Human papillomavirus (HPV) has been identified as the major cause of cervical cancer worldwide.1,2 It is thought that persistent infection with high-risk oncogenic strains of HPV, especially types 16,18, 31, 35 and 45, leads over many years to progressive cervical dysplasia and ultimately to invasive cancer.2 Human immunodeficiency virus (HIV)-infected women have an especially high burden of HPV-associated disease, including cervical and vulvar dysplasia.3–5 The high prevalence of HPV disease in HIV-infected women likely results from a combination of higher rates of incident HPV infection due to sexual risk factors and their decreased ability to clear established infections in the face of HIV-associated immunosuppression.
In animal models, papillomaviruses have been shown to establish a state of latent infection, followed by reactivation of viral replication along with reemergence of clinically apparent papillomas.6 Such a pattern of reactivation of HPV infection has also been characterized in human respiratory epithelia, manifesting as recurrent respiratory papillomatosis.7 Despite evidence that papillomaviruses establish a cycle of latency and reactivation in other settings, little is known about the natural history of HPV latency and reactivation in the human female genital tract. The purpose of this study is to estimate rates of and risk factors for HPV reactivation and recurrence among HIV-infected women and compare those rates to rates among HIV-uninfected women. Demonstration of HPV latency and reactivation in the cervix is a novel finding with implications for cervical cancer screening, prevention, and treatment.
MATERIALS AND METHODS
The HIV Epidemiology Research Study (HERS) is a prospective cohort of HIV-infected women and risk-matched HIV-uninfected women who were recruited and followed longitudinally starting in 1993.7 Baseline specimens from all women were tested for evidence of prior high-risk HPV infection using serology for HPV types 16, 18, 31, 35, and 45.8 We used baseline serology results, along with cervical HPV testing and longitudinal sexual behavior data, to detect the presence and calculate the incidence of high-risk HPV genital tract latency, reactivation, and recurrence in HIV-infected and HIV-uninfected women. Secondary outcomes included an analysis of risk factors for reactivation and recurrence of HPV infection in HIV-infected women.
As outlined in Figure 1, reactivation was defined as evidence of prior infection followed by cervical shedding of a given HPV type after at least two consecutive negative polymerase chain reaction (PCR) tests. The PCR results had to be negative on at least two consecutive visits to account for possible false-negative HPV PCR testing. Reactivation could occur only in patients reporting abstinence over the prior 12 months, as sexually acquired reinfection was assumed not to occur in this group. Recurrence was defined as HPV viral shedding detected by PCR in patients with prior infection and negative PCR tests on at least two prior consecutive visits. Recurrent viral shedding refers to positive PCR testing in sexually active patients with one or more male partners over the prior 12 months because it is not possible to distinguish between reactivation and reinfection in this group.
Eight hundred seventy-one HIV-infected women and 439 women in the control group risk-matched for intravenous drug use and high-risk sexual behavior were recruited in four U.S. cities (Baltimore, MD; Bronx, NY; Providence, RI; and Detroit, MI) between April 1993 and January 1995.7 They were recruited from physicians' offices, community centers, drug treatment centers, and by word of mouth. Human immunodeficiency virus–infected and HIV-unifected women were recruited over the same time period. Women were categorized as injection-drug users if they were currently injecting or ever had injected drugs at screening. High-risk sexual behavior was defined as either having sex with five or more partners in the last year, ever having sex with a male intravenous drug use, ever exchanging sex for money or drugs, or ever having sex with a male who was known to be or suspected of being infected with HIV. Detailed recruitment and eligibility data have been published previously.9 Women with a history of an acquired immunodeficiency syndrome (AIDS)-defining illness were excluded from this prospective study. The study protocol was approved by the institutional review board at each participating site and the Centers for Disease Control and Prevention, and informed consent was obtained from the women.
