Insulin and Glucagon-Like Peptide 1 Receptor Agonist Combination Therapy in Type 2 Diabetes

614 Diabetes Care Volume 40, April 2017

Maria Ida Maiorino,1 Paolo Chiodini,2 Insulin and Glucagon-Like Peptide Giuseppe Bellastella,3 Annalisa Capuano,4 1 Receptor Agonist Combination Katherine Esposito,1 and Dario Giugliano3 Therapy in Type 2 Diabetes: A Systematic Review and Meta-analysis of Randomized Controlled Trials Diabetes Care 2017;40:614–624 | DOI: 10.2337/dc16-1957

OBJECTIVE The combination of basal insulin plus a glucagon-like peptide 1 receptor agonist (GLP-1RA) has been proposed as a treatment option to intensify insulin therapy in type 2 diabetes. We performed a meta-analysis of randomized controlled trials (RCTs) comparing this combination strategy to other injectable antidiabetes treatments on metabolic control in adult patients with type 2 diabetes.

RESEARCH DESIGN AND METHODS 1Diabetes Unit, Department of Medical, Surgical, META-ANALYSIS We conducted an electronic search until November 2016 on many electronic Neurological, Metabolic Sciences and Aging, Uni- versity of Campania “Luigi Vanvitelli,” Naples, databases to identify RCTs assessing changes in HbA1c, proportion of patients at £ Italy HbA1c target 7% (53 mmol/mol), hypoglycemia, and weight change. We used a 2Medical Statistics Unit, University of Campania random-effect model to calculate the weighted mean difference (WMD) or rela- “Luigi Vanvitelli,” Naples, Italy tive risk (RR) with the 95% CI. 3Division of Endocrinology and Metabolic Dis- eases, Department of Medical, Surgical, Neuro- RESULTS logical, Metabolic Sciences and Aging, University “ ” fi – of Campania Luigi Vanvitelli, Naples, Italy We identi ed 26 RCTs, lasting 12 52 weeks, and involving 11,425 patients. When 4Section of Pharmacology, Department of Exper- the combination strategy was compared with other injectable treatments (overall imental Medicine, University of Campania “Luigi Vanvitelli,” Naples, Italy data), there were reductions in HbA1c (WMD = 20.47%, 95% CI 20.59 to 20.35), more patients at HbA1c target (RR = 1.65, 95% CI 1.44–1.88), similar hypoglycemic Corresponding author: Dario Giugliano, dario. events (RR = 1.14, 95% CI 0.93–1.39) and a reduction in weight (WMD = 22.5 kg, [email protected]. 95% CI 23.3 to 21.7), with high heterogeneity (I2 > 89%, P < 0.001) and a signif- Received 10 September 2016 and accepted 26 December 2016. icant publication bias for three outcomes. In preplanned subgroup analyses, the This article contains Supplementary Data online combination treatment was similar to basal-bolus insulin regimens for glycemic at http://care.diabetesjournals.org/lookup/ control, with less hypoglycemia (RR = 0.66, 95% CI 0.46–0.93) and reduced weight suppl/doi:10.2337/dc16-1957/-/DC1. (WMD = 24.7 kg, 95% CI 26.9 to 22.4). Fixed-ratio combinations yielded results This article is featured in a podcast available at similar to the overall analysis (HbA1c WMD = 20.56%, 95% CI 20.72 to 20.40). http://www.diabetesjournals.org/content/ diabetes-core-update-podcasts. CONCLUSIONS © 2017 by the American Diabetes Association. GLP-1RAs alone or as titratable fixed-ratio combinations with basal insulin may Readers may use this article as long as the work represent a promising option to advance basal insulin therapy or to initiate in- is properly cited, the use is educational and not for profit, and the work is not altered. More infor- jectable therapy in patients with type 2 diabetes inadequately controlled on oral mation is available at http://www.diabetesjournals agents. Longer studies are needed to assess durability and tolerability. .org/content/license. care.diabetesjournals.org Maiorino and Associates 615

Current management of hyperglycemia of many trials with different designs, in- prior reviews and meta-analyses were in type 2 diabetes claims that in patients terventions, and study groups are now also manually searched to track relevant with inadequate glycemic control on ini- available, we planned a systematic re- RCTs that were not indexed by normal tial metformin monotherapy, an escala- view and meta-analysis of randomized keywords. tion to a two-drug combination, and controlled trials (RCTs) that evaluated subsequently, to a three-drug combina- the role of GLP-1RA and insulin combi- Study Selection We selected studies if they were RCTs tion is indicated (1). However, because nation in the injectable treatment of pa- performed in adults with type 2 diabe- of the progressive nature of type 2 di- tients with type 2 diabetes. In particular, tes, compared short- and long-acting abetes, many patients eventually re- we sought to answer the following GLP-1RAs administered in association quire insulin therapy, usually started questions: with insulin treatment to another inject- with a long-acting (basal) formulation able treatment strategy, with at least at bedtime (2). When basal insulin has 1. Is this combination therapy similarly 12 weeks’ duration of intervention, been titrated to an acceptable fasting effective in reducing baseline HbA 1c and assessed variation in HbA and/or blood glucose but HbA remains above when compared with basal-plus 1c 1c the proportion of participants with target, therapy may be intensified by (adding one main-meal fast-acting HbA of #7.0% (53 mmol/mol) at the the addition of a mealtime insulin (one insulin to basal insulin once daily) 1c end of the study period or the number of to three injections of a short-acting or full basal-bolus (four insulin injec- participants with hypoglycemic events analog at meals) or by changing to pre- tions daily) insulin regimens? or weight change. mixed insulin formulations (3). These 2. Which is the role of the fixed-ratio intensive insulin regimens (three to combinations of GLP-1RA and basal Data Extraction and Quality four insulin injections daily) are associ- insulin in this scenario? Assessment ated with a higher risk of hypoglycemia 3. Are there additional benefits of the Two investigators (M.I.M., D.G.) used a and weight gain, which in turn may un- combination therapy on other as- standardized tool to independently ab- dermine the achievement of glycemic pects of the diabetes control, includ- stract all data, and disagreements were targets (4). ing percentage of patients at HbA1c resolved by consensus. Any RCTs that The American Diabetes Association’s target, incidence of hypoglycemia, met the inclusion/exclusion criteria Standards of Medical Care in Diabetesd and weight change? were included in the analysis. We 2016 (2) reviews the approach to start- extracted the following data from each fi ing and adjusting insulin in type 2 diabe- RESEARCH DESIGN AND METHODS selected study: 1) author identi cation tes: at the level of two injections daily, a and year of publication; 2)durationof This systematic review was conducted in newer treatment option, consisting in intervention; 3) investigational drug accordance with PRISMA (Preferred Re- the combination of basal insulin plus a with number of patients; 4) compara- porting Items for Systematic Reviews glucagon-like peptide 1 receptor agonist tor drug with number of patients; 5) and Meta-analyses) guidelines (6). The (GLP-1RA), is offered as a trial. This com- background therapy in both study PRISMA checklist and the protocol of this bination shows complementary modes groups; 6)HbA outcome; 7) baseline study are provided in the Supplementary 1c of action in the treatment of type 2 HbA ; 8) study funder; and 9)typeof Data. 1c diabetes, with potential benefits on gly- statistical analysis for the HbA1c out- cemic control and metabolic profile. The Data Sources and Searches come. The relevance of studies was as- meta-analysis of Eng et al. (5) is quoted Bibliographical databases for the litera- sessed with a hierarchical approach on to support the option of adding a GLP- ture search included PubMed, MEDLINE, the basis of title, abstract, and the full 1RA to ongoing insulin treatment be- Cochrane Central Register of Controlled article. After the initial screening of ti- cause this combination yielded an im- Trials, and ClinicalTrials.gov (http:// tles and abstracts, the studies included proved mean reduction in HbA1c of www.clinicaltrials.gov). The last search by both reviewers were compared, and 0.44% compared with other antidiabetes was performed on 7 November 2016. disagreement was resolved by consensus. treatments and was associated with no Our search strategy included the key- We evaluated the risk of bias of the increased hypoglycemia and reduced words ‘‘type 2 diabetes’’ or “type II di- included RCTs according to the Co- weight. However, most of the trials com- abetes”; ‘‘hemoglobin A1c’’ or “HbA1c” chrane Collaboration’stoolforassess- pared the efficacy of the combination or “glucose”; ‘‘biphasic insulin regimen’’ ing the risk of bias (7). We assessed versus placebo, which is far from the or “basal-plus insulin regimen” or risk of bias in random sequence gen- real world of diabetes therapy. Other tri- ‘‘basal-bolus insulin regimen’’; ‘‘glargine’’ eration and allocation concealment (se- als compared the efficacy of the combi- or ‘‘detemir’’ or “degludec” or ‘‘neutral lection bias), blinding of participants and nation versus up-titration of basal insulin, protamine lispro’’ or ‘‘lispro’’ or ‘‘aspart’’ personnel (performance bias), blinding which is unable to cover postprandial or ‘‘glulisine’’; “glucagon-like peptide 1 re- of outcome assessment (detection glucose excursions. ceptor agonist” or “GLP-1” or “exenatide” bias), incomplete outcome data (attri- Before this combination therapy can or “ liraglutide” or “lixisenatide” or “dula- tion bias), and selective reporting (re- be offered to clinicians as the standard glutide” or “albiglutide” or “semaglutide”; porting bias). The risks of bias were of care for patients with type 2 diabetes, ‘‘randomized” or “trial.” A Google search categorized a high, low, and unclear. other questions must be answered that was also conducted to find information on Methodologic quality was also scored can help translate this strategy into clin- RCTs that was unavailable from biblio- using criteria set out by Jadad et al. (8). ical practice. Recognizing that reports graphical databases. Reference lists of This 5-point quality scale includes points 616 Meta-analysis of RCTs of Insulin Plus GLP-1RA Diabetes Care Volume 40, April 2017

