Predictors of Quality of Life and Other Patient-Reported Outcomes in the PANORAMA Multinational Study of People With Type 2 Diabetes

Long-term type 2 diabetes management usually involves diet and lifestyle changes as well as medication, with consequences for quality of life (QoL) and other patient-reported outcomes (PROs). When QoL and/or treatment satisfaction is damaged by a treatment regimen, treatment adherence may be compromised, with adverse consequences for glycosylated hemoglobin (HbA_1c) and complication risk. Insulin use and complications are usually associated with QoL impairment (1,2). Intensive treatment can result in weight gain and increased hypoglycemia and/or involve dietary restrictions, which may damage PROs. Such problems may be avoidable in patients with type 1 and type 2 diabetes when the importance of dietary freedom for QoL is recognized, protected, and measured (3–5).

Studies comparing different treatments have assessed various PRO measures (PROMs). The UK Prospective Diabetes Study (UKPDS) interpreted the finding that the EuroQoL–5 Dimension (EQ-5D) health status questionnaire did not differ between intensified treatment and conventional treatment groups to mean that intensifying treatment did not damage QoL (6). However, QoL was not measured (7). The EQ-5D actually showed that patient-reported health did not improve with intensified treatment. Although efforts have been made to distinguish between health status and QoL (7–9), health status measures are often inaccurately referred to and interpreted as “health-related QoL” or “QoL.” (For example, see ref. 10.)

“QoL” is often used as an umbrella term to refer to any PROM, including satisfaction and symptom measures. However, when the Diabetes Treatment Satisfaction Questionnaire (DTSQ) and Audit of Diabetes-Dependent Quality of Life (ADDQoL) are both included, high treatment satisfaction may be found alongside a marked negative impact of diabetes on QoL (11). To appreciate the importance of differentiating between PROMs and to understand their differential interaction with biomedical outcomes, it is useful to examine predictors of a range of PROMs. Sundaram et al. (1) led the way in investigating predictors of both diabetes-related QoL (measured by ADDQoL) and health status (Short-Form 12-Item Health Survey [SF-12]). Others followed, using the ADDQoL alone or with a health status tool (12,13). However, the small number of PROMs examined and selective samples (e.g., those within 5 years of diagnosis [14], from a specific clinic [1], or from one country [13]) limit the finding’s generalizability.

PANORAMA, a multinational, real-world, cross-sectional study (NCT00916513), aimed to fill the gap in the literature by assessing a range of well-validated and/or widely used PROMs simultaneously in a larger, more representative sample of patients with type 2 diabetes. An earlier study reported univariate associations between hypoglycemia frequency and PROMs (14). Here, we report a different picture obtained from multivariable analyses investigating the contributions of treatment intensity, glycemic control, hypoglycemia, and other variables as predictors of PROMs. This is the first time that health status, QoL, diabetes-specific QoL, diabetes treatment satisfaction, and fear of hypoglycemia as a particular selection of PROMs have been used simultaneously and the first time that their predictors have been examined. Predictors of the various PROMs measuring distinct constructs were expected to differ (15).

The PANORAMA study design has previously been published (15). Patients were enrolled from nine countries: Belgium, France, Germany, Greece, Italy, the Netherlands, Spain, Turkey, and the U.K. The primary objective was to investigate PROMs (ADDQoL, DTSQ, the Hypoglycemic Fear Survey-II [HFS-II] worry subscale, and EQ-5D) in patients with type 2 diabetes with multivariable analyses to examine demographic and clinical predictors of each PROM.

Physicians and patients were recruited between May 2009 and February 2010. In each country, physicians managing type 2 diabetes were randomly selected. In seven countries, primary care physicians were recruited, but in Italy and Greece, hospital diabetologists were selected to reflect country-specific practice.

Eligible patients with type 2 diabetes were randomly (in the Netherlands, Spain, and U.K.) or consecutively (in Belgium, France, Germany, Greece, Italy, and Turkey—countries without electronic health records) selected from primary care or specialist clinics. Inclusion criteria were as follows: age ≥40 years, diagnosis of type 2 diabetes for ≥1 year plus medical records available for ≥1 year. All patients received diet/exercise advice: most were treated with oral antidiabetes drugs (OADs), insulin, and/or glucagon-like peptide 1 receptor agonist (GLP-1) analogs. Treatment was unchanged for 3 months.

