REVIEW Annals of Internal Medicine Benefits and Harms of Once-Weekly -like -1 Receptor Agonist Treatments A Systematic Review and Network Meta-analysis Francesco Zaccardi, MD; Zin Zin Htike, MD; David R. Webb, PhD; Kamlesh Khunti, PhD; and Melanie J. Davies, MD

Background: Once-weekly glucagon-like peptide-1 receptor and for fasting plasma glucose (–0.7 agonists (GLP-1RAs) are new drugs for the treatment of type 2 mmol/L [CI, –1.1 to –0.2 mmol/L]; –12.6 mg/dL [CI, –19.8 to –3.6 diabetes. mg/dL]), and taspoglutide, 20 mg, and , 0.75 mg, for body weight (–1.5 kg [CI, –2.2 to –0.8]). Clinically marginal or no Purpose: To summarize evidence for the cardiometabolic effi- differences were found for blood pressure, blood lipid levels, cacy and adverse effects of once-weekly GLP-1RAs in adults with and C-reactive protein levels. Once-weekly exenatide increased . heart rate compared with albiglutide and dulaglutide (1.4 to 3.2 Data Sources: Electronic databases (PubMed, Web of Science, beats/min). Among once-weekly GLP-1RAs, the risk for hypogly- Cochrane Central Register of Controlled Trials, U.S. Food cemia was similar, whereas taspoglutide, 20 mg, had the great- and Drug Administration, European Medicines Agency, est risk for nausea (odds ratios, 1.9 to 5.9). ClinicalTrials.gov) and congress abstracts from inception Limitation: Data were unavailable for , definitions through 26 September 2015. of outcomes were heterogeneous, the last-observation-carried- Study Selection: Randomized, controlled trials (≥24 weeks of forward imputation method was used in 73% of trials, and pub- follow-up) studying albiglutide, dulaglutide, once-weekly ex- lication bias is possible. enatide, semaglutide, and taspoglutide and reporting a cardio- Conclusion: Compared with other once-weekly GLP-1RAs, du- metabolic (primary outcome, hemoglobin A1c [HbA1c]) or safety laglutide, 1.5 mg; once-weekly exenatide; and taspoglutide, 20 outcome. mg, showed a greater reduction of HbA1c, fasting plasma glu- Data Extraction: Extraction was done in duplicate, and risk of cose, and body weight. Taspoglutide, 20 mg, had the highest bias was assessed. No language restriction was applied. risk for nausea; risk for hypoglycemia among once-weekly GLP- 1RAs was similar. Data Synthesis: 34 trials (21 126 participants) were included. Compared with placebo, all once-weekly GLP-1RAs reduced Primary Funding Source: Sanofi Aventis (grant to the Univer- sity of Leicester). HbA1c and fasting plasma glucose; taspoglutide, 20 mg, once- weekly exenatide, and dulaglutide, 1.5 mg, reduced body weight. Among once-weekly GLP-1RAs, the greatest differences Ann Intern Med. 2016;164:102-113. doi:10.7326/M15-1432 www.annals.org were found between dulaglutide, 1.5 mg, and taspoglutide, 10 For author affiliations, see end of text.

mg, for HbA1c (–0.4% [95% CI, –0.7% to –0.2%]), once-weekly This article was published at www.annals.org on 8 December 2015.

ype 2 diabetes mellitus is a progressive metabolic option in patients receiving with or without Tdisorder associated with an increased risk for vascu- another glucose-lowering treatment if individualized lar diseases and whose main and diagnostic character- metabolic targets are not achieved (7). istic is hyperglycemia (1–3). In the past decade, the The first 2 approved GLP-1RAs are administered as number of available treatments for hyperglycemia has subcutaneous daily injections (twice-daily exenatide increased significantly (4). Glucagon-like peptide-1 re- and once-daily ). A direct comparison of ceptor agonists (GLP-1RAs) are a relatively new class of these “daily” drugs has shown differences in terms of drug that stimulate and inhibit glucagon secre- tolerability, efficacy, and (8). More re- tion, slow gastric emptying, and reduce food intake (5). cently, GLP-1RAs that are administered once weekly In clinical studies, these drugs improve glucose control have emerged, reducing the number of injections and reduce body weight, without an increased risk for and side effects and potentially improving patient hypoglycemia (6). The American Diabetes Association compliance (6). (ADA) and European Association for the Study of Dia- Several randomized, controlled trials (RCTs) have betes (EASD) recommend GLP-1RAs as a therapeutic assessed the efficacy and safety of once-weekly GLP- 1RAs compared with daily GLP-1RAs or other glucose- See also: lowering therapies. However, to date, no direct com- parisons between once-weekly GLP-1RAs have been Editorial comment ...... 127 available. Web-Only In the absence of direct evidence, network meta- Supplement analysis is an increasingly used statistical methodology that allows the estimation of the comparative effective- CME quiz ness of multiple treatments (9, 10). In this context, we

102 © 2016 American College of Physicians Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Comparative Clinical Profiles of Once-Weekly GLP-1RAs REVIEW

conducted a systematic review and network meta- not possible to extract relevant information for the pri- analysis to assess the comparative efficacy and safety of mary outcome from published reports, we searched for the once-weekly GLP-1RAs albiglutide, dulaglutide, ex- trial results in ClinicalTrials.gov and contacted study enatide, semaglutide, and taspoglutide. authors. Study-level quality was assessed by using the items METHODS reported in the Cochrane risk-of-bias tool (random- Data Sources and Searches sequence generation, allocation concealment, blinding This study was done according to a prespecified of participants and personnel and outcome assess- protocol (Supplement 1, available at www.annals.org) ment, and incomplete data and selective reporting) and followed the standard guidelines for the conduct (14). If the 2 independent reviewers disagreed on the and reporting of systematic reviews and network meta- eligibility of an article, extracted information, or quality analysis (11–13). We searched the following databases assessment, consensus was reached by reevaluation of from inception to 26 September 2015: PubMed, ISI the article and consultation with a third reviewer. Web of Science, Cochrane Central Register of Con- trolled Trials, ClinicalTrials.gov, U.S. Food and Drug Data Synthesis and Analysis Administration (www.fda.gov), European Medicines Stata, version 14 (Stata Corp, College Station, Agency (www.ema.europa.eu/ema), and the abstract Texas), was used for all analyses, and results are re- databases of major diabetes conferences (ADA and ported with 95% CIs. We performed pairwise random- EASD from 2012 onward). Reference lists of eligible effects meta-analyses by using the Knapp–Hartung studies, as well as systematic reviews and meta- method with the metareg command (15). We did net- analyses of GLP-1RAs, were manually scanned for addi- work meta-analyses within a frequentist framework us- tional relevant studies. No language restriction was ap- ing the network suite (16). By using the mvmeta com- plied. Detailed information on the search strategy is mand (17), network performs network meta-analysis as provided in Supplement 1. multivariate random-effects meta-analysis and meta- regression, as recently proposed (18); the analysis as- Study Selection sumes that all treatment contrasts have the same heter- Phase 3 RCTs in adults with type 2 diabetes that ogeneity variance (18). We used the network rank lasted 24 weeks or more were included. We included option to estimate the ranking probabilities and the RCTs with at least 1 once-weekly GLP-1RA study group netleague command to report relative treatment effects (albiglutide, dulaglutide, exenatide, semaglutide, and for all pairwise comparisons estimated with the network taspoglutide), regardless of the comparator (placebo meta-analysis (19). or another glucose-lowering drug). We required that Because our goal was to assess differences by RCTs report data on cardiometabolic outcomes (pri- comparing once-weekly GLP-1RAs with each other, we mary outcome, HbA ; secondary outcomes, fasting 1c combined twice-daily exenatide and once-daily liraglu- plasma glucose; body weight; systolic and diastolic tide treatments into one group of daily GLP-1RA thera- blood pressure; heart rate; C-reactive protein; and pies. We also combined and detemir blood lipid values [total, low-density lipoprotein, and (basal insulin) treatments. We defined a single group high-density lipoprotein cholesterol and triglycerides]) for albiglutide because the majority of RCTs were de- or safety outcomes (documented or symptomatic hypo- signed to titrate the drug to 50 mg if necessary, and glycemic events; severe hypoglycemic events; nausea dose-specific data were available only for 2 studies (20, and injection-site reaction events; fatal and nonfatal 21); for these studies, we used data on the 50-mg dose. cardiovascular and cancer events; all-cause mortality). We report the characteristics and summary data of We excluded RCTs involving patients with chronic kid- included studies. For both pairwise and network meta- ney disease. Although the clinical trial program of ta- analyses, we used arm-specific mean differences from spoglutide was stopped in 2010 and development has baseline and odds ratios (ORs) as effect measures for since been suspended, we included RCT data because continuous and dichotomous data, respectively. We they contribute to indirect estimations. added 0.5 when studies reported no events in 1 treat- Data Extraction and Quality Assessment ment group (18). For the primary outcome, we re- Two authors independently performed the litera- ported a random-effect pairwise meta-analysis and es- ture search and extracted study information by using timated heterogeneity across studies by using the I2 standardized predefined forms. Extracted data in- statistic. For each outcome, we summarized the evi- cluded study characteristics and outcome measured dence by using a network diagram (22). We present (group-specific number of participants and mean differ- results against a common comparator (placebo) in for- ence and SE [or SD] for continuous outcomes; total est plots and show comparisons across GLP-1RAs in number of participants and participants with event for forest plots and tables; we also display graphically the dichotomous outcomes). Data were extracted on an ranking probabilities (23). Within the networks, we as- intention-to-treat basis; that is, every participant who sessed consistency between direct and indirect evi- underwent randomization contributed to data accord- dence by using the “design by treatment” interaction ing to their randomized treatment assignment. When model (24). published studies reported outcomes for different du- We performed sensitivity analyses to assess the ro- rations of follow-up, the longest was used. When it was bustness of our results. We considered separate daily

www.annals.org Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 103 Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 REVIEW Comparative Clinical Profiles of Once-Weekly GLP-1RAs

