FormalPara Key Summary Points

Why carry out this study?

People with diabetes on mealtime insulin (MTI) still have unmet needs related to their MTI.

The qualitative and quantitative phases of the present study aimed to identify impacts related to MTI and the relative importance of those impacts.

What has been learned from the study?

People with diabetes reported a variety of unique impacts associated with the administration of MTI, including psychological and social impacts.

Three of the most burdensome impacts related to using MTI were in the diabetes distress domain (e.g., worrying about extreme lows in blood glucose, worrying when not in target range, worrying about extreme highs) followed by impacts related to diabetes management.

The results of the study suggest people with diabetes on MTI experience a wide range of impacts and that there is not a single consistently dominant impact.

Introduction

Diabetes mellitus (DM) is a group of disorders for which the main clinical manifestation is hyperglycemia or raised blood glucose (BG) value [1]. The International Diabetes Federation (IDF) estimates that 463 million adults worldwide are currently living with diabetes, and the total number is predicted to rise to 578 million (10.2%) by 2030 [2]. In the USA, an estimated 34.2 million people have diabetes; this number is approximately 3.2 million people in the UK [3].

Insulin is a required form of treatment for all people with type 1 DM (T1DM) and for people with type 2 DM (T2DM) who cannot achieve glycemic control through exercise, diet, and other anti-hyperglycemic treatments. While there have been continued advancements in diabetes treatments and BG technology since the discovery of insulin 100 years ago, there are still some unmet needs, many of which are related to bolus or mealtime insulin (MTI) [4]. Some of these ongoing challenges are related to the person’s ability to balance insulin dosing and timing against meal content and size to avoid the risk of hypoglycemia or hyperglycemia, and their associated impacts [5]. Due to these ongoing challenges, continued innovation in the treatment experience and in the management of glycemic control is still needed [6]. Additionally, it is important to explore how these impacts are experienced and valued from the patient perspective.

The objective of this study was to understand the importance that people with diabetes place on the impacts of their MTI experience. This information may help inform the selection of patient-relevant endpoints in future clinical research studies, support healthcare providers’ (HCP) interactions with their patients, and also identify concepts relevant to characterizing the value of existing and future treatments in terms of their ability to address patients’ unmet needs. The study was conducted in two phases. The initial, qualitative phase aimed to identify symptoms and impacts related to the MTI experience. The subsequent, quantitative preference phase aimed to characterize the relative burden of different MTI impacts from the patient perspective.

Methods

Study Design

The study included a concept elicitation qualitative interview phase designed to elicit the symptoms and impacts associated with MTI directly from the patient perspective. Data from these interviews were used to develop a best—worst scaling (BWS) survey, i.e., a conjoint-based preference elicitation method [7]. The qualitative interview study included 30 people with T1DM or T2DM who were using MTI. Participants were at least 18 years old, had a self-reported diagnosis of T1DM or T2DM, were residing in the USA or UK, and had been treated either with an insulin pump or multiple daily injections of MTI for at least 1 year.

The subsequent preference survey used BWS case 1 (BWS1) methodology [7] to assess the relative importance (RI) of the impacts associated with MTI that were identified during the qualitative interviews. Participants included people with T1DM and T2DM in the USA and UK who had been treated with ≥ 2 daily injections of MTI for at least 6 months.

The preference survey portion of the study was designed based on guidance of the International Society for Pharmacoeconomics and Outcomes Research (ISPOR), including cognitive pilot interviews and the main data collection phase [8]. During the cognitive interviews, 20 participants completed the survey and provided feedback on the comprehensibility and relevance of each item in the BWS1 task and were also asked whether the list of impacts associated with MTI was comprehensive. In the main data collection phase, 300 participants were targeted for completion of the final preference survey. To evaluate the potential effects of diabetes type and continuous glucose monitoring (CGM) usage on patients’ perceptions of impacts, we targeted an approximately equal split of participants for recruitment from the following four groups of interest, T1DM CGM user, T1DM non-CGM user, T2DM CGM user, and T2DM non-CGM user.

