We conducted a disproportionality analysis on the WHO global database, Vigibase. Vigibase contains over 36 million individual case safety reports (ICSRs) as of February 2024, submitted to the WHO Uppsala Monitoring Center by national pharmacovigilance systems from over 150 countries worldwide since 1967. These ICSRs are spontaneous or solicited accounts of adverse drug reactions describing a suspected causative medicinal product. The case/non-case design of the disproportionality analysis is related to case/control design, but controls are reports from the database for another adverse event.

Data were extracted on 13/02/2024. We extracted all reports with a SGLT-2i (ATC A10BK) or GLP-1 analogues (ATC A10BJ) and DPP-4 inhibitors (ATC A10BH) as suspected drugs between January 1, 2014 (corresponding to the year of approval of the two main SGLT-2i agents, namely empagliflozin and dapagliflozin), and December 31, 2023 and where subjects were aged > 18 years and had age and sex information available. We excluded reports for thyroid and pancreatic cancers from the analysis due to a risk of competition bias with our comparator group [10]. Cases of thyroid and pancreatic cancer were excluded based on preferred terms (PTs) recorded in the SMQ "Malignant tumor" containing the roots "thyroïd" or "pancrea.", using the French version of MedDRA.

Cases were defined as all cases corresponding to the Preferred Terms of the Medical Dictionary for Regulatory Activities (MedDRA®) under the Standardized MedDRA Query "Malignant tumor." Non-cases were reports registered with any other terms. Exposure was defined as SGLT-2i use (ATC A10BK), registered in the database as suspect drugs. We chose other diabetes drugs as the non-exposed group since diabetic individuals are at higher risk of cancer. Specifically, we chose GLP-1 analogues (ATC A10BJ) and DPP-4 inhibitors (ATC A10BH) due to their positioning as second-line treatment for diabetes, like SGLT-2i.

Patient characteristics were described in frequencies and percentages. We performed a disproportionality analysis by calculating the reporting odds ratio (ROR), which is calculated in the same way as an OR. Analyses were adjusted for age (in quartiles), sex, reporting country (USA vs others), co-prescriptions with carcinogenic risk, defined from the list of the International Agency for Research on Cancer [11]. We conducted several sensitivity analyses, the of which was restricted to the period 2016–2023 to limit the temporal effect, also call the Weber effect [12] of newly marketed drugs. Indeed, SGLT2 inhibitors were marketed in 2014, and the Weber effect is characterized by an increase in reports at the time of a drug’s launch due to the rise in exposure to the new drug and its poorly understood safety profile. Therefore, we excluded the first two years, as the Weber effect is generally most significant during the first two years after the introduction of a drug to the market. Other sensitivity analyses were done by restricting to reports from the USA and to reports from physicians or pharmacists. We also conducted several subgroup analyses: by cancer type, by drug (dapagliflozin, empagliflozin or canagliflozin), by sex and age categories (aged < 60 years or ≥ 60 years), and by year of inclusion in the database. The signal detection threshold was a lower bound of the 95% CI of ROR > 1 and in a number of cases ≥ 3, in accordance with the definition used by the European Medicine Agency [13]. All statistical analyses were conducted using SAS software (SAS Institute Inc., Cary, NC, USA). Graphical representations were created using R (version 4.2.2).

Patients’ informed consent was not necessary since data from Vigibase were deidentified. The study protocol was registered in the EU-PAS registry (EUPAS1000000311). This study was conducted in accordance with the READUS-PV guidelines [14].