Comparative effectiveness and safety of glargine 300 U/mL versus degludec 100 U/mL in insulin-naïve patients with type 2 diabetes. A multicenter retrospective real-world study (RESTORE-2 NAIVE STUDY)

Aims This study assessed comparative effectiveness of glargine 300 U/mL (Gla-300) versus degludec 100 U/mL (Deg-100) in insulin-naïve patients with T2D. Methods This is a retrospective, multicenter, non-inferiority study based on electronic medical records. All patients initiating Gla-300 or Deg-100 were 1:1 propensity score-matched (PSM). Linear mixed models were used to assess the changes in continuous endpoints. Incidence rates (IR) of hypoglycemia were compared using Poisson’s regression models. Results Nineteen centers provided data on 357 patients in each PSM cohort. HbA1c after 6 months (primary endpoint) decreased by − 1.70% (95%CI − 1.90; − 1.50) in Gla-300 group and − 169% (95%CI − 1.89; − 1.49) in Deg-100 group, confirming non-inferiority of Gla-300 versus Deg-100. Fasting blood glucose (BG) decreased by ~60 mg/dl in both groups; body weight remained unchanged. In both groups, the mean starting dose was 12U (0.15U/kg) and it was slightly titrated to 16U (0.20U/kg). IR (episodes per patient-months) of BG ≤70 mg/dl was 0.13 in Gla-300 group and 0.14 in Deg-100 group (p=0.87). IR of BG <54 mg/dL was 0.02 in both groups (p=0.49). No severe hypoglycemia occurred. Conclusion Initiating Gla-300 or Deg-100 was associated with similar improvements in glycemic control, no weight gain and low hypoglycemia rates, without severe episodes during 6 months of treatment.


Introduction
Improving glycemic control in type 2 diabetes (T2D) is a major goal of care, to reduce the incidence of micro-and macro-vascular complications [1,2], but it is also a major challenge for patients, physicians and healthcare systems. In fact, inertia in intensifying diabetes therapy and especially in initiating basal insulin in patients with poor metabolic control is a well-recognized problem [3]. Many barriers exist and refer to clinicians, patients and settings; fear of hypoglycemia represents the main limitation to a prompt intensification [4,5].
Second-generation basal insulin analogues (2BI) represent an opportunity to overcome these barriers. According to randomized controlled trials (RCTs), 2BI are non-inferior to first-generation basal insulins (1BI) with regard to the reduction of HbA 1c but safer in terms of hypoglycemia and with lower variability [6].
The randomized, head-to-head, parallel group BRIGHT study [19] involving insulin-naïve patients with T2D demonstrated that Gla-300 and Deg-100 provided similar glycemic control improvements with relatively low hypoglycemia risk. Hypoglycemia rate was comparable with the two basal insulins during the full study period, but lower with Gla-300 during the dose titration period. The CONCLUDE trial, involving people with T2D switching from 1BIs to Gla-300 or the new 200 U/mL formulation of insulin degludec, documented a similar rate of overall symptomatic hypoglycemia during the 36-week maintenance period [20].
Real-world evidence (RWE) is needed to assess effectiveness and safety of 2BI when prescribed in different settings [21], as the three real-world studies currently available [22][23][24], all conducted in US settings, provided controversial results. Therefore, real-world data on 2BI in patients with T2D are relatively scant, and most importantly, data from European countries are missing.
Given these premises, the RESTORE-2 study aimed at assessing the comparative effectiveness and safety of Gla-300 versus Deg-100 in a cohort of insulin-naïve patients with T2D followed under routine care in Italian diabetes outpatient clinics.

Study design and patients
The RESTORE-2 study was a retrospective, comparative, multicenter study. Inclusion criteria were: male or female gender, age ≥18 years, diagnosis of T2D (any disease duration), initiation of Gla-300 or Deg-100 from January 2017 to January 2020 and no previous treatment with basal insulin (naïve cohort) as recorded in the electronic medical records (EMRs). Exclusion criteria were: diagnosis of T1D, more than one type of basal insulin prescribed at index date or prescription of another basal insulin analogue in the six months after initiating insulin Gla-300 or Deg-100.
