The need for effective glycemic control in type 1 (T1DM) and type 2 diabetes mellitus (T2DM) to prevent the development of complications is well understood. However, hypoglycemia and the fear of hypoglycemia, as well as weight gain, impair both patients’ and physicians’ willingness to titrate insulin to the doses required to achieve guideline-recommended target levels of glycemia [13]. With the two basal insulin analogs, insulin glargine (glargine) and insulin detemir (detemir), the risk of hypoglycemia is lower than with older human insulin formulations [4]. Nevertheless, there is still room for improvement in the pharmacokinetics and pharmacodynamics of basal insulin preparations, and consequently the search has continued for a basal analog with less variability in pharmacodynamic effect and a longer duration of action than those currently available to fulfil requirements in all patients with once-daily administration.

Insulin degludec (degludec) is a next-generation basal insulin with an ultra-long and stable action profile and lower pharmacodynamic variability than glargine [5, 6]. Degludec has been designed to form long, soluble multi-hexamer chains upon injection into the subcutaneous tissue; insulin monomers gradually dissociate from these [7]. This mechanism of protraction results in a flat and stable pharmacokinetic and pharmacodynamic profile [6]. Degludec has a half-life of approximately 25 h in patients with T2DM [6] and a glucose-lowering effect at steady state in patients with T1DM beyond 42 h [8].

The efficacy of degludec once daily was examined in a large clinical development program, BEGIN, which included nine 26- or 52-week trials. Three trials were versus glargine in basal–bolus therapy in T1DM [9, 10] and T2DM [11], and four trials were versus glargine in basal–oral therapy in T2DM [1215]. In addition, degludec was compared with sitagliptin in a basal–oral trial in T2DM [16], and with detemir in a basal–bolus trial in T1DM. In all seven trials in which it was compared with glargine, degludec showed non-inferiority with respect to mean decrease in glycosylated hemoglobin (HbA1c). Rates of confirmed hypoglycemia and, in particular, nocturnal confirmed hypoglycemia, were either similar with the two insulins or significantly lower with degludec. A pre-planned meta-analysis of hypoglycemia associated with the two treatments has already been published [17]. This meta-analysis confirmed that degludec is associated with a lower risk of hypoglycemia, in particular nocturnal confirmed hypoglycemia, compared with glargine. The reductions in hypoglycemia were even more marked in the maintenance period (i.e., after 16 weeks, when the initial up-titration was completed). Nocturnal hypoglycemia is a particularly useful outcome for reflecting differences between basal insulins, as it is less likely than daytime hypoglycemia to be confounded by the effect of bolus insulin, meals, and activity.

Previously, confirmed hypoglycemia included both severe and non-severe events. To avoid double counting of events in health economic models, the new meta-analyses reported here regrouped the confirmed events into three mutually exclusive groups: non-severe nocturnal, non-severe daytime, and severe hypoglycemia. The division of non-severe events into daytime and nocturnal was included to provide evidence on whether the documented reduction in nocturnal events with degludec was offset by a change in daytime events.

Individual studies also demonstrate a significantly or numerically lower insulin dose with degludec compared with glargine in T1DM [9, 10] and in T2DM [1215]. In the new meta-analyses reported here, we investigated whether these findings could be confirmed.

Hypoglycemia and HbA1c are the two endpoints generally considered of most interest when evaluating diabetes therapy. However, other endpoints such as fasting plasma glucose (FPG), insulin dose, and health-related quality of life (HRQoL) should also be considered to obtain a global view of the effectiveness of one therapy compared with another. Individual trials (in any field of medicine) are frequently under-powered to show significant differences in secondary endpoints, and meta-analysis is often required to reveal whether such differences exist. The BEGIN program was carefully designed with consistent definitions of outcomes across all trials, thus facilitating subsequent meta-analysis of the data.

This paper reports the results of a comprehensive set of patient-level meta-analyses that were performed to compare degludec and glargine with regard to HbA1c (the primary endpoint in the trials), and hypoglycemia, FPG, and dose (secondary endpoints). The objective was to obtain a comprehensive overview of all relevant differences between degludec and glargine, adding to evidence on hypoglycemia [17] and HRQoL [18, 19].


Trials Included in the Meta-Analysis

Table 1 lists all the trials in the BEGIN program that compared degludec once daily with glargine once daily and explains how they were categorized for the current meta-analysis. Some of the trials included in this analysis were randomized in a 2:1 or 3:1 manner. This unequal randomization is consequently reflected in the meta-analysis with more patients allocated to degludec than to glargine. Trials of degludec against insulin detemir and sitagliptin were not included in the meta-analysis as the purpose was to compare degludec against glargine.

