At the beginning of the HIV/AIDS epidemic in Africa, sex workers (SW) were identified as a high-prevalence core group playing an important role in the dissemination of HIV, especially in urban centres [1–3]. Specialized clinics were set up to provide preventive and curative health services to SW, a very cost-effective intervention [4–6]. Nowadays, the prevalence of HIV within the adult population of West Africa remains generally lower than 10%, while prevalence among SW ranges between 21% and 74% [7–15]. In such circumstances, a sizeable fraction of cases of HIV infection among adult men in the community are presumed to be acquired from SW, but this has never been measured. Simulation modelling suggests that the role of SW in HIV transmission within a community decreases when prevalence increases in its general adult population , as in east and southern Africa. To delineate the extent of HIV infection among clients of SW and the role of SW in the dynamics of HIV in Accra, Ghana, a study of behavioural and epidemiological characteristics of clients and other men in sex work environments was conducted, after ethical review by the Ministry of Health.
There are two categories of SW in Accra: seaters, who work from their homes, and roamers, who find customers in bars, hotels, brothels or on the street . Since 1996, an intervention has aimed to reduce the incidence of HIV and sexually transmitted infections (STI) through STI management, education and condom promotion. The purpose and methods of the study were explained to groups of SW, landlords and bar/hotel owners. Potential study sites had to be located within the Accra or Tema districts and were selected if there were at least 20 SW and if, after the initial visit, it was thought possible to obtain the collaboration of SW and, for the roamers, of the bar/hotel owners and personnel. Visits took place in July to September 2001 in 15 sites of prostitution around Accra. All nine seater communities approached agreed to participate. The number of clients recruited was proportional to the number of seaters in each site. Fifteen bars/hotels where roamers worked were approached, six agreed to participate. Reasons for non-participation were denial of prostitution, fear of losing business and poor frequentation by clients at the time of the initial visit. To facilitate analysis, contiguous prostitution venues were regrouped into six units for the seaters and three for the roamers.
Clients were invited to participate by the SW and by a nurse who sought a verbal consent. It was not possible to estimate reliably the proportion of clients that refused to participate. Consenting clients provided a sample of first-catch urine before the intercourse. After intercourse, the SW took the client to a nearby location where a male field worker administered a questionnaire anonymously. Clients were free to withdraw at any time. The urine was tested with a leukocyte esterase dipstick (Combur Test M, Roche, Mannheim, Germany). Clients with a positive result were given ciprofloxacin (500 mg) and doxycycline (100 mg twice a day for 7 days). Condoms and educational pamphlets were distributed as well as a card allowing clients to receive free STI treatment at Adabraka polyclinic if needed. Information on condom use and price of the intercourse was also obtained from the SW, as well as their age. Boyfriends of SW and employees of prostitution venues were enrolled irrespective of whether they had had a recent intercourse with any SW; information was sought about their last intercourse with a SW.
Two urine samples were frozen at −20°C for the polymerase chain reaction (PCR) detection of sexually transmitted pathogens (results to be reported elsewhere). Two samples were refrigerated at 4°C and tested for anti-HIV antibodies with the Calypte HIV-1 urine enzyme immunoassay (EIA) (Calypte, Carson, California, USA) . Samples where the initial EIA was non-reactive were considered negative. If the initial EIA was positive, it was repeated in duplicate. If both repeat EIA were non-reactive, the sample was considered to be negative. Repeatedly reactive samples were further tested using the Cambridge Biotech HIV-1 urine Western blot (Calypte). Data were analysed with LogXact-4 and StatXact-4. Proportions were compared using the chi-square or Fisher's exact test. Continuous data were compared with the Student t-test or the Kruskal–Wallis test if distribution was not normal. Logistic regression was used for multivariate analysis. Potential interactions were sought in bivariate analyses, using a 0.15 threshold for their subsequent examination in multivariate analyses, where interactions with P < 0.05 were considered significant.