Participant Recruitment

Participants were recruited to the qualitative and quantitative study through a third-party recruitment vendor specializing in health outcomes research. Specifically, participants were identified through patient panels, social media, patient associations, and the recruiter’s database. Prior to enrollment in all phases of the study, all participants were screened via telephone or online and consented to take part in the study via an electronic informed consent form. Participants were remunerated for their time. Prior to initiation, all study protocols received institutional review board approval (qualitative study—Advarra Study Number: Pro00031152, Approval date: 10 December 2018; quantitative study—Ethical and Independent Review Services Study Number: 20083–01, Approval date: 3 June 2020).

Measures

Semi-Structured Interview Guides

A semi-structured interview guide was used in the qualitative concept elicitation interview phase to facilitate the discussion. The purpose of the interview was to gather information directly from the patient perspective on the symptoms and impacts associated with MTI as well as unmet needs. A semi-structured guide was also used during the cognitive pilot interviews in the quantitative study to confirm comprehension and usability of the survey.

BWS1 Preference Survey

BWS1 methodology [7] was used to measure the RI of 14 impacts (Table 1) identified through the qualitative research [9]. The list of impacts was narrowed down based on the cognitive interviews in the BWS study from the 15 that were originally identified. Impacts were categorized a priori based on the results of the qualitative study into four domains (diabetes distress, diabetes management, work productivity, and social).

Table 1 List of impacts for best—worst scaling survey
Full size table

Prior to completing the BWS tasks, participants read a brief description of the purpose of the survey. This section explained that participants would be presented with a list of impacts and would then be asked to select the impacts that were the most and least burdensome across a series of choices. The participants were instructed that each task would comprise a different set of impacts, with five impacts per task. All participants were presented with the same set of 11 BWS tasks (see Fig. 1 for an example choice task). The order of presentation of the impacts was randomized across both participants and tasks. These tasks came from a balanced incomplete blocked design, with each impact appearing 5 times across the tasks and co-appearing twice with each other impact.

Fig. 1
figure 1

Example of best−worst scaling choice task

Full size image

Sociodemographic and Clinical Questionnaire

Participants in all phases of the study completed a questionnaire on their sociodemographic and clinical characteristics. This questionnaire included items on age, gender, race and ethnicity, employment status, education level, current living situation, diabetes history, and medication use.

Analyses

Descriptive statistics were reported for the sociodemographic and clinical characteristics of participants in all phases of the study. Qualitative data were analyzed using ATLAS.ti qualitative data analysis software version 8.0 [10]. Data from the interviews were analyzed using a summative content analysis [11]. A coding dictionary was developed based on the concepts that emerged during the interviews. The qualitative data collected during the interviews on symptoms and impacts were analyzed and grouped according to domains of interest.

For the preference survey data, categorical variables were summarized by frequency statistics (N [%]); continuous variables were summarized using means, standard deviations (SD), and maximum ranges. The BWS1 choice data were analyzed with a multinomial logit (MNL) model [12]. The model estimated the burdensomeness of each impact relative to “feeling embarrassed to inject in public,” which was set as the reference impact. The reference impact was selected post-hoc, such that all the estimates were positive and thus indicated an increase in relative burdensomeness. A second MNL was estimated to explore how the RI differs across diabetes type and CGM use. In addition to the 14 impact variables, this model also included all the interaction terms between the impact variables and diabetes type (i.e., T1DM vs. T2DM) and CGM use (i.e., user vs. non-user). Differences in relative attribute importance by duration of diabetes diagnosis and MTI use were also explored.

Estimating the sample size required for testing effects in BWS is challenging, with Flynn et al. [13] concluding there is no general basis for determining sufficient sample size for a BWS study. The final sample for this study is in line with BWS1 studies published in health literature and allows for meaningful comparisons between the target groups [14].