Table 1 Phase 3a trials comparing insulin degludec once daily with insulin glargine once daily in the BEGIN program: categorization for the meta-analysis [915]

The category T1DMB/B (T1DM treated with basal–bolus therapy) included all patients from the BEGIN BB T1 Long (NCT00982228) and BEGIN Flex T1 (NCT01079234) trials [9, 10], except for patients in one degludec dosing arm in the BEGIN Flex T1 trial. Patients in the excluded arm received degludec at extreme ranges of daily dosing intervals, alternating dose time in the morning and evening; hence, inclusion of these data in the meta-analyses would be potentially confounding. The category T2DMinsulin-naïve included patients, all of whom were insulin-naïve, starting on basal–oral therapy in the BEGIN Once Long (NCT00982644), BEGIN Once Asia (NCT01059799) and BEGIN Low Volume (NCT01068665) trials (U200) [12, 14, 15]. The BEGIN Flex trial (NCT01006291) in type 2 diabetes [13] was excluded from the meta-analysis as it included degludec used at extreme daily dosing intervals and also included patients already on insulin. One trial, BEGIN BB (NCT00972283) [11], compared degludec with glargine in basal–bolus therapy in T2DM (T2DMB/B).

Methodology of the Individual Phase 3a Trials

All of the trials included in the current meta-analyses were randomized, treat-to-target, parallel-group, open-label, non-inferiority trials comparing degludec once daily with glargine once daily. All procedures followed in the trials were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000 and 2008. Informed consent was obtained from all patients for being included in the studies. The analyses in this article do not involve any new studies of human or animal subjects performed by any of the authors.

The treat-to-target design was used, targeted at self-measured blood glucose <5 mmol/L, based on a mean of three consecutive days’ measurements. Treating to a common target, recommended by the FDA [20], allows for interpretation of any between-treatment differences in the frequency and severity of hypoglycemia without being confounded by differences in glycemic control. The primary endpoint was the difference between treatment groups in the change in HbA1c from baseline to study end. Non-inferiority was defined as an upper limit of ≤0.4% points for the 95% confidence interval for the treatment difference.

In all the trials (and for the purposes of meta-analysis), FPG measurements were performed in a central laboratory; hypoglycemia was self-reported; only confirmed hypoglycemic events (plasma glucose <3.1 mmol/L or severe episodes requiring assistance) were included in the analyses; and nocturnal hypoglycemia was defined as episodes with onset from 00:01 to 05:59 am, inclusive.

Statistical Analysis

Meta-analyses were performed on patient-level data. The analyses examined the following endpoints: HbA1c, FPG, insulin dose, and hypoglycemia. A linear model was used to analyze HbA1c and FPG. For insulin dose, the endpoint was log-transformed and analysis of variance (ANOVA) was used. Hypoglycemia was analyzed using a negative binomial regression model. Due to the low number of severe hypoglycemic episodes in T2DMinsulin-naïve, this analysis was performed with a Poisson regression model using a log-link. The a priori level of significance was set as 0.05 and the statistical package used was SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA). All the meta-analysis regression models were adjusted for trial, type of diabetes, antidiabetic therapy at screening, sex, region, and age. Except for hypoglycemia, all the models included baseline value as a covariate.


Glycemic Control

As expected in treat-to-target trials, there was no statistical difference between treatments in reducing HbA1c: degludec was non-inferior to glargine in T1DMB/B, T2DMinsulin-naïve, and T2DMB/B (Table 2). These results confirmed the non-inferiority documented in each individual trial.

Table 2 Meta-analysis comparing insulin degludec once daily with insulin glargine once daily: HbA1c and FPG [912, 14, 15]

End-of-trial reduction in FPG was significantly greater with degludec than with glargine in T1DMB/B and T2DMinsulin-naïve (P < 0.05), and numerically (but not significantly) greater in T2DMB/B (Table 2).

Insulin Dose

In T1DMB/B, the total daily dose of insulin was significantly lower, by 12%, with degludec compared with glargine (P < 0.0001) (Table 3). Statistical analyses were performed for both basal and bolus insulin doses to clarify the relative contribution of each to the observed reduction in total dose. These showed that the daily basal and bolus doses were both lower with degludec, with relative rates (degludec versus glargine) as follows: daily basal dose, 0.87; daily bolus dose, 0.88 (both P < 0.05).