The population attributable fraction (PAF) of prevalent cases of HIV acquired during transactional sex was calculated using data from this survey and four other sources: (i) estimates of the number of transactional sex acts per year per SW from data collected at the Adabraka SW clinic from SW presenting for the first time; ii) estimates of the number of SW in Accra based on these records and on information gathered by peer educators during outreach visits; iii) number of adult men from the Ghana national census of 2000; iv) an estimate of HIV prevalence among pregnant women in 2001, used as a proxy for the prevalence among adult men, from the national surveillance system.
A total of 614 men accepted to participate in at least part of the study; 391 were recruited through seaters (355 clients, 26 boyfriends, 10 members of personnel) and 223 through roamers (167 clients, 5 boyfriends, 51 members of personnel). A questionnaire and a urine sample were obtained from 450 individuals, while only urine was obtained from 164 participants. For 22 participants, there was enough urine to perform the PCR but not enough for HIV serology. Respectively 41, 10, 6 and 0% of clients of seaters, clients of roamers, boyfriends and employees gave urine but then refused to answer the questionnaire. For clients of seaters, this proportion varied from 70% (Korle- Wonko) to 8% (Abeka) and seemed related to geographic dispersion of the SW within sites. Ninety-two per cent (413/447) of men were Ghanaian, 63% (283/446) had never been married, and they paid between 1.000 and 50.000 Cedis (median: 6.000; approximately US$0.85). Among clients, age ranged between 13 and 64 years and the period over which they had frequented SW was 0–31 years (median, 1).
Table 1 shows the demographic and behavioural characteristics of the four categories of men. Compared with clients of roamers, clients of seaters were older, more experienced with SW, reported having had sex with fewer SW other than the current one during the last year but were more likely to have had sex with the same SW before. Forty clients alleged that this was their first intercourse with a SW. Boyfriends reported a median of 11 intercourses per year with the SW which, along with the payment for sexual acts, suggests that they were rather regular, intimate clients of a given SW. Boyfriends and members of personnel were less likely to have used a condom during the last intercourse. In 90% (253/281) of intercourses that were monitored, clients and SW agreed that a condom had been used; they agreed four times (1.4%) that no condom was used. In 22 (8%) occasions, the SW reported that a condom was used while this was denied by the client; in two (0.7%) instances the opposite was seen.
Ten specimens (1.7%) out of the 592 tested for anti-HIV antibodies remained indeterminate and will not be considered further. The highest HIV prevalence was found among boyfriends (9/28, 32.1%) and personnel of prostitution venues (10/57, 17.5%). HIV prevalence was threefold higher among clients of seaters (53/335, 15.8%) than in clients of roamers (8/162, 4.9%) (P = 0.002). It was 33% (6/18) among members of personnel who admitted to previous intercourse with a SW, compared with 10% (4/39) in those who denied it (P = 0.06). None of 27 first-time clients of roamers was infected with HIV compared with 18% (2/11) of first-time clients of seaters (P = 0.07). HIV prevalence was similar in men who completed the questionnaire (57/424, 13%) and those who did not (23/158, 15%) (P = 0.83). As shown in Table 2, non-use of condoms during last intercourse (client-defined) and HIV prevalence varied substantially according to locations. When examining the nine sites of prostitution, there was a strong correlation (r = 0.75; 95% confidence interval, 0.17–0.94; P = 0.02) between HIV prevalence and frequency of non-use of condoms. Among seaters, the number of exposures to HIV per annum (prevalence among clients multiplied by the number of clients multiplied by the proportion of intercourses that were unprotected) varied from 0 in Abeka and Korle-Wonko, six in Adabraka and Labadi/Teshie, 56 in Nima/Accra New-Town, up to 124 in Tema- Ashaiman; among roamers, it was 0.35 in Kwame-Nkrumah Circle, four in Hotels-2 and 51 in Kokomba.
