Diabetes Care 1

Michael A. Nauck,1 Thomas Jon Jensen,2 Reported With Carina Rosenkilde,2 Salvatore Calanna,2 and John B. Buse,3 the LEADER Publication or Placebo in People Committee on behalf of the LEADER Trial With Type 2 Diabetes: Results Investigators* From the LEADER Randomized Trial https://doi.org/10.2337/dc17-1825

OBJECTIVE This study explored risk with liraglutide versus placebo in the LEADER MRIGTCNLGE N THERAPEUTICS AND TECHNOLOGIES EMERGING (Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results) cohort.

RESEARCH DESIGN AND METHODS LEADER (NCT01179048) was an international, phase 3b, randomized, double-blind, controlled trial. Participants aged ‡50 years with type 2 diabetes and high car- diovascular risk were assigned 1:1 to receive liraglutide (£1.8 mg daily; n = 4,668) or placebo (n = 4,672) in addition to standard care and monitored for 3.5–5 years (median follow-up 3.8 years). The occurrence of neoplasms was a prespecified, exploratorysecondaryendpoint.Posthocanalysesofthetime tothefirstconfirmed neoplasms were conducted using a Cox regression model.

RESULTS 1Diabetes Center Bochum-Hattingen, St. Josef- Neoplasm was confirmed in 10.1% of patients with liraglutide versus 9.0% with Hospital, Ruhr-University Bochum, Bochum, placebo (hazard ratio [HR] 1.12 [95% CI 0.99; 1.28]). The HR (95% CI) for liraglutide Germany 2 versus placebo was 1.06 (0.90; 1.25) for malignant neoplasms and 1.16 (0.93; 1.44) Novo Nordisk A/S, Bagsværd, Denmark 3University of North Carolina School of Medicine, for benign neoplasms. Sensitivity analyses excluding neoplasms occurring <1 year Chapel Hill, NC < or 2 years after randomization and analyses by sex provided similar results. Corresponding author: Michael A. Nauck, michael In our main analyses, the 95% CI for the HR included one for all malignant neo- [email protected]. plasms evaluated (including pancreatic and neoplasms), except for pro- Received 31 August 2017 and accepted 29 April state neoplasms, which occurred in fewer liraglutide-treated patients. 2018. Clinical trial reg. no. NCT01179048, clinicaltrials CONCLUSIONS .gov. LEADER was not primarily designed to assess neoplasm risk. Firm conclusions cannot This article contains Supplementary Data online be made regarding numeric imbalances observed for individual neoplasm types at http://care.diabetesjournals.org/lookup/suppl/ (e.g., pancreatic ) that occurred infrequently. LEADER data do, however, ex- doi:10.2337/dc17-1825/-/DC1. clude a major increase in the risk of total malignant neoplasms with liraglutide *A complete list of the LEADER Committee Mem- bers and Trial Investigators can be found in the versus placebo. Additional studies are needed to assess longer-term exposure. Supplementary Data online. © 2018 by the American Diabetes Association. Diabetes and have been identified as risk factors for cancer and cancer-related Readers may use this article as long as the work mortality (1–5). Glucose-lowering medications may also affect the risk of these events is properly cited, the use is educational and not for profit, and the work is not altered. More infor- (1,3). Glucagon-like peptide 1 (GLP-1) receptor agonists (GLP-1RAs) improve glycemic mation is available at http://www.diabetesjournals control with a low risk of hypoglycemia in people with type 2 diabetes (6). GLP-1 .org/content/license. Diabetes Care Publish Ahead of Print, published online June 13, 2018 2 Neoplasms in LEADER Diabetes Care

receptor expression has been reported , surgery, radiotherapy, insufficient evidence (e.g., lack of a pa- in a wide range of rodent and human or palliative therapy in the previous thology report) to confirm an event as a tissues, including the pancreatic islets, 5 years was an exclusion criterion. Also neoplasm. lung, heart, kidney, gastrointestinal tract, excluded were patients with a familial or To confirm the robustness of the anal- pituitary, skin, and several regions of the personal history of multiple endocrine yses based on adjudicated data, a review nervous system (7,8). This widespread neoplasia type 2 or medullary thyroid of investigator-reported adverse events expression of the GLP-1 receptor may cancer (MTC) (11). Patients with intra- of malignant neoplasms was also under- help to explain the range of effects of epithelial squamous carcinoma taken by the sponsor, who performed GLP-1RAs (9–15). of the skin (Bowen’sdisease)treated MedDRA term searches after the un- Data from some preclinical (16–19), with topical 5-fluorouracil and patients blinding of data after database lock. In clinical (15), and/or epidemiological da- with basal cell skin cancer were al- addition, case reviews of all investigator- tabase studies (20,21) have suggested lowed to enter the trial (11). Partici- reported adverse events of malignant an increase in the risk of pants were monitored for 3.5–5 years neoplasms not confirmed by the EAC (16,20,21), (17,18,20, (median follow-up, 3.8) (11). The pri- were performed by the sponsor at this 21), intestinal neoplasms (15,19), and mary end point was time to first occur- time. Because the EAC was blinded, breast neoplasms (15) with the use of rence of CV death, nonfatal (including external, and independent, its assess- incretin-based therapy (GLP-1RAs and/or silent) myocardial infarction, or nonfatal ments were prioritized over the spon- dipeptidylpeptidase-4[DPP-4]inhibitors). stroke (11). sor’s assessments; however, the latter In contrast, other clinical and epidemio- are reported for transparency and com- logical database studies have indicated Evaluation of Neoplasms