Factors Affecting Continued Use of Subcutaneous Depot Medroxyprogesterone Acetate (DMPA-SC): A Secondary Analysis of a 1-Year Randomized Trial in Malawi

Community health workers can adequately provide DMPA-SC directly or train women on self-injection.


BACKGROUND
I njectable contraceptives are increasingly popular in low-and middle-income countries and are the predominant modern method used by women in sub-Saharan Africa. 1 In Malawi, the use of modern contraceptive methods by married women has increased from 7% in 1992 to 58% in 2015-2016; however, unmet need for family planning is considerable at 19%. 2 Injectables are the most commonly used method-of married women in Malawi who use a modern method of contraception, 30% use injectables.Despite their high use, discontinuation rates for injectables are high; 41% of women of reproductive age have reported discontinuing the method in the first year. 2In addition to method-related concerns, travel distance to a nearby health center-over 80% of the approximately 17 million people in Malawi live in rural areas 3 -and frequent contraceptive stock-outs are common barriers to use and continuation. 4he World Health Organization (WHO) has endorsed task sharing as a strategy to bridge the human resource gap in the provision of reproductive health services in low-income countries, noting that "task sharing is envisioned to create a more rational distribution of tasks and responsibilities among cadres of health workers to improve access and cost-effectiveness." 5 When clinicbased providers (CBPs) share tasks with CHWs, the workload of CBPs is reduced, which allows more time for them to provide higher-level care and curative services while increasing access to contraception for women living in hard-to-reach places-thereby helping to address their unmet family planning needs. 6alawi's program for community-based access to injectable contraception started with a pilot in 2008. 7HWs in Malawi (also called health surveillance assistants) provide health services to catchment areas of approximately 1,000 people each; they are the lowest level of paid government workers. 8hey have completed secondary school and received 12 weeks of training-the first 8 weeks are in a classroom, followed by 4 weeks of practical training.As is the case in many other low-and middle-income countries, CHWs in Malawi administer intramuscular depot medroxyprogesterone acetate (DMPA-IM) to clients in community settings as part of the family planning method mix they offer.
A subcutaneous (SC) version of DMPA is delivered in a prefilled, auto-disabled Uniject injection system (Sayana Press with 104 mg of medroxyprogesterone acetate in 0.65 mL suspension for injection).DMPA-SC is steadily gaining popularity among family planning users and providers in sub-Saharan Africa as an easy-to-use and accessible contraceptive option.Studies in Senegal and Uganda found that family planning providers preferred the subcutaneous version over the intramuscular formation-providers indicated that DMPA-SC was easier and faster to administer, would decrease stock-outs (due to its all-in-one presentation compared with DMPA-IM, which requires a vial and syringe that may become separated), and would be less painful and therefore preferable for women. 9Research has also demonstrated that CHWs can safely provide DMPA-SC in community settings. 10,11Moreover, given the simplified delivery system and subcutaneous administration route, a growing body of evidence underscores the feasibility, acceptability, and efficacy of self-injection of DMPA-SC.Self-injection was found to be acceptable and feasible in Senegal and Uganda. 12,13In the current study, a randomized trial recently conducted in Malawi, self-injection improved 12-month continuation rates significantly compared with provider-administered DMPA-SC, by more than 50% (the primary results are reported elsewhere). 11In the Malawi study, both CHWs and CBPs were trained to administer injections and to teach women to self-inject.
The recent trial in Malawi demonstrated that public-sector family planning providers, including CHWs, can safely provide DMPA-SC and train women to self-inject.However, little is known about whether and how outcomes-including continuation, adverse events, side effects, and pregnancy-vary by supply-side factors such as the type of family planning provider (CHW or CBP) who provides DMPA-SC or self-injection training to women.CBPs in the trial were government nurses and midwives with more health care training than the CHWs, which may influence the quality of services provided.Furthermore, the knowledge base is nascent regarding demand-side factors such as the sociodemographic characteristics of women that may influence 12-month DMPA-SC continuation, especially for selfadministered DMPA-SC.0][21][22][23] In contrast, age, marital status, educational level, and parity have not been shown to significantly impact DMPA-IM continuation. 1,17,24,25t is unclear how factors influencing continuation of IM and SC formulations will differ, especially when DMPA-SC is self-administered.Two nonrandomized cohort studies in Burkina Faso and Uganda found no difference in continuation rates between DMAP-SC and DMPA-IM when both were administered by CBPs, but findings showed that increased age and partners' acceptance of family planning increased DMPA continuation in Burkina Faso (no other variables tested with the Uganda data were statistically different). 26In acceptability trials conducted with DMPA-IM clients in Senegal and Uganda, most clients preferred DMPA-SC after trying it; the most common reason for this preference was that clients perceived fewer side effects from DMPA-SC compared with DMPA-IM, 10 though previous safety and effectiveness trials have not demonstrated this difference. 27,28A recent nonrandomized cohort study in Senegal also observed fewer side effects among clients who self-injected DMPA-SC compared with those who received DMPA-IM from a CBP. 29iven the limited research on factors affecting continued use of DMPA-SC, especially for selfinjected DMPA-SC, the aim of this article is to assess the influence of selected supply-and demand-side factors on continued use of DMPA-SC among Malawian women enrolled in a yearlong randomized controlled trial.These data can be used to inform task-sharing decisions and optimize service delivery in Malawi and other lowresource settings.

