Sexually transmitted infections and associated socio-demographic and behavioural factors in women seeking primary care suggest Madagascar’s vulnerability to rapid HIV spread


Frieda Behets, University of North Carolina, Department of Epidemiology, School of Public Health, 2104 G McGavern-Breenberg Building, CB 7400, Chapel Hill, NC 2599-7400, North Carolina, USA. Telephone + 1 919 966 7440; E-mail:


OBJECTIVES To examine sexually transmitted infections (STIs) and associated socio-demographic and behavioural factors in women seeking care for genital discharge syndrome in Antananarivo, Madagascar.

METHODS One thousand and sixty-six consecutive symptomatic women were interviewed and examined; bacterial vaginosis (BV), vulvovaginal candidiasis, trichomoniasis (TV), cervical infection (CI) due to chlamydial or gonococcal infections, and syphilis seroreactivity were determined by laboratory diagnosis. Associations between STIs and individual characteristics were evaluated using bivariate and logistic regression analyses.

RESULTS The prevalence of BV, TV, CI, and syphilis seroreactivity was, respectively, 85%, 16%, 49%, 16% in 94 prostitutes; 70%, 18%, 30%, 13% in 96 occasional sex traders; and 53%, 24%, 17%, and 4% in 876 general women. CI was independently and positively associated with a symptomatic partner, new sex partner in last 3 months, unfaithful partner, prostitution, joblessness and being < 25 years old. Syphilis was associated with low schooling, young age at coital debut, sex trading, and > 1 sex partner in the previous 3 months.

CONCLUSIONS These high STI rates and associated characteristics suggest the local vulnerability to rapid HIV spread and show the need for prevention efforts that involve youth, prostitutes, occasional sex traders, sex clients, and men who have concurrent sexual partnerships.


Relatively few people are infected with HIV in Madagascar, particularly compared to sub-Saharan Africa. Even in groups such as sex workers and patients who seek care for sexually transmitted infections, serosurveys have documented HIV prevalence rates of less than 1% (Harms et al. 1994; Behets et al. 1996; Zeller et al. 1997). However, these surveys also demonstrate high syphilis prevalence rates. Syphilis and other curable STIs seem to pose a substantial public health problem in Madagascar, although available information is scant. Infertility data can provide an indirect measurement of the magnitude of the STI problem in a population, as untreated gonorrhea and chlamydial infections, and possibly bacterial vaginosis, can ascend into the upper female genital tract and cause infertility due to damaged fallopian tubes (Wasserheit & Holmes 1992; Cates & Brunham 1999). In a study that compared patterns and predictors of infertility among women in 27 African nations, Madagascar was found to have infertility rates within the highest quartile (Ericksen & Brunette 1996). Risk factors for female infertility in Madagascar were current unmarried status while exposed to conception (i.e. sexually active) and urban residence.

While many Malagasy clinicians report frequent consultations for STIs, comprehensive, reliable data are lacking in Madagascar. The Malagasy Ministry of Health has estimated that across the island STIs are the sixth most common cause for seeking health care (personal communication, Desire Rasamilalao). The demand for treatment of genital discharge was illustrated in a report portraying the scope of STIs among the first five patients presenting each day at the specialized clinic of the Institut National d’Hygiène Sociale in Antananarivo (Harms et al. 1994): Of 226 women enrolled in the study from November 1992 to the end of April 1993, 93% had cervico-vaginal discharge and 8% had genital ulcers.

Improving STI treatment in patients seeking care in primary health care settings may help to decrease the STI burden in the population and also prevent the spread of HIV (Grosskurth et al. 1995; Anonymous 1997; Hitchcock & Fransen 1999). The Malagasy Ministry of Health resolved to maintain the still relatively low HIV prevalence in Madagascar through better STI control using locally feasible public health approaches. The data reported in this paper are part of a study undertaken to improve primary care of the genital discharge syndrome.


The study was approved by the Committee on the Protection of the Rights of Human Subjects of the University of North Carolina and by the ethical review board of the Malagasy Ministry of Health.

Study subjects were evaluated in Antananarivo, Madagascar’s capital, at the public STI clinic of the Institut d’Hygiène Sociale (IHS) and at the nongovernmental ‘67 Ha’ STI clinic. Eligible women who presented at the Ostie and Jirama private primary health care centres were referred to the IHS clinic. All consecutive women presenting with vaginal discharge syndrome (i.e. who complained of genital discharge, itching, or dysuria), who were at least 18 years old, and presented for a new episode were eligible for enrolment in the study. Patients were enrolled after informed consent was obtained in the local language. Free care and confidentiality was offered to the study patients and to their sexual partners.