Key exclusion criteria included the following: type 1 diabetes and/or history of diabetic ketoacidosis or secondary diabetes (including exocrine-pancreas disease, endocrinopathies) and pregnancy.

Good Clinical Practice was followed throughout.

At the single study visit, participating patients and their physicians completed the study questionnaires; medical records’ data were collected as previously described, including diabetes-related complications, comorbidities, and presence or history of depression (15). No information was collected on how depression was diagnosed or whether it was documented by a mental health provider. HbA_1c levels were measured using the A1CNow SELFCHECK, a point of care–certified device (Bayer Diabetes Care, now manufactured by PTS Diagnostics) (16). Physicians reported hypoglycemia frequency and severity, with severe hypoglycemia defined as symptomatic episodes requiring external assistance owing to severe impairment and recovery after glucose/glucagon administration. Physicians classified patient adherence to medication and lifestyle recommendations as “poor,” “moderate,” or “good.” If patients were not at their HbA_1c target, physicians were asked why: options included “resistance/reluctance of patient to intensify their medication regimen (adding agent or increasing dose)” and “reluctance of physician to intensify treatment.” Subcategories of reasons for physician reluctance were provided, e.g., ‘fear of hypoglycemia.’

Patients completed PROM questionnaires, including ADDQoL (2,17), DTSQ (18,19), the HFS-II worry subscale (20), and EQ-5D (21–23). Linguistically validated (24) PROMs were available in the preferred language. Questionnaires can be accessed as follows: for ADDQoL and DTSQ, go to www.healthpsychologyresearch.com; for EQ-5D, go to www.euroqol.org; and for HFS-II, e-mail lag3g@hscmail.mcc.virginia.edu.

The widely used ADDQoL measures the impact of diabetes on QoL. Two overview items measure generic QoL (7-point scale from excellent [+3] to extremely bad [−3]) and diabetes-dependent QoL (In response to “If I did not have diabetes, my QoL would be:” the answers correspond to a 5-point scale from very much better [−3] to worse [+1]). Five-point impact scale scores relating to 19 specific life domains (e.g., very much better/greater [−3] to worse/less [+1]) are multiplied by related 4-point importance ratings (very important [+3] to not at all important [0]) to produce a weighted impact (WI) score (−9 to +3). For five life domains, preliminary questions determined applicability. Averaging WI scores over applicable domains produce an average weighted impact (AWI) score (−9, maximum negative impact of diabetes, to +3, maximum positive impact).

Recommended by the World Health Organization and International Diabetes Federation to assess diabetes care, the treatment satisfaction score is the sum of six items rated on 7-point scales (6, very satisfied, to 0, very dissatisfied). Two further items, analyzed separately, measure perceived frequency of hyperglycemia/hypoglycemia, (6, most of the time, to 0, none of the time). Physicians also independently completed the DTSQ as if they were each individual patient (see Supplementary Data and Supplementary Table 1 for psychometric properties).

The widely used HFS-II worry subscale measures patient concern about hypoglycemia and its consequences. The 18 items are rated on 5-point Likert scales (0, never, to 4, almost always) which, summed, range from 0, least worry, to 72, most worry.

EQ-5D is a generic measure of health status commonly used in cost-effectiveness analyses and recommended by the National Institute for Health and Care Excellence (NICE) (25). EQ-5D comprises a visual analog scale (EQ-VAS) plus five further questions. Only the EQ-VAS score is reported here.

A sample size of 753 patients per country was determined to provide sufficient precision for the primary outcome in terms of the 95% CIs around each of the country-specific means, given the expected SD (15).

Episodes of nonsevere hypoglycemia were recorded per month; episodes of severe hypoglycemia were recorded per year. Patients with diabetic nephropathy were categorized by the most severe subcategory reported. As in previous PANORAMA articles, patients receiving GLP-1 analogs plus insulin and/or OADs were included in the “on insulin with OADs” (n = 3) or the “on insulin alone” (n = 3) treatment groups; patients on GLP-1 analogs alone (n = 48) were excluded from analyses.

