Introduction
Type 2 diabetes mellitus (T2DM) is a progressive disease often requiring add-on therapy, as many patients remain at suboptimal glycemic control on oral antidiabetic drugs (OADs) alone [
1,
2]. The addition of basal insulin is recommended by guidelines as one of the initial treatment intensification options in these patients [
3‐
6]. Despite recommendations for timely initiation, data indicate that Asian patients with T2DM have suboptimal glycemic control for approximately 6–9 years, with average glycated hemoglobin (HbA
1c) levels of 9.2–10.5% at the point of basal insulin initiation [
7‐
9]. In Asian patients, delays in basal insulin initiation are associated with a lower reduction in HbA
1c, as well as a lower proportion of patients being able to achieve HbA
1c < 7% [
7].
In addition to the issues surrounding timely initiation of basal insulin in Asia, there are several challenges pertaining to the titration of basal insulin. In clinical practice, Asian patients with T2DM who have a lower body mass index (BMI) than their Caucasian counterparts [
10‐
12] are often perceived to be at an increased risk of hypoglycemia, which leads to conservative treatment goals and a cautious approach to dose titration by physicians [
9,
13]. Large, real-world studies conducted in Asia, such as The First Basal Insulin Evaluation (FINE) Asia registry study and the Observational Registry of Basal Insulin Treatment (ORBIT) study, have hypothesized that further active titration of basal insulin can potentially increase the proportion of patients with T2DM achieving adequate glycemic control and improve glycemic outcomes [
9,
14].
There is a growing body of evidence describing the importance of individualizing the management of T2DM based on ethnicity [
15‐
17]. Therefore, when titrating the dose of basal insulin, it is important to consider the ethnic and genetic differences between Asians and Caucasians [
18,
19], which may lead to different insulin requirements in both of these populations [
20]. Understanding the specific insulin needs of Asian populations requires greater understanding of the patient-level factors that impact their required dose of basal insulin and treatment response.
The current study is a pooled analysis of Asian patients included in treat-to-target (TTT) randomized controlled trials (RCTs) of insulin glargine 100 U/mL (Gla-100). It was designed to investigate the relationship between patient baseline covariates, basal insulin dose, and treatment outcomes in insulin-naïve Asian patients with T2DM who have suboptimal glycemic control with OADs. To further explore the relationship between patient characteristics and outcomes, separate analyses were performed for the main population of patients from Asia and for a subpopulation of patients from China. The aim was not to compare both populations, but rather to lay the foundation for future analyses specific to the Chinese population and provide information to guide local clinical practice.
Discussion
Using data from seven published individual TTT RCTs [
1,
21‐
27], this pooled analysis aimed to determine the impact of patient baseline covariates on the dose of Gla-100 and its treatment response at week 24 in an overall population of Asian patients with T2DM and in a subpopulation of these patients from China. Identification of the predictors of dose and treatment response in the Chinese T2DM subpopulation was important to lay the foundation for future research specific to these individuals and to help guide local clinical practice. Results from the Chinese subpopulation were not intended as a comparison with the results obtained in the overall Asian population. For the purpose of this discussion, common baseline predictors for Asian patients and those from China have been elaborated on. Where deemed to be clinically relevant, predictors in the overall Asian population have also been highlighted. However, differences in observed results between the main population and subpopulation may have been affected by several factors, including the number of patients in each group. Further research and clinical validation of the results are therefore required to understand these differences and so are not discussed here.
The dose of Gla-100 at week 24 was negatively correlated with age and positively correlated with BMI and FPG in both the overall Asian population and the Chinese subpopulation. A post hoc analysis of the insulin glargine (Lantus
®) vs. insulin detemir (Levemir
®) Treat-To-Target (L2T3) study, one of the RCTs included in this analysis, also demonstrated a similar relationship between final basal insulin dose (glargine or detemir) and the baseline characteristics of age, BMI, and FPG in patients with T2DM [
28]. In the post hoc analysis, two stepwise regression analyses were performed, the first of which demonstrated that statistically significant predictors of a high final basal insulin dose, amongst others, were high FPG and younger age [
28]. The second regression analysis, in which only physical characteristics that could be assessed “at the bedside” were included, demonstrated that high BMI, high FPG, and younger age were predictors of a high final basal insulin dose [
28]. The study acknowledged that rough estimations of the final basal insulin dose can be made when taking into account a patient’s BMI and age [
28].