Clinical exams and specimen collections took place at the IHS and 67 Ha clinics, where a pre-tested, structured questionnaire in Malagasy was administered to each participant by one of seven trained clinicians. From 17 March 1997 to 28 May 1997, all eligible women presenting consecutively at the participating institutions were evaluated. A total of 1066 women were enrolled. There were no refusals.

All women underwent a pelvic exam using a speculum during which vaginal samples were collected for laboratory analyses. Women were asked to provide a urine sample for gonococcal and chlamydial testing using ligase chain reaction (LCR) (LCx Probe System, Abbott Laboratories, Abbott Park, Ill) following manufacturer’s instructions. Urine samples were transferred into cryovials for storage at –20 °C until transported on dry ice to the laboratory of the University of North Carolina.

Vaginal secretions were microscopically examined for yeast, clue cells, and mobile trichomonads. A venous blood sample was collected from all study participants for syphilis serology using rapid plasma reagin (RPR; Becton Dickinson, Cockeysville, MD). Sera that were found to be RPR reactive were diluted to determine endpoint reactivities and were tested using microhaemagglutination for Treponema pallidum (TPHA; Fujirebio, Tokyo).

The study participants were treated on the same day based on a risk-inclusive algorithm recommended by the World Health Organization (WHO 1994) and following national treatment guidelines for STIs. The women were invited to return to the clinic approximately one week later to ensure that all infections detected in the local laboratory were treated adequately. The women were also encouraged to bring their sexual partner(s) to the clinic for treatment.

The clinical information and the interview data were recorded by the study physicians on pre-numbered data collection forms. The data from the questionnaires, laboratory and clinical forms were entered twice into a database and cleaned using EpiInfo 6.03 (CDC, Atlanta). Data were analysed with SAS 6.12 and 8 (SAS Institute, Cary, NC). The prevalence rates of bacterial vaginosis, candidiasis, trichomoniasis, gonococcal and chlamydial cervical infections were determined. Bacterial vaginosis was diagnosed when at least three of the following four characteristics were present: homogeneous, white, noninflammatory discharge; clue cells found in microscopic examination of vaginal secretions; pH of vaginal fluid >4.5; fishy odour of vaginal discharge after or without adding 10% potassium hydroxyde (KOH) (CDC 1998). Candidiasis was diagnosed when yeast forms were observed in vaginal secretions.

Sociodemographic, behavioural, and clinical characteristics of the study participants were described using standard descriptive statistics (frequencies, mean/standard deviation). Characteristics hypothesized to be associated with the outcomes of interest were examined using Fisher exact test, likelihood ratio Chi-square or Mantel-Haenszel Chi-square test for trend when applicable for categorical variables; and t-test or analysis of variance for continuous variables. Patient records that lacked results for a particular variable under analysis were excluded from that analysis, hence denominators may vary.

The variables hypothesized to be associated with the outcome of interest were entered into a logistic regression model regardless of the bivariate analyses (Sun et al. 1996). First all the sexual behavioural factors in addition to age were entered into the model. Subsequently, covariates that did not contribute to the estimation of cervical infection at the 0.05 significance level were removed from the model one at a time. The sequence of covariate removal from the model was determined by likelihood ratio testing to ensure that the covariate that contributed the least to the prediction of the outcome or the fit of the model was removed first. Variables that did not significantly contribute to the fit of the model were evaluated one at a time to determine whether they acted as confounders for other covariates. Confounding was determined to exist when removal of the potentially confounding variables changed the odds ratio of a risk factor by at least 10%. Confounding variables were maintained in the model. Following this procedure, all the socio-demographic variables were added to the model. First, effect modification was assessed of the socio-demographic variables with the sexual behavioural variables that were still in the model, one at a time. Any interaction term significant at the 0.01 level as well as the main effects of the interacting variables were to be kept in the model. However, no significant interaction terms were found. Subsequently, the final model was built through covariate removal, one at a time, based on likelihood ratio testing and subsequent checking for confounding variables as described.


