Journal Logo

Original Clinical Science—General

Preferences for Risks and Benefits of Islet Cell Transplantation for Persons With Type 1 Diabetes With History of Episodes of Severe Hypoglycemia: A Discrete-Choice Experiment to Inform Regulatory Decisions

Wilson, Leslie PhD1; Kwok, Tiffany PharmD1; Ma, Yanlei PhD2; Wong, Jenise MD1; Ho, Martin MS3; Ionova, Yelena PharmD1; McGrath, Maureen RN, MS, PNP1; Mueller, Monica M. BSN, RN, MPT, PT, PCS1; Gitelman, Stephen E. MD1; Irony, Telba PhD3

Author Information
doi: 10.1097/TP.0000000000004189



Of the 1.25 million Americans living with type 1 diabetes (T1D), approximately 40 000 have “brittle” or suboptimally controlled T1D.1-4 Despite being on optimized insulin regimens, these persons experience unpredictable variability in blood glucose levels, ketoacidosis, and severe hypoglycemia with unawareness, preventing a timely life-saving response.3,4 These episodes disrupt the quality of life (QOL), require frequent hospitalizations, and cost $1.8 to $5.9 billion annually, with 20% to 50% mortality in 27- to 45-y-olds.4-8 Severe hypoglycemia has multiple definitions, but by consensus, the definition used to recruit participants for this study is “episodes requiring assistance from another person to treat.”9-19

In addition to insulin, islet cell transplantation (ICT) is a potential treatment option for hard-to-control diabetes when best management fails.20 The goals of transplantation are to restore glucose-regulated endogenous insulin secretion, reduce long-term diabetes complications, and improve the QOL; however, it also carries significant risks.21 These range from risks of the procedure itself (portal vein thrombosis and hemorrhage) to risks from the need for lifelong immunosuppressive therapy.22

Although ICT has been approved for the treatment of T1D in other countries, it does not have final approval from the US Food and Drug Administration (FDA).23 An authorization requires approval as a biologic from the FDA’s Center for Biologics Evaluation and Research (CBER) rather than approval as a transplant.24,25 In April 2021, an FDA advisory panel endorsed pancreatic ICT therapy for T1D not managed with current therapies, and the FDA is currently considering its approval.26,27

The patient perspective is an important input for CBER across many disease areas, as it weighs if the potential clinical benefits of a new treatment outweigh risks to inform regulatory decisions.28 Research to assess the patient perspective on ICT is limited.29,30 Our objective was to develop a patient preference instrument and use it to determine how people with T1D with severe hypoglycemia weigh the risks and benefits of ICT to inform CBER decision making. We used a stated preference approach based on behavioral economics that models choices a consumer might face when deciding between alternative products and services using discrete-choice experiments (DCEs).31-33 In this study of patient preferences toward ICT, insights are intended to help guide regulatory decisions as well as adoption after approval.


Choice-based conjoint (CBC) experiments elicit stated preferences between 2 or more hypothetical alternatives.31,33 Individuals are presented with a series of choice tasks and asked to choose between alternatives with a common set of attributes representing risks and benefits important to decision makers. Attributes are defined by the levels or values an attribute can take. The theory is that individuals choose the alternative that maximizes their overall utility. Part-worth utilities for each level of each attribute are measured by regression beta-coefficients of patients’ choices. Although there are other methods to help an individual make a choice, such as Kepner-Tregoe, these focus more on problem solving for an individual, whereas CBC is a widely used population-based method, is easy for individuals to understand‚ and better mimics how decisions are actually made.

Attribute Selection

A modified meta-ethnography approach using literature reviews and key-informant interviews identified important concepts of risks and benefits of ICT. Open-ended interviews were conducted with clinicians, adults with T1D, FDA/CBER advisory group experts, and patient advocacy organizations to refine concepts, wording, and definitions based on expertise and daily experiences living with T1D. An initial CBC experiment was then piloted in 12 people with T1D. The final CBC tool approved by the advisory group had 8 attributes of ICT risk and benefits. Five attributes were informed by ICT clinical trial outcomes: (1) chance of achieving clinical treatment success defined by HbA1c,30,34 (2) duration of success,30,35 (3) risk of treatable procedure-related adverse events,36 (4) risk of serious complications requiring hospital treatment and rare death,37 and (5) reduction in the risk of long-term complications.38 Three attributes were informed primarily by patient concerns: (1) extent of insulin independence,30 (2) restrictions and risks of antirejection drugs, and (3) time and support needed if multiple infusions are necessary. Each attribute was defined by 3 to 4 levels39-47 (Table S1, SDC,

Survey Design

We used Sawtooth software (version 9.5.3) to create a random, full-profile, and balanced-overlap experimental design. This software specializes in providing both a design algorithm of choice and a hosting platform for the survey. Different combinations of choice tasks and sample sizes were pretested to generate the most statistically efficient design with a simulated logit SE below 0.05 and with the highest D-efficiency score.48,49

The final CBC design had 18 random and 2 fixed task pairs‚ each using different levels of the 8 attributes.50 The first fixed task presented a choice between the most desirable and least desirable attribute levels to test choice validity. The second fixed task included an option with attribute levels most aligned with ICT outcomes versus an insulin-only alternative. After all tasks were completed, participants self-evaluated their confidence in their responses (Figure 1).

Example of the patient preference discrete choice measure for islet cell transplant. HbA1c, hemoglobin A1c.

