Insulin and Glucagon-Like Peptide 1 Receptor Agonist Combination Therapy in Type 2 Diabetes: A Systematic Review and Meta-analysis of Randomized Controlled Trials

Current management of hyperglycemia in type 2 diabetes claims that in patients with inadequate glycemic control on initial metformin monotherapy, an escalation to a two-drug combination, and subsequently, to a three-drug combination is indicated (1). However, because of the progressive nature of type 2 diabetes, many patients eventually require insulin therapy, usually started with a long-acting (basal) formulation at bedtime (2). When basal insulin has been titrated to an acceptable fasting blood glucose but HbA_1c remains above target, therapy may be intensified by the addition of a mealtime insulin (one to three injections of a short-acting analog at meals) or by changing to premixed insulin formulations (3). These intensive insulin regimens (three to four insulin injections daily) are associated with a higher risk of hypoglycemia and weight gain, which in turn may undermine the achievement of glycemic targets (4).

The American Diabetes Association's Standards of Medical Care in Diabetes—2016 (2) reviews the approach to starting and adjusting insulin in type 2 diabetes: at the level of two injections daily, a newer treatment option, consisting in the combination of basal insulin plus a glucagon-like peptide 1 receptor agonist (GLP-1RA), is offered as a trial. This combination shows complementary modes of action in the treatment of type 2 diabetes, with potential benefits on glycemic control and metabolic profile. The meta-analysis of Eng et al. (5) is quoted to support the option of adding a GLP-1RA to ongoing insulin treatment because this combination yielded an improved mean reduction in HbA_1c of 0.44% compared with other antidiabetes treatments and was associated with no increased hypoglycemia and reduced weight. However, most of the trials compared the efficacy of the combination versus placebo, which is far from the real world of diabetes therapy. Other trials compared the efficacy of the combination versus up-titration of basal insulin, which is unable to cover postprandial glucose excursions.

Before this combination therapy can be offered to clinicians as the standard of care for patients with type 2 diabetes, other questions must be answered that can help translate this strategy into clinical practice. Recognizing that reports of many trials with different designs, interventions, and study groups are now available, we planned a systematic review and meta-analysis of randomized controlled trials (RCTs) that evaluated the role of GLP-1RA and insulin combination in the injectable treatment of patients with type 2 diabetes. In particular, we sought to answer the following questions:

1. Is this combination therapy similarly effective in reducing baseline HbA_1c when compared with basal-plus (adding one main-meal fast-acting insulin to basal insulin once daily) or full basal-bolus (four insulin injections daily) insulin regimens?

2. Which is the role of the fixed-ratio combinations of GLP-1RA and basal insulin in this scenario?

3. Are there additional benefits of the combination therapy on other aspects of the diabetes control, including percentage of patients at HbA_1c target, incidence of hypoglycemia, and weight change?

This systematic review was conducted in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) guidelines (6). The PRISMA checklist and the protocol of this study are provided in the Supplementary Data.

Bibliographical databases for the literature search included PubMed, MEDLINE, Cochrane Central Register of Controlled Trials, and ClinicalTrials.gov (http://www.clinicaltrials.gov). The last search was performed on 7 November 2016. Our search strategy included the keywords ‘‘type 2 diabetes’’ or “type II diabetes”; ‘‘hemoglobin A_1c’’ or “HbA_1c” or “glucose”; ‘‘biphasic insulin regimen’’ or “basal-plus insulin regimen” or ‘‘basal-bolus insulin regimen’’; ‘‘glargine’’ or ‘‘detemir’’ or “degludec” or ‘‘neutral protamine lispro’’ or ‘‘lispro’’ or ‘‘aspart’’ or ‘‘glulisine’’; “glucagon-like peptide 1 receptor agonist” or “GLP-1” or “exenatide” or “ liraglutide” or “lixisenatide” or “dulaglutide” or “albiglutide” or “semaglutide”; ‘‘randomized” or “trial.” A Google search was also conducted to find information on RCTs that was unavailable from bibliographical databases. Reference lists of prior reviews and meta-analyses were also manually searched to track relevant RCTs that were not indexed by normal keywords.

We selected studies if they were RCTs performed in adults with type 2 diabetes, compared short- and long-acting GLP-1RAs administered in association with insulin treatment to another injectable treatment strategy, with at least 12 weeks’ duration of intervention, and assessed variation in HbA_1c and/or the proportion of participants with HbA_1c of ≤7.0% (53 mmol/mol) at the end of the study period or the number of participants with hypoglycemic events or weight change.

