Improved Glycemia and Quality of Life Among Loop Users: Analysis of Real-world Data From a Single Center

Background Despite do-it-yourself automated insulin delivery being an unapproved method of insulin delivery, an increasing number of people with type 1 diabetes (T1D) worldwide are choosing to use Loop, a do-it-yourself automated insulin delivery system. Objective In this study, we aimed to assess glycemic outcomes, safety, and the perceived impact on quality of life (QOL) in a local Edmonton cohort of known Loop users. Methods An observational study of adults with T1D who used Loop was performed. An assessment of glycemic and safety outcomes, HbA1c, time in range, hospital admissions, and time below range compared users most recent 6 months of Loop use, with their prior regulatory approved insulin delivery method. QOL outcomes were assessed using Insulin Dosing Systems: Perceptions, Ideas, Reflections, and Expectations, diabetes impact, and device satisfaction measures (with maximum scores of 100, 10, and 10, respectively) and semistructured interviews. Results The 24 adults with T1D who took part in this study 16 (67%) were female, with a median age of 33 (IQR 28-45) years, median duration of diabetes of 22 (IQR 17-32) years, median pre-Loop HbA1c of 7.9% (IQR 7.6%-8.3%), and a median duration of Loop use of 18 (IQR 12-25) months. During Loop use, the participants had median (IQR) values of 7.1% (6.5%-7.5%), 54 mmol (48-58) for HbA1c and 76.5% (64.6%-81.9%) for time in range, which were a significant improvement from prior therapy (P=.001 and P=.005), with a nonsignificant reduction in time below range; 3.0 to 3.9 mmol/L (P=.17) and <3 mmol/L (P=.53). Overall, 2 episodes of diabetic ketoacidosis occurred in a total of 470 months of Loop use, and no severe hypoglycemia occurred. The positive impact of Loop use on QOL was explored in qualitative analysis and additionally demonstrated through a median Insulin Dosing Systems: Perceptions, Ideas, Reflections, and Expectations score of 86 (IQR 79-95), a median diabetes impact score of 2.8 (IQR 2.1-3.9), and a median device satisfaction score of 9 (IQR 8.2-9.4). Conclusions This local cohort of people with T1D demonstrated a beneficial effect of Loop use on both glycemic control and QOL, with no safety concerns being highlighted.


Introduction
Background Do-it-yourself (DIY) automated insulin delivery (AID) systems are user-designed systems that combine 2 regulated devices, an insulin pump that delivers a continuous subcutaneous insulin infusion and a continuous glucose monitor (CGM) that is controlled by an algorithm. Through this predictive algorithm, coded by the user, these systems facilitate an automated adjustment in insulin delivery, tailored to an individual's requirements [1]. People with type 1 diabetes (T1D) are increasingly using these systems worldwide because the rapidly evolving software with extensive opportunities for customization helps individuals to achieve personalized glucose targets and reduce the burden of diabetes management [2].
DIY AID systems can be subclassified into system types (including AndroidAPS, FreeAPSX, and Loop) depending on the technology and the algorithms on which they run. These systems have not gained regulatory approval; users are effectively hacking licensed technology to run these algorithms and modulate their insulin delivery [3]. Recently, the first randomized controlled trial to highlight both the safety and the efficacy of AndroidAPS has been completed [4]. There are also multiple published studies, single-arm cohort studies, user self-reported pre-post data, and case series, reporting beneficial outcomes in glycemic control, quality of life (QOL), and reassuring safety data with DIY AID use. The studies have reported on individual system types or combinations of these [5][6][7][8][9][10][11][12].
Studies on DIY AID system use consistently report excellent glycemic outcomes, with very high time in range (TIR) and low time below range (TBR). These values far exceed those suggested as clinically recommended targets, achieved by only a minority of people with T1D [13]. Individuals choosing to use DIY AID are a select sample of people with T1D who are highly motivated to engage in self-care. Users are actively involved in optimizing glycemia with the aims of preventing diabetes-related morbidity, increasing life expectancy, and improving sleep quality [12].
Internet resources and social media platforms are currently the mainstay of guidance for DIY AID users [2]. These platforms have been used by enthusiasts in the field to collect outcome data [14]. The average ages of users receiving insulin delivery via a DIY system are reported to be 35.8 years (AndroidAPS), 33 years (OpenAPS), and 28.5 years (Loop), but the extensive benefits of these systems have been reported in studies of both adults and children, with 26% of users in a cross-sectional survey being aged <16 years [11]. Similar benefits have been observed across the 3 DIY AID system types, with the type of system studied usually being dependent on the geographical distribution of system users. Loop is the most commonly used DIY AID system in North America and AndroidAPS is the most commonly used DIY AID system in Europe [12]. To date, there have been no cohort studies performed in Canada to assess user outcomes for DIY AID users.