The survey also included questions about sociodemographics, medical history, diabetes management, and experience with severe hypoglycemia. Additionally, 3 validated QOL instruments were included: (1) the Diabetes QOL questionnaire, (2) the Hypoglycemia Fear Survey II, and (3) Problem Areas in Diabetes (PAID) scale, which indicates severe diabetes distress for scores >40.51-53

Patient Recruitment

English-speaking adults with clinician-confirmed T1D (defined by the American Diabetes Association criteria) with a history of severe hypoglycemia were eligible and recruited from the University of California San Francisco diabetes clinic in person from January to April 2020.54 When clinic visits transitioned to telemedicine during the COVID-19 pandemic, participants were recruited by email and from national diabetes research centers and the study website. Participants who agreed to participate were electronically sent a consent form, a detailed description of the survey attributes, a unique identifier, and a link to the online survey. This study was approved by the University of California San Francisco Institutional Review Board.

High-ris k Diabetes Groups: The Hypoglycemic Scoring System Definition

We defined a scoring method to categorize the potential for higher-risk diabetes for each participant to allow the assessment of preferences in different risk populations. Our full sample was already comprised of persons with high hypoglycemic risk given our recruitment methods intended for high-risk diabetes patients, insulin pump clinics, and email lists of only those with severe hypoglycemic episodes. The scoring method considered other factors of concern for high-risk diabetes: (1) HbA1c >8.0, (2) a severe hypoglycemic episode in 12 mo, (3) diabetes-related hospitalization in 12 mo, and (4) using insulin pumps as a marker in adults to a history of frequent hypoglycemic episodes. Diabetes-related hospitalizations and the use of insulin pumps are shown to be associated with severe hypoglycemic episodes in some populations, although there are also concerns about equity of access to insulin pumps.19 Based on this scoring system, we analyzed 2 risk groups: (1) full sample with at least 1 severe hypoglycemic event at recruitment and (2) patients from the full sample with at least 1 additional diabetes risk element from our scoring systems and factors of concern for higher-risk diabetes (Tables S1 and S2, SDC,

Statistical Analysis

Descriptive statistics were used to summarize demographics, medical history, and QOL data. Analysis of the CBC data used the random parameters multinomial logit model (RPL, also known as the mixed multinomial logit model).55 With the RPL model, the preference parameters are assumed to vary from 1 individual to another and therefore account for the unobserved preference heterogeneity of the population. The RPL model was estimated using the “gmnl” package in R with 10 000 random Halton draws. All attributes were included as dummy coded categorical variables to determine whether each attribute level was statistically significant relative to the reference level. Attribute levels with P < 0.05 were considered statistically significant. To assess differences in preferences between the full sample and in a higher diabetes risk sample, a likelihood ratio goodness-of-fit test compared an unrestricted model‚ where patients with different hypoglycemic risks were allowed to have a distinct preference‚ against a restricted model‚ where the preferences for all patients in the full sample were assumed to be identical. In addition, subgroup analyses were performed using age, gender, HbA1c, and QOL to determine their impact on patient preference.


Study Participants

Of 184 participants consenting, 93 completed the survey. Three were excluded because they responded to the fixed validity question incorrectly, leaving 90 participants included in the full sample. Our analyses included both the 90-participant full sample and a subsample of 77 participants with elements of concern for higher diabetes risk selected from our diabetes risk scoring system. We wanted to test multiple risk groups to determine if preferences differed by diabetes severity (Figure S1, Table S2, SDC,

Tables 1 and 2 summarize the baseline characteristics of the full sample (n = 90) and the higher diabetes risk group (n = 77). In both groups, the participants were balanced in gender with a mean age of 41 to 42 y. A majority of the participants in the full sample and higher diabetes risk group were white (79%, 79%), had at least a bachelor’s degree (76%, 78%), and were employed full time (58%, 62%), respectively. The mean duration of diabetes was 22 and 23 y in the full and high-risk samples, respectively. Both groups scored similarly on the PAID diabetes distress scale (30),56 the Diabetes QOL scale (54 vs 55), and the Hypoglycemia Fear Survey (36).57

TABLE 1. - Patient characteristics: demographics (n = 90 for full sample vs n = 77 for higher-risk sample)a
Full sample, mean (range)/N (%) High hypoglycemic risk sample, mean (range)/N (%) t testP value
N 90 77
Age 42 (20–89) 41 (20–89) 0.76
Female 44 (49) 39 (51) 0.82
White 71 (79) 61 (79) 0.96
Black 6 (7) 5 (6) 0.96
Asian or Pacific Islander 9 (10) 9 (12) 0.92
Native American 2 (2) 2 (3) 0.73
Other 3 (3) 2 (3) 0.88
Latino (any race) 5 (6) 4 (5) 0.78
Education level
High school diploma or General Education Diploma 5 (6) 4 (5) 0.92
Some college 16 (18) 13 (17) 0.88
Bachelor’s degree 38 (42) 34 (44) 0.80
Graduate degree 31 (34) 26 (34) 0.93
Employment status
Employed full-time 52 (58) 48 (62) 0.55
Employed part-time 4 (4) 2 (3) 0.52
Retired 10 (11) 7 (9) 0.67
Homemaker or student 8 (9) 8 (10) 0.75
Disabled 11 (12) 9 (12) 0.92
Unemployed 5 (6) 3 (4) 0.61
Income level
<$50 000 19 (21) 14 (18) 0.64
$50 000–$74 999 11 (12) 9 (12) 0.92
$75 000–$99 999 8 (9) 6 (8) 0.80
$100 000–$199 999 18 (20) 18 (23) 0.87
$200 000 or more 22 (24) 18 (23) 0.68
Prefer not to answer 12 (13) 12 (16)
Health insurance type
Medicare 18 (20) 12 (16) 0.46
Medicaid 10 (11) 7 (9) 0.67
Private plan through work 63 (70) 58 (75) 0.44
Veterans Administration or other military 2 (2) 2 (3) 0.88
Disability insurance 0 (0) 0 (0) 0
Not insured 2 (2) 2 (3) 0.88
aNone were significantly different at the P = 0.05 level or lower

