The assessment of the spread of HIV in sub-Saharan Africa has largely been based on data from women attending antenatal clinics and a limited number of risk groups such as commercial sex workers . More detailed data on trends in HIV prevalence and incidence are available for only a few local populations. The longest running cohort study in Africa, which is located in the Masaka district of rural southwest Uganda, showed a gradual decline in the incidence and prevalence of HIV during the 1990s . In Zambia, two cross-sectional surveys in selected urban and rural populations suggested that the prevalence of HIV declined during 1996–1999 among young women, but not men .
Several studies have considered trends in the prevalence and incidence of HIV in conjunction with trends in AIDS-related knowledge, attitudes and sexual behaviour. Such analyses have been performed using data from antenatal clinics and local population-based surveys [4–6], or from specific populations such as factory workers or sex workers [7,8], and from multi-round or longitudinal general population studies . Also at the national level, attempts have been made to link trends observed in HIV surveillance to behavioural trends obtained from general population or risk group surveys [4,9,10].
An important aspect of trend assessment is the evolution of differentials in the spread of HIV over time. Among the key findings of descriptive studies of the spread of HIV are differences by place of residence. There is substantial evidence of a higher prevalence of HIV in large urban areas compared with rural areas, although there is considerable variation between countries in the magnitude of this difference . Large differences in the prevalence and incidence of HIV have also been observed between smaller urban areas, roadside trading centres, and agricultural villages [12–14]. This paper focuses on trends in HIV transmission and sexual behaviour in an open cohort study in a rural population in Tanzania during 1994–2000.
Background and methods
Survey and laboratory methods
The study was carried out in Kisesa ward in Mwanza Region, Tanzania. A ward is an administrative entity that falls under a division of a district, and can be divided into smaller administrative units, such as village, sub-village and ‘balozi’ (10 household units). Kisesa ward has a population of 20 000 and lies approximately 20 km east of the regional capital Mwanza, along the main road to Kenya. It includes six villages and a trading centre along the main road; these locations have been grouped into trading centre, peri-trading centre and agricultural rural villages for the purpose of this study. Details of the study have been described elsewhere .
A demographic surveillance system was established in 1994, and 13 follow-up rounds of brief household visits were completed during 1994–2000 to estimate the extent of demographic change and to establish the size and composition of the population eligible for survey participation. Eligibility was based on residence status, which was defined as living in the study area at the last demographic round. Detailed surveys of all adults were carried out during 1994–1995, 1996–1997 and 1999–2000, covering the age ranges 15–44, 15–46 and 15 years and older, respectively. All eligible individuals were asked to come to a central point in the village to be interviewed using a structured questionnaire and to give a blood sample for HIV testing. If the individual did not show up, the survey team asked the local village leaders to encourage survey participation further. The purpose of the study was explained and verbal consent was obtained before the interview. The questionnaire was administered in the Swahili language (or in the local vernacular if necessary) and interviews were held in temporarily constructed huts to maximize privacy and ascertain confidentiality. During the first survey whole blood was collected by venapuncture, and during the second and third surveys all blood samples were collected on filter paper.
All study participants were offered free medical treatment for health problems present at the time of the survey, whereas the project also supported a number of community oriented activities during the study period, such as the upgrading of the health facilities, the sponsoring of a secondary school and sports tournaments. During the second and third surveys HIV counselling and testing were offered by a qualified counsellor, who followed up with those wanting to know their HIV test results. The clinicians who provided care during the survey were blind to the HIV status. The study was approved by the national AIDS research scientific and ethical review committee of the Ministry of Health of Tanzania.
HIV testing was performed in a regional reference laboratory in Mwanza. The testing algorithm was based on two independent enzyme-linked immunosorbent assays; Vironostika HIV-MIXT (Organon, Boxtel, the Netherlands) and Enzygnost HIV1/HIV2 (Behring, Marburg, Germany). Only samples with two positive enzyme-linked immunosorbent assay tests were considered HIV positive.
Data were double entered using DBaseIV (Borland International, Scotts Valley, CA, USA) and all analyses were performed using Stata 7.0 (Stata Corporation, College Station, TX, USA). Individuals were matched on the basis of a residential identification number if they had attended more than one survey. The linking of different identification numbers for individuals who had moved residence within the Kisesa ward between survey rounds was only partly successful, as the previous residence identification number was sometimes difficult to trace. Matched cases were checked for consistency by comparing records of age, sex, height, and a number of socioeconomic characteristics between the survey rounds. If these checks showed incompatibilities, the individual was considered a new entry into the study population. Minor discrepancies in age were edited, if necessary, using additional information gathered during the demographic rounds. To compare the prevalence of HIV between the three surveys, the results were standardized using the age distribution of the second survey round. The overall incidence of HIV infection per 100 person-years (PY) of observation was estimated from the number of seroconversions and the total PY of follow-up. The date of seroconversion was assumed to be the mid-point between the two survey dates, except for those who seroconverted between the first and third survey, but had not attended the second survey. Two-fifths of these seroconversions were assumed to have occurred in the 2-year period between the first two surveys, and three-fifths were assumed to have occurred in the 3-year period between the last two surveys. The incidence risk was assumed to be constant over each inter-survey interval and independent between individuals, so that Poisson models could be used for calculating confidence intervals and for testing differences between incidence rates. Even though the study design was a full enumeration of the whole population of the area, statistical testing was considered appropriate, as coverage was not complete. The analyses of the incidence trend were repeated using only respondents who were present both at the beginning and at the end of each inter-survey period, but the results showed little difference with the above approach.
