Determinants of Incidence and Clearance of High-Risk Human Papillomavirus Infections in Rural Rakai, Uganda

Human papillomavirus (HPV) is the etiologic agent for cervical cancer and the most common viral sexually transmitted infection. HPV infections represent a major public health burden worldwide as a cause of cervical cancer in women, genital warts in men and women, other anogenital cancers, and some nasopharyngeal cancers. Acquisition of HPV is especially common among young sexually active adolescents with 3-year cumulative incidence estimated at more than 40% (1-3). Although HPV point prevalence is very high, most clear spontaneously, especially among young women (4, 5); however, persistent infection with high-risk HPV (HR-HPV) types is considered the crucial step in the development of cervical cancer (6).

As a consequence of cervical cytology screening, cervical cancer incidence and mortality has been drastically reduced in developed counties, whereas such a decline has not occurred in less developed countries where 80% of cervical cancers occur. Beyond lack of, or inadequate coverage of screening programs (7, 8) in developing countries, women rarely present for a routine pelvic exam if they are asymptomatic for disease. Hence, to increase screening coverage in settings with low compliance for pelvic exam–based screening, use of self-collected vaginal swabs has been widely advocated and evaluated in the literature (9-23), showing good agreement between self-collected vaginal swabs and clinician-collected cervical swabs for HPV DNA detection. This allows self-collection of vaginal samples to be used as a screening tool for HPV. Furthermore, the possibility of primary prevention of HPV infection and thus cervical cancer via HPV vaccination has been realized. Information on the epidemiology of HPV in understudied regions is fundamental to planning and evaluating future vaccine implementation programs. Finally, the prevention of HPV and cervical cancer is of particular urgency in Africa where there are an estimated 17 million women living with HIV (24).

We previously showed comparable HPV DNA detection from self-collected vaginal and physician-collected cervical swabs in this population (25). To investigate the utility of self-collected vaginal swabs beyond screening, in this analysis, our objective was to investigate the epidemiology (incidence and clearance) of HPV using self-collected vaginal swabs from a general rural African population. If our findings, based on self-collected vaginal samples, are similar to findings from other studies of the natural history of HPV, which relied on clinician-obtained cervical samples, self-sampling could represent an acceptable and valid alternative for epidemiologic studies in the developing world where low compliance with pelvic exams may lead to self-selection of participants and a reluctance to provide repeat samples for research studies. Further, self-sampling may provide an efficient way to monitor vaccine effectiveness in these high-risk populations.

Participants are from the ongoing Rakai Community Cohort Study described in detail elsewhere (26). Briefly, the Rakai Community Cohort Study conducts annual home-based surveys, with comprehensive interviewer-administered questionnaires on sociodemographic, behavioral, and health characteristics, on an average population of 12,000 adults ages 15 to 49 years. Biological samples are collected following the interview. Venous blood is collected for HIV-1, syphilis, and herpes simplex type 2 serology. Starting in 1998, self-collected vaginal swabs were obtained during routine follow-up visits. Consenting women were asked to provide a self-collected vaginal sample at home, using the Digene sampler kit, with a Dacron swab. Women were asked to squat and insert a sterile 20-cm Dacron swab into the vagina up to the vault and to rotate the swab three times in the vaginal vault. The swab was then placed in 1 mL standard transport medium (Digene) and kept on ice until transported back to the field laboratories and frozen at -80°C. All samples were periodically shipped to Johns Hopkins University and stored at -80°C until the time of assay.

Although majority (>85%) of women provided self-collected vaginal swabs at each Rakai follow-up visit, for this longitudinal study, our analytic sample was composed of women who provided three to four self-collected swabs at consecutive follow-up visits.

The study was approved by the institutional review boards in Uganda and at Johns Hopkins University and Columbia University.

