How Well Do Results From Randomized Clinical Trials and/or Recommendations for Implantable Cardioverter‐Defibrillator Programming Diffuse Into Clinical Practice?

Background Inappropriate implantable cardioverter-defibrillator programming can be detrimental. Whether trials/recommendations informing best implantable cardioverter-defibrillator programming (high-rate cutoff and/or extended duration of detection) influence practice is unknown. Methods and Results We measured reaction to publication of MADIT-RIT (Multicenter Automatic Defibrillator Implantation Trial-Reduce Inappropriate Therapy; 2012) and the Consensus Statement (2015) providing generic programming parameters, in a national cohort of implantable cardioverter-defibrillator recipients, using the ALTITUDE database (Boston Scientific). Yearly changes in programmed parameters to either trial-specified or class 1 recommended parameters (≥185 beats per minute or delay ≥6 seconds) were assessed in parallel. From 2008 to 2017, 232 982 patients (aged 67±13 years; 28% women) were analyzed. Prevalence of MADIT- RIT -specific settings before publication was <1%, increasing to 13.6% in the year following. Thereafter, this increased by <6% over 5 years. Among preexisting implants (91 171), most patients (58 739 64.4%) underwent at least 1 in-person device reprogramming after trial publication, but <2% were reprogrammed to MADIT - RIT settings. Notably, prevalence of programming to ≥185 beats per minute or delay ≥6 seconds was increased by MADIT - RIT (57.4% in 2013 versus 40.2% at baseline), but the following publication of recommendations had minor incremental effect (73.2% in 2016 versus 70.8% in 2015). High-rate cutoff programming was favored almost 2-fold compared with extended duration throughout the test period. Practice changes demonstrated large interhospital and interstate variations. Conclusions Trial publication had an immediate effect during 1 year postpublication, but absolute penetration was low, and amplified little with time. Consensus recommendations had a negligible effect. However, generic programming was exercised more widely, and increased after trial publication, but not following recommendations.