(C-144) Prevalent New User Design: Outcome Heterogeneity Based on Different Lengths of Time-Based Exposure Sets

Background: The Prevalent New User (PNU) design evaluates new drug effects when no contemporaneous comparator exists, using time- or prescription-based exposure sets. However, the impact of different time-based exposure set lengths on outcome estimates remains unclear.

Objectives: To assess outcome heterogeneity across varying time-based exposure set lengths using a case study.

Methods: A retrospective cohort study using data from TriNetX Global Collaborative Network compared the risk of major adverse cardiovascular events (MACE) between sodium-glucose co-transporter 2 inhibitors (SGLT2i) and dipeptidyl peptidase-4 inhibitors (DPP4i) under PNU design. Patients treated with DPP4i (2013–2018) were included, excluding those with prior SGLT2i use before DPP4i initiation. Patients who received SGLT2i were classified as SGLT2i users, and were stratified into six groups based on the time since DPP4i initiation: 1–6, 7–12, 13–18, 19–24, 25–30, and 31–36 months. Two time-based exposure set definitions were applied. The first approach matched SGLT2i users who switched from DPP4i, to patients who had been using DPP4i within the same time interval. The second approach additionally required that both SGLT2i and matched DPP4i users had at least one prescription of DPP4i in each preceding interval. Outcomes, including heart failure, stroke, myocardial infarction, atrial fibrillation, and all-cause mortality, were analyzed by Cox proportional hazards models, using DPP4i as the reference group. Results across exposure set lengths were pooled separately for each approach using meta-analysis, with heterogeneity assessed via I² statistics.

Results: In the first approach, pooled hazard ratios (HRs) showed a significantly lower risk of heart failure (HR=0.85; 95% CI: 0.78–0.93) and all-cause mortality (HR=0.76; 95% CI: 0.68–0.84) in SGLT2i users, with no significant differences for stroke, atrial fibrillation, or myocardial infarction. Moderate heterogeneity was found for atrial fibrillation (I²=38%), while other outcomes had low heterogeneity (I² < 25%). The second approach also showed a lower risk of heart failure (HR=0.88; 95% CI: 0.79–0.99) and all-cause mortality (HR=0.71; 95% CI: 0.59–0.85) in SGLT2i users, with no significant differences for other outcomes. Heterogeneity was moderate to high for stroke (I²=50%), all-cause mortality (I²=39%), and atrial fibrillation (I²=22%), but low for others (I² < 20%).

Conclusions: Different time-based exposure set lengths introduced heterogeneity in outcome estimates of PNU design. Despite this, overall pooled HRs remained consistent for both approaches. Future PNU studies might consider incorporating sensitivity analyses using different lengths of time-based exposure sets to enhance the robustness of findings.