Should we be reassured about sitagliptin?

BMJ 2013; 346 doi: http://dx.doi.org/10.1136/bmj.f3051 (Published 14 May 2013) Cite this as: BMJ 2013;346:f3051
  1. Jodi Segal, associate professor of medicine
  1. 1Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA
  1. jsegal{at}jhsph.edu

Remember that the absence of proof of harm is not proof of absence of harm

New drugs for the treatment of diabetes have had to fight hard to prove their safety. The latest class of drugs causing concerns is the dipeptidyl peptidase-4 (DPP-4) inhibitors. Current evidence suggests that these drugs increase the risk of pancreatitis and they are undergoing close study for evidence of sufficient safety to warrant their continued use.1 In a linked paper (doi:10.1136/bmj.f2267), Eurich and colleagues contribute to the growing evidence base regarding sitagliptin and related drugs and provide some evidence of their safety.2

Sitagliptin was approved by the US Food and Drug Administration (FDA) in 2006 as the first in a new class of antidiabetic drugs, the DPP-4 inhibitors. It was approved for use as monotherapy or combination therapy, with diet and exercise, for treatment of type 2 diabetes. During the first three years after approval, the FDA received 88 case reports of acute pancreatitis. This prompted a revision of the prescribing information in February 2009 to include recommendations that prescribers monitor patients for signs and symptoms of pancreatitis at initiation of this drug and at dosage changes. Other labeling revisions followed, including warnings about hypersensitivity reactions and acute renal failure. In March this year the FDA announced that it is actively evaluating the risk of pancreatitis and pancreatic duct metaplasia associated with the DPP-4 inhibitors and with glucagonlike peptide-1 (GLP-1) agonists.3

Eurich and colleagues conducted a retrospective cohort study of patients receiving glucose lowering drugs, using an integrated claims and laboratory database. They found no increased risk of hospital admission, all cause mortality, or death from cardiovascular disease in patients newly exposed to sitagliptin relative to other glucose lowering drugs. They rigorously controlled for the confounding that would be introduced by naively comparing people exposed to sitagliptin with those who were not exposed. The authors had hypothesized that users of sitagliptin would not have more hospital admissions, all cause mortality, or cardiovascular events than those not using sitagliptin. These endpoints were chosen because of their importance to patients and clinicians, and not because signals to date suggest cardiovascular risk. On the contrary, the authors noted that pooled analyses of clinical trials had suggested cardiovascular benefits with DPP-4 inhibitors.4

The authors made many smart analytic choices, including the evaluation of incident users of the drug, careful control for confounding by indication, and ample sensitivity analyses where they varied the choices made in the primary analyses. However, it may be hard to interpret these results in light of their hypothesis of no increased risk. The absence of proof of harm is not proof of absence of harm.5 The authors do not describe the study’s adequacy to demonstrate small increases or decreases in mortality, hospital admissions, or cardiovascular events associated with the drug. As with equivalency trials, a retrospective cohort study that is designed to show an absence of difference between exposed and unexposed people should state what minimally important difference would be detectable in the cohort and what difference may be clinically meaningful.6

Why did Eurich and colleagues not see an increased risk of pancreatitis? The absence of a signal for what is now widely recognized as an outcome that can be attributed to this drug is surprising.7 They report an adjusted hazard ratio of 1.10 (95% confidence interval 0.68 to 1.77) for acute pancreatitis. The absence, in this study, of an association between sitagliptin and acute pancreatitis can be interpreted in two ways: either the study adequately controlled for the confounding that has plagued other studies or the study was underpowered for this outcome and possibly the primary outcomes of interest too.

Eurich and colleagues’ cohort included more than 8000 people exposed to sitagliptin. The mean exposure time to an antidiabetic drug was 2.5 years. This is longer than the follow-up times in the clinical trials that led to the approval of such drugs that were subsequently included in meta-analyses, but it is still not a long time in the life of patients with diabetes. However, for events that are not expected to be associated with the drug (death from cardiovascular disease and all cause mortality), the number of exposed people may have been too few to detect a significant increase in risk. In meta-analyses to date, data were included from 20 312 people exposed to DPP4-inhibitors (not just sitagliptin).

We can certainly admire and learn from the careful methods used in this study, but we should remember that additional considerations may be needed upfront in trials that pose a null question. Perhaps observational studies should be designed like non-inferiority trials, with prespecified minimal detectable differences between groups. Although this study contributes to the evidence base about sitagliptin, it is unlikely to change practice, the current investigation of the safety of DPP4-inhibitors, or regulations.


Cite this as: BMJ 2013;346:f3051


  • Research, doi:10.1136/bmj.f2267
  • Competing interests: I/we have read and understood the BMJ Group policy on declaration of interests and declare the following interests: None.

  • Provenance and peer review: Commissioned; not externally peer reviewed.