Abstract
In a double blind controlled trial, 24 hour DCG records were obtained from 70 patients with ventricular premature contraction (VPC), twice during run-in observation and once at the end of 2 weeks treatment with metoprolol or propranolol.
Each record was summarized in a form of 6-point counts, each point consisting of 4 hours'VPC counts. Then the differences between the pre-and postmedication data of corresponding time points were taken to obtain differential data, which were subjected to principal component analysis (PCA), cluster analysis (CA) and periodic regression analysis (PRA).
The 6-point VPC counts were presented as a set of pre- and postmedication plottings to 5 raters (cardiologists) who independently assessed the degree of improvement in VPC counts on a 6-ordered-categorical scale. The reliability of judgment was 76%. According to the extent of disagreement in the rating, these data were classified into 2 groups of good and poor agreement, and analyzed by means of stepwise regression analysis (SRA) to clarify the policy of judgment.
The results of PCA of the differential data indicated that the size factor (the difference in the total VPC counts) occupied 73% of the overallvariance and that the shape factors (the pattern of the differential plotting) of day-midnight and morning-evening contrasts held 64% of the nonsize variance.
PRA of the differential data gave similar results to those of P CA. With the aid of CA, it was suggested the decrease in VPC counts dominant during day time, although not so clear.
The results of SRA of raters' judgments showed that the judgment policy was largely influenced by the size factor, whose coefficient of determination was 50-60% and further to be modified to some extent by the baseline counts in the control period. In the “good agreement” group the policy seemed to be based on the difference or ratio in the total VP C counts. In the “poor” agreement group, however, the policy was complex, and included shape factors.
