Diabetes Ther (2018) 9:1217–1232 https://doi.org/10.1007/s13300-018-0430-4 ORIGINAL RESEARCH DEVOTE 5: Evaluating the Short-Term Cost-Utility of Insulin Degludec Versus Insulin Glargine U100 in Basal–Bolus Regimens for Type 2 Diabetes in the UK . . . Richard F. Pollock William J. Valentine Steven P. Marso . . . . Jens Gundgaard Nino Halle ´ n Lars L. Hansen Deniz Tutkunkardas John B. Buse On behalf of the DEVOTE Study Group Received: March 7, 2018 / Published online: April 30, 2018 The Author(s) 2018 Methods: A cost-utility analysis was conducted ABSTRACT over a 2-year time horizon using a decision analytic model to compare costs in patients Introduction: The aim of this study was to receiving once daily degludec or glargine U100, evaluate the short-term cost-utility of insulin both as part of a basal–bolus regimen, in addi- degludec (degludec) versus insulin glargine 100 tion to standard care. Clinical outcomes and units/mL (glargine U100) for the treatment of patient characteristics were taken exclusively type 2 diabetes in the basal–bolus subgroup of from DEVOTE, whilst health-related quality of the head-to-head cardiovascular (CV) outcome life utilities and UK-speciﬁc costs (expressed in trial, DEVOTE. 2016 GBP) were obtained from the literature. The analysis was conducted from the perspec- tive of the National Health Service. Results: Degludec was associated with mean cost savings of GBP 28.78 per patient relative to Enhanced Digital Features To view enhanced digital features for this article go to https://doi.org/10.6084/ glargine U100 in patients with type 2 diabetes at m9.ﬁgshare.6128024. high CV risk. Cost savings were driven by the reduction in risk of diabetes-related complica- Electronic supplementary material The online tions with degludec, which offset the higher version of this article (https://doi.org/10.1007/s13300- 018-0430-4) contains supplementary material, which is treatment costs relative to glargine U100. available to authorized users. Degludec was associated with a mean improve- ment of 0.0064 quality-adjusted life-years R. F. Pollock (&) W. J. Valentine (QALYs) compared with glargine U100, with Ossian Health Economics and Communications improvements driven predominantly by lower GmbH, Basel, Switzerland e-mail: firstname.lastname@example.org rates of severe hypoglycemia with degludec versus glargine U100. Improvements in quality- S. P. Marso adjusted life expectancy combined with cost HCA Midwest Health Heart and Vascular Institute, neutrality resulted in degludec being dominant Kansas City, MO, USA over glargine U100. Sensitivity analyses J. Gundgaard N. Hallen L. L. Hansen demonstrated that the incremental cost-utility D. Tutkunkardas ratio was stable to variations in the majority of Novo Nordisk A/S, Søborg, Denmark model inputs. J. B. Buse Conclusion: The present short-term modeling University of North Carolina School of Medicine, analysis found that for the basal–bolus Medicine/Endocrinology, Chapel Hill, NC, USA 1218 Diabetes Ther (2018) 9:1217–1232 subgroup of patients in DEVOTE, with a high diabetes reported an episode of severe hypo- risk of CV events, degludec was cost neutral (no glycemia during the 4-week prospective study additional costs) compared with glargine U100 . Severe hypoglycemic events requiring hos- over a 2-year time horizon in the UK setting. pitalization can pose a signiﬁcant ﬁnancial Furthermore, there were QALY gains with burden on healthcare systems: in the UK, the degludec, particularly due to the reduction in average direct medical cost of an event was the risk of severe hypoglycemia. estimated at over GBP 1300 in patients with Funding: Novo Nordisk A/S. type 2 diabetes . Hypoglycemia has an acute, Trial Registration: ClinicalTrials.gov identiﬁer, negative impact on clinical outcomes, includ- NCT01959529. ing an increased risk of falls, fractures, cardio- vascular (CV) events, coma, dementia, neurological conditions, and death, but can also Keywords: Cardiovascular outcome trial; Cost; have an adverse effect on longer-term diabetes Cost-effective; Diabetes; Hypoglycemia; Insulin management due to the fear of hypoglycemia degludec; QALY; United Kingdom [9–14]. Insulin degludec (degludec) is a basal insulin with an ultralong duration of action and a INTRODUCTION stable glucose-lowering proﬁle . Studies of the pharmacodynamics of degludec and Healthcare systems in many countries are under increasing ﬁnancial pressure due to the demo- another long-acting insulin analog, insulin glargine 100 units/mL (glargine U100), have graphic challenge of aging populations and the growing burden of chronic diseases, including conﬁrmed the lower day-to-day and within-day variability in glucose-lowering effect with diabetes . Globally, one in 11 adults (425 million) has diabetes, of which type 2 diabetes degludec compared with glargine U100 [16, 17]. Phase 3 clinical studies have established that makes up around 90% of cases, with the total cost of diabetes to healthcare systems estimated similar improvements in glycemic control can be achieved with fewer hypoglycemic episodes, to be USD 727 billion in 2017 . The cost of particularly nocturnal hypoglycemia, with treating diabetes-related complications is sub- stantial and in many healthcare systems degludec versus glargine U100 across a broad spectrum of patients with diabetes [18–20]. exceeds the cost of blood-glucose-lowering medication . With increasing constraints on Recently, results from the ﬁrst double-blind, active-comparator cardiovascular outcome trial healthcare budgets, health economic evalua- tions, including cost-utility analyses (CUA), are (CVOT) of a speciﬁc antihyperglycemic therapy were published [21, 22]. DEVOTE playing an increasingly important role in deci- sions to allocate resources between therapy (NCT01959529) evaluated the CV safety of degludec relative to glargine U100 in patients areas and interventions. CUA, a special type of with type 2 diabetes at a heightened risk of CV cost-effectiveness analysis, compares the costs complications . In DEVOTE, degludec was of new interventions with their outcomes noninferior to glargine