Bivariate Pearson correlations were conducted on the ADDQoL overview items and AWI scores.

Mixed-model linear regression was used to analyze variables associated with PROM scores (change in PROM scores with 1-unit change in the independent variable), including physician as a random effect; this model takes account of any clustering of patient and treatment characteristics by physician. Multivariable analyses were adjusted for potential confounding by the other variables included (e.g., patient characteristics, clinical/biological measures), by physician, and physician characteristics. For continuous variables, average differences (95% CIs) were given per additional year (age and diabetes duration), per additional 0.1% (target HbA_1c), per additional kg/m² (BMI), per additional mmHg (systolic/diastolic blood pressure [BP]), and per visit (to general practitioners/specialists). For categorical variables, “yes” was compared with “no,” except for HbA_1c (≥7% vs. <7%) and physician-rated adherence to medication/lifestyle (“good” vs. “moderate”/“poor” adherence); treatment categories were compared with diet/exercise alone.

The multivariable analyses were repeated without treatment covariates to determine whether intensified treatment (including increased propensity for hypoglycemia) was masking effects of hypoglycemia more specifically.

Patients were recruited by 390 physicians, 77.4% of whom (298 of 390) were general practitioners. On average, physicians saw 37.8 patients with diabetes weekly.

PANORAMA enrolled 5,817 patients. Four participants with data-reporting errors, identified after database lock, were excluded from analysis. Overall, the 5,813 patients had a mean age of 65.9 years and diabetes duration of 8.9 years (Table 1). Mean HbA_1c at study visit was 6.9% (52 mmol/mol). The majority was treated with OADs alone (68.9% [3,959 of 5,750]). Microvascular complications were recorded for 35.9% of patients; incidence was higher (P = 0.028) in countries recruiting sequentially (36.9% [1,456 of 3,951]) rather than randomly (33.9% [631 of 1,862]).

Overview item 1 responses gave a mean generic QoL score of 0.93 approximating to “good” (1) as opposed to “very good” (2) or “neither good nor bad” (0) (Table 2; Fig. 1A). However, mean diabetes-related QoL (overview item 2) was −1.26, indicating QoL would be better without diabetes (Fig. 1B). Among the 19 domain-specific items, respondents rated “freedom to eat as I wish” as the most negatively impacted/important (WI score = −3.35)—substantially more so than other items (Fig. 1C). Mean AWI score for the domain-specific items was −1.69, indicating substantial perceived negative impact of diabetes on QoL. AWI correlated more highly (r = 0.60; P < 0.001) with the diabetes-specific overview item, as expected, than with the generic QoL item (r = 0.21; P < 0.001).

The DTSQ score was generally high, as assessed by patients (mean 29.76) and their respective physicians (27.75) (Table 2). Mean physician ratings were lower than mean patient ratings, indicating that physicians rated patients as less satisfied than patients rated themselves (6-item score P < 0.001). However, physicians significantly underestimated hyper- and hypoglycemia compared with their patients (P < 0.001) (Table 2).

The mean HFS-II worry subscale score was 13.27, indicating generally low worry about hypoglycemia (Table 2).

Mean patient-reported health status on the EQ-VAS was 70.55, where 100 is best imaginable health and 0 is worst (Table 2).

Age was significantly associated with ADDQoL AWI score showing less negative perceived impact of diabetes on QoL with increasing age (average difference of AWI score changed 0.01 per year of age [95% CI 0.01–0.02; P < 0.001]) (Table 3; Supplementary Fig. 1A). The strongest negative association with AWI scores was insulin treatment alone (vs. diet/exercise alone), with an average difference of −0.58 (95% CI −0.84 to −0.33; P < 0.001). The two next most intensive treatments, insulin + OADs and three OADs without insulin, were also significantly associated with worse AWI scores. Other associations with worse AWI scores included HbA_1c ≥7%, microvascular complications (but not macrovascular), and depression. Neither measure of hypoglycemia (any “severe”/”nonsevere” episodes) showed any significant relationship with AWI scores in this multivariable analysis, which controlled for other variables, including treatment.

