Short-term Natural History of High-Risk Human Papillomavirus Infection in Mid-Adult Women Sampled Monthly (Short title: Short-term HPV Natural History in Mid-Adult Women)

Abstract

Characterizing short-term HPV detection patterns and viral load may inform HPV natural history in mid-adult women. From 2011–2012, we recruited women aged 30–50 years. Women submitted monthly self-collected vaginal samples for high-risk HPV DNA testing for 6 months. Positive samples were tested for type-specific HPV DNA load by real-time PCR. HPV type-adjusted linear and Poisson regression assessed factors associated with 1) viral load at initial HPV detection and 2) repeat type-specific HPV detection. One-hundred thirty-nine women (36% of 387 women with ≥4 samples) contributed 243 type-specific HR HPV infections during the study; 54% of infections were prevalent and 46% were incident. Incident (versus prevalent) detection and past pregnancy were associated with lower viral load, whereas current smoking was associated with higher viral load. In multivariate analysis, current smoking was associated with a 40% (95%CI:5%–87%) increase in the proportion of samples that were repeatedly positive for the same HPV type, whereas incident (versus prevalent) detection status and past pregnancy were each associated with a reduction in the proportion of samples repeatedly positive (55%,95%CI:38%–67% and 26%,95%CI:10%–39%, respectively). In a separate multivariate model, each log₁₀ increase in viral load was associated with a 10% (95%CI:4%–16%) increase in the proportion of samples repeatedly positive. Factors associated with repeat HPV detection were similar to those observed in longer-term studies, suggesting that short-term repeat detection may relate to long-term persistence. The negative associations between incident HPV detection and both viral load and repeat detection suggest that reactivation or intermittent persistence was more common than new acquisition.

INTRODUCTION

The natural history of human papillomavirus (HPV) infections among mid-adult women has been understudied.¹ Past HPV natural history studies have largely targeted women in their late teens and early 20s, the age group with the highest HPV prevalence and incidence.² Newly detected HPV infections among younger women are associated with reports of new sex partners, indicating new acquisition of the virus.² Conversely, newly detected infections among women in mid-adulthood may represent new acquisition, reactivation of a previous infection from a latent state, or intermittent persistent detection,³ rendering the origin of newly detected infections in this age group unclear. Clarifying the source of newly detected infections may inform whether prophylactic HPV vaccination (currently targeting females 9 to 26 years of age) would be warranted in this age group.¹

While the majority of HPV infections clear without developing clinical manifestations, repeated detection of high-risk (HR) HPV in young women is associated with increased risk of developing cervical neoplasia.⁴ In particular, long-term persistent HR HPV infection is necessary for developing cervical cancer.⁴ Short-term persistence (e.g. for one year) also has been shown to be associated with increased cervical neoplasia risk.^{5, 6} Therefore, understanding the rates of, and factors associated with, short-term persistent infections may be relevant to characterizing infections likely to have clinical significance in the long term. Closely followed cohorts of women with follow-up intervals of one to two weeks have shown that detection of HPV infection may be sporadic or intermittent even within a short period of time.^{7, 8} In addition, previous studies have shown positive correlations between levels of HPV DNA and risk of HPV persistence and carcinogenic progression,^9–15 but whether viral load correlates with frequency of repeat HPV detection over short sampling intervals has not been evaluated. Characterizing the relationship between viral load and repeated HR HPV detection using frequent sampling may aid in understanding both natural history and the origin of newly detected infections in this age group. Further assessments of the relationship between HPV viral load and other factors associated with repeat HPV detection may aid in explaining why fluctuations in HPV positivity tend to be observed in frequently followed cohorts.

We therefore evaluated the relationship between various factors and repeated HR HPV detection in a cohort of mid-adult women aged 30 to 50 years, sampled monthly for up to 6 months. Factors of interest included type-specific HPV viral load and timing of HPV detection (prevalent versus incident), as well as demographic, health, and behavioral characteristics. In addition, we also identified correlates of HPV viral loads.

MATERIAL AND METHODS

Study Population

Between March 2011 and January 2012, we recruited mid-adult women aged 30 to 50 years to participate in a longitudinal study of HPV infections. Enrollment was limited to female students, faculty, and staff at the University of Washington (UW) for convenience of follow-up. Women who were currently pregnant, ever had a hysterectomy, or had any serious medical conditions that would prohibit adherence to the study procedures were ineligible to participate. Women who would not be willing to self-collect vaginal samples or would be unavailable for follow-up procedures within 6 months after enrollment were also excluded. Women were recruited through flyers, advertisements, and letters distributed to students, faculty and staff at the UW. Interested women were screened for eligibility over the telephone, and those meeting the eligibility criteria were scheduled for an enrollment visit. The study coordinator administered informed consent at the enrollment visit. The protocol was reviewed and approved by the UW Institutional Review Board.

