Association Between Myocarditis and Mortality in COVID-19 Patients in a Large Registry

Abstract

Objective

To present a large registry data assessing the association between myocarditis and mortality in patients with severe acute respiratory syndrome coronavirus 2 infection.

Patients and Methods

The researchers identified adult patients aged 18 to 90 years of age with coronavirus disease 2019 (COVID-19) diagnosis in the TriNetX (COVID-19 research network) database between January 20, 2020, and December 9, 2020. These patients were then divided into groups of those who had a positive myocarditis diagnosis and those who did not. The researchers compared all-cause mortality between propensity-matched pairs of patients in both groups.

Results

A total of 259,352 patients with COVID-19 diagnosis were included in the study. Of those patients, 383 (0.2%) had myocarditis diagnosis, whereas 258,969 (99.8%) did not have myocarditis diagnosis during their hospital stay. Patients were predominantly male in the myocarditis group (59.0% vs 45.0%, P<0.001). As to the propensity-matched cohorts, 383 of 383 were matched, and the all-cause mortality was 13.4 % vs 4.2% (P<0.001) at 30 days. A Kaplan-Meier survival analysis was also statistically significant (P<0.001) at 30 days.

Conclusion

In a large multinational database of COVID-19 patients, we observed an association between myocarditis diagnosis and increased mortality. Further prospective studies are recommended to further assess myocarditis outcomes in COVID-19 patients and treatment options.

Coronavirus disease 2019 (COVID-19) continues to be a challenging global pandemic. The total number of COVID-19 cases is approaching 100 million, and 25 million of those cases are in the United States.¹ Recent studies have indicated that this virus often affects the cardiovascular system.2, 3, 4 Myocarditis has been identified as a consequence of the effect of COVID-19 on the cardiovascular system and can even be observed in asymptomatic patients.^2,3 Myocarditis is often underdiagnosed with COVID-19 diagnosis due to lack of diagnostic modalities. Using cardiac magnetic resonance imaging, the incidence of cardiac involvement with COVID-19 diagnosis is variable in the literature ranging from 1.4% in mild to moderate symptomatic athletes to 78.0% in other study by Puntmann et al.^3,5 Several studies have investigated the impact of cardiovascular involvement as a prognostic factor for the morbidity and mortality of COVID-19.6, 7, 8 Based on data from an international registry database, this paper presents the impact of myocarditis as a prognostic factor for the mortality and hospital stays of confirmed COVID-19–positive patients.

Patients and Methods

We used the TriNetX platform to assess COVID-19 patients with a lab-confirmed severe acute respiratory syndrome coronavirus 2 diagnosis and a confirmed myocarditis diagnosis. Additional sensitivity analyses used a negative control of unrelated bleeding with no known relationship. Propensity score matching (PSM) was also used to control for literature-driven covariates.

Data Source

TriNetX Inc (Cambridge, MA) is a global federal research network that combines the real-time data from electronic medical records. The platform interfaces with multiple health care organizations and offers data through a user-friendly platform. The diagnosis of myocarditis in our study population was based on International Classification of Diseases, 10^th Revision (ICD-10) code which was a clinical diagnosis, and not all patients who had a confirmation with cardiac magnetic resonance (CMR) imaging (47.2%) or endomyocardial biopsy (EMB) (3.4%) had a confirmed test associated with the case.

Study Sample

The researchers queried the COVID-19 research network within the TriNetX platform, a collection of 56 health care organizations, from January 20, 2020, to December 9, 2020. The researchers first identified adult patients aged 18 to 90 years of age and then used the COVID-19 diagnosis codes (see Supplement 1, available online at https://mcpiqojournal.org) and myocarditis diagnosis codes.

Exposure

The target exposure is confirmed myocarditis cases among patients with a lab-confirmed diagnosis of COVID-19, excluding those with histories of myocarditis. Myocarditis was clinical diagnosed within these cases.

Statistical Analyses

Descriptive statistics were presented as frequencies with percentages for categorical variables as mean ± standard deviations for continuous measures. Baseline characteristics were compared using Pearson’s χ²test for categorical variables. To account and match for potential differences, TriNetX developed a 1:1 PSM using logistic regression to create two well-matched groups for a comparative analysis (Table 1).

Table 1.