METHODS
We used data collected as part of a randomized controlled trial we conducted to compare A growing body of evidence shows that self-injection of DMPA-SC is feasible and acceptable and that CHWs can safely provide it.continuation rates between women who selfinject DMPA-SC and women who receive the same product from a provider.The trial was conducted from September 2015 to February 2017 in 6 Ministry of Health clinics and surrounding communities in rural Mangochi District, Malawi.During the trial, CBPs and CHWs randomized 731 women seeking family planning services to either receive DMPA-SC administered by the provider or be trained to self-inject DMPA-SC.][32] Women in the self-injection group who successfully self-injected at enrollment (assessed by the provider) received 3 doses of DMPA-SC to take home for subsequent self-injections, whereas women in the provider-administered injection group were asked to return to the provider for injections at 3, 6, and 9 months post-enrollment.Data collectors (not providers) contacted women after the reinjection window at 3, 6, and 9 months to collect data on discontinuation and women's experiences.Twelve months after enrollment or at early discontinuation, women had their final interview, including pregnancy testing.Neither participants nor study staff were blinded after randomization; however, the statistical team remained blinded until key decisions for the primary analysis were made.A detailed description of the methods of the randomized controlled trial has been published elsewhere. 11ur primary outcome was DMPA-SC discontinuation.Women were considered discontinuers if they did not report receiving an injection within the allowable window of time (12 to 14 weeks after the last injection, according to Sayana Press guidelines).Given that reinjection provides 3 months of protection, participants without a DMPA-SC injection within the window or who were lost to follow-up were considered to have discontinued 3 months after the previous injection.Those who had not discontinued by 12 months were censored at 12 months, when the study ended.
We collected data on adverse events, side effects, and pregnancies occurring throughout the 12-month follow-up.In this article, we report outcomes comparing women assisted by CBPs or CHWs.The analysis of safety data included only participants who successfully received or administered a DMPA-SC injection after randomization.
We estimated Kaplan-Meier cumulative probabilities of contraceptive continuation coverage (with 95% confidence interval [CI] at 3, 6, and 9 months) by provider type and compared the distribution of continuation between these groups using a log-rank test stratified by site using a .05significance level for a 2-sided comparison.We present these results separately by the original randomization group (i.e., self-injection and provider-administered DMPA-SC) since the primary results demonstrated a large treatment effect on continuation. 11We also provide discontinuation incidence estimates and incidence rate ratio with 95% CI comparing the 2 provider types.
We also assessed factors that could potentially influence DMPA-SC discontinuation using Cox proportional hazards models with each of the following covariates: treatment group (i.e., selfinjection or provider-administered DMPA-SC), site (i.e., the health facility catchment area where the participant was enrolled), woman's age, marital status, whether she works outside the home, parity, education, religion, previous experience with contraceptives and injectable contraceptives, and whether a CBP or CHW provided DMPA-SC or self-injection training at enrollment.
We assessed each covariate separately and planned to include all covariates found significant at the .05level in the univariate models in a multivariable model.Except for when we analyzed the effect of site specifically, site was used as a stratification variable in the models as consistent with the randomization scheme.Hazard ratios for discontinuation and 95% CI were provided for each covariate modeled.
The study protocol was reviewed and approved by the Protection of Human Subjects Committee at FHI 360, Durham, NC, USA, as well as the College of Medicine Research and Ethics Committee, University of Malawi.All study staff completed training on research ethics, the protocol, and informed consent administration.All trial participants provided their informed consent to participate.The trial was registered with ClinicalTrials.gov (NCT02293694).

