Introduction

While remarkable progress has been made in reducing overall HIV incidence, young women still bear a disproportionate burden of new infections [1,2]. In sub-Saharan Africa (SSA), adolescent girls and young women (AGYW) represent less than 10% of the population yet account for 1 in 5 new HIV infections [1]. Oral pre-exposure prophylaxis (PrEP) with tenofovir disoproxil fumarate/emtricitabine (TDF/FTC) has proved to be a highly effective HIV prevention intervention. Clinical trials and program settings have demonstrated roughly 95% efficacy in men who have sex with men (MSM) who adhere to daily or on-demand dosing [3–7]. In contrast, daily oral PrEP for HIV prevention among women has been less successful, in large part due to low adherence [8–10]. Obstacles including lack of partner support, need to conceal PrEP use from partners, lack of privacy from family members, and low empowerment hinder effective daily oral PrEP use in women, particularly in SSA [11,12].

Injectable long-acting cabotegravir (CAB-LA) provided high-level protection in clinical trials in both MSM and transgender women in the Americas, Asia, and Africa [13] and women in SSA [14] and was found to offer superior HIV prevention relative to oral daily TDF/FTC. In the study in SSA, analysis of dried blood spot (DBS) samples from women in the oral TDF/FTC arm, throughout 1.2 years median follow-up in this trial, showed undetectable levels in 38%, and only 18% with concentrations consistent with 4+ doses/week. That is, women were persistent in visit attendance but not adherent to daily pill use. In contrast, injection coverage was 93% of person-years.

These results expand the possibilities for more effective PrEP use among women. Importantly, many women in clinical trials of daily oral PrEP in SSA continued to attend study visits but did not adhere to daily dosing. Thus, this new modality provides an avenue for more successful HIV prevention, much as implants and injectable contraception offer greater success for pregnancy prevention among women in SSA [15]. A critical next step will be bridging the CAB-LA clinical trial results to real-world settings, especially since prior studies suggest that PrEP visit continuation among AGYW is much higher in clinical trials than in programmatic settings (public and private clinics) [16].

Understanding women who are persistent but not adherent and providing them with long-acting formulations may be a more effective way to reduce population-level HIV incidence. Data from programmatic settings are important for developing effective implementation strategies for these women but such data very rarely include both program persistence, which focuses on ongoing visit attendance [17–19] and objectively measured adherence. To address this, we leveraged data from the PEPFAR-funded DREAMS initiative which offers daily oral PrEP to AGYW within a program that supports HIV prevention activities. Nesting a study in this context allowed collection of data on PrEP persistence as well as self-reported and objectively measured adherence in a program setting.

Our analysis focuses on Kisumu and Homa Bay Counties of western Kenya, where PATH implements the DREAMS Initiative. About 10,500 new HIV infections were reported in young people ages 15 to 24 years in these 2 counties in 2015, accounting for 14% of the national HIV incidence [20]. As part of the DREAMS implementation, an extensive monitoring and evaluation system (DREAMS database) collects comprehensive data on each AGYW and her participation in the DREAMS program, including data on PrEP initiation and refill visits.

By November 2017, an estimated 1,060 AGYW in Kisumu and Homa Bay Counties had enrolled in the DREAMS PrEP program. The current analysis examines the PrEP use cascade as well as predictors of PrEP persistence and adherence (by self-report and drug levels) in a subset of these AGYW enrolled in a prospective study. The main questions examined were the extent of program persistence (a measure focusing on visit attendance) and the level of protection from HIV infection among program attendees as assessed by measuring PrEP metabolites in blood.

Methods

Results

This study recruited 359 AGYW who were enrolled in the DREAMS PrEP program. Of the 359 AGYW, 336 reported currently taking PrEP and provided a DBS at Interview 1; these 336 are the focus of this analysis. Their mean age recorded in DREAMS program data was 22 years; roughly half (45%) resided in Kisumu County, the remainder in Homa Bay County. At Interview 1, AGYW had been in the DREAMS program for 2.5 years on average and in the PrEP program for a median of 112 (IQR: 90,172) days; the majority (176/336, 52.4%) had initiated PrEP use 4 to 6 months earlier, and 42% were living with their partner, 16% reported multiple sexual partners, 37% believed their partner had other partners, 49% believed their partner’s behavior put them at risk, 73% reported inconsistent or no condom use, and 29% reported that their partner’s status was HIV+ or unknown (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of 336 study participants at the first interview.

https://doi.org/10.1371/journal.pmed.1004097.t001

Of the 336 AGYW taking PrEP at enrollment, 302 attended Interview 2, with 18 lost to follow-up and 16 censored. At Interview 2, 105 said they had discontinued PrEP and 197 said they were continuing PrEP (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study flow chart showing study design and the attrition of participants at each step of the analysis.

