(051) Dipeptidyl Peptidase-4 Inhibitors and Inflammatory Bowel Disease Risk: Impact of Study Design Differences on Treatment Effect

Background: Real-world evidence assessing the risk of inflammatory bowel disease (IBD) among patients receiving dipeptidyl peptidase-4 inhibitors (DPP4i) is conflicting. Abrahami et al. compared initiation of DPP4i to all other antidiabetic drugs, modeled DPP4i as a time varying exposure (TVE), and observed a harmful effect in the UK population. Wang et al. conducted an active comparator new-user (ACNU) study and observed a null effect in the US population. Such inconsistency may be due to differences in populations and study design, including choice of active comparators and approach to treatment changes during follow-up.

Objectives: We aimed to estimate the causal effect of DPP4i on IBD risk by an ACNU study in the UK population and assess the impact of study design differences on treatment effect estimates.

Methods: We assembled ACNU cohorts from the Clinical Practice Research Datalink (CPRD) GOLD population from 2007-2022, of adult initiators of DPP4i or one of the active comparators: thiazolidinediones (TZD), sulfonylureas (SU), or sodium-glucose transport protein 2 inhibitors (SGLT2i). New use of each drug class was defined as two prescriptions of the same drug class after a 1-year washout period, during which no use of the comparator is detected. Initiators were followed starting from 180 days post-second prescription of index drug (latency period) to incident IBD, death, or administrative censorship. Propensity scores (PS) were estimated and trimmed asymmetrically. Comparator drug initiators were weighted by the PS odds to estimate the average treatment effect in the treated. We estimated Hazard Ratios (HR) and 95% CIs using weighted Cox proportional hazards models in the ACNU and the TVE design (i.e. allow switching from comparator to DPP4i after cohort entry but not DPP4i to comparator).

Results: In the ACNU design, we identified initiators of DPP4i (105,326), SU (72,515), TZD (14,567), or SGLT2i (25,342). IBD incidence rates ranged from 34-47/100,000 person-years. Over a median follow-up of 2.2-5.5 years, DPP4i initiation was not associated with increased IBD risk across all comparison cohorts. The HRs for DPP4i were 1.05 (95% CI 0.77-1.43) vs SU, 0.95 (0.55-1.66) vs TZD, and 0.84 (0.47-1.50) vs SGLT2i.In the TVE design, we observed elevated HRs for DPP4i and a high proportion of switchers from comparator to DPP4i: DPP4i vs SU 1.13 (0.84-1.51), 41% switchers; TZD 1.33 (0.65-2.75), 11% switchers; and SGLT2i 0.92 (0.49-1.68), 10% switchers.

Conclusions: Initiating DPP4i does not increase short-term IBD risk compared with initiating alternative antihyperglycemic drug classes. Findings from this study emphasize the crucial role of study design in the assessment of treatment effect in pharmacoepidemiology studies.