U100 in terms of the measured in utility-based units, most com- monly the quality-adjusted life-year (QALY). incidence of CV events . Furthermore, the trial demonstrated a signiﬁcant reduction in the This enables the comparison of alternative interventions according to cost per QALY risk of severe hypoglycemia for degludec versus glargine U100 at a similar level of glycemic gained, and assists in the efﬁcient allocation of resources to achieve maximum healthcare gains control . The present evaluation focused on the sub- within the constraints of a limited budget . Episodes of severe hypoglycemia are becom- group of patients who started DEVOTE on a basal–bolus regimen. Guidelines recommend ing increasingly common, particularly as that patients with advanced or very poorly attention focuses on intensive glycemic control [5–7]. In the Hypoglycaemia Assessment Tool controlled diabetes are managed by basal–bolus therapy . Basal–bolus regimens most closely (HAT) study, 8.9% of patients with type 2 Diabetes Ther (2018) 9:1217–1232 1219 mimic the natural pattern of insulin secretion person to actively administer carbohydrate or found in individuals without diabetes but are glucagon or to take other corrective actions complex and costly, as they require multiple . All outcomes, including severe hypo- injections per day and therefore an elevated use glycemia, were conﬁrmed by central, blinded of consumables such as needles and self-moni- review by an independent event adjudication tored blood glucose (SMBG) tests. This subgroup committee. Data on nonsevere hypoglycemia of patients have advanced type 2 diabetes, and were not collected in DEVOTE. are thus at an increased risk of complications relative to patients with a shorter duration of Basal–Bolus Subgroup Analysis disease . Zoungas et al. report that for every 5-year increase in diabetes duration, the risk of This analysis used data on clinical outcomes major adverse cardiovascular events (MACE) derived from the basal–bolus subgroup of and all-cause death increased by 49 and 78%, DEVOTE (Table 1,). The basal–bolus sub- respectively, when accounting for age at diag- group was deﬁned as using basal–bolus (or pre- nosis . Patients on basal–bolus regimens also mix or bolus insulin only) at baseline (before have an increased incidence of hypoglycemia switching to degludec or glargine U100) and relative to those using basal insulin only constituted 46% (n = 3515; degludec, 1760; [25, 26], which can negatively impact quality of glargine U100, 1755) of the patients in life whilst imposing a signiﬁcant economic DEVOTE. Patients on premix regimens were burden through increased healthcare resource switched to appropriate bolus injections at the utilization and loss of productivity [27–29]. The start of the trial. Time to ﬁrst composite MACE, aim of this analysis was therefore to report the which was the primary outcome measure in clinical outcomes for the basal–bolus subgroup DEVOTE and deﬁned as the ﬁrst occurrence of of DEVOTE, and to evaluate the cost-utility of nonfatal MI, nonfatal stroke, or CV mortality, degludec versus glargine U100 in patients trea- was analyzed using a Cox proportional hazards ted with a basal–bolus regimen at baseline, over regression model on the composite endpoint. the 2-year trial duration using a simple and The number of severe hypoglycemic episodes transparent decision analytic model in the UK was analyzed using a negative binomial regres- setting. sion model with a log-link function and log(- duration of observation time) as offset. Death from other causes, i.e., death from causes other METHODS than ﬁrst MACE, was analyzed using a Cox proportional hazards regression model. Insulin The DEVOTE Trial dose (units/kg) was log-transformed and ana- lyzed with a mixed model for repeated measures DEVOTE was a multinational, treat-to-target, (MMRM) within subjects using an unstructured randomized, double-blind, active comparator- residual covariance matrix among visits. Visit controlled CVOT conducted in 20 countries interactions with age, dose at baseline, body . A total of 7637 patients with type 2 dia- mass index, alternative titration target (yes/no), betes and at high risk of CV events were ran- and treatment were included in the model as domly assigned to receive either degludec (100 ﬁxed effect covariates. Change from baseline in units/mL) or glargine U100, both once daily in HbA to the 24-month visit was analyzed using 1c addition to standard care . The event-driven an MMRM within subjects using an unstruc- trial was designed to continue until the occur- tured residual covariance matrix among visits at rence of at least 633 ﬁrst MACE—a combined 6, 12, and 24 months of the trial; interactions endpoint of non-fatal myocardial infarction between visit and treatment and between visit (MI), non-fatal stroke, and CV death . Severe and baseline HbA were included as ﬁxed 1c hypoglycemia was self-recorded, deﬁned effects in the model. according to the American Diabetes Association as an episode requiring the assistance of another 1220 Diabetes Ther (2018) 9:1217–1232 Table 1 Clinical outcomes from the DEVOTE basal–bolus subgroup Degludec/glargine U100 SE 95% CI P value Annual event rate per PYE ratio b Glargine Degludec U100 Complications First MACE 0.81 0.10 [0.66; 0.035 0.0607 0.0489 0.98] c c c Severe hypoglycemia 0.63 0.15 [0.47; 0.002 0.0850 0.0536 0.85] Death from other causes 0.77 0.19 [0.53; 1.12] 0.165 0.0179 0.0179 ETR (degludec/glargine U100) SE 95% CI P value Mean dose (units) Glargine U100 Degludec Insulin dose Basal insulin Baseline N/A 49.0 49.0 12 months 1.03 0.02 [1.00; 1.06] 0.050 65.1 65.1 24 months 1.06 0.02 [1.02; 1.10] 0.007 70.4 74.7 Bolus insulin Baseline N/A 39.4 39.4 12 months 0.95 0.02 [0.91; 1.00] 0.050 59.5 59.5 24 months 0.96 0.04 [0.89; 1.03] 0.204 69.7 69.7 CI conﬁdence interval, ETR estimated treatment ratio, glargine U100 insulin glargine 100 units/mL, MACE major adverse cardiovascular event, N/A not applicable, PYE patient-years of exposure, SE standard error P value refers to two-sided test of degludec/glargine U100 ratio = 1.0 Estimated by applying