We conducted a review of the scientific literature on publicly available (eg, Google Scholar and PubMed) and university-specific (eg, Columbia University libraries) databases to identify barriers and facilitators of contraceptive implant use among cisgender women. We focused our review on contraceptive implants, as this device is most similar to the one in development by SLAP-HIV. Furthermore, there is extremely limited data on cisgender men’s experiences with implants to deliver gender-specific medications (eg, the Testopel testosterone implant); these implants also operate differently than those under study in SLAP-HIV (eg, Testopel is a nonremoveable biodegradable pellet implant) [12]. We limited our search to studies published in the last 10 years (January 1998 to July 2017). Search terms included “contraceptive implant barriers,” “contraceptive implant removal,” “contraceptive implant facilitators,” “contraceptive implant experiences,” and “contraceptive implant continuation.” We limited the search to studies published about US populations and manuscripts on arm-implant contraceptives (eg, studies focused only on intrauterine devices [IUDs] were excluded).

The primary author (CR) and a research assistant (CL) abstracted relevant studies and independently reviewed their full content to determine their applicability to the topic under study (eg, device-related barriers and facilitators to contraceptive implants). CR and CL met in person to discuss the results of the independent reviews, resolve a disagreement about study relevance in concert, and identify studies to move forward with. After the studies to be included in the final review were identified, CR tallied the specific mention of device-related and contraceptive medication−related barriers and facilitators in a word document. There was very little mention in the scientific literature about the device-related effects of contraceptive implants. Thus, stage 2 (summarized below) intended to learn more about users’ device-related subdermal implant experiences, as this will be important for our target population, GBM (stage 4).

A total of 2 members of the research team identified a collection of potentially relevant YouTube videos through a manual search. Search terms included lay and commercial names of contraceptives: “Nexplanon,” “Nexplanon insertion,” “Jadelle,” “Jadelle implant,” “Nexplanon birth control,” “Norplant,” “Implanon,” “Implanon birth control review,” and “Arm bar birth control.” We excluded sponsored testimonials, insertion or removal footage that included no assessment of the patient’s experience, clinical videos (ie, procedural videos on how to insert or remove implants), promotional or educational videos (ie, videos explaining what contraceptive implants are or how they work), and advertisements for any product. Then, spoken word transcripts of potentially relevant videos were automatically generated using YouTube’s state-of-the-art speech recognition algorithms [13].

To identify biomedical terms in YouTube transcripts, we used a publicly available natural language processing system called National Center for Biomedical Ontology Annotator, maintained by the National Library of Medicine [14]. The system recognizes sets of words and phrases using string matching and regular expressions that operate directly on text without preprocessing, such as spelling correction, part-of-speech tagging, and word sense disambiguation [15]. To identify the various side effects associated with subdermal implants, we utilized Systemized Nomenclature of Medicine (SNOMED) [16], a comprehensive terminology that provided term-to-concept mapping for a vast array of conditions, symptoms, and clinical findings.

We restricted our search to several types of concepts or semantic types. We identified relevant semantic types by using SNOMED to annotate a string of words typically associated with unpleasant experiences with drug delivery implants. We then reported the frequency of the specific terms describing these unpleasant experiences (eg, contusion and pain). Terms were paired to ensure clarity in the themes to which they were intended to refer (eg, bleeding, menstrual spotting). We then quantified the frequency (eg, number of videos) with which they appeared.

In stage 3, we recruited potential participants using paid Facebook advertisements and posts on contraceptive implant-themed Facebook pages (eg, Nexplanon and Implanon) and other female-centered Facebook groups (eg, Girls Love Travel), with permission from page moderators. We exclusively sampled cisgender women in this stage because we wanted to ensure that the implant participants had used a design that was consistent with those of the PrEP implants in the development pipeline. Subdermal implants commonly available to men (eg, Testopel testosterone pellets, which are a nonremovable pellet implant, placed in the gluteal region) have a much different design [12,17]. This sampling approach had 2 purposes as it would allow (1) PrEP implant developers to incorporate feedback from past or current users of a similar implant into their device design and (2) the research team to be more precise about the questions asked to potential future PrEP implant users in the stage 4 web-based survey.

Participants were invited to click on a link that took them to a brief screening questionnaire to assess eligibility for telephone interviews about their experiences with implants. Eligible participants were at least 18 years old, identified as cisgender women, and had used an implant for contraception in the last 5 years. To sample a wide range of users’ experiences, we selected those who reported that they liked very much or disliked very much their implant on a 4-point Likert scale (1=disliked very much, 2=disliked a little, 3=liked a little, and 4=liked very much). In all, 24 participants were selected for interviews on a first come first serve basis, 12 who liked very much and 12 who disliked very much their implants and could be reached using the telephone numbers they provided.

