Fourteen years of surveillance of HIV-1 prevalence among pregnant women attending antenatal care clinics in western Uganda

Abstract

This study monitored long-term temporal trends in HIV-1 prevalence in antenatal clinic attendees living in western Uganda. Semi-annual data collection was done from 1991 to 2004. For each woman the following data were recorded: HIV-1 status, age, educational status, marital status, occupation and parity. The results show that the overall HIV-1 prevalence was 15.3% during the entire time period (urban 21.3%, semi-urban 12.7% and rural 7.1%). Between 1991 and 2004, we observed a gradual decline in the HIV-1 prevalence. The decline was most pronounced in urban women aged 15–19 years old and least pronounced in rural women aged 20–24 years. Women above 25 years of age did not show any decline in HIV-1 prevalence over time. The declining HIV-1 prevalence in the younger age groups (15–24 years) likely represents a declining risk for acquiring HIV infection as we have previously shown in the urban sub-sample of this data set.

Keywords

HIV-1 epidemic Uganda anonymous surveillance pregnant women time trends

Introduction

A strongly advocated approach for monitoring HIV trends over time in generalized epidemics such as Uganda is to conduct HIV-1 surveillance in sentinel population groups and sites, e.g. pregnant women attending antenatal care (ANC) clinics.^1–4 This approach has been implemented by the Ugandan Government's National HIV/AIDS Control Program to monitor trends in HIV-1 prevalence. Sites used for ANC clinic surveillance studies of HIV-1 prevalence include Kampala, Jinja, Mbarara, Mbale, Gulu, as well as other townships.^5,6 In all of these studies, a decline in HIV-1 prevalence over time has been described. In addition to these studies, two major cohort studies in Rakai and Masaka districts are ongoing, measuring the number of new HIV-1 infections. Information from the Masaka cohort in 2002 showed a 37% decline in HIV-1 incidence in this population.⁷ Findings from the Rakai cohort (2005) indicate that the HIV-1 prevalence did decline in young adults of 20–24 years, but not in adolescents 15–19 years of age.⁸ The latest national HIV-1 serosurvey from 2004 shows a national average of HIV-1 prevalence of 6.4% in the sexually active population.⁹

As no ANC data on HIV-1 prevalence are available from western Uganda, we analysed existing data from the ANC surveillance programme from 1991 to 2004, which was provided by the Kabarole Health Department and the German Government-sponsored project Basic Health Services, western Uganda in Fort Portal.

Objectives of Data Analysis

(1)

To establish long-term temporal trends in HIV-1 prevalence in pregnant women from western Uganda in three different geographic strata, adjusting for potentially confounding variables such as marital status and education;

(2)

To investigate temporal trends in HIV-1 prevalence separately for different age groups within each geographic stratum.

Methods

Two participant samples were drawn each year as follows: all consecutive pregnant women, who presented themselves to the ANC, were included, beginning on the first of January and the first of July until a sample size of 120 was reached in each. For each woman the following data were recorded: HIV-1 status, age, educational status, marital status, occupation and parity. Blood samples were taken and tested using an enzyme-linked immunosorbent assay at the regional hospital in Fort Portal. Either the Enzygnost HIV 1 + 2 (Behring Diagnostics, Marburg, Germany) or Vironostika anti-HIV1/2+ (Organon Teknika, Boxtel, The Netherlands) was used.

The ANC recruitment sites were divided into three geographical strata: urban, semi-urban and rural. Each stratum was represented by either a hospital (for the urban site) or by a health centre (for the semi-urban and rural sites). The urban site Fort Portal was the only township, and therefore no sampling was required. In the semi-urban and rural stratum, a simple random sample was used to select two ANC clinics in each stratum out of a total of eight ANC clinics.

