Eur J Clin Pharmacol. 2016 Aug;72(8):1013-23. doi: 10.1007/s00228-016-2068-3.

Effect of glucagon-like peptide-1 receptor agonists and dipeptidyl peptidase-4 inhibitors on colorectal cancer incidence and its precursors.

Htoo PT1, Buse JB2, Gokhale M1, Marquis MA3, Pate V1, Stürmer T4.
  • 1Department of Epidemiology, University of North Carolina at Chapel Hill, Gillings School of Global Public Health, Campus Box 7435, 2101 McGavran-Greenberg Hall, Chapel Hill, NC, USA.
  • 2Department of Medicine, University of North Carolina School of Medicine, Chapel Hill, NC, USA.
  • 3Collaborative Studies Coordinating Center, Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA.
  • 4Department of Epidemiology, University of North Carolina at Chapel Hill, Gillings School of Global Public Health, Campus Box 7435, 2101 McGavran-Greenberg Hall, Chapel Hill, NC, USA.



Incretin-based antihyperglycemic therapies increase intestinal mucosal expansion and polyp growth in mouse models. We aimed to evaluate the effect of dipeptidyl peptidase-4 inhibitors (DPP-4i) or glucagon-like peptide-1 receptor agonists (GLP-1ra) initiation on colorectal cancer incidence.


We conducted a cohort study on US Medicare beneficiaries over age 66 from 2007 to 2013 without prevalent cancer. We identified three active-comparator and new-user cohorts: DPP-4i versus thiazolidinediones (TZD), DPP-4i versus sulphonylureas (SU), and GLP-1ra versus long acting insulin (LAI). Follow-up started from 6 months post-second prescription and ended 6 months after stopping (primary as-treated analysis). We estimated hazard ratios (HR) and 95 % confidence intervals (CI) for incident colorectal cancer adjusting for measured confounders using propensity score weighting.


The median duration of treatment ranged 0.7-0.9 years among DPP-4i cohorts. Based on 104 events among 39,334 DPP-4i and 63 events among 25,786 TZD initiators, there was no association between DPP-4i initiation and colorectal cancer (adjusted HR = 1.17 (CI 0.88, 1.71)). There were 73 events among 27,047 DPP-4i and 266 events among 76,012 SU initiators with the adjusted HR 0.98 (CI 0.74, 1.30). We identified 5600 GLP-1ra and 54,767 LAI initiators and the median duration of treatment was 0.8 and 1.2 years, respectively. The adjusted HR was 0.82 (CI 0.42, 1.58) based on <11 events among GLP-1ra versus 276 events among LAI initiators.


Although limited by the short duration of treatment, our analyses based on real-world drug utilization patterns provide evidence of no short-term effect of incretin-based agents on colorectal cancer.


Cohort study; Colorectal cancer; Comparative effectiveness research; Dipeptidyl peptidase-4 inhibitors; Glucagon-like peptide-1 receptor agonists; Pharmacoepidemiology

PMID: 27165664



Incretins are intestinal hormones (e.g., glucagon like peptide 1, GLP-1), which help lower blood glucose by enhancing glucose-dependent insulin secretion, reducing inappropriate glucagon secretion and improving satiety among other actions [1]. Incretin-based antihyperglycemic agents, dipeptidyl peptidase-4 inhibitors (DPP-4i) and glucagon like peptide-1 receptor agonists (GLP-1ra), exert their anti-hyperglycemic actions by increasing incretin levels [2,3]. They are commonly used as second or third line agents in type 2 diabetes mellitus (DM) patients [1]. There are several marketed GLP-1ra which are analogs or mimetics of GLP-1, injected twice-a-day, once-a-day or once-weekly.  DPP-4i are small molecules delivered as once-daily tablets which inhibit DPP-4, a ubiquitous enzyme that normally degrades GLP-1 and other incretin hormones rapidly [3].

In studies of a genetically modified mouse model prone to colon cancer, treatment with incretin-based therapy was associated with intestinal mucosal expansion, crypt fission and increased number and size of polyps [4]. Though incretin-based therapies have been suggested to increase the risk of pancreatic cancer, this has not been supported in a population-based cohort study of older adults [5]. Their effects on initiation or progression of colorectal cancers on the other hand had not previously been hypothesized or explored in population-based studies.

Diabetes mellitus is a chronic disease associated with an increased risk of several types of cancers and the effects of chronic use of anti-hyperglycemic agents on modifying cancer risks are relevant to public health efforts to optimize care of diabetes patients. Data from clinical trials on colorectal cancers are limited by small numbers of patients exposed [6]. Furthermore, patients in clinical trials are often healthier and younger than patients with diabetes in the general populations. Clinicians need information on potential cancer risks associated with chronic medication use and arguably that is best derived from real world data on patients treated with diabetes medications in the community and their health outcomes.   