Telephone interviews were audio-recorded and guided by findings from the YouTube analysis and scientific literature review. Specifically, we focused on the in-depth phone interview (IDI) guide on themes from the scientific literature and those YouTube users referenced frequently (identified using SNOMED). We categorized these themes into 6 broad categories: (1) general questions about participants’ experiences with subdermal implants, (2) decision making around the use of an implant, (3) acceptability of the implant, (4) the implantation process, (5) the removal process, and (6) recommendations for researchers working on a future PrEP implant. Participants were also asked about their reasons for removing the contraceptive implant. Upon completing the interviews, participants were compensated with US $50 Amazon gift cards.

Next, audio-recordings of the interviews were transcribed and cleaned for accuracy. To analyze the data, a codebook was developed jointly by the principal investigator of the project and a research assistant. It incorporated categories and themes from the interview guide and included definitions, inclusion and exclusion criteria, and examples of passages for inclusion. To validate and finalize the codebook, 2 researchers coded 2 transcripts independently, met to reconcile discrepant codes, and modified the codebook as needed to ensure that all themes of interest were being accurately captured. Researchers, then, proceeded to code the remainder of the transcripts in pairs, meeting after every 11 transcripts (half of the remaining data) were coded to reconcile discrepant codes. The finalized codes were analyzed using content analysis [18] to identify prevalent themes, illustrative examples of each theme, and knowledge gaps.

Findings from the IDIs among cisgender women in stage 3 informed the development of a web-based quantitative assessment. This assessment took approximately 3 min to complete and aimed to determine how likely members of key HIV risk demographics (ie, GBM) would be to use an implant for HIV prevention. Specifically, we published text-based pop-up advertisements (known as broadcasts) on Grindr, a dating and hookup app for GBM. Logistically, these advertisements popped up the first time any Grindr user within a specified location opened the app; users who clicked on the advertisement were automatically rerouted to the quantitative assessment. For the purposes of this study, we published 2 broadcast advertisements (per city) in New York City, Baltimore, and Chicago. The quantitative assessment was open to anyone who wished to take it; eligible participants reported identifying as a gay man who was at least 18 years old. We focused our recruitment on GBM (rather than cisgender heterosexual cisgender women) because this population is at comparatively higher risk for HIV exposure and is one of the planned target populations for future PrEP implant studies. To reduce the odds of receiving multiple entries from the same person, participants did not receive compensation for their time, and surveys that had identical Internet Protocol addresses were eliminated. Additionally, for the purposes of the analyses presented in this paper, only HIV-negative GBM who lived in the United States were included.

As the PrEP implant is still hypothetical, we posed a variety of device configurations to participants. Specifically, the quantitative assessment measured how willing participants would be to use a hypothetical HIV prevention implant given a number of different factors: varying lengths of effectiveness (eg, 12 months, 6-8 months, and 2-3 months), number of implantable devices required, potential physical side effects (eg, bruising and scarring) resulting from the implantation or removal processes, implant size, and the convenience of using an implant compared with taking a daily HIV prevention pill (PrEP). All responses were recorded using a 5-point Likert scale, ranging from 1 (I am very unlikely to use it) to 5 (I am very likely to use it). The value 3 on this scale represented a neutral response. Questions were presented in a different random order for each participant to minimize potential bias. Basic demographic, sexual risk, and injection drug use information were also collected using the web-based survey (Multimedia Appendix 1).

To give participants an idea of what a PrEP implant would entail, we embedded photos of the proposed implant next to a dime to help respondents estimate its size. Additionally, we embedded graphics interchange formats (GIFs: a lossless image file format that presents time-lapse video imagery) of the implant insertion and removal processes so that participants could gain an understanding of how they worked.

In the scientific literature, all studies that were identified by the research team focused on cisgender women. Barriers to contraceptive implant uptake and sustained use most frequently included some that do not apply to an HIV prevention implant (eg, irregular or unpredictable menstrual bleeding [19-21]) and others that do apply, such as potential nonmenstrual side effects (eg, headache) [20,21]; perceived or actual high cost of implants [22-29]; patient or provider misconceptions or lack of information about implants (eg, believing that implants are contraindicated for specific patient populations) [21,23,26-32]; and challenges with access to implants [25,27]. Few studies discuss facilitators of the use of contraceptive implants as a birth control method and instead mostly focus on facilitators of implementation of programs designed to encourage implant use [26,33]. However, the scant body of existing evidence on implant facilitators shows that these include low implant cost, same-day insertion [27], and the inherent privacy of implants (eg, you can use this contraceptive method with low risk of others finding out) [29]. We were unable to identify scientific studies focused on user preferences regarding specific implant attributes (eg, size) or device-related effects (eg, bruising and local pain). Consequently, we sought additional sources of information.

