Obesity and Human Papillomavirus Infection in Perimenopausal Women

Abstract

Introduction. Obesity is known to increase susceptibility to certain infections in men. It is unclear whether obesity increases women's risk for human papillomavirus (HPV) infection.

Methods. In a prospective cohort of 696 perimenopausal women enrolled in 2008–2012, we sought to determine whether obesity predicted incident HPV detection or nondetection. Obesity was defined as body mass index (BMI) ≥ 30 kg/m².

Results. Baseline any type HPV prevalence was comparable between obese and nonobese women (18.7% vs 19.1%; P > .05). Over a median follow-up period of 17.9 months (interquartile range: 12.1–24.5), 187 new HPV detections occurred among 123 women, 60 of whom subsequently lost 76 detectable infections. When compared with nonobese participants, obese women had a similar rate of new HPV detection (7.1 vs 7.8 infections per 1000 infection-years; P > .05) or loss of detection (100.3 vs 85.8 infections per 100 infection-years; P > .05). Similar results were found after adjusting for age, menopausal status, smoking habit, and sexual exposure history.

Conclusions. Results from the current analysis suggest little effect of obesity on HPV detection and loss of detection in mid-adult women. More research is needed to determine whether adipokines or cytokines better capture the potential immune modulating effects of obesity on HPV infection.

Over the past decade, an increased detection of human papillomavirus (HPV) infection in mid-adult women, typically aged 45 to 50 years, has been observed in some populations worldwide [1]. With the advent of effective prophylactic vaccines and highly sensitive HPV DNA testing techniques, origins of the second peak in age-specific prevalence and the corresponding clinical management have received much review and discussion [2, 3]. Evidence has suggested that altered host immunity may result in persistent infections or frequent reactivation from latent infections when women transition through menopause, thereby causing an increase in viral detection [4, 5]. In a nested case-control study, Garcia-Pineres et al found a reduced immune response in older women with persistent HPV infection [6]. In the same study population, Kemp et al later found a systemic elevation of inflammatory cytokines, such as interleukin (IL)-6, IL-8, and tumor necrosis factor-alpha (TNF-α), that are correlated with acute-phase response or innate immunity [7].

Basic and clinical research has shown that menopause is often accompanied by redistribution of adipose tissue from the periphery to the abdomen, hip, and thigh [8–10]. While central obesity has been associated with a proinflammatory state in postmenopausal women [11–13], adiposity has been linked to increased susceptibility to viral pathogens, such as herpes simplex virus-1 and -2 [14, 15] and adenovirus-36 [16]. In particular, Baker et al showed increased levels of adipokines in older women with persistent HPV infection [7, 17]. However, it is unknown whether obesity plays an independent role in the increased HPV DNA detection in perimenopausal women. Therefore, we sought to determine whether obesity is independently associated with incident detection of HPV DNA or the subsequent loss of detection among perimenopausal women enrolled in the Natural History of HPV in Peri-menopause (HIP) Study. We also investigated whether the association of obesity and HPV DNA detection varied by recent (prior 6 months) or lifetime sexual exposure to HPV infection.

METHODS

Study Population and Data Collection

We evaluated associations between obesity and HPV infection among women participating in the ongoing HIP study. Details regarding study design, participant recruitment, and interview and follow-up procedures have been described elsewhere [18]. Briefly, in 2008–2010, women aged 35–60 years attending routine obstetrics/gynecology visits were enrolled from 4 clinical sites in Baltimore, Maryland. Eligible women included those who were English speakers, had an intact uterus, and were able to provide informed consent and primary locator information. Current or planned pregnancy within the next 2 years or a history of HIV infection or organ transplantation excluded women from enrollment. Ethical approval for the primary study was obtained from the institutional review boards of the Johns Hopkins University Bloomberg School of Public Health.

