Search strategies were developed after a discussion with two physicians and a clinical epidemiologist (YDT, CLS, and SS). PubMed, Embase, Web of Science, and grey literature sources were searched exhaustively. Two additional websites, e.g., Chest, and the European Heart Journal, were searched to ensure that all current research was included and because of their respective high impact in publishing circulatory and respiratory systems research. A detailed outline of our search strategy has been provided in the Supplementary Materials, as Supplementary Material 1.

Studies identified through the aforementioned databases from inception until 26 November 2021 were initially considered eligible. Eligibility criteria are provided in Fig. 1. CAD was diagnosed and included: (1) existing myocardial infarction; (2) those treated with PCI or a CABG; (3) $>$50% stenosis of at least one of the three major coronary arteries (i.e., left anterior descending, circumflex, or right coronary artery), as observed through coronary angiography.

Fig. 1.

Flowchart according to PRISMA statement. Flowchart of the study. COPD, chronic obstructive pulmonary disease; CAD, coronary artery disease.

COPD was diagnosed and classified according to the pulmonary function test (PFT), International Classification of the Diseases (ICD codes), or through self-reported methods. It is important to note that in most studies, the PFT criteria met the gold standard criteria, although a small number of studies involved various other PFT criteria, which were developed before the gold standard was established. ICD codes indicate that patients might have been diagnosed with COPD prior to admission and should not undergo the PFT. Various studies exhibited divergent self-reported methodologies–for instance, a combination of clinical symptoms and COPD medication usage. Studies that did not report diagnostic methods for either CAD or COPD were excluded. Two reviewers independently screened studies (YTZ and ZLH) and discrepancies were resolved by the third reviewer (SS).

Two reviewers (YTZ and ZLH) independently assessed the risk of bias using two separate tools. For the prevalence of COPD–CAD, we used a customized Newcastle–Ottawa Scale (NOS), to classify studies. Scores $\le$3 were categorized as high-risk. For outcomes according to COPD status in CAD, we used another customized NOS tool related to outcomes, for which a score $\le$6 was thought to indicate a study with a high risk of bias [7]. Details of the customized tools have been provided in Supplementary Material 2.

Two reviewers (YTZ and ZLH) extracted and cross-checked data from studies, including demographics and study designs, such as country or region, study type, age, gender, etc. Detailed information has been provided in Table 1 (Ref. [4, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61,62, 63, 64, 65, 66, 67, 68, 69, 70, 71]) and Supplementary Table 1. Supplementary Material 3 including all supplemental figures and tables.

The random effects model (DerSimonian and Laird method) was implemented due to assumed differences within (and between) populations. Estimates with 95% confidence intervals (CI) have been provided. We adopted two types of proportional transformation, i.e., Logit transformation and Freeman–Tukey double arcsine transformation for sensitivity analysis. Leave-one-out analysis and exclusion analysis were conducted according to the number of participants in specific subgroups.

Subgroup analysis of COPD diagnostics, study type, location of study, economic status (according to World Bank), and risk of bias, were performed to identify potential sources of heterogeneity. Univariate meta-regression analyses were performed, with the prevalence of COPD in CAD as the dependent variable.

Independent variables included the COPD definition, economic status, study type, risk of bias, area, age, male, hypertension, diabetes mellitus, dyslipidemia, atrial fibrillation, stroke, and smoker status. Independent variables with p 0.05 were enlisted for multivariate meta-regression analyses. The proportion of variance in prevalence estimates was explained using R square calculations [72, 73].

For comorbidities and risk factors in the COPD group, odds ratios (ORs) with corresponding 95% CIs were calculated according to COPD status. Values with 95% CIs that did not include one were accepted as statistically significant.

For outcomes related to COPD–CAD status, the primary endpoint was all-cause mortality. Secondary endpoints included cardiac death, stroke, revascularization, myocardial infarction (MI), heart failure, and respiratory failure. The random effect model was also implemented to pool a conservative risk ratio of COPD (compared to non-COPD) in CAD, according to various endpoints.

Subgroup analyses were performed to compare mortality in different groups, specifically the PFT versus ICD codes/self-reported method, and CABG vs. PCI. Leave-one-out analysis was again performed to assess the impact of single studies on pooled risk ratios.