Overall, the number of men and women attending was 5783 in 1994–1995 (round 1), 6392 in 1996–1997 (round 2) and 7438 in 1999–2000 (round 3), when all those individuals aged 15 years and over were eligible for participation. Among individuals aged 15–44 years attendance rates were 78, 80 and 71% in the first, second and third rounds, respectively (Table 1). In all rounds attendance was better among women than men and better in the rural residents than in the trading centre. Young people were more likely to participate in the first two rounds but not in the third round. The drop in attendance in the third round was very pronounced in the trading centre and in the village surrounding the trading centre, but almost no drop occurred in rural attendance.
Re-attendance in the subsequent round was 69 and 52% for the first and second round participants aged 15–44 years, respectively. Among HIV-negative individuals re-attendance was 70 and 53% in the second and third rounds, respectively. The corresponding figures for HIV-positive individuals were 45 and 31%. The main reasons for non-participation in the third survey round among those who attended the second round was migration and short-term absence. Among HIV-negative individuals, 63% of non-attendance was associated with having left the study area, 22% were on short-term travel, and 1.8% had died. Among non-attending HIV-positive individuals, 49% had left the study area, 24% were on short-term travel, and 18% had died since the previous round.
Table 2 presents HIV prevalence by 5-year age group for men and women. HIV prevalence was slightly higher in the 1996–1997 survey than in the first round. In all survey rounds women had a significantly higher prevalence of HIV than men (1.4–1.6 times higher in each round), and reached peak levels at an earlier age than men. The prevalence of HIV increased gradually over time, and for both men and women HIV prevalence levels in the last survey were approximately 1.4 times higher than 5 years earlier.
Fig. 1 presents the age-standardized HIV prevalence trends in three geographical strata of the Kisesa ward. The HIV prevalence estimates for the residents of the rural villages are most reliable because attendance was good and fairly constant over time, whereas the prevalence trends in the trading centre need to be interpreted with most caution because of declining attendance over time. Among rural women, the prevalence of HIV increased gradually during the second half of the 1990s. The twofold gap in HIV prevalence between rural women and women living in the trading centre diminished slightly, but only during the last round. Among men, rural prevalence increased from just under 4% in the first two rounds to 6.7% in the third round, whereas prevalence in the trading centre dropped.
Table 3 shows the annual HIV incidence rates by sex, place of residence and period between the surveys. The incidence of HIV was significantly higher during the second half of the study period for both men and women. The rural incidence rates show the same pattern; almost a doubling among men and an increase by one-third among women. Incidence in and near the trading centre was also considerably higher in the later period, and was of the order of 2% for men and women, but the trends in the separate areas of residence do not attain statistical significance at the 5% level.
Knowledge, attitude and behaviour
In all three surveys, nearly all respondents had heard of HIV/AIDS and knew about the sexual transmission of HIV. Table 4 shows the gradual increase in the proportion of respondents who knew that a healthy person could have HIV. However, respondents were also asked how long an individual could have HIV before falling ill, and the majority of individuals continued to be unaware that this phase generally lasts more than 2 years.
Among men there was an increase in the proportion who knew a person who had died of AIDS or who was living with HIV/AIDS, but no such trend was observed among women. Only a quarter of men and women reported feeling somewhat at risk of contracting HIV, which stands in sharp contrast to the sexual behaviour data. The median age at first sex, derived from current status data, remained approximately 17 years for boys and girls. There was no change in the reported number of sexual partners in the past year among sexually active men and women, and more than half of the men had more than one partner. Lifetime condom use among men increased to 24%, but consistent use in high-risk partnerships was low. Condom use questions addressed to women were changed in the third survey and no trend can be derived.
