Comparison of Frank's and McFee's lead systems using multivariate statistics.

Two widely used orthogonal corrected 3-lead system - the Frank and the McFee systems - were studied in order to evaluate whether a significant difference in diagnostic performance could be observed; in the case of a positive answer, we would be able to advocate one system instead of the other. No such overall difference was noticed. Although +/- 17% discrepancies were found on the individual classification level (a patient correctly classified with system A and missed with system B and conversely), the practical implication is negligible since each system is responsible for +/- half of these discrepancies. Multivariate analysis, as already largely proved by Pipberger's group, drastically ameliorates the diagnostic results. Special caution has been given to the number of selected discriminators in order to enhance repeatability of the results; the reproducibility, using the approach developed by Cornfield, was found excellent. Some particular features as to the choice and location of the best discriminators were found somewhat puzzling by the autors but no deterministic explanation could be offered: the selection of the variables, indeed, rested on statistical bases and not on the (sometimes fragmentary) knowledge of what is going on in the heart and how these events relate to surface waveform patterns.