for randomization (1 point; using tables of was used. In a conservative way, we con- variables in the two groups. Data were random numbers or computer-generated sidered random-effect analysis the main analyzed using Stata 11.2 software randomization, additional 1 point), focus of our meta-analysis. We did pre- (StataCorp LP, College Station, TX). All double-blind (1 point; use of placebo, planned subgroup analyses restricted to statistical tests were two sided, and additional 1 point), and follow-up (stat- trials that compared GLP-1RA + insulin P values of ,0.05 were regarded as ing the numbers and reasons for with- versus insulin or GLP-1RA + placebo; significant. drawal in each group; 1 point) in the GLP-1RA + insulin versus insulin up- report of an RCT. We gave an additional titration; GLP-1RA + insulin versus basal- RESULTS point if the analysis was by intention to plus or basal-bolus insulin regimens; Search Results treat. We considered a score of $4as GLP-1RA + insulin in fixed-ratio combi- The initial search assessed 3,174 cita- good quality. The quality of each RCT nation versus single component; and tions, and 2,202 abstracts were selected was assessed by one reviewer and veri- short- versus long-acting GLP-1RAs. for review after duplicates were re- fied by another reviewer. Disagreement Publication bias was assessed visually moved. By screening the abstracts, we was resolved by consensus. with funnel plots and with the Egger excluded 2,169 citations (observational test (10); a P value of ,0.10 was con- studies, non-RCTs, mostly review arti- Statistical Analysis sidered significant. The trim-and-fill cles, nonhuman studies) (Fig. 1). Of the The decrease of HbA1c from baseline at method was used to estimate the ef- remaining 33 RCTs, 7 were excluded be- the end of treatment was the primary fect of publication bias (11). Sensitivity cause they evaluated type 1 diabetes outcome of this meta-analysis. Second- analyses in which each study was re- (2 studies), were extensions of RCTs ary end points were proportion of patients movedinturntoassesstheinfluence (2 studies), compared 2 basal insulin at HbA1c target #7% (53 mmol/mol), in- of that study on the overall effect size regimens (1 study), or did not include cidence of hypoglycemic events, and were also performed. For descriptive pur- insulin-treated patients (2 studies) weight change. Changes from baseline poses, the median and interquartile range (Supplementary Table 1). Finally, in HbA1c and body weight were analyzed (IQR) was calculated for continuous 26 RCTs (12–37) with 30 comparisons 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 HbA1c targets ,7.0% (53 mmol/mol) and for incidence of hypoglycemic events. For each study, the proportion of participants achieving an HbA1c of #7.0% (53 mmol/mol) and those having any episode of hypoglycemia were used to calculate RR using 2 3 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 Figure 1—Process of study selection. used; otherwise, a fixed-effect model care.diabetesjournals.org Maiorino and Associates 617

