We designed a cohort study of consecutive COVID-19 patients diagnosed by reverse-transcription polymerase chain reaction at Hospital del Mar, Barcelona, Spain, from February 27 to April 7, 2020. Patients were followed up for clinical outcomes until May 1, 2020.

Demographic characteristics (age and sex), comorbidities, laboratory determinations (including cardiac biomarkers), chest radiography, electrocardiographic findings, treatments, complications and outcomes were collected into an electronic data capture system.

A confirmed case of COVID-19 was defined by a positive result on a reverse-transcription polymerase chain reaction assay of a specimen collected from a nasopharyngeal swab. Only laboratory-confirmed cases were included. Reverse-transcription polymerase chain reaction assays were performed in accordance with World Health Organization interim guidance.9

Laboratory tests included blood count, coagulation tests, liver and renal function analyses, electrolytes, C-reactive protein, procalcitonin, lactate dehydrogenase, creatine kinase, D-dimer and the 2 tested cardiac markers, hs-cTnT and NT-proBNP, which were measured within 48hours of admission. An electrochemiluminescence-based immunoanalytical system, Elecsys 2010 (Roche Diagnostics Ltd, Mannheim, Germany), was used to determine plasma levels of NT-proBNP and hs-cTnT.

According to the Fourth Universal Definition of Myocardial Infarction (2018), cardiac injury was diagnosed if serum levels of cardiac biomarkers (eg, hs-cTnT) were above the 99th percentile upper reference limit, (> 14.0 ng/L), as recommended by the manufacturer10,11 and regardless of new abnormalities on electrocardiography and echocardiography. For NT-proBNP, positivity was considered if serum levels were above the limit for ruling out heart failure in the acute setting, which is <300 pg/mL.12 Thus, the cutoff was set at ≥ 300 pg/mL.

Acute respiratory distress syndrome was diagnosed according to the Berlin criteria as acute-onset hypoxemia (ratio of arterial oxygen partial pressure to fractional inspired oxygen expressed as a fraction <300) associated with bilateral pulmonary opacities on chest imaging that were not fully explained by congestive heart failure or other forms of volume overload.13

The primary endpoint was the composite of death or the need for MV at 30 days after COVID-19 diagnosis, as also used in previous studies to assess the severity of COVID-19 infectious disease.5 In addition, we analyzed the capacity of the biomarkers to predict mortality alone.

Categorical variables are summarized as counts and percentages, and continuous variables as the number of nonmissing observations, the mean and standard deviation (SD), or the median and interquartile range [IQR], depending on the distribution of the variable. Normality of distributions was tested by normal Q-Q plots. Patient characteristics were compared between hs-cTnT (cutoff point> 14 ng/L), NT-proBNP (cutoff point> 300 pg/L) and outcome status categories (composite endpoint including death or MV) by the Student t test or Mann-Whitney U test for continuous variables, and by the Pearson chi-squared test for categorical variables.

Kaplan-Meier survival curves for death or the composite endpoint were plotted, and the log-rank test was computed to assess differences between groups of hs-cTNT and NT-pro-BNP.

The adjusted hazard ratio of death and the composite endpoint for hs-cTnT and NT-proBNP status was analyzed using Cox proportional hazard models. The models were adjusted for potential confounders selected by stepwise backward elimination, among patient characteristics that were significantly (P <.10) associated with an hs-cTnT- or NT-proBNP-positive status as well as with the composite endpoint. Age was excluded from the list of potential confounders because it was among the criteria used to give patients access to an intensive care unit and MV, which is part of the composite endpoint outcomes.

The assumption of proportionality of hazards from the Cox models was checked. The hazard ratio for laboratory determinations was calculated by 10 or 100 measurement units change. The diagnostic test accuracy for hs-cTnT and NT-proBNP and mortality and death or mechanical ventilation is illustrated in table 1 of the supplementary data. Kaplan-Meier and Cox models took into account the delay between symptom onset and admission by the left-truncation approach. The c-statistic was calculated to analyze the discriminatory ability of the adjusted models. Hosmer-Lemeshow test taking into account right-censoring was computed to assess model calibration. Continuous, categorical (into 3 risk groups defined by tertiles) and clinical net reclassification indexes were computed to assess whether the inclusion of NT-proBNP in a model with hs-cTnT and confounders improved the classification of individual outcomes. P values <.05 were considered statistically significant. All tests were performed with R (3.5.3) (R: A language and environment for statistical computing; R Foundation for Statistical Computing, Vienna, Austria).14

This study was performed in accordance with the provisions of the Declaration of Helsinki, ISO 14155, and clinical practice guidelines. The study protocol was approved by the Institutional Ethics Committee and the research commission of our hospital. The need for written informed consent was waived in light of the urgent need to collect data and the infectious disease hazard.

