|Emerging Patterns of HIV Incidence in Uganda and other East African Countries (International Center for Migration and Health - ICMH, 1997, 97 pages)|
|AN ASSESSMENT OF EMERGING PATTERNS OF HIV INCIDENCE IN UGANDA AND OTHER EAST AFRICAN COUNTRIES|
|3.0 HIV prevalence declines in antenatal sentinel surveillance sites in Uganda|
|3.1 External validation of apparent declining HIV prevalence in Uganda: Identification and analyses of collateral data.|
Causal associations between other selected factors and observed declines in HIV prevalence in pregnant women attending antenatal sentinel surveillance sites merit consideration, and potential biases in selection would be deemed less probable if a number of conditions were met. Similar declines, for example, would need to be observed in younger age groups entering their sexual careers in other populations which are not subject to the potential biases inherent in surveying HIV prevalence in pregnant women, or in sentinel surveillance sites. Databases and reports that lend themselves to such further analysis were analyzed to the extent possible, and are reported on below.
The National Blood Transfusion Service (NBTS) collected approximately 42,000 units of blood nation-wide in 1996 from volunteer donors. The annual number of units collected has steadily increased from about 8,000 in 1989. Data are available on donor HIV status and the following donor characteristics: personal identifier code, demographic profile (age, site, and date of donation), and donation type (replacement, school, and non-school). A large proportion of all blood donations is from secondary school attendees (approximately 18-24) who have been consistently shown to have lower prevalence of HIV than the general population. Secular trends in HIV among these donations show consistent declines in the prevalence of HIV. Despite some potential for selection bias in this population, trends by time, age and region for first-time blood donors tend to support the same "cohort effect" as seen in ANC clinic HIV prevalence rates, and indicate that fewer younger people are becoming infected.
The Nakasero Blood Bank provided data on HIV prevalence trends on blood donations from secondary schools (n= +/- 17,000 per year). These were stratified according to age, school and region in order to provide age-specific trends for different regions of the country. Logistical problems and limited technical capacity for data management prevented the full utilisation of these data during the time period of the study. Nevertheless, data were provided for HIV by donation status, for the period 1989-96.
Secondary school donors are generally in the age range of 18-24. All donors received pre-blood-donation risk counselling and were excluded, as were people with the following characteristics: prior transfusion; more than 1 sex partner in last year; sex with a prostitute; use of un-sterilized needles; fever and weight loss; spouse with any of aforementioned risks; liver disease or hepatitis; or on medication. Previously identified HIV positive donors were also excluded from donating blood according to a pre-1989 donor exclusion policy. HIV prevalence, by year and by donor type is displayed in Figures 3a and 3b.
The prevalence of HIV among blood donors declined from 14.4% in 1989, to 4.3% in 1996. Among secondary or post secondary school students (approximate age, 18-24) HIV prevalence declined from 5.2% in 1989, to 1.2% in 1996. Despite some inherent biases in data from volunteer donations, these prevalence trends are consistent with those observed in ANC clinics. Data stratified by age and region are available for further analysis.
The Joint Clinical Research Centre (JCRC) is collaborating with a number of partners (WHO, UNAIDS, NIH, Case Western Reserve Medical School) in vaccine cohort trials in the military. The JCRC has also performed HIV prevalence surveys among military conscripts (age 19-22) who were in basic training and underwent physical examinations in 1991 and the autumn of 1996. The 1991 (n=300) sample was randomly selected, while the 1996 (n=1000) sample was a consecutive one in which all recruits underwent physical examinations over a 3-day period in October. The results show that the prevalence of HIV in recruits declined from 18.57% (95% CI 14.1-22.9) in 1991 (n=300), to 8% (95% CI 6.3-9.7) in 1996 (n=1000). (Figure 4).
A third survey which was scheduled for late December 1996, and which would have involved 2000 new conscripts has been postponed. Data limitations include the fact that there was no stratification by geographic region or by education. The 1991 recruits were from the "South"; those in 1996 were from a broader geographic area, and were generally of low educational background. More recent recruits are likely to be better educated (at least primary or "0" level). These potential biases in 1996 would possibly have favoured higher prevalence differences since HIV risk in Uganda in the past has tended to be positively associated with higher education and social and economic status. After discussions about the importance of cross-sectional data quality, Dr Mugenyi agreed to review all records from 1991 and 1996 surveys for data on the residential and educational backgrounds of recruits. These data will be gathered on the 2000 recruits set for testing in the spring, and will offer comparative trend data by region and education level. Based on discussions in February 1997, JCRC was attempting to identity demographic data from all recruits from the 1991 survey, but it seemed unlikely that they would be able to locate all records. In addition. HIV incidence studies in vaccine cohorts of military personnel suggest annual incidence rates in the range of 3 to 4%. not inconsistent with modelling scenarios of pregnant women in Kampala.
