Some theory of reference values. I. Stratified (categorized) normal ranges and a method for following an individual's clinical laboratory values

Abstract

The conventional population-based normal range has recently been shown to be a generally defective reference criterion for assessing individual laboratory test results. Applying a previously derived formula to published data, we find that the use of age-, sex-specific normal ranges may fail to produce a substantial improvement in sensitivity over nonspecific ranges, even when age-sex differences in mean values are statistically significant. This occurs when the difference in means is not accompanied by a sufficient reduction in the variation among individuals within a given class. Turning therefore to comparison of an individual's current measurement with his own previous value(s), I suggest a simple statistical model that leads to sequential testing of each new observation against an exponentially weighted moving average of previous results. Estimates of biological and analytical components of variance are required. The ability of this method to detect trends in very short series is explored with the aid of computer-simulated laboratory data. A sample of these data is also used to illustrate the application of these estimation and testing procedures by means of a graph.