Combination antiretroviral therapy reduces the detection risk of cervical human papilloma virus infection in women living with HIV
Download the PDF here
"cART significantly reduced the detection risk for cervical hrHPV (OR 0.33, 95% CI 0.24-0.44) and HPV-16 (OR 0.50, 95% CI 0.37-0.67)."
"From this study, we learned that cART exposure reduced the risk of detection of any HPV type by 77%. Adjusting for other covariates, a longer time of exposure to cART significantly reduced the odds of HPV detection. For every additional month of cART since initiation, the risk of detection of any HPV type decreased by 9%. The effect was also significant on HPV-16 genotype alone. We also found that lrHPV genotypes were less influenced by cART as compared to HPV-16 genotypes. Furthermore, not surprisingly, the effect of cART on the detection risk for the most prevalent hrHPV genotype HPV-16 was not different to the effect of cART on the detection risk of the other high-risk oncogenic genotypes grouped together. The effect of cART seemed to be immunology-driven. There was an increased risk for any HPV detection at CD4+ T-cell count below 200 cells/μl (OR 3.78, CI 1.87-7.64), but when adjusted, the time of cART exposure again remained the stronger predictor of risk. There was a weak association of HPV-16 with very low CD4+ T-cell count."
Combination antiretroviral therapy reduces the detection risk of cervical human papilloma virus infection in women living with HIV
AIDS: Jan 2 2015
Zeier, Michele D.a; Botha, Matthys H.b; Engelbrecht, Susanc,d; Machekano, Rhoderick N.e; Jacobs, Graeme B.c; Isaacs, Shahiedac; van Schalkwyk, Marijea; van der Merwe, Haynesb; Mason, Deidreb; Nachega, Jean B.a,f,g
aDepartment of Medicine and Centre for Infectious Diseases (CID)bDepartment of Obstetrics and GynecologycDepartment of Pathology, Division of Medical Virology, Stellenbosch University, Faculty of Health SciencesdNational Health Laboratory Services (NHLS), Western Cape Region, Tygerberg Hospital (Coastal)eDepartment of Interdisciplinary Health Sciences, Centre for Evidence-Based Healthcare, Biostatistics Unit, Stellenbosch University, Faculty of Health Sciences, Cape Town, South AfricafDepartment of Epidemiology, Pittsburgh Graduate School of Public Health, Infectious Diseases Epidemiology Research Program, Pittsburgh, PennsylvaniagDepartments of Epidemiology and International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA.
Data on the effect of combination antiretroviral therapy (cART) on cervical human papilloma virus (HPV) infection are both limited and conflicting. We aimed to determine the effect of the initiation of cART for HPV genotype detection on cervical samples in HIV-infected South African women.
Prospective cohort study.
Generalized estimating equation was performed to estimate parameters of mixed-effects logistic regression models of cART on HPV cervical detection risk, adjusting for time-dependent covariates CD4+ T-cell count, sexual activity and excision treatment. Ratio of odds ratios (ORs) was computed to compare the pooled cART effect on lower vs. high-risk HPV genotype groups, to the effect of cART on the risk of HPV-16 detection.
Of the 300 patients, 204 (68%) were commenced on ART during follow-up, as they met the criteria for cART initiation. cART significantly reduced the risk for detection of HPV by 77% [OR 0.23, 95% confidence interval (CI) 0.15-0.37]. cART significantly reduced the risk of HPV-16 detection (OR 0.50, 95% CI 0.37-0.67). Every month on cART significantly reduced the detection risk of any HPV type by 9% (OR 0.91, 95% CI 0.89-0.94). The protective effect of cART on the detection risk for the low-risk HPV genotype group was significantly less than the protective effect of cART on the detection risk of HPV-16 (ratio of ORs 1.35, 95% CI 1.22-1.50).
cART significantly reduced cervical HPV infection. This effect was dependent on the duration of exposure to cART and is the mechanism by which cART may improve the outcome of dysplasia in HIV-infected women.
