Investigation of the Effect of Canagliflozin on the Disposition Index, a Marker of Pancreatic Beta Cell Function, in Patients with Type 2 Diabetes

Introduction

Type 2 diabetes mellitus is characterized by progressive impairments in pancreatic beta cell function,^1–4 with an estimated decrease of about 25% over 5 years.² The deterioration in glycemic control itself also aggravates pancreatic beta cell function, so-called glucotoxicity, leading to a vicious cycle.⁵ As a consequence, the proportion of patients who achieve their glycated hemoglobin (HbA1c) target decreases with longer diabetes duration, despite increases in the number of prescribed antihyperglycemic agents.^6–8 However, once hyperglycemia is appropriately corrected, it may be possible to eliminate glucotoxicity and ameliorate pancreatic beta cell function.^9,10 Therefore, it is important to preserve beta cell function using appropriate antihyperglycemic agents.

Metformin, dipeptidyl peptidase-4 (DPP-4) inhibitors, and sulfonylureas (especially glimepiride) are commonly used in clinical practice.^11–13 If patients taking these three agents require further treatment intensification, increasing the sulfonylurea dose is an option. However, sulfonylureas are associated with increased risk of hypoglycemia,^14,15 especially at high doses,¹⁶ necessitating continued vigilance for hypoglycemia. By comparison, sodium–glucose cotransporter-2 (SGLT2) inhibitors, with a different antihyperglycemic mechanism, have the potential to be combined with commonly used antihyperglycemic agents.^17–21 Adding an SGLT2 inhibitor may be an effective alternative to adjusting the glimepiride dose in such patients.

Some studies have suggested that SGLT2 inhibitors may confer improvements in beta cell function.²² It is desirable that beta cell function is evaluated as insulin secretion adjusted for whole-body insulin sensitivity, because insulin secretion has a hyperbolic relationship with whole-body insulin sensitivity.²³ The disposition index (DI), the product of the insulin response to glucose and insulin sensitivity indices, has emerged as a valuable tool for the purpose of evaluating beta cell function.²⁴ In fact, a short-term pilot study recently demonstrated that administration of an SGLT2 inhibitor improved beta cell function measured as DI.²⁵

Currently, however, it is unknown whether adjusting the sulfonylurea dose or adding an SGLT2 inhibitor is more beneficial in terms of preserving beta cell function and glycemic control in patients with type 2 diabetes on triple therapy. Therefore, we performed a randomized, open-label, parallel-group trial to compare the effects of add-on canagliflozin with glimepiride dose adjustment and glimepiride dose adjustment on pancreatic beta cell function measured in terms of the DI, in patients with type 2 diabetes with inadequate glycemic control despite stable triple therapy (metformin, teneligliptin, and glimepiride) plus diet and exercise therapy. We also explored which factors are related to the improvement in DI.

Methods

Ethics

The study was conducted in compliance with the Declaration of Helsinki and the Clinical Trials Act (Act No. 16 of April 14, 2017). The study was approved by the Institutional Review Board and was registered with the University Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR, UMIN000030208) on March 28, 2018. However, due to the new Clinical Research Act enforced on April 1, 2018, in Japan, studies not completed by March 31, 2019, were required to be re-reviewed by a Certified Review Board and re-registered on the Japan Registry of Clinical Trials (jRCT). Therefore, this study was re-reviewed by Osaka University Clinical Research Review Committee, and re-registered with the jRCT on January 29, 2019 (jRCTs051180036).

Patients

Patients aged 20–74 years who provided informed consent and who had not achieved their individual glycemic target according to Japanese guidelines²⁶ despite unchanged diet and exercise therapy over 12 weeks and stable doses of metformin (0–2250 mg/day), teneligliptin (20 mg/day), and glimepiride (0.5–2 mg/day) over 12 weeks were eligible for the study.

Patients with any of the following were excluded: type 1 diabetes mellitus, need for insulin therapy, history of hypersensitivity to canagliflozin, history of heart failure (New York Heart Association class IV), estimated glomerular filtration rate <45 mL/min/1.73 m², severe hepatic dysfunction, pregnancy, nursing or planning to become pregnant during the study, suspected or diagnosed malignant tumors, participating in another interventional study, and considered by the investigator to be inappropriate for the study for any other reason.

