Treatment with antiretroviral therapy is not associated with increased sexual risk behavior in Kenyan female sex workers

McClelland, R Scotta,b,c,f; Graham, Susan Ma,f,g; Richardson, Barbra Ad; Peshu, Norbertg; Masese, Linnet Nb; Wanje, George Hf; Mandaliya, Kishorchandra Nh; Kurth, Ann Ee; Jaoko, Walterf; Ndinya-Achola, Jeckoniah Of

doi: 10.1097/QAD.0b013e32833616c7
Epidemiology and Social

Objective: The objective of this study was to test the hypothesis that sexual risk behavior would increase following initiation of antiretroviral therapy (ART) in Kenyan female sex workers (FSWs).

Design: Prospective cohort study.

Setting: FSW cohort in Mombasa, Kenya, 1993–2008.

Subjects: Eight hundred and ninety-eight women contributed HIV-1-seropositive follow-up visits, of whom 129 initiated ART.

Intervention: Beginning in March 2004, ART was provided to women qualifying for treatment according to Kenyan National Guidelines. Participants received sexual risk reduction education and free condoms at every visit.

Main outcome measures: Main outcome measures included unprotected intercourse, abstinence, 100% condom use, number of sexual partners, and frequency of sex. Outcomes were evaluated at monthly follow-up visits using a 1-week recall interval.

Results: Compared with non-ART-exposed follow-up, visits following ART initiation were not associated with an increase in unprotected sex [adjusted odds ratio (AOR) 0.86, 95% confidence interval (CI) 0.62–1.19, P = 0.4]. There was a nonsignificant decrease in abstinence (AOR 0.81, 95% CI 0.65–1.01, P = 0.07), which was offset by a substantial increase in 100% condom use (AOR 1.54, 95% CI 1.07–2.20, P = 0.02). Numbers of sex partners and frequency of sex were similar before versus after starting ART. A trend for decreased sexually transmitted infections following ART initiation provides additional support for the validity of the self-reported behavioral outcomes (AOR 0.67, 95% CI 0.44–1.02, P = 0.06).

Conclusion: In the setting of ongoing risk reduction education and provision of free condoms, initiation of ART was not associated with increased sexual risk behavior in this cohort of Kenyan FSWs.

Author Information

aDepartment of Medicine, USA

bDepartment of Epidemiology, USA

cDepartment of Global Health, USA

dDepartment of Biostatistics, USA

eSchool of Nursing, University of Washington, Seattle, USA

fDepartment of Medical Microbiology, University of Nairobi, Nairobi, Kenya

gKenya Medical Research Institute, Kilifi, Kenya

hCoast Provincial General Hospital, Mombasa, Kenya.

Received 5 September, 2009

Revised 25 November, 2009

Accepted 3 December, 2009

Correspondence to R. Scott McClelland, MD, MPH, International AIDS Research and Training Program, University of Washington, Box 359909, 325 Ninth Avenue, Seattle, WA 98104, USA. Tel: +1 206 543 4278; fax: +1 206 543 4818; e-mail:

Article Outline
Back to Top | Article Outline


Increased access to antiretroviral therapy (ART) in resource-limited settings has improved the health and survival of millions of people [1]. Because individuals with low plasma viral load may be less infectious [2], widespread use of ART might also influence the course of the global HIV-1 epidemic [3]. However, the net effect of the ART rollout on global HIV-1 epidemiology will also depend on long-term effects of treatment on sexual risk behavior [4].

Ninety percent of people infected with HIV-1 live in developing countries [1], but few studies have examined sexual risk behavior after ART initiation in these settings. A key finding of a 2007 systematic review was the lack of data on this topic [5]; only three studies were identified. Two cross-sectional analyses, from Uganda and Cote d'Ivoire, found no significant difference in risk behavior in ART-treated versus ART-naive adults [6,7]. The only prospective study demonstrated that integrated provision of home-based ART with prevention services was associated with 70% lower reported risky sexual behavior over the first 6 months of treatment in a Ugandan cohort [8].