The ADDQoL Generic QoL score had nearly twice as many significant associations as the AWI score (Table 3; Supplementary Fig. 1B vs. A). Although lower HbA_1c was significantly associated with higher (better) AWI score, it had no significant relationship with generic QoL. Several factors were significantly associated with higher (better) ADDQoL generic QoL scores: the strongest was male sex (average difference was 0.16 [95% CI 0.11–0.22; P < 0.001), followed by physician-reported good adherence and self-monitoring of blood glucose (SMBG). The strongest negative association was having depression, average difference −0.38 (95% CI −0.46 to −0.30; P < 0.001), followed by sleep disorders, −0.18 (95% CI −0.26 to −0.10; P < 0.001), and being on insulin with OADs −0.21 (95% CI −0.37 to −0.06; P = 0.005) or without (vs. diet/exercise)—but not being on three OADs (significant for the ADDQoL AWI score). Being unemployed (vs. not unemployed) predicted generic QoL but not AWI score (−0.20 [95% CI −0.33 to −0.07]; P = 0.002). Other factors negatively associated with ADDQoL Generic QoL scores included increasing age (positively associated with AWI scores), living alone, higher BMI (unrelated to AWI scores), macrovascular (but not microvascular) complications, comorbidities, and more physician visits. Whereas neither hypoglycemia nor BP was related to AWI score, any nonsevere (but not severe) hypoglycemia was associated with worse generic QoL, and higher systolic (but not diastolic) BP was associated with better generic QoL.

Lower HbA_1c was associated with greater satisfaction measured by the DTSQ 6-item score: having HbA_1c ≥7% (vs. <7%) was most strongly associated with worse DTSQ patient scores (−1.25 [95% CI −1.63 to −0.87]; P < 0.001) (Table 3; Supplementary Fig. 1C). Other factors associated with less treatment satisfaction included depression, weight gain, abdominal pain, physician-reported patient reluctance to intensify treatment, and treatment with insulin + OADs. The strongest positive association with patients’ treatment satisfaction was physician-reported good medication adherence (vs. moderate/poor adherence) (1.07 [95% CI 0.66–1.48]; P < 0.001), followed by physician-reported good adherence to lifestyle changes. Unlike ADDQoL scores, no patient characteristics (including age) significantly predicted DTSQ scores.

Fear of hypoglycemia was most strongly predicted by reported hypoglycemia: those experiencing any severe or nonsevere hypoglycemia reported greater hypoglycemia worry (Table 3; Supplementary Fig. 1D). Another strong predictor was being on insulin treatment alone (vs. diet/exercise alone), with a difference of 3.35 (95% CI 0.92–5.79; P = 0.007). Other factors associated with increased worry about hypoglycemia included depression and SMBG use. Men worried less about hypoglycemia than women (−2.56 [95% CI −3.45 to −1.67]; P < 0.001), as did those with higher HbA_1c targets (−1.29 [95% CI −2.52 to −0.07]; P = 0.039) and older people (−0.11 [95% CI −0.16 to −0.07]; P < 0.001). HbA_1c did not predict hypoglycemia worry.

Depression was the strongest predictor of lower EQ-VAS scores (worse self-rated health), with a difference of −6.44 (95% CI −7.85 to −5.03; P < 0.001). EQ-VAS health status scores, like generic QoL scores, were worse with age and better in men, while ADDQoL AWI scores improved with age without sex associations. HbA_1c had no association with EQ-VAS. Physician-reported good adherence to lifestyle (2.84 [95% CI 1.70 to 3.97]; P < 0.001) and medication were associated with better EQ-VAS scores. All medical disorders examined, including micro- and macrovascular complications, were associated with worse health status, as were higher BMI and increased visits to physicians. However, higher systolic BP was marginally associated with better health status (Table 3; Supplementary Fig. 1E). No treatment variable was associated with health status.

