Millions of women worldwide use hormonal contraception, the most widely used being oral contraceptives (OCs) and progestin-only injectables, such as depot medroxyprogesterone acetate (DMPA).1 The 150 mg aqueous injection of DMPA is marketed as Depo-Provera. A recent systematic review of 22 prospective studies, including studies on female sex workers, concluded that OCs do not increase HIV acquisition risk, but that the evidence for progestin-only injectables is conflicting.2 Several biological mechanisms have been postulated, but it is unclear which, if any, are clinically relevant. The proposed mechanisms include direct effects of DMPA, or the lack of estrogen induced by high levels of DMPA, on genital immune activation (including the attraction of CD4+/CCR5+ target cells for HIV) and/or epithelial and mucosal remodeling, or indirect effects: DMPA or lack of estrogen might increase vulnerability to other sexually transmitted infections (STIs) or change the composition of the cervicovaginal microbiome (VMB), which in turn might cause genital immune activation or mucosal remodeling.3–5
A 2009 review concluded that OC use is associated with an increased risk of Chlamydia trachomatis, but the evidence for DMPA is weaker.4 Furthermore, reports of OC use and Neisseria gonorrhoeae are inconsistent, and DMPA does not seem to increase risk.4–6 Studies on hormonal contraception and Trichomonas vaginalis and syphilis are scarce, but most show (nonsignificant) trends toward protective effects.4 Long-term use of OCs has been associated with an increased risk of cervical cancer, but there is no proven association between OCs and DMPA and human papillomavirus (HPV) acquisition.7 Finally, herpes simplex virus type 2 (HSV-2) was identified as an important effect modifier of the relationship between DMPA and HIV acquisition in one pivotal study,8 but this has not been confirmed by others.9 Prospective data on hormonal contraception and HSV-2 acquisition risk are scarce.
Two recent systematic reviews showed that hormonal contraception reduces bacterial vaginosis (BV; Nugent score 7–10) and intermediate microbiota (Nugent score 4–6), but that OC use may increase vaginal candidiasis.10,11 Only 4 studies were identified that had used molecular techniques to characterize the VMB.10 These confirmed that high estrogen levels favor a VMB composition dominated by “healthy” Lactobacillus species, but the effects of progesterone were not well studied. A similar association between pregnancy (i.e., a progesterone-dominated state) and a Lactobacillus-dominated VMB has been described.12
Several epidemiological studies have shown an association between STIs and increased risk of HIV acquisition, with the highest population-attributable risk for HSV-2 (reviewed in Ref. 13). Similarly, several studies have shown associations between BV/intermediate Nugent score, candidiasis, and HIV acquisition.13 A 2014 systematic review of all 63 studies that had used molecular methods between 2008 and 2013 to characterize the VMB found 11 cross-sectional studies comparing the VMB of HIV-positive and HIV-negative women: most found trends toward decreased lactobacilli (and particularly Lactobacillus crispatus).14 No prospective studies were identified.
The original aim of our study was to determine HIV and STI incidence in Rwandan female sex workers. In this secondary analysis, we evaluated the effects of hormonal state (i.e., hormonal contraception use or pregnancy) on the prevalence and incidence of STI pathogens, BV by Nugent score category, and phylogenetic VMB composition. We hypothesized that the presence of STI pathogens or VMB bacteria that activate the immune system is an intermediate step on the causal pathway between hormonal state and HIV acquisition.
The Kigali HIV Incidence Study was a prospective cohort study conducted in 2007 to 2008 at Rinda Ubuzima, a nongovernmental research clinic in Kigali, Rwanda.15 The study was approved by the National Ethics Committee, Rwanda, and the Columbia University Medical Center Institutional Review Board, USA. All participants provided written informed consent.
Eight hundred adult female sex workers with unknown HIV status were tested for HIV, HSV-2, and pregnancy in a screening survey (Fig. 1).15 Subsequently, a subset of 397 HIV-negative nonpregnant women was enrolled in a prospective cohort study. Cohort participants visited the study clinic quarterly for 1 year (referred to as the month 3 [M3], M6, M9, and M12 visits), and one additional time thereafter (the year 2 [Y2] visit; 3.5–16 months after M12). An additional 141 women who tested HIV positive in the screening survey attended the Y2 visit. At each visit, women were interviewed and counseled (including family planning counseling and provision of condoms free of charge), provided blood samples, and were tested for HIV and pregnancy. Women self-sampled vaginal swabs for diagnostic STI testing at enrollment and M12, and a pelvic examination (including cervicovaginal sampling by a clinician) was performed at M6 and Y2.
