Long-term cardiac symptoms following COVID-19: a systematic review and meta-analysis

Search strategy and selection criteria

The initial search was conducted from January 1, 2020, to May 10, 2021 in seven electronic databases (PubMed, Embase, Web of Science, CINAHL Cochrane Library, Scopus, and PsycINFO), and two preprint servers (medRxiv and bioRxiv). We searched for both preprints and peer-reviewed articles that were published. The following broad search terms and keywords were used: (“COVID-19” OR “SARS-CoV-2” OR “COVID” OR “COVID19” OR “2019-nCoV” OR “Coronavirus” OR “SARS-CoV-2” OR “severe acute respiratory syndrome coronavirus-2”) AND (”post-acute COVID-19 syndrome” OR “post-acute” OR “sequelae” OR “long-term symptoms” OR “long-COVID” OR “post-COVID syndrome” OR “post-COVID symptoms” OR “hauler” OR “long-haul” OR “lingering” OR “chronic covid” OR “chronic symptoms” OR “persistent” OR “recurrent” OR “recurring” OR “complication” OR “subacute”). To include the up-to-date articles and have a manageable number of articles to screen, we conducted a second-round article search in PubMed, and searched for peer-reviewed articles published on PubMed from May 11, 2021, to January 8, 2022, using the same search terms as the previous data pull and including cardiac related terms: “heart” OR “chest” OR “cardiovascular” OR “myocardial” OR “cardiac” OR “palpitation” OR “tachycardia” OR “arrhythmia” OR “myocarditis”.

There were no language restrictions for articles included. The exclusion criteria were as follows: 1) reviews, commentaries, abstracts only, posters, or conference proceedings; or 2) studies having a follow-up duration of fewer than four weeks (28 days) or reporting no information on follow-up duration; or 3) studies with a sample size of less than 30 participants; or 4) cases with long-term cardiac symptoms among individuals who had not tested positive for COVID-19; or 5) non-human studies; or 6) study participants who were not alive at the time of the study (ex. post-mortem examination); or 7) studies that only presented data from modeling outputs or did not use primary data (i.e., secondary data, or summarize findings from previously published papers); or 8) no relevant symptoms or only reported biomarkers rather than symptoms. When several studies were based on the same or overlapping study participants, the study with the largest sample size was used as the representative study.

Screening process and data extraction

The article screening process was conducted using Covidence, a web-based collaboration software platform that streamlines the production of systematic and other literature reviews.²² At least two reviewers independently screened the title and abstract of each study for full-text examination. Two reviewers further conducted the full-text examination to evaluate if the study met the eligibility criteria. Any disagreements among reviewers were resolved through group discussion. For articles meeting the inclusion criteria, the following information was extracted by a single reviewer to a spreadsheet and checked by another reviewer: 1) author and publication information; 2) study characteristics (study country, study period, patient diagnosis/recruitment period, study design, COVID-19 diagnosis tools, study setting, study population, sample size, start of follow-up time, outcome assessment time points, outcome assessment method); 3) participants characteristics (age, sex, comorbidities or symptoms at baseline, COVID-19 treatment, vaccination status, hospitalization status, severity); and 4) results (prevalence of the outcome, outcome duration) (Table S2 and S3).

Quality assessment

A modified Newcastle–Ottawa scale (NOS)²³ was developed to evaluate quality of seven items for each included study: sampling representativeness, sample size, exposure assessment, outcome assessment, covariate assessment, follow-up, and statistics analysis (Table S4). The quality of each item was scored as ‘good’ (2), ‘fair’ (1), or ‘poor’ (0), and a total score was calculated (range: 0-14). A study with a total score of 11 to 14 was classified as of ‘high’ quality, a score of 7 to 10 was classified as of ‘medium’ quality, and a score of 0 to 6 as ‘low’ quality.

Two reviewers appraised each study independently, and disagreements were resolved through consensus.

