Factors contributing to the clinical effectiveness of imeglimin monotherapy in Japanese patients with type 2 diabetes mellitus

Abstract

Aims/Introduction

To investigate the effect of patient characteristics on imeglimin effectiveness in Japanese patients with type 2 diabetes mellitus.

Materials and Methods

Data were pooled from two randomized, placebo‐controlled, 24‐week, double‐blind studies of imeglimin monotherapy in Japanese adults with type 2 diabetes mellitus, with the proportion of responders (glycated hemoglobin [HbA1c] < 7.0%) and sustained responders (i.e., achieved and maintained response) in the imeglimin 1,000 mg twice daily group calculated at each visit. Patient factors significantly (P < 0.05) correlated with response were explored through multivariate logistic regression. Subgroup analyses compared the efficacy of imeglimin in patients with a HbA1c improvement less than or equal to −0.3% (early responders) versus greater than −0.3% (early non‐responders) at week 4.

Results

A total of 38.0% of imeglimin‐treated patients and 7.2% of placebo‐treated patients were responders (P < 0.001, number needed to treat = 4). The proportion of sustained responders at weeks 4, 8, 12, 16 and 20 was 10.6, 19.0, 24.0, 25.7 and 29.1%, respectively (>70% of responders at each visit). Improvements in HbA1c and fasting glucose were significantly greater in early responders versus early non‐responders from week 4; between‐group differences remained significant to week 24. Older age (odds ratio 1.09, 95% confidence interval 1.04–1.14; P < 0.001); treatment‐naïve status vs previous treatment (odds ratio 3.70, 95% confidence interval 1.55–8.82; P = 0.003), and lower baseline HbA1c (odds ratio 0.06, 95% confidence interval 0.02–0.16; P < 0.001) predicted response.

Conclusions

A significantly higher proportion of patients receiving imeglimin 1,000 mg twice daily monotherapy were responders versus placebo. Most (>70%) were sustained responders, suggesting that response is fairly predictable. Older age, treatment‐naïve status and early treatment response significantly predicted imeglimin effectiveness.

It is important to investigate the effect of various patient characteristics on the clinical effectiveness of imeglimin in Japanese patients with type 2 diabetes mellitus. In this post‐hoc analysis, the majority (>70%) of patients who received imeglimin 1,000 mg twice daily as monotherapy and showed a treatment response were sustained responders, suggesting that response was fairly predictable. Older age, treatment‐naïve status and early treatment response significantly predicted imeglimin efficacy.

INTRODUCTION

Imeglimin ([6R]‐[+]‐4‐dimethylamino2‐imino‐6methyl‐1,2,5,6 tetrahydro1,3,5, triazine hydrochloride) is a new oral antidiabetic agent for the treatment of type 2 diabetes mellitus, and the first in a new class of tetrahydrotriazine‐containing drugs, known as the ‘glimins’ ¹ . The antihyperglycemic action of imeglimin has been shown in preclinical studies to be attributable to the suppression of hepatic gluconeogenesis ² , ³ , increased glucose uptake in skeletal muscle ² , ⁴ and improved glucose‐stimulated insulin secretion ⁵ , ⁶ , ⁷ . Furthermore, imeglimin appears to correct the underlying mitochondrial dysfunction and associated defects in cellular energy metabolism underpinning type 2 diabetes mellitus pathophysiology ⁷ , ⁸ . To date, most randomized controlled studies have confirmed the efficacy and safety of imeglimin as either monotherapy or as add‐on therapy with either metformin or insulin for type 2 diabetes mellitus ⁹ , ¹⁰ , ¹¹ . However, given its distinct modes of action compared with existing therapeutics, imeglimin has also shown potential for use in combination with other drugs routinely used to treat type 2 diabetes mellitus, such as insulin sensitizers, insulin secretagogues and β‐cell protective drugs ¹² , ¹³ .

