Adenovirus 36 Antibodies Associated With Clinical Diagnosis of Overweight/Obesity but Not BMI Gain: A Military Cohort Study

Abstract

Context:

Obesity is a public health priority, which also threatens national security. Adenovirus 36 (Adv36) increases adiposity in animals and Adv36 antibody status is associated with human obesity, but it is unknown whether infection predicts the development of human adiposity.

Objective:

The objective of the study was to assess infection status and subsequent weight gain.

Design:

The study had a retrospective cohort design.

Setting:

The study was conducted at Air Force fitness testing and clinical encounters.

Participants:

Participants included Air Force male enlistees, aged 18–22 years, with a baseline body mass index (BMI) of 20–30 kg/m² followed up from enlistment (beginning in 1995) until 2012 or separation from the Air Force.

Exposure:

Exposure included Adv36 infection status at the time of entry.

Main Outcome Measure:

Follow-up BMI, the primary outcome, and diagnosis of overweight/obesity by the International Classification of Diseases, ninth revision V85.25+ and 278.0* series (secondary outcome) were recorded.

Results:

The last recorded follow-up BMI was similar among infected and uninfected, 26.4 and 27.2 kg/m², respectively (P > .05). However, infected individuals had a higher hazard of a medical provider's diagnosis of overweight/obese over time (hazard ratio 1.8, 95% confidence interval 1.0–3.1, P = .04), adjusted for baseline BMI. Additionally, infected individuals who were lean at baseline (BMI of 22.5 kg/m²) had a 3.9 times greater hazard of developing an overweight/obese clinical diagnosis (95% confidence interval 1.5–9.7, P = .004) compared with uninfected lean individuals after adjusting for interaction (P = .03) between infection and baseline BMI.

Conclusions:

The presence of Adv36 antibodies was not associated with higher BMI at baseline or follow-up within this military population. However, being infected was associated with developing a clinical diagnosis of overweight/obesity, especially among those lean at baseline.

Obesity has become a global epidemic with pervasive public health implications. For instance, the national security threat posed by the obesity epidemic has been described by the First Lady, the Institute of Medicine, academia, and retired military leaders (1–4). Specifically, excess body weight among service members and recruits is associated with lower physical performance, medical ineligibility for service, earlier separation, and billions in attributable Department of Defense (DoD) costs (5–8). Accordingly, DoD Directive 1308.1 instructs each military service to maintain professional health, fitness, and appearance standards for medical entry into the military (accession) and promotion, which includes body composition standards (9). These regulations provide unique incentive for military personnel to maintain a lean body composition.

Although often lumped together as a single disease, obesity has diverse phenotypes (subtypes of functional and clinical consequences) (10) and body mass index (BMI) does not always reflect adiposity. Adenovirus 36 (Adv36) infection is one of several potential explanations for these differences because it is known to alter adiposity and other metabolic outcomes in animals. Among humans, two meta-analyses of cross-sectional data have shown infected individuals have 1.6 [95% confidence interval (CI) 1.1–2.3] (11) or 1.9 (95% CI 1.0–3.6) (12) times higher odds of obesity than uninfected individuals. Meta-regression indicates the association is stronger in children (odds ratio 2.0, 95% CI 1.2–3.1) (11). Additionally, a longitudinal study demonstrated an association with higher body fat persisted for a decade in older Mexican Americans (13).

Although nonhuman primates gain weight after either natural or experimental Adv36 infection (14), no studies have demonstrated an association with new-onset adiposity in humans. This leaves open the possibility the cross-sectional associations were explained by people with obesity subsequently acquiring Adv36 infection. In this study, we evaluate body weight and new diagnoses of overweight/obesity over time by Adv36 antibody status at entry into the Air Force.

Materials and Methods

The study design was a retrospective cohort sampled from a range of prespecified time and baseline BMI quartiles. Serum samples (0.5 mL) are taken at time of medical in-processing into the Air Force for all recruits and stored in continuously frozen conditions at −30°C in the DoD Serum Repository (Silver Spring, Maryland). Service members were chosen based on inclusion criteria of active duty Air Force, male, aged 18–22 years, enlisting in 1995, 1999, 2001, 2003, and 2005 with accession BMI of 20–22.4, 22.5–24.9, 25–27.4, and 27.5–29.9 kg/m². For each of the BMI quartiles, 25 accession samples were selected for each year. Thus, there were 20 strata of BMI category by enlistment year combinations with n = 25, for a total of n = 500. Within each stratum, samples were randomly selected using numeric randomized sampling by the Armed Forces Health Surveillance Center (AFHSC).

AFHSC deidentified the selected samples with an alphanumeric code, and subsequently delivered the blinded samples to Obetech, LLC, who provided testing in duplicate using a previously published Adv36 ELISA test (15). The ELISA used a standard curve based on serum from a rabbit vaccinated with Adv36 that produces a high antibody titer. The cutoff for a positive test was at the 1:320 dilution of the positive control serum. Equivocal readings were declared by forced choice based on the average value of the duplicates.

