Evaluating Comparative Effectiveness of Empagliflozin in Type 2 Diabetes Population With and Without Chronic Kidney Disease

Incidence Rate of the 40% Decline in Estimated Glomerular Filtration Rate (eGFR)

40% decline in eGFR: at least 2 measurements (second measurement must be within 2 years) during follow-up of at least a 40% decline relative to baseline separated by \>= 28 days. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of End-stage Kidney Disease (ESKD)

ESKD definition: at least 1 kidney transplant or ESKD diagnosis/procedure or at least 2 dialysis diagnoses/procedures separated by \>= 28 days or eGFR\<15 on 2 measurements separated by \>= 28 days. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Dialysis

Incident dialysis, given dialysis in the 12 months preceding index date is a disqualifying diagnosis/procedure. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Kidney Transplant

Kidney transplant: any procedure associated with healthcare encounters, including hospitalizations and specialist outpatient and primary care encounters. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Composite Outcome Including Acute Hospitalization for Heart Failure and All-cause Death

Composite outcome including acute hospitalization for heart failure and All-cause death: Any diagnosis of heart failure associated with hospital admission, including inpatient and emergency department (ED) to inpatient encounters. From death records provided by sites, supplemented with linkage using Datavant Software. Post-LASSO overlap weighting was used. A LASSO penalized regression model was created with covariates of interest, subgroup variables, and all pairwise interactions. Variables chosen by the LASSO model were refit to a logistic model in order to calculate the PS estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Acute Hospitalization for Heart Failure

Acute hospitalization for heart failure: Any diagnosis of heart failure associated with hospital admission, including inpatient and emergency department (ED) to inpatient encounters. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of All-cause Death

All-cause death, from death records provided by sites, supplemented with linkage using Datavant Software. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. a LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of the Composite Outcome Including MI, Stroke, All-cause Death and Coronary Revascularization Procedure

Composite outcome including MI, Stroke, All-cause death and Coronary revascularization procedure: For MI and stroke, any inpatient diagnosis associated with healthcare encounters, including inpatient and ED to inpatient For all-cause death, death records provided by sites, supplemented with linkage using Datavant Software. Post-LASSO overlap weighting was used. A LASSO penalized regression model was created with covariates of interest, subgroup variables, and all pairwise interactions. Variables chosen by the LASSO model were refit to a logistic model in order to calculate the PS estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of the Composite Outcome Including MI, Stroke, All-cause Death (MACE)

Composite outcome including MI, Stroke, All-cause death: For MI and stroke, any inpatient diagnosis associated with healthcare encounters, including inpatient and ED to inpatient For all-cause death, death records provided by sites, supplemented with linkage using Datavant Software. Post-LASSO overlap weighting was used. A LASSO penalized regression model was created with covariates of interest, subgroup variables, and all pairwise interactions. Variables chosen by the LASSO model were refit to a logistic model in order to calculate the PS estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Diabetic Ketoacidosis

Any diabetic ketoacidosis diagnosis associated with healthcare encounters in the inpatient setting. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Severe Hypoglycemia

Any severe hypoglycemia diagnosis associated with healthcare encounters in the inpatient or ED setting. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Urinary Tract Cancer

Two or more urinary tract cancer diagnoses associated with healthcare encounters within 2 months. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Severe Urinary Tract Infections (UTI)

Any severe Urinary Tract Infections (UTI) diagnosis associated with healthcare encounters in the inpatient or ED setting for Pyelonephritis or Urosepsis. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Acute Kidney Injury (AKI) That Requires Dialysis

Any Acute kidney injury diagnosis associated with inpatient healthcare encounters for AKI plus at least one inpatient encounter indicating dialysis within 28 days of the AKI encounter. Two or more dialysis encounters separated by 28 days or more was excluded from this definition. Post-LASSO overlap weighting was used. A LASSO penalized regression model was created with covariates of interest, subgroup variables, and all pairwise interactions. Variables chosen by the LASSO model were refit to a logistic model in order to calculate the PS estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.

Incidence Rate of Genital Mycotic Infection

Any genital mycotic infection diagnosis associated with healthcare encounters, including hospitalizations and outpatient encounters, or prescription for fluconazole. Post-LASSO (least absolute shrinkage and selection operator) overlap weighting was used. A LASSO penalized regression model was created with the covariates of interest, subgroup variables, and all pairwise covariate-subgroup interactions. Outcome for this model was treatment group (DPP4i as the reference). The variables chosen by the LASSO model were refit to a logistic model in order to calculate the propensity score (PS) estimates. The overlap weights were then created such that the weights were equal to the PS for participants prescribed the reference treatment (DPP4i) and 1-PS for participants prescribed empagliflozin. Results are presented for the overall cohort and separately for the CKD and non-CKD cohorts.

Time frame: Starting the day after the index date (date of initiation of empagliflozin or DPP4i) and continuing until the first occurrence of the outcome of interest, or end of study date (date of death, date of study end, 2 years after index date). Up to 2 years.