Corneli, Amy L. PhD, MPH^*; McKenna, Kevin MPH^*; Perry, Brian MPH^*; Ahmed, Khatija MBChB, MMed Micro^†; Agot, Kawango PhD, MPH^‡; Malamatsho, Fulufhelo BSocSci^†; Skhosana, Joseph BTechBioSci^†; Odhiambo, Jacob MA^‡; Van Damme, Lut MD, MS, PhD^§

INTRODUCTION

FEM-PrEP was a phase III, randomized, double-blind, placebo-controlled trial to assess the safety and effectiveness of once-daily, oral emtricitabine/tenofovir disoproxil fumarate (FTC/TDF) as pre-exposure prophylaxis (PrEP) for the prevention of HIV in women in Bondo, Kenya; Bloemfontein and Pretoria, South Africa; and Arusha, Tanzania.¹ Adherence to the study pill was low,^1,2 leading the trial to close early because of futility.³

Analyses of plasma and intracellular drug concentrations among a subcohort of the study population demonstrated that 12% of the subcohort achieved good adherence to the study pill (ie, tenofovir in plasma exceeding 10 ng/mL and intracellular tenofovir diphosphate in upper layer packed cells exceeding 100,000 fmol/mL) throughout their participation; 23% of participants rarely took FTC/TDF, if ever; and 60% of participants had fluctuating adherence throughout the trial.² Yet, self-reported adherence during the trial was high, with 95% of participants reporting that they had usually or always taken their assigned study pill. Moreover, pill-count data were consistent with participants having taken the study pill on 88% of the days in the trial.¹ Further analyses comparing drug concentrations with self-reported adherence and pill-count data provided evidence that participants overreported adherence and confirmed that these participant-dependent adherence measurements were unreliable in FEM-PrEP.⁴ Together, these findings suggested that FEM-PrEP participants may have counted and then possibly kept, discarded, given away, or traded study pills before study visits.⁵ The iPrEx and MTN-003 (Vaginal and Oral Interventions to Control the Epidemic—VOICE) PrEP clinical trials have also reported inconsistencies between participant-dependent adherence measures and drug concentrations.^6,7

A better understanding of the reasons trial participants purposefully overreport their adherence is needed to improve the use of participant-dependent measures in future HIV prevention clinical trials.^4,6,8 In this article, we describe the findings of a follow-up study we conducted with former FEM-PrEP participants in 2 of the sites—Bondo, Kenya, and Pretoria, South Africa—to understand factors that had influenced participants' overreporting of adherence and to learn the whereabouts of unused study pills.

METHODS

FEM-PrEP Clinical Trial

A description of the trial can be found elsewhere.^1,2,4 A total of 2120 women (764 in Pretoria, 739 in Bondo, 554 in Bloemfontein, and 63 in Arusha) were enrolled. Participants were asked to take their assigned study pill—either FTC/TDF or placebo—daily for 52 weeks and attend a study visit every 4 weeks. At each visit, participants reported their adherence from the previous 4 weeks, were asked to return any unused study pills, and left with 30–37 pills (each new bottle contained 30 pills, and up to 7 previously returned pills could be redispensed). Adherence counseling was also provided at each visit by staff separate from those who assessed the participant-dependent adherence measures.⁹ We announced that the trial was closing early in April 2011. All regular follow-up visits were completed by August 2011, community activities were completed by December 2011, and follow-up visits for participants who seroconverted were completed by July 2012.

