Explaining the Efficacy of Pre-Exposure Prophylaxis (PrEP) for HIV Prevention: A Qualitative Study of Message Framing and Messaging Preferences among US Men Who Have Sex with Men

Abstract

We investigated message comprehension and message framing preferences for communicating about PrEP efficacy with US MSM. We conducted 8 focus groups (n=38) and n=56 individual interviews with MSM in Providence, RI. Facilitators probed comprehension, credibility, and acceptability of efficacy messages, including percentages, non-numerical paraphrases, efficacy ranges vs. point estimates, and success- vs. failure-framed messages. Our findings indicated a range of comprehension and operational understandings of efficacy messages. Participants tended to prefer percentage-based and success-framed messages, although preferences varied for communicating about efficacy using a single percentage versus a range. Participants reported uncertainty about how to interpret numerical estimates, and many questioned whether trial results would predict personal effectiveness. These results suggest that providers and researchers implementing PrEP may face challenges in communicating with users about efficacy. Efforts to educate MSM about PrEP should incorporate percentage-based information, and message framing decisions may influence message credibility and overall PrEP acceptability.

INTRODUCTION

Clinical trials have now demonstrated the safety and efficacy of antiretroviral pre-exposure prophylaxis (PrEP) for preventing HIV (1–6), and formative research has found that PrEP may be feasible and acceptable for a range of at-risk populations (7,8). PrEP acceptability studies and early implementation efforts in the US have focused on men who have sex with men (MSM) (9–26), as MSM transmission continues to account for the majority of new HIV infections (27). Several sources of guidance are now available to assist providers in implementing PrEP for MSM and other populations (28–33), including new guidelines from the US Centers for Disease Control and Prevention (CDC) (34). The manufacturer of tenofovir co-formulated with emtricitabine—the only medication currently FDA-approved for use as PrEP (35)—has also released both provider-directed and consumer-directed information to aid PrEP uptake (32,33). Despite these implementation efforts, many gaps remain in our understanding of how best to communicate with at-risk individuals about PrEP, both before and during PrEP use. The CDC guidelines discuss the importance of “simple explanations and education” in discussing medication dosage and adherence (34), but specific messaging and message framing strategies for PrEP outreach and patient education remain untested.

A high-priority question for PrEP messaging is how best to communicate facts about PrEP efficacy. Prior research has found that willingness to use PrEP depends in part on perceived efficacy, with higher perceived efficacy predicting greater acceptability (26,36). Trials have yielded efficacy estimates ranging from 44% to well above 90% depending on medication adherence (1–4,37). But individuals’ expectations of PrEP efficacy will be shaped by the messages that they receive, both through media outreach and through discussing PrEP with clinicians. Studies of vaccine acceptability have suggested that people interpret efficacy information in divergent ways; for example, participants told that a vaccine is 70% effective believed this to mean that 70% of people were 100% immune, or that every vaccinated individual obtained 70% protection (38). Because individuals use efficacy information to form expectations about how well an HIV prevention strategy will work for them personally, these varying interpretations may influence intervention uptake, medication adherence, and risk compensation behavior (i.e., increased risk-taking due to the perception that PrEP confers protection from HIV). Language choices and framing decisions may also influence how messages are understood and interpreted. Several types of message framing are particularly applicable for communicating about PrEP efficacy, including “different consequences framing” (describing efficacy in terms of success or failure rates) (39–41), loss- and gain-framing (emphasizing averted harms compared to health gains) (39,42), and one- or two-sided framing (providing information about positive attributes, negative attributes, or both) (43).

Two studies have examined message framing in the context of PrEP education, and both enrolled US MSM. One study found that MSM who received information about PrEP in conjunction with information about other methods (either condoms, or condoms along with rectal microbicides and post-exposure prophylaxis) reported a significantly higher likelihood of using PrEP, compared to MSM who received information solely about PrEP (16). A second study tested three separate messaging features: loss- and gain-framing (i.e., emphasizing the risk-reduction features of PrEP, compared to its health promotion features); presenting efficacy via percentages compared to “gist” messaging (without numerical values); and presenting information through a computer compared to a health educator (44,45). Analyses are ongoing, but preliminary findings suggest that message framing may indeed influence perceptions of PrEP. Motivation to use PrEP was highest among participants who received gist-based efficacy messages from health educators, and risk compensation intentions were lower among young MSM who received information emphasizing PrEP’s health promotion benefits, compared to young MSM who received information emphasizing PrEP’s risk-reduction benefits (44).

In addition to quantitative approaches that measure the impacts of messaging strategies, qualitative research can help to explain how individuals interpret and use information about PrEP efficacy. This study uses a qualitative approach to explore how MSM understand PrEP efficacy messages, the links between messaging and PrEP acceptability, and preferences for receiving efficacy information.