Most (246/302, 81%) AGYW interviewed at follow-up had attended at least 1 PrEP dispensation visit between the 2 interviews. Fifty-six had no record of attendance at interim PrEP refill visits, including 21 of the 197 AGYW who reported they were continuing PrEP. This left 176 of the AGYW who attended Interview 2 who met our definition of PrEP persistence (176/302, 58.3%, [95% CI 52.7–63.8] (Figs 1 and 2A and Table A in S1 Table). AGYW, adolescent girls and young women; CI, confidence interval; DBS, dried blood spot; PrEP, pre-exposure prophylaxis.

https://doi.org/10.1371/journal.pmed.1004097.g001

The median time between the 2 interviews was 93 (IQR [91, 99]) days. Among AGYW whose TFV-DP levels at enrollment indicated they were taking at least some PrEP (>10 fmol/punch), a high proportion were persistent (52/78: 66.7%, [95% CI 56.2 to 77.1]) (Fig 2B, Table B in S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. PrEP Cascade.

(A) All participants. (B) Participants with detectable TFV-DP levels (10+ fmol/punch) at the first Interview. T1: first interview visit (Time point 1). T2: second interview visit (Time point 2). The numbers of participants in the categories of TFV-DP levels are shown in S1 Table. PrEP, pre-exposure prophylaxis; TFV-DP, tenofovir-diphosphate.

https://doi.org/10.1371/journal.pmed.1004097.g002

While visit attendance suggested high persistence (Figs 1 and 2B) and self-reported PrEP use was high, actual adherence was low with very few AGYW taking sufficient PrEP to be protected from HIV acquisition. At Interview 1, 298/336 (89%, [95% CI 85 to 92.1]) AGYW reported taking 4 or more (4+) doses in the past 7 days. However, only 21/336 (6.3%, [95% CI 3.7 to 8.8]) had TFV-DP levels consistent with 4+ doses per week (700+ fmol/punch). In fact, although only 20 AGYW (6%, [95% CI 3.4 to 8.5]) reported taking 1 or no doses in the past week, 258/336 (76.8%, [95% CI 72.3 to 81.3]) had TFV-DP levels <10 fmol/punch, which is consistent with no doses in the recent past. Results at Interview 2 were similar. Although most (166/197, 84%, [95% CI 79.2 to 89.3]) of the AGYW who had samples taken at this visit reported taking 4+ doses in the past 7 days, only 9 (4.6%, [95% CI 1.6 to 7.5]) had TFV-DP levels consistent with this level of adherence and 163 (82.7%, [95% CI 77.4 to 88]) had levels consistent with no PrEP use.

PrEP use at the second interview was highly correlated with PrEP use at Interview 1 (S1 Fig). TFV-DP levels greater than 199 fmol/punch (the lower limit of quantification) at Interview 2 were seen almost exclusively among AGYW with levels >199 fmol/punch at Interview 1. Very few AGYW who were not taking PrEP at baseline (TFV-DP <10 fmol/punch) achieved quantifiable levels at Interview 2, and almost all the AGYW who were lost-to-follow up were not taking PrEP at baseline.