Wet mounts and rapid tests were conducted in real time (see further); all other samples were processed the same day and stored at −80°C until testing. Women were tested for HIV and pregnancy at all visits; syphilis, N. gonorrhoeae, C. trachomatis at enrollment, M6, M12; T. vaginalis, BV, and candidiasis at enrollment, M6, M12, Y2; HSV-2 at screening (and those who tested negative were tested again at M12 and Y2, with retroactive testing of stored samples if seroconverted); and HPV types at M6 and Y2 (Fig. 1).
Testing was conducted on-site at Rinda Ubuzima in Kigali, Rwanda, unless stated otherwise. Whole blood was tested for HIV using the First Response rapid test (Premier Medical Corporation, Nani Daman, India), followed by Uni-Gold rapid test (Trinity Biotech Plc, Bray, Ireland) when the first test was positive, and the Capillus HIV-1/HIV-2 rapid test (Trinity Biotech Plc) as tie-breaker if needed. A serum pregnancy test (Fortress Diagnostics hCG serum pregnancy test, Antrim, UK) was used to screen for pregnancy. Plasma specimens were tested for HSV-2 (HerpeSelect 2 ELISA; Focus Diagnostics Inc, Cypress, CA; with index ≥3.5 defined as positive) and syphilis serology (Spinreact Rapid Plasma Reagin test with confirmation by Spinreact T. pallidum Haemagglutination test, Girona, Spain). An endocervical swab was used for N. gonorrhoeae and C. trachomatis testing using the Amplicor CT/NG PCR test (PCR Roche Diagnostic Corp, Indianapolis, IN; testing done at the Institute of Tropical Medicine, Antwerp, Belgium), and a vaginal swab was used for T. vaginalis InPouch (Biomed Diagnostics, White City, OR) testing. Vaginal swabs were also used to prepare a wet mount and Gram stain slide to determine the presence of more than 20% clue cells and trichomonads (a participant was considered positive for T. vaginalis if she tested positive on either the wet mount or the InPouch test), and for Gram stain Nugent scoring. KOH was added to the wet mount slide to detect an amine smell and visualize yeasts. The vaginal pH was measured by pressing a pH paper strip against the vaginal wall (pH range 2–9 with 0.5 increments). Spatulas and cytobrushes were rinsed in Preservcyt transport medium (ThinPrep Pap Test; Cytyc Corporation, Boxborough, MA), which was stored at −80 °C until batched testing at the end of the study in specialized laboratories for HPV genotyping (Linear Array HPV Genotyping Test, Roche Molecular Systems; testing done at the Institute of Tropical Medicine) and phylogenetic microarray analysis (TNO, Zeist, the Netherlands).
M6 and Y2 Preservcyt samples of 174 women were analyzed using a DNA hybridization microarray containing 251 probes targeting urogenital microorganisms. Women with HIV and STIs were overrepresented in this subsample by design. Microarray design, sample preparation, amplification, and hybridization were described elsewhere.16 Of the 251 probes, 66 16S probes were bacterial species-specific, 56 16S probes targeted multiple bacterial species within one genus, 105 16S probes were specific at a higher bacterial taxonomic level, 5 were groEL probes, 16 were 18S probes, and 3 were viral probes. We focused our clustering analyses on all 251 probes, and we performed additional analyses on 20 16S probes representing the most abundant bacterial species/genera. Microarray data were normalized by calculating signal (S) over background (B) ratios per spot, setting the ratio to one when signals were not confidentially above background, and by Lowess smoothing.17 Women were assigned to a VMB cluster if they had at least 70% probability of belonging to a cluster identified by neighborhood co-regularized multiview spectral clustering.
Data were double-entered and analyzed using Stata version 12.0 (StataCorp, College Station, TX), MATLAB version R2012a (The MathWorks, Natick, MA), and R i386 (R Foundation for Statistical Computing, Vienna, Austria). The hormonal status of all women was determined at each visit: currently pregnant by pregnancy test (regardless of self-reported use of contraception), self-reported current use of OCs or injectables, or nonexposed controls (women who did not report hormonal contraception and were not pregnant). When women reported to use a certain family planning method at a study visit during which an outcome of interest was measured, we assumed that they used the same method during the entire interval between this visit and the previous visit during which the same outcome was measured.
