(206) Trials Replication through Observational study across OHDSI (TROY)

Background: Randomized controlled trials (RCTs) are considered the gold standard for evaluating medical products. In certain trials, researchers enhance internal validity by setting overly restrictive selection criteria, aiming to showcase intervention effectiveness. Yet, this may not reflect the intervention's effect on all patients adhering to approved indications (i.e., external validity).

Objectives: We selected thirteen pivotal trials (type 2 diabetes, atrial arrhythmias, and acute coronary syndromes) and evaluated the generalizability of clinical practice by closely replicating the populations included in each trial.

Methods: This study aims to replicate settings as close as possible to the eligibility criteria of pivotal trials in clinical practice, using two Korean electronic medical record databases standardized in the Observational Medical Outcome Partners-Common Data Model (OMOP-CDM) format. A new-user cohort (target-comparator) design was performed for each treatment in each pivotal clinical trial. Replication was conducted from the date the target drug of each trial received FDA approval through 2023. In clinical practice settings, the proportion of patients meeting inclusion and exclusion criteria among those meeting FDA-approved indications is calculated to identify patients believed to be eligible for the treatment effect shown in each clinical trial.

Results: As a result of database replication of the selected RCTs, the observed proportion of the group satisfying the eligible criteria and the indication group approved by the FDA is as follows: LEADER (liraglutide): 5,516/48,015 (11%), DECLARE-TIMI 58 (dapagliflozin): 647/26,135 (2%), EMPA-REG OUTCOME (empagliflozin): 5,761/44,025 (13%), CANVAS (canagliflozin): 4,053/44,315 (9%), CARMELINA (linagliptin): 2,066/38,983 (5%), TECOS (sitagliptin): 1,244/45,264 (3%), SAVOR-TIMI 53 (saxagliptin): 6,586/33,525 (20%), CAROLINA (linagliptin): 0/32,916 (NA), TRITON-TIMI 38 (prasugrel): 1,120/10,048 (11%), PLATO (ticagrelor): 9,221/10,664 (86%), ROCKET AF (rivaroxaban): 32/6,972 (0%), ARISTOTLE (apixaban): 2,966/7,017 (42%), ENGAGE AF-TIMI 48 (edoxaban): 861/4,384 (20%).

Conclusions: The proportion of patients meeting eligibility compared to indications in clinical practice changes depending on the stringency of clinical trial eligibility conditions, resulting in the need for more intensive external verification. As confoundings may occur due to inaccuracy or chance during replication using the database, we will introduce a more systematic evaluation and recruit multiple OMOP-CDM data partners.