Table 3 displays clients’ characteristics that were correlated with non-use of a condom during last intercourse (client-defined). As there was borderline (P = 0.05–0.15) interaction between type of SW and other variables in bivariate analyses, and since it was known a priori that there were important differences between the two categories of SW , stratified analyses are presented. The frequency of intercourse with any SW during the last year was higher among non-users of condoms than in users, both in clients of seaters (median, 12 versus 5; P = 0.002) and clients of roamers (median, 9 versus 3; P = 0.02). The median price did not vary according to whether or not a condom had been used. Frequency of non-use of condom increased with age of the SW, from 5% (5/94) to 10% (11/108) and 16% (14/89) in the 16–24, 25–34 and 35+ year age groups (P = 0.07). In a multivariate analysis with stratification by type of client, the independent risk factors of non-use of condom during the last intercourse were older age of client (P = 0.01), current urethritis (P = 0.02) and frequency of intercourse with any SW during the last year (P < 0.001). There was no significant interaction.
HIV prevalence increased dramatically with age of clients (Table 4) but also with age of SW, from 2% (2/99) to 15% (19/128) and 18% (32/177) in clients of SW aged 16–24, 25–34 and 35+ years, respectively (P < 0.001). HIV infection was associated with repeated intercourse with the current seater but not with the total number of sexual contacts with any SW during the last year. In multivariate analysis stratified by type of client, there were three independent risk factors for HIV among clients: older age of the client (P = 0.002), older age of the SW (P = 0.06) and past episode of urethritis (P = 0.03).
To calculate the PAF value of HIV infections acquired from SW, it was estimated that, on average, seaters worked 245 days per year (excluding 60 days for menstrual periods and 60 days to attend festivals or funerals, to visit relatives, to grow crops in their villages and for sickness) and that roamers work 266 days (most do not attend festivals nor grow crops). Seaters and roamers had in 2001, an average of 4.5 and 2.5 intercourses, respectively, with clients each working day, which represented the mean of the self-reported number of clients the day before and of their estimation of the number of clients per day in general. These numbers, obtained from a questionnaire routinely used at the Adabraka clinic, were similar to those measured in 1997–1999 . As clients of seaters and of roamers had on average 18.0 and 9.1 intercourses with SW in the last year, there must be an average of 61.25 clients for each seater and 73.1 clients for each roamer. Based on our 7-year experience of community interventions with SW, it was estimated that there were approximately 500 seaters and 4500 roamers in Accra/Tema. In such circumstances, there would be 30.625 men having used the services of seaters, 4.845 of them being HIV seropositive, and 328.950 men having visited roamers, 16.244 of them being HIV seropositive. The Accra and Tema districts had in 2000 a population of 661.718 males aged 15–59 years ; given a 4.4% annual growth , this reached 690.833 in 2001. HIV prevalence among antenatal clinics attendees of Accra/Tema was 3.3% in 2001 . Presuming that the prevalence is similar in men, a large majority (21.089/22.797; 92.5%) of HIV-infected adult men had been clients of SW. It can be calculated that HIV prevalence among men who did not buy sex was 0.5%, while it was 5.9% among clients of both categories of SW combined, for a relative risk of 11.8. Therefore, the PAF was 84%. In other words, four-fifths of prevalent cases of HIV infection among males aged 15–59 years in Accra/Tema could be attributed to transactional sex. Table 5 shows a sensitivity analysis of the PAF. The number of seaters was kept constant at 500, as this estimate seems reasonably accurate. For roamers, it was presumed that their true number must lie within ±1000 of the best estimate of 4500. For HIV prevalence among clients, the lower and upper limits of 80% confidence intervals of the measured prevalence were used. For the prevalence among the adult male population, 1% was arbitrarily added or subtracted to account for sampling variation and biases inherent to extrapolating from antenatal clinic attendees.