degludec/glargine U100 ratio if statistically signiﬁcant, otherwise glargine U100 value Previously reported in the DEVOTE primary manuscript supplementary appendix  Model Overview cycles capturing ﬁrst MACE, death from other causes, severe hypoglycemic events, and insulin dose. HbA was not captured in the analysis 1c A decision analytic model (Fig. 1,) was given the short time horizon and the treat-to- developed to evaluate the cost-utility of deglu- target trial design of DEVOTE, resulting in dec versus glargine U100 using clinical data similar end-of-trial HbA levels in both treat- from the basal–bolus subgroup of patients with 1c ment arms . The model used relative rates type 2 diabetes from DEVOTE. The analysis was (hazard, rate, and dose ratios) from regression conducted for the UK setting from the per- analyses to estimate differences between the spective of the National Health Service (NHS). degludec and glargine U100 treatment arms Baseline characteristics for the simulated (Table 1). In line with previous evaluations of patient cohort were based on the DEVOTE degludec, the model only included treatment basal–bolus subgroup (Table S1 in the Electronic effects for which a statistically signiﬁcant dif- supplementary material, ESM). The model was a ference between treatment arms was short-term cohort model with two annual Diabetes Ther (2018) 9:1217–1232 1221 the inter-arm differences in cost and QALYs, and the incremental cost-utility ratio (ICUR; cost per QALY gained). Cost Data and Resource Use Treatment unit costs were based on UK list pri- ces (Table 2,[32–37]). It was assumed that four injections per day were administered for the basal–bolus regimen (same in both arms) and a new needle and SMBG test strip and lancet were used per injection. Mean doses of basal and bolus insulin were estimated for each of the annual cycles and adjusted for survival. Insulin treatment costs were modeled as the unit cost multiplied by the mean annual dose, calculated as the arithmetic mean of the start- and end-of- year doses captured from the clinical trial data to approximate the average under the curve (Table 1b). Costs of complications were derived from the literature and inﬂation-adjusted to 2016 prices using the hospital and community health services index from the Personal Social Services Research Unit  (Table 2). For non- fatal MI and nonfatal stroke, an event cost was captured in the cycle in which the event took place and a state cost was captured in the sub- sequent year, reﬂecting ongoing excess health- Fig. 1 Schematic of the cost-utility model. Diagram shows care resource use after the event (only relevant the model as run for each treatment arm (degludec and for ﬁrst-year events given the 2-year time-hori- glargine U100). The model used relative rates (hazard, rate, zon). Costs of severe hypoglycemia were cap- and dose ratios) from regression analyses to derive the rates tured exclusively in the year of the event, and and doses in the degludec arm (see the ’’Model Overview’’ were a summary of direct costs for patients who section). * Severe hypoglycemia was deﬁned according to were treated by a family member or friend, or the American Diabetes Association as an episode requiring received emergency treatment from a para- the assistance of another person to actively administer medic or general practitioner, or received carbohydrate or glucagon or to take other corrective treatment in a hospital . Cost estimates of actions . CV cardiovascular, glargine U100 insulin severe hypoglycemia included immediate direct glargine 100 units/mL, MACE major adverse cardiovascu- treatment costs (attendance by a healthcare lar event, MI myocardial infarction, U units professional, ambulance callout, and adminis- tration of medications) and follow-up treatment documented, and assumed that all other dif- costs (additional visits to see a general practi- ferences were due to random variation (i.e., the tioner, extra blood glucose test strips, and null hypothesis could not be rejected). How- patient/family education). ever, nonsigniﬁcant differences were explored in the sensitivity analyses. An annual discount Event Rates rate of 3.5% was applied to the costs and clinical beneﬁts in the second cycle . Model outputs Rates from DEVOTE for ﬁrst MACE, death from include the average cost in pounds sterling other causes, and severe hypoglycemic events (GBP), utility outcomes in QALYs for each arm, 1222 Diabetes Ther (2018) 9:1217–1232 Table 2 Input parameters Parameter Value Unit Source Treatment costs Glargine U100 unit price 0.0277 GBP MIMS  Degludec unit price 0.0311 GBP MIMS  IAsp unit price 0.0204 GBP MIMS  Needle unit price 0.0969 GBP MIMS  SMBG test strip unit price 0.3678 GBP MIMS  Complication costs MI, nonfatal 7718.85 GBP Alva et al. 2015  (inﬂation-adjusted) Stroke, nonfatal 8301.28 GBP Alva et al. 2015  (inﬂation-adjusted) CV death 0 GBP Year 2 MI 1918.67 GBP Alva et al. 2015  (inﬂation-adjusted) Year 2 stroke 1977.55 GBP Alva et al. 2015  (inﬂation-adjusted) Severe hypoglycemia 414.09 GBP Hammer et al. 2009  (inﬂation-adjusted) Utilities Base 0.785 Utility Clarke et al. 2002  MI - 0.055 Disutility Clarke et al. 2002  Stroke - 0.164 Disutility Clarke et al. 2002  Severe hypoglycemia - 0.057 Disutility Evans et al. 2013  CV cardiovascular, GBP pounds sterling, IAsp insulin aspart, glargine U100 insulin glargine 100 units/mL, MI myocardial infarction, MIMS Monthly Index of Medical Specialities, SMBG self-monitored blood glucose Lantus (in Solostar pen) £41.50 for 1500 units Degludec (in FlexTouch pen) £46.60 for 1500 units Insulin aspart (in FlexPen ) £30.60 for 1500 units BD MicroFine 5 mm: £9.69 per 100 needles SMBG test costs based on the use of one Accu-Chek Aviva test strip (£15.96 for 50 units) and one Accu-Chek FastClix lancet (9.92 for 204 units) per test Inﬂation-adjusted to 2016 prices using the hospital and community health services index from the Personal Social Services Research Unit  were captured in the model. Adopting a con- rates were applied to the surviving proportion servative approach, subsequent CV events after of the cohort in the