HIV-infected women have an increased risk for the development of invasive cancer and the cancer seems to occur at a younger age [1-4]. It is thought to be due to the increased incidence and persistence of cervical human papilloma virus (HPV) infection in HIV-positive women, the recurrence thereof after excision treatment [5-14], and possible reactivation of latent HPV infection [5,14-16]. Also, HIV-positive women may be more likely to be infected with HPV from their male partners than HIV-negative women . Early local cervical immune dysfunction after HIV acquisition may contribute to the increased susceptibility .
The additional risk of progression of cervical lesions from pre-cancer to cancer can be attributed to the specific oncogenic HPV infection and also likely to the increased immunosuppression triggered by HIV infection. In a Johannesburg study by Firnhaber et al. of HIV-infected women, 54% had abnormal smears [of which 66% were low-grade squamous intraepithelial lesions (LSILs) and 33% were high-grade squamous intraepithelial lesions (HSILs)]. Ninety-five percent of the 148 women had HPV detected and 83% more than one subtype, with increased risk with CD4+ T cells below 200 . This was also seen in a Cape Town study by Denny et al., in which abnormal cytology and high-risk HPV (hrHPV) positivity were strongly correlated with low CD4+ T-cell counts and high HIV viral loads.
Treatment of HIV infection with combination antiretroviral therapy (cART) is associated with at least some immune recovery, as is observed in a rise in the CD4+ T-cell count measured in the blood. An assumption may be made that treatment with cART and the associated immune restoration, even if partial, could also reduce the incidence and duration of HPV infection. However, several studies from as early as 1998 examined the effect of cART on cervical HPV infection and reported conflicting findings. One of the first studies saw early regression of cervical lesions in 49 women initiating ART, but clearance of the HPV infection was not always observed in those cases . In another study that enrolled 45 MSM, a high prevalence of anal HPV and HSIL remained despite immune restoration under cART . When looking more specifically at incident HPV detection, two more recent study reports also failed to show a protective effect of cART [22,23].
In contrast, in a study by Minkoff et al. of women initiating cART, study participants were 40% more likely to experience regression of their cervical lesions after adjustment for CD4+ T-cell counts. cART altered the course of HPV-related disease, reducing progression and increasing regression of lesions . Paramsothy et al. included 537 women in their study and found that cART was associated with improved HPV clearance in women with pre-existing cervical squamous intraepithelial lesions [hazard ratio 4.5, 95% confidence interval (CI) 1.2-16.3], but not with abnormal Pap test regression. Also, it did not lead to improved clearance of HPV infection in women with normal Pap smears . Similarly, another larger study by Konopnicki et al. saw sustained suppression of HIV-RNA plasma load for more than 40 months and CD4+ T-cell count above 500 cells/μl for more than 18 months associated with reduced risk of hrHPV persistence.
Minkoff did another prospective study of 286 women which was published in 2010. His team again found that cART use in women was significantly associated with a reduced burden of HPV infection and SILs. The effect was more pronounced if the women were adherent to the medication. He did not find a difference if a stricter definition - that of undetectable viral loads - was used for adherence . In our setting, HPV prevalence is among the highest ever reported in HIV-infected women, at 95% , with the risk for hrHPV infection estimated at five times that of HIV uninfected women . Recently, in South Africa, cART use has again been reported to be associated with decreased risk for progression, persistence and post-excision recurrence of HPV-associated SIL in women [1-3].
The lack of consensus on the effect of cART on cervical HPV infection provided us with the impetus for this research question to be revisited. We, therefore, conducted a large prospective cohort study which enrolled women with a wide range of immunological as well as cervical cytological status. Using advanced and more robust statistical techniques, we chose to compare the effect of cART on each individual HPV genotype to the effect cART has on HPV-16 (which causes more than half of cases of cervical cancer ).
Study design, population, setting, inclusion criteria and outcomes definitions
In this prospective cohort study, women known with HIV infection were approached for enrolment provided they were cART-naive. Patients known with any cancer (including cervical) and pregnant women were excluded. The study was approved by the Stellenbosch University Human Research Ethics Committee (N09/04/1065).
Follow-up visits were conducted twice a year. At each visit, the clinical, obstetric and sexual behaviour data were recorded and a cervical Pap test was performed. Referral for colposcopy evaluation occurred after a single abnormal cytology result.
