Intraoperative testing of implantable defibrillators is necessary, but compromises between patient safety and adequate information to predict performance must be made. Determining the threshold characteristics for each patient is not ethical, but data aggregated from clinical trials combined with limited intraoperative information from a particular patient can be used to characterize the patient's responses with measurable probability. In this article, defibrillation is modeled as a stochastic event in which energy (or, equivalently under square-root transformation, voltage) determines the probability of defibrillation at that output. All other factors are treated as though they varied randomly. The shape (or 'slope') of an individual's probability curve mimics that of the aggregate population, and varies only in location. The curve for the aggregate population gives the distribution of the locations; e.g., the voltage at which 90% of the population is defibrillated is also the output below which 90% of individual curves lie. A probability theorem allows the location of an individual's curve to be inferred probabilistically by the results of a limited number of trials. A study of human defibrillation involving an investigational device (model 2376, Medtronic, Minneapolis, Minnesota) provided data from which an aggregate curve was estimated, to exercise and illustrate the model. The estimated, individual probability of success at 500 V, based on a procedure utilizing a limited number of intraoperative trials at the same voltage, increases as the number of successes increases, as one would expect. The procedure can be compared to a 'step-down', in which successively lower voltages are attempted until a failure occurs. At 300 V the step-down produces a lower probability than the repeated-trial method at 300 V with the same number of trials; a similar result holds for 500 V.
|Original language||English (US)|
|Number of pages||5|
|State||Published - Dec 1 1988|