pleteness. that incretin-based therapy does not in- The occurrence of neoplasms was a pre- crease the risk of these events (22–25). specified, exploratory secondary end Evaluation of Cause of Death Results from large, prospective clinical point. The investigators were asked to All deaths were adjudicated by a CV trials may help clarify a potential risk report all types of suspected neoplasms, subcommittee to identify potential CV for neoplasms with GLP-1RAs or other including malignant neoplasm, in situ deaths. The subcommittee classified glucose-lowering medications (26). neoplasm, and neoplasm of uncertain the cause of deaths as “known” or “un- The LEADER (Liraglutide Effect and or unknown behavior. known” and further, for a known cause, Action in Diabetes: Evaluation of Cardio- Potential neoplasms identified via in- as CV or non-CV deaths. This required vascular Outcome Results) trial was de- vestigator reports, Medical Dictionary for formal agreement/adjudication. Based signed to assess the cardiovascular (CV) Regulatory Activities (MedDRA) search, on comments that were not subject to safety of liraglutide but provided the the event adjudication committee (EAC), reconciliation between adjudicators, the opportunity to evaluate other impor- or the contract research organization, sponsor further categorized the non-CV tant safety end points in a population were sent for adjudication (Supplemen- deaths according to plausible cause, which of 9,340 participants monitored for tary Fig. 1). All potential neoplasms were could include neoplastic disease. 3.5–5 years (median follow-up, 3.8) adjudicated in a blinded manner by the (11). The primary analysis showed neoplasms subcommittee of the exter- Statistical Methods liraglutide treatment was associated nal, independent EAC. The clinical eval- Post hoc analyses of the time to the first with a lower rate of major adverse CV uation of neoplasms by the EAC could be EAC-confirmed neoplasm events were events and total mortality (11). The pres- based on diagnostic tests, pathology conducted using a Cox regression model ent report explores intermediate-term reports, specialist consultations, related to compare liraglutide treatment with neoplasm risk with liraglutide versus imaging reports, and/or biomarkers. For placebo. Exploratory statistical testing placebo, based on detailed analyses of thyroid neoplasms, operative reports without correction for multiplicity was data from LEADER. and relevant laboratory findings (e.g., performed if one or more events oc- tumor markers) were also used as di- curred in both treatment groups. The RESEARCH DESIGN AND METHODS agnostic criteria. The EAC used a prespe- main analyses used an intention-to- Trial Design cified definition for adjudication of treat approach, including all first events The design of the LEADER trial (clinicaltrials potential neoplasms, and a pathologic collected from randomization until the .gov, NCT01179048) has been published diagnosis by histology or cytology was end-of-trial follow-up visit. previously (11,27). LEADER was a phase considered of foremost importance for Plots were prepared showing the cu- 3b, randomized, double-blind, controlled confirmation. After database lock, EAC- mulative incidence probability of con- trial, conducted between 2010 and confirmed neoplasm events categorized firmed neoplasm index events with 2015 in 32 countries (11,27). During the as tissue of origin “other” were classified liraglutide or placebo over time. The trial, 9,340 participants aged $50 years according to the organ system affected cumulative incidence was estimated with type 2 diabetes and high CV risk by medically qualified personnel at Novo using the Aalen-Johansen method with were assigned, in a 1:1 ratio, to receive Nordisk. death as a competing risk. Definitions liraglutide (up to 1.8 mg daily) or The EAC was not required to specify its of neoplasm index events, first events, matched placebo as a subcutaneous reason for not confirming a neoplasm and recurrent events are provided in the injection, in addition to their standard- event. Possible reasons for nonconfir- Supplementary Data. of-care treatment (11,27). The pres- mation could include sufficient evidence Sensitivity analyses were conducted ence of malignant neoplasms requiring that an event was not a neoplasm or that excluded neoplasm events occurring care.diabetesjournals.org Nauck and Associates 3

,1yearand,2 years after randomi- (interquartilerange 1.54–1.79),including H)or,2 years after randomization to zation to treatment, in case of an induc- periods during which the participants treatment (Figs. 1C and 2C, F,andI) tion or latency time of these durations. did not receive liraglutide (11). Overall, provided similar results to the main In addition to Cox regression, a Fine and 96.8% of all randomized participants analysis for the overall neoplasm cate- Gray method that adjusted for death completed a final visit, died, or expe- gories evaluated (“all neoplasms,”“all as a competing risk was also used to rienced a component of the primary malignant neoplasms,” and/or “all be- analyze the time to the first neoplasm composite outcome, and 99.7% of the nign neoplasms”). The subdistribution events. participants had known vital status at HRs estimated using the Fine and Gray Separate summaries were prepared the end of the trial (11). method were consistent with the Cox to investigate the occurrence of neo- regression analyses. plasms in sex (male and female) and age Adjudication Flow for Neoplasms Results from the sensitivity analyses of fi (,65 and $65 years) subgroups. A