After laboratory testing was completed by Obetech (by R.L.A.) and reported to the principal investigator (J.D.V.), AFHSC provided the deidentified coded baseline characteristics of each subject such as BMI and demographics, and the corresponding merged outcome data (International Classification of Diseases, ninth revision, codes, fitness test, and personnel records) taken from the Defense Medical Surveillance System and the Air Force Fitness Management System. The Defense Medical Surveillance System contains all diagnoses since 1996, and the Air Force Fitness Management System recorded fitness data since 1997. The study was identified as exempt by the Uniformed Services University Infectious Disease Institutional Review Board.

Data analysis was performed using Stata (version 12.0–13.0; StataCorp), and P < .05 was considered significant. The primary outcome was the final recorded BMI value across all fitness tests for each individual. Similarly, maximum BMI was assessed as the highest BMI from all the recorded tests over the time frame of the study as long as height was within 3 in. of mode height. Follow-up (final and maximum) BMI was available for 64% of the study population.

A new clinical diagnosis of overweight/obesity was evaluated as a secondary outcome. This outcome was based on clinical International Classification of Diseases, ninth revision, codes (V85.2-4, V85.53-4, and 278.00-02). For those who were lean at baseline, an incident case was any new overweight diagnosis, whereas those who were already overweight at baseline required an obese diagnosis (codes 278.00, 278.01, V85.30-4, V85.53-4) to be considered an incident case. The medical providers assigning these diagnoses were not aware of this study or the infection status.

Analysis of the dichotomous outcome used Cox proportional hazards survival analysis, with time beginning at recruitment and censoring at the time of separation and failure at the time of the first diagnosis. Cox regression with survey commands was used to account for the sampling design (primary sampling unit: each individual; strata: BMI category by entry year combinations; weight: inverse sampling probability based on aggregate recruit numbers obtained from the AFHSC). The same method was used for analyzing time to separation.

Specific dates beyond year were not disclosed to maintain subject confidentiality. AFHSC provided the specific number of days from the beginning of one's service (d 0) to the day of diagnosis or the day of separation. Person time for those who remained on active duty was based on the difference in years from year of entry to year data collection ended (2012). Robustness analysis was conducted using logistic regression, adjusting for year of entry and separation status as covariates (avoiding sensitivity to specific dates).

Results

The overall prevalence of Adv36 seropositivity at entry to military service within this sample was 20.8%. Overall demographics by serostatus were not different by age, race/ethnicity, BMI group, or entry year as presented in Table 1. The 500 subjects served an average of 5.2 person years before separation, study conclusion, or first-time clinical diagnosis of overweight/obesity.

Table 1.

Baseline Characteristics by Serostatus (Weighted Column Proportions)

	Adv36−	Adv36+	P Value (Weighted)
Total
500	396 (79.2%)	104 (20.8%)
Baseline age, y
18	98 (25.0%)	26 (24.5%)	.77 (.39)
19	130 (33.7%)	33 (31.0%)
20	88 (22.9%)	21 (21.0%)
21	54 (12.5%)	19 (19.1%)
22	26 (5.9%)	5 (4.5%)
Race/ethnicity
Asian	18 (4.5%)	2 (2.3%)	.63 (.53)
AA/Black	49 (11.7%)	12 (10.7%)
Hispanic	23 (6.2%)	7 (6.0%)
American Indian	1 (0.3%)	1 (1.0%)
Other	3 (0.4%)	2 (2.2%)
White	289 (73.9%)	78 (75.6%)
Unknown	13 (3.0%)	2 (2.2%)
BMI group, kg/m2
20.0–22.4	95 (29.7%)	30 (34.4%)	.45 (.63)
22.5–24.9	98 (31.1%)	27 (29.7%)
25.0–27.4	98 (28.9%)	27 (28.6%)
27.5–30.0	105 (10.4%)	20 (7.3%)
Entry year
1995	77 (18.2%)	23 (22.8%)	0.85 (0.61)
1999	78 (20.3%)	22 (22.0%)
2001	78 (20.1%)	22 (20.9%)
2003	80 (24.0%)	20 (21.3%)
2005	83 (17.4%)	17 (12.9%)

Abbreviation: AA, African American. Baseline characteristics by serostatus. Counts (of 500) are presented in rows, and in parentheses is the weighted proportion based on the inverse probability each subject's strata was selected (number of eligible recruits in strata divided by 25).