Women were followed with physical examinations, laboratory testing, and structured interviews at 6-month intervals for up to 6 years. The acceptable time window for sampling was 14 days from initial enrollment for the first visit, and ±42 days from the 6-month target date for all future visits. Baseline serology for type-specific immunoglobulin G–recognizing HPV-like particles has been described previously and was performed using specimens collected at the initial visit.8 Immunoglobulin G titer was categorized as negative or reactive by enzyme-linked immunosorbant assay.8 At each visit, Pap tests and cervicovaginal lavage specimens were obtained, and PCR testing for infection with HPV using generic (not type-specific) probes was carried out on frozen cervicovaginal lavage samples, as detailed previously.7 Human papillomavirus typing was then performed by dot blotting the PCR products using type-specific probes.7 The PCR testing is highly sensitive and specific, and because it is the gold standard for detection, false-positive and false-negative rates cannot be estimated. Patients had HIV viral load testing at each visit using a third-generation branched DNA signal amplification assay (Chiron Corp, Emeryville, CA) with a lower quantification limit of 50 copies/mL, and CD4 counts were measured by flow cytometry at each visit. Sexual history was recorded at each visit, detailing the number of male and female sexual partners, frequency of sexual intercourse, and contraceptive methods used since the last study visit.7 The structured core interview was conducted by trained interviewers using the same standardized forms at each site. Use of antiretroviral medication was categorized as none, highly active antiretroviral therapy (HAART), or suboptimal therapy. Women categorized as receiving HAART met Department of Health and Human Services guidelines for HAART, and those categorized as receiving suboptimal therapy received either monotherapy or combination therapy not qualifying as HAART.10
The analysis included 898 women. Women seropositive for HPV type 16, 18, 31, 33, or 45 antigens on enzyme-linked immunosorbent assay at enrollment were included in the analysis. Additionally, those who were serologically negative at enrollment but then acquired and cleared an HPV infection during the course of the study were also eligible for inclusion in the analysis. Other inclusion criteria included having an intact cervix, attending at least three consecutive study visits with HPV testing, and for HIV-uninfected women, remaining HIV-uninfected throughout follow-up.
Human papillomvirus results were missing for approximately 10% (1,116) of the study visits in the dataset. If a visit was missed but the woman attended the previous and next visits, or a woman's final visit was missed but she attended the previous visit, values for relevant variables were imputed. Multiple imputation11 was implemented 10 times with the multiple imputation procedure in SAS version 9.1 (SAS Inc, Cary, NC), which uses a Markov chain Monte Carlo procedure to create 10 analysis data sets. Values for missing variables (both covariates and HPV status) were imputed using known values for HIV status, age, race, intravenous drug use risk at enrollment, and study site. Missing values for viral load and CD4 count were imputed based on regression models using Gaussian errors with log transformations, while other covariates were imputed from logistic regression models. The imputation models used the values at previous and next visits or at the previous two visits if the missing value was at the last visit. Values were imputed for 613 (6.0%) of 10,296 total woman-visits. The SAS procedure MIANALYZE was used to combine the results from the analysis of each of the 10 data sets. The statistic of interest was estimated as the mean of this statistic from the 10 analyses; its variance was estimated using the within and between analysis variances.11 Fractional numbers of events reflect variation in the numbers of events in the 10 analysis data sets.
Initially, baseline demographic and health characteristics were compared between HIV-infected and HIV-uninfected women. For all women seropositive for HPV type 16, 18, 31, 33, or 45 at study enrollment, HPV serology type by HIV status was compared. Differential distributions of variables by HIV status were assessed using the Pearson χ2.
Proportional hazards models with repeated measures12 were implemented using the PHREG procedure in SAS version 9.1 to compute unadjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for reactivation and recurrence by potential risk factors. Separate models were run for reactivation and recurrence by HPV type and risk factor. Due to the low rates of reactivation and recurrence, adjusting for covariates was not possible. Univariable analysis of factors associated with reactivation of individual HPV types was limited to HIV-infected women because of the small number of reactivation events (0–0.1 per 100 woman-years) in HIV-uninfected women. Hazard ratios for individual risk factors were not calculated for HPV types 31 and 35 because of the low number of events. Finally, rates of total HPV shedding (recurrence and reactivation) during follow-up were assessed by HIV status, and rate ratios were calculated comparing HIV-infected women with HIV-uninfected women.
A total of 1,310 women were enrolled in HERS. We excluded from this analysis 11 (0.8%) women who experienced HIV seroconversion during the study, 48 (3.7%) with missing baseline serologic data, and 172 (13.1%) with inadequate follow-up. Another 65 (5.0%) women with no HPV infection detected by serology or PCR at baseline and throughout follow-up and 111 (8.5%) women without a cervix were also excluded. Therefore, the final analysis included a total of 898 women, 634 (70.6%) HIV-infected women and 264 (29.4%) HIV-uninfected women. Values were imputed for 613 (6.0%) of 10,296 total woman-visits. The large majority of imputed values for the outcome were negative test results.