Studies that reported comparisons in outcomes related to CABG and PCI in COPD patients were enlisted for meta-analysis. Outcomes related to revascularization methods, such as all-cause death, myocardial infarction, stroke, and revascularization were established as endpoints. Additionally, publication bias was assessed using Egger’s test, with results presented in the form of a funnel plot.

All statistical analyses were performed using Stata (version 13.0, StataCorporation, Austin, TX, USA) and R (version 4.0.5, R Foundation for Statistical Computing, Vienna, Austria). This study was registered with PROSPERO (CRD #42021293270) and in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Please see Supplementary Material 4 for further details.

Study and participant characteristics, such as study type, research location, economic status, etc., are provided in Table 1 and Supplementary Table 1. Forest plots suggested that the pooled prevalence of COPD in CAD patients is 14.2% (95% CI: 13.3–15.1). Please see the Supplementary Materials, Supplementary Fig. 1, for further details. Supplementary Fig. 2 showed the publication bias of each study.

Sensitivity analysis was conducted using the inverse variance and Logit transformation methods and a similar prevalence was reported for each (Supplementary Table 2). Leave-one-out analysis suggested that there was no significant impact by a single study on the pooled COPD–CAD prevalence (Supplementary Fig. 3). However, during the sensitivity analysis, and by excluding studies according to sample size, some heterogeneity was found to exist. By initially excluding the smallest sample of studies, we found that heterogeneity was closely related to studies using the PFT as a diagnostic method (Supplementary Tables 3,4).

The pooled prevalence of COPD–CAD across different countries or regions was presented as a visualized version of the world map, with different colors indicating the extent of the COPD–CAD prevalence (Fig. 3). From the heat map, one can see that the prevalence appears highest in North America, followed by Asia, Europe, and South America. For countries in Africa and Oceania, evidence of COPD prevalence in CAD is lacking, with only one study from the African continent reporting on prevalence (Fig. 3).

Fig. 3.

Prevalence of COPD in CAD in various subgroups. COPD, chronic obstructive pulmonary disease; ICD, International Classification of the Diseases; N/A, not applicable.

The analysis of subgroups also helped to uncover potential factors that may influence prevalence. For example, the prevalence in the PFT group was significantly higher than the rate observed in the ICD codes group or in the self-reported group (21.3% vs. 14.6% vs. 8.8%) and was also significant (p 0.0001) (Fig. 3). There was no obvious difference between high-income countries and upper-middle-income countries, although there was statistical significance when comparing these with “undetermined income” countries, which included multinational clinical trials (15.3% vs. 15.0% vs. 6.7%, p 0.0001).

Univariate and multivariate meta-regression analyses were performed to identify potential sources of heterogeneity. COPD diagnostics appears to be the main source of heterogeneity, followed by study type, economic status, and diabetes mellitus (all p 0.05; R${}^2$: 28.39% vs. 15.01% vs. 6.32% vs. 5.76%). See Supplementary Table 3 for details. After imputing these factors into the multivariate model, statistical significance (p = 0.0016; R${}^2$ = 23.64%) remained (Supplementary Table 5).

Information on OR related to patient characteristics, according to COPD status, is provided in Table 2. A total of 23 studies reported the number of men in the COPD group and non-COPD groups, with no obvious differences noted according to gender (OR = 1.001, 95% CI: 0.87–1.15). A total of 17 studies also reported dyslipidemia rates in the two groups, although, again, no statistically significant differences were observed (OR = 1.03, 95% CI: 0.89–1.19).

Further comorbidities and risk factors, including hypertension, diabetes mellitus, atrial fibrillation, stroke, smoking, dyspnea, wheezes, and chronic bronchitis were all reported to be significantly higher in the COPD–CAD group, compared with the non-COPD–CAD group (all OR $>$1, with 95% CI: beyond 1).

The OR in the COPD group was nearly twice that in the non-COPD group (OR: 1.94, 95% CI: 1.57–2.4). Moreover, a higher incidence of atrial fibrillation and a history of stroke were both observed in the group with comorbid COPD. Atrial fibrillation provided an OR of 1.64 (95% CI: 1.14–2.36), while a history of stroke generated an OR of 1.72 (95% CI: 1.35–2.18). See Table 2 for further details.