Data Collection and Study Population

The FEM-PrEP follow-up study was implemented from March to June 2013. We conducted qualitative semistructured interviews (SSIs) with 88 former FEM-PrEP participants from Bondo (n = 43) and Pretoria (n = 45) who had been assigned FTC/TDF during the trial. These participants were purposefully selected to represent a range of adherence levels, based on their drug concentration data. Eighty-three of the 88 participants were selected from a list of 133 FEM-PrEP participants from Bondo and Pretoria who had been assigned FTC/TDF during the trial, who had their blood samples previously analyzed for drug concentrations as part of other FEM-PrEP analyses, and who had given their permission at the time the trial closed to be contacted for future research; an additional 5 participants who had samples previously analyzed, were included as a result of community-wide promotion of the follow-up study. We included women who had had drug concentrations that suggested that they rarely took FTC/TDF (n = 32), that they took FTC/TDF some of the time (n = 31), or that they took it regularly at several or multiple points during the trial (n = 25). Details of the drug concentration analyses are described elsewhere.^2,4,10 During the SSIs, participants were asked to describe what they had done and what they believed other participants had done with the unused study pills and why; specific probes were asked as necessary (eg, shared, sold, discarded, stockpiled). All SSIs were audio-recorded with the participant's permission. Other adherence-related topics, such as reasons for taking and not taking the study pill based on the participant's drug concentration data, were also explored and are described elsewhere.^11,12

We also conducted quantitative, audio computer-assisted self-interviews (ACASI) with 224 former FEM-PrEP participants who had been assigned FTC/TDF or placebo. The group of ACASI participants was composed of (1) 86 of the 88 SSI participants described above (Bondo, n = 43; Pretoria, n = 43), (2) a convenience sample of 50 participants assigned placebo during FEM-PrEP (Bondo, n = 25; Pretoria, n = 25), and (3) 88 of the 90 former FEM-PrEP participants who participated in a separate SSI on risk perceptions that we conducted as part of the same follow-up study (Bondo, n = 44; Pretoria, n = 44). Participants in the risk perception group were purposefully selected and recruited based on their perceived risk (ie, none, some/moderate, high) reported at their last study visit during the trial. The placebo group was identified and recruited through a variety of approaches: direct recruitment, community-wide study promotion, word-of-mouth by other former FEM-PrEP participants, and hearing about the study after being at the research clinic for another reason. ACASI was conducted after participants in the FTC/TDF group and the risk perception group completed their SSIs; participants in the placebo group completed only ACASI.

During ACASI, participants were asked whether they had ever told trial staff that they had taken a study pill when they knew they had not. For those who reported that they had, a series of questions followed with yes/no responses on the reasons for overreporting their adherence. Response options were informed primarily from the adherence-related data collected in qualitative interviews with participants during the trial and at study closure.⁴ Participants were also asked a series of yes/no questions about the whereabouts of their unused study pills.

Previous FEM-PrEP findings demonstrated that participants overreported their adherence during the clinical trial, in both face-to-face quantitative and face-to-face qualitative interviews.⁴ We therefore did not want to limit data collection in the follow-up study to only face-to-face methods. ACASI has been previously used as a method to reduce the reporting of socially desirable responses and was therefore included as a data collection method.¹³ Additionally, through discussions with site staff and the sites' community advisory boards, we identified and used other methods to reduce the potential for socially desirable responses: conducting the SSIs and ACASI at a location other than the FEM-PrEP study clinic, having non–FEM-PrEP staff conduct the SSIs, not having staff present in the room when participants completed the ACASI, showing participants a graph of their drug concentration data, and informing participants that their ACASI responses would not be linked to them or to their demographic information. Last, we asked indirect questions (ie, what participants believed other participants had done) in addition to direct questions (ie, what participants had done themselves) in the SSIs because it is a valid method for reducing the likelihood of socially desirable responses and for projecting undesirable behaviors.¹⁴

Data Analysis

We used descriptive statistics to analyze the quantitative data. For the qualitative data, we used applied thematic analysis.¹⁵ Four analysts coded transcripts using NVivo 10.¹⁶ Two structural codes were applied to the text describing what participants had done and what they believed other participants had done with their unused study pills. Intercoder reliability was assessed on 10% of the transcripts. At each assessment, discrepancies were identified and resolved, and transcripts were recoded as needed. After all transcripts were coded, coding reports were produced and reviewed for the 2 structural codes. One analyst created a data reduction matrix for each code and identified the themes and subthemes of each code subthemes of each code, the frequency of each them and sub-theme, and relevant quotes. A second analyst verified the themes, subthemes, and frequencies. A summary memo was written that described all themes, subthemes, and frequencies and percentages based on the number of participants who discussed a particular topic. The memo also provided quotes that best exemplified the themes and subthemes.