Differences in baseline prevalence of outcomes by hormonal status were tested by multivariable logistic regression for STIs, Nugent score 7–10 or 4–6 (each compared to 0–3), and candidiasis by wet mount; unadjusted Fisher exact tests for VMB clusters; and Kruskal-Wallis 1-way analysis of variance tests for normalized S/B ratios of individual VMB species/genera. Differences in incidence were determined by multivariable logistic regression with random effects. This model was chosen over Cox or Poisson proportional hazards models due to the relatively low frequency of follow-up visits. Sensitivity analyses with Cox or Poisson regression were conducted, showing similar results (not shown). In all multivariable prevalence and incidence analyses, adjustments were made for age, educational level, years worked as sex worker, and breast-feeding (to account for differences in endogenous hormone levels) as reported at baseline, consistent condom use as reported at the current or nearest visit, antibiotic use in the past 14 days, and ever having used antibiotics before the outcome assessment date. In all incidence analyses, an additional adjustment was made for the differences in the duration of the interval between repeated assessments. Missing data were imputed using the mode (binary variables) or median (categorical variables). When the use of antibiotics was unknown and no previous STIs were detected, we assumed no antibiotics were used.
Study Participant Characteristics
Of the 800 screening survey participants, 61 (7.6%) tested positive on the pregnancy test, 49 (6.1%) reported to use OCs, 97 (12.1%) injectables, and 3 a hormonal implant. Intrauterine devices or sterilization were not reported. Of the remaining 590 women, 439 reported to use condoms only, 88 reported periodic abstinence or coitus interruptus with or without condoms, 60 reported not to use family planning, and 3 had missing data. The women using a hormonal implant or having missing data (n = 6) were excluded from all analyses, resulting in a control group of 587 women.
The age of the participants ranged from 18 to 49 years with no differences between hormonal status categories (Table 1). Most women were never married and had primary school or less education. Pregnant women were less likely to currently breastfeed, and injectable users had a slightly higher median number of lifetime pregnancies. All but 5 women were working as a sex worker at the time of the survey with a median duration of 3 years. Most women reported to wash inside the vagina at least once daily, and a total of 74.1% reported to use condoms. At screening, the HIV prevalence was 24.0% and the HSV-2 prevalence was 60.6%.
Of the 397 HIV-negative, nonpregnant women who enrolled in the cohort study, 338 attended all follow-up visits, 53 were lost to follow-up, 4 withdrew early, and 2 died15 (Fig. 1). The 59 women who were lost were not significantly different from the 338 women who completed follow-up, except that their median income was lower (8000 vs. 12,000 RWF/mo, respectively; data not shown). Contraceptive switching was common. Of the 49 women who reported to use OCs at screening, 24 were enrolled in the cohort: 12 consistently used OCs until study exit, but 5 switched to injectables and 1 to an implant, and 6 became pregnant. Of the 97 women who reported to use injectables at screening, 48 were enrolled in the cohort: 26 consistently used injectables until study exit, but 4 switched to OCs and 1 to an implant, and 17 became pregnant. Thirty-two women started using OCs, and 78 women injectables, after enrolment into the cohort study.
Prevalence of STIs
In multivariable logistic regression analysis, pregnancy was associated with a lower HIV prevalence (adjusted odds ratio [aOR], 0.45; 95% confidence interval, 0.22–0.92), injectable use with a higher HSV-2 prevalence (aOR, 2.13; 1.26–3.59), and OC use with a higher HPV (any type) prevalence (aOR, 3.10; 1.21–7.94) (Table 2). No statistically significant associations between hormonal status and prevalence of bacterial STIs, Nugent score 7-10/4-6, or candidiasis were found (Table 2).
Incidence of STIs
In multivariable logistic regression analysis with random effects, OC use was associated with a significantly higher C. trachomatis incidence (aOR, 6.1; 1.6–23.8; Table 3). Although T. vaginalis incidence was not associated with hormonal status in comparisons with controls, a significantly higher incidence was found during pregnancy than during injectable use (aOR, 0.29; 0.11–0.78). Pregnancy was also associated with a higher candidiasis incidence (aOR, 2.1; 1.1–4.1).