Similar to the primary outcome, the dose of Gla-100 expressed in units per kilogram per day at week 24 was also shown to be negatively correlated with age and positively correlated with BMI and FPG in the Asian population. Post hoc analysis of the L2T3 study also reported a positive correlation between BMI and the final basal insulin dose expressed in terms of weight; however, data supporting this outcome were not shown [
28].
These results suggest that a patient with higher FPG at baseline requires a higher dose of Gla-100 to reduce their FPG level, which is in accordance with clinical expectations. The relationship between BMI and dose of Gla-100 has a physiological basis. Higher BMI has been shown to be associated with decreased insulin sensitivity [
29,
30], and a higher dose may be required in patients with higher BMI in order to stimulate insulin absorption and decrease glycemic levels. Indeed, titration based on a patient’s weight (an important factor in determining BMI) is thought to have a stronger physiological basis than glucose level or dose-based regimens, especially given the relationship between body weight and insulin sensitivity [
31]. As mentioned above, the post hoc analysis of the L2T3 study also demonstrated a negative correlation between age and the final dose of basal insulin; however, the authors of the study acknowledged that there are no straightforward explanations for this finding [
28]. The authors of the current study hypothesize that younger patients, who are in need of stricter glycemic control [
4,
5,
32] and who have been shown to be at a lower risk of hypoglycemia [
33‐
35], are likely to have had their doses of Gla-100 titrated to a greater extent than older patients, leading to the observed relationship between the dose of Gla-100 at week 24 and age.
The above analyses indicate that a higher dose of basal insulin is likely to be required in younger patients or patients with higher baseline BMI or FPG, and physicians must therefore titrate the insulin doses of these patients to an adequate level to achieve glycemic targets. The opposite must also be taken into consideration in older patients or patients with lower baseline BMI or FPG, for whom lower doses of basal insulin may be required to achieve glycemic targets. In these patients, a more cautious approach to the titration of basal insulin is warranted.
The nomogram developed in this study can help clinicians to determine the dose of Gla-100 likely to be required by a patient to achieve target glycemic control at week 24, thereby providing guidance for the appropriate titration of Gla-100 according to a TTT algorithm. Further details on interpretation of the nomogram are included in Fig.
2. The nomogram developed in this study is a representation of the results of the multivariate analysis of the primary outcome as a whole and includes all the patient baseline covariates included in the analysis, regardless of whether they were significant or not. Therefore, although not significant, baseline HbA
1c, duration of diabetes, and gender have also been included in the nomogram. With regard to baseline HbA
1c, post hoc analysis of the L2T3 study demonstrated similar results, with multivariate analyses demonstrating that the baseline level of HbA
1c was not a significant determinant of the final basal insulin dose [
28]. It should be noted that the dose of Gla-100 predicted by the nomogram developed in this study is only an estimation of the required dose to achieve target glycemic control at week 24 and does not enable calculation of the initial therapeutic dose of Gla-100. In general, Asian country-specific guidelines recommend a dose of 0.1–0.3 U/kg/day for the initiation of basal insulin [
6,
36‐
38].
A formula for determining the total optimal daily dose of Gla-100 at 24 weeks, based on both the optimal starting and incremental doses, has been previously developed using data from a 24-week observational study of Japanese patients with T2DM [
39]. The objective and parameters included in the formula differ, however, from those of the nomogram developed in the present analysis; furthermore, the formula was derived from observational data from a single study wherein physicians were free to titrate according to their clinical practice [
39]. The nomogram in the current study was developed using data from seven RCTs, each with its own starting doses and TTT algorithms based on FPG [
1,
21‐
27]. The patient populations used in both analyses also differ; the aforementioned formula was derived from a population that achieved target HbA
1c [
39], whereas the nomogram in the current study was constructed using data from the FAS populations of the individual TTT RCTs. Hence, direct comparisons between the formula and the nomogram developed in the current study cannot be made.
The multivariate analyses demonstrated a statistically significant positive correlation between HbA
1c value at baseline and HbA
1c value at week 24 in both the Asian population and the Chinese subpopulation. In a study by Fujita et al. [
40], multiple linear regression analysis was conducted to determine the characteristics influencing the effectiveness of Gla-100 treatment in insulin-naïve T2DM Japanese patients with suboptimal glycemic control on OADs. HbA
1c at baseline was shown to be a statistically significant predictor of HbA
1c at week 24 (
P = 0.006), with an increase at baseline predicting a higher endpoint HbA
1c value [
40].
