This study aimed at assessing the extent to which decline in HIV infection prevalence reflects decline in incidence in 3 areas with contrasting initial exposure to the HIV epidemic in the Kagera region of Tanzania. A population sample was recruited for the baseline study in 1987 through a multistage cluster sampling technique to determine HIV prevalence. Seronegative individuals identified in the baseline and subsequent studies were followed up for 3 years to yield trends in incidence that were compared for the 3 areas. The overall age-adjusted HIV-1 prevalence in the high-prevalence area of Bukoba urban declined significantly from 24.2% in 1987 to 18.2% in 1993 and later to 13.3% in 1996 (P = 0.0001). In the medium-prevalence area of Muleba, overall age-adjusted prevalence declined significantly from 10.0% in 1987 to 6.8% in 1996 and later to 4.3% in 1999 (P = 0.0003), whereas in the low-prevalence area of Karagwe the prevalence declined from 4.5% in 1987 to 2.6% in 1999 (P = 0.01). In all 3 areas, the most significant decline was consistently observed among women in the age group 15-24 years. No age group exhibited a significant upward prevalence trend. The HIV-1 incidence for Bukoba urban declined from 47.5 to 9.1 per 1000 person-years of observation in 1989 and 1996, respectively, whereas in Muleba it decreased from 8.2 to 3.9 in 1989 and 2000, respectively. Sex-specific estimates indicated a significant decline among women in the high-prevalence area of Bukoba urban from 51.5 to 9.2 per 1000 person-years at risk (P = 0.001). It is concluded that the HIV-1 epidemic in Kagera is on the decrease as reflected by the decline in HIV-1 incidence and prevalence trends particularly among the 15-24 year olds. The decline in the 3 areas of differing magnitude implies that the HIV/AIDS epidemic may be arrested early without necessarily peaking to saturation levels.
From the *Department of Epidemiology and Biostatistics and †Department of Microbiology/Immunology, Muhimbili University College of Health Sciences, Dar es Salaam, Tanzania; ‡Department of Anthropology, University of New Hampshire, Durham, NH; §Epidemiology and Public Health Science, Department of Public Health and Clinical Medicine, Umea University, Umea, Sweden; ∥Department of Kiswahili, University of Dar es Salaam, Tanzania; and ¶Swedish Institute for Infectious Disease Control, Karolinska Institute, Stockholm, Sweden.
Received for publication June 24, 2004; accepted April 20, 2005.
This study was funded by the Department for Research Cooperation (SAREC) at the Swedish International Development Cooperation Agency (SIDA) based in Stockholm, Sweden.
Reprints: Gideon Kwesigabo, Department of Epidemiology and Biostatistics, Muhimbili University, College of Health Sciences, PO Box 65015, Dar es Salaam, Tanzania (e-mail: firstname.lastname@example.org).
HIV-1 infection prevalence levels and trends in sub-Saharan Africa have varied in different periods. Some have been reported as stable or starting to decrease1 while others have been reported as increasing.2-5 Despite these observations, however, population-based studies in certain areas have indicated a downward trend, particularly among individuals aged 15-24 years.6-8 Though some have used prevalence data to monitor the epidemic, skepticism has been expressed about the use of such data as they may not reflect corresponding changes in incidence. In particular, earlier studies in Uganda have indicated a downward prevalence trend that did not reflect corresponding trends in incidence.9,10 Studies with more statistical power conducted later in the same country have shown significant declines in both prevalence and incidence.11,12 Such downward trends, conversely, have been observed in areas that had originally experienced high prevalence or incidence rates that were associated with so-called maturation states of the epidemic.9 In areas or situations where prevalence decline does not reflect a corresponding decline in incidence, several factors could be responsible. These include excess mortality among HIV-positive individuals, nonrandom migration, temporary absenteeism of the infected, and differential or selective refusal to participate in the intended studies. In areas where these explanatory factors can be ruled out, behavioral change in sex relationships has been singled out as being the most consistent explanation for the observed decline in prevalence.10,13,14
The Kagera AIDS Research Project was initiated in the Northwestern part of Tanzania in 1987 to study the epidemiology, spread, and dynamics of HIV infection transmission in the study area to identify appropriate entry points for HIV control.15 The study base established since then has facilitated monitoring of the epidemic and has provided information for use in evaluation of interventions. Findings from a previous study on HIV prevalence trends in the area have indicated a downward trend in the prevalence of HIV-1 infection in the Bukoba urban area and the immediate surrounding districts of Bukoba rural and Muleba, particularly among women in the age group 15-24 years.8,16 However, no assessment has been made on whether this trend in prevalence decline has reflected a corresponding decline in incidence trends in the same areas and what factors are likely to be responsible for these changes. Such assessment would be helpful in determining the true direction of the epidemic and the possible role of interventions.