The baseline characteristics of the population are summarized in table 1 by admission levels of hs-TnT and NT-proBNP. Patients with cardiac injury as defined by hs-TnT levels, were significantly older, with higher prevalence rates of cardiovascular risk factors and a previous history of cardiovascular disease. Patients with cardiac injury as defined by baseline NT-proBNP levels had a similar clinical profile to the corresponding hs-cTnT-defined population.

Laboratory findings are summarized in table 1 by admission levels of hs-TnT and NT-proBNP. Patients with positive hs-cTnT and NT-proBNP levels showed more severe inflammatory responses.

Management details and hospital outcomes are summarized in table 2 by admission levels of hs-TnT and NT-proBNP. Almost two thirds (62%) of patients required some degree of oxygen therapy, with 55 patients (7%) requiring MV. Patients with cardiac injury received more hydroxychloroquine, azithromycin, systemic glucocorticoids, and low-molecular-weight heparin. In addition, they required more support from either a high-flow nasal cannula or invasive mechanical ventilation, as well as vasopressors. Median length of hospital stay was 9 [IQR 3–19] days. One hundred and four patients (12%) died during admission (figure 1).

Figure 1.

Flowchart of patient selection. hs-TnT, high-sensitivity troponin T; NT-proBNP, N-terminal pro-B-type natriuretic peptide.

(0.18MB).

Patients with cardiac injury had longer hospital stays. The rates of mortality and of the composite endpoint including mortality or the need for MV were significantly higher among patients with vs without cardiac injury, as shown in table 2 and the Kaplan-Meier survival curves in figure 2.

Figure 2.

Kaplan-Meier 50-day survival curves for mortality during the time from symptom onset by 4 combinations of high-sensitivity troponin T (hs-TnT) and N-terminal pro-B-type natriuretic peptide (NT-proBNP) levels (A), and for the composite of mortality or mechanical ventilation by 4 combinations of hs-cTnT and NT-proBNP levels (B).

(0.18MB).

As illustrated in table 2 of the supplementary data, hs-cTnT levels were significantly associated with COVID-19 severity. Patients with mild symptoms and those who were discharged to the hospital-at-home program had no cardiac injury. One-quarter of patients needing hospitalization but no respiratory support showed cardiac injury. Patients with respiratory support but no mechanical ventilation constituted one-third of cases and nearly half of patients requiring MV had cardiac injury. More than 80% of patients who died had positive levels of hs-cTnT at admission. We found similar results for the NT-proBNP values.

NT-proBNP showed a fair correlation with hs-cTnT when considered as continuous data (Spearman's R=0.64, P-value <.001), as shown in figure 1 of the supplementary data.

The Cox proportional hazard regression models (table 3) showed that serum levels of hs-cTnT> 14 ng/L and NT-proBNP> 300 pg/mL were significant independent predictors of mortality and of the composite of mortality or need for MV. Other factors contributed independently to mortality, namely, history of coronary heart disease, creatine phosphokinase levels, C-reactive protein and D-dimer. Glycemia and C-reactive protein were significantly associated with the composite endpoint, in addition to hs-TnT and NT-proBNP.

In the Cox proportional hazard regression models presented in table 3, the inclusion of NT-proBNP (cutoff value, 300 pg/mL) significantly improved the discrimination (C-index) and continuous net reclassification indexes for both mortality alone and the composite endpoint (table 3).

Our study has some limitations. First, asymptomatic patients were not included in this registry, which confers a selection bias. However, we focused on patients who were diagnosed in hospital, some of whom were discharged to home for care there. Second, because this is an observational study, no causal inference for the association between cardiac injury and severity of COVID-19 infection can be drawn. Clinical trials to demonstrate whether this association is useful in guiding treatments are needed to further comprehend the significance of our findings. Third, this is a single-centre study with a limited number of patients relative to the magnitude of this pandemic. Larger studies in upcoming months should add to our results.