(183) Feasibility of Constructing an External Control Arm with Real-World Data to Augment Subgroups in Cardiorenal Trials: A Matched Cohort Study Focusing on Trial Subgroups of Sodium-Glucose Cotransport 2 Inhibitor Users at Baseline

Background: Finerenone demonstrated a risk reduction of kidney and cardiovascular (CV) events in patients with chronic kidney disease (CKD) and type 2 diabetes in the FIDELIO-DKD and FIGARO-DKD phase 3 randomized controlled trials (RCTs). Pooled subgroup analyses reported consistent cardiorenal benefits of finerenone, independent of sodium-glucose cotransport 2 inhibitor (SGLT2i) co-therapy. However, subgroup analyses in clinical trials often lack power to detect significant treatment effects.

Objectives: We aimed to evaluate the feasibility of building an external control arm (ECA) by matching patients from real-world data (RWD), to augment the pooled SGLT2i subgroups of FIDELIO-DKD and FIGARO-DKD.

Methods: We adapted the RCT selection criteria to identify eligible ECA patients in US Optum electronic health records data from 01/2013 to 09/2021. Patients had at least one qualifying measurement of estimated glomerular filtration rate (25-90 ml/min/1.73m2) and urine albumin-creatinine ratio (≥30 mg/g) separated by 90 to 540 days (second qualifying value = index date), and an SGLT2i prescription during the 365-day baseline period. Using a mixed linear integer programming approach with a set of 42 medically informed baseline covariates (i.e., demographics, comorbidities, comedication, labs), we matched patients from the eligible ECA cohort to those in the pooled RCT subgroup. Standardized mean differences (SMD) assessed matching quality; summary statistics of baseline characteristics and incidence rates of cardiorenal outcomes were compared between the ECA and the RCT controls (n=439) to determine data homogeneity.

Results: From 8,272 eligible ECA patients, we successfully matched 877 patients to the pooled RCT subgroup, leading to a 1:2 ratio of internal vs external controls. Median SMD was 0.000 (IQR 0.000, 0.004) across matching parameters, indicating good model performance, and achieving balance of baseline characteristics across groups. Furthermore, incidence rates of study outcomes overlapped between the ECA and pooled RCT controls, with 4.12 events (95% CI, 3.35-5.02) per 100 patient years vs 4.08 (3.05-5.35) for the CV composite endpoint, 2.15 (1.61-2.81) vs 1.68 (1.05-2.55) for hospitalization due to heart failure, 2.06 (1.54-2.70) vs 2.23 (1.50-3.18) for all-cause mortality, and 1.04 (0.68-1.53) vs 1.37 (0.80-2.19) for the kidney composite endpoint.

Conclusions: High alignment in baseline characteristics and outcome incidences are suggestive of a good model fit and indicate adequate homogeneity in RCT controls vs the matched ECA. Our results demonstrate the feasibility of creating an ECA in a large indication such as CKD, opening the possibility to evaluate clinical outcomes with an ECA-augmented subgroup in the future.