In June 2021, imeglimin received its first global approval for the treatment of type 2 diabetes mellitus in Japan ¹⁴ . Approval was granted based on the results of extensive preclinical and clinical trial data, including the pivotal phase III Trials of IMeglimin for Efficacy and Safety (TIMES) studies carried out in Japanese patients with type 2 diabetes mellitus ¹⁰ , ¹³ , ¹⁵ . Across these studies, imeglimin alone and in combination with insulin therapy was associated with significantly greater reductions in HbA1c compared with placebo ¹⁰ , ¹⁵ . The safety profile of imeglimin was similar to placebo and no treatment‐related serious adverse events (AEs) were reported; thus, its efficacy and safety profile was considered favorable ¹⁰ , ¹⁵ .

Treatment goals for the management of type 2 diabetes mellitus typically consist of glycemic control and modification of risk factors to prevent the development and progression of diabetic complications ¹⁶ , ¹⁷ . Guidelines generally recommend a glycated hemoglobin (HbA1c) target of <7.0% (<53 mmol/mol) for most patients ¹⁶ , ¹⁷ , although a more stringent HbA1c target of ≤6.5% (≤48 mmol/mol) is endorsed by the American Association of Clinical Endocrinology, provided that it can be achieved safely without significant hypoglycemia or other AEs ¹⁸ . However, there has been a shift in these treatment guidelines, as well as those of the Japan Diabetes Society, toward a patient‐centered approach, with individualized treatment goals based on patient characteristics, such as their age, weight, comorbidities, preferences/priorities and disease duration ¹⁶ , ¹⁷ , ¹⁸ . Nevertheless, data evaluating the effect of various patient characteristics on the effectiveness of various antidiabetic therapies are limited, particularly for recently launched agents. We therefore carried out a post‐hoc analysis of two similarly designed, randomized, double‐blind, placebo‐controlled trials of imeglimin in type 2 diabetes mellitus patients to investigate the effect of various patient characteristics on the clinical effectiveness of imeglimin in Japanese patients with type 2 diabetes mellitus ¹⁰ , ¹¹ .

MATERIALS AND METHODS

Data sources

Data were pooled from two similarly designed, randomized, double‐blind, placebo‐controlled trials of imeglimin in type 2 diabetes mellitus, the phase IIb (Protocol PXL008‐014; JAPIC number: JapicCTI‐153,086) ¹¹ and phase 3 (Protocol PXL008‐018; JAPIC number: JapicCTI‐173,769) ¹⁰ studies.

Studies were carried out in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines, and were approved by the appropriate institutional review boards and regulatory agencies. Patients provided written informed consent to participate in each study.

Study design

Full details regarding the study design and eligibility criteria for each study have been published previously ¹⁰ , ¹¹ . Briefly, both studies were carried out in Japan, and consisted of a screening and drug washout/run‐in period, followed by a 24‐week double‐blind treatment period, and a 1‐week safety follow‐up period.

Patients in PXL008‐014 were randomized in a 1:1:1:1 ratio to receive either imeglimin 500 mg, 1,000 mg, 1,500 mg or a matched placebo, administered orally twice daily (b.i.d.). Both the imeglimin 500 mg and 1,500 mg b.i.d. dosage groups were excluded from this post‐hoc analysis, as they are not within the dosage range recommended for imeglimin.

Patients in PXL008‐018 were randomized in a 1:1 ratio to receive either imeglimin 1,000 mg or matched placebo b.i.d..

The primary efficacy endpoint in both studies was the change in HbA1c from baseline to week 24.

Patients

Japanese adults in the PXL008‐014 study (aged 20–75 years) and PXL008‐018 study (aged ≥20 years) were eligible if they had type 2 diabetes mellitus that was managed with diet and exercise ± a single oral antidiabetic agent (at a stable dose) for ≥12 weeks, and had a HbA1c level between 7.0 and 10.0% (53–86 mmol/mol).

Detailed exclusion criteria have been published previously ¹⁰ , ¹¹ . The main exclusion criteria included prior treatment with insulin or any injectable glucose‐lowering drug in the 3 months (PXL008‐014) or 30 days (PXL008‐018) before screening, an estimated glomerular filtration rate (eGFR; using the Japanese Modification of Diet in Renal Disease equation specified in the Japanese Modification of Diet in Renal Disease trial) ¹⁹ of <45 mL/min/1.73 m², heart failure (New York Heart Association class III or IV), or any acute coronary or cerebrovascular events in the 24 weeks before screening.

