Diabetes Care 1

Danielle K. Longmore,1,2,3 and Overweight/ Jessica E. Miller,1,4 Siroon Bekkering,1,5 Christoph Saner,1,6 Edin Mifsud,1,7 Are Independent, Nonadditive Yanshan Zhu,8 Richard Saffery,1,4 Alistair Nichol,9,10,11 Graham Colditz,12 Risk Factors for In-Hospital Kirsty R. Short,8 and David P. Burgner,1,3,4,13 on behalf of the International BMI-COVID Severity of COVID-19: An consortium* International, Multicenter 1Murdoch Children’s Research Institute, The Royal Retrospective Meta-analysis Children’s Hospital, Parkville, Victoria, Australia 2Menzies School of Health Research, Charles RESEARCH SERVICES EPIDEMIOLOGY/HEALTH https://doi.org/10.2337/dc20-2676 Darwin University, Darwin, Australia 3Infectious Diseases Unit, Department of General Medicine, The Royal Children’s Hospital, Park- ville, Victoria, Australia 4Department of Paediatrics, Melbourne Univer- sity, Parkville, Victoria, Australia 5Department of Internal Medicine, Radboud In- stitute for Molecular Life Sciences, Radboud Uni- versity Medical Center, Nijmegen, the Netherlands 6Pediatric , Diabetology and Me- tabolism, Department of Pediatrics, University Hospital Inselspital, University of Bern, Bern, Switzerland OBJECTIVE 7World Health Organization Collaborating Cen- Obesity is an established risk factor for severe coronavirus disease 2019 (COVID-19), tre for Reference and Research on Influenza, but the contribution of overweight and/or diabetes remains unclear. In a mul- Doherty Institute, Melbourne, Australia 8School of Chemistry and Molecular Biosciences, ticenter, international study, we investigated if overweight, obesity, and diabetes The Universityof Queensland, Brisbane, Australia were independently associated with COVID-19 severity and whether the BMI- 9Department of Intensive Care, Alfred Health, associated risk was increased among those with diabetes. Melbourne, Australia 10Australian and New Zealand Intensive Care RESEARCH DESIGN AND METHODS Research Centre, Monash University, Melbourne, Australia We retrospectively extracted data from health care records and regional databases 11University College Dublin Clinical Research Cen- of hospitalized adult patients with COVID-19 from 18 sites in 11 countries. We used tre, St Vincent’s Hospital, Dublin, Ireland fi fi 12Washington University, St. Louis, MO standardized de nitions and analyses to generate site-speci c estimates, modeling 13 the odds of each outcome (supplemental oxygen/noninvasive ventilatory support, Department of Paediatrics, Monash University, Clayton, Victoria, Australia invasive mechanical ventilatory support, and in-hospital mortality) by BMI category Corresponding authors: David P. Burgner, david. (reference, overweight, obese), adjusting for age, sex, and prespecified comorbid- [email protected], and Danielle K. Longmore, ities. Subgroup analysis was performed on patients with preexisting diabetes. Site- [email protected] specific estimates were combined in a meta-analysis. Received 29 October 2020 and accepted 14 January 2021 RESULTS This article contains supplementary material online Among 7,244 patients (65.6% overweight/obese), those with overweight were at https://doi.org/10.2337/figshare.13616024. more likely to require oxygen/noninvasive ventilatory support (random effects D.K.L. and J.E.M. contributed equally as first adjusted odds ratio [aOR], 1.44; 95% CI 1.15–1.80) and invasive mechanical authors, and K.R.S. and D.P.B. contributed equally as senior authors. ventilatory support (aOR, 1.22; 95% CI 1.03–1.46). There was no association between overweight and in-hospital mortality (aOR, 0.88; 95% CI 0.74–1.04). *A complete list of International BMI-COVID con- sortium members is included in the supplemen- Similar effects were observed in patients with obesity or diabetes. In the subgroup tary material online. analysis, the aOR for any outcome was not additionally increased in those with This article is part of a special article collection diabetes and overweight or obesity. available at https://care.diabetesjournals.org/ collection/diabetes-and-COVID19. CONCLUSIONS © 2021 by the American Diabetes Association. In adults hospitalized with COVID-19, overweight, obesity, and diabetes were Readers may use this article as long as the work is associated with increased odds of requiring respiratory support but were not properly cited, the use is educational and not for profit, and the work is not altered. More infor- associated with death. In patients with diabetes, the odds of severe COVID-19 were mation is available at https://www.diabetesjournals not increased above the BMI-associated risk. .org/content/license. Diabetes Care Publish Ahead of Print, published online April 15, 2021 2 Diabetes, Increased BMI, and COVID-19 Severity Diabetes Care

In the first 6 months of the coronavirus RESEARCH DESIGN AND METHODS Data Collection $ disease 2019 (COVID-19) pandemic (until Study Design Study participants were aged 18 years, June 30, 2020), .10 million people had We conducted an international, multi- admittedtohospitalwithCOVID-19 been infected with severe acute respi- center, retrospective analysis of hospi- (confirmed by PCR for SARS-CoV-2), had ratory syndrome coronavirus 2 (SARS- talized patients with COVID-19 from a height and weight recorded on admission CoV-2) and .500,000 COVID-19–related total of 69 hospitals in 11 countries (Sup- to participating sites with local approval deaths had been recorded (1), but strik- plementary Table 1) from 17 January to participate. The period for data col- ing variation in clinical severity and out- 2020 to 2 June 2020. Data from 69 lection from individual sites varied (Sup- comes remains. Identifying risk factors hospitals were collated to form 18 sites plementary Table 1). associated with more severe COVID-19 is that each provided site-specific outcomes Information