Evidence of changes in sexual behaviours among male factory workers in Ethiopia

Objective: To assess changes in sexual behaviours among male factory workers in Ethiopia.

Design: Open cohort studies in two factories near Addis Ababa.

Data and methods: At intake and biannual follow-up visits, data were collected on sexual behaviours including casual sex, sex with commercial sex workers (CSW), condom use, and history of sexually transmitted diseases (STDs) as indicated by genital discharge and genital ulcer. Health education, HIV testing, and counselling were offered to all participants.

Results: Between February 1997 and December 1999, 1124 males were enrolled in the two cohort studies. At intake, the prevalence of casual sex in the past year, sex with CSWs, condom use with the last casual partner, history of genital discharge in the past 5 years, and history of genital ulcer in the past 5 years were 9.7, 43.4, 38.8 (Akaki site only), 10.6 and 2.1%, respectively. At the Wonji site, the intake prevalence of casual sex, sex with CSW, and history of genital discharge decreased significantly by calendar year between 1997 and 1999. At both sites combined, between the first and the fourth follow-up visits, there was a decline in the proportion of males reporting recent casual sex (from 17.5 to 3.5%, P < 0.001), sex with CSWs (from 11.2 to 0.75%, P < 0.001), and genital discharge (from 2.1 to 0.6%, P = 0.004).

Conclusion: There was a decline over time in risky sexual behaviours reported by cohort participants. Part of this decline occurred independently of cohort interventions.

Introduction

Changes in sexual behaviours have been advocated to reduce the spread of HIV in the developing world, where heterosexual sex is the most common mode of HIV transmission [^1-3]. These changes mainly involve reduction in the number of sexual partners [^4,5] and increase in condom use during casual sex [⁶]. The effectiveness of these changes is supported by data from Thailand and Uganda, where both HIV prevalence estimates and prevalence of risky sexual behaviours decreased in the past decade [^6-9]. However, few longitudinal studies have been conducted within the same population group to correlate reduction in behavioural risk with lower incidence rates of HIV infection or other sexually transmitted diseases (STD) [^6,8].

It is now almost 16 years since the HIV epidemic started in Ethiopia. The epidemic has spread throughout the country, as shown by HIV prevalence levels of 5-20% among blood donors in 11 major towns in 1994 [¹⁰]. By the end of 1999, an estimated 3 million adults and children were living with HIV/AIDS in Ethiopia [¹¹]. Recent surveillance data collected among antenatal care attendants in Addis Ababa, the capital city, suggest a decline in HIV prevalence among young pregnant women of the inner city [¹²]. This decline may be related to intervention programmes focusing on risk behaviour reduction, condom promotion, and STD control, which have been underway since 1987 in the country [¹⁰]. However, these programmes have been seriously constrained by years of political instability, and their effectiveness has not been evaluated in longitudinal studies. In this paper, we present the first evidence of change in sexual behaviours among male factory workers who participated in a cohort study on HIV incidence and disease progression in Ethiopia.

Materials and methods

Study population and procedures

The Ethio-Netherlands AIDS Research Project (ENARP) established two cohort sites in 1997 to study the incidence and progression of HIV infection in Ethiopia. One cohort is located in a Fibre Products factory at Akaki, a suburb of Addis Ababa, and the other at Wonji, a sugar estate 107 km south-east of Addis Ababa. Both cohorts followed the same procedures, which are detailed elsewhere [¹³], and summarized below. The two cohort studies are still ongoing, and have been planned to last for 8 to 10 years.

Factory workers of both sexes were invited to join the study during general information meetings. After signing an informed consent form at the project's clinics, each participant was given pre-test counselling by an experienced counsellor. During the session, the counsellor evaluated the participant's knowledge of HIV transmission, prevention, and course of infection, as well as his/her perception of HIV infection risk and the potential consequences of a positive or negative result on HIV antibody testing. After pre-test counselling, interviewers matched by sex with subjects collected data on socio-demographic characteristics, sexual behaviours and medical history, using a structured questionnaire. Each interview was followed by a clinical examination performed by a medical doctor, before blood was drawn for laboratory analyses. Finally, all individuals were informed that their HIV test results would be available 30 days later at the project's clinic. Attending post-test counselling (PTC) was optional, although an appointment date was given in writing to everyone. Participants were thereafter seen every 6 months at the study clinic for follow-up, and the above procedures were repeated at each visit.