Table. Baseline Characteristics of the Included Studies

Study, Year (Reference) Study Acronym Background Therapy Once-Weekly GLP-1 Comparator Wysham et al, 2014 (51) AWARD-1 MET + TZD DUL 0.75/1.5 mg PLA, exenatide twice daily Giorgino et al, 2015 (52) AWARD-2 MET ± OADs; SU ± OADs DUL 0.75/1.5 mg Insulin glargine Umpierrez et al, 2014 (49) AWARD-3 Diet + exercise DUL 0.75/1.5 mg MET Blonde et al, 2015 (53) AWARD-4 Insulin (basal/basal + prandial/premixed) ± OADs DUL 0.75/1.5 mg§ Insulin glargine§ Weinstock et al, 2015 (41) AWARD-5 MET DUL 0.75/1.5 mg Dungan et al, 2014 (34) AWARD-6 MET DUL 1.5 mg Liraglutide once daily Drucker et al, 2008 (33) DURATION-1 MET, SU, TZD, or any two combination EOW 2 mg Exenatide twice daily Bergenstal et al, 2010 (27) DURATION-2 MET EOW 2 mg PIO, sitagliptin Diamant et al, 2014 (31) and DURATION-3 MET ± SU EOW 2 mg Insulin glargine 2010 (32)͉͉ Russell-Jones et al, 2012 (48) DURATION-4 Diet + exercise EOW 2 mg PIO, sitagliptin, MET Blevins et al, 2011 (28) DURATION-5 MET, SU, TZD, or any combination EOW 2 mg Exenatide twice daily Buse et al, 2013 (29) DURATION-6 MET, SU, MET + SU, MET + PIO EOW 2 mg Liraglutide once daily Reusch et al, 2014 (45) HARMONY 1 PIO ± MET ALB 30 mg PLA Nauck et al, 2013 (20) HARMONY 2 Diet + exercise ALB 30/50 mg¶ PLA Ahrén et al, 2014 (25) HARMONY 3 MET ALB 30 to 50 mg PLA, sitagliptin, Weissman et al, 2014 (50) HARMONY 4 MET ± SU ALB 30 to 50 mg Insulin glargine Home et al, 2015 (37) HARMONY 5 MET + SU ALB 30 to 50 mg PLA, PIO Rosenstock et al, 2014 (47) HARMONY 6 Insulin glargine/detemir/NPH insulin ± OADs ALB 30 to 50 mg Pratley et al, 2014 (42) HARMONY 7 MET, SU, TZD, or any combination ALB 30 to 50 mg Liraglutide once daily Raz et al, 2012 (44) T-emerge 1 Diet + exercise TAS 10/20 mg PLA Rosenstock et al, 2013 (46) T-emerge 2 MET, TZD, MET + TZD TAS 10/20 mg Exenatide twice daily Henry et al, 2012 (35) T-emerge 3 MET + PIO TAS 10/20 mg PLA Bergenstal et al, 2012 (26) T-emerge 4 MET TAS 10/20 mg PLA, sitagliptin Nauck et al, 2013 (40) T-emerge 5 MET + SU TAS 10/20 mg Insulin glargine Pratley et al, 2013 (43) T-emerge 6 SU ± MET TAS 10/20 mg PIO Hollander et al, 2013 (36) T-emerge 7 MET TAS 20 mg PLA Miyagawa et al, 2015 (54) – Diet + exercise DUL 0.75 mg PLA, liraglutide once daily Araki et al, 2015 (55) – SUs, BIGs, SUs + BIGs DUL 0.75 mg Insulin glargine NCT01648582 (56) – MET, SU, MET + SU DUL 0.75/1.5 mg Insulin glargine Wang et al, 2015 (57) – Diet + exercise ± OAD DUL 0.75/1.5 mg Glimepiride Davies et al, 2013 (30) – MET ± SU EOW 2 mg Inagaki et al, 2012 (38) – BIG, BIG + TZD, BIG + SU, BIG + TZD + SU EOW 2 mg Insulin glargine Ji et al, 2013 (39) – MET, SU, TZD, MET + SU, MET + TZD, SU + TZD EOW 2 mg Exenatide twice daily NCT01733758 (21) – Diet + exercise, OAD ALB 30/50 mg¶ PLA, liraglutide once daily ALB = albiglutide; BIG = ; DUL = dulaglutide; EOW = once-weekly exenatide; GLP-1 = glucagon-like peptide-1; MET = metformin; NPH = neutral protamine Hagedorn; OAD = oral antihyperglycemic drug; PIO = ; PLA = placebo; SU = ; TAS = taspoglutide; TZD = . * When not reported for the overall population, values have been estimated as weighted means. † 26 wk for body weight, fasting plasma glucose, and side effects. ‡ 26 wk for body weight. § Both groups also added insulin lispro. ͉͉ Data from reference 31 (156 wk; year 2014) for continuous outcomes, nausea, and injection-site side effects; data from reference 32 (26 wk; year 2010) for hypoglycemic events. ¶ Data and analyses are reported for the higher dose (50 mg). ** 24 wk for fasting plasma glucose and body weight; 104 wk for side effects. †† 156 wk for side effects.

GLP-1RA drugs and basal and restricted the and 2 abstracts [20, 57]) based on 34 unique RCTs ful- analyses for studies with similar duration of follow-up filled our inclusion criteria (Appendix Figure 1, avail- (24 to 26 weeks). For hypoglycemic events, we also ex- able at www.annals.org). No RCT for semaglutide was cluded trials in which once-weekly GLP-1RAs were found. Included RCTs were published between Octo- combined with either insulin or sulfonylurea. ber 2008 and September 2015 and had a total of 21 126 participants with type 2 diabetes and study du- Role of the Funding Source rations of 24 to 156 weeks (Table). Dr. Zaccardi is a Clinical Research Fellow funded by Overall, the risk of bias for the domains included in an unrestricted educational grant from Sanofi Aventis the Cochrane tool of risk assessment were judged to to the University of Leicester. The funding source had be low, high, and unclear in 51.0%, 24.5%, and 24.5% no role in the design, conduct, or analysis of the study or the decision to submit the manuscript for of the cases, respectively (Supplement 2, available at publication. www.annals.org). In 24 of 34 RCTs (71%), risk of bias was high for incomplete outcome data and unclear for blinding of outcome assessment, whereas 21 RCTs RESULTS (62%) had a high risk of bias in the domain of blinding Study Characteristics and Quality of participants and personnel. Conversely, the risk of Of 1065 identified records, 35 reports (31 full-text bias for random-sequence generation, allocation con- articles [25–55], 2 studies in ClinicalTrials.gov [21, 56], cealment, and selective reporting was considered low.

104 Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 www.annals.org Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Comparative Clinical Profiles of Once-Weekly GLP-1RAs REVIEW

Table —Continued

Study Duration, wk Sample Size, n Women, % Age, y* Duration of Diabetes, y* Hemoglobin A1c Value, %* 52† 976 41.6 55.6 9.0 8.1 78 810 48.7 56.7 9.0 8.1 52‡ 807 55.9 55.5 3.0 7.6 52 884 46.5 59.4 12.7 8.5 104 921 52.7 54.1 7.0 8.1 26 599 52.1 56.7 7.2 8.1 30 295 46.8 55.0 6.0 8.3 26 491 47.3 52.3 5.7 8.5 156 456 46.7 58.0 7.9 8.3

26 820 42.1 53.9 2.7 8.5 24 252 42.5 55.5 7.0 8.5 26 911 45.2 57.0 8.5 8.5 52 301 40.2 55.0 7.9 8.1 52 296 46.0 53.0 4.0 8.1 104 1012 53.5 54.4 6.0 8.1 52 745 43.9 55.4 8.7 8.3 52 663 46.8 55.2 8.9 8.2 26 563 52.9 55.5 11.0 8.5 32 812 49.6 55.6 8.3 8.2 24 368 60.3 55.0 2.4 7.6 52** 1149 47.4 56.9 6.7 8.3 24 313 46.7 54.2 7.7 8.1 52†† 546 44.6 55.9 5.9 8.0 24 1028 48.3 57.7 9.5 8.3 24 740 50.4 56.4 8.8 8.3 24 292 59.2 53.5 5.1 7.5 26 487 18.7 57.4 6.6 8.1 26 361 28.5 56.8 8.8 8.0 52 770 45.4 54.9 – – 26 807 46.1 52.8 3.7 7.9 26 216 33.8 58.5 7.5 8.4 26 427 32.1 56.8 9.0 8.5 26 678 45.9 55.5 8.1 8.7 24 330 25.1 57.8 – 8.1

The completion rate ranged from 60% to 97%, with indirect evidence were combined, the network meta-

25 RCTs using the last-observation-carried-forward im- analysis showed a mean reduction in HbA1c values putation for incomplete data outcome and 9 using a compared with placebo of –1.4% (95% CI, –1.6% to mixed-effects model for repeated measures (Supple- –1.2%) for dulaglutide, 1.5 mg; –1.3% (CI, –1.5% to ment 3, available at www.annals.org). Thirteen RCTs –1.1%) for once-weekly exenatide; –1.2% (CI, –1.4% used the last observation before hyperglycemia rescue, to –1.0%) for dulaglutide, 0.75 mg; –1.1% (CI, –1.3% to and data handling for rescued patients was not re- –0.9%) for taspoglutide, 20 mg; –1.0% (CI, –1.2% to ported in 21 studies. Other characteristics of the in- –0.8%) for albiglutide; and –0.9% (CI, –1.2% to –0.8%) cluded studies are reported in Supplements 4 through for taspoglutide, 10 mg (Appendix Figure 3, available 9 (available at www.annals.org). at www.annals.org). Statistical inconsistency for the whole network was not significant. Comparisons across Meta-analyses once-weekly GLP-1RAs showed a greater reduction in Details of available data are shown in Supplements HbA values with dulaglutide, 1.5 mg, compared with 4 through 6 for cardiometabolic outcomes and Supple- 1c dulaglutide, 0.75 mg, albiglutide, and taspoglutide, 10 ments 7 through 9 for safety outcomes; the results of mg and 20 mg, whereas no difference was observed in network-specific inconsistency tests are presented in Supplement 10 (available at www.annals.org). Net- comparison with once-weekly exenatide (Figure 2 and works of evidence are graphically displayed in Figure 1. Supplement 11, available at www.annals.org). The ranking probabilities and the mean rank for each drug included in the analysis are shown in Appendix Figure

Primary Outcome: HbA1c 4 (available at www.annals.org) and Supplement 12 Data on HbA1c were available from all RCTs. Direct (available at www.annals.org), respectively. pairwise random-effects meta-analyses showed differ- Sensitivity analyses considering separate daily GLP- ent effects of once-weekly GLP-1RAs compared with 1RAs (once-daily liraglutide and twice-daily exenatide) placebo or other glucose-lowering drugs, from a 1.6% and basal insulins (detemir and glargine) showed re-

reduction to a 0.3% increase in HbA1c (Appendix Fig- sults consistent with the main analysis (Supplement 13, ure 2, available at www.annals.org). When direct and available at www.annals.org). The estimates from stud-

www.annals.org Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 105 Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 REVIEW Comparative Clinical Profiles of Once-Weekly GLP-1RAs

Figure 1. Network maps for cardiometabolic and safety outcomes.