Anonymous patient data were derived from the same electronic chart system adopted at all participating centers (Smart Digital Clinic software-property of METEDA s.r.l.) Centers recorded data on EMRs according to their clinical practice, taking into consideration that patients with T2D who need intensification are generally seen by the diabetologist on a 3-6 months basis. The date of the first prescription of the 2BI was considered as the index date (T0, baseline). All data relative to the period before and after the index date (± 6 months) for each patient were extracted and analyzed. Data recorded in the 6 months previous T0 were used to identify the baseline characteristics of patients, whereas data collected after 6 months (T6) represented the follow-up data. When more values of the same parameters were available during the before and after T0 periods, those recorded in the nearest date to T0 and T6 were considered. Data relative to the efficacy endpoints were considered only if values were recorded at T0±30 days and T6±30 days. The following characteristics were considered to describe the baseline patient profile: age, gender, diabetes duration, HbA1c, fasting blood glucose (FBG), weight/body mass index (BMI), total, basal and short-acting insulin dose, number of insulin injections, glucose-lowering drugs other than insulin, blood pressure, lipid profile, diabetes complications (low glomerular filtration rate, albuminuria, cardiovascular complications-i.e., myocardial infarction, coronary revascularization, coronary artery bypass, stroke, lower limb complications, peripheral artery disease-by ICD-9 CM codes). Efficacy endpoints were: changes at 6 months (T6) in HbA 1c , insulin doses, FBG and body weight (continuous endpoints); and frequency and proportion of patients with HbA1c <7% and HbA1c <8% at T0 and T6 (categorical endpoints). Changes in HbA 1c at 6 months from the Gla-300 or Deg-100 initiation represented the primary endpoint.
Safety endpoints were: episodes of hypoglycemia ≤70 mg/dl or <54 mg/dl from self-monitoring blood glucose tests (SMBG) downloaded in EMR (cutoffs recommended by ADA guidelines 2017) during 6 months, severe hypoglycemia (defined as "need of assistance by a third party" and reported in a dedicated module of the EMR) during 6 months.
In addition, the change in HbA 1c at 12 months was evaluated as a post hoc analysis in the subgroup of patients with available data.

Statistical analysis
Sample size estimation was based on the primary endpoint, represented by the change in HbA1c levels after 6 months from the initiation of Gla-300 or Deg-100. We calculated that 296 patients per group were needed in order to achieve 80% power to detect non-inferiority using a one-sided, twosample t-test. The margin of non-inferiority was set at 0.3%, which is generally considered a clinically meaningful difference. The true difference between the means was assumed to be 0.0. The significance level (alpha) of the test was 0.025. The standard deviation of HbA1c was assumed to be 1.3 in both groups.
Baseline patient characteristics according to the initiation of Gla-300 or Deg-100 were compared using the unpaired t-test or the Mann-Whitney U-test in case of normal and skewed continuous variables, respectively, and the Chisquare test or the Fisher exact test for categorical variables, as appropriate.
To allow for an unbiased comparison between patients initiating Gla-300 vs. Deg-100, a propensity score (PS) matching algorithm on a one-to-one basis was applied. To compute PS, we performed a logistic regression model taking into consideration age, gender, diabetes duration, baseline HbA1c, FBG, BMI, basal insulin dose, glucoselowering drugs other than insulin as covariates. Variables included in the logistic model were those showing a statistically significant between-group difference at baseline [25]. A five-to-one greedy matching algorithm was used to identify a unique matched control in the initial Deg-100 group for each Gla-300 patient according to the individual PS. Adequacy of balance for the covariates in the matched sample was assessed via standardized mean difference between the 2 groups, considering differences less than 10% (absolute value) as indicative of a good balance.
PS matching was performed separately in the efficacy population and in the safety population.