were included for quantitative synthesis combination, and four trials (25,27,28,31) (IQR 6.4–6.7% [47–50 mmol/mol]), and meta-analysis. compared the combination with intensified respectively. The mean reduction of insulin regimens (basal-plus or basal-bolus). HbA1c was 20.32% with short-acting Study Characteristics One trial (32) tested the superiority and five and 20.60% with long-acting GLP- The characteristics of the included trials (33–37) tested the noninferiority of 1RAs; the final (median) HbA1c levels 26RCTsaresummarizedinTable1. the fixed-ratio GLP-1RA and insulin combi- achieved with short-acting and long- The participants in all RCTs were pa- nation over basal insulin intensification or acting GLP-1RAs were 7% (53 mmol/mol) tients with type 2 diabetes (.18 years GLP-1RA alone. (IQR 6.5–7.2% [48–55 mmol/mol]) and old). Most trials were multinational and According to the Cochrane Collabora- 6.9% (52 mmol/mol) (IQR 6.6–7.1% sponsored by industry (11 trials by Novo tion’s tool for assessing risk of bias, the [49–54 mmol/mol]), respectively. Hetero- Nordisk,3trialsbyEliLilly,8trialsbySa- three common biases were blinding of geneity remained high in all subgroup nofi, and 1 trial by GlaxoSmithKline). The participants (16 of 26 trials, performance analyses (Table 2). trials were published between 2011 and bias, high risk of bias), blinding of out- In the overall analysis of 28 compari- 2016, with 9 studies (19–21,26,31,34–37) come assessment (16 of 26 trials, detec- sons, the percentage of patients achiev- published in 2016. All trials were of par- tion bias, high risk of bias), and allocation ing an HbA1c target of #7% at the end of allel-group design, 10 were double-blind concealment (all trials, selection bias, un- the intervention was higher with the (12,14–21,32), and the remaining were of clear risk of bias) (Supplementary Fig. 1 GLP-1RA and insulin combination com- open-label design. Trial duration ranged and Supplementary Table 2). The median pared with other treatments: the overall from 12 to 52 weeks. score of methodologic quality was 3.0 likelihood of achieving the HbA1c target The trials had different designs: 10 (IQR 3.0–5.0), and many studies (n =12) was 65% higher in favor of the combina- trials (12–21) compared GLP-1RA with had a quality score $4, indicating high tion (RR = 1.65, 95% CI 1.44–1.88), with placebo on a background of basal insu- quality (Supplementary Table 3). high heterogeneity (I2 = 91.1%, P , lin therapy, 3 RCTs (22–24) compared 0.001) (Supplementary Fig. 2 and Table 2) GLP-1RA with intensification of insulin Outcomes and evidence of publication bias (P = on a background of different insulin Intheoverallanalysisof30comparisons, 0.002). The trim-and-fill method indi- regimens, 2 RCTs (25,26) compared GLP- the combination of GLP-1RA and insulin cated that this publication bias did not 1RA and basal insulin versus a basal- led to a mean HbA1c decrease signifi- change the statistical significance of the plus insulin regimen, 5 trials (27–31), cantly greater than comparator groups estimate (RR = 1.33, 95% CI 1.16–1.53). with 7 comparisons, compared GLP-1RA (20.47%, 95% CI 20.59 to 20.35%, P , The likelihood of achieving the HbA1c tar- and basal or prandial insulin versus full 0.001), with high heterogeneity be- get with GLP-1RA in combination with basal-bolus insulin regimen (1 injection tween studies (I2 =93.7%,P , 0.001) insulin compared with other treatments of basal insulin, for most a long-acting (Fig. 2 and Table 2). There was no evi- was greatest versus placebo (RR = 3.16, insulin analog, plus 3 injections of a dence of publication bias (Egger test, P = 95% CI 2.20–4.54) and lowest versus short-acting insulin analog at meals), 0.138). In sensitivity analyses, the re- basal-bolus insulin regimen (RR = 1.07, and 6 trials (32–37), with 8 comparisons, moval of each study in turn did not 95% CI 0.95–1.20). The likelihood of compared fixed-ratioGLP-1RAandin- change the overall effect size. Preplanned achieving the HbA1c target compared sulin combination versus insulin inten- subgroup analyses showed that most of with respective control subjects was sification (32–37) or GLP-1RA alone the estimate favoring the combination of 49% and 80% higher with short- and (33,37). GLP-1RA and insulin over comparators de- long-acting GLP-1RAs, respectively. Ex- rived from the comparison with placebo/ cept for the seven comparisons versus Intervention and Risk of Bias insulin titration: in the 10 comparisons ver- intensified insulin regimens (basal-plus The trials evaluated 11,425 patients for sus placebo, the difference for HbA1c was or basal-bolus), heterogeneity remained the primary end point of this meta-analysis greatest (20.76%, 95% CI 20.96 to 20.57, high in all subgroup analyses (Table 2). (HbA1c change at the end of the trial), P , 0.001) (Fig. 2 and Table 2). There was In the overall analysis of 25 compari- 5,689 in the intervention groups and 5,736 no difference in the mean overall HbA1c sons, the relative risk of any hypoglyce- in the comparator groups (range of pa- reduction between GLP-1RA and basal in- mia was not different between GLP-1RA tients in each arm, 21–834) (Table 1). The sulin versus full basal-bolus (seven com- in combination with insulin compared baseline HbA1c level was identical between parisons) or basal-plus (two comparisons) with other treatments (RR = 1.14, 95% groups, with a median of 8.3% (67 mmol/ insulin regimens. The comparison of fixed- CI 0.93–1.39, P = 0.214), with high het- mol) in the GLP-1RA and insulin groups (IQR ratioGLP-1RAandbasalinsulincombina- erogeneity (I2 = 89.3%) (Supplementary 7.7–8.5% [60.7–69.4 mmol/mol]), and 8.1% tion with intensification of single therapies Fig. 3 and Table 2) and evidence of pub- (65 mmol/mol) in the comparator groups showed a mean HbA1c reductioninfavor lication bias (Egger test, P = 0.047). The (IQR 7.5–8.5% [58.5–69.4 mmol/mol]). of the fixed-ratio combination (eight com- trim-and-fill method indicated that this Four trials (12,13,15,20) tested the superi- parisons, WMD = 20.56%, 95% CI 20.72 publication bias did not change the sta- ority of the combination (GLP-1RA and in- to 20.40). The final (median) HbA1c levels tistical significance of the estimate (RR = sulin) over placebo, six trials (14,16–19,21) achieved with both combinations (insulin 0.92, 95% CI 0.73–1.16). There was a compared the combination with placebo, degludec and liraglutide [IDegLira] or of higher RR of hypoglycemia when GLP- three trials (22–24) compared the combina- glargine and lixisenatide [IGlarLixi]) were 1RA and insulin combination treatment tion with insulin intensification, three trials 6.5% (48 mmol/mol) (IQR 6.4–6.7% [47– was compared with placebo (RR = 1.47, (26,29,30) tested the noninferiority of the 50 mmol/mol]) and 6.5% (48 mmol/mol) 95% CI 1.16 –1.88) and a lower risk when 1 eaaayi fRT fIslnPu GLP-1RA Plus Insulin of RCTs of Meta-analysis 618

Table 1—Characteristics of RCTs included in the meta-analysis

Baseline HbA1c (%) (mmol/mol) Study design/ Study drug/ Comparator/ interventional/ Author, year duration (weeks) patients (n) patients (n) Background therapy HbA1c outcome comparator Study funder

Buse et al., 2011 R, DB, P/30 Exenatide 10 mg Placebo/123 Glargine 6 metformin, pioglitazone, HbA1c levels 8.3/8.5 (67/69) Eli Lilly-Amylin twice daily/138 or both at week 30 Pharmaceuticals

DeVries et al., 2012 R, O, P/26 Detemir/162 No detemir/161 Metformin + liraglutide Change in HbA1c 7.6/7.6 (60/60) Novo Nordisk to week 26

Seino et al., 2012 R, DB, P/24 Lixisenatide Placebo/157 Basal insulin 6 sulfonylurea Change in HbA1c 8.5/8.5 (69/69) Sanoﬁ 20 mg/154 to week 24

Riddle et al., 2013 R, DB, P/24 Lixisenatide Placebo/167 Basal insulin 6 metformin Change in HbA1c 8.4/8.4 (68/68) Sanoﬁ 20 mg/328 to week 24

Riddle et al., 2013 R, DB, P/24 Lixisenatide Placebo/223 Glargine plus metformin, 6 pioglitazone Change in HbA1c 7.6/7.6 (60/60) Sanoﬁ 20 mg/223 to week 24

Ahmann et al., 2015 R, DB, P/26 Liraglutide Placebo/225 Basal insulin 6 metformin Change in HbA1c 8.2/8.3 (66/67) Novo Nordisk 1.8 mg/226 to week 26