Baseline Characteristics (PSM Match)^a,b

Baseline characteristic	Unmatched cohorts			Standardized mean difference	Matched cohorts			Standardized mean difference
	Myocarditis (n=383)	No - myocarditis (n=258,969)	P		Myocarditis (n=383)	No - myocarditis (n=383)	P
Age at index, y	47.5±21.4	47.9± 18.6	0.67	0.02	47.5±21.4	47.8±20.5	0.85	0.01
Male	59.0	45.0	<0.01	0.28	59.0	55.9	0.42	0.06
White	52.2	53.5	0.63	0.02	52.2	52.5	0.94	0.01
Female	41.0	55.0	<0.01	0.28	41.0	44.1	0.42	0.06
Black or African American	19.0	15.2	0.07	0.09	19.0	22.2	0.21	0.09
Hispanic or Latino	14.1	13.3	0.66	0.02	14.1	13.0	0.52	0.05
D-dimer	13.0	4.3	<0.01	0.31	13.0	17.0	0.13	0.11
TCF	3.0	0.1	<0.01	0.22	3.0	3.0	1.00	0.00
AC	48.0	35.4	<0.01	0.25	48.0	49.4	0.62	0.04
Azithromycin	22.5	17.0	0.02	0.15	22.5	21.2	0.66	0.03
Dexamethasone	22.2	16.0	0.04	0.17	22.2	26.4	0.17	0.10
Methylprednisolone	22.0	15.0	0.01	0.18	22.0	25.0	0.30	0.07
Hydrocortisone	11.0	8.0	0.03	0.10	11.0	12.3	0.50	0.05
Hydroxychloroquine	7.0	2.0	<0.01	0.26	7.0	8.1	0.41	0.06
Hypertension	37.0	27.0	<0.01	0.21	37.0	36.0	0.82	0.02
Diabetes mellitus	24.0	14.1	<0.01	0.25	24.0	26.0	0.50	0.05
CHF	22.2	5.4	<0.01	0.50	22.2	21.0	0.60	0.04
CAD	14.0	6.4	<0.01	0.25	14.0	16.0	0.48	0.05
Smoking history	14.0	8.0	<0.01	0.19	14.0	16.0	0.41	0.06
Cerebrovascular diseases	9.1	5.3	0.08	0.15	9.1	7.1	0.29	0.08
COPD	8.0	4.5	0.02	0.14	8.0	8.1	0.89	0.01
Old myocardial infarction	8.0	3.3	<0.01	0.25	7.6	8.1	0.79	0.02
Alcohol dependence, uncomplicated	3.0	1.00	<0.01	0.14	3.0	2.9	0.82	0.02
BMIc	34.5	31.1	0.63	0.04	34.5	39.0	0.32	0.12

^aAC = adrenal corticosteroid; BMI = body mass index; CAD = coronary artery disease; CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease; PSM = propensity score matching; TCF = transfusion of convalescent plasma.

^bValues shown are percentages.

^cObesity was defined as a BMI ≥ 30 kg/m².

The platform uses a logistic regression to obtain listed PSM scores for each of the selected covariates, as well as the Python libraries (NumPy and sklearn). The PSM platform also runs the final results in R to compare and verify them. The final step of verification uses a nearest neighbor function set to a tolerance level of 0.01 and a difference of value greater than 0.1. Mortality for the PSM cohorts was determined using the Kaplan-Meier method and the statistical significance of the difference risk factors and differing measures of association for 30-day all-cause mortality. We also examined the time interval at 6 months for all-cause mortality between the cohorts.

Sensitivity Analyses

Establishing whether different health conditions drive mortality might allow researchers to understand and test potential confounders. The researchers thus performed two sensitivity analyses of the primary endpoint of all-cause mortality to test these differing combinations. The researchers first created two cohorts that excluded all the patients except those with coronary artery disease (CAD), hypertension, or diabetes and myocarditis and compared these patients to those without myocarditis, creating two propensity-matched cohorts. Finally, given the possibility of residual confounders, the researchers used the falsification endpoint of bleeding that would not likely be affected by myocarditis.

Results

Overall, 259,352 patients were included in the study. Of those patients, 383 (0.2%) had a positive myocarditis diagnosis following a COVID-19 diagnosis; 93% were in the United States, and 7% were outside of the United States. Patients were predominantly male in the myocarditis group (59.0% vs 45.0%, P<0.001) and had a higher prevalence of key comorbidities, including atherosclerotic heart disease, hypertensive disease, diabetes mellitus, heart failure, chronic obstructive pulmonary disease, cerebrovascular disease, past myocardial infarction, a history of smoking, and a history of alcohol use, than those in the latter group. This distribution of differing used steroids was much higher in the myocarditis group compared to the nonmyocarditis group with use of dexamethasone (22.2% vs 16.0%, P=.04), methylprednisolone (22.0% vs 15.0%, P=.01), and hydrocortisone (11.0% vs 8.0%, P=.03).