Data Collection

Enrollment and 6-month exit visits took place at the Hall Health Primary Care Center (HHPCC) of the UW. The enrollment visit was comprised of multiple components, in order: (1) face-to-face interview on cervical cancer screening and HPV vaccination history administered by the study coordinator, (2) vaginal self-sampling for HPV DNA testing (oral and fingernail samples for HPV DNA testing and a blood draw for HPV antibody testing were also collected, but not included in this analysis), (3) online self-administered survey on demographics, health status, pregnancy history, contraceptive use, hormone use, smoking habits, and sexual history, and (4) online self-administered sexual behavior diary covering the two-weeks prior to enrollment. Survey instruments were similar to those used in our previous HPV natural history studies.^{16, 17} During the 6-month follow-up, women were asked to self-collect additional vaginal samples each month (resulting in a potential maximum of 7 total samples per woman) and complete online sexual behavior diaries every two weeks. The exit visit included the same components as the enrollment visit except the questionnaires were modified to capture new information occurring since enrollment. Only the enrollment visit survey data were included in this analysis.

At the enrollment visit, the study coordinator verbally outlined the procedures for self-collecting vaginal samples and also provided illustrated written instructions.¹⁸ Self-collection kits contained two sterile Dacron swabs (to enhance sensitivity for HPV detection¹⁹), a covered tube containing 1.5 mL of Specimen Transport Medium (Qiagen, Gaithersburg, MD), and nitrile gloves.¹⁶ Women were sent monthly reminders to self-collect vaginal samples and return mail or hand-deliver them to HHPCC on the same day the sample was collected, and biweekly reminders to complete the sexual behavior diaries. Each week, batched vaginal samples were transported by courier from HHPCC to the laboratory, and either stored at −20°C or prepared for immediate testing.

HPV DNA Testing

Genomic DNA was isolated from self-collected vaginal specimens and HPV genotyped using the PCR-based Roche Linear Array assay (Roche Molecular Systems, Alameda, CA), which uses a β-globin control. K562 (negative for HPV) was included in each set for sample extraction to monitor for cross-contamination. The vaginal specimens (including swabs) were digested with 20 μg/mL proteinase K at 37°C for one hour to dislodge the cells, and DNA was isolated from 200 μL (about 12%) of the digested sample using the QIAamp DNA blood mini kit, following the protocol of the manufacturer (Qiagen, Cat. No.51104). The DNA was re-suspended in 100 μL of EB butter. The volume that was tested by the Roche Linear Array was 2 μL. Specimens negative for β-globin were considered insufficient, and specimens positive for β-globin but negative for HPV were considered negative for HPV infection. This analysis included 19 HR HPV types (16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82, IS39) considered to be carcinogenic, probably carcinogenic, or possibly carcinogenic.^{20, 21}

HPV Viral Load Testing

Viral load testing, or quantification of HPV DNA, was conducted on samples that tested positive for type-specific HR HPV by the Roche assay. Viral load testing was conducted for 16 HR HPV types, including HPV-16, 18, 31, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, and 82. Real-time PCR was used for simultaneous quantification of HPV E7 DNA and cellular DNA (β-actin). An optimized duplex reaction condition was used to ensure no interference between the detection of HPV E7 and cellular DNA. Each sample was assayed in triplicate. HPV type-specific primers/probes for the E7 gene were designed using Primer Express (Applied Biosystems) (Supplemental Table 1). The primers and probe for β-actin gene are commercially available (Applied Biosystems, Foster City, CA). The assay was set up in a reaction volume of 5 μl with the TaqMan® Universal PCR Master Mix kit (Applied Biosystems). Amplification was carried out on Applied Biosystems 7900 HT Sequence Detection System with a cycling program of holding at 50°C for 2 minutes and then at 95°C for 10 minutes followed by a two-step cycle of 15 seconds at 95°C and 1 minute at 60°C for 40 cycles. The known copy number of type-specific HPV genome and the known amount of cellular DNA were used as standards for absolute quantification. The number of viral copies was normalized according to the input amount of cellular DNA and expressed as the number of HPV copies per nanogram of cellular DNA. These corrected copy numbers were log₁₀ transformed, and a mean of the three measures was used for analysis. If one of the three measures was negative or differed by more than two standard deviations, the mean of the two remaining measures was used. Samples positive for β-actin but negative for HPV in two or more runs were assigned a value of 1 copy number per nanogram of cellular DNA if they tested positive by the Roche assay (n = 133/773 type-specific positive samples; 17.2%). For all statistical analyses using viral load data, sensitivity analyses were conducted excluding samples with undetectable HPV on viral load testing.

HPV Variant Characterization

Samples from women with the same HPV type detected in ≥2 samples separated by ≥1 intercurrent negative sample were selected for HPV variant characterization. In general, the first sample of the initially detected infection and the last sample of the subsequently redetected infection were selected and sequenced. If any selected sample had weak PCR genotyping results, it was replaced with the chronologically closest sample positive for the same HPV type (maintaining the criteria that both samples selected for variant characterization be separated by ≥1 intercurrent negative sample); if no such alternate sample was available, then the originally selected sample was sequenced.

