The search resulted in 226 titles. After duplicate removal of and title screening, we screened the full text of 108 articles (52 from weekly alerts) that yielded 88 articles reporting PM tissue pathology. Among those, 41 studies reported PM heart examination and included 336 cases (Figure 1). Studies were mostly case reports (n = 13) or case series (n = 24), while three studies compared cases to controls (6–10, 13–48) (Table 2, Figure 2). Authors reported cases from 14 countries, mostly developed westernized ones (Supplementary Table 1). Two studies reported on the same population, with one mainly focusing on PM cardiac examination (22, 23). The quality of the included studies was mostly moderate (Supplementary Table 2).

We analyzed the PM cardiac histopathology for 316 cases [after excluding cases unconfirmed as COVID-19 (n = 6) or with no PM cardiac tissue examination (n = 14)].

Study characteristics and pathological findings are detailed in Supplementary Tables 1, 2, respectively. Cases were predominantly male (172/275, 62%). The deceased were mostly elderly [median: 75 years; interquartile range (IQR), 63–84 years; range, 22–97 years, n = 228] and overweight [body mass index (BMI): median, 27; IQR, 22.9–34.7 kg/m²; range = 15.4–61.2 kg/m², n = 148).

Cardiovascular comorbidities were prevalent, most commonly hypertension (n = 152, 48.1%), coronary artery disease (CAD) (n = 105, 33.2%), cardiomyopathy and heart failure (n = 68, 21.5%), and atrial fibrillation (AF) (n = 35, 11.1%). Other comorbidities included chronic respiratory diseases (n = 91, 28.7%), diabetes mellitus (n = 81, 25.6%), chronic kidney disease (CKD) (n = 53, 16.7%), dementia (n = 40, 12.7%), and cancer (n = 39, 12.3%).

The median duration of prehospital symptoms (n = 82) and hospital stay (n = 158) were 5 (IQR, 2–7) and 6 days (IQR, 3–10), respectively. In total, the median duration from the onset of symptoms to death was 12 days (range, 0–52 days, n = 98). The median time interval between death to PM autopsy was 1.2 days (n = 31).

Cardiac abnormalities either on gross pathology or histology were identified in almost all cases. Most autopsies demonstrated chronic cardiac pathologies [hypertrophy (n = 85), fibrosis (n = 72), and amyloidosis (n = 11)], which may have contributed to the increased heart weight where this was reported (median, 455 g; IQR, 399–576 g; range, 250–1,070 g, exceeded normal range in 39/44 (normal reference: male, 270–360 g; female, 200–280 g)] (47).

While myocardial fibrosis was identified in only 72 cases, in a series where this was specifically reported, the prevalence was high (9, 10). Myocyte and ventricular wall hypertrophy were reported in 85 cases, again highly prevalent where specifically reported (18). Significant cardiac dilatation/cardiomegaly was described in 66 cases (10, 14, 15, 21, 24, 27, 30).

Overall changes consistent with cardiac ischemia and thrombosis were the most frequently reported acute findings. Acute myocardial ischemia was evident in 25 cases either in the form of acute myocardial infarction (MI) (n = 11) or microscopic evidence of acute or early ischemia (n = 14). Moreover, fibrin microvascular thrombi were identifiable in 27 cases (6, 8, 35, 36, 42, 47, 48). Thrombi in cardiac veins were described in three cases (8, 10). Lastly, there was eight cases with mural thrombi including the heart valves (n = 3) and the right atrium (RA) (n = 1) (10, 35, 47, 48).

