Using Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA), we conducted a systematic review of studies looking at the relationship of COVID-19 cardiac sequelae and cardiovascular magnetic resonance imaging (CMR) (Fig. 1). Cross-sectional, case-control, and cohort studies were included in the analyses.

Fig. 1.

PRISMA flow chart.

A PubMed database search from January 1, 2021 to February 23, 2022 was performed. Observational studies, including cross-sectional, case-control, and cohort studies that examined post-COVID-19 cardiac symptomatology with CMR were included in the analyses. Studies that were excluded included: (1) case reports, case series, and conference abstracts; (2) papers not written in English; (3) protocol papers, letters to the editor, and healthcare provider surveys without data; and (4) papers that did not use CMR as a data metric. Search terms, PRISMA checklists and additional information regarding the search strategy can be found in the Supplementary Materials. This protocol was not registered online.

Study characteristics, including study type, origin, follow-up period and qualitative findings were systematically categorized and MRI findings, including T1, T2, late gadolinium enhancement (LGE) assessments, and the presence or absence of myocarditis, were described.

Twenty-one studies fulfilled the inclusion criteria (Table 1, Ref. [19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39]). Seventeen were prospective, 3 were retrospective, and 1 cross-sectional. Three studies [24, 25, 30] were performed in North America, 7 in Asia [19, 23, 27, 31, 34, 37, 39] and 11 in Europe [20, 21, 22, 26, 28, 29, 32, 33, 35, 36, 38]. Male predominance was present in 10 studies [19, 21, 24, 27, 28, 31, 32, 35, 36, 38], women were more represented in 9 [20, 22, 23, 25, 26, 29, 33, 37, 39] and in 2 studies gender was not available [30, 34]. The main CMR characteristics are summarized in Table 2 (Ref. [19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39]).

Three studies reported [24, 30, 38] CMR results during the acute COVID-19 hospitalization. The mean time from COVID-19 diagnosis to follow-up was 20 days, representing 375 patients. In all three studies at least one sign of myocardial injury and myocarditis was noted. Only two out of the three studies reported T1, T2, extracellular volume (ECV), and LGE data, with elevations of T1, T2 and ECV in addition to LGE present in both studies.

Ten studies [19, 20, 21, 22, 25, 27, 28, 34, 35, 36] reported CMR results three weeks to three months after SARS-CoV-2 infection, encompassing a total of 565 patients. Abnormal T1 values were noted in 3/6 studies, abnormal T2 values in 3/6 studies, elevated ECV in 2/5 studies, LGE in 9/10 studies, and myocarditis in 6/8 studies.

Eleven studies [21, 23, 25, 26, 29, 30, 31, 32, 33, 37, 39] reported CMR results beyond three months after SARS-CoV-2 infection. The average follow-up time was 15 weeks to 9.6 months with a total of 1566 patients represented. Abnormal T1 was reported in 3/10 studies, abnormal T2 in 1/7 studies, elevated ECV in 1/7 studies, LGE in 3/8 studies, and myocarditis in 1/4 studies.

Tabulating available data, we found the rate of myocarditis via CMR to be 18% (154/868) across all studies.

Cardiac and inflammatory biomarkers were reported in 18 out of the 21 studies, the most common of which include troponin, N-terminal pro b-type natriuretic peptide (NT-proBNP), and C-reactive protein (CRP). Additional biomarkers reported included IL-6, D-dimer, LDH, creatinine kinase (CK), procalcitonin (PCT), myoglobin (MYO), ferritin, and myeloperoxidase.

A direct association between these biomarkers and CMR findings at either hospital admission or follow-up was reported in 8 of the 18 studies (Table 3, Ref. [19, 23, 24, 27, 30, 32, 37, 39]). In these studies, an elevated troponin was associated with CMR pathology in 5/6 studies, CRP in 3/5 studies, NT-proBNP in 1/2 studies, and LDH and D-dimer in a single study. Although the remaining 10 studies did measure biomarkers, they were not included in the analysis because they did not analyze them in relation to CMR pathology.