two annual cycles (Fig. 1). ﬁrst MACE were not incorporated into the CUA. The model used CV death from ﬁrst MACE as Utility Data well as death from other causes to estimate annual survival, which was applied in each of Baseline utility and disutility values were iden- the two annual cycles. Annual nonfatal MI, tiﬁed in the literature (Table 2). For nonfatal MI nonfatal stroke, and severe hypoglycemic event and nonfatal stroke, the utilities were half- Diabetes Ther (2018) 9:1217–1232 1223 cycle-corrected in the ﬁrst year and applied fully RESULTS in the subsequent year (only relevant for ﬁrst- year events given the 2-year time horizon). The Basal–Bolus Subgroup Analysis severe hypoglycemia disutility was an annual- ized disutility for the year of the event. The The risk of experiencing the primary composite model was developed to optionally capture an outcome of 3-component MACE was 19% lower ‘‘extension period,’’ allowing additional costs with degludec versus glargine U100 (hazard and quality of life estimates over a longer time ratio, 0.81; 95% CI 0.66–0.98; P = 0.035) horizon to be attached to the events that (Table 1a). This hazard ratio was applied to the occurred in the within-trial period, i.e., four annual rate of ﬁrst MACE observed in the glar- scenarios based on a nonfatal MI, nonfatal gine U100 arm of 0.0607 events/patient-year of stroke, no MACE, or death in the trial. Extend- exposure (PYE) to estimate a rate of 0.0489 ing the analysis with additional costs was events/PYE for the degludec arm. The distribu- explored in a sensitivity analysis. tion of components within the composite MACE endpoint for degludec versus glargine Sensitivity Analyses U100 (events/PYE) was nonfatal MI (0.0208 [42%] vs. 0.0292 [48%]), nonfatal stroke (0.0099 Deterministic sensitivity analyses (of the effects of [20%] vs. 0.0124 [20%]), and CV death (0.0182 individual changes in model parameters on clin- [37%] vs. 0.0191 [31%]). There was a 37% lower ical and cost outcomes) were performed based on risk of severe hypoglycemia with degludec ver- no discounting, the inclusion of nonsigniﬁcant sus glargine U100 (estimated rate ratio, 0.63; differences between treatment arms, the use of 95% CI 0.47–0.85; P = 0.002) (Table 1a). This the same MACE distribution for degludec and ratio was applied to the annual rate of severe glargine U100, alternative unit costs and disutili- hypoglycemia observed in the glargine U100 ties, the inclusion of a utility for ﬂexible dosing, arm of 0.085 to result in an estimated annual and the inclusion of long-term additional costs rate of 0.0536 episodes of severe hypoglycemia and effects based on scenarios run in the IQVIA for degludec. There was no signiﬁcant differ- CORE Diabetes Model version 9.0 (IQVIA, Basel, ence in the incidence of death from other cau- Switzerland) [using updated UKPDS 82 CV risk ses (excluding ﬁrst MACE) between treatment equations], assuming the same rescue medication arms (Table 1a). The only signiﬁcant difference in both arms (Fig. S1 in the ESM) and extension in insulin dose between treatment arms was for parameters detailed in Table S2 of the ESM. basal insulin dose at 24 months, which was A probabilistic sensitivity analysis (PSA) was higher in the degludec arm (Table 1b). There performed to quantify the effect of statistical was no signiﬁcant difference in change from uncertainty around the input parameters on the baseline in HbA after 24 months between 1c model results. Uncertainty was captured based treatment arms (Table S4 of the ESM). on standard errors around the DEVOTE out- comes, which were used to inform the shape of Base Case a series of normal and lognormal distributions around model parameters (Table S3 of the ESM). In our evaluation of discounted costs, the mean PSA outcomes were based on 1000 model iter- cost per patient was GBP 4002.36 in the deglu- ations, sampling from all modeled distributions dec arm compared with GBP 4031.13 in the in each iteration without capturing covariance. glargine U100 arm. This resulted in a negligible cost saving of GBP 28.78 per patient with Compliance with Ethics Guidelines degludec over the 2-year time horizon. Savings were driven by a reduction in the risk of dia- This article does not contain any new studies betes-related complications with degludec, par- with human or animal subjects performed by ticularly nonfatal MI, which more than offset any of the authors. the higher acquisition costs compared with 1224 Diabetes Ther (2018) 9:1217–1232 Table 3 Short-term cost-utility outcomes of treatment with degludec versus glargine U100 (base case analysis) Degludec Glargine U100 Difference (Degludec 2 glargine U100) Costs (GBP) Total costs 4002.36 4031.13 - 28.78 Treatment costs Basal insulin 1361.02 1191.06 169.96 Bolus insulin 802.01 801.21 0.80 Basal needles 67.12 67.06 0.06 Bolus needles 201.37 201.18 0.19 Routine SMBG test 1019.17 1018.22 0.95 Costs of complications Nonfatal MI 336.64 470.81 - 134.18 Nonfatal stroke 172.42 214.02 - 41.60 Severe hypoglycemia 42.61 67.58 - 24.96 Effects (QALYs) Total QALYs 1.4778 1.4715 0.0064 QALY breakdown Baseline 1.4888 1.4874 0.0014 Nonfatal MI - 0.0021 - 0.0030 0.0009 Nonfatal stroke - 0.0031 - 0.0038 0.0008 Severe hypoglycemia - 0.0057 - 0.0091 0.0034 Incremental cost-utility ratio ICUR (cost/QALY) Dominant Costs and QALYs are discounted by 3.5% GBP pounds sterling, glargine U100 insulin glargine 100 units/mL, ICUR incremental cost-utility ratio, MI myocardial infarction, QALY quality-adjusted life-year, SMBG self-monitored blood glucose Dominant, improved quality of life at lower or similar cost glargine U100 (Table 3). Total discounted Sensitivity Analyses QALYs were estimated at 1.4778 and 1.4715 for degludec and glargine U100, respectively, with The deterministic sensitivity analyses demon- an increment of 0.0064 in favor of degludec. strated that the base-case ﬁndings remained The main drivers of this were the risk reductions largely unchanged despite variation in a range for severe hypoglycemia and (to a smaller of model input values, with ICURs showing extent) nonfatal MI and nonfatal stroke with dominance