Few prospective studies have been published since the 2007 review. In Kenyan and South African adults, unprotected intercourse was less frequently reported 1 year after ART initiation compared with baseline [9,10]. In contrast, a study in Cote d'Ivoire found that unprotected intercourse increased significantly during the first 6 months of ART [11]. The differing results highlight both the complexity of the question of how ART influences risk behavior and the need to develop a broad evidence base [5]. Studies with extended follow-up are needed to determine how ART is likely to influence the long-term course of HIV-1 epidemics.

This report presents analyses of the association between ART and risk behavior in a cohort of Kenyan female sex workers (FSWs). The prespecified hypothesis was that ART would be associated with increased sexual risk behavior.

Back to Top | Article Outline


Population and procedures

This open cohort study of risk factors for HIV-1 acquisition in FSWs was established in 1993 [12]. Women who acquired HIV-1 were invited to continue with follow-up. All participants were asked to return for monthly follow-up including a standardized interview and physical examination. They were asked to report the number of sex partners, sexual contacts, and sexual contacts with condoms in the past week. Genital specimens were collected for diagnosis of sexually transmitted infections. At all visits, participants received individualized risk reduction education and free condoms. Beginning in 1998, HIV-1-seropositive women in the cohort had CD4 cell counts performed every 3 months. Women who were HIV-1-seropositive at screening were invited to enroll starting in 2001. In 2004, ART was introduced for those eligible according to Kenyan Guidelines (CD4 cell count <200 cells/μl or AIDS-defining illness). This analysis includes HIV-1-seropositive follow-up accrued from February 1993 through April 2008. Women receiving ART from other sources were excluded, as it was not possible to precisely ascertain the timing of treatment initiation. The protocol was approved by Human Subjects Committees at the Kenya Medical Research Institute, Kenyatta National Hospital, and University of Washington. All participants provided informed consent.

Back to Top | Article Outline
Serology and microbiology

HIV-1 screening was performed by ELISA (Detect-HIV; BioChem ImmunoSystems, Allentown, Pennsylvania, USA). HIV-positive status was confirmed using a second ELISA (Recombigen; Cambridge Biotech, Worcester, Massachusetts, USA) or Vironostika (bioMérieux, Marcy l'Etoile, France) [13]. Culture for Neisseria gonorrhoeae was performed on modified Thayer–Martin media. A vaginal saline wet mount was examined microscopically for motile Trichomonas vaginalis parasites. The presence of sperm was determined by microscopic examination of the vaginal wet mount and cervical Gram-stain. CD4 lymphocyte quantitation was performed using a manual system (Cytosphere; Coulter, Hialeah, Florida, USA) from 1998 until 2004, and thereafter by an automated method (FACSCount; Becton Dickinson, Franklin Lakes, New Jersey, USA).

Back to Top | Article Outline
Data analysis

Reported sexual risk behaviors were used to define five outcomes based on a 1-week recall period [14]. Women were considered to have unprotected intercourse if they were not abstinent and did not have 100% condom use. They were classified as abstinent if they reported no sexual intercourse during the past week. Among those reporting intercourse, 100% condom use was defined as having the number of sexual contacts with a condom equal to the total number of sexual contacts. We evaluated the number of sex partners in the past week and number of sexual encounters in the past week among women who reported any sexual intercourse. Both outcomes were dichotomized at their medians.

Our primary comparison was sexual risk behavior during ART-naive versus ART-exposed follow-up, including visits from all 898 participants. Data were analyzed according to the intent-to-treat principle; women presenting for an ART-initiation visit were considered ART-exposed at all subsequent visits. To gain a greater understanding of how immunosuppression influenced risk behavior, we performed additional analyses stratifying visits at which CD4 cell count was measured into three categories (≥500, 200–499, and <200).

We also analyzed changes in two biological outcomes: presence of a sexually transmitted infection (STI) (surrogate marker for unprotected sex since last visit) and presence of sperm in genital secretions (surrogate marker for recent unprotected sex). Women were considered to have an STI if they had gonorrhea, trichomoniasis, or both. Sperm were considered present if they were identified on vaginal saline wet mount, cervical Gram stain, or both.

