Introduction

Since December 2019, the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused a global outbreak of coronavirus disease 2019 (COVID-19). Currently, the pandemic has affected more than 127,319,002 people in more than 200 countries and killed more than 2,785,838 people (https://www.who.int/emergencies/diseases/novel-coronavirus-2019). Previous studies have reported that several pre-existing medical conditions, such as hypertension, diabetes and so on, might accelerate disease progression of COVID-19 [1,2,3]. Cardiovascular disease (CVD), one of the most common comorbidities of COVID-19, has been observed to be associated with adverse outcomes among COVID-19 patients by Li et al. in a meta-analysis study [4]. Nevertheless, it is worth noting that the results of Li et al.’s study were based on the unadjusted effect estimates [4]. It is reported that age, sex, and co-existing diseases are known to affect the outcomes of COVID-19 patients [5,6,7], which may modulate the association between CVD and adverse outcomes in COVID-19 patients. Moreover, Zhou et al. observed that coronary heart disease (CHD), one of CVD, was strongly correlated with an increased risk of in-hospital mortality among COVID-19 patients in univariable analysis (odds ratio (OR) = 21.4, 95% confidence interval (CI): 4.64-98.76), but no significant correlation was observed in multivariable analysis (OR = 2.14, 95% CI: 0.26-17.79) [8]. The similar results were also observed by Robilotti et al. [9] and Louapre et al. [10]. Therefore, it is necessary to clarify whether pre-existing CVD was an independent risk factor associated with adverse outcomes in COVID-19 patients. In this study, we performed a quantitative meta-analysis on the basis of adjusted effect estimates.

Methods

This is a quantitative meta-analysis on the basis of adjusted effect estimates. Admittedly, our study was not registered, but our meta-analysis was made in strict accordance with the process of systematic evaluation (Fig. 1). Moreover, our study is less likely to be biased by artificial bias because this study was carried out rigorously in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines (Online supplemental Table A1) [11].

Fig. 1
figure 1

Flow diagram of selection process

Full size image

Literature search strategy

The databases of PubMed, Web of Science, MedRxiv, Scopus, Elsevier ScienceDirect, Cochrane Library and Embase were searched to obtain a complete data source up to January 7, 2021. The search strategies were as follows: (“COVID-19” OR “coronavirus disease 2019” OR “SARS-CoV-2” OR “2019-nCoV”) AND (“cardiovascular disease” OR “coronary heart disease” OR “cardiac disease” OR “heart disease” OR “heart failure” OR “coronary artery disease”) AND (“outcome” OR “severe” OR “critical” OR “severity” OR “fatality” OR “mortality” OR “death” OR “adverse outcome” OR “poor outcome” OR “clinical characteristics”). All the terms matched the MesH browser. Beyond that, the relevant references of preceding studies were also taken into account.

Eligibility criteria

The criteria for including studies were: (1) Subjects should be laboratory-confirmed COVID-19 patients; (2) Studies should report the correlation between CVD and COVID-19 patients and the data are available; (3) Studies should be published in English; (4) Studies should include the multivariate analysis. The studies with the largest sample size were selected for inclusion when studies were conducted in the same hospital and the overlapping period. There was no restriction for region of study. The exclusion criteria included case reports, review papers, comments, errata, repeated studies, studies only reporting the characteristics of COVID-19 patients with CVD, and studies without available full text.

Data extraction and quality assessment

Data were extracted independently by two investigators (J.X. and W.X.), including the following information: the first author, source of data, country, date of data collection, number of patients, mean/median age, the percent of males, study design, the percent of COVID-19 patients with CVD, adjusted effect estimates (hazard ratio (HR) or OR) and adjusted risk factors. When both OR and HR existed in the same article, it was preferred to include HR because cox regression took time into account. Two researchers negotiated to resolve it in case of any issues not covered by the criteria and Y.W. acted as arbiter. The quality of the included studies was evaluated by investigators according to the Newcastle-Ottawa Scale [12]. High-quality studies referred to studies with a score above 7.

