Clinical, Immunological, and Virological SARS-CoV-2 Phenotypes in Obese and Nonobese Military Health System Beneficiaries.

BACKGROUND: The mechanisms underlying the association between obesity and coronavirus disease 2019 (COVID-19) severity remain unclear. After verifying that obesity was a correlate of severe COVID-19 in US Military Health System (MHS) beneficiaries, we compared immunological and virological phenotypes of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection in both obese and nonobese participants.
METHODS: COVID-19-infected MHS beneficiaries were enrolled, and anthropometric, clinical, and demographic data were collected. We compared the SARS-CoV-2 peak IgG humoral response and reverse-transcription polymerase chain reaction viral load in obese and nonobese patients, stratified by hospitalization, utilizing logistic regression models.
RESULTS: Data from 511 COVID-19 patients were analyzed, among whom 24% were obese and 14% severely obese. Obesity was independently associated with hospitalization (adjusted odds ratio [aOR], 1.91; 95% confidence interval [CI], 1.15-3.18) and need for oxygen therapy (aOR, 3.39; 95% CI, 1.61-7.11). In outpatients, severely obese had a log10 (1.89) higher nucleocapsid (N1) genome equivalents (GE)/reaction and log10 (2.62) higher N2 GE/reaction than nonobese (P = 0.03 and P < .001, respectively). We noted a correlation between body mass index and peak anti-spike protein IgG in inpatients and outpatients (coefficient = 5.48, P < .001).
CONCLUSIONS: Obesity is a strong correlate of COVID-19 severity in MHS beneficiaries. These findings offer new pathophysiological insights into the relationship between obesity and COVID-19 severity.

(

Worldwide, nearly 114 million people have been infected by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), and more than 2.5 million people have died from coronavirus disease 2019 (COVID-19) [1]. Manifestations of COVID-19 vary widely, from asymptomatic infection to critical illness. Increased disease severity has been associated with age, race, and medical comorbidities such as diabetes, hypertension, asthma, chronic kidney disease, cancer, and neurological disease [2-9].

Many studies have reported that obesity is one of the strongest risk factors for severe COVID-19 and mortality. Early in the pandemic, investigators in Wuhan, China reported that each 1-unit increase in body mass index (BMI; kg/m2) was associated with a 12% increase in the risk of severe COVID-19, and obesity was associated with a 3-fold increase in risk of severe COVID-19 compared with nonobesity [10]. A French study reported that, among 124 patients admitted to an intensive care unit for COVID-19, 69% were obese or severely obese [11]. Obesity is associated with complex interactions between genetic, behavioral, metabolic, hormonal, and environmental influences. Obesity is known to be a risk factor for conditions associated with severe COVID-19 (eg, diabetes [12, 13], cardiovascular disease [14, 15], cancer [16-18], and other causes of mortality [19]). Determining the independent effect of obesity on COVID-19 severity requires consideration of these comorbidities, which may act as confounders or mediators of this relationship [11, 20–22].

The mechanisms of this association between obesity and severe COVID-19 remain unclear and are likely multifactorial. Obesity-related impairments in cardiovascular, respiratory, metabolic, and thrombotic pathways may decrease a given patient’s physiologic reserve and ability to recover from COVID-19 [23]. Obese individuals may have amplified or dysregulated immune responses that lead to greater viral replication, which may potentiate inflammatory immune responses [24-26]. In addition, obesity has been associated with a higher level of angiotensin-converting enzyme 2 (ACE2) receptor expression [27, 28], the membrane-bound host cell protein that mediates SARS-CoV-2 attachment and entry. Some posit that this difference in ACE2 receptor expression may potentially lead to higher SARS-CoV-2 viral loads among obese individuals [25, 26].