HPV DNA fragments covering the 3′ part of the long control region and the entire E6 and E7 region were generated by PCR with a set of type-specific external primers (Supplemental Table 1).²² PCR products were then sequenced with a pair of external primers and a pair of internal primers using BigDye^™ Sequencing kit (Applied Biosystems, Foster City, CA).²³ The sequencing reaction was run from both directions. DNA sequences were analyzed using Sequencher^™ package (Gene Codes Corp., Ann Arbor, MI). A viral isolate in one patient sample was defined as a distinct variant if, as compared to the isolates detected in the other patient sample, there was one or more nucleotide alterations in the region analyzed.

Statistical Analyses

Samples were numbered according to the sample window during which the samples were received (Supplemental Figure 1). A sample window was defined as 14 to 45 days after the previous sample was received or a 30-day period after the end of the previous sample window if no sample was received. The sample window numbering system ensured that samples were spaced at least 14 days apart and accounts for the fact that samples were not always submitted on a monthly schedule. Additional samples that were received during the same sample window were dropped at random (n = 123/2755 samples dropped; 4.5%). Samples received ≥ 8 sample windows after the enrollment sample were excluded (n = 153/2755 samples excluded; 5.6%). As a result of these exclusions (and loss to follow-up), not all women analyzed had 7 available samples. Women who provided fewer than 4 self-collected vaginal samples were excluded from all analyses.

Type-specific detections were classified as prevalent if detected at enrollment and incident if detected during follow-up but negative at enrollment (Supplemental Figure 2). Redetection was defined as the first type-specific positive detection following a period of intercurrent negativity for the same type. Mean type-specific HPV viral loads were calculated separately for prevalent, incident, and redetected infections. Mean values were calculated for each individual HPV type and also calculated overall by pooling across all types.

We evaluated factors associated with type-specific HR HPV viral load (log₁₀ copies of HPV DNA per nanogram of cellular DNA) at the first positive detection by the Roche assay using linear regression with robust variance estimates to account for correlation within subjects. Analyses were adjusted for HPV type. Potential viral load risk factors assessed at enrollment included timing of initial detection (incident vs. prevalent), demographic characteristics, women’s health history (abnormal Pap test history, prior pregnancies, hormonal contraceptive use history, STD history), immunosuppression (defined as HIV positivity or currently taking immunosuppressive medications), smoking history, lifetime number of male sex partners, and recent sex with male partners. Variables for which p<0.10 by Wald test in the univariate analyses were included in a multivariate model.

To evaluate factors associated with repeat type-specific HR HPV detection over follow-up, we used Poisson regression to model the mean number of each woman’s follow-up samples that were repeatedly positive for the same HR HPV type. Each woman contributed an observation for each type detected during the study, with the exception that types first detected in the last available sample were excluded (because subsequent repeat detection could not be assessed). The number of type-specific HR HPV detections per woman was used as the outcome of interest, and the number of available samples per woman used as the offset variable (thus, we will describe the results in terms of the proportion of follow-up samples that are repeatedly positive). Robust variance estimates were used to account for correlation within subjects and deviations from the Poisson assumption. All models were adjusted for HPV type. The risk factors assessed were similar to those described above, with the addition of type-specific HPV viral load at the first positive sample by the Roche assay. Factors for which p<0.10 by Wald test in univariate analyses were included in multivariate models. Two multivariate models were constructed, including and excluding HPV viral load at first positive detection. Comparing the associations between potential health or sexual behavior factors and repeat HPV detection in the two models enabled an informal assessment of whether these associations are mediated by viral load. Sensitivity analyses were conducted restricting to a subset of women with at least 6 available samples.

RESULTS

Characteristics of the study population

Of the 409 women enrolled, 387 (94.6%) returned at least 4 self-collected samples and 345 (84.4%) returned at least 6 self-collected samples. Analyses included the 387 women who returned at least 4 self-collected vaginal samples (Figure 1). Of the 387 women, their mean age at enrollment was 38.4 (SD = 6.1) years (Table 1). The majority of women were white (79.1%), and approximately half were married (48.1%). The median lifetime number of male sex partners reported was 8 (interquartile range = 3 – 15), and 30.7% reported sex with at least one new male sex partner within the year prior to enrollment. Out of 2,432 vaginal samples collected within a valid sample window, 2 (0.1%) were insufficient for HPV DNA testing.

Figure 1.

Enrollment and follow-up status of mid-adult women in Seattle, WA (2011–2012)

Table 1.

Enrollment characteristics of 387^a mid-adult women in Seattle, WA (2011–2012)