Table 3 Total daily insulin dose [basal + bolus (if relevant)] in U/kg (adjusted for covariatesa)

The total daily insulin dose was also significantly lower (by 10%, P = 0.0004) with degludec in T2DMinsulin-naïve. In T2DMB/B, total daily insulin dose did not differ statistically between treatments (Table 3).

Hypoglycemia Analyzed in Mutually Exclusive Groups

The actual event rates for the mutually exclusive groups used in the current meta-analysis are shown in Table 4. Event rates for the individual trials are available in the earlier papers [11, 17]. The estimated hypoglycemia rate ratios for the current meta-analysis are shown in Table 5. These results showed significantly lower rates of nocturnal non-severe hypoglycemia with degludec in T1DMB/B, T2DMinsulin-naïve, and T2DMB/B (by 17, 36, and 25%, respectively; all P < 0.05). With degludec, rates of daytime non-severe hypoglycemia were significantly lower than with glargine (by 17%; P < 0.05) in T2DMB/B, with no statistical differences in T1DMB/B and T2DMinsulin-naïve. Rate ratios for severe events, analyzed in the previous meta-analysis (and not re-analyzed, as the definition remained unchanged) [17], were 1.12 [not significant (ns)] (T1DMB/B); 0.14 (P < 0.05) (T2DMinsulin-naïve); and not analyzed (T2DMB/B) due to too few events. Rate ratios for daytime and nocturnal non-severe events in the full trial and maintenance (week 16 onwards) periods are shown in Table 5.

Table 4 Observed daytime and nocturnal non-severe hypoglycemic events in the current meta-analysis [912, 14, 15]
Table 5 Hypoglycemia rate ratios in the current meta-analysis during the full trial period, and in the maintenance period


This meta-analysis showed that across subgroups of patients with diabetes, those treated with degludec achieve similar or significantly better results than those treated with glargine in terms of FPG and rates of hypoglycemia, with similar reductions in HbA1c. These results are achieved with lower mean total insulin doses.

Non-inferiority of degludec with respect to HbA1c lowering was confirmed (Table 2). Non-inferiority is expected for treat-to-target trials and was indeed seen in each trial. Nevertheless, it is useful to confirm that, across different subgroups, based on patient-level data, the lower rates of hypoglycemia seen with degludec are not obtained at a cost of inferior glycemic control.

The individual trials had shown numerically [9, 11, 14] or significantly greater reductions in FPG [12, 15] with degludec. The current meta-analysis showed significantly greater reductions in FPG at trial end with degludec in both T1DMB/B and T2DMinsulin-naïve (0.61 and 0.34 mmol/L, respectively) (Table 2).

In the current meta-analysis of hypoglycemia analyzed in mutually exclusive groups (Table 4), rates of confirmed hypoglycemia in T2DM were low and were generally in line with rates observed in other trials. For example, among insulin-naïve patients, the non-severe confirmed hypoglycemia rates with glargine were 2.05 and 3.0 events per patient-year, respectively, in the current meta-analysis (Table 4) and in the treat-to-target trial (which used a similar definition, with confirmed events of ≤3.1 mmol/L) [21]. Corresponding nocturnal events were 0.51 and 1.3 per patient-year, respectively.

It is striking that despite the low hypoglycemic event rates observed in trials, in the current meta-analysis, rates of nocturnal non-severe hypoglycemia were significantly lower with degludec than with glargine in all subgroups (Table 5). Rates of daytime non-severe hypoglycemia were numerically lower in T2DMinsulin-naïve and significantly lower in T2DMB/B with degludec (Table 5). Thus, the lower rates of nocturnal non-severe events observed with degludec do not occur at a cost of higher daytime rates in T2DM. In T1DMB/B, daytime non-severe rates were numerically but not statistically higher with degludec (Table 5). These results may have been confounded by the unequal dose adjustment in the T1DMB/B trials for patients randomized from twice-daily (BID) insulin. For patients previously on BID insulin who were randomized to glargine, a 20–30% reduction in starting dose was recommended, whereas patients randomized to degludec maintained the same dose. This possibility is further supported by the fact that during the predefined maintenance phase (>16 weeks), there was no increase in the risk of daytime non-severe hypoglycemia with degludec.

Furthermore, these lower rates of nocturnal non-severe hypoglycemia with degludec were observed together with significantly or numerically greater reductions in FPG values. Lower FPG values would normally be expected to be accompanied by higher rates of nocturnal hypoglycemia, but with degludec, the reverse was observed. This can most likely be attributed to the stable and consistent profile of degludec, with its long duration of action and lower day-to-day pharmacodynamic variability compared with glargine [5, 6].