Blood samples were transported on the day they were obtained to ENARP's laboratory at the Ethiopian Health and Nutrition Research Institute (EHNRI). Plasma samples were tested for HIV-1 antibodies by HIVSPOT (Genelabs Diagnostics, Singapore) and Vironistika enzyme-linked immunosorbent assay (Organon, Boxtel, The Netherlands). Positive results found with either or both tests were confirmed by Western blot testing (HIV Blot 2.2 Genelabs Diagnostics). Plasma samples were also tested for syphilis using the Treponema pallidum hemaglutination test (TPHA; Serodia-TP, Fujirebio, Tokyo, Japan) and, if results were positive, a further sample was tested using the Rapid Plasma Reagin (RPR) test (RPR Slide-Test; Biomerieux, Marcy-Etoile, France).

As an incentive to participate, factory workers and their families were offered free medical care for the duration of the cohort study. All study information was coded and kept confidential. The EHNRI Ethics Committee and the National Ethical Clearance Committee approved the study protocol.

Statistical methods

Due to the low number of women participating in the study, and the likely under-reporting of risky sexual behaviours by women, we have not included in the analysis data obtained among female participants. Analysis was also restricted to participants enrolled before 31 December 1999 and includes data from follow-up visits up to 31 December 2000. Individuals who missed visits for one year or more were considered lost to follow-up (dropouts).

Indicators used to assess changes in sexual behaviours were reports of casual sex, sex with commercial sex worker (CSW), condom use, history of genital discharge and history of genital ulcer. Casual sex was defined as having sex with someone who is not your spouse or your steady sexual partner. Sex with CSW was defined as having sex with a CSW when going to bars, the usual place to meet CSWs for these factory workers. Condom use was queried in relation to the last sexual act with a casual partner or CSW but these data were available only from Akaki. History of genital discharge and genital ulcer was based only on participant's recollection, not medical records. At intake, the questions on casual sex referred to the past year, and the questions on genital discharge and ulcer to the past 5 years. There was no specific time-frame for the questions regarding sex with CSW at intake, so that it was not possible to assume that persons reporting sex with CSW had casual sex in the past year. During follow-up visits, these questions referred to the time-period since the last visit. Any person reporting sex with CSWs during follow-up visits was assumed to have had casual partners since the last visit.

Background characteristics of the study participants who completed follow-up were compared with those of the dropouts, using χ² tests of independence. Test for trends were performed for ordered categories (age groups, numbers of sexual partners in lifetime). Furthermore, a multivariate analysis was performed using Cox proportional hazards model to identify independent predictors of the risk of dropping out from the cohort. Unlike the cohort at Akaki, where enrolment took place mostly in 1997, the Wonji cohort recruited participants throughout the period 1997-1999, allowing comparison of the intake prevalence of reported sexual behaviours and STDs by calendar year of enrolment (χ² test for trends). To study whether changes in intake prevalence of the various behavioural indicators might reflect diversity of background characteristics among the populations enrolled in successive years, the effect of calendar year on each indicator was examined in a logistic regression model, controlling for socio-demographic characteristics.

The incidence of HIV infection was calculated by dividing the total number of seroconversions by the total follow-up time of HIV-negative participants. For seroconverters, follow-up time until the seroconversion visit was included in the denominator of the calculation of the HIV incidence rate. The 95% confidence intervals (CI) were estimated assuming a Poisson distribution of events. The observed incidence rate was compared with that expected under the steady-state assumption. Under this assumption, prevalence is stable over time, and the incidence rate is the ratio of the odds of infection [prevalence/(1 - prevalence)] divided by the duration of infection [¹⁴]. The prevalence used for our calculation was the cohort intake prevalence. The duration of infection was estimated at 10 years in an African context [¹⁵]. The 95% CI of the expected incidence rate was calculated by applying the same formula as above and using for prevalence the 95% CI limits of the intake prevalence estimates.