Three large surveys of adults aged 15–44 years in a rural community cohort study in northwest Tanzania show that the incidence of HIV increased from 0.8 to 1.3 per 100 PY during 1994–1997 and 1997–2000, respectively. The prevalence of HIV increased gradually from 5.9% in 1994–1995 to 6.6% in 1996–1997 and 8.1% in 1999–2000. In spite of a modest increase in knowledge during the study period, most individuals continued to feel that they were not at risk of HIV, and sexual risk behaviour remained largely unchanged, except for a small increase in condom use. HIV transmission levels continued to be higher in the trading centre than in the nearby rural villages within this small geographical area, although differences became smaller over time. The incidence of HIV in the trading centre was 2.2 per 100 PY, significantly higher than in the rural and intermediate villages (1.0 and 1.7, respectively), but the ratio of prevalence in the trading centre to prevalence in rural villages fell from 3 to 2 between 1994 and 2000.
Before going into a substantive discussion of these results, a number of biases need to be considered. In all three rounds non-response was higher in trading centres than in rural areas, and trading centre residents were also more mobile, which may have led to an underestimation of the rates of HIV infection in the Kisesa population in general and especially in the trading centre population. On the other hand, the free treatment of common conditions with the support of a small field clinic and laboratory during the surveys may have attracted more sick people, and could have led to an overestimation of the prevalence and incidence of HIV in the sero-surveys, as sick individuals are more likely to be HIV infected. HIV incidence rates have to be interpreted with caution, because the overall follow-up rate of HIV-negative individuals in round 3 was only 53%. Attendance rates in the rural villages remained over 80% in all three rounds, and these estimates of trends in the prevalence of HIV and sexual behaviour patterns are least likely to suffer from changes in participation bias over time. However, the estimates of trends in the trading centre need to be treated cautiously because non-participation rates were high and there is no way to adjust for the potential bias. The differences in participation rates between the trading centre and rural area suggest that the overall trends are biased downwards, and that the rate ratio of HIV prevalence in trading villages to rural villages may be underestimated.
HIV testing algorithm and quality control procedures were similar during all surveys. The main difference was that during the first survey whole blood was collected, whereas the second and third surveys relied on blood-spotted filter paper, but this is unlikely to have affected the comparability of HIV test results. In case of a change in HIV status an extensive comparison of individual demographic, socioeconomic and anthropometric (height) characteristics was conducted to make sure that no administrative mistakes had been made.
Several studies have recorded limitations in the validity of self-reported survey data on sexual behaviour [15–17]. Desirability bias in particular is thought to affect assessments of trends in behaviour over time. The three surveys produced similar results, despite the presence of AIDS awareness campaigns aimed at limiting multiple partnerships and promoting condom use, suggesting that in this case desirability bias was not a problem (only if it masked a true increase in risk behaviour, but that is unlikely). However, this is no proof of validity. A detailed analysis of partnership data from the same population showed that the sexual behaviour data are of relatively good quality in terms of internal consistency .
The lack of change in sexual behaviour is striking, but is consistent with trends observed in Tanzania as a whole . During the study period the Kisesa population was exposed to the district HIV prevention programme. The district programme, which was coordinated by a multi-sectoral AIDS action committee, included community mapping of high-risk places, the establishment of village AIDS committees, community campaigns against AIDS, the formulation of village bylaws to reduce high-risk sexual behaviour, and school interventions. During 1996–1997 the TANESA project helped the district implement these community interventions to promote safer sexual behaviour in the villages in Kisesa ward. There were initially three (and from 1996, four) government dispensaries and one private health facility in Kisesa ward. Improved sexually transmitted disease services, using the syndromic approach, regular supervision and improved drug supplies for sexually transmitted disease treatment, became available in one clinic in late 1994, and in two other clinics in mid-1996.
The combination of these interventions has been insufficient to turn the tide of the AIDS epidemic in Kisesa. The low-cost intervention package does not appear to be adequate to stem the growth of the epidemic, and more intensive AIDS control efforts are needed. Risk perception levels are still low, and stigma associated with AIDS is still very common. For example, very few individuals wanted to know their HIV status, and very few families were willing to attribute a family death openly to AIDS . In this population as many as 40% of adult deaths are associated with HIV/AIDS, but less than half of the respondents said that they knew someone who had died of AIDS or was living with HIV/AIDS. It is possible that the interventions prevented a more rapid spread of HIV in this population, but the lack of changes in self-reported behaviours suggests that their impact has been limited. Clearly, under-funded district and community programmes may only have a limited impact on HIV transmission, and there is an urgent need to intensify the interventions in the next few years.
The study also shows the importance of long-running observational cohort studies. The gradual increase in the prevalence of HIV, the steady spread into the rural population from the more urban settlements, the unexpected increase in the incidence of HIV in the second half of the study, and the striking lack of change in sexual behaviour present important information for AIDS programmes. Data derived from surveillance among pregnant women have shown that the spread of HIV can take a multitude of courses, from extremely rapid increases, such as those observed in several southern African countries, to dramatic declines, as shown in Uganda and the neighbouring Kagera region in Tanzania . Longitudinal studies such as those in Masaka district in Uganda and this Tanzania cohort study often show much slower and more protracted declines or increases over time. Such data are necessary to complement the attempt to track changes in the prevalence of HIV among surveillance groups and to monitor behaviour in risk groups or in the general population.