This study was therefore conducted with the main objective of assessing whether the observed HIV prevalence trends in these areas with contrasting initial exposure to the AIDS epidemic reflect incidence trends in the same areas.
Study Area and Study Population
Kagera region is situated on the western side of Lake Victoria in Tanzania with its 6 administrative districts of Bukoba urban, Bukoba rural, Muleba, Karagwe, Ngara, and Biharamulo. The region has been the focus of HIV research and interventions since the first AIDS cases in Tanzania were reported in 1983. The interventions have included HIV prevention education, promotion of condom use, and support activities for affected individuals. According to projections from the 1988 population census,17 the total population of the region is about 1.8 million, of whom 63,000 are resident in Bukoba urban. Districts in the region are further divided into divisions, wards, and villages or streets. Villages are mainly rural and are further subdivided into “vitongoji,” which are the smallest administrative units in the country having households varying between 25 and 200. Urban streets, the equivalent of rural vitongoji, may have a similar number of households. The majority of the people in the region are Christians, followed in a decreasing order by Muslims and traditional religions, whereas Buddhists and Hindus are the minority.
Selection of Study Areas
To obtain a representative sample for the 1987 region-wide cross-sectional study, the region was divided into 4 zones on the basis of presumed contrasting exposure to the HIV epidemic as indicated by the differing number of reported AIDS cases from the region. Findings from the 1987 cross-sectional study showed that HIV infection prevalence varied between the study zones as had earlier been predicted, being highest in the urban area and lowest in the rural areas.15 The prevalence was 24.2% in Bukoba urban, 10.0% in Bukoba rural and Muleba, 4.5% in Karagwe, and 0.6% in Ngara and Biharamulo. For the purpose of monitoring HIV infection trends, 3 contrasting exposure areas were identified, with Bukoba urban representing a high HIV prevalence area, Muleba representing medium prevalence, and Karagwe, Ngara, and Biharamulo representing low HIV prevalence areas.
The Cross-Sectional Studies
Following the region-wide baseline cross-sectional study in 1987, subsequent cross-sectional studies were done in Bukoba urban district in 1993 and 1996, in Bukoba rural and Muleba districts in 1996 and 1999, and in Karagwe district in 1999. All studies involved the collection of blood specimens from newly selected probability population samples. Details of implementation of the 1987, 1993, and 1996 cross-sectional studies have been reported earlier.8,16,18 The cross-sectional studies conducted in Karagwe and Bukoba rural/Muleba districts in 1999 used the same multistage cluster sampling methods as used in the earlier studies. This means that 3 divisions were randomly selected from each district and subsequently 2 wards were randomly selected from the divisions. For each selected ward a number of clusters (vitongoji in rural settings and streets in urban settings) were selected with probability proportional to ward size. Although the number of households in the vitongoji ranged from 25 to 200, all the households in the selected vitongoji were included in the study to avoid possible stigmatization that might affect those selected for the study. However, only one adult aged 15-54 years was randomly selected from each household for the study.
