We have derived the mathematical relationship between the coefficient of variation associated with repeated measurements from quantitative assays and the expected fraction of pairs of those measurements that differ by at least some given factor, i.e., the expected frequency of disparate results that are due to assay variability rather than true differences. Knowledge of this frequency helps determine what magnitudes of differences can be expected by chance alone when the particular coefficient of variation is in effect. This frequency is an operational index of variability in the sense that it indicates the probability of observing a particular disparity between two measurements under the assumption that they measure the same quantity. Thus the frequency or probability becomes the basis for assessing if an assay is sufficiently precise. This assessment also provides a standard for determining if two assay results for the same subject, separated by an intervention such as vaccination or infection, differ by more than expected from the variation of the assay, thus indicating an intervention effect. Data from an international collaborative study are used to illustrate the application of this proposed interpretation of the coefficient of variation, and they also provide support for the assumptions used in the mathematical derivation.Although assay variability is well recognized as pertinent to the interpretation of quantitative bioassays such as the enzyme-linked immunosorbent assay (ELISA), few tools that link assay precision with interpretation of results are readily available. In our investigations, we have expanded on previous studies that evaluated the relationship between assay precision and the capabilities and limitations of a given assay system. In this article we develop a simple procedure to determine the probability that an assay will accurately discern whether two samples have the same analyte concentration or not based on a knowledge of the assay variability as measured by the coefficient of variation (CV).In many laboratories, the variability of the ELISA and other methods of chemical assay that produce continuous-type values is summarized not by the standard deviation (SD) but by the CV, which is defined as the SD divided by the mean, with the result often reported as a percentage. The main appeal of the CV is that the SDs of such assays generally increase or decrease proportionally as the mean increases or decreases, so that division by the mean removes it as a factor in the variability. The CV is therefore a standardization of the SD that allows comparison of variability estimates regardless of the magnitude of analyte concentration, at least throughout most of the working range of the assay.In serological assays a twofold difference in measurements of the same sample has been widely regarded as the upper limit on acceptable variability, and the frequency of such differences among pairs of repeated measurements has been proposed as an apt index for assay variability (5). Wood (4) showed the mathema...

No abstract