METHODS

We conducted a series of 8 focus groups (n=38) to identify initial themes, followed by n=56 individual interviews to gather in-depth information from MSM in Providence, RI. All participants were English-speaking adult cisgender males who self-reported their HIV status to be negative or unknown, and who reported condomless unprotected anal sex in the past 6 months with a male partner of positive or unknown HIV status. No participants had been in a PrEP efficacy trial. Focus groups occurred in February-June 2012, and individual interviews occurred in April 2013–April 2014. Recruitment included outreach and advertising in local entertainment venues, community-based organizations, clinics, and local media. Focus group methods have been reported elsewhere (17).

Data collection took place in private rooms in clinic settings, and we stopped interviewing when preliminary analyses showed data saturation for main themes. Focus group participants were permitted to enroll in interviews. We anonymized data collection during the interview stage, but based on interviewer familiarity with participants, we estimate that approximately 10 participants enrolled in both stages. Participants in both stages completed written questionnaires reporting demographics and risk behavior. We used written informed consent procedures for focus groups and verbal informed consent for interviews. We also obtained a Certificate of Confidentiality from NIH. Procedures were approved by the Yale Human Subjects Committee and the Miriam Hospital IRB.

Facilitators for focus groups and interviews used written agendas to guide discussions. These agendas included specific modules on PrEP efficacy messaging. Specific probes discussed interpretations of percentage-based information, paraphrased information (e.g., “highly effective,” “usually effective,” “sometimes fails”), efficacy ranges compared to single point estimates (e.g., “PrEP ranges from 44%–92% effective depending on whether you take pills every day” vs. “PrEP is around 92% effective when you take it every day”), success- and failure-framed information (e.g., “PrEP is 92% effective” vs. “PrEP is 8% ineffective”), and messaging preferences. Table I provides sample messages for each of these distinctions. Efficacy estimates were based on the iPrEx trial of tenofovir with emtricitabine for HIV prevention in MSM (4). We used the word “effective” instead of “efficacious” in spoken messages to encourage comprehension among participants with less formal education. Facilitators provided participants with additional information about PrEP, including results of other PrEP trials, FDA approval, side effects, and dosing.

Table I.

Summary of Message Frames and Example Messages

Percentage-based versus paraphrased efficacy information
Percentage-based PrEP is 92% effective when used every day. PrEP can range from 44% to 92% effective depending on adherence. When used daily, PrEP has an 8% failure rate.	Paraphrased PrEP is highly effective. PrEP can range from somewhat effective to highly effective depending on adherence. When used daily, PrEP has a low failure rate.
Range-based versus point estimates of efficacy
Range-based The effectiveness of PrEP can range from 44% to 92% depending on adherence.	Point estimate The effectiveness of PrEP is 44% when it is taken imperfectly. The effectiveness of PrEP is 92% when it is taken every day.
Success-framed versus failure-framed efficacy information
Success-framed PrEP is 92% effective when it is used every day. PrEP is 44–92% effective depending on adherence.	Failure-framed PrEP is 8% ineffective when it is used every day. PrEP has an 8–56% failure rate depending on adherence.

All sessions were audio-recorded, transcribed, entered into NVivo 9 (46), and thematically coded. Coding structures were developed separately for each stage and included codes derived from our interview and focus group agendas, as well as codes derived from unanticipated findings. We used an interpretive description paradigm to analyze coded data: we describe participants’ experiences of framed messages, and we used an inductive approach to develop conceptual understandings of these experiences (47–50). Two independent coders double-coded all focus groups and resolved discrepancies by discussion. A subset of 8 interviews was initially double-coded by the principal investigator and 1 of 2 research assistants, who were trained through group coding and discussion of 3 transcripts. The purpose of this double-coding approach was to encourage discussion and refinement of the coding structure (51), particularly for identifying codes derived from unanticipated findings. Disagreements in double-coding of the interview transcripts were resolved by discussion, consensus, and modifications of the coding structure; because codes were not independently applied, an inter-coder reliability estimate was not computed (52). After finalizing the coding structure, the lead investigator coded the remaining transcripts.

RESULTS

Table II reports sample characteristics, and Table III reports a summary of results. Illustrative quotes are included in text below. We identified seven thematic categories related to PrEP efficacy messaging: (1) comprehension of efficacy estimates from clinical trials, (2) reactions to efficacy estimates from clinical trials, (3) differences between generalized estimates of efficacy and personal effectiveness expectations, (4) preferences for numerical information versus paraphrased information, (5) preferences for receiving efficacy ranges versus point estimates, (6) preferences for success-framed or failure-framed messages, and (7) suggestions for tailoring efficacy messages.

Table II.