Results

Qualitative Interview Study

Participant Characteristics and Self-Reported Clinical Information

The mean age among participants was 45.6 (SD 13.4) years (Table 2). The majority of the participants were female (63.3%) and were employed full-time (63.3%). The average duration of diabetes among the entire participant study population was 16.9 (SD 11.7) years, among whom patients with T1DM reported having diabetes for an average of 22.2 (SD 13.2) years and patients with T2DM reported having diabetes for an average of 11.5 (SD 6.9) years. Patients with T1DM reported using MTI on average for 14.3 (SD 8.1) years and patients with T2DM reported using MTI for an average of 8.0 (SD 5.0) years)

Table 2 Key sociodemographic and clinical characteristics of patients—qualitative interview study participants
Full size table

Qualitative Interview Results

Impacts Associated with MTI

Participants were asked to describe the impact of their MTI experience on their daily lives. Figure 2 displays the impacts of MTI categorized by domain, with the more commonly reported impacts appearing in a larger font in the word cloud. Participants most commonly reported psychological/mood impacts (n = 18/25, 72.0%). Most frequently, participants shared that it is frightening when their blood sugar is not in range, and they worry about the rises and falls of their blood sugar (n = 9/25, 36.0%).

Fig. 2
figure 2

Impacts of MTI: word cloud. Single asterisk refers to blood glucose (BG) not being in range. Double asterisk (**) refers to BG being in range

Full size image

Participants also described social impacts (n = 17/27, 63.0%), including a sense of embarrassment associated with injecting or changing the pump in public (n = 11/27, 40.7%) and avoiding social functions (n = 2/27, 7.4%). Additional impacts reported were related to work/school (n = 14/26, 53.8%) and sleep (n = 15/29, 51.7%). Some participants commented that their symptoms affect work productivity (n = 5/26, 19.2%), while others shared that they forget to take MTI at work (n = 2/26, 7.7%), or take MTI and forget to eat (n = 2/26, 7.7%). Example quotes are provided in Electronic Supplementary Material (ESM) Table 1. Based on the results of the qualitative research, the reported impacts of MTI were condensed into 15 impacts for the BWS1 preference study.

BWS1 Preference Study Results

Participant Characteristics and Self-Reported Clinical Information

A total of 336 people with diabetes (N = 167 in USA; N = 169 in UK) completed the online survey (Table 3). The mean age among these participants was 56.0 (SD 14.1) years, of whom the majority were male (n = 216, 64.3%) and White (n = 316, 94.0%). Approximately equal proportions of participants were users of CGM (n = 171, 50.9%) and non-CGM users (n = 165, 49.1%).

Table 3 Key sociodemographic and clinical characteristics of patients—best–worst scaling 1 preference study sample
Full size table

People with T1DM (N = 161) self-reported being diagnosed with diabetes an average of 27.3 years ago (SD 16.1 years) and starting insulin an average of 25.2 years ago (SD 16.7 years). People with T2DM (N = 175) self-reported being diagnosed an average of 17.2 years ago (SD 13.0 years) and starting insulin an average of 9.6 years ago (SD 7.7 years). The majority of participants had been on MTI for > 5 years (n = 253, 75.3%), and reported typically taking their MTI > 5 min before a meal (34.2%) or at mealtime (48.8%) (Table 3).

Preference Results

The marginal utility and RI of each impact for the overall sample estimated using the MNL model are shown in Fig. 3. Higher marginal utility values indicate a greater perceived burden by patients; the impacts are presented in order from the most to least burdensome.

Fig. 3
figure 3

Results of the multinomial logit model and relative importance of mealtime insulin impacts: overall sample

Full size image

The three most burdensome impacts included “worrying about experiencing extreme lows in BG” (RI = 14.2%), “worrying when BG levels were not in the target range” (RI = 11.9%), and “worrying about experiencing extreme highs in BG” (RI = 11.6%) (Fig. 3) All three of these impacts were in the diabetes distress domain. The next most burdensome impacts were all part of the diabetes management domain and included “having to calculate the insulin dose for the meal” (RI = 11.0%), “having to check BG regularly around mealtimes” (RI = 8.7%), and “adjusting MTI dose and timing as needed when exercising” (RI = 8.5%). Although the three most burdensome impacts were within the diabetes distress domain, the cumulative RI for the diabetes management domain was slightly higher than that for diabetes distress (RI 47.1% vs. 43.9%).