Ethics

The Ethics Review Committee at the Kenya Medical Research Institute (Bondo), the Pharma-Ethics Review Board (Pretoria), and the Protection of Human Subjects Committee at FHI 360 in the United States reviewed and approved the research. Participants provided either verbal (Bondo) or written (Pretoria) informed consent to participate in the follow-up study, per local ethics requirements.

RESULTS

Participant Demographics

Using the demographic information reported at screening for FEM-PrEP, former participants who participated in the follow-up study in Bondo were generally older than the larger trial population in Bondo (mean age, 29.3 years compared with 26.5 years), and a higher percentage were married (89.3% compared with 72.9%). We found no meaningful differences in age and marital status between participants in the follow-up study and the larger trial population in Pretoria, or in education and occupation between the 2 populations in either Bondo or Pretoria.

Overreporting Adherence

During ACASI, 70 (31%) of the 224 participants reported that they had overreported their adherence to the study pills during the trial. A greater percentage of participants in Pretoria (43%, n = 48) than in Bondo (20%, n = 22) acknowledged overreporting. Table 1 lists the reasons participants acknowledged for providing inaccurate information about their adherence.

TABLE 1

The most common reason reported overall by participants during ACASI was the concern that they would be terminated from the trial for nonadherence (69%, n = 48). This was the most frequent reason given by participants from Pretoria (83%, n = 40), together with acknowledging that overreporting was easier than providing accurate reports of their adherence (83%, n = 40), and the second most frequent reason given by participants from Bondo (36%, n = 8). Among the other reasons, anticipated negative reactions from trial staff were generally reported more often by participants from both sites than were reasons related to study procedures, although site differences remained. Eighty-one percent of participants from Pretoria who acknowledged overreporting (n = 39) said that they had not accurately reported their adherence because they did not want to disappoint staff, and 73% (n = 35) thought staff might scold them if they reported nonadherence. These reasons were notably less common among participants from Bondo (23% and 18%, respectively). Concerning the reasons related to study procedures, 43% of participants from both sites (n = 30) acknowledged that they had overreported their adherence because they did not want to be reminded to take their study pills; 23% (n = 16) believed that they would have needed to stay longer at the study clinic if they had reported missing a pill, although the latter was acknowledged by only 1 participant in Bondo. Wanting to avoid being reminded to take the study pills was the most frequent reason given by participants in Bondo (50%, n = 11).

Whereabouts of the Study Pills

Returning the unused study pills to the study clinic was the most frequent answer given during ACASI on what had been done with the unused study pills (83%, n = 185) (Table 2); it was also volunteered by 40 (45%) of the participants in the SSI on what they had personally done with their unused study pills. Discarding or throwing away the unused study pills was the second most frequent action reported in ACASI (35%, n = 78). It was also commonly described by participants from both sites during the SSIs on what they had personally done with their unused study pills (63%, n = 45/72) and what they had heard or seen other participants do with their unused pills (84%, n = 72/86). A higher percentage of participants in the FTC/TDF ACASI group (55%, n = 47) than in the placebo ACASI group (30%, n = 15) reported discarding their pills (Table 3). In particular, 43% (n = 37) in the FTC/TDF group had thrown their unused study pills in a toilet/pit latrine, compared with 28% (n = 14) in the placebo group. In the SSIs, throwing pills into the toilet, pit latrine, or a urinating bucket was mentioned far more often than other methods of disposal (eg, trash, ground). Disposal in toilets and latrines was also mentioned by some participants to be a safer means of disposal, as children would be unable to access the pills. A participant from Pretoria described what she had heard others did with their unused study pills:

What we heard a lot when we were all sitting there [at the study clinic] is flushing them. It's a simple way. If you put them there like this [e.g., on a countertop], you'll find that kids will eat them. They will think they're sweets, you see. So, when you flush them, it's simple. They are gone.

