 points), comparability of study groups (0–2 points), and determination of outcome of interest (0–3 points). A higher score represented better methodological quality.

All analyses were conducted using RevMan 5.4 and StataSE 15 software. Statistical significance was set at p ≤ 0.05 (2 tailed).

The statistical heterogeneity was assessed using the I² test (p < 0.10 was considered to be statistically significant) and the I² statistic (no statistical heterogeneity, I² < 25%; low heterogeneity, I² ≥ 25% and <50%; moderate heterogeneity, I² ≥ 50% and <75%; high heterogeneity, I² ≥ 75%). The pooled ORs and HRs for all-cause mortality (unless only cardiac or cardiovascular mortality was reported) and MACE were calculated using the random effects model using inverse variance weighting [16].

The measurement modality of CMD, as well as population characteristics such as sex, diabetes, age, were potential clinical factors contributing to heterogeneity. Therefore, we conducted a subgroup analysis to evaluate the certainty of the evidence. The study was stratified into four groups based on the measurement modality of CMD: PET, ECHO, and CMR, and invasive groups. Additionally, we divided the studies into female and male groups, diabetes and non-diabetes groups, and elder (mean age of population <60 years old) and younger (mean age of population ≥60 years old) groups.

Sensitivity analyses were performed to evaluate the robustness of the association between isolated CMD and endpoints. We assessed the effect of each individual study on the overall risk estimate by leaving-one-out analysis. Planned sensitivity analysis also included the stratified analyses to assess modification for angiographic exclusion of obstructive CAD and adjusted HRs for outcomes.

The potential publication bias was examined by a funnel plot and Egger’s test. The asymmetry of the plot was estimated visually and quantitatively using Egger’s linear regression test [17].

The results of the database search and study selection process are summarized in Figure 1. Of the 5,193 studies initially identified, the electronic search identified 3,710 citations that were screened by reviewing the title and abstract. A total of 146 articles were assessed in full text and 27 studies were included in the meta-analysis.

Table 1 displays the characteristics of the studies included in the meta-analysis [5, 6, 18‒42]. All studies were prospective cohort studies published between 2004 and 2023, with a total of 27 studies (n = 18,204) included. Nine and six articles were included in the meta-analysis for the calculation of ORs and HRs for mortality, respectively. For the calculation of ORs and HRs for MACE, 18 and 16 articles were included in the meta-analysis, respectively. The mean age of subjects ranged from 36 to 74.8 years. The majority of studies included patients with suspected ischemic heart disease (22 studies including 17,538 subjects). Other patient groups included: myocardial infarction with no obstructive coronary arteries (MINOCA) (1 study including 109 subjects), asymptomatic patients with or without type 2 diabetes (1 study including 200 subjects), heart failure with preserved ejection fraction (HFpEF) (1 study including 163 subjects), and autoimmune disease (2 studies including 194 subjects). Indices of coronary microvascular function measurement included caIMR, HMR, MFR, MPRI, CFR, and CFVR measuring via various modalities including PET (8 studies), ECHO (11 studies), CMR (2 studies), and invasive measurement (6 studies). Various cutoffs were utilized in the included studies to distinguish between normal and abnormal coronary microvascular function.

A total of 4,852 subjects were included in the 9 studies that reported the number of deaths, with 247 deaths reported. Of these 9 studies, 6 reported cardiac or cardiovascular mortality only [6, 18, 23, 27, 30, 34], while the remaining 3 reported all-cause mortality [28, 29, 36]. The median follow-up period ranged from 1.3 to 8.5 years. Among the 2,876 patients without CMD, there were 95 deaths (3.3%), while among the 1,976 patients with CMD, there were 152 deaths (7.7%). The OR for mortality in patients with CMD compared to those without CMD was 2.97 (95% CI, 1.91–4.60; p < 0.0001; I² = 46%) (Fig. 2a).

Six studies, encompassing 9,507 subjects, presented HRs for mortality [5, 28, 31, 33, 34, 37]. The median follow-up period ranged from 1.6 to 12.5 years. Meta-analysis of the HRs revealed a significantly increased hazard of mortality (HR, 3.38, 95% CI, 1.77–6.47; p = 0.0002, I² = 87%) in the CMD population compared to those without CMD (Fig. 2b).