Analyses and statistical considerations

Treatment responders

Responders were defined as patients who achieved a treatment target goal of HbA1c <7.0% (<53 mmol/mol) at week 24, based on the recommendations by the American Diabetes Association ²⁰ , ²¹ and Japan Diabetes Society ¹⁶ . Sustained responders were defined as those patients who met responder criteria and maintained their response until the end of study treatment.

Analysis of responders by visit

The proportion of responders and sustained responders were summarized by visit, and compared between imeglimin and placebo groups using logistic regression analysis, with treatment group and study as fixed effects, and baseline HbA1c as a continuous covariate. The number needed to treat (NNT) was also calculated from the difference in the proportion of responders versus non‐responders.

Multivariate logistic regression analysis

A multivariate logistic regression model was constructed to identify the explanatory variables that had a significant correlation with treatment response to imeglimin at week 24. The stepwise method was used for variable selection, with criteria P < 0.05 for both entry and stay. The explanatory variables included in the model were age, sex, treatment status (treatment naïve vs previous treatment), baseline HbA1c, duration of diabetes, baseline body mass index (BMI) and baseline eGFR. The results were further stratified by baseline HbA1c level (<8.0%/≥8.0%) and previous treatment status (treatment naïve/previous treatment).

Receiver operating characteristic curve analysis

Receiver operating characteristic (ROC) curves were plotted to identify a HbA1c and fasting plasma glucose (FPG) cutoff point at week 4 that was predictive of response to imeglimin 1,000 mg b.i.d. at week 24. The area under the curve (AUC) was used as an indicator, and was calculated with 95% confidence intervals and the P‐value compared with the reference AUC = 0.5 (null hypothesis). In the case where AUC was significant, the optimal cutoff point was determined by Youden's index (sensitivity + specificity − 1) ²² .

Subgroup analyses based on ROC analysis

Analyses of patient subgroups were carried out according to the cutoff point for HbA1c and FPG improvement at week 4 identified by the ROC analysis. Patients with a HbA1c or FPG improvement on or above the cutoff point at wee 4 were defined as early responders, and patients with a HbA1c or FPG improvement below the cutoff point at week 4 were defined as early non‐responders. The efficacy and safety of imeglimin was investigated across three patient subgroups (placebo, early responders and early non‐responders). For efficacy, all pairwise comparisons by visit on changes from baseline in HbA1c and FPG were analyzed using a mixed‐effects model for repeated measures with treatment group, study, visit and treatment by visit interaction as fixed effects, and baseline value as a covariate.

Safety

Safety was assessed by the incidence of AEs and treatment discontinuations due to AEs in each patient subgroup. AEs were categorized by the preferred term of the Medical Dictionary for Regulatory Activities, version 20.1.

Other statistical considerations

Baseline demographics and safety variables were summarized descriptively for each patient subgroup, with the mean and standard deviation (SD) calculated for continuous variables, and the frequency number and proportion calculated for categorical variables.

Two‐sided P‐values <0.05 were considered statistically significant. No multiplicity adjustment was made, because post‐hoc analyses were exploratory in nature.

All statistical analyses were carried out using SAS® software version 9.4 (SAS Institute, Cary, NC, USA).

RESULTS

A total of 362 patients (imeglimin 1,000 mg b.i.d., n = 180; placebo, n = 182) were included in the pooled analysis population. Baseline characteristics were similar between treatment groups (Table 1).

Table 1.

Clinical characteristics of patients at baseline

	Placebo (n = 182)	Imeglimin
		500 mg b.i.d. † (n = 75)	1,000 mg b.i.d. (n = 180)	1,500 mg b.i.d. † (n = 75)
Sex, male, n (%)	127 (69.8)	49 (65.3)	138 (76.7)	54 (72.0)
Mean age, years (SD)	61.2 (9.91)	58.7 (8.51)	61.3 (9.78)	57.7 (10.76)
Mean BMI, kg/m2 (SD)	25.52 (4.29)	25.24 (4.60)	25.47 (4.08)	26.75 (4.35)
Mean HbA1c, % (SD)	7.91 (0.68)	7.94 (0.68)	7.94 (0.72)	7.92 (0.62)
Mean duration of diabetes, years (SD)	6.86 (6.22)	7.20 (6.27)	7.12 (5.56)	5.29 (5.12)
Mean FPG, mg/dL (SD)	159.80 (27.07)	164.51 (31.75)	163.76 (29.51)	165.21 (31.29)

^† These treatment groups were excluded from the analysis.

b.i.d., twice daily; BMI, body mass index, FPG, fasting plasma glucose, HbA1c, glycated hemoglobin.