Sample Characteristics

Characteristic	Focus groups (n=38)a	Individual interviews (n=56)
Mean age (range)	38.5 (21–61)	34.8 (21–70)
Race
African American	21%	16%
White	79%	77%
Native American	0%	4%
Asian	0%	2%
Refused	0%	2%
Hispanic/Latino	11%	16%
Education
Did not complete high school	18%	25%
High school or GED only	37%	32%
Some college	26%	29%
Completed college	18%	14%
Employment
Disabled	26%	9%
Unemployed	45%	45%
Part-time or seasonal work	10%	23%
Full-time work	18%	16%
Annual income less than $12,000	48%	36%
Housing
Homeless	11%	18%
Staying with friends/family	18%	32%
Renting	66%	41%
Owns	5%	9%
Health insurance
No insurance	42%	52%
Public insurance	39%	23%
Private insurance	18%	25%
Sexual identity
Gay or ”mostly gay”	37%	30%
Bisexual	53%	41%
Straight or “mostly straight”	5%	23%
Other	0%	4%
Did not know	0%	2%
Did not respond	5%	0%
Self-reported HIV status
Unknown	26%	27%
Negative	74%	73%
Diagnosed with STI other than HIV in past 6m	18%	4%
Mean number of male and female sex partners for oral, anal, or vaginal sex in the past 6m (range)	22 (range 2–150)	18 (range 1–150)
Had both male and female sex partners in the past 6m	39%	68%
Mean number of male oral sex partners in the past 6m	19.7 (range 1–135)	14.1 (range 1–149)
Mean number of male anal sex partners in the past 6m	10.7 (range 1–45)	8.5 (range 1–80)
Had a primary partner in the past 6m	42%	48%
Sex under the influence of alcohol in past 6m	66%	46%
Sex under the influence of drugs in past 6m	66%	50%

^aWe report focus group and interview samples separately because several focus group participants also enrolled in individual interviews, and the data collection time periods were distinct.

Table III.

Summary of Qualitative Themes

Comprehension of PrEP Efficacy Estimates Varying operational understandings of efficacy rates from trials (e.g., 92% efficacy for all users vs. 100% efficacy for 92% of users). PrEP failures due to user behavior or unique physiological features.

Reactions to PrEP Efficacy Estimates from Clinical Trials 92% efficacy (iPrEx trial among MSM with detectable drug levels) was acceptable. Lower efficacy estimates (e.g. 44%) often unacceptable given side effects and cost. PrEP was acceptable at several lower boundaries of efficacy, ranging from 50% to 100%.

Development of Personal Effectiveness Expectations Generalized estimates of efficacy do not apply to all PrEP users due to behavioral and biological differences. Many expected personal effectiveness of 90–100% due to high predicted adherence, low predicted risk behaviors, “average” physiology. Others said they could not use efficacy estimates from clinical trials to predict their own personal effectiveness.

Numerical Information versus Paraphrases Most preferred percentage-based information for credibility, ease of comprehension, ease of comparison, and as proof of scientific basis. Some preferred paraphrases to improve understanding and limit concern about failure. Offering both message types will ensure that messages reach a diverse audience.

Ranges versus Point Estimates for Communicating Efficacy Most preferred hearing a full range of efficacy for completeness and as a warning against poor adherence. Others were deterred by ranges because lower-bound efficacy estimates reduced acceptability, multiple percentages may be confusing, and ranges may encourage excessive dosing to get “extra” effectiveness.

Success-Framed versus Failure-Framed Messages Many preferred success-framed messages for increased optimism, comprehension, and easier comparisons. Others preferred both message frames, particularly during clinical consultation, for increased message credibility and strengthening of adherence and condom use motivations. A few participants preferred failure-framed messages.

Suggestions for Tailoring Efficacy Messages to Users Visual aids and multimedia are needed to communicate efficacy to individuals who are not skilled with percentages. Different message framing strategies may be needed for different stages of PrEP use (e.g., initial consultation vs. ongoing use).

Comprehension of PrEP efficacy estimates from clinical trials

Facilitators provided participants in both focus groups and individual interviews with efficacy results from clinical trials, most notably results of the iPrEx trial among MSM. Participants reported several operational interpretations of percentages. Many suggested that an efficacy rate of 92% meant that 92% of the users were 100% protected, while the remaining 8% were not at all protected. When asked what would happen to the unprotected 8% of PrEP users, these participants sometimes said that that those individuals would certainly get HIV, and sometimes said that they would simply be at risk for getting HIV.

• I think when you look at it sociologically … one out of the ten [PrEP users] would be unprotected from HIV, but when you think personally, um, you think 90% of the time that you’re gonna be protected. (INT115)

Participants often explained the PrEP failure rate by saying that this 8% of PrEP users must not have taken PrEP as directed, or they must have been engaging in riskier behaviors. Sometimes men also offered a biological explanation for the 8% failure rate, suggesting that PrEP does not work for some users due to physiological differences.