Sponsorship: The TANESA project and this study were funded by the Minister of Development Cooperation of the Netherlands.
1. Walker N, Garcia-Calleja JM, Heaton L, Asamoah-Odei E, Poumerol G, Lazzari S, et al
. Epidemiological analysis of the quality of HIV sero-surveillance in the world: how well do we track the epidemic?AIDS
2. Mbulaiteye SM, Mahe C, Whitworth JA, Ruberantwari A, Nakiyingi JS, Ojwiya A, et al
. Declining HIV-1 incidence and associated prevalence over 10 years in a rural population in south-west Uganda: a cohort study.Lancet
3. Fylkesnes K, Musonda RM, Sichone M, Ndhlovu Z, Tembo F, Moze M. Declining HIV prevalence and risk behaviours in Zambia: evidence from surveillance and population-based surveys.AIDS
4. Asiimwe-Okiror G, Opio AA, Musinguzi J, Madraa E, Tembo G, Carael M. Change in sexual behaviour and decline in HIV infection among young pregnant women in urban Uganda.AIDS
5. Okware S, Opio A, Musinguzi J, Waibale P. Fighting HIV/AIDS: is success possible?Bull WHO
6. Kilian AH, Gregson S, Ndyanabangi B, Walusaga K, Kipp W, Sahlmuller G, et al
. Reductions in risk behaviour provide the most consistent explanation for declining HIV-1 prevalence in Uganda.AIDS
7. Ng'weshemi JZ, Boerma JT, Pool R, Barongo L, Senkoro K, Maswe M, et al
. Changes in male sexual behaviour in response to the AIDS epidemic: evidence from a cohort study in urban Tanzania.AIDS
8. Ghys PD, Diallo MO, Ettiegne-Traore V, Kale K, Tawil O, Carael M, et al
. Increase in condom use and decline in HIV and sexually transmitted diseases among female sex workers in Abidjan, Cote d'Ivoire, 1991–1998.AIDS
9. Meda N, Ndoye I, M'Boup S, Wade A, Ndiaye S, Niang C, et al
. Low and stable HIV infection rates in Senegal: natural course of the epidemic or evidence for success of prevention?AIDS
10. National AIDS Control Program, Tanzania
, and MEASURE. AIDS in Africa during the nineties: Tanzania.
Chapel Hill, USA: Carolina Population Center, University of North Carolina; 2001.
11. Carael M. Urban–rural differentials in HIV/STDs and sexual behaviour.
In: Herdt G, editor. Sexual cultures and migration in the era of AIDS: anthropological and demographic perspectives
. Oxford: Oxford University Press; 1997. pp. 107–126.
12. Boerma JT, Urassa M, Senkoro K, Klokke A, Ngweshemi JZ. Spread of HIV infection in a rural area of Tanzania.AIDS
13. Wawer M, Serwadda D, Musgrave SD, Konde-Lule JK, Musagara M, Sewankambo NK. Dynamics of the spread of HIV-1 infection in a rural district of Uganda.BMJ
14. Barongo LR, Borgdorff MW, Mosha F, Nicoll A, Grosskurth H, Senkoro KP, et al
. The epidemiology of HIV-1 infection in urban areas, roadside settlements and rural villages in Mwanza Regon, Tanzania.AIDS
15. Buve A, Lagarde E, Carael M, Rutenberg N, Ferry B, Glynn JR, et al
. Interpreting sexual behavior data: validity issues in the multicentre study on factors determining the differential spread of HIV in four African towns.AIDS
2001, 15 (Suppl. 4)
16. Dare OO, Cleland JG. Reliability and validity of survey data on sexual behaviour.Health Trans Rev
1997, 4 (Suppl.)
17. Gregson S, Zhuwau T, Ndlovu J, Nyamukapa CA. Methods to reduce social desirability bias in sex surveys in low-development settings: experience in Zimbabwe.Sex Transm Dis
18. Nnko S, Boerma JT, Urassa M, Mwaluko G, Zaba B. Secretive females or swaggering males: an assessment of the quality of sexual partnership reporting in rural Tanzania.Soc Sci Med
2003, in press.
19. Urassa M, Boerma JT, Isingo R, Ngalula J, Ng'weshemi JZL, Mwaluko G, Zaba B. The impact of HIV/AIDS on mortality and household mobility in rural Tanzania.AIDS
20. Kwesigabo G, Killewo J, Urassa W, Mbena E, Mhalu F, Lugalla JL, et al
. Monitoring of HIV-1 infection prevalence and trends in the general population using pregnant women as a sentinel population: 9 years experience from the Kagera region of Tanzania.J Acquir Immune Defic Syndr