Cardiovascular disease as a risk factor of disease progression in COVID-19: The corrected effect size and forest plot

Dear Editor,

While the world is in lockdown for months since December of 2019 due to novel coronavirus disease (COVID-19) pandemic, a lot of research is in progress to find out various risk factors associated with COVID-19 progression and related mortalities.¹

We read with great interest, the recent and very informative article by Zheng et al.,² who performed a meta-analysis to identify various risk factors such as; demographical (male, age, current smoking), comorbidities (diabetes, hypertension, malignancy, respiratory disease and cardiovascular disease), and other laboratory variables for the progression of COVID-19.

Firstly, we have a concern related to the result on the presence of cardiovascular disease (CVD) in association with COVID-19. The reported result for this pooled-outcome based on ten included studies has been shown to be statistically significant with a higher proportion of CVD in critical/mortal group compared to the non-critical group of COVID-19 patients. The pooled effect size for this association has been reported as odds ratio (OR) with its 95% confidence interval (CI) levels (OR=5.19, 95% CI=3.25–8.29, P<0.00001). However, upon proper examination of the reported forest plot and the included studies, we observed that the input data for one included study by Shi Y et al.,³ was wrong and therefore the outcome ‘effect size’ for this study has been shown to be ‘Not Estimable’ (Fig. 3. of Zheng et al.).²

It is very much important for proper data inputs and thorough checks for its correctness by multiple authors while performing ameta-analysis. In the reported meta-analysis, the CVD events and total number under critical/mortal group from a study by Shi Y et al.,³ have been recorded for meta-analysis as 49 and 4, respectively. It is for this reason, the effect-size has been found to be ‘Not Estimable’ in the respective forest plot (Fig. 3. of Zheng et al.).² We, upon a thorough review of the study by Shi Y et al., noticed that the actual values are 4 and 49 for CVD events and total number, respectively.³

Secondly, by further inspecting the forest plot, in addition to inappropriate data inputs, the study weights have been noticed to be very disproportionate to each other (minimum 1.7 & maximum 28.7).² It is a known myth to choose either fixed or random effects model for a meta-analysis based on the heterogeneity statistics, particularly fixed effects model is not a viable method when objective is to measure the effect size of group level variables.⁴

Therefore, in this letter we updated the forest plot and the effect-size characteristics for the relationship between CVD and COVID-19 progression (including the missing data for a study & random effects model which considers both within and between study variability). According to the random effects model used (Fig. 1 ), the study weights were found to better distributed than in the fixed effects model (minimum 3.1 & maximum 18.0), and the results showed a significantly higher proportion of CVD in critical/mortal group compared to the non-critical group of COVID-19 patients (OR=4.78, 95% CI=2.71–8.42, P<0.00001). Considering the limitation that the COVID-19 patients with underlying CVD may also have other comorbidities, the use of random effects model would provide a better statistic and the obtained result should be interpreted with a caution to the limitation of other overlapping comorbidities.

Fig. 1.

Forest plot for cardiovascular disease comorbidity between critical/mortal and non-critical COVID-19 patients.