The decision to initiate ART was not determined by the study protocol, but was based on the South African Government Protocol for the Initiation of ART. Adherence to the prescribed regimen was ascertained by a combination of the pharmacy pick-up and patient self-report methods. Although interruptions were recorded, distinction was not made between those that were physician-initiated and those that were patient-initiated.
Response to ART was measured by testing the CD4+ T-cell count and plasma HIV-RNA viral load determination at study enrolment and at each six monthly follow-up visits. The viral load determination done at baseline is referred to as the viral set point.
Cervical HPV samples were collected with a Cervexbrush at baseline and 6 monthly intervals, placed in PreservCyt (Hologic, Bedford, Massachusetts, USA) solution and stored at -80°C for later analysis by Roche linear array, as previously described . Test strip results were read independently by two molecular scientists. Both were blinded as to the ART and cytological status detected in any of the study participants. Proviral DNA was isolated from sample specimens using the Qiagen DNA Blood and Tissue Kit (Qiagen, Hilden, Germany). HIV proviral DNA copy numbers were measured by the real-time quantitative PCR method (qPCR) on the DNA extracted from the cervical specimens for the HPV analysis, as previously described . Primers and probes were obtained from Integrated DNA Technologies (South Africa) and targets part of the HIV-1 LTR-gag fragment for amplification . The cell numbers were calculated by using pc-CCR5 (obtained from the NIH AIDS reference reagent laboratory) as a reference gene for cellular copy number, while pMJ4 was used as a positive control to determine HIV-1 copy number . Negative and positive-control DNAs were included in all reactions. Samples were run in duplicate on a CFX96 Touch Real-Time PCR Detection system (Bio-Rad Ltd., Hercules, California, USA) and analyses completed using CFX Manager software.
A SQL database (Microsoft SQL Server 2008) was created to capture, store and retrieve study data. Once extracted, the data were analysed using Microsoft Excel 2010, XLSTAT (version 2013.3.01) and Stata (version 11; Stata Corp., College Station, Texas, USA). Baseline demographic variables were assessed using Pearson's chi-square test or Fisher's exact test to compare proportions, and Student's t test to compare means between groups.
For the purpose of this analysis, we defined ART in two ways. The first variable, 'cART treatment status', was used as a binary variable to indicate if the patient was using cART at a visit. A patient who had as yet not commenced treatment, or had an interruption of more than 1 month, was labelled 'not on treatment' at that visit. For the other variable, 'time on cART', the cART start date was recorded and the time that had elapsed since the first date of administration was calculated at each visit. For women not initiated to cART, the value remained 0 for the time until they started taking antiretrovirals. This variable would not be influenced by adherence to the cART regimen.
We modelled the probability of positive detection of HPV using mixed-effects logistic regression as a function of cART treatment status or time on cART.
Mixed models incorporate a random effect to account for within-participant correlation due to the repeated measurements . Generalized estimating equation (GEE) procedure, assuming an exchangeable correlation structure, was used to estimate the population average effect of ART expressed as an odds ratio (OR) and the associated 95% CIs. All models were adjusted for 'excision treatment of cervical neoplasia sexual activity' and the CD4+ T-cell count as time-dependent variables. Age was included as a non-time-dependent variable due to the relatively short follow-up time. We compared the effect of cART on the overall risk of detection of each HPV genotype to the effect of cART on HPV-16 genotype by estimating the ratio of OR for each genotype to the OR for HPV-16 genotype using the GEE models. Thereafter, using meta-analysis techniques of combining results across subgroups, we estimated the overall effect of cART on hrHPV genotypes (HPV-16, 18, 31, 33, 35, 39, 45, 52, 58) and all others as low-risk HPV (lrHPV) genotypes compared to the effect of cART on HPV-16 genotype alone. These results were expressed as the ratio of ORs and the associated 95% CIs. For all analyses, P values reported were two-tailed at the 5% alpha level.