In 1997–1999, HIV prevalence among seaters and roamers was 74% and 26%, respectively, and substantial differences were found between the two groups with regard to age, duration of prostitution, number of clients, price per intercourse and condom use . This mirrors the threefold higher prevalence of HIV among clients of seaters (15.8%) compared with clients of roamers (4.9%). In Cotonou, HIV prevalence among clients of SW was 8%, while it was 2–3% among the general adult population and 41–55% in SW [9,11,20,21].
Our estimate that 84% of prevalent cases of HIV among males in Accra can be attributed to transactional sex is based on a number of assumptions. Estimation of the number of clients per day used data gathered among first-time attendees at the Adabraka clinic; it seems unlikely that this number differed substantially between these SW and those who had attended before. Estimation of the number of seaters must be fairly accurate as they are easy to enumerate. Our sample of their clients, although not a random one, was probably representative, as they were recruited after having had sex with 203 different SW, almost half of all seaters. Clients who refused to participate may have differed from those who accepted for some characteristics, including HIV prevalence, but HIV prevalence did not differ between men who, after giving urine, accepted or did not accept to be interviewed, so a strong selection bias seems unlikely. Our sample of clients of roamers may have been less representative; since it was more difficult to obtain the collaboration of roamers and personnel of prostitution venues, we had to focus on larger downmarket bars/hotels, where number of clients per day might be higher than elsewhere. It is more difficult to estimate accurately the total number of roamers, who are mobile and poorly organized, but 4.500 seems a conservative minimum. A study of four African cities estimated that there were between 10.1 (Cotonou) and 19.5 (Kisumu) SW per 1.000 men aged 15–59 years . Our total of 5.000 SW corresponds to 7.2 SW per 1000 men. HIV prevalence among adult men was extrapolated from surveys of antenatal clinics attendees. Such measures are sometimes higher than prevalence among men, who are less susceptible to HIV than women [7,22]. This bias as well as an underestimation of the number of SW would result in an underestimation of the proportion of HIV-infected men who are clients of SW, and of the PAF of transactional sex. In contrast, if we overestimated the number of clients per day and if we were to use instead the number of clients they had the previous day (seaters, 4.2; roamers, 2.2), the PAF would be 67%. The sensitivity analysis displayed in Table 5 shows that our measure is robust. Univariate estimates of PAF assume that there is no confounding. Men using the services of SW may be more promiscuous and have more non-SW partners than others but, given the low HIV prevalence among non-SW women, exposure through them would be infrequent, resulting in little confounding.
Sex workers and their clients remain at the very centre of the HIV/AIDS epidemic in Ghana. Roughly four-fifths of prevalent cases of HIV among adult males would have been avoided if transmission from SW to their clients had been stopped with 100% condom use. The prevalence measured among clients reflected their behaviours over several years; for part of this period, condom use was lower than currently. The PAF of incident cases of HIV infection attributable to transactional sex cannot be determined, but it seems likely that it must slowly become lower than the PAF of prevalent cases as the virus progressively spreads outside core groups and as condom use during transactional sex increases.
Condom use reported by clients is more reliable than self-reports by SW (social desirability bias). In Zimbabwe, Gambia and Bénin, estimates of condom use by SW were 10–24% higher than client reports [9,20,21,23–25]. The 90% condom usage (reported by client and corroborated by SW) in Accra is comparable to the 93% reported in Thailand after the ‘100% condom’ policy was implemented . Admittedly, we might have overestimated the frequency of condom use with SW in Accra if non-participants were less likely to use condoms than participants. Condom use among well-organized and informed groups of SW may not reflect what occurs among isolated and vulnerable SW. Nevertheless, our data show that high rates of condom use in core groups can be achieved in sub-Saharan Africa. This was probably a consequence of sustained efforts for education of SW, their networking and the constant availability of condoms, but it may also reflect mass-media marketing for the general population, which could enhance the desire of clients to use condoms. Frequency of condom use was lower when clients had sex with other types of partner, specially their wives.