Concomitant use of the following drugs was prohibited from 12 weeks before the study to the end of the study: insulin, sulfonylureas other than glimepiride, glinides, biguanides other than metformin, DPP-4 inhibitors other than teneligliptin, GLP-1 receptor agonists, SGLT2 inhibitors except for canagliflozin during the treatment period and newly launched antidiabetic drugs or investigational products. Patients could use α-glucosidase inhibitors or thiazolidinediones provided that they were started more than 12 weeks before the run-in period and was planned to continue administration at the same dose until the end of the study.

Study Design

This study was performed at Shiraiwa Medical Clinic (Osaka, Japan). It comprised a 2–4 week run-in period, a 24-week treatment period, and a 1-week wash-out period. At the start of the run-in period, the participants underwent a 75 g oral glucose tolerance test (OGTT). After the run-in period, they were randomized 1:1 to the glimepiride group (glimepiride dose adjustment) or the canagliflozin group (add-on canagliflozin 100 mg and glimepiride dose adjustment) using a dynamic allocation method stratified by the OGTT-derived DI and the glimepiride dose at the start of the run-in period. Randomization was performed by central registration with an online registration system. In both groups, teneligliptin and metformin were continued at the same doses. Patients visited the clinic every 4 weeks. Twenty-four-hour blood glucose levels in daily life were monitored for 2 weeks after each visit by continuous glucose monitoring (CGM; FreeStyle Libre Pro; Abbott). The glimepiride dose was adjusted according to the prescribed algorithm based on the CGM data (Table S1) at the next visit. At the end of the treatment period, the patients returned to the regimen they had used in the run-in period for 1 week (wash-out period). The patients underwent a 75 g OGTT at the end of the wash-out period.

Prior to the OGTTs, the patients fasted for more than 10 h and were asked to refrain from taking the study drugs, concomitant medications, or restricted medications (α-glucosidase inhibitors, thiazolidines). Patients drank Trelan-G solution (=75 g of glucose) within 5 min, and blood samples were obtained at 0 min, 30 min (±5 min), 60 min (±10 min), and 120 min (±10 min).

The primary endpoint of this study was the change in DI from baseline to week 25. DI was calculated as (IAUC_Ins0–120/IAUC_Glu0–120) × (Matsuda index), where IAUC_Ins0–120 and IAUC_glu0–120 are the incremental areas under the insulin and glucose curves, respectively, from 0 to 120 min.^10,24,25,27 As the insulin secretion index and Matsuda index (index of insulin resistance) tend to be influenced by SGLT2 inhibitors,²⁵ wash-out of the SGLT2 inhibitor is needed to evaluate the change in beta cell function after treatment. The effect of canagliflozin (100 mg) on urinary glucose excretion is mostly abolished by 3 days after stopping administration.²⁸ Therefore, we assessed the post-treatment DI after a 1-week wash-out to exclude the influence of canagliflozin on the calculation of DI from an OGTT. Secondary endpoints were changes in HbA1c, fasting plasma glucose (FPG), body weight, glycemic parameters derived from CGM, and glimepiride doses. Glycemic parameters derived from CGM data were the mean glucose, glucose standard deviation (SD), mean amplitude of glucose excursions (MAGE), total area for the range of glucose variability, and the proportion of time spent with glucose in hypoglycemic (≤70 mg/dL, ≤3.89 mmol/L; <54 mg/dL, <3.00 mmol/L), normoglycemic (>70 to <180 mg/dL, >3.89 to <9.99 mmol/L), and hyperglycemic (≥180 mg/dL, ≥9.99 mmol/L) ranges. Safety evaluations include assessments of adverse events (AEs) described using the Medical Dictionary for Regulatory Activities/Japanese version 21.0 and laboratory variables. Hypoglycemia was defined as a blood glucose level of ≤70 mg/dL (≤3.89 mmol/L) measured using an in-hospital blood sample, a CGM level of ≤70 mg/dL (≤3.89 mmol/L), or the presence of symptoms of hypoglycemia.

Determination of Sample Size

For the change in DI from baseline to the week 25, we estimated a mean difference in the log(DI) of 0.48 (i.e. 1.62-fold of intergroup ratio; difference ratio 62%) with an SD of 0.49 in the canagliflozin add-on group versus the glimepiride group based on an unpublished sub-group analysis using data for patients with an HbA1c of 7.0–8.0% (53.01–63.94 mmol/mol) in the TA-7284-05 study.^17,29 We estimated that a sample size of 17 patients per group was needed to achieve a power of 80%. Taking into consideration the number of withdrawals, the target sample size was set at 40 patients (20 patients/group).