or close to cost neutrality in most degludec compared with glargine U100 cases. Using the same MACE distribution (the (Table 3). An ICUR is not shown, per conven- glargine U100 MACE distribution) in both tion, as the result was ‘‘dominant,’’ meaning treatment arms changed the results, lowering that an improvement in health is observed CV mortality at the expense of the nonfatal MI together with cost savings/cost neutrality. reduction, which increased the QALYs gained in Diabetes Ther (2018) 9:1217–1232 1225 the degludec arm but lowered the cost offsets. Previous health economic evaluations from a Changing the treatment costs (a 15% lower NHS perspective have reported similar results, glargine U100 cost) resulted in an ICUR in the with degludec considered dominant or cost-ef- lower range of £20,000–£30,000 per QALY fective relative to glargine U100 in the UK set- gained, although when long-term extension ting across diabetes types and regimens (basal costs and effects were included, degludec was only or basal–bolus) [49–51]. These results were still highly cost-effective (Table 4,[38–45]). The based mainly upon differences in event rates probabilistic sensitivity analyses showed that (and associated costs) of severe and nonsevere degludec is likely to be cost-effective with 96.5 hypoglycemia, whilst CV events and death from and 97.7% probabilities for willingness-to-pay other causes (excluding ﬁrst MACE) were not thresholds of £20,000 and £30,000 per QALY captured by the models [49–51]. With the pre- gained, respectively (Fig. 2). sent analysis using data from a CVOT, CV out- comes and death from other causes were included. However, data on nonsevere hypo- DISCUSSION glycemia, a large driver of QALYs in other reported evaluations, were not collected and In this analysis of the basal–bolus subgroup of therefore could not be included in our evalua- DEVOTE, there was a 19% lower risk of MACE tion. Hypoglycemia exerts signiﬁcant societal and a 37% lower risk of severe hypoglycemia costs through lost productivity related to with degludec versus glargine U100. The results absences from work  that were not reﬂected of the CUA showed that for the basal–bolus by our CUA from a NHS perspective. subgroup of patients in DEVOTE, with a high In the overall trial population of DEVOTE, risk of CV events, degludec was cost neutral (no degludec was noninferior to glargine U100 with additional cost) compared with glargine U100 respect to the incidence of ﬁrst MACE (hazard over a 2-year trial horizon in the UK setting. The ratio, 0.91; 95% CI 0.78–1.06) . Here, we higher acquisition costs with degludec were report that the risk of ﬁrst MACE was signiﬁ- more than offset by the lower costs associated cantly lower with degludec versus glargine with a reduced risk of diabetes-related compli- U100 in the basal–bolus subgroup of DEVOTE. cations, particularly MI, with degludec versus The more pronounced treatment difference for glargine U100. Furthermore, there were QALY MACE in the basal–bolus subgroup of DEVOTE gains with degludec, particularly due to the compared with the full trial population may be reduction in risk of severe hypoglycemia. explained by the more advanced disease in Limiting the risk of hypoglycemia is impor- these patients and thus a higher risk of diabetes- tant for both patients and physicians [46, 47]. related complications such as MACE . This is Severe hypoglycemia is associated with an evidenced by the higher proportion of patients increased frequency of various adverse out- that experienced ﬁrst MACE who started comes in patients with diabetes, including CV DEVOTE on a basal–bolus regimen (9.8% with disease, MACE, dementia (in older patients), degludec vs. 12.0% with glargine) compared major microvascular events, and death [10–14]. with the full trial population (8.5% with Hypoglycemia can have considerable negative degludec vs. 9.3% with glargine U100) . effects on patient quality of life , with an Recently published and currently ongoing increasing frequency and severity of events CVOTs of antihyperglycemic therapies are usu- associated with a reduced quality of life . ally placebo-controlled, noninferiority trials The long-term effects of hypoglycemia may lacking an active comparator . DEVOTE is include behavioral changes and signiﬁcant the ﬁrst published double-blind CVOT that anxiety or fear of future episodes . Fear of compared the drug of interest head-to-head hypoglycemia can adversely affect diabetes with an active comparator—in this case deglu- management and clinical outcomes by com- dec versus glargine U100, both once daily in promising adherence to medications . addition to standard care [21, 22]. This strategy allows a direct comparison of treatment effects, 1226 Diabetes Ther (2018) 9:1217–1232 Table 4 Results of the deterministic sensitivity analyses Parameter Sensitivity analysis DCosts DQALY ICUR (GBP per QALY (GBP) gained) Base case - 28.78 0.0064 Dominant No discounting Discount rate = 0% - 28.85 0.0065 Dominant Insigniﬁcant difference All RRs and HRs - 19.65 0.0126 Dominant applied regardless of P value Same MACE distribution Glargine U100 27.71 0.0102 2705 MACE distribution applied in the degludec arm CV costs  MI Y1 5687 5.66 0.0064 891 MI Y2? 639 Stroke Y1 9567 Stroke 2572 Y2? CV disutilities MI - 0.047 - 28.78 0.0058 Dominant Stroke - 0.060 CV disutilities MIY1 - 0.129 - 28.78 0.0074 Dominant MI Y2?