Finally, predictors of behavioural indicators were examined using a generalized estimating equation (GEE) model for binary outcome with each follow-up visit as the unit of analysis and an exchangeable within-individual correlation structure. Data from the enrolment visit were ignored in this analysis, since the referent time period for behavioural indicators differed at enrolment compared to the follow-up visits. To explore potential biases related to the differential lost to follow-up of subjects with certain characteristics, two separate models were run: one including all participants and one including only participants not lost to follow-up. Variables included as potential predictors were socio-demographic characteristics (age, educational status, marital status), cohort site, attendance at PTC, HIV antibody test result, and time. Time was defined as the number of follow-up visits. Since the median time between two consecutive visits was 191 days, or about 6 months, the odds-ratio (OR) associated with the variable 'visit' gives the odds of reporting risky sexual behaviours per additional 6 months of follow-up. For those who attended PTC, the variable 'effect of PTC' was coded as a time-dependent variable, 0 until PTC took place, and 1 thereafter. As those attending PTC might have differed from those not attending by characteristics potentially associated with the outcome variables (because subjects were not randomly allocated to receive PTC), a variable called 'ever attended PTC' was added to the model to control for such differences. 'Effect of PTC' measures the changes in sexual behaviours occurring after PTC, while 'ever-attended PTC' controls for the background characteristics which differentiate those who attended and those who did not. All associations examined in univariate analysis were adjusted for time (visit number). Multivariate analysis was performed through stepwise modelling including all variables with a univariate P-value < 0.25. Statistical analysis was performed using the STATA computer package (Stata Statistical Software, Stata Corporation, College Station, Texas, USA).

Results

Between February 1997 and 31 December 1999, 1483 participants were enrolled in the two cohort studies. Of these, 1124 (75.8%) were males (418 in Akaki and 706 in Wonji), and are considered in this study. At enrolment, 874 (77.8%) participants were 30 years or older and 899 (80.1%) reported to be married. Risky sexual behaviours were common in this population, with 677 (60.2%) acknowledging more than five sexual partners in their lifetime, 115 (10.2%) reporting a history of genital discharge in the past 5 years, and 24 (2.1%) reporting a genital ulcer in the past 5 years. Although only 109 (9.7%) reported having had casual sexual partners in the past year, 488 (43.4%) acknowledged having sex with CSWs when going to bars (this question was asked without any specific time frame). In Akaki, 19 (38.8%) of the participants who had casual sex in the past year reported using a condom during last sexual act with a casual partner. The prevalence of HIV infection was 47/418 (11.2%) in Akaki and 49/706 (6.9%) in Wonji (P = 0.01), and PTC attendance was 311(74.4%) in Akaki and 564 (79.9%) in Wonji (P = 0.03).

As of 31 December 2000, 921 of the 1124 (81.9%) men were still in follow-up, whereas 203 (18.1%) had not been seen for more than 1 year and thus considered dropouts. Most of the dropouts (118 = 58%) left the study after the enrolment visit, and another 85 (28%) after the first follow-up visit. Of all dropouts, 30 (14.8%) had died, 20 (9.3%) had left the factories and 153 (75.4%) were still working there. Table 1 compares the background characteristics of the dropouts and the remaining participants. Dropouts did not differ from other participants by past sexual behaviours. However, in comparison with participants still in follow-up, dropouts were more likely to be HIV-positive (18.2 versus 6.5%, respectively, P < 0.001), less likely to have attended PTC (55.7 versus 82.7%, respectively, P < 001), more likely to report sex with CSWs (50.7 versus 40.6%, respectively, P < 0.001) and tended to be older. In multivariate analysis, only HIV status remained an independent predictor of being lost-to-follow-up (data not shown).

Table 1

Change in reported sexual behaviour and STDs by year of enrolment in Wonji

In Akaki, most (95%) participants were enrolled in 1997, whereas in Wonji, the proportions enrolled in 1997, 1998, and 1999 were 156 (22.1%), 316 (44.8%), and 234 (33.1%), respectively, allowing comparison of behavioural indicators by year of enrolment (Table 2). The proportions of Wonji men who reported risky sexual behaviours and past history of STDs at the intake visit declined between 1997 and 1999 from 10.3 to 3.9% for casual sex in the past year (P = 0.02), from 53.2 to 20.2% for sex with CSW (P < 0.001), and from 15.4 to 5.1% for history of genital discharge in the past 5 years (P < 0.01). The effect of calendar year remained statistically significant after controlling for socio-demographic characteristics in logistic regression models, suggesting that the declines were not related to changes in the type of population enrolled at different years.