Twelve studies explored the presence of SARS-CoV-2 within the myocardium using different techniques (Table 2) (8, 9, 17, 19, 23, 24, 29, 30, 39, 41, 44, 48). In those studies, SARS-CoV-2 was detected in 50 of 105 hearts (47%). However, clear myocarditis meeting the Dallas criteria was described in only five cases (6, 9, 17, 22). In an additional 35 cases, minimal lymphocytic (n = 33) or mononuclear infiltration (n = 2) not meeting the criteria for myocarditis was identified (13, 15, 27, 28). In three cases, authors attributed those changes as consistent with ischemic damage response (28). Overall, lymphocytic infiltration was scarce but can be detected in any of the pericardium, myocardium, epicardium, or endothelium. Lastly, pericardial affection was described in the form of pericardial effusion (n = 8) and pericarditis (n = 5, one had chronic pericarditis).

The cause of death was reported for 190 cases and, for the majority of these, was respiratory in origin (Supplementary Tables 1, 2). However, cardiac contribution to death was mentioned for 15 cases while pulmonary embolism (PE) was mentioned in eight cases.

The most alarming finding is the intracardiac, coronary arterial, and venous thrombosis, which may be in part explained by the COVID-19-associated coagulopathy (CAC). Myocardial ischemia can be further aggravated by the frequent pre-existing CAD and myocardial supply–demand mismatch.

By means of its receptor, SARS-CoV-2 can directly invade the endothelium leading to endothelial cell (EC) inflammation (i.e., endothelialitis), dysfunction, and death (49). Endothelial dysfunction can also result from an inappropriate immune and cytokine response. Endothelialitis, and hence EC dysfunction, subsequently induces a procoagulant state (CAC), loss of barrier function, inflammatory tissue infiltration, edema, and injury (49, 50). Cardiovascular comorbidities are usually associated with chronic EC dysfunction, which can explain the worse outcome when further acute insult is superadded.

However, endothelialitis was not a consistent finding in our reviewed studies but, when detected, was associated with microthrombi and had multiorgan distribution. Varga et al. showed multiorgan endotheliitis in all three studied cases (7). Ackermann et al. showed widespread endotheliitis and capillary thrombosis in COVID-19-affected lungs in a much more common prevalence than in non-COVID acute respiratory distress syndrome (ARDS) lungs (51). In contrast, Bradley et al. concluded not only no evidence of endothelialitis but also little evidence of cardiac microthrombi (9). Rapkiewics et al. noted no endothelial abnormalities but a platelet-rich microthrombi in all seven hearts examined, despite anticoagulation (8). It appears that alternate mechanisms of ischemia overlap, and while anticoagulation may be highly relevant in limiting pulmonary thrombosis, this may be less likely to significantly ameliorate any cardiac contribution to poor outcomes. Nicolai et al. highlighted thrombi to be rich in platelets, fibrin, and neutrophil extracellular traps (NETs), while Jensen et al. described platelet-rich cerebral microangiopathy (36, 46). The role of NET and platelets may be significant and could support other potential therapies (e.g., antiplatelet therapy).

Heart weight exceeded the normal range in 90% of cases reflecting a combination of chronic pathologies (e.g., hypertrophy), myocardial edema (marker of injury), and chambers dilatation. The observed cardiac dilatation (especially of the right heart) may be long standing or acute and hence relate to preload or afterload (pulmonary hypertension) changes occurring during the acute illness and its treatment.

ME reflects myocardial tissue response to most types of injury and hence its nonspecificity. Ischemia, septic cardiomyopathy, viral, or inflammatory infiltration can all contribute to it. Schmittinger et al. showed ME in 90% of PM septic hearts in a patchy distribution (median of 25% of tissue sections) (52). Of note, ME can reflect an early tissue change after insult (as early as 3 min in the setting of ischemia due to the disruption of the Na+/K+ pump) (53). Detecting ME has therapeutic implications, as it causes less energetic efficiency, arrhythmias, and reduced cardiac wall compliance. All of these are expected to impair systolic and diastolic function and can ultimately lead to fibrosis (53, 54). While cardiac MRI (CMR) can detect it in vivo, histological diagnosis remains technically challenging (53). This challenge, combined with the lack of standardized protocol guidelines for PM cardiac pathology reporting, may mean that ME was overlooked in many of the published reports.