Multivariable regression analysis without treatment covariates found small changes in the average difference and 95% CI of the PROMs. Few associations changed in significance; nonsevere hypoglycemia became associated with less treatment satisfaction (−0.57 [95% CI −1.04 to −0.10]; P = 0.018), and its association with greater perceived negative impact of diabetes on QoL became close to significant (−0.13 [95% CI −0.26 to −0.00]; P = 0.050).

PANORAMA provided an overview across nine countries of five PROMs validated and widely used in diabetes. Simultaneous collection of PROMs and clinical data allowed exploratory analyses to identify predictors of outcomes that matter to patients. Numbers and patterns of predictors for different PROMs supported the hypothesis that they would differ in important ways (Table 3).

Nearly three-quarters of patients indicated that their QoL would be better without diabetes. The overall mean ADDQoL AWI score (−1.69) showed that QoL was impaired by diabetes to a degree similar to that found in other unselected samples (1,11,12). In the ADDITION Europe study (Anglo-Danish-Dutch Study of Intensive Treatment In People with Screen-Detected Diabetes in Primary Care), patients reported lesser negative impact of screen-detected type 2 diabetes 5 years postdiagnosis (median AWI −0.32) (13). As in most other studies using the ADDQoL, the item reflecting the greatest negative impact of diabetes on QoL was “freedom to eat as I wish” (WI −3.35) and higher HbA_1c levels were associated with greater negative impact of diabetes on QoL (ADDQoL AWI score) (1,2,12,13,17).

AWI score was significantly and negatively associated with other important variables such as presence of microvascular complications, depression, and sleep disorders, as well as by more intensive treatments. As found previously (1,2,11–13,17), insulin use was associated with more negative impact of diabetes on QoL versus oral treatment or diet/exercise alone. Treatment with three OADs without insulin was also significantly associated with more negative AWI scores. These findings suggest that intensifying treatment for type 2 diabetes above two OADs is likely to damage QoL and, given that “freedom to eat as I wish” is the item most negatively impacted by diabetes, greater damage can be expected when dietary freedom (including having to eat when not wanting to) is limited. Future research would usefully categorize treatments by their dietary restrictions to test this hypothesis and highlight treatments that protect QoL.

Univariate analyses of the PANORAMA data had indicated a relationship between nonsevere hypoglycemia and impact of diabetes on QoL. However, this relationship was no longer present when multivariable analyses controlled for treatment, suggesting that nonsevere hypoglycemia does not add further negative impact of diabetes on QoL over and above that of intensified treatments. Nonetheless, the ADDQoL generic QoL item and EQ-5D VAS were significantly associated with nonsevere hypoglycemia regardless of whether treatment variables were included in the model. Insulin treatment with or without OADs was predictive of worse generic QoL. It may be that patients attribute some symptoms of preprandial or nocturnal hypoglycemia to malaise unrelated to diabetes, which reduces their ratings of health status and generic QoL, but not their ADDQoL AWI scores. If so, hypoglycemia symptom monitoring (26) alongside SMBG would help identify and avoid hypoglycemia.

Older age was associated with a small positive difference in AWI score; however, the opposite was found for generic QoL and EQ-VAS health status, indicating that while older patients felt that diabetes impacted less negatively on QoL than younger patients did, they perceived their overall health as worse and, perhaps as a result, their overall QoL. These PROMs also differed in their associations with HbA_1c; higher HbA_1c was associated with worse diabetes-specific AWI scores but not generic QoL or health status. This may be due to difficulties and conflicts experienced by those with elevated HbA_1c, which are reflected in responses to ADDQoL domains, contributing to AWI scores. Health status and generic QoL are not directly associated with HbA_1c, only becoming related via diabetes complications.

