Background. Herpes simplex virus (HSV) type 2 increases the risk of human immunodeficiency virus (HIV) infection, and, in regions with high prevalence of both viruses, control of HSV-2 may be an effective method of HIV prevention. Identification of modifiable factors for prevention of HSV-2 infection is essential. We conducted this study among female bar and hotel workers in Moshi, Tanzania.
Methods. Factors associated with prevalent infection were examined among 1039 women. Predictors of incident infection were examined among 360 women initially HSV-2 negative, with at least 1 follow-up visit.
Results. HSV-2 prevalence was 56.3% (95% confidence interval [CI], 53.3%–59.3%). Only 2.5% of women able to name a sexually transmitted infection named herpes. Incidence was 14.2 cases/100 person-years (95% CI, 10.5–18.8 cases/100 person-years). Incident HSV-2 infection was independently associated with HIV infection, younger age of sexual initiation, ethnicity, alcohol consumption, and having a male partner with other sexual partners.
Conclusions. The occurrence of HSV-2 is high in this population, but knowledge is low. Development of education programs to increase awareness of HSV-2 is critical. The control of both HSV-2 and HIV infections is a major public health priority in Moshi. Prevention interventions in this and other high prevalence populations might most effectively target younger women, before initiation of sexual activity.
Herpes simplex virus (HSV) type 2 is the most common cause of genital ulcers worldwide [1–3]. The prevalence of HSV-2 varies from <15% in most European countries  to 40% and higher in some resource-poor countries [5, 6]. In Tanzania, the prevalence of HSV-2 among women in the general population is 42% in the Mwanza region  and 44% in Moshi Urban District . The high prevalence of HSV-2 has implications for HIV transmission dynamics, particularly in mature HIV epidemics such as the one in Tanzania . The synergy between HSV-2 and HIV has been recently highlighted, with HSV-2 increasing both susceptibility to and transmission of HIV infection . In a recent meta-analysis of 18 longitudinal studies, the summary estimate for HSV-2 on HIV acquisition among women was 3.1 (95% confidence interval [CI], 1.7–5.6) .
There is limited information about risk factors for acquisition and transmission of HSV-2 infection. Several longitudinal studies have identified HIV infection, being female, inconsistent condom use, number of sex partners, frequency of sexual activity, and other sexually transmitted infection (STIs) as predictors of HSV-2 acquisition [5, 6, 12–19]. Evidence from in vivo studies shows that genital shedding of HSV-2 is higher in HIV-positive individuals [20, 21], suggesting a potential role that HIV infection plays in HSV-2 transmission.
Female bar and hotel workers in resource-poor settings are often at higher risk for STIs because they engage in high-risk behaviors, including accepting money or gifts in exchange for sex [22, 23]. In Tanzania, the prevalence of HSV-2 among female bar and hotel workers ranges from 53% to 87%, and HIV prevalence is also high [24–26]. This population might continue to serve as a core group for the transmission of HIV, HSV-2, and other STIs in Tanzania, where the HIV epidemic is generalized . Thus, improved knowledge of factors associated with HSV-2 infection in this population is important for the design of interventions. The aims of this study were to determine risk factors for incident HSV-2 infections among female bar and hotel workers in Moshi, Tanzania.
Participants and Methods
Study design and data collection. We conducted analyses of data from a prospective study designed primarily to determine the predictors of HIV seroconversion in a cohort of female bar and hotel workers in Moshi Urban District, Tanzania. Study procedures and methods have been described elsewhere  and will be reviewed here briefly. Women working in the hotels and bars located in 7 of the 15 local administrative wards in the central part of Moshi were eligible to participate in this study. These wards were selected as areas where most of the hotels and bars are located. The owners of all registered bars and hotels were visited and given brief information about the study. Most bar owners are aware of the severity of the HIV situation in Moshi and agreed to participate in the study. After these initial contacts, outreach workers met with eligible female bar and hotel workers to provide more study information. Women were invited to visit the study clinic, where they received detailed information about study aims and procedures. Those who agreed to participate signed a written consent form and were enrolled.
At baseline, we collected information about sociodemographic characteristics, work and reproductive history, STI symptoms, sexual behaviors, and alcohol consumption. This information was obtained during interviews with trained female nurses in private rooms at the study clinic. Pretest HIV/STI counseling was provided after the interviews, and blood was drawn for HIV, syphilis, and HSV-2 testing. A female physician performed clinical examinations, and genital samples were collected for diagnosis of STIs and other genital infections. All samples were tested in the Department of Clinical Laboratories at Kilimanjaro Christian Medical Center in Moshi. Women returned to the study clinic after 7–10 days for results and posttest counseling; subjects with treatable STIs received free treatment based on the guidelines of the Tanzania Ministry of Health. Women who tested positive for HIV infection were given information about medical and social support services available in the study area for HIV-infected individuals.