The AIC is a USAID sponsored organisation that has been providing HIV testing and counselling services in Uganda since 1990. The AIC had provided counselling and testing to over a third of a million clients by the end of 1995. The demand for these services remains high; the 1995 DHS survey indicated that about two thirds of men and women in both urban and rural areas would like to be tested for HIV. Testing services are available in urban centres of Kampala, Jinja, Mbale, Mbarara, Kumi, and Soroti.
Since 1991, data have been available on HIV status, demographic characteristics of those tested including educational status, and reasons for seeking testing. More recently, they have included information on residential status. Recent analyses of AIC data reveal that the age distributions of clients remained stable over time; that the majority of persons reported that marriage was the main reasons for testing; and that less than 20% sought testing because of symptoms of illness.
Although the data are from a self-selected population who chose to be HIV tested, the data set on persons tested for the first time includes over a quarter of a million people, and thus provides an "N" which is sufficiently large to provide some extrapolation to the general population, and to ANC trends. Data were first analyzed by age, sex, educational level, and according to the following sites: Kampala, other urban locations (Mbarara, Jinja, Mbale), and less urban locations. Trends in HIV prevalence for first time visits, by sex and age in Kampala and rural sites, and by education from 1992-96 are presented in the following figures.
Figure 5: Trends in HIV prevalence (%) by age and sex among clients tested at the AIDS Counseling and Testing Center in Kampala, 1992-96.
Figure 6: Trends in HIV prevalence (%) trends by age and sex among clients tested at the AIDS Counselling and Testing Centres in less urban, areas of Uganda, 1992-96.
Declining prevalence in the 15-19 and 15-24 age groups (males and females) is especially evident in the Kampala data, and less so in more rural areas. The stability in prevalence among older age groups is consistent with their having been infected at an earlier period, and before they could benefit from the prevention interventions that came later. Findings from other urban areas were similar to trends in Kampala. Biases in self-selection are difficult to account for in these different trends.
Analyses of the data according to educational status are shown in Figure 7, and show differential trends over time.
Figure 7: Trends in HIV prevalence (%) by educational status, age and sex, AIDS Testing and Counselling Centre, Uganda, 1992-96.
The differential declines according to age were most evident in women aged between 20-24 and 25-29 and with primary and secondary education. Among women with no education there was a tendency for the prevalence of HIV to be high and increasing (though not statistically significant). These differences warrant further analysis, especially given the importance of education as a potential factor in education and information interventions. Furthermore, these findings add support to the idea that significant behavioural changes may have taken place in response to AIDS prevention initiatives. The consistency of these data with those from sentinel sites-urban declines, rural stability, and differential trends by educational status - also suggest that major difference may be emerging between the behaviour patterns of urban and rural populations, and that education may be one of the key factors involved.
The Department of OB/GYN, at Makarere University in collaboration with Case Western Reserve University, WHO, and Johns Hopkins University have been involved in cohort studies of HIV and reproductive health among women attending clinics at Mulago Hospital since 1989. Patient selection methods differ from those used at antenatal clinic sentinel surveillance in that study entry is voluntary, involves informed consent, is restricted to women living within 15 km of Kampala, and involves women who agree to HIV testing (notification of results is voluntary). Study participation rates are high and there appear to be no notable differences between participants and non-participants.
This study population provides a comparative ANC population to that at Nsambya and Rubaga, which are the Kampala ANC sentinel sites. The study population may be less influenced by the potential of changing population denominator characteristics such as residence (in migration of persons from areas with lower HIV prevalence rates which could result in a "dilution effect" and hence an apparent decline in HIV prevalence). Figures 8 and 9 show trends in ANC sentinel sites in Kampala according to time and age, and also provide a comparison of age-specific prevalence over time at the Kampala ANC sentinel and the Mulago ANC clinic population.
The data illustrate decreasing prevalence rates among 15-19 and 20-24 year olds that are similar to those observed in the sentinel sites. Only data for Mulago were available for the 1994-96 period. Data for Mulago for 1991-93 are currently being analysed by WHO/HRP in Geneva.
Differences in prevalence in the age group 25-29 are statistically significant and most likely reflect differences in patient selection, the Mulago cohort being a non-random selection from stable residents. The fact that the observed trends are in all other respects similar to those of ANC sentinel sites suggest that migration into Kampala is not influencing HIV prevalence declines.
