The Effect of SMS Reminders on PrEP Adherence in Young Kenyan Women: Findings from a Randomized Controlled Trial

Background

HIV prevention for young women in sub-Saharan Africa is a key priority for the global strategy to end the HIV epidemic. This population is particularly vulnerable to HIV acquisition based on a multitude of factors, including disproportionate power relationships with sexual partners, limited access to sexual health resources, and the biology of HIV transmission as a receptive partner during sex (1). Typical neurodevelopment associated with adolescence and young adulthood also promotes the pursuit of immediate needs (e.g., food, friends, and love) over abstract concepts (e.g., disease prevention) (2), and assessment of HIV risk is complex (3). With this backdrop, 25% of new HIV infections in sub-Saharan Africa occur among adolescent girls and young women (1). Concerted efforts are therefore needed to support use of effective HIV prevention tools in this population.

Pre-exposure prophylaxis (PrEP) is a highly effective means for preventing HIV acquisition when consistent adherence aligns with risk (4). However, adherence has been challenging for many young women. In phase III clinical trials (i.e., FEM-PrEP and VOICE), adherence was low enough that PrEP was not effective for this population (5, 6). Similar experiences have been seen in more recent studies and demonstration projects, despite adherence interventions such as tailored counseling, PrEP champions, and drug level feedback (7, 8).

SMS reminders and outreach have been shown to improve adherence to HIV antiretroviral therapy (ART) in some, but not all, contexts (9). One study in Uganda showed significant, sustained improvement in adherence among adults initiating ART when SMS reminders were combined with a real-time adherence monitor (10); participants reported that both tools helped establish the habit of adherence and provided a sense of connection and support from the clinic (11). SMS reminders have been less well studied for adherence to PrEP, although initial experience with men who have sex with men (MSM) in the United States has been mixed (12–14). Importantly, SMS may be attractive, particularly for youth, given the high and rising rates of mobile phone usage in sub-Saharan Africa (15).

To determine the impact of SMS reminders on PrEP adherence, we conducted a randomized controlled trial among young Kenyan women at high risk for HIV acquisition who were offered PrEP over a two-year period.

Methods

Results

We screend 642 participants for study participation; the primary reason for non-participation was lack of interest in the study. As shown in Figure 1, we enrolled and followed 348 participants. Median follow-up was 22.1 months (interquartile range [IQR] 22.0, 23.0) per participant.

Participant characteristics at enrollment are presented in Table 1. Demographic and behavioral characteristics were largely similar (i.e., <10% difference) between the two arms with the exception of problematic alcohol use, which was more common in the SMS reminder arm (41 [24%] versus 25 [14%]).

Of the 348 women enrolled in the study, we assigned 173 to SMS reminders and 175 to no reminders; 97% (N=69,291/71,791) of all SMS reminders were sent as planned. Technical difficulties early in implementation prevented the transmission of the remaining 3% (N=2,500) of SMS reminders. Twenty-four participants opted for “as needed” reminders (14% overall and 16% of the 153 participants taking PrEP from Month 1 onward). Table 2 indicates the most common SMS content selected by participants. No cross-over of study arms occurred.

Figure 2 shows electronically monitored adherence and pharmacy refill data. We considered 4,543 participant-months in the analysis of prevention-effective adherence (i.e., during periods when PrEP was picked up) and 7,848 participant-months in the intention-to-treat analysis. Fifty-nine (34.1%) participants in the SMS reminder arm and 58 (33.1%) participants in the no SMS reminder arm chose to permanently stop PrEP during the course of the study. The availability of PrEP adherence monitoring data was similar by study arm. We censored electronically monitored adherence data for 12.8% (15,165/118,456) and 11.2% (13,445/120,550) of participant-days in the SMS reminder and no SMS reminder study arms, respectively, due to non-functional monitors; the primary cause of non-functional devices was participants not bringing them to the study site for charging. We censored an additional 111/3,600 (3%) and 71/3,638 (2%) participant-months primarily due to breastfeeding and seroconversion (N=4) in the SMS reminder and no SMS reminder study arms, respectively. We did not detect drug resistance among three participants who acquired HIV; the test for the fourth participant acquiring HIV could not be amplified.