- 0.078 Stroke Y1 - 0.181 Stroke - 0.269 Y2? CV disutilities MI - 0.026 - 28.78 0.0056 Dominant Stroke - 0.099 a,b Cost of CV death  3238 - 34.51 0.0064 Dominant Flex utility  0.006 - 28.78 0.0177 Dominant Flex utility  0.013 - 28.78 0.0310 Dominant Hypoglycemia cost  93.12 - 9.35 0.0064 Dominant Hypoglycemia disutility  0.0118 - 28.78 0.0037 Dominant Glargine U100 price - 15% 149.88 0.0064 23579 Additional long-term costs and Extension - 76.45 0.0344 Dominant effects Diabetes Ther (2018) 9:1217–1232 1227 Table 4 continued Parameter Sensitivity analysis DCosts DQALY ICUR (GBP per QALY (GBP) gained) Combination of above two Extension and 102.21 0.0344 2971 glargine U100 - 15% cost D difference in, CV cardiovascular, GBP pounds sterling, HR hazard ratio, ICUR incremental cost-utility ratio, glargine U100 insulin glargine, MACE major adverse cardiovascular event, MI myocardial infarction, RR risk ratio, QALY quality- adjusted life-year, Y1 year one, Y2? year two onwards Inﬂation-adjusted to 2016 prices using the hospital and community health services index from the Personal Social Services Research Unit  Average cost of fatal MI, fatal stroke, and fatal ischemic heart disease  Flex utility refers to the convenience of ﬂexible dosing times with degludec Adjusted to 1-year time horizon based upon one severe event in the past 3 months causing a 4.7% loss of utility See Table S2 of the ESM for long-term modeling extension parameters and represents an improvement in the design of setting, in this case the UK cost setting, and are CVOTs . To the authors’ knowledge, this is always associated with some uncertainty arising the ﬁrst health economic analysis of a speciﬁc from the model design and input parameters. antihyperglycemic therapy to use data sourced Many of the factors contributing to the uncer- from a double-blind, active-comparator CVOT. tainty have been explored in sensitivity analy- This enabled clinical data for the base-case ses, which supported the results of the base-case analysis for both treatment arms to be derived model and suggest that the ﬁndings are robust. from a single homogeneous source (the A further limitation is the omission of basal–bolus subgroup of the DEVOTE study microvascular complications from the analysis. population) including observed event rates for While microvascular complications can have a MACE, death from other causes, and severe sizeable effect on diabetes-related treatment hypoglycemia. Additional strengths of this CUA costs and patient quality of life, no microvas- include the exclusive use of endpoints with cular endpoints were included in the DEVOTE signiﬁcant differences between treatment arms; trial. The rationale for their exclusion was dri- that the base-case ﬁndings remained largely ven primarily by the desire to maintain the unchanged despite variation in a range of homogeneity of the clinical data [33, 54]. model input values in the sensitivity analyses; Finally, the time horizon in this analysis was and the conservative assumptions of the base- limited by the duration of the trial. It could be case analysis, which did not include long-term perceived as being overly conservative to limit a improvements. CV beneﬁt to only two years. However, a This study also has several limitations. Fore- longer-term (lifetime) perspective was explored most, the evaluation is based on a subgroup of in a sensitivity analysis which assumed rescue patients, which affects the generalizability of medication in both arms and no continued the results beyond patients at high risk of CV beneﬁts after the trial period beyond differences events and treated with basal–bolus regimens. in the modeled history of events in the two Notwithstanding, the internal validity for this model arms. Further analyses would be required group of patients is high. Decision analytic to quantify the potential long-term costs and models, such as the one used in the present beneﬁts based on possible scenarios of the pro- analysis, bring together evidence from various tective effects of degludec on CV risk and severe sources to model cost-utility in a speciﬁc hypoglycemia. 1228 Diabetes Ther (2018) 9:1217–1232 Fig. 2a–b Probabilistic sensitivity analysis results: a cost- expectancy. GBP pounds sterling, glargine U100 insulin utility scatterplot; b cost-utility acceptability curve. In a, glargine 100 units/mL, QALY quality-adjusted life-years the orange square represents the average value for incremental cost and incremental quality-adjusted life CONCLUSION ACKNOWLEDGEMENTS Based on this short-term modeling analysis, We thank the trial investigators, staff, and degludec provides improvements in clinical patients for their participation in the DEVOTE outcomes at no additional cost in the UK set- trial. ting, as compared with glargine U100 in patients with type 2 diabetes at high risk of CV Funding. Sponsorship for the DEVOTE clin- events who are treated with a basal–bolus ical trial, this study, and article processing insulin regimen. charges were funded by Novo Nordisk A/S. All Diabetes Ther (2018) 9:1217–1232 1229 authors had full access to all of the data in this consultation to UNC under contract and travel/ study and take complete responsibility for the meals/lodging for contracted activities from integrity of the data and accuracy of the data Hoffmann-La Roche, Bristol-Myers Squibb, analysis. Liposcience, Elcelyx, Metavention, Dance Bio- pharm Inc, and Quest; grants from Medtronic Medical Writing and Editorial Assis- Minimed, Tolerex, Osiris, Halozyme, Pﬁzer, tance. Medical writing support was provided by Johnson & Johnson, Andromeda, Boehringer Anna Campbell, Ph.D., and editorial assistance Ingelheim, GlaxoSmithKline, Astellas, was provided by Beverly La Ferla, of Water- MacroGenics, Intarcia Therapeutics, and Scion meadow Medical, an Ashﬁeld company, part of NeuroStim; stock options in PhaseBio outside UDG Healthcare plc, and was funded by Novo the submitted work; and is or has been a Nordisk. member of a variety of nonproﬁt boards: American Diabetes Association, DiabetesSisters, Authorship. All named authors meet the Taking Control of Your Diabetes, AstraZeneca International Committee of Medical Journal Healthcare Foundation, Bristol-Myers Squib Editors (ICMJE) criteria for authorship for this Together on Diabetes Foundation, and the manuscript, take responsibility for the integrity National Diabetes Education Program. of the work as a whole, and have given ﬁnal approval to the version to be published. Compliance with Ethics Guidelines. This article does not contain any new studies with Disclosures. Richard F. Pollock is a full-time human or animal subjects performed by any of employee of Ossian Health Economics and the authors. Communications GmbH, which received con- sultancy fees from Novo Nordisk to construct Data Availability. The datasets generated