Responder and sustained responder

Baseline characteristics between imeglimin responders and imeglimin non‐responders, and imeglimin sustained responders and imeglimin non‐sustained responders are presented in Tables S1 and S2, respectively. Compared with imeglimin non‐responders, imeglimin responders were older, had a slightly lower BMI, and lower baseline HbA1c and FPG (Table S1). Similarly, compared with imeglimin non‐sustained responders, sustained responders were slightly older, had a slightly lower BMI, and lower baseline HbA1c and FPG, and had a shorter disease duration (Table S2). Sustained responders/non‐sustained responders differed to responders/non‐responders with respect to patient background characteristics in that there is likely to be a difference in disease duration. The proportions of responders and sustained responders at each study visit for the imeglimin 1,000 mg b.i.d. and placebo treatment groups are presented in Figure 1. Both proportions in the imeglimin group increased over time. Responder proportions were significantly higher in the imeglimin group compared with the placebo group at each time point; that is, at week 4 (P < 0.01) and weeks 8, 12, 16, 20 and 24 (P < 0.001 each) of administration. For sustained responders, between‐group differences were significant for the imeglimin group as early as week 4 (the earliest time point evaluated; P < 0.001). Sustained responders accounted for >70% of responders at each time point evaluated (Figure 1). The NNT at 24 weeks of administration was four.

Figure 1.

Proportion of responders and sustained responders to imeglimin 1,000 mg twice daily. **P < 0.01, ***P < 0.001 versus responder in placebo. †††P < 0.001 versus sustained responder in placebo. Response was defined as a glycated hemoglobin value <7.0% at end‐of‐treatment (week 24). Number needed to treat was presented with the 95% confidence interval. P‐value was calculated with logistic regression analysis. LOCF, last observation carried forward, NNT, number needed to treat.

Multivariate logistic regression analysis: Patient background factors associated with response

The results of the multivariate logistic regression analysis evaluating patient background factors associated with response are presented in Table 2 and Table S3. Explanatory variables that were significantly correlated with response were age, prior treatment status, baseline HbA1c and baseline BMI. Specifically, older age (odds ratio [OR] 1.09, 95% confidence interval [95% CI] 1.04–1.14; P < 0.001); treatment‐naïve versus previously treated patients (OR 3.70, 95% CI 1.55–8.82; P = 0.003) and lower baseline HbA1c (OR 0.06, 95% CI 0.02–0.16; P < 0.001) were significantly associated with imeglimin response.

Table 2.

Multivariate logistic regression analysis evaluating factors associated with treatment response to imeglimin after 24 weeks

Variables	Unit	Coefficient (95% CI)	Odds ratio (95% CI)	P‐value
Age	1 year	0.08 (0.04, 0.13)	1.09 (1.04, 1.14)	<0.001
Previous treatment status	Treatment naïve vs previously treated	1.31 (0.44, 2.18)	3.70 (1.55, 8.82)	0.003
HbA1c at baseline	1%	−2.79 (−3.74, −1.84)	0.06 (0.02, 0.16)	<0.001

CI, confidence interval; HbA1c, glycated hemoglobin.