Diabetes Type

To compare the ranking of impacts for people with T1DM and T2DM, we used an interacted MNL (IMNL) (Fig. 4). We found that there were some differences in the RI each group placed on impacts. People with T1DM placed a higher RI on “worrying about experiencing extreme highs in BG,” “adjusting MTI dose and timing as needed when exercising,” and “my MTI routine interfering with work or school” than did people with T2DM (all p < 0.05). People with T2DM placed a higher RI on “having to check my BG regularly around mealtimes” and “planning when to inject MTI dose when eating out” compared to people with T1DM (all p < 0.05).

Fig. 4
figure 4

Relative importance of impacts by diabetes type. Asterisks indicate a significant difference in the specified impact between people with T1DM and those with T2DM at ***p < 0.001, **p < 0.01, and *p < 0.05. T1DM Type 1 diabetes mellitus, T2DM type 2 diabetes mellitus

Full size image
CGM Use

The relative importance of each of the impacts was similar across CGM users and non-CGM users, with the exception of five impacts. “Worrying about experiencing extreme lows in BG” and “worrying when BG levels are not in my target range” were significantly more important to CGM users, whereas “having to check my BG regularly around mealtimes,” “having to manage when I inject my MTI around meals,” and “worrying about the possibility of missing an insulin dose” were significantly more burdensome to non-CGM users (Fig. 5; all p < 0.05).

Fig. 5
figure 5

Relative importance of mealtime insulin impacts by CGM use. Asterisks indicate a significant difference in the specified impact between users of CGM and non-users (i.e., according to CGM use group) at ***p < 0.001, **p < 0.01, and *p < 0.05. CGM Continuous glucose monitoring

Full size image
Diabetes and CGM Use

The interaction between diabetes type and CGM use was further evaluated by comparing the RI, with the results for each group shown in ESM Fig. 1. The results are consistent with the univariate analyses of these covariates, with diabetes type most strongly associated with the perceived burden of “planning when to inject MTI dose when eating out” and “my MTI routine interfering with work or school.” Similarly, the difference in relative importance of “worrying about experiencing extreme lows in BG” is primarily driven by CGM use, as seen in the univariate results.

Duration of Diabetes Diagnosis and MTI Use

The effects of time since being diagnosed with diabetes and duration of MTI use on perceived importance of impacts were also examined. When examined by duration of diabetes, there were also minor but significant trends (some p < 0.05) for “having to calculate insulin dose for a meal,” “worrying about missing a dose,” and “planning when to inject MTI dose when eating out” (ESM Fig. 2). The relative importance of impacts was similar regardless of duration on MTI except for one impact “planning when to inject MTI when eating out,” which was more burdensome for those on MTI for ≤ 5 years (ESM Fig. 3).

Discussion

This study was designed to identify and characterize the impacts associated with MTI for people living with diabetes in the USA and UK. Many people living with diabetes experience significant unmet needs, including impacts associated with their diabetes and MTI, beyond changes in glycated hemoglobin (HbA1c) [15]. In the qualitative portion of the current study, a large majority of participants reported experiencing psychological and social impacts. These results are supported by prior research that suggests that people with diabetes who spend more time in range and less time in severe hyperglycemia [16, 17] or hypoglycemia are more likely to have better mental and emotional health [18]. Future research might be aimed at exploring differences in emotional well-being by diabetes type and treatment regimen.