the model and conduct the analyses. William J. during and/or analyzed during the current Valentine is also a full-time employee of Ossian study are available from the corresponding Health Economics and Communications author on reasonable request. GmbH, which received consultancy fees from Novo Nordisk to construct the model and con- Open Access. This article is distributed duct the analyses. Steven P. Marso has received under the terms of the Creative Commons personal fees from Abbott Vascular, Novo Nor- Attribution-NonCommercial 4.0 International disk, University of Oxford, AstraZeneca, and License (http://creativecommons.org/licenses/ Bristol-Myers Squibb, and research support from by-nc/4.0/), which permits any noncommercial Novo Nordisk, The Medicines Company, and use, distribution, and reproduction in any Terumo Medical. Jens Gundgaard is an medium, provided you give appropriate credit employee of Novo Nordisk and also holds to the original author(s) and the source, provide shares/stocks in Novo Nordisk. Nino Halle ´nis a link to the Creative Commons license, and an employee of Novo Nordisk. Lars L. Hansen is indicate if changes were made. an employee of Novo Nordisk. Deniz Tutkunkardas is an employee of Novo Nordisk. John B. Buse reports grant and consultation fees REFERENCES to the University of North Carolina (UNC) under contract and travel/meals/lodging for 1. NCD Risk Factor Collaboration. Worldwide trends contracted activities from Novo Nordisk during in diabetes since 1980: a pooled analysis of 751 the conduct of the study; grants and fees for population-based studies with 4.4 million partici- consultation to UNC under contract and travel/ pants. Lancet. 2016;387:1513–30. meals/lodging for contracted activities from Eli Lilly, GI Dynamics, Amylin, Orexigen, Merck, 2. International Diabetes Federation. IDF diabetes atlas. Brussels: IDF; 2017. Novo Nordisk, Transtech Pharma, AstraZeneca, Takeda, Sanoﬁ, and Lexicon; fees for 1230 Diabetes Ther (2018) 9:1217–1232 3. Kanavos P, Van den Aardweg S, Schurer W. Diabetes 15. Heise T, Nosek L, Bottcher SG, Hastrup H, Haahr H. expenditure, burden of disease and management in Ultra-long-acting insulin degludec has a ﬂat and 5 EU countries. London: London School of Eco- stable glucose-lowering effect in type 2 diabetes. nomics; 2012. Diabetes Obes Metab. 2012;14:944–50. 4. Robinson R. Cost-utility analysis. BMJ. 16. Heise T, Kaplan K, Haahr HL. Day-to-day and 1993;307:859–62. within-day variability in glucose-lowering effect between insulin degludec and insulin glargine (100 5. Holman RR, Paul SK, Bethel MA, Matthews DR, Neil U/mL and 300 U/mL): a comparison across studies. HA. 10-year follow-up of intensive glucose control J Diabetes Sci Technol. 2018;12(2):356–63. in type 2 diabetes. N Engl J Med. 2008;359:1577–89. 17. Heise T, Hermanski L, Nosek L, Feldman A, Ras- 6. Stratton IM, Adler AI, Neil HA, et al. Association of mussen S, Haahr H. Insulin degludec: four times glycaemia with macrovascular and microvascular lower pharmacodynamic variability than insulin complications of type 2 diabetes (UKPDS 35): glargine under steady-state conditions in type 1 prospective observational study. BMJ. diabetes. Diabetes Obes Metab. 2012;14:859–64. 2000;321:405–12. 18. Ratner R, Gough SC, Mathieu C, et al. Hypogly- caemia risk with insulin degludec compared with 7. Holbrook T, Tang Y, Das R, et al. Direct medical costs of severe hypoglycaemic events in patients insulin glargine in type 2 and type 1 diabetes: a pre- with type 2 diabetes in England: a retrospective planned meta-analysis of phase 3 trials. Diabetes database study. Int J Clin Pract. 2017;71:e12958. Obes Metab. 2013;15:175–84. 8. Khunti K, Alsifri S, Aronson R, et al. Rates and 19. Wysham C, Bhargava A, Chaykin L, et al. Effect of predictors of hypoglycaemia in 27 585 people from insulin degludec vs insulin glargine U100 on 24 countries with insulin-treated type 1 and type 2 hypoglycemia in patients with type 2 diabetes: the diabetes: the global HAT study. Diabetes Obes SWITCH 2 randomized clinical trial. JAMA. Metab. 2016;18:907–15. 2017;318:45–56. 9. Leiter LA, Yale JF, Chiasson J-L, Harris SB, Klein- 20. Lane W, Bailey TS, Gerety G, et al. Effect of insulin stiver P, Sauriol L. Assessment of the impact of fear degludec vs insulin glargine U100 on hypoglycemia of hypoglycemic episodes on glycemic and hypo- in patients with type 1 diabetes: the SWITCH 1 glycemic management. Can J Diabetes. randomized clinical trial. JAMA. 2017;318:33–44. 2005;29:186–92. 21. Marso SP, McGuire DK, Zinman B, et al. Design of 10. Goto A, Arah OA, Goto M, Terauchi Y, Noda M. DEVOTE (trial comparing cardiovascular safety of Severe hypoglycaemia and cardiovascular disease: insulin degludec vs insulin glargine in patients with systematic review and meta-analysis with bias type 2 diabetes at high risk of cardiovascular analysis. BMJ. 2013;347:f4533. events)—DEVOTE 1. Am Heart J. 2016;179:175–83. 11. Johnston SS, Conner C, Aagren M, Ruiz K, Bou- 22. Marso SP, McGuire DK, Zinman B, et al. Efﬁcacy chard J. Association between hypoglycaemic events and safety of degludec versus glargine in type 2 and fall-related fractures in Medicare-covered diabetes. N Engl J Med. 2017;377:723–32. patients with type 2 diabetes. Diabetes Obes Metab. 2012;14:634–43. 23. American Diabetes Association. Standards of medi- cal care in diabetes 2017. Diabetes Care. 12. Zoungas S, Patel A, Chalmers J, et al. Severe hypo- 2017;40:S64–74. glycemia and risks of vascular events and death. N Engl J Med. 2010;363:1410–8. 24. Zoungas S, Woodward M, Li Q, et al. Impact of age, age at diagnosis and duration of diabetes on the risk 13. Whitmer RA, Karter AJ, Yaffe K, Quesenberry CP Jr, of macrovascular and microvascular complications Selby JV. Hypoglycemic episodes and risk of and death in type 2 diabetes. Diabetologia. dementia in older patients with type 2 diabetes 2014;57:2465–74. mellitus. JAMA. 2009;301:1565–72. 