Additional mixed‐effects model for repeated measures and logistic regression analysis were also carried out to evaluate the effect of baseline HbA1c (<8.0%/≥8.0%) and treatment status (treatment naïve/previous treatment) on imeglimin response, with the results presented in Figure [Link], [Link]. HbA1c improved irrespective of the baseline HbA1c level, and the effect size was not significantly different between subgroups (Figure S1). The proportion of responders and sustained responders was high in patients with a baseline HbA1c <8.0% (Figure S2a), low in patients with a baseline HbA1c level ≥8.0% and not significantly different from placebo (Figure S2b). HbA1c improvements were observed in patients treated with imeglimin 1,000 mg b.i.d. irrespective of whether they were treatment naïve or had received previous treatment, with minimal differences in effect size observed between treatment naïve and pretreated groups (Figure S3). Regarding the proportion of responders and sustained responders, there was a significant difference between treatment‐naïve and pretreated patient groups, with a significantly higher proportion of responders and sustained responders to imeglimin being treatment naïve (Figure S4).

Early responder versus early non‐responder

ROC analysis was carried out to identify a HbA1c and FPG cutoff point at week 4 that was predictive of response to imeglimin 1,000 mg b.i.d. at week 24. The best predictor of response at week 24 was determined to be a HbA1c improvement of less than or equal to −0.3% at week 4 (Figure 2); therefore, the optimal HbA1c cutoff point was determined to be −0.3% from baseline at week 4. This ROC analysis yielded an AUC of 0.703, which is significantly different from a curve obtained by chance (P < 0.001). Based on this cutoff point, 67 patients (37.4%) were classified as early responders and 112 patients (62.6%) were classified as early non‐responders. For the ROC analysis to determine an FPG cutoff point at week 4, the outcome was defined as a HbA1c <7% at week 24, and the threshold for FPG decline at week 4 that best predicted the outcome was determined. The best prediction was −21.6 mg/dL at week 4, but the AUC of the ROC curve was approximately 0.5, and there was no significant difference from a curve obtained by chance (Figure 3).

Figure 2.

Receiver operating characteristic analysis evaluating glycated hemoglobin cutoff at week 4 predicting treatment response. The red line indicates the receiver operating characteristic curve of the target variable (change from baseline in glycated hemoglobin at week 4); the green line indicates the receiver operating characteristic curve where the variable would have no predictive value (area under the curve 0.5). The arrow indicates cutoff at which optimal sensitivity/specificity is reached. The cutoff point is −0.3%, area under the curve 0.703, 95% confidence interval 0.626–0.780, P < 0.001 compared with by chance.

Figure 3.

Receiver operating characteristic analysis evaluating fasting plasma glucose cutoff at week 4 predicting treatment response. The red line indicates the receiver operating characteristic curve of the target variable (change from baseline in fasting plasma glucose at week 4); the green line indicates the receiver operating characteristic curve where the variable would have no predictive value (area under the curve 0.5). The cutoff point is −21.6 mg/dL, area under the curve 0.545, 95% confidence interval 0.459–0.631, P = 0.306 compared with by chance.

Subgroup analyses by HbA1c cutoff point

Patient demographic data and disposition stratified by early response by HbA1c cutoff point are summarized in Tables S4 and S5. Baseline demographics and characteristics were similar between early responders and early non‐responders, with the exception of mean baseline HbA1c, which was higher in early responders (8.13% [SD 0.74]) versus non‐responders (7.82% [SD 0.69]), and mean duration of type 2 diabetes mellitus, which was shorter in early responders (6.22 years [SD 5.27]) versus early non‐responders (7.64 years [SD 5.70]).

Subgroup analyses examined the least squares mean changes from baseline in HbA1c over 24 weeks for placebo, early responder and early non‐responder subgroups by cutoff point (Figure 4). A significant improvement in HbA1c was observed at all evaluation points from week 4 to week 24 in patients receiving imeglimin 1,000 mg b.i.d. compared with placebo. Differences were significant at each time point for early responders and early non‐responders compared with placebo (P < 0.001, respectively); however, the magnitude of effect was significantly greater in early responders (P < 0.001). The difference from placebo in HbA1c change from baseline at week 24 was −1.06% for early responders compared with −0.37% for early non‐responders, with an effect size of 1.92 compared with 0.91, respectively. Similar results were observed for FPG (Figure S5). No significant differences were observed between early responders and early non‐responders with respect to treatment discontinuation rates (Table S5) and the incidence of adverse events (Table S6).

Figure 4.