TABLE 2. - Patient clinical characteristics (n= 90 for full sample vs n = 77 for high-risk sample)
Full sample, mean (range)/N (%) High hypoglycemic risk sample, mean (range)/N (%)
N 90 77
Diabetes status
 Years with diabetes 22 (1–61) 23 (1–61)
 Latest HbA1c
 <6.5% 18 (20) 15 (19)
 6.5%–8% 57 (63) 47 (61)
 >8% 14 (16) 14 (18)
 Do not know 1 (1) 1 (1)
Severe hypoglycemia episodes
 In the past 6 mo 0.8 (0–20) 0.9 (0–20)
 In the past 12 mo 1.6 (0–40) 1.9 (0–40)
Diabetes management
 Insulin 90 (100) 77 (100)
 Oral medications 3 (3) 3 (4)
 Diet 36 (40) 28 (36)
 Physician activity 30 (33) 25 (32)
 Pancreas transplant 1 (1) 0 (0)
 Islet cell transplant 0 (0) 0 (0)
 Other 3 (3) 2 (3)
Ever considered islet cell transplant 25 (28) 22 (29)
Device used for insulin injection
 Vial and syringe 9 (10) 8 (10)
 Pen (self-injecting) 37 (41) 25 (32)
 Insulin pump single loop 22 (24) 22 (29)
 Insulin pump closed loop 21 (23) 21 (27)
 Self-programmed insulin pump 8 (9) 8 (10)
 Other 3 (3) 3 (4)
Frequency of blood glucose check
 Rarely or never 7 (8) 6 (8)
 Several times a wk 18 (20) 14 (18)
 1–3 times/d 19 (21) 15 (19)
 4–6 times/d 18 (20) 14 (18)
 >6 times/d 28 (31) 28 (36)
Diabetes-specific emotional distress
 Problem Areas in Diabetes Scale 29.8 (0–87.5) 30.0 (0–87.5)
 Diabetes Quality of Life 54.3 (36.9–81.5) 55.0 (36.9–81.5)
 Hypoglycemia Fear Survey-II 35.7 (11–70) 35.5 (12–70)
HbA1c, hemoglobin A1c.

Preference Weights

The RPL estimation of preference weights showed that persons in both samples had similar preference priorities, although the higher hypoglycemic risk group had stronger preferences toward the highest levels relative to the reference levels for certain attributes.

In the full sample (n = 90), 5 attributes, "chance of clinical treatment success," "duration of treatment success," "duration of insulin independence," "avoiding treatable adverse events," and "avoiding serious complications," were statistically significant across all levels (Figure 2; Table S3a, SDC, RPL beta-coefficients (β) reflected preference scores aligned with the natural ordering of expected risks and benefits: individuals preferring to avoid 15% over 1% risk of serious complications and preferring 90% over 40% chance of treatment success. The strongest preference overall, a negative preference, was for avoiding the highest chance (15%) of a serious procedure-related complication (β = −2.03, P < 0.001). The strongest positive preference was for gaining 5-y insulin independence (β = 1.75, P < 0.001). There was also strong preference for 5-y clinical treatment success (β = 1.39, P < 0.001), 2-y insulin independence (β = 1.10, P < 0.001)‚ and a 90% chance of achieving HbA1c ≤7.0%, which eliminates severe hypoglycemia (β = 1.09, P < 0.001). The next strongest negative preferences were to avoid a 5% chance of serious complications (β = −1.02, P < 0.001) and to avoid 40% risk of treatable adverse events (β = −1.01, P < 0.001) (Figure 2; Table S3a, SDC,

Preference utility scores and relative importance of treatment attributes for full and higher-risk samples of persons with diabetes and risk of severe hypoglycemia. AE, adverse event; ICT, islet cell transplantation; LT, liver transplantation; TX, transplantation.

Attribute levels with the same or similar beta-coefficients indicate a preference for acceptable direct trade-offs of risks and benefits. The full sample preference scores showed a willingness to trade the highest levels of risks (15% and 5% chance of serious complications) for the highest levels of benefit (5-y insulin independence and 90% chance of clinical treatment success). Still, the preference utility for avoiding 15% serious complications was 16% higher than the highest benefit for 5-y insulin independence. Individuals were almost equally willing to trade a 5% risk of serious complications for a 90% chance of clinical treatment success. Overall, individuals were only willing to accept high risks for high, not moderate, benefits.

Not significant in our analysis was a preference for avoiding the time and support needed if multiple ICT procedures are required, each taking 3 mo of concentrated recovery. Also, of little concern was for a decrease in risk of long-term complications and for avoiding restrictions related to immunosuppression required with ICT. The only significant restriction participants significantly preferred to avoid was renal monitoring to avoid renal failure risks (β = −0.61, P < 0.001). There were significant SD estimates across at least some levels of all the attributes, showing heterogeneity across respondents.