Table 2

HIV incidence and changes in sexual behaviours during follow-up

The incidence HIV infection was 6/952 = 0.63 (95% CI, 0.29-1.40) and 2/1125 = 0.18 (95% CI, 0.04-0.71) per 100 person-years in Akaki and Wonji, respectively. The expected incidence under the steady-state assumption (see Materials and methods) was 1.26 (95% CI, 0.92-1.72) and 0.74 (95% CI, 0.55-1.00) per 100 person-years in Akaki and Wonji, respectively. There was no change in incidence by calendar year or by duration of follow-up. Due to the low number of seroconversions (n = 8), we could not identify risk factors associated with incident HIV infection. However, we were able to look at longitudinal trends in sexual behaviours and STDs in our two cohorts of male factory workers.

Figure 1 displays sexual behaviours and STDs by follow-up visit among study participants. The median time-period between two visits was 191 days, with little variation over time, namely from 189 days between second and third follow-up visit to 194 days between first and second. At both sites combined, between the first and the fourth follow-up visits, there was a decline in the proportion of males reporting recent casual sex (from 17.5 to 3.5%, P < 0.001), sex with CSWs (from 11.2 to 0.75%, P < 0.001), and genital discharge (from 2.1 to 0.6%, P = 0.004). The decline over time was statistically significant in GEE models for all indicators except genital ulcer, for which the decline was significant only for HIV-negative participants (Tables 3 and 4). Models including data from all participants or data from only those who remained in follow-up gave very similar results (data not shown). This was as expected, since most losses to follow-up occurred just after enrolment, and only data from first follow-up visit and thereafter were included in the longitudinal analysis. There was no significant change in reported condom use during last casual sex in Akaki, which was 57.5% at the first follow-up visit and 50% at the fourth visit.

Fig. 1

Table 3

Table 4

Factors associated with behavioural indicators are shown in Table 3 (sexual behaviours) and Table 4 (STDs). In multivariate analysis, several factors were significantly associated with lower reporting of casual sex (time, Wonji residence, being married and effect of PTC), sex with CSW (time, Akaki residence, being married, lower educational status and effect of PTC), genital discharge (time, age ≥ 40 years, and being HIV-negative), and genital ulcer (higher educational status and being HIV-negative). The association between casual sex and PTC was complex: those reporting casual sex were more likely to attend PTC during follow-up (non-significant), but the reporting of casual sex decreased after PTC attendance. The same association was described for sex with CSWs and PTC.

Discussion

This paper presents evidence of changes in risky sexual behaviours among factory workers who received health education and HIV testing and counselling in Ethiopia. These behavioural changes mainly involve reduction in both commercial and non-commercial casual sex. The decline was particularly pronounced for sex with CSWs, reported by 11.2% at the first follow-up visit and by less than 1% at the fourth follow-up visit. A decline in sex with CSWs would have a strong impact on the spread of HIV in this community. As in many other African countries, CSWs are an important core group for HIV transmission in Ethiopia, as shown by the high prevalence of HIV infection among CSWs in the early days of the epidemic (17.0% among 6234 CSWs sampled nation-wide in 1988) [¹⁶] and the 73.4% prevalence documented in 1998 among CSWs of Addis Ababa [¹⁷].

Part of the decline in risky behaviours can be attributed to the prevention interventions (health education and HIV testing and counselling) provided to the workers through their participation in our study. Indeed, the decline in casual sex for an individual was significantly associated with his PTC attendance and the number of follow-up visits. Males with risky sexual behaviours were more likely to attend PTC compared to those without; however, casual sex and sex with CSWs decreased after attending PTC, suggesting a positive effect of PTC on participants' behaviours. This result is in line with a recent report on the effect of HIV voluntary testing and counselling (VTC) in Kenya, Tanzania, and Trinidad, where individuals receiving VTC were less likely to report unprotected intercourse with non-primary partners than were individuals receiving only health information [¹⁸]. However, we suspect that the decline in casual sex may not be related only to cohort participation but also to a more general reduction in risky sexual behaviours in the surrounding community. Indeed, the proportion of factory workers who reported risky sexual behaviours at enrolment declined by calendar year in Wonji. However, the HIV and TPHA prevalence at intake, which reflect lifetime exposure to sexually transmitted diseases, were not yet affected by these recent changes in sexual behaviours. Information about prevention provided to our cohort may well have had repercussions on other workers among their contacts who modified their behaviours as a result. More general prevention programmes provided nation-wide may also have influenced the Wonji community in recent years. Programmes focusing on risk-behaviour reduction, condom promotion, and STD control have been underway since 1987 in Ethiopia and, although hampered by the political instability of the late 1980s and early 1990s, they have raised the general awareness of HIV/AIDS. During the 1993 general population survey on prevention indicators carried out by the World Health Organization in four urban areas of Ethiopia, 94.1% of 2712 males and 83.8% of 4173 females aged 15-48 years demonstrated a good knowledge of HIV preventive practices [¹⁹]. Similar findings were reported among Akaki factory workers at the intake of the cohort study [¹³]. Such knowledge may have translated into behavioural changes in a population confronted by the rising number of AIDS cases in the past decade.

