Nathan Geffen, TAC

The South African Department of Health released the 2008 antenatal survey results in October 2009, later than usual. Nevertheless substantial improvements on last year’s report made up for the delay. The survey found that HIV prevalence in pregnant women aged 15 to 49 using public antenatal facilities is 29.3% (95% CI: 28.5%-30.1%). This is statistically unchanged from the 2007 survey which calculated a prevalence of 29.4% (95% CI: 28.5-30.1). [1]

Just under 34,000 women at about 1,450 facilities in all 52 health districts were anonymously tested for HIV and syphilis, a sample size almost identical to the 2007 survey. The survey sample size has increased dramatically in recent years; in 2005 only 16,500 people in just under 340 clinics were sampled. About 130 samples were excluded because of missing data. Testing was unlinked and carried out using an ELISA assay (Abbot Axsym System for HIV-1 and HIV-2). Active syphilis was screened using a rapid plasma reagin test.

Incorrect estimates of 2007 antenatal survey corrected

The presentation of the survey is much improved. It provides a breakdown of prevalence at national, provincial and district levels (as it did last year). It also provides details of district prevalence for 2007 and 2006 making comparison relatively easy. Critically, it corrects a problem from last year’s survey raised by Dorrington and Bourne in a letter to the South African Medical Journal (SAMJ). [2]

The problem was as follows: In 2007 survey age weights were introduced, so that the age profile of the weighted sample would match that of the general female population, not the population of pregnant women. Pregnant women have a different age profile from that of women in the general population, and since the antenatal survey is a survey of pregnant women, it is illogical to introduce age weights to match the general population. Furthermore, the Department of Health did not explain that they had changed the weights, so it was not apparent that the reduction in prevalence was actually due to the introduction of the age weights.

Consequently the 2007 provincial and national estimates were wrong. As Dorrington and Bourne pointed out, the survey stated, “South Africa may be making some real progress in its response to the HIV epidemic” and the “South African HIV epidemic is on a downward trend”, even though it provided no evidence to support this because its comparison with data from previous surveys was invalidated by the incorrect calculation of the 2007 survey prevalence.

In December 2008 Dorrington and Bourne published another letter with corrected calculations of the provincial and national prevalence rates. The corrected prevalence was higher; while the survey reported a national prevalence of 28%, Dorrington and Bourne calculated that it was 29.4% up from 29.1% reported in 2006. This debunked the notion that there was evidence from the antenatal survey that HIV prevalence was “on a downward trend.” [3]

The publication of the first Dorrington and Bourne letter coincided with the appointment of Barbara Hogan as health minister, replacing Manto Tshabalala-Msimang. Dorrington and Bourne’s concerns were taken seriously and there was a commitment to correct the antenatal report. This was done by publishing the corrected calculations for the 2007 survey in this year’s report.

Main results

Table 1 lists HIV and syphilis prevalence by province and nationally for 2008 and 2007.

Table 2 provides a breakdown by age group of HIV infection. Note that 138 people under 15 and five people over 49 were tested.

ADD TABLE

Table 1: Antenatal HIV and syphilis prevalence for 2007 and 2008 by province and nationally. [1]

Table 2. Antenatal HIV prevalence for 2007 and 2008 by age. [1]

The survey includes a discussion on estimating population prevalence. It presents estimates from UNAIDS models, the Actuarial Society of South Africa’s ASSA2003 model and the Human Sciences Research Council (HSRC; see our review of this survey in HTB-South April-June 2009).

The Department of Health uses the UNAIDS model to calculate prevalence for the total population. This actually consists of two models, the Estimation and Projection Package (EPP) and Spectrum. EPP estimates prevalence based on antenatal and national population survey data. Spectrum uses the output of EPP (ie the population HIV prevalence) and other demographic data to estimate incidence, AIDS deaths, orphans due to AIDS and HAART and PMTCT needs. Some key outputs from all three sources are presented in Table 3.

ADD TABLE

Table 3: HIV demographic data for 2008 from three sources. Values in square brackets are 95%CI. [1]

COMMENT

The correction of the 2007 results is welcome and emphasises the change in health politics since the appointment of Hogan and subsequently Aaron Motsoaledi. Interestingly there has been no explicit acknowledgement of the correction; the new report simply used the corrected data and last year’s report appears to have been removed from the Department of Health’s website.

The close correspondence in prevalence estimates between UNAIDS, ASSA and HSRC is notable, though this has been the case for several years. This raises confidence that the estimates of HIV prevalence in South Africa are accurate.

Until 2001, the antenatal survey was the only large population sample of HIV prevalence in South Africa. Consequently it was the main data source for estimating HIV indicators including prevalence and incidence for the whole population. Since then three HSRC household surveys have been published in 2002, 2005 and 2008. Despite some problems, they have added considerably to our understanding of the South African epidemic.

While prevalence is an important measure, incidence is at least as important. For one thing the HIV/AIDS National Strategic Plan (NSP) commits to reducing incidence by 50% between 2007 and 2011. Unfortunately incidence is extremely difficult to calculate (see our review of the HSRC survey in HTB South April-June 2009). The relationship between population incidence, population prevalence and antenatal prevalence is complex, even more so since the wide-scale introduction of HAART in 2004. These factors mean that the antenatal surveys are not quite as critical as they once were. Furthermore, the doubling of the antenatal survey sample size in 2006 and the consequent change of methodology means we have to treat comparisons between surveys pre-2006 and since 2006 with caution. Nevertheless, they continue to be a useful source of data.

The one group in the antenatal survey that can be used to gauge incidence, at least partially, is 15-19 year-olds. Since they are likely to be pregnant for the first time the women constituting this part of the sample are unlikely to overlap significantly with a previous survey. HIV-positive women in this age group generally constitute a group of recent sero-converters. Here the news is not promising. There is no statistical change in prevalence in this group between 2002 and 2008. There is also no change in the proportion of their contribution to the sample (19.4% in 2002, 19.4% in 2008). Again, comparisons between surveys before 2006 with those since must be treated cautiously.

References
1. Department of Health. National antenatal sentinel HIV & syphilis prevalence survey 2008.
http://www.doh.gov.za/docs/nassps-f.html
2. Dorrington R and Bourne D. Has HIV prevalence peaked in South Africa? – Can the report on the latest antenatal survey be trusted to answer this question? SAMJ. October 2008, Vol. 98, No. 10.
http://www.samj.org.za/index.php/samj/article/view/2885/2153
3. Dorrington R and Bourne D. Re-estimated provincial HIV antenatal survey prevalence for 2007 and a reinterpretation of the national trend. SAMJ. December 2008, Vol. 98, No. 12.
http://www.samj.org.za/index.php/samj/article/view/3024/2154

Thanks to Leigh Johnson for reviewing this article.