It remains unclear whether the association between obesity and COVID-19 severity is due to an individual’s putative increased viremia, an aberrant immune response, related comorbidities, or other, yet to be identified, risk factors. Therefore, we leveraged a prospective cohort of 511 Military Health System (MHS) beneficiaries with documented anthropometry and extensively characterized SARS-CoV-2 infection. Clinical and demographic data were collected through prospective methods, and for some participants anthropometry data were collected through routinely collected MHS electronic medical report (EMR). We evaluated whether obesity was independently associated with COVID-19 severity after adjusting for a wide range of covariates, particularly obesity-related comorbidities. Additionally, we compared the SARS-CoV-2 humoral responses in those with and without obesity, adjusted for clinical severity. Finally, we compared sampling-time–adjusted peak SARS-CoV-2 viral load by BMI strata. Collectively, this study clarifies the prognostic associations of obesity and SARS-CoV-2 infection outcomes in MHS beneficiaries, a population characterized by a high prevalence of obesity, and offers new insights into the possible mechanisms underlying the association of obesity and severe COVID-19 [29-31].

METHODS

Population and Setting

The Epidemiology, Immunology, and Clinical Characteristics of Emerging Infectious Diseases with Pandemic Potential (EPICC) study is a longitudinal cohort study conducted in 7 military treatment facilities (Brooke Army Medical Center, Fort Belvoir Community Hospital, Madigan Army Medical Center, Naval Medical Center Portsmouth, Naval Medical Center San Diego, Tripler Army Medical Center, and Walter Reed National Military Medical Center) [32]. Briefly, the EPICC study enrolled participants who were eligible for health care within the MHS and who met 1 of the following criteria: (1) confirmed SARS-CoV-2 infection; (2) COVID-like illness; or (3) high risk of developing COVID-19 due to a recent exposure. Demographic and clinical data were collected at enrollment and periodically during follow-up. Upper respiratory tract swabs were collected on a weekly basis (including day 0 and day 3) for inpatients during hospitalization (and day 14 if discharged), and on days 0 and 14 postenrollment for outpatients. Sera were collected at the same time points plus at 28 days and 6 months for inpatients, and 7 days, 28 days, and 6 months postenrollment for outpatients. Additional clinical records pertaining to hospitalization were reviewed in the MHS EMR. We excluded from the analysis those participants with negative SARS-CoV-2 reverse-transcription polymerase chain reaction (RT-PCR) tests (n = 101) and those for whom we did not have height and weight data. In addition, children (aged < 18 years) were excluded from the analysis due to the small number (n = 19) enrolled in the study and the qualitatively different categorization of obesity in pediatric populations (Supplementary Figure 1).

Diagnosis of SARS-CoV-2 Infection and Case Definition

SARS-CoV-2 infection was diagnosed by (1) SARS-CoV-2 RT-PCR of clinical specimens (results from swabs taken at the time of initial diagnosis, retrieved from the MHS electronic medical record and with a range of diagnostic platforms used); (2) SARS-CoV-2 RT-PCR on research specimens (nasopharyngeal, oropharyngeal, nasal, and rectal swabs) collected within 17 days after symptom onset [33-35]; and (3) serological testing, which detects resolving or prior SARS-CoV-2 infection by measuring the humoral immune response to the virus by targeting spike glycoprotein (S) in research serum samples collected within 35 days after symptom onset.

The SARS-CoV-2 RT-PCR assay used in this research is described in detail elsewhere [36]. Briefly, we utilized the SARS-CoV-2 (2019-nCoV) Centers for Disease Control and Prevention (CDC) quantitative PCR (qPCR) Probe Assay research use only kits (catalogue No. 10006770) manufactured by Integrated DNA Technologies, Inc. and consistent with the most recent revision of the emergency use authorization issued to the CDC on 1 December 2020. The assay targets 2 regions of the nucleocapsid (N) gene with an additional primer/probe set to detect the RNase P gene (RP) in specimens. A cycle threshold value of less than 40 for both N gene targets was considered positive for SARS-CoV-2 infection.

Serological testing was performed by a SARS-CoV-2 spike protein-based multiplex microsphere immunoassay, described in detail elsewhere [32]. For the detection of SARS-CoV-2 spike reactive immunoglobulin G (IgG) from serum samples collected 10–60 days after the onset of symptoms, this immunoassay has a sensitivity of 99% and specificity of 100%. Briefly, prefusion stabilized spike protein ectodomain trimers were coupled to magnetic microspheres (Bio-Rad). Samples for serology and spike-coupled microspheres were added to 96-well plates. Serum samples were diluted 1:400, added to 96-well plates, and tested in technical duplicates. Antibodies in serum samples were detected with biotinylated, cross-absorbed anti-human secondary antibodies. Antigen-antibody complexes were then incubated with streptavidin-phycoerythrin and quantified with a Bio-Plex 200 HTF multiplexing system (Bio-Rad). IgG levels are reported as median fluorescence intensity (MFI).