The Incidence Studies
HIV-seronegative individuals who were identified during the cross-sectional studies formed the population at risk for subsequent follow-up studies designed to estimate the incidence of HIV infection in the closed cohorts. Consequently, the study population recruited in 1987 was followed up in 1989 to obtain baseline incidence estimates in the high-prevalence area of Bukoba urban and the medium-prevalence area of Bukoba rural and Muleba districts. The study population recruited from Bukoba urban in 1993 was followed up to 1996 and that recruited in Muleba in 1996 was followed up in 2000 to obtain incidence estimates that would be compared with the baseline incidence to assess trends. Trained interviewers-cum-counselors working on the project in previous years were reoriented to the study methods and trained to administer the research instruments. Using addresses recorded during the previous cross-sectional studies, and having obtained permission from the respective public administrative levels, house-to-house visits were made to identify individuals who had participated in those studies. Cards bearing the respondents' names and study numbers issued during the cross-sectional studies and kept by the respondents themselves were used to facilitate recognition of study participants during the follow-up. In addition, efforts were made to ensure that the same research assistants used in the earlier studies were also used to locate and interview the same study individuals during the incidence study.
Following a verbal consent, individuals were interviewed about sociodemographic characteristics and HIV infection risk using a structured questionnaire. A blood specimen was collected from each individual after pretest counseling and was later tested for HIV antibodies. Seroconversion among seronegative persons was assumed to have occurred halfway between the dates of the first test and the last test results in the respective follow-up period, and the resulting interval was taken to be the follow-up time. For those who did not seroconvert, the full interval between the 2 test results was taken to be the follow-up time. Confidentiality was strictly observed at all stages of the study. To minimize the risk of HIV-positive individuals being stigmatized in the community, efforts were made to revisit all individuals studied during the cross-sectional studies regardless of their HIV status. Members of the studied population who wished to have their HIV test results disclosed to them were given posttest counseling before disclosure by the same interviewer. These follow-up studies were performed following ethical approval from the research committee of the National AIDS Control Program, Ministry of Health, Tanzania, and the ethical committee of Umea University in Sweden.
Details of laboratory procedures used in the follow-up studies were the same as those used in the cross-sectional studies and have been reported elsewhere.8 Accordingly, HIV testing was supervised and validated by microbiologists/immunologists from the Swedish Institute for Infectious Disease Control (World Health Organization Reference Laboratory) or from the Muhimbili University College of Health Sciences, which houses the National HIV Reference Laboratory for Tanzania. During the process of conducting these studies, internal and external quality assurance was strictly adhered to.
Data Management and Analysis
Using Epi Info (Centers for Disease Control and Prevention, Atlanta, 1999) and Quest and Cohort (Umea University, Umea) computer software packages, data were entered, cleaned, and analyzed. Age-adjusted prevalence estimates were calculated using the regional population census figures as standard. Differences between proportions and statistical significance were assessed using the χ2 test, and the Fisher exact test was used whenever the cell expected values were <5. Trends over time were assessed using the χ2 test for trend. Incidence estimates were calculated using person-years of observation as denominator. When the exact day and month of birth were missing, the midyear date of July 1 was used as an approximation for age calculation. The 95% CIs around the incidence estimates were calculated using the Poisson method.
In the high-prevalence area of Bukoba urban district, totals of 553, 653, and 1276 individuals were studied during the cross-sectional studies in 1987, 1993, and 1996 respectively (Table 1). The response rates (individuals participating fully and giving a blood sample) of the study populations were 86%, 66%, and 74%, respectively. The age-adjusted prevalence for the high-prevalence area declined from 24.2% in 1987 to 18.2% in 1993 and to 13.3% in 1996 (P < 0.001). The decline was more pronounced among women (29.1% in 1987, 18.7% in 1993, and 14.9% in 1996; P < 0.001) than among men (17.1% in 1987, 17.1% in 1993, and 10.3% in 1996; P = 0.01). For both sexes, the decline was most pronounced in the age group 15-24 years.
In the medium-prevalence area of Bukoba rural and Muleba districts, totals of 538, 1728, and 1639 individuals were studied during 1987, 1996, and 1999, respectively, and the corresponding response rates were 90%, 70%, and 95%, respectively. The age-adjusted prevalence declined from 10.0% in 1987 to 6.8% in 1996 and to 4.3% in 1999 (P < 0.001) (Table 2). As in the high-prevalence area, the decline in the medium-prevalence area was more pronounced among women (10.1% in 1987, 7.6% in 1996, and 3.7% in 1999; P < 0.001) than among men (9.9% in 1987, 5.7% in 1996, and 5.5% in 1999; P = 0.05).