Objectives
We sought to explore the experiences of adults using Loop at a single center in Canada. We aimed to assess quantitative outcomes in the form of glycemic, QOL, and safety data and also used a qualitative approach to gain a greater understanding of the lived experiences of Loop users.

Methods
A cross-sectional study of current glycemia, experiences of Loop use, and QOL was performed in adults with T1D who were attending the Kaye Edmonton Clinic, which is part of the University of Alberta Hospital in Edmonton, Alberta, and were known to be currently using any form of DIY AID system.

Ethics Approval
This study was approved by the University of Alberta Research Ethics Board (Study ID pro00111577).

Participants
Prospective participants were identified and contacted by a member of their clinical team at the Kaye Edmonton Clinic. All participants were adults (aged ≥18 years) with T1D who were using a DIY AID system at the time of data collection. We arranged a semistructured interview with a member of the study team for participants after obtaining informed consent from them to take part in the study.

Outcome Measures
Up to 6 months of most recent glucose data, while using Loop, were collected from the participants' CGM download data, to record mean TIR 3.9 to 10.0 mmol/L (70-180 mg/dL), TBR 3.0 to 3.9 mmol/L (54-70 mg/dL), TBR; <3.0 mmol/L (<54 mg/dL), and time above range: >10.0 mmol/L (180 mg/dL). Where available, the same data were collected retrospectively from the participants' glucose sensor data for the 6-month period before commencing Loop, while they were using their previous mode of insulin delivery. The participants' laboratory HbA 1c readings (%) were collected from hospital records, including the most recent value with Loop use, in addition to the participant's last reading before commencing Loop. The hospital records of all participants were reviewed for hospital admissions, specifically assessing the occurrence of severe hypoglycemia (SH) and diabetic ketoacidosis (DKA) throughout the duration of the participants' Loop use. The University of Alberta Hospital uses an integrated medical record system, enabling data capture of admissions to any facility in the province.
Semistructured interviews were arranged via telephone or through the use of the Zoom videoconferencing service (Zoom Video Communications, Inc) [15] between July and September 2021. A full interview transcript guide is available in Figure 1. Each interview was conducted by researchers AM and KC, with one asking questions while the other transcribed responses. During the interview process, demographic data were collected, including age, type of DIY AID system used, duration of DIY AID use, duration of diabetes, sex, ethnicity, occupation, and highest level of educational attainment. Participants were asked to report any episodes of SH that required the assistance of another person to treat and any occurrence of DKA during Loop use. Qualitative questions were related to participants' reasons for commencing, challenges in its use, and support mechanisms with regard to using a DIY AID system as well as the benefits and barriers that they experienced with DIY AID use.
After the interviews, the participants electronically completed 2 validated questionnaires, Diabetes Impact and Device Satisfaction (DIDS) [16,17] and Insulin Dosing Systems: Perceptions Ideas Reflections and Expectations (INSPIRE) [18], evaluating their perceived impact of using DIY AID on their QOL. Full copies of these questionnaires are available in the appendices.