TABLE 2

TABLE 3

Some participants in the SSIs also said they knew or had heard of participants discarding pills on the ground (14%, n = 12/86; more common in Bondo) or in dustbins (12%, n = 10/86; all from Pretoria). Narratives that described throwing the pills on the ground also included descriptions of throwing the pills in the bush, along the roadside, or in pits in the ground; some of these participants described that pills were tossed on the way to or from the clinic. A participant from Pretoria described:

Others threw them. [Interviewer question: Where?] Like when we were on our way home from Setshaba [study clinic]. She [Interviewer note: Meaning the other participant] tells you that, “aah, when I pass through the bush, I just throw them away. By doing this, at home they won't be able to know where I am coming from.”

Some participants in the SSIs (n = 11) also described how other participants discarded pills while at the clinic, primarily by flushing them down the toilet.

Stockpiling pills (ie, keeping the pills to use later) also occurred among some participants. Twenty-two (49%) of the 45 participants in the SSIs who discussed whether or not they stockpiled their own pills said that they had stored their unused study pills, and 15 (7%) of the participants in ACASI said that they had stored their pills to use them later (Table 2). Similarly, 45 participants in the SSIs also discussed stockpiling among other participants, and approximately half (n = 22/45) said that they knew of others who had held on to their unused study pills during the trial. Of those participants in the SSIs who stored their own pills, some (41%, n = 9) said that they disposed of their stashed pills after the trial closed. Other participants in the SSIs said that they (11%, n = 5/45) or others (18%, n = 8/45) had kept their unused study pills to use later, if FTC/TDF was shown to be effective or if they became HIV-positive. A participant from Bondo explained:

So, most were throwing them in the pit latrine. But, some were storing them in their houses, thinking that after some years, the drugs will help them. So they will not be buying those drugs, they will just take them. They stored hoping just in case the study result comes back positively, they will use them later.

Another participant from Bondo said:

Some participants always kept their pills that they will use in the future. [Interviewer question: Why were they doing that?] Because they thought that they would need them later when they are sick.

Sharing or selling the study pills was infrequently mentioned by participants during ACASI and also during the SSI when participants described what they had done with their own study pills. In ACASI, few reported giving the pills to someone else (4%, n = 10), and even fewer reported giving them to someone with HIV (2%, n = 5) (Table 2). Only 5% of the participants in the SSIs (n = 3/61) said they had ever shared their study pills with others (all specified sharing with individuals who were HIV positive). Only 2 (1%) participants in ACASI said they had sold or traded pills, and none of the 53 participants in the SSIs who were asked about selling their pills said they had sold them. A few participants in the SSIs (n = 5) described why they had not shared their pills; reasons included that they were instructed not to share, were unsure of their pill assignment, were unsure of pill effectiveness, or had heard negative rumors about the pills. A participant from Bondo described:

I could not do that [share pills] because I did not know the type of pill I was taking during that time

Participants provided similar reasons for why they did not sell the pills. A participant from Bondo said:

How can you sell to someone a pill that you do not know if it works?

More instances of sharing and selling were discussed when participants in the SSIs described what other participants had done with their unused study pills, although only a minority of participants described these occurrences: 18% (n = 15/84) said pills were shared, primarily with individuals who were HIV-positive, and 9% (n = 8/85) said pills were sold, half of the time to individuals who were HIV positive or suspected to be infected. Only 2 participants said they had heard of participants sharing pills with other participants. Sharing and selling the study pills were mentioned more often among participants in the SSIs from Pretoria than among participants in the SSIs from Bondo. A participant from Pretoria said:

I once heard somebody saying her boyfriend was sick, so she gives him this pill. He then gets better. She never drank them. She gave them to her boyfriend. [Interviewer question: Was her boyfriend HIV-positive?] Ja, he was HIV-positive. He was a person that when he drank them, he got strength.

Counting Pills

In describing the whereabouts of the unused study pills or in answering other adherence-related questions in the SSIs, a considerable minority of participants (34%, n = 30/88) volunteered that they or other participants would count their study pills before discarding them or returning them to the clinic. Although no direct counting questions were asked (out of concern that such questions could be offensive), participants from both sites described that this was a common method used to seem adherent to trial staff. Participants described that they or other participants who counted pills were aware of approximately how many study pills should be left in their pill bottle by their next study visit, so they would ensure that their bottle contained the appropriate amount. A participant from Pretoria explained:

Let's just talk about me. I would sort them at home. Let's say today is the 18th and I came to Setshaba. After the 18th, maybe I drank two times or three times. Then I go back next month on the [15th]. So I count those days, like [Interviewer note: Participant demonstrated by counting on with her fingers]. Yes, I count those days then I sort out the pills, like “okay, I was supposed to have drank this many pills, this much must be left over.” So that's what I did.

