Effectiveness, safety, initial optimal dose, and optimal maintenance dose range of basal insulin regimens for type 2 diabetes: A systematic review with meta‐analysis

Abstract Aims To investigate the effectiveness, safety, optimal starting dose, optimal maintenance dose range, and target fasting plasma glucose of five basal insulins in insulin‐naïve patients with type 2 diabetes mellitus. Methods MEDLINE, EMBASE, Web of Science, and the Cochrane Library were searched from January 2000 to February 2022. The Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines were followed and the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) approach was adopted. The registration ID is CRD42022319078 in PROSPERO. Results Among 11 163 citations retrieved, 35 publications met the planned criteria. From meta‐analyses and network meta‐analyses, we found that when injecting basal insulin regimens at bedtime, the optimal choice in order of most to least effective might be glargine U‐300 or degludec U‐100, glargine U‐100 or detemir, followed by neutral protamine hagedorn (NPH). Injecting glargine U‐100 in the morning may be more effective (ie, more patients archiving glycated hemoglobin < 7.0%) and lead to fewer hypoglycemic events than injecting it at bedtime. The optimal starting dose for the initiation of any basal insulins can be 0.10–0.20 U/kg/day. There is no eligible evidence to investigate the optimal maintenance dose for basal insulins. Conclusions The five basal insulins are effective for the target population. Glargine U‐300, degludec U‐100, glargine U‐100, and detemir lead to fewer hypoglycemic events than NPH without compromising glycemic control.


| INTRODUCTION
Type 2 diabetes mellitus (T2DM) is a disease associated with a state of chronic hyperglycemia that results in a significant increase in the risk of microvascular and macrovascular complications, and eventually, leads to diabetesrelated mortality. 1 About 537 million individuals globally were affected by diabetes mellitus in 2021, and T2DM accounted for over 90% of these cases. 2 Thus, appropriately controlling glucose and at the same time avoiding hypoglycemia is essential for patients with T2DM, as it is linked to their quality of life in the long run. When hyperglycemia cannot be properly controlled, injectable glucose-lowering therapy such as glucagon-like peptide-1 receptor agonist or basal insulin is added. To date, there are five common basal insulins available for use: glargine U-300, degludec U-100, glargine U-100, detemir, and neutral protamine hagedorn (NPH) insulin. After a literature search for English publications, there is no systematicreview-based clinical practice guideline to date on these five basal insulin regimens for adult T2DM insulin-naïve patients with inadequately controlled glucose (ie, glycated hemoglobin [HbA1c] > 7.0%) treated with one or more oral glucose-lowering drugs in the Asian-Pacific region. Therefore, the authors aim to fill this knowledge gap by focusing on answering the following questions: Q1. What are the differences in the effectiveness and safety among five basal insulin regimens after the initiation of insulin therapy in the target population?
Q2. What are the optimal starting dose (U/kg/day) and time of administration (morning vs. bedtime) of each basal insulin?
Q3. Among the target population, what is the optimal maintenance dose (U/kg/day) that achieves target fasting plasma glucose (FPG)?
Q4. After initiation of any of the five basal insulins, what range of target FPG can lead to the ideal HbA1c level (ie, <7.0%) in the target population?
After a quick PubMed search, we found three relevant systematic reviews. [3][4][5] However, none of them answered all four questions. Therefore, a new systematic review was worthwhile.

| Search strategy and selection criteria
This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting checklist. 6 MEDLINE, EMBASE, Web of Science, and the Cochrane Library were searched from January 2000 to February 2022. Search terms included the names of the five basal insulins, T2DM, randomized controlled trials (RCTs), and combinations with their alternatives, respectively (Supplementary Table S1). One reviewer screened the references retrieved. A second reviewer audited and resolved any disagreements with a third reviewer. The protocol was registered in PROS-PERO (CRD42022319078).

| Inclusion Criteria
The study recruited adult insulin-naïve patients with T2DM who required basal insulin therapy and had ≥1 oral glucose lowering drug; was an RCT; had a treatment duration of ≥12 weeks; reported any of the following outcomes: HbA1c, incidence of hypoglycemia, incidence of severe hypoglycemia, incidence of nocturnal hypoglycemia, time in range, FPG, cost-effectiveness, weight gain, patient-reported outcomes (eg, quality of life); and was published after 1 January 2019 for a conference abstract.