Analysis
Descriptive statistical analysis and normality testing via the Shapiro-Wilk test were performed using GraphPad Prism (version 9.2.0 for macOS; GraphPad Software). A normal distribution was seen in both TIR and HbA 1c data before but not after Loop use, with additional skewed distributions being seen in age and QOL outcome measures. Therefore, data are reported as median (IQR) in the analysis of this cohort with nonparametric tests being used and statistical significance being defined as P<.05. Paired groups were compared using the Wilcoxon signed-rank test and unpaired data were compared using the Mann-Whitney U test, in addition to the correlation of variables using the Spearman correlation coefficient.
Qualitative interview data were coded deductively by the research team using NVivo 12 (QSR International) [19], after the data-driven inductive generation of the code structure (Multimedia Appendix 1). This deductive code structure was developed inductively from our data in addition to the consideration and inclusion of common themes identified in previous DIY AID user interview studies [20][21][22]. Overarching themes were constructed from the participants' viewpoints and reflexive thematic analysis was performed by AM [23].

Overview
A total of 24 adults with T1D participated in this cross-sectional study, with a median age of 33 (IQR 27.5-44.8) years and median duration of diabetes of 21.5 (IQR 17.3-32.0) years. All 24 participants were using the Loop subtype of DIY AID as their method of insulin delivery for a median duration of 18 (IQR 12-25) months, with a total of 470 months or 39.2 years of Loop use in the cohort. The demographic characteristics of this cohort of Loop users are described in Table 1. Of the 24 participants, the majority (n=16, 67%) were female and (n=22, 92%) White and over one-third (n=9, 38%) were employed in health care professions.

Glycemic Outcomes
HbA 1c values were available both before and after commencing Loop use for all participants. CGM data were available for 6 months before commencing Loop use for 71% (17/24) of the study participants, with a mean of 5.8 (SD 0.66) months of CGM data with Loop use being reviewed per participant. No significant differences in age, duration of diabetes, duration of Loop use, baseline HbA 1c , or QOL outcome measure scores were seen between those participants with and without pre-Loop CGM data. Before Loop, median HbA 1c was 7.9% (IQR 7.6%-8.3%) or 63 (IQR 60-67) mmol/mol, and median TIR was 58% (IQR 52.3%-64.0%). A statistically significant improvement in these parameters was seen with Loop (P=.001 and P=.005). A median increase of 15% (IQR 6.3%-23.8%) in TIR was seen in 82% (20/24) of Loop users. Before Loop, 17% (4/24) of users achieved the clinical target of 70% TIR, in comparison with 67% (16/24) of users who achieved it with Loop use. HbA 1c reduction was seen in 79% (19/24) of users with Loop; the median rate of improvement was 0.8% (IQR 0.28%-1.18%). In addition, a significant reduction in time above range was demonstrated with the introduction of Loop (P=.008). Glycemic data are shown in Table 2

Safety
Of the 24 participants, 2 (8%) experienced an episode of DKA, and no episodes of SH occurred in the cohort with Loop use. One episode of DKA was euglycemic and was associated with gastrointestinal infection and sodium-glucose cotransporter-2 inhibitor use; it required hospital admission, including intensive care unit stay for 4 days and was completely resolved. The other one was documented to be associated with a urinary tract infection; intensive care unit stay was not required and no insulin pump or Loop system failure was identified. These episodes of DKA occurred 15 and 11 months following starting Loop, respectively.

QOL Measures
The QOL measures collected following participant interviews by using the DIDS and INSPIRE questionnaires are shown in Figure 3 and Table 3. The median diabetes impact score was 2.8 (IQR 2.1-4.8) out of a maximum of 10, with lower scores indicating better outcomes. The median device satisfaction score was 9.0 (IQR 8.2-9.4) out of 10, with higher scores indicating better outcomes. The median INSPIRE score was 86.0 (79.5-94.6), with 100 being the maximum and optimal score. An examination of these QOL scores and of glycemic variables showed no significant positive correlations with TIR (r=0.024, r=0.007, and r=0.207; P=.41 or with HbA 1c (r=−0.163, r=−0.287, and r=−0.254; P=.38). A moderate correlation was seen between increased duration of Loop use and lower diabetes impact scores (r=−0.420, P=.04).

Overview
The analysis of semistructured interview data highlighted frequent topics that participants had expressed as important in their lived experiences of Loop use. Overarching themes were constructed from these viewpoints, comprising empowerment and control, the daily impact of living with diabetes with Loop use, quantification of risk, and society's understanding and awareness of Loop (Textbox 1).