In the low-prevalence area of Karagwe district, 671 and 1381 individuals were recruited in 1987 and 1999 with response rates of 98% and 93%, respectively (Table 3). The age-adjusted prevalence declined from 4.5% in 1987 to 2.6% in 1999 (P = 0.01). This downward trend was statistically significant among women (5.3% to 2.6%; P = 0.02) but not among men (3.3% to 2.7%; P = 0.6). Apart from women aged 15-24 years showing a pronounced decline in prevalence in all 3 areas, (Fig. 1) a significant decline was also seen across the 3 areas among Christians and peasants. Individuals who had attained educational levels of primary class 5 and above, as well as those married monogamously, also showed a significant decline. However, this observation was limited to the high- and medium-prevalence areas (Tables 1, 2, and 3).
A total of 535 HIV-seronegative individuals were eligible for follow-up from 1993 to 1996 in the high-prevalence area of Bukoba urban district. Of these, 238 individuals were fully followed up until 1996. Individuals eligible for follow-up in the medium-prevalence area of Muleba district in 1996 numbered 1610, and of these, 1185 individuals were fully followed up until 2000. Reasons for nonparticipation varied between the urban and rural settings. In the urban setting, those responding fully were 45.5%, whereas 9.4% answered the questions but did not give a blood sample, 3.8% refused to participate, 5.7% could not be traced, 8.3% were presumed dead, and 27.4% had migrated. For the rural area, 74.1% responded fully, whereas 16.6% answered the questionnaire but did not give a blood sample, 0.9% refused to answer, and 8.4% could not be traced. Table 4 shows sex-specific incidence of HIV-1 infection in Bukoba urban and Bukoba rural/Muleba districts, for the respective follow-up periods during 1987-2000. In Bukoba urban district, the overall incidence declined significantly from 47.5/1000 person-years at risk in 1989 to 9.1/1000 person-years at risk in 1996. Compared with men, women had a significant decline (P = 0.001). For the medium-prevalence area, Muleba district, the overall incidence declined from 8.2 per 1000 person-years at risk in 1989 to 3.9 per 1000 person-years at risk in 2000.
Population-based cross-sectional and follow-up studies conducted in Kagera over a period of 13 years involving probability samples of the general population have indicated a downward trend in the overall HIV-1 prevalence, which was also reflected by a downward trend in incidence. This downward trend was noted in all of the 3 areas defined as differentially exposed to the AIDS epidemic in 1987. Analysis of the age-specific prevalence trends showed that the decline was significant among young women aged 15-24 years in all 3 areas, as well as men of the same age in the high-prevalence area.
We have used repeat cross-sectional studies to monitor HIV-1 prevalence in this area of Kagera. Despite the importance of such studies in assessing the direction of trends, we have not accounted for the possible effects of excess mortality due to AIDS or migration because our study did not collect such information. Due to the high cost of conducting population-based cross-sectional and follow-up studies, it was not possible to recruit large enough sample sizes to allow for multiple stratification. However, as time passed additional resources became available, making it possible to increase sample sizes and follow-up periods to compensate for the decreasing seroconversions. It is also worth noting that prevalence data from young people reflect infections that had occurred recently, and for this reason prevalence data from young people approximates new infections.19 Because infection in this age group is recent, it is unlikely to cause immediate death and therefore mortality is substantially lower in this than in the other groups.11,20
One of the possible explanations for the observed decline in both prevalence and incidence could be a result of the study participants modifying their sexual behavior positively after having received counseling and preventive messages, which were being provided by the research team during the previous cross-sectional and subsequent follow-up studies. However, although this effect of behavioral modification on prevalence decline is unlikely among those not previously exposed to the research team, it may not be ruled out in the case of incidence studies because the same individuals were in contact with members of the research team during each survey round and received counseling and preventive education.12
This series of studies in Kagera has shown an overall and a specific decrease in prevalence among young people. A study in Uganda showed slightly different findings; despite a decline among the young age groups, no change was observed in the overall prevalence.6 Studies in Uganda have further demonstrated a reduced prevalence of HIV infection in men and women not reflected by a similar decrease in incidence.9,10 This observed decline in prevalence without a corresponding decline in incidence could be explained by excess mortality or migration in the study areas.