HIV-1 prevalence data from the two sample collections made each year were pooled to give mid-year point prevalence estimates. Descriptive, univariate and multivariate statistical methods were used to analyse the data. Temporal trends in HIV-1 prevalence were presented unadjusted as well as adjusted for educational status, marital status, occupation and study site by applying univariate and multivariate statistical analysis. In the logistic regression model with HIV-1 status as the outcome of interest, age, education, marital status and occupation were entered as categorical variables. Time was considered as a continuous variable, that is 1 for 1991, 2 for 1992, so on, and 14 for 2004. We considered both linear trends and non-linear trends by using quadratic terms of the time covariate. To study age-specific and site-specific temporal trends we considered interaction terms of time with age and time with site. The statistical significance level was selected at 0.05. The statistical analysis was performed using SPSS version 15.0 (SPSS Inc, Chicago, IL, USA).

The study was approved by the University of Alberta's Health Research Ethics Board (Panel B). In Uganda, further approval was obtained from the Uganda National Council of Science and Technology, Kampala.

Limitations

The following are the limitations of the study. (1) The study consisted of a series of repeated cross-sectional studies. Any true cause–effect relationship between HIV-1 prevalence and the other variables cannot be established. (2) HIV data in pregnant women attending antenatal services can over-represent or under-represent the general population. (3) Pregnant women attending ANC clinics are a selected group of the general population. This limits the generalizability of the study.

Results

A total of 14,748 pregnant women attending ANC clinics in the three geographical strata participated. The median annual sample size for urban, semi-urban and rural pregnant women in the 14 years of observation was 491, 300 and 276, respectively. The mean age of participants was 24 years (SD 6.1 years, range 15–58 years). Between the years 1991 and 2004, the mean age of HIV-positive pregnant women increased significantly by 3.5 years, while the mean age of HIV-negative pregnant women in the sample remained the same.

Overall HIV-1 prevalence in all participants was 15.3% (urban 21.3%, semi-urban 12.7% and rural 7.1%). For demographic details see Table 1. Figure 1 shows the time trend of the HIV-1 prevalence with fitted regression lines from 1991 to 2004, stratified for age and geographical location. We used the age groups 15–19 years, 20–24 years and 25 years and older. This was justified because we found that in the age groups above 25 years no significant HIV-1 decline existed, when these age groups were examined separately, e.g. 25–29 years, 30–34 years, 35–39 years and 40 years and older. This was a consistent finding in all three geographical strata.

Figure 1

Temporal trends in the HIV-1 prevalence, stratified for age and geographical location, from 1991 to 2004 in western Uganda

Table 1

Odds ratio (OR) and 95% confidence interval (95% CI) for HIV prevalence – univariate analysis

Variable	n = 14,748 (%)	% HIV positive	OR (95% CI)	P value
Age in years*
15–19	3988 (27.0)	13.1	1.00 (Reference)
20–24	4866 (33.0)	16.4	1.30 (1.15–1.48)	< 0.001
25 +	5890 (48.0)	15.7	1.24 (1.10–1.40)	< 0.001
Marital status^†
Single	3062 (28.0)	16.3	1.00 (Reference)
Married	11527 (78.4)	14.9	0.90 (0.80–1.00)	0.057
Widowed/divorced	120 (0.8)	17.3	1.07 (0.63–1.83)	0.792
Education^‡
Illiterate	5521 (37.6)	11.6	1.00 (Reference)
Literate – no schooling	2986 (20.3)	12.8	1.11 (0.97–1.28)	0.122
Primary	4877 (33.2)	20.2	1.93 (1.72–21.6)	<0.001
Secondary/tertiary	1307 (8.9)	19.9	1.89 (1.59–2.24)	<0.001
Occupation
Farmer/peasant	9115 (61.8)	13.7	1.00 (Reference)
Business owners	889 (6.0)	23.4	1.93 (1.61–2.30)	< 0.001
Other	2831 (19.2)	20.3	1.60 (1.42–1.80)	< 0.001
Unknown/missing	1913 (13.0)	12.5	0.90 (0.78–1.05)	0.168
Site
Urban	7574 (51.4)	21.3	1.00 (Reference)
Semi-urban	3626 (24.6)	12.7	0.54 (0.48–0.60)	< 0.001
Rural	3548 (24.1)	7.1	0.28 (0.24–0.33)	< 0.001
Time^§
Time			0.96 (0.95–0.97)	< 0.001

