Exploring Why People With Type 2 Diabetes Do or Do Not Persist With Glucagon-Like Peptide-1 Receptor Agonist Therapy: A Qualitative Study

This noninterventional, cross-sectional, qualitative analysis used face-to-face interviews with adults with type 2 diabetes who had been treated with at least one GLP-1 receptor agonist formulation. The objective of the analysis was to enhance current understanding of participants’ experiences, motivations, and challenges relevant to their persistence with GLP-1 receptor agonist therapy.

The interviews were conducted between 24 January and 1 June 2018 at six clinical sites in the United States (in the states of Alabama, Indiana, North Carolina, Ohio, Texas, and Washington). Inclusion criteria were ≥18 years of age, diagnosed with type 2 diabetes, and currently treated with a GLP-1 receptor agonist for at least 1 month at the time of screening or discontinued use of a GLP-1 receptor agonist within 1 year of screening but with a total of at least 1 month of treatment. The study protocol was approved by a central institutional review board (Quorum Review, Seattle, WA) and performed in accordance with Good Clinical Practice and applicable regulatory requirements. All eligible participants provided written informed consent and demographic data before being scheduled for their interview visit.

Interviews were conducted using a semi-structured qualitative interview guide that included open-ended questions and probes to obtain both spontaneous and prompted input from participants about their treatment experiences with GLP-1 receptor agonist therapy. The guide included discussion items designed to elicit information regarding participants’ reasons for treatment continuation, adherence/nonadherence, or discontinuation. Key items covered initiation of GLP-1 receptor agonist therapy, interactions with their health care team, and psychosocial contributors (facilitators) to continuing or discontinuing therapy. Table 1 presents a sample of the questionnaire items. In response to these questions, “spontaneous offered” was selected on the interview grid when participants offered a response concept on their own. “Recognized probe” was selected if the interviewer asked a probing, follow-up question to elicit a concept response. “No effect” was selected when participants stated that they had not experienced the response concept.

The interview tool was developed by an experienced qualitative research group (Health Research Associates, Inc. [HRA], Mountlake Terrace, WA) in collaboration with the study sponsor. Four interviewers reviewed the content of the interview guides and participated in mock interview sessions with each other (led by senior research staff) to test the flow of the questions to find any problematic, slow, or awkward areas and to test the general timing of the interview.

A training day of observed practice interviews was organized with the HRA and study sponsor team and the study interviewers. Willing volunteers with type 2 diabetes were recruited from a local volunteer list to take part in the training interviews; no participant information or results of the practice interviews were recorded or retained. After the practice interviews, the interview guide was finalized, and changes were submitted to the institutional review board for approval.

Four trained HRA research staff members conducted the interviews in a private room at the enrolled clinic sites. Each interview visit was audio-recorded and lasted ∼90 minutes. On arrival at the scheduled interview session, participants were given a brief introduction to the interview purpose and process. They were reassured about the confidential nature of the interview contents and the personal information they provided. All interviews were conducted in English and directed by the semi-structured qualitative interview guide.

Digital audio files of subject interviews were transcribed and entered into the ATLAS.ti software program v. 7.1.0 (19) for coding. The coding process was structured for thematic analysis, identifying patients’ expressions of concepts, highlighting quotations to tag them in the coding program, and assigning a code stem to them so the data could be organized by similar content and used to build thematic pictures of patients’ responses. As new concepts appeared in patients’ responses in the transcripts, new categories were made to capture them in the coding framework. Some sections of the interview were used to compile actual full responses from patients against specific questions in instances in which showing the spectrum of replies against a research question was important. The interviews included a few rating exercises for symptom severity or symptom-related bother, for which patients were asked to give a rating on a numerical scale ranging from 0 to 10. Data from these exercises were used to develop descriptive quantitative tables.

Saturation of concept was assessed to determine whether a sufficient number of interviews had been conducted. Saturation of concept is reached when no new concepts are being identified in the interview data (20).

Transcripts were ordered chronologically based on interview completion date and then grouped into five groups of seven to eight transcripts each. Saturation was then evaluated by comparing the codes that were derived from the second transcript group with the codes that appeared in the transcripts from the first group to identify whether new information was still forthcoming. These evaluations and comparisons were repeated for each subsequent group. Once new information was no longer forthcoming, it was considered unlikely that further interviews with similar participants would have contributed any additional information or understandings (Figure 1).