We previously described 6 VMB clusters in this study population (Fig. 216;). Briefly, clusters R-I and R-II had the lowest total bacterial load and diversity and were dominated by L. crispatus and Lactobacillus iners, respectively. Cluster R-V had an intermediate bacterial load and diversity, and clusters R-III, R-IV, and R-VI had the highest bacterial loads and diversity. Clusters R-III to R-VI were not dominated by any one species but consisted of different combinations of (facultative) anaerobic bacteria, including high abundance of Gardnerella, Prevotella, and Atopobium spp. and lower abundance of Dialister, Megasphaera, Mobiluncus, and BV-associated bacterium type I. Clusters R-III to R-VI had a varying abundance of L. iners, with the lowest abundance in cluster R-III. The association between hormonal status and VMB clusters was not statistically significant (Fisher exact test: P = 0.72; Table 4). However, it is interesting to note that none of the OC users and pregnant women were assigned to the R-I L. crispatus cluster (compared with 7.3% of controls and 7.9% of injectable users) and that a higher proportion of them were assigned to the R-II L. iners cluster (52.4%–57.1% vs. 29.0%–36.5%; Table 4).
Of the 20 microarray probes targeting the most abundant VMB bacteria in our study population, none showed an overall significant association with hormonal status (Fig. 3). Of the 4 probes that were associated with hormonal status with an overall P < 0.1, Prevotella uncultured bacterium, Sneathia/Leptotrichia amnionii, and Mycoplasma species showed lower normalized S/B ratios in the OC users compared with controls (P = 0.03, P = 0.05, and P = 0.02, respectively).
In this population of African sex workers highly exposed to STIs, OC use was associated with increased HPV prevalence and C. trachomatis incidence and a lower semiquantitative abundance of bacteria associated with BV (Prevotella, Sneathia/L. amnionii, and Mycoplasma species). Injectable use was associated with an increased HSV-2 prevalence, but no other clear differences with controls were identified. Pregnancy was associated with a lower HIV prevalence but a higher candidiasis incidence.
Our data confirm that OC use is associated with C. trachomatis and may also be associated with N. gonorrhoeae, although the evidence for the latter remains conflicting.4–6 Furthermore, the evidence for a relationship between OCs and HPV is rapidly accumulating and should be carefully reviewed.7,18–20 We did not find an association between OC use and HPV incidence, perhaps due to inadequate statistical power. We therefore do not know whether the association between OCs and HPV prevalence in our study population can be explained by higher rates of acquisition or by infection persistence. Only one study has investigated the latter and found no association.21 The dominant hypothesis to explain the relationships between OCs, cervical infections, and HPV is that estrogen facilitates cervical ectopy, thereby exposing a larger area of single-layer columnar epithelium to pathogens.22,23
We found an association between injectable use and HSV-2 prevalence, but not with HSV-2 incidence, in our study population. Although we did not record the type of injectable used, family planning programs in Rwanda mostly offer DMPA, and only occasionally norethisterone enanthate (Association Rwandaise pour le Bien-Etre Familial, personal communication, February 2013). Cross-sectional epidemiological studies reporting on the relationship between hormonal contraception and HSV-2 have been inconclusive,4 but 2 studies have shown that DMPA increases HSV-2 genital shedding in those already infected.24,25 Furthermore, laboratory studies have shown that DMPA treatment resulted in a 100-fold increase in genital HSV-2 susceptibility in mice26 and prevented a protective immune response in HSV-2-positive mice.27
Pregnant women had a significantly lower HIV prevalence in our study, but this may be due to reverse causality: HIV-positive women are known to be less fertile.28 Pregnancy was also associated with vaginal candidiasis, as has been described before.29 We found a slightly higher incidence of T. vaginalis during pregnancy, but this only reached significance when comparing to injectable use (which was associated with a lower T. vaginalis incidence than in controls) and not when comparing to controls.