regarding participant de- essential for optimizing individual treat- and estimates. We modeled the odds of mographic variables (i.e., age, sex), BMI, ment, resource allocation, and prioritiz- in-hospital respiratory support (ie, sup- pre-existing medical conditions, clinical ing immunization distribution. Obesity plemental oxygen/noninvasive ventilatory variables including intensive care unit has emerged as an important risk factor support, invasive mechanical ventilatory admission, and treatment including ox- for severe COVID-19 (2), but several key support) and in-hospital mortality by ygen and noninvasive ventilatory sup- questions remain unanswered (3). BMI category, adjusting for age, sex, port and mechanical ventilatory support First, most studies to date have fo- and prespecified comorbidities, as de- were identified. Supplemental oxygen cused on individuals with obesity (BMI scribed later in this section. A protocol was defined as the provision of oxygen 2 $30 kg/m ) (4), but the specific contri- was finalized on 20 April 2020 (see Sup- via nasal canulae or face mask. Nonin- bution of overweight (BMI between $25 plementary Material) prior to com- vasive ventilatory support was defined and ,30) to severe COVID-19 has only mencement of the study. The study was as the use of a device providing contin- been investigated in a few studies, which conducted in accordance with Good Clin- uous positive airway pressure or bilevel have reported inconsistent results (5– ical Practice guidelines, local regulations, positive airway pressure. Cardiovascular 10). This is a significant knowledge gap, and the ethical principles described in disease was defined as preexisting, phy- because 40% of the global population is the Declaration of Helsinki. Ethical ap- sician-diagnosed coronary disease, overweight, in addition to the 13% living proval was obtained at the coordinating ischemic stroke, heart failure, and/or with obesity (11). Second, most studies center (Murdoch Children’sResearch peripheral vascular disease. Diabetes are single-center analyses and are un- Institute [MCRI], Royal Children’sHos- was defined as preexisting diabetes (in- likely to be globally representative, given pital, Melbourne, Australia; approval cludingtype1or2).Inallcountries,type2 the marked intercountry variation in no. HREC 63887), and local approvals diabetes was diagnosed according to the overweight and obesity (12). This short- were obtained at participating sites, American Diabetes Association guide- coming has partially been addressed by depending on local regulations. Informed lines or local guidelines with the same meta-analyses, but these rarely include consent was not required. diagnostic criteria as the American Di- individuals who are overweight (2). Fi- abetes Association guidelines. For three nally, both overweight and obesity fre- Data Source sites only (Cape Town, South Africa; quently occur with other comorbidities, We analyzed deidentified data from ex- Lausanne, Switzerland; and Ticino, Swit- particularly (13). How- isting collections of hospital data and zerland), a small number of patients ever, many studies have either not ad- regional databases, including the Norwe- were included who were first diagnosed justedforthesecovariablesorthe gian Intensive Care and Pandemic Reg- withdiabetesduringtheiradmissionwith regression models used did not allow istry,Norway;and Washington University, COVID-19. Preexisting respiratory condi- clinical translation of findings (3). Spe- St. Louis, Missouri (see Supplementary tions and were defined cifically, a key clinical question is whether Appendix for participating sites and in- as physician-diagnosed and currently patients with both diabetes and higher vestigators). Data from smaller contrib- on treatment. Data cleaning was per- BMI have a higher risk of severe COVID- uting hospitals were collected for clinical formed for out-of-range values, incon- 19 compared with those with diabetes auditing processes approved by local sistent data, and repeated participant and BMI in the normal range. hospitals and in accordance with local entries. Central source data verification In the present study, we aimed to regulations. Each site followed a stan- was not feasible for this study, because address these unresolved questions by dardized protocol for data coding and coding was performed by the individual performing an international, retrospec- analysis to generate site-specific esti- participating centers. tive, multisite analysis of 7,244 hospi- mates for each study population (Supple- talized patients with COVID-19 from 18 mentary Material). Deidentified data Statistical Analysis sites in 11 countries. We used common from hospitals in Austria, Singapore, All analyses were conducted as outlined definitions and analyses to delineate the Netherlands, Switzerland, and Indo- in our protocol (Supplementary Material). whether overweight, obesity, and di- nesia were exported to the coordinating Participant data are presented as fre- abetes are independent risk factors center (MCRI) for generation of site- quency, reported as percentage. Each for respiratory support and in-hospital specific estimates. All transfer of data site (or the MCRI) followed a standard- mortality. In patients with diabetes, we and site-specific estimates to the MCRI ized protocol for data coding and analysis also investigated the association be- was subject to a data transfer agreement. to generate site-specific estimates from tween BMI category and COVID-19 Statisticians at the MCRI completed the each study population, modeling the severity. meta-analyses. odds of each outcome by BMI (calculated care.diabetesjournals.org Longmore and Associates 3

