PubMed1975
For elucidation of the type of population distribution of Frank-vectorcardiographic items of normal Japanese population, forty-eight measurements were obtained from 364 healthy Japanese men in the fourth decade. These measurements were studied with Fisher's g test of normality of population distribution, which showed that normal population distribution could be assumed in only six items (12.5%)-the maximal QRS vector angle and the directional QRS-T angle in the frontal plane, RX and TZ height, OZ duration, and anteorior accession time. The frontal maximal T vector angle was regarded to have an approximately normal population distribution. In the remaining forty-one items logarithmic normalization of distribution was attempted with a transformation formula, Y=Log10 (kX+c). Being applied to Y's, the variate logarithm-transformed with an appropriate value of c, Fisher's g test accepted the null hypothesis of normality of population distribution in 27 items (56.3%); in other words, the population distribution of 27 items was lognormal. These included the magnitudes of maximal QRS and T vectors and T/QRS ratios in the three projection planes, and some of vector angles. The population distribution was usually lognormal in scalar amplitudes and ratios except for items related to Q and S waves in leads X and Y. Bimodality of distribution was seen in six items (12.5%), which consisted of the maximal QRS vector angle and the QRS-T angles in the horizontal and left sagittal planes. In these planes the maximal QRS vector and the directional QRS-T angles showed biomodality attributed to combination of two normal distributions. Three of the four QRS-T angles in these planes and also the horizontal maximal QRS vector angle were lognormalized, although a lesser degree of bimodality still remained after seemingly successful logarithmic transformation in respect to the tg statistics. Chi-square test accepted the null hypothesis of exponential population distribution in four items (8.4%), which included the depth of QX, SX, and SY, and S/(S+R) ratio in lead Y. Exponential approximation appeared much better than normal one in the rest of measurements related to Q and S waves in leads X and Y. The following conclusions were drawn from the present study: i) statistical tests based on normal population distribution should be carefully applied to Frank-vectorcardiographic measurements of healthy Japanese men in the fourth decade and, presumably, of healthy Japanese adults regardless of age and sex; and ii) logarithmic transformation is an efficient means to normalize a skewed distribution in the majority of measurements.