Changes in HbA1c, FBG, body weight and insulin dose were assessed using mixed models for repeated measurements. Results are expressed as estimated mean or estimated mean difference from T0 with their 95% confidence interval (95% CI). Paired and unpaired t-test derived from linear mixed models for repeated measurements were applied for within-group and between-group comparisons, respectively.
As secondary outcomes, the proportions of patients with HbA1c <7.0% and HbA1c <8.0% at T0 and T6 were evaluated. Both within-group (McNemar test for change vs. baseline) and between-group (Chi-square test) statistical comparisons were applied.
Incidence rates (IR) of hypoglycemic events were calculated and expressed as numbers of events per patient-month with their 95% CI. Incidence of hypoglycemic events was compared between groups using Poisson's regression model with correction for overdispersion.
The main analysis was conducted on the intention-to-treat (ITT) population, including all patients meeting eligibility criteria. The post hoc population included the subgroup of PS matched patients with a HbA1c value available at baseline and after 12 months.
For the evaluation of severe hypoglycemia, the safety population was represented by the ITT post-PS matching population (data derived from EMRs). For the evaluation of glycemic values ≤70 mg/dl and <54 mg/dl, a subsample of the safety population (PS matched patients having at least 1 SMBG value available) was considered.

Patient disposition and characteristics
Data were extracted from 19 centers ( Fig. 1), yielding initial information from 1,166 patients (808 initiating Gla-300 and 358 initiating Deg-100), who were eligible insulin-naïve patients with first prescription of either 2BI as recorded in EMRs from January 2017 to January 2020. Centers with SMBG tests downloaded in the EMRs were 14 out of 19 (73.7%).
After PS matching, 357 subjects were included in each group (post-PSM ITT). The standardized differences of the 1 3 variables before and after the PS matching are shown in Appendix 1. All selected variables had an absolute standardized mean difference >10% before matching and <10% after matching. The complete baseline characteristics of the ITT matched population are reported in Table 2.
Results of longitudinal models relative to secondary continuous endpoints are shown in Table 2.
At baseline, mean levels of FBG were 212.6 mg/dl in Gla-300 group and 201.6 mg/dl in Deg-100 group. In both groups, statistically significant reductions from baseline to 6 months were shown: − 63.23 mg/dl in Gla-300 group and − 61.14 mg/dl in Deg-100 group, with no between-group differences (p=0.74) ( Table 2). Not significant changes in body weight were documented in both groups after 6 months ( Table 2).
Mean basal insulin dose was titrated during 6 months and statistically significant within-group increases were observed at T6 in both groups ( Table 2). In the Gla-300 group, the estimated mean starting dose (T0) was 11.79 U and increased on average by +4.45 U at T6; in Deg-100 group, the estimated mean starting dose (T0) was 12.45 U and increased by +3.54 U at T6. In both groups, perkg basal insulin dose significantly increased during the follow-up; at T6, the dose was of 0.20 U/kg in both groups. No between-group differences were found in insulin dose changes over time.
At the post hoc analysis, in the subgroup of patients with a HbA 1c value at 12 months (121 in Gla-300 group and 133 in Deg-100 group), a marked reduction in HbA 1c mean levels was maintained in both groups; the reduction was − 1.71% in Gla-300 versus − 1.44% in Deg-100 group, although the between-group difference did not reach the statistical significance (p=0.052) (Appendix 5).

Comparative safety analysis
The safety population (i.e., ITT patients with at least 1 SMBG downloaded in the EMRs) was PS matched for the following unbalanced variables at T0: diabetes duration, HbA 1c , number of glucose-lowering drugs other than insulin (<2 or >=2), use of metformin and secretagogues. Each PS matched group included 123 subjects (Appendix 3 and 4). No difference in the incidence of episodes of BG≤70 mg/ dl or <54 mg/dl was present before insulin initiation in the two groups (Appendix 6). Overall, 18,353 SMBG tests were available in Gla-300 group, and 19,621 SMBG tests were available for Deg-100 group. The incidence of BG events ≤70 mg/dl and <54 mg/ dl during 6-month follow-up was very low and similar in Gla-300 group and in Deg-100 group. No between-group differences emerged (Table 3).