Systolic Blood PressureDiastolic Blood Pressure Heart Rate C-Reactive Protein Level

DUL 1.5 DUL 1.5 DUL 1.5 EOW EOW DUL 0.75 PIO DUL 0.75 DUL 0.75 EOW EOW GLIM GLIM BASAL BASAL BASAL PLA MET MET GLIM ALB ALB ALB BASAL

PIO PIO MET dGLP1 dGLP1 dGLP1 TAS 10 PLA PLA TAS 20 PIO dGLP1 SITA SITA TAS 20 SITA TAS 10 TAS 10 PLA TAS 20

Total Cholesterol LevelLDL Cholesterol Level HDL Cholesterol Level Triglyceride Level

DUL 1.5 DUL 1.5 DUL 1.5 DUL 1.5 EOW DUL 0.75 EOW DUL 0.75 EOW DUL 0.75 EOW DUL 0.75 GLIM BASAL GLIM BASAL GLIM BASAL GLIM BASAL

PIO PIO PIO PIO ALB ALB ALB ALB

PLA dGLP1 PLA dGLP1 PLA dGLP1 PLA dGLP1 SITA TAS 20 SITA TAS 20 SITA TAS 20 SITA TAS 20 TAS 10 TAS 10 TAS 10 TAS 10

HbA1c ValueFasting Plasma Glucose Level Body Weight

EOW DUL 1.5 EOW DUL 1.5 EOW DUL 1.5 GLIM DUL 0.75 GLIM DUL 0.75 GLIM DUL 0.75 MET BASAL MET BASAL MET BASAL

PIO ALB PIO ALB PIO ALB

PLA dGLP1 PLA dGLP1 PLA dGLP1

RAPID TAS 20 RAPID TAS 20 RAPID TAS 20 SITA TAS 10 SITA TAS 10 SITA TAS 10

Documented or Nausea Symptomatic Hypoglycemia

EOW DUL 1.5 EOW DUL 1.5 GLIM DUL 0.75 GLIM DUL 0.75

MET BASAL MET BASAL

PIO ALB PIO ALB

PLA dGLP1 PLA dGLP1 RAPID RAPID TAS 20 TAS 20 SITA SITA TAS 10 TAS 10

Nodes represent the treatments being compared; their size is proportional to the number of participants. Edges represent the available direct comparisons between pairs of treatments, and their width is proportional to the number of trials comparing every pair. ALB = albiglutide; BASAL = basal insulin; dGLP1 = daily glucagon-like peptide-1 receptor agonists; DUL 0.75 = dulaglutide, 0.75 mg; DUL 1.5 = dulaglutide, 1.5 mg; EOW =

exenatide once-weekly; GLIM = glimepiride; HbA1c = hemoglobin A1c; HDL = high-density lipoprotein; LDL = low-density lipoprotein; MET = metformin; PIO = pioglitazone; PLA = placebo; RAPID = rapid insulin; SITA = sitagliptin; TAS 10 = taspoglutide, 10 mg; TAS 20 = taspoglutide, 20 mg.

ies with a follow-up duration of 24 to 26 weeks (28 Secondary Cardiometabolic Outcomes

RCTs) similarly showed a greater HbA1c reduction with Data on fasting plasma glucose were available dulaglutide, 1.5 mg; no difference was found com- from 31 RCTs. Compared with placebo, once-weekly pared with albiglutide, possibly owing to the small exenatide reduced fasting plasma glucose levels by number of albiglutide studies in this analysis compared –2.2 mmol/L (CI, –2.6 to –1.8 mmol/L) (–39.6 mg/dL [CI, with the main analysis (3 vs. 8) (Supplement 14, avail- –46.8 to –32.4 mg/dL]), followed by dulaglutide, 1.5 mg able at www.annals.org). (–2.2 mmol/L [–2.6 to –1.7 mmol/L]; –39.6 mg/dL [CI,

106 Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 www.annals.org Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Comparative Clinical Profiles of Once-Weekly GLP-1RAs REVIEW

Figure 2. Comparison of once-weekly glucagon-like peptide-1 receptor agonists for cardiometabolic outcomes.

Comparison

DUL 1.5 DUL 0.75 0.19 (0.04 to 0.34) 0.95 (0.44 to 1.46) 0.42 (0.09 to 0.75) 0.03 (−0.19 to 0.25) ALB 0.40 (0.17 to 0.63) 0.56 (−0.16 to 1.28) 0.65 (0.18 to 1.12) 0.08 (−0.19 to 0.35) TAS 20 0.27 (0.03 to 0.51) −0.52 (−1.26 to 0.22) 0.30 (−0.21 to 0.81) −0.02 (−0.29 to 0.25) TAS 10 0.41 (0.17 to 0.65) 0.11 (−0.65 to 0.87) 0.52 (−0.01 to 1.05) 0.02 (−0.26 to 0.30) EOW 0.09 (−0.12 to 0.30) −0.04 (−0.69 to 0.61) −0.02 (−0.45 to 0.41) −0.01 (−0.26 to 0.24)

DUL 0.75 ALB 0.21 (−0.01 to 0.43) −0.38 (−1.06 to 0.30) 0.23 (−0.22 to 0.68) 0.05 (−0.22 to 0.32) TAS 20 0.09 (−0.14 to 0.32) −1.46 (−2.17 to −0.75) −0.12 (−0.62 to 0.38) −0.05 (−0.32 to 0.22) TAS 10 0.22 (−0.02 to 0.46) −0.83 (−1.56 to −0.10) 0.11 (−0.41 to 0.63) −0.01 (−0.29 to 0.27) EOW −0.10 (−0.30 to 0.10) −0.99 (−1.62 to −0.36) −0.44 (−0.85 to −0.03) −0.04 (−0.29 to 0.21)

ALB TAS 20 −0.13 (−0.36 to 0.10) −1.08 (−1.81 to −0.35) −0.35 (−0.85 to 0.15) −0.10 (−0.26 to 0.06) TAS 10 0.01 (−0.22 to 0.24) −0.45 (−1.21 to 0.31) −0.13 (−0.65 to 0.39) −0.06 (−0.23 to 0.11) EOW −0.32 (−0.54 to −0.10) −0.61 (−1.32 to 0.10) −0.68 (−1.13 to −0.23) −0.09 (−0.24 to 0.06)

TAS 20 TAS 10 0.14 (−0.03 to 0.31) 0.63 (0.07 to 1.19) 0.23 (−0.15 to 0.61) 0.04 (−0.08 to 0.16) EOW −0.19 (−0.42 to 0.04) 0.47 (−0.25 to 1.19) −0.32 (−0.82 to 0.18) 0.01 (−0.14 to 0.16)

TAS 10 EOW −0.33 (−0.57 to −0.09) −0.16 (−0.89 to 0.57) −0.55 (−1.07 to −0.03) −0.03 (−0.19 to 0.13)

−0.5 0 0.5 −2 −1 0 1 2 −2 −1 0 1 −0.50 −0.25 0 0.25 0.50

HbA1c Value, % Body Weight, kg Fasting Plasma Glucose Level, mmol/L* Total Cholesterol Level, mmol/L†

Comparison

DUL 1.5 DUL 0.75 −0.02 (−0.18 to 0.14) −0.05 (−0.09 to −0.01) 0.30 (0.09 to 0.51) 0.51 (−0.52 to 1.54) ALB 0.15 (−0.03 to 0.33) −0.03 (−0.08 to 0.02) 0.14 (−0.14 to 0.42) −0.53 (−2.48 to 1.42) TAS 20 0.02 (−0.18 to 0.22) −0.05 (−0.10 to −0.00) 0.12 (−0.17 to 0.41) −0.65 (−2.28 to 0.98) TAS 10 0.04 (−0.15 to 0.23) −0.06 (−0.11 to −0.01) 0.16 (−0.13 to 0.45) −0.29 (−1.91 to 1.33) EOW 0.00 (−0.17 to 0.17) −0.03 (−0.07 to 0.01) 0.25 (−0.15 to 0.65) −1.17 (−2.59 to 0.25)

DUL 0.75 ALB 0.17 (−0.02 to 0.36) 0.02 (−0.03 to 0.07) −0.16 (−0.42 to 0.10) −1.04 (−2.93 to 0.85) TAS 20 0.04 (−0.16 to 0.24) 0.00 (−0.05 to 0.05) −0.18 (−0.45 to 0.09) −1.16 (−2.72 to 0.40) TAS 10 0.06 (−0.15 to 0.27) −0.01 (−0.06 to 0.04) −0.14 (−0.42 to 0.14) −0.80 (−2.35 to 0.75) EOW 0.02 (−0.16 to 0.20) 0.02 (−0.02 to 0.06) −0.05 (−0.44 to 0.34) −1.67 (−3.05 to −0.29)

ALB TAS 20 −0.13 (−0.26 to −0.00) −0.02 (−0.05 to 0.01) −0.02 (−0.17 to 0.13) −0.12 (−2.13 to 1.89) TAS 10 −0.10 (−0.23 to 0.03) −0.03 (−0.06 to −0.00) 0.02 (−0.14 to 0.18) 0.24 (−1.77 to 2.25) EOW −0.15 (−0.26 to −0.04) 0.00 (−0.03 to 0.03) 0.11 (−0.20 to 0.42) −0.63 (−2.60 to 1.34)

TAS 20 TAS 10 0.02 (−0.06 to 0.10) −0.01 (−0.04 to 0.02) 0.04 (−0.05 to 0.13) 0.36 (−0.83 to 1.55) EOW −0.02 (−0.15 to 0.11) 0.02 (-0.01 to 0.05) 0.13 (−0.16 to 0.42) −0.51 (−2.13 to 1.11)

TAS 10 EOW −0.04 (−0.17 to 0.09) 0.03 (0.00 to 0.06) 0.09 (−0.21 to 0.39) −0.87 (−2.49 to 0.75)

−0.25 0 0.25 −0.10 −0.05 0 0.05 0.10 −0.5 0 0.5 −3 −2 −1 0 1 2 3 LDL Cholesterol, mmol/L† HDL Cholesterol Level, mmol/L† Triglyceride Level, mmol/L‡ Systolic Blood Pressure, mm Hg

Comparison

DUL 1.5 DUL 0.75 0.00 (−0.62 to 0.62) 0.02 (−0.91 to 0.95) – ALB −0.27 (−1.52 to 0.98) −1.75 (−3.59 to 0.09) – TAS 20 −0.90 (−1.91 to 0.11) – – TAS 10 −0.68 (−1.70 to 0.34) – – EOW −0.55 (−1.41 to 0.31) 1.42 (0.01 to 2.83) –

DUL 0.75 ALB −0.27 (−1.50 to 0.96) −1.77(−3.56 to 0.02) – TAS 20 −0.91 (−1.88 to 0.06) – – TAS 10 −0.68 (−1.65 to 0.29) – – EOW −0.55 (−1.41 to 0.31) 1.40 (0.01 to 2.79) –

ALB TAS 20 −0.64 (−1.97 to 0.69) – – TAS 10 −0.41 (−1.74 to 0.92) – – EOW −0.28 (−1.60 to 1.04) 3.17 (1.18 to 5.16) –

TAS 20 TAS 10 0.23 (−0.48 to 0.94) – −1.59 (−7.84 to 4.66) EOW 0.36 (−0.69 to 1.41) – −2.55 (−16.10 to 11.00)

TAS 10 EOW 0.13 (−0.93 to 1.19) – −0.96 (−14.42 to 12.50)

−2 −1 0 1 2 −6−3 0 3 6 −20−10 0 1 0 2 0 Diastolic Blood Pressure, mm Hg Heart Rate, beats/min C-Reactive Protein Level, nmol/L

Data are reported as mean differences (95% CIs) and indicate differences versus the reference drug (for example, compared with dulaglutide, 1.5

mg, albiglutide increases HbA1c by 0.40% [CI, 0.17% to 0.63%]). ALB = albiglutide; DUL 0.75 = dulaglutide, 0.75 mg; DUL 1.5 = dulaglutide, 1.5 mg; EOW = exenatide once-weekly; HbA1c = hemoglobin A1c; HDL = high-density lipoprotein; LDL = low-density lipoprotein; TAS 10 = taspoglutide, 10 mg; TAS 20 = taspoglutide, 20 mg. * To convert values to mg/dL, divide by 0.0555. † To convert values to mg/dL, divide by 0.0259. ‡ To convert values to mg/dL, divide by 0.0113.

www.annals.org Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 107 Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 REVIEW Comparative Clinical Profiles of Once-Weekly GLP-1RAs

Figure 3. Comparison of once-weekly glucagon-like peptide-1 receptor agonists for safety outcomes.