Two investigators (M.I.M., D.G.) used a standardized tool to independently abstract all data, and disagreements were resolved by consensus. Any RCTs that met the inclusion/exclusion criteria were included in the analysis. We extracted the following data from each selected study: 1) author identification and year of publication; 2) duration of intervention; 3) investigational drug with number of patients; 4) comparator drug with number of patients; 5) background therapy in both study groups; 6) HbA_1c outcome; 7) baseline HbA_1c; 8) study funder; and 9) type of statistical analysis for the HbA_1c outcome. The relevance of studies was assessed with a hierarchical approach on the basis of title, abstract, and the full article. After the initial screening of titles and abstracts, the studies included by both reviewers were compared, and disagreement was resolved by consensus.

We evaluated the risk of bias of the included RCTs according to the Cochrane Collaboration’s tool for assessing the risk of bias (7). We assessed risk of bias in random sequence generation and allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), and selective reporting (reporting bias). The risks of bias were categorized a high, low, and unclear. Methodologic quality was also scored using criteria set out by Jadad et al. (8). This 5-point quality scale includes points for randomization (1 point; using tables of random numbers or computer-generated randomization, additional 1 point), double-blind (1 point; use of placebo, additional 1 point), and follow-up (stating the numbers and reasons for withdrawal in each group; 1 point) in the report of an RCT. We gave an additional point if the analysis was by intention to treat. We considered a score of ≥4 as good quality. The quality of each RCT was assessed by one reviewer and verified by another reviewer. Disagreement was resolved by consensus.

The decrease of HbA_1c from baseline at the end of treatment was the primary outcome of this meta-analysis. Secondary end points were proportion of patients at HbA_1c target ≤7% (53 mmol/mol), incidence of hypoglycemic events, and weight change. Changes from baseline in HbA_1c and body weight were analyzed as continuous variables, and weighted mean differences (WMDs) were used as the summary measure. The difference between the two groups of the mean decrease of the continuous variable and its SD was extracted from each study. If not reported, the SD of the difference was estimated by standard equations from the reported standard error, CI, or P value. Risk ratios (RRs) were used as the meta-analytic measure of association for patients at HbA_1c targets <7.0% (53 mmol/mol) and for incidence of hypoglycemic events. For each study, the proportion of participants achieving an HbA_1c of ≤7.0% (53 mmol/mol) and those having any episode of hypoglycemia were used to calculate RR using 2 × 2 table. For studies with three groups and a shared intervention group, to include each pairwise comparison separately, the shared group was split into two groups with the sample size halved. Heterogeneity between studies was assessed by using the Q statistic and I², which is the proportion of total variance observed between the trials attributed to the differences between trials rather than to sampling error. I² < 25% was considered as low in heterogeneity, I² > 75% as high in heterogeneity, and a Q statistic P value of <0.10 was considered significant (9). If overall heterogeneity was significant, a random-effect model was used; otherwise, a fixed-effect model was used. In a conservative way, we considered random-effect analysis the main focus of our meta-analysis. We did preplanned subgroup analyses restricted to trials that compared GLP-1RA + insulin versus insulin or GLP-1RA + placebo; GLP-1RA + insulin versus insulin up-titration; GLP-1RA + insulin versus basal-plus or basal-bolus insulin regimens; GLP-1RA + insulin in fixed-ratio combination versus single component; and short- versus long-acting GLP-1RAs. Publication bias was assessed visually with funnel plots and with the Egger test (10); a P value of <0.10 was considered significant. The trim-and-fill method was used to estimate the effect of publication bias (11). Sensitivity analyses in which each study was removed in turn to assess the influence of that study on the overall effect size were also performed. For descriptive purposes, the median and interquartile range (IQR) was calculated for continuous variables in the two groups. Data were analyzed using Stata 11.2 software (StataCorp LP, College Station, TX). All statistical tests were two sided, and P values of <0.05 were regarded as significant.

The initial search assessed 3,174 citations, and 2,202 abstracts were selected for review after duplicates were removed. By screening the abstracts, we excluded 2,169 citations (observational studies, non-RCTs, mostly review articles, nonhuman studies) (Fig. 1). Of the remaining 33 RCTs, 7 were excluded because they evaluated type 1 diabetes (2 studies), were extensions of RCTs (2 studies), compared 2 basal insulin regimens (1 study), or did not include insulin-treated patients (2 studies) (Supplementary Table 1). Finally, 26 RCTs (12–37) with 30 comparisons were included for quantitative synthesis and meta-analysis.

