Heparin-bonded versus standard polytetrafluoroethylene arteriovenous grafts: A Bayesian perspective on a randomized controlled trial for comparative effectiveness

Background Heparin-bonded polytetrafluoroethylene grafts were marketed to improve hemodialysis access outcomes but are twice the cost of standard polytetrafluoroethylene. We launched a randomized trial of heparin-bonded polytetrafluoroethylene versus standard polytetrafluoroethylene for hemodialysis access to compare patency. Since the trial began, additional studies were published with heterogeneous findings. We performed an interim analysis by Bayesian methods using prior probability from meta-analysis of existing literature. Methods NCT01601873 is a randomized, blinded trial of heparin-bonded polytetrafluoroethylene versus standard polytetrafluoroethylene for dialysis access at 5 sites. Planned sample size was 200 with 1-year primary patency as the primary endpoint. At interim analysis (50% of sample size at 1 year), we also performed a meta-analysis for 1-year primary patency with a random effects model to compute summary rate ratio and standard-error estimates. Meta-analysis estimates formed a prior probability for a Bayesian Cox regression model, and trial data were reanalyzed to develop posterior probability of heparin-bonded polytetrafluoroethylene effectiveness at our hypothesized effect size. Futility analysis was conducted using posterior probability estimates. Results One hundred and five patients were enrolled at the time of interim analysis. One-year primary patency was 34.9% in the heparin-bonded-polytetrafluoroethylene group vs 32.7% in the standard-polytetrafluoroethylene group (P = .884). Summary rate ratio from the meta-analysis (1,209 patients) was 0.87 favoring heparin-bonded polytetrafluoroethylene (P = .33). Posterior hazard ratio from Cox regression was 0.90 (credible interval 0.70–1.13) favoring heparin-bonded polytetrafluoroethylene, which was not significant. Bayesian posterior probability of the a priori hypothesized 20% better patency with heparin-bonded polytetrafluoroethylene was 24%. Sample size to detect superiority with the small observed effect size would require about 3,800 subjects. Conclusion Current evidence does not demonstrate sufficiently large benefit of heparin-bonded polytetrafluoroethylene over standard polytetrafluoroethylene for dialysis access to justify higher cost. Given similar 1-year patency rates, a conclusive finding of superiority was judged to be infeasible, and the trial was stopped for futility.