Response to Comment on Cai et al. Obesity and COVID-19 Severity in a Designated Hospital in Shenzhen, China. Diabetes Care 2020;43:1392–1398

In this issue of Diabetes Care, a comment letter by He and Li showed that obesity was consistently a significant risk factor for coronavirus disease 2019 (COVID-19) severity (1). Furthermore, they also reported 17 risk factors affecting the disease severity in 98 patients with COVID-19. Beyond the study findings themselves, this study is of interest in its use of four machine learning classification algorithms, although most of the predictors of COVID-19 severity have been reported recently (2). Similar predictors from different patient series identified by using different statistical methods might strengthen causal inference and shed light on the underlying mechanisms. However, there are some concerns that warrant attention. The first lies in the small sample size. There were 98 patients included in the machine learning algorithm, and of them, 26 (36.5%) were severe cases. As for the description that there were “17 key factors affecting the severity of COVID-19,” at least 17 clinical parameters might have been included in the machine learning process, although the exact number of parameters was unclear. The sample size for machine learning was less than six times the number of parameters included, which may be substantially underpowered. Second, whether these 98 patients were all adults or included both children and adults was unclear. Thus, the results should be interpreted with caution. It should be noted that pediatric patients, who usually have lower BMI than adults, were reported to have better prognosis than adult patients (3), a finding that, if ignored, might lead to a false association between BMI and COVID-19 prognosis. Third, determining the “importance” of the predictive capability by using coefficients without taking into account data types as well as units of the variables could be misleading. Also, the definition of “importance” needs to be specified more explicitly, i.e., important as determined by the variance explained, P values, correlation coefficients, or other factors. Our study of all adult patients from Shenzhen, one of the Big Four cities in China, with comprehensive assessment of adiposity and COVID-19 severity, should have provided the best available evidence to date (4). Further studies investigating the mechanistic pathways by which obesity exaggerated the severity of COVID-19 are needed.

Given that obesity is a well-documented risk factor for a wide range of noncommunicable chronic diseases as well as infectious diseases including severe acute respiratory syndrome coronavirus 2 infection, public health professionals and health care providers need to advocate urgently for keeping a healthy body weight.