as weight [kg] divided by height squared Boston, MA) (16). Meta-analyses were CI 0.92–1.64; P 5 0.17). The low number of [m2]) category (Supplementary Material). performed in Stata SE, version 16.0 (17). participants in the group (n 5 BMI was categorized as underweight 162) precluded calculation of robust odds ($12 to ,18), normal ($18 to ,25 [the RESULTS ratios. The I2 statistic, which describes the referent]),overweight ($25to,30), and Characteristics of Patients Included in percentage of variation across studies that obese ($30). In sensitivity analyses for the Study is due to heterogeneity rather than chance, Asian populations, respective BMI cate- A total of 7,244 patients from 18 sites was 43.6% and 53.7% among the obese gories were based on the following (n 5 69 hospitals) in 11 countries were groups for invasive mechanical ventila- ranges: $12 to ,18.5, $18.5 to ,24 included in this study of hospitalized tory support and in-hospital mortality, (referent), $24 to ,28, and $28 (14). patients with COVID-19 (Supplementary respectively, suggestingmodest hetero- Logistic regression was used to model the Tables 1 and 3). Among these, 60.1% geneity across studies. Unadjusted site- association between BMI category and were male and 51.7% were older than specific odds ratios are presented in Figs. the use of in-hospital respiratory thera- 65 years. Overall, 34.8% were over- 1–3. pies (listed previously in this section) and weight and 30.8% obese; however, there For the Chinese, Indonesian, and Sin- in-hospital mortality. was considerable variability across dif- gaporean sites, odds ratios varied slightly All models estimated crude (unad- ferent individual countries and sites. depending on whether the standard or justed) and adjusted odds ratios. Two Prevalence of obesity for each site coun- Asian country–specificBMIcategories levels of adjustments were made. The try is provided in Supplementary Table 2. were used. The variation did not mean- first level, available for all sites, included The rates of comorbidities and the fre- ingfully affect the summarized meta- age, sex, preexisting cardiovascular dis- quency of outcomes varied across sites analysis estimates (Supplementary Table ease, diabetes, preexisting respiratory (Supplementary Tables 3 and 4). Preva- 4). Additional adjustments for current conditions, and hypertension. The sec- lence of diabetes varied from 7% in Guang- and race/ethnicity, where data ond level of adjustments included the dong Province, China, to 46% in St. Louis, were available, had little impact on the first level of adjustments plus current Missouri. Prevalence of diabetes among odds ratios (Supplementary Table 5). smoking status and/or race/ethnicity, those of normal weight ranged from 6% depending on data availability. The sec- in Milan Sacco, Italy, to 39% in Cape Association of Diabetes With ond level was available for only five sites. Town, South Africa. Prevalence of dia- Supplemental Oxygen/Noninvasive The crude and adjusted (first-level) mod- betes among those who were overweight Ventilatory Support, Mechanical els were run on data from a subgroup of or obese ranged from 7% to 5%, respec- Ventilatory Support, and In-Hospital patients with preexisting diabetes. No tively, in Guangdong Province, China, to Death adjustment was made for multiple com- 47% and 53%, respectively, in St. Louis, Compared with patients without diabe- parisons. Covariables had few missing Missouri. tes, those with diabetes had an increased dataandnoimputationswerewar- odds of needing mechanical ventilatory ranted. Site-specificadjustedestimates Association of Overweight, Obesity, support in random effects models ad- for BMI category, each independent and Supplemental Oxygen/ justed for all covariables, including BMI covariable included in the adjusted mod- Noninvasive Ventilatory Support, category and comorbidities (aOR 1.21; els, and the diabetes subgroup estimates Mechanical Ventilatory Support, and 95% CI 1.03–1.41; P 5 0.02) (Supplementary were combined in meta-analyses. In-Hospital Death Fig. 1). There was no increase in odds of Summarized estimates included fixed Compared with normal weight, over- noninvasive respiratory support or in- and random effects models (15). Ran- weight and obesity were associated with hospital mortality in those with diabetes dom effects estimates are presented in increased odds of supplemental oxygen/ (Supplementary Fig. 1). In addition to di- the text. Of note, the Los Angeles, New noninvasive ventilatory support (random abetes, other host factors previously asso- York, and Cape Town sites were not in- effects adjusted odds ratio (aOR), 1.44 ciated with severe COVID-19 (i.e., increased cluded in analysis of supplemental oxygen/ [95% CI 1.15–1.80], P 5 0.02; and aOR, age, male sex, preexisting cardiovascular noninvasive ventilatory support, because 1.75 [95% CI 1.33–2.30], P , 0.01), re- disease, and chronic respiratory disease) nearly all hospitalized patients received spectively(Fig.1).Obesitywasassociated (18) were each independently associated supplemental oxygen per local policies. with a 73% increase in odds for invasive with an increased risk of one or more of the Data on supplemental oxygen were not mechanical ventilatory support (aOR, 1.73; selected study outcomes (Supplementary available for Austria, Norway, or Amphia 95% CI 1.29–2.32; P , 0.01) (Fig. 2), and a Figs. 2–4). (the Netherlands). Variations to the pre- more modest association was observed for planned analysis were made because overweight (aOR, 1.22; 95% CI 1.03–1.46; Among Patients With Diabetes, there were insufficient data available P 5 0.02). Data on this outcome were not Increased BMI Did Not Increase the from the majority of sites. The outcomes available from Amphia (the Netherlands). Risk of Severe COVID-19 Outcomes not analyzed included intensive care unit Overweight was not associated with To further inform patient care, we next length of stay, hospital length of stay, an increase in odds for in-hospital mor- performed a subgroup analysis of indi- and extracorporeal membrane oxygen- tality (aOR, 0.88; 95% CI 0.74–1.04; P 5 viduals with diabetes. Specifically, we ation use (Supplementary Material). Site- 0.13) (Fig. 3). Obesity was also not as- investigated if BMI category among those specific analyses were performed in SAS sociated with an increase in odds of with diabetes was associated with (SAS Institute, Cary, NC), Stata (StataCorp, in-hospital mortality, with confidence the selected COVID-19 outcomes. Strik- College Station, TX) or R studio (PBC, limits including the null (aOR, 1.23; 95% ingly, there was no association between 4 Diabetes, Increased BMI, and COVID-19 Severity Diabetes Care