No severe hypoglycemic episodes (evaluated in the safety population) were reported on EMRs in both groups.

Conclusions
In this real-world study, comparative effectiveness analyses showed that initiating Gla-300 or Deg-100 in uncontrolled insulin-naïve patients with T2D was followed by statistically significant and clinically relevant HbA 1c reductions (− 1.70%) after 6 months, without significant differences between groups. In addition, after 6 months, in both groups the proportion of patients with HbA1c <7.0% increased from a small minority to almost one-third, while about 70% achieved levels <8%. Given the very high HbA 1c and FBG levels at initiation of 2BI, this result can be considered clinically meaningful.
In the subgroup of patients with HbA 1c values available at 12 months, HbA 1c reduction was sustained in the two groups and numerically greater with Gla-300.
After 6 months, FBG was reduced by about 60 mg/dl in both groups and only minor changes in body weight were recorded. Insulin dose up titration was modest (+4 U/day) but statistically significant; during the first 6 months of treatment, 0.2 U/kg of basal insulin was used in both groups.
Concerning safety, we found a similar profile of the two 2BIs. Indeed, incidence of hypoglycemic episodes ≤70 mg/ dl and <54 mg/dl during 6 months was very low and similar in the two groups, and no severe hypoglycemic episodes were recorded on EMRs.
This study adds important insights into the understanding of clinical profile of T2D patients initiating Gla-300 or Deg-100 in the real world. Before PSM, socio-demographic and clinical characteristics of patients initiating Gla-300 or Deg-100 were similar. Only small between-group differences emerged in the concomitant use of some classes of diabetes therapies (secretagogues, glitazones and SGLT2i); these differences are likely to reflect the evolving scenario of T2D pharmacotherapy when the two 2BI were made available in Italy.
Our results underline the effectiveness and safety of both 2BI. Furthermore, the very high baseline HbA 1c (>9.0%) and FBG (>200 mg/dl) levels at the time of insulin initiation highlight once more the well-known problem of clinical inertia [3], suggesting a late treatment intensification. After 6 months, despite relevant improvements, many patients were still above the recommended HbA 1c and FBG targets in both groups. Low doses of basal insulin were adopted at 6-month follow-up of this real-world context (0.2 U/Kg) as compared to RCTs (0.3-0.8 U/Kg) [11][12][13][14][15][16][17][18][19], also considering that 0.2 U/kg should be the starting dose according to the Gla-300 label. However, a titration beyond 6 months from starting insulin therapy cannot be excluded in a reallife setting.
Results of our study were comparable to those of RCTs including insulin-naïve T2D cohorts treated with Gla-300 or Deg-100. The EDITION 3 study [13] showed that in the Gla-300 group, HbA 1c at T0 was of 8.51%, and at T6, it was reduced by 1.42%; the proportion of patients reaching HbA1c <7.0 at T6 was 43.1%. In the BEGIN Once-long [18], in the arm treated with Deg-100, HbA 1c decreased from 8.2% at T0 by 1.06% after 6 months. In the first headto-head RCT (BRIGHT) [19], HbA 1c improved similarly in the two groups from 8.6 to 8.7% to 7.0% after 6 months. In the same study, comparable rates of hypoglycemia in the Gla-300 vs. Deg-100 group were documented (9.3 and 10.8 events per patient-year for hypoglycemia <=70 mg/dl and 0.6 versus 0.9 events per patient-year for hypoglycemia <54 mg/dl). A weight gain of about 2 Kg was documented in both groups.