Socio-demographic characteristics

The average age of the 1066 evaluated women was 27.6 years (median: 26 years). 2% complained spontaneously of dysuria, 15% of genital pruritus, 81% of genital discharge, and when asked, 98% complained of genital discharge. Half of the study women reported having no live children; 235 (22.0%) had one child; 132 (12.4%) had two children, and the remaining 171 (16.0%) had three or more children. Other socio-demographic characteristics are given in Table 1. Employment status of the steady partner was obtained from 842 women: 17.0% reported that their partners were factory workers; 11.6% drivers; 10.4% employees or professionals; 10.8% traders; 7.9% students; 6.6% military or police; 3.1% unemployed, and 32.4% in unspecified ‘other’ employment.

Table 1.   Socio-demographic characteristics of Malagasy women seeking primary care for genital discharge syndrome by self-reported sexual behaviour Thumbnail image of

Women with the lowest socioeconomic status (SES) were defined as those who lacked tap water and a flush toilet in their home and who had no electricity or no radio in their home. Consistent with this definition, the 283 poorest women were more likely to be less educated than the 779 women of higher SES as measured by reported years of schooling (P < 0.001). The 282 women from the lowest SES reported a mean number of 2.9 occupants per room compared to 2.5 for women of a higher SES (P < 0.001).

Sexual behavioural characteristics

In addition to the 94 women who gave prostitution as their profession, 96 (9.0%) of the 1066 subjects answered ‘yes’ when asked if they had ever traded sex for money or goods. We labelled these 96 women occasional sex traders to distinguish them from the 94 professional prostitutes and from the remaining 876 ‘general’ women who did not report these behaviours. The study participants were thus categorized into one of three groups depending on the self-reported transactional nature of their sexual behaviour: either professional sex trading, occasional sex trading, or no acknowledged sex trading. While these categories were not hypothesized per se or conceived as explanatory, presenting the various characteristics of the study women by self-reported sexual behaviour was judged to provide a more informative description.

Fewer general women had the lowest SES than occasional sex traders, who in turn were better off than prostitutes (P=0.001) (Table 2). There was a similar pattern for the level of schooling: prostitutes had the lowest education levels, followed by sex traders and general women (P=0.001). Prostitutes were younger and had started sexual intercourse at a younger age than occasional sex traders who were in turn younger and had started having sex at a younger age than general women (Table 2). General women reported on average fewer sexual partners in the last 3 months than sex traders who reported fewer partners than prostitutes (P=0.001).

Table 2.   Selected socio-demographic and sexual behavioural characteristics in women seeking primary care for the genital discharge syndrome by self-reported sexual behaviour Thumbnail image of

General women were more likely to report multiple partners during the past 3 months when their partner had other sex partners than women whose partner did not have sex with others (62 (17.8%) of 349 vs. 37 (7.1%) of 524 women, respectively; OR: 2.8; P < 0.001). However, 287 (32.9%) of 873 general women reported that their partner had sex with others while they were monogamous during the preceding 3 months. On the other hand, 37 (4.2%) of 873 general women reported that they had more than one sexual partner in the preceding 3 months while their partner had no outside sex partner. Of 96 occasional sex traders, 35 (36.5%) reported that their partner had sex with others while they were monogamous during the preceding 3 months.

Overall, 211 (19.8%) of the 1066 patients reported a history of spontaneous abortion, and 503 (47.2%) underwent induced abortion. The prevalence of induced abortion did not differ significantly among the three groups but a history of spontaneous abortion was reported by 18.4% of the general women, 29.2% of the sex traders, and 23.4% of the prostitutes (P=0.004).

Bacterial vaginosis

Bacterial vaginosis was found in 601 (56.4%) women and was most prevalent in prostitutes followed by sex traders and general women (Table 3). Women with bacterial vaginosis reported a higher number of sexual partners during the last 3 months: 9.1 vs. 2.5 reported by the women who did not have bacterial vaginosis (P < 0.001). The prevalence of bacterial vaginosis increased from 47.0% in women who reported no sexual partner during the last 3 months, to 53.9% in women with one partner, to 55.7% in women with 2–5 partners, to 75.0% in women with 6–10 partners, and 86.6% in women with more than 10 partners (P < 0.001).

