Body mass index and severity/fatality from coronavirus disease 2019: A nationwide epidemiological study in Korea

Objective Obesity has been reported as a risk factor for severe coronavirus disease 2019 (COVID-19) in recent studies. However, the relationship between body mass index (BMI) and COVID-19 severity and fatality are unclear. Research design and methods This study included 4,141 COVID-19 patients who were released from isolation or had died as of April 30, 2020. This nationwide data was provided by the Korean Centers for Disease Control and Prevention Agency. BMI was categorized as follows; < 18.5 kg/m2, 18.5–22.9 kg/m2, 23.0–24.9 kg/m2, 25.0–29.9 kg/m2, and ≥ 30 kg/m2. We defined a fatal illness if the patient had died. Results Among participants, those with a BMI of 18.5–22.9 kg/m2 were the most common (42.0%), followed by 25.0–29.9 kg/m2 (24.4%), 23.0–24.9 kg/m2 (24.3%), ≥ 30 kg/m2 (4.7%), and < 18.5 kg/m2 (4.6%). In addition, 1,654 (41.2%) were men and 3.04% were fatalities. Multivariable analysis showed that age, male sex, BMI < 18.5 kg/m2, BMI ≥ 25 kg/m2, diabetes mellitus, chronic kidney disease, cancer, and dementia were independent risk factors for fatal illness. In particular, BMI < 18.5 kg/m2 (odds ratio [OR] 3.97, 95% CI 1.77–8.92), 25.0–29.9 kg/m2 (2.43, 1.32–4.47), and ≥ 30 kg/m2 (4.32, 1.37–13.61) were found to have higher ORs than the BMI of 23.0–24.9 kg/m2 (reference). There was no significant difference between those with a BMI of 18.5–22.9 kg/m2 (1.59, 0.88–2.89) and 23.0–24.9 kg/m2. Conclusions This study demonstrated a non-linear (U-shaped) relationship between BMI and fatal illness. Subjects with a BMI of < 18.5 kg/m2 and those with a BMI ≥ 25 kg/m2 had a high risk of fatal illness. Maintaining a healthy weight is important not only to prevent chronic cardiometabolic diseases, but also to improve the outcome of COVID-19.

Introduction temporarily accessed the data from the KDCA through an encrypted remote system that contained anonymized data of 5,628 patients who were released from isolation or died during the aforementioned period. After excluding 1,487 patients (272 patients under the age of 20, 19 pregnant women, 26 without clinical severity information, 4 with no information regarding comorbidity, and 1,166 without available BMI value), 4,141 patients were included in the analysis for this study.
The Ewha Womans University Mokdong Hospital Institutional Review Board deemed this study exempt from ethical review and waived the requirement for informed consent (EUMC2020-07-002), since it was completely anonymized and without personal information.  was diagnosed based on the detection of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) by real-time polymerase chain reaction test with respiratory specimens [18]. The investigation of comorbid diseases was performed depending on whether the patients were previously diagnosed with specific diseases or not. Body temperature and BMI were the initial findings of hospital admission.
Disease severity was defined according to the guidelines of the KDCA and World Health Organization (WHO) [22,23]. Briefly, we defined a patient as having a "critical illness" if they required more than invasive mechanical ventilation and "fatal illness" if they died. Critical illnesses included patients requiring invasive mechanical ventilation, those with multi-organ failure, those requiring extracorporeal membrane oxygenation therapy, and/or those who died. The severity evaluation was based on the most severe condition during the hospital stay. Critical illness was defined as severe COVID-19.

Statistical analysis
Descriptive statistics were described using frequencies and proportions for categorical variables. The chi-square test was used to compare the variables. The values of continuous variables were expressed as mean and standard deviation. Analysis of variance was performed for the body temperature.
Multivariable logistic regression was used to analyze the independent risk for COVID-19 critical illness and COVID-19 fatal illness after adjusting for age, sex, BMI, and five comorbid diseases: diabetes mellitus (DM), hypertension, chronic kidney disease (CKD), cancer (excluding cured cases) and dementia. The age variable was given as a categorical variable in units of 10 years, but we calculated the odds ratio of every 10 years as a continuous variable. Since dementia had missing data (n = 314) that were not replaced, 3827 participants were included in the multivariable analysis. Although the BMI of 23.0-24.9 kg/m 2 was classified as overweight rather than normal-weight according to the Asia-Pacific cutoff values [21], we used it as a reference. Since this group is of normal weight according to the WHO classification [24], and a previous study reported a BMI cutoff value of Koreans as 24.2 kg/m 2 [25].