During periods when participants chose to pick up PrEP, electronically monitored adherence was 38.5% of doses over 6 months of follow-up and was similar by arm (40.2% among women receiving SMS reminders; 36.8% without SMS). Over 24 months of follow-up, electronically monitored adherence was 26.8% of doses and was again similar by arm (27.0% among women receiving SMS; 26.6% without SMS). When adjusting for site and study quarter, no statistically significant difference was present at 6 months (aIRR 1.14 [95% CI 0.96, 1.34; p=0.13] or at 24 months (aIRR 1.16 [95% CI 0.92, 1.45; p=0.19]). We found significant differences in electronically monitored adherence levels between the two study sites (aIRR 0.65 [95% CI 0.54, 0.84; p<0.001]); however, the effect of the intervention did not differ statistically by site (p=0.86 for the interaction term). No intervention effect was present when assessing electronically monitored adherence regardless of PrEP refills (aIRR 1.14 [95% CI 0.90, 1.43; p=0.28]). Additionally, we observed no difference in pharmacy pick-up between the two randomization arms over 6 months (aIRR 1.01 [95% CI 0.95, 1.07]; p=0.86) or 24 months (aIRR 1.10 [95% CI 0.97, 1.24; p=0.13]). As shown in Figure 3, the correlation between electronic adherence monitoring and DBS had a rho of 0.59 overall (0.56 in the SMS reminder arm and 0.66 in the no SMS reminder arm; p=0.41).

No serious adverse events were related to study participation. The only adverse drug event was vomiting in one participant in the SMS reminder arm. Social harms probably or definitely related to study participation were also the same in both study arms (5 each) and consisted of verbal and physical abuse with a sexual partner. These harms were primarily related to PrEP use, although one also involved the SMS. No participants declined SMS reminders or left the study because of social harms. Forty-six and 56 participants became pregnant in the SMS reminder and no SMS reminder arms, respectively, of whom 32 (70%) and 34 (61%) chose to continue PrEP during the pregnancy.

Discussion

In this randomized controlled trial, SMS reminders did not improve PrEP adherence among young women in Kenya. This finding was consistent for both electronically monitored adherence and pharmacy refill data and did not vary over time or by study site. One of the primary analyses purposefully focused on the women choosing to take PrEP, given that executed adherence is most relevant for those choosing to take it. Moreover, pharmacy refill may serve as a proxy for a desire and/or need for HIV prevention. This approach is consistent with prevention-effective adherence, or the alignment of adherence with periods of risk (18), although other factors may be relevant (e.g., inaccurate self-assessment of risk). When considering an intention-to-treat approach, results were similar.

Adherence by both electronic adherence monitoring and pharmacy refill decreased considerably after a few months of PrEP use, as has been seen in several other PrEP studies and demonstration projects (8, 24). Pharmacy refill indicates the maximal possible adherence, and the consistently lower adherence by electronic monitoring suggests that participants may not have taken many of the doses available to them. Alternatively, some women may have taken PrEP without opening the monitor for each dose (i.e., “pocket doses”); however, the moderately high correlation between electronically monitored adherence and TFV-DP levels gives some confidence in the accuracy of the former. In any case, even if misclassification occurred by either measure, bias between the study arms would be unlikely and would not impact the null findings of the trial.

We considered SMS reminders as a promising means to promote adherence among young women for several reasons. First, they have been shown in prior research to help establish a daily habit of adherence (11). Most young women do not take medication on a daily basis, especially for disease prevention, and have had similar adherence challenges with taking daily oral contraceptives (25). Although widely available, oral contraceptives constitute a small fraction of utilized family planning options and may reflect lack of desire or ability to take a daily pill. Long acting injectables and implantable contraception are clearly preferred, particularly in sub-Saharan Africa, and have higher rates of success. Similar formulations for PrEP are currently in development (26) and may serve young (and older) women well if they have rapidly changing or difficult to assess HIV risk, have trouble with accessing the clinic, and/or simply desire the convenience. Morover, choice is empowering and the ability to switch between methods depending on life circumstances could lead to higher adherence.