After the baseline interview, women were followed every 6 months for 2 years. At each visit, an experienced interviewer collected detailed demographic information, reproductive and medical history, and information regarding sexual behaviors and medication use. In addition, a physician performed a cervical exam to collect cervical secretion specimens, vaginal swabs, and cervical swabs for HPV testing. Pap smears were only collected when indicated by clinical protocol. Viral DNA was extracted from the cervical swab and then genotyped using the Roche HPV linear array (LA) assay according to the manufacturer's instructions. The detection limit of the LA kit was 10–100 viral copies for each genotype tested. Each run of a polymerase chain reaction (PCR) test was comprised of 81 study specimens in a batch, including 4 positive controls (25 and 100 copies of HPV 16 and 18 plasmid DNA in a background of roughly 30 000 human cellular DNA equivalents); the extraction controls from the LA kit: HPV-DNA negative (containing 1.0 × 10⁶ K562 cells/mL), high HPV-DNA positive (1.25 × 10⁵ SiHa cells and 1.0 × 10⁶ K562 cells/mL); low HPV-DNA positive controls (5.0 × 10³ SiHa cells and 1.0 × 10⁶ K562 cells/mL) as well as 6 negative (no DNA) controls [19]. Results of a PCR batch were valid when all positive controls were positive for human β-globin and the intended PCR genotype (and no others) and the negative controls were all negative. The batch would be repeated if any deviation occurred. All runs in the current study were valid.

For the current analysis, we documented all 37 types of incident HPV detection and loss of detection for each participant during follow-up through February 2012. Incident detection was defined as a positive type-specific test result following a single negative test of the same genotype in the previous visit. Observed loss of detection was determined by a negative test result following a previous positive one. We investigated the effect of obesity as quantified by body mass index (BMI) at baseline categorized according to the World Health Organization classification system (obese: BMI ≥ 30 kg/m²) [20]. Waist-to-hip ratio and waist circumference (WC) have been suggested as better indices for central obesity than BMI in older populations [21]. However, we only considered baseline WC > 88 cm as an alternative in sensitivity analysis because we did not have hip measurements for women enrolled in the study. We treated both chronological age and reproductive age (by the STRAW [Stages of Reproductive Aging Workshop] criteria) at baseline as potential confounders for adjustment [22]. Additionally, other possible confounders that we considered included history of smoking (never, former, and current), use of oral contraceptive or hormonal replacement therapy, and race/ethnicity (white, African American, and others) as well as lifetime and recent (prior 6 months) sexual exposure history.

Statistical Analysis

We performed descriptive analysis comparing baseline characteristics and HPV status at baseline and at follow-up by obesity status. Assuming an event occurred between 2 return visits, we calculated a midinterval date for the occurrence of an incident event (detection or loss of detection) between a positive and a negative test, as is conventionally employed in interval-censored survival data [23]. We then estimated incidence rates and the corresponding 95% confidence intervals (CIs) for the first incident detection and its subsequent loss of detection of HPV DNA overall and by baseline characteristics. We used the jackknife method to account for the intraperson clustering of HPV testing results of all 37 genotypes while estimating these event rates.

Applying discrete-time survival analysis, we determined univariate associations between obesity measures and incident detection (or loss of detection) of HPV DNA. Considering the 6-month follow-up scheme of the HIP study, we employed complimentary log–log models with subject-specific random effects (shared frailty models) [24]. Normally distributed random effects (frailty terms) not only accounted for individual susceptibility to type-specific infections but also correlations among events of all genotypes recorded in the same woman [24]. For incident detection, the time origin was the baseline visit or the first visit with a negative test after a prevalent HPV DNA detection at baseline. For incident nondetection, the time origin was the baseline visit for women with prevalent HPV detection at enrollment. For women who were negative at baseline, risk for incident nondetection began at the time of the first positive test during follow-up. For each genotype, observations were censored from the risk sets at the time of the outcome, when lost to follow-up, or administratively on 6 February 2012. The time metric used was the estimated interval between the calculated event dates of detection and nondetection as conventionally employed [23].