Similarly, a participant from Bondo described:

There were some who just kept the drug. On the day of their study visit, they would just count the days from the [last] day they went [to the clinic] till the day they went back. Then, they throw [the extra pills] in the pit latrine and carry the remaining ones and return them back.

DISCUSSION

Participants acknowledged several perceived negative consequences of nonadherence that made it difficult for them to accurately report their nonadherence during the trial. Concern about being terminated from the trial was the most commonly acknowledged reason. Yet, participants' uneasiness during the trial occurred despite our best efforts to create a supportive environment where participants could feel comfortable reporting instances of nonadherence. For example, trial staff was to communicate standardized messages during each adherence assessment and each adherence counseling session that stressed they would not negatively judge participants for nonadherence and that nonadherence would not upset them.^4,9 However, we have no data on the fidelity of staff communicating these messages to participants or on how they were communicated. If these messages were regularly communicated, our data suggest that they were not effective. Additionally, unclear or mixed messages could have been communicated and influenced participants' overall perceptions of negative consequences for nonadherence. For example, although we have no data suggesting this, the standard message that is frequently included in clinical trial consent forms—“you may be asked to leave the research if you are not able to follow the research procedures”—could have been interpreted by the trial participants to mean taking their study pills, which was not the intent of the statement. Of note in this regard, no participants were terminated from the trial for reasons related to noncompliance.

Participants in the follow-up study seemed to be somewhat more upfront about acknowledging that they had overreported their adherence than were participants during the actual trial, in which very few admitted to overreporting.⁴ Many reasons may explain this differential reporting during and after the trial, such as the large time gap between the closure of FEM-PrEP and this study, the use of ACASI, interviews being conducted at a location other than the study clinic, interviews being conducted by non–FEM-PrEP staff, participants being shown their drug concentration data, or participants having no fear that they would lose the benefits of being in the clinical trial because it was over. Nonetheless, for many participants, their discomfort in self-reporting a likely perceived negative adherence behavior continued even after the trial; 69% of participants (n = 154) reported in ACASI that they had never overreported their pill adherence during the trial. Although it is possible that some of these participants may have reported their adherence accurately during the trial (and correctly reported in the follow-up study that they had never overreported their adherence), the previously documented extent of overreporting in FEM-PrEP⁴ suggests that many of these follow-up participants felt uncomfortable acknowledging that they had previously overreported their adherence to trial staff.

Social desirability bias in self-reported data is well documented.^17–20 Because of the weaknesses associated with self-reported measures, including social desirability bias, a considerable amount of research and dialog has occurred on improving and identifying best practices for assessing self-reported adherence.^21–23 One such proposed best practice is to minimize respondents' perceptions of negative consequences.²² Based on our findings, in moving forward with collecting self-reported data within the context of HIV prevention clinical trials, we suggest that future research explore this best practice in more detail by (1) identifying effective methods to reduce participants' concerns about disappointing staff when reporting nonadherence, and (2) developing persuasive messages and strategies to reduce participants' fears that negative consequences, such as being terminated from the trial or scolded by trial staff, will occur after reports of nonadherence. Empirical evidence is also needed on the extent to which we can in fact minimize social desirability bias in self-reported data in clinical trials, as biomedical HIV prevention is not the only field grappling with nonadherence and overreporting.²⁴ Ultimately, the motivation among trial participants to provide socially desirable responses may be grounded within societal norms and expectations geared toward building and reinforcing social relationships. For this reason, when we designed this study, we included a social desirability scale^25,26 to gauge the inclinations of former FEM-PrEP participants to report favorable responses; these data will be reported elsewhere.