Comparison Odds Ratio (95% CI) Odds Ratio (95% CI) DUL 1.5 DUL 0.75 0.93 (0.69 to 1.26) 0.56 (0.45 to 0.69) ALB 0.68 (0.43 to 1.06) 0.33 (0.23 to 0.47) TAS 20 0.63 (0.38 to 1.04) 1.92 (1.28 to 2.89) TAS 10 0.72 (0.43 to 1.20) 1.43 (0.94 to 2.18) EOW 0.78 (0.52 to 1.16) 0.55 (0.40 to 0.75)

DUL 0.75 ALB 0.72 (0.45 to 1.16) 0.58 (0.40 to 0.85) TAS 20 0.68 (0.40 to 1.13) 3.45 (2.24 to 5.31) TAS 10 0.77 (0.46 to 1.30) 2.56 (1.67 to 3.93) EOW 0.83 (0.55 to 1.27) 0.97 (0.69 to 1.36)

ALB TAS 20 0.93 (0.56 to 1.57) 5.88 (3.83 to 9.04) TAS 10 1.06 (0.63 to 1.79) 4.35 (2.85 to 6.64) EOW 1.15 (0.73 to 1.82) 1.67 (1.16 to 2.38)

TAS 20 TAS 10 1.14 (0.77 to 1.67) 0.74 (0.59 to 0.92) EOW 1.23 (0.74 to 2.05) 0.29 (0.19 to 0.43)

TAS 10 EOW 1.08 (0.64 to 1.81) 0.38 (0.25 to 0.58)

0.3 0.5 1 2 3 0.10 0.25 0.50 1 2 4 8 Documented or Nausea Symptomatic Hypoglycemia

Data are reported as odds ratios (95% CIs) and indicate differences versus the reference drug (for example, compared with dulaglutide, 1.5 mg, treatment with dulaglutide, 0.75 mg, is associated with an odds ratio for nausea of 0.56 [CI, 0.45 to 0.69]). ALB = albiglutide; DUL 0.75 = dulaglutide,

0.75 mg; DUL 1.5 = dulaglutide, 1.5 mg; EOW = exenatide once-weekly; HbA1c = hemoglobin A1c; HDL = high-density lipoprotein; LDL = low-density lipoprotein; TAS 10 = taspoglutide, 10 mg; TAS 20 = taspoglutide, 20 mg.

–46.8 to –30.6 mg/dL]); taspoglutide, 20 mg (–1.9 taspoglutide, 10 mg (–0.6 kg [CI, –1.3 to –0.1 kg]), and mmol/L [CI, –2.3 to –1.5 mmol/L]; –34.2 mg/dL [CI, no weight reduction for albiglutide (–0.2 kg [CI, –0.8, –41.4 to 27.0 mg/dL]); dulaglutide, 0.75 mg (–1.7 0.4 kg]) and dulaglutide, 0.75 mg (0.2 kg [CI, –0.4 to mmol/L [CI, –2.1 to –1.3 mmol/L]; –30.6 mg/dL [CI, –37.8 0.8 kg]) (Appendix Figure 3). Statistical inconsistency to 23.4 mg/dL]); taspoglutide, 10 mg (–1.6 mmol/L [CI, for the network was not significant. –2.1 to –1.2 mmol/L]; –28.8 mg/dL [CI, –37.8 to –21.6 Comparisons among once-weekly GLP-1RAs mg/dL]); and albiglutide (–1.5 mmol/L [CI, –1.9 to –1.1 showed greater reduction in body weight with taspo- mmol/L]; –27.0 mg/dL [CI, –34.2 to –19.8 mg/dL) glutide, 20 mg, compared with taspoglutide, 10 mg, (Appendix Figure 3). Statistical inconsistency for the albiglutide, and dulaglutide, 0.75 mg, and nonsignifi- network was not significant. cant differences compared with dulaglutide, 1.5 mg, When once-weekly GLP-1RAs were compared, ex- and once-weekly exenatide (Figure 2 and Supplements enatide was associated with a greater fasting plasma 11 and 12). glucose reduction than albiglutide, dulaglutide, 0.75 Data availability for all other cardiometabolic out- mg, and taspoglutide, 10 mg, and there were nonsig- comes ranged from 7 RCTs for C-reactive protein to 24 nificant differences from taspoglutide, 20 mg, and du- RCTs for systolic blood pressure. Comparisons among laglutide, 1.5 mg (Figure 2 and Supplements 11 and once-weekly GLP-1RAs were not significant for diastolic 15, available at www.annals.org). blood pressure, total cholesterol, or C-reactive protein Data on body weight were available from 32 RCTs. (Figure 2 and Supplements 11 and 15). Marginal differ- Network meta-analysis results showed a significant re- ences were found for low-density lipoprotein choles- duction of –1.3 kg (CI, –1.9 to –0.7 kg) for taspoglutide, terol (maximum difference, 0.15 mmol/L [5.80 mg/dL]), 20 mg, compared with placebo, followed by once- high-density lipoprotein cholesterol (0.06 mmol/L [2.32 weekly exenatide (–0.8 kg [CI, –1.5 to –0.1 kg]), dulaglu- mg/dL]), triglycerides (0.30 mmol/L [26.57 mg/dL]), tide, 1.5 mg (–0.8 kg [CI, –1.4 to –0.1 kg]), and and systolic blood pressure (1.7 mm Hg). Once-weekly

108 Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 www.annals.org Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Comparative Clinical Profiles of Once-Weekly GLP-1RAs REVIEW

exenatide increased heart rate compared with albiglu- dulaglutide, 0.75 mg; 6.9% for taspoglutide, 10 mg; tide (3.2 beats/min); dulaglutide, 0.75 mg (1.4 beats/ 7.1% for taspoglutide, 20 mg; 9.9% for albiglutide; and min); and dulaglutide, 1.5 mg (1.4 beats/min). Network 18.3% for once-weekly exenatide). Similarly, few cases inconsistency was significant for heart rate (Supplement of severe hypoglycemia, fatal and nonfatal cardiovascu- 10). lar and cancer events, and all-cause mortality events were reported (Supplements 7, 18, and 19, available at www.annals.org). Documented or Symptomatic Hypoglycemia and Nausea On the basis of 30 RCTs, network meta-analysis DISCUSSION showed an increased risk for documented or symptom- Although many studies have investigated the effi- atic hypoglycemia for albiglutide (OR, 1.82 [CI, 1.05 to cacy and safety of once-weekly GLP-1RAs compared 3.15]), taspoglutide, 10 mg (OR, 1.94 [CI, 1.03 to 3.62]), with other commonly used glucose-lowering drugs (in- once-weekly exenatide (OR, 2.08 [CI, 1.14 to 3.82]), du- cluding metformin, glimepiride, pioglitazone, sitaglip- laglutide, 0.75 mg (OR, 2.51 [CI, 1.39 to 4.54]), and tin, and insulin), to date no direct head-to-head com- dulaglutide, 1.5 mg (OR, 2.69 [CI to 1.51, 4.82]), but not parison between GLP-1RAs has been reported. Using a for taspoglutide, 20 mg (OR, 1.69 [CI, 0.92 to 3.14]), network meta-analysis, we aimed to assess the compar- compared with placebo (Appendix Figure 5, available ative efficacy and safety of once-weekly GLP-1RAs at www.annals.org). No differences were found among across a wide range of cardiometabolic and safety out- once-weekly GLP-1RAs (Figure 3 and Supplement 16, comes. The results suggested similar effects of these available at www.annals.org). Sensitivity analyses with- drugs on several cardiometabolic outcomes, such as out grouping of basal insulins and daily GLP-1RAs that blood pressure, blood lipids, and C-reactive protein, excluded studies with background therapy including and a modest increase in heart rate with once-weekly sulfonylurea or insulin showed similar results (Supple- exenatide versus albiglutide (mean difference, 3 beats/ ments 13 and 17, respectively [available at www.annals min). Conversely, we found appreciable differences for .org]). Owing to the presence of a disconnected net- HbA , fasting plasma glucose, and body weight, with work (only HARMONY 6 reported data on albiglutide 1c greater reductions for dulaglutide 1.5 mg, once-weekly and rapid insulin), sensitivity analysis limited to studies exenatide, and taspoglutide, 20 mg, compared with al- with a follow-up duration of 24 to 26 weeks was not biglutide and other once-weekly GLP-1RAs at lower possible. doses. We also found differences in the risk for nausea, Data on nausea were available from all RCTs. Com- which was highest for taspoglutide, 20 mg and 10 mg, pared with placebo, risk for nausea was increased with and dulaglutide, 1.5 mg. Conversely, the risk for docu- albiglutide (OR, 1.43 [CI, 1.02 to 2.01]), once-weekly mented or symptomatic hypoglycemic did not differ exenatide (OR, 2.38 [CI, 1.59 to 3.54]), dulaglutide, among once-weekly GLP-1RAs. 0.75 mg (OR, 2.44 [CI, 1.62 to 3.66]), dulaglutide, 1.5 Previous network meta-analyses (PubMed search, mg, (OR, 4.35 [CI, 2.93 to 6.47]), taspoglutide, 10 mg, September 2015) have investigated the efficacy and (OR, 6.19 [CI, 4.20 to 9.11]), and taspoglutide, 20 mg safety of GLP-1RAs in patients with type 2 diabetes (58– (OR, 8.32 [CI, 5.74 to 12.07]) (Appendix Figure 5). Ta- 67), with limited data on once-weekly GLP-1RAs. More- spoglutide, 20 mg, showed the greatest risk compared over, single or few outcomes were reported, making it with all other once-weekly GLP-1RAs, with odds ratios difficult to formulate a balanced overall assessment of ranging from 1.92 (CI, 1.28 to 2.89) versus dulaglutide, GLP-1RA therapies. Our study included data from re- 1.5 mg, to 5.88 (CI, 3.83 to 9.04) versus albiglutide (Fig- cent RCTs and compared once-weekly GLP-1RAs ure 3 and Supplements 12 and 16). A sensitivity analy- across a wide range of clinically relevant outcomes. Its sis without grouping basal insulins and daily GLP-1RAs aim is to assist decision makers in providing patient- showed the same results (Supplement 13). Data from centered care by balancing the potential risks and ben- studies with durations of 24 to 26 weeks confirmed a efits of individual drugs within the class of once-weekly higher risk for nausea for taspoglutide vs. once-weekly GLP-1RAs. Beyond HbA , therapeutic decisions should exenatide and dulaglutide, 0.75 mg, whereas no differ- 1c be based on other outcomes, including body weight, ences were found versus dulaglutide, 1.5 mg, and albi- risk for hypoglycemia, and gastrointestinal disorders. In glutide, possibly owing to the small number of studies this context, the results of our analyses could help cli- compared with the main analysis (4 vs. 8 for dulaglu- nicians to follow the ADA and EASD recommendations tide, 1.5 mg; 2 vs. 8 for albiglutide) (Supplement 14). (7), because several outcomes have been assessed simultaneously. Other Safety Outcomes Similar to other glucose-lowering agents (7), once-