Characteristics	Mean	(SD)
Age (years)	38.4	(6.1)
	Median	(IQR)
Age at first sexual intercourse (years)b	18	(16 – 21)
Lifetime number of male sex partners	8	(3 – 15)
	nc	(%)
Race
African American	9	(2.3)
Asian	45	(11.6)
White	306	(79.1)
Otherd	27	(7.0)
Current marital status
Unmarried, not living with partner	139	(36.1)
Unmarried, living with a partner	54	(14.0)
Married	185	(48.1)
Separated	7	(1.8)
Education
Some college or less	65	(16.8)
College bachelor’s degree	145	(37.5)
College master’s or doctoral degree	177	(45.7)
Ever had a non-HPV-related sexually transmitted diseasee
No	308	(80.0)
Yes	77	(20.0)
Ever had genital warts
No	343	(88.6)
Yes	44	(11.4)
Ever had an abnormal Pap test
No	218	(56.3)
Yes	169	(43.7)
Ever received ≥1 dose of HPV vaccine
No	357	(92.5)
Yes	29	(7.5)
Ever been pregnant
No	158	(40.8)
Yes	229	(59.2)
Currently using hormonal birth control methodsf
No	255	(65.9)
Yes	132	(34.1)
Currently have an immunosuppressive conditiong
No	379	(97.9)
Yes	8	(2.1)
Smoking statush
Never	287	(74.4)
Former	83	(21.5)
Current	16	(4.2)
Sexual activity within one year prior to enrollment
No sexual activity with male partners	55	(14.4)
Sex with non-new male partners only	209	(54.9)
Sex with ≥ 1 new male partner(s)	117	(30.7)

^aTwenty-two women with fewer than 4 samples were excluded.

^bRestricted to 376 women who have ever had sex with a male partner

^cNumbers may not add up to total due to missing data.

^dIncludes individuals indicating the following: American Indian/Alaska Native, Native Hawaiian/other Pacific Islander, other race, or multiple races

^eIncludes chlamydia, gonorrhea, genital herpes, and HIV

^fIncludes birth control pills, hormonal patches, vaginal rings, implanted contraception, injectable contraception, and hormonal intrauterine devices

^gIncludes HIV positivity (n=1) or currently taking immunosuppressive medications (n=7)

^hSmoking was defined as smoking at least one cigarette a day for one month or longer; former smokers reported ever smoking but not currently smoking, and current smokers reported currently smoking.

HR HPV detections at enrollment and over follow-up

At enrollment and over follow-up, 139 (35.9%) women tested positive for at least one HR HPV type. Overall, a total of 243 type-specific HR HPV infections were observed, with HPV types 53 (16.1%), 51 (14.4%), and 16 (8.6%) most commonly detected (Table 2). Of these infections, 132 (54.3%) were prevalently detected and 111 (45.7%) were incidently detected. Eighty-four type-specific infections (34.6%) were transiently detected (i.e. single-time positive), 62 (25.5%) were intermittently detected (i.e., at least one HPV-negative sample between two HPV-positive samples), and 97 (39.9%) were repeatedly detected without any intercurrent negative samples. To determine whether redetection of type-specific HPV DNA following a period of intercurrent negativity represented the same variant as the initial detection, we selected infections that were positive for one of the 12 types with available primers and probes for variant sequencing (45 of 62 intermittently detected infections, representing 90 individual samples). Thirteen (14.4%) of 90 samples tested HPV-negative by PCR-based DNA sequencing. As a result, 35 (77.8%) intermittently detected infections had two viable samples for variant comparison. One (2.9%) infection (HPV-16) displayed different HPV variants in the pre- and post-intercurrent negative samples. This HPV-16 infection (with a detection pattern of + − − − − − +) was therefore reclassified as transient (with a pattern of + − − − − − −) for all remaining analyses described below.

Table 2.

High-risk (HR) HPV detection at enrollment and during follow-up among mid-adult women in Seattle, WA (2011 – 2012) (N = 387)^a

	Type-specific HR HPV infections n (%)	Prevalent detectionsb n (%)	Incident detectionsc n (%)
All HR HPV types	243d	132	111
HPV-16	21 (8.6)	15 (11.4)	6 (5.4)
HPV-18	9 (3.7)	5 (3.8)	4 (3.6)
HPV-26	0 (0.0)	0 (0.0)	0 (0.0)
HPV-31	13 (5.4)	5 (3.8)	8 (7.2)
HPV-33	1 (0.4)	1 (0.8)	0 (0.0)
HPV-35	6 (2.5)	3 (2.3)	3 (2.7)
HPV-39	15 (6.2)	8 (6.1)	7 (6.3)
HPV-45	10 (4.1)	4 (3.0)	6 (5.4)
HPV-51	35 (14.4)	16 (12.1)	19 (17.1)
HPV-52	13 (5.4)	8 (6.1)	5 (4.5)
HPV-53	39 (16.1)	27 (20.5)	12 (10.8)
HPV-56	14 (5.8)	8 (6.1)	6 (5.4)
HPV-58	12 (4.9)	8 (6.1)	4 (3.6)
HPV-59	12 (4.9)	6 (4.6)	6 (5.4)
HPV-66	19 (7.8)	9 (6.8)	10 (9.0)
HPV-68	7 (2.9)	3 (2.3)	4 (3.6)
HPV-73	10 (4.1)	4 (3.0)	6 (5.4)
HPV-82	6 (2.5)	1 (0.8)	5 (4.5)
HPV-IS39	1 (0.4)	1 (0.8)	0 (0.0)

^aTwenty-two women with fewer than 4 samples were excluded.