Measurement of Demographic, Clinical, and Anthropometric Subject Characteristics

The primary outcome of interest in this analysis was SARS-CoV-2 infection-related hospitalization. Hospitalization status was obtained from the case report forms (CRF) (completed at the site using the participant’s medical record) and directly using the MHS EMR. Because hospitalization may be a limited proxy of COVID-19 severity (eg, hospital admissions may be for nonmedical reasons and thresholds for admission may vary by provider and setting), we also used an alternative severity outcome based on whether subjects required treatment with supplemental oxygen. Anthropometric measures were obtained either through collection of height and weight during a prior medical visit or from the participant via survey. BMI was calculated using weight and height values, and individuals were classified as normal/underweight (<24.9 kg/m2), overweight (25–29.9 kg/m2), obese (30–34.9 kg/m2), and severely obese (≥35 kg/m2). In addition, the CRF assessed whether the subject had a diagnosis of obesity along with a range of comorbidities, many of which were possible confounders or mediators of the relationship between obesity and severe COVID-19 (Table 1). To measure the burden of comorbidities in a subject, we calculated the Charlson comorbidity index (CCI) [37], which is a standardized method of categorizing comorbidities of patients; for this study, we used the updated version age-adjusted CCI [38]. Age, sex, race, and ethnicity were reported by the participant.

Table 1.

Distribution of Demographic Data by Weight Strata in 511 SARS-CoV-2 Infections in Military Health System Beneficiaries

Characteristic	Normal (n = 140)	Overweight (n = 178)	Obese (n = 121)	Severely Obese (n = 72)	Total (n = 511)	P Valuea
Sex male	82 (58.6)	122 (68.5)	82 (67.8)	42 (58.3)	328 (64.2)	.16
Age group						<.01
18–44	106 (75.7)	120 (67.4)	66 (54.5)	25 (34.7)	317 (62.0)
45–64	24 (17.1)	39 (21.9)	38 (31.4)	32 (44.4)	133 (26.0)
65+	10 (7.1)	19 (10.7)	17 (14.0)	15 (20.8)	61 (11.9)
Race						<.01
Black	15 (10.7)	27 (15.2)	33 (27.3)	19 (26.8)	94 (18.4)
Hispanic	39 (27.9)	53 (29.8)	39 (32.2)	30 (42.3)	161 (31.6)
Others	17 (12.1)	23 (12.9)	9 (7.4)	3 (4.2)	52 (10.2)
White	69 (49.3)	75 (42.1)	40 (33.1)	19 (26.8)	203 (39.8)
Military status						<.01
Active military	89 (63.6)	111 (62.4)	49 (40.5)	14 (19.4)	263 (51.5)
Dependent	37 (26.4)	25 (14.0)	33 (27.3)	27 (37.5)	122 (23.9)
Retired military	14 (10.0)	42 (23.6)	39 (32.2)	31 (43.1)	126 (24.7)
DOD						.09
Air force	16 (11.4)	27 (15.2)	21 (17.4)	21 (29.2)	85 (16.6)
Army	49 (35.0)	70 (39.3)	45 (37.2)	26 (36.1)	190 (37.2)
Coast guard	3 (2.1)	1 (0.6)	1 (0.8)	0 (0.0)	5 (1.0)
Marines	12 (8.6)	21 (11.8)	11 (9.1)	1 (1.4)	45 (8.8)
Navy	56 (40.0)	56 (31.5)	42 (34.7)	23 (31.9)	177 (34.6)
Other	4 (2.9)	3 (1.7)	1 (0.8)	1 (1.4)	9 (1.8)
Site						<.01
BAMC	18 (12.9)	45 (25.3)	34 (28.1)	32 (44.4)	129 (25.2)
FBCH	6 (4.3)	9 (5.1)	1 (0.8)	5 (6.9)	21 (4.1)
MAMC	26 (18.6)	19 (10.7)	10 (8.3)	1 (1.4)	56 (11.0)
NMCP	3 (2.1)	1 (0.6)	1 (0.8)	0 (0.0)	5 (1.0)
NMCSD	35 (25.0)	47 (26.4)	29 (24.0)	14 (19.4)	125 (24.5)
TAMC	16 (11.4)	14 (7.9)	8 (6.6)	4 (5.6)	42 (8.2)
WRNMMC	36 (25.7)	43 (24.2)	38 (31.4)	16 (22.2)	133 (26.0)

