Patterns of Adherence to a Dapivirine Vaginal Ring for HIV-1 Prevention Among South African Women in a Phase III Randomized Controlled Trial

Abstract

Background

Persistent use of HIV prevention methods can be a challenge, particularly for some younger women. The long-acting, discreet, woman-centric dapivirine vaginal ring offers promise as a prevention method with less user burden, which could support continued use. We assessed dapivirine vaginal ring use to understand adherence patterns and identify characteristics influencing patterns.

Setting

Participants enrolled in South Africa in the MTN-020/ASPIRE randomized placebo-controlled trial.

Methods

We used group-based trajectory modeling to identify clusters of participants with similar longitudinal patterns of adherence in last year of participation and potential predictors of group membership. Women with at least one year of follow-up were included (n=626).

Results

Five adherence patterns were identified: (1) consistently high, 34%, (2) consistently moderate, 34%, (3) consistently low, 16%, (4) decreasing, 9%, and (5) increasing, 7%. Women <22 years (adjusted odds ratio [AOR] 1.8, 95% Confidence Interval [CI]: 1.0, 3.0), those using an intrauterine device [AOR 3.3, 95% CI: 1.4, 7.8] or oral contraceptives [AOR 3.9, 95% CI: 1.7, 8.9], experiencing menses [AOR 1.8, 95% CI: 1.1, 3.0] and who reported inconsistent condom use [AOR 1.8, 95% CI: 1.0, 3.3] were more likely to be classified as consistently low compared to consistently high.

Conclusion

Most South African women successfully persisted with moderate or high-level of use. Encouraging ring replacement with completion of menses may help to decrease concerns about hygiene and improve persistence. Associations between contraception and persistent low adherence suggest efforts may be needed to ensure contraceptive method choice does not interfere with ring use.

Introduction

Although currently available HIV prevention methods such as male and female condoms and oral preexposure prophylaxis (PrEP) have proven effective when used as prescribed, persistent use is often cited as a challenge for some women in Sub-Saharan Africa [1,2]. Noted barriers include male partner disapproval and unacceptable dosing frequency [3,4]. A monthly vaginal ring containing 25 mg of dapivirine is a new biomedical HIV prevention technology that may help women overcome difficulties seen with other prevention methods. The dapivirine vaginal ring’s (DVR’s) long-acting, discreet, woman-centric features offer promise for less user burden, which could support persistent use. It has been found to reduce the risk of HIV infection by at least 27% with no safety concerns, and higher adherence is associated with greater protection [5–7]. The DVR has received a positive scientific opinion from the European Medicines Agency and a recommendation from the World Health Organization, and is being submitted for regulatory reviews in sub-Saharan Africa [8,9].

Exploring how women use the DVR over time and gaining a better understanding of challenges with continual use may help inform counseling and implementation programs as the DVR prepares to enter the market in South Africa and other countries globally, pending regulatory approvals. There have been previous reports describing correlates of DVR adherence with demographic, behavioral, and partnership characteristics [10–15]. However, these analyses have been either cross-sectional or with repeat measures and have not explored longitudinal patterns of DVR use. To further understand women’s ability to persist with the DVR it is important to assess use patterns over an extended period.

Our objective was to use group-based trajectory modeling to classify distinct patterns of DVR use over the last year of participation in the MTN-020/ASPIRE trial. In addition, we explored how demographic and behavioral characteristics were associated with adherence trajectories. We focused specifically on the last year to understand how patterns of ring use manifested once a participant had sufficient time to learn and adjust to using a novel prevention product. We hypothesized that groups of women would display identifiable patterns of DVR adherence, and these patterns could be related to sociodemographic characteristics as well as ongoing HIV risk behavior and vaginal practices.