Results of our study are also comparable with those obtained in the US real-world setting among insulin-naïve patients with T2D [23]. RESTORE-2 study shows very similar results to DELIVER D naïve study [23], where HbA 1c reduction was comparable in the Gla-300 and Deg-100 cohorts (− 1.67% vs. − 1.58%; p = 0.51), as was HbA 1c target attainment (HbA 1c <7%: 23.8% and 27.4%; p = 0.20; HbA1c <8%: 55.0% and 57.1%; p = 0.63). Furthermore, similarly low rates of hypoglycemia were reported in the two groups. In the LIGHTING study [24], data were collected from the Optum Humedica US electronic health records database. In the naïve matched cohort, a HbA 1c reduction of 1.5% was detected in patients treated with Gla-300 and Deg-100. Furthermore, severe hypoglycemic event rates with Gla-300 (estimated though a predictive modeling approach) were not different from those with Deg-100, whereas predicted rates of non-severe hypoglycemia were significantly lower with Gla-300 versus all comparators (first-and second-generation basal insulins). In the CONFIRM study [22], significantly greater HbA 1c reduction, larger reductions in hypoglycemia rates and lower risk of treatment discontinuation were demonstrated with Deg-100 versus Gla-300. However, these results should be interpreted with caution for at least two reasons: i) Both HbA 1c and hypoglycemia were analyzed in subgroups of the PS matched cohort without any further check on between-group imbalance and ii) pattern of baseline medications and an initial daily insulin dose of 40 U are poorly consistent with the expected profile of insulin-naïve patients [23]. Furthermore, there was no match with regard to hypoglycemia rates prior to insulin initiation and there was an imbalance that could per se drive the differences of the changes in hypoglycemia rates with Deg-100 versus Glar-300 [26].
A study on T1D (RESTORE-1) based on the same methodology described here has recently been published [27]. In agreement with RESTORE-1 findings, the take-home message from our new RESTORE-2 is the confirmation of a similar effectiveness and safety of the two 2BI. Yet, the need to overcome clinical inertia is still not met. Late insulin initiation and slow titration are likely due at least in part to the fear of hypoglycemia and weight gain. From a methodological standpoint, both studies confirm the importance of the secondary use of preexisting data for clinical research purposes. In this respect, the Italian network of diabetes centers adopting the same EMRs system represents a unique opportunity to conduct large, real-world effectiveness studies.
Among notable strengths of this study, we underline that this is the first RWE comparative study conducted outside the USA on the effectiveness and safety of 2BIs in T2D. Thanks to the large sample of patients with T2D routinely cared for by centers located in different areas of Italy, and the efficient use of EMR for research purposes, our data have good generalizability to the population of individuals with T2D followed under diabetologist care in Italy. The comparison of hypoglycemic episodes was based on documented events, with the caveat that only a subset of patients had their SMBG downloaded on EMRs. Indeed, the main limitation of this retrospective analysis was the lack of information on SMBG tests for a relevant proportion of patients, although the performed analysis is robust due to the large number of SMBG tests analyzed, the post-matching balance between the two groups, and the lack of difference in baseline risk of hypoglycemia. The download of SMBG values from glucose meters on EMRs was not a common practice in participating centers, suggesting the need to implement the systematic revision of SMBG data through EMRs.
Furthermore, between-group comparison was at least partially limited by the modest dose titration, preventing a head-to-head comparison between the two insulins when optimally used. Nevertheless, data reflect real-life practice and the barriers to optimize insulin doses, especially when patients or diabetologists are concerned by the risk of hypoglycemia. This is a key point deserving consideration in view of the importance of bringing more patients to their HbA 1c target.
In conclusion, in this comparative real-world study with PS matched cohorts of adult patients with T2D, initiating Gla-300 or Deg-100 was associated with similar improvements in glycemic control both in medium and long term, without weight gain and with low rate of hypoglycemia, with no severe episodes observed during an average of 6-month follow-up. The RESTORE-2 study confirms the effectiveness and safety of Gla-300, with comparable results with respect to Deg-100.   Human and animal rights All procedures followed 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 2008. The study protocol was approved by all local ethics committees of the participating centers.

Appendix 1: Variables used in PS matching-pre-and post-PSM baseline patients' characteristics (ITT population)
Informed consent Informed consent was obtained from all patients for being included in the study.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.