Table 3.   Prevalence of vaginosis, vaginitis, and STIs in Malagasy women seeking primary care for the genital discharge syndrome by self-reported sexual behaviour Thumbnail image of

Trichomonas vaginitis

The prevalence of trichomoniasis was lowest among prostitutes, followed by the general women and highest among sex traders (Table 3). Of 291 women who were at least 30 years old, 72 (19.8%) had trichomoniasis. Women who were 25–29 years old and women who were less than 21 years old were more likely to have trichomoniasis than the women who were at least 30 years old (OR: 1.5; P=0.2 and OR: 1.8; P=0.005, respectively). The following factors were positively associated with trichomoniasis in bivariate analyses: not living with a steady partner (OR: 1.4; P=0.01); living regularly without the steady partner (OR: 1.6; P=0.002); ever having a used a condom during sexual intercourse (OR: 0.7; P=0.04). The sexual behavioural and sociodemographic variables found to be associated with trichomoniasis using logistic regression are presented in Table 4.

Table 4.   Sexual behavioural and socio-demographic characteristics associated with trichomoniasis using logistic regression analysis Thumbnail image of

Cervical infection due to gonorrhoea or chlamydia

Cervical infection due to gonorrhoea or chlamydia was most prevalent among professional prostitutes, followed by the occasional sex traders and least prevalent among general women (Table 3). Cervical infection due to Chlamydia or gonorrhoea was positively associated in bivariate analyses with the following characteristics: lowest SES (OR: 1.7; P < 0.001); jobless (OR:2.0; P=0.03); not living with a steady partner (OR: 1.7; P < 0.001); lived regularly without a steady partner (OR: 1.8; P < 0.001); ever exchanged sex for money or goods (OR: 3.3; P < 0.001); reported prostitution as profession (OR: 4.4; P < 0.001); nonconsensual sex in the past 3 months (OR: 1.7; P=0.06); sex partner with urethral discharge (OR: 2.2; P < 0.001); sex partner with STI symptoms, i.e. either genital discharge or ulcer (OR: 2.3; P < 0.001); more than one sex partner during the last 3 months (OR: 2.9; P < 0.001); partner had other sexual partners (OR: 2.3; P < 0.001); ever having used a condom (OR: 1.4; P=0.03). The mean number of sexual partners was 14.9 among the women infected with gonorrhoea or Chlamydia compared with 4.0 partners reported by the women who were not found to be infected (P < 0.001).

Low educational status was significantly associated with cervical infection: 26.4% of women who never completed primary school had cervical infection, 28.3% of the women who completed six years of schooling, 20.6% of the women who had nine years of schooling, 15.9% of the women who had 12 years of schooling, and 5.5% of the women who completed higher education had cervical infection (P=0.001). The behavioural and socio-demographic factors found to be associated with cervical infection using logistic regression are presented in Table 5.

Table 5.   Sexual behavioural and socio-demographic characteristics associated with cervical infection due to gonorrhoea or Chlamydia using logistic regression Thumbnail image of

Syphilis seroreactivity

Of 992 sera that were evaluated by RPR, 71 (7.1%) were reactive. Of 70 RPR-reactive sera, 33 (47.1%) had endpoint reactivity titres of 1 : 8 or greater. Sixty (89.5%) of 67 RPR-reactive sera had a positive MHATP test result. The prevalence of syphilis seroreactivity, i.e. RPR (+) and MHATP (+), was higher in prostitutes than in sex traders and in general women (Table 3). Syphilis seroreactivity was positively associated with the following characteristics: lowest SES (OR: 1.8; P=0.03); young age at coital debut (P < 0.001); partner had other sexual partners (OR: 2.0; P=0.006); partner with genital ulcers (OR: 3.5; P=0.03); nonconsensual sex in the last 3 months (OR: 3.4; P=0.003); new sex partner during the last 3 months (OR: 3.0; P < 0.001); more than one sex partner during the last 3 months (OR: 3.4; P < 0.001).

The average age of syphilis seroreactive women was 25.1 years compared with 27.8 years for syphilis seronegative women (P=0.005). The prevalence of syphilis was inversely associated with years of schooling in a linear fashion: 15 (17.0%) of 88 women with less than six years of schooling were syphilis-seroreactive; 26 (10.4%) of 250 women who completed primary school; 16 (4.8%) of 335 women with nine years of schooling; 1 (0.4%) of 245 women with 12 years of schooling; and 1 (0.9%) of 112 women with higher education (P=0.001). The sexual behavioural and socio-demographic variables associated with syphilis seroreactivity using logistic regression are reported in Table 6.