RESULTS
Participants' sociodemographic characteristics are shown in Table 1.Over 70% were enrolled in the study by a CHW.The mean age was 27 years, and 75% had no schooling or did not complete primary school.Over half were Muslim.Almost all were married or had a sexual partner, and 20% said that their husband or partner did not know about their appointment to receive family planning.Almost all had previously given birth and had 3 living children, on average.The large majority (93%) had previously used contraception, primarily injectables.One-quarter did not want additional children.
Cumulative probabilities of continuation and 95% CI for each quarter by type of provider at enrollment and treatment group are presented in Table 2.Among women in the self-injection group, the continuation rate through 12 months of contraceptive use was not significantly different for women who received DMPA-SC selfinjection training from a CBP [0.79 (95% CI, 0.70 to 0.86)] than those who received the training from a CHW [0.70 (95% CI, 0.64 to 0.75)] (P=.77).Though the continuation rates were much lower in the provider-administered group (the self-administered and provider-administered groups had 99 and 199 discontinuations, respectively), we did not find a significant difference between women who received DMPA-SC from a CBP [0.48 (95% CI, 0.39 to 0.57) and those who received the method from a CHW [0.44 (95% CI, 0.38 to 0.50)] (P=.78).The incidence rate of discontinuation for those who received self-injection training from a CHW was 9 per 100 injection cycles (95% CI, 7 to 11) compared with 6 per 100 injection cycles (95% CI, 4 to 9) among those who were trained by a CBP.For the provider-administered group, the incidence rate for those who received DMPA-SC from a CHW was 21 per 100 injection cycles (95% CI, 18 to 25) compared with 19 per 100 injection cycles (95% CI, 14 to 25) among those who received the method from a CBP.
The distribution of reasons for discontinuation did not differ significantly by provider type for selfinjectors (P=.49) or for those in the provideradministered group (P=.26).The most common reason for discontinuing was due to missing the reinjection window (data are reported elsewhere 11 ).Other reasons for discontinuing (in order of decreasing frequency) included loss to follow-up; by the woman's request, mostly related to side effects of DMPA-SC; and less commonly, by the provider's request for medical reasons.The reasons for discontinuation may underestimate the role of side effects during the trial.This is because after women discontinued, they were no longer counted in the estimates of side effect occurrence as the trial moved forward.
Data from pregnancy tests were incomplete due to refusals, loss to follow-up, and data The percentage of continuing women who experienced side effects decreased over time across all groups (Table 3 and Table 4).The differences in percentages of women experiencing side effects among those trained to self-inject by CBPs compared with those trained by CHWs were not statistically significant: 3 months-20% vs. 28% (P=.21), 6 months-15% vs. 18% (P=.74), and 9 months-12% vs. 14% (P=.70).Similarly, there were no statistically significant differences between those who received DMPA-SC from a CBP compared with a CHW: 3 months-34% vs. 31% (P=.61), 6 months-23% vs. 22% (P>.99), and 9 months-15% vs. 19% (P=.69).Among women who reported side effects, the majority across all groups reported little to no effect on daily life.
Twenty related or possibly related adverse events were reported by 10 women in the selfadministration group (data not shown).Nine of these events were reported by 3 women who received self-injection training by CBPs and 11 of these events were reported by 7 women who received training by CHWs.These differences by type of provider were not statistically significant (P=.73 for the differences in proportion of women experiencing adverse events).Twentyeight related or possibly related adverse events were reported by 17 women in the provideradministered group (data not shown).Nine of these events were reported by 7 women who received DMPA-SC from CBPs and 19 of these events were reported by 10 women who received DMPA-SC from CHWs; these differences were not statistically significant (P=.28 for the differences in proportion of women experiencing adverse events).Furthermore, there were no significant differences between the groups in the types of adverse events reported.There were 5 serious adverse events reported during the trial by 4 different women.Two events related to DMPA-SC (menorrhagia and anemia requiring hospitalization) were reported by the same woman in the provider-administered group who was enrolled by a CHW and resolved without sequalae.The other serious adverse events, including 1 death (suspected liver cirrhosis), were unrelated to DMPA-SC.The percentage of women experiencing side effects were not significantly different by provider type.
The results of the Cox model are presented in Table 5.Only treatment group and health facility catchment site were statistically significant predictors of continuation; therefore, no additional multivariable analyses were conducted.Consistent with the primary analysis reported elsewhere, we found that women in the self-injection group were significantly less likely to discontinue compared with women in the provider-administered group (hazard ratio, 0.43; P<.001).Risk for discontinuation was also different among clinics (P<.001).

