(CSPOT15) Impact of antithrombotics on the association between proton pump inhibitors and ischaemic stroke: standard and modified self-controlled case series

Background: Previous studies have yielded conflicting associations between proton pump inhibitor (PPI) use and the risk of ischaemic stroke, which could be due to unmeasured confounding.

Objectives: This study aimed to investigate the association between PPI use and the risk of ischaemic stroke using self-controlled case series (SCCS) models.

Methods: Adult PPI users with an incident ischaemic stroke were identified from 1/1/2003 to 31/12/2014 in the Clinical Data Analysis and Reporting System (CDARS) in Hong Kong. The exposure window was pre-defined as days 1-15, 16-30, 31-60, 61-90, and 91 to the prescription end date, since the prescription start date. All other periods were referent windows. We assessed the assumptions for standard SCCS and estimated incidence rate ratios (IRR) using conditional Poisson regression. We also conducted modified SCCS using estimating equations to address the violation of assumptions of standard SCCS and stratified the analysis by the use of antithrombotics.

Results: We included 9,976 and 19,976 patients in the standard and modified SCCS models respectively. The assumptions for standard SCCS were violated, with a reduced risk of ischaemic stroke during days 91 to prescription end [0.35, 95% Confidence Interval (95% CI): 0.29-0.43] but no differences in risks in other periods versus the referent window. Using modified SCCS, the IRRs for ischaemic stroke were 1.22 (0.86-1.74) for days 0-30, 1.28 (0.85-1.94) for days 31 to 60, 1.55 (1.00-2.42) for days 61-90, 1.51 (95% CI: 1.14-2.00) days 91 till prescription end, versus referent window. In stratified analysis, increased risks were observed in patients co-prescribed PPIs with antithrombotics [1.89 (1.23-2.91) for days 0-30, 1.90 (1.17-3.08) for days 31-60, 2.12 (1.26-3.55) for days 61-90, 2.26 (1.61-3.16) for days 90 to prescription end] versus referent window. No differences in risks was observed in those receiving PPI without antithrombotics [0.69 (0.37-1.31) for days 0-30, 0.63 (0.27-1.51) for days 31-60, 0.93 (0.38-2.27) for days 61-90, 0.37 (0.18-0.78) till prescription end] versus referent window.

Conclusions: We observed an increased risk of ischaemic stroke associated with PPI use in people co-prescribed PPIs and antithrombotics, which could be due to the high baseline risk of stroke, but not in users prescribed PPI only. Modified SCCS should be considered when there is a violation of the standard SCCS assumption.