This study explores the changes in glucose-lowering drug (GLD) use before and after cancer diagnosis among patients with diabetes.

New GLD users (1998–2011) living in the Dutch ECR–PHARMO catchment area were selected from the PHARMO Database Network (n=52,228). Those with a primary cancer diagnosis were considered cases (n=3281) and matched with eligible controls (n=12,891) without cancer during follow-up. Conditional logistic regression analysis was used to assess changes in GLD use, such as treatment add-ons, treatments drops and initiation of insulin, for cases compared with controls associated with specific cancer types in four time windows (6–3 and 0–3months before cancer diagnosis; 0–3 and 3–6months after cancer diagnosis).

In the 3months before cancer diagnosis, patients with upper gastrointestinal (GI) cancers (oesophageal, stomach, pancreatic, liver cancers) had higher odds of initiating insulin (OR: 9.3; 95% CI: 3.6–24.1); to a lesser extent, this was also observed in the 3months prior to that (at 6months, OR: 3.9; 95% CI: 1.3–12.1). Diagnosis of colorectal (OR: 3.4; 95% CI: 1.4–8.4), pulmonary (OR: 2.5; 95% CI: 1.1–5.4) and upper GI (OR: 13.6; 95% CI: 5.0–36.9) cancers was associated with increased odds of initiating insulin in the 3months after cancer diagnosis. During all study time windows, the odds of treatment drops were higher for patients with upper GI cancers whereas, for most other cancers, these odds were higher only after a diagnosis of cancer.

The greater odds of initiating insulin during the 6months prior to diagnosis of upper GI cancers suggest reverse causation. After cancer diagnosis, drops in use of GLDs was commonly seen.