In multivariable models, we retained factors that were hypothesized a priori as important confounders or that were univariately associated with HPV detection or its loss of detection. We constructed several models to infer the role of obesity in HPV detection and loss of detection separately. To determine whether the obesity–HPV association varied by levels of recent (last 6 months) or past (lifetime) sexual risk, we conducted stratified analysis by women's lifetime number of sex partners (≥5 vs <5) and by recent sex exposure (no sex vs had sex in last 6 months). Planned subgroup analyses also included stratification by race/ethnicity and by women's menopausal status. We compared the goodness of fit of various models by Akaike information criteria. We conducted all analysis using Stata 11 at a 2-sided significance level of 0.05 [25].

RESULTS

Population Characteristics

In the current analysis, we included 696 (78.6%) of 885 women in the HIP cohort; 22.4% of these 696 women completed 24 months of follow-up (Figure 1); overall, the median follow-up length was 17.9 months (interquartile range [IQR], 12.1–24.5). Table 1 shows that of these 696 individuals, the majority were aged 40–55 years (mean: 46.7; standard deviation [SD]: 6.7 years), predominantly white (75.1%), premenopausal (42.2%), and reported former use of oral contraceptive (68.4%). While obese women were more likely to be African American (31.3% vs 13.1%; P < .001), to have self-reported high blood pressure (42.4% vs 17.9%; P < .001), to have self-reported diabetes or to be actively taking medications for diabetes (12.6% vs 1.6%; P < .001), and to be abstinent in the prior 6 months (27.3% vs 18.1%; P = .007) compared with their nonobese counterparts, both groups were fairly comparable in age distribution, smoking and contraceptive use history, menopausal status, and past HPV testing experience as well as number of lifetime sex partners (Table 1).

Figure 1.

Study population and analytic sample. *Including women whose baseline body mass index was less than 18.5 (Kg/m²) (N = 3).

Table 1.

Comparison of Selected Characteristics by Obesity Status, Human Papillomavirus in Perimenopausal Women Cohort (N = 696), 2008–2012

	All		Nonobese (BMI: 18.5–29.9 kg/m2)		Obese (BMI: ≥ 30 kg/m2)
	N = 696 (100%)		N = 498 (71.6%)		N = 198 (28.4%)
Characteristic	N	%	N	%	N	%	P Value*
Age, mean (SD)	46.7	(6.7)	46.6	(6.9)	46.9	(6.5)	.528
35–40	132	19.0	101	20.3	31	15.7	.230
40–45	144	20.7	95	19.1	46	23.2
45–50	170	24.4	124	24.9	46	23.2
50–55	150	21.6	100	20.1	50	25.3
55–60	103	14.8	78	15.7	25	12.6
Race
White	523	75.1	395	79.3	128	64.6	<.001
Black	127	18.2	65	13.1	62	31.3
Other	46	6.6	38	7.6	8	4.0
Smoking history
Never	486	69.8	360	72.3	126	63.6	.077
Former	140	20.1	93	18.7	47	23.7
Current	70	10.1	45	9.0	25	12.6
Medical history
High blood pressure	173	24.9	89	17.9	84	42.4	<.001
Diabetes	33	4.7	8	1.6	25	12.6	<.001
Menopausal stage
Premenopause	294	42.2	218	43.8	76	38.4	.310
Perimenopause	212	30.5	144	28.9	68	34.3
Postmenopause	190	27.3	136	27.3	54	27.3
Contraceptive use
Never	61	8.8	43	8.6	18	9.1	.899
Former	476	68.4	339	68.1	137	69.2
Current	159	22.8	116	23.3	43	21.7
Ever having human papillomavirus test before enrollment
No	377	54.2	265	53.2	112	56.6	.173
Yes	213	30.6	162	32.5	51	25.8
Don't know	106	15.2	71	14.3	35	17.7
Lifetime sex partners
≥5	431	61.9	309	62.0	122	61.6	.916
Had sex within last 6 mo
No	144	20.7	90	18.1	54	27.3	.007
Yes	552	79.3	408	81.9	144	72.7
BMI
Mean (SD)	27.7	(6.3)	24.4	(2.9)	36.0	(4.7)	<.001
Median (IQR)	26.3	(22.9–31.3)	24.3	(22.1–26.8)	34.7	(32.6–39.0)	<.001
Range	(18.6–60.2)		(18.6–30.0)		(30.0–60.2)
Waist circumference	N = 617		N = 450		N = 167
Median (IQR)	23	(16–31)	19	(14–25)	38	(33–42)	<.001
>88 cm (35 in)	122	19.8	14	3.1	108	64.7	<.001

Abbreviations: BMI, body mass index; IQR, interquartile range; SD, standard deviation.