Lind et al., 2015 R, DB, P/24 Liraglutide Placebo/60 MDI 6 metformin Change in HbA1c 9.0/9.0 (75/75) Novo Nordisk 1.8 mg/64 to week 24

Aroda et al., 2016 R, DB, P/26 Degludec/174 Placebo/172 Metformin + liraglutide Change in HbA1c 7.6/7.6 (60/60) Novo Nordisk to week 26

Seino et al., 2016 R, DB, P/36 Liraglutide Placebo/130 Basal or premixed or basal-bolus therapy Change in HbA1c 8.8/8.8 (73/73) Novo Nordisk 0.9 mg/127 to week 16

Vanderheiden et al., 2016 R, DB, P/24 Liraglutide Placebo/36 High-dose insulin regimen Change in HbA1c 9.0/8.9 (74/75) Novo Nordisk 1.8 mg/35 to week 24

Li et al., 2012 R, O, P/12 Liraglutide Up-titration of Basal insulin or premixed insulin 6 metformin, HbA1c levels 8.8/8.7 (73/72) National Nature 1.2 mg/42 insulin therapy/42 sulfonylurea, thiazolidinedione, glinide, at week 12 Science Fundation or a-glucosidase inhibitor of China de Wit et al., 2014 R, O, P/26 Liraglutide Intensiﬁcation of Basal insulin 6 bolus insulin or metformin, Change in HbA1c 7.2/7.5 (55/58) Novo Nordisk 1.8 mg/26 insulin therapy/24 sulfonylurea, or both to week 26

Lane et al., 2014 R, O, P/24 Liraglutide Up-titration of CSII or MDI 6 metformin HbA1c at week 24 7.8/7.8 (62/62) NS 1.8 mg/21 insulin therapy/16

Mathieu et al., 2014 R, O, P/26 Liraglutide Aspart at largest Degludec + metformin Change in HbA1c 7.7/7.7 (61/61) Novo Nordisk Care Diabetes 1.8 mg/88 meal/89 to week 26

Rosenstock et al., 2016 R, O, P/26 Lixisenatide Glulisine once Glargine 6 metformin Change in HbA1c 7.7/7.8/7.7 Sanoﬁ 20 mg/298 daily/298, to week 26 (61/62/61) glulisine thrice daily/298 2017 April 40, Volume

Shao et al., 2014 R, O, P/12 Exenatide 10 mg Aspart thrice Glargine HbA1c at week 12 7.6/7.7 (60/61) NS twice daily/30 daily/30

Rosenstock et al., 2014 R, O, P/26 Albiglutide Lispro thrice daily/ Glargine 6 metformin, Change in HbA1c 8.5/8.4 (69/70) GlaxoSmithKline 30 mg/282 281 pioglitazone, or both to week 26 Continued on p. 619 care.diabetesjournals.org

Table 1—Continued

Baseline HbA1c (%) (mmol/mol) Study design/ Study drug/ Comparator/ interventional/ Author, year duration (weeks) patients (n) patients (n) Background therapy HbA1c outcome comparator Study funder

Diamant et al., 2014 R, O, P/30 Exenatide 10 mg Lispro thrice daily/ Glargine + metformin Change in HbA1c 8.3/8.2 (67/66) Eli Lilly-Amylin twice daily/315 312 to week 30 Pharmaceuticals, Bristol-Myers Squibb, AstraZeneca

Blonde et al., 2015 R, O, P/52 Dulaglutide Glargine Prandial lispro 6 metformin Change in HbA1c 8.5/8.4/8.5 Eli Lilly 1.5 mg/295, titration/296 to week 26 (69/68/69) dulaglutide 0.75/293

FLAT-SUGAR, 2016 R, O, P/26 Exenatide Short-acting Glargine + metformin Change in HbA1c 7.9/7.9 (63/63) Sanoﬁ, Bristol-Myers 5–10 mg twice insulin analogs to week 26 Squibb, AstraZeneca daily/52 thrice daily/50

Buse et al., 2014 R, DB, P/26 IDegLira + Degludec + Basal insulin + metformin 6 Change in HbA1c 8.7/8.6 (72/71) Novo Nordisk metformin/207 metformin sulfonylurea/glinide to week 26 titration/206

Gough et al., 2014 R, O, P/26 IDegLira/834 Degludec Metformin 6 pioglitazone Change in HbA1c 8.3/8.3/8.3 Novo Nordisk titration/414, to week 26 (67/67/67) liraglutide 1.8 mg/415

Lingvay et al., 2016 R, O, P/26 IDegLira/278 Glargine Glargine + metformin Change in HbA1c 8.4/8.2 (68/66) Novo Nordisk titration/279 to week 26

Aroda et al., 2016 R, O, P/30 IGlarLixi/367 Glargine Glargine + metformin Change in HbA1c 8.1/8.1 (65/65) Sanoﬁ titration/369 to week 30

Rosenstock et al., 2016 R, O, P/24 IGlarLixi/161 Glargine Metformin Change in HbA1c 8.1/8.0 (65/64) Sanoﬁ titration/162 to week 24

Rosenstock et al., 2016 R, O, P/30 IGlarLixi/469 Glargine Metformin Change in HbA1c 8.1/8.1/8.1 Sanoﬁ titration/467, to week 30 (65/65/65) lixisenatide m 20 g/234 619 Associates and Maiorino CSII, continuous subcutaneous insulin infusion; DB, double-blind; MDI, multiple daily injections; NS, not speciﬁed; O, open; P, parallel; R, randomized. 620 Meta-analysis of RCTs of Insulin Plus GLP-1RA Diabetes Care Volume 40, April 2017

Figure 2—Forest plots of meta-analysis for the primary end point (HbA1c decrease from baseline) in all 30 comparisons. There are 13 comparisons of the combination therapy vs. placebo/insulin titration, 9 comparisons vs. basal-plus/basal-bolus insulin regimens, and 8 comparisons of the ﬁxed- ratio combination vs. single component. The results are expressed as the WMD (HbA1c decrease in the GLP-1RA and insulin combination arm minus HbA1c decrease in the comparator arm). ES, effect size.

compared with intensified insulin regi- treatment (26.3 kg, 95% CI 28.0 to placebo or up-titration of one compo- mens (RR = 0.68, 95% CI 0.52–0.89). 24.6), and the least difference was nent, mainly basal insulin. The consider- The relative risk of hypoglycemia was seen with the fixed-ratio GLP-1RA and able number of new studies published 5% and 27% higher, compared with con- insulin combination versus insulin in- in the past 2 years allowed us to assess trol arms, with short- and long-acting tensification (21.0 kg, 95% CI 22.4 to the comparative value of GLP-1RAs added GLP-1RAs, respectively. Heterogeneity 0.5, P = 0.181). The weight decrease was to basal insulin versus the addition of was high for all subgroup comparisons 2.3 kg and 2.7 kg with short or long-acting prandial insulin as well as the role of (Supplementary Fig. 3 and Table 2). GLP-1RAs, respectively. Heterogeneity fixed-ratio combinations. We could not In the overall analysis of 27 compari- was high for all subgroup comparisons find any significant advantage on gly- sons, the combination of GLP-1RA and (Table 2). cemic control (HbA1c reduction from insulin led to a mean weight decrease baseline and percentage of patients CONCLUSIONS significantly greater than compara- at HbA1c target ,7% [53 mmol/mol]) tor groups (22.5 kg, 95% CI 23.3 The overall results of our meta-analysis when the combination therapy with to 21.7), with high heterogeneity (I2 = of RCTs show beneficial effects of the GLP-1RAs and basal insulin was com- 97.3%) (Supplementary Fig. 4 and Table 2) combination treatment with a GLP-1RA pared with basal-plus (2 comparisons, and evidence of publication bias and insulin, compared with other inject- HbA1c decrease in favor of combination (Egger test, P = 0.003). The trim-and-fill able treatments, on HbA1c reduction therapy = 20.2%, P = 0.057) or basal-bolus method indicated that this publication and target and on weight reduction (7 comparisons, HbA1c decrease = 20.08%, bias did not affect the estimate. The among individuals with type 2 diabetes. P = 0.171) insulin regimens. On the one greatest difference in weight favoring As expected, the greatest metabolic hand, the risk of hypoglycemia was lower GLP-1RA and insulin combination treat- effects of the combination therapies and the weight decrease was higher with ment was seen versus insulin intensification were seen during the comparison with the combination therapies. On the other care.diabetesjournals.org Maiorino and Associates 621