25. Elliott L, Fidler C, Ditchﬁeld A, Stissing T. Hypo- 14. Bloomﬁeld HE, Greer N, Newman D, et al. Predic- glycemia event rates: a comparison between real- tors and consequences of severe hypoglycemia in world data and randomized controlled trial popu- adults with diabetes—a systematic review of the lations in insulin-treated diabetes. Diabetes Ther. 2016;7:45–60. evidence. VA-ESP Project #09-009. Washington, DC: Department of Veterans Affairs; 2012. 26. Meneghini L, Mersebach H, Kumar S, Svendsen AL, Hermansen K. Comparison of 2 intensiﬁcation Diabetes Ther (2018) 9:1217–1232 1231 regimens with rapid-acting insulin aspart in type 2 uc2016/sources-of-information.pdf. Accessed Octo- diabetes mellitus inadequately controlled by once- ber 2017. daily insulin detemir and oral antidiabetes drugs: the step-wise randomized study. Endocr Pract. 38. Hunt B, Ye Q, Valentine WJ, Ashley D. Evaluating 2011;17:727–36. the long-term cost-effectiveness of daily adminis- tered GLP-1 receptor agonists for the treatment of 27. Barendse S, Singh H, Frier BM, Speight J. The impact type 2 diabetes in the United Kingdom. Diabetes of hypoglycaemia on quality of life and related Ther. 2017;8:129–47. patient-reported outcomes in type 2 diabetes: a narrative review. Diabet Med. 2012;29:293–302. 39. Sullivan PW, Ghushchyan VH. EQ-5D scores for diabetes-related comorbidities. Value Health. 28. Brod M, Christensen T, Thomsen TL, Bushnell DM. 2016;19:1002–8. The impact of non-severe hypoglycemic events on work productivity and diabetes management. Value 40. Palmer AJ, Roze S, Valentine WJ, et al. The CORE Health. 2011;14:665–71. diabetes model: projecting long-term clinical out- comes, costs and cost-effectiveness of interventions 29. Bron M, Marynchenko M, Yang H, Yu AP, Wu EQ. in diabetes mellitus (types 1 and 2) to support Hypoglycemia, treatment discontinuation, and clinical and reimbursement decision-making. Curr costs in patients with type 2 diabetes mellitus on Med Res Opin. 2004;20(Suppl 1):S5–26. oral antidiabetic drugs. Postgrad Med. 2012;124:124–32. 41. Hayes A, Arima H, Woodward M, et al. Changes in quality of life associated with complications of 30. Seaquist ER, Anderson J, Childs B, et al. Hypo- diabetes: results from the ADVANCE study. Value glycemia and diabetes: a report of a workgroup of Health. 2016;19:36–41. the American Diabetes Association and the Endo- crine Society. Diabetes Care. 2013;36:1384–95. 42. Boye KS, Matza LS, Walter KN, Van BK, Palsgrove AC, Tynan A. Utilities and disutilities for attributes 31. National Institute for Health and Care Excellence. of injectable treatments for type 2 diabetes. Eur J NICE process and methods guides. Guide to the Health Econ. 2011;12:219–30. methods of technology appraisal. 2013. Available at: https://www.nice.org.uk/process/pmg9/chapter/ 43. Evans M, Jensen HH, Bogelund M, Gundgaard J, the-reference-case. Accessed November 2017. Chubb B, Khunti K. Flexible insulin dosing improves health-related quality-of-life (HRQoL): a 32. MIMS. MIMS monthly index of medical specialities. time trade-off survey. J Med Econ. 2017. Available at: https://www.mims.co.uk/. 2013;16:1357–65. Accessed October 2017. 44. Waugh N, Cummins E, Royle P, et al. Newer agents 33. Alva ML, Gray A, Mihaylova B, Leal J, Holman RR. for blood glucose control in type 2 diabetes: sys- The impact of diabetes-related complications on tematic review and economic evaluation. Health healthcare costs: new results from the UKPDS Technol Assess. 2010;14:1–248. (UKPDS 84). Diabet Med. 2015;32:459–66. 45. Currie CJ, Morgan CL, Poole CD, Sharplin P, Lam- 34. Hammer M, Lammert M, Mejias SM, Kern W, Frier mert M, McEwan P. Multivariate models of health- BM. Costs of managing severe hypoglycaemia in related utility and the fear of hypoglycaemia in three European countries. J Med Econ. people with diabetes. Curr Med Res Opin. 2009;12:281–90. 2006;22:1523–34. 35. Clarke P, Gray A, Holman R. Estimating utility 46. Peyrot M, Barnett AH, Meneghini LF, Schumm- values for health states of type 2 diabetic patients Draeger PM. Insulin adherence behaviours and using the EQ-5D (UKPDS 62). Med Decis Making. barriers in the multinational global attitudes of 2002;22:340–9. patients and physicians in insulin therapy study. Diabet Med. 2012;29:682–9. 36. Evans M, Khunti K, Mamdani M, et al. Health-re- lated quality of life associated with daytime and 47. Fidler C, Elmelund CT, Gillard S. Hypoglycemia: an nocturnal hypoglycaemic events: a time trade-off overview of fear of hypoglycemia, quality-of-life, survey in ﬁve countries. Health Qual Life Out- and impact on costs. J Med Econ. 2011;14:646–55. comes. 2013;11:90. 48. Nicolucci A, Kovacs Burns K, Holt RI, et al. Diabetes 37. Personal Social Services Research Unit. The hospital Attitudes, Wishes and Needs second study and community health services index. 2016. (DAWN2): cross-national benchmarking of dia- Available at: http://www.pssru.ac.uk/pub/uc/ betes-related psychosocial outcomes for people with diabetes. Diabet Med. 2013;30:767–77. 1232 Diabetes Ther (2018) 9:1217–1232 49. Evans M, Wolden M, Gundgaard J, Chubb B, 52. Brod M, Wolden M, Christensen T, Bushnell DM. Christensen T. Cost-effectiveness of insulin deglu- Understanding the economic burden of nonsevere dec compared with insulin glargine in a basal-bolus nocturnal hypoglycemic events: impact on work regimen in patients with type 1 diabetes mellitus in productivity, disease management, and resource the UK. J Med Econ. 2015;18:56–68. utilization. Value Health. 2013;16:1140–9. 50. Evans M, Chubb B, Gundgaard J. Cost-effectiveness 53. Schnell O, Ryden L, Standl E, Ceriello A. Current of insulin degludec versus insulin glargine in adults perspectives on cardiovascular outcome trials in with type 1 and type 2 diabetes mellitus. Diabetes diabetes. Cardiovasc Diabetol. 2016;15:139. Ther. 2017;8:275–91. 54. Bagust A, Beale S. Modelling EuroQol health-related 51. Evans M, Wolden M, Gundgaard J, Chubb B, utility values for diabetic complications from Christensen T. Cost-effectiveness of insulin deglu- CODE-2 data. Health Econ. 2005;14:217–30. dec compared with insulin glargine for patients with type 2 diabetes treated with basal insulin— from the UK health care cost perspective. Diabetes Obes Metab. 2014;16:366–75.
– Springer Journals
Published: Apr 30, 2018