Data synthesis and statistical analysis

All analyses were performed using R (version 4.2.2). We described the characteristics of included studies and presented proportions of commonly reported cardiac symptoms, calculated as the number of COVID-19 survivors who reported a specific cardiac symptom divided by the total number of COVID-19 survivors. Meta-analysis was performed using the meta package in R using both fixed-effects and random-effects models to estimate the pooled proportion of long-term cardiac symptoms following COVID-19 and 95% confidence intervals (CI). Heterogeneity between studies was assessed using the τ² (between-study variance) and I² (total proportion of variance owing to heterogeneity) statistics. The Cochrane Q-test was used to determine statistical significance. Studies that did not report the number of COVID-19 survivors with long-term symptoms and/or size of the total study population were excluded from meta-analysis, and were only included in the systematic review. Case-control studies were not included in the meta-analysis because the proportion of each symptom was determined by study design. Additional stratified analyses were performed based on quality assessment score (low, medium, high), sample size (30-99, 100-999, ≥ 1000), sampling representativeness (online survey and single hospital, multiple hospitals, national studies), and study design (cross-sectional, retrospective cohort, prospective cohort) to examine how each long-term cardiac symptom was distributed by these study characteristics. Funnel plots were plotted using the log transformation of the proportion of the symptom against the standard error of proportion to explore the presence of publication bias, followed by Egger’s test to assess funnel plot asymmetry.

Findings

Study selection

The initial search of studies published between January 2020 and May 2021 yielded 40,215 records from electronic databases (Figure 1). After the removal of 17,176 duplicates and the screening of titles and abstracts of 23,039 unique records, 321 studies underwent full-text examinations, and 65 studies met the inclusion criteria. The second-round search of studies published between May 2021 and January 2022 yielded 3,830 new records, among which a total of 36 studies met the inclusion criteria. In total, 101 studies on long-term cardiac symptoms following COVID-19 were included in this systematic review, and 92 studies provided data that could be used in the meta-analysis.

• Download figure
• Open in new tab

Figure 1. PRISMA flow diagram for systematic reviews

Quality assessment

Figure 2A shows the number of studies by the total score of quality assessment. Only 16 studies out of 101 studies received a total score of 11 to 14, considered to be of ‘high’ quality; 66 studies received a total score of 7 to 10, which were of ‘medium’ quality; and the remaining 19 studies received a total score of 6 and below, considered of ‘low’ quality. Figure 2B summarizes the proportion of studies receiving a score of ‘good’ (2), ‘fair’ (1), and ‘poor’ (0) for each of the seven quality assessment items. Across seven quality assessment items, the proportion of studies scoring ‘good’ was highest for exposure assessment and lowest for sampling representativeness and follow-up. A detailed quality assessment for each study by domain is presented in Figure S1.

• Download figure
• Open in new tab

Figure 2. Quality assessments of the included studies.

Meta-analysis

In total, ten cardiac symptoms were reported by five and more studies, including chest pain (n=70), arrhythmia (n=36), hypertension (n=10), cardiac abnormalities (n=10), myocardial injury (n=7), thromboembolism (n=6), stroke (n=6), heart failure (n=6), coronary disease (n=5), and myocarditis (n=5) (Table S5). Fixed-effects and random-effects meta-analyses were conducted for these symptoms (Figure S2). Substantial heterogeneity across studies was observed for all symptoms. The pooled proportion for each of the symptoms using a random-effects model was as following: chest pain (10.06%, 95% CI: 6.40-15.47), arrhythmia (9.80%, 95% CI: 5.39-17.18), hypertension (6.07%, 95% CI: 3.42-10.53), cardiac abnormalities (0.05%, 95% CI: 0.03-0.08), myocardial injury (2.63%, 95% CI: 0.13-36.30), thromboembolism (2.68%, 95% CI: 0.83-8.37), stroke (0.71%, 95% CI: 0.18-2.84), heart failure (1.23%, 95% CI: 0.41-3.59), coronary disease (0.41%, 95% CI: 0.23-0.70), and myocarditis (0.62%, 95% CI: 0.08-4.44).