Four missing values

†

Thirty-nine missing values

‡

Fifty-nine missing values

Time is defined as 1: 1991, 2:1992, 3: 1993, …, 14: 2004 and used as a continuous variable

In the age group 15-19 years, there was a decline in HIV-1 prevalence in all three geographical strata. It was most pronounced between 1991 and 1997. In the age group 20–24 years, similar observations were made but the trends were less clear. In contrast, temporal trends in HIV-1 prevalence in the age groups above 25 years in the urban and semi-urban areas were best explained by a curved relationship, indicating an increase in the HIV-1 prevalence in the 1990s and then a slight decline from 2000 onwards. In the rural areas, participants above 25 years did not show any upward or downward trend, i.e. HIV-1 prevalence graph was more or less horizontal. Noteworthy was that in all three age groups studied, HIV-1 prevalence was highest in the youngest age groups, followed by the age group 20–24 years and then the older age groups.

Results from the multivariate logistic regression model also showed that there was a significant downward trend in the risk of acquiring HIV over time. In contrast, when time square was used in the model, a slight increase of the risk of acquiring HIV over time was found. However, this finding must be interpreted with caution, as we found three significant interaction terms: time × age, time × site and time² × age (Table 2).

Table 2

Odds ratio (OR) and 95% confidence interval (95% CI) for HIV prevalence – multivariate analysis

Variable	OR (95% CI)	P value
Age in years
15–19	1.00 (Reference)	.
20–24	0.68 (0.43–1.07)	0.099
25 +	0.34 (0.22–0.54)	< 0.001
Marital status
Single	1.00 (Reference)
Married	0.80 (0.70–0.90)	< 0.001
Widowed/divorced	1.16 (0.63–2.11)	0.639
Education
Illiterate	1.00 (Reference)
Literate but no schooling	1.18 (1.02–1.36)	0.027
Primary	1.25 (1.09–1.42)	0.001
Secondary/tertiary	1.16 (0.96–1.41)	0.120
Occupation
Farmer/peasant	1.00 (Reference)
Business owners	1.28 (1.06–1.55)	0.011
Other	0.91 (0.78–1.07)	0.253
Unknown/missing	0.87 (0.71–1.08)	0.209
Site
Urban	1.00 (Reference)
Semi-urban	0.83 (0.64–1.07)	0.142
Rural	0.36 (0.26–0.49)	< 0.001
Time ^*
Time	0.83 (0.72–0.90)	< 0.001
Time²	1.01 (1.00–1.02)	0.006
Time × site
Time × semi-urban	0.95 (0.91–0.98)	0.003
Time × rural	0.99 (0.95–1.02)	0.428
Time × age
Time × age 20–24	1.30 (1.13–1.50)	< 0.001
Time × age 25+	1.47 (1.28–1.69)	< 0.001
Time² × age
Time² × age 20–24	0.98 (0.97–0.99)	< 0.001
Time² × age 25+	0.98 (0.97–0.99)	< 0.001

Time is defined as 1: 1991, 2: 1992, 3: 1993, …, 14: 2004 and used as a continuous variable

In order to account for the interactions, we developed a logistic regression model for each age group and each geographical stratum separately. These findings are presented in Table 3.

Table 3

Odds ratio (OR) and 95% confidence interval (95% CI) of time^* variable for HIV, stratified by site and age

	Age in years
	15–19		20–24		25+
Site	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
Urban	0.91 (0.88–0.95)	<0.001	0.96 (0.93–0.98)	< 0.001	1.17 (1.04–1.30)	0.008
					0.99 (0.98–1.00)	0.021
Semi-urban	0.92 (0.87–0.97)	0.002	0.90 (0.86–0.94)	< 0.001	1.23 (1.00–1.52)	0.047
					0.98 (0.97–1.00)	0.030
Rural	0.94 (0.87–1.00)	0.060	0.96 (0.90–1.01)	0.099	1.00 (0.96–1.05)	0.955

Time is defined as 1: 1991, 2: 1992, 3: 1993, …, 14: 2004. The OR and 95% CI in the first line are for the time variable and second line (if applicable) for time-square variable

