Risk of hypoglycaemia with insulin degludec versus insulin glargine U300 in insulin-treated patients with type 2 diabetes: the randomised, head-to-head CONCLUDE trial

Detailed methods of CONCLUDE have been described previously [22]. Briefly, this was a treat-to-target, randomised, open-label, active comparator-controlled trial that was conducted at 229 sites in 11 countries. The original 58 week trial duration comprised 52 weeks of active treatment with designation of the first 16 weeks as the titration period and the remaining 36 weeks as the maintenance period (hereafter referred to as the ‘variable maintenance period’). In February 2018, a protocol amendment led to the extension of the trial, resulting in a total trial duration of up to 94 weeks with up to 88 weeks of active treatment, including a new maintenance period (hereafter referred to as the ‘maintenance period’) of 36 weeks. A detailed rationale for this amendment has been published previously [22] and the key reasons for the amendment are outlined in ESM Fig. 1. In brief, routine monitoring of blinded data showed an unusual pattern in the reporting of glycaemic variables and hypoglycaemic events. Specifically, the glycaemic data were inconsistent between central-laboratory-measured variables (HbA_1c and fasting plasma glucose [FPG]) and patient-reported fasting self-measured blood glucose (SMBG) values. Data available from SMBG monitoring indicated to the patient that the blood glucose levels were higher than they actually were, potentially increasing the risk of hypoglycaemia as a result of unnecessary insulin up-titration. At the time of the amendment, the number of patient-reported hypoglycaemic events confirmed by blood glucose was low while the number of pseudo-hypoglycaemic events (blood glucose >3.9 mmol/l with symptoms) was high compared with the SWITCH 2 trial (comparing the effect of insulin degludec vs insulin glargine U100 on in individuals with type 2 diabetes) [10]. These observations, seen in general across the entire trial population, were related to the glycaemic data collection system (MyGlucoHealth blood glucose meter and electronic diary). Therefore, because of these safety concerns, the glycaemic data collection system was discontinued during the variable maintenance period. This system was replaced with an Abbott blood glucose meter and paper diary to be used for the remainder of the trial. To accommodate these changes, preserve the scientific integrity of the trial and ensure sufficient data collection for the confirmatory endpoints using the same glycaemic data collection system (Abbott blood glucose meter and paper diary), a new 36 week maintenance period was included in the trial. At the time of the amendment, recruitment had been finalised and all participants on treatment had completed the titration period. The duration of the variable maintenance period was dependent on each participant’s individual randomisation date and/or approval of the amended protocol by health authorities and local ethics committees, if applicable. After implementation of the amended protocol, participants were asked to come in and initiate the maintenance period as soon as the resources were available at the trial site, irrespective of the next planned visit. Thus, all participants were not required to have all visits scheduled between weeks 16 and 52. The trial data remained blinded at the point of discovering the issue with the glycaemic data collection system and the implementation of the protocol amendment. No unplanned interim analysis of the trial data from the titration period was conducted. The primary endpoint (number of severe or blood-glucose-confirmed symptomatic hypoglycaemic events) at the completion of the maintenance period was evaluated utilising the same analysis duration and statistical methods as the original protocol. Changes were implemented to maintain participant safety and protect the scientific integrity of the trial.

CONCLUDE is registered with ClinicalTrials.​gov no. NCT03078478. The trial was conducted in accordance with the Declaration of Helsinki and ICH Good Clinical Practice Guideline [23, 24]. The protocol was approved by independent ethics committees or institutional review boards for each centre; written informed consent was obtained from each participant before any trial-related activities.

Eligible participants included adults aged >18 years with type 2 diabetes with HbA_1c ≤80 mmol/mol (9.5%), BMI ≤45 kg/m² and treated with basal insulin (once or twice daily; NPH insulin, insulin detemir, glargine U100) with or without oral glucose-lowering drugs (OADs) at stable doses (any combination of metformin, dipeptidyl peptidase-4 inhibitor, α-glucosidase inhibitor, thiazolidinedione and sodium–glucose cotransporter 2 inhibitor) for at least 90 days. In addition, participants had to fulfil at least one risk criterion for hypoglycaemia [22]. The main exclusion criteria were treatment with bolus or premixed insulin or with sulfonylureas/glinides within 90 days before the screening visit, severe renal impairment (eGFR <30 ml min⁻¹ [1.73 m]⁻²), or impaired liver function (alanine aminotransferase or aspartate aminotransferase ≥2.5 times the upper limit of normal).