• When you say 90% effective [it] could mean that … 10% of the community has something in their system that it’s not going to work, not going to allow it to work. (INT126)

• 92% of [PrEP users] will not get HIV infected. [The other 8%] either didn’t use it properly or they’re [engaging in] super risky behavior. (INT151)

Besides this group who believed that PrEP would work for only 92% of men, an equal share of participants suggested that PrEP would protect all users 92% of the time, or in 92% of their HIV exposures, or with 92% of their sexual partners. One participant interpreted efficacy to mean that PrEP would reduce population-level HIV incidence by the given percentage among a PrEP-using population.

• [If 10 people take PrEP at 90% effectiveness,] it’s effective 90% of the time for, for all ten of the people. (INT111)

• [I[f you told me it was 90% effective… I would personally look at like, okay, well, 90% of the behaviors that I do, uh, is covered. (INT102)

Reactions to PrEP efficacy estimates from clinical trials

When we discussed the overall iPrEx efficacy level of 44%, many men in our sample stated that they would be unwilling to use PrEP without higher efficacy. Unwillingness to use PrEP at 44% efficacy was particularly pronounced when we discussed PrEP’s side effects and potential cost, which we estimated at approximately $700 per month for out-of-pocket payment; these facts sometimes led men to compare PrEP unfavorably to condoms, noting that condoms would be both cheaper and more effective.

• If it gets below 50%, it’d be a problem. ‘Cause then … I might as well save my money or co-pay or whatever for the pills ‘cause … you’re taking a 50/50 chance…. I would say, like, above 70[%] I would take it. (INT114)

But when facilitators then told all participants that PrEP was 92% effective in the iPrEx trial among the subset of MSM with detectable drug levels, participants almost always said that a 92% efficacy level made PrEP acceptable for personal use. Although cost would still be a barrier to PrEP use, many men noted that the risk-reduction benefits of PrEP at 92% efficacy would outweigh the burdens of side effects and daily adherence.

• If they said 80% I would still … use a condom every time just because … 20% is a good chance to fail, so. But, you know, 90% … although there's still a 10% [failure rate], it's highly unlikely that you'll be in that 10%. (INT129)

When we asked men to describe the lowest level of efficacy at which they would use PrEP, estimates ranged from 50% to 100%, although most clustered at 50%, 75%, or 90% (most common).

• I know nothing is 100%, but [if] you reduce your risk 75% by taking this drug, that to me would be what I’d be looking for… at least give me a 50/50 benefit. (FG101)

People hear anything over a good 80%… maybe a good 75% … they wanna try it … [O]nce you hit me with that 90, I’m willing to go for it. (INT123)

Distinctions between clinical trial efficacy and personal effectiveness

Facilitators asked participants whether they believed PrEP would be equally effective for all users, and if not, probed for explanations. We also asked participants to estimate how effective PrEP would be for them personally, including effectiveness for long-term use. A majority of participants believed that generalized estimates of PrEP efficacy would not apply to all PrEP users due to a variety of factors. These included variations in adherence and sexual risk behaviors, differences in personal biology, racial differences, the user’s age and other health conditions (e.g., allergies or hepatitis C), potential interactions between PrEP and other medications, and interactions between PrEP and recreational drugs or alcohol.

• I think it would work differently for different people. As does every medication work differently for different people's body chemistry, you know, physical makeup, and, you know, allergies, and everything else that creates factors. (INT102)

• When it did fail, then what were the circumstances?…. What kind of lifestyle was the person living, you know? …. [W]ere they on alcohol, were they on drugs, were they taking… other medication? (FG134)

In keeping with this logic, some participants said that they could not use efficacy estimates from clinical trials to predict their own personal PrEP effectiveness, citing conditions such as hepatitis C or use of other medications.

• Um I don’t know [how effective PrEP would be for me]…. As of right now, with the hep C, I, I don’t have an answer. I wouldn’t know ‘cause I don’t know if, you know, if my body would process it properly. (INT143)

• Okay, so if I take this drug it’s, it’s gonna [be] effective 90% of the time…. Do I know if I’m that 90 or do I know if I’m that 10? I don’t until I look behind and say, “Oh, fuck, I got [HIV],” you know, or I don’t. (INT130)

In contrast, many participants drew on the iPrEx findings to develop personal effectiveness expectations of 90% to 100%. These participants gave several reasons for expecting high effectiveness, including the belief that they have “average” physiological characteristics, successful use of medications in the past, and predictions about their own high adherence or low sexual risk-taking.

• Everyone's biology is somewhat different…. [My personal PrEP effectiveness would] probably [be] 90%. Um, unless there is data that says [people of my race] react differently to this drug…. I probably would fit into the average profile. (INT156)

• In my sexual life I think it would be very, very, very effective… because I… I don’t sleep around…. [My personal effectiveness would be] 98%, 99%. (INT112)

Only one participant in this study was using PrEP, and had been using it for almost one year. When we discussed PrEP efficacy, this participant did not recall the efficacy estimates from clinical trials; he estimated his own personal effectiveness, however, to be close to 100% because he reported never having missed a dose.