The characteristics of the included 26 RCTs are summarized in Table 1. The participants in all RCTs were patients with type 2 diabetes (>18 years old). Most trials were multinational and sponsored by industry (11 trials by Novo Nordisk, 3 trials by Eli Lilly, 8 trials by Sanofi, and 1 trial by GlaxoSmithKline). The trials were published between 2011 and 2016, with 9 studies (19–21,26,31,34–37) published in 2016. All trials were of parallel-group design, 10 were double-blind (12,14–21,32), and the remaining were of open-label design. Trial duration ranged from 12 to 52 weeks.

The trials had different designs: 10 trials (12–21) compared GLP-1RA with placebo on a background of basal insulin therapy, 3 RCTs (22–24) compared GLP-1RA with intensification of insulin on a background of different insulin regimens, 2 RCTs (25,26) compared GLP-1RA and basal insulin versus a basal-plus insulin regimen, 5 trials (27–31), with 7 comparisons, compared GLP-1RA and basal or prandial insulin versus full basal-bolus insulin regimen (1 injection of basal insulin, for most a long-acting insulin analog, plus 3 injections of a short-acting insulin analog at meals), and 6 trials (32–37), with 8 comparisons, compared fixed-ratio GLP-1RA and insulin combination versus insulin intensification (32–37) or GLP-1RA alone (33,37).

The trials evaluated 11,425 patients for the primary end point of this meta-analysis (HbA_1c change at the end of the trial), 5,689 in the intervention groups and 5,736 in the comparator groups (range of patients in each arm, 21–834) (Table 1). The baseline HbA_1c level was identical between groups, with a median of 8.3% (67 mmol/mol) in the GLP-1RA and insulin groups (IQR 7.7–8.5% [60.7–69.4 mmol/mol]), and 8.1% (65 mmol/mol) in the comparator groups (IQR 7.5–8.5% [58.5–69.4 mmol/mol]). Four trials (12,13,15,20) tested the superiority of the combination (GLP-1RA and insulin) over placebo, six trials (14,16–19,21) compared the combination with placebo, three trials (22–24) compared the combination with insulin intensification, three trials (26,29,30) tested the noninferiority of the combination, and four trials (25,27,28,31) compared the combination with intensified insulin regimens (basal-plus or basal-bolus). One trial (32) tested the superiority and five trials (33–37) tested the noninferiority of the fixed-ratio GLP-1RA and insulin combination over basal insulin intensification or GLP-1RA alone.

According to the Cochrane Collaboration’s tool for assessing risk of bias, the three common biases were blinding of participants (16 of 26 trials, performance bias, high risk of bias), blinding of outcome assessment (16 of 26 trials, detection bias, high risk of bias), and allocation concealment (all trials, selection bias, unclear risk of bias) (Supplementary Fig. 1 and Supplementary Table 2). The median score of methodologic quality was 3.0 (IQR 3.0–5.0), and many studies (n = 12) had a quality score ≥4, indicating high quality (Supplementary Table 3).

In the overall analysis of 30 comparisons, the combination of GLP-1RA and insulin led to a mean HbA_1c decrease significantly greater than comparator groups (−0.47%, 95% CI −0.59 to −0.35%, P < 0.001), with high heterogeneity between studies (I² = 93.7%, P < 0.001) (Fig. 2 and Table 2). There was no evidence of publication bias (Egger test, P = 0.138). In sensitivity analyses, the removal of each study in turn did not change the overall effect size. Preplanned subgroup analyses showed that most of the estimate favoring the combination of GLP-1RA and insulin over comparators derived from the comparison with placebo/insulin titration: in the 10 comparisons versus placebo, the difference for HbA_1c was greatest (−0.76%, 95% CI −0.96 to −0.57, P < 0.001) (Fig. 2 and Table 2). There was no difference in the mean overall HbA_1c reduction between GLP-1RA and basal insulin versus full basal-bolus (seven comparisons) or basal-plus (two comparisons) insulin regimens. The comparison of fixed-ratio GLP-1RA and basal insulin combination with intensification of single therapies showed a mean HbA_1c reduction in favor of the fixed-ratio combination (eight comparisons, WMD = −0.56%, 95% CI −0.72 to −0.40). The final (median) HbA_1c levels achieved with both combinations (insulin degludec and liraglutide [IDegLira] or of glargine and lixisenatide [IGlarLixi]) were 6.5% (48 mmol/mol) (IQR 6.4–6.7% [47–50 mmol/mol]) and 6.5% (48 mmol/mol) (IQR 6.4–6.7% [47–50 mmol/mol]), respectively. The mean reduction of HbA_1c was −0.32% with short-acting and −0.60% with long-acting GLP-1RAs; the final (median) HbA_1c levels achieved with short-acting and long-acting GLP-1RAs were 7% (53 mmol/mol) (IQR 6.5–7.2% [48–55 mmol/mol]) and 6.9% (52 mmol/mol) (IQR 6.6–7.1% [49–54 mmol/mol]), respectively. Heterogeneity remained high in all subgroup analyses (Table 2).

