A Comparison of the Natural History of HPV Infection and Cervical Abnormalities among HIV-Positive and HIV-Negative Women in Senegal, Africa

Human papillomavirus (HPV) is the necessary cause of squamous cervical cancer (1), is highly transmissible, and generally acquired closely after sexual debut (2, 3). Persistent infection can lead to the development of precancerous lesions which, in the absence of treatment or an effective immune response, can progress to cancer. Approximately 40 genotypes infect the genital tract and are classified on the basis of oncogenic potential with high-risk types 16 and 18 accounting for roughly 70% of cervical cancer (4). There is evidence of an interaction between human immunodeficiency virus (HIV) and HPV, with HIV-positive women at an increased risk for HPV detection, precancerous lesions, and cervical cancer compared with HIV-negative women (5–13). Studies have shown that HIV-positive women are 1.5 to 8 times more likely to have cervical cancer than HIV-negative women (9, 11, 14–17).

Despite numerous studies describing the increased burden of HPV and cervical disease among HIV-positive women, few have longitudinally examined the natural history within a single study population and provided direct comparisons to HIV-negative women. Thus, estimates for the probability of transitioning between each natural history stage (i.e., progression and regression) are limited, as well as our understanding of the point at which the natural history of HPV diverges for HIV-positive and HIV-negative women. We estimate and compare the probability of transitioning between three clinically relevant natural history stages [Normal, HPV, and high-grade precancerous lesions (HSIL)] for HIV-positive and HIV-negative women using data from multiple cohort studies conducted in Senegal, West Africa.

Data from six studies conducted from 1994 to 2010 in Senegal were used for the present analysis (Table 1). Protocols have been described elsewhere (9, 10, 18–25). Women age ≥15 at outpatient clinics were screened for participation in longitudinal research with HIV testing at baseline, as well as cervical cytology and HPV DNA sampling roughly every four months for at least two years. Recruitment occurred at two infectious disease, two family planning, and two sexually transmitted disease clinics in or around Dakar. Studies were approved by the University of Washington and Senegalese Human Subjects Review Boards with informed consent obtained from each subject. Data were deidentified and shared variables across the studies were pooled for analysis.

Blood samples were tested for HIV-1 and HIV-2 antibodies using an enzyme-linked immunosorbent assay (ELISA; Genetic Systems), a microwell plate enzyme immunoassay (HIV 1/2 EIA; Sanofi Diagnostics Pasteur), or by rapid HIV testing (Determine Alere, Inc.). HIV-1 and HIV-2 infections were distinguished by a peptide-based assay, although the assay varied by study (Genie II, Genetic Systems; Multispot, Sanofi Diagnostics Pasteur; or Immunocomb II, bispot, Orgenics). For HIV-positive women, peripheral blood was analyzed with a FACSCount analyzer (Becton Dickinson Biosciences) to determine number of CD4⁺ cells per microliter.

Conventional Pap smears were used and evaluated in Dakar until 1998; thereafter, the Thin Prep monolayer cell preparation system (Cytyc Corp.) was used and evaluated by a cytopathologist in Seattle, WA. Results were classified according to the Bethesda system (atypical squamous cells of undetermined significance, ASCUS; low-grade squamous intraepithelial lesion, LSIL; high-grade squamous intraepithelial lesions, HSIL; ref. 26). All slides obtained before 1998, those classified as LSIL or worse, and a random subset of negative slides, were re-read in Seattle, WA.

Protocols for all six parent studies called for colposcopically directed biopsies and treatment for women with evidence of HSIL or invasive cervical cancer (ICC). Biopsy/treatment participation was low in earlier studies, in part, due to delays resulting from samples being sent to and re-read in Seattle. Representative hematoxylin and eosin–stained slides were prepared from paraffin-embedded biopsy specimens and reviewed by a blinded cytopathologist. World Health Organization (WHO) pathology criteria were used to classify specimens (19).

