Trial Design
This was a 24-week, single-center, open-label, parallel-group exploratory trial with 1:1 randomization comparing self-titration of Gla-300 versus NPH insulin on GV in insulin-naïve T2D patients. The study was conducted at the Prince of Wales Hospital, Hong Kong Special Administrative Region, between January 2018 and December 2019. The frequency of study visits was in line with usual practice following insulin initiation in the local setting. We selected NPH as a comparator since this remains the first-choice insulin in the publicly funded care setting in the absence of risk factors for hypoglycemia including advanced chronic kidney disease, history of severe hypoglycemia or significant comorbidities. The study protocol was approved by the Joint New Territories East Cluster and Chinese University of Hong Kong Clinical Research Ethics Committee. The study was conducted according to the Good Clinical Practice guidelines and Declaration of Helsinki. All participants provided written informed consent (ClinicalTrials.gov Identifier: NCT03389490).
Participants
Insulin-naïve T2D individuals were included if they were aged between 18 and 75 years, treated with three or fewer OGLDs, had a body mass index (BMI) < 40 kg/m2 and were capable and willing to perform regular self-monitored blood glucose at home (SMBG). Patients with HbA1c level > 7.0 and < 11% and fasting plasma glucose (FPG) > 7.5 and < 15 mmol/l at screening were included. Key exclusion criteria included known hypoglycemia unawareness or recurrent severe hypoglycemia, concomitant medications known to interfere with glucose metabolism, a change in dose of non-insulin GLDs or initiation of new OGLDs in the 8 weeks prior to screening, acute concurrent illness during the 3-month period prior to study, hepatic disease, end-stage kidney disease, pregnant or breastfeeding women and those who were unable to comply with follow-up visits.
Randomization
Eligible patients were randomized to Gla-300 insulin (Sanofi-Aventis, Paris France) or NPH insulin (either Humulin N, Eli Lilly, Indianapolis, IN, USA, or Protaphane HM, Novo Nordisk A/S, Bagsvaerd, Denmark) in a 1:1 ratio. Assignment to treatment group was by a computer-generated random sequence by personnel not involved in the study.
Interventions
Each patient was randomized to either Gla-300 or NPH insulin to be administered once daily between 9 p.m. and 12 midnight at approximately the same time every day. The starting doses of Gla-300 and NPH were both 0.2U/kg for comparison. Insulin was self-titrated weekly based on an average of three fasting SMBG readings per week, aiming for a mean fasting SMBG target of 4.4–6.0 mmol/l. Insulin was increased by 1U per week if the mean value was greater than the pre-set target. Patients were asked to contact the study team for adjustment of insulin doses if the average SMBG was ≥ 3.3 and < 4.4 mmol/l. The study team would decrease insulin by ≥ 2 U (according to investigator discretion) if SMBG < 3.3 mmol/l or severe or multiple symptomatic hypoglycemic events occurred.
The study consisted of seven pre-specified on-site visits at weeks − 1, 0, 4, 11, 12 and 24 and three telephone visits. Following randomization at week 0, phone visits were conducted at week 1, 2 and 8 to report on SMBG, insulin titration, and any hypoglycemia or adverse events. Patients were instructed on the self-titration algorithm at randomization, reinforced as necessary during the first 12 weeks of self-titration. Masked CGMs for determination of GV and TIR were collected for 7 consecutive days at baseline (week – 1), week 11 and week 24 using the Medtronic iPro2™ with Elite sensor (Medtronic, Northridge, CA, USA) with frequency of calibrations as recommended by the manufacturer. Raw CGM data were inspected and only patients with at least 70% valid sensor data were included in the analysis. Other secondary outcomes were measured at weeks 0, 12 and 24.
Rescue Medications
The doses and regimen of concomitant OGLDs remained stable throughout the study unless the patient fulfilled the rescue criteria. After week 12, rapid-acting acting prandial insulin (Actrapid or Humulin R) was added before the largest meal at the investigators’ discretion, if FPG could not be lowered below < 11 mmol/l or HbA1c < 8.5% (< 69.4 mmol/mol). Participants were continued on OGLDs throughout the study, with doses of sulfonylureas/glinides adjusted if ≥ 2 symptomatic or one severe hypoglycemic episode occurred, even after adjustment of insulin.
Outcome Measures
The primary outcome was described by the standard deviation (SD) of CGM glucose. The coefficient of variation of CGM glucose (CV%) was defined as SD/average glucose × 100%. Secondary CGM-based endpoints included average sensor glucose (SG), mean amplitude of glucose excursions (MAGE), glucose management index (GMI), mean percentage time in range (< 3.0 mmol/l, < 3.9 mmol/l 3.9–10.0 mmol/l, > 10.0 mmol/l, > 13.9 mmol/l) [7]. GMI, a correlate of HbA1c based on CGM mean glucose, was defined according to the formula GMI (%) = 3.31 + 0.02392 × [mean glucose in mg/l] [8]. Other glycemic endpoints included mean change in laboratory HbA1c and FPG at week 24, daily fasting SMBG and 7-point SMBG profiles at week 24, insulin dose (units per kilogram) and changes in body weight. The proportion of patients achieving target HbA1c < 7%, with and without biochemically confirmed hypoglycemia, respectively, was calculated.
Self-reported hypoglycemic events were collected throughout the on-treatment study period. The number of biochemically confirmed level 1 (< 3.9 mmol/l) and level 2 (< 3.0 mmol/l) hypoglycemic events was reported [9]. A severe hypoglycemic event was denoted by severe cognitive impairment requiring assistance for recovery. The incidence and number of hypoglycemic events per patient-year were calculated using the above thresholds; 00:00–05:59 h was categorized as nocturnal and between 0600 to 2359 h as diurnal. We also analyzed events during titration (weeks 0–12) and maintenance (weeks 13–24) periods.
Statistical Methods
Data were expressed as mean (SD) unless otherwise stated. All efficacy endpoints were assessed in the intention-to-treat (ITT) population (all randomized patients who received at least one dose of study insulin, analyzed according to the treatment group allocated by randomization). Change in GV and secondary endpoints (average SG, TIRs, HbA1c, FPG) during the 24-week period were analyzed by a mixed effect model with repeated measures (MMRM) using a missing at random framework, with treatment group, visit and treatment-by-visit interaction as fixed effects, participant as random effect, and baseline efficacy variable and baseline efficacy variable-by-visit interaction as fixed continuous covariates. From the model, relevant treatment differences were estimated as least squares (LS) means with standard errors (SEs) and LS mean difference with 95% confidence interval [CI]. Twenty-four-hour glucose profiles were generated by calculating the mean and 95% CI pooled across all patients within each treatment group. The proportion of patients with one or more self-reported hypoglycemic episodes was estimated. Annualized hypoglycemia event rates were calculated as events per patient-year, and the relative risk between treatment groups was estimated using the overinflated Poisson regression model.
Based on estimated mean scores for anytime glucose SD and a standard deviation of 0.5 mmol/l, a total sample size of 44 (22 in each arm) would have a power of ~ 80% at alpha 0.05 to detect a 0.5 mmol/l difference in the primary outcome at week 24 using sample size estimation for a multilevel design with repeated measures [10]. Fifty patients were enrolled allowing for a 12% drop-out rate. Data were analyzed with R version 3.3.2 (R Foundation for Statistical Computing) and SPSS software package version 25.0 (IBM Corp., Armonk, NY, USA). A p value < 0.05 was considered statistically significant.