Change from baseline in glycated hemoglobin (mixed‐effects model for repeated measures). ***P < 0.001 versus placebo. †††P < 0.001 versus early non‐responder (ENR). Early responder (ER): participants in the imeglimin group with a glycated hemoglobin improvement of less than or equal to −0.3% at week 4. Effect sizes versus placebo are shown. Effect size of ER versus ENR was 1.01. ES, effect size, LS, least square.

DISCUSSION

The present post‐hoc analyses investigated the effect of various patient characteristics on the clinical effectiveness of imeglimin using pooled data from two similarly designed, randomized, double‐blind, placebo‐controlled trials of imeglimin in Japanese patients with type 2 diabetes mellitus. A significantly higher proportion of patients receiving imeglimin 1,000 mg b.i.d. were treatment responders compared with placebo at each time point evaluated up to 24 weeks of administration. The NNT was four; thus, not only was the observed effect statistically significant, it was also clinically significant ²³ . The proportion of responders tended to increase over time, >70% of whom achieved sustained therapeutic response at each time point evaluated, suggesting that the efficacy of imeglimin is fairly predictable (i.e., once response is achieved, it is maintained) in most patients.

Type 2 diabetes mellitus encompasses a heterogeneous patient population with diverse clinical features, including varied clinical symptoms, multiple comorbidities and unpredictable responses to treatment. Therefore, medications that are universally and consistently effective are particularly attractive for this indication. Furthermore, characterizing the effectiveness of individual treatments across specific patient subgroups with type 2 diabetes mellitus is needed to allow clinicians to tailor specific therapeutic interventions to those patients who will derive the greatest benefit from them. In this present post‐hoc analysis, the clinical effectiveness of imeglimin monotherapy was shown across a broad spectrum of type 2 diabetes mellitus patients, irrespective of age, sex, previous treatment status, baseline HbA1c, duration of diabetes, baseline BMI or baseline eGFR. These results extend on previous findings showing the overall efficacy of imeglimin in patients with type 2 diabetes mellitus ¹⁰ , ¹³ , ¹⁵ . Of the explanatory variables assessed through multivariate logistic regression, older age (P < 0.001), treatment‐naïve status versus previous treatment (P = 0.003) and lower baseline HbA1c (P < 0.001) were significantly correlated with imeglimin response. These findings are consistent with previous findings, including the results of the pivotal TIMES 1 study, which reported greater mean reductions in HbA1c in elderly patients aged ≥65 years (−0.75%, 95% CI −0.96, −0.54) compared with younger patients aged <65 years (−0.70%, 95% CI −0.92, −0.49) ¹⁰ , as well as the results of an earlier post‐hoc analysis by our group ²⁴ , which found consistently greater reductions in HbA1c in elderly patients (aged ≥65 years) versus younger patients (aged <65 years) receiving imeglimin 1,000 mg b.i.d. monotherapy compared with placebo, although differences from placebo were statistically significant for both age groups at each time point assessed.

To assess whether response to imeglimin was affected by prior treatment for type 2 diabetes mellitus, treatment history (treatment naïve vs previous treatment) was also included as an exploratory variable in the multivariate logistic regression model. Our analyses showed that reductions in HbA1c were greater among patients who were treatment‐naïve compared with those who had received prior treatment for type 2 diabetes mellitus. This finding is broadly consistent with the results of the TIMES 1 study, which found that, although there was no clear difference between treatment‐naive and previously treated patients with respect to placebo‐adjusted HbA1c changes, the absolute reduction in HbA1c from baseline was greater in the treatment‐naive group compared with previously treated patients ¹⁰ . Taken together, these findings, coupled with the known mechanism of action of imeglimin ¹² , suggest that imeglimin might be more effective in treatment‐naïve patients, likely due to its ability to improve glucose‐stimulated insulin secretion ¹⁰ .