DISCUSSION
Contraceptive continuation is important for reducing unintended pregnancies.This is one of the first studies to explore factors that affect continued use of DMPA-SC through 12 months, Very little 0 (0.0) 5 (6.9) 5 (5.5) 0 (0.0) including self-administered DMPA-SC.A retrospective study of 2015-2016 Demographic and Health Survey data in Malawi found a 12-month discontinuation rate of 41% for injectable contraceptive users. 2 In this prospective trial, the discontinuation rates of DMPA-SC through 12 months were 52% and 56%, for clients who received the injections from CBPs and CHWs, respectively.The discontinuation rates for self-injecting clients trained by CBPs and CHWs were substantially lower, 21% and 30%, respectively.The differences in the continuation rates by provider type (CBPs and CHWs) were not statistically different for either self-administered or provider-administered DMPA-SC.We did not find evidence that the type of provider influenced the risk of discontinuing, pregnancy, or safety, which suggests that CHWs-not only CBPs-can provide DMPA-SC or training on self-injection in low-resource settings without  Don't know 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) (0.0) hampering continuation.Wait times at health facilities are often long and health facilities are often overcrowded and understaffed.Permitting CHWs to train women on DMPA-SC self-injection in community settings would enable women to circumvent the long lines and alleviate some of the work load at these health facilities.CHWs are based in rural and low-income areas where there is often high unmet family planning need, and they are more likely to remain in their communities once trained. 33In 2009, WHO concluded that CHWs can safely and effectively administer injectable contraceptives in non-clinical settings. 34CHW provision of injectable contraception was once innovative but is now a standard of practice.Our results add to the body of evidence supporting task sharing and CHWs' potential to increase access to contraception and reduce unmet family planning needs, despite lower levels of training. 6Based on the evidence, self-administered and provider-administrated DMPA-SC should be scaled up in community settings using CHWs.
Of the factors explored, treatment group and health facility catchment site were the only factors that significantly influenced the risk of discontinuation, with self-injection leading to a reduced risk of discontinuation compared with provider administration.Consistent with previous studies of DMPA-IM, 1,17,24,25 we did not find evidence that sociodemographic factors influenced DMPA-SC continuation.Importantly, we found that education levels did not affect women's ability to selfinject.Most women enrolled in the study had very little education and could inject on time and continue using DMPA-SC during the year-long trial.
Our findings are consistent with findings from nonrandomized prospective cohort studies in Senegal and Uganda, which observed that clients who self-injected DMPA-SC had a lower risk of discontinuing relative to clients who received DMPA-IM from CBPs. 29,35However, our findings differed from these studies in that they observed several other variables-some that we included in our model and some we did not-that influenced DMPA continuation.In Uganda, rural location and being younger increased discontinuation risk, whereas having a primary or greater education (versus no education), more children, and partner support for family planning increased continuation.In Senegal, paying for travel to the clinic and experiencing side effects increased discontinuation risk, whereas having more education, children, and household assets increased continuation.It may be that our site variable encompasses other underlying factors, such as rurality, which are not otherwise included in our model.