Baseline characteristics of the women are shown in Table 1. By study design, history of intravenous drug use and high-risk sexual behavior did not vary by HIV status. A larger percentage of HIV-uninfected women were white (27.7%) compared with HIV-infected women (19.2%) (P=.02). Women infected with HIV were more likely to report sexual abstinence during the 6 months before study enrollment (P=.001). Of the HIV-infected women, 66% were not taking any antiretroviral medication at the time of enrollment. However, as the study progressed, increasing numbers of women reported taking antiretroviral medications, including HAART. Study site was not associated with the outcomes and was not included in the final analysis.
Of the HIV-infected women, 98.3% (623 of 634) had detectable immunoglobulin G specific for HPV 16, 18, 31, 35, or 45 at baseline, whereas 93.2% (246 of 264) of the HIV-uninfected women were seropositive for at least one high-risk HPV type at baseline (P<.001). Women infected with HIV were also significantly more likely to have been infected with HPV types 18 and 31 at baseline (Table 2). Recurrence of HPV shedding among sexually active women in this cohort was uncommon, with rates of 2.1–2.8 events per 100 woman-years of follow-up for each HPV type (Table 3). Women infected with HIV appeared to be at increased risk of recurrence for all five high-risk HPV types when compared with HIV-uninfected women. This difference was statistically significant for HPV types 18 (HR 4.9, 95% CI 1.1–22), 35 (HR 6.7, 95% CI 1.6–28), and 45 (HR 5.0, 95% CI 1.5–17). Among HIV-infected women, rates of HPV recurrence tended to be lower among women with higher CD4 counts, but the association of recurrent HPV infection with CD4 count was not statistically significant for any HPV type.
For types 16 and 18, we observed decreased rates of reactivation among women aged more than 35 years compared with women aged 35 years and less, but this association was not seen for HPV 45 (Table 4). The differences among age groups were not statistically significant and were limited by the small sample size. For HPV 16, 18, and 45, reactivation rates were lower among women with CD4 counts equal to or greater than 200/mm3 but this association was statistically significant only for HPV 18 (HR 0.2, 95% CI 0.1–0.9). Rates of HPV cervical shedding were higher among HIV-infected women for every HPV type, and this was statistically significant for all types except HPV 31 (Table 5). Women infected with HIV were 1.8 to 8.2 times more likely than HIV-uninfected women to have recurrent HPV shedding.
Studying HPV reactivation in cervical infection is complicated by the inability to distinguish reactivation of existing infections from clearance of HPV infection followed by re-infection with the same HPV type through sexual contact with an infected partner. Distinguishing between reactivation and re-infection requires knowledge of prior infection, an intervening period without detectable HPV disease or viral shedding, and the subsequent recurrence of active infection with the same HPV type. Accurate interim sexual exposure data must be available to exclude the possibility of re-infection. Given these criteria, and considering that HPV disease regression or progression may occur over many years, our study is among the first with adequate longitudinal data to examine HPV reactivation. The HERS cohort also allows analysis of the natural history of HPV infection in HIV-infected women before use of HAART. To our knowledge, only one other study observed new incident HPV infections in HIV-infected women reporting abstinence for 18 months or longer.13 However, that analysis did not include serologic evidence of prior infection, leaving uncertainty as to whether these events were due to reactivation of latent infections. The incident infections did, however, occur more frequently in abstinent women with CD4 counts less than 200/mm3.
Our results show that recurrent viral shedding occurs at higher rates in HIV-infected women compared with HIV-uninfected women. These results also demonstrate that HPV latency and reactivation seem to occur in HIV-infected women, with the implication that HPV clearance, as measured by PCR testing in the absence of further sexual exposure, does not guarantee these women a future free of active HPV infection. Because of the apparent low rate of reactivation and the small number of woman-years of follow up available for HIV-uninfected women, we cannot determine whether HPV reactivation also occurs in the absence of immunosuppression. Higher CD4 counts were consistently associated with lower rates of reactivation, further supporting the hypothesis that immunocompromised women are susceptible to reactivation of latent HPV infection.
Strengths of this study include the longitudinal sexual history data and twice-yearly HPV PCR and CD4 counts. To account for possible false negative HPV PCR testing, we required two consecutive negative tests over at least 12 months before considering a woman to be at risk of recurrence or reactivation. While improving the reliability of the data, this requirement likely lowered the observed rates of HPV latency and reactivation.