Data synthesis

The major information such as study design and effect estimates were directly extracted from original articles. The research type of some articles was not clear and some articles provided both OR and HR. Besides, the calculation methods of HR and OR are different. The calculation of HR takes into account the concept of time, and OR is the approximate value of risk ratio. Therefore, pooled HR, OR and 95% confidence intervals (CIs) were separately calculated to address the risk of adverse outcomes in COVID-19 patients with a history of CVD. Heterogeneity was assessed by Cochran’s Q-statistic and I2 test, if no significant heterogeneity was observed (I2 ≤ 50%, P > 0.1), a fixed-effects model was adopted; otherwise, a random-effects model was applied [13]. In addition, we also provided the prediction interval, which was helpful for assessing whether the variation across studies was clinically significant [14, 15]. The robustness of the results was evaluated by sensitivity analysis which omitted one study at a time. Publication bias was assessed by Begg’s test [16], Egger’s test [17] and trim-and-fill method [18]. Subgroup analysis and meta-regression were conducted to determine the source of heterogeneity. Data analyses were conducted using Stata, version 12.0 (meta-program) and R, version 3.6.1 (netmeta package). A two-tailed P-value < 0.05 was regarded as significant.

Results

The flow chart of selection process is shown in Fig. 1. 5,025 records were retrieved after removing 23,826 duplicates, of which 245 studies were full-text assessed. Eventually, a total of 203 eligible studies with 24,032,712 patients were enrolled in our meta-analysis [2, 3, 8, 9, 19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199,200,201,202,203,204,205,206,207,208,209,210, 212,213,214,215,216,217,218]. 81 studies originated from Europe, 54 studies came from North America, 61 from Asia, 2 from Australia, and the remained 5 were not just from one country (Table 1). Among these studies, cardiac disease was mentioned in 63 studies, HF was involved in 35 studies, and CAD was involved in 35 studies (Table 2). Adjusted HR was reported in 65 studies and adjusted OR was reported in 138 studies (Table 2). The main characteristics of the selected studies are summarized in Table 1.

Table 1 Main characteristics of the included studies
Full size table
Table 2 The results of subgroup analysis
Full size table

Totally, our results revealed that COVID-19 patients who suffered from CVD tended more to adverse outcomes (pooled ORs = 1.41, 95% CIs: 1.32-1.51, prediction interval: 0.84-2.39; pooled HRs = 1.34, 95% CIs: 1.23-1.46, prediction interval: 0.82-2.21 Fig. 2). Subgroup analysis by sample size showed consistent results (pooled HRs = 1.16, 95% CIs: 1.03-1.32, prediction interval: 0.66-2.04; pooled ORs = 1.41, 95% CIs: 1.32-1.51, prediction interval: 0.84-2.39 for sample size >= 1000; pooled HRs = 1.63, 95% CIs: 1.41-1.88, prediction interval: 0.86-3.10; pooled ORs: 1.57, 95% CIs: 1.40-1.77, prediction interval: 0.84-2.95 for sample size < 1000; Table 2 and Fig. A1). The positive association between pre-existing CVD and adverse outcomes in COVID-19 patients was also observed in subgroup analysis by disease types (Table 2 and Fig. A2): cardiac disease (pooled HRs = 1.40, 95% CIs: 1.17-1.69, prediction interval: 0.68-2.90; pooled ORs = 1.43, 95% CIs: 1.25-1.64, prediction interval: 0.80-2.55), HF (pooled HRs = 1.23, 95% CIs: 1.05-1.44, prediction interval: 0.63-2.39; pooled ORs = 1.46, 95% CIs: 1.31-1.62, prediction interval: 1.01-2.10), and CAD (pooled HRs = 1.48, 95% CIs: 1.14-1.93, prediction interval: 0.67-3.29; pooled ORs = 1.17, 95% CIs:1.02-1.35, prediction interval: 0.75-1.83). In addition, subgroup analyses stratified by age, the proportion of males, region, disease outcomes and study design supported the above positive associations (Table 2 and Fig. A3-7). Sensitivity analysis indicated that our result was robust (Fig. 3A and B). There was no publication bias was detected by Begg’s test (OR: P = 0.233, HR: P = 0.054; Fig. 4A and B), while significant publication bias was found by Egger’s test (OR: P = 0.000, HR: P = 0.000; Fig. 4C and D). Therefore, the trim-and-fill method was adopted for further analysis. The results for HR showed that with the addition of 21 more studies, the results of the meta-analysis would be more robust but not reversed (pooled HRs = 1.11, 95% CIs: 1.01-1.14, fixed-effects model; pooled HRs = 1.16, 95% CIs: 1.06-1.26, random-effects model), and the OR results (pooled ORs: 1.18, 95% CIs: 1.16-1.20, fixed-effects model; pooled ORs: 1.21, 95% CIs: 1.12-1.30, random-effects model) showed that the results would be equally robust after adding 29 studies. However, there was high heterogeneity in our study. To find sources of heterogeneity, we conducted a meta-regression. However, adjustments for multivariate regression coefficients for sample size, age, proportion of males, study design, region, disease types, disease outcomes were not statistically significant (Table 3), suggesting that these were not sources of heterogeneity identified.