Cox According to the broad study de nition, events captured by the sponsor’sMedDRA regression model was used to conduct 3,802 potential neoplasms were identi- searches for malignant tumors were gen- fi post hoc analyses to assess the inter- ed and sent for adjudication. The EAC erally consistent with the adjudicated fi action between treatment group and con rmed 1,477 (38.8%) of the potential outcome (Supplementary Table 1). Addi- thesesubgroups.AP value of,0.05was neoplasms sent for adjudication as neo- tional data regarding pancreatic cancer taken to indicate a statistically signifi- plasms (Supplementary Fig. 1). are provided in Supplementary Table 2. cant difference. However, because the Frequency of Confirmed Neoplasms fi interaction analyses were exploratory, Frequency of Con rmed Neoplasms by Excluding duplicate events and events Sex and Age Subgroups the results of these analyses should be occurring before randomization or after interpreted with caution. The proportions of patients experiencing the end-of-trial follow-up, 1,123 con- confirmed overall neoplasms, malignant In this report, increased attention is firmed neoplasms occurred. A total of paid to specific types of neoplasms, in- neoplasms, or benign neoplasms with 595 neoplasms (in 470 of 4,668 patients) liraglutide versus placebo were similar with- cluding pancreatic, thyroid, colorectal, in the liraglutide group and 528 neo- breast, prostate, nonmelanoma skin, in male and female subgroups (Table 1). plasms (in 419 of 4,672 patients) in the The frequencies of overall neoplasms melanoma, and lung and bronchus neo- placebo group were confirmed. Overall, plasms. This selection of neoplasms of (HR 1.24 [95% CI 1.04; 1.47]; P = 0.10 the proportion of patients who had for interaction), as well as combined interest was based on several consider- fi con rmed neoplasms was 10.1% (3.34 malignant neoplasms (HR 1.15 [95% CI ations, including GLP-1 receptor tissue ex- events per 100 patient-years of obser- fi 0.94; 1.41]; P = 0.24 for interaction) and pression (7,8), published ndings from vation [PYO]) in the liraglutide group and LEADER (11), and other previously re- combined benign neoplasms (HR 1.21 9.0% in the placebo group (2.98 events [95% CI 0.89; 1.64]; P = 0.72 for inter- ported data. The sponsor also conducted per 100 PYO; estimated hazard ratio [HR] a post hoc review of clinical narratives action) were numerically higher with 1.12 [95% CI 0.99; 1.28]) (Fig. 1A). The liraglutide versus placebo in the sub- to categorize confirmed malignant neo- fi proportion of patients with con rmed group aged $65 years (Table 1). plasms of the liver and gallbladder malignant neoplasms was 6.2% (1.92 or bile duct. The investigator-reported events per 100 PYO), and a benign neo- Deaths Among Individuals With terms for these malignancies were plasm was confirmed in 3.4% of patients Malignant Neoplasm reviewed. (1.03 events per 100 PYO). The HR (95% Among patients who had a confirmed CI) for liraglutide versus placebo was malignant neoplasm, the proportions RESULTS 1.06 (0.90; 1.25) for “all malignant neo- that died of any cause (including deaths Baseline Characteristics plasms” and 1.16 (0.93; 1.44) for “all not related to cancer) were similar in the The baseline characteristics for the LEADER benign neoplasms” (Fig. 1A). liraglutide and placebo groups (25.3% vs. trial population (n =9,340)havebeen The cumulative incidence of confirmed 25.1%, respectively) (Table 2). In both published previously (11). The mean du- overall neoplasms, malignant neoplasms, treatment groups, most of these deaths ration of diabetes (mean glycated hemo- and benign neoplasms was similar with were classified as non-CV and attributed globin [HbA1c] 8.7% [72 mmol/mol]) was liraglutide treatment and placebo during to malignancy (19.6% vs. 19.7% of pa- 12.8 years, and 81.3% of the participants the randomized treatment period (Fig. 2). tients with an EAC-confirmed malignant had established CV disease (11). Partic- A small separation in the cumulative in- neoplasm treated with liraglutide vs. ipants were a mean age of 64.3 years, cidence of confirmed benign neoplasms placebo, respectively) (Table 2). and 64.3% were men. At baseline, 12.1% appeared after ;14 months (Fig. 2G)and of participants were smokers, 46.5% were was also observed for the cumulative Neoplasms of Interest previous smokers, and 41.4% had never incidence of confirmed overall neoplasms Results for pancreatic, thyroid, colorec- smoked. (Fig. 2A), but thereafter and for the re- tal, breast, prostate, nonmelanoma skin, mainder of the trial, these events occurred melanoma, and lung and bronchus neo- Exposure to Randomized Treatment at constant and similar rates in both plasms are presented in Fig. 1, Table 2, and Follow-up treatment groups. and the Supplementary Data. Findings As reported previously, the median du- from the analyses of malignant hepatic ration of exposure to liraglutide or pla- Sensitivity Analyses and biliary neoplasms are presented in cebo was 3.5 years (11). The median Sensitivity analyses excluding neoplasms the Supplementary Data. Demographics dose of liraglutide was 1.78 mg daily occurring ,1 year (Figs. 1B and 2B, E, and and baseline characteristics of patients 4 Neoplasms in LEADER Diabetes Care