To analyze the relationship between BMI and fatality by sex, we performed a multivariable analysis after adjusting for age and the five comorbid conditions. In addition to sex, we analyzed the relationship between BMI and fatality according to whether they had DM or HTN. We performed multivariable analysis after adjusting for age, sex, and four comorbidities other than the target disease (DM or HTN). Then, the relationships between BMI and fatality in the groups with and without the diseases (DM or HTN) were evaluated separately.
The threshold for statistical significance was set at P <0.05. Statistical analyses were performed using SAS software (version 9.4, SAS Institute, Cary, NC, USA).

Results
Baseline characteristics according to BMI are presented in Table 1. A BMI of 18.5-22.9 kg/m 2 was the most common, with 1,741 (42.0%) patients, followed by BMI of 25.0-29.9, 23.0-24.9 kg/m 2 as 1,011 (24.4%), and 1,005 (24.3%), respectively. Moreover, BMI of < 18.5, and � 30 kg/m 2 were less than 5%. In the BMI < 18.5, 18.5-22.9, and � 30 kg/m 2 groups, the most patients were in their 20s, whereas in the BMI of 23.0-24.9, and 25.0-29.9 kg/m 2 groups, most patients were in their 50s. The BMI of 18.5-22.9 kg/m 2 was the most common both in male and female, whereas the proportions of the BMI ranging 23.0-24.9 and 25.0-29.9 kg/m 2 were higher in male than in female (raw %, male: 27.3, 32.0 vs. female: 22.1, 19.0%, respectively). Those with BMI � 25.0 kg/m 2 had higher body temperature than others (P < 0.0001). Dyspnea was a more frequent complaint in patients with a BMI of < 18.0 and � 25.0 kg/m 2 than in others (P = 0.0135). Hypertension and DM were more common in the BMI of 23.0-24.9 and � 25.0 kg/m 2 than in others (P < 0.0001). S1 Table shows the patient characteristics according to the fatal illness. We noted that 41.7% were male. There were no deaths in patients aged < 40 years. The overall proportions of patients in their 20s and 50s were 942 (22.7%) and 873 (21.1%), respectively. Systolic blood pressures of < 120 mmHg and � 140 mmHg were more related to fatal illness than those between 120 mmHg and 140 mmHg (P < 0.0001).
Independent risk factors associated with severe COVID-19 are shown in Table 2. With every 10 years of increase in age, the odds ratio of multivariable analyses for critical and fatal illness increased to 2.64 (95% CI: 2.20-3.15) and 3.04 (2.45-3.78), respectively. Male sex was also found to be an independent risk factor. When the BMI of 23.0-24.9 kg/m 2 group was used as a reference, the risk of BMI ranging 18.5-22.9 kg/m 2 was as high as 1.59 (0.88-2.89) for fatal illness without statistical significance. All five comorbidities (DM, hypertension, CKD, cancer, and dementia) were related to a higher critical illness rate compared to the absence of comorbidities. Hypertension had an odds ratio of 1.34 in fatal illness, but had no statistical power (95% CI: 0.96-2.30) in multivariable analysis. S1 Fig shows the number of critical and fatal illnesses according to the BMI group and the fatality rate of the study population. The BMI of 23.0-24.9 kg/m 2 group was found to have the lowest fatality. Fatality rate increased in both BMI of < 18.5 and 25.0-29.9 kg/m 2 groups, but slightly decreased in the BMI of � 30.0 kg/m 2 group. As shown in S1 Fig, patient fatality was the highest in the BMI of < 18.5 kg/m 2 , followed by the BMI of 25.0-29.9, and � 30.0 kg/m 2 . After multivariable analysis with adjusting covariates, a similar result was shown in the group with BMI < 18.5 kg/m 2 (Fig 1). However, the result according to the degree of obesity showed that the odds ratio of BMI � 30.0 kg/m 2 was higher in patients with critical and fatal illness than in those with a BMI of 25.0-29.9 kg/m 2 . Hence, the risk was increased in the BMI < 18.5 kg/m 2 and BMI of � 25.0 kg/m 2 groups, resulting in a U-shaped curve for critical and fatal illness (Fig 1). Fig 2 shows the association between BMI and sex for critical and fatal illnesses. This was expressed as an odds ratio after adjusting for age and comorbidities. Fig 2 shows similar U-shaped curves in both males and females, although it seems to be slightly different in detail depending on sex. Women showed a higher risk in their BMI < 18.5 kg/m 2 population than men. While, men showed a higher risk in their BMI � 25.0 kg/m 2 population than women (Fig 2A and 2B). However, the P interaction for among males and females were not significant.