^bType-specific HR HPV DNA detected at enrollment

^cType-specific HR HPV DNA detected during follow-up but not at enrollment

^dAmong 139 women with HR HPV detected

Quantification of HR HPV viral load in type-specific prevalent, incident, and redetected infection

Pooling across types, the mean log₁₀ HPV viral load per nanogram of cellular DNA measured at the time of first positive detection by the Roche assay was 2.12 (SD = 1.67) for prevalent detections and 0.70 (SD = 1.14) for incident detections. The mean viral load for incident detections was similar when restricting to first positive detections preceded by ≥2 (versus ≥1) negative samples (data not shown). For infections that were intermittently detected, the mean log₁₀ viral load per nanogram of cellular DNA measured at the time of first type-specific redetection was 0.90 (SD = 1.06). The mean viral loads for each individual HPV type showed consistent and similar differences between prevalent and incident detections (Supplemental Table 2). Mean viral load by HPV type is not presented for redetection due to the small number of redetections.

Factors associated with HR HPV viral load

In univariate analyses adjusted for HPV type, incident detection status, being married or living with a partner, and ever having been pregnant were negatively associated with HPV viral load while current smoking was positively associated with HPV viral load (Table 3). In multivariate analysis, incident detection status and current smoking remained independently associated with HPV viral load. Incident detection status (adjusted β = −1.30 log₁₀ HPV DNA copies per nanogram of cellular DNA; 95% CI: −1.76, −0.84) was associated with lower HPV viral load while current smoking was associated with higher HPV viral load (adjusted β = 0.70 log₁₀ HPV DNA copies per nanogram of cellular DNA; 95% CI: −0.03, 1.43). A sensitivity analysis excluding samples with undetectable viral load results generated similar findings, with slightly attenuated results observed for smoking status (data not shown).

Table 3.

Associations between various factors and mean type-specific log₁₀ HR HPV viral load (per nanogram of cellular DNA) at first positive detection among mid-adult women in Seattle, WA (2011 – 2012) (N = 241)^a

Model Term	n	Univariate Analysisb			Multivariate Analysisb
		β	(95% CI)	p-value	β	(95% CI)	p-value
Intercept	--	--	--	--	2.06	(1.26, 2.85)	<0.001**
HPV detection status
Prevalent	130	0.00	--	--	0.00	--	--
Incident	105	−1.42	(−1.88, −0.97)	<0.001**	−1.30	(−1.76, −0.84)	<0.001**
Age
≤ 40 years	165	0.00	--	--	--	--	--
> 40 years	76	−0.09	(−0.63, 0.45)	0.736	--	--	--
Marital status
Unmarried or separated	139	0.00	--	--	0.00	--	--
Married or living with partner	101	−0.46	(−0.99, 0.08)	0.095*	−0.20	(−0.68, 0.27)	0.401
Abnormal Pap before enrollment
No	107	0.00	--	--	--	--	--
Yes	134	−0.06	(−0.58, 0.47)	0.826	--	--	--
Ever had genital warts
No	217	0.00	--	--	--	--	--
Yes	24	−0.08	(−0.77, 0.60)	0.807	--	--	--
Ever been pregnant
No	123	0.00	--	--	0.00	--	--
Yes	118	−0.63	(−1.14, −0.12)	0.015**	−0.31	(−0.78, 0.15)	0.186
Currently using hormonal birth controlc
No	140	0.00	--	--	--	--	--
Yes	101	−0.21	(−0.69, 0.27)	0.396	--	--	--
Current immunosuppressive conditiond
No	233	0.00	--	--	--	--	--
Yes	8	−0.27	(−1.05, 0.52)	0.506	--	--	--
Smoking statuse
Never	165	0.00	--	--	0.00	--	--
Former	64	0.22	(−0.35, 0.79)	0.450	−0.08	(−0.58, 0.42)	0.747
Current	12	1.01	(0.23, 1.79)	0.012**	0.70	(−0.03, 1.43)	0.060*
Lifetime no. of male sex partnersf
0–2	10	0.00	--	--	--	--	--
3–6	48	−0.25	(−1.83, 1.33)	0.756	--	--	--
7–14	78	−0.30	(−1.82, 1.23)	0.700	--	--	--
15+	102	−0.54	(−2.02, 0.94)	0.472	--	--	--
Sexual activity within 1 year prior to enrollment
No sexual activity with male partners	18	0.00	--	--	--	--	--
Sex with non-new male partners only	130	0.24	(−0.55, 1.04)	0.549	--	--	--
Sex with ≥ 1 new male partner(s)	87	0.30	(−0.50, 1.10)	0.457	--	--	--

^*p-value < 0.1;

^**p-value < 0.05

^aHR HPV types tested for viral load include HPV-16, 18, 31, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82. Therefore, infections with HPV types 26, 33, and IS39 were excluded (n=2).

^bAll models were HPV type-adjusted.

^cIncludes birth control pills, hormonal patches, vaginal rings, implanted contraception, injectable contraception, and hormonal intrauterine devices

^dIncludes HIV positivity or currently taking immunosuppressive medications

^eSmoking was defined as smoking at least one cigarette a day for one month or longer; former smokers reported ever smoking but not currently smoking, and current smokers reported currently smoking.

^fLifetime number of male sex partners categorized based on approximate quartiles.