Myocardial fibrosis was reported in nearly a quarter of cases. It is the end result of cardiac injury arising from different acute or chronic mechanisms. Cytokines were also implicated in cardiac fibroblast activation (55, 56).

The interpretation in COVID-19 is difficult and depends on many factors. It can reflect a chronic or a de novo subacute process. Aging and many reported comorbidities are strongly associated with fibrosis (56). Of note, amyloidosis (a pathology associated with fibrosis) was described in 11 cases and was significantly more prevalent when compared to a historical age-matched cohort (18, 37, 48).

Myocardial fibrosis can be divided into two types: interstitial fibrosis and replacement fibrosis, with considerable overlap between the two (55). While interstitial fibrosis is considered reactive and potentially reversible, replacement fibrosis is not (55). Interstitial fibrosis was previously detected in 100% of PM septic hearts but in a patchy nature (52). Such focal nature means that an extensive PM cardiac pathological examination is necessary. In fact, CMR may be superior as a diagnostic modality despite the difficultly to perform in unstable patients (55, 57).

Myocardial fibrosis represents the structural equivalent of heart failure. While ME is expected in the “reversible” septic cardiomyopathy, increased fibrous deposition (i.e., replacement fibrosis) would not be a likely finding in such reversible pathology (52, 57, 58).

Studies investigating the presence of SARS-CoV-2 within the myocardium were positive in about half the cases. In 1986, The Dallas criteria were proposed for the histopathological categorization and diagnosis of myocarditis based on endomyocardial biopsies. The “Dallas criteria” defines acute myocarditis as “an inflammatory infiltrate associated with myocyte necrosis or damage not characteristic of myocardial ischemia.” Borderline myocarditis requires a less intense inflammatory infiltrate with no light microscopic signs of myocyte destruction (59). In COVID-19 PM studies, inflammatory infiltrate (mainly lymphocytic) was observed in a minor proportion (about 10%) and was limited in extent for the majority of cases. As such, when interstitial edema and inflammatory infiltrate were observed, they did not meet the diagnostic criteria of myocarditis, except in five cases. In fact, some authors attributed such inflammatory infiltrate to an ischemic process (28). This suggests that contrary to early conjectures, acute and fulminant myocarditis are rare during the acute illness.

Correlating the histopathological data to the clinical, imaging, and investigational data can provide more insights into the likely mechanisms of cardiac involvement in COVID-19. Clinical presentation varies from ST elevation MI due to thrombotic occlusion of epicardial coronaries, to ischemia and/or infarction without obstructive coronary disease, through to tachy and brady arrhythmias, depressed left and right ventricular function, and occasional pericardial involvement (60). A review of published literature suggests that elevated Troponin and heart failure dominate the clinical presentations (61).

Echocardiography is readily performed in the acute setting but provides limited insights into the cause when compared to CMR. In a large multinational survey, Dweck et al. reported the echocardiographic findings in 1,216 studies performed over 17 days (62). Fifty-five percent of scans were abnormal. Impaired LV function or dilation (39%) followed by RV abnormalities (33%) dominated. These findings are non-specific, but clear wall motion abnormalities suggesting infarction were rare (3%). The RV abnormalities most likely relate to increased afterload given the high prevalence of pulmonary thromboembolism and extensive lung damage associated with COVID-19 infection (63). The LV abnormalities are non-specific but provide further evidence of the high prevalence of cardiac damage.

CMR-based studies have focused on patients post recovery (too late for confirmation of myocarditis) but have shown a high prevalence of abnormalities. The largest to date is a German study of relatively young patients (mean age, 49 years), largely managed at home (67 of 100), studied a median of 71 days post infection. Seventy-eight percent were reported to show abnormalities, including reduction in LV function, elevated T1 and T2 (the latter suggesting ME) and late gadolinium enhancement (LGE) (non-ischemic pattern in 20, ischemic in 12). Three patients with very elevated T2 were referred for endomyocardial biopsy and typical features of myocarditis reported. The T1 and T2 abnormalities suggest ongoing myocardial edema, and the LGE enhancement suggests fibrosis—both of which are common in the autopsy data (64).

