ABC | Volume 115, Nº2, August 2020

Brief Communication Martins-Filho et al. Cardiac biomarkers in patients with COVID-19 Arq Bras Cardiol. 2020; 115(2):273-277 hemodynamic instability and/or ventricular fibrillation, and clinically significant bradycardia on electrocardiography. We excluded publications with potentially overlapping reports based on data collection and setting and studies from which data extraction was not possible. In the event of potentially overlapping data, we selected the study with the most complete information. Reports were screened in two stages, screening of titles and abstracts followed by the retrieval and screening of full-text articles. Searches were performed from January 1, 2020 to April 14, 2020, without language restrictions. The reference lists of all eligible studies and reviews were also evaluated to identify additional studies for inclusion. The following search terms were used: “COVID-19”, “SARS-CoV-2”, and “coronavirus”. All COVID-19 reports, irrespective of cardiovascular topic, were reviewed. Data from publications were extracted by two authors and crosschecked for accuracy. Our outcome of interest was in-hospital death. Clinical data (age, sex, and existing comorbidities), myocardial injury biomarkers (hs-cTnI, NT- proBNP, and creatine kinase-MB [CK-MB]), and cardiac complications (acute cardiac injury, heart failure, and arrhythmias) were considered independent variables. The Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies of the National Institutes of Health (https://www.nhlbi.nih.gov/health-topics/study-quality- assessment-tools) was used to grade the quality of each study. This tool is composed of 14 items that evaluate the representativeness and selection of the sample, description and measurement of exposure, follow-up of participants, and treatment of confounding factors. The findings were discussed qualitatively. Disagreements were resolved by discussion. Effect sizes were reported as mean difference (MD) or standardized mean difference (SMD) for continuous variables and risk ratio (RR) for dichotomous variables with 95% confidence intervals (CI). To calculate MD and SMD, means and standard deviations (SD) of myocardial injury biomarkers were obtained for each study. If the means and SD were not directly reported in the publication, indirect methods of extracting estimates were used. 5,6 When data were not presented in tables or in the text and the authors could not be reached, data were extracted using WebPlotDigitizer graph digitization software (available at http://arohatgi. info/WebPlotDigitizer). Not all studies reported data on all predictor variables, and the pooled analysis was estimated from the data available for each variable. A random effects model was used to pool the results, and 2-tailed p < 0.05 was used to determine significance. Cohen’s classification was used to interpret magnitude of the effect size for myocardial injury biomarkers. SMD > 0.8 was considered a large effect size. Statistical heterogeneity was quantified by the I 2 index, and potential for publication bias was analyzed for hs-cTnI using Egger’s regression test and visual inspection of funnel plots. Because of the small number of studies reporting data for NT-proBNP and CK-MB, analysis of publication bias was not performed. Analyses were conducted using Review Manager 5.3 (Cochrane IMS, Copenhagen, Denmark). Results After screening 8,091 titles and abstracts, 31 full-text articles were assessed for eligibility and 25 studies were excluded, seven of which were due to potentially overlapping data. Six retrospective studies 1,7-11 were included, providing data from 1,141 patients (633 male and 508 female) with confirmed SARS-CoV-2 infection, 832 survivors and 309 non-survivors. Details of included studies are shown in Table 1. The risk of bias of the studies is showed in e-Table 1 in the supplemental digital content. All studies had clear objectives and eligibility criteria, recruited subjects from the same population, and described the definitions of exposure factors and outcomes. However, studies have not been able to determine whether the sample size was representative for Table 1 - Characteristics of included studies and clinical data of patients with COVID-19, including in-hospital deaths Authors Design Setting Data collection Sample size Age* Sex In-hospital deaths Survivors Male Female Zhou et al, 2020 1 Retrospective cohort Jinyintan Hospital and Wuhan Pulmonary Hospital Dec 29, 2019 to Jan 31, 2020 191 56.3 (15.6) 119 72 54 137 Cao et al, 2020 7 Retrospective cohort Zhongnan Hospital of Wuhan University Jan 3, 2020 to Feb 1, 2020 102 52.7 (22.2) 53 49 17 85 Chen et al, 2020 8 Retrospective cohort Tongji Medical College of Wuhan Jan 13, 2020 to Feb 28, 2020 274 58.7 (19.3) 171 103 113 161 Guo et al, 2020 9 Retrospective cohort Seventh Hospital of Wuhan Jan 23, 2020 to Feb 23, 2020 187 58.5 (14.7) 91 96 43 144 Wang et al, 2020 10 Retrospective cohort Renmin Hospital of Wuhan University Jan 1, 2020 to Feb 6, 2020 339 70.0 (8.2) 166 173 65 274 Zhang et al, 2020 11 Retrospective cohort Wuhan No.1 Hospital Dec 25, 2019 to Feb 15, 2020 48 70.6 (13.4) 33 15 17 31 *Data reported in mean and standard deviation. 274