Trends in HIV prevalence and incidence sex ratios in ALPHA demographic surveillance sites, 1990 2010 Zaba B 1, Calvert C 1, Marston M 1, Isingo R 2, Nakiyingi Miiro J 3, Lutalo T 4, Crampin A 1,5, Nyamukapa C 6, 7, Todd J 1, Reniers G 1 1 London School of Hygiene and Tropical Medicine, United Kingdom 2 National Institute for Medical Research, Tanzania 3 MRC/UVRI Uganda Research Unit on AIDS, Uganda 4 Rakai Health Sciences Program, Uganda 5 Karonga Prevention Study, Malawi 6 Imperial College, United Kingdom 7 Biomedical Research and Training Institute, Zimbabwe Short abstract Antiretroviral Therapy (ART) has two well described beneficial effects. First, it drastically prolongs the life expectancy of those receiving treatment, and second, it reduces HIV transmission to uninfected partners. The expansion of ART programs is thus expected to elevate HIV prevalence rates and reduce HIV incidence at the same time. These expectations are largely confirmed in a pooled dataset of five African demographic surveillance sites with HIV status information. Further, we find that the F/M sex ratio of prevalent infections increases over time and that suggests that ART coverage is better among women than men. This is corroborated by an increasing F/M sex ratio of incidence, suggesting that although the pool of HIV infected women is increased, less are infectious compared to the pool of HIV positive men. Our results thus indicate that higher ART coverage rates benefit HIV positive women (more than men) in term of increased survival, but benefit HIV negative men (more than women) because of a greater reduction in new infections.
Extended abstract Background and research questions Antiretroviral Therapy (ART) has two well described beneficial effects. First, it drastically prolongs the life expectancy of those receiving treatment [1], and second, it reduces HIV transmission to uninfected partners [2 3]. The expansion of ART programs is thus expected to inflate HIV prevalence rates and reduce HIV incidence at the same time. We explore gender differences in these epidemic markers because a number of studies have suggested that women make greater use of AIDS care services than men (and often at an earlier stage of disease progression) [4 5]. One study suggested that gender differences in the uptake of ART have become more explicit ever since antriretrovirals have been made available free of charge to the end user [6]. If ART coverage is indeed higher for HIV positive women compared to men, then the sex ratio of prevalent as well as incident infections should increase with the expansion of ART programs. Data and methods We seek evidence for these hypotheses in a pooled dataset of five demographic surveillance sites with HIV status information (Table 1). 1 These projects are located in East and Southern Africa and revolve around a regular census of the population under surveillance and community based testing for HIV. The population in these sites ranges from 7,000 32,000. Table 1: ALPHA demographic surveillance site attributes Study site Karonga Kisesa Manicaland Masaka Rakai Country Malawi Tanzania Zimbabwe Uganda Uganda Demographic surveillance 2002-11 1994-11 1998-08 1989-11 1994-10 HIV surveillance 2002-09 1994-10 1998-08 1989-11 1994-08 Other linked HIV testing 1982, 88 1990 ART roll-out in community 2005-07 2005-07 2006-08* 2003-05 2005-07 Population size (2008) 32,000 28,000 30,000 7,000 30,000 2008 HIV prevalence 7% 7% 16% 6% 11% Total PYO (ages 15+) 147,000 240,000 57,000 150,000 265,000 Age eligibility for testing ** 15+ 15+ (15-44 pre- 1999) Women:15 49 Men: 18 59 13+ 15+ (15 49 in some rounds) Notes: * ART rollout is still incomplete in Zimbabwe ** These age-eligibility criteria are not always strictly enforced 1 All of these sites are member of the ALPHA Network: http://www.lshtm.ac.uk/eph/dph/research/alpha/
Because HIV surveys are not conducted on an annual basis and not everyone is tested in each survey round, we developed a set of rules for imputing missing values in order to maximize the utilization of the available data. These imputation rules are detailed in Table 2. Table 2: Imputation rules for annual HIV status information HIV negative HIV positive HIV status unknown notes Before first neg test Between first and last neg. test Included Included After last negative test Included for X years Included after X years X = 5 years Eastern Africa X = 2 years Southern Africa Before first pos. test Between first and last pos. test Included for Y years included Included before Y years Y = 5 years Eastern Africa Y = 2 years Southern Africa After last pos. test included Between last neg and first pos test never had HIV test Share after random allocation of infection date if interval < Z years Included if interval > Z years Included Z = 4 years Classified ineligible in the Manicaland study Figure 1: Proportion with HIV status unknown by study site, sex and year (adults aged 15 and above) Proportion HIV status unknown by sex and calendar year Karonga Kisesa Manical propunk 0.2.4.6 0.2.4.6 Masaka Rakai 1990 1995 2000 2005 2010 1990 1995 2000 2005 20101990 1995 2000 2005 2010 calendar year males females Graphs by study_name
