A Retrospective Cohort Study of Acute Pancreatitis in Relation to Use of Exenatide and Other Antidiabetic Agents

Limitations of the study: The results provided below should be interpreted in light of the following limitations:

• Misclassification of acute pancreatitis may have distorted our estimates of absolute and relative IRs. In cases where the degree of misclassification is non-differential with respect to the exposure cohort, as is likely the case in administrative data, the RR would be biased toward the null value, although the magnitude of bias will depend on the amount of misclassification.
• Lack of information on important potential confounders, like obesity and alcohol use, is another limitation of the present analysis. Although we adjusted for propensity scores of exenatide initiation, which included a large number of factors derived from the claims data, it is likely that the present estimates are somewhat inaccurate due to residual confounding.
• Our definition of current use in the time-on-drug analysis, which extended 31 days past the nominal end of the last dispensing of the cohort-defining drug, may be too long in duration and thus misclassify exposure during the relevant etiologic period.
• Additionally, these analyses assume that when pharmacies submit a claim for a medication that patients receive and consume the medication. While it is possible that misclassification of exposure by non-adherence to the medications dispensed occurred, prior work showed that pharmacy claims are valid for ascertaining medication exposure.

Inclusion criteria

• Exenatide Initiators: The date of the first dispensing of exenatide under which a potential cohort member might qualify will be 1 June 2005, and the latest date will be 31 December 2007. We will note the first initiation of exenatide, and also note subsequent initiation of other antidiabetic drugs. Eligible exenatide initiators will be included in the exenatide initiator cohort on the first eligible dispensing following at least 9 months of continuous health plan enrollment.
• Other Antidiabetic Drug Initiators: The date of the first dispensing of an antidiabetic drug other than exenatide under which a potential cohort member might qualify will be 1 June 2005, and the latest date will be 31 December 2007. We will note the earliest date of other antidiabetic drug dispensings within their cohort membership, and also note subsequent initiation of other antidiabetic drugs. We will choose a subset of this comparator cohort randomly selected to be approximately 9 times larger than the exenatide initiators and will oversample patients receiving certain drugs to the extent necessary to ensure inclusion of at least as many persons initiating these drugs as initiating exenatide. Specifically, oversampling will be performed as needed on initiators of metformin, TZDs, SUs, sitagliptin and insulin glargine to allow for further sub-analysis.
• Exenatide and Other Antidiabetic Drug Initiators: For all patients in these two study cohorts, the date of cohort entry (the date ranged between January 1, 2005 and December 31, 2007) marked the beginning of observation for study outcomes (follow-up). The end of observation for a given patient happened on the earliest of occurrence of likely acute pancreatitis, end of the study period (March 31, 2008), or disenrollment from the health plan.
• Non-Diabetes Cohort: A third cohort will consist of randomly-selected Ingenix Research Data Mart members who have no claim associated with a diabetes diagnosis or antidiabetic drug dispensing in the baseline continuous enrollment period. This cohort will enter as of the later of 1 June 2005 or completion of 9 months continuous baseline enrollment and will provide follow-up through end of enrollment or 31 March 2008, or the receipt of an antidiabetic drug dispensing or diabetes diagnosis, at which point they may enter one of the other exposure cohorts.

Exclusion criteria

• The 3 cohorts (exenatide initiators, other antidiabetic initiators, and those without diabetes) will be subject to minimal exclusions in order to observe acute pancreatitis across a wide spectrum of patient characteristics. We will apply a baseline enrollment requirement of 9 months prior to cohort entry so that the first day of follow-up (on which acute pancreatitis could occur) will be characterized with the same level of detail (based on 9 months of preceding health insurance claims) as any other day during the study.
• Consistent with the principle of minimal exclusions, we will only exclude from the cohorts people who have baseline claims associated with pancreatitis (in the 9-month period preceding cohort entry) as these people either had pre-existing pancreatitis or had a pre-existing suspicion of pancreatitis. We will not exclude persons with a type I diabetes diagnosis as we intend to observe the spectrum of clinical practice with respect to antidiabetic drug initiation.