Evaluating SGLT2i Impact on AKI Risk

We performed propensity score matching to construct a comparative cohort for SGLT2i users from the study population. Propensity scores (PS) were calculated as a probability dependent on a vector of observed covariates associated with receipt of treatment with SGLT2i. We conducted a logistic regression analysis for estimating PS with adjustment for age, sex, comorbidities, and related clinical medications. A 1:1 PS-matched cohort was constructed using greedy nearest neighbor matching, and the caliper width was within 0.2.^{17 (link)} Moreover, standardized mean differences with a cutoff value of 0.10 were used to observe the fitness of covariate comparisons between the propensity score–matched groups.^{18 (link),19 (link)} Continuous variables are presented as mean and SD and categorical variables as numbers and frequencies. Age was stratified into 10-year groups of younger than 25 years, 25 to 34 years, 35 to 44 years, and so on. All participants were followed up from the index date until AKI diagnosis, death, or the study end date (December 31, 2018), whichever occurred first. Kaplan-Meier survival curves were expressed and compared for the risk of AKI or AKI-D incidence using log-rank tests. We used conditional Cox proportional hazard regressions to determine the crude and adjusted hazard ratios (HRs) and 95% CIs for risk of AKI or AKI-D after SGLT2i administration. SGLT2is, including dapagliflozin, empagliflozin, and canagliflozin, were evaluated separately for AKI or AKI-D risk using stratification analysis. We then added a multiplicative interaction term to the regression models to calculate the interactions between comorbidities and the use of SGLT2i on AKI risk. Finally, we analyzed the associations between SGLT2i use and concomitant diseases with AKI using conditional Cox proportional hazard regression analysis. The prognostic outcomes of AKI, including advanced CKD, ESKD, and death, were compared using χ² tests. All hypothesis tests were 2-sided. Significance was defined as α = 0.05. Statistical analysis was performed using the SAS statistical software version 9.4 (SAS Institute). Data analysis was conducted from October 15, 2021, to January 30, 2022.