In the overall analysis of 28 comparisons, the percentage of patients achieving an HbA_1c target of ≤7% at the end of the intervention was higher with the GLP-1RA and insulin combination compared with other treatments: the overall likelihood of achieving the HbA_1c target was 65% higher in favor of the combination (RR = 1.65, 95% CI 1.44–1.88), with high heterogeneity (I² = 91.1%, P < 0.001) (Supplementary Fig. 2 and Table 2) and evidence of publication bias (P = 0.002). The trim-and-fill method indicated that this publication bias did not change the statistical significance of the estimate (RR = 1.33, 95% CI 1.16–1.53). The likelihood of achieving the HbA_1c target with GLP-1RA in combination with insulin compared with other treatments was greatest versus placebo (RR = 3.16, 95% CI 2.20–4.54) and lowest versus basal-bolus insulin regimen (RR = 1.07, 95% CI 0.95–1.20). The likelihood of achieving the HbA_1c target compared with respective control subjects was 49% and 80% higher with short- and long-acting GLP-1RAs, respectively. Except for the seven comparisons versus intensified insulin regimens (basal-plus or basal-bolus), heterogeneity remained high in all subgroup analyses (Table 2).

In the overall analysis of 25 comparisons, the relative risk of any hypoglycemia was not different between GLP-1RA in combination with insulin compared with other treatments (RR = 1.14, 95% CI 0.93–1.39, P = 0.214), with high heterogeneity (I² = 89.3%) (Supplementary Fig. 3 and Table 2) and evidence of publication bias (Egger test, P = 0.047). The trim-and-fill method indicated that this publication bias did not change the statistical significance of the estimate (RR = 0.92, 95% CI 0.73–1.16). There was a higher RR of hypoglycemia when GLP-1RA and insulin combination treatment was compared with placebo (RR = 1.47, 95% CI 1.16 –1.88) and a lower risk when compared with intensified insulin regimens (RR = 0.68, 95% CI 0.52–0.89). The relative risk of hypoglycemia was 5% and 27% higher, compared with control arms, with short- and long-acting GLP-1RAs, respectively. Heterogeneity was high for all subgroup comparisons (Supplementary Fig. 3 and Table 2).

In the overall analysis of 27 comparisons, the combination of GLP-1RA and insulin led to a mean weight decrease significantly greater than comparator groups (−2.5 kg, 95% CI −3.3 to −1.7), with high heterogeneity (I² = 97.3%) (Supplementary Fig. 4 and Table 2) and evidence of publication bias (Egger test, P = 0.003). The trim-and-fill method indicated that this publication bias did not affect the estimate. The greatest difference in weight favoring GLP-1RA and insulin combination treatment was seen versus insulin intensification treatment (−6.3 kg, 95% CI −8.0 to −4.6), and the least difference was seen with the fixed-ratio GLP-1RA and insulin combination versus insulin intensification (−1.0 kg, 95% CI −2.4 to 0.5, P = 0.181). The weight decrease was 2.3 kg and 2.7 kg with short or long-acting GLP-1RAs, respectively. Heterogeneity was high for all subgroup comparisons (Table 2).