One thousand fifty women were enrolled in the study between December 2002 and November 2003. Follow-up visits were conducted at the clinic every 3 months for 1 year. At each visit, a structured follow-up questionnaire was administered to update baseline information. Most information collected during these visits was limited to the previous 3 months (or since the last clinic visit if subjects did not return within 3 months). After the interviews, pre- and posttest counseling and a medical examination, including a blood draw for HIV/STI testing, were conducted. The study protocol was approved by the Ethics Committee of Kilimanjaro Christian Medical Centre and the Institutional Review Board of Harvard School of Public Health.
Laboratory methods. HSV-2 antibodies were detected using an IgG type-specific ELISA in accordance with the manufacturer's instructions (Focus Technologies). The median time to detection of HSV-2 antibodies after infection using this assay is 2–3 weeks . HIV infection was determined using 2 HIV ELISAs, and discordant tests results were resolved by Western blot (Genetic Systems; Bio-Rad Laboratories). As part of the quality assurance system, all samples testing positive for HSV-2 and HIV and 10% of randomly selected negative samples were retested to confirm results. Of the 693 samples initially testing positive for HSV-2 at baseline, 689 retested as positive; 453 samples tested negative initially, and all of the 50 randomly selected samples retested as negative (Cohen's κ, 0.96). Of the 231 samples testing positive for HIV, 229 retested as positive; 845 tested negative initially, and all of the 90 randomly selected samples retested as negative (Cohen's κ, 0.98). Individuals were considered to have recent or untreated syphilis if their serum was positive by both the rapid plasma reagin card test (Becton-Dickinson) and the Treponema pallidum hemagglutination assay test (Murex Biotech). A wet mount of vaginal fluid was prepared and examined microscopically for motile Trichomonas vaginalis. Chlamydia trachomatis antigen was detected by use of an antigen-detection enzyme immunoassay, with positive samples confirmed by use of a blocking assay from the same manufacturer (Murex Biotech). Candida albicans was identified by isolation of the gram-negative yeast-like cells on Saboraud's dextrose agar and was confirmed by use of the germ tube test (Remel). Neisseria gonorrhoeae isolation was done by inoculation of endocervical swabs on modified Thayer-Martin medium. However, because of the extremely low prevalence of this infection in the study population, this procedure was discontinued after the first 9 months of the study. Vaginal flora disturbances and bacterial vaginosis were diagnosed using 4 clinical criteria, as previously proposed by Amsel et al. .
Statistical analysis. The time of an incident infection was defined as the midpoint between the last negative test and the first positive test. The incidence rate was calculated as the number of seroconversions among individuals HSV-2 negative at baseline divided by the total person-time at risk. Stratum-specific incidence rates were also calculated. Exact 95% CIs were calculated using the Poisson distribution. Cox proportional hazard models were used to calculate crude and adjusted hazard ratios (HRs) and 95% CIs. The proportional hazards assumption for various covariates was checked graphically and then tested using interaction terms. Modification of the effect of HIV on HSV-2 by sexual behavior variables was assessed using stratified analyses and likelihood ratio tests. We also performed several sensitivity analyses of our assumption about the timing of HSV-2 infection.
To assess whether loss to follow-up might be differential, we compared women who returned for at least 1 follow-up visit with those with no follow-up, by all baseline factors. We also made comparisons between different categories of women contributing follow-up time: we compared women lost to follow-up without becoming infected to (1) the remaining individuals who contributed follow-up time (those who became infected and those followed for 12 months without becoming infected) and to (2) those followed for 12 months without becoming infected.
In the incidence analysis, 1 woman was missing baseline age information, and her age was set to the average age for women with her covariate pattern. One woman missing baseline HIV information was assumed to be negative. Results were unchanged when these participants were excluded or when their values for age or HIV status were changed. All analyses were conducted using SAS for Windows (version 8.2).
There were 1050 women enrolled in the study. Five women were missing baseline HSV-2 status and were excluded from all analyses.
The baseline prevalence of HSV-2 was 56.3% (95% CI, 53.3%–59.3%). Prevalence peaked between the ages of 46 and 50 years, at 89.5% (figure 1). Of 912 women able to name at least 1 disease or infection transmitted through sexual intercourse, 96.2% named gonorrhea and 91.6% named syphilis, whereas only 7.6% named genital ulcers and 2.5% named herpes (figure 2). Only 9.9% of HSV-2-positive women reported having had a genital ulcer during the past 12 months, and only 3.9% had a clinically diagnosed genital ulcer at baseline. Twenty-three of the 27 women with clinically diagnosed ulcers (85.2%) were HSV-2 positive; 1 woman (3.7%) had syphilis. Factors associated with prevalent infection were consistent with those observed previously by us and other investigators [12, 25, 30].
