Fig 5. Primary and secondary endpoints, according to study cohorts.

Hazard Ratio calculated with confounder adjustment and inverse probability weighting (panel a). Hazard Ratio calculated without any adjustment nor weighting (panel b). Shown is the primary endpoint of the study -estimated glomerular filtration rate (estimated by the CKD EPI formula) of more than 30% from the baseline- and the secondary endpoint of the study -first observation of the albuminuria (defined by the urine albumin creatinine ratio greater or equal to 30mg/mmol)- according to the study cohorts. High bilirubin denotes the patient subpopulation that has above or equal cohort median of the baseline serum bilirubin concentration (9μmol/L for SAVOR, 10μnil/L for CPRD-DM2 and HT), Low bilirubin denotes the patient subpopulation that has below cohort median of the baseline serum bilirubin concentration, HR hazard ratios in panel a, inverse probability weighted and adjusted for confounding covariates (age, sex, race if available, baseline body mass index, baseline hemoglobin, baseline alanine transaminase, baseline aspartate transaminase, smoking), HR hazard ratios in panel b, no inverse probability weighted and no adjustment, CI confidence interval calculated by fitting a normal distribution to the 1000 bootstrap sample of the HR.