This 52-week, randomized, controlled, parallel-group, open-label, multinational, treat-to-target, noninferiority trial compared the efficacy and safety of once daily insulin degludec with those of once daily insulin glargine, both administered subcutaneously in combination with metformin, in insulin-naive participants requiring intensification of their therapy for type 2 diabetes inadequately controlled with OADs. Dipeptidyl peptidase-4 (DPP-4) inhibitors could be continued as adjunct therapy, but fewer than 2% of patients (distributed similarly between treatment groups) continued use of a DPP-4 inhibitor throughout the trial.
The trial took place between 1 September 2009 and 17 January 2011 and was conducted in accordance with the Declaration of Helsinki (20
) and the Good Clinical Practice guidelines (21
) and approved by institutional review boards and independent ethics committees before initiation. Signed informed consent was obtained from participants before trial entry. The trial was conducted at 166 sites in 12 countries: Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France, Germany, Norway, Serbia and Montenegro, Spain, and the U.S.
Participants included adults ≥18 years of age diagnosed with type 2 diabetes for ≥6 months, with A1C 7−10% (inclusive) and BMI ≤40 kg/m2, treated with unchanged doses and dosing frequency of OADs (metformin monotherapy or metformin in any combination with insulin secretagogues [sulfonylurea or glinide, DPP-4 inhibitor] or α-glucosidase-inhibitor) for ≥3 months before screening. Participants were excluded if they received thiazolidinediones, exenatide or liraglutide within 3 months of screening or if they had clinically significant cardiovascular, hepatic, renal or oncologic disease; recurrent severe hypoglycemia; hypoglycemia unawareness; or proliferative retinopathy.
Eligible participants were randomized 3:1 to receive once daily degludec (100 U/mL, 3 mL PDS290; Novo Nordisk, Bagsværd, Denmark) or glargine (Lantus, 100 U/mL, 3 mL SoloStar; sanofi-aventis, Paris, France) by means of a centralized, computer–generated, interactive voice and web response system that generated randomization blocks. Randomization (3:1) ensured adequate exposure to degludec in accordance with regulatory guidelines (22
). Investigators and participants were not blinded to treatment. Treatment group assignment was blinded for individuals involved as titration surveillance monitors, internal safety committee members, external committee members responsible for cardiovascular event adjudication, and personnel involved in defining analysis sets until data were locked for statistical analysis. An independent ad hoc group was to be established to maintain blinding if the internal safety committee members requested unmasking; however, this did not occur during the trial.
At randomization (week 0), eligible participants discontinued all OADs, with the exception of metformin and a DPP-4 inhibitor (the latter was continued if country-specific approved labeling allowed combining a DPP-4 inhibitor with insulin), maintaining their pretrial dose and dosing frequency, and were randomized to treatment with degludec or glargine in parallel groups. Insulin degludec was administered once daily, with the main evening meal, and glargine was administered once daily at the same time every day, as chosen by patient and investigator, in accordance with approved labeling. The starting dose for both insulins was 10 units. In the subsequent 52 treatment weeks, each participant’s insulin dose was titrated on the basis of the average of prebreakfast self-measured blood glucose (SMBG) values of 3 consecutive days preceding a visit, ensuring titration toward a predefined prebreakfast plasma glucose (PG) target of 3.9−4.9 mmol/L. Participants measured blood glucose with a glucose meter (Abbott Diabetes Care, Abbott Park, IL), with test strips calibrated to plasma values to obtain PG readings. The frequent visit schedule for first 26 weeks and treat-to-target approach were chosen to ensure optimal titration.
The primary end point was change in A1C from baseline after 52 weeks. Other efficacy assessments included change from baseline in central laboratory–measured fasting PG (FPG), SMBG, A1C <7% responders, and functional health status (assessed by the 36-item short-form health survey version 2.0 [SF-36]).
Safety assessments included adverse events (AEs), hypoglycemic episodes, insulin dose, body weight, injection site reactions, abnormal findings related to physical examination, vital signs, fundoscopy, electrocardiogram (ECG), and laboratory tests (including insulin antibodies). Confirmed hypoglycemic episodes included either episodes confirmed by SMBG corresponding to PG value <3.1 mmol/L or severe episodes requiring assistance (no SMBG confirmation) (4
). Hypoglycemic episodes occurring from 0001 to 0559 h (inclusive) were classified as nocturnal. Treatment-emergent events were described as occurring on or after the first day of exposure to treatment and no later than 7 days after the last day of treatment with insulin degludec or insulin glargine.
After 52 weeks, participants switched to NPH insulin and continued with OADs for a 1-week washout period for accurate assessment of anti-insulin antibody levels by a subtraction radioimmunoassay method (23
). Laboratory analyses were performed by Quintiles Laboratories Europe (West Lothian, U.K.) and Quintiles Laboratories Limited (Marietta, GA). ECG central reading was performed at Quintiles ECG Services (Mumbai, India). Insulin antibodies were analyzed at Celerion Switzerland AG (Fehraltorf, Switzerland).
The trial’s primary objective was to confirm noninferiority of insulin degludec to glargine, as assessed by change in A1C from baseline after 52 weeks, with a noninferiority limit of 0.4% for the treatment difference (22
). For secondary confirmatory end points, type I error rate (false-positive results) was controlled by means of a hierarchical (fixed-sequence) testing procedure (Supplementary Fig. 1
values provided for hypothesis testing outside this procedure were not controlled for multiplicity. Sample size was determined on basis of the primary objective with a t
statistic under the assumption of a one-sided t
test of size 2.5%, a zero mean treatment difference, and a 1.3% SD for A1C. In total, 984 participants were to be randomized for ≥95% power in the per protocol analysis set.
In line with the intention-to-treat principle, statistical analyses of all efficacy end points, hypoglycemia, and body weight included the full analysis set, comprising all randomized participants. Safety end points were evaluated with the safety analysis set, comprising all participants exposed to treatment. Missing values were imputed with the last observation carried forward method. The last observation carried forward approach was selected for the primary analysis on the basis of U.S. Food and Drug Administration guidance (22
), and its robustness was ensured by excluding early withdrawals, as in the per protocol analysis (including only participants treated for ≥12 weeks), and by sensitivity analyses (described in Supplementary Data
Treatment differences in A1C, FPG, SMBG, functional health status, insulin dose (post hoc), and body weight after 52 weeks were analyzed with ANOVA, with treatment, antidiabetic therapy at screening, sex, and region (European Union or North America) as fixed factors and age and baseline value as covariates. The numbers of treatment-emergent confirmed hypoglycemic episodes per patient-year of exposure (PYE) were analyzed with a negative binomial regression model that included treatment, antidiabetic therapy at screening, sex, and region as fixed factors and age as covariate for all reported treatment-emergent episodes. A similar model was used for post hoc analysis of episodes in the maintenance period (weeks 16−52), when stable insulin dose and glycemic control were achieved for most participants. For severe hypoglycemia, the negative binomial model could not be fitted to the sparse data, and a simpler Poisson regression model was used with the same covariates as originally intended. The 9-point SMBG profile was analyzed with a repeated-measures ANOVA model (Supplementary Data