*P value for t test or rank sum test in comparing metric variables; for χ² or Fisher exact test in comparing categorical variables.

Table 2 summarizes women's HPV status at baseline and follow-up according to their obesity status. At baseline, 132 women (19.0%) were positive for at least 1 type of HPV, with 62 (8.9%) having 1 or more high-risk types. There were no differences in baseline prevalence for any-type (18.7% vs 19.1%; P = .906) or high-risk–type HPV (10.6% vs 8.2%; P = .321) between obese and nonobese women. Among 579 women with baseline HPV serology results available, 58.9% of obese women (n = 158) were positive, whereas 62.2% of nonobese women (n = 421) were seropositive (P = .458). At follow-up, 167 women (24.0%) were positive for at least 1 type of HPV, whereas 92 (13.2%) were positive for at least 1 high-risk type of HPV infection. Overall, women seemed to have an equal risk of HPV infection regardless of their obesity status (Table 2).

Table 2.

Human Papillomavirus Status at Baseline and at Follow-up by Obesity Status, Human Papillomavirus in Perimenopausal Women Cohort (N = 696), 2008–2012

	All		Nonobese (BMI: 18.5–29.9 kg/m2)		Obese (BMI: ≥ 30 kg/m2)
	N = 696		N = 498		N = 198
Human Papillomavirus Infection	N	%	N	%	N	%	P Value
Baseline
Any type positive	132	19.0	95	19.1	37	18.7	.906
High-risk type positive	62	8.9	41	8.2	21	10.6	.321
Among ever positive women at baseline	N = 132		N = 95		N = 37
Number of types positive, median (IQR)	1	(1–2)	1	(1–2)	1	(1–2)	.876
Number of high-risk types positive, median (IQR)	0	(0–1)	0	(0–1)	1	(0–1)	.214
Human papillomavirus seropositivity at baseline	N = 579		N = 421		N = 158
	355	61.3	262	62.2	93	58.9	.458
Follow-up
Any type positive	167	24.0	119	23.9	48	24.2	.923
High-risk type positive	92	13.2	60	12.0	32	16.2	.148
Among ever positive women at follow-up	N = 167		N = 119		N = 48
Number of types positive, median (IQR)	1	(1–2)	1	(1–2)	1	(1–2)	.834
Number of high-risk types positive, median (IQR)	1	(0–1)	0	(0–1)	1	(0–1)	.147

Abbreviations: BMI, body mass index; IQR, interquartile range

Predictors of HPV New Detection

During 25 600 infection-years of follow-up, 123 women had 195 type-specific new detections, with only 8 being redetections after at least 1 negative test in the previous visits. For the 187 first-incident detections, the overall incidence rate (IR) was 7.3 infections per 1000 infection-years (95% CI, 6.1–8.9). As shown in Table 3, obese women had a risk of incident HPV detection comparable to that of nonobese women at follow-up (IR, 7.8 vs 7.1 per 1000 infection-years). Likewise, women with central obesity (WC > 88 cm) had a 14% lower risk of new HPV detection than those with normal WC (crude HR [cHR], 0.86 [95% CI, .50–1.48]); however, this difference was not statistically significant (Table 3). While postmenopausal women had a 44% lower risk of incident HPV detection than their premenopausal counterparts (cHR, 0.56; P = .025), women who were current smokers (cHR, 1.86 [95% CI, 1.01–3.43]) or who were HPV-seropositive (cHR, 2.68 [95% CI, 1.61–4.47]) at baseline experienced a significantly higher likelihood of HPV infection over the study period. In addition, women with 5 or more lifetime sex partners were at a 3.5-fold increased risk for detection of new types of HPV as compared with those with 4 or fewer lifetime sex partners (cHR, 4.62 [95% CI, 2.79–7.64]). However, only a high number of lifetime sex partners (≥5) conferred a substantial risk of HPV infection (adjusted HR [aHR], 4.42 [95% CI, 2.66–7.35]) after adjusting for women's age, smoking history, menopausal stage, and recent sexual activity as well as obesity status, which was not statistically associated with HPV incidence (aHR, 1.07 [95% CI, .70–1.65]).