A striking finding was the important variation in condom use and HIV prevalence among clients according to locations, especially among the seaters. The strong correlation between HIV prevalence and non-use of condoms suggests that this behaviour had been going on for years in selected venues. Men who did not want to use condoms knew where they could obtain such services and went, time after time, to the same location, where they repeatedly had unprotected sex with the same SW. The solidarity, or lack thereof, between SW in a given location must be important determinants of condom use. At the individual level, a SW and her regular client may become so familiar that condom use is considered unnecessary. Many ‘boyfriends’ paid for the intercourse and were presumably regular clients who developed friendship with a SW. Overall, 39% of boyfriends of seaters were HIV infected. Their role in infecting SW might be important in areas of high condom use.
Age of the client and age of the SW with whom the last intercourse occurred were independent risk factors for HIV infection among clients, reflecting cumulative exposure. HIV infection among clients was independently associated with a past history of urethritis, underscoring the importance of non-ulcerative STI as a co-factor of HIV transmission among core groups in West Africa. A past history of genital ulcer was not associated with HIV, in line with recent findings that the rarity of genital ulcers is one of the determinants of the relatively low HIV prevalence in West Africa .
Clients should be targeted by educational campaigns in prostitution sites and mass-media campaigns. For some SW, non-users of condoms represented a third of their income; these women were older, presumably less attractive and economically vulnerable. Nevertheless, if solidarity between SW could be strengthened, it seems plausible that recalcitrant men would eventually have no other choice than using a condom. Efforts to reach SW not yet using the services available should be given the highest priority. Less-fatalistic approaches to HIV control in West Africa are needed. National programmes aiming to enhance condom use during transactional sex would lead to a reduction in HIV and STI incidence and prevalence in the general population. Preventive and curative services for SW and their clients should be organized with the same goal of nationwide access as for other public health priorities such as immunization and the management of malaria.
We are indebted to the field workers and the project peer educators who helped with the field work and with data collection.
Sponsorship: This work was supported by the Canadian International Development Agency, which funds the West Africa Project to Combat AIDS and STI, the public health intervention around which this study was organized, and has provided a scholarship to Anne-Marie Côté through the Canadian Bureau for International Education. CIDA had no role in study design, data collection, analysis, interpretation or in the writing of this article.
Note: There are no conflicts of interest.
1. Kreiss JK, Koech D, Plummer FA, Holmes KK, Lightfoote M, Piot P, et al.AIDS virus infection in Nairobi prostitutes. Spread of the epidemic to East Africa.N Engl J Med
2. Piot P, Plummer FA, Rey MA, Ngugi EN, Rouzioux C, Ndinya-Achola JO, et al.Retrospective seroepidemiology of AIDS virus infection in Nairobi populations.J Infect Dis
3. Plummer FA, Nagelkerke NJD, Moses S, Ndinya-Achola J, Bwayo J, Ngugi E. The importance of core groups in the epidemiology and control of HIV-1 infection.AIDS
(Suppl 1): S169–S176.
4. Merson MH, Dayton JM, O'Reilly K. Effectiveness of HIV prevention interventions in developing countries.AIDS
5. Marseille E, Hofmann PB, Kahn JG. HIV prevention before HAART in sub-Saharan Africa.Lancet
6. Creese A, Floyd K, Alban A, Guinness L. Cost-effectiveness of HIV/AIDS interventions in Africa: a systematic review of the evidence.Lancet
7. UNAIDS. Report on the Global HIV/AIDS Epidemic
. Geneva: UNAIDS, 2002.