Analyses were performed using SPSS version 15.0 (SPSS Inc., Chicago, Illinois, USA) and Stata version 9.2 (Stata Corporation, College Station, Texas, USA). Generalized estimating equations with a logit link and exchangeable correlation structure were used to quantify the effect of ART on risk behaviors, allowing for multiple observations per individual. Multivariate models controlled for time-dependent changes in risk behavior in the cohort by adjusting for years since enrolment in all analyses [14,15]. Other potential confounding factors were considered for inclusion in adjusted analyses based on known or suspected associations with risk behavior. Analyses were adjusted for time-varying cofactors including calendar year category (1993–1996, 1997–2000, 2001–2004, and 2005–2008), age, and contraceptive method. Further adjustment for baseline educational level, marital status, workplace (bar versus nightclub), alcohol use, and Karnofsky score did not substantially change the associations between risk behaviors and ART, so these variables were not retained in the final adjusted model.

Back to Top | Article Outline


Between February 1993 and April 2008, 966 women contributed HIV-seropositive follow-up visits. Of these, 68 (7%) received ART from other clinics and were excluded from further analyses. The remaining 898 women are the focus of this study. Of these, 298 (33%) acquired HIV-1 during follow-up, whereas the remainder were HIV-1-seropositive at screening. There were 15 926 HIV-1-seropositive follow-up visits during 2404 woman-years of follow-up. The median number of follow-up visits per participant was 10 [interquartile range (IQR) 3–27] and the median interval between visits was 33 days (IQR 29–48). One hundred and twenty-nine women (14%) initiated ART, contributing a median of 24.6 (IQR 11.4–32.9) months following ART initiation.

Participants' baseline characteristics are shown in Table 1. These women had a median age of 31 years (IQR 26–36), and the majority had completed at least some primary education. Over 40% were using a contraceptive method other than condoms alone. The women reported a relatively low number of sex partners per week (median 1, IQR 1–2) and number of sexual encounters per week (median 2, IQR 1–2). The majority of the women reported using condoms during 100% of sexual encounters during the past week. At baseline, the 129 women who started ART during the study were similar to the other 769 women in the cohort with respect to the variables shown in Table 1, except that they were more likely to have ever been married (92, 71.3% versus 471, 61.2%; P = 0.003), had higher parity (3, IQR 2–4 versus 2, IQR 1–3; P < 0.001), and were less likely to be abstinent (5, 3.9% versus 122, 15.9%; P < 0.001). Median age at ART initiation was 36 years (IQR 32–40).

Back to Top | Article Outline
Changes in risk behavior

Unprotected intercourse was reported at 2308 of 13 025 (17.7%) non-ART visits versus 258 of 2901 (8.9%) visits following ART initiation (Table 2). After adjusting for years since enrolment [14,15], there was no evidence of increased unprotected intercourse after ART initiation [adjusted odds ratio (AOR) 0.80, 95% confidence interval (CI) 0.58–1.12]. Further adjustment for potential confounding factors did not substantially alter these results (AOR 0.86, 95% CI 0.62–1.19). The absence of significant change in unprotected sex reflected divergent results in the rates of abstinence and condom use following ART initiation. Specifically, there was a trend suggesting that women on ART were less likely to be abstinent, but this was offset by a highly significant increase in 100% condom use. At visits where women reported that they were sexually active, frequency of reporting multiple sex partners and frequency of sex did not increase on ART. Restricting to the subset of 129 women who initiated ART did not substantially alter these results (Table 3).

Women on ART may be more likely to adhere to monthly visits because they require medication refills. Thus, a prespecified secondary analysis was conducted including only the first visit for each participant in each quarter. After controlling for potential confounding factors, changes in risk following ART initiation were similar to our primary analysis for unprotected sex (AOR 0.77, 95% CI 0.54–1.10, P = 0.2), abstinence (AOR 0.78, 95% CI 0.61–0.99, P = 0.05), 100% condom use (AOR 1.56, 95% CI 1.09–2.22, P = 0.01), more than 1 sex partner (AOR 0.74, 95% CI 0.51–1.08, P = 0.1), and more than two sexual encounters (AOR 1.04, 95% CI 0.77–1.42, P = 0.8).