Factors associated with repeatedly detected HR HPV infections

Prevalently detected infections were more likely than incidently detected infections to be repeatedly detected. On average, the likelihood of a prevalent infection being repeatedly detected in any given follow-up sample was 67% (range: 60 – 75%) compared to 27% (range: 16 – 36%) for incident infections, with little variation over time (Figure 2). In univariate analyses, type-specific HR HPV viral load (at first positive detection by the Roche assay) and both former and current smoking were positively associated with the proportion of follow-up samples that were repeatedly positive for the same HR HPV type (Table 4). Incident detection status, a history of genital warts, and ever having been pregnant were negatively associated with the proportion of HR HPV detections. In the multivariate analysis without HPV viral load, incident detection status, pregnancy, and smoking status remained statistically significantly associated with the proportion of HR HPV detections. Incident detection status was associated with a 55% reduction in the proportion of HR HPV detections compared to prevalent detection status (adjusted RR = 0.45; 95% CI: 0.33, 0.62). Compared to never having been pregnant, a history of pregnancy was associated with a 26% reduction in the proportion of HR HPV detections over the course of follow-up (adjusted RR = 0.74; 95% CI: 0.61, 0.90). Compared to never smoking, former and current smoking were associated with 19% (RR = 1.19; 95% CI: 0.99, 1.42) and 40% (RR = 1.40; 95% CI: 1.05, 1.87) increases in the proportion of HR HPV detections, respectively. After adding HPV viral load at first positive detection to the model (to explore associations with repeat detection independent of potential effects on viral load), the effects of incident detection, pregnancy, and smoking were only slightly attenuated. In the same model, type-specific HPV viral load remained positively associated with the proportion of HR HPV detections after adjustment for other factors (adjusted RR = 1.10; 95% CI: 1.04, 1.16; per one log₁₀ increase in viral load). Neither lifetime number of male sex partners nor recent sexual activity with male partners was significantly associated with repeat detection of HR HPV. Results were similar when restricting to women with at least 6 available vaginal samples (data not shown). Similar findings were also found when excluding samples with undetectable viral load, with slightly attenuated results for both former and current smoking (data not shown).

Figure 2. Type-specific HR HPV positivity over time among mid-adult women in Seattle, WA (2011 – 2012).

Data are presented separately for prevalent (n=132) and incident (n=111) infections. The x-axis denotes the sample window. In this figure, sample window 0 denotes the initial type-specific positive sample; for prevalent infections, this is always the baseline sample, whereas for incident infections, this is always a subsequent sample. Sample windows 1 through 6 are numbered in relation to the first type-specific positive sample (e.g., sample window 1 denotes the first sample collected after the initial positive, sample window 2 denotes the second sample collected after the initial positive, etc). The y-axis denotes the proportion of samples positive for the specific HR HPV type at the given sample window, and the error bars represent 95% confidence intervals for each estimate (excluding sample window 0). For example, a prevalent infection with the pattern (+ − + − + + −) would contribute to the prevalent denominator at all sample windows, and to the numerator at windows 0, 2, 4, and 5; an incident infection with the pattern (− − − + − + +) would contribute to the incident denominator at sample windows 0, 1, 2, and 3 only, and to the numerator at windows 0, 2, and 3. The table depicts the numerators (n) and denominators (N) at each sample window for incident and prevalent infections. This figure includes HPV infections with 19 HR types (HPV-16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82 and IS39).

Table 4.

Association between various characteristics and proportion of type-specific HR HPV detections after first HPV-positive detection among HPV-positive mid-adult women in Seattle, WA (2011 – 2012) (N = 222)^a