The Basic Health Services, Western Uganda (GTZ) project has been responsible for sentinel HIV surveillance in urban and rural ANC populations in Fort Portal and Kaborole District since 1990-91. Both HIV and TPHA prevalence from sentinel ANC populations by age is compared for the periods over time and for the periods 1991/92, 1993-94, and 1996/96, by educational status. The results shared by Dr Kilian (GTZ) are presented below in figures 10-12:
Figure: 10: HIV trends by age among antenatal clinic attendees by residence, Kabaorle district, Uganda, 1991-96.
Figure 12: HIV by age and educational status in ANC pregnant women, Kabaorle, 1991-96.
Decreases in HIV prevalence, TPHA, and active syphilis (not shown, but reported) were apparent over time in the 15-20 old age group. Although a syphilis treatment program in ANC clinic populations is on-going, it is not believed to be associated with the declines beginning in the early 1990s (personal communication Dr A Kilian, GTZ). Stratification of the study population by educational level revealed an inverse relationship between current HIV prevalence in 15-20 year olds and educational level; decrease were even more evident among people with secondary education than those with only primary education.
Behavioural survey data were not available, but these findings when coupled with behavioural survey data, could provide insights into the reasons for the less impressive decline in HIV prevalence in rural areas. Prevalence by age from population based sero-survey data provides reasonable comparability to ANC populations, particularly in younger age groups and in urban areas (data not shown).
These findings further corroborate MOH/ANC sentinel data that indicate that a decline in prevalence in younger age groups (15-20) in urban and semi-urban populations. The differential according to educational status also provides additional support to the premise that behaviour change has been a major determinant in the observed declines in HIV prevalence. It is also consistent with the different prevalence by educational status in the AIC data. The lack of major declines in younger cohorts in rural areas may be indicative of lower out-reach coverage by prevention campaigns and less change in high-risk behaviours.
The MRC-supported observational cohort study undertaken in 15 villages in rural Masaka District, and started in 1989, has provided considerable information on HIV dynamics in rural settings. This study provided the first evidence of decreasing HIV-1 sero-prevalence in young adults. From 1990-94 HIV sero-prevalence in males 13-24 decreased from 3.4% to 1% and in females 13-24, from 9.9% to 7.2%, while overall cohort prevalence was stable, 8.2% in 1990 and 7.2% in 1994 (BMJ 31 30 Sept. 1995). More recent cohort data (Lancet. Sept. 1996) on sero-incidence suggest an increase in HIV infections in females aged 13-34 from 0.6% in 1990 to 1.2% in 1994, but declines in older females and in males. The incidence rates are so low that meaningful trends in age-specific rates are illusive (13-34 is a very broad age group), but in discussion with Dr Whitworth in late February, he agreed to supply these data as well as the latest prevalence data for the cohort by mid-April.
The apparent stability (or even increase) in prevalence and possibly incidence rates, in this rural population contrasts with the declines observed among urban and semi-urban populations and suggests that distinct HIV behavioural dynamics are at work in rural populations. Behavioural surveys suggest that greater behaviour changes have occurred in urban than rural people, and also supports the contention that sexual behaviour patterns may be more amenable to aggressive promotion of preventive strategies and interventions than had been previously thought possible or likely. Further population based KABP surveys of this population would offer an important and much more precise link with data on sero-incidence and prevalence.
NB. Data from Masaka was not available by mid-May, but should be pursued by USAID Uganda.
This study offers critically important insights into HIV dynamics in a predominantly rural population. It is unique in that it is the likely site of initial epidemic spread of HIV in the early 1980s in Uganda, It has some of the highest HIV prevalence rates ever recorded in the general population (45%) together with empirical evidence of negative population growth at village level in Rakai district (Low-Beer, Stoneburner, Mukulu: Nature Medicine, May 1997 appended). The direction of recent secular HIV trends from 1993 onward is not yet clear from current data, but Dr David Serwaada believes data relevant to this could be released by the summer of 1997. They will be of critical importance in interpreting differential HIV incidence and prevalence trends in Uganda, particularly with respect to heterogeneity of HIV risk behaviours. A major STD experimental intervention trial is also underway there, and the understanding of current and previous HIV incidence dynamics before and during the trial will be critical to any interpretation of the effect of interventions. This,, like the MRC cohort, could be an important site for a population-based KABP survey of HIV risk behaviours. A population-based HIV survey to replicate earlier surveys would similarly allow a better longitudinal evaluation of HIV incidence and prevalence.