Data are No. (%).

Abbreviations: BAMC, Brooke Army Medical Center; FBCH, Fort Belvoir Community Hospital; MAMC, Madigan Army Medical Center; NMCP, Naval Medical Center Portsmouth; NMCSD, Naval Medical Center San Diego; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; TAMC, Tripler Army Medical Center; WRNMMC, Walter Reed National Military Medical Center

aObesity category defined by body mass index; individuals were classified as normal/underweight (<24.9 kg/m2), overweight (25–29.9 kg/m2), obese (30–34.9 kg/m2), and severely obese (≥35 kg/m2).

bn × k Fisher exact test.

Quantitation of SARS-CoV-2 Viral Load and IgG Response

The viral load (genome equivalents [GE]/reaction) of each specimen was calculated for each N gene target (N1 and N2) by constructing plate specific standard curves from 3 dilutions of a known SARS-CoV-2 gene copy number standard (10 000, 1000, 100, and 10 copies/µL) on each RT-PCR assay run. The viral load calculations were log transformed for normalization such that a rise in log10 quantity GE/reaction equated to a rise in viral load (Supplementary Figure 2). We defined peak viral load as the highest log10 GE/reaction measured for the subject (from any time point or any specimen type). We used the SARS-CoV-2 spike protein-based multiplex microsphere immunoassay IgG MFI read-out as a quantitative measurement of anti-S IgG antibody against the SARS-CoV-2 spike protein [32].

Statistical Analysis

Descriptive statistics were calculated for the demographic characteristics, comorbidities, and illness severity by BMI category with P values computed using Fisher exact test. Univariate logistic regression was performed to evaluate whether COVID-19 severity was significantly associated with other independent variables, and then multivariable regression was performed, adjusting for other putative risk factors for COVID-19 including sex, age group, race, smoking history, and comorbidities (diabetes, hypertension, chronic kidney disease, asthma, chronic pulmonary disease, chronic neurological disorder, peripheral vascular disease, venous thromboembolism, ischemic heart disease, and cancer). Model fit was estimated by the Akaike information criterion and Bayesian information criterion, with the best fitting model used to present the adjusted odds ratio of obesity diagnosis and BMI strata on an outcome of hospitalization. This model was then fit to an alternative outcome of COVID-19 severity, as defined by requirement of supplemental oxygenation. We further carried out sensitivity analyses to minimize the risk of model misspecification by presenting comparative models to ensure that an estimate of the independent association of obesity and severe COVID-19 were robust.

Univariate and multivariate linear regression models were fit to evaluate whether peak viral load and/or peak anti-S IgG antibody response was associated with obesity and BMI strata (Supplementary Figure 3). These analyses aimed to determine whether obesity was associated with a difference in virological and humoral immune response phenotypes. These regression analyses were further stratified by disease severity (inpatient vs outpatient). All statistical analyses were conducted using R version 4.0.2 [39].

Ethics

This study was approved by the Uniformed Services University Human Research Protection Office under protocol IDCRP-085; all participants provided informed consent.