BMI, the primary outcome, was assessed during fitness testing. The final follow-up BMI averaged 26.4 kg/m² among infected and 27.2 kg/m² among uninfected individuals, which was not statistically different (by linear regression adjusting for baseline BMI and time in service β = −.13 kg/m², P = .71). Evaluation of maximum BMI showed similar nonsignificant differences between infected and uninfected recruits over time (27.2 kg/m² vs 28.1 kg/m², respectively; by linear regression β = −.12 kg/m², P = .70, adjusted for baseline BMI and time in service).

In contrast, the evaluation of a new diagnosis of overweight/obesity (our secondary outcome) did show statistically significant differences between Adv36 infected and uninfected individuals. A Cox proportional hazards model showed infected individuals had 1.8 times the hazard of receiving a new clinical diagnosis of overweight/obesity (P = .04) as compared with uninfected individuals when adjusting for baseline BMI as a covariate. The association between infection status and clinical diagnoses showed interaction with baseline BMI (P < .03). The interaction is visualized by the varying hazard ratios according to baseline BMI (Figure 1). The model is less reliable at the extremes of baseline BMI because there were no subjects with a baseline BMI at the exact values of 20.0 or of 30.0 kg/m². Therefore, we describe the hazard ratio for the midpoint of the lean (BMI 22.5 kg/m²) and the midpoint of the overweight (BMI 27.5 kg/m²). The infected individuals at a baseline BMI of 22.5 kg/m² had a 3.9-fold (P = .004) higher hazard of receiving a diagnosis of overweight/obesity as compared with uninfected individuals (who began at a BMI of 22.5 kg/m²), after adjusting for baseline BMI. These data demonstrate the overall association with incident clinical diagnoses was due to the higher ratio among the infected individuals who were leaner at baseline.

Figure 1.

Hazard ratio and 95% CI for the diagnosis of overweight/obesity among the infected (seropositive for Adv36) compared with the uninfected (seronegative for Adv36) by baseline BMI. The x-axis represents baseline BMI.

Additional analyses evaluated sensitivity to the modeling technique. Logistic regression adjusting for year of entry as a categorical variable, separation status as a dichotomous variable, and BMI showed nearly identical results as Cox regression. There was interaction by baseline BMI (P < .05) with this method of analysis as well, showing that lean individuals infected at baseline were more likely to be diagnosed with overweight/obesity over time than uninfected lean individuals (at baseline BMI 22.5 kg/m², odds ratio 3.9, 95% CI 1.39–11.0, P = .01).

Cox regression results were also similar when only analyzing the 64% of individuals with follow-up fitness data available. Assuming no interaction with baseline BMI, the hazard ratio among the infected was 1.9 (95% CI 1.1–3.4, P = .02) compared with uninfected individuals, adjusting for baseline BMI. Even in this smaller sample, the interaction between baseline BMI and infection status showed a similar pattern but was no longer significant (P = .09). At a baseline BMI of 22.5 kg/m², the infected individuals had 4.0 (95% CI 1.5–10.8, P = .006) times the hazard of diagnosis as compared with the uninfected individuals.

Discussion

We report the longest evaluation of natural Adv36 infection in humans. The Air Force population and settings (clinical and occupation) for outcome assessment were also novel. In contrast to our hypothesis, Adv36 infection was not associated with BMI at baseline, nor did it predict higher BMI at follow-up (our primary outcome). A previous cross-sectional study in a military population also showed no association between BMI and Adv36 serum neutralization (16). Thus, BMI associations with Adv36 appear to be different in the military and other populations. A military-specific result for BMI would not be unique to Adv36. For instance, sibutramine causes body fat and body weight loss in civilians, (17), but among soldiers it causes only the former due to higher lean mass retention (18). Because there are occupational repercussions for excess adiposity in the military, individuals have unique incentives to maintain a healthy body composition. In the Air Force, this is enforced by annual or semiannual fitness testing, which was the setting for the outcome BMI assessment in this study. Other data have shown that military members gain weight around the time of military separation (19). Because we did not have data on BMI after separation, we could not exclude the possibility that differential loss to follow-up contributed to the null BMI results.

Despite the similar BMI, those infected at military entry were more likely to be subsequently labeled with an overweight/obesity diagnosis by clinicians blinded to infection status. This association persisted when only considering those with available outcome BMI (indicating the difference between these outcomes cannot be explained by loss to follow-up alone). Due caution is appropriate in interpreting secondary outcomes, but these data suggest Adv36 serostatus might be investigated in military recruits to help predict the potential risk of clinical overweight/obesity diagnoses and allow appropriate preventive intervention.

Conclusions

We found Adv36 antibody did not predict BMI gain but did predict new-onset clinical diagnoses of overweight/obesity among Air Force enlistees, particularly among those who were lean at baseline. Because military personnel have a strong incentive to maintain a lean body composition, BMI may not be a suitable outcome variable for Adv36 studies. Additional studies are required to determine the magnitude of the problem of Adv36 and overweight/obesity within the military and among other populations.