Higher Hypoglycemic Diabetes Risk Sample (n = 77)

The RPL analysis of the higher hypoglycemic group (which includes part of the full sample) (Figure 2; Table S3b, SDC, shows preference priorities similar to the full sample with exception of a 1% risk of serious complications (β = −0.29, P = 0.11) now not a significant negative preference. The higher-risk group also generally exhibited stronger preference utilities for both avoiding serious risks (−2.51 vs −2.03) and preferring 5-y of insulin independence benefits (1.96 vs 1.75) for full and higher-risk samples, respectively, and also for other highest-level attributes.

The strongest preference in the higher hypoglycemia risk sample was for avoiding the highest risk (15%) for serious complications from the ICT procedure (β = −2.51, P < 0.001). The magnitude of preference weight for avoiding 15% serious complications was 72% larger than the second strongest preference overall and the strongest positive preference to gain 5 y of insulin independence (β = 1.96, P < 0.001). The magnitude of beta-coefficients (Table 3) shows that participants in the higher hypoglycemic risk group were more risk averse, as their willingness to accept the risk of serious complications decreased and more benefit was necessary to accept the risk trade-off. For example, for the most important risk attribute, serious complications, an additional utility of −1.52 was expressed by the higher-risk group, an additional 0.31 utility was expressed for insulin independence, and an additional 1.11 utility for gaining 5-y clinical success.

TABLE 3. - Preference weights for low versus additional weights for high relative to low hypoclycemic risk patients
Benchmark preference weights for low hypoglycemic risk patients Additional preference weights for high relative to low hypoglycemic risk patients
Mean estimates SD estimates Mean estimates SD estimates
Attribute and level Coef. SE T value P value Coef. SE T value P value Coef. SE T value P value Coef. T value P value
Chance of clinical success
60% 0.79** 0.38 2.08 0.04 0.56* 0.23 2.43 0.02 −0.10 0.42 −0.24 0.81 0.12 0.56 0.57
90% 1.55*** 0.41 3.75 0.00 0.47 0.30 1.57 0.12 −0.03 0.47 −0.06 0.95 1.30*** 5.65 0.00
Duration of clinical success
0.5 y
1 y 0.71 0.47 1.51 0.13 0.57** 0.24 2.43 0.02 0.33 0.52 0.63 0.53 0.50** 2.13 0.03
2 y 1.01** 0.43 2.36 0.02 0.13 0.23 0.58 0.56 0.24 0.49 0.48 0.63 1.20*** 4.43 0.00
5 y 1.30** 0.56 2.32 0.02 1.33*** 0.27 5.02 0.00 1.11 0.64 1.73 0.08 0.09 0.44 0.66
Insulin independence
2 y 1.39*** 0.40 3.47 0.00 0.60** 0.23 2.56 0.01 0.36 0.44 0.83 0.41 0.31 0.97 0.33
5 y 2.27*** 0.41 5.52 0.00 0.23 0.23 1.04 0.30 0.31 0.44 0.69 0.49 1.35*** 5.68 0.00
Reduction of T1D complications
Kidney 1.15*** 0.35 3.30 0.00 0.02 0.21 0.10 0.92 −1.17 0.39 −2.98 0.00 0.70*** 3.22 0.00
Nerve −0.06 0.36 −0.16 0.87 0.28 0.21 1.33 0.18 0.37 0.41 0.88 0.38 1.30*** 4.94 0.00
Treatable adverse effects
5% −0.49 0.48 −1.02 0.31 0.86*** 0.21 4.11 0.00 −0.33 0.52 −0.63 0.53 0.15 0.70 0.48
15% −1.22*** 0.44 −2.76 0.01 0.44** 0.20 2.17 0.03 0.31 0.49 0.62 0.53 0.29 1.38 0.17
40% −1.75*** 0.49 −3.57 0.00 0.55* 0.28 1.95 0.05 0.42 0.54 0.77 0.44 1.10*** 3.94 0.00
Serious complications
1% −0.26 0.41 −0.63 0.53 0.06 0.38 0.15 0.88 −0.32 0.46 −0.70 0.48 0.09 0.38 0.71
5% −0.97* 0.53 −1.82 0.07 1.08*** 0.23 4.77 0.00 −0.54 0.58 −0.93 0.35 0.34 1.47 0.14
15% −1.76*** 0.47 −3.78 0.00 0.31 0.36 0.88 0.38 −1.52 0.57 −2.69 0.01 2.74*** 7.39 0.00
Immunosupressant restrictions
Mouth sores, anemia
Renal monitoring −0.14 0.49 −0.28 0.78 1.01*** 0.24 4.21 0.00 −1.03 0.56 −1.83 0.07 2.00*** 6.71 0.00
Higher infection risk 0.00 0.42 0.00 1.00 0.11 0.21 0.52 0.60 −0.07 0.47 −0.15 0.88 1.04*** 4.14 0.00
Cancer 0.48 0.42 1.13 0.26 0.08 0.23 0.33 0.74 −0.31 0.48 −0.64 0.52 1.05*** 4.14 0.00
Time needed for ICTinfusion(s)
3 mo
6 mo 0.69** 0.35 1.97 0.05 0.08 0.19 0.43 0.66 −0.96 0.39 −2.43 0.01 0.38** 2.37 0.02
9 mo 0.40 0.35 1.15 0.25 0.20 0.20 0.96 0.33 −0.67 0.39 −1.74 0.08 0.05 0.23 0.82
*P ≤ 0.05.
**P ≤ 0.01.
***P ≤ 0.001.
ICT, islet cell transplantation; T1D, type 1 diabetes.