Interrater agreement was used to assess consistency of code assignment. Two randomly selected interview transcripts were independently dual-coded and compared for percentage of agreement in the assignment of codes to concepts expressed by participants in the interview process.

Thirty-six participants completed the qualitative interviews. Participants were predominantly older, female, and non-Hispanic White (Table 2). Participants who were currently on GLP-1 receptor agonist therapy at the time of the interviews (continuers) tended to have higher household incomes, shorter durations of diabetes, and lower A1C levels at the start of GLP-1 receptor agonist therapy than those who had stopped therapy (discontinuers). The 16 continuers were being treated with once-daily liraglutide, once-weekly exenatide, twice-daily exenatide, or once-weekly dulaglutide. Among the 20 discontinuers, which included discontinuation to one or more of all of the available GLP-1 receptor agonist formulations, 15 had discontinued GLP-1 receptor agonist therapy >6 months before their interview. The study did not include two recent additions to the GLP-1 receptor agonist class: once-weekly semaglutide and once-daily oral semaglutide.

The number and percentages of continuers (n = 16) who identified primary facilitators for continuing GLP-1 receptor agonist use are presented in Table 3. The most commonly identified group of facilitators reflected participants’ perception that the treatment was contributing to tangible, positive results (most importantly, glycemic control and weight loss). Some of the verbatim responses recorded during the interviews with GLP-1 receptor agonist therapy continuers are shown below.

The other major category of facilitators represented an appreciation of reduced or absent treatment burden (e.g., “no side effects,” “available or low cost,” and “ease of use”).

The most commonly identified challenges leading to treatment discontinuation reflected participants’ disappointment with the same two categories—in this case, lack of treatment efficacy and/or enhanced treatment burden (e.g., medication side effects or high costs) (Table 3).

In general, both continuers and discontinuers broadly reported supportive experiences with their health care provider team when initiating a GLP-1 receptor agonist. However, it appeared that continuers were more likely than discontinuers to receive clinically relevant information from their health care team, including facts about GLP-1 receptor agonist medications, likely treatment benefits (e.g., how the new treatment would contribute to better glucose control), the importance of gradual dose titration, and the need to adjust diet after initiation), as well as ongoing support (i.e., the health care team making proactive calls to check on their progress).

Table 4 presents the percentage of participants who confirmed specific statements included in the questionnaire. In most areas, the majority of continuers, in contrast to discontinuers, confirmed that they had good communication with their health care team and that their questions about medication were answered and they were provided information about the importance of gradual dosage titration.

Although recent studies have provided initial insights regarding various reasons for discontinuation of GLP-1 receptor agonist therapy, the contributors to therapy persistence have not been well defined or understood (15). To our knowledge, this is the first qualitative exploration of factors that may affect an individual’s decision to discontinue or continue GLP-1 receptor agonist therapy.

In this qualitative study, we interviewed 36 individuals with type 2 diabetes who shared personal accounts of their experiences with GLP-1 receptor agonist therapy. Despite the uniqueness of their individual circumstances, we identified two broad participant-perceived factors associated with medication persistence or nonpersistence: perceived treatment efficacy and perceived treatment burden.

Regarding the first factor, we see that the majority of participants who continued therapy cited aspects of treatment efficacy, including “numbers improved” (50.0%), “weight loss” (25.0%), and “numbers controlled” (18.8%) as primary facilitators of persistence with their therapy. Conversely, efficacy concerns appeared to influence discontinuation, with 25.0% of discontinuers reporting that “numbers did not improve.” At the heart of this behavioral concept, it is apparent that, when individuals sense that their treatment is contributing to positive, tangible benefits, they are more likely to feel motivated to continue to use it (21).

Regarding the second factor, continuers reported a reduction or absence of treatment burden, with as many as 18.8% of participants citing such factors as “no side effects,” “available or low cost,” and “easy to use” as crucial to why they continued on GLP-1 receptor agonist therapy. Not surprisingly, we see the converse among discontinuers, with “side effects” (55.0%) and “high cost” (50.0%) as the most common primary challenges to therapy persistence.