Among AGYW who said they discontinued PrEP at Interview 2, primary reasons for discontinuation included stock outs of PrEP, no longer being at risk for HIV (mainly attributed to currently trusting their partner), and returning to school or being away from home (S2 Table). Side effects were also cited fairly frequently as a reason for discontinuation, but reports of discontinuation due to health effects, lifestyle demands, difficulty taking daily pills, pill characteristics, or pressure from partners were not common. Among those who had a late or missed refill visit (N = 49), the most common reasons offered were traveling away from home, forgot the appointment, or no provider was available. Almost all AGYW (179/197, 90.9%) who reported continuing PrEP at Interview 2 said their strong commitment to preventing HIV infection made it easy to take PrEP, and the majority (106/197, 53.8%) said that they had had no difficulty taking PrEP (S3 Table). Only a few (25/197, 13%, [95% CI 8 to 17]) of these AGYW reported not taking any PrEP in the past week. Most (166/197, 84.3%, [95% CI 79.1 to 89.3]) reported taking sufficient (4+) pills to provide protection.

Interestingly, although non-protective TFV-DP levels were detected in almost all participants, most (179/197, 91%) AGYW who said they were continuing PrEP indicated that they were taking PrEP to avoid HIV infection and believed themselves to be protected. At the second interview, participants were asked, “What do you think your chance is of getting HIV?”; only 11/197 (5%) who said they were continuing PrEP answered moderate-to-high chance, in contrast to 38/105 (36%) among discontinuers (S3 Table). AGYW who reported continuing PrEP were then asked, “What would your chance of getting HIV be if you were not taking PrEP?”; 167/197 (85%) answered moderate-to-high.

Despite attending the follow-up and PrEP refill visits and stating that they were continuing PrEP, most PrEP persisters were not adherent. Only 28/176 (16%) had detectable levels at Interview 2, with only 7 (4%) having levels consistent with protection (≥700 fmol/punch). While interventions to improve adherence may benefit these AGYW, it is unclear that such interventions will help the larger group of AGYW who persist in the program without taking PrEP at all. Compared to AGYW who discontinued PrEP, this persistent but non-adherent group (N = 144) was older (adjusted OR 2.13 [95% CI 1.01 to 4.49], p = 0.048), more likely to be active in the DREAMS program (adjusted OR 3.85, [95% CI 1.07 to 13.82], p = 0.039), think they would be at moderate-to-high risk for HIV if not taking PrEP (adjusted OR 10.17, [95% CI 5.14 to 20.13], p < 0.001), and more likely to use injectable contraceptives (adjusted OR 3.28, [95% CI 1.4 to 7.67], p = 0.006) (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Factors associated with Persistence despite poor adherence (Discontinuers vs. Persisters who did not take PrEP).

https://doi.org/10.1371/journal.pmed.1004097.t002

Among AGYW who persisted in the program, no significant differences were seen between those with TFV-DP levels above versus below 10 fmol/punch in general (S4 Table), likely due to the small number (N = 32) with levels above 10 fmol/punch. Models comparing all persistent AGYW (N = 176) to (1) AGYW who discontinued PrEP (N = 105); and (2) AGYW who attended the second interview but were non-persistent (N = 105 + 21) gave similar results, although the association with partners who were supportive was no longer significant (S5 and S6 Tables).

Discussion

In this study designed to examine PrEP adherence, persistence, and factors associated with persistence, nested within a real-world PrEP program in western Kenya, AGYW had moderate retention in the program, and most reported high adherence and continuation of oral PrEP. However, a minority had detectable TFV-DP at interview visits, and only a small percentage achieved or sustained sufficient drug concentrations to prevent HIV acquisition. Most (85%) continued to receive HIV prevention support through the DREAMS initiative. AGYW who stopped taking PrEP were more likely to have stopped participation in DREAMS (27% versus 8%) and gave a variety of reasons for discontinuation, most of which were not directly related to the PrEP product itself (stockouts, no longer being at risk, returning to school, travel away from the site, etc.). Notably, AGYW who persisted in the PrEP program had a higher self-perceived risk of HIV infection. Among AGYW who reported they were continuing PrEP at the second interview, over 90% indicated the reason was to prevent HIV, although almost all had non-protective TFV-DP levels. This study documents the disconnect between intention and action in a real-world PrEP program and identifies AGYW who may be optimal candidates for long-acting injectable PrEP.