The second CMR-based study included only patients in whom Troponin had been elevated during hospital admission. Fifty-one patients were studied 27 days post hospital discharge. In 22 patients, pulmonary embolism and/or coronary ischemia were identified before scanning as the most likely cause of troponin leakage. Among 29 patients (mean age, 64 years) with no clinically identified cause for myocardial injury, an ischemic pattern injury (LGE) was identified in 5, dual pathology (ischemic and non-ischemic) in 4, and non-ischemic in 11. Intriguingly, T1 and T2 were not abnormal in this study. This study thus also supports the histological finding of significant myocardial fibrosis but suggests that edema clears fairly quickly in those that recover (2). Again, Rajpal et al. performed CMR on 26 athletes with a history of mild COVID-19 infection. Four of them (15%) had criteria of myocarditis despite mild or no symptoms, and 30% showed signs of previous cardiac injury (65).

Merging the clinical, investigational, and autopsy data, we are presented with a picture that demonstrates a high prevalence of cardiac abnormalities, in part due to exacerbation of underlying cardiac pathology and partly coagulation disorders affecting the pulmonary and coronary vessels. Direct cardiac involvement mainly takes the form of non-coronary myocyte death, myocyte dysfunction, and interstitial fibrosis without substantial inflammatory infiltration or clear ischemia.

The role of direct viral cellular damage remains to be fully explored, and if this is the driving force, it is intriguing that the inflammatory response appears muted. However, it is possible that while the virus is rarely causing a fulminant or acute myocarditis, it can cause a persistent chronic myocardial inflammation with significant long-term implications. It is also important to note the reporting of a delayed immune response in the form of Kawasaki's disease in pediatric patients supporting the issue of long-term sequelae of the SARS-CoV-2 infection (66). Whether immunosuppressive treatment (e.g., dexamethasone and Tocilizumab) during the acute illness is of benefit or causes more harm to the heart should await randomized controlled studies including long-term follow-up.

Thus, on balance, the data strongly suggest significant viral replication in the myocardium without true acute myocarditis in most instances, with frequent non-MI pattern fibrosis—consistent with microvascular ischemia/thrombi and, in some cases, endothelial inflammation. Given the frequent presence of fibrosis associated with cell death, it is likely that complete recovery is unlikely—a clear distinction from septic cardiomyopathy. In addition, the exacerbation of underlying disease would appear to frequently unmask coronary disease, further increasing the benefit of careful cardiological follow-up.

As the vast majority of studied patients in this review died during the acute illness and cardiac abnormality was prevalent in the population studied, we can conclude that myocarditis was not a dominant cause of cardiac dysfunction identified premortem in COVID-19 patients, while the role of endothelialitis needs further clarification.

Our work delineates the importance of PM to guide the understanding of COVID-19. However, the small number of published PM cases in a disease, which has caused more than 1 million fatalities, highlights a hugely missed opportunity. Cardiac pathological changes are more likely to be focal in nature and hence easily missed if the heart is not examined in its entirety. Furthermore, the high prevalence of myocardial fibrosis, myocyte damage, or viral RNA in some studies but not others suggest a need to standardize histological reporting to establish common ground between pathologists and clinicians. There is also a genuine need for an international case register to gather the largest possible data in the shortest interval.

While our work is limited by the quality and small number of cases per study, we think it can contribute to a better understanding of COVID-19-associated cardiac injury. Other limits include the probable selection and reporting bias. PM is performed for patients who died during the acute illness and for certain subgroups of patients due to clinical or legal reasons. The longest duration of illness in our cohort is 52 days, which means that the long-term evolution or complications of the disease cannot be covered by this review.