We also saw SMS reminders as a promising means of social support for young women. Prior research has indicated that SMS are perceived as an extension of the clinic, indicating clinicians’ concern for their well-being (11, 27). The high retention in this study speaks to the comfort the participants had with the study staff. The high levels of pharmacy refills may also reflect a desire to reciprocate on their support, rather than to actually take the pills. Qualitative research to understand the meaning of SMS reminders, feelings of social support, and any associated social desirability bias will be important to explore.

Finally, we saw SMS reminders as a possible way to harness the pervasive presence of digital technology in the lives of young women. In our study, all but one young woman approached for study participation had her own cell phone and none required training. Similar experience with technology can be found across sub-Saharan Africa with particularly high rates of cell phone ownership in Kenya, South Africa, and Nigeria (15). However, the ubiquitous presence of technology may have been a detriment rather than an advantage. When the landmark Wel-Tel study first showed SMS reminders to be supportive of adherence to ART in Kenya a decade ago, SMS were novel and may have been experienced as more special and meaningful than they are today. SMS are now commonplace and often a nuisance (e.g., uninvited and undesirable marketing messages, or SPAM). Indeed, a study of SMS reminders for ART adherence among adolescents in Uganda similarly found no benefit in 2017 (28), and the authors concluded that simple reminder messages may not be enough to capture the attention of recipients. We tried to anticipate SMS fatigue by offering the option of “as needed” reminders, but only a minority of participants selected them over the daily reminders for unclear reasons. Given the rapidly evolving nature of digital technology, on-going assessment of the latest apps and devices will be important for future intervention development.

In light of our trial findings and the ever more crowded space of digital technology, we must reenvision the role for mHealth in supporting medication adherence. Some promise may be seen in another study among Ugandan adolescents living with HIV in which SMS were used to inform participants not only of their own adherence, but that of their peers (29). That approach also involved a real-time adherence monitor and built on “reference dependence bias”—the tendency of individuals to want to equal or surpass the performance of their peers. Similar efforts could be tried with PrEP. Drug level feedback has shown some potential impact for PrEP adherence among MSM in the US (30). Similar benefit was not seen with young African women (8), although feedback of results was delayed by several weeks. Real-time adherence feedback could be achieved through text messages associated with electronic or ingestion monitoring or through immediate display of results with point-of-care urine-based adherence monitoring (31); these novel approaches may be considered for future research.

Alternatively, mHealth tools could be used to trigger and facilitate engagement with PrEP beyond reminders. Structural barriers (e.g., distance to clinic, adolescent-unfriendly services) and stigma are known barriers to PrEP uptake and persistence (32). Digital communication platforms could be used to request and support PrEP delivery for community-based care in preferred settings. The PrEP@Home intervention uses a web portal and video to support PrEP use for MSM in the US and has been shown to be feasible and acceptable (33). A similar intervention is underway for young women in Kenya that uses SMS or WhatsApp to arrange convenient PrEP delivery with supporting video guidance on smart phones (NCT04408729). These approaches reflect an evolution in mHealth that aims to keep pace with its target audience.

This study has a number of strengths and limitations. The randomized, controlled nature of the study design supports the rigor of the findings, and the populations in the two study arms were quite similar. This study was also well-powered and had high levels of retention. Additionally, contamination was minimized through the automated SMS delivery in the study, although sharing of messages among participants may have occurred. One limitation is reduced generalizability through the use of electronic adherence monitors. Although these monitors enhanced understanding of adherence, effects of SMS reminders on PrEP adherence without concurrent monitoring is unknown. Additionally, risk for HIV acquisition is complex and may not have been adequately captured through our use of the VOICE risk score. Use of PrEP and the effect of SMS reminders may vary in other groups of young women.

In conclusion, our randomized, controlled trial found no benefit for SMS reminders in supporting PrEP adherence among young women in Kenya. Most women expressed prolonged interest in PrEP based on continued pharmacy refills, yet a minority persisted in taking PrEP over the study period according to the electronic monitoring data. Future work should consider alternate PrEP formulation options that are less reliant on daily behaviors, which may appeal to many women. For those who desire a daily pill and/or struggle with engagement in their sexual and reproductive health, alternative strategies to promote adherence will be needed. Novel uses of technology to mitigate challenges in both the use and delivery of PrEP may also be beneficial.

Supplementary Material