Consenting participants were randomised using a trial-specific, interactive-voice, web-response system. Participants were randomised 1:1 to receive degludec U200 (Novo Nordisk, Bagsvaerd, Denmark; the 100 U/ml and 200 U/ml concentrations of degludec are bioequivalent and interchangeable [25‐27]) or glargine U300 (Sanofi, Paris, France) administered once daily. Within each treatment arm, participants were randomised 1:1 to administer basal insulin either in the morning (from waking to breakfast) or in the evening (from main evening meal to bedtime). The same dosing time was maintained for each participant throughout the trial. When initiating degludec U200, the pre-trial daily basal insulin dose was reduced by 20%, as per the protocol, irrespective of prior insulin type. Glargine U300 was initiated according to its label: unit-to-unit switch for participants on once-daily basal insulin; 20% reduction for those on twice-daily NPH insulin (US patients) or any twice-daily basal insulin (European and Canadian patients). The insulin dose was titrated similarly for both insulins: once-weekly titration was based on the mean of three pre-breakfast SMBG measurements, with a fasting blood glucose target of 4.0–5.0 mmol/l. The insulin dose was adjusted in multiples of 2 U ranging from −4 U to +8 U depending on the mean pre-breakfast SMBG level [22]. The type and dose of pre-trial OADs remained unchanged throughout the trial unless safety reasons required a change.

The primary endpoint was the rate of overall symptomatic hypoglycaemic events (defined as severe [an event requiring third-party assistance [28]] or confirmed blood glucose <3.1 mmol/l [with symptoms]) during the maintenance period. Secondary confirmatory hypoglycaemia endpoints included the rate of nocturnal symptomatic hypoglycaemic events (severe or blood-glucose-confirmed with symptoms, occurring between 00:01 and 05:59 h) and the rate of severe hypoglycaemic events during the maintenance period. Overall symptomatic, nocturnal symptomatic and severe hypoglycaemic events were also assessed during the total treatment period (up to 88 weeks) as secondary endpoints. Other secondary endpoints included change from baseline to end of treatment in HbA_1c level and FPG level, basal insulin dose at the end of treatment, pre-breakfast SMBG level and body weight. The composite endpoints HbA_1c <53 mmol/mol (7.0%) with no overall symptomatic hypoglycaemia and HbA_1c <53 mmol/mol (7.0%) with no nocturnal symptomatic hypoglycaemia were assessed during the maintenance period. The number of adverse events between the two treatment arms was also assessed during the trial period. An independent external event adjudication committee validated the following selected adverse events in a blinded manner: fatal events and severe hypoglycaemia.

The statistical analyses of the primary and secondary endpoints have been described previously [22]. Endpoints related to hypoglycaemia and safety endpoints were summarised using the safety analysis set; efficacy endpoints were summarised using the full analysis set. Statistical superiority testing of the primary and confirmatory secondary endpoints was performed following a hierarchical testing procedure to control the family-wise type I error rate in the strong sense and has been described previously [22]. The sample size was calculated to ensure at least 80% power for the primary endpoint analysis.

A negative binomial model with pre-trial OADs, region, sex and dosing time as fixed effects, age as covariate and logarithm of the exposure time as offset was used to estimate the rate ratio (RR) of hypoglycaemic events during the maintenance and total treatment periods. Participants with no on-treatment data during the maintenance period had values imputed for the maintenance period analyses based on participants discontinuing treatment during the maintenance period. Multiple imputations were performed using standard methods aligned with the analyses and planned to create 1000 complete datasets. The results were then combined using Rubin’s methods [29]. The proportion of participants experiencing hypoglycaemic events was analysed post hoc using a logistic regression model. The model included treatment, pre-trial OADs, region, sex and dosing time as fixed effects and age as a covariate, and logarithm of the exposure time as offset. Change from baseline to end of treatment in HbA_1c levels, FPG levels, SMBG and body weight were analysed post hoc using mixed models for repeated measures (MMRM) with treatment, pre-trial OADs, region, sex and dosing time as fixed effects, and age and baseline HbA_1c/FPG as covariates. Pre-specified sensitivity analyses were also conducted to test the primary and protocol-specified confirmatory secondary hypoglycaemia endpoints without imputed data as well as capping the number of hypoglycaemic events at three. Further post hoc sensitivity analyses controlling for variation across sites were conducted for HbA_1c and FPG.