Figure 1—Frequency of all confirmed neoplasms, malignant neoplasms, benign neoplasms, and neoplasms of interest. A: Main analysis. Events occurring later than 1 year (B) and later than 2 years (C) after randomization (sensitivity analysis) (11). Full analysis set (FAS) HRs are derived from a Cox proportional hazard regression model adjusted for treatment. C: Firth correction was used for malignant pancreatic neoplasms occurring later than 2 years after randomization. Proportions for breast neoplasms are calculated based on the number of female participants. Proportions for prostate neoplasms are calculated based on the number of male participants. Reprinted from Marso et al. (11) with permission from the Massachusetts Medical Society. Copyright © 2016 Massachusetts Medical Society. care.diabetesjournals.org Nauck and Associates 5

Figure 2—Cumulative incidence plots for confirmed neoplasms. A: Confirmed neoplasm index events. Confirmed neoplasm index events occurring later than 1 year (B) and later than 2 years (C) after randomization (sensitivity analysis). D: Confirmed malignant neoplasm index events. Confirmed malignant neoplasm index events occurringlater than 1 year (E) and later than 2 years (F) after randomization (sensitivity analysis). G: Confirmed benign neoplasmindexevents.Confirmedbenignneoplasmindexeventsoccurringlaterthan1year (H)andlaterthan2years(I)afterrandomization (sensitivity analysis). Cumulative incidence was estimated using the Aalen-Johansen method with death as a competing risk. A cumulative incidence probability of 0.1 is equivalent to 10%. HRs are derived from a Cox proportional hazard regression model adjusted for treatment and are for the proportion of patients with an event with liraglutide vs. placebo. who experienced a malignant neoplasm included one for all neoplasms of interest (n = 5 [0.11%] for each type of neoplasm of interest are presented in Supplemen- evaluated, except for malignant prostate [HR 2.59 {95% CI 0.92; 7.27}] in the main tary Table 3. neoplasms, which were experienced by a analysis) (Fig. 1A). No patients in the The most frequently occurring type of lower proportion of patients receiving placebo group had a confirmed malig- confirmed malignant neoplasm in the liraglutide (HR 0.54 [95% CI 0.34; 0.88]) nant pancreatic neoplasm after 2 years of overall study population was malignant (Fig. 1A). Malignant pancreatic neoplasm randomized treatment compared with nonmelanoma skin neoplasm (1.5% of or malignant melanoma events were five patients (0.11%) in the liraglutide patients [n = 140]). In the main analyses confirmed infrequently in both groups group (HR 10.90 [95% CI 0.46; 260.03]), comparing the occurrence of confirmed and for more patients in the liraglutide and only one patient (0.02%) had a con- neoplasms with liraglutide treatment group (n = 13 [0.28%] for each type of firmed malignant melanoma after 1 year versus placebo, the 95% CI for the HR neoplasm) than in the placebo group of receiving placebo compared with 6 Neoplasms in LEADER Diabetes Care

fi Table 1—Frequency of confirmed neoplasms by sex and age subgroups It is therefore dif cult to evaluate how P value for the overall cancer rates in LEADER compare interaction with those in the general population of between patients with type 2 diabetes. Among Liraglutide Placebo treatment and those patients in LEADER with a con- n (%) n (%) HR (95% CI) subgroup firmed malignant neoplasm, most deaths Full analysis set 4,668 (100.0) 4,672 (100.0) were due to malignancy. The occurrence All neoplasms 470 (10.1) 419 (9.0) 1.12 (0.99; 1.28) of deaths attributable to malignancy Female 127 (7.7) 123 (7.3) 1.04 (0.81; 1.34) Sex: 0.50 was similar with liraglutide and placebo; Male 343 (11.4) 296 (9.9) 1.16 (0.99; 1.35) hence, cause of death seemed to reflect Aged ,65 years 178 (7.1) 177 (7.1) 0.99 (0.80; 1.22) Age: 0.10 the prognosis of underlying malignant $ Aged 65 years 292 (13.5) 242 (11.1) 1.24 (1.04; 1.47) disease, irrespective of treatment group. All malignant neoplasms 296 (6.3) 279 (6.0) 1.06 (0.90; 1.25) There was no suggestion of an asso- Female 86 (5.2) 77 (4.6) 1.13 (0.83; 1.54) Sex: 0.60 ciation between liraglutide exposure Male 210 (7.0) 202 (6.8) 1.03 (0.85; 1.25) Aged ,65 years 101 (4.0) 106 (4.2) 0.94 (0.71; 1.23) Age: 0.24 and development of malignant neoplasms Aged $65 years 195 (9.0) 173 (8.0) 1.15 (0.94; 1.41) in the liraglutide clinical development All benign neoplasms 168 (3.6) 145 (3.1) 1.16 (0.93; 1.44) program for treatment of type 2 diabetes Female 41 (2.5) 42 (2.5) 0.98 (0.64; 1.51) Sex: 0.40 (at doses up to 1.8 mg daily for up to Male 127 (4.2) 103 (3.4) 1.22 (0.94; 1.59) 22 months) (14). Slight imbalances in Aged ,65 years 79 (3.1) 70 (2.8) 1.11 (0.80; 1.53) Age: 0.72 rates of and $ Aged 65 years 89 (4.1) 75 (3.5) 1.21 (0.89; 1.64) prostate cancer were identified with Full analysis set. HRs are derived from a Cox proportional hazard regression model adjusted for liraglutide versus a pooled comparator treatment. group consisting of an active compar- ator and placebo; however, rates were similar compared with placebo alone (14). 