In the age group 15–19 years, there was a significant decline in the odds ratio (OR) for acquiring HIV infection in urban and semi-urban sites, and a borderline significant (P = 0.060) decline in the rural site. This pattern was similar in the age group 20–24 years, that is, a significant decline of the risk of HIV-1 infection over time in the urban and semi-urban sites, and a borderline significant (P = 0.099) decline in the rural sites were found. In contrast, in the age groups above 25 years we observed an increasing trend over time in the 1990s in the urban and semi-urban sites, which means that the risk of HIV-1 infection increased in those sites between 1991 and 2000. We also observed a significant decline in the OR for HIV-1 infection in pregnant women above 25 years in the early years after 2000, as shown by the significant time² variable in Table 3. In contrast, the rural stratum of pregnant women over 25 years did not show any change in HIV-1 prevalence rates during the entire period of study.

Discussion

This article adds to the growing body of evidence supporting a declining HIV-1 prevalence in the Uganda. It also presents new and important information for the following reasons: (1) most ANC surveillance studies in Uganda have shorter time spans for their observations; (2) we present data from three geographical strata (urban, semi-urban and rural) within the same region, while most studies have only examined pregnant women in one geographical stratum, e.g. Kampala, Jinja, Gulu, Mbarara, etc.; and (3) no comprehensive HIV-1 surveillance data have been published from this part of western Uganda, which is distinct in its geographical terrain, its ethnicity and cultural characteristics of its population.

The most important finding from this analysis is that in the age group 15–24 years, age group HIV-1 prevalence has significantly declined in urban and semi-urban areas and has shown a downward trend in rural areas as well with a borderline significance (P = 0.060, see Figure 1). In an earlier, separate study, we analysed HIV-1 prevalence in the urban sub-sample of this data set for the period of 1991–1997. Our earlier study concluded that the decline in HIV-1 prevalence from 31% in 1991 to 9% in 1997 in pregnant women aged 15–19 years was best explained by sexual behaviour changes. This conclusion was supported using computer simulation methods.¹⁰ If the most likely explanation for the decline in HIV-1 prevalence in the urban sub-sample of our data from 1991 to 1997 was sexual behaviour change, there is no reason not to believe that the same is true for the semi-urban and rural samples in our data. In a second study that we published analysing sexual behaviour change data from 2000 to 2005, we confirmed that sexual behaviour change had taken place in young people. The two key sexual behaviour changes we reported were increased condom use and delayed onset of first sexual intercourse.¹⁶

HIV-1 prevalence decline has been observed in other parts of Uganda.^5,11–13 Another long-term study on ANC clinic attendees was conducted in Mbeya, Tanzania, where the HIV-1 prevalence in the age group 15–24 years increased from 1988 to 1995, but then decreased in all age groups between 1995 and 2000.¹⁴ This later decline in HIV-1 prevalence from 1995 to 2000 and not in the early 1990s, as we observed, may be caused by a less mature HIV epidemic in Tanzania compared with Uganda. Our argument that the observed decline in HIV-1 prevalence in younger pregnant women in Kabarole District is most likely the result of sexual behaviour change was also confirmed in the Tanzanian study.

Of some concern is the finding that in the period 2000–2004, the proportions of pregnant women over 25 years who were infected with HIV has increased compared with the early 1990s. The shift of the HIV epidemic into the older age groups has been observed in other parts of Uganda, e.g. in the Masaka study where a new HIV incidence peak was found in older men aged 40–44 years between 2000 and 2004.¹⁵ The spread of the HIV epidemic into the older age groups (as we observed) is likely a consequence of a slower and delayed sexual behavioural change in older persons, who are more determined in their sexual behaviour and are less amenable to safer sex interventions than the younger age groups.

This shift of the HIV-1 infection from younger to the older age groups has to be addressed by the District Health Services. The more vulnerable population for acquiring HIV/AIDS are now older women who often have less exposure and access to health services including HIV/AIDS treatment and prevention. Therefore, our recommendation to the District Health Services is to focus their HIV/AIDS treatment and prevention efforts specifically on this more recently identified risk population.