| Exclusion Criteria
The study focused on pregnant women, investigated basal insulin that was initiated twice daily, was published in a language other than English, or was a non-RCT.

| Data analysis
One reviewer extracted data, which was subsequently audited by an independent auditor. The risk of bias per outcome for each included study was assessed using the Cochrane Collaboration Risk of Bias 2.0 tools. 7 We used RevMan 5.3 to analyze the data. For binary outcomes, we calculated the risk ratio (RR) and its 95% confidence interval (CI). For continuous outcomes, where possible, we calculated mean difference (MD) with its 95% CI. When clinically and methodologically homogeneous results from two or more studies were available, a meta-analysis was conducted. We employed I 2 > 50% as a general guide to identify statistical heterogeneity in the pooled analysis.
The certainty of the evidence per outcome for each comparison, considering the risk of bias, inconsistency, indirectness, imprecision, and publication bias, was assessed using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach. 8 3 | RESULTS

| Risk of bias assessment and evidence certainty
Results of the risk of bias assessment for the 35 studies are shown in Supplementary Table S2. Overall, the risk of bias ranged from "some concerns" to "high risk" in the analysis in Q1 and Q4, and "some concerns" in Q2. The certainty of evidence for each comparison of interventions was high to low after considering four other factors from the GRADE approach (Supplementary Table S3). The clinical thresholds defined as small, moderate, and large effects for relevant outcomes were made by the working group members based on their clinical experience for judgment of imprecision of certainty of evidence and balancing the effect magnitudes of desirable and undesirable outcomes based on the GRADE approach (listed in Supplementary  Table S4). 44
(3) one basal insulin injected at morning time versus another basal insulin injected at bedtime; and (4) one basal insulin injected at morning time versus the same insulin injected at bedtime. All results of the relative and absolute effects between any comparison are shown in Appendix Table S3   37 (fewer per 1000, 95% CI 74 fewer-13 fewer); and hypoglycemia <3.9 mmol/L at 3 months (71 fewer/1000, 95% CI 125 fewerÀ5 fewer). But nocturnal hypoglycemia <3.9 mmol/L at 3 months (36 more/1000, 95% CI 11 few-erÀ100 more) favored degludec U-100. Except for these outcomes, the results of other critical outcomes at three or 6 months are similar between two insulins, such as HbA1c change and nocturnal hypoglycemia <3.0 mmol/L at 3 months; severe hypoglycemia, nocturnal hypoglycemia <3.0 mmol/L, and nocturnal hypoglycemia <3.9 mmol/L at 6 months, etc. Subgroup analyses of the BRIGHT trial by age (<65 years and ≥ 65 years) showed consistent results as described, 15 whereas subgroup analyses by renal function (<60, ≥60 but <90, ≥90 mL/min/1.73 m 2 ) showed heterogeneity of treatment effects. 16 For patients with estimated glomerular filtration rate (eGFR)≥90 mL/min/1.73 m 2 , the results from critical outcomes were consistent with the overall results in the BRIGHT trial. Degludec U-100 induced less harm in patients with eGFR 60-90 mL/ min/1.73 m 2 (n = 365) on nocturnal hypoglycemia <3.0 mmol/L at 6 months (28 more/1000, 95% CI 12 few-erÀ125 more), and nocturnal hypoglycemia <3.9 mmol/L at 6 months (64 more/1000, 95% CI 24 fewerÀ186 more) with similar results of HbA1c. The data from patients with eGFR <60 mL/min/1.73 m 2 indicated that the incidence of nocturnal hypoglycemia slightly favored degludec U-100 but the results of event rate/patient-year for nocturnal hypoglycemia slightly favored glargine U-300, and HbA1c change at 6 months was 0.43% lower (95% CI 0.74% lower to 0.12% lower) favored glargine U-300.
The results of direct head-to-head comparison among five basal insulins are summarized in Table 4.

| Network meta-analyses
We performed network meta-analyses for the critical outcomes. The statistically significant results are shown in Table 5. Six comparisons showed statistically significant results (ie, p value ≤.05) that are consistent with the results from the direct paired comparison in Supplementary Table S3.

| Detemir
One RCT with 60 patients reported that, compared with twice daily, injecting detemir 0.12 U/kg/day once daily We defined the small sample size as <100.