Limited data for other safety outcomes were avail- weekly GLP-1RAs reduced HbA1c from 0.9% to 1.4% able, possibly because RCTs were designed to assess compared with placebo. Among once-weekly GLP-

differences in HbA1c (primary outcome) or outcomes 1RAs, the highest difference was found in favor of du- were selectively reported (publication bias). Overall, laglutide, 1.5 mg, versus taspoglutide, 10 mg (0.4%). there were 1668 heterogeneously defined cases of All once-weekly GLP-1RAs significantly reduced fasting injection-site side effects, from 22 cases of “bruising” to plasma glucose compared with placebo (reductions of 473 “reactions” (1.1% for dulaglutide, 1.5 mg; 1.9% for 1.5 mmol/L to 2.2 mmol/L [27 mg/dL to 39.6 mg/dL]);

www.annals.org Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 109 Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 REVIEW Comparative Clinical Profiles of Once-Weekly GLP-1RAs

the greatest difference among once-weekly GLP-1RAs Fourth, the small number of events and heteroge- was 0.7 mmol/L (12.6 mg/dL) in favor of once-weekly neous definitions did not allow analyses for severe hy- exenatide versus albiglutide. With the exception of al- poglycemic and injection-site side effects. The different biglutide and the lower doses of dulaglutide and ta- formulations of once-weekly GLP-1RAs (as injectable spoglutide, all other once-weekly GLP-1RAs signifi- microspheres [33], in combination with human albumin

cantly reduced body weight (maximum, 1.3 kg for [45], bound to modified human immunoglobulin G4 taspoglutide, 20 mg; minimum, 0.7 kg for dulaglutide, [51], or with a modified amino acid sequence [69]) has 1.5 mg) compared with placebo; the greatest differ- implications in terms of preparation for injection and ence was 1.5 kg in favor of taspoglutide, 20 mg, vs. results in a different risk for and nature of injection site dulaglutide, 0.75 mg. Our results would therefore sug- reactions. gest clinically significant differences on 3 key indicators Fifth, the fewer injections with once-weekly versus of metabolic control. Of note, comparisons among li- daily GLP-1RAs should increase quality of life, patient censed drugs showed no differences between once- satisfaction, and possibly adherence to therapy. How- weekly exenatide and the maintenance dose of dula- ever, we did not collect data to explore this hypothesis. glutide (1.5 mg) for all 3 metabolic outcomes, and both Sixth, the absence of significant differences for sev-

treatments reduced HbA1c to a better extent than eral cardiometabolic risk factors does not necessarily albiglutide. mean the absence of a difference for “hard” cardiovas- The results also evidenced differences in the risk cular outcomes. We collected data on cardiovascular, for nausea. Compared with placebo, all once-weekly cancer, and all-cause mortality events, although the GLP-1RAs increased significantly the risk for nausea, small numbers prevented any analysis. Published from 1.4-fold for albiglutide to 8.3-fold for taspoglu- individual-level meta-analysis (PubMed search, Sep- tide, 20 mg. Comparisons among once-weekly GLP- tember 2015) (Supplement 20, available at www.annals 1RAs showed a 1.9-fold to 5.9-fold greater risk with .org) suggested a neutral profile of once-weekly GLP- taspoglutide, 20 mg, versus all other once-weekly GLP- 1RAs drugs on hard cardiovascular end points. 1RAs. Discrepancies in the risk for nausea could be re- Ongoing RCTs will further elucidate the comparative lated to outcome definition, study design, a true phar- effects of these drugs on cardiovascular risk factors and macologic difference among drugs, or a combination their long-term cardiovascular safety (Supplement 21,

of these factors. In contrast to HbA1c, fasting plasma available at www.annals.org). glucose, and body weight, comparison of once-weekly Seventh, 73% of RCTs used the last-observation- exenatide and dulaglutide, 1.5 mg, showed a signifi- carried-forward imputation method for handling miss- cant difference in the risk for nausea (1.8-fold higher for ing data. This can result in biased estimations, because dulaglutide). it assumes the last measure as the final measure re- Our study has limitations. First, we performed a gardless of the “true” direction (increase or decrease of study-level meta-analysis based only on available arti- the outcome value) (70). cles, abstracts, and Web documents. These are more Eighth, in 3 studies with zero events in one group, likely than unpublished reports to report positive find- we added the standard 0.5 continuity correction. Al- ings. This risk should be low for RCTs and, for the pri- though the influence of this correction on summary es- mary outcome and nausea (the most important side ef- timates has been investigated in the context of pairwise fect), data were available from all trials. meta-analysis (71), little is known about the degree to

Second, the magnitude of HbA1c reduction could which zero-event groups and the 0.5 correction affect depend on the baseline HbA1c value, because higher network meta-analysis estimates. Results should there- reductions are typically associated with higher baseline fore be interpreted with caution, and further research is

HbA1c values (68). Most of the studies, however, re- needed in this area. ported baseline-adjusted HbA1c differences, and each Finally, RCTs are not independent because they once-weekly GLP-1RA has been evaluated in a wide “cluster” within the same sponsoring company; that is,

range of patients and HbA1c values. RCTs sponsored or conducted by the same pharma- Third, we selected the longest study duration for ceutical company are more similar than are RCTs from the main analysis to better reflect “real-world” condi- 2 different companies in terms of follow-up duration, tions, where these drugs are used for month to years. definition and assessment of outcomes, design of hy- We performed sensitivity analyses to assess the effect perglycemia rescue, data analysis and reporting, and of study duration, although the limited number of RCTs sharing of results. Although it is difficult to avoid this for albiglutide made it difficult to derive definitive con- limitation, it should be considered while interpreting

clusions. The effects of albiglutide on HbA1c, fasting combined data from RCTs. plasma glucose, body weight, and nausea were consis- To our knowledge, our study is the first attempt to tent and, along with available direct evidence, would summarize available data on once-weekly GLP-1RAs for suggest a lower efficacy of this drug. This is in line with a wide range of outcomes. Given the small number of the pharmacologic properties of albiglutide, whose safety end points in some studies, we also added de- large molecular weight (about 73 kDa) reduces cross- tailed data on study, drug, and outcome-specific num- ing of the blood–brain barrier and speculatively modu- ber of participants to help the reader interpret the lates central nervous system effects (37). results.

110 Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 www.annals.org Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Comparative Clinical Profiles of Once-Weekly GLP-1RAs REVIEW

In conclusion, available data suggest differences in References cardiometabolic outcomes and safety among once- 1. Inzucchi SE. Diagnosis of diabetes [Letter]. N Engl J Med. 2013; weekly GLP-1RAs. Further RCTs with direct compari- 368:193. [PMID: 23301749] sons of once-weekly GLP-1RAs can help better clarify 2. Seshasai SR, Kaptoge S, Thompson A, Di Angelantonio E, Gao P, their comparative tolerability and efficacy and inform Sarwar N, et al; Emerging Risk Factors Collaboration. Diabetes mel- the choice among these newly available glucose- litus, fasting glucose, and risk of cause-specific death. N Engl J Med. 2011;364:829-41. [PMID: 21366474] lowering agents. 3. Sarwar N, Gao P, Seshasai SR, Gobin R, Kaptoge S, Di Angelan- tonio E, et al; Emerging Risk Factors Collaboration. Diabetes melli- From the Diabetes Research Centre, University of Leicester, tus, fasting blood glucose concentration, and risk of vascular dis- Leicester, United Kingdom. ease: a collaborative meta-analysis of 102 prospective studies. Lancet. 2010;375:2215-22. [PMID: 20609967] 4. Majumdar SK, Inzucchi SE. Investigational anti-hyperglycemic Disclaimer: The views expressed are those of the authors and agents: the future of type 2 diabetes therapy? Endocrine. 2013;44: do not necessarily reflect those of the United Kingdom's Na- 47-58. [PMID: 23354728] tional Health Service, National Institute for Health Research, or 5. Holst JJ, Knop FK, Vilsbøll T, Krarup T, Madsbad S. Loss of Department of Health. effect is a specific, important, and early characteristic of type 2 dia- betes. Diabetes Care. 2011;34 Suppl 2:S251-7. [PMID: 21525464] 6. Drucker DJ, Sherman SI, Gorelick FS, Bergenstal RM, Sherwin RS, Acknowledgment: The authors thank the National Institute for Buse JB. Incretin-based therapies for the treatment of type 2 diabe- Health Research, Collaboration for Leadership in Applied tes: evaluation of the risks and benefits. Diabetes Care. 2010;33:428- Health Research and Care—East Midlands and Diet, Lifestyle & 33. [PMID: 20103558] 7. Inzucchi SE, Bergenstal RM, Buse JB, Diamant M, Ferrannini E, Physical Activity Biomedical Research Unit, based at the Uni- Nauck M, et al. Management of hyperglycaemia in type 2 diabetes, versity Hospitals of Leicester and Loughborough University, 2015: a patient-centred approach. Update to a position statement of United Kingdom, for their support of this work. They also the American Diabetes Association and the European Association thank Sarah Sutton (University of Leicester, United Kingdom) for the Study of Diabetes. Diabetologia. 2015;58:429-42. [PMID: for her assistance with the literature search and Dimitris Ma- 25583541] vridis (University of Ioannina, Greece) for his valuable 8. Buse JB, Rosenstock J, Sesti G, Schmidt WE, Montanya E, Brett suggestions. JH, et al; LEAD-6 Study Group. Liraglutide once a day versus ex- enatide twice a day for type 2 diabetes: a 26-week randomised, parallel-group, multinational, open-label trial (LEAD-6). Lancet. 2009; Grant Support: From Sanofi Aventis to the University of 374:39-47. [PMID: 19515413] Leicester. 9. Sutton A, Ades AE, Cooper N, Abrams K. Use of indirect and mixed treatment comparisons for technology assessment. Phar- macoeconomics. 2008;26:753-67. [PMID: 18767896] Disclosures: Dr. Khunti reports having acted as a consultant, 10. Jansen JP, Trikalinos T, Cappelleri JC, Daw J, Andes S, Eldes- speaker, and advisory board member for AstraZeneca, No- souki R, et al. Indirect treatment comparison/network meta-analysis vartis, Novo Nordisk, Sanofi-Aventis, Lilly, Merck Sharp & Do- study questionnaire to assess relevance and credibility to inform hme, Janssen Pharmaceuticals, and Boehringer Ingelheim; health care decision making: an ISPOR-AMCP-NPC Good Practice grants from AstraZeneca, Novartis, Novo Nordisk, Sanofi- Task Force report. Value Health. 2014;17:157-73. [PMID: 24636374] 11. Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Pre- Aventis, Lilly, Boehringer Ingelheim, Merck Sharp & Dohme, ferred reporting items for systematic reviews and meta-analyses: the and Roche. Dr. Davies reports having acted as a consultant, an PRISMA statement. J Clin Epidemiol. 2009;62:1006-12. [PMID: advisory board member, and speaker for Novo Nordisk, 19631508] Sanofi-Aventis, Lilly, Merck Sharp & Dohme, Boehringer Ingel- 12. Hoaglin DC, Hawkins N, Jansen JP, Scott DA, Itzler R, Cappelleri heim, AstraZeneca, and Janssen Pharmaceuticals and as a JC, et al. Conducting indirect-treatment-comparison and network- speaker for the Mitsubishi Tanabe Pharma Corporation, as meta-analysis studies: report of the ISPOR Task Force on Indirect well as grants in support of investigator and investigator- Treatment Comparisons Good Research Practices: part 2. Value initiated trials from Novo Nordisk, Sanofi-Aventis, and Lilly, Health. 2011;14:429-37. [PMID: 21669367] outside the submitted work. Authors not named here have 13. Hutton B, Salanti G, Caldwell DM, Chaimani A, Schmid CH, Cam- eron C, et al. The PRISMA extension statement for reporting of sys- disclosed no conflicts of interest. Forms can be viewed at tematic reviews incorporating network meta-analyses of health care www.acponline.org/authors/icmje/ConflictOfInterestForms interventions: checklist and explanations. Ann Intern Med. 2015;162: .do?msNum=M15-1432. 777-84. [PMID: 26030634] 14. Higgins JP, Altman DG, Gøtzsche PC, Ju¨ ni P, Moher D, Oxman AD, et al; Cochrane Bias Methods Group. The Cochrane Collabora- Reproducible Research Statement: Study protocol: Available tion's tool for assessing risk of bias in randomised trials. BMJ. 2011; as Supplement 1 at www.annals.org. Statistical code and data 343:d5928. [PMID: 22008217] set: Available from Dr. Zaccardi (e-mail, [email protected]). 15. Harbord RM, Higgins JP. Meta-regression in Stata. Stata J. 2008; 8:493-519. 16. Medical Research Council Biostatistics Unit. Ian White's Stata Requests for Single Reprints: Francesco Zaccardi, MD, Diabe- software list. 2015. Accessed at www.mrc-bsu.cam.ac.uk/software tes Research Centre, Leicester General Hospital, Gwendolen /stata-software on 1 October 2015. Road, Leicester LE5 4PW, United Kingdom; e-mail, [email protected] 17. White IR. Multivariate random-effects meta-regression: updates .uk. to mvmeta. Stata J. 2011;11:255-70. 18. White IR, Barrett JK, Jackson D, Higgins JP. Consistency and inconsistency in network meta-analysis: model estimation using mul- Current author addresses and author contributions are avail- tivariate meta-regression. Res Synth Methods. 2012;3:111-25. [PMID: able at www.annals.org. 26062085]