The interplay between the perceived value of a medication and the perceived concerns (or sense of burden) about that medication is a well-studied phenomenon in the medication adherence literature (22), and both of these factors are seen as crucially important. Although they are typically considered to be modifiable factors, it is well recognized that the biological response to GLP-1 receptor agonist therapy can vary significantly depending on individuals’ unique physiological characteristics, current level of glycemic control, and specific drug formulation (23). Therefore, lack of glycemic improvement, lack of weight loss, and/or intolerability of side effects may simply be an unmodifiable class effect in some participants, which could explain why 25.0% of discontinuers reported “numbers did not improve” as their primary reason for discontinuation. Although “high cost” was reported as a primary factor affecting therapy persistence, this issue can only be resolved through changes in payer reimbursement policies.

Nevertheless, our results suggest that several of these other key issues could potentially be addressed through more collaborative patient-clinician interactions before initiating therapy. For example, it appears that participants who continued with GLP-1 receptor agonist treatment were more likely than discontinuers to receive more comprehensive preparation and instruction at the start of treatment. As reported, 93.8% of continuers (vs. 75% of discontinuers) confirmed receiving information about GLP-1 receptor agonist medications and having their questions answered when starting therapy. Importantly, only 50% of discontinuers versus 75% of continuers confirmed receiving information about the importance of gradual dosage titration. This finding suggests that continuers may have been more likely than discontinuers to be started on a “low-dose” regimen at the time of initiation, which could account for the higher percentage of continuers versus discontinuers who reported “minimal or no side effects” with treatment: 87.5 versus 55.0%, respectively. Moreover, continuers were more likely than discontinuers (62.5 vs. 40%) to receive proactive phone contacts from their health care team during the first few months so that any concerns could be addressed. Through these interactions, participants likely received feedback on their progress, making them more aware of any tangible benefits (e.g., improved glycemic control, weight loss, cardiovascular benefits, and improved well-being) that might be accruing. It could also be that these patients were able to discuss with their health care team any issues they had with the therapy, which could have helped them overcome minor adverse effects.

This was an average sample size for qualitative research studies, and saturation of concept was achieved, indicating that a sufficient number of interviews had been conducted. However, given the large amount of individual variability found in this therapeutic area, the relatively small sample size limits the generalizability of these findings to the larger diabetes population. Additionally, participants’ perceptions and continuation of therapy may have been affected by differences in GLP-1 receptor agonist formulations (e.g., frequency of administration and drug-specific side effects).

A key limitation is that no insurance coverage data were collected. Given that “high cost” was a common obstacle among discontinuers, it is reasonable to assume that more continuers versus discontinuers had insurance coverage for their medications. Additionally, because our population was heavily White/Caucasian, and there were no Black/African American participants in the discontinuer group, our findings cannot be generalized to the larger population of people with type 2 diabetes. Other limitations include self-selection for participation, dependence on participants’ memories and/or articulation of their perceptions and experiences, and inability to exactly determine the range of previous GLP-1 receptor agonist medications that discontinuers may have tried.

Although larger, quantitative studies are clearly needed to confirm our findings and identify characteristics of each GLP-1 receptor agonist formulation that may affect treatment persistence, we believe the information and insights provided by the study participants may be useful to clinicians when initiating GLP-1 receptor agonist therapy, as these results point to strategies that could be adopted by health care providers to enhance perceived treatment efficacy while addressing and potentially resolving treatment burden issues. More specifically, these improvements could be accomplished by thoroughly informing patients about the potential benefits, limitations, and side effects of the medication. Doing this would prepare them with information and strategies to minimize any side effects that may arise. It would also provide ongoing support and feedback to highlight the tangible benefits of their therapy and help address any obstacles they may be encountering. Finally, these findings highlight the importance of insurance providers’ coverage policies and ongoing access to GLP-1 receptor agonist medications, which will likely be a key driver of enhanced therapy persistence.

The authors thank Christopher Parkin, of CGParkin Communications, Inc., for editorial assistance.

Funding for this study was provided by Novo Nordisk.

W.P. has served as a consultant for Astra Zeneca, Dexcom, Eli Lilly, Intarcia, Mannkind, Merck, Novo Nordisk, Onduo LLC, and Sanofi. M.M. is an employee of Health Research Associates, Inc. C.G., N.I., and C.H. are employees of Novo Nordisk.

All authors designed the study, reviewed the data, and wrote the manuscript. M.M. provided statistical analysis. W.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.