Characteristics	n	Univariate Analysisb			Multivariate Analysis 1bc			Multivariate Analysis 2bc
		RR	(95%CI)	p-value	RR	(95%CI)	p-value	RR	(95%CI)	p-value
HPV viral load at 1st positive sampled	--	1.18	(1.13, 1.24)	<0.001**	--	--	--	1.10	(1.04, 1.16)	<0.001**
HPV detection status
Prevalent	130	1.00	--	--	1.00	--	--	1.00	--	--
Incident	92	0.39	(0.28, 0.53)	<0.001**	0.45	(0.33, 0.62)	<0.001**	0.52	(0.37, 0.72)	<0.001**
Age
≤ 40 years	151	1.00	--	--	--	--	--	--	--	--
> 40 years	71	1.00	(0.78, 1.27)	0.970	--	--	--	--	--	--
Marital status
Unmarried or separated	130	1.00	--	--	--	--	--	--	--	--
Married or living with a partner	91	0.86	(0.67, 1.11)	0.244	--	--	--	--	--	--
Abnormal Pap before enrollment
No	94	1.00	--	--	--	--	--	--	--	--
Yes	128	0.99	(0.79, 1.24)	0.903	--	--	--	--	--	--
Ever had genital warts
No	200	1.00	--	--	1.00	--	--	1.00	--	--
Yes	22	0.77	(0.60, 0.99)	0.044**	0.91	(0.67, 1.23)	0.523	0.90	(0.67, 1.21)	0.475
Ever been pregnant
No	114	1.00	--	--	1.00	--	--	1.00	--	--
Yes	108	0.63	(0.50, 0.79)	<0.001**	0.74	(0.61, 0.90)	0.006**	0.80	(0.66, 0.96)	0.019**
Currently using hormonal birth controle
No	127	1.00	--	--	--	--	--	--	--	--
Yes	95	1.02	(0.81, 1.28)	0.894	--	--	--	--	--	--
Current immunosuppressive conditionf
No	215	1.00	--	--	--	--	--	--	--	--
Yes	7	0.79	(0.49, 1.29)	0.348	--	--	--	--	--	--
Smoking statusg
Never	150	1.00	--	--	1.00	--	--	1.00	--	--
Former	61	1.36	(1.12, 1.66)	0.002**	1.19	(0.99, 1.42)	0.064*	1.21	(1.01, 1.44)	0.039**
Current	11	1.63	(1.21, 2.18)	0.001**	1.40	(1.05, 1.87)	0.021**	1.26	(0.97, 1.64)	0.085*
Lifetime no. of male sex partners
0–2	8	1.00	--	--	--	--	--	--	--	--
3–6	45	0.64	(0.63, 1.12)	0.119	--	--	--	--	--	--
7–14	71	0.81	(0.49, 1.33)	0.401	--	--	--	--	--	--
15+	95	0.68	(0.41, 1.11)	0.123	--	--	--	--	--	--
Sexual activity within 1 year prior to enrollment
No sexual activity with male partners	18	1.00	--	--	--	--	--	--	--	--
Sex with non-new male partners only	119	0.97	(0.68, 1.40)	0.878	--	--	--	--	--	--
Sex with ≥ 1 new male partner(s)	79	0.95	(0.64, 1.39)	0.774	--	--	--	--	--	--

^*p-value < 0.1;

^**p-value < 0.05.

^aNineteen single-time positives detected in the last available sample were excluded from this analysis. In addition, HPV types 26, 33, and IS39 (representing two infections) were dropped due to low numbers.

^bRR = risk ratios and were HPV type-adjusted.

^cMultivariate analyses included variables with p < 0.1 in the univariate analyses; model 1 did not include viral load as a covariate while model 2 did. Both models were restricted to samples with HPV viral load measurements available (N = 217).

^dHPV types tested for HPV viral load include HPV - 16, 18, 31, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82.

^eIncludes birth control pills, hormonal patches, vaginal rings, implanted contraception, injectable contraception, or hormonal intrauterine devices

^fIncludes HIV positivity or currently taking immunosuppressive medications

^gSmoking was defined as smoking at least one cigarette a day for one month or longer; former smokers reported ever smoking but not currently smoking, and current smokers reported currently smoking.

DISCUSSION

In this study of mid-adult women, we utilized monthly sampling intervals to characterize the short-term fluctuations in HPV detection over a 6-month follow-up. We sought to characterize the relationships between repeat HR HPV detection, type-specific viral load, and demographic, health, and behavioral characteristics.

During follow-up, we found that the majority of type-specific HR HPV infections were detected either transiently (35%) or intermittently (26%). Temporal detection of different type-specific variants within a woman was rare, suggesting that type-specific redetection after a short period of intercurrent negativity most likely represents intermittent persistent detection rather than new infection. Intermittent detection (or re-detection) of type-specific HPV has been described in previous studies, with varying frequency that tends to correlate with sampling frequency and length of follow-up.^{7, 8, 15, 24–27} It has been hypothesized that fluctuations in HPV detection could be caused by changes in HPV viral load or HPV DNA replication over time.^{8, 28} Positive associations between increased HPV viral load and persistent infection have been reported in some^{12–15, 29}, but not all^{30, 31} previous studies. However, to our knowledge HPV viral load has not been measured in studies using monthly or more frequent follow-up. In our study, we found that initial type-specific HPV viral load was positively associated with repeat detection over 6 months, which is similar to the associations observed in studies with longer sampling intervals.^{12–15, 29} These findings suggest that the relationship between HPV viral load and repeat detection may be similar for short-term and long-term persistent detection.

We observed notable differences between HR HPV infections that were first detected at enrollment (prevalent) versus at a follow-up visit (incident). First, prevalently detected infections were more likely than incidently detected infections to be repeatedly positive at subsequent follow-up visits. Second, significantly higher HPV viral load at first detection was observed among prevalent versus incident detections. This finding is consistent with our previous study that followed a cohort of mid-adult female online daters with triannual sampling for one year.¹⁵ As noted in that study, this result suggests that prevalent detections represent a mix of incident and persistent infections (and that persistent infections tend to be of greater viral load)^{12–14, 29} whereas newly detected infections among mid-adult women may represent a mix of new acquisition, reactivation of previous infections, or intermittently detected infections.¹⁵ We also observed that HPV viral loads in infections that were redetected after a period of intercurrent negativity (measured at the time of redetection) were lower than in prevalently detected infections but slightly higher than in incidently detected infections. In contrast, in our previous high-risk cohort of mid-adult women, HPV viral load was significantly higher for newly detected infections versus redetected infections.¹⁵ Although sampling intervals differed between these two studies, the results nonetheless suggest that the majority of incident detections in this lower-risk cohort likely represented reactivation of previous infection or intermittent persistent detection, whereas a higher proportion of incident detections in the previous higher-risk cohort were newly acquired infections.