We decided to study the development of colorectal cancers in patients who started treatment with incretin-based anti-hyperglycemic agents compared to other second-line anti-hyperglycemic agents using a population-based healthcare database covering a 20% sample of Medicare beneficiaries. Medicare is the largest insurance program in the U.S., covering over 98% of adults over the age of 65. The Medicare database includes information on the use of inpatient, outpatient and pharmacy services.

We identified new medication users of incretins and comparison drugs – thiazolidinediones (TZD), sulfonylureas (SU) and long-acting insulins (LAI). We limited our study population to new medication users so that we could synchronize follow-up at initiation in each patient and thereby cancer risks could be examined in a time varying manner [7]. Our study compared rates of new cases of colorectal cancer in three sets of patients: (i) DPP-4i vs. TZD, (ii) DPP-4i vs. SU, and (iii) GLP-1ra vs. LAI. These comparisons were chosen to best represent the similar stage of diabetes progression with incretin and non-incretin treatment and thus minimize the chance of confounding biases [1]. We required all study participants to be free of any cancer diagnoses or cancer-diagnostic procedures prior to the start of the follow-up. Detailed schematic of cohort construction is presented in Fig. 1.


figure-1-study-timelineWe started the follow-up for new colorectal cancer diagnoses 6 months after the second filled prescription for each drug to increase the likelihood that the patient was taking the drug chronically and to minimize noise related to cancers diagnosed in the first few months after a new prescription which are unlikely to be caused by the treatment (Figure 1). This ‘lag’ period before which we started counting cancer outcomes is called the ‘empirical induction period’ and we excluded patients with colorectal cancer outcomes during this period [8]. We varied this period to various lengths in additional analyses to check the robustness of our findings. The follow-up ended on 2013 December 31st, or the diagnosis date of any cancers, or the study outcome (colorectal cancer diagnosis identified by the reimbursed claims submitted by clinicians), or death. Other diabetes medication changes resulted in censoring for further events in the ‘as-treated’ analyses while in the ‘intent-to-treat’, we followed the patient regardless if the patient stayed on drugs of the same class or stopped or switched drug classes during follow-up. Thus our new user active comparator cohorts mimic population-based ‘trials’ in which patients initiated medications of interest and were followed up for cancer outcomes [7]. Though our study is not a randomized trial, this ‘new user active comparator’ design has been shown to minimize bias associated with drug indications and underlying disease severity [7].

We evaluated time to colorectal cancer outcomes between our comparison cohorts and conducted multivariable statistical analyses (propensity score weighted Cox proportional hazards models) [9]. We present the incidence of colorectal cancer over the course of follow-up in Fig. 3 (Kaplan-Meier curves).






Due to the choice of our comparison groups, our cohorts were similar in terms of baseline characteristics (demographics, clinical comorbidities, co-medications use and healthcare system utilization) even prior to statistical adjustments or weighting. For example, the age distributions were similar between our cohorts (mean age: 76 vs. 74 in DPP-4i vs. TZD, 76 vs. 75 in DPP-4i vs. SU and 72 vs. 75 in GLP-1ra vs. LAI). The proportions of white race in DPP-4i vs. TZD, DPP-4i vs. SU and GLP-1ra vs. LAI were 74% vs. 70%, 72% vs. 76% and 87% vs. 74% respectively. The prevalence of baseline comorbid conditions was also more or less similar in our cohorts with some minor differences: (e.g. congestive heart failure of 25% in DPP-4i vs. 16% in TZD; approximately 23% each in DPP-4i vs. SU cohort, and 15% vs. 29% in GLP-1ra vs. LAI). All these differences were removed after statistical weighting. Cumulative incidence functions show that the curves are parallel and close to each other indicating that cancer incidences are similar between incretin-based therapies and comparator drugs even after adjusting for baseline covariates (Fig. 3). Multivariable hazard ratios (and 95% confidence intervals) from statistical models also indicate that colorectal cancer incidence is similar between our comparison cohorts, which holds true even after varying our initial assumption of empirical induction period (Fig. 4). Hazard ratios are consistent irrespective of as-treated and intent-to-treat analyses.


figure-4-induction-periodsSince the genetically modified mouse study demonstrated an increased number and size of polyps, we also evaluated the risk of colorectal polyps and adenomas between incretin-based therapies and comparator drugs. We combined invasive and benign tumor to increase the number of events. Results were similar to the invasive colorectal cancer analysis. There was no association between initiation of incretin-based therapies and the short term risk of composite colorectal tumor outcomes.