After we discussed the iPrEx study, some participants suggested that if they took PrEP personally, they would ask their doctors to test their own blood for detectable drug levels to verify that they were protected from HIV. This suggests uncertainty about personal PrEP effectiveness, even among men who reported high expectations.

• FG121: I would get a follow-up within that three-month period [after starting PrEP] as far as a blood test, because they can run a blood test and see, you know, the levels. Facilitator: The levels of the drug in your blood? FG121: Yeah. FG129: Oh, I would definitely do that too. I would want to see my levels.

When we discussed long-term PrEP efficacy, participants generally believed that PrEP would maintain its high efficacy rates over long-term use, assuming no scientific studies had shown otherwise. This question also prompted several participants to ask whether PrEP users would build up a tolerance to the medication over time, which they assumed would require increased drug dosing or a change of medication. Facilitators informed participants that clinical trials had not demonstrated the development of tolerance to PrEP.

• If this is new, how do you know it's gonna work five years down the road? …. You hope that it would be effective. (INT151)

Numerical information versus verbal paraphrases

Participants had a range of interpretations of verbal paraphrases about drug efficacy. For example, participants estimated that a “usually effective” drug would have an efficacy level ranging from 50% to 90%, with most estimates at approximately 60% to 70%. Participants estimated that a “highly effective” drug would be approximately 80% to 95% effective, with most estimates at 90% or higher. A drug that “sometimes fails” was estimated to have a failure rate of approximately 20% to 80%, and many participants noted that this messaging would make them unwilling to use a medication. When asked to create a verbal description of efficacy for PrEP, given the iPrEx estimate of 92% efficacy among men with detectable drug levels, participants said that they would describe the drug as “highly,” “very,” “very highly,” or “extremely” effective.

Facilitators asked participants whether they preferred to receive percentage-based information or verbal paraphrases about PrEP efficacy. Most participants preferred percentage-based information. These participants believed that percentages were more credible than verbal paraphrases, that percentages were easier to comprehend, and that percentages facilitated comparisons of PrEP with other prevention strategies such as condoms. Participants also treated percentages as proof that PrEP recommendations are based on clinical trials, reassuring them that PrEP is an evidence-based strategy for prevention.

• I think I like the percentage figure, I guess. I feel like it's more specific, I guess, and, and, uh, I feel like it seems more credible … It sounds more based on stats and figures and, and like trials and all that kind of stuff.… There’s some research behind it. And I wouldn't want to feel like a guinea pig or whatever for, for something like this, I guess. (INT145)

A smaller number of participants preferred to receive efficacy information as verbal paraphrases. These participants noted that numbers could be difficult to understand, particularly for less educated individuals. Some also suggested that hearing an efficacy rate that was less than 100% would make them worry about the possibility that PrEP would fail; to avoid these worries, they would prefer to simply hear a verbal paraphrase that did not call attention to failure rates. One participant also advocated paraphrases on the grounds that describing PrEP’s high efficacy of 92% could encourage increased sexual risk-taking among PrEP users.

• Some [people] don’t wanna hear the numerical part. They like to hear, “Oh, it’s highly effective,” …. And some … can’t even figure when you say 90%. They don’t get it. They just want you to talk about it and break it down, you know, verbally. (INT154)

• I have more a sense of a confidence in “highly effective” rather than hearing 90% ‘cause when I hear 90% … I’m automatically like, “[W]hat about the other 10%?” But then [when I hear] “highly effective” … it doesn’t leave that doubt in your mind, where you’re wondering, like, where the other percent is. (INT111)

Finally, several participants emphasized the need to make information available in both numerical and verbal formats to ensure message comprehension by a range of potential PrEP users.

• I think if I was hearing words I’d probably want numbers, and if I [heard] numbers, I’d probably want it in words, you know what I mean…. I think it’s, I think it’s a marriage of the two that gets the best result. (INT106)

Efficacy ranges versus point estimates

Facilitators asked participants if they preferred to receive efficacy information in a range (i.e., 44% to 92% efficacy depending on adherence) or if they would prefer to receive a single point estimate (i.e., 92% with high adherence). Most participants preferred to hear the entire range of PrEP efficacy, including estimates for both typical use and use with high adherence. Participants with this preference said that range-based information was more complete and even more honest. They also believed that hearing a lower bound for PrEP efficacy would serve as a cautionary message against poor adherence, and that range-based information would help people to understand that PrEP users have control over the amount of benefit they obtain from PrEP use.

• INT144: The range works because you know that’s, that’s how much of a window you got so like it’s something…. If, if you do what you’re supposed to do, one, all, every day, then it’s here percentage. And then from there down, that’s what you have to worry about if you, if you don’t do it every day. Facilitator: So it kinda shows them that they have control over it. INT144: Yeah. Yeah, exactly. You have control over what you're doing to your body.