After imputation, HIV status information is known for 60 to 85 percent of the adults (aged 15 and above), and these figures vary considerably by location, period and sex (Figure 1). For the Manicaland site, we do not (currently) have information on who has been approached for an HIV test and cannot produce this statistic. The proportion with unknown HIV status is generally higher for men than for women and that is consistent with other studies that have identified higher non participation rates in serosurveillance studies among men [7], and, more generally, lower coverage of HIV testing and counseling (HTC) [8]. Over time, the proportion with unknown HIV status tends to increase (the Karonga site is an exception in this regard. HIV status information availability is generally lower in old age (not shown), but that is largely due to the age eligibility criteria for HIV testing in these studies. In the dataset for studying HIV incidence, we only include individuals with at least one negative HIV test result, and the analysis intervals are bounded by the first and last test (no assumptions about prepositive or post negative status are required). We randomly distribute sero conversions between the date of the last negative and first positive test in order to avoid spurious incidence peaks between serosurveys. Sero conversion intervals in excess of 4 years are excluded because these would not allow us to attribute the sero conversions to the pre or post ART period. For studying gender differences in HIV prevalence over time, we use logistic regression to model the odds that an individual is HIV positive. We are particularly interested in an interaction effect between sex and dummies pertaining to the stage of ART availability in the community. We distinguish three phases: the pre ART phase, the rollout phase and the phase where ART is available to everyone. For modelling HIV incidence, we rely on a piece wise exponential regression model. Again, we are interested in an interaction effect between sex and the stage of ART availability. Preliminary results Figure 2: Trends in sex ratio of HIV prevalence by age group, pooled data 5 ALPHA sites HIV prevalence sex ratios by age group and calendar year HIV status data for all studies combined Prevalence sex ratio.5 1 1.5 2 2.5 1995 2000 2005 2010 calendar year 15-29 30-44 45-59 60+
(a) HIV prevalence We first present trends in the F/M sex ratio of HIV prevalence by age in all sites combined (Figure 2). This illustration highlights that the F/M prevalence ratio is particularly high in young adults, and that is a direct result of the younger age pattern of infections among women. More interesting for the current discussion is the fact that the sex ratio of prevalent infections slowly increases in the 2000s. That could indeed result from the fact that women have privileged access to or simply make better use of the available AIDS care services. These interpretations are confirmed in a (preliminary) multivariable analysis whereby we model HIV status (Table 3). The first model, without interactions, indicates that HIV prevalence is higher for women than for men, that it varies by age and by site, and also that it increased with the expansion of ART programs. The model with interactions adds that the F/M sex ratio of prevalence is significantly higher in the period that ART has been available (even after accounting for a year by year linear increase in the sex ratio of infections), and that could indeed indicate that HIV positive women survive longer because of higher ART utilization rates. Table 3: Logistic regression of HIV status on background characteristics (ALPHA sites, 1990 2010) MODEL 1: NO INTERACTIONS MODEL 2: SEX INTERACTIONS Odds Ratio sig Odds Ratio sig Male (ref) 1 Male (ref) 1 fem 1.34 *** fem 1.65 *** Karonga 1.14 *** Karonga 1.13 *** Kisesa 0.76 *** Kisesa 0.76 *** Masaka (ref) 1 Masaka (ref) 1 Manical 3.01 *** Manical 3.07 *** Rakai 1.70 *** Rakai 1.68 *** sex interaction 15 19 0.14 *** 15 19 0.09 *** 1.98 *** 20 24 0.48 *** 20 24 0.35 *** 1.60 *** 25 29 (ref) 1 25 29 (ref) 1 1 30 34 1.39 *** 30 34 1.57 *** 0.83 *** 35 39 1.44 *** 35 39 1.93 *** 0.62 *** 40 44 1.25 *** 40 44 1.96 *** 0.48 *** 45 49 0.97 45 49 1.65 *** 0.42 *** 50 54 0.70 *** 50 54 1.23 *** 0.39 *** 55 59 0.54 *** 55 59 0.86 ** 0.44 *** 60+ 0.23 *** 60+ 0.43 *** 0.28 *** none (ref) 1 none (ref) 1 1 roll out 0.98 roll out 0.95 * 1.06 * available 1.05 * available 0.98 1.11 ** year (cont) 0.99 *** year (cont) 0.99 ** 1.00
