Is there an exposure–effect relationship between body mass index and invasive mechanical ventilation, severity, and death in patients with COVID‐19? Evidence from an updated meta‐analysis

We read with great interest the article by Fööldi et al. about the effect of obesity on COVID‐19. The authors evaluated the risk of different levels of obesity on COVID‐19 and found obesity was a significant risk factor for invasive mechanical ventilation (IMV) and intensive care unit (ICU) admission for patients with COVID‐19. This is a timely and exhaustive study. However, we have several concerns. First, by combing limited studies (N = 6 for IMV and N = 5 for ICU) in the forest plot, the results were not stable in the sensitive analysis by omitting one study each time. Moreover, some studies had greater than a 40%‐70% weighting for the primary outcomes, which suggested a large selection of studies. For example, in the results of ICU admission, the weight of Lighter et al. ¹ was 74%, and the results were not significant (OR: 1.07, 95% confidence interval (CI): 0.74–1.56) when the Lighter et al. ¹ was excluded. Also, the phenomenon was observed in the results of IMV when Simonnet et al. ² was excluded in the sensitively analysis. The above mentioned reduced the robustness of present results. Second, an association was observed with obesity when body mass index (BMI) was considered as a categorical variable. However, whether there is an exposure–effect association between BMI and COVID‐19 was not clear. Based on available data presented in Földi's meta‐analysis, we further systematically searched the databases (Pubmed, medRxiv, Embase) and added nine studies ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ and have evaluated pooled exposure–effect odd ratios (ORs) for the impact of increasing BMI on COVID‐19. Exposure–effect risk was calculated by the robust error meta‐regression method. ¹¹ As shown in Figure 1, BMI was associated with IMV, severe COVID‐19, and hospitalized death in a non‐linear manner (P _nonlinearity = .018 for IMV, P _nonlinearity = .031 for severe COVID‐19, P _nonlinearity = .0067 for hospitalized death). The summary ORs for a 5‐unit increment in BMI was 1.43 (95% CI: 1.17–1.76) for IMV, 1.22 (95% CI: 1.13–1.31) for severe COVID‐19, and 1.06 (95% CI: 1.03–1.09) for hospitalized death.

FIGURE 1.

Exposure–effect impact of body mass index (BMI) on invasive mechanical ventilation (IMV), severity, and hospitalized death in patients with COVID‐19. Solid line in red represents fitted linear trend and corresponding upper/lower 95% confidence interval (CI); lines with long dashes in black represent fitted non‐linear trend and corresponding upper/lower 95% CI. (A) Exposure–effect association between BMI and IMV. P _nonlinearity = .018, OR = 1.43 (95% CI: 1.17–1.76) per 5 kg/m² of BMI increment; (B) exposure–effect association between BMI and severe COVID‐19. P _nonlinearity = 0.031, OR = 1.22 (95% CI: 1.13–1.31) per 5 kg/m² of BMI increment; (C) exposure–effect association between BMI and hospitalized death. P _nonlinearity = 0.0067, OR = 1.06 (95% CI: 1.03–1.09) per 5 kg/m² of BMI increment

Third, the authors used funnel plots for detecting potential publication bias and found possibility of publication bias. However, funnel plots are under powered when only few studies (N < 10) were included in the meta‐analysis according to the guidelines. ¹² , ¹³ Therefore, it could be more appropriate to mention the known limitations of these methods because of inadequate numbers of included trials to properly assess a funnel plot, rather than claiming use of any test and further possibility of publication bias.

FUNDING INFORMATION

This work was supported in part by the National Natural Science Foundation of China (No. 81760050, 81760048) and the Jiangxi Provincial Natural Science Foundation for Youth Scientific Research (No. 20192ACBL21037).

CONFLICT OF INTEREST

No conflict of interest was declared.