The overall results of our meta-analysis of RCTs show beneficial effects of the combination treatment with a GLP-1RA and insulin, compared with other injectable treatments, on HbA_1c reduction and target and on weight reduction among individuals with type 2 diabetes. As expected, the greatest metabolic effects of the combination therapies were seen during the comparison with placebo or up-titration of one component, mainly basal insulin. The considerable number of new studies published in the past 2 years allowed us to assess the comparative value of GLP-1RAs added to basal insulin versus the addition of prandial insulin as well as the role of fixed-ratio combinations. We could not find any significant advantage on glycemic control (HbA_1c reduction from baseline and percentage of patients at HbA_1c target <7% [53 mmol/mol]) when the combination therapy with GLP-1RAs and basal insulin was compared with basal-plus (2 comparisons, HbA_1c decrease in favor of combination therapy = −0.2%, P = 0.057) or basal-bolus (7 comparisons, HbA_1c decrease = −0.08%, P = 0.171) insulin regimens. On the one hand, the risk of hypoglycemia was lower and the weight decrease was higher with the combination therapies. On the other hand, fixed-ratio combinations (IDegLira or IGlarLixi) were significantly better than each single component in reducing HbA_1c levels and improving the percentage of patients at target (<7% [53 mmol/mol]) at the same level of weight change and hypoglycemia. The overall results suggest that the combination of GLP-1RA and insulin may be considered a promising therapeutic strategy to improve the clinical management of type 2 diabetes.

In patients with type 2 diabetes who are not achieving glycemic goals, insulin therapy should not be delayed (2), considering regimen flexibility for the initiation and intensification of therapy. Insulin regimens containing long- and short-acting preparations are widely used; however, there is still reluctance to intensifying insulin treatment (38) because of fear of hypoglycemia and weight gain that undermines its capacity to reach the HbA_1c target in clinical practice. Furthermore, postprandial hyperglycemia may become the main contributor to hyperglycemic exposure, necessitating the timely initiation of prandial treatment. Just increasing the dose of basal insulin more than 0.5 units/kg does not result in further improvements in glycemic control and is associated with increased weight gain and risk of hypoglycemia (39).

GLP-1RAs evaluated in combination with basal insulin included exenatide, liraglutide, lixisenatide, albiglutide, and dulaglutide; except for dulaglutide, all have U.S. Food and Drug Administration– approved indication for concomitant use with basal insulin. The GLP-1 family is usually divided in short-acting (exenatide and lixisenatide) and long-acting (exenatide extended release, liraglutide, albiglutide, and dulaglutide) receptor agonists, which may have greater advantages in reducing postprandial (40) or fasting (41) hyperglycemia, respectively. In the subgroup analysis comparing short-acting versus long-acting GLP-1RAs of combination therapies, short-acting agonists reduced HbA_1c by 0.32% and long-acting agonists by 0.60% (Table 2), compared with respective control subjects, which seems in line with their overall effect on HbA_1c. However, the decrease of HbA_1c during treatment depends on many variables, among which are the baseline HbA_1c level (42), study design with or without a lead-in period, and the intensification of single therapies. Accordingly, the final HbA_1c levels achieved with short- or long-acting GLP-1RAs as groups were quite similar (7.0 vs. 6.9% [53 vs. 52 mmol/mol], respectively).

The Standards of Medical Care in Diabetes—2016 (2) also invites consideration of the use of a GLP-1RA added to basal insulin as a newer treatment option for adjusting insulin therapy in type 2 diabetes. Although the meta-analysis of Eng et al. (5) was the first attempt to meta-analyze trial data about this new combination therapy, and their results have influenced current therapeutic guidelines (2), there is some imprecision about the HbA_1c outcome at the end of treatment. Supplementary Table 4 provides the WMD for HbA_1c as reported in the original article from Eng et al. (5) and those we extracted from the 15 original trials and also reported in the present meta-analysis. In contrast with the previous meta-analysis, which found a significant albeit small greater HbA_1c reduction (0.1%) with the combination therapy versus basal-bolus insulin regimens, we did not find any significant difference between these two therapeutic strategies (WMD = −0.08%, P = 0.177), probably owing to the greater number of comparisons we did in our analysis (seven vs. three comparisons). Taken globally, the results of the nine comparisons between combination therapies versus intensified insulin regimens (basal-plus and basal-bolus) in type 2 diabetes suggest quite similar glycemic control associated with a lower risk of hypoglycemia and a greater reduction in weight.