(b) HIV incidence Figure 3: Age specific patterns of HIV incidence by ART availability, pooled data 3 ALPHA sites HIV infection hazard by age, sex and ART availability pooled data for Kisesa, Masaka and Rakai pre-art ART introduction ART available 0.02.04.06.08.1 0.02.04.06.08.1 0.02.04.06.08.1 15 25 35 45 55 65 75 age 15 25 35 45 55 65 75 age 15 25 35 45 55 65 75 age male female male female male female Figure 4: Trends in sex ratio of crude HIV incidence rate, by study site sexratio_crude.5 1 1.5 2 2.5 Trends in sex ratio (F/M) of crude incidence rate, by site 1995 2000 2005 2010 calendar year Karonga Manical Rakai Kisesa Masaka
Figure 3 illustrates the decline in HIV incidence rates in recent years: whereas incidence rates in young adults were still around 1% in the pre ART period, the peak incidence is less than half a percent in recent years. ART probably contributes to the reduction in incidence rates, but it is unlikely to be the sole factor of importance. The age pattern of incidence rates becomes progressively older (especially for women), and there is increasing similarity in the age pattern of HIV incidence between men and women. Figure 4 suggests that the sex ratio of the crude incidence rate has generally increased over time, although Rakai and Karonga do not maintain this trend throughout. The differences in age pattern of incidence shown in figure 3, suggest the sex ratio will change at different rates in different age groups. Determinants of the sex ratio of new infections are further explored in the analysis presented in Table 4. This analysis shows us that the incidence rate declined over time, particularly after the expansion of the ART programs in the sites. However, that decline is much more pronounced for men compared to women (the interaction between sex and ART availability is positive and significant) and that has led to an increase in the F/M sex ration of new infections. Table 4: Piece wise exponential regression of incident HIV infections on background characteristics characteristics (ALPHA Network sites, 1990 2010) MODEL 1: NO INTERACTIONS MODEL 2: WITH SEX INTERACTIONS Hazard Ratio sig Hazard Ratio sig Male (ref) 1 Male (ref) 1 fem 1.07 *** fem 0.99 Karonga 0.87 ** Karonga 0.84 *** Kisesa 1.56 *** Kisesa 1.57 *** Masaka (ref) 1 Masaka (ref) 1 Manical 2.17 *** Manical 2.17 *** Rakai 1.84 *** Rakai 1.82 *** sex interaction 15 19 0.52 *** 15 19 0.35 *** 1.92 *** 20 24 1.02 20 24 0.93 * 1.19 *** 25 29 (ref) 1 25 29 (ref) 1 1 30 34 0.89 *** 30 34 0.92 * 0.95 35 39 0.74 *** 35 39 0.85 *** 0.78 *** 40 44 0.60 *** 40 44 0.67 *** 0.82 ** 45 49 0.44 *** 45 49 0.48 *** 0.89 50 54 0.44 *** 50 54 0.49 *** 0.81 55 59 0.37 *** 55 59 0.45 *** 0.63 ** 60+ 0.23 *** 60+ 0.34 *** 0.39 *** none (ref) 1 none (ref) 1 1 roll out 0.90 *** roll out 0.86 *** 1.08 * available 0.54 *** available 0.45 *** 1.34 ** year (cont) 0.96 *** year (cont) 0.95 *** 1.02 **
Concluding remarks The sex ratio of prevalent as well as incident infections increased in conjunction with the expansion of ART programs in these sites and that suggest that women have higher ART enrollment rates than men. We assert that higher ART coverage rates benefit HIV positive women (more than men) in term of increased survival, but benefit HIV negative men (more than women) because of a greater reduction in new infections. In the full paper we will extend the evidence presented here with more detailed statistical analysis and will complement our data with information on ART enrollment in the study sites. References 1. Life expectancy of individuals on combination antiretroviral therapy in high income countries: a collaborative analysis of 14 cohort studies. Lancet, 2008. 372(9635): p. 293 9. 2. Cohen, M.S., et al., Prevention of HIV 1 infection with early antiretroviral therapy. N Engl J Med, 2011. 365(6): p. 493 505. 3. Smith, K., et al., HIV 1 treatment as prevention: the good, the bad, and the challenges. Curr Opin HIV AIDS, 2011. 6(4): p. 315 25. 4. Muula, A.S., et al., Gender distribution of adult patients on highly active antiretroviral therapy (HAART) in Southern Africa: a systematic review. BMC Public Health, 2007. 7: p. 63. 5. Braitstein, P., et al., Gender and the use of antiretroviral treatment in resource constrained settings: findings from a multicenter collaboration. Journal of Women's Health, 2008. 17(1): p. 47 55. 6. Reniers, G., et al., Steep declines in population level AIDS mortality following the introduction of antiretroviral therapy in Addis Ababa, Ethiopia. AIDS, 2009. 23(4): p. 511 8. 7. Reniers, G. and J. Eaton, Refusal bias in HIV prevalence estimates from nationally representative seroprevalence surveys. AIDS, 2009. 23(5): p. 621 9. 8. Venkatesh, K.K., et al., Who gets tested for HIV in a South African urban township? Implications for test and treat and gender based prevention interventions. Journal of Acquired Immune Deficiency Syndromes, 2011. 56(2): p. 151 65.