Incidence may be affected by high attrition rates during follow-up and differential migration or mortality of individuals with differing risks of acquiring HIV infection. In these series of studies, the migratory patterns of individuals at various risks of infection did not vary significantly.
We have assumed that >90% of HIV infections in most of Africa are acquired through heterosexual contacts,21,22 although controversy is increasing among researchers about the routes of HIV transmission, suggesting lower proportions attributable to sexual transmission.23,24 For situations of high sexual transmission rates, a decrease in the proportion of new infections, especially among the younger generation, is most likely a result of a behavioral change.10,25 This could be a result of diminished compensatory increase in the supply of new susceptibles with a high risk or lack of increase in the risk behavior of others in the population. This will lead to a fall in the average level of risk, which will lead in turn to a reduced incidence of infection.4
Although it is difficult to attribute reduction in HIV prevalence and incidence to the effects of intervention activities, studies in Thailand have indicated a possible association between intervention activities and a fall in HIV prevalence.4,26 Conversely, it has been reported that prevalence may fall without the direct effects of interventions.27 In our study it was not possible to exclude the possible effects of migration and excess mortality due to AIDS on the observed decline in prevalence. However, the finding of significant decline of both prevalence and incidence in a population with no known selection factors for newly infected persons to migrate out, and where independent reports of similar findings have been published in neighboring Uganda, provide sufficient evidence for a real decline. Nevertheless, further studies are necessary to determine the effect, if any, of selective migration or mortality.
In Kagera, upon recognition that there was an epidemic, the government, nongovernmental organizations, individuals, and the community instituted measures to mitigate its effects in the area. The type, nature, frequency, magnitude, mode of delivery, and coverage of such interventions or measures varied from one area to another and from organization to organization. A recent study conducted in the area has indicated that there have been significant changes in sexual behaviors, norms, values, and customs that are considered high risk for HIV transmission. Specific changes in behavior reported to have occurred in the area include increase in condom use, sticking to one sexual partner, abstinence, and increased uptake of voluntary testing and counseling. Conversely, polygamy, widow inheritance, excessive alcohol consumption, and having multiple sexual partners were reported to have declined. Furthermore, grief and agony concretized by personal experiences of caring for the sick and increased mortality due to AIDS may also have influenced people to change their behaviors.28
In conclusion, despite possible limitations, these sets of studies have provided evidence that there has been a decline in the prevalence and incidence of HIV-1 in the general population of the Kagera region most significant among women aged 15-24 years and to a lesser extent among men of the same age. This decline was observed in all the 3 areas classified as high, medium, or low prevalence, implying that prevalence declined regardless of the epidemic phase documented by the 1987 study.
These findings should encourage program managers to intensify prevention strategies even in places where the epidemic has not been severe. This is especially so in some parts of Tanzania and in countries where the epidemic is still in its infancy. Because the general population continues to be at risk for HIV infection regardless of the epidemic phase, intervention programs should continue unabated and efforts made to sustain them to prevent a resurgence of the epidemic.
We would like to acknowledge the study population for their participation; the Kagera AIDS Research Project data collection team based in Kagera, especially Dr. S. Joel for his coordinating role in the field and Mr. Audax Mbakileki for his assistance in laboratory work; and Mr. Faustin Mayunga, Mrs. Tatu Mwanoga, Dr. Esther Mtumbuka, and the late Mr. Joseph Mtui for their assistance with data processing.
1. Mertens TE, Low Beer D. HIV and AIDS: where is the epidemic going? Bull World Health Organ
2. Mbizvo MT, Mashu A, Chipato T, et al. Trends in HIV-1 and HIV-2 prevalence and risk factors in pregnant women in Harare, Zimbabwe. Cent Afr J Med
3. Wilkinson D, Abdool Karim SS, Williams B, et al. High HIV incidence and prevalence among young women in rural South Africa: developing a cohort for intervention trials. J Acquir Immune Defic Syndr
4. UNAIDS. Trends in HIV Incidence and Prevalence: Natural Course of the Epidemic or Results of Behavioural Change
? UNAIDS in collaboration with Wellcome Trust Centre for Epidemiology of Infectious Diseases; Geneva: 1999:4-15.