HIV Serology. The HIV status of each participant was determined using two enzyme immunoassays (Vironostika HIV-1; Organon Teknika and Cambridge Biotech), and discordant enzyme immunoassay results or new seroconversions were confirmed by Western blot (HIV-1 Western Blot; Bio-Merieux-Vitek).

HPV Detection and Genotyping. HPV status was determined using a two-stage approach. First, using hybrid capture 2 (HC2; Digene), all samples were screened for the presence of 13 carcinogenic HPV genotypes most clinically relevant to cervical premalignant and malignant lesions. Subsequently, all HC2 HPV-positive samples and 10% random sample of HC2 HPV-negative samples were genotyped using the PGMY09/11 L1 consensus primer PCR to amplify HPV DNA and reverse hybridization technique (Roche Molecular Systems) to detect 37 HPV genotypes (27, 28).

HPV DNA Testing by HC2. HC2 is a Food and Drug Administration approved, commercially available HPV test that collectively detects 13 carcinogenic HPV types (HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68) without distinguishing the HPV genotype. HC2 is a signal amplification assay that uses antibody capture of HPV DNA and RNA probe hybrids and chemiluminescent signal detection. The HC2 assay was done according to the manufacturer's instructions on a 100 μL aliquot of each specimen, a strategy that preserved sufficient sample volume for the additional PCR-based analyses.

HPV DNA Testing by Roche-PCR Line Blot Assay: DNA Extraction. HPV genotyping was done using the reverse line-blot assay as described previously (28). Briefly, 50 μL standard transport medium sample was denatured with 10× digestion buffer for 1 h at 65°C followed by heat denaturation at 95°C for 10 min. DNA was precipitated with ethanol and ammonium acetate at -20°C overnight. After centrifugation at 21,000 × g for 30 min at 4°C, DNA pellet was dried and resuspended in 25 μL Tris-EDTA and stored at -20°C (13).

DNA (5 μL) was amplified using the PGMY09/11 system that coamplifies HPV genotypes and human β-globin internal control target in a single reaction (13, 27, 28). Genotypes were discriminated by hybridizing the biotinylated PCR products to a probe array (RMS Linear Array) developed by Roche Molecular Systems, as described earlier, which can differentiate between 37 HPV types (27).

Appropriate quality-control samples were imbedded during the DNA purification, PCR, and genotype discrimination steps.

Incidence of HR-HPV Infection. Genotype-specific HPV incident infection was defined as a positive test result for that genotype not present at study baseline. Subsequently, the 13 HR-HPV (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68) genotypes most associated with cervical cancer were combined into one outcome variable of “incidence of new carcinogenic HPV” infection. Person-time methods were used to calculate the type-specific incidence density (29).

Univariate and multivariate Poisson regression was used to evaluate the effect of exposure variables on the risk of new HR-HPV infection. Incidence of any of the 13 HR-HPV was measured as the number of new infections in a given time interval divided by the population at risk during the interval. The log of the HPV incidence rate is assumed to be linearly associated with the risk factors, and the model parameters, after exponentiation, can be interpreted as relative incidence, which is similar to the relative risk. Behavioral and demographic factors measured at every visit were entered into the models as time-dependent covariates. When subjects missed study visits, the exposures were treated as missing. The independent contribution of each exposure variable, after adjustment for the effects of the other covariates, was expressed as the adjusted relative incidence rate or rate ratio (RR).

Clearance of HR-HPV Infection. Subjects included in the clearance analysis had to have been positive at least once for any of the 13 HR-HPV genotypes during their follow-up visits.

Methods for multiple failure-time data were used to evaluate the effect of covariates on the risk of type-specific HR-HPV clearance. The unit of analysis for the carcinogenic HPV clearance is the HPV infection. Hence, individuals infected with multiple HPV genotypes could clear one but not another type, and one or more clearance events could occur for the same individual. We first assigned clearance for each HPV genotype and then constructed an overall variable for combined clearance. We assumed clearance events to be unordered, which means that clearing one HPV type was independent on clearing another HPV type. We then used survival analysis methods to evaluate factors associated with type-specific HPV clearance. Time was measured from the first time a woman tested positive for a specific HPV type, and the event, type-specific HPV clearance, was defined as the first time a woman tested negative following a positive measurement. Those who never cleared the specific HPV-type infection were administratively censored at the end of the study follow-up. Because it is not reasonable to assume that clearance of each HPV type should have the same baseline hazard, the Cox model was stratified by each HPV type. Relative hazards <1 indicate lower clearance.