Some participants, however, found range-based information confusing, overwhelming or even discouraging when considering PrEP use. For these participants, hearing lower-bound efficacy estimates undermined their confidence in PrEP, particularly among those who wanted PrEP to be “perfect.” Participants who were suspicious of range-based information preferred to receive a single point estimate—namely, 92%—but with the caveat that this efficacy is only possible with high adherence.

• [Giving the whole range] is kind of misleading. Not misleading, but I think it diminishes the effectiveness of it…. That would puzzle me. So I think if I take this I’m, it's only gonna work 44% of the time, but maybe go up to 92%. Don't you think that's a big, big variance? I feel [that's too confusing]…. Put a positive spin on it….You're trying to get [PrEP] to the right people. (INT151)

• [I want to] hear one number. ‘Cause a range, I mean it’s not guaranteed if you have a range, I don’t know. I want it to be perfect. (INT142)

One participant believed that range-based information may encourage men to misuse PrEP by exceeding the recommended dosage, on the theory that “if a little’s good, a lot must be better.”

• If a little's good, a lot must be better, so if you tell me it's 40% effective when I don't take it regularly, 80% effective if I take it regularly, well, I'll take an extra pill and then I'll get it tonight, and it’s effective…. [There] really has to be some -- some focus on making sure people understand that this isn't something you take extra. (FG106)

Success-framed versus failure-framed messaging

Facilitators offered participants examples of a success-framed and a failure-framed message about PrEP efficacy, then asked participants to discuss which type of message they thought would be more useful to them. Many participants in both focus groups and individual interviews described a strong preference for success-framed messages or two-sided messages (describing both a success percentage and a failure percentage), rather than failure-framed messages alone. Participants who preferred the success framing alone suggested that it increased their optimism about PrEP’s potential benefits, and they described optimism about PrEP efficacy as a necessary condition for making the commitment to obtain and adhere to PrEP. They also suggested that success-framed messages were common for other HIV prevention strategies (e.g., condoms), and that framing PrEP efficacy in terms of success rates would facilitate comparisons with condoms. These participants also noted that it is common to see success-framed messages elsewhere (e.g., gambling, efficacy of other medications), so it would be counterintuitive to present efficacy in any manner other than success framing.

• I want the, the effectiveness rate, not the defectiveness rate…. Both of them are the numbers and they mean the same thing, but you know I, I’d rather hear it in the positive you know…. Why, why would you wanna hear the negative things you know. (INT130)

• If it’s gonna be one I’d say success…. [Success-framed messages are] how I’m used to hearing that type of [information], you know, with medications. That’s how they normally say it. (INT105)

Many participants preferred only success-framed messaging about PrEP efficacy, but a second group wanted a two-sided message that incorporated complementary failure rates in addition to success rates. These participants emphasized that providing two-sided messaging would be more honest (particularly in conjunction with information about side effects). Some of these MSM also believed that PrEP users who received information specifically reporting PrEP failure rates would be more likely to adhere to the medication schedule or to continue using condoms, motivated by the fear that PrEP will fail.

• I like hearing both … [Without both] you could be misguided…. If you only hear good things about something… you might become thinking you’re invincible. You might not buy condoms anymore, you know…. [Hearing the failure rate] lets you know where you stand… Just know to be cautious. (INT135)

Participants differed in their preferences for two-sided messages, however. Some said that two-sided messaging is unnecessary when people can calculate failure rates themselves, and one participant thought a two-sided message would be offensive because it would imply that he is unable to calculate the percentage himself. A few participants also suggested that failure-framed messages could deter some people from using PrEP, even when counterbalanced with success-framed messages.

Very few participants reported a preference for receiving solely failure-framed information. Reasons for preferring failure-framed messaging included the belief that failure rates were more honest or explanatory than success rates, and the belief that lower absolute numbers were easier to understand (i.e., it was easier to comprehend an 8% failure rate than a 92% success rate).

• I would definitely prefer the “didn't work” [message] because the percentage is going to be lower then, you know?…. I'd rather … know what's not going to work than what is. (INT139)

Suggestions for tailoring efficacy messages

A number of participants suggested that a range of messaging strategies would be important to ensure comprehension among MSM and other potential PrEP users during both outreach and clinical consultation. Participants emphasized the need for flexible messaging throughout discussions about message comprehension, numerical information versus paraphrases, range-based information versus single point estimates, and success- or failure-framed messages. Facilitators also separately asked participants to suggest alternative messaging strategies to communicate about PrEP efficacy. Many participants suggested that people will differ in their preferences for each of the message frames described above, and a recurring theme was that potential PrEP users will have difficulty understanding numerical information. Participants often noted that they personally found numerical information hard to use, or if not, that their peers may have difficulty interpreting percentages. They suggested that PrEP educators and prescribing clinicians should offer visual representations to demonstrate drug efficacy, such as pie charts, bar graphs, or pictures of stick figures in different colors. Some suggested videos to explain efficacy in a range of ways for people who learn best through combinations of auditory and visual information, and others believed that testimonials from PrEP users would be the best means of instilling confidence about drug efficacy. Others noted the need to verify that individuals understand efficacy information, and to offer messaging information in a variety of ways over time until comprehension is complete.

