Fig. 2

Results of subdistribution hazard model analyses for GLP-1RA versus LAI therapies on study outcomes (primary and sensitivity analyses). Black vertical dotted lines are 95% CIs of study outcomes obtained from primary analyses in propensity-score-matched cohorts. SDHRs subdistribution hazard ratios, PSM propensity score matching, IPTW inverse probability of treatment weighting, SMRW Standardized mortality ratio weighting, hdPS high-dimensional propensity score, ESRD end-stage renal disease. ^*Composite renal outcome includes stable use of ESA, dialysis-dependent ESRD, and renal death. ^†Renal insufficiency referred to estimated glomerular filtration rate < 15 mL/min/1.73 m² and was determined by the stable use of erythropoiesis stimulating agents (ESAs) (i.e., at least two prescriptions of darbepoetin alfa or methoxy polyethylene glycol-epoetin beta, or four prescriptions of erythropoietin within three months), given that the reimbursement policy of Taiwan’s National Health Insurance program restricts the use of ESAs only to patients with stage 5 chronic kidney disease. This operational definition was also confirmed with clinical nephrologists. ^‡We removed patients whose propensity scores were either more than 0.95 or less than 0.05; the weights were then estimated based on the trimmed populations. Weights in the IPTW approach were estimated as follows: Weight_GLP1RAs = 1/PS and Weight_LAIs = 1/(1-PS). Weights in the stabilized IPTW approach were estimated as follows: Weight_GLP1RAs = Prevalence of GLP1RA users (%)/PS and Weight_LAIs = Prevalence of LAI users (%)/(1-PS). Weights in the SMRW approach were estimated as follows: Weight_GLP1RAs = 1 and Weight_LAIs = PS/(1-PS)