One thousand seventy-nine women contributed 2,954 visits for the incidence analysis. The median number of visits per person was 3, with an interquartile range of 2 to 4 visits. There were 24 women with only 1 visit who were excluded from this analysis, leaving 1,055 women in the analytic sample.

When we compared the analytic sample with all women in the Rakai cohort (n = 6,520) at the HPV study baseline visit, we observed significant differences in certain characteristics suggesting that the analytic sample had a lower risk profile for HPV infection. Women in the analytic sample were slightly younger (median age, 28 years; interquartile range, 23-36) compared with those not in the analytic sample (median age, 29 years; interquartile range, 22-40; P = 0.005), less likely to be HIV positive compared with others (14% versus 21%; P < 0.0001), more likely to be currently married (75% versus 63%; P < 0.0001), and with fewer lifetime sexual partners (P < 0.0001; data not shown). Thus, the analytic sample had a lower risk profile than the general Rakai population.

Table 1 shows the univariate associations with new HR-HPV infections. Incidence was constant over time of observation, indicating that there were no temporal trends over the study duration. HR-HPV incidence was highest among younger women and decreased with age. There were no associations with education. Married women had a lower incidence of HR-HPV infection compared with single women. Incidence was lower among gravid women compared with nulligravid women. Incidence was lower among never-pregnant women; however, this was due to confounding by single marital status.

HIV-positive women were at a significantly elevated risk of new HR-HPV infection compared with HIV-negative women [RR, 2.48; 95% confidence interval (95% CI), 1.79-3.43]. Self-reported numbers of current sexually transmitted disease (STD) symptoms or self-reported general health symptoms were not associated with new HR-HPV infection (Table 1).

Age at first sex was not associated with new HR-HPV infection (Table 1), whereas HR-HPV incidence was significantly elevated among women who reported 2 and ≥3 lifetime sexual partners [RR (95% CI), 1.86 (1.19-2.91) and 2.06 (1.36-3.13), respectively] compared with women with 0 or 1 lifetime partners. Incidence was marginally elevated among women with ≥2 sex partners in the past year. Women who perceived themselves or their partner to be at high risk for HIV/AIDS infection were more likely to have new HR-HPV infection. Drinking alcohol was not associated with incident HR-HPV.

Risk factors associated with new HR-HPV infections in the univariate analyses were evaluated in multivariate models (Table 1). After adjustment, risk of incident HR-HPV was significantly increased with younger age, HIV infection, unmarried women, 2 and ≥3 lifetime sex partners, and ≥2 partners in the past year. Women who perceived their recent partner to be at high risk for HIV/AIDS infection were more likely to have new HR-HPV infection.

Table 2 shows incidence rates for the 13 HR-HPV individually and combined. HPV-51 had the highest incidence rate followed by HPV-16, whereas HPV-31 and HPV-35 had the lowest incidence rates. A total of 178 new HR-HPV infections were observed during 2,054 person-years of observation, reflecting an incidence rate of 8.7 per 100 person-years. Stratifying by HIV serostatus, among 9 of the 13 HR-HPV types, the risk of a new infection was significantly greater among HIV-positive compared with HIV-negative women (types 16, 18, 35, 45, 51, 56, 58, 59, and 68). The highest incidence RR comparing HIV-positive with HIV-negative women was observed for HPV-58. Overall, the risk of a new HPV infection was significantly higher among HIV-positive women (17.3 per 100 person-years) compared with HIV-negative women (7.0 per 100 person-years; P < 0.0001).