Stage 2 summarizes our systematic review of YouTube videos. In July 2017, 272 videos about users’ experiences with contraceptive implants (that did not meet the exclusion criteria specified in the Methods section) had been uploaded to YouTube. Table 1 reports the frequency of videos in which the most often cited terms related to undesirable implant effects appear. These terms include the following: (1) pain, stinging, burning, or soreness and tenderness unrelated to menstruation (51 videos); (2) menstrual bleeding or spotting (33 videos); (3) contusion (32 videos); (4) scarring (25 videos); and (5) bleeding unrelated to menstruation (24 videos). We do not report demographic data as terms originated from videos posted on YouTube, and in the sample of videos accessed for the purposes of this study, posters rarely reported their demographic characteristics.

Participants in IDIs were cisgender women who were aged, on average, 24 years and had used an implant for contraception in the last 5 years. Coders identified several themes within the topics previously recognized using SNOMED and covered in the IDI guide. Topics included the following: (1) specific attributes of implant products that attracted participants to these devices; (2) concerns about the implant; (3) feelings about the insertion process; (4) participants’ observed likes and dislikes about the implant, once it was inserted; and (5) recommendations for future implant designs and placement.

Participants (N=304) in the stage 4 web-based survey were aged, on average, 38 years and predominantly self-identified as gay or homosexual (204/238, 85.7%) cisgender men (231/238, 97.1%); about half (125/238, 52.5%) of them identified as white and a quarter (58/238, 24.4%) identified as Latino. Over two-thirds of the sample were college graduates (89/238, 37.4%) or had attended graduate school (77/238, 32.4%) and reported ever (101/174, 43.4%) or currently (73/174, 30.7%) using oral PrEP. Condomless sex with an HIV-positive or unknown status partner (104/126, 43.7%) was the most frequently reported HIV risk behavior in the past year (Table 2).

Table 3 shows the participants’ preferences related to specific PrEP implant characteristics. Of these characteristics, not having to take a daily pill increased the likelihood of PrEP implant use (mean 4.13). Requiring more than 1 device to achieve 1 year of protection (mean participant rating range:1.79-2.94) mildly discouraged implant use. Participants did not mind the following: moderate bruising, a small scar, tenderness, or bleeding after insertion or removal, or having an implant slightly longer or fatter than a matchstick.

In stage 2, we may have missed some references to specific participant experiences if participants used descriptive words not contained in SNOMED’s database. The interviews we conducted as part of stage 3 were partially meant to address this potential gap by more fully fleshing out users’ experiences with contraceptive implants. As is the case with all research on acceptability and feasibility of hypothetical clinical products, it is difficult to estimate the gap between participants’ perceived likelihood of use and actual use of implants for PrEP. However, data from past and current users of similar implants for contraception in stages 2 and 3 help to mediate this issue.

Additionally, because we were concerned with ensuring that users’ experiences with existing implants were as similar as possible to potential future PrEP implants, we sampled only women in stage 3. In stage 4, we targeted only GBM (as they have a higher risk for HIV exposure compared with heterosexual cisgender women) with web-based surveys on their preferences for PrEP implants. Given the lack of formative work specific to GBM, there may be a gap in the preferences and perceptions of this implant product between these 2 groups. Future studies should conduct qualitative interviews with GBM on their preferences around PrEP implant products to limit this shortcoming. Despite these limitations, given the dearth of research on this topic, this work makes an important and unique contribution to the literature on implants for PrEP.

In all stages of research, participants reported experiencing (stages 2 and 3) or anticipating (stage 4) implant-related side effects. These effects were related to the device itself (bleeding, bruising, and pain after insertion) or, among contraceptive implant users, medication-related side effects (menstrual changes, weight gain, and acne). Although it is highly unlikely that PrEP implant users will experience similar hormonal side effects to contraceptive implant users, it is critical to point out the difference between these 2 products to potential PrEP users. Icon arrays could be a useful way to communicate serious risks related to implant devices. Future studies should monitor for the presence of beneficial side effects related to implant medication, as this has been evidenced with other medications and may improve uptake and sustained use. Furthermore, small implants need only 1 device, last for at least six months, are unintrusive, and require little to no user effort to optimize uptake. Current and potential future users appear comfortable with the currently used procedures for implant insertion and removal. Finally, future work should include more formative work specific to GBM, as this will allow researchers to address their concerns with greater precision.