Addressing the most common reason given by participants for overreporting their adherence—the fear that they would be terminated from the trial–is complex. As we have shown elsewhere,^11,27 many participants reported continuing in the FEM-PrEP trial because of the indirect benefits and ancillary care they received (eg, HIV testing, medical care for common illnesses). Furthermore, counting the study pills, as described here, demonstrates the extent of what participants will do and can do (ie, correctly determine the number of pills that should remain in their pill bottle when the number may differ at each visit) to stay enrolled in a clinical trial even when they are not motivated to use the study product. The perceived potential loss of such benefits may have likely influenced the responses on adherence that the participants provided. These findings suggest that women value their health and want access to better health care. From a trial perspective, the appropriate level of ancillary care to be provided in biomedical HIV prevention clinical trials has been a long-standing ethical discussion in the field.^28–32 Individuals may continue to enroll in clinical trials as a means to obtain ancillary health care until these needs can be otherwise sufficiently met.

Concerning the whereabouts of the unused study pills, participants were likely referring to only some (ie, not all) of their unused study pills when they said they had returned pills to the clinic, the most commonly acknowledged location of the unused study pills. This response may also be a reflection of what participants had been asked to do with their unused pills. We have described elsewhere the analyses of pill-count and drug concentration data collected during the trial, which demonstrated that most participants did not return all of their unused study pills to the study clinic.⁴ Our findings here also show that only a small number of participants gave their unused study pills to other people. This may suggest that the study-specific counseling messages on reasons for not sharing or selling the study pills (provided to participants during adherence counseling at enrollment, at each 4-week visit during the first quarter, and thereafter as needed⁹) may have helped limit occurrences of pill sharing and selling.

By design, we used a flexible, purposeful, and multipronged sample selection and recruitment approach to (1) identify and recruit individuals who had various experiences in taking the study pill and (2) maximize the number of participants we could locate given the large time gap between the closing of FEM-PrEP and the initiation of this study. We therefore did not expect this sample to be representative of the larger FEM-PrEP population. Although the follow-up participants in Bondo were generally older and many were married compared with the larger trial population in Bondo, we previously found no association between age, education, and marital status and (1) adherence to the study pills² or (2) misreporting.⁴ The consistency of participants' responses about other participants' behaviors suggests that the overall themes described here may be extrapolated to FEM-PrEP participants not in this follow-up study.

We also included participants whose drug concentrations suggested that they regularly took the study pill at several or many of their visit intervals. Therefore, the overall frequencies of misreporting and of reporting on the whereabouts of unused study pills may have been different if we had included only participants who had had consistent, suboptimal adherence. We were unable to stratify the ACASI data by adherence group because we did not link responses to participant identifiers, as a method to reduce socially desirable responses. Participants who had evidence of regular adherence were included in the sample because they likely did not have perfect adherence (ie, did not take their pills every day) throughout the trial and therefore possibly had some unused study pills and missed pills to report. Last, participants in the FTC/TDF and risk perception interviewing groups may have been more, or less, comfortable answering the ACASI questions than participants in the placebo group, because the FTC/TDF group had just discussed adherence issues in an SSI and the risk perception group had just discussed other trial-related issues in an SSI; participants in the placebo group did not participate in an SSI and may have answered differently in ACASI if they had previously participated in an SSI.

In conclusion, although biological specimens provide an objective measurement of adherence, they are expensive and their ability to fully describe participant adherence (eg, patterns of use, reasons for nonuse) is limited. Other relatively objective measures of adherence, such as unannounced pill counts and electronic systems to measure medication events, exist and have been used in PrEP-related research.^33,34 Yet, reliable participant-dependent adherence measures, especially participant self-report, are needed; self-report is the most common adherence measure²² and the least expensive. The validity of participant-dependent measures, however, cannot likely be enhanced without consideration of the broader environment (perceived or real) in which the data are collected. Efforts to improve such measures should include identifying alternative methods for creating supportive environments that allow participants to feel comfortable reporting actual adherence.

ACKNOWLEDGMENTS

The authors are grateful to the women who participated in the FEM-PrEP follow-up study, to the study staff, and to the communities who partnered with us to conduct the study. The authors thank Kate MacQueen for feedback on the study design and manuscript.

REFERENCES