Figure 1—Meta-analysis odds ratios for requirement of supplemental oxygen/noninvasive ventilatory support by BMI category. Models were adjusted for age (,65, $65 years),sex (male/female), preexisting (yes/no), diabetes (yes/no), preexistingrespiratory conditions(yes/no), and hypertension (yes/no). The reference BMI is in the normal range (i.e., $18 to ,25). The 95% CIs of the odds ratios were not adjusted for multiple testing and should not be used to infer definitive effects. Data from Norway; Amphia (in the Netherlands); Austria; South Africa; University of California, Los Angeles, California; Cornell University, Ithaca, New York, were not included in this model, because data were either not available for this outcome or all patients received the therapy. D1L, DerSimonian and Laird random effects model; FG&V, Franciscus Gasthuis & Vlietland; I-V, inverse-variance weighted fixed effects model; MC, medical center.

overweight or obesity and supplemental COVID-19, overweight was associated disease were associated with worse out- oxygenuse/noninvasiveventilatorysup- overall with an increased requirement comes with COVID-19 (19). In the present port (aOR 1.04 [95% CI 0.54–2.00], P 5 of respiratory support. The association study, neither overweight/obesity nor 0.91; and 1.29 [95% CI 0.68–2.46], P 5 between overweight and in-hospital diabetes was associated with in-hospital 0.44, respectively), invasive mechanical mortality was not statistically significant. mortality. Although previous analyses ventilatory support (aOR 0.67 [95% CI Similar trends were observed in patients have suggested that obesity increases 0.40–1.12], P 5 0.10; and 1.25 [95% CI with obesity. In addition to the associ- the mortality risk associated with COVID- 0.62–2.53], P 5 0.73, respectively), or ations with BMI, diabetes was indepen- 19 (4,8,20), these studies did not neces- in-hospital mortality (aOR 0.79 [95% CI dently associated with increased COVID- sarily make adjustments for age, sex, and 0.52–1.20], P 5 0.28; and 1.14 [95% CI 19 severity but not death. Importantly, other comorbidities as we did in the 0.61–2.13], P 5 0.52, respectively) in amongpatientswithdiabetes,overweight/ present study, or only found an effect on those with preexisting diabetes (Fig. 4). In obesity were not associated with an in- death for those with more severe obe- this subgroup analysis, the sample size creased risk of severe COVID-19. sity (BMI .35) (21). The data presented was reduced and resulted in wide CIs. The data presented here are consis- here are consistent with previous find- tent with those of previous studies that ings that an elevated BMI is associated CONCLUSIONS reported not only obesity but also ad- with an increased requirement for re- In this large, international, multicen- vanced age, male sex, and preexisting spiratory support (5,22–24) and that di- ter study of patients hospitalized with cardiovascular,metabolic,andrespiratory abetes in patients with COVID-19 is not care.diabetesjournals.org Longmore and Associates 5

Figure 2—Meta-analysis odds ratios for invasive mechanical ventilatory support by BMI category. Models were adjusted for age (,65, $65 years), sex (male/female), preexisting cardiovascular disease (yes/no), diabetes (yes/no), preexisting respiratory conditions (yes/no), and hypertension (yes/no). The reference BMI is in the normal range (i.e., $18 to ,25). The 95% CIs of the odds ratios have not been adjusted for multiple testing and should not be usedtoinferdefinitiveeffects.DatafromAmphia(intheNetherlands)werenotavailableforinvasivemechanicalventilatorysupport.D1L,DerSimonian and Laird random effects model; FG&V, Franciscus Gasthuis & Vlietland; I-V, inverse-variance weighted fixed effects model; US UCLA, University of California, Los Angeles, California. significantly associated with in-hospital presentation (28), complement activation data from suggest that each mortality after appropriate adjustment (29), and/or suboptimal T-cell responses 1 mmol/L increase in plasma glucose (25). (30). Moreover, these immunomodula- level is associated with a 6% increased The mechanisms underlying the asso- tory effects may be compounded by the risk of hospitalization for pneumonia ciation between BMI and COVID-19 se- reduced functional respiratory capacity (32). Elevated blood glucose levels are verity likely reflect a dysregulated host in individuals with overweight/obesity, also associated with altered activity of response, resulting in heightened inflam- which may lead to a lower threshold for transporters responsible for clearing mation and/or a suboptimal antiviral noninvasive or invasive respiratory sup- the lung of interstitial edema and for response. There are a number of relevant port (31). maintaining the integrity the pulmonary immunomodulatory effects of overweight/ It is likely that the independent role epithelial–endothelial barrier (33–35), obesity, including chronic systemic in- identified for diabetes in COVID-19 se- both of which are likely to be important flammation (26), reduced production of verity reflects the role of hyperglycemia in determining the clinical outcome of type I interferons (27), reduced antigen in severe respiratory disease. Indeed, SARS-CoV-2 infection. This hypothesis is 6 Diabetes, Increased BMI, and COVID-19 Severity Diabetes Care