www.annals.org Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 111 Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 REVIEW Comparative Clinical Profiles of Once-Weekly GLP-1RAs

19. Multiple-Treatments Meta-Analysis. 2015. Using STATA for net- domised, open-label, phase 3, non-inferiority trial. Lancet. 2014;384: work meta-analysis. Accessed at www.mtm.uoi.gr/index.php/stata 1349-57. [PMID: 25018121] -routines-for-network-meta-analysis on 1 October 2015. 35. Henry RR, Mudaliar S, Kanitra L, Woloschak M, Balena R; 20. Nauck M, Stewart M, Perkins C, Jones-Leone A, Yang F, Perry C, T-emerge 3 Study Group. Efficacy and safety of taspoglutide in pa- et al. HARMONY 2 wk 52 results: albiglutide monotherapy in drug tients with type 2 diabetes inadequately controlled with metformin naı¨ve patients with type 2 diabetes mellitus [Abstract]. Presented at plus pioglitazone over 24 weeks: T-emerge 3 trial. J Clin Endocrinol American Diabetes Association 73rd Scientific Session, Chicago, Metab. 2012;97:2370-9. [PMID: 22539590] Illinois, 21–25 June 2013. Abstract no. 55-LB. 36. Hollander P, Lasko B, Barnett AH, Bengus M, Kanitra L, Pi-Sunyer 21. A monotherapy study to evaluate the efficacy and safety of 2 FX, et al. Effects of taspoglutide on glycemic control and body dose levels of albiglutide in Japanese subjects with type 2 diabetes weight in obese patients with type 2 diabetes (T-emerge 7 study). mellitus (T2DM). ClinicalTrials.gov: NCT01733758. Accessed at Obesity (Silver Spring). 2013;21:238-47. [PMID: 23404788] https://clinicaltrials.gov/ct2/show/NCT01733758 on 26 September 37. Home PD, Shamanna P, Stewart M, Yang F, Miller M, Perry C, 2015. et al. Efficacy and tolerability of albiglutide versus placebo or piogli- 22. Chaimani A, Higgins JP, Mavridis D, Spyridonos P, Salanti G. tazone over 1 year in people with type 2 diabetes currently taking Graphical tools for network meta-analysis in STATA. PLoS One. 2013; metformin and glimepiride: HARMONY 5. Diabetes Obes Metab. 8:e76654. [PMID: 24098547] 2015;17:179-87. [PMID: 25406730] 23. Salanti G, Ades AE, Ioannidis JP. Graphical methods and numer- 38. Inagaki N, Atsumi Y, Oura T, Saito H, Imaoka T. Efficacy and ical summaries for presenting results from multiple-treatment meta- safety profile of exenatide once weekly compared with insulin once analysis: an overview and tutorial. J Clin Epidemiol. 2011;64:163-71. daily in Japanese patients with type 2 diabetes treated with oral an- [PMID: 20688472] tidiabetes drug(s): results from a 26-week, randomized, open-label, 24. Jackson D, Barrett JK, Rice S, White IR, Higgins JP. A design-by- parallel-group, multicenter, noninferiority study. Clin Ther. 2012;34: treatment interaction model for network meta-analysis with random 1892-908. [PMID: 22884767] inconsistency effects. Stat Med. 2014;33:3639-54. [PMID: 24777711] 39. Ji L, Onishi Y, Ahn CW, Agarwal P, Chou CW, Haber H, et al. 25. Ahre´ n B, Johnson SL, Stewart M, Cirkel DT, Yang F, Perry C, et al; Efficacy and safety of exenatide once-weekly vs. exenatide twice- HARMONY 3 Study Group. HARMONY 3: 104-week randomized, daily in Asian patients with type 2 diabetes mellitus. J Diabetes In- double-blind, placebo- and active-controlled trial assessing the effi- vestig. 2013;4:53-61. [PMID: 24843631] cacy and safety of albiglutide compared with placebo, sitagliptin, 40. Nauck M, Horton E, Andjelkovic M, Ampudia-Blasco FJ, Parusel and glimepiride in patients with type 2 diabetes taking metformin. CT, Boldrin M, et al; T-emerge 5 Study Group. Taspoglutide, a once- Diabetes Care. 2014;37:2141-8. [PMID: 24898304] weekly glucagon-like peptide 1 analogue, vs. insulin glargine titrated 26. Bergenstal RM, Forti A, Chiasson JL, Woloschak M, Boldrin M, to target in patients with type 2 diabetes: an open-label randomized Balena R. Efficacy and safety of taspoglutide versus sitagliptin for trial. Diabet Med. 2013;30:109-13. [PMID: 22937895] type 2 diabetes mellitus (T-emerge 4 trial). Diabetes Ther. 2012;3:13. 41. Weinstock RS, Guerci B, Umpierrez G, Nauck MA, Skrivanek Z, [PMID: 23138449] Milicevic Z. Safety and efficacy of once-weekly dulaglutide versus 27. Bergenstal RM, Wysham C, Macconell L, Malloy J, Walsh B, Yan sitagliptin after 2 years in metformin-treated patients with type 2 di- P, et al; DURATION-2 Study Group. Efficacy and safety of exenatide abetes (AWARD-5): a randomized, phase III study. Diabetes Obes once weekly versus sitagliptin or pioglitazone as an adjunct to met- Metab. 2015;17:849-58. [PMID: 25912221] formin for treatment of type 2 diabetes (DURATION-2): a ran- 42. Pratley RE, Nauck MA, Barnett AH, Feinglos MN, Ovalle F, domised trial. Lancet. 2010;376:431-9. [PMID: 20580422] Harman-Boehm I, et al; HARMONY 7 study group. Once-weekly al- 28. Blevins T, Pullman J, Malloy J, Yan P, Taylor K, Schulteis C, et al. biglutide versus once-daily liraglutide in patients with type 2 diabe- DURATION-5: exenatide once weekly resulted in greater improve- tes inadequately controlled on oral drugs (HARMONY 7): a ran- ments in glycemic control compared with exenatide twice daily in domised, open-label, multicentre, non-inferiority phase 3 study. patients with type 2 diabetes. J Clin Endocrinol Metab. 2011;96: Lancet Diabetes Endocrinol. 2014;2:289-97. [PMID: 24703047] 1301-10. [PMID: 21307137] 43. Pratley RE, Urosevic D, Boldrin M, Balena R; T-emerge 6 Study 29. Buse JB, Nauck M, Forst T, Sheu WH, Shenouda SK, Heilmann Group. Efficacy and tolerability of taspoglutide versus pioglitazone in CR, et al. Exenatide once weekly versus liraglutide once daily in pa- subjects with type 2 diabetes uncontrolled with sulphonylurea or tients with type 2 diabetes (DURATION-6): a randomised, open-label sulphonylurea-metformin therapy: a randomized, double-blind study. Lancet. 2013;381:117-24. [PMID: 23141817] study (T-emerge 6). Diabetes Obes Metab. 2013;15:234-40. [PMID: 30. Davies M, Heller S, Sreenan S, Sapin H, Adetunji O, Tahbaz A, 22958426] et al. Once-weekly exenatide versus once- or twice-daily insulin det- 44. Raz I, Fonseca V, Kipnes M, Durrwell L, Hoekstra J, Boldrin M, emir: randomized, open-label, clinical trial of efficacy and safety in et al. Efficacy and safety of taspoglutide monotherapy in drug-naive patients with type 2 diabetes treated with metformin alone or in type 2 diabetic patients after 24 weeks of treatment: results of a combination with . Diabetes Care. 2013;36:1368-76. randomized, double-blind, placebo-controlled phase 3 study [PMID: 23275363] (T-emerge 1). Diabetes Care. 2012;35:485-7. [PMID: 22301126] 31. Diamant M, Van Gaal L, Guerci B, Stranks S, Han J, Malloy J, et al. 45. Reusch J, Stewart MW, Perkins CM, Cirkel DT, Ye J, Perry CR, Exenatide once weekly versus insulin glargine for type 2 diabetes et al. Efficacy and safety of once-weekly glucagon-like peptide 1 re- (DURATION-3): 3-year results of an open-label randomised trial. Lan- ceptor agonist albiglutide (HARMONY 1 trial): 52-week primary end- cet Diabetes Endocrinol. 2014;2:464-73. [PMID: 24731672] point results from a randomized, double-blind, placebo-controlled 32. Diamant M, Van Gaal L, Stranks S, Northrup J, Cao D, Taylor K, trial in patients with type 2 diabetes mellitus not controlled on piogli- et al. Once weekly exenatide compared with insulin glargine titrated tazone, with or without metformin. Diabetes Obes Metab. 2014;16: to target in patients with type 2 diabetes (DURATION-3): an open- 1257-64. [PMID: 25155146] label randomised trial. Lancet. 2010;375:2234-43. [PMID: 20609969] 46. Rosenstock J, Balas B, Charbonnel B, Bolli GB, Boldrin M, Ratner 33. Drucker DJ, Buse JB, Taylor K, Kendall DM, Trautmann M, R, et al; T-emerge 2 Study Group. The fate of taspoglutide, a weekly Zhuang D, et al; DURATION-1 Study Group. Exenatide once weekly GLP-1 receptor agonist, versus twice-daily exenatide for type 2 dia- versus twice daily for the treatment of type 2 diabetes: a randomised, betes: the T-emerge 2 trial. Diabetes Care. 2013;36:498-504. [PMID: open-label, non-inferiority study. Lancet. 2008;372:1240-50. [PMID: 23139373] 18782641] 47. Rosenstock J, Fonseca VA, Gross JL, Ratner RE, Ahre´ n B, Chow 34. Dungan KM, Povedano ST, Forst T, Gonza´ lez JG, Atisso C, Sealls FC, et al; HARMONY 6 Study Group. Advancing basal insulin re- W, et al. Once-weekly dulaglutide versus once-daily liraglutide in placement in type 2 diabetes inadequately controlled with insulin metformin-treated patients with type 2 diabetes (AWARD-6): a ran- glargine plus oral agents: a comparison of adding albiglutide, a