5. Mwaluko G, Urassa M, Isingo R, et al. Trends in HIV and sexual behaviour in a longitudinal study in a rural population in Tanzania, 1994-2000. AIDS
6. Mulder D, Nunn A, Kamali A, et al. Decreasing HIV-1 seroprevalence in young adults in a rural Ugandan cohort. BMJ
7. Larsen O, da Silva Z, Sandstrom A, et al. Declining HIV-2 prevalence and incidence among men in a community study from Guinea-Bissau. AIDS
8. Kwesigabo G, Killewo J, Godoy C, et al. Decline in the prevalence of HIV-1 infection in young women in the Kagera region of Tanzania. J Acquired Immune Defic Syndr Hum Retrovirol
9. Wawer MJ, Serwadda D, Gray RH, et al. Trends in HIV-1 prevalence may not reflect trends in incidence in mature epidemics: data from the Rakai population-based cohort, Uganda. AIDS
10. Kamali A, Carpenter LM, Whitworth JA, et al. Seven-year trends in HIV-1 infection rates, and changes in sexual behaviour, among adults in rural Uganda. AIDS
11. Whitworth JA, Mahe C, Mbulaiteye SM, et al. HIV-1 epidemic trends in rural south-west Uganda over a 10-year period. Trop Med Int Health
12. Mbulaiteye SM, Mahe C, Whitworth JA, et al. Declining HIV-1 incidence and associated prevalence over 10 years in a rural population in south-west Uganda: a cohort study. Lancet
13. Kilian AH, Gregson S, Ndyanabangi B, et al. Reductions in risk behaviour provide the most consistent explanation for declining HIV-1 prevalence in Uganda. AIDS
14. Tarantola D, Schwartlander B. HIV/AIDS epidemics in sub-Saharan Africa: dynamism, diversity and discrete declines? AIDS
. 1997;11(Suppl B):S5-21.
15. Killewo J, Nyamuryekunge K, Sandstrom A, et al. Prevalence of HIV-1 infection in the Kagera region of Tanzania: a population-based study. AIDS
16. Kwesigabo G, Killewo JZ, Urassa W, 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
17. Bureau of Statistics, Ministry of Finance, Economic Affairs and Planning. National Population Census
. Dar es Salaam, Tanzania; 1988.
18. Killewo JZ, Sandstrom A, Bredberg Raden U, et al. Incidence of HIV-1 infection among adults in the Kagera region of Tanzania. Int J Epidemiol
19. Mertens T, Carael M, Sato P, et al. Prevention indicators for evaluating the progress of national AIDS programmes. AIDS
20. Fabiani M, Accorsi S, Lukwiya M, et al. Trend in HIV-1 prevalence in an antenatal clinic in North Uganda and adjusted rates for the general female population. AIDS
21. Chin J, Sato PA, Mann JM. Projections of HIV infections and AIDS cases to the year 2000. Bull World Health Organ
22. Brundtland GH. From the World Health Organization: reducing risks to health, promoting healthy life. JAMA
23. Gisselquist D, Potterat JJ. Heterosexual transmission of HIV in Africa: an empiric estimate. Int J STD AIDS
24. Schmid GP, Buve A, Mugyenyi P, et al. Transmission of HIV-1 infection in sub-Saharan Africa and effect of elimination of unsafe injections. Lancet
25. Fylkesnes K, Musonda RM, Sichone M, et al. Declining HIV prevalence and risk behaviours in Zambia: evidence from surveillance and population-based surveys. AIDS
26. Nelson KE, Celentano DD, Suprasert S, et al. Risk factors for HIV infection among young adult men in northern Thailand. JAMA
27. Bouckenooghe A, Shandera W. HIV trends in African blood donors. J Infect
28. Lugalla J, Emmelin M, Mutembei A, et al. Social, cultural and sexual behavioral determinants of observed decline in HIV infection trends: lessons from the Kagera Region, Tanzania. Soc Sci Med
Keywords:© 2005 Lippincott Williams & Wilkins, Inc.
HIV-1 infection; prevalence; incidence; trends; Tanzania