Before PSM, the all-cause mortality for the myocarditis group was 13.4% (n=51 of 383), whereas the all-cause mortality for the control COVID-19 group was 2.4% (n=6150 of 258,969) (Figure 1). The post-PSM all-cause mortality was 383 of 383 (13.4% vs 4.2%, P≤.001). A log-rank test also supported the findings (82.0% vs 93.8%, P<.001) at 30 days. We also examined the time interval at 6 months for all-cause mortality between the cohorts. The results suggest a similar all-cause mortality (15.1% vs 5.2%, P<.001) at 6 months. A log-rank test also supported the findings (75.7% vs 90.3%, P<.001) (Figure 2).

Figure 1.

Kaplan Meier curves showing all-cause mortality for myocarditis (unmatched) for (A) 30 days and (B) 6 months.

Figure 2.

Kaplan Meier curves showing all-cause mortality for myocarditis) (matched) for (A) 30 days and (B) 6 months.

In a subset of PSM cases that excluded all patients except those with CAD, hypertension, or diabetes, the researchers created two similar cohorts and included 71,463 patients with a matched PSM of in 201 of 201 and an all-cause mortality of 19.4% vs. 14.0% (P=.141). A log-rank test showed 77.2%. vs 83.0% (P=.26) at 30 days (Figure 3). The subset suggests that CAD, hypertension, or diabetes comorbidities influence all-cause mortality between the two groups less than myocarditis. Finally, no difference in the falsification of the total cohort was observed (0% vs 0%), suggesting the absence of a significant unmeasured confounder that would affect the primary outcome.

Figure 3.

Kaplan Meier curves showing all-cause mortality for myocarditis and coronary artery disease (CAD), hypertension, and diabetes/sensitivity analysis.

Discussion

Whereas cardiac involvement in COVID-19 diagnosis is well known and proven, the pathogenesis of myocardial involvement is not well understood. Theories have been suggested that include direct viral insult to cardiomyocytes and the human’s immune response to virally infected myocardium.⁹ Lack of guidelines that suggest routine screening for myocarditis and lack of well-known treatment led to underdiagnoses of cardiac involvement. Researchers, such as Halushka et al,¹⁰ have confirmed that myocarditis occurred in 20% of cases of 277 autopsied hearts in 22 separate publications. The disease ecology of this condition is believed to be more common.¹⁰

Our understanding of the COVID-19 pandemic has been dynamic and constantly evolving. Many poor prognostic factors have been identified with increased age on top of the list, and cardiac injury was one of the poor prognostic factors.^11,12 Traditional risk scores underestimated the risk of COVID-19 outcomes and the need for COVID-19–specific prognostic scores was emphasized.¹³ New prognostic scores have been developed but did not include cardiac injury as a prognostic factor.^14,15 In addition to prior reports, we emphasize in our study the poor prognostication of cardiac involvement in COVID-19 patients.6, 7, 8^,16 Our study population was diverse, from different ethnicities and countries.

Whereas all-cause mortality was statistically significantly more in the myocarditis group, this group also had more comorbidities. The association between myocarditis and mortality was evident after matching the groups with all known major comorbidities. Although these results were expected, it is essential to reproduce them in prospective trials to fully understand the association and suggest certain treatments. Certain treatment for myocarditis has been debatable because of a lack of large, randomized control trials. Sawalha et al¹⁷ reported favorable outcomes in myocarditis cases with COVID-19 treated with steroids. In our matched cohort, the use of dexamethasone and hydroxychloroquine were the same in both groups (22% and 7%, respectively).

Study Limitations

Our study has some limitations. The diagnosis of myocarditis in our study population was based on an ICD-10 code which was a clinical diagnosis, and not all patients who had a confirmation with CMR (47.2%) or EMB (3.4%) also had a confirmed test associated with the case. The lack of advanced technology in many hospitals, the efforts to limit the spread of infection, and the lack of clear guidelines limited the number of patients who had CMR or EMB. We also followed the general practice of applying 2013 European Society of Cardiology guidelines in cases of clinically diagnosed myocarditis.¹⁸ Our study was a retrospective chart review, which has the inherited limitations of selection bias and the inability to assess incidence. Finally, our study included patients from January 2020 to December 2020, and much of our understanding of COVID-19 infection and treatment modality has changed in this time frame which may have affected the outcomes. These data show association between myocarditis and mortality in COVID-19 diagnosis.¹⁹ Large prospective studies are recommended to define the association more and suggest treatment options. Other potential avenues of identifying myocarditis can be found in the identification of differing biomarkers and should be further explored in follow-up research studies.^20,21

Conclusion

We present a large registry data that shows association between myocarditis and mortality in COVID-19 diagnoses. Large prospective studies are recommended to define the association more and suggest treatment options.