Overall treatment satisfaction in PANORAMA, measured by the DTSQ, showed that patients were generally satisfied with their treatment; their physicians slightly but significantly underestimated this satisfaction (while underestimating patients’ experience of hyper- and hypoglycemia). In PANORAMA, physician-reported adherence was a crude estimate measured on a single 3-point scale but was associated with greater patient treatment satisfaction, an association observed elsewhere using patient-reported adherence, and treatment satisfaction measures (27), suggesting that this physician rating has some validity. Physician-rated adherence was not attributable to HbA_1c level, which was controlled for in the multivariable analysis. However, lower HbA_1c predicted greater patient treatment satisfaction. Treatment satisfaction was negatively associated with insulin treatment + OADs (vs. diet/exercise),but was not significantly associated with other treatments. The treatment group definitions in PANORAMA were too broad for discernment of differences between insulin regimens, which have previously been associated with differences in treatment satisfaction and other PROMs (28), or classes of OADs, which have different tolerability profiles (29,30). Treatment satisfaction was negatively associated with physician-reported patient reluctance to intensify treatment. Perceived reluctance (reported for 11% of patients) may be overcome if physicians share with patients research showing that satisfaction improves once individuals with suboptimal diabetes control start insulin (28).

The mean HFS-II worry score showed that hypoglycemia worry was not high, although worry was greater with use of insulin alone vs. diet/exercise. Sex associations were observed: women worried more about hypoglycemia than men, as reported in patients with type 1 diabetes (20). Hypoglycemia worry was greater in younger patients and those with physician-reported depression. Patients with sleep disorders also worried more about hypoglycemia, perhaps reflecting experience of nocturnal hypoglycemia. Having “severe” or “nonsevere” hypoglycemic episodes strongly predicted HFS-II worry scores while not predicting other diabetes-specific PROMs studied here.

Overall PANORAMA EQ-VAS health status scores were generally good (mean 70.55 and median 70). Indeed, EQ-VAS scores were higher than those recently reported for Polish patients with type 2 diabetes (54.9 for 55- to 64-year-old patients; 50.2 in those ≥65 years old) (31). A recent cross-sectional German general population study reported a mean EQ-VAS score of 84.9 in 60–69 year olds not reporting health problems; people with diabetes of all ages had a mean −24.4 reduction in EQ-VAS (32).

Depression was the strongest negative predictor of health status and the only predictor significant for all five PROMs. Despite significant negative associations of insulin treatments with other PROMs, treatment was not significantly associated with EQ-VAS health status, suggesting that this PROM is unlikely to be a good choice for inclusion in clinical trials comparing diabetes treatments. As expected, patients with macro- and/or microvascular complications reported worse health than those without, but other factors, including sleep disorders, showed greater differences. Patients rarely consider their eyes when rating their health while reporting negative impact of eye conditions on QoL using eye condition–specific QoL measures modeled on the ADDQoL (33). Retinopathy will be associated with generic and diabetes-specific QoL rather than perceived health. In PANORAMA, higher systolic BP was associated with better health status (marginally significant) and better ADDQoL generic QoL, possibly due to side effects of BP-lowering medications.

The PANORAMA study has strengths and limitations. A key strength is the use of a selection of well-validated PROMs. Use of these tools in validated translations in a very large multicountry sample of patients with type 2 diabetes increases the generalizability of the findings. However, as with all cross-sectional data, caution is needed when making causal inferences. One limitation was lack of clarity in the Case Report Form questions about why physicians were reluctant to intensify treatment. Few physicians endorsed the initial question to indicate their reluctance to intensify treatment, but many ticked subsequent reasons for reluctance suggesting that patient reluctance was being considered as well as/instead of their own. Responses to the initial question about physician reluctance used in the models did not significantly predict any PROM. Improving the design of this question may in the future reveal associations between PROMs and particular reasons for physician reluctance to intensify treatment. We acknowledge that sequential recruitment based on clinic attendance (in six countries), rather than randomization from eligible preselected patients (three countries), led to a small but significant bias toward patients with more microvascular complications.

Our findings in the PANORAMA population show that individual PROMs performed in much the same ways as in prior reports of each individually. However, by assessing them simultaneously in this large multinational population we confirmed the hypothesis that predictors of the various PROMs studied would vary widely. Notably, while QoL and the negative impact of diabetes on QoL (measured by the ADDQoL) were worse for patients on more intensified treatment, health status (EQ-5D VAS) was unrelated to treatment intensity. To reach conclusions about QoL we must measure QoL; health status or another PROM is no substitute. The PANORAMA findings demonstrate the importance of appreciating the differences between QoL, health status, and other PROMs and interpreting them appropriately. Genuine measures of QoL and diabetes-dependent QoL are needed, perhaps alongside other PROMs, when evaluating and choosing between diabetes treatments if we are to protect QoL, increase regimen adherence, and thereby improve glycemic control in the long term.