Hypoglycemia and fear of hypoglycemia remain barriers to achieving target levels of control [3]. In addition, hypoglycemia has negative health-economic consequences such as additional contacts with healthcare professionals and absence from work following an event, and these are particularly marked for nocturnal hypoglycemia [22]. The possibility of achieving lower FPG together with lower risk of nocturnal hypoglycemia with degludec is, therefore, valuable both to patients and in terms of overall costs.

The value to patients of the lower risk of hypoglycemia can be expressed in terms of numbers needed to treat, as reported previously [17]. In T2DMinsulin-naïve, for every 100 people treated with degludec instead of glargine for 1 year, 50 confirmed hypoglycemic episodes (of which 20 are nocturnal) and two severe episodes will be avoided. In T2DMB/B, for every 100 people treated with degludec instead of glargine for 1 year, 326 confirmed hypoglycemic episodes (of which 71 are nocturnal) will be avoided. In T1DMB/B, for every 100 subjects treated with degludec instead of glargine for 1 year, 130 nocturnal confirmed episodes will be avoided once the initial titration phase has been completed.

These results are unlikely to have arisen through bias. The trials in the BEGIN program were open-label because blinding of treatment would have been extremely difficult, given the different insulin-delivery devices used. This open-label design could have given rise to a reporting bias in the patient-reported outcomes of hypoglycemia, which would constitute a limitation of the current study. There is no reason to believe that such a bias was present, and the requirement for patients to report only confirmed episodes of hypoglycemia should have limited any subjective influences. Furthermore, the same titration algorithm was used consistently across all trials for both basal insulins to ensure that differences in titration would not confound results.

Glargine was given according to its product labeling (i.e., administered at any time of day as advised by the investigator, at the same time each day), whereas degludec was administered once daily with the main evening meal, except in BEGIN Once Asia, where it could be given from the start of the evening meal to bedtime. However, any effect of possible different timing of administration of the two insulins is unlikely to change the conclusions of the meta-analysis. If glargine had systematically been given earlier in the day than degludec, nocturnal hypoglycemia would have been expected to be lower with glargine. If, on the other hand, it had been given later than degludec, any increase in nocturnal hypoglycemia should have been accompanied by greater decreases in FPG with glargine.

End-of-trial total daily insulin doses were significantly lower with degludec versus glargine in T1DMB/B and T2DMinsulin-naïve (Table 3). This observation of lower doses with degludec confirms the findings from individual trials, with significantly lower end-of-trial mean total insulin doses as follows: in T1DMB/B, 11% lower in both the BEGIN BB T1 Long trial (P < 0.001) [9] and in the BEGIN Flex T1 trial (statistical significance not reported) [10]; and in T2DMinsulin-naïve, 20% lower in BEGIN Once Asia (P = 0.0004) [14] and 11% lower in BEGIN Low Volume (P < 0.05) [12]. In the third T2DMinsulin-naïve trial, BEGIN Once Long [15], mean end-of-trial doses were not statistically different for degludec and glargine.

With the exception of BEGIN Flex T1, HRQoL was measured in all of the trials included in this meta-analysis using the SF-36 questionnaire [23], which patients completed themselves. As HRQoL results have already been published, HRQoL was not included in the current meta-analysis. For T1DMB/B, there were no significant between-treatment differences in change from baseline in HRQoL in the BEGIN BB T1 Long trial [9]. For T2DMinsulin-naïve, a meta-analysis examining HRQoL changes—which included BEGIN Once Long, BEGIN Once Asia and BEGIN Low Volume—reported that, at endpoint, the overall physical health component score was significantly better with degludec versus glargine, due primarily to a better score for degludec in the bodily pain domain [19]. In the mental domains, the vitality score was significantly better with degludec. For T2DMB/B, HRQoL scores were significantly better with degludec for the domain of bodily pain [11]. A further meta-analysis evaluated HRQoL expressed in terms of health utility score (a value for estimating quality of life) across all six BEGIN trials in which HRQoL was measured [18]. Degludec was associated with a modest but statistically significant improvement in health utility compared with glargine.


Compared with glargine, degludec is associated with equivalent HbA1c control and significantly lower nocturnal hypoglycemia rates. In patients with T1DM and insulin-naïve patients with T2DM, degludec is also associated with significantly greater reductions in FPG versus glargine. Patients with T1DM and insulin-naïve patients with T2DM also required lower total doses of insulin and all groups reported improvements in HRQoL. It is possible that these advantages—in particular, efficacious lowering of FPG values together with lower rates of nocturnal hypoglycemia—could encourage physicians and patients to titrate insulin regimens more rigorously to reach glycemic target values.