Table 3.

Incidence Rates of Human Papillomavirus and Results of Discrete Time Survival Analysis: Human Papillomavirus in Perimenopausal Women Cohort, 2008–2012

Characteristics	No. Events	Infection-Years	Incidence Rate	95% CI		Hazard Ratio	95% CI
Obesity (BMI ≥ 30 kg/m2)		(×1000)	(per 1000 infection-years)
No	129	18.2	7.10	5.69	8.98	REF
Yes	58	7.4	7.79	5.58	11.2	1.10	.71	1.70
Waist circumference >88 cma
No	135	18.1	7.48	5.98	9.47	REF
Yes	28	4.9	5.76	3.87	8.93	0.86	.50	1.48
Age						0.98	.95	1.01
35–40	40	5.0	7.93	5.31	12.4	REF
40–45	35	5.0	6.95	4.86	10.3	0.92	.50	1.72
45–50	59	6.5	9.14	6.45	13.4	1.15	.65	2.05
50–55	34	5.5	6.21	4.03	10.1	0.75	.40	1.40
55–60	19	3.6	5.29	2.85	10.9	0.59	.28	1.24
Race
White	137	20.0	6.87	5.48	8.71	REF
Black	38	4.1	9.26	6.62	13.3	1.46	.88	2.42
Other	12	1.5	7.77	3.62	19.5	1.14	.50	2.60
Smoking history
Never	123	18.2	6.74	5.33	8.65	REF
Former	37	5.1	7.25	4.92	11.1	1.13	.68	1.87
Current	27	2.2	12.0	7.65	19.8	1.86*	1.01	3.43
Medical history
Hypertension
No	140	19.3	7.24	5.84	9.08	REF
Yes	47	6.3	7.49	5.19	11.2	1.05	.66	1.66
Diabetes
No	180	24.6	7.32	6.05	8.94	REF
Yes	7	1.0	6.83	3.43	15.1	1.00	.38	2.65
Menopausal stage
Premenopause	101	10.5	9.64	7.42	12.7	REF
Perimenopause	48	8.6	5.60	4.05	7.97	0.63*	.39	1.00
Postmenopause	38	6.5	5.80	3.82	9.22	0.56*	.34	.93
Contraceptive use
Never	11	2.2	5.04	2.64	10.8	REF
Former	117	17.7	6.63	5.25	8.47	1.30	.59	2.84
Current	59	5.8	10.2	7.28	14.8	2.07	.90	4.76
Lifetime sex partners ≥5
No	24	10.3	2.33	1.49	3.87	REF
Yes	163	15.3	10.6	8.74	13.1	4.62**	2.79	7.64
Had sex within last 6 mo
No	37	5.0	7.46	4.93	11.8	REF
Yes	150	20.6	7.27	5.91	9.04	1.04	.63	1.71
Ever having HPV test before enrollment
No	99	15.1	6.54	5.11	8.49	REF
Yes	66	7.0	9.45	6.79	13.5	1.27	.82	1.97
Don't know	22	3.5	6.33	3.84	11.2	0.86	.47	1.59
HPV serostatus at baselineb
Negative	35	9.2	3.79	2.36	6.44	REF
Positive	121	13.1	9.21	7.36	11.7	2.68**	1.61	4.47

Abbreviations: BMI, body mass index; CI, confidence interval; HPV, human papillomavirus; REF, reference group with which comparison in risk of HPV incidence was made.

^a Excluded 79 women whose waist measurements were missing.

^b Excluded 117 women whose serological data were missing.