Footnotes

Acknowledgements

We thank Jean Kipp for her comments on the first draft of the manuscript. We thank the Health Department, Kabarole District, Uganda and the Basic Health Services Project, supported by the Deutsche Gesellschaft fuer Technische Zusammenarbeit (GTZ) for providing the data for our analysis.

Financial support: The HIV-1 surveillance program was financially supported by the Federal Republic of Germany through the projects PN 1993.2027.6 (AIDS Control in Uganda), PN 1994.2195.9 (Basic Health Services Western Uganda) and PN 2000.2072.7 (AIDS-Control in Developing Countries). Our study (analysis of the data) was financed through the Fund for Support of International Development Activities (FSIDA), University of Alberta, Edmonton, Canada. All authors declare no conflict of interest.

References

Chin

Public health surveillance of AIDS and HIV infections.

Bull World Health Organ 1990; 68: 529–36

Kwesibago

, Killewo

, Sandstrom

Sentinel surveillance and cross sectional surveys on HIV infection prevalence: a comparative study.

East Afr Med J 1996; 73: 298–302

Borgdorff

, Barongo

, van Jaarsweld

Sentinel surveillance for HIV-1 infection: how representative are blood donors, outpatients with fever, anemia or sexually transmitted diseases and antenatal clinic attendees in Mwanza region, Tanzania?

AIDS 1993; 7: 567–72

Kigadye

, Klokke

, Nicoll

Sentinel surveillance for HIV-1 among pregnant women in a developing country: 3 years’ experience and comparison with a population serosurvey.

AIDS 1993; 7: 849–53

Fabiani

, Accorsi

, Lukwiya

Trend in HIV prevalence in an antenatal clinic in North Uganda and adjusted rates for the general female population.

AIDS 2001; 15: 97–103

STD/HIV/AIDS Surveillance Report. STD/AIDS Control Programme, Ministry of Health, Kampala, Uganda, June 2003

Mbulaiteye

, Mahe

, Whitworth

Declining HIV incidence and associated prevalence over 10 years in a rural population in south-west Uganda: a cohort study.

Lancet 2002; 360: 41–5

Wawer

, Gray

, Serwadda

, Namukwaya

, Makumbi

, Sewankambo

Declines in HIV Prevalence in Uganda: Not as Simple as ABC. 12th Conference on Retroviruses and Opportunistic Infections, Boston, 2005. http://www.retroconference.org/2005/Home.htm

Uganda HIV/AIDS Sero-behavioral Survey 2004-2005. Ministry of Health/Macro International, Kampala, Uganda, July 2006

10.

Kilian

, Gregson

, Ndyanabangi

Reductions in risk behavior provide the most consistent explanation for declining HIV-1 prevalence in Uganda.

AIDS 1999; 13: 391–8

11.

Asiimwe-Okiror

, Opio

, Musinguzi

, Madraa

, Tembo

, Carael

Change in sexual behaviour and decline in HIV infection among young pregnant women in urban Uganda.

AIDS 1997; 11: 1757–63

12.

Fabiani

, Ayella

, Ble

Increasing HIV prevalence among pregnant women living in rural areas of the Gulu district (North Uganda).

AIDS 2001; 15: 2330–1

13.

Kirungi

, Musinguzi

, Madraa

Trends in antenatal HIV prevalence in urban Uganda associated with uptake of preventive sexual behaviour.

Sex Transm Infect 2006; 82 (S1): 36–41

14.

Jordan-Harder

, Maboko

, Mmbando

Thirteen years HIV-1 sentinel surveillance and indicators for behavioral change suggest impact of programme activities in south-west Tanzania.

AIDS 2004; 18: 287–94

15.

Shaffer

LA.

HIV prevalence and incidence no longer falling in south-west Uganda: evidence from a rural population cohort 1989-2005.

AIDS 2008; 22: 1641–49

16.

Kilian

, Kipp

, Saunders

Trends in HIV-prevention related attitudes and behaviour among secondary school students in Western Uganda.

J Acquir Immun Defic Syndr 2007; 44: 586–93