A sophisticated meta-analysis, pooling individual-level data
on 37,000 women, has found that the use of DMPA injectable hormonal
contraception is linked with a higher rate of new HIV infections in women, the 20th International AIDS Conference (AIDS
2014) in Melbourne heard today. However, the conference was also told that a
large Zambian study has not found any link between contraceptive use and HIV
infection. Furthermore, the World Health Organization (WHO) announced at the
same session that its guideline supporting the provision of this contraceptive
to women at risk of HIV infection remains unchanged.
WHO reviewed research on the issue earlier this year,
concluding that the data that were available at the time did not support
restrictions on the use of injectable contraceptives, and agreed a new version
of its guidance just a few days ago. However, they could not take into
consideration the two analyses released today, nor a number of new studies
which are expected to issue findings within the next year. WHO’s guidance will
likely need to be reconsidered when those data are available.
The potential for some hormonal contraceptives to raise a woman’s risk of HIV infection has been the subject of scientific debate for several years. Around half the available studies suggest an increased risk, but half do not. Moreover, accurately assessing associations between contraceptive use and HIV infection with data from observational studies is challenging and analytically complex. Studies have used inconsistent approaches and generated a body of evidence that is complicated and difficult to interpret.
Furthermore, HIV is not the only consideration when it comes to contraceptive use. The broader health impact of fewer women using effective, long-acting contraceptives could be significant. In settings where maternal mortality is high due to complications of pregnancy or childbirth, effective contraception contributes to better health outcomes in women. Infants of mothers who die in childbirth have poorer health outcomes; closely spaced pregnancies contribute to infant mortality.
Charles Morrison of FHI 360 presented the results of a meta-analysis, which aimed to determine whether the use of specific hormonal contraceptives increases the risk of HIV acquisition compared to women not using hormonal contraception, or using a different form of contraceptive.
Most meta-analyses take the overall results from previously published research and combine each study’s results. The figures used are “aggregate data” – the overall average or estimate derived from each study.
In contrast, this was an individual participant data meta-analysis, a more sophisticated and complex approach. It involved collecting the raw data on all individual study participants from the original researchers and pooling it. This ensures greater consistency in the way the data are analysed (e.g. attention to confounders) and a greater ability to analyse the findings for different subgroups of participants.
Eighteen prospective observational studies were included, with data on 37,124 women. All were conducted in southern or eastern Africa. Reported contraceptive use included 28% using the injectable depot medroxyprogesterone acetate (DMPA, Depo Provera), 8% using the injectable norethisterone enanthate (NET-En, Noristerat), 19% using a combined oral contraceptive pill, and 43% not using any form of hormonal contraceptive. A total of 1830 women acquired HIV while in a study.
Compared to non-users, women using DMPA had an elevated risk of infection (hazard ratio 1.56, 95% CI 1.31-1.86), as did women using NET-En (1.51, 95% CI 1.21-1.90). There was no increased risk for women using oral contraceptives.
Similarly, comparing women using injections with those using oral contraceptives, there was an elevated risk associated with DMPA (1.43, 95% CI 1.23-1.67) and NET-En (1.30, 95% CI 0.99-1.71).
The results were consistent in several subgroup and sensitivity analyses. However, when only studies which were judged to be methodologically more reliable were included, the increased risk appeared smaller.
Charles Morrison did not use his findings to argue for a restriction on the use of injectable contraception in settings with a high prevalence of HIV. He acknowledged the limitations of all observational data and argued that what is needed is a well-conducted randomised controlled trial – something he is seeking funding for.
The findings of the meta-analysis were not replicated in a
study from Zambia, presented by Kristin Wall of Emory University. While the
findings from systematic reviews and meta-analyses are usually considered to
carry more weight than single studies, the Zambian study was methodologically
rigorous and is one of only two large studies to have recruited serodiscordant
The cohort included 1393 couples, each with an HIV-positive
man and HIV-negative woman. The female partner provided information about
sexual behaviour and contraceptive use, and was tested for HIV, every three
months. Couples stayed in the cohort for an average of two years.
During this time, 252 women acquired HIV. Incidence of HIV
infection was 8.4% for women who did not use hormonal contraception, 10.7% for
women using injectable contraceptives, 11.5% for women using oral
contraceptives and 7.3% for women using a contraceptive implant. However, these
apparent differences were not in fact statistically significant and appear to
have been driven by differences in sexual behaviour – for example, women using injectables and oral
contraceptives reported more unprotected sex than women who did not use
Moreover, in several multivariate models which took into
account a range of factors which could skew the results, the risk of acquiring
HIV did not vary according to the contraceptive used. The study did not find
any association between the use of hormonal contraceptives and HIV infection.