Methods

Study Design

MTN-020/ASPIRE was a phase III randomized, double-blind, placebo-controlled clinical trial conducted across 15 sites in Malawi (2 sites), South Africa (9 sites), Uganda (1 site), and Zimbabwe (3 sites) to evaluate DVR safety and efficacy (ClinicalTrails.gov NCT01617096). Overall, 2,629 women aged 18-45 were randomized in a 1:1 ratio to use either a monthly vaginal ring that contained 25 mg of dapivirine or placebo. Participants completed study visits monthly and were followed for a minimum of 12 months (and for maximum of 33 months). At each visit, participants returned the ring used in the prior month, completed HIV-1 serologic and pregnancy testing, received HIV risk reduction counseling and individualized adherence counseling, completed safety monitoring, and received a new ring. Clinic staff counseled participants to leave the ring in place until the next monthly visit and provided additional rings if participants were not able to attend a regularly scheduled visit. Testing for other sexually transmitted infections (e.g., trichomoniasis, gonorrhea, chlamydia) occurred semi-annually. Additional details of trial procedures have been published elsewhere [5]. Seven months after trial initiation, numerous site-specific participant engagement activities were implemented to encourage high adherence and retention. In addition, blinded monthly aggregate summary data of participants’ adherence were provided to study teams to motivate staff to continue counseling participants to strive toward high adherence [16]. Data collection occurred between August 2012 through June 2015. All sites received institutional review board approval, and all participants provided written informed consent.

Sample Population

Because enrollment initiated at different times by country and sociodemographic characteristics differed by country, we focused our analysis on the country with the largest sample as well as highest burden of HIV—South Africa [17]. Hence, this analysis included participants enrolled at a South African site who were randomized to DVR and had at least 12 months of follow-up.

Outcome: Monthly Adherence

For women randomized to DVR, adherence was measured by residual dapivirine in returned rings (monthly) and by dapivirine levels in blood plasma (quarterly). Our outcome was based on residual drug levels because of the relative frequency of measurements and for its more accurate representation of ring use throughout the entire month. Residual dapivirine was measured using acetone extraction and high-pressure liquid chromatography [18]. Measurements started approximately one year after trial initiation because of procedural changes, and data were not available from the first year of the trial.

We estimated the amount of dapivirine released per day by taking the ratio of dapivirine released (manufacturing load level minus residual drug level) to number of days since dispensation. The release rate was then normalized to represent the estimated amount of dapivirine released during a standard 28-day period. From previous studies, continuous ring use has been equated with 4.0–5.0 mg dapivirine released over a 28-day period [18]. Rings reported lost or not returned (e.g., because of a missed study visit) were assumed nonadherent and to have a release rate of 0 mg/month. If a participant was placed on a product hold (e.g., because of pregnancy), the release rate was considered missing.

Predictors of Trajectory Group Membership

Sociodemographic and risk behavior characteristics, contraceptive method, menstruation, and vaginal practices (insertion of tampons, cloth, etc.) were hypothesized to be associated with adherence patterns based on existing evidence about challenges to ring use [19,20]. Demographic information, including participant’s age, highest level of education, employment status, and marital status was collected on a standardized questionnaire at enrollment. More sensitive topics related to risk behavior and vaginal practices were assessed using an audio computer-assisted self-interview (ACASI). Measures collected via ACASI included engagement in transactional sex (defined by receiving money, material goods, gifts, drugs, or shelter in exchange for vaginal or anal sex in the past year), number of sex partners in the past 3 months, frequency of alcohol use, and insertion of tissue/toilet paper/cloth inside the vagina in the past 3 months. Regular alcohol use was defined by seven or more drinks per week or drinking more than one day per week. Participants’ contraceptive method use was recorded at each monthly visit. At enrollment and each quarterly visit, clinic staff assessed participants’ sexual activity, including number of sexual partners in the past 3 months and whether male condoms were used during every sex act in the past week, as well as if participants had any menstrual bleeding or spotting in the last 3 months. Based on prior findings [15,19,21], we hypothesized that use of the copper intrauterine device (IUD) and oral contraceptive pills (compared to all other methods) may relate to DVR adherence pattern. Given that participants were counseled at every visit to use condoms for HIV prevention as DVR efficacy was unknown, we considered report of condomless sex an indicator of sexual risk behavior. All predictors were included as binary indicators in models and were created from the most recent assessment of each measure prior to the start of the trajectory period.