11 patients (0.24%) in the liraglutide subtypes. In our study we have limited In LEADER, the frequency of confirmed group (HR 10.95 [95% CI 1.41; 84.82]) power to examine any specific cancer malignant was similar (Fig. 1). Details for individual cases of subtype. The primary end point power in the liraglutide and placebo groups, and malignant or premalignant pancreatic calculation for LEADER determined that the only observed case of MTC occurred neoplasm or melanoma are provided in 611 events were required to exclude a in a patient assigned to placebo. Fewer Supplementary Tables 2 and 4. Several HR .1.3 with 90% power. The cancer participants experienced malignant pros- patients who developed malignant pan- subtype with the highest number of tate neoplasms in the liraglutide group creatic neoplasm experienced weight events in LEADER was nonmelanoma than in the placebo group (11). In vitro loss before their diagnosis, and some skin cancer, affecting 140 patients studies have shown that liraglutide and experienced increases in HbA1c before (liraglutide, 78 patients [1.7%]; placebo, significantly inhibit prolifera- diagnosis, but there was no universal 62 patients [1.3%]) during the entire tion in human prostate cancer cell lines pattern. study period. This number of events through GLP-1 receptor–dependent mech- gives a power of 34% to exclude a HR anisms (28,29), giving some plausibility CONCLUSIONS .1.3 for nonmelanoma skin cancer; to to our observation. Alternatively, greater LEADER has provided information about exclude a HR .1.3% with a power of urinary frequency among placebo-treated intermediate-term neoplasm risk with 90%, the number of patients required is patients (due to greater hyperglycemia liraglutide versus placebo (both in com- ;36,000 (assuming the observed rate of and more diuretic use) and consequen- bination with standard care) in 9,340 nonmelanoma skin cancer events of 0.4% tly more frequent investigation, could patients with type 2 diabetes monitored subjects/year and $3.5 years’ observa- account for the small difference. for 3.5–5 years. There was a greater total tion time). Beyond these limitations We report a numeric imbalance in number of neoplasms with liraglutide around the power of the trial to detect the frequency of adjudication-confirmed compared with placebo (470 vs. 419 differences in cancer incidence, the lim- malignant pancreatic neoplasms in the total [benign, premalignant, or malignant] ited time frame (median follow-up of liraglutide group versus placebo group neoplasms, respectively; HR 1.12 [95% 3.8 years) is inadequate to exclude the (0.28% [n = 13] vs. 0.11% [n = 5], re- CI 0.99; 1.28]). However, our overall carcinogenic potential of a drug. spectively; HR 2.59 [95% CI 0.92; 7.27]). analyses do not suggest a major increase The proportion of patients with con- Additional details on the capture and (.25% based on the 95% CI) in the risk firmed malignant neoplasms during adjudication of these cases can be found of malignant neoplasms with liraglutide, LEADER was 6.2% (1.92 events per in the Supplementary Data (Neoplasms in general terms, versus placebo (HR 1.06 100 PYO), which is lower than that for of interest: Pancreatic neoplasms). The [95% CI 0.90; 1.25]). the primary composite end point for development of clinically evident pan- Overall cancer rates, although of clin- major adverse CV events (11). To our creatic cancer is a process that takes ical interest, do not have a known bi- knowledge, evidence from global epide- many years (30). A substantial number of ological relevance because all carcinogens miological studies on cancer prevalence our patients with confirmed malignant reported to date affect specificcancer in people with type 2 diabetes is lacking. pancreatic neoplasms had advanced care.diabetesjournals.org Nauck and Associates 7

Table 2—Confirmed deaths, including deaths not related to cancer, among therapy (median follow-up 15 months) patients with confirmed malignant neoplasm (22). Monitoring of pancreatic cancer in Liraglutide Placebo CV outcome trials (such as LEADER) has n (%) n (%) been cited as a strategy for increasing All malignant neoplasms 296 (100.0) 279 (100.0) knowledge in this area (26). To date, All-cause mortality 75 (25.3) 70 (25.1) CV outcome trials have consistently Known cause of death 73 (24.7) 64 (22.9) reported pancreatic cancer occurs in CV death 8 (2.7) 1 (0.4) similar or fewer numbers of patients Non-CV death 65 (22.0) 63 (22.6) receiving GLP-1RAs than placebo: Malignancy 58 (19.6) 55 (19.7) semaglutide (SUSTAIN-6 [Trial to Evalu- Malignant pancreatic neoplasms 13 (100.0) 5 (100.0) ate Cardiovascular and Other Long-term All-cause mortality 11 (84.6) 5 (100.0) Outcomes with Semaglutide in Subjects Known cause of death 11 (84.6) 5 (100.0) CV death 0 (0.0) 0 (0.0) with Type 2 Diabetes]), with a mean Non-CV death 11 (84.6) 5 (100.0) observation time of 2.1 years; lixisena- Malignancy 11 (84.6) 5 (100.0) tide (ELIXA [Evaluation of Lixisenatide in Malignant thyroid neoplasms 5 (100.0) 3 (100.0) Acute Coronary Syndrome]), with a me- All-cause mortality 0 (0.0) 0 (0.0) dian follow-up of 25 months; and exena- Malignant colorectal neoplasms 28 (100.0) 28 (100.0) tide once weekly (EXSCEL [Exenatide All-cause mortality 6 (21.4) 6 (21.4) Study of Cardiovascular Event Lower- Known cause of death 6 (21.4) 4 (14.3) ing]), with a median follow-up of 3.2 CV death 0 (0.0) 0 (0.0) years (12,31,32). A recent meta-analysis Non-CV death 6 (21.4) 4 (14.3) of all GLP-1RA CV outcome trials (includ- Malignancy 3 (10.7) 4 (14.3) ing LEADER) also reported an absence of Malignant breast neoplasms (female only) 21 (100.0) 20 (100.0) fi All-cause mortality 1 (4.8) 1 (5.0) a signi cant effect for GLP-1RA on rates Known cause of death 1 (4.8) 1 (5.0) of pancreatic cancer and MTC, albeit CV death 1 (4.8) 0 (0.0) with significant heterogeneity between Non-CV death 0 (0.0) 1 (5.0) the four eligible trials (33). Malignancy 0 (0.0) 1 (5.0) We also report a numeric imbalance in Malignant prostate neoplasms (male only) 26 (100.0) 47 (100.0) the occurrence of adjudication-confirmed All-cause mortality 0 (0.0) 6 (12.8) malignant melanoma with liraglutide – Known cause of death 6 (12.8) (0.28% [n = 13]) vs. placebo (0.11% CV death 0 (0.0) 0 (0.0) Non-CV death 0 (0.0) 6 (12.8) [n = 5], HR 2.59 [95% CI 0.92; 7.27]). Only Malignancy 0 (0.0) 3 (6.4) one patient developed malignant mela- Malignant nonmelanoma skin neoplasm 78 (100.0) 62 (100.0) noma after 1 year of receiving placebo. All-cause mortality 3 (3.8) 3 (4.8) The low numbers of patients led to wide Known cause of death 3 (3.8) 3 (4.8) CIs for the HRs for malignant melanoma, CV death 1 (1.3) 0 (0.0) thus precluding firm conclusions. The Non-CV death 2 (2.6) 3 (4.8) HR for skin neoplasms with semaglutide Malignancy 1 (1.3) 1 (1.6) versus placebo in SUSTAIN-6 was 1.41 Malignant melanoma 13 (100.0) 5 (100.0) (95% CI 0.76; 2.63), though the relevance All-cause mortality 2 (15.4) 1 (20.0) Known cause of death 2 (15.4) 0 (0.0) of comparison with our analysis is very CV death 0 (0.0) 0 (0.0) limited, with only one of the six skin Non-CV death 2 (15.4) 0 (0.0) neoplasms reported in SUSTAIN-6 being Malignancy 2 (15.4) 0 (0.0) classified as malignant melanoma (12). Malignant lung and bronchus neoplasms 28 (100.0) 33 (100.0) The primary reports for ELIXA and EXSCEL All-cause mortality 21 (75.0) 22 (66.7) did not include data specifictoskin Known cause of death 20 (71.4) 20 (60.6) neoplasms (31,32). A meta-analysis of CV death 2 (7.1) 0 (0.0) randomized clinical trials of 24 weeks’ Non-CV death 18 (64.3) 20 (60.6) ’ Malignancy 17 (60.7) 20 (60.6) to 5 years duration showed that DPP-4 inhibitors were not associated with a fi The EAC CV subcommittee classi ed deaths with a known cause into CV or non-CV deaths. The significantly increased risk of developing sponsor further categorized the deaths according to plausible cause. malignant melanoma (25). Preclinical data suggest that GLP-1RAs stages, suggesting the tumors originated (18,20,21), a recent large, international may have GLP-1 receptor-dependent in- well before the patients were recruited registry study did not confirm an in- testinotrophic effects (19). Treatment into LEADER. creased risk of pancreatic cancer versus with exenatide for 4 weeks significantly Although earlier studies, subject to sulfonylureas (median follow-up 1.3–2.8 increased polyp number and size, pri- criticism on methodological grounds, years) (23), in line with a U.S. insur- marily in the distal small bowel (19), an suggested an increased risk for pancre- ance claims database study that com- atypical location for human gastrointes- atic tumors with incretin-based therapy pared liraglutide with non–GLP-1–based tinal adenomas. 8 Neoplasms in LEADER Diabetes Care

In clinical trials of liraglutide at higher sensitivity. To address this, we also ex- Sharp&Dohme,andNovoNordisk,travel doses (up to 3.0 mg daily for the weight amined cases based on clinical reporting, support from Berlin-Chemie, Eli Lilly and Co., management indication), small imbalan- which generally confirmed the adjudi- Merck Sharp & Dohme, Novo Nordisk, and In- tarcia Therapeutics/Servier, and grant support ces between treatment groups were cated findings. (to his institution) from AstraZeneca, Eli Lilly noted in the number of patients with Fourthly, although of scientificand and Co., GlaxoSmithKline, Intarcia Therapeutics/ benign colorectal neoplasm events (n = clinical interest, our subgroup analyses Servier, Merck Sharp & Dohme, Novartis, and 17 [0.52%] with liraglutide vs. n =4 and sensitivity analyses excluding neo- Novo Nordisk. T.J.J., C.R., and S.C. are employees , , and shareholders of Novo Nordisk. J.B.B. reports [0.22%] with placebo; odds ratio 2.39 plasms occurring 1 year or 2 years receiving contracted consulting fees, paid to his [95% CI 0.78; 9.76]) (15). These findings after randomization were based on lower institution, and travel support from Adocia, were not reproduced in LEADER, where numbers of patients than the main AstraZeneca, Dance Biopharm, Dexcom, Elcelyx the liraglutide dose was lower (up to analyses, further reducing the precision Therapeutics, Eli Lilly and Co., Fractyl, GI Dynam- 1.8 mg daily). of the HRs and complicating data inter- ics, Intarcia Therapeutics, Lexicon, Merck, Meta- vention, NovaTarg, Novo Nordisk, Orexigen, Furthermore, in clinical trials of liraglutide pretation. In addition, LEADER was con- PhaseBio, Sanofi, Shenzhen HighTide, and vTv for weight management (up to 3.0 mg ducted in a population at high CV risk Therapeutics, grant support from AstraZeneca, daily), an imbalance was reported in the and excluded patients based on a history , Boehringer Ingelheim, Eli Lilly, GI Dy- number of women with malignant or of malignancy, which could potentially namics, GlaxoSmithKline, Intarcia Therapeutics, premalignant breast neoplasms in the limit the generalizability of the findings Johnson & Johnson, Lexicon, Medtronic, Merck, Novo Nordisk, Orexigen, Sanofi, Scion Neuro- liraglutide (n =14[0.59%])versusplacebo (37). Stim, Takeda, and Theracos. J.B.B. also reports group (n = 3 [0.23%]; odds ratio 2.56 [95% Finally, the generalizability of our find- receiving fees and holding stock options in In- CI 0.71; 13.91]) (15). It was speculated ings to GLP-1RAs other than liraglutide sulin Algorithms and PhaseBio, and reports serv- that weight loss may have led to en- up to doses of 1.8 mg daily is unclear. ing on the board of the AstraZeneca Healthcare hanced detection of breast neoplasms in In conclusion, overall safety analyses Foundation. Author Contributions. M.A.N., T.J.J., C.R., S.C., liraglutide-treated women (15). Similar of data from the LEADER trial do not and J.B.B. participated in analysis and interpre- findings were reported from an obser- suggest a major increase in the risk of tation of the data, edited the manuscript for vational study in GLP-1RA users (34). malignant