Statistical Analyses

Efficacy was analyzed using the full analysis set, which included all patients in whom DI was calculated in both the run-in and wash-out periods. Safety was assessed using the safety analysis set, which included all randomized patients, excluding any patients who did not receive any dose of canagliflozin or glimepiride.

Data are presented as the least squares mean (LS mean) with 95% CIs, geometric mean with 95% CI, or mean with SD. Because DI was expected to show a right-skewed distribution, the values were log-transformed prior to data analysis. Some secondary variables demonstrating a right-skewed distribution were also log-transformed. Comparisons between the canagliflozin and glimepiride groups for the change from baseline were made using ANCOVA with the baseline level as the covariate. Anti-log transformation (exponential transformation) was performed to calculate the post/pre-dose ratios and the intergroup ratios (canagliflozin group/glimepiride group) with 95% CIs. The post/pre-percent change (%) or intergroup difference (%) was calculated as one subtracted from these values and multiplied by 100. Similar analyses were made for the DI-related items.

For intergroup comparisons of HbA1c, FPG, and body weight, the change at the last time point (last observation carried forward, LOCF) was analyzed with the baseline level as a covariate, and the LS mean changes with 95% CIs are shown. For the glimepiride dose and CGM-derived parameters (mean, SD, MAGE, and total area for the range of glucose variability), the LS mean changes from baseline with 95% CIs are shown. For the changes at 24 weeks after administration or at 1 week after the treatment period, ANCOVA was performed using the baseline level as the covariate. Similarly, the changes at each time point were also analyzed as the reference. The mean and SD for the time course of the proportion of time in hypoglycemic (≤70 mg/dL, ≤3.89 mmol/L; <54 mg/dL, <3.00 mmol/L), normoglycemic (>70 to <180 mg/dL, >3.89 to <9.99 mmol/L), and hyperglycemic (≥180 mg/dL, ≥9.99 mmol/L) ranges were determined from CGM data. The correlations between log(DI) and secondary endpoints using Pearson’s correlation coefficient were determined in the canagliflozin group, glimepiride group, and in the overall cohort. The proportions of time in hypoglycemic, normoglycemic, and hyperglycemic ranges on CGM, and the correlations between log(DI) and secondary endpoints in the overall cohort were assessed in post hoc analyses.

All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). Data management and statistical analyses were conducted by EP-CRSU Co., Ltd. (Tokyo, Japan).

Results

Patients

Between April 24 and August 29, 2018, a total of 40 patients were recruited, enrolled, and randomized to the two groups, with 20 patients per group (Figure S1). Thirty-nine patients completed the treatment period. One patient allocated to the canagliflozin group was withdrawn because this patient was found to meet an exclusion criterion (diagnosis of a malignant tumor or suspected malignant tumor) following randomization. Therefore, the canagliflozin group comprised 19 patients for efficacy analyses and 20 patients for safety analyses. Both groups were well balanced in terms of baseline characteristics (Table 1).

Table 1 Patient Characteristics (Full Analysis Set)

Βeta Cell Function and Glycemic Control

The mean daily glimepiride dose decreased steadily in both groups, from about 0.8 mg per day at baseline to 0.069 mg in the canagliflozin group and 0.188 mg in the glimepiride group (Figure 1A). Overall, 15/19 and 12/20 patients in the canagliflozin and glimepiride groups, respectively, stopped glimepiride use during the treatment period. The between-group difference of glimepiride dose in the LS mean at week 24 was −0.119 mg (95% CI −0.286 to 0.048, P = 0.158).

Figure 1 Glimepiride doses over the preceding 4 weeks (A) and changes in HbA1c (B), FPG (C), body weight (D), and percent change in DI (E) from baseline.

As shown in Figure 1B, HbA1c decreased from baseline to week 24 in the canagliflozin group but increased in the glimepiride group, with a significant difference in LS mean of −0.72% at the LOCF (95% CI −1.20% to −0.24%, P = 0.005) (−7.9 [−13.1 to −2.6] mmol/mol). A significant reduction in FPG was observed in the canagliflozin group, with a difference in LS mean of −27.0 mg/dL at the LOCF (95% CI −47.3 to −6.6 mg/dL, P = 0.011) (−1.50 [−2.63 to −0.37] mmol/L) (Figure 1C). Although body weight decreased in both groups, the magnitude of the decrease at LOCF was greater in the canagliflozin group (LS mean: −4.03 vs −1.88 kg; difference in LS mean: −2.15 kg, 95% CI −3.46 to −0.84 kg, P = 0.002; Figure 1D).