Acknowledgments. The authors thank Mónica Tafalla, from AstraZeneca, for overall management of the trial; Pierre Maheux, from AstraZeneca, for contributions to the study design and critical review of the data; Ian Wood, formerly of Bristol-Myers Squibb, for critical review of the data and manuscript; and Richard Cairns (Worldwide Clinical Trials, Nottingham, U.K.) for statistical analysis. Sally Cotterill and Nikki Kendrick (QXV Communications, an Ashfield Company, Macclesfield, U.K.) provided medical writing support.Duality of Interest. PANORAMA was funded by AstraZeneca and Bristol-Myers Squibb. Medical writing support and statistical analysis were also funded by AstraZeneca and Bristol-Myers Squibb. AstraZeneca and Bristol-Myers Squibb were involved in the design and conduct of the study and data collection, analysis, and interpretation. After the divestiture of diabetes projects from Bristol-Myers Squibb to AstraZeneca, AstraZeneca was involved in the preparation and review of the manuscript and the decision to submit the manuscript for publication. C.B., P.d.P.-V., and D.S. are co-chairs of the PANORAMA steering committee. All non-company authors have received honoraria/consulting fees for participation in the PANORAMA advisory committee. All authors received honoraria and expenses for attending steering group meetings and conferences where PANORAMA results were presented. C.B. is copyright owner of the ADDQoL and DTSQ used in PANORAMA and a director and majority shareholder in the spin-off company, Health Psychology Research (HPR) Ltd., which licenses C.B.’s questionnaires. C.B. receives royalties when the questionnaires are licensed to commercial companies and advises others on the use of the questionnaires. HPR Ltd. pays C.B.’s university for a percentage of her time at full economic costs. HPR Ltd. licenses C.B.’s questionnaires for others to use and supports the linguistic validation and further development of her questionnaires. The study sponsors (AstraZeneca and Bristol-Myers Squibb) licensed the ADDQoL and DTSQ from HPR Ltd. C.B. has also received honoraria and expenses for speaking about the measurement of PROs at meetings organized by AstraZeneca and Bristol-Myers Squibb and by other pharmaceutical companies including Lilly Global and Novo Nordisk. She has advised most of the major pharmaceutical companies on the use of PROMs, including the ADDQoL and DTSQ, in clinical trials. P.d.P.-V. has received speaker fees from Bristol-Myers Squibb, Novartis, Sanofi, and Takeda. K.G.P. has received consulting fees and/or speaker fees from AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, and Merck Sharp & Dohme. D.S. has served on speakers’ bureaus for LifeScan, Lilly France, and Sanofi; has served on advisory panels for AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Janssen, Merck Sharp and Dohme, Novartis, and Takeda; and has participated as an investigator in studies conducted by Lilly France, Novartis, and Novo Nordisk. H.V. was a full-time employee of AstraZeneca at the time of study conduct. This employment has ceased. L.G.-F. is co-copyright owner of the HFS-II and receives licensing fees, which support research programs in hypoglycemia at the University of Virginia, from corporations that use the survey in their research. The study sponsors (AstraZeneca and Bristol-Myers Squibb) licensed the HFS-II from the University of Virginia. L.G.-F. has also received honoraria and expenses for consulting from AstraZeneca, Bristol-Myers Squibb, Johnson & Johnson, Abbott Diabetes Lab, and Dexcom, Inc. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.B., E.E., P.d.P.-V., K.G.P., D.S., and L.G-F. assisted in the design of the study. C.B., E.E., P.d.P.-V., K.G.P., D.S., H.V., and L.G.-F. interpreted study data, critically reviewed the manuscript, and approved the final version for submission. All of the authors had final responsibility for the decision to submit for publication. C.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.