| DISCUSSION
As demonstrated by the synthesized results of a direct comparison between the five basal insulins in Table 4 and the network meta-analyses in Table 5, we found that when injecting basal insulin at bedtime, the optimal choice in descending order might be glargine U-300 or degludec U-100, glargine U-100 or detemir, and lastly NPH. Injecting glargine U-100 in the morning may be more effective and lead to fewer hypoglycemic events than injecting it at bedtime. However, future high-quality research is needed to confirm these because of the low quality of the evidence. The current evidence shows the starting dose for initiation of any of the five basal insulins to be from 0.10 U/kg/day to 0.20 U/kg/day according to the individual patient's characteristics, such as age, weight, morbidities, kidney function, etc. It appears that an FPG range of 3.9-6.1 mmol/L for any basal insulin may be an acceptable range for achieving a target HbA1c level of <7%. However, for individuals at high risk of hypoglycemia, such as with serious diseases that affect life, severe hypoglycemia history, acute cerebrovascular disease, or severe chronic renal failure leading increased risk of hypoglycemia, the target range of FPG should be higher.
This study is the first systematic review that defined thresholds for outcomes that correspond to trivial/none, small, moderate, or large effects and used them to rate the imprecision domain when assessing quality of evidence and balancing the magnitudes of the desirable and undesirable outcomes' effects according to the GRADE approach in the diabetes community. Thus, this paper provides the references for future researchers to set up clinical thresholds for these outcomes in their own research according to their individual experiences and different contexts.
It should be noted that statistically significant and clinically significant findings are different. 45 For example, when degludec U-100 was compared with glargine U-100, there were 41 fewer patients/1000 who had hypoglycemic events (<3.0 mmol/L) at 6 months (95% CI, 89 few-erÀ11 more) in the degludec U-100 group, which showed there was no statistical significance. However, the point estimate of 41 patients is higher than our clinical threshold of the small effect (20 patients/1000 in Supplementary  Table S4). The upper boundary of 95% CI was 11 patients, which did not reach the small clinical threshold (20 patients/1000), and the lower boundary of 95% CI was higher than our clinical threshold of a large effect (>80 patients/1000) to favor degludec U-100. Hence, we can consider that the result may be clinically significant favoring degludec U-100. Setting up clinical thresholds instead of only considering statistical significance is very important in conducting systematic reviews, as it helps clinicians make appropriate clinical decisions. This point is emphasized in the 2022 version of the Cochrane Handbook for Systematic Reviews of Interventions. 46 This systematic review has some limitations. First, the literature search was restricted to English language publications, which can potentially lead to missing references published in other languages. Second, we included only RCTs. There might be moderate-quality nonrandomized studies beyond our search that could potentially answer some of our research questions.
Among the included studies, there is no subgroup analysis by ethnicity, which should be considered by investigators in their future research. Finally, individualized patient care is the key in clinical practice, and so, treatment plans should always be discussed in consideration of the individual patient's values and preferences.
AUTHOR CONTRIBUTIONS Linong Ji, Yingying Luo, Jun Xia, and Xiaomei Yao conceived and designed this study. Zhan Zhao and Chenchen Xu conducted the database search and reviewed the reference lists of articles included in screening. Zhan Zhao and Chenchen Xuperformed initial screening and review of full texts for eligibility. Zhan Zhao and Chenchen Xu extracted the data and completed quality assessment. Xiaomei Yao resolved any conflicts in quality assessment. Yaping Chang and Zhan Zhao prepared the tables and figures, and conducted the data analysis. Yingying Luo, Jun Xia, Xiaomei Yao, and Linong Ji conducted data interpretation and drafted the first draft of the manuscript. All authors approved the project plan, and reviewed and revised the final manuscript before submission. in scientific meetings from Servier, Boehringer Ingelheim and Eisai. Soo Lim received research funding from MSD and CKD, and honoraria for lectures from Novo Nordisk, Sanofi, Boehringer Ingelheim, AstraZeneca, and MSD. Linong Ji has received consulting and lecture fees from Eli Lilly, Novo Nordisk, Merck, Bayer, Sanofi-Aventis, Roche, MSD, Metronics AstraZeneca, Boehinger Ingelheim, and Abbott. Other authors declare no competing interest.

DATA AVAILABILITY STATEMENT
Most of the systematic review data are available in the supplementary materials. Additional requests can be provided by contacting authors.