Empowerment and Control
The principle of autonomy, with individual choice in selecting an optimal management regimen for their condition that was best suited to and most beneficial for them, was a prominently featured theme in why participants had chosen Loop. The feeling of dissatisfaction with a prior treatment option was described, with the need to make an individual choice to optimize their lifestyle:

The Daily Impact of Living With Diabetes With Loop Use
Participants discussed the psychological impact of living with diabetes both before and since using Loop, with notable improvements expressed in the time spent thinking about diabetes and diabetes-related distress: Participants reported burnout as a result of day-to-day demands, and despite the beneficial impact of Loop on psychological well-being expressed by them, they noted that starting to use the system and the initial setup required a significant investment of time and energy:

I felt burnout in managing my diabetes, spending all my time managing diabetes or filling out insurance forms for my diabetes, it was a real mental challenge to think about and set up a new system, a lot of mental energy. [29 years, female; 27 years DM; 7 months of Loop use]
Significant financial investments, both initial and ongoing, were reported by Loop users. They required both the component technology (an insulin pump and a CGM) and appropriate devices on which to set up and use the app-an iPhone with iOS 12.4 or newer operating system and a Mac computer-as well as a communicating device (RileyLink, OrangeLink, or EmaLink) and an Apple developers' license [25]. Access to and cost of this hardware were the most commonly perceived barriers to Loop use in this cohort:

I would recommend everyone to try it. It is quite a bit of work getting it setup and getting it ready but is pretty minimal effort for upkeep. The access to the devices is the one thing that makes it difficult (especially coverage for it). The peace of mind makes it worth it because it makes so much of a difference.
[24 years, female; 14.5 years DM; 5 months of Loop use] Consequently, the use of a system such as Loop comprising multiple devices requires users to ensure that all necessary components are carried around with them and have sufficient battery charge. The devices must be in constant communication with each other to effectively use the app. Some participants reported these day-to-day aspects of Loop use to be challenging at times:

Quantification of Risk
Because Loop is unregulated and therefore unsupported, there may be perceptions of risk. When asked about this, none of the participants considered that using Loop was any more of a risk than an alternate option in diabetes management. Indeed, most participants deemed it to be of much lower risk: Many participants were using older and out-of-warranty pumps (because many newer in-warranty pumps were incompatible with the Loop app), which they identified as a potential risk in itself:

I was worried about using the older pump but I have recently acquired both a backup pump and RileyLink so have more confidence in this. My pump looks really rough and I do worry occasionally about button errors especially in the heat. [29 years, female; 27 years DM; 7 months of Loop use]
Participants reported dissatisfaction with the alternate diabetes management options that are currently available, including Commercial AID systems. A participant had used the Medtronic MiniMed 670G system but had struggled with the Enlite sensor, especially with its alarms. Another user was dissatisfied with Tandem Control-IQ as a result of the lack of customizable glucose targets, with the system providing fixed thresholds that some people felt were too high. Many expressed that they did not wish to consider any other options now that they had experienced Loop:

Society's Understanding and Awareness of Loop
Owing to the unregulated nature of Loop, some participants expressed concern in discussing the use of Loop with others, including with health care providers. All participants in this study were seen in the same diabetes clinic, although with multiple different care providers practicing within the clinic. Most participants expressed positive interactions in the health care setting, frequently describing "passive encouragement" to consider and use Loop. A participant explained that because of the lack of support, with the discouragement of Loop by her previous health care team, she had moved to a new provider as she wished to continue using Loop. Another described being discouraged from continuing to use Loop while seeing a different endocrinologist during pregnancy, despite finding it very beneficial. All other participants felt they could discuss Loop with their clinical team without concern and that health care providers were largely keen to learn more about Loop: Most participants felt that their family and friends were supportive of Loop, although several noted that they had reservations at first, before seeing the benefits of the system for themselves: Users expressed frustration at the lack of industry support for Loop and the fact that it had required people with diabetes and their families to build this system. However, they also expressed concerns relating to future industry involvement with Loop and the potential changes in the system that this may involve:

