Methods

In February 1993, a prospective cohort study of women attending a municipal sex worker clinic was initiated in Mombasa, Kenya. One of the primary aims of this cohort has been to determine risk factors for HIV-1 and STIs. Detailed procedures have been described previously.6 Briefly, women were offered confidential HIV-1 counselling and testing, and HIV-1-seronegative women were invited to enroll. At enrollment and monthly follow-up visits, women completed face-to-face interviews using standardised questionnaires to ascertain medical, gynaecological and sexual history. A physical examination was performed, including a pelvic examination with collection of specimens for laboratory diagnosis of STIs. The presence of genital ulcers was recorded. A blood sample was obtained for HIV-1 serological testing and aliquots from each sample were archived at −80°C.

Women with STI symptoms were provided with syndromic management in accordance with Kenya Ministry of Health guidelines. Women were invited to return 1 week after each monthly visit for laboratory results, and additional treatment was provided for infections detected on laboratory testing that had not been recognised at the time of the examination. At each visit, HIV-1 risk reduction counselling and free condoms were provided. All participants provided informed consent.

Screening for HIV-1 infection was performed by ELISA (Detect HIV Biochem ImmunoSystems, Montreal, Canada). Positive samples were confirmed using a second ELISA (Recombigen, Cambridge Biotech, Galway, Ireland or Vironostika, Biomerieux, Marcy l′Etoile, France). Serological testing for HSV-2 was performed on archived samples using a type-specific HSV-2 gG based ELISA (HerpeSelect 2, Focus Diagnostics, Cypress, California, USA). An index value of >1.1 (the ratio of the optical density (OD) of the sample to the OD of the standard calibrator) was considered to be positive for HSV-2 as per the manufacturer’s instructions. While early data suggested that index values between 1.1 and 3.5 had reduced specificity for defining HSV-2 serostatus in African specimens,7 a recent study comparing the HerpeSelect and Kalon assays showed that low-positive index values generally correspond to true HSV-2 infection, particularly in cases with recent HSV-2 acquisition in the settings of prior HSV-2 seronegativity.8 Thus, we felt that a cut-off index value of 1.1 was appropriate for this study of HSV-2 acquisition.

A staged protocol was used in testing samples for HSV-2. First, a sample from each participant’s enrollment visit was tested. Next, women with negative enrollment results had their last follow-up samples tested to define their HSV-2 serostatus at the end of the study interval. Finally, for those who were HSV-2 seropositive at their last follow-up visit, intervening samples were tested to determine the timing of seroconversion. The archived enrollment samples were serum and follow-up samples were plasma. The HSV-2 ELISA used in this study provides similar results with both sample types.9

Vaginal wet preparations were examined at 40× magnification for the presence of Trichomonas vaginalis and yeast. Bacterial vaginosis (BV) was diagnosed based on Gram stain criteria.10 Gram stained cervical secretions were examined microscopically and cervicitis was defined as an average of ⩾30 polymorphonuclear leucocytes in three non-adjacent oil immersion fields. Endocervical secretions were inoculated on modified Thayer-Martin medium for isolation of Neisseria gonorrhoeae.

Data from February 1993 through December 2006 were used. Women who were HSV-2 seronegative at baseline were selected. All visits at which HSV-2 serostatus could be determined were included. Analyses were performed using SPSS version 10 and S-Plus 2000 (Mathsoft). Women who seroconverted to HSV-2 were censored at that visit. For time-to event analyses, HSV-2 seroconversion was considered to have occurred on the date HSV-2 seropositivity was detected. Since our aim was to assess risk factors for HSV-2 acquisition among HIV-1 seronegative women, those who became infected with HIV-1 were censored at the time of HIV-1 seroconversion.

Several potential risk factors for HSV-2 acquisition were analysed. Information concerning educational level, parity, workplace (bar vs nightclub), alcohol and tobacco use, and vaginal washing practices were based on data collected at enrollment. Time-dependent modelling was used for other covariates, including age, duration of sex work, contraceptive method and the presence of other genital tract infections (gonorrhoea, cervicitis, BV, vaginal candidiasis, trichomoniasis and genital ulcer disease). As we have done previously for analyses of the effect of genital infections on risk of HIV-1 seroconversion,3 we assumed an effect window of 60 days to capture the influence of genital tract infections on HSV-2 seroconversion risk. Two measures of sexual behaviour, number of sex partners per week and percentage condom use were analysed as time-dependent covariates. To capture usual behaviour over time, averages were calculated for these variables for each year of follow-up.11 We modelled condom use in three categories (<25%, 25–75% and >75%) to remain consistent with prior studies.12

Cox proportional hazard models were used to calculate univariate and multivariate hazard ratios (HRs) and 95% confidence intervals (CIs) for potential risk factors related to HSV-2 acquisition. The final multivariate model was constructed using reverse stepwise selection of variables. Starting with a model including all covariates from the univariate analyses, one variable at a time was excluded. The variable removed was selected based on having the p value closest to 1.0. This process was repeated until all variables remaining in the model had p⩽0.10. For variables with multiple categories (for example, contraceptive method, each method as a separate category vs none), the lowest p value for any category was considered when selecting variables for exclusion. Genital ulcer disease was not included in the multivariate model as this was felt to be a consequence of HSV-2 infection rather than a risk factor.