8. Asamoah-Adu C, Khonde N, Avorkliah M, Bekoe V, Alary M, Mondor M, et al.HIV infection among sex workers in Accra: need to target new recruits entering the trade.J Acquir Immune Defic Syndr
9. Alary M, Mukenge-Tshibaka L, Bernier F, Geraldo N, Lowndes CM, Méda H, et al.Decline in the prevalence of HIV and sexually transmitted diseases among female sex workers in Cotonou, Bénin, 1993–1999.AIDS
10. Lankoande S, Méda N, Sangaré L, Compaore IP, Catraye J, Sanou PT, et al.Prevalence and risk of HIV infection among female sex workers in Burkina Faso.Int J STD AIDS
11. Morison L, Weiss HA, Buvé A, Carael M, Abega SC, Kaona F, et al.Commercial sex and the spread of HIV in four cities in sub-Saharan Africa.AIDS
12. Ghys PD, Fransen K, Diallo MO, Ettiegne-Traore V, Coulibaly IM, Yeboue KM, et al.The associations between cervicovaginal HIV shedding, sexually transmitted diseases and immunosuppression in female sex workers in Abidjan, Côte d'Ivoire.AIDS
13. Buvé A, Carael M, Hayes RJ, Auvert B, Fery B, Robinson NJ, et al.Multicentre study on factors determining differences in rate of spread of HIV in sub-Saharan Africa: methods and prevalence of HIV infection.AIDS
14. Esu-Williams E, Mulanga-Kabeya C, Takena H, Zwandor A, Aminu K, Adamu I, et al.Seroprevalence of HIV-1, HIV-2 and HIV-1 group O in Nigeria: evidence for a growing increase of HIV infection.J Acquir Immune Defic Syndr
15. Ettiègne-Traoré V, Ghys PD, Maurice C, Hoyi-Adonsou YM, Soroh D, Adom M-L, et al.Evaluation of an HIV saliva test for the detection of HIV-1 and HIV-2 antibodies in high-risk populations in Abidjan, Côte d'Ivoire.Int J STD AIDS
16. Robinson NJ, Mulder D, Auvert B, Whitworth J, Hayes R. Type of partnership and heterosexual spread of HIV infection in rural Uganda: results from simulation modelling.Int J STD AIDS
17. Berrios DC, Avins AL, Haynes-Sanstad K, Eversley R, Woods WJ. Screening for human immunodeficiency virus antibody in urine.Arch Pathol Lab Med
18. Anonymous. 2000 Population and Housing Census. Accra: Ghana Statistical Service, 2002.
19. Anonymous. HIV Sentinel Surveillance 2001. Accra: National AIDS Control Programme, 2002.
20. Lowndes CM, Alary M, Gnintoungbé CAB, Bédard E, Mukenge L, Geraldo N, et al.Management of sexually transmitted diseases and HIV prevention in men at high risk: targeting clients and non-paying sexual partners of female sex workers in Benin.AIDS
21. Lowndes CM, Alary M, Meda H, Gnintoungbe CA, Mukenge-Tshibaka L, Adjovi C, et al.Role of core and bridging groups in the transmission of HIV and STI in Cotonou, Benin, West Africa.Sex Transm Infect
22. Glynn J, Buvé A, Carael M, Musonda RM, Kahindo M, Macauley I, et al.Factors influencing the difference in HIV prevalence between antenatal clinic and general population in sub-Saharan Africa.AIDS
23. Wilson D, Chiroro P, Lavelle S, Mutero C. Sex worker, client sex behaviour and condom use in Harare, Zimbabwe.AIDS Care
24. Pickering, H. Quigley M, Hayes R, Todd J, Wilkins A. Determinants of condom use in 24000 prostitute/client contacts in The Gambia.AIDS
25. Labbé AC, Gnintoungbé CAB, Méda H, Lowndes C, Batona G, Belleau M, et al.Clients of sex workers in Cotonou, Bénin: risk factors and prevalence of sexually transmitted infections.XIV International Conference on AIDS.
Barcelona, July 2002 [abstract ThPeC7582].
26. Nelson KE, Celentano DD, Eiumtrakol S, Hoover DR, Beyrer C, Suprasert S, et al.Changes in sexual behaviour and a decline in HIV infection among young men in Thailand.N Engl J Med