Table 6.   Sexual behavioural and socio-demographic characteristics associated with syphilis seroreactivity using logistic regression Thumbnail image of


Almost one in two women who sought primary care for the genital discharge syndrome had at least one curable STI in this study. This high STI prevalence rate confirms anecdotal and partial evidence that STIs are an important public health problem in Madagascar. Intriguingly, trichomoniasis was least common among self-proclaimed prostitutes. Potential reasons for this phenomenon include previous treatment and immunity, although the protection provided by immune response to infection appears unresolved (Krieger & Alderete 1999). We believe that the prostitutes in this study population might have sought reproductive health care more often than the other women. Since trichomoniasis can easily be diagnosed locally by microscopy, unlike gonococcal and chlamydial infections, prostitutes might thus have been more likely to have been previously treated. The low trichomoniasis prevalence in prostitutes might also be due to a measurement issue since detection of trichomonads using wet mount microscopy might be less sensitive in chronic cases with few organisms (Krieger & Alderete 1999).

The overall prevalence rate of bacterial vaginosis was high compared to studies in other countries. For instance, bacterial vaginosis was found in 15% of care-seeking patients in Brazil (Moherdaui et al. 1998); in 23% of symptomatic women in Malawi (Costello Daly et al. 1998); in 37% of nonpregnant women and in 21% pregnant women seeking care in Tanzania (Mayaud et al. 1998); in 30% of women seeking care in Peru (Sanchez et al. 1998), and in 24% of women attending primary healthcare centres in Morocco (Ryan et al. 1998). In contrast, an overall prevalence of 51% was found in a large community-based survey in Uganda, where bacterial vaginosis was carefully assessed (Sewankambo et al. 1997). Measurement issues may play a meaningful role in the widely varying reported prevalence rates. Bacterial vaginosis has been diagnosed in some studies based on laboratory results only and using varying criteria while in other studies a combination of clinical and laboratory characteristics were used. African-American women have also been found to be less colonized by lactobacilli than Caucasian women in the US (Hillier 1999), suggesting the possibility of differences in vaginal flora by ethnic or social groups.

The risk of having bacterial vaginosis clearly increased with the reported number of sex partners in the last 3 months. In a recent population-based study in rural Uganda, frequency of bacterial vaginosis increased with mounting numbers of reported partners but less than 4% of the women reported more than one partner in the past year (Sewankambo et al. 1997) compared to 22% of the patients who reported multiple partners in the previous 3 months in this study. Bacterial vaginosis has been hypothesized as a biological risk factor for increased susceptibility to STIs, and is associated with a decreased presence of lactobacilli, particularly those that produce hydrogen peroxide with microbicidal activities in vitro (Hillier 1999). A recent in vitro study documented the inhibition of Neisseria gonorrhoeae by certain species of H2O2-producing lactobacilli (Ruffner & Jerse 1999). In Uganda, HIV-1 infection was associated with bacterial vaginosis in women < 40 years irrespective of the number of sex partners (Sewankambo et al. 1997), suggesting that bacterial vaginosis may increase the susceptibility to HIV-1 infection.

The overall prevalence of cervical infection among the study patients was 21%. In a study conducted in 1994 in Jamaica, 34% of women seeking care for vaginal discharge syndrome had gonococcal or chlamydial infection (Behets et al. 1995). Recent studies in developing countries have found gonococcal or chlamydial infection prevalence rates among female symptomatic patients ranging from 8% in Morocco (Ryan et al. 1998) and Benin (Alary et al. 1998) to 11% in rural Tanzania (Mayaud et al. 1998); 15% in Peru (Sanchez et al. 1998); 17% in Brazil (Moherdaui et al. 1998); to 19% in Malawi (Costello Daly et al. 1998). The probability of cervical infection in symptomatic women seeking primary care for genital discharge syndrome in these Malagasy clinics was thus quite high and, together with the other sexual behavioural and biological study findings, suggests a considerable disease burden in the community. Although population-based data are needed to evaluate the magnitude of the STI problem in the community and direct policy, this study confirms that attention to STI case management as part of primary health care is warranted.

