Recalls and advisories of implanted cardioverter-defibrillators (ICDs) have become an unfortunate reality of cardiac rhythm management. With a paucity of data available on which to base replacement decisions, our goal is to model the potential risks and benefits of ICD generator replacement. The estimated risks are varied through a wide range to determine the potential range of outcomes.

Using initial estimates of risk derived from real data on 2915 advisory devices from 17 implanting centers, a decision analysis and Markov model were used to estimate survival according to device replacement decision. Survival rates at 5 years with and without device replacement were estimated at 60.38% and 60.66%, respectively. This difference was not significantly different on comparative analysis, using variability determined by Monte Carlo simulation. One-way and two-way sensitivity analyses are presented, demonstrating the minimal effect of varying estimates of risk. Only variation in risk of device failure had a differential effect on survival, with a survival benefit at 7 years if annual risk of device failure is at least 1.8%. Little differential effect on survival was demonstrated by variation of estimates of arrhythmia risk, nonarrhythmic mortality, and postprocedure infection rate.

Survival rates with a generator replacement or nonreplacement strategy in response to ICD recalls are similar and decrease nearly in parallel over time. The main factor with differential effect on survival is risk of device failure, although the level of this risk required to confer a survival advantage to a replacement strategy is quite large.