Normalized Attribute Scores

Although levels are not directly comparable across attributes, we compared mean relative attribute preferences using a normalized score‚ where the attribute with the highest difference and that was affecting preference the most was assigned a score of 10 and other attributes were scored relative to that attribute (Figure 2). Avoiding serious complications (10) was most important, followed by the benefits of duration of insulin independence (8.6) and clinical success (6.8). Gaining clinical treatment success (5.4) was almost equal in importance to avoiding treatable adverse events (5.0).

Descriptive Subgroup Analysis

Subgroup analysis of the full sample can further specify characteristics contributing to heterogeneity of preference seen in the 2 samples. There were few differences in preference scores across age (not shown). Gender differences in the preference score seem more pronounced‚ with women showing much higher utility scores for all benefits: a 68% higher preference for 5-y insulin independence and also for 90% chance of clinical treatment success and more than twice higher preference for 2-y insulin independence and 87% higher preference for 2-y duration of clinical treatment success. Men, on the other hand, showed 2 times stronger preference than women to avoid a 40% chance of treatable adverse event risks, although women had 30% stronger preference than men for avoiding the highest risk (15%) of serious complications (Figure 3; Table S4, SDC,

Attribute levels preference scores and relative attribute importance scores for islet cell transplant procedure by gender. AE, adverse event; ICT, islet cell transplantation; LT, liver transplantation; TX, transplantation.

Preferences also differed between diabetes distress scores when divided by PAID scores <40 (indicating low diabetes distress) and >40 (indicating clinically important high diabetes distress) (Figure 4). Those with high diabetes distress have almost 3 times stronger preference for 5-y insulin independence, more than twice stronger preference for 2-y insulin independence, and more than twice stronger preference for 90% chance of clinical treatment success than those with low diabetes distress. Those with high diabetes distress have more than twice stronger preference to avoid both 5% and 15% risks of serious complications from ICT and also showed a high desire to avoid the time and support required if needing 3 ICT procedures (Figure 4; Table S4, SDC,

Attribute levels preference scores and relative attribute importance scores for islet cell transplant procedure by high versus low PAID distress level. AE, adverse event; ICT, islet cell transplantation; LT, liver transplantation; PAID, Problem Areas in Diabetes ; TX, transplantation.


This is the first study to evaluate patient preferences of ICT, and it has been helpful for realizing the FDA goal to include the patient voice in its ICT decision making.

We used discrete-choice methodology to quantify preference weights for ICT risks and benefits. The results indicate that respondents most importantly preferred to avoid the highest risk (15%) of serious complications, followed by their strong desire to gain the maximum (5-y) insulin independence and duration and chance of clinical success. Although respondents are actively weighing treatment risk/benefit trade-offs, they are trading at the extremes—wanting high benefits for low risks.

The study indicates the most important benefit attribute was duration of insulin independence (mean relative attribute preference = 8.6), despite this not being the main clinical trial outcome for ICT‚ which is usually normalizing HbA1c. There was a stronger preference for 5-y insulin independence than for a 90% chance of clinically defined treatment success (normal HbA1c level), more often the primary outcome in these clinical trials. Together, this suggests that the most burdensome aspect for those with diabetes is the need to give exogenous insulin. Because of the constant demand for maximum attention to their exogenous insulin dosing needs throughout the day, accompanied by their high risk for hypoglycemia, these respondents would be most likely to find benefit from an ICT if it can diminish their insulin management needs.

Currently one of the inclusion criteria for ICT clinical trials is metabolic instability, and most likely to benefit from transplant are those who cannot achieve metabolic stability without experiencing hypoglycemia, similar to our T1D sample with severe hypoglycemia.56 We developed a diabetes risk scoring rubric to analyze preferences of a second group, which has an expected higher diabetes risk than the full sample. Both samples have broader inclusion criteria than those of ICT clinical trials. We defined 2 groups that expressed T1D with erratic glycemic control for studies unable to gather multiple blood tests‚ as do ICT clinical trials. Although those with highest-risk diabetes did show stronger preferences for both benefits and risk avoidance, this was not strong, and their preference ordering was the same. Therefore, we think the full sample well represents the preferences across persons with T1D and severe hypoglycemia.

Patient preference information (PPI) studies help us understand how viewpoints of patients, who are the end-users of treatments and devices, differ from those of clinicians and field experts. This was apparent in the DCE design process when deciding that we should include both the attributes derived from clinical trial primary outcomes and also attributes identified as important from patient interviews. We found that duration of insulin independence, a patient-derived attribute, was the most important benefit attribute and the second-most important attribute overall. Our study also demonstrated that attributes derived from clinical trial outcomes may not be ones that patients value the most. The main clinical benefit from ICT clinical trial outcomes is the reduction in HbA1c, but in this study, insulin independence was more important to respondents. Furthermore, although one of the main potential clinical benefits of ICT is reduction in progression of long-term microvascular and macrovascular complications, our study found this was one of the least important attributes to patients.21 The preference for “reduction in risks of long-term complications” was far less important relative to other benefit attributes, although realization of this risk may depend on one’s experience with long-term complications.

This study has important implications for discrete-choice methodology. Although FDA guidance currently recommends that attributes be “based on existing literature, clinical data, and/or expert” input, it does not explicitly recommend incorporating patient-derived attributes.58 Although clinical trials can pinpoint important risks and benefit attributes, it is important to also include patient perspectives in DCE design. Failing to consider patient-derived attributes may result in a DCE that does not accurately reflect real-world decision making. Another patient-derived attribute, “avoidance of time and support that may be required if multiple ICT procedures are required,” however, was significantly important for only a specific subgroup with high diabetes distress.