Discussion
In this nationwide study, we showed that BMI � 25.0 kg/m 2 was associated with an increased risk for fatal illness as well as critical illness from COVID-19. Furthermore, BMI < 18.5 kg/m 2 also increased the risk of critical and fatal illness caused by COVID-19, similar to that seen with BMI � 25.0 kg/m 2 . The frequency data of S1 Fig shows that fatality decreased in those with BMI � 30.0 kg/m 2 , and it appeared as if the obesity paradox existed. However, the odds ratio after adjusting for covariates showed a U-shaped curve, as shown in Fig 1. Hence, BMI � 25.0 kg/m 2 and < 18.5 kg/m 2 were found to have an association with increased fatality from COVID-19 independent of combined comorbidities. Although the male sex showed a higher risk of fatal illness from COVID-19, there was no definite evidence that BMI had a more detrimental effect on men than on women in this study.  The results of the current study showed that maintaining a healthy weight in the general population, regardless of underlying comorbidities, is still important in the COVID-19 pandemic era. In this study, normal weight (18.5 < BMI <25 kg/m 2 ) according to the WHO criteria, especially 23-25 kg/m 2 in the South Korean population, is considered a healthy weight in the COVID-19 pandemic era.
Obesity, defined using BMI, has different cutoff values between the global and Asia-Pacific standards. The WHO announced the Asia-Pacific perspective: redefining obesity and its treatment' in 2000 on the evidence that the Asian population has an increased risk of diseases at BMI > 23 kg/m 2 [21]. The Asia-Pacific BMI was categorized as followings; underweight (< 18.5 kg/m 2 ), normal-weight (18.5-22.9 kg/m 2 ), overweight (23.0-24.9 kg/m 2 ), obesity-1 (25.0-29.9 kg/m 2 ), and obesity-2 (� 30 kg/m 2 ). Proposed mechanisms of this inter-population discrepancy of defined obesity by BMI include differences in percentage and distribution of body fat across populations, the interplay of genetic susceptibility, and environmental factors related to diet and sedentary lifestyle [19,20].
However, the debate on whether racial differences exist in defining obesity based on BMI is still ongoing. Despite the differences in BMI cutoff points for obesity in these populations, a meta-analysis of four continents in 239 prospective studies showed that BMI and all-cause mortality were broadly consistent across the four continents [3]. In 2004, WHO expert consultation recommended that the use of global standards is appropriate because the BMI standards for obesity do not differ greatly by race, and it is not appropriate to stipulate different standards for the Asia-Pacific region due to small differences [19]. In a study comparing the BMI cutoff value of the South Koreans with the global standard in 2015, the cutoff point of Korean BMI was 24.2 kg/m 2 (sensitivity 78%, specificity 71%, as the result of receiver operating characteristic cure, based on the body fat percentage) and it was only 1.3 kg/m 2 lower than the global standard [25]. The result supports the recommendation of the WHO expert consultation in 2004. Nonetheless, the Korean government continues to follow the 2000 Asian-Pacific cut-off values. Hence, to avoid confusion, we did not express obesity or normal for the BMI group, but only used numeric values in this study. There are limited studies on the clinical implications of BMI, from underweight to obesity, in viral pandemics. In the case of influenza-associated pneumonia, an increase in risk has been observed in underweight and obesity [4]. However, regarding obesity, contradictory resultsobesity paradox-were reported in previous studies [15]. Similarly, there was a recent study on 'an obesity survival paradox' that authors showed the projected rates of COVID-19 infection and mortality drop with elevated prevalence of obesity in United States [26]. However, in the current COVID-19 pandemic, the majority of studies reported obesity as a risk factor for progressing to severe COVID-19 and were associated with the need for hospitalization and admission to the critical care unite due to COVID-19 [7,17,27,28]. Importantly, our study also showed increased BMI as a risk factor for critical and fatal illness of COVID-19 and cannot prove the 'obesity survival paradox.' In contrast to obesity, the relationship between underweight and COVID-19 is unknown. A recent study showed a J-shaped (non-linear) relationship between BMI and risk for COVID-19 related hospitalization, ICU admission, and death [28]. In our study, subjects with a BMI � 30 kg/m 2 presented higher fatality with an odds ratio of 4.32, compared to subjects of a 23 �BMI < 25 kg/m 2 . In addition, BMI < 18.5 kg/m 2 also showed an odds ratio of 3.97 (1.77-8.92) for fatal illness.
One recent remarkable study is the analysis of the UK biobank, even though preliminary data are available [29,30]. They showed that BMI was strongly associated with positive results in the COVID-19 test and the risk of death related to COVID-19 [29]. Interestingly, both BMI and waist circumference were associated with testing positive for COVID-19 in a doseresponse fashion [30]. However, another study using a two-sample multivariable Mendelian randomization failed to show the impact of body composition (waist circumference, trunk fat ratio) on COVID-19 susceptibility and severity; only BMI was significant [31]. Our study has a limitation in that it did not include body composition.