Results

A total of 1786 HIV-1 seronegative women were enrolled between February 1993 and December 2006, of whom 1527 (85.5%) returned for at least one follow-up visit. For determination of HSV-2 serostatus, 1506 (98.6%) samples were available at enrollment—302 (20.1%) from women who were initially HSV-2 seronegative. Of these, 297 (98.3%) had follow-up samples available for HSV-2 testing and served as the study population for this analysis. Baseline characteristics are provided in table 1. The median duration of follow-up was 335 days (interquartile range (IQR) 92–862), reflecting a total of 499 person-years of follow-up over 2819 visits. The median time between visits was 33 days (IQR 28–48). Altogether, 115 women seroconverted to HSV-2 (incidence 23 cases/100 person-years). There were 10 women who experienced HSV-2 and HIV-1 seroconversion at the same visit.

In univariate analyses, recent entry into sex work, heavier alcohol use (>7 drinks per week) and BV were significantly associated with increased risk for HSV-2 (table 2). More frequent condom use was protective, although this association was only statistically significant for the 25–75% condom use category. Acquisition of HSV-2 was associated with >10-fold increase in the risk of GUD (HR 10.40; 95% CI 4.57 to 23.80; p<0.001).

In the multivariate analyses, more recent entry into sex work remained significantly associated with increased HSV-2 risk. The association between condom use and reduced risk for HSV-2 also remained essentially unchanged, with significantly lower HSV-2 risk in the 25–75% condom use group. After adjustment for potential confounding factors in the multivariate model, higher number of sexual partners and bar work both achieved statistically significant associations with HSV-2 acquisition. There were statistical trends for increased risk of HSV-2 among women with BV (p = 0.07) and for decreased risk of HSV-2 among women using oral contraceptive pills (p = 0.08).

Discussion

Nearly 80% of the participants in this cohort of women at high risk for HIV-1 were seropositive for HSV-2 upon entry into the study. Among those who were initially seronegative, the rate of new HSV-2 infections was 23% per year. To our knowledge, this is among the highest incidence rates for HSV-2 observed anywhere in the world.13 14 Recent entry into sex work, bar work (as opposed to nightclub work), higher number of sexual partners and less frequent condom use were associated with elevated HSV-2 risk.

Our findings are comparable with a recent study of female bar and hotel workers in Tanzania.13 That study found an overall HSV-2 incidence of 14.2 cases/100 person-years. There was increased risk of HSV-2 acquisition in women who reported sexual debut at a younger age.13 The high incidence of HSV-2 in these cohorts of East African women suggests an urgent need for interventions aimed at delaying sexual debut, encouraging fewer partners after initiation of sex, and promoting correct and consistent condom use.

In this population of female sex workers from Mombasa, we found a higher risk of acquiring HSV-2 among women working in bars as opposed to nightclubs. This is consistent with our previously reported findings of risk factors for HIV-1 in this cohort.6 Most of the women work in bars and supplement their income as barmaids with transactional sex. As such, they tend to have relatively lower numbers of sex partners. While the specific workplace association is unlikely to be widely generalisable, it highlights the fact that network-level factors may influence STI risk in a way that is not captured by adjustment for individual-level sexual risk behaviours such as number of partners, frequency of intercourse and condom use.

We found that condom use reduced the risk of HSV-2 acquisition, but this association was statistically significant only for the 25%–75% condom use category. Surprisingly, the protective effect was smaller and was not statistically significant for the >75% condom use category. Given the use of participants’ self reported behaviour, the highest category of condom use may represent over-reporting due to social desirability bias. In addition, women with the highest reported condom use may have more sex partners or may be exposed to partners who are at higher risk for transmitting HSV-2. Nonetheless, it is encouraging to see that condom use appeared to provide some reduction in the risk of HSV-2 in this cohort.

In the present study, BV was associated with a significantly increased risk for HSV-2 acquisition in unadjusted analyses, although this was reduced to a statistical trend in the multivariate model. Prior studies have found an association between BV and HSV-2.15 16 The relatively small number of HSV-2 seroconversions in our cohort may have limited our ability to achieve statistical significance for this modest association. Oral contraceptive use was associated with a 50% decreased risk for HSV-2 acquisition in multivariate analysis, although this did not achieve statistical significance. Few data have been reported concerning the effect of hormonal contraception on HSV-2 risk, although interactions between contraceptive use, HSV-2 and HIV-1 risk have been proposed.3 Future studies should continue to explore the relationship between contraception and HSV-2 acquisition.

Ultimately, a safe and effective vaccine would provide the most effective means of reducing women’s risk for acquiring HSV-2. Until such a vaccine is available, studies that help us to understand risk factors for HSV-2, such as higher partner number, lack of condom use and potentially BV, may help to inform the development of intervention strategies to reduce HSV-2 transmission.