Importantly, life changes related to the need to take immunosuppressants after an ICT were not a major concern for these participants. Of all risks, this was least important and only rose to a significant preference when considering renal complications for the subgroup with high diabetes distress. Although immunosuppressant risks are consistently an important concern for clinicians and regulators, these participants showed a strong willingness to trade these risks for all levels of the benefits of an ICT. Although patients may not be able to fully appreciate the potential risks of immunosuppressants without experiencing them, they will be asked to make their decision to weigh this risk with expected benefits without having this experience. Further studies should also measure patient preferences after ICT to determine the stability of these preexperience preferences.

There were other differences across our subgroups, indicating differences in willingness to trade ICT risks and benefits. For example, the “duration of insulin independence” was an even stronger attribute, with the chance for serious complications less important, for those with higher diabetes distress level and Hypoglycemia Fear Survey II scores, indicating they had more concerns related to living with diabetes and willingness to trade that high risk for long relief from insulin management. The high diabetes distress group cared much more about immunosuppressant restrictions than those with low distress. For those with low diabetes distress, avoiding serious complications remained the most important attribute‚ and achieving insulin independence was second in importance, and they were generally more risk averse than the high diabetes distress group.

Patient preference studies provide invaluable input in regulatory decision making. Ho et al.59 conducted a choice-experiment survey to understand patient preferences regarding weight-loss devices, the results of which were the first to be incorporated into approval of a new weight-loss device, the Maestro Rechargeable System. Another preference study that evaluated kidney patients’ risk tolerance led to an expanded indication for a home dialysis machine, allowing patients to use the machine without a care taker present.60,61

This study has several limitations to consider. We chose to recruit patients using broader inclusion and exclusion criteria than the very strict criteria used in current ICT clinical trials, so not all subjects included in our study are currently eligible for ICT. We mitigated this potential concern by analyzing 2 risk groups using our diabetes risk tool to better understand how high-risk diabetes affects patient preference for ICT; however, our definitions of hypoglycemia risk are limited. For example, pump therapy use is not equitably distributed‚ which will diminish its value as an indicator of hypoglycemia as used in this study. We also use limited criteria to best define the hypoglycemia risk; however, using our risk tool, we showed that, even at these lower risk levels than those currently eligible for a transplant, patients are willing to undergo the risks for a sufficient amount of benefit that ICT can provide. This provides evidence of patient interest in potential adoption of this treatment for themselves across risk levels. Lastly, as in all DCEs, we present a simplified overview of ICT risks and benefits‚ and there may be unobserved attributes affecting patient preference that were not captured in this study. For example, patients may have many different conceptions of “insulin independence” and although we provided definitions, interpretations could still vary. Additionally, clinical success defined by reaching HbA1C goals and the duration of that success were separate attributes in our CBC, so each could result in a lower preference score than if combined, as in ICT clinical trial end points.

PPI studies are used to identify treatment risks and benefits that are most important to patients, to clarify patients’ willingness to make trade-offs, to understand how acceptance of risks and benefits vary within the population, and to discern differences in preference decision making between subgroups.58 The use of PPI in regulatory decision making becomes especially important when a treatment presents a wide range of benefits and potential risks, as with ICT.61 This study, conducted in collaboration with the FDA, provides essential evidence to support the concept that those with the risk of severe diabetes at 2 different risk levels are similarly willing to weigh for themselves the current levels of risks and benefits necessary to undergo ICT. These results show patient support for the introduction of ICT as a treatment modality for high-risk T1D patients. As the safety of ICT improves over time, patients may be more willing to take those more minor risks for the larger benefits, especially benefits that reduce their need for insulin. Additionally, decision-making preferences around ICT can help us understand how patients might accept other new modalities such as beta cell replacement with similar risk benefit profiles as they become more widely available.


