Functional disability after clinically significant extracranial bleeding: a secondary analysis of the Aspirin in Reducing Events in the Elderly (ASPREE) Trial

Figures

• 

  FIGURE 1. Cumulative incidence of ADL dependence stratified by clinically significant extracranial bleeding. Cumulative incidence of ADL dependence for patients with a prior bleed or no bleed was estimated using extended Kaplan–Meier curves for time-dependent covariates. Patients are initially all at risk in the “No Bleeding” group but are switched to be at risk in the “Clinically Significant Extracranial Bleeding” group when they experience a bleed. ADL, activities of daily living.

• 

  FIGURE 2. Association of clinically significant bleeding with new ADL dependence. Associations are estimated based on a Cox model with a time-dependent covariate for bleeding adjusting for potential confounders including participant demographics, comorbidities, smoking status, alcohol use, concomitant medications, and performance on tests for cognitive abilities (Modified Mini-Mental State Test), and physical functioning (12-Item Short Form Health Survey physical component score). Separate models are fitted to estimate the associations specifically with clinically significant GI and non-GI bleeds. An extended model that additionally includes a main effect for randomization arm and its interaction with bleeding is used to estimate associations among aspirin- and placebo-randomized patients. ADL, activities of daily living; CI, confidence interval; GI, gastrointestinal.

• 

  FIGURE 3. Sensitivity analysis on the association between clinically significant bleeding with new ADL dependence. Associations are estimated using linear probability models for ADL dependence in each semiannual period during trial follow-up with participant-level fixed effects, an indicator for a prior clinically significant extracranial bleed, and a linear time trend. Separate models are fitted to estimate associations specifically with clinically significant GI and non-GI bleeds. An extended model that additionally includes an interaction effect between prior bleed and treatment arm was fitted to estimate associations among aspirin- and placebo-randomized patients. Cluster-robust standard errors clustered at the participant-level were used for calculating confidence intervals. ADL, activities of daily living; GI, gastrointestinal.