Statistical Analysis

We used group-based trajectory modeling to identify latent clusters of participants with similar longitudinal patterns of adherence during their last year of study participation. Group-based trajectory modeling is a data-driven approach, whereby the optimal number of trajectories as well as the shape of each trajectory is determined by best model fit to the data [22–24]. First, we fit a series of unconditional trajectory models sequentially increasing the number of groups to a maximum of 10. All trajectory groups were modeled using cubic terms during step one, to allow for greater flexibility in determining the shape of each trajectory. Each model included a covariate for months of study participation prior to beginning of the trajectory period to control for differing follow-up times. Following best practices, model fit was determined by (1) Bayesian Information Criteria (BIC, greater values indicating better fit); (2) average posterior probability of class membership (minimum >0.80); (3) entropy (index of classification based on averaging the largest posterior probabilities for each individual, with values closer to 1 indicating greater precision); and (4) size of smallest trajectory group (>5%) [24]. Second, we considered constant, linear, quadratic, and cubic specifications for each trajectory group, selecting a final model through visual inspection, estimated parameter value significance, and model fit statistics. After selecting the number and shape of trajectories based on the best-fitting model, we then examined whether sociodemographic and behavioral variables were related to membership in a particular adherence trajectory group. Each predictor was individually included in the model, along with months of prior study participation and city of enrollment. Factors found to be associated with group membership (at p<0.05) were then included in a final multivariable model. The largest group characterized by highest adherence pattern was the referent in all models. Chi-squared tests were used to compare characteristics of those with and without 12-months of follow-up. All analyses were completed with Stata version 16 using the traj command [25].

Results

Of the 713 women enrolled in South Africa and randomized to the DVR, 626 (88%) had at least 12 months of follow-up and were included in the analysis. The 87 participants with fewer than 12 months of follow-up did not differ by baseline demographic and behavioral characteristics (all p≥0.36) but were more likely to have seroconverted (n=35; 40% vs. 3%, p<0.001). Most (n=60, 69%) had < 5 rings with available adherence data (primarily because they were enrolled early in the trial, before adherence measurements began).

Table 1 outlines the characteristics of the analytic sample. On average, participants were age 25 at enrollment, with 31% between the ages of 18 and 21 years. About half had completed secondary school (55%), and 38% were earning an income. Nearly all had a primary partner (98%) and were unmarried (93%). Most were parous (84%). The most common form of contraception was injectable (62%); 16% were using a copper IUD, and 12% were using oral contraceptive pills. The majority (62%) had experienced menstrual bleeding or spotting in the past few months. Tampons were not commonly used to manage menstrual bleeding (6%). Before enrollment in the trial, nearly one-fourth of participants’ vaginal practices included inserting paper, cloth, tissue, or rags into the vagina (24%). Sexual behaviors associated with increased risk for HIV were uncommon; very few reported having engaged in transactional sex in the year prior to enrollment (6%) or had multiple sex partners in the 3 months prior to the trajectory period (5%). Likewise, regular alcohol use was rare (7%).

Table 1.

Characteristics of analytic sample: women in South Africa enrolled in the MTN-020/ASPIRE Trial with at least 12 months of follow-up.

	N	(%)
Total	626	(100)
Sociodemographic characteristics
Age at enrollment, years – mean median (IQR)	25.6, 24	(21–29)
Younger, aged 18–21 at enrollment	192	(31)
Secondary education completed	347	(55)
Earns income	239	(38)
Had primary sex partner in past 3 months1	612	(98)
Married	46	(7)
Parous	528	(84)
Body Mass Index1,2
Underweight: ≤18.5	17	(3)
Normal weight: 18.5–24.9	176	(28)
Overweight: 25–29.9	190	(30)
Obese: ≥30	241	(39)
Risk behavior/history
Regular alcohol use3	46	(7)
Concurrent partnership(s), in past 3 months1	32	(5)
Condomless sex1,4	143	(23)
Tested positive for sexually transmitted infection1,5	90	(14)
Transactional sex in year prior to enrollment2	34	(6)
Contraceptive method and vaginal practices
Current contraceptive method1
Injectable	383	(61)
Copper IUD	98	(16)
Oral contraceptive	74	(12)
Implant	28	(5)
Male condom only	7	(1)
Sterilization	25	(4)
None	11	(2)
Menstruation in past 3 months1	390	(62)
Tampons used to manage menstrual bleeding, prior to enrollment	35	(6)
Vaginal practices: tissue/toilet paper/cloth inside the vagina, prior to enrollment	150	(24)
Operational measures
Months of study participation prior to trajectory period – mean, median (IQR)	12, 15	(6-18)
Enrollment location
Cape Town	72	(12)
Durban	456	(73)
Johannesburg	98	(16)

¹Reported at most recent study visit prior to start of trajectory period.

²Missing data: body mass index (N=2); transactional sex (N=8).

³Defined by 7 or more alcoholic drinks per week or drinking alcohol 2 or more times per week.

⁴At least one vaginal sex act without use of a male or female condom in week prior to most recent study visit.

⁵Positive for chlamydia, gonorrhea, or trichomoniasis at most recent study visit.

Half of participants had been participating in the trial for ≥15 months before the start of the trajectory analysis period (interquartile range [IQR] 6–18 months). The analysis included 7,512 monthly study visits, of which 7,165 (96%) had residual dapivirine data available. A five-group trajectory model was selected as the best fit model to describe adherence patterns during participants’ last year of follow-up (BIC=12,256.23; Smallest group N=44; Entropy= 0.88; Table 2). The five groups, depicted in Figure 1, were characterized by (1) consistently high adherence with an average DPV release rate of approximately 4 mg/month (labeled “High,” 34% of participants); (2) consistently moderate adherence (“Moderate,” 34%); (3) adherence that increased from low to moderate (“Increasing,” 7%); (4) adherence that steadily declined from moderate to low (“Decreasing,” 9%); and (5) consistently low adherence (“Low,” 16%).

Table 2:

Fit statistics for trajectory model. Bold text indicates selected model.

K	BIC (N=7165)	BIC (N=626)	AIC	LL	Entropy	Smallest group (%)	Smallest group (N)	Average posterior probability, per group
2	11577.64	11591.05	11615.47	11626.47	0.97	29%	184	0.99	0.99
3	11981.20	12001.92	12039.66	12056.66	0.90	22%	135	0.98	0.92	0.96
4	12086.31	12114.35	12165.40	12188.40	0.87	13%	83	0.95	0.89	0.90	0.95
5	12220.88	12256.23	12320.60	12349.60	0.88	7%	44	0.93	0.93	0.88	0.90	0.94
6	12245.55	12288.21	12365.89	12400.89	0.89	3%	20	0.94	0.93	0.85	0.91	0.90	0.94
7	12245.30	12295.27	12386.28	12427.28	0.84	7%	41	0.93	0.88	0.87	0.91	0.92	0.77	0.83
8	12265.28	12322.56	12426.89	12473.89	0.83	4%	23	0.89	0.83	0.85	0.92	0.77	0.84	0.92	0.81
9	12274.22	12338.82	12456.46	12509.46	0.82	3%	19	0.88	0.83	0.91	0.93	0.86	0.92	0.79	0.81	0.79
10	12254.64	12326.55	12457.51	12516.51		3%	19	0.87	0.84	0.92	0.79	0.85	0.90	0.80	0.78	0.78	0.24

K: Number of groups; BIC: Bayesian information criterion; AIC: Akaike information criterion; LL: Log likelihood

Figure 1.

Estimated group-based trajectories of dapivirine (DPV) vaginal ring adherence during the last year of follow-up among women enrolled in South Africa, with 95% confidence intervals (N=626).

Several participant demographic and behavioral characteristics were associated with adherence trajectory group membership (Table 3). Younger women were more likely to be in the Low or Moderate adherent group compared to High. Those who completed secondary school were more like to be in the Moderate or Decreasing group. Parous women were significantly less likely to be in the Decreasing group and less likely to be Moderate or Low adherers, although this trend was inconclusive (p>0.05). Women who reported having condomless sex were more likely to be in the Increasing, Decreasing, or Low adherence group compared to High. Contraception method was strongly associated with lower adherence, with women who were using an IUD significantly more likely to be in the Moderate, Low, or Decreasing groups, and women using oral contraceptives significantly more likely to be in the Low or Increasing adherence groups. Contraceptive method choice was not consistent over the last year [26], although 85% of women used the same method for at least 9 of the last 12 months of follow-up. In a sensitivity analysis, we assessed how the contraceptive method used most often in the last year related to adherence trajectory group membership. Overall, we found the same associations with slightly decreased effect estimates (data not shown). Women who experienced menstruation were more likely to be Moderate or Low adherers compared to High adherers. Because the proportion of women who had concurrent partnerships was small, model estimates did not converge, and membership probabilities are not presented.

Table 3.

Adjusted odds ratios (AOR) and 95% confidence intervals (CI) for the association between each participant characteristic and adherence trajectory group membership.

	High (34%)	Moderate (34%)		Increasing (7%)		Decreasing (9%)		Low (16%)
		AOR†	(95% CI)	AOR†	(95% CI)	AOR†	(95% CI)	AOR†	(95% CI)	p-value
Younger, age 18-21	(Ref)	1.54	(0.95, 2.51)	0.74	(0.30, 1.83)	0.83	(0.39, 1.79)	1.78	(1.05, 3.01)	0.08
Completed secondary school		1.65	(1.06, 2.57)	1.47	(0.71, 3.04)	2.45	(1.14, 5.26)	1.02	(0.63, 1.65)	0.04
Earns income		0.82	(0.52, 1.30)	1.17	(0.56, 2.47)	1.32	(0.69, 2.53)	1.05	(0.64, 1.73)	0.64
Parous		0.62	(0.33, 1.18)	1.71	(0.38, 7.76)	0.29	(0.13, 0.64)	0.67	(0.33, 1.35)	0.03
Obese: BMI ≥ 30		1.28	(0.82, 1.99)	1.71	(0.82, 3.57)	0.61	(0.28, 1.33)	1.21	(0.74, 1.97)	0.26
Regular alcohol use		1.51	(0.62, 3.68)	2.68	(0.88, 8.21)	1.56	(0.45, 5.43)	1.38	(0.51, 3.76)	0.55
STI positivity		1.00	(0.52, 1.93)	0.76	(0.25, 2.31)	1.17	(0.47, 2.91)	1.42	(0.72, 2.80)	0.78
Transactional sex prior to enrollment		0.17	(0.04, 0.82)	0.61	(0.07, 5.12)	1.30	(0.41, 4.13)	1.49	(0.63, 3.53)	0.11
Condomless sex		1.40	(0.80, 2.45)	3.67	(1.71, 7.90)	2.39	(1.12, 5.14)	1.80	(1.00, 3.25)	0.006
Contraceptive method - IUD		8.17	(3.79, 17.62)	2.51	(0.58, 10.82)	3.58	(1.13, 11.35)	3.33	(1.42, 7.81)	<0.001
Contraceptive method - Pills		2.03	(0.85, 4.84)	4.59	(1.61, 13.07)	0.82	(0.14, 4.91)	3.92	(1.73, 8.91)	0.003
Menstruation in past 3 months		2.29	(1.44, 3.64)	1.89	(0.88, 4.10)	1.23	(0.60, 2.51)	1.78	(1.07, 2.95)	0.007
Vaginal practices prior to enrollment		0.69	(0.41, 1.16)	0.80	(0.37, 1.76)	0.69	(0.28, 1.71)	0.67	(0.37, 1.21)	0.55

Bold P<0.05

^†Each model controlled for months of participation prior to trajectory period and enrollment city

Nearly all significant relationships between participant characteristics and adherence pattern in univariable models held in multivariable analysis, with small changes in the strengths of associations. The only exception was menstruation; odds of membership in the Moderate, Increasing, or Low groups were higher among those who experienced menstruation, although this relationship was no longer significant in the multivariable model. Because of the potential correlation between IUD use and menstruation, we re-estimated the multivariable model without the IUD use indicator and found women who experienced menstruation had greater odds of membership in the Moderate group compared to High adherence group, adjusting for other characteristics (Adjusted Odds Ratio 2.18, 95% Confidence Interval: 1.36, 3.51; p=0.001; data not shown).

Discussion

During the last year of follow-up in the ASPIRE trial, we found evidence of five distinct trajectories of DVR use among women enrolled from South Africa. In general, most women successfully persisted with at least some ring use, with either consistently high or moderate adherence. In addition, we identified a small group with an increasing adherence trend. However, we also identified two groups with lower adherence, one with declining and another with steady low to no ring use. There were several significant associations between demographic characteristics, sexual risk behaviors, menstruation, contraceptive use, and adherence trajectories. Specifically, those who were using a copper IUD or oral contraceptive pills, reported condomless sex, and had menses were less likely to have a high adherence pattern. In addition, as seen in other analyses [5,15], younger women were more likely to be in a lower adherence group. Identifying patterns of ring use over time and the facilitators and barriers of persistent ring use can provide insight in how to support women to continuously wear the ring as directed.

Contraceptive method and menses were strongly associated with adherence pattern. Women using a copper IUD were significantly less likely to be consistently high adherers. Additionally, women who had menses were more like to be moderate or low adherers. From qualitative analyses, women shared that management of menstrual blood and desire to keep the ring clean interfered with their ability to leave the ring inserted [20,27]. The prolonged and increased bleeding during menses and irregular bleeding side effects of the copper IUD [28,29] may make it more challenging for women to wear the ring consistently as directed. Notably, one IUD user described that she removed the ring during menses because her vagina was “overloaded”; with the ring, IUD, and menses, there were “a lot of things down there” [19]. It may be that regardless of menses, women felt the vagina was overburdened with devices, with both the ring and IUD inserted. As we found menstruation was no longer significantly associated with adherence patterns in multivariable analyses, this suggests that correlation between menstrual side effects and contraceptive method may explain the relationship between IUD use and DVR adherence pattern. While our findings are consistent with prior findings in the ASPIRE trial [15], this trajectory analysis highlights the potential long-term influence of contraceptive methods and/or menses on women’s ability to successfully persist with ring use for a year, and therefore be provided with sufficient protection against infection. Strategies to facilitate consistent ring use, such as pairing acceptable contraceptive methods with the ring and encourage timing of ring replacement with the end of menses to decrease concerns of hygiene, may be worth exploring to promote long term ring use.