Fig. 2
figure 2

Forest plot of adjusted pooled effects for adverse outcomes associated with CVD in patients with COVID-19. A) Pooled OR; B) Pooled HR

Full size image
Table 3 The result of meta-regression
Full size table
Fig. 3
figure 3

Sensitivity analysis for pooled OR (A) and HR (B)

Full size image
Fig. 4
figure 4

Publication bias for pooled OR (A and B) and HR (C and D)

Full size image

Discussion

Many countries have been hit by the pandemic caused by SARS-CoV-2, numerous people lost their lives because of this. Meanwhile, health systems in every country were under so unprecedented strain that it was very important to find an effective marker to help implement bed grading management. What called for special attention was that earlier studies have shown COVID-19 patients with at least one underlying conditions, such as chronic kidney disease, HIV, diabetes and other comorbidities, have a poor disease course [2, 29, 211, 219, 220], which means that those patients with underlying diseases should be monitored more carefully in case of disease getting worse. Furthermore, it was reported that the risk of primary respiratory syndrome severity and adverse outcomes was increased in Middle East respiratory syndrome (MERS) patients with pre-existing CVD. The research by Li et al. [8] with unadjusted effect estimates showed that there was a positive association between CVD and adverse outcomes in patients with COVID-19, but the association might be confounded by other factors such as age, gender and comorbidities. Thus, we performed a quantitative meta-analysis on the basis of adjusted effect estimates to clarify whether pre-existing CVD was an independent risk factor associated with adverse outcomes in COVID-19 patients.

Our results based on adjusted effect estimates revealed that pre-existing CVD was significantly related to adverse outcomes in COVID-19 patients on the basis of 203 eligible studies with 24,032,712 cases. The significant association between pre-existing CVD and adverse outcomes in COVID-19 patients was still existent in further subgroup analyses stratified by the proportion of males, study design, disease types, sample size, region and disease outcomes, which suggests that our findings are relatively stable.

Similar to other meta-analyses, several limitations should be acknowledged in this present study. Firstly, data on drug and supportive treatments are not clear in the selected studies presently, thus, we could not evaluate the effects of treatments on the association between co-existing CVD and adverse outcomes in COVID-19 patients. Secondly, statistically significant results were more likely to be accepted and published than non-statistically significant results in similar studies, but in fact, the data of the meta-analysis mainly derived from the studies which have been published, which may lead to publication bias. Thirdly, the causal relationship of CVD and adverse outcomes in patients with COVID-19 cannot be confirmed on account of the inherent limitation of the observational study. Therefore, well-designed studies with larger sample sizes are needed for further verification.

Conclusions

In conclusion, our findings indicated that pre-existing CVD was an independent risk factor associated with adverse outcomes among COVID-19 patients. COVID-19 patients with a history of CVD might need more attention.