On ROC analysis, a HbA1c improvement of less than or equal to −0.3% at week 4 was found to be significantly correlated with response to imeglimin 1,000 mg b.i.d. at week 24 (P < 0.001). Based on a cutoff point of −0.3%, 37.4% (n = 67) were early responders and 62.6% (n = 112) of patients were early non‐responders. Baseline demographics and characteristics were generally similar between early responders and early non‐responders, with the exception of mean baseline HbA1c, which was higher in early responders (8.13% [SD 0.739]) versus non‐responders (7.82% [SD 0.687]), and mean (SD) duration of type 2 diabetes mellitus, which was shorter in early responders (6.22 years [SD 5.266]) versus early non‐responders (7.64 years [SD 5.702]). This is in contrast with the results of our multivariate analysis, which found that lower HbA1c was significantly correlated with response to imeglimin (P < 0.001). This is not unexpected considering that patients with HbA1c levels closer to 7.0% at baseline required less of a reduction to reach target HbA1c compared with those with higher baseline HbA1c, and were therefore more likely to be considered responders in multivariate analysis, whereas patients with higher HbA1c levels at baseline were more likely to experience greater HbA1c improvements with treatment; thus, increasing the likelihood that these patients would be early responders. When treatment responders were defined as patients with a HbA1c reduction of ≥1.0% from baseline in multivariate analysis (Table S3), factors significantly associated with response were treatment‐naïve status versus previously treated (P < 0.001), higher baseline HbA1c (P < 0.001) and lower baseline BMI (P = 0.004). Regarding the ROC analysis to determine the threshold for FPG decline at wee 4 that best predicted the outcome (HbA1c <7% at week 24), the best prediction was −21.6 mg/dL at week 4, but the AUC of the ROC curve was approximately 0.5, and there was no significant difference from a curve obtained by chance (Figure 3). Unfortunately, given these results, it is difficult to predict the threshold for response based on the degree of change in FPG.

Overall, the incidence of AEs was generally similar between imeglimin and placebo groups, and early responder and early non‐responder subgroups. Nasopharyngitis was the most common AE reported with imeglimin across subgroups in our analysis, which was consistent with the pivotal TIMES 2 and TIMES 3 phase III studies ¹³ , ¹⁵ . No other AEs occurred with a ≥5% incidence. There was a slightly higher incidence of discontinuations due to AEs in the imeglimin treatment groups compared with placebo, but no noticeable difference between early responders versus early non‐responders. Hypoglycemia and diarrhea occurred more frequently in early non‐responders (3.6% each vs 1.5% each, respectively), whereas muscle spasms, pharyngitis, abdominal discomfort and dyslipidemia occurred more frequently in early responders (3.0% each vs 0.9, 0.9, 0.9, 0%, respectively). No new safety signals were identified during our analyses, and AEs reported were consistent with the known safety profile of imeglimin ¹⁰ , ¹³ , ¹⁵ .

This was a post‐hoc analysis of randomized controlled trials, the results of which are hypothesis‐generating in nature. However, the study was subject to several potential biases, which are inherent in post‐hoc analyses and must be considered. First, the sample size of responder/non‐responder patient subgroups was relatively small, which might increase statistical variability. Second, explanatory variables included in the multivariate analysis included age, sex, treatment status (treatment naïve vs previous treatment), baseline HbA1c, duration of diabetes, baseline BMI and baseline eGFR. However, we cannot discount the possibility of other confounding factors that might have influenced the results, such as underlying comorbidities, which are present in approximately 90% of patients ²⁵ . Nevertheless, the results of our earlier analyses showed that imeglimin was consistently effective in patients with and without hypertension, dyslipidemia and hepatic fibrosis ²⁴ .

In the present post‐hoc analyses of two similarly‐designed, randomized, controlled trials, a significantly higher proportion of patients were responders to imeglimin 1,000 mg b.i.d. compared with placebo at each time point up to 24 weeks, most (>70%) of whom were sustained responders. Older age, treatment‐naïve status and lower baseline HbA1c were significantly correlated with treatment response on multivariate analysis, as was a HbA1c improvement of less than or equal to −0.3% at week 4 on ROC analysis, suggesting that these are the patient populations most likely to derive benefit from treatment. These findings build on the wealth of clinical evidence showing the efficacy and safety of imeglimin for the treatment of type 2 diabetes mellitus ¹⁰ , ¹³ , ¹⁵ , ²⁶ , and additionally identify those patients most likely to respond to imeglimin monotherapy in clinical practice.

