The establishment of normal limits for serum proteins measured by the rate nephelometer. Concepts of normality revisited.

A study of the distribution of nephelometrically determined data for IgG, IgM, IgA, C3, C4, and haptoglobin was made in two geographically separate, healthy populations. Evaluation of these data by several tests of statistical normality revealed that the tests vary considerably in power; i.e., data that one rejects as gaussian may be accepted by the other. No rules exist to enable one to decide which is the best test to apply in a given situation. This study indicates that the uncritical application of X +/- 2 S.D. for the reference range of clinical laboratory tests may be fraught with error, which may not be obvious and may not provide clinically useful information in many cases. A large scale clinical investigation designed to empirically obtain limits of different laboratory tests with desired levels of specificity and sensitivity for different disease states may prove rewarding and yield knowledge more useful for clinical purposes. Until such an investigation is carried out, however, we recommend determination of reference ranges by a nonparametric percentile estimate (2.5 to 97.5 per cent), which makes no a priori assumptions about the distribution of the data in the population.