In this systematic review, we found that CMR abnormalities associated with SARS-CoV-2 infection resolved over time. T1 (indicative of myocardial fibrosis and/or edema) and T2 (specific for myocardial edema [40]) values, in addition to ECV, were elevated in the acute COVID-19 hospital setting. Compared to the acute-disease phase, the 0–3 months phase showed a decrease in T1 and T2 elevations consistent with a decline in myocardial edema, and a decreased ECV elevation compatible with decreased myocardial inflammation. Compared to the 3 week–3-month phase, fewer studies demonstrate elevated T1, T2, and ECV in the $>$3-month phase, further suggesting that myocardial edema and inflammation of COVID-19 infection resolves over time. The persistence of LGE in the 3 week to 3 month phase might be consistent with myocardial scar formation and regional myocardial fibrosis. However, the decrease in LGE in the $>$3-month studies suggests that in addition to myocardial fibrosis, LGE is detecting reversible myocardial injury, consistent with studies of non-COVID-19 myocarditis and with some cases of myocardial infarction [41]. Likewise, CMR-defined myocarditis also showed timewise resolution.

The hypothesis that CMR abnormalities resolve over time is supported by the four prospective studies within the overall cohort that analyzed COVID-19 patients at two different follow-up times [21, 25, 30, 36]. Cassar et al. [21] found evidence of decreased T1, ECV, and LGE at a 6-month follow-up compared to an earlier 2–3-month follow-up of the same patients. Leslie Li et al. [30] found that 75% of patients showed complete resolution of myocardial edema/inflammation at a 138-day follow-up relative to a 36-day follow-up, as well as a significant improvement of LGE burden over that same time. Hanneman et al. [25] showed increased LVEF, decreased T1, T2, and ECV at a 119-day follow-up compared to a 67-day follow-up in those with FDG uptake. Using a different analysis, Ulloa et al. [36] reported that patients analyzed via CMR at 8 weeks from COVID-19 diagnosis had significantly less circumferential and radial strain compared to those analyzed at $>$8 weeks. Together, there is strong evidence in support of the hypothesis that cardiac abnormalities found on CMR resolve over time. Whether the persistence or resolution of CMR findings offer prognostic value for long-term adverse cardiovascular events remains to be determined.

Quantitative analysis of a panel of biomarkers may be useful in predicting patients who are more likely to have COVID-19-associated CMR abnormalities and, thus, who should be referred for CMR analysis. Myhre et al. [32] showed that elevated levels of NT-proBNP and troponin on admission were associated with positive cardiac MRI findings during long-term follow-up, but inflammatory markers were not. Galea et al. [24] report that post-acute COVID-19 patients with elevated troponin-T values have higher T2 and ECV values than patients below the upper limit troponin cutoff with a sensitivity of 92.9% and specificity of 76.9%. Multiple authors have reported an association between CMR abnormalities and admission troponin I levels, even in the absence of clinical symptoms [19, 27, 37, 39], including the presence of LGE many months after infection. This suggests that patients with elevated troponin-I levels at admission should be monitored closely for many months after discharge.

Although a meta-analysis by Fu et al. [42] supports the association of severity of the acute illness phase of COVID-19 and biomarker-defined cardiac injury (a seven-fold higher prevalence of biomarker-defined cardiac injury in severe compared to non-severe patients), others reported that disease severity during the acute illness is not associated with CMR pathologic findings [24, 27, 32]. This dichotomy may be explained by the fact that biomarkers such as troponin-I can be elevated in non-cardiac injuries, including renal insufficiency and pulmonary embolism, which are also a result of COVID-19.

Multiple risk factors, such as acute COVID-19 illness severity, older age, male sex, preexisting cardiovascular disease, hypertension, and COPD have been previously suggested to be associated with higher rates of myocardial injury as defined by elevated biomarkers [43, 44]. In our review, some studies found evidence in support of an association with older age [23, 32, 35] while others did not [27]. This may be due, in part, to the higher disease severity of COVID-19 seen in those studies that show an association [23, 32, 35]. Furthermore, it is unknown whether the CMR changes seen in older patients are preexisting rather than COVID-19 related. Similarly, the relationship between male sex and CMR evidence of injury is unclear [27, 32, 35, 45, 46]. In a pooled meta-analysis of 4 studies, Zou et al. [43] found no statistical significance between male sex to the appearance of cardiac injury.