Limitations of this study include missing data and the small number of HIV-uninfected women. Based on the low rate of HPV shedding among HIV-uninfected women compared with HIV-infected women, a much larger study would be needed to adequately address reactivation in the HIV-uninfected group. We used repeated measures survival analysis to accurately assess rates of reactivation and recurrence. However, this technique required separate analytic models for each HPV type, limiting our ability to make conclusions about rates of reactivation and recurrence among all types combined. Similarly, we cannot evaluate the effect of co-infection with multiple HPV types on the incidence of HPV reactivation. The limited number of reactivation and recurrence events observed precluded multivariate analysis, and our analysis did not evaluate the correlation between HPV reactivation and development of dysplasia. Additionally, women self-reported their sexual activity, and no objective tools are available to validate the truthfulness of self-reporting. We attempted to minimize reporting bias by using structured interviews by trained interviewers. Finally, the low number of events do not allow us to consider correlation within site (cluster effect).
Although limited in power, this analysis challenges the current understanding of the natural history of cervical HPV infection. An updated model, shown in Figure 2, includes a latent state of viral infection, followed by reactivation of viral replication in some cases. Further longitudinal studies are needed to better characterize clinical variables associated with increased risk of HPV reactivation. As suggested by our data, reduced immune surveillance secondary to HIV progression may trigger viral reactivation, and may ultimately lead to progression of dysplasia. In this model, abstinence among currently HPV-negative, previously HPV-infected women may not be completely protective against future disease. Surveillance and counseling for such patients should reflect this updated model of HPV infection.
1. Bonnez W. Papillomavirus. In: Richman DD, Whitley RJ, Hayden FG, editors. Clinical virology. 3rd ed. Washington, DC: ASM Press; 2002. p. 603–44.
2. Munoz N. Human papillomavirus and cancer: the epidemiological evidence. J Clin Virol 2000;19:1–5.
3. Harris TG, Burk RD, Palefsky JM, Massad LS, Bang JY, Anastos K, et al. Incidence of cervical squamous intraepithelial lesions associated with HIV serostatus, CD4 cell counts, and human papillomavirus test results. JAMA 2005;293:1471–6.
4. Jamieson DJ, Duerr A, Burk R, Klein RS, Paramsothy P, Schuman P, et al. Characterization of genital human papillomavirus infection in women who have or who are at risk of having HIV infection. Am J Obstet Gynecol 2002;186:21–7.
5. Duerr A, Kieke B, Warren D, Shah K, Burk R, Peipert JF, et al. Human papillomavirus-associated cervical cytologic abnormalities among women with or at risk of infection with human immunodeficiency virus. Am J Obstet Gynecol 2001;184:584–90.
6. Selvakumar R, Schmitt A, Iftner T, Ahmed R, Wettstein FO. Regression of papillomas induced by cottontail rabbit papillomavirus is associated with infiltration of CD8+ cells and persistence of viral DNA after regression. J Virol 1997;71:5540–8.
7. Smith DK, Warren DL, Vlahov D, Schuman P, Stein MD, Greenberg BL, et al. Design and baseline participant characteristics of the Human Immunodeficiency Virus Epidemiology Research (HER) Study: a prospective cohort study of human immunodeficiency virus infection in US women. Am J Epidemiol 1997;146:459–69.
8. Viscidi RP, Snyder B, Cu-Uvin S, Hogan JW, Clayman B, Klein RS, et al. Human papillomavirus capsid antibody response to natural infection and risk of subsequent HPV infection in HIV-positive and HIV-negative women. Cancer Epidemiol Biomarkers Prev 2005;14:283–8.
9. Phelps RM, Smith DK, Heilig CM, Gardner LI, Carpenter CC, Klein RS, et al. Cancer incidence in women with or at risk for HIV. Int J Cancer 2001;94:753–7.
10. Schuman P, Ohmit SE, Klein RS, Duerr A, Cu-Uvin S, Jamieson DJ, et al. Longitudinal study of cervical squamous intraepithelial lesions in human immunodeficiency virus (HIV)-seropositive and at-risk HIV-seronegative women. J Infect Dis 2003;188:128–36.
11. Little RJ, Rubin DB. Statistical analysis with missing data. 2nd ed. Hoboken (NJ): Wiley Interscience; 2002.
12. Therneau TM, Grambsch PM. Modeling survival data: extending the Cox model. New York (NY): Springer-Verlag; 2000.
13. Strickler HD, Burk RD, Fazzari M, Anastos K, Minkoff H, Massad LS, et al. Natural history and possible reactivation of human papillomavirus in human immunodeficiency virus-positive women. J Natl Cancer Inst 2005;97:577–86.