The results of the preference survey on impacts of the MTI experience on the participants provide insights into the RI of the unmet needs of people living with diabetes. Specifically, the most burdensome impacts were reported to be related to diabetes distress (e.g., worrying about extreme lows, BG not in target range, and extreme highs) and diabetes management (e.g., having to calculate insulin dose, checking BG, and adjusting MTI dose). These results are consistent with the findings from the qualitative interviews, which found that the majority of participants reported experiencing specific impacts related to managing the MTI process as well as psychological impacts (i.e., anxiety and distress associated with BG levels). These findings are also aligned with those of previous studies in which participants reported that time-in-range was ranked as the highest impact on daily life for both people living with T1DM and those living with T2DM [19]. Participants reported low levels of success for their current treatments at reducing stress or worry related to their glucose numbers [19]. Additionally, studies found that greater diabetes distress is associated with difficulty managing diabetes [20,21,22].

While people with T1DM and T2DM ranked the relative burden of each of the impacts similarly, there were minor differences between the two groups, regardless of CGM use. In this sample of people with diabetes, those with T1DM reported experiencing hypoglycemia more often and being more likely to be diagnosed with anxiety compared to people with T2DM, which may be either a consequence and/or a contributor to people with T1DM having strong preferences. In a prior study evaluating burden and unmet needs related to MTI, people with T1DM reported more moderate to severe burden related to their current MTI experience than people with T2DM [4]. While the current study cannot determine if the absolute burden is higher with T1DM or T2DM, the results do confirm the difference between the two patient groups and how they evaluate the relative burden of the impacts associated with MTI.

CGM and non-CGM users also ranked the relative burden of the impacts similarly, albeit with a few key differences in that non-CGM users ranked two impacts related to diabetes management (i.e., having to check BG around mealtimes, and having to manage when to inject MTI around meals) and one related to diabetes distress (i.e., worrying about the possibility of missing an insulin dose) as more burdensome than CGM users. This finding was expected given that CGM users are able to monitor their glucose values more closely and, therefore, could conceivably be less concerned with planning and calculating MTI doses and with checking their glucose. Consistent with this finding, in a previous study patients using CGM reported fewer emotional impacts due to being able to always check their glucose and more comfort calculating dosing due to being able to see trends through their use of CGM [23]. While CGM has been shown to help with clinical outcomes, including glycemic control, reduced risk of hypoglycemia, and more time spent in glucose range [24], CGM can also be beneficial for diabetes management by providing patients with confidence when calculating dosing and with constant assessments of BG levels [25]. CGM is being perceived by people with diabetes as a tool that could help the diabetes management as people with diabetes refer to being less worried about their insulin dose calculation and constant assessments on glucose values.

Limitations

The results of the current study should be interpreted with consideration for the following limitations. First, all clinical data were self-reported by the participants, with possible recall bias due to the duration of diabetes. Second, as with all preference elicitation studies, the results of the BWS can only be interpreted relative to the set of impacts that were presented to participants. However, as the selected impacts were identified through a dedicated qualitative phase among people with diabetes, and the BWS instrument was further pilot tested to assess whether participants perceived any relevant impacts that were missing, it is likely that the list of impacts was comprehensive and reflected the most relevant burdens that people with diabetes face related to MTI. The BWS1 preference study was conducted during the coronavirus disease 2019 (COVID-19) pandemic, which resulted in stay-at-home orders and fewer social activities. As a result, the RI of the social impacts evaluated in this study may have been lower. Lastly, panel recruitment was used for both studies, which may have implications for the generalizability of the results; for example, the BWS1 sample had more males than females and was mostly White.

Conclusion

People living with diabetes have a range of unmet needs associated with their use of MTI. These primarily include a high level of worry regarding their glucose values and fluctuations and ongoing impacts associated with the management of their condition. These findings may allow physicians to better identify factors that may be affecting the accurate management of the patient’s condition, as well as to allow more specific guidance that increases patient confidence and decreases the burden or distress of some of the aspects related to the use of MTI. In addition, this study identified CGM as a tool for providing information to help alleviate the burdens associated with insulin management for people with diabetes. Endpoints that more accurately capture these aspects of the patient experience (beyond postprandial glucose and time in range) could also be incorporated into future diabetes research that is more directly relevant to people living with diabetes.