Multiple mechanisms of cardiac injury due to SARS-CoV-2 infection have been postulated. These include ischemic and non-ischemic pathways [47]. Ischemic mechanisms, which include acute coronary artery and arteriole occlusion, involve pathological endothelial activation and thrombosis. Non-ischemic mechanisms include myocarditis, systemic hyperinflammation, hypoxic injury due to severe respiratory infection, and down regulation of ACE2 receptors. All nine studies which separated cardiac injury into non-ischemic and ischemic causes reported a greater prevalence of non-ischemic injury [20, 22, 24, 28, 29, 32, 35, 36, 38] suggesting that thrombosis and endothelial damage driven cardiac injury is less likely in COVID-19 associated cardiac injury. While Thornton et al. [35] reported the prevalence ischemic injury in their cohort to be 17%, they concluded that this most likely reflected pre-existing comorbidities. Interestingly, Saricam et al. [34] report that COVID-19 patients with markers of cardiac injury have higher nitric oxide levels than those with no markers. Given that nitric oxide is a potent anti-inflammatory molecule and vasodilator released by endothelial cells, this suggests that endothelial dysfunction and a subsequent prothrombotic state may be a mechanism of COVID-19 cardiac injury. Though it should be noted that this is a relatively minor finding, and that the majority of papers in this review reported a mainly non-ischemic mechanism of injury.

Cytokine storm in severe COVID-19 results in multiorgan damage including the lung, heart, kidney, testis, and liver [48, 49, 50, 51, 52, 53, 54] due to vascular hyperpermeability, edema, and hypercoagulation, has been shown to be accompanied by elevation of the inflammatory markers IL-6, IL-8, and CRP in those with cardiac injury [55, 56, 57]. Normalization of these biomarkers over time is accompanied by a trend towards normalization of CMR abnormalities [25]. Moreover, it has been demonstrated that COVID-19 patients with multisystem inflammatory syndrome (MIS), characterized by a systemic hyperinflammatory state, have significantly elevated native T1, whereas only regional inflammation and edema was noted in the non-MIS group [30]. Likewise, Cassar et al. [21] and Ulloa et al. [36] also conclude that CMR abnormalities are likely the result of cytokine-mediated injury.

Myocarditis is another mechanism of cardiac injury in COVID-19 patients. Given the patient populations analyzed in these studies, the 18% (154/868) incidence we report is likely an overestimation of the true prevalence of myocarditis in COVID-19 infected patients. This means that although myocarditis may be a mechanism of cardiac injury in COVID-19 patients, it is not the only mechanism of injury. This conclusion is further supported by endomyocardial biopsies which, in addition to confirming myocarditis through the presence of lymphocytic infiltrate with associated myocyte necrosis, have also confirmed macrophage-dominated inflammation without any myocyte necrosis, which is inconsistent with myocarditis-mediated cardiac injury [38, 58].

The studies included in this systematic review are heterogeneous in nature. Variability in the study design, size, use of controls, CMR parameters, and patient populations resulted in patient selection bias. Many studies did not include proper control groups and thus the incidence of cardiac abnormalities could not be compared with their natural incidence. Due to the heterogenous nature of the studies, it was impossible to report more granular detail beyond the presence and absence of CMR abnormalities. Some CMR findings could be pre-existing and were not caused or associated with SARS-CoV-2. The notion that these CMR findings were observed were at least likely exacerbated by SARS-CoV-2 infection. Arrythmias, which are a reported complication of COVID-19, could not be analyzed [59, 60]. Despite COVID-19 being a systemic disease, mechanisms of cardiac injury were only analyzed in the context of the cardiovascular system. Lastly, there could be unintentional reporting biases in published literature as many studies were retrospective studies and case series reports.