Figure 3—Meta-analysis odds ratios for in-hospital mortality by BMI category. Models were adjusted for age (,65, $65 years), sex (male/female), preexisting cardiovascular disease (yes/no), diabetes (yes/no), preexisting respiratory conditions (yes/no), and hypertension (yes/no). The reference BMI is inthenormalrange(i.e.,$18 to ,25).The95% CIsoftheoddsratioswerenotadjustedformultipletestingandshouldnotbeusedtoinferdefinitiveeffects. Datafrom GuandongProvince, China,andSingaporewerenotavailablefor in-hospitalmortality.D1L, DerSimonian and Laird random effects model; FG&V, Franciscus Gasthuis & Vlietland; I-V, inverse-variance weighted fixed effects model; US UCLA, University of California, Los Angeles, California.

consistent with observations that well- Currently, it is estimated that ;90% of respiratory support or death among pa- controlled blood glucose levels corre- patients with type 2 diabetes are over- tients with both COVID-19 and diabetes. lated with improved clinical outcomes in weight or obese (39). Previous studies Larger studies will be needed to confirm patients with COVID-19 who also had have suggested that among patients with these findings; however, this finding may diabetes (36). More severe COVID-19, COVID-19 who have diabetes, nonsurvi- reflect a “threshold effect” of suscepti- however, may also be associated with vors had a greater prevalence of comor- bility to severe COVID-19 in these con- elevated glucoselevels (37).Withrespect bidities compared with survivors (25). ditions. This hypothesis will require to the relationship between longer-term Given the clear independent role of BMI clinical and experimental evaluation. glucose control and COVID-19 severity, in COVID-19 severity, we reasoned that Our data will inform public policy, an elevated hemoglobin A1c is associated patients with both diabetes and an ele- particularly for risk-stratification of se- with increased risk of hospital admission vated BMI may be at increased risk of vere COVID-19 disease. The U.S. Centers due to COVID-19 in those with diabetes severe disease outcomes compared with for Disease Control and Prevention iden- (38). Additional studies investigating the patients with diabetes and a BMI in the tifies individuals with obesity (BMI $30) mechanistic roles of both BMI and diabe- normal range. Surprisingly, BMI was not as being at increased risk for severe dis- tes in COVID-19 severity are warranted. associated with the risk of in-hospital ease, as well as those with cardiovascular care.diabetesjournals.org Longmore and Associates 7

Figure 4—Meta-analysis odds ratios for supplemental oxygen/noninvasive ventilatory support (A), invasive mechanical ventilatory support (B), and in-hospital mortality by BMI category in patients with preexisting diabetes (C). Models were adjusted for age (,65, $65 years), sex (male/female), preexisting cardiovascular disease (yes/no), preexisting respiratory conditions (yes/no), and hypertension (yes/no). The reference BMI is in the normal range (i.e., $18 to ,25). The 95% CI of the odds ratios have not been adjusted for multiple testing and should not be used to infer definitive effects. Data from New York were not available for this subgroup analysis. D1L, DerSimonian and Laird random effectsmodel; FG&V,Franciscus Gasthuis& Vlietland; I-V, inverse-variance weighted fixed effects model.

disease, and has recently outlined that We acknowledge limitations of our Given this analysis involves patients individuals who are overweight may be at study. Data on socioeconomic status admitted to hospital with COVID-19 increased risk (40). Similarly, the most were not available, limiting the inter- only, we also were unable to assess recent guidelines from pretation of these findings, particularly whether patients with diabetes and England consider overweight and obesity because there may be important rela- obesity were more likely to experience asriskfactorsforsevereCOVID-19(41),in tionships among BMI, race/ethnicity, out-of-hospital death due to COVID-19 contrast to more conservative guidelines and socioeconomic status (44). Adjust- infection. These patients were not cap- from the UK National Health Service that ment for confounders including smok- tured in the data, and this may have suggested anincreased riskonly foraBMI ing and race/ethnicity was only possible resulted in an underestimation of over- of $40 (42). The World Health Organi- for five sites, with no difference in odds all mortality. At some sites, BMI was not zation now considers obesity a risk factor ratio observed. Supplemental oxygen consistently recorded during the study for severe COVID-19 disease (43). Incon- use varied; oxygen was administered period, which may have introduced sistent recommendations may impede to all hospitalized patients at a limited site-specific bias. Because of the rela- optimal patient care and compromise number of sites, affecting our ability to tively small numbers of patients at clear public health messaging. To our determine the influence of host comor- some sites, we were unable to stratify knowledge, there is currently no clinical bidities on this outcome. There may also BMI categories to include underweight guidance on the role of BMI in COVID- be varying and unmeasurable differen- (BMI ,18.5) or BMI .40, so we were 19 risk stratification of patients with ces in thresholds for escalating care unable to report specificoddsratiosfor diabetes. in those with overweight and obesity. these groups. We were unable to 8 Diabetes, Increased BMI, and COVID-19 Severity Diabetes Care

Figure 4dContinued.