PCRs assays for detection of HPV DNA were performed. Laboratory methods evolved over time with expansion of type-specific probes (Table 1). Testing was initially performed by use of HPV L1 consensus primers, HPV type-specific oligonucleotide probes, and a generic probe (27). This method yielded type-specific identification of HPV 6/11, 16, 18, 31/33/35/39, 45/56, and 51/52. With new probes available, in 1998 HPV detection and typing analyses were performed via a PCR-based reverse-line strip test method (Roche Molecular Systems) with probes for types 6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51, 52, 53, 54, 55, 56, 57, 58, 59, 66, 68, 73, 82, 83, and 84 (28). In 2000, a Luminex-based testing approach was adopted with additional probes for 61, 62, 64, 67, 69, 70, 71, 72, 81, IS39, and CP6108 (29).

In combination with HPV DNA results, cervical histology was used when available, with cytology in all other cases, to classify women at each visit as: Normal (HPV-negative), HPV (HPV-positive with one or more type in the absence of HSIL), or HSIL (HPV-positive with one or more type in the presence of HSIL). HSIL was defined on the basis of the presence of HSIL, CIN2, CIN3, or CIS. Because of a demonstrated lack of reproducibility for cytologic and histologic interpretations of low-grade lesions (i.e., ASCUS, LSIL, and CIN1), these results were not used to inform stage classifications (30). Transitions to ICC were not examined due to small numbers and possible HSIL treatment effects.

Competing risk modeling was used to estimate cumulative incidence functions and the probability of a specific transition over time (31). Follow-up time was set as the midpoint between visits (32). At each visit, natural history stage was classified as either incident (i.e., the stage differed from the preceding visit) or prevalent (i.e., the baseline visit or the stage was the same as the preceding visit). As HPV can be transient, restricting analyses to women with incident classification to eliminate left-censoring may produce bias. For example, eliminating women who are classified as Normal throughout follow-up (prevalent classification), leads to overestimation of the probability of transitioning from Normal to HPV as only women with prior HPV detection who become HPV-negative during follow-up contribute to incident Normal. Thus, to bound results, analyses were calculated separately for prevalent and incident natural history stages.

The following exploratory analyses were conducted on covariates of interest: age (≤25 vs. >25), CD4⁺ count (<200/mm³ vs. ≥200/mm³), HPV type (16/18 vs. all others), HIV type (1 vs. 2), and concurrent infection with multiple HPV types (>1 vs. 1 type). These analyses were restricted to transitions from Normal and HPV due to limited numbers for HSIL. For HPV-16/18 type-specific analyses, women were classified on the basis of the presence of either one of both of these types (33). Women dually infected with HIV-1 and HIV-2 were removed from HIV type-specific analyses due to small numbers (n = 52).

The Gray's method was used to test for statistical differences in cumulative incidence functions (34). Robust variances were used when multiple observations from the same woman were included when estimating a transition. Schoenfeld residuals were examined to assess the proportional hazards assumption. Analyses were conducted using SAS 9.4 software (SAS Institute).

Of 1,320 women, the majority were in their thirties, Muslim, with no more than a primary school education (Table 2). Forty-percent of married women were in a polygynous relationship. Forty-seven percent of women were HIV-positive, of which roughly two thirds were infected with HIV-1. Antiretroviral therapy (ART) was reported by 31% of HIV-positive women. Twenty-five percent were classified as having AIDS. Median CD4⁺ count at baseline was 406 cells/μL. HIV-positive women had a lower level of education, and were less likely to be currently married or report using contraception than HIV-negative women. HIV-positive women were more likely to be commercial sex workers (CSW). Most women were followed for over two years (median, 796 days; IQR, 377–1,224 days) with clinic visits approximately every 4 months. At baseline, roughly 36%, 59%, and 5% of women were classified as Normal, HPV, and HSIL, respectively (Table 2).