From November 2009 to October 2011, a total of 304 participants were consented for participation in the study. Two women were excluded, as they did not meet the eligibility criteria. Also, one woman immediately withdrew consent, leaving 1115 specimens from 301 women available for laboratory analysis. Four cervical samples were discarded due to unsuccessful DNA extraction. One of these smears represented the only available specimen from one woman, which led to further exclusion of this participant from the analysis. From the final 300 women included, we thus had 1111 available DNA smears. Lastly, a further 10 smears had no associated CD4+ T-cell count or HIV-RNA viral load determination available, leading to 1101 cervical smear specimens from 300 women being included in the analysis.
Table 1 summarizes the characteristics of women included in the study analysis. Of the 300 patients enrolled, 204 (68%) were initiated on ART at some point during study follow-up time. The other 96 (32%) did not meet the cART initiation criteria and remained ART-naive. Of the 204 women who had to be initiated on cART, 132 did so within the first month of study enrolment. Participants not initiated on cART during follow-up time had higher CD4+ T-cell count at study enrolment and a lower viral set point, were younger and less likely to report abstinence from sexual intercourse.
Also of note is that cART status had no bearing on management of cervical dysplasia. Indications for treatment were not different in women according to the cART status. Excision treatment during study follow-up was equal both in proportion, completeness and indication whether a woman had to be initiated on cART or not. In this study, only one patient underwent cold-knife cone (CKC) excision treatment. Distinction as to the type of excision treatment was thus not included in the analyses of factors influencing cervical HPV infection.
Exposure to cART was a sum of 98 442 days for the 204 of 300 study participants who were initiated on cART. All patients treated were initiated on treatment containing a non-nucleoside analogue reverse transcriptase inhibitor (NNRTI), which is a standard regimen at the time. Only seven patients were exposed to a lopinavir-containing regimen for a total of 2435 days during the study, which represents 2.48% of the total time of regimen exposure. The type of cART regimen was therefore not included in analysis.
Number of human papilloma virus genotype detections
Among all the subtypes, HPV-16 was the most frequently detected oncogenic subtype (230 positive detections on 1101 smears, 20.9%). HPV-84 was the most common non-oncogenic subtype (303 positive detections, 27.5%). There were fewer than 20 positive detections for HPV-40 and HPV-64.
Human papilloma virus prevalence in study participants during follow-up time
The prevalence of HPV types was calculated as the number of study participants who ever had a positive detection of a given subtype. Of the 300 participants analysed, 287 (94.3%) had at least one hrHPV subtype during follow-up. Of the oncogenic subtypes, HPV-16 was detected in at least one smear from 127 women (42.3%), whereas HPV-84 was detected in at least one smear from 183 women (61.0%).
Influence of combination antiretroviral therapy status on cervical human papilloma virus infection
In the unadjusted GEE population-averaged logistic regression models of the detection of any HPV genotype, and cART status (receiving cART or not) as the only predictor, the risk for detection of HPV was reduced by 77% (OR 0.23, 95% CI 0.15-0.37). cART significantly reduced the detection risk for cervical hrHPV (OR 0.33, 95% CI 0.24-0.44) and HPV-16 (OR 0.50, 95% CI 0.37-0.67). After adjusting for the other covariates, time on cART (months that had elapsed since cART was first started), the cervical HIV DNA proviral load (pVL) and whether the patient was sexually active were significantly associated with detection of any HPV. The plasma HIV-RNA level was not a predictor of risk for cervical HPV detection (Table 2).
In unadjusted models, CD4+ T-cell count less than 200 cells/μl was significantly associated with detection of any HPV genotype (OR 3.78, 95% CI 1.87-7.64). The effect was not as strong for hrHPV, and not significant if HPV-16 detection was the only outcome. We therefore included CD4+ in our adjusted model, but found that time since cART was started had a stronger effect on HPV detection risk than CD4+ T-cell count, and this also applied to HPV-16 detection risk (Table 2).
In an ad-hoc subgroup analysis, we also compared the effect of cART on the lrHPV genotype group to the effect of cART on HPV-16, and found that the protective effect of cART on the detection risk for the lrHPV genotype group was significantly less than the protective effect of cART on the detection risk of HPV-16 (ratio of ORs 1.35, 95% CI 1.22-1.50). The pooled effect of cART on the detection risk of hrHPV genotypes was not significantly different from the effect of cART on HPV-16 (ratio of ORs 1.05, 95% CI 0.86-1.25) (Fig. 1).