In the present study, multivariate analyses also demonstrated a negative association between achieving target glycemic control (HbA
1c < 7%) at week 24 and baseline HbA
1c in the Asian population and Chinese subpopulation. The negative association between achieving target glycemic control and baseline HbA
1c has been reported in several studies aimed at identifying the characteristics associated with glycemic response to newly initiated insulin therapy in both Asian and Western populations with T2DM [
41‐
43]. In particular, a subject-level meta-analysis of 12 RCTs that used Gla-100 in a TTT titration regimen in patients with T2DM, including some of the RCTs involved in the current pooled analysis, demonstrated that baseline HbA
1c was negatively associated with achieving HbA
1c ≤ 7.0% [
43]. In an observational study of newly initiated insulin therapy in patients with T2DM by Nichols et al. [
42], multivariate analyses demonstrated that HbA
1c prior to insulin initiation was the dominant factor in predicting treatment goal attainment (HbA
1c < 7%), and that a 1% increase in HbA
1c prior to insulin initiation reduced the probability of attaining target glycemic control by 26% [
42]. In the current analysis, a 1% increase in HbA
1c prior to initiation of Gla-100 reduced the probability of attaining target glycemic control by 63.1% in the Asian population and by 43.7% in the Chinese population.
In addition to baseline HbA
1c, duration of diabetes was also negatively associated with achieving target glycemic control at week 24 in Asian patients. The aforementioned studies by Fujita et al. [
40] and Nichols et al. [
42] have also reported that a longer duration of diabetes is associated with a lower likelihood of achieving target glycemic control.
Validating the above, multivariate analyses conducted in this study demonstrated a negative correlation between HbA
1c response (reduction from baseline to week 24) and baseline HbA
1c in both the Asian population and the Chinese subpopulation. Predictors of change in HbA
1c were also investigated using data from the large, 24-week, observational A
1chieve study, which involved patients with T2DM initiating insulin therapy. In both predictor and explanatory analyses, HbA
1c level at baseline was negatively associated with change in HbA
1c from baseline to the endpoint [
44].
This above evidence indicates that baseline HbA
1c is an important factor in determining the level of glycemic control attained. In the study by Nichols et al. [
42], HbA
1c prior to starting insulin therapy accounted for 95% of the discriminatory ability to predict the probability of attaining target glycemic control and 96% of the explainable variance in HbA
1c change. These observations are expected, since a patient with HbA
1c closer to 7% should achieve target glycemic control more easily following treatment initiation with insulin. This, however, does not eliminate the fact that in several studies, patients with T2DM who achieved target glycemic control also had a greater reduction in HbA
1c after initiating insulin therapy, despite a lower mean HbA
1c at baseline [
41,
42].
Baseline predictors for FPG value at week 24 were also investigated in our study. To the best of our knowledge, other studies demonstrating similar results have not been conducted, and hence these novel findings require further investigation. Common baseline predictors in the Asian population and Chinese subpopulation were BMI and FPG, both being positively associated with FPG at week 24.
There are several limitations to this study. No methods were used to assess risk of bias within the individual studies or across the included studies. Additionally, selection of the patient baseline covariates included in the analyses was limited by the available data in the individual RCTs. This pooled analysis, however, provides results from a larger group of Asian and Chinese patients compared with each of the individual seven studies. Additionally, this study performed separate analyses for patients from Asia and for the subpopulation of patients from China, in order to explore the relationship between patient characteristics and outcomes in each of the populations. Understanding the reasons behind differences between the populations’ results requires further research involving larger patient populations and validation of these results. The nomogram developed in this study requires further independent clinical validation. Future studies could also look to develop and assess nomograms based on a wider range of patient characteristics, such as the presence of diabetes complications or comorbidities. Additional research involving a larger pool of baseline covariates, including threshold homeostasis model assessment (HOMA) of insulin resistance levels, HOMA of β-cell function levels, and C-peptide may provide more insight into the impact of insulin resistance and secretion capacity on the dose of Gla-100 and treatment response. Investigating different patient profiles could also explain and establish more in-depth and clinically relevant correlations, such as those associated with the risk of occurrence of hypoglycemia.