There were 457 women who were HSV-2 negative at baseline. The mean age was 25 years (SD, 6.7 years; range, 14–64 years). Sixty-two percent of the women had never been married. Ninety percent had at least 7 years of formal education, and 6% were HIV positive. Three hundred sixty women (78.8%) returned for at least 1 follow-up visit. There were some differences in baseline characteristics between women lost to follow-up and those who returned for at least 1 visit. Women who did not return were significantly more likely to have initiated sexual activity at a younger age and less likely to be currently married (15.5% vs. 26.1%). Thirty-five percent were ≤20 years old, compared with 23% of women followed. Women not followed had also lived in Moshi and had worked in bars for shorter periods of time. When comparing those lost to follow-up without infection with the other individuals contributing follow-up time, the only significant differences were that women lost to follow-up without infection had fewer births and a shorter stay in Moshi and fewer expected to remain in Moshi.
The total person-time at risk was 337.8 person-years. Forty-eight women seroconverted during follow-up, for an overall HSV-2 incidence rate of 14.2 cases/100 person-years (95% CI, 10.5–18.8 cases/100 person-years). Incidence rates and univariate HRs are summarized in table 1. Women HIV positive at baseline had a >4-fold increase in the hazard of HSV-2 acquisition. Other baseline or incident STIs and genital infections were not associated with an increased hazard. The hazard of infection increased with increasing number of sex partners over the past 5 years, although this was significant only among women reporting ≥4 partners (HR, 2.6 [95% CI, 1.0–6.3]). Women whose age at sexual initiation was 21 years or older had a lower rate of infection than women with younger ages. Effect sizes for younger age categories were similar and were combined in the final model. Infection was also associated with ethnicity and alcohol consumption but not with education or religion.
Multivariate HRs are summarized in table 2. HIV infection at baseline or follow-up was a strong predictor of HSV-2 infection, with HIV-infected women having 3.4 times the HR of uninfected women. (Three of the 8 incident HIV infections occurred after or in the same interval as HSV-2 seroconversion; 5 occurred without a subsequent HSV-2 seroconversion.) Women whose partners had other sex partners during the follow-up period also had an increased risk of infection, as did women who initiated sex at ≤21 years old. Women belonging to ethnic groups other than the Pare ethnic group had more than double the incidence of HSV-2 than women in the predominant Chagga ethnic group. The number of alcoholic drinks consumed on drinking days was also predictive of HSV-2 infection.
In all sensitivity analyses, there were no material changes in any effect estimates. There was no significant effect modification detected. The proportional hazards assumption was met for all predictors.
This population of female bar and hotel workers had a high prevalence of HSV-2 and very low knowledge about genital herpes. The rate of new infections was also high, compared with most other rates observed among female populations [5, 14–16, 19]. The high prevalence and incidence, along with the relatively lower prevalence of curable STIs, confirm that HSV-2 infection is the most important STI and a major public health problem in this high risk population in northern Tanzania. Strategies to control or suppress prevalent HSV-2 infections, along with prevention of new infections in currently uninfected girls and women, are urgently needed. Given the low knowledge about HSV-2 in this population, strategies to raise awareness about genital herpes should be included when developing prevention programs for this infection.
Our findings add to the evidence of a synergistic relationship between HIV and HSV-2 infections. There are a number of mechanisms that have been proposed to account for the association between HIV infection and subsequent HSV-2 acquisition. One potential mechanism is that there is increased biological susceptibility to HSV-2 infection among HIV-positive individuals who are immunocompromised. Another mechanism is that women who are HIV infected become infected with HSV-2 at higher rates because they are having sexual contact with men who are infected with both viruses, since coinfection with HIV and HSV-2 results in increased infectiousness for both viruses . To distinguish between these 2 potential mechanisms, studies will need to have information on the immune status of HIV-positive individuals. If HSV-2 acquisition rates vary by CD4 cell count, for example, this would suggest that the link between HIV and HSV-2 is a direct biological association. If these rates do not vary, or if comparisons between individuals newly infected with HIV and HIV-negative individuals result in HRs similar to what the present study and other longitudinal studies have observed, then this would instead suggest that the association is due to sexual intercourse with coinfected individuals. One study among HIV-infected prostitutes found that HSV-2 acquisition rates did not vary by CD4 cell count ; future studies should attempt to confirm these findings.