Other studies have observed vast discrepancies between self-reported and objective PrEP adherence measures in clinical trial settings [8,10,33,34]. Importantly, adherence in our study participants was much lower than adherence in AGYW enrolled in clinical trials or in a recent trial of support strategies for oral PrEP in South Africa and Zimbabwe (89% retention at 3 months; 84% of participants tested had detectable TFV-DP) [35,36]. This suggests that oral PrEP adherence among AGYWs in real-world programmatic settings in SSA may be lower than predicted by trial settings and may not assure the drug levels necessary to prevent HIV acquisition. A recent evaluation of the DREAMS Initiative in Kenya and South Africa reported a decline in HIV incidence in AGYW which began prior to DREAMS and which did not accelerate during the first 3 years of the program [37]. Although the reasons for the failure of DREAMS to impact HIV incidence in these locales are unknown, low uptake of interventions by participants may have contributed.

The results of this observational study also provide additional data on the experiences of AGYW using PrEP in a programmatic setting in SSA. Despite their commitment to preventing HIV infection, few AGYW achieved high-level adherence. AGYW who discontinued PrEP use reported multiple reasons for doing so. Although stockouts were most commonly cited, this reason was cited by only a quarter of AGYW who discontinued. Many of the other reasons cited pertained to the PrEP program but had little to do with the PrEP product itself. Few AGYW cited fear of or experience with side effects (15.2% of AGYW who discontinued and 7.6% of those who said they were continuing PrEP). Some AGYW may have engaged in the HIV prevention program, while not adhering to a daily pill regimen, due to social desirability. However, their responses to questions suggest that they were motivated at least in part by perceived HIV risk. The relationship between perceived HIV risk and actual risk in discontinuers versus continuers is complex, as both groups had similar proportions who reported having a sexual partner, living with a partner, using condoms, etc., although continuers were more likely to have multiple partners (7% versus 15%) or to believe that their partner’s behavior put them at risk (29% versus 44%). Only 36% of AGYW who discontinued PrEP reported their current HIV risk as moderate-to-high, although they were not taking PrEP. Only 6% of AGYW who said they were continuing PrEP assessed their current risk as moderate-to-high; a striking 79% said their current risk was low but would be moderate-to-high if they were not taking PrEP. Thus, most AGYW who said they were continuing PrEP acknowledged moderate-to-high HIV risk and the vast majority said they were using PrEP to prevent HIV. The fact that almost none had protective levels implies that these AGYW did not understand or acknowledge that only high-level adherence to daily oral PrEP affords protection.

Overall, these results reveal a group of AGYW who need tailored HIV prevention, including methods such as long-acting injectable PrEP. In our cohort, use of injectable contraception or implants was common. At the second interview, nearly half (45%) of all AGYW and 88% of those using a highly effective contraception method reported use of injectables or implants, which suggests that injections or implants may be acceptable for HIV prevention as well. While attention has focused on improving adherence to oral PrEP, understanding women who are persistent but not adherent and providing them with long-acting formulations may be a more effective way to reduce population-level HIV incidence. In addition, newer prevention modalities will need appropriate behavioral counseling support to assure understanding and correct use.

Limitations of this study include both size and generalizability, as we included a relatively small sample size from a large programmatic project, which may be most generalizable to AGYW in western Kenya. In addition, all AGYW in the study were recruited from the DREAMS initiative and were receiving support for HIV prevention that would likely not be available to them otherwise. Other limitations include potential social desirability bias, recall error in the self-reported data, and possibly remaining confounding bias. A strength was the testing of TFV-DP by LC-MS/MS as an objective measure of adherence. All AGYW participated in face-to-face interviews with a trained counselor. In addition to the self-reported data on factors that might influence PrEP use, we also had access to programmatic data, for example, on pharmacy refills.

In conclusion, our study provides insight into oral PrEP use in a real-world programmatic setting, as well as predictors of PrEP persistence in this context. The results reveal that most AGYW may be better protected by long-acting injectable PrEP than by oral daily PrEP. Future research is needed to clarify whether persistence without adequate adherence is as common among AGYW in other settings, and whether new long-acting PrEP formulations, adequately supported by facilitators identified during use of existing PrEP agents, can afford higher level protection to this very vulnerable population.

Supporting information

Acknowledgments

The authors would like to thank Sahar Zangeneh, Eline Appelmans, Siavash Pasalar, Lili Peng, and Jinru Tao for their early inputs into this research plan and design. We also thank all AGYW who participated in the study and all the staff involved in the data collection for the project.