Mathematical approach to some methodological difficulties in settling newborn IgE reference levels (study in a...

The authors used statistical procedures to attempt to solve the usual methodological difficulties inherent in the determination of neonatal IgE reference values. The problem of un detectable values was overcome by theoretical estimation of their distribution, through the regression of the probits of the cummulative frequencies of the determinations on the natural logarithms of the known concentrations. Thus, the geometrical mean and 95% confidence limits were established, and population selection was replaced by exclusion of aberrant observations in successive steps, with recalculation of reference values. The final geometrical mean obtained was 0.21 I.U./ml with 95% confidence limits of 0.03 and 1.54 I.U./ml. These results have been compared with the work presented by other authors, observed differences probably reflecting distinct methodologies.