differentiate between type 1 and type 2 19 surges in different countries. Given from these populations can be added to diabetes from the data available. Not- that improvements in clinical care did our ongoing analysis; potential collab- withstanding, the majority of patients not occur uniformly in all countries, orators are encouraged to contact the with diabetes included would be ex- however, we were unable to adjust corresponding authors. pected to have type 2 diabetes, given for this in our analysis. Finally, al- In conclusion, our findings high- the expected prepandemic relative though we enrolled multiple sites, light the importance of maintaining a prevalence (25). Moreover, type 1 di- our findings should not be considered healthy BMI, because patients with abetes has not been associated with regionally or globally representative either overweight or obesity are at increased severity of COVID-19 (45); and the study population was under- risk for severe COVID-19. Although re- therefore, we believe the findings for represented for low- and middle-in- ducing the current levels of overweight patients with diabetes predominantly come countries, which may limit and obesity is unlikely in the short term represent those with type 2 diabetes. It generalizability. to have an impact on the COVID-19 is important to note that the modest Notwithstanding, to our knowledge, pandemic, doing so may contribute to sample size of this study precludes pre- this study remains one of the largest reducing disease burden in future viral cise estimates of risk, particularly with multinational study to date on the pandemics (41,46). Furthermore, the respect to the associations among the risk factors associated with severe absence of an association between subgroup of patients with diabetes. COVID-19. Inclusion of individuals from overweight/obesity and COVID-19 se- We acknowledge that the number of low- and middle-income countries and verity among those with diabetes deaths likely decreased over the pe- disadvantaged or higher-risk populations should guide additional exploration riod of the study, which may have in such analyses is essential, and it is of mechanistic pathways and may in- altered results dependent on dates of hoped that as the pandemic progresses form risk stratification and appropriate data collection and the timing of COVID- and more data become available, data patient treatment. Finally, our findings care.diabetesjournals.org Longmore and Associates 9

Figure 4dContinued. may inform immunization prioritization St. Louis, Missouri, acknowledge Drs. Albert Lai The funders had no role in study design, data for higher-risk groups. and Randi Foraker of the Institute for Informatics collection, data analysis, data interpretation, or at Washington University School of Medicine. writing of the report. The main writing group acknowledges patients Duality of Interest. No potential conflicts of Acknowledgments. Authors from several sites and their families and health care providers interest relevant to this article were reported. wish to extend acknowledgments: Those at Aux- worldwide. Author Contributions. Authors had full access ilogico, Milan, Italy, acknowledge Drs Irene Campi, Funding. There was no specific project funding totheircorresponding site’sdata in the studyand Iacopo Chiodini, Luca Giovanelli, Giovanni Perego, for thestudy.Individualinvestigators were funded had final responsibility for the decision to submit Francesca Heilbron, Roberto Mene,` Andrea Cas- as follows: J.E.M. was supported by a fellowship for publication. D.K.L., E.M., Y.Z., K.R.S., and D.P.B. cella, Stefano Vicini, and all nurses; those at the from the DHB Foundation, Australia; S.B. is contributed to data interpretation and wrote University of California, Los Angeles, acknowledge supported by the Dutch Heart Foundation the first draft of the manuscript. J.E.M., S.B., and Dr. Paul C. Adamson; those in Cape Town, South (Dekker grant 2018-T028); E.M. is supported by C.S. performed the data analysis, contributed to Africa, acknowledge doctors and nurses working the World Health Organization Collaborating data interpretation, and reviewed the manu- in COVID-19 inpatient service; those in Indonesia Centre for Reference and Research on Influ- script. R.S., A.N., and G.C. contributed to the data thank the patients, doctors, nurses, pharmacists, enza, funded by the Australian Commonwealth interpretation and reviewed the manuscript. other health care workers, and research admin- Government, Department of Health. K.R.S. was J.E.M. and D.P.B. are the guarantors of this work istrators at all the participating sites. Authors in supported by the Australian Research