In infectious diseases, research on whether there are gender differences in the effect of BMI on mortality is limited. In a UK study [32], with the national mortality data of 3.6 million adults, the mortality specific outcome related to respiratory infection, the hazard ratio (HR) of underweight increased similarly to obesity, showing a U-shaped curve, which is consistent with our study results. Further, the UK study showed the difference between men and women in all-cause mortality among never-smokers. Underweight women were associated with a greater increase in HR than men, while obese men were associated with a greater increase in HR than women (P interaction < 0.0001). In other words, underweight and obesity were associated with an increase in the HR of all-cause mortality in both men and women, but the increased risk was greater in underweight women and obese men. This is consistent with our results shown in Fig 2. However, there are some differences between the study populations and the target outcomes, since our study aimed at evaluating the association between the fatality of COVID-19 and BMI. Additionally, we did not have information regarding smoking history. Furthermore, the sample size in our study was relatively small, and when divided by subgroup analysis, a similar pattern was shown as the UK study between men and women, although the P interaction did not show a significant value. Therefore, it is necessary to verify this through large-scale research. A Chinese study of 383 patients found that obesity, especially in men, significantly increased the risk of developing severe COVID-19 [9]. In this study, we found that there was a slight difference in the odds ratio and increasing pattern in those with BMI of < 18.5 and � 25.0 kg/m 2 according to sex. Male sex was associated with an increased odds ratio of fatal illness (2.92 [1.9-4.48]) than the female sex. However, there was no definite evidence that men had a higher risk of severe COVID-19 due to BMI than women.
All-cause mortality due to obesity is attenuated by aging. Large-scale epidemiological studies have reported that heathy BMI shifts to the right in old age compared to young age [3,32,33]. Among the those with COVID-19, young obese patients (< 60 years old and BMI > 35 kg/m 2 ) were reported to be 3.6 times more likely to be admitted to the critical care unit than those with BMI < 30 kg/m 2 [34]. However, in our study, the incidence of severe COVID-19 in patients aged < 60 years was low (fatal illness was only 8 cases and critical illness was as small as 15 cases), so there was a limit to whether obesity was related to critical illness at a young age.
Obesity is regarded as an important public health issue, mainly related to chronic diseases. Obesity potentiates multiple cardiovascular risk factors and causes atherosclerosis. It also causes insulin resistance and reduces beta cell function [35]. It can cause functional immunological deficits through dysregulation of the immune system, resulting in diseases mainly related to cardiometabolic problems and chronic inflammation [2]. However, in the recent SARS-CoV2 pandemic, obesity was reported to be associated with severe COVID-19, and its clinical significance in acute infections may be receiving new attention from researchers. In addition to chronic metabolic conditions, proposed mechanisms for the detrimental effect of obesity in severe COVID-19 are related to the cardiorespiratory system, increased thrombogenicity, and immune hyper-reactivity [7,36,37]. Beyond these proposed mechanisms, COVID-19 can induce direct target organ damage, regardless of the underlying disease. Direct invasion or inflammation of cardiomyocyte is related to poor disease outcomes [18,38,39]. In the case of SARS, induced by SARS-CoV, the possibility of disruption of pancreatic beta cell function through pancreatic invasion through binding to the cellular entry, ACE-2, was confirmed in an autopsy report [40]. However, there is paucity of evidence in COVID-19 pancreatitis through direct beta cell involvement of SARS-CoV-2 [41]. Furthermore, there is no available data regarding whether obesity or underweight can augment the virus infection related to direct target organ damage by altering the susceptibility or ACE-2.
The present study has some limitations. Despite the advantage of nationwide data compared to previous studies related to COVID-19 and BMI, this study was limited to Koreans and did not include a diverse population. Although the fatality of BMI and COVID-19 was evaluated by adjusting for sex, age, and comorbid conditions, there was an issue in subgroup analysis because the sample was insufficient in terms of specific diseases. Furthermore, detailed information about comorbidities was not provided. The number of patients with BMI < 18.5, or � 30 kg/m 2 , was small; 191 (4.6%) or 193 (4.7%), respectively. Therefore, there is a chance of being statistically underpowered or having an increased margin of error. Furthermore, we could not check the distribution of continuous BMI according to fatal illness status. The KCDA provided BMI data in a categorized form. Finally, it was an observational study and could not clearly explain causal relationships. Although the characteristics of acute infectious disease act as a limitation in the study of COVID-19, a large-scale prospective cohort study of various populations is of utmost important.

Conclusions
We analyzed the relationship between BMI and criticality/fatality from COVID-19. In particular, those with a BMI of < 18.5 kg/m 2 and � 25.0 kg/m 2 were found to have a higher association with critical and fatal illness compared to those with a BMI of 23.0-24.9 kg/m 2 . Maintaining a healthy weight is important not only to prevent chronic cardiometabolic diseases, but also to improve the outcome of COVID-19.