• [T]here's the people out there … they're not good with uh percentages and can't figure all that out so you kind have to [use] layman’s terms for ‘em … Pictures, pies, the graphs, you know like they show in junior high or elementary school when you first start learning the fractions … that would probably help. (INT149)

Some people are visual … they need to see a visual … a picture versus getting words…. Some people are auditory and they have to hear it. (INT119)

The one participant who was using PrEP believed that separate messaging strategies would be necessary to promote initial PrEP uptake compared to long-term adherence and safer sex among PrEP users. This participant suggested that people merely considering PrEP use should receive messages using success framing and single point estimates of high efficacy to motivate initial interest and uptake. But once people decide to use PrEP, this participant recommended using failure framing and emphasizing that PrEP can have a range of efficacy rates depending on adherence. The participant suggested that this staged messaging strategy might maximize uptake, but also optimize adherence and condom use behaviors among PrEP users who want to obtain maximum efficacy and protect against HIV in the event of medication failures. Several other participants also emphasized the need for separate messaging strategies during initial outreach compared to clinical consultation, on the basis that uniformly positive (success-framed) messages are needed to motivate initial interest, but that clinicians have an obligation to provide more complete information to individuals considering personal PrEP use.

• I think to not deter people from it, I think giving the 90 percentile … to those who haven’t taken the medication would probably be more useful because that just describes the fact that, you know, it is highly effective. But I think for those who are on it… I think that that 10% [failure rate] needs to be addressed … because I think that people who are on it really feel like it’s the ultimate safety and protection … [and] it kinda leads to the condom use situation where I think, you know, you should probably, um, heed the warning… and try to use condoms as much as possible. (INT115)

• When you wanna get people’s attention on TV or advertising, then you just stick to the pros [success-framed messages] just to get, you know, people who might be scared … or people who are just not easily entertained, you know. You give ‘em all the positive stuff, you know, to get their attention …. But with the information that you’re provided along with the prescription, then you should have all the information. (INT135)

DISCUSSION

This qualitative study explored message comprehension and preferences for framed messages to communicate about PrEP efficacy with MSM, who are a key user group for PrEP. Most participants reported that PrEP would be unacceptable for personal use if it were only 44% effective for HIV prevention, but that PrEP would be acceptable at a lower bound of approximately 75% efficacy; this made TDF-FTC highly acceptable at the 92% efficacy level reported among iPrEx participants with detectable drug levels. This result aligns with a prior study that found high PrEP acceptability with efficacy hypothesized at 80% (53), and a study finding that most MSM would require 80% efficacy to use PrEP during unprotected sex (36). Participants offered differing interpretations of percentage-based information about PrEP efficacy, and many were uncertain about how to use trial results to predict PrEP effectiveness for personal use (often citing concern about biological differences or interactions with other medications or conditions). Where participants made predictions about their personal PrEP effectiveness, expected efficacy was approximately 90–100%. Given that the overall estimate for efficacy among MSM in the iPrEx trial was 44%, these high expectations may reflect optimism bias (54), which may be stronger for men who perceive PrEP effectiveness to be controllable through adherence and other behaviors (55).

Participants reported preferences for numerical information about drug efficacy, which differs from preliminary findings that men were more motivated to use PrEP after receiving “gist”-based information from health educators (44). Where verbal paraphrases are used, participants suggested using the language “highly effective” to describe 92% efficacy. Participants were divided on whether they would prefer to hear an efficacy range or a single point estimate. This may support prior findings on risk communication, which have suggested no difference in credibility and perceived accuracy when risks are presented as a point estimate, range, or both (56). Participants generally preferred success-framed and two-sided messages compared to failure-framed messages about efficacy, citing the need for optimism in order to motivate initial PrEP uptake. This finding is consistent with prior research on condom messaging, which found that success-framed messages increased condom acceptability (40). Similar findings have been observed in other fields, such as messaging about the benefits of surgery (57). Many participants also emphasized the need for a variety of messaging strategies. They suggested that giving information in a range of formats may improve comprehension, increase PrEP interest among potential users, improve message credibility and completeness during clinical consultation, and increase adherence and condom use among current users.

These findings have several implications for PrEP implementation, including PrEP outreach, education, and prescription. First, the high levels of efficacy observed in PrEP trials among users with detectable drug levels may indeed be acceptable to potential users, even when overall efficacy estimates (including non-adherent users) are less acceptable. PrEP education should incorporate upper-bound efficacy estimates to maximize initial PrEP interest. Second, our findings highlight the importance of providing percentage-based information about drug efficacy. Even if numeracy among potential users is limited, they may value percentage-based information as proof that PrEP recommendations are based on scientific research; potential users may also see percentages as more credible and unbiased than paraphrases, particularly when these messages are provided by clinicians. Third, we note that trial results alone may not provide adequate information for many MSM trying to formulate their own expectations of PrEP effectiveness; many men in our study had many concerns about predicting personal effectiveness, with questions about interactions with other medications or substances, racial differences in biology, or interactions with other conditions (e.g., hepatitis C, allergies). Clinicians should be prepared to address these questions to the fullest extent possible when prescribing PrEP.