*P ≤ .05, **P < .001.

In subgroup analysis, BMI ≥ 30 kg/m² was not associated with incident HPV infection when women were stratified by race/ethnicity, menopausal status, or sexual risk (lifetime or recent). In the multivariable model adjusting for obesity and other factors, we found that the increased risk for HPV infection due to a higher number of lifetime sexual partners was particularly prominent in white participants (aHR, 6.24 [95% CI, 3.21–12.1]; P for interaction, .01). In comparison, for African American women and women of other race/ethnicity, the risk of incident HPV infection was not affected by their self-reported number of lifetime sexual partners.

Predictors of HPV Loss of Detection

Seventy-six events of incident HPV detection in 60 women were subsequently lost to detection (or cleared) over 79.8 infection-years of follow-up, resulting in a clearance rate of 95.2 (95% CI, 80.6–112.8) per 100 infection-years. Only 1 woman had a recurrent clearance of the same genotype while under observation. The median time from incident detection to its subsequent nondetection, or equivalently the duration of persistent detection, was 10.0 months (IQR, 8.9–11.6). Table 4 presents the incident rates and cHRs of loss of detection by women's baseline characteristics. Neither being obese (BMI ≥ 30 kg/m²) nor having a large waist (88 cm) was associated with loss of detection of HPV DNA in the bivariate (Table 4) or multivariable analysis after adjusting for age and smoking history as well as lifetime and recent sex exposure (data not shown). The only significant predictor for HPV loss of detection was self-reported history of smoking. Compared with those who never smoked, former smokers had a 55% reduction in their risk of loss of detection (cHR, 0.45; P = .017). This heightened risk of HPV persistent detection among former smokers changed little in the multivariable model adjusting for BMI, age, and past and recent sexual partnerships (aHR, 0.43 [95% CI, .22–084]).

Table 4.

Clearance Rates of Human Papillomavirus DNA and Results of Discrete Time Survival Analysis: Human Papillomavirus in Perimenopausal Women Cohort, 2008–2012

Characteristics	No. Events	Infection-Years	Clearance Rate	95% CI		Hazard Ratio	95% CI
Obesity (BMI ≥ 30 kg/m2)			(per 100 infection-years)
No	52	51.8	100.3	84.1	120.5	REF
Yes	24	28.0	85.8	60.6	121.5	0.85	.49	1.46
Waist circumference >88 cma
No	57	56.8	100.4	85.7	118.1	REF
Yes	16	13.3	120.2	80.3	179.9	1.41	.43	4.63
Age						1.00	.96	1.05
35–40	15	16.2	92.7	63.4	143.2	REF
40–45	15	15.4	97.1	61.8	154.3	1.14	.54	2.38
45–50	30	29.5	101.8	83.2	124.5	1.28	.68	2.42
50–55	8	11.9	67.5	32.5	148.0	0.70	.29	1.69
55–60	8	6.9	116.2	82.0	164.2	1.58	.65	3.9
Race
White	58	59.4	97.6	80.4	118.8	REF
Black	12	13.7	87.5	55.9	140.9	0.89	.45	1.77
Other	6	6.7	89.8	56.3	168.1	0.86	.33	2.24
Smoking history
Never	54	51.8	104.3	87.0	125.3	REF
Former	11	18.5	59.6	33.9	110.5	0.45*	.23	0.86
Current	11	9.6	114.8	86.0	153.9	1.34	.66	2.70
Medical history
Hypertension
No	55	61.4	89.6	74.4	108.5	REF
Yes	21	18.5	113.7	79.1	164.9	1.28	.70	2.34
Diabetes
No	74	77.7	95.2	80.4	113.2	REF
Yes	2	2.1	94.4	43.5	228.6	0.82	.19	3.61
Menopausal stage
Premenopause	42	38.8	108.3	88.8	132.4	REF
Perimenopause	23	24.6	93.4	68.5	128.3	0.85	.50	1.45
Postmenopause	11	16.4	67.1	40.3	114.9	0.55	.28	1.10
Contraceptive use
Never	5	5.5	91.1	59.9	137.9	REF
Former	43	51.2	83.9	67.5	104.8	0.87	.32	2.32
Current	28	23.1	121.2	92.6	159.9	1.29	.47	3.6
Lifetime sex partners ≥5
No	12	11.0	108.9	78.1	155.6	REF
Yes	64	68.8	93.0	77.2	112.5	0.64	.33	1.26
Recent sex within past 6 mo
No	13	17.1	75.9	50.2	113.8	REF
Yes	63	62.7	100.5	84.3	120.5	1.56	.84	2.88
Ever having HPV test before enrollment
No	39	38.7	84.3	69.6	102.4	REF
Yes	25	29.0	86.1	61.3	122.6	0.70	.41	1.19
Don't know	12	12.1	98.9	61.9	164.4	0.81	.41	1.61
HPV serostatus at baselineb
Negative	11	16.1	68.2	45.1	109.6	REF
Positive	58	53.2	108.9	89.9	132.3	2.10	.98	4.48

