Incretin-based Drugs and Pancreatic Cancer

The study objective is to determine whether the use of incretin-based drugs, compared with the use of sulfonylureas, is associated with an increased risk of pancreatic cancer (PC) in routine clinical practice. The investigators will use a common-protocol approach to conduct retrospective cohort studies using administrative health care data from five jurisdictions (the Canadian provinces of Alberta, Manitoba, and Ontario, as well as United States (US) MarketScan, and the United Kingdom (UK) Clinical Practice Research Datalink [CPRD]). Briefly, the Canadian databases include population-level data on physician billing, diagnoses and procedures from hospital discharge abstracts, and dispensations for prescription drugs. Ontario data will be restricted to patients aged 65 years and older as prescription data are not available for younger patients. The CPRD is a clinical database that is representative of the UK population and contains the records for patients seen at over 680 general practitioner practices in the UK; these data will be linked to the Hospital Episode Statistics (HES) database, which contains in-hospital diagnosis and procedure data. US MarketScan includes individuals and their dependents covered by large U.S. employer health insurance plans, and government and public organizations.

Study population

In each jurisdiction, the investigators will assemble a base cohort that includes all patients with a first-ever prescription for a non-insulin anti-diabetic drug, including biguanides, sulfonylureas, thiazolidinediones, DPP-4 inhibitors, GLP-1 analogs, alpha-glucosidase inhibitors, meglitinides, or combinations of these drugs from the earliest availability of data at each site to the last date of availability of data. The date of prescription (for the CPRD) or dispensation (for all other sites) of the first-ever non-insulin anti-diabetic drug will define the date of base cohort entry. From this base cohort, a study cohort will be created including all patients who initiated a new anti-diabetic drug class during the year in which incretin-based drugs entered the market in each jurisdiction or any time thereafter. These new users consist of both those who are newly-treated for diabetes, as well as those who switch to or add on a new anti-diabetic drug class not included as part of their previous treatment history. The date of study cohort entry is defined by the prescription date of the newly-prescribed drug class. Patients in the study cohort will be followed from the date of study cohort entry + 365 days until an event (defined below) or censoring due to death, departure from the database, loss of continuous health plan or drug plan enrolment, entry into a long-term care facility, or the end of the study period (June 30, 2014 or the last date of data availability at that site), whichever occurs first.

Case-control selection

The cohort defined above will be analyzed using a nested case-control analysis, where cases are defined as a hospitalization for PC. Risk set sampling will be used to randomly select up to 20 controls for each case, matched on sex, age (± 365 days), date of study cohort entry (± 180 days), duration of treated diabetes (± 90 days), and duration of follow-up in days.

Exposure assessment

Ever-use of an anti-diabetic drug will be defined as any prescription for an anti-diabetic agent between base cohort entry and the index day -365 days. This 365-day lag period will be applied to account for disease latency and potential protopathic bias by only considering prescriptions received from (and including) the Base Cohort Entry Date until (and including) the date one year prior to the Index day. Ever-use of exposure will be classified hierarchically based on the following three mutually-exclusive categories: 1) incretin-based drugs; 2) sulfonylureas; 3) other antidiabetic agents. Sulfonylureas will serve as the primary reference category as incretin-based drugs are second- to third-line therapy and thus used at a comparable point in the disease management.

Statistical analyses

Conditional logistic regression will be used to estimate odds ratios (ORs) and corresponding 95% confidence intervals (CIs) of the association of hospitalization for PC, comparing ever-use of incretin-based drugs to ever-use of sulfonylureas. This is considered the primary analysis. Secondary analyses will include sub-classifying ever-users of incretin-based drugs by type (i.e., DPP-4 inhibitor vs GLP-1 analog), cumulative duration of ever-use (≤ 365 days, 366-729 days, and ≥730 days), and time since initiation of treatment among ever-users. In addition, four sensitivity analyses will be conducted, all defined a priori, to assess the robustness of the results. Finally, all site-specific estimates will be meta-analyzed using random-effects models with inverse variance weighting, with fixed-effects analyses conducted as sensitivity analyses. The amount of between-site heterogeneity will be estimated using the I square statistic.