The primary endpoint, the change in DI from baseline to week 25, is shown in Figure 1E and Table S2. DI increased by 5.1% in the canagliflozin group and decreased by 11.0% in the glimepiride group. Although the ratio for the change in DI was 18.0% greater in the canagliflozin group, this difference was not statistically significant (95% CI −16.0% to 65.9%, P = 0.330).

As DI is calculated from OGTT-associated indices (Matsuda index, IAUC_Ins0–120 and IAUC_Glu0–120), we examined the changes in the indices differently. However, we found no significant differences in the changes in these indices between the two groups (Table S2).

We also examined the changes in glycemic parameters derived from CGM (Figure 2). As illustrated in Figure 2A–D, the measures of mean glucose and mean glucose variability (SD, MAGE, and total area for the range of glucose variability) showed tendencies to decrease in the canagliflozin group within 4–8 weeks after baseline and remained at lower levels compared with the glimepiride group. We observed trends toward increased time spent in the normoglycemic range and decreased time spent in the hyperglycemic range in the canagliflozin group, whereas the time spent in the hypoglycemic range decreased in both groups (Figure 2E–H). The proportion of time with nocturnal hypoglycemia (0:00–5:59) decreased in both groups (Figure S2).

Figure 2 Continued.

Figure 2 (A–D) Changes in mean glucose (A), glucose SD (B), MAGE (C), and total area for the range of glucose variability (D) determined by CGM. (E–H) The proportion of time with glucose levels in hyperglycemic (≥180 mg/dL, ≥9.9 mmol/L, E), normoglycemic (>70 to <180 mg/dL, >3.89 mmol/L to <9.99 mmol/L, F), and hypoglycemic (≤70 mg/dL, ≤3.89 mmol/L, G; <54 mg/dL, <3.00 mmol/L, H) ranges, as determined by CGM.

We next determined the correlations between change in log(DI) and efficacy endpoints in the overall cohort (n = 39) and in both treatment groups separately (Figure 3). In the overall cohort, the change in log(DI) was significantly correlated with the changes in HbA1c (r = −0.448, P = 0.004), FPG (r = −0.486, P = 0.002), mean glucose (r = −0.588, P < 0.001), MAGE (r = −0.477, P = 0.003) and the proportions of time spent in the hyperglycemic (r = −0.638, P < 0.001) and normoglycemic (r = 0.647, P < 0.001) ranges. These correlations were observed in the canagliflozin and glimepiride groups individually, except for the change in FPG in the canagliflozin group and the changes in HbA1c and MAGE in the glimepiride group. The change in log(DI) was not correlated with the change in body weight in the overall cohort or in either group (Figure 3C).

Figure 3 Continued.

Figure 3 Correlations between the changes in log(DI) and the changes in HbA1c (A), FPG (B), body weight (C), mean glucose (D), MAGE (E), and proportions of time with glucose level in hyperglycemic (≥180 mg/dL, ≥9.99 mmol/L, F) and normoglycemic (>70 to <180 mg/dL, >3.89 to <9.99 mmol/L, G) ranges.

Safety

Safety was assessed in terms of vital signs, laboratory data, and AEs. The changes in laboratory values are shown in Table S3. AEs occurred in 20 (100%) and 18 (90%) patients in the canagliflozin and glimepiride groups, respectively, and included serious AEs in one patient in each group (hepatic neoplasm in the canagliflozin group and hydronephrosis in the glimepiride group) and adverse drug reactions in 15 patients in each group (Table S4). Hypoglycemia was the most frequent AE, with 44 episodes occurring in 12 patients (60%) in the canagliflozin group and 79 episodes in 14 patients (70%) in the glimepiride group. Only three episodes of hypoglycemia with symptoms were reported in the canagliflozin group and were not severe; all other episodes were detected by CGM, in other words, unconscious hypoglycemia. The AEs blood ketone body increased, and pruritus genital were observed in six and three patients in the canagliflozin group, respectively; all of these events were not severe.