A total of 18 studies [18, 20‒27, 29, 30, 32, 34‒36, 38, 39, 41] comprising 8,402 subjects reported the number of events for MACE. The median follow-up period ranged from 1 to 8 years. The event rate for MACE was 20.5% among patients with CMD, as compared to 6.2% among those without CMD. The OR for MACE in patients with CMD compared to those without CMD was 5.82 (95% CI, 3.65–9.29; p < 0.00001; I² = 86%) (Fig. 3a).

Sixteen studies, including 9,422 subjects, presented HRs for MACE [18‒24, 26, 27, 32‒35, 40, 41, 43]. The median follow-up period ranged from 1 to 5.7 years. The summary HR for MACE among patients with CMD was 4.01 (95% CI, 2.59–6.20; p < 0.00001; I² = 91%) compared to those without CMD (Fig. 3b).

In the subgroup analysis of studies grouped according to measurement modalities, a consistent association of CMD with mortality was observed, with no inter-group heterogeneity seen between the subgroups (χ² test = 1.86, p = 0.39, I² = 0%; χ² test = 2.98, p = 0.23, I² = 32.8%, respectively) (Fig. 4a, b). The associated OR (3 studies including 887 subjects) and HR (2 studies including 2,062 subjects) with invasive measurement were 2.27 (95% CI, 1.09–4.72; p = 0.03; I² = 45%), and 2.79 (95% CI, 0.51–15.17; p = 0.23; I² = 91%), respectively. With PET, the OR (3 studies including 1,400 subjects) and HR (1 study including 2,175 subjects) were 2.48 (95% CI, 1.18–5.23; p = 0.02; I² = 13%) and 2.23 (95% CI, 1.31–3.78; p = 0.003), respectively. With ECHO, OR (3 studies including 2,565 subjects) and HR (3 studies including 5,270 subjects) were 4.51 (95% CI, 2.09–9.75; p = 0.0001; I² = 46%) and 4.55 (95% CI, 2.45–8.46; p < 0.00001; I² = 55%), respectively. A similar effect was observed for the outcome of MACE, with moderate-to-high inter-subgroup heterogeneity and high heterogeneity seen in studies using ECHO (Fig. 5a, b).

In the subgroup analysis of studies categorized by age, similar effects for mortality and MACE were observed between the two groups (online suppl. Fig. 1a, b, 2a, b). Similarly, the pooled ORs and HRs for MACE were consistent and comparable between female and male subjects, with no inter-subgroup heterogeneity (online suppl. Fig. 3a, b). In the subgroup analysis of studies adjusted for diabetic status, pooled ORs for MACE were calculated in 3 studies including 558 diabetes subjects and 2 studies including 668 non-diabetes subjects, while pooled HR for MACE was calculated in 3 studies including 927 diabetes subjects and 1 study including 451 non-diabetes subjects. The associated OR for MACE was greater in diabetes subjects (OR, 15.26; 95% CI, 4.50–51.77, p < 0.0001, I² = 71%) compared with non-diabetes subjects (OR, 3.65; 95% CI, 1.84–7.24, p = 0.0002, I² = 0%), with high inter-subgroup heterogeneity (χ² test = 4.01; p = 0.05; I² = 75.1%) (online suppl. Fig. 4a), while the associated HR for MACE was 5.52 (95% CI: 3.68–8.28) in diabetes subjects and 3.16 (95% CI: 0.56–17.72) in non-diabetes subjects, with no inter-subgroup heterogeneity (χ² test = 0.38; p = 0.54; I² = 0%) (online suppl. Fig. 4b).

The leave‐one‐out sensitivity analysis indicated that no individual study significantly influenced the summary risk estimate (online suppl. Table 3). Sensitivity analysis stratified by the use of angiography to exclude obstructive CAD did not alter the overall results (online suppl. Fig. 5a–d). Sensitivity analysis was performed after excluding studies that presented unadjusted HRs for mortality and MACE. The excluding of these studies did not significantly affect the pooled HRs (online suppl. Fig. 6a, b).