neoplasms with liraglutide, in important intellectual content, approved the In LEADER, with longer exposure in a general terms, versus placebo. Based on final version, and take full responsibility for its contents. M.A.N. and J.B.B. participated in the larger population with more cases re- small numbers, imbalances with uncer- development of the protocol and in the conduct ported, the overall incidence of con- tain significance were observed for neo- and oversight of the protocol. M.A.N. is the firmed malignant breast neoplasms was plasms of some tissue types, without guarantor of this work and, as such, had full similar across the liraglutide and placebo allowing firm conclusions. We cannot access to all the data in the study, and takes groups. comment on the longer-term neoplasm responsibility for the integrity of the data and the accuracy of the data analysis. Our study has limitations. Firstly, risk beyond a median follow-up time of Prior Presentation. Parts of this study were LEADER was not primarily designed or 3.8 years. Only long-term, larger, pro- presented as oral presentations at the 76th Sci- powered to provide conclusive data re- spective, pragmatic, or observational entific Sessions of the American Diabetes Asso- garding occurrence of neoplasms with registry studies may help to answer ciation, New Orleans, LA, 10–14 June 2016, and liraglutide. A relatively low frequency questions about the longer-term safety at the 52nd Annual Meeting of the European Association for the Study of Diabetes, Munich, of neoplasms was observed for most of liraglutide and other GLP-1RAs. Germany, 12–16 September 2016. tissues, precluding firm conclusions. In addition, the median follow-up time in References LEADER was 3.8 years (11), and longer- Acknowledgments. The authors thank Helle 1. Cohen DH, LeRoith D. Obesity, type 2 di- term neoplasm risk could not be as- Frimer-Larsen (Novo Nordisk A/S) for statistical abetes, and cancer: the insulin and IGF connec- sessed. support and for reviewing the manuscript. Med- tion. Endocr Relat Cancer 2012;19:F27–F45 Secondly, neoplasms in some tissue ical writing assistance was provided by Water- 2. Tsilidis KK, Kasimis JC, Lopez DS, Ntzani EE, meadow Medical, an Ashfield Company, part of Ioannidis JP. Type 2 diabetes and cancer: um- types may have been subject to detec- UDG Healthcare plc, funded by Novo Nordisk brella review of meta-analyses of observational tion bias (e.g., due to gastrointestinal ad- A/S. Editorial support was provided by Helen studies. BMJ 2015;350:g7607 verse events or weight loss observed Marshall of Watermeadow Medical, funded by 3. Garc´ıa-Jimenez´ C, Gutierrez-Salmer´ on´ M, with liraglutide), and a role for detection Novo Nordisk A/S. The authors are grateful to Chocarro-Calvo A, Garc´ıa-Martinez JM, Castano~ A, or surveillance bias in the relationship the LEADER trial participants, investigators De la Vieja A. From obesity to diabetes and (available in the Supplementary Data), and cancer: epidemiological links and role of ther- between diabetes and cancer has been trial-site staff, and the leadership, employees, apies. Br J Cancer 2016;114:716–722 suggested previously (35,36). and contractors of the sponsor who were in- 4. Kyrgiou M, Kalliala I, Markozannes G, et al. Thirdly, a relatively low proportion of volved in the conduct of the trial. Adiposity and cancer at major anatomical sites: potential neoplasms captured were con- Funding. J.B.B. received support from the U.S. umbrella review of the literature. BMJ 2017;356: firmed by adjudication. These events National Institutes of Health award UL1-TR- j477 001111. 5. Chen Y, Wu F, Saito E, et al. Association were reported and sent for adjudication This study was supported by Novo Nordisk. between type 2 diabetes and risk of cancer according to a broad study definition to Duality of Interest. M.A.N. reports receiving mortality: a pooled analysis of over 771,000 help ensure a robust process. The em- fees for serving on advisory boards from Astra- individuals in the Asia Cohort Consortium. phasis on a histological or cytological Zeneca, Berlin-Chemie, Eli Lilly and Co., Fractyl, Diabetologia 2017;60:1022–1032 Hanmi, Merck Sharp & Dohme, Novo Nordisk, 6. Drucker DJ, Nauck MA. The incretin system: diagnosis during neoplasm adjudication and Intarcia Therapeutics/Servier, lecture fees glucagon-like peptide-1 receptor agonists and provided high specificity in confirmed from AstraZeneca, Berlin-Chemie, Boehringer dipeptidyl peptidase-4 inhibitors in type 2 di- casesbutmayalsohavedecreased Ingelheim, Eli Lilly and Co., Medscape, Merck abetes. Lancet 2006;368:1696–1705 care.diabetesjournals.org Nauck and Associates 9

7. Baggio LL, Drucker DJ. Biology of incretins: 16. Bjerre Knudsen L, Madsen LW, Andersen S, 26. Egan AG, Blind E, Dunder K, et al. Pancreatic GLP-1 and GIP. Gastroenterology 2007;132: et al. Glucagon-like peptide-1 receptor agonists safety of incretin-based drugs--FDA and EMA 2131–2157 activate rodent thyroid C-cells causing assessment. N Engl J Med 2014;370:794–797 8. Pyke C, Heller RS, Kirk RK, et al. GLP-1 receptor release and C-cell proliferation [published cor- 27. Marso SP, Poulter NR, Nissen SE, et al. De- localization in monkey and human tissue: novel rection appears in 2012;153: sign of the liraglutide effect and action in dia- distribution revealed with extensively validated 1000]. Endocrinology 2010;151:1473–1486 betes: evaluation of cardiovascular outcome monoclonal antibody. Endocrinology 2014;155: 17. Matveyenko AV, Dry S, Cox HI, et al. Ben- results (LEADER) trial. Am Heart J 