Stratified analyses were conducted using the following characteristics: total quality assessment score, sample size, sampling representativeness, and study design. Heterogeneity remained high for most symptoms within the stratum after stratification. Figure 3 summarizes results of stratified analyses for chest pain and arrhythmia. For both symptoms, studies with the lowest quality score, smallest sample size, poorest sampling schemes, and cross-sectional designs reported the highest proportion of the symptom. For example, the proportion of chest pain among individuals survived COVID-19 was 21.32% (95% CI: 10.50-38.50) for low quality score studies, 9.26% (95% CI: 6.04, 13.95) for medium quality score studies, and 4.04% (95% CI: 1.28, 11.98) for high quality score studies. We observed similar patterns for other symptoms, although a small number of studies within some strata precluded formal analyses (Figure S3).

• Download figure
• Open in new tab

Figure 3. Meta-analysis results of top two most reported symptoms: chest pain and arrhythmia, using a random effects model*

* Figure 3A: Chest pain; 3B: Arrythmia.

For stratified analysis of each characteristic from left to right: quality score (low, medium, high), sample size (30-99, 100-999, ≥ 1000), sampling representativeness (online survey and single hospital, multiple hospitals, national studies), and study design (cross-sectional, retrospective cohort, prospective cohort).

Interpretation

Our systematic review and meta-analysis provide a comprehensive overview of evidence for the long-term cardiac symptoms following COVID-19. Up to January 2022, there were 101 studies that examined 49 different long-term cardiac symptoms. These studies varied substantially in design, and a small proportion of them (16%) were of high quality. Chest pain and arrhythmia were the two most studied long-term cardiac symptoms, and results from meta-analysis show that around 10% of participants had them. Meta-analysis identified high heterogeneity across studies for all cardiac symptoms and subgroup analyses showed systematic differences in reported proportions by selected study characteristics, including total quality assessment score, sample size, sampling representativeness, and study design.

It is evident that many COVID-19 survivors may experience chronic cardiovascular symptoms, even those without previous cardiovascular disease, comorbidities, and who have a baseline low risk of cardiovascular disease. To date, multiple reviews have examined the long-term cardiac symptoms following COVID-19 (Table S6).^{10–12,14–16,18–20} For example, several reviews quantified chest pain and arrhythmia as two of the most common long-term symptoms, with proportion estimates for chest pain ranging from 5.0% to 16.0% and proportion estimates for arrhythmia ranging from 9.7% to 11%.^10,13,16,18 Our proportion estimates of these two symptoms broadly agree with findings from these previous reviews. A probable source of differences in estimates across these systemic reviews is that they used different inclusion criteria to select studies.

Although pathophysiological mechanisms underlying COVID-19 cardiac symptoms remain unclear, studies suggest that the chronic inflammatory response may be hyperactivated by persistent viral reservoirs in the initial acute phase, which may lead to post-acute COVID-19 cardiovascular sequelae.^19,20 Studies have shown that over 20% of patients with acute COVID-19 had evidence of cardiac injury, even if they did not have underlying cardiovascular diseases or pre-existing comorbidities.^24–27 It is hypothesized that viral invasion through binding angiotensin-converting enzyme-2 (ACE-2) causes a cytokine storm and triggers systemic hyper-inflammation, which can affect multiple organ systems and induce cardiac injury as one of the severe complications.^13,14 Persistent chest pain and arrythmia may be indicative of underlying cardiac abnormalities and damage resulting from the systematic hyper-inflammation and/or viral yocarditis affecting the cardiac conduction system. It is critical for clinicians to thoroughly examine patients with long-term cardiac symptoms following COVID-19, especially for survivors with pre-existing cardiac conditions and other high-risk comorbidities.