112 Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 www.annals.org Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Comparative Clinical Profiles of Once-Weekly GLP-1RAs REVIEW

weekly GLP-1 receptor agonist, versus thrice-daily prandial insulin 2 diabetes: a systematic review and network meta-analysis. J Diabe- lispro. Diabetes Care. 2014;37:2317-25. [PMID: 24898300] tes Res. 2015;2015:157201. [PMID: 25688373] 48. Russell-Jones D, Cuddihy RM, Hanefeld M, Kumar A, Gonza´ lez 59. Sun F, Chai S, Yu K, Quan X, Yang Z, Wu S, et al. Gastrointestinal JG, Chan M, et al; DURATION-4 Study Group. Efficacy and safety of adverse events of glucagon-like peptide-1 receptor agonists in pa- exenatide once weekly versus metformin, pioglitazone, and sitaglip- tients with type 2 diabetes: a systematic review and network meta- tin used as monotherapy in drug-naive patients with type 2 diabetes analysis. Diabetes Technol Ther. 2015;17:35-42. [PMID: 25375397] (DURATION-4): a 26-week double-blind study. Diabetes Care. 2012; 60. Sun F, Wu S, Guo S, Yu K, Yang Z, Li L, et al. Effect of GLP-1 35:252-8. [PMID: 22210563] receptor agonists on waist circumference among type 2 diabetes 49. Umpierrez G, Tofe´ Povedano S, Pe´ rez Manghi F, Shurzinske L, patients: a systematic review and network meta-analysis. Endocrine. Pechtner V. Efficacy and safety of dulaglutide monotherapy versus 2015;48:794-803. [PMID: 25115635] metformin in type 2 diabetes in a randomized controlled trial 61. Sun F, Wu S, Wang J, Guo S, Chai S, Yang Z, et al. Effect of (AWARD-3). Diabetes Care. 2014;37:2168-76. [PMID: 24842985] glucagon-like peptide-1 receptor agonists on lipid profiles among 50. Weissman PN, Carr MC, Ye J, Cirkel DT, Stewart M, Perry C, et al. type 2 diabetes: a systematic review and network meta-analysis. Clin HARMONY 4: randomised clinical trial comparing once-weekly albi- Ther. 2015;37:225-241. [PMID: 25554560] glutide and insulin glargine in patients with type 2 diabetes inade- 62. Sun F, Yu K, Wu S, Zhang Y, Yang Z, Shi L, et al. Cardiovascular quately controlled with metformin with or without sulfonylurea. Dia- safety and glycemic control of glucagon-like peptide-1 receptor ago- betologia. 2014;57:2475-84. [PMID: 25208756] nists for type 2 diabetes mellitus: a pairwise and network meta- 51. Wysham C, Blevins T, Arakaki R, Colon G, Garcia P, Atisso C, analysis. Diabetes Res Clin Pract. 2012;98:386-95. [PMID: 23020934] et al. Efficacy and safety of dulaglutide added onto pioglitazone and 63. Sun F, Yu K, Yang Z, Wu S, Zhang Y, Shi L, et al. Impact of GLP-1 metformin versus exenatide in type 2 diabetes in a randomized con- receptor agonists on major gastrointestinal disorders for type 2 dia- trolled trial (AWARD-1). Diabetes Care. 2014;37:2159-67. [PMID: betes mellitus: a mixed treatment comparison meta-analysis. Exp Di- 24879836] abetes Res. 2012;2012:230624. [PMID: 23365557] 52. Giorgino F, Benroubi M, Sun JH, Zimmermann AG, Pechtner V. 64. Sun F, Wu S, Guo S, Yu K, Yang Z, Li L, et al. Impact of GLP-1 Efficacy and safety of once-weekly dulaglutide versus insulin glargine receptor agonists on blood pressure, heart rate and hypertension in patients with type 2 diabetes on metformin and glimepiride among patients with type 2 diabetes: a systematic review and net- (AWARD-2). Diabetes Care. 2015. [PMID: 26089386] work meta-analysis. Diabetes Res Clin Pract. 2015;110:26-37. [PMID: 53. Blonde L, Jendle J, Gross J, Woo V, Jiang H, Fahrbach JL, et al. 26358202] Once-weekly dulaglutide versus bedtime insulin glargine, both in 65. Zhong X, Zhang T, Liu Y, Wei X, Zhang X, Qin Y, et al. Effects of combination with prandial insulin lispro, in patients with type 2 three injectable antidiabetic agents on glycaemic control, weight diabetes (AWARD-4): a randomised, open-label, phase 3, non- change and drop-out in type 2 diabetes suboptimally controlled with inferiority study. Lancet. 2015;385:2057-66. [PMID: 26009229] metformin and/or a sulfonylurea: a network meta-analysis. Diabetes 54. Miyagawa J, Odawara M, Takamura T, Iwamoto N, Takita Y, Res Clin Pract. 2015;109:451-60. [PMID: 26233934] Imaoka T. Once-weekly glucagon-like peptide-1 receptor agonist 66. Zintzaras E, Miligkos M, Ziakas P, Balk EM, Mademtzoglou D, dulaglutide is non-inferior to once-daily liraglutide and superior to placebo in Japanese patients with type 2 diabetes: a 26-week ran- Doxani C, et al. Assessment of the relative effectiveness and tolera- domized phase III study. Diabetes Obes Metab. 2015;17:974-83. bility of treatments of type 2 diabetes mellitus: a network meta- [PMID: 26179187] analysis. Clin Ther. 2014;36:1443-53. [PMID: 25109773] 55. Araki E, Inagaki N, Tanizawa Y, Oura T, Takeuchi M, Imaoka T. 67. Scott DA, Boye KS, Timlin L, Clark JF, Best JH. A network meta- Efficacy and safety of once-weekly dulaglutide in combination with analysis to compare glycaemic control in patients with type 2 diabe- sulphonylurea and/or biguanide compared with once-daily insulin tes treated with exenatide once weekly or liraglutide once daily in glargine in Japanese patients with type 2 diabetes: a randomized, comparison with insulin glargine, exenatide twice daily or placebo. open-label, phase III, non-inferiority study. Diabetes Obes Metab. Diabetes Obes Metab. 2013;15:213-23. [PMID: 22958381] 2015;17:994-1002. [PMID: 26179754] 68. DeFronzo RA, Stonehouse AH, Han J, Wintle ME. Relationship of 56. A study comparing the effects and safety of dulaglutide with baseline HbA1c and efficacy of current glucose-lowering therapies: a insulin glargine in type 2 diabetes mellitus. ClinicalTrials.gov: meta-analysis of randomized clinical trials. Diabet Med. 2010;27: NCT01648582. Accessed at https://clinicaltrials.gov/ct2/show 309-17. [PMID: 20536494] /NCT01648582 on 26 September 2015. 69. Kapitza C, Heise T, Birman P, Jallet K, Ramis J, Balena R. Phar- 57. Wang W, Huang CN, Young MC, Li P, Gu L, Yang J. The efficacy macokinetic and pharmacodynamic properties of taspoglutide, a and safety of once-weekly, subcutaneous dulaglutide monotherapy once-weekly, human GLP-1 analogue, after single-dose administra- compared to glimepiride in Asian patients with type 2 diabetes mel- tion in patients with type 2 diabetes. Diabet Med. 2009;26:1156-64. litus. Presented at 51st EASD Meeting, Stockholm, Sweden, 15–18 [PMID: 19929995] September 2015. Abstract no. 778. Accessed at www.easdvirtual 70. Stack CB, Localio AR, Griswold ME, Goodman SN, Mulrow CD. meeting.org/resources/the-efficacy-and-safety-of-once-weekly-sub Handling of rescue and missing data affects synthesis and interpre- cutaneous-dulaglutide-monotherapy-compared-to-glimepiride-in tation of evidence: the sodium-glucose cotransporter 2 inhibitor ex- -asian-patients-with-type-2-diabetes-mellitus–2 on 26 September ample. Ann Intern Med. 2013;159:285-8. [PMID: 24026261] 2015. 71. Sweeting MJ, Sutton AJ, Lambert PC. What to add to nothing? 58. Sun F, Chai S, Li L, Yu K, Yang Z, Wu S, et al. Effects of glucagon- Use and avoidance of continuity corrections in meta-analysis of like peptide-1 receptor agonists on weight loss in patients with type sparse data. Stat Med. 2004;23:1351-75. [PMID: 15116347]

www.annals.org Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 113 Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Annals of Internal Medicine

Current Author Addresses: Drs. Zaccardi, Htike, Webb, Khunti, Author Contributions: Conception and design: F. Zaccardi, and Davies: Diabetes Research Centre, Leicester General Z.Z. Htike, D.R. Webb, M.J. Davies. Hospital, Gwendolen Road, Leicester LE5 4PW, United Analysis and interpretation of the data: F. Zaccardi, D.R. Kingdom. Webb. Drafting of the article: F. Zaccardi, Z.Z. Htike, D.R. Webb, M.J. Davies. Critical revision of the article for important intellectual con- tent: F. Zaccardi, Z.Z. Htike, D.R. Webb, M.J. Davies. Final approval of the article: F. Zaccardi, Z.Z. Htike, D.R. Webb, K. Khunti, M.J. Davies. Statistical expertise: F. Zaccardi. Obtaining of funding: D.R. Webb. Collection and assembly of data: F. Zaccardi, Z.Z. Htike, K. Khunti.

Appendix Figure 1. Summary of evidence search and selection.

1065 records identified through database searching on 26 September 2015 50 additional records identified through ClinicalTrials.gov, PubMed: 244 fda.gov, and ema.europa.eu/ema ISI Web of Science: 235 CENTRAL: 381

Identification Abstracts: 205

1056 records removed after exclusion of duplicates and title/abstract selection

Articles and abstracts screened (n = 59)

Screening Articles and abstracts excluded (n = 24) Specific patient groups (i.e., renal impairment): 3 Pharmacokinetics studies: 4 Irrelevant drugs or outcomes: 16 Follow-up <24 weeks: 1

31 articles, 2 studies in ClinicalTrials.gov, and 2 abstracts were included in quantitative synthesis (meta-analysis)

Unique RCTs in quantitative synthesis (meta-analysis) (n = 34)

Inclusion Articles: 31 Studies in ClinicalTrials.gov: 2 Abstracts: 2

For PubMed, the search was “Exenatide” OR “Taspoglutide” OR “Albi- glutide” OR “Dulaglutide” OR “Semaglutide”; limits: Humans and Ran- domized Controlled Trial. The search strategy was specifically trans- lated for other databases. RCT = randomized, controlled trial.

www.annals.org Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Appendix Figure 2. Pairwise random-effects meta-analysis for the primary outcome (HbA1c).