RESULTS

Obesity, Obesity-Associated Comorbidities, and Severe COVID-19 Outcomes Are Prevalent in Those With SARS-CoV-2 Infection in the US Military Health System

Among 619 COVID-19–positive participants who were enrolled in EPICC from 20 March 2020 through 15 September 2020, 511 (69%) were included in this analysis, as they were classified as COVID-19 cases and had anthropometric data available (Supplementary Figure 1). Over half of our study sample was male (64.2%), 18–44 years of age (62%), active-duty military (51.5%), and 48.6% were dependents and retired military; 72.6% were overweight, obese, and/or severely obese (Table 1). Overall, 25% of our participants were inpatients, incrementally from 14% to 52% in normal/underweight to severely obese participants, respectively (Supplementary Table 1). When considered as a continuous variable, average BMI values were higher in inpatients when compared with outpatients, and BMI was higher in those who received supplemental oxygen when compared with those who did not (Figure 1). Thirty-five percent of the participants had at least 1 other comorbidity (range 25%–68% in normal/underweight to severely obese participants, respectively). The most common additional comorbidity in the sample was hypertension (20.1%), followed by diabetes (12%), both of which were more common in increasing categories of obesity (Supplementary Table 1).

Figure 1.

Body mass index (BMI) distribution by severity, stratified into inpatient and outpatient (A), and medical oxygen requirements, stratified into oxygen supplement yes and oxygen supplement no, respectively (B). Statistically significant differences by nonparametric t tests are noted. Each dot represents a subject. Boxplots denote median, first quartile (25th percentile), and third quartile (75th percentile); statistical significance was determined by Wilcoxon rank sum test.

Obesity Is Independently Associated With Severe COVID-19 in US Military Health System Beneficiaries

We evaluated both reported and measured obesity in separate models. Logistic regression demonstrated that obesity was associated with an approximately 3-fold (odds ratio [OR] = 2.63; 95% confidence interval [CI], 1.72–4.02) increased odds of hospitalization, and remained significant after controlling for sex, age group, race, and a number of comorbidities (Table 2). Ordinal measured BMI categories were associated with increasing probability of hospitalization, although the ORs were statistically significant in the severely obese (OR = 3.1; 95% CI, 1.39–6.89) category only after multivariate adjustment. Similar results were observed when the need for supplemental oxygen was the outcome of interest (Supplementary Table 4).

Table 2.

Crude and Adjusted Association of Covariates With Hospitalization in 511 SARS-CoV-2 Infections in Military Health System Beneficiaries

Covariatesa	Unadjusted OR (95% CI)	P Value	aORb (95% CI)	P Value	aORc (95% CI)	P Value
Sex male	1.16 (.76–1.76)	.5	1.33 (.79–2.23)	.28	1.39 (.82–2.35)	.22
Age group 45–64	7 (4.19–11.72)	<.001	5.95 (3.46–10.25)	<.001	5.49 (3.17–9.52)	<.001
Age group 65+	27.93 (14.06–55.49)	<.001	22.53 (10.66–47.6)	<.001	22.19 (10.5–46.88)	<.001
Race Black	1.64 (.93–2.89)	.09	1.38 (.69–2.74)	.36	1.36 (.68–2.73)	.39
Race Hispanic	1.68 (1.03–2.73)	.04	1.73 (.96–3.14)	.07	1.68 (.92–3.06)	.09
Race Others	1.5 (.74–3.03)	.26	2.21 (.94–5.16)	.07	2.27 (.98–5.29)	.06
Obesity	2.63 (1.72–4.02)	<.001	1.91 (1.15–3.18)	.01
BMI class overweight	1.85 (1.01–3.36)	.04			1.39 (.68–2.81)	.37
BMI class obese	2.59 (1.39–4.83)	<.001			1.71 (.82–3.57)	.15
BMI class severely obese	6.02 (3.09–11.76)	<.001			3.1 (1.39–6.89)	.01
Ischemic heart disease	7.05 (2.81–17.71)	<.001	0.92 (.29–2.88)	.89	0.84 (.27–2.66)	.77
Chronic pulmonary disease	8.69 (3.54–21.33)	<.001	3.16 (1.09–9.2)	.03	3.36 (1.14–9.91)	.03
Chronic neurological disease	4.5 (1.68–12.09)	<.001	3.06 (.88–10.63)	.08	2.93 (.85–10.06)	.09

Obesity was defined by the medical diagnosis. BMI class was computed by participants height and weight.