Based on our recent review of the literature on hormonal contraception and the VMB,10 we expected to find more pronounced associations between hormonal status and VMB composition than we did. However, none of our findings contradict the conclusions of our review and of a recent molecular VMB study in pregnant women30: we found nonsignificant trends toward protection against incidence of BV or intermediate microbiota by Nugent scoring in OC users and pregnant women, and OC users had significantly lower semiquantitative abundance of several BV-associated bacteria than controls. Our findings may have been biased toward the null due to nondifferential misclassification of hormonal exposure status (see below) or because other determinants of microbiome composition masked the hormonal effects in this group of women with high coital frequency and high STI prevalence. In the case of pregnant women, reverse causality cannot be ruled out. Clinical BV is known to be a risk factor for various adverse pregnancy outcomes, and it is therefore possible that women with a lactobacilli-dominated VMB are more fertile or better able to carry a pregnancy to term.31,32
Some limitations of our study should be noted. Selection bias is common in all observational studies and was likely present in our study as well. Although pregnancy was ascertained by serum hCG test at each visit, hormonal contraceptive (and condom) use was self-reported, switching between hormonal exposure groups was common, data on adherence and phase of the menstrual cycle were lacking, and timing of conception was imprecise, all of which could have resulted in nondifferential exposure misclassification. Sexually transmitted infection, BV, and VMB outcomes were laboratory confirmed, but the timing of STI or BV acquisition was imprecise. We could not control all analyses for potential confounders due to the small sizes of some hormonal exposure categories, and even when adjustments were made, residual confounding cannot be ruled out. Strengths and limitations of the phylogenetic microarray are described elsewhere.16 Despite these limitations, our study does provide exploratory insights into relationships that have not been studied previously and cannot easily be investigated in study populations with lower STI rates.
In conclusion, use of OCs and injectables was associated with prevalence and incidence of several STIs in our study population, but not with hormone-related variations in the VMB that are likely to increase HIV risk. The increased HSV-2 prevalence in injectable users might explain a potentially higher HIV risk in these women, but further research is needed to confirm these results and unravel biological mechanisms.
1. United Nations Population Fund. The global programme to enhance reproductive health commodity security: annual report 2013. Available at: http://www.unfpa.org/public/home/publications/pid/17481
. Accessed November 23, 2014.
2. Polis CB, Phillips SJ, Curtis KM, et al. Hormonal contraceptive methods and risk of HIV acquisition in women: A systematic review of epidemiological evidence. Contraception 2014; 90: 360–390.
3. Blish CA, Baeten JM. Hormonal contraception and HIV-1 transmission. Am J Reprod Immunol 2011; 65: 302–307.
4. Morrison CS, Turner AN, Jones LB. Highly effective contraception and acquisition of HIV and other sexually transmitted infections. Best Pract Res Clin Obstet Gynaecol 2009; 23: 263–284.
5. Pettifor A, Delany S, Kleinschmidt I, et al. Use of injectable progestin contraception and risk of STI among South African women. Contraception 2009; 80: 555–560.
6. Gursahaney PR, Meyn LA, Hillier SL, et al. Combined hormonal contraception may be protective against Neisseria gonorrhoeae
infection. Sex Transm Dis 2010; 37: 356–360.
7. Veldhuijzen NJ, Snijders PJ, Reiss P, et al. Factors affecting transmission of mucosal human papillomavirus. Lancet Infect Dis 2010; 10: 862–874.
8. Morrison CS, Richardson BA, Mmiro F, et al. Hormonal contraception and the risk of HIV acquisition. AIDS 2007; 21: 85–95.
9. Baeten JM, Benki S, Chohan V, et al. Hormonal contraceptive use, herpes simplex virus infection, and risk of HIV-1 acquisition among Kenyan women. AIDS 2007; 21: 1771–1777.
10. van de Wijgert JH, Verwijs MC, Turner AN, et al. Hormonal contraception decreases risk of bacterial vaginosis but oral contraceptives may increase risk of vaginal candidiasis: implications for HIV transmission. AIDS 2013; 27: 2141–2153.
11. Vodstrcil LA, Hocking JS, Law M, et al. Hormonal contraception is associated with a reduced risk of bacterial vaginosis: A systematic review and meta-analysis. PLoS One 2013; 8: e73055.
12. Romero R, Hassan SS, Gajer P, et al. The composition and stability of the vaginal microbiota of normal pregnant women is different from that of non-pregnant women. Microbiome 2014; 2: 4.