Of 1,079 cases classified as Normal, 56.3% were incident. HIV-positive Normal women had a 1.58 times higher rate of HPV detection than HIV-negative women [95% confidence interval (CI), 1.32–1.89] (Table 3). The 24-month predicted probability of HPV detection for HIV-positive women was 0.66 (95% CI, 0.60–0.73) and 0.82 (95% CI, 0.78–0.87) for prevalent and incident classification of Normal, respectively, compared with 0.50 (95% CI, 0.44–0.56) and 0.66 (95% CI, 0.60–0.73) for HIV-negative women (Fig. 1A). Women identified as incident Normal had a 1.58 times higher rate of HPV detection than prevalent cases (95% CI, 1.33–1.88). HIV-positive Normal women also had a 1.53 times higher rate of progression to HSIL than HIV-negative women, although this association was not significant (95% CI, 0.73–3.21). The 24-month predicted probability of HSIL for HIV-positive women was 0.04 (95% CI, 0.02–0.08) and 0.07 (95% CI, 0.04–0.12) for prevalent and incident classification of Normal, respectively, compared with 0.03 (95% CI, 0.01–0.06) and 0.04 (95% CI, 0.02–0.08) for HIV-negative women (Fig. 1B). Women identified as incident Normal had an increased rate of progression to HSIL than prevalent cases, although this was not significant (Table 3).

Of 1,265 cases of HPV detection, 39.4% were incident. HIV-positive women had a 0.46 times lower rate of regression from HPV to Normal than HIV-negative women (95% CI, 0.39–0.54). The 24-month predicted probability of regression to Normal for HIV-positive women was 0.56 (95% CI, 0.50–0.62) and 0.78 (95% CI, 0.74–0.83) for prevalent and incident HPV detection, respectively, compared with 0.83 (95% CI, 0.80–0.87) and 0.96 (95% CI, 0.94–0.99) for HIV-negative women (Fig. 1C). Women with incident HPV detection had a 1.87 times higher rate of regression to Normal than women with prevalent detection (95% CI, 1.60–2.18). HIV-positive women with HPV had a 2.55 times higher rate of progression to HSIL than HIV-negative women (95% CI, 1.69–3.86). The 24-month predicted probability of progression to HSIL for HIV-positive women was 0.18 (95% CI, 0.13–0.23) and 0.20 (95% CI, 0.16–0.25) for incident and prevalent HPV detection, respectively, compared with 0.07 (95% CI, 0.05–0.11) and 0.08 (95% CI, 0.06–0.12) for HIV-negative women (Fig. 1D). Those with prevalent HPV detection had a slightly increased rate of progression to HSIL than incident cases, although this was not significant.

Of 179 cases of HSIL, 60.9% were incident. HIV-positive women had a 0.57 times lower rate of regression from HSIL to Normal than HIV-negative women (95% CI, 0.26–1.29). The 12-month predicted probability of regression to Normal for HIV-positive women was 0.09 (95% CI, 0.04–0.19) and 0.14 (95% CI, 0.08–0.24) for prevalent and incident HSIL, respectively, compared with 0.16 (95% CI, 0.07–0.34) and 0.23 (95% CI, 0.14–0.39) for HIV-negative women (Fig. 1E). Women with incident HSIL had a 1.57 times higher rate of regression to Normal than women with prevalent classification, although this association was not significant. HIV-positive and HIV-negative women with HSIL had a similar rate of regression to HPV (relative rate, 1.06; 95% CI, 0.71–1.59). For HIV-positive women the 12-month predicted probability of regression to HPV ranged from 0.70 (95% CI, 0.60–0.79) to 0.68 (95% CI, 0.56–0.84) for incident and prevalent HSIL, respectively, compared with 0.67 (95% CI, 0.52–0.86) and 0.67 (95% CI, 0.54–0.81) for HIV-negative women (Fig. 1F). The rate of regression to HPV was similar for both prevalent and incident cases of HSIL.