1. American Diabetes Association. Statistics about diabetes. Available at Accessed January 16, 2020.
2. Editor. Brittle diabetes (Labile diabetes). the global diabetes community, 2019. Available at Accessed October 21, 2021.
3. Vantyghem MC, Press M. Management strategies for brittle diabetes. Ann Endocrinol (Paris). 2006;67:10–10.
4. Barrou Z, Seaquist ER, Robertson RP. Pancreas transplantation in diabetic humans normalizes hepatic glucose production during hypoglycemia. Diabetes. 1994;43:11–11.
5. Kendall DM, Rooney DP, Smets YF, et al. Pancreas transplantation restores epinephrine response and symptom recognition during hypoglycemia in patients with long-standing type I diabetes and autonomic neuropathy. Diabetes. 1997;46:11–11.
6. Cottrell DA, Henry ML, O’Dorisio TM, et al. Hypoglycemia after successful pancreas transplantation in type I diabetic patients. Diabetes Care. 1991;14:11–11.
7. Cantarovich D, Karam G, Hourmant M, et al. Steroid avoidance versus steroid withdrawal after simultaneous pancreas-kidney transplantation. Am J Transplant. 2005;5:11–11.
8. DeVries JH, Bailey TS, Bhargava A, et al. Day-to-day fasting self-monitored blood glucose variability is associated with risk of hypoglycaemia in insulin-treated patients with type 1 and type 2 diabetes: a post hoc analysis of the SWITCH Trials. Diabetes Obes Metab. 2019;21:11–11.
9. Zinman B, Marso SP, Poulter NR, et al.; DEVOTE Study Group. Day-to-day fasting glycaemic variability in DEVOTE: associations with severe hypoglycaemia and cardiovascular outcomes (DEVOTE 2). Diabetologia. 2018;61:11–11.
10. Kovatchev B, Cobelli C. Glucose variability: timing, risk analysis, and relationship to hypoglycemia in diabetes. Diabetes Care. 2016;39:11–11.
11. Rodbard D. Glucose variability: a review of clinical applications and research developments. Diabetes Technol Ther. 2018;20:11–11.
12. Foster NC, Beck RW, Miller KM, et al. State of type 1 diabetes management and outcomes from the T1D exchange in 2016-2018. Diabetes Technol Ther. 2019;21:11–11.
13. Robertson RP, Sutherland DE, Kendall DM, et al. Metabolic characterization of long-term successful pancreas transplants in type I diabetes. J Investig Med. 1996;44:11–11.
14. Robertson RP, Sutherland DE, Lanz KJ. Normoglycemia and preserved insulin secretory reserve in diabetic patients 10-18 years after pancreas transplantation. Diabetes. 1999;48:11–11.
15. Pozza G, Traeger J, Dubernard JM, et al. Endocrine responses of type 1 (insulin-dependent) diabetic patients following successful pancreas transplantation. Diabetologia. 1983;24:11–11.
16. Katz H, Homan M, Velosa J, et al. Effects of pancreas transplantation on postprandial glucose metabolism. N Engl J Med. 1991;325:11–11.
17. Diem P, Abid M, Redmon JB, et al. Systemic venous drainage of pancreas allografts as independent cause of hyperinsulinemia in type I diabetic recipients. Diabetes. 1990;39:11–11.
18. Osei K, Henry ML, O’Dorisio TM, et al. Physiological and pharmacological stimulation of pancreatic islet hormone secretion in type I diabetic pancreas allograft recipients. Diabetes. 1990;39:11–11.
19. Schuetz C, Markmann JF. Islet cell transplant: update on current clinical trials. Curr Transplant Rep. 2016;3:11–11.
20. Thompson DM, Meloche M, Ao Z, et al. Reduced progression of diabetic microvascular complications with islet cell transplantation compared with intensive medical therapy. Transplantation. 2011;91:11–11.
21. Venturini M, Angeli E, Maffi P, et al. Technique, complications, and therapeutic efficacy of percutaneous transplantation of human pancreatic islet cells in type 1 diabetes: the role of US. Radiology. 2005;234:11–11.
22. Chang CA, Lawrence MC, Naziruddin B. Current issues in allogeneic islet transplantation. Curr Opin Organ Transplant. 2017;22:11–11.
23. U.S. Food & Drug Administration. Overview of device regulation. FDA. 2019. Available at Accessed May 23, 2020.
24. U.S. Department of Health and Human Services. Public workshop - The patient preference initiative: incorporating patient preference information into the medical device regulatory processes, September 18-19, 2013, Available at Accessed May 23, 2020.
25. U.S. Food & Drug Administration. Advisory committee meeting: Cellular, Tissue, and Gene Therapies Advisory Committee April 15, 2021 meeting announcement. 2021. Available at Accessed May 17, 2022.
26. U.S. Food & Drug Administration. Cellular, Tissue, and Gene Therapies Advisory Committee Meeting CellTrans briefing document and session on product authorization BLA125734 Lantidra (donislecel) applicant for treatment of brittle type 1 diabetes mellitus 2021. Available at Accessed May 17, 2022.
27. U.S. Food & Drug Administration. Patient preference-sensitive areas: using patient preference information in medical device evaluation. 2019. Available at Accessed January 16, 2020.
28. Kawabe A, Matsumoto S, Shimoda M. Patient and family expectations of beta-cell replacement therapies in type 1 diabetes. Islets. 2018;10:11–11.
29. Barton FB, Rickels MR, Alejandro R, et al. Improvement in outcomes of clinical islet transplantation: 1999-2010. Diabetes Care. 2012;35:11–11.
30. Bekker-Grob EW de, Ryan M, Gerard K. Discrete choice experiments in health economics: a review of the literature. Health Econ. 2012;21:11–11.
31. Wardman M. A comparison of revealed preference and stated preference models of travel behaviour. J Transp Econ Policy. 1988;22:11–11.
32. McLeod C, Norman R, Schultz A, et al. Discrete choice experiment to evaluate preferences of patients with cystic fibrosis among alternative treatment-related health outcomes: a protocol. BMJ Open. 2019;9:11.