To determine the extent to which behavioral changes following ART initiation were independent of changes in CD4 cell count, we conducted an analysis adjusting for CD4 cell count in the 4957 visits for which these data were available (Table 4). This analysis demonstrated significantly lower likelihood of unprotected intercourse following ART initiation. Moreover, the previously observed trend suggesting decreased abstinence after ART initiation was eliminated.

To further explore the effect of disease stage on risk behavior in the setting of ART, analyses were repeated after stratifying by CD4 cell count (<200, 200–499, and ≥500 cells/μl). Similar to the primary analysis, there was no evidence of increased risk following ART initiation (Table 5). Among women with the lowest CD4 cell counts, there were significant reductions in both sex with more than one partner and more than two sexual encounters in the past week.

Back to Top | Article Outline
Biological outcomes

To complement self-reported behavioral risk data, we evaluated STI incidence and the incidence of sperm detection in genital secretions. STIs were identified at 780 of 13 024 (6.0%) pre-ART versus 116 of 2901 (4.0%) post-ART visits. After adjustment for years since enrolment, there was a lower likelihood of STIs after ART initiation (AOR 0.62, 95% CI 0.42–0.91, P = 0.01). Results were similar after additional adjustment for calendar year category, age, and contraceptive use (AOR 0.67, 95% CI 0.44–1.02, P = 0.06). The presence of sperm in genital secretions, a marker for recent unprotected intercourse, was identified at 684 (5.0%) pre-ART visits versus 141 (4.9%) post-ART visits (AOR 1.01, 95% CI 0.75–1.34, P = 1.0). Findings were similar following adjustment for calendar year category, age, and contraceptive use (AOR 0.88, 95% CI 0.65–1.19, P = 0.4).

Back to Top | Article Outline


In this population of high-risk Kenyan women, we found no increase in sexual risk behaviors following ART initiation. On the contrary, a highly significant more than 50% increase in consistent condom use was reported. Among those with advanced immunosuppression (CD4 cell count <200 cells/μl), ART use was associated with significant reductions in both partner numbers and frequency of sex.

Although this study did observe a statistical trend (P = 0.07) for decreased abstinence following ART initiation, this finding must be interpreted in the context of an earlier study in this cohort, which demonstrated that women with advanced immunosuppression were significantly more likely to be abstinent compared with HIV-1-seropositive women with higher CD4 cell counts [14]. As health and quality of life improve following ART initiation [16], it seems plausible that sexual activity might increase to parallel the rates of sexual encounters in healthier women. In support of this hypothesis, analyses adjusting for CD4 cell count showed similar rates of abstinence before and after ART initiation.

These data add to the evidence base on changes in sexual risk behavior following ART initiation in resource-limited settings. Most prospective studies have found that sexual risk behavior is either unchanged, or that risky behavior decreases after ART initiation [8–10]. One exception was a study from Cote d'Ivoire [11], which reported an increase in unprotected intercourse in the ART group. Methodological differences including the selection and measurement of outcomes may contribute to differences in the results of these epidemiological studies [17]. Variations in population and program characteristics including sex, socioeconomic status, cultural beliefs, health status, duration of ART, treatment setting, and the type, quality, and intensity of risk-reduction services are also likely to influence the direction and magnitude of changes in risk behavior following ART initiation.

This study had several strengths. First, the prospective cohort design provided ample baseline data on risk behavior prior to ART initiation. Second, the long period of follow-up, including a median of more than 2 years following ART initiation, is useful for understanding how risk behavior may change with prolonged ART. Third, in this high-risk cohort, risk-reduction messages and provider time devoted to counseling were similar before and after ART initiation, reducing the potential for bias. Fourth, the study's size permitted careful control for multiple potential confounding factors. Finally, several relevant outcomes were used to assess a range of risk behaviors. The results highlight the importance of evaluating multiple outcomes to gain a comprehensive understanding of changes in sexual risk.