HIV-positive women with CD4⁺ counts <200 had a 2.86 times higher rate of transitioning from Normal to HSIL than women with higher CD4⁺ counts (95% CI, 0.95–8.58; Table 4). Low CD4⁺ count was also associated with a higher rate of progression from HPV to HSIL (1.83; 95% CI, 1.07–3.14), and a lower rate of regression from HPV to Normal (0.57; 95% CI, 0.39–0.83).

HIV-positive Normal women with HPV-16/18 had a 4.62 times higher rate of progression to HSIL than those infected with other HPV types (95% CI, 1.10–19.42). Similarly, those with HPV-16/18 infection had a higher rate of progressing from HPV to HSIL (2.20; 95% CI, 1.34–3.62) and a lower rate of regression to Normal (0.35; 95% CI, 0.23–0.54) compared with those infected with other HPV types. Rates for acquisition of HPV (Normal to HPV) were similar across HPV type sub-groupings.

HIV-positive women with multiple HPV types had a higher rate of progression to HSIL (2.33; 95% CI, 1.40–3.88) than those with a single HPV type. HIV-positive women with multiple HPV types also had a 0.34 times lower rate of regression from HPV to Normal than women with a single HPV type (95% CI, 0.25–0.47). In contrast, progression rates from Normal to HPV and Normal to HSIL were similar for those who acquired multiple HPV types and those who acquired a single type.

Among Normal HIV-positive women, those age ≤25 years had a 1.56 times higher rate of incident HPV detection than those >25 (95% CI, 1.14–2.14). Age associations for the other transitions examined were nonsignificant.

Among Normal HIV-positive women, those with HIV-1 had a 1.85 times higher rate of incident HPV detection than those with HIV-2 (95% CI, 1.38–2.49). Similarly, women with HIV-1 had a 1.69 times higher rate of progression from HPV to HSIL compared with women with HIV-2 (95% CI, 0.94–3.03). Women with HIV-1 also had a lower rate of regression from HPV to Normal than those with HIV-2 (0.83; 95% CI, 0.64–1.08), although this association was not statistically significant. In contrast, those with HIV-1 had a lower rate of progression from Normal to HSIL (0.48; 95% CI, 0.19–1.21). Mutual adjustment for baseline age and CD4⁺ count (as surrogate measures of duration and severity of HIV infection), did not affect these findings.

As HIV-positive women are at an increased risk of cervical cancer, understanding the distinct natural history within this population is essential for informing targeted prevention efforts. We found that HIV-positive women had higher rates of incident HPV detection and progression to HSIL, as well as lower rates of regression from HSIL and HPV infection when compared with HIV-negative women. The most notable difference between HIV-positive and HIV-negative women was the more than doubled rate of progression from HPV to HSIL (although absolute risk remained low). True transition probabilities likely lie between the incident and prevalent curves as a function of age and sexual activity.

Although analyses of potential effect modifiers should be interpreted with caution due to limited sample sizes, overall results suggest that HIV-positive women with baseline CD4⁺ counts <200 (cells/μL) or infection with HPV-16/18 had higher rates of progression and lower rates of regression. Consistent with other research (35, 36), this study found increased incident detection of multiple HPV types in HIV-positive women compared with HIV-negative women, but also demonstrated that infection with multiple types increased progression and decreased regression rates from HPV. This is particularly relevant as >1 type was present in 57% of visits in which HIV-positive women had HPV detection, in comparison with 28% among HIV-negative women.

Evidence regarding the role of HIV type on the natural history of HPV is conflicting (9, 10, 24, 37–41), with some research indicating that HIV-2 is more strongly associated with HPV-related disease than HIV-1. In contrast, the present analysis found HIV-1 to be more strongly associated with adverse transitions than HIV-2 (with the notable exception of a lower probability of transitioning from Normal to HSIL). HIV-1 has a shorter incubation period, higher transmissibility, and more rapid development of immunosuppression compared with HIV-2 (22, 42). Thus, it is biologically plausible that women with HIV-1 are at greater risk due to more severe immunosuppression. However, longer survival among women with HIV-2 (43) may result in higher lifetime risk due to extended time to develop cervical cancer.

