Menu
Menu
Menu
Menu
Introduction: Some observational data suggest that the progestogen injectable contraceptive depot medroxyprogesterone acetate (DMPA) may increase a woman’s risk of HIV acquisition but a randomized clinical trial did not find a statistically significant increase in HIV risk for women using DMPA compared to two other methods. However, it could not rule out up to 30% increased HIV risk for DMPA users. We evaluate changes to contraceptive method mix in South Africa under different assumptions about the existence and strength of a possible undetected relationship between DMPA use and HIV risk. Methods: A mathematical model was developed to simulate the ongoing HIV epidemic and contraceptive method mix in South Africa to estimate how changes in method mix could impact HIV- and reproductive health-related outcomes. We made different assumptions about the relationship between DMPA use and HIV risk, from no relationship to a 30% increase in HIV risk for women using DMPA. Scenario analyses were used to investigate the impact of switching away from DMPA predominance to new patterns of contraceptive use. Results: In South Africa, the HIV-related benefits of reduced DMPA use could be as great as the harms of increased adverse reproductive health outcomes over 20 years, if DMPA did increase the risk of HIV acquisition by a relative hazard of infection of 1.1 or greater. A reduction in DMPA use among HIV-positive women would have no benefit in terms of HIV infections, but would incur additional negative reproductive health outcomes. The most important driver of adverse reproductive health outcomes is the proportion of women who switch away from DMPA to no contraceptive method. Conclusions: If there is any real increased HIV risk for DMPA users that has not been detected by the recent randomized trial, a reduction in DMPA use could reduce the ongoing number of new HIV infections. However, such a change would place more women at risk of adverse reproductive health effects. It is imperative that these effects are minimized by focusing on expanding access to safe, effective and acceptable alternative contraceptive methods for all women.
We calibrated a deterministic dynamic transmission model to represent the ongoing HIV epidemic and changing contraceptive method mix over time in South Africa using data on age‐ and sex‐specific HIV prevalence, total HIV incidence, incidence among high‐risk women, and age‐specific contraceptive method mix [18, 19]. The model has been described in detail elsewhere and full details are provided in the Supporting Information [20, 21]. Key features of the model include representation of the population age structure, sexual behaviour (sex acts, condom use, partner change rates), HIV transmission and natural history, rollout of the antiretroviral therapy (ART) cascade, prevention interventions (male circumcision, expanded condom use), contraceptive method mix and reproductive health outcomes. We assigned women to a type of contraception based on current method mix by age, and assumed that this pattern of use has held constant since the beginning of the HIV epidemic (Table 1). Many methods, including injectables, oral pills and female sterilization, have been used at broadly similar levels since the first Demographic and Health Survey in 1998, for example injectables (both DMPA and NET‐EN) were used by 27% of all women in 1998 and 23% in 2016 [16, 22]. We parameterized the effectiveness of each method assuming the quoted “typical use” contraceptive efficacies and one‐year continuation rates, recognizing the limitation that in reality “typical use” could be country specific [18, 23]. Model contraceptive parameters Model outputs include demographic, HIV‐related and reproductive health outcomes encompassing morbidities and mortality stemming from both HIV infection and unintended pregnancy. All modelled health outcomes are jointly summarized as disability‐adjusted life‐years (DALYs). For HIV‐related outcomes, these include the different stages of HIV infection and ART use, and for reproductive health outcomes, these include morbidities associated with unsafe abortion and complications of labour (haemorrhage, puerperal sepsis, eclampsia, obstructed labour), and mortality resulting from unsafe abortion and maternal deaths. All model simulations were run for 20 years and the results summed for that period without discounting. Two sources of uncertainty are included in the model outputs. First, several parameters defining the HIV epidemic were fitted (per sex act probability of HIV transmission, sexual mixing rates between different behavioural risk groups, the size of these risk groups and the start time of the epidemic) by running the model 20,000 times and using a filtration method to select the 100 most acceptable epidemic fits, and the analysis was repeated sampling from each of these parameter sets. Second, we sample the maternal mortality ratio (MMR) in South Africa from a log‐normal distribution constructed using a point estimate and associated 95% CI [24]. The association between DMPA use and HIV acquisition risk in the model was varied between an HR of 1.0 (representing no association, consistent with the ECHO trial results) to 1.3 (representing a 30% increase in HIV risk for women using DMPA, consistent with both the ECHO trial results and a meta‐analysis of observational data [5, 6]) at increments of 0.1, and all analyses were repeated under each assumption. We do not consider the possibility that HR < 1 because this would not result in any undetected ongoing excess HIV risk. In the absence of data, we assume that the IUD and levonorgestrel implant have no effect on risk of acquiring HIV. Similarly, we assume that NET‐EN does not affect HIV acquisition risk. The analysis compares the HIV‐ and reproductive health‐related outcomes in a 20‐year period that could result from changes in the contraceptive method mix from 2019. Eighteen different scenarios, described in Table 2, were defined by stakeholders that vary in respect of: Analysis plan Eighteen different scenarios are constructed by combining each of the three magnitudes of migration options (Panel A) with the six contraceptive replacement options (Panel B). DMPA is replaced over three years from 2019 onwards. Baseline: no change in DMPA use. These scenarios are not intended to predict what will happen, but rather to illustrate key relationships and their potential outcomes over a wide range of assumptions. For example it is unlikely that DMPA availability would ever be restricted for HIV‐positive women, who have limited alternative contraceptive options [25]. Finally, we used an analysis of the 2016 South African HIV Investment Case to infer the cost at which each scenario would be cost‐effective [26]. That analysis implied that HIV interventions that lead to saving life‐years at a cost of $547 to 872 per life‐year saved (LYS) would be at the margin of cost‐effectiveness based on the set of interventions being funded under the current budget. We therefore multiplied the net health impact (in LYS) of each scenario by the cost range above to give the total affordable cost of an intervention to implement each switching scenario.