One hundred ninety-five HPV-positive women contributed 389 observations (the unit of analysis is type-specific HPV infections) and 465 person-times for analysis of viral clearance. Two hundred forty-nine of the infections cleared over the follow-up period (53.55%).

Table 3 presents findings of factors associated with HPV clearance. Women ages >30 years were less likely to clear infections compared with women ages 15 to 29 years. Women with higher education were more likely to clear infection compared with those with no education, as were currently married women. Ever-pregnancy and gravidity were not associated with clearance in univariate analyses.

HIV-positive women were significantly less likely to clear infection compared with HIV-negative women (RR, 0.70; 95% CI, 0.56-0.87). Clearance was not associated with symptoms suggestive of STD, AIDS, opportunistic infections, or infection with multiple HPV types (Table 3). Women with greater HPV shedding as indicated by the higher relative light units (based on HC2 assays) were less likely to clear their infection compared with those in the lowest relative light units.

Clearance rates were lower in women who reported 2 or ≥3 lifetime partners compared with women with 1 lifetime sex partner (RR, 0.77 and 0.79, respectively). There were no associations with age at first sex, recent sex partners, AIDS perception, or drinking.

In multivariate analysis (Table 3), HIV-positive women were less likely to clear their infection (adjusted RR, 0.75; 95% CI, 0.59-0.96) compared with HIV-negative women, as were women ages >30 years compared with women ages ≤29 years (adjusted RR, 0.84; 95% CI, 0.68-1.02). Clearance was also reduced in women reporting ≥2 lifetime partners and among women with higher HPV viral burden. Women with secondary or higher education were more likely to clear their infection. Stratified analysis by HIV status showed that among HIV-positive women, higher HPV viral burden, and higher number of lifetime sexual partners and age (marginal effect) were associated with lower clearance. However, among HIV-negative women, age was associated with a marginal decrease in clearance, but education was associated with higher clearance (data not shown).

Table 4 presents clearance rates for the 13 different HR-HPV types. Highest clearance rates were observed for HPV-18 and HPV-68 followed by HPV-39. Lowest clearance occurred for HPV-31 and HPV-35 followed by HPV-16. Stratifying by HIV status, for almost all viral types (with the exception of HPV-56), we observed lower clearance rate among HIV-positive compared with HIV-negative women. HPV-33 achieved statistical significance after adjusting for age and lifetime number of sexual partners.

Data on the epidemiology of HPV from resource poor countries are sparse, mainly due to lack of longitudinal studies. A constraint on performing these studies has been recruiting and retaining participants to studies, which require pelvic examination for collection of cervical specimens. This analysis based on self-collected vaginal swabs showed that HR-HPV incidence and clearance from a rural community in Uganda mirrors studies of the epidemiology of HPV based on cervical samples. These findings have important implications for future longitudinal studies in settings where resource constraints limit pelvic exams and, most importantly, where women are reluctant to participate in studies that rely on pelvic exams for data collection (25, 30-32).

We were comfortable using self-collected vaginal swab to determine the epidemiology of HPV, because in an earlier paper we validated HPV DNA detection from self-collected vaginal swabs to physician-collected cervical swabs showing very good agreement between vaginal and cervical DNA detection (25).

Similar to other studies, we found younger age and higher number of recent and lifetime sex partners to be significant risk factors for new HPV infection (1, 33). Younger age reflects onset of sexual activity as well as limited acquired immunity. The association of recent numbers of sexual partners with incident detection of HR-HPV suggests that part of the new infections are due to sexual behavior; however, the association of higher numbers of lifetime sexual partners with incident infection suggests that part of the new infections may be due to reactivated past infections. Married women had lower risk of new HPV infection compared with single women and women with high perception of their partner's AIDS risk were more likely to be at risk of new HPV infection. This may be a marker of the participant or her partner's sexual behaviors that were not directly measured or were misreported.

