Depot contraception and HIV: an exercise in obfuscationBMJ 2019; 367 doi: https://doi.org/10.1136/bmj.l5768 (Published 07 October 2019) Cite this as: BMJ 2019;367:l5768
- C Sathyamala, academic researcher
In June this year, the Lancet published results of a randomised trial by the Evidence for Contraceptive Options and HIV Outcomes (ECHO) consortium.1 The main purpose of this trial, according to the consortium, was to settle the long controversy over whether girls and women using intramuscular depot medroxyprogesterone acetate as a contraceptive are at increased risk of HIV infection.
The open label trial, conducted in South Africa, Kenya, Zambia, and Eswatini (Swaziland), compared the risk of HIV associated with three contraceptive methods: depot medroxyprogesterone, a copper intrauterine device (IUD), and a levonorgestrel subdermal implant.2 The authors noted that throughout the trial, HIV incidence, the primary endpoint of the study, was “alarmingly high.” Yet, because the differences were not judged statistically significant—taken in this study as P<0.04—the authors concluded that there was “no link between HIV and the methods tested.”3
In keeping with this interpretation, the World Health Organization’s guideline development group recommended that intramuscular depot medroxyprogesterone be placed in category 1 of its medical eligibility criteria: “safe for use without restrictions.”4 It had previously been given the more cautious category 2 status: “advantages … generally outweigh the possible increased risk of HIV acquisition.”5
However, closer scrutiny of the trial results does not support the new guidance. The overall incidence of HIV among participants was 3.81/100 woman years. HIV incidence was high irrespective of the contraceptive method used: 4.19/100 woman years in the medroxyprogesterone group, 3.94/100 woman years in the copper IUD group, and 3.31/100 women years in the implant group.
Furthermore, when intramuscular medroxyprogesterone was compared with the levonorgestrel implant, the hazard ratio for HIV acquisition on continuous use analysis was 1.29. This increase was dismissed as “not significant” because the P value was 0.06, though just above the conventionally accepted 0.05 threshold for significance and the more unusual 0.04 cut-off used in this trial.
This reliance on P values runs counter to recent debates on the value of significance tests in interpreting causality.67 The term “dichotomania” has been coined to describe the practice of dividing trial findings into statistically significant or not; a confidence interval is more informative because confidence intervals “that contains the null value will often also contain non-null values of high practical importance.”8 In this trial, though the 96% confidence interval (0.98 to 1.71) around the hazard ratio comparing intramuscular medroxyprogesterone and levonorgestrel implant includes 1, the direction of effect indicates the probability of substantially increased risk. It is unsurprising that this difference was not found to be significant because the ECHO trial was not powered to detect differences less than 50%.9
The consortium justified a 50% difference as the cut-off most meaningful for policy. However, a modelling exercise showed that even with a risk difference of 20% (relative risk 1.2), reducing the use of intramuscular medroxyprogesterone would benefit public health in southern Africa.10 A correction to the ECHO study acknowledges that, “even a relatively small effect might be important in contraceptive and HIV prevention decision making.”11 Inexplicably, WHO’s new guidance fails to report the difference in risk between medroxyprogesterone and the levonorgestrel implant.
The design of the ECHO trial assumed that the two control treatments—the copper IUD and levonorgestrel implant—did not increase HIV risk based on evidence rated as low quality by the guideline development group. Given high rates of HIV in the three randomised groups, and the lack of a barrier method control, one possible interpretation is that all three methods are associated with an increased risk of HIV in high prevalence populations, with intramuscular medroxyprogesterone having an edge over the other two.
Basing their recommendations on the ECHO trial’s conclusions suggests a lapse in scientific rigour by WHO’s guideline development group.12 Not taking account of ethical concerns about this trial is another opportunity lost.13
WHO’s current guidance dilutes its previous advice to “also always use condoms” while using intramuscular medroxyprogesterone14 and instead recommends offering “pre-exposure prophylaxis (PrEP) in settings where the incidence of HIV is above 3%.”15 This recommendation lends credence to the charge made by women’s leaders from 12 African countries, who observed in an open letter to WHO in July: “It seems clear to us that for African women and girls, it has been decided that a pregnancy is worse than an HIV infection.”16
Evidence from the ECHO trial, together with that from earlier observational studies, strengthens the contention that using intramuscular medroxyprogesterone increases girls’ and women’s risk of acquiring HIV. It also raises questions about the other two contraceptive methods. The wider scientific community now has a responsibility to call for a review and reanalysis of all data from the ECHO trial. Until then, WHO’s new guidance should not be accepted.
Competing interests: The BMJ has judged that there are no disqualifying financial ties to commercial companies. The author declares the following other interests: I have written extensively and advocated against the use of injectable contraceptives in the national family planning programme of India. I helped draft a petition against the injectable contraceptive NET-EN that was filed in the Supreme Court of India in the 1980s on behalf of women’s organisations. These have been evidence based and none were paid activities. I interact on an informal basis with women health advocates from outside India as well, including some of the signatories whose letter to WHO has been quoted in this editorial. The BMJ’s policy on financial interests is here: https://www.bmj.com/sites/default/files/attachments/resources/2016/03/16-current-bmj-education-coi-form.pdf.”
Provenance and peer review: Commissioned; not externally peer reviewed.