Abbreviations: aOR, adjusted odds ratio; BMI, body mass index; CI, confidence interval; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

aOnly includes covariates included in multivariable models; referent: severity (inpatient vs. outpatient).

bAdjusted for obesity, sex, age groups, race, ischemic heart disease, chronic pulmonary disease, chronic neurological disease

cAdjusted for BMI category, sex, age groups, race, ischemic heart disease, chronic pulmonary disease, chronic neurological disease

Viral Load and Anti-S IgG Responses Correlate With Weight Strata

In upper respiratory tract specimens, we observed peaks in SARS-CoV-2 viral load at the time of or shortly after the onset of symptoms in the early stages of infection and decreasing in the convalescent stage (Figure 2). Conversely, we observed rising serum IgG soon after the onset of symptoms and plateauing thereafter, with persistence out to 3 months after symptom onset (Figure 3).

Figure 2.

Viral load as measured by qPCR N1 GE/reaction (A) and N2 GE/reaction (B), log10 transformed and plotted by symptom day and stratified by obesity status. Each dot represents a subject. Local polynomial regression curves were fit to nonobese and obese groups; 95% confidence intervals are shaded for nonobese and obese groups; statistical significance was determined by Wilcoxon rank sum test. Abbreviations: Ct, cycle threshold; GE, genome equivalent; N, nucleocapsid; qPCR, quantitative polymerase chain reaction.

Figure 3.

Anti-spike IgG MFI plotted by sampling day and stratified by obesity status. Each dot represents a subject. Local polynomial regression curves were fit to nonobese and obese groups; 95% confidence intervals are shaded for nonobese and obese groups; statistical significance was determined by Wilcoxon rank sum test. Abbreviations: IgG, immunoglobulin G; MFI, median fluorescence intensity.

When pooling inpatient and outpatient participants, we observed no association between viral load and BMI category (Table 3). However, when we stratified by COVID-19 severity (inpatient versus outpatients), outpatients in the severely obese group had a log10 1.89 higher N1 GE/reaction and a log10 2.62 N2 GE/reaction increase in peak viral load compared to those with normal weight, adjusting for sampling time (Supplementary Table 3). We also observed that obese or severely obese participants had a higher IgG antibody response compared to normal-weight participants, with an increasing coefficient by weight strata (Table 4). When stratified on inpatients and outpatients, there was an increased IgG response in severely obese subjects compared to normal-weight participants that was statistically significant (Table 4).

Table 3.

Association of Obesity Category With Peak Viral Load, Adjusted for Sampling Day

Covariates	β Coefficient (95% CI)	P Value	Adjusted β Coefficient (95% CI)a	P Value	Adjusted β Coefficient (95% CI)b	P Value
log10 N1 GE/reaction, inpatients and outpatients (n = 114)
Obesity	1.2 (.48 to 1.92)	<.001	0.72 (.04 to 1.41)	.04
BMI class overweight	1.13 (.09 to 2.16)	.03			0.72 (−.23 to 1.67)	.14
BMI class obese	0.96 (−.03 to 1.95)	.06			0.55 (−.37 to 1.47)	.24
BMI class severely obese	1.61 (.58 to 2.65)	<.001			0.83 (−.17 to 1.83)	.1
Symptom day	−0.07 (−.1 to −.05)	<.001	−0.06 (−.09 to −.04)	<.001	−0.07 (−.09 to −.04)	<.001
log10 N2 GE/reaction, inpatients and outpatients (n = 114)
Obesity	1.25 (.51 to 1.98)	<.001	0.81 (.11 to 1.51)	.02
BMI class overweight	1.21 (.15 to 2.27)	.03			0.94 (−.02 to 1.9)	.06
BMI class obese	1.08 (.07 to 2.1)	.04			0.79 (−.14 to 1.72)	.1
BMI class severely obese	1.67 (.61 to 2.72)	<.001			0.96 (−.05 to 1.97)	.06
Symptom day	−0.07 (−.1 to −.05)	<.001	−0.07 (−.09 to −.04)	<.001	−0.07 (−.09 to −.04)	<.001

Obesity was defined by the medical diagnosis. BMI class was computed by participants height and weight.