Table 2—Preplanned subgroup analysis Estimate (95% CI) Patients/control Parameter Comparisons subjects WMD P value I2 P value of Q test

HbA1c (%) All comparisons 30 5,689/5,736 20.47 (20.59, 20.35) ,0.001 93.7 ,0.001 GLP-1RA/Ins vs. Placebo/intensification 13 1,720/1,536 20.68 (20.86, 20.50) ,0.001 88.8 ,0.001 Placebo 10 1,631/1,454 20.76 (20.96, 20.57) ,0.001 89.0 ,0.0001 Intensification 3 89/82 20.37 (20.70, 20.04) 0.029 75.9 0.016 B-B/B-P 9 1,653/1,654 20.11 (20.20, 20.01) 0.031 69.6 0.001 B-B 7 1,416/1,267 20.08 (20.19, 0.04) 0.177 70.7 0.002 B-P 2 237/387 20.20 (20.40, 0.01) 0.057 68.3 0.076 FC vs. intensification 8 2,316/2,546 20.56 (20.72, 20.40) ,0.001 92.0 ,0.001 IDegLira 4 1,319/1,314 20.68 (20.87, 20.48) ,0.001 84.9 ,0.001 IGlarLixi 4 997/1,232 20.44 (20.70, 20.19) ,0.001 94.9 ,0.001 Short-acting GLP-1RAs 13 2,535/2,890 20.32 (20.49, 20.15) ,0.001 94.4 ,0.001 Long-acting GLP-1RAs 17 3,154/2,846 20.60 (20.75, 20.44) ,0.001 90.7 ,0.001 RR

HbA1c #7% All comparisons 28 5,607/5,656 1.65 (1.44, 1.88) ,0.001 91.1 ,0.001 GLP-1RA/Ins vs. Placebo/intensification 13 1,720/1,536 2.67 (1.89, 3.79) ,0.001 89.7 ,0.001 Placebo 10 1,631/1,454 3.16 (2.20, 4.54) ,0.001 87.9 ,0.001 Intensification 3 89/82 1.43 (0.72, 2.83) 0.303 75.6 0.017 B-B/B-P 7 1,571/1,574 1.09 (0.99, 1.20) 0.076 28.8 0.209 B-B 5 1,334/1,187 1.07 (0.95, 1.20) 0.085 41.9 0.142 B-P 2 237/387 1.18 (0.98, 1.41) 0.272 0.0 0.419 FC vs. intensification 8 2,316/2,546 1.54 (1.30, 1.81) ,0.001 92.8 ,0.001 IDegLira 4 1,319/1,314 1.55 (1.27, 1.89) ,0.001 90.3 ,0.001 IGlarLixi 4 997/1,232 1.52 (1.09, 2.12) 0.013 95.6 ,0.001 Short-acting GLP-1RAs 11 2,453/2,810 1.49 (1.22, 1.83) ,0.001 91.9 ,0.001 Long-acting GLP-1RAs 17 3,154/2,846 1.80 (1.49, 2.17) ,0.001 90.8 ,0.001 Hypoglycemia All comparisons 25 4,928/5,275 1.14 (0.93, 1.39) 0.214 89.3 ,0.001 GLP-1RA/Ins vs. Placebo/intensification 11 1,635/1,460 1.41 (1.11, 1.80) 0.006 68.0 0.001 Placebo 9 1,567/1,394 1.47 (1.16, 1.88) 0.002 67.1 0.002 Intensification 2 68/66 0.85 (0.19, 3.82) 0.829 85.6 0.008 B-B/B-P 6 977/1,269 0.68 (0.52, 0.89) 0.005 78.6 ,0.001 B-B 5 828/971 0.66 (0.46, 0.93) 0.018 83.2 ,0.001 B-P 1 149/298 0.76 (0.60, 0.98) 0.032 dd FC vs. intensification 8 2,316/2,546 1.26 (0.85, 1.87) 0.253 94.2 ,0.001 IDegLira 4 1,319/1,314 1.19 (0.57, 2.46) 0.643 96.7 ,0.001 IGlarLixi 4 997/1,232 1.33 (0.84, 2.12) 0.226 88.8 ,0.001 Short-acting GLP-1RAs 13 2,535/2,890 1.05 (0.82, 1.34) 0.697 86.6 ,0.001 Long-acting GLP-1RAs 12 2,393/2,385 1.27 (0.89, 1.82) 0.192 91.9 ,0.001 WMD

Weight (kg) All comparisons 27 4,927/5,268 22.5 (23.3, 21.7) ,0.001 97.3 ,0.001 GLP-1RA/Ins vs. Placebo/intensification 12 1,546/1,364 22.6 (23.8, 21.5) ,0.001 96.3 ,0.001 Placebo 9 1,457/1,282 21.5 (22.3, 20.6) ,0.001 92.5 ,0.001 Intensification 3 89/82 26.3 (28.0, 24.6) ,0.001 73.8 0.022 B-B/B-P 7 1,065/1,358 24.1 (25.7, 22.5) ,0.001 96.4 ,0.001 B-B 5 828/971 24.7 (26.9, 22.4) ,0.001 97.3 ,0.001 B-P 2 237/387 22.7 (24.7, 20.7) 0.008 91.1 0.001 FC vs. Intensification 8 2,316/2,546 21.0 (22.4, 0.5) 0.181 98.1 ,0.001 IDegLira 4 1,319/1,314 21.4 (24.1, 1.3) 0.323 99.0 ,0.001 IGlarLixi 4 997/1,232 20.6 (22.1, 0.9) 0.455 96.1 ,0.001 Short-acting GLP-1RAs 13 2,535/2,890 22.3 (23.3, 21.3) ,0.001 96.7 ,0.001 Long-acting GLP-1RAs 14 2,392/2,378 22.7 (24.0, 21.3) ,0.001 97.8 ,0.001 B-B, basal-bolus; B-P, basal-plus; FC, fixed-ratio combination; GLP-1RA/Ins, GLP-1RA plus insulin. 622 Meta-analysis of RCTs of Insulin Plus GLP-1RA Diabetes Care Volume 40, April 2017