Our results are also similar to a study of provider-administered DMPA-SC in urban Nigeria that found no differences in continuation at 3 months according to the place women received DMPA-SC. 36In that study, data were collected from a convenience sample of users who obtained DMPA-SC from selected privatesector providers working in hospital, clinics, and retail drug outlets, as well as licensed Community Health Extension Workers.Unlike our study, the Nigeria study found differences in sociodemographic characteristics: women with some college education or more and those with 4 or more children were more likely to obtain another dose at 3 months.They also found that quality of counseling and side effects influenced continuation.
One limitation of our study is that women's reported outcomes may have been influenced by social desirability bias.Another challenge we faced was missing data for the pregnancy outcome.Given this, our pregnancy data should not be used for estimating the DMPA-SC failure rate.The study was also not designed to assess whether women with different characteristics or being assisted by various types of providers had different risks of discontinuation; therefore, the sample size for some of the comparisons may be too small to be conclusive.Furthermore, these are non-randomized comparisons and may be affected by selection biases.Lastly, there are numerous other variables and combinations of variables that we did not explore but which may influence continuation.
Although we did not find any of the sociodemographic factors associated with DMPA-SC discontinuation to help us target future efforts, the differences observed across sites may indicate the presence of other underlying factors that would be interesting to explore in future studies.For example, providers' management of clients who would like to continue using injectables but arrived late for their scheduled reinjections (although still within the grace period) has been documented to vary and to directly affect clients' continued use of contraception. 37Understanding the context and other characteristics of the populations served by these sites is important, but further exploration is not possible in our study due to our sample size and data contents.Despite the site differences, the positive effects of selfinjection were present in all sites, which speaks of the robustness of this finding across contexts and further supports our recommendations for scaling up DMPA-SC.Implementation challenges will need to be addressed to make this recommendation possible, including resources and planning for training and advanced provision of commodities for self-administration; however, the introduction and scale-up of this new evidence-based approach addresses the severe shortage of family planning providers and the persistent problem of DMPA discontinuation.We urge WHO and the global health community to expand their endorsements of CHW provision of injectables to include CHW provision of DMPA-SC for self-injection.

TABLE 1 .
Continued Abbreviations: No., number; SD, standard deviation.a This question purposefully excludes 25 women who were married but were not living with their husband and had no other regular sexual partner.

TABLE 2 .
Cumulative Probability of Continuation Among Self-Administered and Provider-Administered Clients, Stratified by Type of Provider at Enrollment Abbreviation: CI, confidence interval.

TABLE 3 .
Experience With Side Effects in Last 3 Months Among Self-Administered Participants, Stratified by Type of Provider at Enrollment, No. (%)

TABLE 4 .
Experience With Side Effects in Last 3 Months Among Provider-Administered Participants, Stratified by Type of Provider at Enrollment, No.(%)

TABLE 5 .
Baseline Factors That May Influence DMPA-SC Discontinuation (N=731) Abbreviations: CI, confidence interval; DMPA-SC, subcutaneous depot medroxyprogesterone acetate; vs., versus.a Sample size for each factor varied due to missing values.b Except for site, the univariable models for all other factors were stratified by site.