From this study, we learned that cART exposure reduced the risk of detection of any HPV type by 77%. Adjusting for other covariates, a longer time of exposure to cART significantly reduced the odds of HPV detection. For every additional month of cART since initiation, the risk of detection of any HPV type decreased by 9%. The effect was also significant on HPV-16 genotype alone. We also found that lrHPV genotypes were less influenced by cART as compared to HPV-16 genotypes. Furthermore, not surprisingly, the effect of cART on the detection risk for the most prevalent hrHPV genotype HPV-16 was not different to the effect of cART on the detection risk of the other high-risk oncogenic genotypes grouped together. The effect of cART seemed to be immunology-driven. There was an increased risk for any HPV detection at CD4+ T-cell count below 200 cells/μl (OR 3.78, CI 1.87-7.64), but when adjusted, the time of cART exposure again remained the stronger predictor of risk. There was a weak association of HPV-16 with very low CD4+ T-cell count.
We were curious to know why previous reports on the effect of cART on cervical HPV infection have been so conflicting, with several studies failing to show any effect. Firstly, we postulated that HPV subtype differences could have contributed to the confusion. It has become possible to test for an increasing number of subtypes over the past 15 years, and genotypes differ in their oncogenic potential. HPV-16 and HPV-18 together cause only about 3.8% of cervical infections, but 70% of cases of cervical cancer . Because we were not just interested in the prevention of cervical cancer alone, but the impact on cervical dysplasia, we applied the method which can test for the largest number of HPV genotypes. Although cervical infection with oncogenic HPV types was not decreased more after initiation with cART than was HPV-16, we saw that infection with the non-oncogenic types was reduced significantly less than HPV-16. We conclude that subtype differences of the effect of cART on HPV, that is oncogenic vs. non-oncogenic, may have been one of the possible contributing factor in conflicting previously published results.
We included in the analysis two time-dependent variables as indicators of ART administration. The first, a binary variable, was indicative of receiving cART, which was determined by self-reporting of ART use. Because the effect of cART may be longer-lived than the actual administration status of the medication, we chose to include another time-dependent variables, time since cART first started, which depicts the months that had passed since the cART initiation date. This variable would therefore be independent of cART adherence or interruptions (and the HIV-RNA level). This variable was chosen to reflect a possible immune reconstitution memory effect, maybe even on local cervical level of indicators not measured.
We were intrigued by the apparent lack of independent association of plasma-HIV-RNA levels with the risk for HPV detection and therefore included cervical HIV-DNA pVL in the adjusted model. There was an increased risk of detection of any HPV type with higher HIV-DNA pVL from the cervix. This remained significant when adjusting for the time that had passed since cART was started.
Our study has several strengths: use of well defined and documented outcome (positive detection of HPV) and exposure (e.g. cART treatment status or time on cART) variables and adjusted for other factors such as CD4+ T-cell count in a time-dependent manner; the comparison of the effect of cART on HPV genotypes (to include hrHPV genotypes vs. lrHPV genotypes as separate groups) to the effect cART has on HPV-16; the use of advanced and robust statistical procedures to quantify the probability of positive detection of HPV using mixed-effects logistic regression (with a random effect) as a function of cART treatment status or time on cART and GEE which estimate correctly the model parameters under specific correlation structures. As possible limitations, it should be remembered that given the observational nature of our cohort study design, other unknown (and therefore not measured/adjusted for) confounding factors may be at play to account for some of our findings; our results does not necessarily infer causality between detection of cervical HPV and HIV, but may both be time-dependent result of initiation of cART. Indeed, it appears that an immune reconstitution effect on the level of the cervix may be slower than what occurs in the plasma. Further research to investigate mucosal immune restoration after initiation of cART and cellular localization of cervical HIV should be considered.
In conclusion, cART reduces cervical HPV infection in a time-dependent and immunology-driven manner. Reducing cervical HPV co-infection in women living with HIV infection should be seen as one of the many benefits of initiating the combination anti-HIV medication. This opinion is further supported by the strength of ART programmes as part of the prevention strategies, especially in countries of high HIV prevalence.