The quality of each study was assessed using the Newcastle-Ottawa Scale quality assessment criteria for cohort studies, and the results are presented in online supplementary Table 4. Funnel plots and Egger’s test were used to examine possible publication bias. No visual asymmetry was observed in the funnel plot for the association between CMD and MACE among studies presenting with OR. However, the association between CMD and MACE exhibited asymmetry among studies presenting with HR (online suppl. Table 5). Publication bias for the meta-analysis of the mortality was difficult to estimate because a limited number of datasets was included. In addition, Egger’s regression tests were unable to perform since fewer than ten datasets were insufficient to draw confident conclusions or to perform statistical tests of symmetry.

In this study, we expanded on previous meta-analysis [10, 11] by including more recent studies [18, 19, 25, 26, 31‒40]. We have expanded our study population to include a greater number of Asian populations, such as those from China and Japan. Additionally, we have included subjects with different disease presentations, such as MINOCA, HFpEF, and autoimmune diseases, which increase the representativeness of our study. Furthermore, we have included studies that utilize emerging technologies, such as caIMR and CMR. With the inclusion of more studies, we have been able to perform subgroup analysis to explore whether differences exist in the predictive value of isolated CMD diagnosed by different measurement modalities for prognosis. Our findings demonstrated that patients with isolated CMD exhibited a 3-fold higher risk of death and over a 4-fold higher risk of MACE comparing to those without CMD. The pooled ORs and HRs for mortality or MACE remained unchanged when studies presenting unadjusted HRs or studies not using coronary angiography to exclude obstructive CAD were omitted.

Notably, approximately 50% of women and 30% of male patients undergoing invasive coronary angiography suspected of having obstructive CAD do not present significant epicardial coronary stenosis [44‒46]. For lacking evidence of obstructive lesions, some patients with chest pain are often misdiagnosed as non-cardiac, leading to under-diagnosis and under-treatment. CMD, a disorder affecting the structure or function of the coronary microcirculation, has been increasingly recognized recently. Even if the epicardial coronary arteries are not obstructive, CMD can still lead to impaired myocardial perfusion [47]. It is reported that CMD can cause myocardial ischemia which affects up to 50% of subjects with chronic coronary syndromes (CCS) and up to 20% of those with acute coronary syndromes (ACS), resulting in adverse events [48, 49]. CMD is classified into four types according to its clinical characteristics. Prior studies have revealed that patients with isolated CMD (type 1 CMD), i.e., no evidence of myocardial disease and obstructive CAD, showed poorer prognosis than those without CMD [10, 11]. However, because this group of patients does not have any obvious organic heart disease, they are indeed clinically easily underestimated. Assessing and organizing patients with CMD, especially those without obstructive CAD or any other identifiable cardiac pathologies, hold significant clinical implications for public health [50]. Therefore, to generate attention towards the isolated CMD, we conducted this meta-analysis.