Council (grant and, as such, had full access to all the data in the Ticino, Switzerland,acknowledgeLorenzoRuinelli; DE180100512); A.N. is supported by a Health study and take responsibility for the integrity of thoseinLausanne,Switzerland,acknowledgeOriol Research Board of Ireland Clinical Trail Network the data and the accuracy of the data analysis. Manuel, Desgranges Florian, Filippidis Paraskevas, award (grant CTN-2014-012); and D.P.B. was Kampouri Eleftheria-Evdokia, Tschopp Jonathan, supported by a National Health and Medical References and Viala Benjamin; those at Amphia, the Nether- Research Council Australian Investigator grant 1. World Health Organization. Coronovirus Dis- lands, acknowledge A.G. Loman, B.W. Driessen, (GTN1175744). Research at the Murdoch ease (COVID-19) Situation Report -162. Geneva, and Franciscus Gasthuis; those in Vlietland, Rotter- Children’s Research Institute is supported World Health Organization, 2020, p. 17 dam, the Netherlands, acknowledge Dr. Bianca M. by the Victorian Government’sOperational 2. Popkin BM, Du S, Green WD, et al. Individuals Boxma-de Klerk; those at Washington University, Infrastructure Support Program. with obesity and COVID-19: a global perspective 10 Diabetes, Increased BMI, and COVID-19 Severity Diabetes Care

on the epidemiology and biological relationships. 17. Statacorp. Stata Statistical Software: release 32. Benfield T, Jensen JS, Nordestgaard BG. Obes Rev 2020;21:e13128 16. College Station, TX, StataCorp LLC, 2019 Influence of diabetes and hyperglycaemia on 3. Selvin E, Juraschek SP. Diabetes epidemiology 18. Zaki N, Alashwal H, Ibrahim S. Association of infectious disease hospitalisation and outcome. in the COVID-19 pandemic. Diabetes Care 2020; hypertension, diabetes, stroke, , kidney Diabetologia 2007;50:549–554 43:1690–1694 disease, and high-cholesterol with COVID-19 33. Yamashita T, Umeda F, Hashimoto T, et al. 4. Rottoli M, Bernante P, Belvedere A, et al. How disease severity and fatality: a systematic review. Effect of glucose on Na, K-ATPase activity in importantisobesityasariskfactorforrespiratory Diabetes Metab Syndr 2020;14:1133–1142 cultured bovine aortic endothelial cells. Endo- failure, intensive care admission and death in 19. Williamson EJ, Walker AJ, Bhaskaran K, crinol Jpn 1992;39:1–7 hospitalised COVID-19 patients? Results from a et al. Factors associated with COVID-19-related 34. Rivelli JF, Amaiden MR, Monesterolo NE, single Italian centre. Eur J Endocrinol 2020;183: death using OpenSAFELY. Nature 2020;584:430– et al. High glucose levels induce inhibition of 389–397 436 Na,K-ATPase via stimulation of aldose reductase, 5. Simonnet A, Chetboun M, Poissy J, et al.; 20. Anderson MR, Geleris J, Anderson DR, et al. formation of microtubules and formation of an LICORN and the Lille COVID-19 and Obesity study and risk for intubation or death acetylated tubulin/Na,K-ATPase complex. Int J group. High prevalence of obesity in severe acute in SARS-CoV-2 infection: a retrospective cohort Biochem Cell Biol 2012;44:1203–1213 respiratory syndrome coronavirus-2 (SARS-CoV- study. Ann Intern Med 2020;173:782–790 35. Hulme KD, Yan L, Marshall RJ, et al. High 2) requiring invasive mechanical ventilation. 21. Klang E, Kassim G, Soffer S, Freeman R, Levin glucose levels increase influenza-associated dam- Obesity (Silver Spring) 2020;28:1195–1199 MA, Reich DL. Severe obesity as an independent age to the pulmonary epithelial-endothelial bar- 6. Halasz G, Leoni ML, Villani GQ, Nolli M, Villani risk factor for COVID-19 mortality in hospitalized rier. eLife 2020;9:e56907 M. Obesity, overweight and survival in critically ill patients younger than 50. Obesity (Silver Spring) 36. Zhu L, She Z-G, Cheng X, et al. Association of patients with SARS-CoV-2 pneumonia: is there an 2020;28:1595–1599 blood glucose control and outcomes in patients ? Preliminary results from Italy. 22. Watanabe M, Caruso D, Tuccinardi D, et al. with COVID-19 and pre-existing type 2 diabetes. Eur J Prev Cardiol. 