[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Views 6,828
Citations 0
Comment & Response
August 25, 2020

Errors in Statistical Numbers and Data in Study of Cardiovascular Magnetic Resonance Imaging in Patients Recently Recovered From COVID-19

Author Affiliations
  • 1Institute for Experimental and Translational Cardiovascular Imaging, DZHK Centre for Cardiovascular Imaging, University Hospital Frankfurt, Frankfurt am Main, Germany
JAMA Cardiol. Published online August 25, 2020. doi:10.1001/jamacardio.2020.4661

To the Editor On behalf of our coauthors, we write to explain errors that were included in the Original Investigation, “Outcomes of Cardiovascular Magnetic Resonance Imaging in Patients Recently Recovered From Coronavirus Disease 2019 (COVID-19),”1 published in JAMA Cardiology on July 27, 2020. In this study of an unselected cohort of 100 German patients recently recovered from coronavirus disease 2019 (COVID-19) infection, cardiovascular magnetic resonance (CMR) imaging revealed cardiac involvement in 78 patients and ongoing myocardial inflammation in 60 patients, independent of preexisting conditions, severity, and overall course of the acute illness and time from the original diagnosis.

We were made aware of the errors in our original report as they were discussed on Twitter through a journalist, who was covering the publication of the article. We immediately studied the Twitter discussion, which made note of 2 problems: the use of inaccurate metrics for the data analysis as well as inconsistencies between the reported data in the legend of Figure 1 and the data points provided for the patients with COVID-19 in Figure 2. As a result, we have reviewed the data and repeated the analysis. Herein, we summarize the errors and corrections needed.

First, some of the metrics were not reported correctly (eg, mean vs median) for the values in Table 1. We have recalculated all data according to data type. Now, we correctly report means (SDs) or medians (interquartile ranges). We have made sure to use the correct statistical test for the type of the data and explained this in the Methods section. In the corrected Table 1, the following were associated with COVID-19 diagnosis vs the healthy controls: body mass index, hypertension, diabetes, hypercholesterolemia, known coronary artery disease, and chronic obstructive pulmonary disease or asthma. There were no differences between those with COVID-19 and the risk factor–matched patients.

In addition, in Figure 1, the data for the time between diagnosis and the CMR image (78 days vs 67 days) and troponin level (17.8 pg/mL vs 16.7 pg/mL) were incorrectly reported, leading to an inconsistency with Figure 3, which did not change. Also, the legend for Figure 1 did not indicate that the representative images are from 2 different patients in the cohort.

We discovered a few other errors. The initial calculations were based on an early version of the data set that included only 54 instead of 57 patients in the risk factor–matched cohort. We have recalculated all percentages against the correct final group of 57 risk factor–matched controls and updated Figure 2 accordingly. Also, some counts in the risk factor–matched control groups were erroneously based on clinical assignments rather than standard deviation–based cutoffs. We have recounted all data using the research cutoff values. Consequently, the counts in the risk factor–matched group are now different than originally presented.

During the correction and recalculation process, we were able to provide some missing data from the original CMR scans as well as correct some data entry errors. Consequently, some minor changes are needed in the area under the receiver operating characteristic curve (AUC) values in Table 2. However, only the comparison for left ventricular mass index between those with COVID-19 and healthy controls changed from a significant to a nonsignificant association.

In addition, there may have been a lack of clarity on how the counts for T1 and T2 were obtained and which cutoff values were used. We clarify in the Methods section that the cutoff values for abnormal native T1 and T2 values were based on previously derived sequence-specific standard deviations above the respective means in a healthy population and not based on the current healthy control group, as this was a selected sample to match for age and sex. To avoid confusion with cutoff values determined from the current data set, we removed these AUC-based cutoff values from Table 2.

We are pleased to confirm that reanalysis of the data has not led to a change in the main conclusions of the study. As we originally reported, compared with healthy controls and risk factor–matched controls, patients recently recovered from COVID-19 had lower left ventricular ejection fraction, higher left ventricle volume, and elevated values of T1 and T2. However, the corrected findings no longer show higher left ventricular mass in these patients. We confirm that there are no other errors. The errors and corrections affect the Abstract, Methods and Results sections, Tables, and Figures, and the article has been corrected online.2 We most sincerely apologize to the readers and editors of JAMA Cardiology for any confusion, and we appreciate the opportunity to correct the original publication.

Back to top
Article Information

Corresponding Author: Eike Nagel, MD, Institute for Experimental and Translational Cardiovascular Imaging, DZHK Centre for Cardiovascular Imaging, University Hospital Frankfurt, Theodor-Stern-Kai 7, Frankfurt am Main 60590, Germany (eike.nagel@cardiac-imaging.org).

Published Online: August 25, 2020. doi:10.1001/jamacardio.2020.4661

Conflict of Interest Disclosures: Dr Nagel has received grants from Bayer, the German Ministry for Education and Research, Deutsche Herzstiftung e.V., Neosoft Technologies, and Cardio-Pulmonary Institute and personal fees from Bayer. No other disclosures were reported.

References
1.
Puntmann  VO, Carerj  ML, Wieters  I,  et al.  Outcomes of cardiovascular magnetic resonance imaging in patients recently recovered from coronavirus disease 2019 (COVID-19).   JAMA Cardiol. Published online July 27, 2020. doi:10.1001/jamacardio.2020.3557PubMedGoogle Scholar
2.
 Errors in statistical numbers and data.  Correction.  JAMA Cardiol. Published online August 25, 2020. doi:10.1001/jamacardio.2020.4648Google Scholar
×