13. Hayes RJ, Watson-Jones D, Celum C, et al. Treatment of sexually transmitted infections for HIV prevention. AIDS 2010; 24( suppl 4): S15–S26.
14. van de Wijgert JHHM, Borgdorff H, Verhelst R, et al. The vaginal microbiota: What have we learned after a decade of molecular characterization? PLoS One 2014; 9: e105998.
15. Braunstein SL, Ingabire CM, Kestelyn E, et al. High human immunodeficiency virus incidence in a cohort of Rwandan female sex workers. Sex Transm Dis 2011; 38: 385–394.
16. Borgdorff H, Tsivtsivadze E, Verhelst R, et al. Lactobacillus
-dominated cervicovaginal microbiome associated with reduced HIV/STI prevalence and genital HIV load in African women. ISME J 2014; 8: 1781–1793.
17. Quackenbush J. Microarray data normalization and transformation. Nat Genet 2002; 32: 496–501.
18. Oliveira FA, Ehrig V, Lang K, et al. Human papillomavirus genotype distribution and risk factors for infection in women from a small municipality in north east Brazil. Int J STD AIDS 2012; 23: e5–e10.
19. Ozcan ES, Taskin S, Ortac F. High-risk human papillomavirus prevalence and its relation with abnormal cervical cytology among Turkish women. J Obstet Gynaecol 2011; 31: 656–658.
20. Longatto-Filho A, Hammes LS, Sarian LO, et al. Hormonal contraceptives and the length of their use are not independent risk factors for high-risk HPV infections or high-grade CIN. Gynecol Obstet Invest 2011; 71: 93–103.
21. Maucort-Boulch D, Plummer M, Castle PE, et al. Predictors of human papillomavirus persistence among women with equivocal or mildly abnormal cytology. Int J Cancer 2010; 126: 684–691.
22. Bright PL, Norris Turner A, Morrison CS, et al. Hormonal contraception and area of cervical ectopy: A longitudinal assessment. Contraception 2011; 84: 512–519.
23. De Villiers EM. Relationship between steroid hormone contraceptives and HPV, cervical intraepithelial neoplasia and cervical carcinoma. Int J Cancer 2003; 103: 705–708.
24. Mostad SB, Kreiss JK, Ryncarz AJ, et al. Cervical shedding of herpes simplex virus in human immunodeficiency virus–infected women: effects of hormonal contraception, pregnancy, and vitamin A deficiency. J Infect Dis 2000; 181: 58–63.
25. Cherpes TL, Melan MA, Kant JA, et al. Genital tract shedding of herpes simplex virus type 2 in women: Effects of hormonal contraception, bacterial vaginosis, and vaginal group B Streptococcus
colonization. Clin Infect Dis 2005; 40: 1422–1428.
26. Kaushic C, Ashkar AA, Reid LA, et al. Progesterone increases susceptibility and decreases immune responses to genital herpes infection. J Virol 2003; 77: 4558–4565.
27. Gillgrass AE, Ashkar AA, Rosenthal KL, et al. Prolonged exposure to progesterone prevents induction of protective mucosal responses following intravaginal immunization with attenuated herpes simplex virus type 2. J Virol 2003; 77: 9845–9851.
28. Glynn JR, Buve A, Carael M, et al. Decreased fertility among HIV-1–infected women attending antenatal clinics in three African cities. J Acquir Immune Defic Syndr 2000; 25: 345–352.
29. Bauters TG, Dhont MA, Temmerman MI, et al. Prevalence of vulvovaginal candidiasis and susceptibility to fluconazole in women. Am J Obstet Gynecol 2002; 187: 569–574.
30. Aagaard K, Riehle K, Ma J, et al. A metagenomic approach to characterization of the vaginal microbiome signature in pregnancy. PLoS One 2012; 7: e36466. Doi: 10.1371/journal.pone.0036466.
31. Leitich H, Kiss H. Asymptomatic bacterial vaginosis and intermediate flora as risk factors for adverse pregnancy outcome. Best Pract Res Clin Obstet Gynaecol 2007; 21: 375–390.
32. Low N, Chersich MF, Schmidlin K, et al. Intravaginal practices, bacterial vaginosis, and HIV infection in women: Individual participant data meta-analysis. PLoS Med 2011; 8: e1000416. doi: 10.1371/journal.pmed.1000416.