hand, fixed-ratio combinations (IDegLira or therapies. Accordingly, the final HbA1c subgroup analysis, IDegLira reduced IGlarLixi) were significantly better than levels achieved with short- or long-acting HbA1c by 0.68% and IGlarLixi by 0.44% each single component in reducing HbA1c GLP-1RAs as groups were quite similar compared with respective control sub- levels and improving the percentage of pa- (7.0 vs. 6.9% [53 vs. 52 mmol/mol], jects. This may reflect the different tients at target (,7% [53 mmol/mol]) at respectively). value of liraglutide and lixisenatide in the same level of weight change and hy- The Standards of Medical Care in improving glycemic control when given poglycemia. The overall results suggest Diabetesd2016 (2) also invites consid- as an add-on to metformin in patients that the combination of GLP-1RA and in- eration of the use of a GLP-1RA added to with type 2 diabetes (43). However, the sulin may be considered a promising ther- basal insulin as a newer treatment option final HbA1c levels achieved were identi- apeutic strategy to improve the clinical for adjusting insulin therapy in type 2 cal (6.5% [48 mmol/mol]) with both management of type 2 diabetes. diabetes. Although the meta-analysis fixed-ratio combinations, suggesting In patients with type 2 diabetes who of Eng et al. (5) was the first attempt that the different HbA1c decrease was are not achieving glycemic goals, insulin to meta-analyze trial data about this mainly related to the different baseline therapy should not be delayed (2), con- new combination therapy, and their re- HbA1c levels. sidering regimen flexibility for the initi- sults have influenced current therapeutic Until now, no reported study has ation and intensification of therapy. guidelines (2), there is some imprecision compared fixed-ratio combinations Insulin regimens containing long- and about the HbA1c outcome at the end of with intensified insulin regimens, either short-acting preparations are widely treatment. Supplementary Table 4 pro- basal-bolus or basal-plus. At the pre- used; however, there is still reluctance vides the WMD for HbA1c as reported in sent, the fixed-ratio combinations may to intensifying insulin treatment (38) be- the original article from Eng et al. (5) and be of particular value in patients on in- causeoffearofhypoglycemiaand those we extracted from the 15 original sulin therapy in whom HbA1c is not suf- weight gain that undermines its capacity trials and also reported in the present ficiently reduced, when there is no need to reach the HbA1c target in clinical meta-analysis. In contrast with the pre- to increase the number of daily injec- practice. Furthermore, postprandial hy- vious meta-analysis, which found a sig- tion. There is also evidence (44) that perglycemia may become the main con- nificant albeit small greater HbA1c IDegLira reduces HbA1c more than pla- tributor to hyperglycemic exposure, reduction (0.1%) with the combination cebo (1.02%) in people with diabetes necessitating the timely initiation of therapy versus basal-bolus insulin regi- uncontrolled on an oral agent (sulfonyl- prandial treatment. Just increasing the mens, we did not find any significant dif- urea with or without metformin), with dose of basal insulin more than 0.5 units/kg ference between these two therapeutic more weight gain (1.5 kg) and hypogly- does not result in further improvements strategies (WMD = 20.08%, P = 0.177), cemia (RR = 3.7). in glycemic control and is associated probably owing to the greater number The strengths of this review include with increased weight gain and risk of of comparisons we did in our analysis the comprehensive systematic search hypoglycemia (39). (seven vs. three comparisons). Taken that considered all of the RCTs pub- GLP-1RAs evaluated in combination globally, the results of the nine compar- lished until November 2016, the use with basal insulin included exenatide, isons between combination therapies of a prespecified subgroup analyses liraglutide, lixisenatide, albiglutide, and versus intensified insulin regimens and double-checking of data extraction, dulaglutide; except for dulaglutide, all (basal-plus and basal-bolus) in type 2 di- and the quality of studies above the have U.S. Food and Drug Administration– abetes suggest quite similar glycemic averaged12 studies had good quality. approved indication for concomitant use control associated with a lower risk of Finally, the different designs of the stud- with basal insulin. The GLP-1 family is hypoglycemia and a greater reduction ies included in the analysis (at least usually divided in short-acting (exena- in weight. three) makes it comprehensive of the tide and lixisenatide) and long-acting Titratable fixed-ratio combination many situations occurring in clinical (exenatide extended release, liraglutide, products that administer both agents practice. albiglutide, and dulaglutide) receptor ag- in a single injection are available and Our study has several limitations. The onists, which may have greater advan- will offer additional options for clini- main limitations relate to the high de- tages in reducing postprandial (40) or cians and patients. The fixed-ratio com- gree of between-study heterogeneity fasting (41) hyperglycemia, respectively. binations of IDegLira and of IGlarLixi and the evidence of publication bias In the subgroup analysis comparing offer a simpler way to use this therapeu- for three outcomes (proportion of pa- short-acting versus long-acting GLP- tic strategy in patients with type 2 dia- tients at target, incidence of any hypo- 1RAs of combination therapies, short- betes where basal insulin needs to be glycemia, and weight change), which acting agonists reduced HbA1c by 0.32% intensified. In our meta-analysis com- may limit the generalizability across clin- and long-acting agonists by 0.60% (Table prising six trials with eight different ical settings of the metabolic benefits of 2), compared with respective control comparisons (four for IDegLira and the combination therapies, giving more subjects, which seems in line with their four for IGlarLixi), the fixed-ratio combi- credit to the characteristics of each sin- overall effect on HbA1c. However, the nation therapy was associated with gle trial. Reasons for the high heteroge- decrease of HbA1c during treatment de- a0.56%lowerHbA1c level than com- neity may include, although not limited pends on many variables, among which parator therapies, mainly insulin up- to patients’ characteristics, the GLP-1RA are the baseline HbA1c level (42), study titrations, and with a not statistically preparations under study, the back- design with or without a lead-in period, different hypoglycemic risk (26% higher) ground antidiabetes therapy, and the and the intensification of single and weight change (1.0 kg lower). In the trials’ design. Although the trim-and-fill care.diabetesjournals.org Maiorino and Associates 623