Discussion

In this study, patients with type 2 diabetes on triple combination therapy were randomized to undergo either glimepiride dose adjustment or glimepiride dose adjustment and the addition of canagliflozin to ongoing treatment. During the 24-week treatment period, the mean glimepiride dose was decreased in both groups due to low glucose levels detected by CGM in accordance with the prespecified algorithm. The change in DI was −11% in the glimepiride group and 5% in the canagliflozin group, with an intergroup difference ratio of 18%, although this was not statistically significant. HbA1c and FPG levels were significantly lower in the canagliflozin group than in the glimepiride group. The CGM data, including the mean, SD and proportion of time with glucose levels in hyperglycemic, normoglycemic and hypoglycemic ranges, suggest that canagliflozin improved the daily blood glucose profile. A total of 44 and 79 hypoglycemic episodes were observed in the canagliflozin group and the glimepiride group, respectively. In addition, we found that the change in DI was correlated with improvements in glycemic control in terms of glucose levels and variability.

The TA-7284-05 study^17,29 was used as a reference to determine the sample size for the present study. The change in DI in the present study was not statistically significant, indicating that DI was not improved by treatment with canagliflozin in the current population to the extent expected from the TA-7284-05 study (i.e. an improvement by 62%). However, the patients enrolled in the present study had a longer duration of diabetes, of about 10 years, and were on the triple therapy (metformin, teneligliptin, and glimepiride) at baseline, in contrast to the TA-7284-05 study (duration of diabetes: about 5 years; diet/exercise therapy only). Therefore, it may have been difficult to achieve the level of improvement in DI seen in the TA-7284-05 study in the present study.

We previously reported that medical nutrition therapy without pharmacotherapy aided the recovery of DI in patients with newly diagnosed type 2 diabetes presenting with extreme hyperglycemia and also show the correlation between DI and HbA1c.¹⁰ Consistently, we found a correlation between the improvements in DI and glucose control in a population of patients with more advanced diabetes. A number of cross-sectional studies and an interventional study have documented an association between beta cell function and glycemic variability.^30–32 However, in those studies, glycemic variability was evaluated by self-monitored blood glucose^30,31 or CGM under controlled meal conditions.³² Unlike these earlier studies, glycemic levels and variability were assessed by CGM in real daily-life conditions over 2-week periods in the present study. The results of our interventional study provide further, novel evidence for the association between mean blood glucose or glycemic variability and beta cell function in daily life. Our previous in vitro studies, which explored the underlying mechanisms in beta cell glucotoxicity, revealed that alleviating glucotoxicity and lipotoxicity is associated with improvements in islet gene expression of Pdx1, Mafa, and Slc2a2, key factors involved in beta cell function.^33,34 Thus, it is possible that reducing glucotoxicity may contribute to the improvement in beta cell function and better glycemic control observed in the current study.

CGM in daily life in the current study revealed frequent asymptomatic hypoglycemia despite low doses of glimepiride in patients presenting with hyperglycemia and no obvious signs of hypoglycemia at baseline. Accordingly, and somewhat unexpectedly, it was necessary to lower the dose of glimepiride, not increase it, in many patients, even in the glimepiride group. Hypoglycemia is associated with increased cardiovascular risk, cognitive dysfunction, dementia, and vision disorders.^35,36 In patients with recurrent hypoglycemia, the brain adapts to hypoglycemia such that subsequent symptoms manifest at lower plasma glucose concentrations, reducing awareness of hypoglycemia.³⁶ Therefore, it is important to adjust the dose of glimepiride to avoid the risk of hypoglycemia. The current findings indicate that glimepiride is associated with increased risk of hypoglycemia, even when administered at low doses, in patients with inadequate glycemic control on the triple therapy and that canagliflozin may improve glycemic control without increasing the risk of hypoglycemia in these patients.

AEs other than hypoglycemia that were observed in the canagliflozin group only included blood ketone body increased and pruritus genital. These AEs have previously been documented in patients treated with SGLT2 inhibitors, including canagliflozin, in clinical practice.^{17–20,29,37} These AEs were mild in severity in this study. Our results demonstrate the tolerability of adding canagliflozin to ongoing triple combination therapy.

Limitations of this study include its small sample size, the short-term treatment period, and open-label design. The sample size was small and did not have sufficient power to detect a significant difference in DI in this population. The longer-term effects of canagliflozin await confirmation.

In conclusion, adding canagliflozin to the triple therapy improved beta cell function by 18%, but it did not reach statistical significance. Treatment with canagliflozin was associated with improvements in overall glycemic control and glycemic variability, with a relatively lower risk of hypoglycemia compared with glimepiride dose adjustment. This study also revealed correlations between the improvement in beta cell function (change in DI) and changes in markers of glycemic control.