A total of 23 studies reported all-cause mortality, 7 studies focused on cardiac death, 9 on myocardial infarction, and 6 focused on revascularization and stroke. Pooled all-cause mortality in the COPD group was triple that in the non-COPD group (risk ratio [RR] = 2.81, 95% CI: 2.40–3.29), see Fig. 4 and Supplementary Table 6.

Fig. 4.

Forest plot for risk of all-cause mortality according to COPD status. COPD, chronic obstructive pulmonary disease.

Pooled cardiac death, myocardial infarction, stroke, heart failure, and respiratory failure were significantly higher in the COPD group than in the non-COPD group (all RR $>$1). However, pooled revascularization was lower in the COPD group than in the non-COPD group (RR = 0.86, 95% CI: 0.75–0.97), see Supplementary Figs. 4–9. Publication bias for different outcomes was also assessed using Egger’s test (Supplementary Fig. 10) and described below.

Sensitivity analyses, using the leave-one-out method, were conducted to determine whether any single study impacted the overall pooled RR (Supplementary Fig. 11). There was no obvious impact of a single study on all-cause death, cardiac death, myocardial infarction, stroke, heart failure, and respiratory failure. However, after excluding the study by Lin et al. [12] from the pooled RR for revascularization, the pooled RR increased, although, ultimately, remained lower than the null hypothesis, i.e., one (Supplementary Fig. 11).

A total of two studies directly compared the outcome of COPD–CAD patients who underwent CABG vs. PCI therapy. These two studies were both large, high-quality multicenter randomized clinical trials [8, 9]. However, they were also both post hoc, non-prespecified explorations. Therefore, we pooled the outcomes of these two studies, to compare the impact of the revascularization method on the COPD–CAD patients. As shown in Supplementary Fig. 12, an obvious reduction in revascularizations was observed after therapy (OR: 0.43, 95% CI: 0.20–0.94), along with a reduction in myocardial infarction (OR: 0.62, 95% CI: 0.18–2.11). However, no obvious benefit was observed from PCI for all-cause mortality (OR: 0.97, 95% CI: 0.54–1.74). The risk of stroke after revascularization increased in the CABG group (OR: 2.00, with 95% CI: 0.50–7.94).

Further subgroup analysis was conducted to investigate the differences in the RR between the PFT and ICD codes/self-reported method groups, and the CABG vs. PCI groups (Supplementary Figs. 13,14). Slightly higher mortality was observed in the PFT group compared with the ICD-codes/self-reported method group, although this was not considered significant (3.08 vs. 2.94, p value for subgroup differences = 0.833) (Supplementary Fig. 14). We also found no significant decrease in mortality in the CABG group compared to the PCI group (2.97 vs. 3.43, p for subgroup difference = 0.427) (Supplementary Fig. 13).

Studies reporting on revascularization in COPD–CAD patients were systematically reviewed. A total of seven studies reported the PCI rate for COPD–CAD patients. The pooled OR of the prescription rate in the COPD group was 0.68, with 95% CI: 0.56 to 0.83 when compared with the non-COPD group. A total of six studies reported the CABG rate for COPD–CAD patients and also indicated a reduced prescription rate for CABG in the COPD group, with a pooled OR equal to 0.93 and 95% CI: 0.75 to 1.15 (Supplementary Fig. 15). No significant difference was observed in strictly corrected COPD group according to GOLD criteria when compared with not strictly corrected COPD group (Supplementary Fig. 16).

Biases associated with the prevalence and outcomes were assessed separately, including publication bias and quality assessment. Prevalence-based publication bias was assessed using Begg’s and Egger’s tests; both at p $>$ 0.05 (Supplementary Fig. 2). Publication bias for different outcomes was also assessed using Egger’s test (Supplementary Fig. 10). The main publication bias appeared to relate to the cardiac death and revascularization studies (cardiac death: Egger’s p = 0.027; revascularization: Egger’s p = 0.026). The risk of bias for prevalence and outcomes was independently assessed according to the methods described (Supplementary Tables 7,8).