computation did not change the signifi- and A.C. contributed to the data analysis and to 14. Seino Y, Min KW, Niemoeller E, Takami A; cance of the results, this method may writing the manuscript. K.E. and D.G. reviewed EFC10887 GETGOAL-L Asia Study Investigators. inappropriately adjust for publication and edited the manuscript. Randomized, double-blind, placebo-controlled trial of the once-daily GLP-1 receptor agonist bias when the studies are highly hetero- References lixisenatide in Asian patients with type 2 diabe- geneous (45). Moreover, trial durations tes insufficiently controlled on basal insulin with 1. Inzucchi SE, Bergenstal RM, Buse JB, et al. or without a sulfonylurea (GetGoal-L-Asia). were too short for assessing the long- Management of hyperglycemia in type 2 diabe- Diabetes Obes Metab 2012;14:910–917 term durability of the combination, tes, 2015: a patient-centered approach: update 15. Riddle MC, Aronson R, Home P, et al. Add- most studies (23 of 26) were industry to a position statement of the American Diabe- ing once-daily lixisenatide for type 2 diabetes tes Association and the European Association funded, of open-label design (16 of inadequately controlled by established basal for the Study of Diabetes. Diabetes Care 2015; 26), and had comparison with placebo insulin: a 24-week, randomized, placebo- 38:140–149 controlled comparison (GetGoal-L). Diabetes (10 of 26). The 10 double-blind trials 2. American Diabetes Association. Approaches Care 2013;36:2489–2496 were placebo-controlled, which ex- to glycemic treatment. Sec. 7. In Standards of 16. Riddle MC, Forst T, Aronson R, et al. Adding plains why they produced larger a Medical Care in Diabetesd2016. Diabetes Care once-daily lixisenatide for type 2 diabetes inad- 2 2016;39(Suppl. 1):S52–S59 HbA1c decrease (WMD HbA1c = 0.83, equately controlled with newly initiated and 3. Maiorino MI, Bellastella G, Esposito K, 95% CI 21.04 to 20.62, P , 0.001) com- continuously titrated basal insulin glargine: a GiuglianoD.Premixedinsulinregimensin 24-week, randomized, placebo-controlled study pared with the 16 trials with an open type 2 diabetes: pros. Endocrine 2017;55:45–50 (GetGoal-Duo 1). Diabetes Care 2013;36:2497– design (WMD HbA = 20.31, 95% 4. Inzucchi SE, Bergenstal RM, Buse JB, et al.; 1c 2503 CI 20.43 to 20.19, P , 0.001). Finally, American Diabetes Association (ADA); Euro- pean Association for the Study of Diabetes 17. Ahmann A, Rodbard HW, Rosenstock J, no trial was specifically designed for as- fi (EASD). Management of hyperglycemia in et al.; NN2211-3917 Study Group. Ef cacy and sessing clinical outcomes (microvascular type 2 diabetes: a patient-centered approach: safety of liraglutide versus placebo added to and macrovascular complications and position statement of the American Diabetes basal insulin analogues (with or without metfor- death). However, a meta-analysis (46) Association (ADA) and the European Association min) in patients with type 2 diabetes: a randomized, placebo-controlled trial. Diabetes Obes of more than 300 available clinical trials for the Study of Diabetes (EASD). Diabetes Care – Metab 2015;17:1056–1064 in type 2 diabetes concluded for limited 2012;35:1364 1379 5. Eng C, Kramer CK, Zinman B, Retnakaran R. 18. Lind M, Hirsch IB, Tuomilehto J, et al. Lira- evidence that any glucose-lowering Glucagon-like peptide-1 receptor agonist and glutide in people treated for type 2 diabetes drug prolonged life expectancy or pre- basal insulin combination treatment for the with multiple daily insulin injections: randomised vented cardiovascular disease. management of type 2 diabetes: a systematic clinical trial (MDI Liraglutide trial). BMJ 2015;351: h5364 GLP-1RAs alone or as titratable fixed- review and meta-analysis. Lancet 2014;384: – 19. Aroda VR, Bailey TS, Cariou B, et al. Effect of ratio combinations with basal insulin 2228 2234 6. Liberati A, Altman DG, Tetzlaff J, et al. The adding insulin degludec to treatment in patients may represent a promising option to ad- PRISMA statement for reporting systematic re- with type 2 diabetes inadequately controlled vance basal insulin therapy or to initiate views and meta-analyses of studies that evalu- with metformin and liraglutide: a double-blind injectable therapy in patients with ate health care interventions: explanation and randomized controlled trial (BEGIN: ADD TO elaboration. Ann Intern Med 2009;151:W65– GLP-1 Study). Diabetes Obes Metab 2016;18: type 2 diabetes inadequately controlled 663–670 on oral agents. In the clinical manage- W94 7. Higgins JP, Altman DG, Gøtzsche PC, et al.; 20. Seino Y, Kaneko S, Fukuda S, et al. Combi- ment of these patients, this combina- Cochrane Bias Methods Group; Cochrane Statis- nation therapy with liraglutide and insulin in tion treatment may offer the same tical Methods Group. The Cochrane Collab- Japanese patients with type 2 diabetes: a glycemic control of intensified insulin oration’s tool for assessing risk of bias in 36-week, randomized, double-blind, parallel- group trial. J Diabetes Investig 2016;7:565–573 regimens (basal-plus or basal-bolus), randomised trials. BMJ 2011;343:d5928 8. Jadad AR, Moore RA, Carroll D, et al. Assess- 21. Vanderheiden A, Harrison L, Warshauer J, Li X, but with less hypoglycemia and reduced ing the quality of reports of randomized clinical Adams-Huet B, Lingvay I. Effect of adding weight. Moreover, the use of titratable trials: is blinding necessary? Control Clin Trials liraglutide vs placebo to a high-dose insulin reg- fixed-ratio combinations of GLP-1RAs 1996;17:1–12 imen in patients with type 2 diabetes: a random- plus basal insulin may improve meta- 9. Higgins JPT, Green S, Eds. Cochrane Hand- ized clinical trial. JAMA Intern Med 2016;176: 939–947 bolic control more than each single com- book for Systematic Reviews of Interventions Version 5.1.0. Oxford, The Cochrane Collabora- 22.LiCJ,LiJ,ZhangQM,etal.Efficacy and ponent. Longer studies are needed to tion, 2011 safety comparison between liraglutide as add-on assess tolerability, effectiveness, and 10.EggerM,DaveySmithG,SchneiderM, therapy to insulin and insulin dose-increase in costs of this combination before it can Minder C. Bias in meta-analysis detected by a Chinese subjects with poorly controlled type 2 be favored as standard of care for pa- simple, graphical test. BMJ 1997;315:629–634 diabetes and abdominal obesity. Cardiovasc fi Diabetol 2012;11:142 tients with type 2 diabetes when basal 11. Duval S, Tweedie R. Trim and ll: a simple funnel-plot-based method of testing and adjust- 23. de Wit HM, Vervoort GM, Jansen HJ, de insulin therapy is failing. ing for publication bias in meta-analysis. Bio- Grauw WJ, de Galan BE, Tack CJ. Liraglutide re- metrics 2000;56:455–463 verses pronounced insulin-associated weight 12. Buse JB, Bergenstal RM, Glass LC, et al. Use gain, improves glycaemic control and decreases Duality of Interest. K.E. received a consultancy of twice-daily exenatide in basal insulin-treated insulin dose in patients with type 2 diabetes: a fee from Eli Lilly and has held lectures for Eli Lilly, patients with type 2 diabetes: a randomized, 26 week, randomised clinical trial (ELEGANT). Sanofi, and Novo Nordisk. D.G. received a con- controlled trial. Ann Intern Med 2011;154: Diabetologia 2014;57:1812–1819 sultancy fee from Eli Lilly and has held lectures 103–112 24. Lane W, Weinrib S, Rappaport J, Hale C. The for Eli Lilly and Sanofi. No other potential 13. DeVries JH, Bain SC, Rodbard HW, et al.; effect of addition of liraglutide to high-dose in- conflicts of interest relevant to this article Liraglutide-Detemir Study Group. Sequential tensive insulin therapy: a randomized prospective were reported. intensification of metformin treatment in trial. Diabetes Obes Metab 2014;16:827–832 Author Contributions. M.I.M. and D.G. con- type 2 diabetes with liraglutide followed by 25. Mathieu C, Rodbard HW, Cariou B, et al.; ducted the literature search, data extraction, randomized addition of basal insulin prompted BEGIN: VICTOZA ADD-ON (NN1250-3948) study and data analysis and wrote the manuscript. by A1C targets. Diabetes Care 2012;35:1446– group. A comparison of adding liraglutide P.C. and D.G. did the statistical analyses. G.B. 1454 versus a single daily dose of insulin aspart to 624 Meta-analysis of RCTs of Insulin Plus GLP-1RA Diabetes Care Volume 40, April 2017