7 July 2020 [Epub ahead of Visceral shows the strongest association with Cell Metab 2020;31:1068–1077.e3 print]. DOI: 10.1177/2047487320939675 theneedofintensivecareinpatientswithCOVID- 37. Coppelli A, Giannarelli R, Aragona M, et al.; 7. Hamer M, Gale CR, Kivimaki¨ M, Batty GD. 19. Metabolism 2020;111:154319 Pisa COVID-19 Study Group. Hyperglycemia at Overweight, obesity, and risk of hospitalization 23. Monteiro AC, Suri R, Emeruwa IO, et al. hospital admission is associated with severity of for COVID-19: a community-based cohort study Obesity and smoking as risk factors for invasive the prognosis in patients hospitalized for COVID- of adults in the United Kingdom. Proc Natl Acad mechanical ventilation in COVID-19: a retrospec- 19: the Pisa COVID-19 Study. Diabetes Care 2020; Sci U S A 2020;117:21011–21013 tive, observational cohort study. PLoS One 2020; 43:2345–2348 8. Tartof SY, Qian L, Hong V, et al. Obesity and 15:e0238552 38. Merzon E, Green I, Shpigelman M, et al. mortality among patients diagnosed with COVID- 24. Cariou B, Hadjadj S, Wargny M, et al.; CO- Haemoglobin A1c is a predictor of COVID-19 19: results from an integrated health care orga- RONADO Investigators. Phenotypic characteris- severity in patients with diabetes. Diabetes Metab nization. Ann Intern Med 2020;173:773–781 tics and prognosis of inpatients with COVID-19 Res Rev. 27 August 2020 [Epub ahead of print]. 9. Nakeshbandi M, Maini R, Daniel P, et al. The and diabetes: the CORONADO study. Diabeto- DOI: 10.1002/dmrr.3398 impact of obesity on COVID-19 complications: logia 2020;63:1500–1515 39. Public Health England. Adult Obesity and a retrospective cohort study. Int J Obes 2020;44: 25. Shi Q, Zhang X, Jiang F, et al. Clinical Type 2 Diabetes. London, Public Health England, 1832–1837 characteristics and risk factors for mortality 2014, pp. 39 10. Petrilli CM, Jones SA, Yang J, et al. Factors of COVID-19 patients with diabetes in Wuhan, 40. Centers for Disease Control and Prevention. associated with hospital admission and critical China: a two-center, retrospective study. Di- People with certain medical conditions, 2020. illness among 5279 people with coronavirus abetes Care 2020;43:1382–1391 Accessed26 October2020.Availablefrom https:// disease 2019 in New York City: prospective 26. Suganami T, Tanaka M, Ogawa Y. Molecular www.cdc.gov/coronavirus/2019-ncov/need- cohort study. BMJ 2020;369:m1966 mechanisms underlying obesity-induced chronic extra-precautions/people-with-medical-conditions 11. World Health Organization. Obesity and inflammation. In Chronic Inflammation: Mecha- .html overweight, 2020. Accessed 17 October 2020. nisms and Regulation. Miyasaka M, Takatsu K, 41. Blackshaw J, Feeley A, Mabbs L, et al. Excess Availablefromhttps://www.who.int/news-room/ Eds. Tokyo, Springer Japan, 2016, pp. 291–298 Weight and COVID-19: Insights from New Evi- fact-sheets/detail/obesity-and-overweight 27. Siegers JY, Novakovic B, Hulme KD, et al. A dence. London, Public Health England, 2020 12. World Health Organization. Prevalence of high-fat increases influenza A virus-associ- 42. National Health Service. Coronavirus (COVID- obesity among adults, BMI .5 30 (crude esti- ated cardiovascular damage. J Infect Dis 2020; 19): Shielded patients list, 2020. Accessed 13 mate) (%). 2017. Accessed 19 October 2020. 222:820–831 July 2020. Available from https://digital.nhs.uk/ Available from https://www.who.int/data/gho/ 28. SmithAG,SheridanPA, TsengRJ,SheridanJF, coronavirus/shielded-patient-list#risk-criteria data/indicators/indicator-details/GHO/prevalence- Beck MA. Selective impairment in dendritic cell 43. World Health Organization. COVID-19: vul- of-obesity-among-adults-bmi-5-30-(crude- function and altered antigen-specificCD81 T-cell nerable and high risk groups, 2020. Accessed estimate)-(-) responses in diet-induced obese mice infected 12 July 2020. Available from https://www.who 13. Smith KB, Smith MS. Obesity statistics. Prim with influenza virus. Immunology 2009;126:268– .int/westernpacific/emergencies/covid-19/ Care 2016;43:121–135, ix 279 information/high-risk-groups 14. World Health Organization Expert Consul- 29. Lockhart SM, O’Rahilly S. When two pan- 44. McLaren L. Socioeconomic status and obe- tation. Appropriate body-mass index for Asian demics meet: why is obesity associated with sity. Epidemiol Rev 2007;29:29–48 populations and its implications for policy and increasedCOVID-19mortality?Med(NY)2020;1: 45. Vangoitsenhoven R, Martens PJ, van Nes F, intervention strategies. Lancet 2004;363:157– 33–42 et al. No evidence of increased hospitalization 163 30. Paich HA, Sheridan PA, Handy J, et al. Over- rate for COVID-19 in community-dwelling pa- 15. Harris R, Bradburn M, Deeks J, Harbord R, weight and obese adult humans have a defective tients with . Diabetes Care 2020; Altman D, Sterne J. metan: Fixed- and random- cellular immune response to pandemic H1N1 43:e118–e119 effects meta-analysis. Stata J 2008;8:3–28 influenza A virus. Obesity (Silver Spring) 2013;21: 46. Short KR, Kedzierska K, van de Sandt CE. Back 16. RStudio. Integrated Development for R. 2377–2386 to the future: lessons learned from the 1918 in- RStudio, 2020. Accessed 3 September 2020. 31. Kassir R. Risk of COVID-19 for patients with fluenza pandemic. Front Cell Infect Microbiol Available from https://www.rstudio.com/ obesity. Obes Rev 2020;21:e13034 2018;8:343