As elsewhere, young age was an independent risk factor for cervical infection. Why a woman who reported to be jobless was at greater risk of cervical infection is not obvious. The study data showed that the stated lack of a job was not significantly associated with a reported history of trading sex for money or goods, but was associated in general women with having no steady partner; a new sex partner in the last 3 months, and with reporting two to five sexual partners in the last 3 months rather than one or no sex partners. Women who stated that they had no job thus were likely to report some sexual behaviour that increased the risk of cervical infection while they reported that their sexual relations were nonmercantile. However, the measurement of trading sex for goods as well as other sexual behavioural factors through self-reporting may be imprecise. For instance, the perception of a sexual relationship as one in which sex was traded for goods or money may vary by individual and context. The study findings suggest that women who reported to have no job had other unmeasured risk factors in addition to those reported.

An independent risk factor for cervical infection was the woman’s opinion that her partner had other sex partners. Ryan et al. (1998) found in a study conducted in Morocco that women who believed that their partner was unfaithful were at increased risk of having cervical infection. Of the four variables that were associated in multivariate analysis with cervical infection in that study, the belief that the partner was unfaithful was the strongest predictor.

About half of the study participants reported a history of induced abortion, showing the need to improve local family planning policies and programs. While integrated care has been promoted in recent years, it usually implies the introduction of STI services in family planning clinics (Wilkinson et al. 1997; Temmerman et al. 1998). STI care providers should also be encouraged to enquire about birth control practices and needs and appropriately educate their patients about options and referrals. As an alternative strategy, the provision of birth control services could be part of reproductive health services together with STI care (Miller & Rosenfield 1996).

Although the syphilis seroprevalence rates in the care-seeking women were not as high as in other local surveys, they are high enough to warrant screening as part of routine clinical care. Coital debut at <17 years was independently and positively associated with syphilis. This may be due to longer exposure, associated sexual risk behaviour, or increased susceptibility associated with young age. The negative association of syphilis with years of schooling was found previously in Madagascar in prostitutes, care-seeking patients, and prenatal women (Behets et al. 1996). Why this particular socioeconomic factor was independently associated with syphilis cannot be understood from the study data. However, general education may influence attitudes regarding health, illness, and health seeking; it may change beliefs regarding alternative explanations for illness; it may enable an individual to interact with health services more competently; it may move the locus of control from elsewhere, such as in the family or supernatural, to the individual and it may also influence health workers’ attitudes and practices (Caldwell 1993). People with little education may have social and sexual mixing patterns that favour sexual intercourse with high-risk partners.

The women who reported prostitution as their occupation were the youngest of the three categories of women; they were predominantly single without a steady partner. They were significantly poorer and less educated than the sex traders who themselves were younger and of lower SES than the general women. While the study design does not allow any empirical explanatory interpretation, it appears plausible that prostitution was one of the few options for young women of lower SES with restricted educational opportunities. Improved general education, including nursery school, results in better health and social outcomes (Chard et al. 1999). Better access to education for girls and improved conditions for women in general encompassing availability of alternatives to prostitution and trading sex for goods or money might be among the most effective STI control strategies (Cohen et al. 1996).

Self-reported professional prostitutes started sexual intercourse at a significantly younger age than the sex traders, who in turn started earlier than the general women. Early coital debut has been described as a marker for high-risk sexual behaviour (Greenberg et al. 1991) and has been linked with a higher risk for STIs (Horn et al. 1990). While the prostitutes had the most high-risk sexual behaviours in general, they were most likely to have practised protected sexual intercourse albeit in an insufficient way, as indicated by their infections. This may be a result of local efforts targeted at professional sex workers to promote behaviours that reduce sexual risk, the prostitutes’ awareness of occupational risks, and the sex clients’ attitudes and behaviours.

Interestingly, prostitutes reported less frequently than sex traders to have a partner with penile discharge or genital ulcers. However, prostitutes were much more likely to ignore STI-related symptoms in their sexual partners than sex traders who in turn were more likely than general women to report not knowing if their partner(s) had STI-related symptoms. While sex traders might have had sex with partners who were more often symptomatic than the partners of the prostitutes, the probability of acquiring infection depends on several components of sexual behaviour, i.e. the number of sex partners and the number of sexual encounters as well as sexual practices such as the use of condoms, in addition to the choice of a high-risk, i.e. infected, partner.

The fact that women were willing to report prostitution as their occupation is encouraging from an STI control policy perspective since it allows identification and access to an important core group. In many settings, prostitution may be frequent but interventions are impeded by poor access (Day & Ward 1997). Moreover, the prostitutes sought care, thus offering an opportunity to provide needed services including the treatment of infections, learn more about their needs and obstacles to safer sexual behaviour, and promote consistent protected sex.