Abbreviations: BMI, body mass index; CI, confidence interval; HPV, human papillomavirus; REF, references group from with which comparison in risk of HPV incidence was made.

^a Excluded 3 women whose waist measurements were missing.

^b Excluded 7 women whose serological data were missing.

* P < .05.

DISCUSSION

In this cohort of mid-adult women, we did not find evidence supporting a role of obesity (as defined by BMI ≥ 30 kg/m²) in HPV detection. Among a subgroup of individuals whose waist measures were available for this analysis, central obesity (as defined by WC > 88 cm) did not appear to increase the risk for incident HPV detection or loss of detection. To our knowledge, this was one of the first investigations that prospectively examined associations between obesity and specific infections.

Early work has demonstrated that chronic low-grade inflammation may predispose obese individuals to various infections [16]. In a small group of healthy, middle-aged males (N = 74), Fernandez et al found a correlation between fat mass (as measured by bioelectrical impedance) and infections with multiple pathogens including HSV1, HSV2, enteroviruses, and Chlamydia pneumonia [15], yet it is unknown whether similar associations can be found in women. Recently, in another cross-sectional study, the authors reported that obesity (as defined by BMI 30–39.9 kg/m²) was strongly associated with HSV1 infection among a large representative population in the United States [14]. During the recent pandemic influenza (H1N1), clinical observations also recognized that obese individuals appeared to have a greater risk for influenza infection [26–28].

The increased susceptibility to infection in obese populations has been attributed to altered production of adipokines, such as leptin, adiponectin, TNF-α, IL-6, and others [16]. In particular, leptin may be a key element involved due to its potent proinflammatory properties, including activation of neutrophils and T lymphocytes, regulation of monocytes/macrophages, and cytokine secretion [16, 29]. While animal models have shown that leptin deficiency may lead to an increased susceptibility to infections [29], an early clinical report found that serum concentrations of leptin were positively correlated with BMI, suggesting that leptin resistance in obese subjects may similarly result in susceptibility to certain infections [30, 31]. Nevertheless, we were unable to determine the relationship between leptin and HPV infection due to the lack of leptin measurements in this study.

Despite the possibility of a truly null association between obesity and HPV infection, we reasoned that our null findings might have resulted from misclassification of the underlying infection status. Previous work has demonstrated that detectability of HPV DNA is highly variable when multiple swabs are repeatedly taken from the same women over intervals shorter than the 6- or 12-month intervals in conventional natural history studies [32–39]. By following women every 6 months, we ignored the interim variation in detectable HPV DNA and missed events of short-lived incident detection between 2 follow-up visits, resulting in an underestimated viral incidence. Similarly, we might have mislabeled 2 episodes of short-term infections separated by 1 negative test as a single event of long, persistent infection, resulting in an underestimated loss of detection. Moreover, anthropometric measures are relatively stable over a course of days or weeks, rendering them less favorable for a timely reflection of subtle changes associated with inflammatory and immune perturbations among obese subjects. Future research should explore the utility of biomarkers, such as adipokines and other inflammatory cytokines, in elucidating associations between obesity and specific infections.