2013;166:823– 1280–1290 eficial endocrine but adverse exocrine effects 830.e5 9. Vilsbøll T, Christensen M, Junker AE, Knop FK, ofsitagliptininthehumanisletamyloidpoly- 28. Nomiyama T, Kawanami T, Irie S, et al. Gluud LL. Effects of glucagon-like peptide-1 peptide transgenic rat model of type 2 diabetes: Exendin-4, a GLP-1 receptor agonist, attenuates receptor agonists on weight loss: systematic interactions with metformin. Diabetes 2009; prostate cancer growth. Diabetes 2014;63: review and meta-analyses of randomised con- 58:1604–1615 3891–3905 trolled trials. BMJ 2012;344:d7771 18. Butler AE, Campbell-Thompson M, Gurlo T, 29. Li XN, Bu HM, Ma XH, et al. Glucagon-like 10. Nauck M. Incretin therapies: highlight- Dawson DW, Atkinson M, Butler PC. Marked peptide-1 analogues inhibit proliferation and ing common features and differences in the expansion of exocrine and endocrine increase apoptosis of human prostate cancer modes of action of glucagon-like peptide-1 with incretin therapy in humans with increased cells in vitro. Exp Clin Endocrinol Diabetes 2017; receptor agonists and dipeptidyl peptidase-4 exocrine pancreas dysplasia and the potential 125:91–97 inhibitors. Diabetes Obes Metab 2016;18:203– for glucagon-producing neuroendocrine tumors. 30. Yachida S, Jones S, Bozic I, et al. Distant 216 Diabetes 2013;62:2595–2604 occurs late during the genetic evo- 11. Marso SP, Daniels GH, Brown-Frandsen K, 19. Koehler JA, Baggio LL, Yusta B, et al. GLP-1R lution of pancreatic cancer. Nature 2010;467: et al.; LEADER Steering Committee; LEADER Trial agonists promote normal and neoplastic intes- 1114–1117 Investigators. Liraglutide and cardiovascular out- tinal growth through mechanisms requiring Fgf7. 31. Pfeffer MA, Claggett B, Diaz R, et al.; ELIXA comes in type 2 diabetes. N Engl J Med 2016;375: Cell Metab 2015;21:379–391 Investigators. Lixisenatide in patients with type 2 311–322 20. Elashoff M, Matveyenko AV, Gier B, Elashoff diabetes and acute coronary syndrome. N Engl J 12. MarsoSP,Bain SC, Consoli A,etal.;SUSTAIN-6 R, Butler PC. Pancreatitis, pancreatic, and thyroid Med 2015;373:2247–2257 Investigators. Semaglutide and cardiovascular cancer with glucagon-like peptide-1-based ther- 32. Holman RR, Bethel MA, Mentz RJ, et al.; outcomes in patients with type 2 diabetes. apies. Gastroenterology 2011;141:150–156 EXSCEL Study Group. Effects of once-weekly ex- N Engl J Med 2016;375:1834–1844 21. Butler PC, Elashoff M, Elashoff R, Gale EA. A enatide on cardiovascular outcomes in type 2 13. Dalsgaard NB, Brønden A, Vilsbøll T, Knop FK. critical analysis of the clinical use of incretin- diabetes. N Engl J Med 2017;377:1228–1239 Cardiovascular safety and benefits of GLP-1 re- based therapies: are the GLP-1 therapies safe? 33. Bethel MA, Patel RA, Merrill P, et al.; EXSCEL ceptor agonists. Expert Opin Drug Saf 2017;16: Diabetes Care 2013;36:2118–2125 Study Group. Cardiovascular outcomes with 351–363 22. Funch D, Gydesen H, Tornøe K, Major- glucagon-like peptide-1 receptor agonists in 14. Novo Nordisk. Liraglutide (injection) for the Pedersen A, Chan KA. A prospective, claims- patients with type 2 diabetes: a meta-analysis. treatment of patients with type 2 diabetes. NDA based assessment of the risk of pancreatitis Lancet Diabetes Endocrinol 2018;6:105–113 22-341 Briefing Document. Endocrine and Meta- and pancreatic cancer with liraglutide compared 34. Hicks BM, Yin H, Yu OH, Pollak MN, Platt RW, bolic Drug Advisory Committee, 2 April 2009 to other antidiabetic drugs. Diabetes Obes Azoulay L. Glucagon-like peptide-1 analogues [Internet], 2009. Available from https://wayback. Metab 2014;16:273–275 andriskofbreastcancerinwomenwithtype2 archive-it.org/7993/20170405221128/https:// 23. Azoulay L, Filion KB, Platt RW, et al.; Cana- diabetes: population based cohort study using www.fda.gov/downloads/AdvisoryCommittees/ dian Network for Observational Drug Effect the UK Clinical Practice Research Datalink. BMJ CommitteesMeetingMaterials/Drugs/Endocrinologic Studies Investigators. Incretin based drugs and 2016;355:i5340 andMetabolicDrugsAdvisoryCommittee/UCM148659 the risk of pancreatic cancer: international multi- 35. Dankner R, Boffetta P, Balicer RD, et al. Time- .pdf. Accessed 6 April 2018 centre cohort study. BMJ 2016;352:i581 dependent risk of cancer after a diabetes di- 15. Novo Nordisk. Liraglutide 3.0 mg for weight 24. Htoo PT, Buse JB, Gokhale M, Marquis MA, agnosis in a cohort of 2.3 million adults. Am J management. NDA 206-321 Briefing Document. Pate V, Sturmer¨ T. Effect of glucagon-like peptide- Epidemiol 2016;183:1098–1106 Endocrinologic and Metabolic Drug Advisory 1 receptor agonists and dipeptidyl peptidase-4 36. Lega IC, Wilton AS, Austin PC, Fischer HD, Committee, September 11, 2014 [Internet], inhibitors on colorectal cancer incidence and its Johnson JA, Lipscombe LL. The temporal relation- 2014. Available from https://wayback.archive-it precursors. Eur J Clin Pharmacol 2016;72:1013– ship between diabetes and cancer: a population- .org/7993/20170405215636/https://www.fda 1023 based study. Cancer 2016;122:2731–2738 .gov/downloads/AdvisoryCommittees/Committees 25. Zhao M, Chen J, Yuan Y, et al. Dipeptidyl 37. Rinde LB, Smabrekke˚ B, Hald EM, et al. MeetingMaterials/Drugs/Endocrinologicand peptidase-4 inhibitors and cancer risk in patients Myocardial infarction and future risk of cancer MetabolicDrugsAdvisoryCommittee/UCM413318 with type 2 diabetes: a meta-analysis of ran- in the general population-the TromsøStudy. Eur J .pdf. Accessed 6 April 2018 domized clinical trials. Sci Rep 2017;7:8273 Epidemiol 2017;32:193–201