Our results also have implications for framing PrEP efficacy messages. We have identified potential advantages and disadvantages for communicating about efficacy using a percentage range (e.g., “44–92% depending on adherence”) instead of a point estimate (e.g., “92% with good adherence”). Range-based information may strengthen users’ adherence motivations and may be perceived as more honest, but it may also deter some individuals from using PrEP if it is confusing or undermines confidence in drug efficacy. We observed a similar effect of communicating with failure-framed, success-framed, and two-sided messages. These results suggest that optimism about drug efficacy plays an important role in willingness to use PrEP, and leading with success-framed messaging can increase PrEP acceptability. We suggest that PrEP educators initially use success-framed messages to encourage interest in PrEP uptake. But during clinical consultation, we suggest that clinicians present both success-framed and complementary failure-framed messages in order to maximize credibility and highlight the need for drug adherence and continued risk-reduction behaviors. Two-sided and range-based messages may be particularly useful for follow-up with PrEP users, given the need for medication adherence. Prior research on two-sided messaging suggests that these messages may lead to increased certainty about decisions (43), which could be useful in the PrEP context. We also suggest providing visual information about drug efficacy (e.g., pictures, pie charts, video images) in order to facilitate comprehension by a range of potential users.

This is the first qualitative study to explore PrEP efficacy message comprehension and messaging preferences among MSM. Our use of focus groups followed by individual interviews allowed us to capture both initial themes and in-depth narratives about participants’ preferences, and our sample sizes for both qualitative stages allowed us to reach data saturation on all themes related to efficacy messaging. We have highlighted emergent (unanticipated) findings, which are considered particularly robust because they are unprompted by facilitators. These findings included the concern that some PrEP users will exceed the recommended drug dosage if they interpret range-based messages to mean that higher dosing will yield more protection; the possibility that user testimonials may be a more powerful way to communicate about drug efficacy than describing trial results; and the potential need for different messaging strategies during initial PrEP outreach, clinical consultation, and PrEP user follow-up.

Our results also have limitations. Our sample tended to be low-income, high-school educated, white, non-Latino, English-speaking, and non-gay-identified, which may limit generalizability. A small number of focus group participants also enrolled in interviews, and the focus groups occurred before FDA approved TDF-FTC for PrEP in 2012 (although off-label PrEP prescription was permissible at the time). We used consensus methods to resolve differences in coding, but the lack of an inter-coder reliability statistic may also be a limitation of our approach. We did not evaluate numeracy skills, and future research should consider numeracy as a moderator of message comprehension and messaging preferences. This qualitative study was not designed to identify differences by education level, race, preferred language, or any other characteristic, and future research should consider longitudinal or cross-sectional survey approaches to clarify these differences. Additional areas for further research include how range-based messaging, point estimate messaging, and success- or failure-framed messaging may actually influence PrEP uptake, adherence behavior, and risk behavior; how best to design and evaluate a staged messaging approach, with different message frames for initial PrEP outreach, clinical consultation, and long-term follow-up; and how message framing might influence PrEP acceptability and PrEP user behaviors in other high-priority populations, including at-risk women, serodiscordant couples, and people who inject drugs.

These findings also indicate the need for additional research with a careful focus on language for communicating about PrEP. Future investigations should also consider the extent to which individuals’ self-reported message preferences predict PrEP uptake, expected efficacy, adherence, and ongoing risk behavior; prior work in message framing suggests that it may be difficult for individuals to predict the messaging strategies most likely to influence them (58,59). For example, consistent with the elaboration likelihood model of persuasion (60), messages framed in an unexpected way can increase the extent to which people scrutinize them carefully (“elaboration”), which may in turn make such messages more persuasive (58,59). If individuals at risk for HIV infection expect PrEP efficacy to be described using success-framed and percentage-based information, using the opposite message frames may counterintuitively be more persuasive. Prior research also suggests that the effects of presenting range-based or point estimates may interact with success- and failure-framing, such that ranges may be more persuasive in the context of failure-framed messages (61). Research on the communication of risks has shown that providing a point estimate along with a range can influence risk perceptions more than providing either alone (56), but additional research is needed to understand how this approach may influence the receipt of messages about benefits. Randomized trials are needed to identify the effect of framed messages describing PrEP. Message framing can play an important role in maximizing PrEP acceptability and optimizing PrEP user behaviors, with implications for the entire generation of emerging biomedical HIV prevention strategies.