Titratable fixed-ratio combination products that administer both agents in a single injection are available and will offer additional options for clinicians and patients. The fixed-ratio combinations of IDegLira and of IGlarLixi offer a simpler way to use this therapeutic strategy in patients with type 2 diabetes where basal insulin needs to be intensified. In our meta-analysis comprising six trials with eight different comparisons (four for IDegLira and four for IGlarLixi), the fixed-ratio combination therapy was associated with a 0.56% lower HbA_1c level than comparator therapies, mainly insulin up-titrations, and with a not statistically different hypoglycemic risk (26% higher) and weight change (1.0 kg lower). In the subgroup analysis, IDegLira reduced HbA_1c by 0.68% and IGlarLixi by 0.44% compared with respective control subjects. This may reflect the different value of liraglutide and lixisenatide in improving glycemic control when given as an add-on to metformin in patients with type 2 diabetes (43). However, the final HbA_1c levels achieved were identical (6.5% [48 mmol/mol]) with both fixed-ratio combinations, suggesting that the different HbA_1c decrease was mainly related to the different baseline HbA_1c levels.

Until now, no reported study has compared fixed-ratio combinations with intensified insulin regimens, either basal-bolus or basal-plus. At the present, the fixed-ratio combinations may be of particular value in patients on insulin therapy in whom HbA_1c is not sufficiently reduced, when there is no need to increase the number of daily injection. There is also evidence (44) that IDegLira reduces HbA_1c more than placebo (1.02%) in people with diabetes uncontrolled on an oral agent (sulfonylurea with or without metformin), with more weight gain (1.5 kg) and hypoglycemia (RR = 3.7).

The strengths of this review include the comprehensive systematic search that considered all of the RCTs published until November 2016, the use of a prespecified subgroup analyses and double-checking of data extraction, and the quality of studies above the average—12 studies had good quality. Finally, the different designs of the studies included in the analysis (at least three) makes it comprehensive of the many situations occurring in clinical practice.

Our study has several limitations. The main limitations relate to the high degree of between-study heterogeneity and the evidence of publication bias for three outcomes (proportion of patients at target, incidence of any hypoglycemia, and weight change), which may limit the generalizability across clinical settings of the metabolic benefits of the combination therapies, giving more credit to the characteristics of each single trial. Reasons for the high heterogeneity may include, although not limited to patients’ characteristics, the GLP-1RA preparations under study, the background antidiabetes therapy, and the trials’ design. Although the trim-and-fill computation did not change the significance of the results, this method may inappropriately adjust for publication bias when the studies are highly heterogeneous (45). Moreover, trial durations were too short for assessing the long-term durability of the combination, most studies (23 of 26) were industry funded, of open-label design (16 of 26), and had comparison with placebo (10 of 26). The 10 double-blind trials were placebo-controlled, which explains why they produced larger a HbA_1c decrease (WMD HbA_1c = −0.83, 95% CI −1.04 to −0.62, P < 0.001) compared with the 16 trials with an open design (WMD HbA_1c = −0.31, 95% CI −0.43 to −0.19, P < 0.001). Finally, no trial was specifically designed for assessing clinical outcomes (microvascular and macrovascular complications and death). However, a meta-analysis (46) of more than 300 available clinical trials in type 2 diabetes concluded for limited evidence that any glucose-lowering drug prolonged life expectancy or prevented cardiovascular disease.

GLP-1RAs alone or as titratable fixed-ratio combinations with basal insulin may represent a promising option to advance basal insulin therapy or to initiate injectable therapy in patients with type 2 diabetes inadequately controlled on oral agents. In the clinical management of these patients, this combination treatment may offer the same glycemic control of intensified insulin regimens (basal-plus or basal-bolus), but with less hypoglycemia and reduced weight. Moreover, the use of titratable fixed-ratio combinations of GLP-1RAs plus basal insulin may improve metabolic control more than each single component. Longer studies are needed to assess tolerability, effectiveness, and costs of this combination before it can be favored as standard of care for patients with type 2 diabetes when basal insulin therapy is failing.

Duality of Interest. K.E. received a consultancy fee from Eli Lilly and has held lectures for Eli Lilly, Sanofi, and Novo Nordisk. D.G. received a consultancy fee from Eli Lilly and has held lectures for Eli Lilly and Sanofi. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.I.M. and D.G. conducted the literature search, data extraction, and data analysis and wrote the manuscript. P.C. and D.G. did the statistical analyses. G.B. and A.C. contributed to the data analysis and to writing the manuscript. K.E. and D.G. reviewed and edited the manuscript.