33. Foster ED, Bridges ND, Feurer ID, et al.; Clinical Islet Transplantation Consortium. Improved health-related quality of life in a phase 3 islet transplantation trial in type 1 diabetes complicated by severe hypoglycemia. Diabetes Care. 2018;41:11–11.
34. Rickels MR, Peleckis AJ, Markmann E, et al. Long-term improvement in glucose control and counterregulation by islet transplantation for type 1 diabetes. J Clin Endocrinol Metab. 2016;101:11–11.
35. Villiger P, Ryan EA, Owen R, et al. Prevention of bleeding after islet transplantation: lessons learned from a multivariate analysis of 132 cases at a single institution. Am J Transplant. 2005;5:11–11.
36. Barton FB, Stablein D. Scientific summary of the Collaborative Islet Transplant Registry Tenth Annual Report. 2017. Available at Accessed October 2021.
37. Shapiro AMJ. Islet transplantation in type 1 diabetes: ongoing challenges, refined procedures, and long-term outcome. The review of diabetic studies: review. Chapter V.1. 2013. Available at Accessed May 17, 2022.
38. Shapiro AMJ, Lakey JRT, Ryan EA, et al. Islet transplantation in seven patients with type 1 diabetes mellitus using a glucocorticoid-free immunosuppressive regimen. N Eng J Med. 2000:343:11–11.
39. Vantyghem MC, Chetboun M, Gmyr V, et al.; Members of the Spanish Back Pain Research Network Task Force for the Improvement of Inter-Disciplinary Management of Spinal Metastasis. Ten-year outcome of islet alone or islet after kidney transplantation in type 1 diabetes: a prospective parallel-arm cohort study. Diabetes Care. 2019;42:11–11.
40. Lablanche S, Borot S, Wojtusciszyn A, et al.; GRAGIL Network. Five-year metabolic, functional, and safety results of patients with type 1 diabetes transplanted with allogenic islets within the Swiss-French GRAGIL Network. Diabetes Care. 2015;38:11–11.
41. Lablanche S, Vantyghem MC, Kessler L, et al.; TRIMECO trial investigators. Islet transplantation versus insulin therapy in patients with type 1 diabetes with severe hypoglycaemia or poorly controlled glycaemia after kidney transplantation (TRIMECO): a multicentre, randomised controlled trial. Lancet Diabetes Endocrinol. 2018;6:11–11.
42. Brennan DC, Kopetskie HA, Sayre PH, et al. Long-term follow-up of the edmonton protocol of Islet transplantation in the United States. Am J Transplant. 2016;16:11–11.
43. Williams J, Jacus N, Kavalackal K, et al. Over ten-year insulin independence following single allogeneic islet transplant without T-cell depleting antibody induction. Islets. 2018;10:11–11.
44. Ortiz F, Harjutsalo V, Helanterä I, et al. Long-term mortality after kidney transplantation in a Nationwide Cohort of Patients With Type 1 Diabetes in Finland. Diabetes Care. 2019;42:11–11.
45. Luu QF, Villareal CJ, Fritschi C, et al. Concerns and hopes of patients with type 1 diabetes prior to islet cell transplantation: a content analysis. J Diabetes Complications. 2018;32:11–11.
46. Piemonti L. Islet transplantation. Available at 2000. Accessed March 16, 2021.
47. Hill AR. The CBC System for Choice-Based Conjoint Analysis. Version 9. Sawtooth Software, Inc. Published online August 2017:11. Accessed May 17, 2022.
48. Haaijer R, Wedel M. Conjoint choice experiments: general characteristics and alternative model specifications. Gustafsson A, Herrmann A, Huber F, eds. In: Conjoint Measurement: Methods and Applications. Springer; 2007:12–12.
49. Cunningham CE, Deal K, Chen Y. Adaptive choice-based conjoint analysis: a new patient-centered approach to the assessment of health service preferences. Patient. 2010;3:12–12.
50. Burroughs TE, Desikan R, Waterman BM, et al. Development and validation of the diabetes quality of life brief clinical inventory. Diabetes Spectrum. 2004;17:12–12.
51. Cox DJ, Irvine A, Gonder-Frederick L, et al. Fear of hypoglycemia: quantification, validation, and utilization. Diabetes Care. 1987;10:12–12.
52. McGuire BE, Morrison TG, Hermanns N, et al. Short-form measures of diabetes-related emotional distress: the Problem Areas in Diabetes Scale (PAID)-5 and PAID-1. Diabetologia. 2010;53:12–12.
53. American Diabetes Association. 2. Classification and diagnosis of diabetes: standards of medical care in diabetes--2021. Diabetes Care. 2021;44(Suppl 1):12–12.
54. Sarrias M, Daziano R, Croissant Y. Gmnl: Multinomial logit models with random parameters. 2020. Available at Accessed March 17, 2021.
55. Maffi P, Secchi A. Islet transplantation alone versus solitary pancreas transplantation: an outcome-driven choice? Curr Diab Rep. 2019;19:12.
56. Snoek F, Welch G. Problem Areas in Diabetes Questionnaire (PAID). Diabetes Attitudes Wishes & Needs. Available at Accessed April 1, 2022.
57. Gonder-Frederick LA, Schmidt KM, Vajda KA, et al. Psychometric properties of the hypoglycemia fear survey-ii for adults with type 1 diabetes. Diabetes Care. 2011;34:12–12.
58. U.S. Food & Drug Administration. Patient preference information - voluntary submission, review in premarket approval applications, humanitarian device exemption applications, and de novo requests, and inclusion in decision summaries and device labeling. 2016. Available at Accessed March 22, 2021.
59. Ho MP, Gonzalez JM, Lerner HP, et al. Incorporating patient-preference evidence into regulatory decision making. Surg Endosc. 2015;29:12–12.
60. Walker RC, Morton RL, Palmer SC, et al. A discrete choice study of patient preferences for dialysis modalities. Clin J Am Soc Nephrol. 2018;13:12–12.
61. Hauber AB, González JM, Groothuis-Oudshoorn CG, et al. Statistical methods for the analysis of discrete choice experiments: a report of the ISPOR Conjoint Analysis Good Research Practices Task Force. Value Health. 2016;19:12–12.

Supplemental Digital Content

Copyright © 2022 Wolters Kluwer Health, Inc. All rights reserved.