Principal Findings
In this cohort of adults with T1D at a single center, we have highlighted improved glycemic outcomes with Loop use. With this glucose management system, 67% (16/24) of users achieved the clinically recognized TIR target of 70% [27]. In this first described Canadian cohort of Loop users, we have identified high QOL scores with Loop. The Loop users demonstrated superior glycemic outcomes relative to the general population of people with T1D, with 42% (10/24) of them achieving an HbA 1c of <7%, in comparison with the reported average of 21% [13]. The users noted that the removal of an emotive decision-making component in diabetes management was an overwhelmingly favorable aspect of Loop and felt that it aided in the achievement of individualized glucose targets. The safety features of Loop were particularly felt to be important by our participants overnight, with associated improved sleep. Reduction in hypoglycemia (frequency and severity), improved overnight glycemic control, and improved sleep have been widely reported for all DIY AID system types [2,8,28].
We have demonstrated a strikingly similar TIR reported to that in a large prospective observational study of 558 residents of the United States, with a mean age of 23 (SD 13) years who had been new Loop users for 6 months [10]. In this large cohort, with a maximum of 7 days Loop experience at baseline, mean TIR at 6 months was 73% (SD 13%), compared with 71% (SD 16%) in the most recent 6 months of Loop use in our local cohort of 24 users. In comparison to the participants in this prospective study, our study participants were relatively experienced Loop users, with a median of 18 (IQR 12-25) months of Loop use. These results suggest that the benefits of Loop can occur early in its implementation and are somewhat durable, a desirable characteristic for a therapeutic intervention in a chronic condition such as T1D.
No adverse safety outcomes related to hypoglycemia because of Loop use were reported in our data; there was an improvement in TBR, with time <3.0 mmol/L and no admissions related to SH. However, 2 episodes of DKA occurred, both of which were associated with underlying infections. In people with T1D, the estimated incidence of DKA is reported to be 4.6 to 8.0 events per 1000 patient years [29]. Lum et al [10] reported no episodes of DKA with 6 months of Loop in 558 individuals (279 years); they reported 51 episodes of SH, with only one of these episodes being attributed to Loop use [10]. This larger prospective study used weekly electronic messages (with an 89% response rate) for data collection to maximize user recall but was dependent on self-reporting for these likely memorable and significant events for a person with diabetes [30]. Our data were reported based on retrospective recall from participants at the time of interview but were verified by reviews of their medical records. Only 1 of the 2 participants in our cohort self-reported the occurrence of an episode of DKA. All participants in our study reported that they perceived Loop to be safe when the correct settings were in place. Interview responses relating to risk in this cohort were similar to those described by Schipp [20], highlighting a conscious weighing of risks against benefits for DIY AID users. With a detailed understanding of risk, including the use of unregulated and potentially out-of-warranty devices, using a DIY system was felt to be the best glucose management option available to them at this moment in time [19]. DIY AID systems were primarily designed for safety, initially targeting the avoidance of hypoglycemia. This concept of risk reduction through AID system use has been discussed by Lewis [24], highlighting the importance of taking the level of risk in AID use into context, with the risk faced by a person with diabetes who is manually dosing insulin representing the most appropriate comparator and not the risk faced by a person without diabetes. The use of AID systems removes a proportion of this total risk and provides an overall net risk reduction for people with T1D [24].
In terms of quantitative QOL outcomes, we found low diabetes impact and high device satisfaction and INSPIRE scores with Loop use for a median of 18 months in our cohort. The scores were very similar to DIDS outcomes of 2 months of Tandem Control-IQ use (Commercial AID) in 1435 people with T1D aged ≥14 years [15], with a median diabetes impact score of 2.7 (2.8 in this cohort) and median device satisfaction score of 9.1 (9.0 in this cohort). The INSPIRE outcomes of this study were also comparable with those reported with 6 months of Tandem Control-IQ use in another cohort of 112 users with a mean of 87 (IQR 77.6-96.5), in comparison with 86 in this cohort [26]. The median TIR achieved with Tandem Control-IQ was similar to that in our cohort, 79.2% (IQR 70.3%-86.2%) with a shorter duration of AID use, but closer to target glycemia at baseline; with a mean HbA 1c of 6.9% (SD 0.9%) [15]. These studies of Commercial AID [17,31] were conducted with substantially greater supervision and support, as would be expected in a randomized controlled trial, in comparison with the real-world experiences collected from our Loop users.