DISCLOSURE

KH and KK (Kochi) are employees of Sumitomo Pharma Co., Ltd. HW has received honoraria for lectures for Mitsubishi Tanabe Pharma, Sumitomo Pharma, Sanwa Kagaku, Takeda Pharmaceuticals, Sanofi, Kowa, MSD, Nippon Boehringer Ingelheim, Eli Lilly, Novo Nordisk, AstraZeneca, Ono Pharmaceutical, Astellas, Kyowa Kirin, Terumo, Taisho Pharmaceutical, Abbott and Kissei Pharmaceutical, and research activities for Takeda Pharmaceuticals, Nippon Boehringer Ingelheim, Kissei Pharmaceutical, Novo Nordisk, Mitsubishi Tanabe Pharma, Lifescan Japan, Kyowa Kirin, Sumitomo Pharma, Eli Lilly Japan, Teijin Pharma, Taisho Pharmaceutical, Abbott Japan, Daiichi Sankyo, Astellas, Ono Pharmaceutical Co. Ltd., Sanofi, MSD, Soiken Inc., Sanwa Kagaku and Kowa. KK (Kaku) has been an advisor to Sanwa Kagaku, and received honoraria for lectures from Astellas, AstraZeneca, Daiichi Sankyo, Kowa, Sumitomo Pharma Co. Ltd., MSD, Ono Pharmaceutical Co. Ltd., Sanwa Kagaku, Novo Nordisk, Nippon Boehringer Ingelheim, Eli Lilly Japan, Taisho Pharmaceutical, Takeda Pharmaceuticals and Mitsubishi Tanabe Pharma, and received scholarship grants from Nippon Boehringer Ingelheim, Taisho Pharmaceutical and Kowa. KU has received honoraria for lectures from AstraZeneca, Taisho Pharmaceutical, Novo Nordisk, Sumitomo Pharma, Kowa, Mitsubishi Tanabe Pharma and Ono Pharmaceutical; research grants from Sumitomo Pharma, Novo Nordisk, Eli Lilly Japan, Sanofi, Abbott Japan, MSD and Nippon Boehringer Ingelheim; and scholarship grants from Nippon Boehringer Ingelheim, Mitsubishi Tanabe Pharma, Sumitomo Pharma, Takeda Pharmaceuticals and Sanofi. HW, KK and KU are Editorial Board members of the Journal of Diabetes Investigation and co‐authors of this manuscript. To minimize potential for bias, they were excluded from all editorial decision‐making related to the acceptance of this manuscript for publication.

This post‐hoc analysis was carried out using data from the phase IIb (JAPIC number: JapicCTI‐153086) and phase III (JAPIC number: JapicCTI‐173769) studies.

Approval of the research protocol: Studies were carried out in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines, and were approved by the appropriate institutional review boards and regulatory agencies.

Informed consent: Patients provided written informed consent to participate in each study.

Registry and the registration no. of the study/trial: N/A.

Animal studies: N/A.

Supporting information

Figure S1. | Change from baseline in HbA1c (MMRM) by baseline HbA1c.

Figure S2. | Proportion of responders and sustained responders to imeglimin 1,000 mg BID by baseline HbA1c.

Figure S3. | Change from baseline in HbA1c (MMRM) by previous treatment status.

Figure S4. | Proportion of responders and sustained responders to imeglimin 1,000 mg BID by previous treatment status.

Figure S5. | Change from baseline in FPG (MMRM).

Table S1. | Clinical characteristics of patients in the imeglimin non‐responder and imeglimin responder subgroups at baseline.

Table S2. | Clinical characteristics of patients in the imeglimin non‐sustained responder and imeglimin sustained responder subgroups at baseline.

Table S3. | Multivariate logistic regression analysis evaluating factors associated with treatment response to imeglimin after 24 weeks.

Table S4. | Clinical characteristics of patients in the placebo, early responder, and early non‐responder subgroups at baseline.

Table S5. | Reasons for treatment discontinuation in the placebo, early responder, and early non‐responder subgroups.

Table S6. | Common (≥2% incidence) TEAEs in the placebo, early responder, and early non‐responder subgroups.