Abbreviations: BMI, body mass index; CI, confidence interval; GE, genome equivalent; N, nucleocapsid.

aModel contains obesity and sampling day.

bModel contains BMI classification and sampling day.

Table 4.

Association of Weight Strata With Peak Anti-Spike IgG MFI, Adjusted for Sampling Day and Stratified by Hospitalization Status

Covariates	Coefficient (95% CI)	P Value	Adjusted Coefficienta (95% CI)	P Value	Adjusted Coefficientb (95% CI)	P Value
Inpatients and outpatients (n = 314)
Obesity	4.09 (2.35 to 5.83)	<.001	4.06 (2.33 to 5.79)	<0.001
BMI class overweight	1.69 (−.54 to 3.92)	.14			1.56 (−.65 to 3.78)	.17
BMI class obese	3.18 (.68 to 5.67)	.01			3.21 (.73 to 5.7)	.01
BMI class severely obese	5.53 (2.64 to 8.41)	<.001			5.48 (2.61 to 8.34)	<.001
Symptom day	−0.02 (−.04 to 0)	.03	−0.02 (−.04 to 0)	0.03	−0.02 (−.04 to 0)	.03
Inpatients (n = 79)
Obesity	3.42 (1.19 to 5.65)	<.001	3.43 (1.24 to 5.61)	<.001
BMI class overweight	2.15 (−1.13 to 5.44)	.19			1.95 (−1.25 to 5.15)	.23
BMI class obese	3.31 (−.22 to 6.85)	.07			4.17 (.65 to 7.7)	.02
BMI class severely obese	3.97 (.54 to 7.4)	.02			3.53 (.17 to 6.89)	.04
Symptom day	−0.03 (−.07 to 0)	.04	−0.03 (−.07 to 0)	.03	−0.04 (−.07 to 0)	.03
Outpatients (n = 235)
Obesity	3.12 (.99 to 5.24)	<.001	3.13 (1 to 5.25)	<.001
BMI class overweight	0.98 (−1.62 to 3.58)	.46			0.92 (−1.68 to 3.52)	.49
BMI class obese	2.46 (−.49 to 5.41)	.1			2.42 (−.52 to 5.37)	.11
BMI class severely obese	3.91 (0 to 7.82)	.05			4.09 (.18 to 8)	.04
Symptom day	−.01 (−.03 to .01)	.27	−0.01 (−.03 to .01)	.25	−0.01 (−.03 to .01)	.22

Obesity was defined by the medical diagnosis. BMI class was computed by participants height and weight.

Abbreviations: BMI, body mass index; CI, confidence interval.

aModel contains obesity and sampling day.

bModel contains BMI classification and sampling day.

Discussion

The findings of our study corroborate those of previous studies; categories of higher BMI are associated with increased odds for severe COVID-19 [20-24]. Implicating obesity as a risk factor for severe COVID-19 can be challenging, given the multitude of obesity-associated comorbidities that may confound or mediate this association. To account for these, we conducted several sensitivity analyses to examine a range of obesity-associated comorbidities and to reduce the risk of model misspecification (Supplementary Table 2). Our analysis of BMI strata revealed a “dose-response” association between increasing BMI categories and odds of severe COVID-19, thereby strengthening causal inference. These findings are of concern to MHS beneficiaries given the known prevalence of obesity in US military beneficiaries, including active-duty service members who comprised a substantive proportion of our study population [29-31]. Indeed, among the study participants included in this analysis, we noted that over three-quarters of SARS-CoV-2 infections were in overweight MHS beneficiaries, and over one-third of subjects were obese or severely obese (Table 1). These findings also have relevance in other populations where obesity is prevalent.

To elucidate potential underlying mechanisms to which the association between obesity and severe COVID-19 might be attributed, we examined both virological and immunological characteristics of SARS-CoV-2 infection in the study population. In outpatients, we noted that obesity (BMI ≥ 30) was associated with a substantively higher viral load. Interestingly, this was not seen in inpatients. It is unknown whether this reflects antiviral drug use in inpatients or sampling of inpatient illness relatively later in their course of illness when viral loads may be expected to fall more precipitously. To date, there has been only limited other data that have reported a positive correlation between obesity and SARS-CoV-2 viral load. Maltezou et al described a higher viral load in obese versus nonobese patients but did not adjust for sampling time or disease severity and only used a dichotomous categorization of obesity [40]. The association between BMI and viral load has also been demonstrated for influenza [41, 42].