insulin degludec in subjects with type 2 diabetes 33. Gough SC, Bode B, Woo V, et al.; NN9068- 39. Reid T, Gao L, Gill J, et al. How much is too (BEGIN: VICTOZA ADD-ON). Diabetes Obes 3697 (DUAL-I) trial investigators. Efficacy and much? Outcomes in patients using high-dose Metab 2014;16:636–644 safety of a fixed-ratio combination of insulin insulin glargine. Int J Clin Pract 2016;70:56–65 26. Rosenstock J, Guerci B, Hanefeld M, et al.; degludec and liraglutide (IDegLira) compared 40. Meier JJ, Rosenstock J, Hincelin-Mery´ A, GetGoal Duo-2 Trial Investigators. Prandial op- with its components given alone: results of a et al. Contrasting effects of lixisenatide and lira- tions to advance basal insulin glargine therapy: phase 3, open-label, randomised, 26-week, glutide on postprandial glycemic control, gastric testing lixisenatide plus basal insulin versus in- treat-to-target trial in insulin-naive patients emptying, and safety parameters in patients with sulin glulisine either as basal-plus or basal-bolus with type 2 diabetes. Lancet Diabetes Endocri- type 2 diabetes on optimized insulin glargine with in type 2 diabetes: the GetGoal Duo-2 Trial. Di- nol 2014;2:885–893 or without metformin: a randomized, open-label abetes Care 2016;39:1318–1328 34. Lingvay I, Perez´ Manghi F, Garc´ıa- trial. Diabetes Care 2015;38:1263–1273 27. Shao N, Kuang HY, Hao M, Gao XY, Lin WJ, Hernandez´ P, et al.; DUAL V Investigators. Effect 41. Karagiannis T, Liakos A, Bekiari E, et al. Ef- Zou W. Benefits of exenatide on obesity and of insulin glargine up-titration vs insulin degludec/ ficacy and safety of once-weekly glucagon-like non-alcoholic fatty liver disease with elevated liraglutide on glycated hemoglobin levels in pa- peptide 1 receptor agonists for the manage- liver enzymes in patients with type 2 diabetes. tients with uncontrolled type 2 diabetes: the ment of type 2 diabetes: a systematic review Diabetes Metab Res Rev 2014;30:521–529 DUAL V randomized clinical trial. JAMA 2016; and meta-analysis of randomized controlled tri- 28. Rosenstock J, Fonseca VA, Gross JL, et al.; 315:898–907 als. Diabetes Obes Metab 2015;17:1065–1074 Harmony 6 Study Group. Advancing basal insu- 35. Aroda VR, Rosenstock J, Wysham C, et al.; 42. Esposito K, Chiodini P, Bellastella G, lin replacement in type 2 diabetes inadequately LixiLan-L Trial Investigators. LixiLan-L Trial Inves- Maiorino MI, Giugliano D. Proportion of pa- controlled with insulin glargine plus oral agents: tigators. Efficacy and safety of LixiLan, a titrat- tients at HbA1c target ,7% with eight classes a comparison of adding albiglutide, a weekly able fixed-ratio combination of insulin glargine of antidiabetic drugs in type 2 diabetes: system- GLP-1 receptor agonist, versus thrice-daily plus lixisenatide in type 2 diabetes inadequately atic review of 218 randomized controlled trials prandial insulin lispro. Diabetes Care 2014;37: controlled on basal insulin and metformin: the with 78 945 patients. Diabetes Obes Metab 2317–2325 LixiLan-L randomized trial. Diabetes Care 2016; 2012;14:228–233 29. Diamant M, Nauck MA, Shaginian R, et al.; 39:1972–1980 43. Nauck M, Rizzo M, Johnson A, Bosch- 4B Study Group. Glucagon-like peptide 1 recep- 36. Rosenstock J, Diamant M, Aroda VR, et al.; Traberg H, Madsen J, Cariou B. Once-daily tor agonist or bolus insulin with optimized basal LixiLan PoC Study Group. Efficacy and safety of liraglutide versus lixisenatide as add-on to met- insulin in type 2 diabetes. Diabetes Care 2014; LixiLan, a titratable fixed-ratio combination of formin in type 2 diabetes: a 26-week random- 37:2763–2773 lixisenatide and insulin glargine, versus insulin ized controlled clinical trial. Diabetes Care 2016; 30. Blonde L, Jendle J, Gross J, et al. Once- glargine in type 2 diabetes inadequately con- 39:1501–1509 weekly dulaglutide versus bedtime insulin glargine, trolled on metformin monotherapy: the LixiLan 44. Rodbard HW, Bode BW, Harris SB, et al.; both in combination with prandial insulin lispro, in proof-of-concept randomized trial. Diabetes Dual Action of Liraglutide and insulin degludec patients with type 2 diabetes (AWARD-4): a Care 2016;39:1579–1586 (DUAL) IV trial investigators. Safety and efficacy randomised, open-label, phase 3, non-inferiority 37. Rosenstock J, Aronson R, Grunberger G, of insulin degludec/liraglutide (IDegLira) added study. Lancet 2015;385:2057–2066 et al.; LixiLan-O Trial Investigators. Benefits of to sulphonylurea alone or to sulphonylurea and 31. FLAT-SUGAR Trial Investigators. Glucose LixiLan, a titratable fixed-ratio combination of metformin in insulin-na¨ıve people with type 2 variability in a 26-week randomized comparison insulin glargine plus lixisenatide, versus insulin diabetes: the DUAL IV trial. Diabet Med 2017; of mealtime treatment with rapid-acting insulin glargine and lixisenatide monocomponents in 34:189–196 versus GLP-1 agonist in participants with type 2 type 2 diabetes inadequately controlled with 45. Terrin N, Schmid CH, Lau J, Olkin I. Adjusting diabetes at high cardiovascular risk. Diabetes oral agents: the LixiLan-O randomized trial. for publication bias in the presence of hetero- Care 2016;39:973–981 Diabetes Care 2016;39:2026–2035 geneity. Stat Med 2003;22:2113–2126 32. Buse JB, Vilsbøll T, Thurman J, et al.; 38. Khunti K, Nikolajsen A, Thorsted BL, 46. Palmer SC, Mavridis D, Nicolucci A, et al. NN9068-3912 (DUAL-II) Trial Investigators. Andersen M, Davies MJ, Paul SK. Clinical inertia Comparison of clinical outcomes and adverse Contribution of liraglutide in the fixed-ratio with regard to intensifying therapy in people events associated with glucose-lowering drugs combination of insulin degludec and liraglutide with type 2 diabetes treated with basal insulin. in patients with type 2 diabetes: a meta-analysis. (IDegLira). Diabetes Care 2014;37:2926–2933 Diabetes Obes Metab 2016;18:401–409 JAMA 2016;316:313–324