To our knowledge, only a few studies [25, 31, 32] have considered the role that alcohol plays in HSV-2 infection. We believe the present study is the only longitudinal study to observe an effect. This association could be due to confounding by other unmeasured sexual behaviors (choice of partner or specific sexual practices), differences in risk-seeking personality, or more contextual factors such as individuals' social and sexual networks. Alternatively, the association could be causal and could increase adverse sexual outcomes by affecting behavior or sexual arousal . Alcohol use is high in this population. If the association we observed between the number of drinks per day and HSV-2 infection reflects heavier or more frequent drinking before sexual intercourse, then it could be that women who drink before sex are practicing riskier sexual behaviors that increase their risk of infection with HSV-2 and other sexually transmitted infections. Strategies aimed at reducing alcohol consumption could be important components of behavioral interventions designed to reduce HSV-2 incidence in this population.
There was a strong association between younger age of sexual initiation and subsequent risk of HSV-2 acquisition, a finding previously observed . The increased hazard could be due to the riskier sexual behaviors and partner characteristics associated with younger age of sexual initiation [35–37]. Together with the very high prevalence in all but the youngest age group in the present study, this result suggests that interventions to reduce HSV-2 incidence might be most effective if they focus on young women, particularly individuals who have not yet initiated sexual activity [38, 39].
We found no effect of increasing number of sex partners and history of exchanging sex for gifts or money. Women who reported no sexual partners in the past 5 years experienced no infections, but, for all other women, infection rate did not vary by number of partners. Because HSV-2 prevalence is so high in this population, it could be that sexually active women are at similarly increased risk for infection regardless of their individual sexual behaviors. Because women whose male partners had other sex partners had a higher infection rate, other partner characteristics, such as partner type and age, could potentially be more important predictors of infection than number of partners . Unlike previous studies [19, 40], use of condoms was not protective against HSV-2 acquisition. This may be due to misreporting of condom use resulting in misclassification. In addition, condoms may not provide full protection against HSV-2 because the virus can be shed from mucosal surfaces and on the skin surface not covered by condoms during sexual contact .
We also found a significant association between ethnicity and infection. Women of Chagga ethnicity had the lowest incidence, were less likely than women of “other” ethnicity to report a history of exchanging sex for gifts or money, and were more likely to report a younger age of sexual initiation. Therefore, the association could be due to other unmeasured sexual behaviors or partner characteristics or to differences in sexual mixing patterns or other sexual behaviors between the ethnic groups in Moshi.
There are 2 important limitations to the present study. To detect HSV-2 antibodies, we used an index cutoff value of 1.1 for the HerpesSelect IgG ELISA. Studies published after our lab analyses were under way have reported that this cutoff value resulted in reduced specificity when used in Ugandan and Kenyan populations [42, 43]. We may therefore have overestimated the HSV-2 incidence rate. All samples were tested in a lab that provides results indicating whether samples are negative or positive, and information on the actual index values was not available. We were therefore not able to redo our analyses using the recommended higher cutoff value.
Also, our follow-up rate was consistent with those of other published African studies [5, 15], with 79% returning for at least 1 follow-up visit and substantial numbers missing interim visits. Our estimate of the rate of infection is limited by this loss to follow-up. Differential loss to follow-up is also possible. Our sensitivity analyses resulted in consistent effect estimates, suggesting that most effects were not largely affected by differential loss to follow-up. However, younger age of sexual initiation was predictive of higher loss to follow-up. The likely effect is an underestimate of the effect of age of sexual initia-tion on HSV-2 infection.
The lack of HSV-2 knowledge and low prevalence of symptoms, together with the high prevalence of both HIV and HSV-2 infection in this population, suggests many opportunities to control HSV-2 infection and thereby reduce HIV incidence. Education to raise awareness about herpes and increased knowledge among HSV-2-positive women, particularly regarding the association between HSV-2 and HIV infection, could reduce both HSV-2 transmission and HIV acquisition. In addition to behavioral interventions, aggressive HSV-2 control programs, focusing on episodic or suppressive therapy, will be essential to reduce the rate of HSV-2 transmission by HSV-2-positive women, as well as HIV acquisition among these women . The high incidence of new infections also means that prevention methods are urgently needed. Prevention of HSV-2 incidence would be most effectively targeted toward girls and young women. Interventions that focus on delaying sexual initiation and reducing alcohol consumption could be very beneficial.
We would like to thank the women who participated in this study; research and administrative staff, for their efforts in the implementation of the study; and the Kilimanjaro Christian Medical Centre, the Harvard School of Public Health, and the Moshi Municipal Council, for providing institutional support. We also thank Mary Solomon, Coleta Mbuya, Esther Mchome, Grace Mhango, Basidi Bamba, Lulu Oguda, Uzodinma Ndibe, Elisante Masenga, George Suleman, Sarah Chiduo, and Christopher Mtamakaya, for their dedication and support.