In inpatients, we noted an increase in anti-S IgG MFI by BMI strata (BMI ≥ 30, coefficient = 4.17, P = 0.02; BMI ≥ 35 coefficient = 3.53, P = 0.04). As higher levels of anti-S IgG have been described during hospitalization in patients with more severe disease [43, 44], this may reflect increasing severity in obese inpatients. However, a similar magnitude of effect was seen in severely obese outpatients (coefficient = 4.09, P = 0.04; Table 4). Further analyses leveraging this study population will examine whether nonneutralizing antibody, T-cell, and innate immune responses are quantitatively or qualitatively different in obese versus nonobese subjects, within and between strata of clinical severity.

Our study had several strengths. EPICC employs a comprehensive collection of demographic, epidemiologic, clinical, and laboratory data across the time course of infection. This enabled adjustment for important confounders and mediators. The prospective, longitudinal designed allowed ascertainment of peak viral load and anti-S IgG magnitude across illness time, as well as careful ascertainment of COVID-19 case status. Our sample was also relatively heterogeneous for age, sex, and race, as it draws from the population of MHS beneficiaries comprised of active-duty service members, retirees, and their dependents.

There were several limitations of our study. The paucity of early illness sampling was a limitation across this study, with the first study specimen being collected, on average, 34 (interquartile range, 20–54) days after symptom onset. While sampling time was adjusted in regression models this delay likely resulted in many subjects with resolving or resolved viral load. Another possible limitation was use of BMI as one set of criteria for obesity. We determined that our alternative methods of measuring obesity—BMI and a diagnosis of obesity as indicated in the EMR and/or CRF—were highly correlated. In total, 511 participants had both BMI and obesity reported on the EMR and CRF form respectively, among whom 404 (79%) were found to be obese using both data sources (Cohen κ = 0.58; P < .001). Nevertheless, elevated BMI in young muscular active-duty soldiers may be a result of increased muscle mass, and not an overweight status per se [45]. Future studies in this population could use waist circumference, neck circumference, and/or body fat percentage, which is an indicator of abdominal obesity and would resolve the possible misclassification [46]. However, it is important to note that such a misclassification of obesity would be expected to underestimate the magnitude of the association between obesity and severe COVID-19.

Although the current study examined humoral IgG responses in obesity and severe COVID-19, further virological and immunological exploration are necessary for a comprehensive understanding. Obesity is associated with chronic low-grade systemic inflammation, including higher levels of interleukin 6 (IL-6) [47, 48]. Investigation of inflammatory cytokine biomarkers (eg, IL-6, tumor necrosis factor-α [TNF-α], C-reactive protein [CRP]), adipokines, innate immunity such as natural killer cells, memory T cells, macrophages, and inflammasome signaling would allow us to understand complex pathophysiology of obesity and severe COVID-19 [47, 49]. Further studies are also needed to explore the role of obesity-associated insulin resistance and microbiome derangement in COVID-19 outcomes.

Other putative mechanisms that may explain the association between obesity and disease severity that were observed in this study could include excess soft tissue in the upper respiratory tract that results in obstruction of the airway [22], or decreased diaphragm contractility and poor pulmonary mechanics that result in hypoxia or contribute to type II respiratory failure, which may complicate the outcomes of some severe COVID-19 cases [23]. Such physiological pathways are more challenging to measure but could drive future studies in this [47] and other populations at risk of COVID-19.

In conclusion, we recapitulate the epidemiological association between obesity and severe COVID-19 seen in other studies, including using alternative definitions of severity. We further show an increase in SARS-CoV-2 viral load in outpatient COVID-19 with obesity, and a more robust anti-IgG response in obese SARS-CoV-2 infections, even when stratified to outpatients. These findings prompt further study into the mechanisms of severe COVID-19 in those with obesity, including examination of other facets of host response to SARS-CoV-2 such as innate and T-cell immunity.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.