The assessment of CMD can be performed using various diagnostic modalities, both invasive and noninvasive. Although invasive wire-based indices (such as CFR and IMR) are gold standards for diagnosing CMD, some derived noninvasive indicators via ECHO, CMR, PET, etc. also show similar performance in previous studies. Widespread implementation of wire-based CFR and IMR in clinical practice is hampered by the requirement for a specific intracoronary wire and the time- and cost-consuming nature of the techniques. More recently, with the technical development, coronary angiography-derived indices, such as angiography-derived index of microcirculatory resistance (IMRangio) [51], nonhyperemic angiography-derived index of microcirculatory resistance (NH-IMRangio) [42], caIMR [13], and angiographic microvascular resistance (AMR) [52], have emerged as potential alternatives to pressure-derived indices and have been proved to show good diagnostic accuracy in assessing CMD. In addition, these indices are advantageous as they do not require the use of pressure wire, hyperemic agents, or thermodilution methods, enabling the assessment of coronary microvascular function more efficiently, conveniently, and cost-effectively. Novel technologies such as caIMR, based on an optimized computational fluid dynamics model and aortic pressure waves, were added into this meta-analysis, indicating screening for CMD has significant value for predicting disease prognosis whatever diagnostic modalities. One of the subgroup analyses of studies calculating HRs for MACE revealed no significant hazard for MACE in the invasive measurement group. However, the heterogeneity between studies in this group was high. After excluding the study [43] that presented unadjusted HR for MACE, the pooled HR was 3.10 (95% CI, 1.56–6.15, p = 0.001, I² = 0%). The consistent association between CMD and poorer prognosis among different diagnostic modalities suggested that various measurement modalities could be used for assessing prognosis in the isolated CMD population. Besides, this meta-analysis shows a promising implication that due to the limitation of invasive procedure, noninvasive examination techniques, such as ECHO, CMR, and PET, would be performed as the priority to screen for CMD in the future clinical practice, which would contribute to identification and intervention at the earlier stage, and reduce the risk of adverse events as well. Furthermore, we noted high overall heterogeneity in this meta-analysis which was similarly seen in the high-quality meta-analysis published before [10, 11]. The heterogeneity can be partly attributed to the fact that all included studies were observational in nature and recruited subjects with a wide range of risk profiles. Studies that utilized ECHO exhibited higher heterogeneity, which may be due to difference in the risk profiles of the recruited subjects, as this technique is commonly used as a handy screening tool. Moreover, ECHO is highly operator-dependent, which may lead to variations in the values of CFR among the included studies. In fact, comparing with various diagnostic modalities, we attach more importance on the prognosis of isolated CMD. Almost all of the studies, no matter which indices were used, showed significant adverse outcomes among isolated CMD population, which places a pivotal thought of detecting CMD rather than macro-vessel lesions only.

There are several limitations of our study. First, all included studies published observational data and were necessarily at high risk of bias. Therefore, we conducted the random effects model and subgroup analysis to minimize the potential heterogeneity. In addition, the pooled HR for MACE and mortality did not alter when we only included studies presenting HRs adjusting for clinical confounders. Second, the cutoffs for CMD measurement and definition of MACE varied across included studies. Third, studies are at risk of publication bias especially in studies for calculating the pooled HR for MACE, because only the positive studies are likely to be reported. When we excluded part of studies of relative smaller sample size [22, 24, 43], the regression tests for funnel plot asymmetry did not show publication bias in studies for pooled HR for MACE, and the overall finding did not alter (HR for MACE, 4.35, 95% CI 2.96–6.41, p < 0.00001, Egger’s test: p = 0.077). Fourth, the majority of included studies were performed at single centers and part of the studies consisted of small sample size. Finally, it is worth noting that some patients with CMD may exhibit positive stress myocardial ischemia. Therefore, we may have excluded some patients who had positive stress myocardial ischemia in the absence of coronary angiography. However, our sensitivity analysis showed a similar outcome when we only included studies that employed coronary angiography to screen for obstructive CAD.

An increased risk of death and MACE was revealed in isolated CMD patients compared to those without CMD. Various measurement modalities can be used to screen for CMD. More large-scale studies are needed to explore the management and treatment for CMD to improve prognosis.

An ethics statement is not applicable because this study is based exclusively on the published literature.

The authors have no conflicts of interest to declare.

This work is supported by the National Key Research and Development Programme of China 2021YFA1000200 and 2021YFA1000204.

Research idea and study design: Xingyu Luo, Yaokun Liu, and Bo Zheng. Data acquisition: Xingyu Luo and Yaokun Liu. Statistical analysis: Xingyu Luo. Data analysis and interpretation: Xingyu Luo, Yaokun Liu, Bo Zheng, and Yanjun Gong. Results discussion: Xingyu Luo, Yaokun Liu, Bo Zheng, Yanjun Gong, Jiahui Liu, Jin Zhang, Yanyan Zhang, Songyuan Gao, Zuoyi Zhou, Haotai Xie, Weijie Hou, Yan Zhang, and Jianping Li. Manuscript drafting: Xingyu Luo and Liu YK. Manuscript review: Bo Zheng. Supervision or mentorship: Zheng B and Gong YJ.

Xingyu Luo and Yaokun Liu are coauthors.

All data generated or analyzed during this study are included in this article and online supplementary material. Further inquiries can be directed to the corresponding authors.