We did not see a strong correlation between device perception and satisfaction outcome measures (DIDS and INSPIRE) or glycemic outcomes in this cohort. This may be a result of the small sample size with a narrow spectrum in these outcomes, but our qualitative data highlight a strong benefit of Loop use on QOL. After improved glycemic outcomes, enhanced QOL was the most frequently reported benefit of Loop use in our cohort. This concept comprises a reduction in the psychological impact of living with diabetes including time spent thinking about diabetes, diabetes-related distress, and burnout, in addition to greater flexibility in day-to-day life, notably related to diet and activity. The reduced mental burden of diabetes and less reliance on the accuracy of carbohydrate counting are consistently reported positive outcomes with DIY AID system use [2].
Another common theme identified was the financial resources required for Loop use, which restricted the availability of this beneficial system. We did not collect data relating to income or index of deprivation, but our participants' educational attainment and occupations indicated higher-than-average socioeconomic status [32]. Access and coverage of insulin pumps across Canada remains unequal, with varying provincial health care funding models in place; insulin pump therapy is more commonly used in areas with reimbursement programs in place [33].
All except 1 Loop user in this cohort used both an insulin pump and a CGM device at the time of deciding to commence Loop. Having access to, but frequently experiencing dissatisfaction with these devices was a contributing factor to the process of behavior change in these users. For effective behavior change to occur, such as the initiation and continuation of Loop, there are key components for the user and their environment according to the capability, opportunity, motivation, behavior model of behavior change. These include capability (both physical and psychological), physical (including financial and material) and social opportunity (considering social and cultural norms) as well as motivation for change [34]. The components of this model are apparent in the lived experiences that we have described. Loop users highlighted the importance of this physical opportunity, with availability and access to technological devices being a potential limiting factor in the initiation of Loop. Most participants found their health care providers to be relatively supportive toward commencing Loop, despite the system being unregulated. This "social opportunity" enabled reassurance for users, this being an acceptable behavior change. This positive interaction is by no means guaranteed, with varied experiences reported with DIY AID use in health care settings [35].
This study had some strengths and weaknesses. The cohort were recruited from a single center with an integrated medical record that would capture admissions to any facility in the province. Objective collection of these data was performed by the health care team, rather than through self-reporting by users themselves, which has been a weakness in most previous reports of DIY AID systems that describe glycemic outcomes [2,8,11,12,28,[36][37][38]. This study did not include a comparator control group. We collected qualitative data in addition to quantitative data, with Loop users being able to compare their own lived experiences, both with and without Loop use. The sample size for the collection of quantitative outcome data was small, limited by the number of Loop users locally. Selection bias, as a result of the inclusion of individuals who chose to use Loop, must be considered in the generalizability of our findings to the wider population of people with T1D. We have only included current Loop users and therefore, have not been able to explore the reasons behind why users may decide to stop using this form of glucose management system. The fear of disapproval of Loop use from a diabetes care provider as well as barriers to acquiring the component devices have been reported as reasons for Loop discontinuation [22], although we cannot estimate whether these are significant factors in our cohort of individuals who had shared their DIY AID use with their health care providers.

Conclusions
DIY AID use in this local cohort of individuals who have chosen to start and continue to use Loop has been associated with notable improvements in glycemic outcomes and excellent QOL. Through a combination of quantitative data collection and qualitative interview analysis, we have gained a greater understanding of the lived experiences of the Loop users in this cohort, including the common challenges and extensive benefits. What is most striking is the ability for motivated individuals to further increase their success in achieving glycemic targets while simultaneously experiencing reduced burden and distress from diabetes. Although most DIY users who have been studied to date have been those who were already successful in achieving glycemic targets, future studies should focus on the potential benefits of DIY AID for people who have found it difficult to achieve glycemic targets because of this goal being excessively burdensome or beyond their capacity, as a result of limited financial, social, or educational resources. It is hoped that the experience of Loop users described in this cohort, in combination with further broader user experience, may aid many other future users to access and experience the benefits of Loop use.