Examination of an exponential model of conduction through the human atrioventricular node

The atrioventricular node (AVN) has been modeled by relating output (A2H2 or H1H2) to input (A1A2) where A and H are atrial and His bundle electrograms during fixed rate atrial pacing (A1A1) or with an extrastimulus (A2). (Formula: see text) This study examined this model in 61 nonselected patients, specifically for AVN (in)stability and the possibility of multiple pathways. After programmed atrial stimulation at two basic cycle lengths of 600 ms and 462 ms, A1H1, A2H2 and H1H2 were digitized and plotted as a function of A1A2. Seven of 104 trials were rejected as SD. A1H1 was greater than 15 ms, suggesting AVN instability. Another 26 and 34 plots, respectively, of A2H2 and H1H2 were rejected because of inadequate data. In the remainder, goodness of fit of the single exponentials was tested statistically in three ways: R2, the runs test, and the Kendall rank coefficient test. Results were compared with an electrophysiologist who examined plots for one or more pathways (either discontinuous curves or slope change in a continuous curve). Single exponentials were successfully fitted (by runs test) in 44/71 and 34/63 of A2H2 and H1H2 plots, respectively, usually in accordance with the cardiologist. Discordance between computations and the cardiologist could be attributed to data scatter and lack of a sufficiently rigid stimulation protocol. The identification of bifurcation points in the presence of multiple pathways, particularly when manifest as a change in slope (approximately 6% of trials) rather than discontinuity of plots (approximately 20% of trials) remains an outstanding problem.