Another plausible reason for the null association found in the current study is that the infection outcomes were inadequately defined in our analysis. Although animal models have demonstrated that HPV infection usually causes persistence, as do most DNA viruses [40–42], epidemiological studies have analyzed HPV infection as if it were acute infection that completely clears once the viral DNA is undetectable [3, 37]. As such, so-called incident HPV detection can be a mixture of new acquisition and redetection of a persistent or latent infection that merely fluctuates around the assay detection limit [3, 43]. However, the ability to distinguish type-specific new acquisitions from redetection of persistent infections has remained a major laboratory challenge in both research and clinical practice [3, 37]. Our ability to determine women's “true” infection status was limited by the fact that the LA assay we used could not make a distinction between established viral infection and transient deposition of viral particles from a recent sex act [44]. This may have contributed to the recurrent nondetection observed over the study period.

Another conceivable explanation for the null associations found may be an underrepresentation of obese women in the current study. However, this is unlikely when prevalence of obesity in the state of Maryland or in Baltimore is considered. According to the Center for Disease Control and Prevention's (CDC's) Behavioral Risk Factor Surveillance System report for 2010, the proportion of obese women in our study population (28.4% [95% CI, 25.1%–31.8%]) was similar to that for the state of Maryland (including Baltimore: 28.0% for both sexes [95% CI, 25.8%–30.1%]) and to that of Baltimore alone (29.2% for both sexes [95% CI, 23.9%–34.4%]) but was slightly higher than that of the District of Columbia metropolitan area (25.2% for both sexes [95% CI, 22.4%–27.9%]) [45]. On the other hand, the prevalence of obesity among African American women enrolled (48.8% [95% CI, 40.0%–57.6%]) was slightly higher than that in African American women of Maryland who responded to the CDC's survey in 2010 (38.1% [95% CI, 34.3%–41.9%]) [45]. In this regard, we might have been at risk for selection bias while performing subgroup analysis by race/ethnicity, though we did not find evidence supporting this conjecture.

Several limitations were noted while interpreting our findings. First, because we proposed to examine a non-a priori association in the HIP study, lack of sufficient power was a major concern while determining a small to moderate effect of obesity and the potential modifying effect of lifetime and recent sexual exposure on the association of obesity and HPV infection. In our ad hoc power calculation assuming 80% power, we would have required at least 454 and 863 events of nondetection in the obese and nonobese group, respectively, at follow-up if we were to detect an HR of 0.85 (as we actually observed) [46]. As such, we viewed our study as exploratory rather than confirmatory. Likewise, the prevalence of self-reported diabetes was low (4.7% overall; 12.6% in the obese vs 1.6% in the nonobese group), rendering the test for a possible differential effect of diabetes on the obesity–HPV association as nonsignificant.

Second, although the WHO classification for overweight and obesity based on BMI is a widely accepted anthropometric measure [20], it is well recognized that within each category, individual variation in the distribution of visceral adiposity and in the resultant metabolic changes does exist [21]. If obesity differentially affects other risk factors for detectable HPV infection, for example, number of lifetime sex partners or adherence to cervical cancer screening guidelines, an attenuated effect of obesity on HPV detection could be expected [47, 48]. We were also unable to evaluate the time-varying effect of weight change or weight cycling between follow-up visits, as well as the misclassification bias that might result. Last, as revealed in others' and our earlier work on the short-term variability in detectable HPV DNA [49, 50], the current study may fail to capture the interim fluctuations in HPV detection that are affected by hormonal changes in during a menstrual cycle.

In summary, we did not find evidence suggesting a role of obesity in HPV gain or loss of detection in the current analysis. Understanding the origins of the increased detection of HPV in mid-adult women unaccounted for by new sexual exposures [43] is essential for developing the next generation of prevention programs for cervical cancer. As a growing body of evidence has suggested that obesity may be involved in the pathogenesis of many infections, delineating the relationship between HPV infection and obesity, particularly in mid-adult women, may also help identify an at-risk population for HPV infection.