This study has limitations. The present analysis focused on three stages of natural history: Normal (HPV-negative), HPV (HPV-positive with one or more type and the absence of HSIL), or HSIL (HPV-positive with one or more type and the presence of HSIL). If a woman went from HPV-positive with type 16 to HPV-positive with type 18 in consecutive visits this was classified as a continuation of the HPV stage. This situation was unusual as many women had multiple HPV types such that one type was persistent during consecutive HPV visits. Histology was available for 10% of all clinic visits. Misclassification resulting from cytology likely leads to undetected cervical abnormalities (e.g., false negatives), such that some women are classified as HPV when in fact they are HSIL. Thus, detection of HSIL may be underestimated. However, it is important to note that histology is an imperfect gold standard with demonstrated low reproducibility (30). Data on ART were limited for several studies included in the analysis; therefore, the impact of CD4⁺ count was examined to indirectly capture some of these treatment effects. Furthermore, this sample is largely comprised of women in their thirties who may have been previously infected with HPV, such that a positive HPV test represents incident DNA detection rather than incident infection (i.e., potential reactivation effects).

This study has a number of strengths, most notably a large longitudinal sample to directly compare HIV-positive and HIV-negative women. These data allowed for the examination of the impact of HIV type, as both HIV-1 and HIV-2 are endemic to West Africa. HPV-16 and HPV-18 were tested for in each study included in the analysis; therefore, we were able to examine the roles of these highly oncogenic types separate from other types. Further, this sample includes registered CSWs. As CSW are estimated to represent roughly 25% of new HIV cases in Africa, this population is particularly relevant for coinfection research (44). Finally, factors known to impact HPV natural history were minimal in this sample (e.g., oral contraception use, smoking, prior cervical cancer screening/treatment, and HPV vaccination).

Our findings suggest that targeted screening programs for the high-risk population of HIV-positive women are needed, in addition to HPV vaccination of youth before HIV infection. However, optimal screening modality, frequency, triage, and treatment threshold remain unknown, requiring both expert clinical review and cost-effectiveness examination.

Several studies and pilot projects have demonstrated the feasibility of implementing cervical cancer screening and treatment in low-resource settings (45), although significant economic and infrastructural challenges remain (46). Although screening with cytology has been widely adopted in high-resource settings resulting in major declines in cervical cancer incidence over the past 60 years, significant laboratory, equipment, and clinical expertise requirements have limited the capacity to establish cytology in low-resource settings (46, 47). Furthermore, cytology has low reproducibility and sensitivity in comparison with other strategies, necessitating frequent screening to achieve high effectiveness (30, 47). As such, a substantial effort has been made to identify and evaluate screening strategies that are more contextually relevant for the high-risk setting of sub-Saharan Africa. Hybrid Capture 2 (HC2) HPV testing is highly sensitive and reproducible with the potential for self-collection of samples (48), yet requires significant investment in laboratory equipment and technician expertise (46). Both cytology and HPV testing involve laboratory processing time leading to delays in obtaining results and an additional clinic visit if treatment is needed. Visual inspection with acetic acid (VIA) may be most suitable to low-resource settings as it involves naked eye inspection of the cervix (yielding immediate results), and requires little clinical expertise and no laboratory equipment for processing samples. However, inter-observer differences in subjectively determining positivity, concerns regarding over treatment, and the potential for small lesions to remain undetected may reduce the effectiveness of VIA (46). The more recently developed careHPV test offers the benefits of standard HPV testing, yields rapid results (roughly 2 hours of processing time), uses mobile battery operated processing equipment, and requires minimal technical training (46, 49). Both VIA and rapid HPV testing can be implemented as part of a same day “screen and treat” approach, minimizing losses to follow-up and delayed treatment. Immediate treatment with cryotherapy has minimal infrastructure requirements and can be provided by nurses and midwives (46). In contrast, surgical treatment with conization or loop electrosurgical excision procedure (LEEP) require significant infrastructure and technical training, and are further limited by the lack of pathology services in sub-Saharan settings (50). These various factors for testing and treatment must be considered to provide contextually relevant screening recommendations.