We found that studies with a low-quality assessment score, small sample size, non-systematic sampling methods, and cross-sectional designs were more likely to report high proportions of long-term cardiac symptoms. For example, the pooled proportion of chest pain among low-quality score studies (21%) was over four times higher than that among high quality score studies (4%). Such patterns were also observed for arrhythmia and other long-term cardiac symptoms. This strong relationship between the proportion of long-term cardiac symptoms following COVID-19 and study quality and design characteristics should be considered when interpreting existing studies.

In our quality assessment, we determined that around 16% of included studies were of high quality^6,7,28–41 and the remaining 84% were of medium or low quality. The small number of high-quality studies demonstrates the urgent need to improve the quality of studies investigating the long-term cardiac symptoms following COVID-19. A key feature among most included studies of medium or low quality was that they were predominantly based on clinical or hospital samples. These studies usually had small sample sizes and had no or only one point of follow-up. While relatively small clinical or hospital-based studies, which are often easier to conduct in shorter time periods, can be useful to establish preliminary evidence of an association especially in an emergent situation such as a pandemic, they often lack population representativeness and statistical power to make broader conclusions about the hypothesized relationships. Furthermore, cross-sectional studies cannot establish the temporality required to infer any causal relationship and also prohibit examinations of changes in symptoms over time.

Based on the above findings, we formulated recommendations for the design and analysis of future studies on long-term cardiac symptoms of COVID-19 (Table 2). Many studies adopted convenience sampling schemes, which hinders the interpretability and generalizability of their findings. Therefore, it is important to conduct systematic sampling, which can facilitate a continuing and meaningful exploration of data collected and underpin clinical research. The majority of included studies only assessed long-term symptoms at a single time point. It is therefore challenging to examine how the long-term symptoms may change over time. Many studies did not distinguish between long-term symptoms following COVID and pre-COVID symptoms at baseline level.

Our study has several limitations. First, studies included in our systematic review and meta-analysis are highly heterogenous. We therefore performed subgroup analyses by multiple characteristics, and we believe that existing heterogeneity across studies makes it difficult to generalize our results to the general population. Second, many reported outcomes had a small sample size, which increases the possibility for publication bias. Third, we were unable to stratify our meta-analysis by length of follow-up because of widely varying follow-up times across studies. Finally, we were not able to stratify our analyses by prior comorbidities, previous cardiovascular disease, history of treatment or medication use for cardiac symptoms, or COVID-19 vaccination status, mainly because this information was underreported, especially in the articles published in the early phase of the pandemic. As the pandemic continues to progress, the virus mutates, treatment strategies for acute and long COVID-19 evolve, and as the uptake of vaccine increases, it is possible that the epidemiology of long-term cardiovascular manifestations in COVID-19 might change over time.

Our systematic review and meta-analysis also have multiple strengths. First, to our knowledge, this is the first systematic review focusing on long-term cardiac symptoms following COVID-19 that included preprints and articles published in all languages and in the global network. Second, in an effort to ensure our results were current, we updated our search to capture articles published from the early phases of the pandemic to the most recently published studies. Third, we assessed the quality of included articles from the perspective of study design and epidemiologic principles and provided detailed recommendations on future long-COVID epidemiologic research. Fourth, we performed meta-analysis and subgroup analyses to examine patterns of reported findings.

In conclusion, this systematic review and meta-analysis synthesized the evidence on the long-term cardiac symptoms following COVID-19. We found there were diverse manifestations of cardiac symptoms, and many can last for months and even over a year. There is substantial heterogeneity in terms of study design and systematic differences in reported proportion of symptoms by study characteristics. Specifically, we found that studies with low-quality, small sample size, unsystematic sampling method, or cross-sectional design were most likely to report high estimates of symptoms among individuals survived COVID-19. We believe that a deeper understanding of long COVID is currently prevented by the limitations of the published literature. Our study underscores the need to conduct high-quality studies on long COVID and the importance of long-term cardiac surveillance of COVID-19 survivors.