Our study has a major strength in that it is based on the real world drug utilization data with a new user active comparator study design which allows us to conduct population ‘trials’ that follows patients from the date of their drug initiation, synchronize their baseline cancer risks and reduce potential bias often associated with observational studies. As a result of our study design, our comparison cohorts were similar to each other in most measured baseline characteristics, although they were not randomized. We still could not have controlled for unmeasured characteristics such as smoking, and physical activity but the choice of comparison drug users who are likely at a similar stage of diabetes progression reduces the potential for bias. We were also able to control for baseline diagnoses of cardiovascular or lung diseases which reduce unmeasured bias by smoking, obesity or physical activity.

Our study has limitations: it is based on data on dispensed and reimbursed prescriptions therefore lacking data on 4$ generics and samples. Sample drug use is usually temporary and therefore its effects on our findings are likely minimal. The short follow-up of our study (median duration without induction period: 8-14 months for as treated and 24-39 months for intent-to-treat) implies that we were only able to rule out short-term cancer effects and potential long-term cancer effects by incretin-based treatments cannot be excluded. Our cancer definition is based on diagnostic claims submitted by physicians rather than pathology reports. These definitions have been shown to have high validity (sensitivity and specificity) in prior validation studies [10]. We were not able to separate colon from rectum cancer and polyps from adenomas.

            In conclusion, our population-based cohort study is the first to examine the potential association between incretin-based anti-hyperglycemic agents for the treatment of type 2 diabetes and colorectal cancer. Fortunately for patients, there does not seem to be an increased short-term risk for colorectal cancer or polyps in patients treated with incretin-based therapy as compared to alternative anti-hyperglycemic therapies.  


Conflict of interest:  This study was not funded.



1.      Inzucchi SE, Bergenstal RM, Buse JB, Diamant M, Ferrannini E, Nauck M, Peters AL, Tsapas A, Wender R, Matthews DR (2015) Management of hyperglycemia in type 2 diabetes, 2015: a patient-centered approach: update to a position statement of the American Diabetes Association and the European Association for the Study of Diabetes. Diabetes Care 38(1): 140­—149.  

2.      Doyle M, Egan JM (2007) Mechanisms of Action of GLP-1 in the Pancreas. Pharmacol Ther 113(3): 546—593. DOI:10.1016/j.pharmthera.2006.11.007

3.      Thornberry NA, Gallwitz B (2009) Mechanism of action of inhibitors of dipeptidyl-peptidase-4 (DPP-4). Best Pract Res Clin Endocrinol Metab 23(4): 479—486. DOI:10.1016/j.beem.2009.03.004

4.      Koehler JA, Baggio LL, Yusta B, et al (2015) GLP-1R agonists promote normal and neoplastic intestinal growth through mechanisms requiring Fgf7. Cell Metab 21(3): 379—391. DOI:10.1016/j.cmet.2015.02.005

5.      Gokhale M, Buse JB, Gray CL, et al (2014) Dipeptidyl peptidase-4 inhibitors and pancreatic cancer: a cohort study. Diabetes Obes Metab 16(12): 1247—1256. DOI:10.1111/dom.12379

6.      Green JB, Bethel MA, Armstrong PW, Buse JB, Engel SS, Garg J, Josse R, Kaufman KD, Koglin J, Korn S, Lachin JM, McGuire DK, Pencina MJ, Standl E, Stein PP, Suryawanshi S, Van de Werf F, Peterson ED, Holman RR, TECOS Study Group (2015) Effect of Sitagliptin on Cardiovascular Outcomes in Type 2 Diabetes. N Engl J Med 373(3): 232—242. DOI:10.1056/NEJMoa1501352

7.      Lund JL, Richardson DB, Stürmer T (2015) The active comparator, new user study design in pharmacoepidemiology: historical foundations and contemporary application. Current Epidemiology Reports 2:221–228. DOI 10.1007/s40471-015-0053-5

8.      Rothman KJ. Induction and latent periods (1981) Am J Epidemiol 114(2): 253—259.

9.      Rosenbaum PR, Rubin DB (1983) The central role of the propensity score in observational studies for causal effects. Biometrika 70(1): 41—55. DOI:10.1093/biomet/70.1.41

10.  Setoguchi S, Solomon DH, Glynn RJ, et al (2007) Agreement of diagnosis and its date for hematologic malignancies and solid tumors between Medicare claims and cancer registry data. Cancer Causes Control 18(5): 561—569. DOI:10.1007/s10552-007-0131-1



Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann