News from the HC-HIV Front: It’s raining meta (analyses)!

January 27, 2015

The body of evidence related to hormonal contraception and HIV risk is having a growth spurt in 2015. In the beginning of January, Lauren Ralph and her colleagues published a meta-analysis of existing observational data on different contraceptive methods and rates of HIV acquisition among women. We described the approach used in this study in an AVAC blog post and a piece for RH Reality Check.

Less than two weeks later, another meta-analysis has arrived. This one comes from a team led by Charles Morrison of FHI 360 in PLoS Medicine and includes a range of collaborators including some of the researchers involved with the planned ECHO trial that proposes to directly evaluate the relationship between three different contraceptive methods and HIV risk.

The bottom line from this new study is that its findings line up with all of the available data, though with the slight nuances and variations that come with different approaches and data sets (read on for more on this). Overall, they found no evidence that oral hormonal contraceptive pills increase women’s risk of HIV. Once again, their data indicated that Depo Provera (DMPA) may potentially increase women’s risk of HIV acquisition. In this study, the findings were:

  • Across all the data analyzed, DMPA was associated with an increase in women’s risk HIV acquisition by about fifty percent compared to women who were using condoms, diaphragms, sterilization, non-hormonal IUDs or no method of contraception.
  • DMPA was associated with an increase in women’s risk of acquisition by about thirty to forty percent compared to women using combined oral contraceptives or the injectable NET-EN.
  • Most importantly, in studies that the authors evaluated as having less methodological bias, the risk related to DMPA use was lower—only about 20 percent higher than women condoms, diaphragms, sterilization, non-hormonal IUDs or no method of contraception. Methodological bias is a major challenge for observational data, which can measure but not control for factors that might skew the outcome (e.g., the reasons a woman chooses one method versus another might also put her at increased or decreased risk of HIV).

This was one of the handful of evaluations of the relationship between NET-EN, a lower dose injectable contraceptive—and there was a slight association between NET-EN use and increased HIV risk, but it was lower than DMPA, and not statistically significant. This finding may perhaps supporting the theory that alternate dosages and formulations of DMPA might not be associated with an increase in HIV risk.

For advocates who want to get deep into the data to date, here are a few things to note about the new study (Morrison et al) compared to Ralph et al from earlier this month.

Methods and data sources:

  1. The Morrison et al meta-analysis analyzes individual participant data. The researchers who did this review were able to access the data for individual participants from a range of studies. Ralph et al looked at study data and overall findings. This is the way trial data are most commonly presented—e.g., the overall findings across all participants, average ages and so on. Morrison et al, via agreement with various study investigators, went back to the original by-participant data and re-analyzed these findings. (All of the women who had participated in the original studies have given consent for this type of additional analysis to be done, though not this specific study.)
  2. Morrison et al included participant data from a large study of women’s vaginal practices. This study had never published data on the relationship between hormonal contraception and HIV, so it was not been included in the other meta-analyses. Many of the other studies included have also been examined in Ralph et al and in previous systematic reviews. (Our previous blog described the difference between a meta-analysis and a systematic review.)


  1. Morrison et al found that use of DMPA was associated with increased risk of HIV acquisition, but that both the magnitiude and the statistical significance of this finding (eg the degree of confidence that it was real and not a coincidence) risk varied by the quality of the studies considered. Studies with less bias showed less associated risk, and included the possibility of no associated risk. Ralph et al found an overall association but  studies that involved women in the “general population” (not sex workers or women in serodiscordant couples) had lower risk.
  2. Both Morrison et al and Ralph et al found a moderate increase in risk of HIV acquisition associated with DMPA, with a range between 20 percent and 50 percent depending on the subset of data considered. However these findings carry caveats. All conclusions to date have been drawn from observational data and have included the finding that there’s a great deal of variation depending on study quality and/or the population considered.

So, where does this leave us?

Evidence but no certainty, according to Morrison et al, who argue that these data confirm the need for a randomized trial that would seek to eliminate bias and get more clarity on three methods: DMPA, the Jadelle implant and the copper IUD.

AVAC and partners continue to work together to understand these findings, communicate them and to advocate for research, policies and programs that expand women’s range of contraceptive choices in the context of informed choice.