2 Group 1 Group 2 Comparison, n Participants, n HbA1c Reduction, % I Value, % Reference (95% CI)

Dulaglutide, 0.75 mg Placebo 1 348 −1.57 (−1.79 to −1.35) – 54 Albiglutide Placebo 5 1506 −1.03 (−1.31 to −0.75) 89.7 20, 21, 25, 37, 45 Taspoglutide, 20 mg Placebo 3 751 −0.92 (−1.14 to −0.69) 77.5 35, 36, 44 Taspoglutide, 10 mg Placebo 2 439 −0.91 (−1.07 to −0.76) 0 35, 44 Dulaglutide, 1.5 mg Sitagliptin 1 619 −0.67 (−0.84 to −0.50) – 41 Dulaglutide, 1.5 mg Glimepiride 1 531 −0.60 (−0.79 to −0.41) – 57 Taspoglutide, 20 mg Sitagliptin 1 364 −0.52 (−0.69 to −0.35) – 26 Once-weekly exenatide Sitagliptin 2 737 −0.48 (−0.69 to −0.27) 43.1 27, 48 Dulaglutide, 0.75 mg Sitagliptin 1 617 −0.39 (−0.57 to −0.21) – 41 Taspoglutide, 10 mg Sitagliptin 1 359 −0.37 (−0.54 to −0.20) – 26 Dulaglutide, 1.5 mg Basal insulins 3 1638 −0.36 (−0.51 to −0.21) 35.9 52, 53, 56 Once-weekly exenatide Basal insulins 3 1099 −0.35 (−0.50 to −0.20) 43.8 30, 31, 38 Albiglutide Sitagliptin 1 604 −0.35 (−0.53 to −0.17) – 25 Dulaglutide, 0.75 mg Glimepiride 1 527 −0.31 (−0.50 to −0.12) – 57 Dulaglutide, 1.5 mg Daily GLP-1RAs 2 1154 −0.30 (−0.79 to 0.19) 92.9 34, 51 Taspoglutide, 20 mg Pioglitazone 1 500 −0.30 (−0.52 to −0.08) – 43 Once-weekly exenatide Daily GLP-1RAs 4 2136 −0.29 (−0.66 to 0.09) 93.3 28, 29, 33, 39 Albiglutide Glimepiride 1 609 −0.27 (−0.45 to −0.09) – 25 Dulaglutide, 1.5 mg Dulaglutide 0.75 7 3875 −0.25 (−0.33 to −0.17) 4.4 41, 49, 51–53, 56, 57 Taspoglutide, 20 mg Daily GLP-1RAs 1 765 −0.24 (−0.49 to 0.01) – 46 Dulaglutide, 0.75 mg Basal insulins 4 1996 −0.23 (−0.46 to 0.01) 86.9 52, 53, 55, 56 Taspoglutide, 10 mg Daily GLP-1RAs 1 757 −0.22 (−0.47 to 0.03) – 46 Dulaglutide, 1.5 mg Metformin 1 537 −0.19 (−0.38 to 0.00) – 49 Taspoglutide, 20 mg Taspoglutide 10 6 2803 −0.18 (−0.28 to −0.08) 32.7 26, 35, 40, 43, 44, 46 Dulaglutide, 0.75 mg Daily GLP-1RAs 2 972 −0.17 (−0.33 to −0.00) 30.1 51, 54 Albiglutide Rapid insulin 1 563 −0.16 (−0.33 to 0.01) – 47 Taspoglutide, 20 mg Basal insulins 1 667 −0.14 (−0.28 to −0.00) – 40 Once-weekly exenatide Pioglitazone 2 736 −0.09 (−0.49 to 0.30) 82.8 27, 48 Once-weekly exenatide Metformin 1 494 −0.05 (−0.24 to 0.14) – 48 Dulaglutide, 0.75 mg Metformin 1 538 −0.04 (−0.23 to 0.15) – 49 Albiglutide Daily GLP-1RAs 2 1065 0.03 (−0.29 to 0.34) 83.6 21, 42 Taspoglutide, 10 mg Basal insulins 1 680 0.07 (−0.07 to 0.21) – 40 Albiglutide Basal insulins 1 745 0.12 (−0.10 to 0.34) – 50 Taspoglutide, 10 mg Pioglitazone 1 495 0.12 (−0.10 to 0.34) – 46 Albiglutide Pioglitazone 1 533 0.25 (0.08 to 0.42) – 37

Favors Group 1 Favors Group 2

−2.0 −1.5 −1.0 −0.5 0 0.5

HbA1c Reduction, %

Mean differences are reported for comparisons with at least 1 once-weekly GLP-1RA group. GLP-1RA = glucagon-like peptide-1 receptor agonist;

HbA1c = hemoglobin A1c.

Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 www.annals.org Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Appendix Figure 3. Differences versus placebo ( dashed line) for cardiometabolic outcomes.

29 3 Level, Protein C-Reactive 4 4 5.0

nmol/L 19 2 mm Hg mm mm Hg mm 2 2.5 2 9 1 0 beats/min 0 0 0 0 −2 −2.5 −9 −1 mg/L Heart Rate, −4 −2 −5.0 −19 −2 C-Reactive Protein Level, Systolic Blood Pressure, Diastolic Blood Pressure, PIO PIO PIO PIO ALB ALB ALB SITA SITA SITA MET MET MET EOW EOW EOW EOW GLIM GLIM GLIM dGLP1 dGLP1 dGLP1 dGLP1 BASAL BASAL BASAL BASAL TAS 10 TAS 20 TAS TAS 20 TAS TAS 20 TAS 10 TAS 10 TAS DUL 1.5 DUL 1.5 DUL 1.5 DUL DUL 0.75 DUL DUL 0.75 DUL DUL 0.75 DUL Total Cholesterol Level, HDL Level, Cholesterol 0.4 15 0.4 15 LDL Cholesterol Level, 0.2 8 0.50 44 Triglyceride Level, mmol/L mmol/L 0.2 8 mmol/L 0.2 8 22

mmol/L 0.25 0.1 4

0 0 0 0 0 0

0 0 −0.2 −8 −0.2 −8 −22 mg/dL

mg/dL −0.25 mg/dL mg/dL

−0.4 −15 −0.4 −15 −0.1 −4 Level, Triglycerides −0.50 −44 LDL Cholesterol Level, Total Cholesterol Level, HDL Cholesterol Level, HDL PIO PIO PIO PIO ALB ALB ALB ALB SITA SITA SITA SITA EOW EOW EOW EOW GLIM GLIM GLIM GLIM dGLP1 dGLP1 dGLP1 dGLP1 BASAL BASAL BASAL BASAL TAS 20 TAS TAS 10 TAS TAS 10 TAS 20 TAS TAS 10 TAS 20 TAS 20 TAS TAS 10 TAS DUL 1.5 DUL DUL 1.5 DUL DUL 1.5 DUL DUL 1.5 DUL DUL 0.75 DUL DUL 0.75 DUL 0.75 DUL DUL 0.75 DUL

0 0 0 0 HbA 4 FPG Level, 1c Value, kg % −0.5 −5 −1 −18 2 mmol/moL Value, mg 1c −1.0 −11

−2 −36 /

dL 0 HbA Body Weight, FPG Level, mmol/L −1.5 −16 −3 −54 −2 PIO PIO PIO ALB ALB ALB SITA SITA SITA MET MET MET EOW EOW EOW GLIM GLIM GLIM dGLP1 RAPID dGLP1 dGLP1 RAPID RAPID BASAL BASAL BASAL TAS 20 TAS TAS 10 TAS TAS 20 TAS 10 TAS TAS 10 TAS TAS 20 TAS DUL 1.5 DUL DUL 1.5 DUL DUL 1.5 DUL DUL 0.75 DUL DUL 0.75 DUL DUL 0.75 DUL

ALB = albiglutide; BASAL = basal insulin; dGLP1 = daily glucagon-like peptide-1 receptor agonists; DUL 0.75 = dulaglutide, 0.75 mg; DUL 1.5 =

dulaglutide, 1.5 mg; EOW = exenatide once-weekly; FPG = fasting plasma glucose; GLIM = glimepiride; HbA1c = hemoglobin A1c; HDL = high-density lipoprotein; LDL = low-density lipoprotein; MET = metformin; PIO = pioglitazone; PLA = placebo; RAPID = rapid insulin; SITA = sitagliptin; TAS 10 = taspoglutide, 10 mg; TAS 20 = taspoglutide, 20 mg.

www.annals.org Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Appendix Figure 4. Rank probabilities for the primary outcome (hemoglobin A1c).

Albiglutide Basal Insulins Dulaglutide, 0.75 mg Dulaglutide, 1.5 mg Once-Weekly Exenatide

100 100 100 100 100

80 80 80 80 80 % % % % %

60 60 60 60 60

40 40 40 40 40 Probability, Probability, Probability, Probability, Probability,

20 20 20 20 20

0 0 0 0 0 1234567891011121314 1 2 3 4 5 6 7 8 9 10 11 12 1314 1 2 3 4 5 6 7 8 9 10 11 12 1314 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Rank Rank Rank Rank Rank

Glimepiride Metformin Pioglitazone Placebo Rapid insulin

100 100 100 100 100

80 80 80 80 80 % % % % %

60 60 60 60 60

40 40 40 40 40 Probability, Probability, Probability, Probability, Probability,

20 20 20 20 20

0 0 0 0 0 1234567891011121314 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1234567891011121314 Rank Rank Rank Rank Rank

Sitagliptin Taspoglutide, 10 mg Taspoglutide, 20 mg Daily GLP-1RAs

100 100 100 100

80 80 80 80 % % % %

60 60 60 60

40 40 40 40 Probability, Probability, Probability, Probability,

20 20 20 20

0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 1314 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Rank Rank Rank Rank

GLP-1RA = glucagon-like peptide-1 receptor agonist.

Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 www.annals.org Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016 Appendix Figure 5. Differences versus placebo (dashed line) for safety outcomes.

Documented or Symptomatic Hypoglycemia

20 10

4 2 1

0.3 Odds Ratio (95% CI)

0.1 PIO ALB SITA MET EOW GLIM dGLP1 RAPID BASAL TAS 20 TAS 10 DUL 1.5 DUL 0.75

Nausea

20 10

4 2 1

0.3 Odds Ratio (95% CI)

0.1 PIO ALB MET SITA EOW GLIM dGLP1 RAPID BASAL TAS 10 TAS 20 DUL 1.5 DUL 0.75

ALB = albiglutide; BASAL = basal insulin; dGLP1 = daily glucagon-like peptide-1 receptor agonists; DUL 0.75 = dulaglutide, 0.75 mg; DUL 1.5 = dulaglutide, 1.5 mg; EOW = exenatide once-weekly; GLIM = glimepiride; MET = metformin; PIO = pioglitazone; PLA = placebo; RAPID = rapid insulin; SITA = sitagliptin; TAS 10 = taspoglutide, 10 mg; TAS 20 = taspoglutide, 20 mg.

www.annals.org Annals of Internal Medicine • Vol. 164 No. 2 • 19 January 2016 Downloaded From: http://annals.org/ by Kevin Rosteing on 03/14/2016