Importantly, both test and treatment effectiveness is reduced when applied to HIV-positive women adding uncertainty regarding optimal cervical cancer prevention strategies for this high-risk population (51–56). For instance, given the high prevalence of HPV among HIV-positive women, HPV testing as a stand-alone strategy may be inefficient for identifying women for targeted follow-up (due to a low positive predictive value such that a positive HPV test is not in itself actionable and other intermediate steps may be required). Furthermore, although HIV-positive women have significantly higher rates of adverse transitions than HIV-negative women, absolute risk remains low that could indicate that testing frequencies currently used among the general population may be extended to HIV-positive women. However, increased false negatives among the HIV-positive population may indicate the need for more frequent testing, and potential poor retention in care for certain settings (i.e., rural with limited access to follow-up care) may necessitate aggressive treatment of precancer lesions.

In summary, there is a pressing need for targeted screening in the high-risk population of HIV-positive women. These efforts can subsequently be used to inform the eventual implementation of comprehensive population-based screening paradigms that are not yet widely available in the limited resource setting of sub-Saharan Africa (57). Given the distinct natural history of HPV among HIV-positive women (as demonstrated in our analyses), as well as varying test and treatment attributes, it is important to quantitatively compare the potential impact of screening strategies to understand how these factors collectively impact HPV-associated morbidity and mortality. In the absence of comprehensive cost-effectiveness analyses specific to low-resource settings of sub-Saharan Africa, frequent screening with VIA beginning at the time of HIV diagnosis with conservative triage of positive cases may be worth examining. Identifying optimal targeted prevention strategies is essential given modeling evidence, suggesting that cervical cancer among HIV-positive women may increase in the future due to extended life expectancy resulting from ART (58, 59).

No potential conflicts of interest were disclosed.

Conception and design: H.K. Whitham, S.E. Hawes, J.M. Oakes, A.R. Lifson, N.B. Kiviat, P.S. Sow, G.S. Gottlieb, S.L. Kulasingam

Development of methodology: H.K. Whitham, N.B. Kiviat, S.L. Kulasingam

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.E. Hawes, N.B. Kiviat, P.S. Sow, G.S. Gottlieb, M.P. Sy, S.L. Kulasingam

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.K. Whitham, H. Chu, J.M. Oakes, A.R. Lifson, N.B. Kiviat, S. Ba, S.L. Kulasingam

Writing, review, and/or revision of the manuscript: H.K. Whitham, S.E. Hawes, H. Chu, A.R. Lifson, N.B. Kiviat, G.S. Gottlieb, S. Ba, M.P. Sy, S.L. Kulasingam

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.K. Whitham, S.E. Hawes, N.B. Kiviat, P.S. Sow, G.S. Gottlieb, S. Ba

Study supervision: J.M. Oakes, N.B. Kiviat, S. Ba, S.L. Kulasingam

Other (study nurse and follow patients on the daily basis, send data to investigators, and share thoughts and suggestions regarding patients health and concerns): M.P. Sy

We thank the research teams from the University of Washington and Senegal for undertaking this research, as well as the many women who participated in these studies over the course of the past 20 years.

This work was supported by the University of Minnesota's Doctoral Dissertation Award. Studies were funded by grants from the NIH: CA62801 and AI48470 (to N.B. Kiviat), AI49755 (to G.S. Gottlieb), CA97275 (to N.B. Kiviat), AI60466 (to G.S. Gottlieb), and CA111187 (to N.B. Kiviat).

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.