HIV-positive women were twice as likely to have a new HR-HPV infection in accord with published studies from the United States and Europe (34, 35). With the exception of HPV-31 and HPV-33, the incident RR for the other 11 HR-HPV ranged from 1.5 to 6 times higher among HIV-positive women compared with HIV-negative women (HPV-52 and HPV-58, respectively). The incidence RR for HPV-16 was 3.09 among HIV-positive compared with HIV-negative women, similar to findings by Strickler et al. (36), suggesting that HPV-16 may be weakly associated with immune status in HIV-positive women compared with other types and be better at evading the immune system, which can lead to higher persistence and its stronger association with cervical cancer.

HPV-31, HPV-35, and HPV-16 had the lowest clearance rates as reported in other studies (37, 38). Similar to other studies (35, 34, 39), we observed lower HPV clearance rates among HIV-positive women compared with HIV-negative women. We found HPV viral burden to be associated with lower clearance, and this was highly significant among HIV-positive women. This may be an indication of the damage by HIV to the immune system and its inability to perform normally. Furthermore, higher HPV viral load is also associated with worse cytopathology and hence could be an indication of higher-grade cervical lesions among HIV-positive women. However, although we did not have pathology on all women, among those with colposcopy (and biopsy if it was indicated), we observed more lesions among HIV-positive women than HIV-negative women. Additionally, we found that age affected HPV clearance: women ages >30 years were less likely to clear an infection compared with women ages <30 years, consistent with findings of other studies (1, 4, 34). We also found that women with more lifetime sex partners were less likely to clear their HPV infection. A similar finding was observed by Shew et al. (40) among adolescents. This is suggestive of multiple past exposures or reactivated infections acquired earlier, which may be less prone to clearance.

Interpretations of our results should take into account some limitations. Our application of the HC2 assay to all samples, followed by genotyping of all HC2-positive samples and 10% randomly selected HC2-negative samples, assumed that HC2 detected all 13 HPV types in its probe with reasonable accuracy. We found 8% HC2-negative samples were PCR positive, which could have resulted in an underestimation of the true incidence and clearance of HR-HPV. Differential virus quantity used in HC2 and line-blot assays or variations in the laboratory techniques could be possible explanation for the discordant results between HC2 and line-blot assay. To minimize differences due to laboratory practices and ensure quality control, one person (M.S.) performed all laboratory assays, and negative control cell lines were included for DNA isolation and positive and negative controls were included in the PCR and genotyping procedures. Reasons for differences between HC2 and line-blot warrant further investigation.

Further, all factors found to be associated with HPV infection in other studies were not measured in this study. We could not assess the effect of CD4 cell counts and HIV viral load on incidence and clearance of HPV. Additionally, because no women were using oral contraceptives or smoking cigarettes in this sample, we could not assess their effects on incidence or clearance.

Specific to the incidence and clearance analysis, some women may have acquired or lost their infection in between-study visits, which could lead to underestimation of incidence or overestimation of duration of infection. With regard to the latter, our objective was not to report on the duration of infection but rather to characterize women who cleared their infection.

Because the epidemiology of HPV as measured by incidence and clearance in this representative rural Ugandan population was similar to studies that relied on cervical sampling, recruitment and retention of women for natural history studies of HPV may be increased by self-collection as an alternate to clinician-collected cervical specimens. In settings where obtaining the necessary data for the natural history of HPV poses a challenge for researchers, self-collection can provide a reliable method for obtaining samples and may be adopted in trials to increase compliance and reduce costs associated with conducting pelvic exams. Moreover, determining the effectiveness of vaccines in resource poor and HIV endemic countries is dependent on knowledge of natural history and epidemiology of HPV and can provide an important tool for future research or vaccine evaluation.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank the Roche Diagnostics for providing the reagents and line blots for the genotyping assay and Digene for providing the HC2 probe B kits at reduced costs.