Of the 2008 eligible participants screened, 1609 were randomised to receive either degludec U200 (n = 805) or glargine U300 (n = 804) (Fig. 1). A total of 1467 participants (91.2%) completed the trial of whom 1409 (87.6%) completed the trial on treatment. The proportion of participants withdrawing from the trial and discontinuing treatment prematurely was similar for both treatment groups. The protocol amendment did not have an apparent impact on participant retention rates (Fig. 1).

The characteristics of the participants at baseline were similar in the two treatment groups (Table 1) and did not differ between the randomised population and those entering the maintenance period (ESM Table 1). The mean age was 62.8 years, the mean duration of diabetes was 15.1 years, and the mean±SD HbA_1c level was 59.2 ± 10.5 mmol/mol (7.6 ± 1.0%). At screening, most participants were using glargine U100 (65.0%) and were treated with metformin (77.5%).

The observed mean HbA_1c values at the end of the titration period were 50.2 mmol/mol (6.8%) with degludec U200 and 50.9 mmol/mol (6.8%) with glargine U300. At the end of treatment, the observed mean HbA_1c was 52.8 mmol/mol (7.0%) and 54.1 mmol/mol (7.1%) for degludec U200 and glargine U300, respectively. An analysis of HbA_1c demonstrated a reduction in the HbA_1c level from baseline to the end of treatment with degludec U200 compared with glargine U300: observed mean −5.90 mmol/mol (−0.54%) vs −5.04 mmol/mol (−0.46%); estimated treatment difference −1.07 mmol/mol (95% CI −1.94, −0.20) (−0.10% [95% CI −0.18, −0.02]), post hoc analysis (Fig. 6a). At the end of treatment, the observed mean FPG was 5.9 mmol/l and 6.5 mmol/l for degludec U200 and glargine U300, respectively. There was also a reduction in FPG from baseline to the end of treatment with degludec U200 compared with glargine U300: observed mean −1.97 mmol/l vs −1.43 mmol/l; estimated treatment difference −0.62 mmol/l (95% CI −0.82, −0.43), post hoc analysis (Fig. 6b). Sensitivity analyses controlling for variation across study sites were conducted for HbA_1c and FPG and showed similar results to the main analysis (ESM Table 3). Over 88 weeks, pre-breakfast SMBG values were similar in the two treatment groups, decreasing during the titration period and then levelling off (estimated treatment difference −0.18 mmol/l [95% CI −0.37, 0.01], post hoc analysis) (Fig. 6c).

At the end of the maintenance period, 35.3% of participants treated with degludec U200 vs 30.0% of participants treated with glargine U300 achieved a composite endpoint of HbA_1c <53 mmol/mol (7.0%) with no overall symptomatic hypoglycaemia (OR 1.31 [95% CI 1.04, 1.65], post hoc analysis). Similarly, 47.4% of participants treated with degludec U200 achieved an HbA_1c <53 mmol/mol (7.0%) with no nocturnal symptomatic hypoglycaemia compared with 39.3% of participants treated with glargine U300 (OR 1.23 [95% CI 0.99, 1.54], post hoc analysis).

The observed mean±SD baseline insulin dose for the degludec U200 and glargine U300 treatment arms was 42.7 ± 29.5 U and 42.2 ± 29.1 U, respectively. At the start of treatment, the observed mean±SD basal insulin dose was 35.1 ± 23.8 U in the degludec U200 group and 42.4 ± 29.2 U in the glargine U300 group. At the end of treatment, the observed mean±SD dose was 66.6 ± 48.5 U for the degludec U200 group and 73.0 ± 48.5 U for the glargine U300 group (Fig. 7).

The number of adverse events per 100 person-years of exposure (PYE) was 367.3 in the degludec U200 group and 365.4 per 100 PYE in the glargine U300 group; the corresponding rate of serious adverse events was 27.3 per 100 PYE vs 25.7 per 100 PYE, respectively (ESM Table 4). The most frequent adverse events (≥5%) reported were nasopharyngitis, upper respiratory tract infection and diarrhoea. Serious adverse events attributed to hypoglycaemia during the trial were relatively infrequent (8 events [0.72 events/100 PYE] for degludec U200 and 21 events [1.89 events/100 PYE] for glargine U300). A total of seven participants (all on treatment) died in the degludec U200 group and nine (six on treatment) died in the glargine U300 group, of which no deaths were attributed to hypoglycaemic events or the trial products.

At the end of treatment, the observed change from baseline in body weight was higher in the degludec U200 group compared with the glargine U300 group (mean±SD: 2.9 ± 5.2 kg vs 1.7 ± 5.8 kg), with an estimated treatment difference of 1.18 kg (95% CI 0.60, 1.75; post hoc analysis).