The Ethics Committee of Jichi Medical University has approved this study and waived the requirement for informed consent owing to the retrospective design and use of anonymized data (approval number, 21–198). This study was performed in accordance with the relevant guidelines and regulations. The data used in this study were obtained from the JMDC Claims Database, which contains anonymized individual-level, de-identified data from multiple health insurance associations. All personal information was encrypted and anonymized to ensure patient confidentiality. The JMDC Claims Database is a commercially available database. Permission was obtained from the data provider (JMDC Inc.) to access and use the data. The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

We conducted a retrospective cohort study using data from the JMDC Claims Database. This database comprises inpatient, outpatient, and dispensing receipt data obtained from health insurance associations (employee health insurance systems in large companies in Japan) [15]. Therefore, the study predominantly included the working-age population (under 65 years old) and did not encompass self-employed individuals, pensioners, or those aged 75 years or older. As of February 2022, approximately 14 million persons were registered in this database. Each patient was assigned a unique ID, facilitating continuous patient follow-up even if they underwent transfers between hospitals [16]. The database encompasses patient demographics (unique ID, year and month of birth, and sex), diagnostic codes following the 10th edition of the International Classification of Diseases (ICD-10), medical procedures (tests and treatment), medication codes based on the Anatomical Therapeutic Chemical (ATC) classification, prescription durations, healthcare utilization details, and dates of diagnosis, procedures, prescriptions, and hospitalizations. The coding system employed for electronic claims aligns with the system developed by the Japanese Ministry of Health, Labor and Welfare (Japanese electronic claims codes). Most diagnostic and medication codes correspond to the ICD-10 and ATC classification systems, respectively.

We included all patients aged 18 years or older diagnosed with type 2 diabetes mellitus (identified by diagnostic codes E11 or E14) and prescribed new SGLT2 inhibitors or dipeptidyl peptidase 4 (DPP-4) inhibitors between April 1, 2014, and August 31, 2020. The index date was set as the day of the first prescription of an SGLT2 or DPP-4 inhibitor, following a 180-day look-back period. Exclusions encompassed patients with type 1 diabetes (E10), secondary diabetes (E12 and E13), and individuals undergoing hemodialysis. Additionally, those prescribed SGLT2 or DPP-4 inhibitors only once and individuals diagnosed with necrotizing fasciitis (M72.6) during the look-back period were excluded from the study. Individuals who were prescribed the comparator drugs during the look-back period were also excluded. SGLT2 and DPP-4 inhibitors were identified based on the ATC classifications A10BK and A10BH, respectively.

The outcome measures included perineal soft tissue, genital bacterial, and urinary tract infections. Individuals were categorized as having a perineal soft tissue infection if they were hospitalized and underwent debridement or related surgical procedures within 7 days following a diagnosis of necrotizing fasciitis, cellulitis, or a skin abscess in the perineal area. Genital bacterial infections and UTIs were defined as cases with the respective diagnostic codes and where any antibiotic was prescribed within 7 days of diagnosis. The diagnostic codes for these definitions, using the Japanese electronic claims codes, are provided in Supplementary Table S1.

The follow-up time was calculated from the index date until the occurrence of specific events: patients were censored if they discontinued therapy (with a treatment gap of > 30 days), added or switched to the comparator, reached the date of the last claim record, or reached the end of the study period (August 31, 2021). The treatment gap was defined as the duration between the end of one prescription period (prescription date plus the prescribed days) and the start of the subsequent prescription. If a patient was censored due to a treatment gap, the follow-up time was defined as the period from the index date to the last prescription date. Patients with a follow-up duration of 0 days were excluded from the study.

We employed propensity score (PS) models adjusted for 61 variables, including age, sex, comorbidities, co-medications, procedures, and healthcare utilization during the look-back period (Supplementary Table S2). The look-back period was defined as the 180 days preceding the index date, which was set as the day of the first prescription of an SGLT2 or DPP-4 inhibitor. Both one-to-one and nearest-neighbor matching without replacement were conducted with a caliper width set at 0.20 times the standard deviation of the logit of the PS. The absolute standardized mean difference (ASD) was calculated to assess the balance in variable distribution between the SGLT2 and DPP-4 inhibitor groups, with an ASD greater than 0.1 considered indicative of imbalance.

We conducted separate survival time analyses for three outcomes (UTIs, genital bacterial infections, and perineal soft tissue infections) using the Kaplan–Meier method and log-rank test. Additionally, Cox proportional hazard regression was used to calculate hazard ratios (HRs) for each outcome independently. The threshold for significance was set at P ≤ 0.05. Subgroup analyses for all outcomes were also conducted, considering sex, age group, comedication (biguanides, insulin, and sulfonylureas), and history of major cardiovascular disease. All analyses were performed using R version 3.5.3 (R Foundation, Vienna, Austria).