An important limitation in this type of study is the potential for underreporting of risky sexual behavior. In this context, it is reassuring that results evaluating biological outcomes were consistent with results based on self-reported behavior. Several factors should be considered in relation to the generalizability of the findings. These women were enrolled based on self-reported transactional sex. Most supplemented income from work as barmaids with occasional payment for sex in cash or in kind. Although sex workers are a special population, they also represent an important core-transmitter group [18]. As a final point, risk-reduction services in research cohorts may be more intensive than in other settings, and participants' risk behavior is likely to decrease over time [15]. To address this bias toward a finding of decreased risk with ART, all analyses were presented with adjustment for time in the cohort.

The magnitude of an HIV-1 epidemic is a function of infectivity, rate of partner change, duration of infection [19], and network-level influences [20]. As ART prolongs life without eradicating HIV-1, the premise that treatment might decrease transmission rests on the hope that treatment will decrease infectivity without resulting in behavioral disinhibition. A modeling study based on the population in Rakai, Uganda, found that the modest decreases in HIV-1 incidence resulting from decreased infectivity could be counterbalanced if ART resulted in behavioral disinhibition [4]. In the Mombasa Cohort, prior studies have demonstrated rapid and sustained suppression of plasma and genital HIV-1 following ART initiation [21,22]. The present findings provide strong evidence that risk behavior did not increase following ART initiation in the same cohort. Taken together, these results support the potential importance of ART as one of the tools for reducing sexual HIV-1 transmission. This study also underscores the need for prevention targeting those who do not currently qualify for ART, as their viral loads and transmission risk may be considerably higher.

Back to Top | Article Outline


This study was supported by National Institutes of Health (grant R01-AI-58698) and by the Fogarty International Center (grant 5D43-TW000007 to S.M.G., L.N.M., and G.H.W.). Additional support for the Mombasa Field Site was received from the University of Washington Center for AIDS Research (grant P30-AI-27757).

The authors wish to acknowledge the study participants, who contributed their time and effort to make this study a success. We also wish to recognize the contributions made by our clinical, laboratory, and administrative staff. We thank the Mombasa Municipal Council for providing clinical space and Coast Provincial General hospital for providing laboratory space. This study was approved for publication by the Director, Kenya Medical Research Institute.

Author contributions: R.S.M., B.A.R., N.P., K.N.M., A.E.K., W.J., and J.O.N.-A. conceived the question and designed the study. R.S.M. obtained funding for the study. R.S.M., S.M.G., B.A.R., L.N.M., and G.H.W. participated in collection and interpretation of the data. B.A.R. conducted the data analyses. All authors participated in preparation of the manuscript and have approved the final draft for submission.

The contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH

Preliminary data from this study have been presented at the 5th International AIDS Society Conference on HIV Pathogenesis, Treatment, and Prevention, Cape Town, South Africa, 19–22 July 2009.