In the absence of a comparison group, it is unknown whether the cohort interventions resulted in a decline in HIV incidence. In comparison with the expected incidence rates under the steady-state assumption, the observed HIV incidence rates were lower but not significantly, and there was no trend of declining HIV incidence with longer duration of follow-up. It could well be that HIV transmission continued unabated between regular partners of discordant serological status, whereas the incidence of STDs with shorter transmission periods dramatically decreased in relation to the decline in casual sex. In favour of that scenario is the zero incidence of syphilis observed among factory workers in Akaki between February 1997 and March 1999 (0 per 100 person-years, 97.5% one-sided CI, 0-0.5) [²⁰] and the decline in genital discharge among male workers at both sites, from 2% at the first follow-up visit to 0.6% at the fourth follow-up visit (P < 0.01). Recent genital ulcers also declined, but only among HIV-negative participants.

The findings of this study are pertinent largely to the group that was studied, who were mainly middle-aged and married males working in Ethiopian factories. Whether they would apply to young, unmarried males, who are particularly vulnerable to HIV infection, is not known. Another study limitation is that health education and HIV testing and counselling was offered to participants and linked to benefits such as free family medical care, perhaps pre-selecting participants whose motivation to get tested and counselled might differ from that seen in ordinary VTC centres. Moreover, since a substantial proportion of the study participants were lost to follow-up (close to 20%) our conclusions apply mainly to those who remained in the cohort. However, HIV-positive individuals, particularly those who learned of their HIV status through PTC, were over-represented among the dropouts; therefore we would like to know how the knowledge of their HIV status may have influenced their subsequent sexual behaviours.

Finally, one may question the validity of the behavioural indicators used in this study, particularly if self-reported [²¹]. In males, genital discharges and genital ulcers are quite specific of STDs, and can be easily recognized by the patients themselves, so that they are effective entry points for the syndromic approach of STDs treatment [²²]. However, genital ulcers caused by herpes simplex infections may reflect recurrences of past infections rather than new infections, and thus may not be a valid marker of behavioural changes. This is particularly true among HIV-positive individuals who are more subject to recurrences because of immunosuppression [²³], explaining why no decline was observed among HIV-positive individuals in the reporting of genital ulcers in this study. Despite an assurance of privacy and confidentiality during the interview, some cohort participants will misreport sexual behaviour. Some participants may indeed prefer the socially acceptable answers when replying to the questionnaire, particularly those who have attended PTC and extensively discussed the issues of safe sex with the counsellor. However, behavioural trends are of greater interest to the present study than the exact levels of risk behaviour. Even when there is misreporting, repeated behavioural surveys will show changes in trends over time, provided that the magnitude or direction of misreporting does not change significantly [²⁴].

In conclusion, this study provides substantial evidence that risky sexual behaviours and the incidence of STDs have decreased in this cohort of factory workers provided with health education and HIV testing and counselling. This programme offers an encouraging view of the interventions that can be offered to developing countries to minimize the spread of HIV infection.

Acknowledgements

This study was carried out by ENARP. ENARP is a collaborative effort of the EHNRI, Addis Ababa, the Municipal Health Service, Amsterdam, the Department of Human Retrovirology of the Academic Medical Centre (University of Amsterdam) and the Central Laboratory of the Netherlands Red Cross Blood Transfusion Service. An earlier version of the manuscript was edited by Lucy D. Phillips.

Sponsorship: ENARP is financially supported by the Dutch Ministry for Development Co-operation, the Ethiopian Ministry of Health, and by a grant from the World Health Organization.