Women who were using oral contraceptives were more likely to have an increasing or low adherence pattern. Although participants were encouraged to use a long-acting contraceptive while in the trial, women may have chosen to use oral contraception because it is more easily reversible. Indeed, women using oral contraceptive pills were more likely to discontinue contraceptive use altogether during the trial [26]. This might have been because of lower perceived risk of pregnancy or change in pregnancy intentions. In fact, women who were parous were more likely to be in the consistently high adherence group, suggesting that future pregnancy intentions may influence women’s adherence pattern. In a Phase IIIb open-label extension trial of the dapivirine vaginal ring (MTN-025/HOPE) with former ASPIRE participants, where participants were allowed to start or stop using the ring at any time, initial and continued acceptance of the ring was lower among those using oral contraceptives or other short term contraceptive methods [31]. It may be that women with a shorter-acting prevention mindset or with intent to conceive may find oral PrEP a more amenable HIV prevention method since it is approved for use during pregnancy [30]. Women using oral contraceptives were also more likely to be in the increasing adherence group. If oral contraceptive users include women with near-term pregnancy intentions, accumulating ring experience and adherence support intervention during the trial may have reassured women that the ring was safe and may not impact fertility or other health outcome, as we previously described [11]. Thus this may lead to improved ring use over time in this group. Additional investigation is warranted to understand the association between oral contraceptive use and ring adherence patterns, and how fertility intentions may influence women’s persistent and consistent use of the ring.

Inconsistent condom use was related to DVR adherence pattern as well, where women who self-reported not using condoms with every sex act were more likely to be in any of the four lower adherence groups. Although condom use may be highly variable during participants’ lives, particularly over the 1-year trajectory period, we considered report of condomless sex a potential reflection of prevention behavior. In ASPIRE, participants were counseled at every visit to always use condoms because of the unknown protection they may be receiving from the ring. Noncompliance with prescribed condom use may reflect difficulty to comply with instructions, low motivation for HIV prevention, or low personal agency to use a prevention method (e.g., because of partner’s lack of support or fear of intimate partner violence). In a household survey among couples living in Durban, women who perceived themselves to be at risk of HIV infection from their partner were more likely to report condom use [32]. Alternatively, inconsistent condom use could be an indicator for lower autonomy [33], suggesting the need for additional support and/or counseling for how to negotiate ring use with a partner. While the DVR is relatively discreet and about 80% of ASPIRE participants said they never felt the ring during normal daily activities or during sex [34], how perceived HIV risk and/or relationship dynamics may influence women’s ability to use the ring long-term should be explored further.

There are several limitations of this analysis to consider. First, we assessed patterns of adherence within the context of a clinical trial and use patterns may be different in “real-world” settings. The likelihood of persistence may be higher in the trial provided participants received regular counseling to reinforce proper ring use, which may differ from clinical practice where visits may be less frequent. Alternatively, the unknown efficacy of the ring may have influenced women’s desire to use the ring and consequently their adherence pattern. Patterns of DVR use may be different with known efficacy. Nonetheless, we identified several unique groups of adherence patterns, and their correlation with participant characteristics are helpful in understanding who and how women used the ring. Second, we only evaluated trajectory groups within one country. It is possible that patterns of adherence would be different in other countries or regions. However, we decided to focus our analysis within one country to decrease enrollment cohort bias. We also had missing data from the first year of the trial, which impacted our ability to assess how patterns might have differed initially or evaluate a longer follow-up period. Finally, because our analysis only included those who had at least one year of follow-up, we may have excluded participants with lower adherence. This may potentially underestimate the size of adherence patterns or have influenced the number or shape of patterns. Yet, the excluded sample was small (87 participants), so it is unlikely they would have drastically altered the patterns we observed, and adherence data were largely unavailable so we are not able to estimate what impact they might have had on observed patterns.

Overall, we found that most women in South Africa were able to persist with using the DPV in their last year of participation in the ASPIRE trial. However, there were groups of women with patterns that reflected inconsistent or rare ring use. This poor adherence occurred in spite of several months of prior trial participation, during which concerted efforts were made to dispel worries and rumors about the ring and encourage proper use through multiple adherence-related study activities and support clubs [16,35]. While the DVR is a promising new women-centric prevention method, continuous use is critical to its effectiveness. The variable patterns of adherence identified among women using the monthly DVR and the factors associated with such patterns may help to inform targeted counselling messages, to help women select and use an HIV prevention method best aligned with their needs, intentions, partnership dynamics, and life stage.