The occupational hazards of prostitutes were not limited to a high risk of having an STI. Almost one in four reported having been raped during the preceding 3 months. In the National Health and Social Life Survey conducted in the US in 1992, almost 22% of the surveyed women reported ever having been forced perform a sex act with a man (Laumann et al. 1994). But as the authors of this study noted, comparing rates of forced sex is fraught with problems since the definitions as well as the perceptions by the research subjects may vary considerably. From an STI control and prevention policy perspective, this problem may warrant attention. Rape during the past 3 months was not independently associated with STIs in multivariate analysis in this study, possibly because of the frequency of unprotected consensual sex. However, rape is a potential source of STI transmission and might be likely to occur without the use of mechanical barrier protection despite intensive condom use promotion interventions.

The high STI rates among prostitutes demonstrate the inadequacies of current STI prevention interventions and call for more effective approaches to sexual health promotion. Although the policy of efficiency-based targeting has been criticized as prone to stigmatizing and victimizing (Decosas & Pedneault 1996), STI prevention efforts directed at individuals who have a high number of sexual partners, and are thus likely to acquire and transmit STIs, can be expected to have a greater impact on the STI prevalence and incidence in the community than prevention programs aimed at the general public (Over & Piot 1996). The targeting of services to core groups as part of national STI control policy does not preclude a need-based approach that includes genuine attention to the inherent social dimensions.

The sex traders were, with few exceptions, situated between the prostitutes and the general women in terms of sexual behaviour, socioeconomic and demographic characteristics, and biological conditions including STIs. From an STI control policy perspective, sex traders as well as prostitutes are likely to play an important role in the spread of STIs and thus warrant specific attention and interventions.

The ‘general’ women were on average older, better educated, and the least poor of the study women. Almost 90% had a steady partner, and only about 11% reported more than one sex partner in the last 3 months. However, 40% reported that their steady partner had other sex partners and about 12% said that their partner had STI-related symptoms. Having a partner who is not monogamous and having a partner with STI-related symptoms were both independently and positively associated with cervical infection. About one third of the general women reported that their partner had sex with others while they had no outside sex partners during the preceding 3 months. The study data suggest that a significant proportion of STIs in the general women was due to the steady partners’ sexual behaviour. Similar results, indicating male promiscuity, were found in a study in Ethiopia that showed high STI rates in women still married to their first and only sex partner (Duncan et al. 1994). Treating women whose risk is attributable to their only sex partner’s behaviour will have little impact on the STI epidemic in the community (Over & Piot 1996). However, treating the women’s partners and promoting STI prevention among those multiple transmitters might have an effect.

The high STI prevalence rates in women seeking primary care for the vaginal discharge syndrome and the associated patterns of socio-demographic and behavioural characteristics suggest favourable conditions for rapid HIV spread in Madagascar. Although published data are currently not available, many Malagasy women seek care for vaginal discharge syndrome, and trading sex for money or goods appears to be common. Our study subjects, while not representative of the general population, are likely to be representative of a large group of Malagasy women. This vulnerability to an explosive HIV epidemic calls for prevention efforts that include youth, professional prostitutes, occasional sex traders, and sex clients, as well as men who have concurrent sexual partnerships.


We are grateful for assistance from the following individuals: Dr Vololona Rasolofo, Dr Arline Sahondra Vololona, Dr Mariette Ralalaharimanga, Dr Jean Razafindramonjy, Dr Aime Ramambazafy, Dr Rene Randriamanga, Dr Solofoson Rakotonandrasana, Dr Mamitiana Andrianalinera, Dr Simon Razafindrabe and Dr Claire Rasoazananoro, Dr Charles Rakotonirina, Dr Martin Rakotomanga, Dr Marie Rasoa, Dr Kotozafitsimiheloka, Dr Elise Rasoanjanahary, Dr Charlotte Razanasolonambinina, Dr Jules Randriamanana, Dr Joasephine Rasoamialinoro, and Dr Solange Razanamalala. Thanks to Richard Steen for help with clinical training.

This study was funded by the US Agency for International Development through Futures Group International and Family Health International. The contents of this report do not necessarily reflect the views or policies of the funding agencies.