Both former and current smoking were positively associated with repeated HR HPV detection in our study. Results from past studies on the relationship between smoking and the persistence of HR HPV infections have been equivocal.^{30, 32–35} Possible mechanisms through which smoking may facilitate acquisition or persistence of HPV may be the reduction of Langerhans cells and CD4 lymphocytes,³⁶ as well as the inactivation of natural killer cells.³⁷ Decreased immune responses in conjunction with higher HPV viral load may lead to increased risk of viral acquisition or persistence.³⁸ We also observed that current smoking, but not former smoking, was positively associated with HPV viral load. These findings are consistent with a previous study measuring HPV-16 and HPV-18 viral load, where the enrollment viral load was found to be greater among current smokers when compared with never smokers,³⁹ but inconsistent with our previous study among high-risk mid-adult women, where viral load was higher in never or former smokers compared with current smokers.¹⁵ A case control study conducted in Sweden also found a synergistic effect between current smoking, HPV-16 DNA positivity, and HPV-16 viral load on cervical cancer in situ development.⁴⁰ While we are unaware of any mediation analysis methods for correlated data, we did an informal comparison of models adjusted versus unadjusted for viral load. The point estimate for the association between current smoking and repeat HPV detection was only slightly attenuated after adjusting for viral load, suggesting that at least part of the effect of smoking on repeat HPV detection is via a pathway independent of an effect on HPV viral load.

A history of pregnancy was negatively associated with repeat HR HPV detection. Studies on the relationship between gravidity or parity and HPV infection have been equivocal.^{16, 41–46} Differences in sexual risk behavior profiles could explain the apparent protective effect of gravidity on repeat detection, whereby women who have been pregnant may be less prone to high-risk sexual behavior than women who have never been pregnant.^{16, 46} Newly detected infections among women with past pregnancies may therefore be more likely to represent reactivation of previously acquired infection (which tend to have lower viral load and are less likely to be repeatedly detected¹⁵) than newly acquired infection. We adjusted for age and sexual behavior in our study, but residual confounding may still have been present.

Several limitations of our study should be considered. First, our study population is a convenience sample of generally well-educated mid-adult women affiliated with the University of Washington, and therefore may not be representative of women in the general population or higher-risk groups of women. Within one year prior to enrollment, 14% were not sexually active and 55% did not engage in sexual activity with new sex partners. For a comparison, population based data from the 2006–2008 National Survey of Family Growth indicated that 8% of 25 to 44 year old women in the US were not sexually active in the past year.⁴⁷ However, the median lifetime number of male sex partners reported by the women in our cohort (8 partners) was higher than that reported by 25 to 44 year old women in the same national survey (3.6 partners),⁴⁷ though data from the National Health and Examination Surveys (1999–2012) show that the median number of lifetime partners increases across birth cohorts.⁴⁸ Second, the duration of follow-up in this study was only 6 months. Our study provides a snapshot of frequent sampling during a short period of time but may not inform long-term persistence of HR HPV. Third, HPV detection and viral load measurements may be dependent on sampling method and assay sensitivity. In addition, all behaviors were self-reported and may be prone to biases. However, we utilized a computer-assisted self-administered questionnaire for the majority of questions, which has been shown to minimize social desirability bias.⁴⁹ Finally, our Poisson regression analysis only accounts for the proportion of HPV positive detections but does not account for the patterns of infection. Our choice of analysis methods was supported by a previous study which found that long-term persistent infection was associated with the number of sporadic HPV detections rather than consecutive detections.⁵⁰

In conclusion, our study utilized frequent sampling to investigate the role of HPV viral load and other characteristics in the short-term natural history of HPV among mid-adult women. Our data suggest that while incident HPV detection likely represented a mixture of new acquisition, reactivation, and intermittent persistent detection, new acquisition was likely less common than reactivation or intermittent persistence in this mid-adult cohort. Therefore, prophylactic HPV vaccines may have limited benefit among average-risk mid-adult women due to the dominance of pre-existing infections. Furthermore, clinical counseling for this group of women should reflect the possibility of multiple sources of newly detected infection, with new acquisition only one of several possibilities. Future studies are needed to clarify the clinical significance of these different infection patterns.

Novelty and impact.

With monthly vaginal self-sampling, we characterized short-term patterns of detecting high-risk HPV in mid-adult women. Greater HPV viral load and smoking were each associated with an increased likelihood of repeat detection over a short interval, whereas incident (versus prevalent) detection and pregnancy history were negatively associated with repeat detection. These risk factors are similar to those previously observed for long-term persistence, suggesting that short-term persistence may inform the likelihood of long-term persistence.

Abbreviations used

HPV

human papillomavirus

HR

high-risk

UW

University of Washington

HHPCC

Hall Health Primary Care Center