Design, setting, and data sources

Real-world, active comparator, new user retrospective cohort study combining several population-wide databases from the Valencia Health System Integrated Database (VID). VID is a set of multiple, public, population-wide electronic databases for the Valencia Region, the fourth most populated Spanish region, with approximately 5 million inhabitants. VID provides exhaustive longitudinal information including sociodemographic and administrative data, clinical, pharmaceutical prescription and dispensing linked at the individual level (including brand and generic name, formulation, strength and dosing schedule/regimen), and healthcare utilization data from public hospital care, emergency departments, specialized care, primary care and other public health services.⁶

We included new users of GLP-1 RA agonists and an active comparator, sodium-glucose cotransporter 2 inhibitors (SGLT-2i), in the period Jan 1, 2015 to December 31, 2021 (see ^{Table S1 in online Supplementary Material} for Anatomical Therapeutic Chemical Classification of drugs included). Patients receiving a prescription for either GLP-1 RA or SGLT-2i medication and not having received a prescription for an antidiabetic drug (except for metformin) in the previous 365 days were considered new users. Considering that SGLT-2i and GLP-1RA are recommended as therapeutic alternatives by many clinical guidelines, usually as second-line treatments after first-line metformin or other non-insulin treatment, new users of SGLT-2i appear to be a suitable active comparator group. The index date was the day of first prescription of a study drug. In Spain GLP-1RA for diabetes mellitus are only authorised for individuals with obesity. Accordingly, we only included new SGLT-2i or GLP-1RA users with body mass index (BMI) ≥30 when initiating therapy (see operational definition in Table 1, footnote), in order to produce groups as similar as possible at baseline. Patients in both groups were required to use metformin in the index date (defined by having a prescription for metformin issued in the 60 days before the index date).

Variables

We included the following patient sociodemographic, lifestyle and clinical variables at the individual-level: age, sex assigned at birth, income level, comorbidities at baseline (heart failure, dementia, hypertension, liver disease, renal disease, coronary heart disease, congestive obstructive pulmonary disease, malignancies, depression, sleep disorders, anxiety, substance abuse, personality disorders, psychotic disorders, other psychiatric disorders), BMI strata at baseline, previous acute pancreatitis and previous biliary disease (see Table 1 and ^{Table S2 in online Supplementary Material} for ICD codes employed).

Outcomes

Primary outcomes were acute pancreatitis (ICD code: 577; ICD-10 codes: K85, K86) and biliary disease, including cholelithiasis (ICD-9 codes; 574-575-576; ICD-10 codes: K80-K81-K82-K83). Outcomes were retrieved from different real-world databases in VID (primary care electronic records, in-hospital and outpatient hospital records and emergency room records), via ICD-10 codes registered during healthcare encounters. In the main per-protocol analyses, participants were followed from the index date to first mutually exclusive incidence of the outcome of interest, discontinuation of study drug (defined as more than 60 days without drug supply; days of drug supply was estimated using dosing information and dispensation data), initiation of other study drug, or switch to other glucose-lowering drug (defined as receiving a prescription for any glucose-lowering drug other than the baseline treatment or metformin and not refilling the assigned drug for 60 days), death, or end of follow-up. In secondary, intention-to-treat analyses, participants were followed from the index date to first event, death, or end of follow-up (Dec 31, 2021), and were analysed according to the treatment assigned at baseline, irrespective of actual exposure during follow-up.

Analysis

First, we described patient characteristics of GLP-1RA and SGLT-2i users included in the main analysis of acute pancreatitis and of biliary disease. For each analysis, new users that had no previous event of interest registered in VID before the index date were included. Second, as main analyses, we used a per-protocol approach to estimate the risk of pancreatitis and biliary disease in GLP-1RA vs. SGLT-2i. We carried out multivariable Cox regression modelling using our set of covariates. To adjust for potential confounding, we used inverse probability of treatment weighting based on propensity scores (PS). PS for each outcome were calculated based on the probability of initiating treatment with GLP-1RA taking into account the same covariates, to generate patient-specific stabilised weights. Covariate balance between the weighted exposure cohorts was assessed using standardised mean differences, with standardised differences <0.10 suggesting adequate balance.⁷ As secondary analyses, we provided estimates for patients with a previous outcome of interest before the index date, as well as intention-to-treat analyses. Statistical significance was defined as 0.05. Analyses were performed using STATA version 14 and R version 3.6.0. Ethics approval was obtained from the Valencia Hospital Clinic Universitari's clinical research ethics board (2022/163) with a waiver of informed consent.