Back to Top | Article Outline


1. UNAIDS. Report on the Global AIDS Epidemic. Geneva, Switzerland: WHO; 2008. pp. 1–357.
2. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li C, Wabwire-Mangen F, et al. Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group. N Engl J Med 2000; 342:921–929.
3. Velasco-Hernandez JX, Gershengorn HB, Blower SM. Could widespread use of combination antiretroviral therapy eradicate HIV epidemics? Lancet Infect Dis 2002; 2:487–493.
4. Gray RH, Li X, Wawer MJ, Gange SJ, Serwadda D, Sewankambo NK, et al. Stochastic simulation of the impact of antiretroviral therapy and HIV vaccines on HIV transmission; Rakai, Uganda. AIDS 2003; 17:1941–1951.
5. Kennedy C, O'Reilly K, Medley A, Sweat M. The impact of HIV treatment on risk behaviour in developing countries: a systematic review. AIDS Care 2007; 19:707–720.
6. Moatti JP, Prudhomme J, Traore DC, Juillet-Amari A, Akribi HA, Msellati P. Access to antiretroviral treatment and sexual behaviours of HIV-infected patients aware of their serostatus in Cote d'Ivoire. AIDS 2003; 17(Suppl 3):S69–S77.
7. Bateganya M, Colfax G, Shafer LA, Kityo C, Mugyenyi P, Serwadda D, et al. Antiretroviral therapy and sexual behavior: a comparative study between antiretroviral-naive and -experienced patients at an urban HIV/AIDS care and research center in Kampala, Uganda. AIDS Patient Care STDS 2005; 19:760–768.
8. Bunnell R, Ekwaru JP, Solberg P, Wamai N, Bikaako-Kajura W, Were W, et al. Changes in sexual behavior and risk of HIV transmission after antiretroviral therapy and prevention interventions in rural Uganda. AIDS 2006; 20:85–92.
9. Luchters S, Sarna A, Geibel S, Chersich MF, Munyao P, Kaai S, et al. Safer sexual behaviors after 12 months of antiretroviral treatment in Mombasa, Kenya: a prospective cohort. AIDS Patient Care STDS 2008; 22:587–594.
10. Eisele TP, Mathews C, Chopra M, Lurie MN, Brown L, Dewing S, Kendall C. Changes in risk behavior among HIV-positive patients during their first year of antiretroviral therapy in Cape Town South Africa. AIDS Behav 2009; 13:1097–1105.
11. Diabate S, Alary M, Koffi CK. Short-term increase in unsafe sexual behaviour after initiation of HAART in Cote d'Ivoire. AIDS 2008; 22:154–156.
12. Martin HL Jr, Nyange PM, Richarson BA, Lavreys L, Mandaliya K, Jackson DJ, et al. Hormonal contraception, sexually transmitted diseases, and risk of heterosexual transmission of human immunodeficiency virus type 1. J Infect Dis 1998; 178:1053–1059.
13. WHO, CDC. Guidelines for appropriate evaluations of HIV testing technologies in Africa. Harare, Zimbabwe: CDC; 2001. pp. 1–48.
14. McClelland RS, Hassan WM, Lavreys L, Richardson BA, Mandaliya K, Ndinya-Achola J, et al. HIV-1 acquisition and disease progression are associated with decreased high-risk sexual behaviour among Kenyan female sex workers. AIDS 2006; 20:1969–1973.
15. Baeten JM, Richardson BA, Martin HL, Nyange PM, Lavreys L, Ngugi EN, et al. Trends in HIV-1 incidence in a cohort of prostitutes in Kenya: implications for HIV-1 vaccine efficacy trials. J Acquir Immune Defic Syndr 2000; 24:458–464.
16. Jelsma J, Maclean E, Hughes J, Tinise X, Darder M. An investigation into the health-related quality of life of individuals living with HIV who are receiving HAART. AIDS Care 2005; 17:579–588.
17. van der Elst EM, Okuku HS, Nakamya P, Muhaari A, Davies A, McClelland RS, et al. Is audio computer-assisted self-interview (ACASI) useful in risk behaviour assessment of female and male sex workers, Mombasa, Kenya? PLoS ONE 2009; 4:e5340.
18. Cowan FM, Langhaug LF, Hargrove JW, Jaffar S, Mhuriyengwe L, Swarthout TD, et al. Is sexual contact with sex workers important in driving the HIV epidemic among men in rural Zimbabwe? J Acquir Immune Defic Syndr 2005; 40:371–376.
19. Anderson RM, May RM. Infectious diseases of humans: dynamics and control. Oxford, England: Oxford University Press; 1991.
20. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS 1997; 11:641–648.
21. Graham SM, Holte SE, Peshu NM, Richardson BA, Panteleeff DD, Jaoko WG, et al. Initiation of antiretroviral therapy leads to a rapid decline in cervical and vaginal HIV-1 shedding. AIDS 2007; 21:501–507.
22. Graham SM, Masese LN, Gitau R, Richardson B, Peshu N, Mandaliya K, et al. Correlates of genital HIV-1 shedding among antiretroviral naive women initiating therapy. 16th Conference on Retroviruses and Opportunistic Infections; Montreal, Canada; 8–12 February 2009.

Africa; antiretroviral therapy; human immunodeficiency virus type 1; sexual risk behavior; sexually transmitted infection

© 2010 Lippincott Williams & Wilkins, Inc.