A new look at the limits of detection (LD), quantification (LQ) and power of definition (PD).

The relationship between the concentration of the analyte and the imprecision of an analytical method can be displayed by the precision profile in which the coefficient of variation (relative standard deviation) is plotted against the concentration of the analyte. The function of the curve of the profile and its confidence limits can easily be assessed by a computer program developed by W.A. Sadler & M.H. Smith (Clin. Chem. 36 (1990), 1346-1350). For the assessment of limits of detection and of quantification the following procedure is proposed: The lower (and upper) limit of the measuring interval is defined by the point at which an acceptable CV-line intersects the confidence limit. If, in the variance function one sets the concentration to zero, the normal distribution of the random errors of the blank will result. The mean of the next adjacent normal distribution, following the variance formula and overlapping the "zero-distribution" by a defined amount, represents the limit of detection. Within the described measuring interval, or within a fraction of it, one might construct overlapping normal distributions in an analogous manner. Their number represents the "power of definition" (PD) (instead of the "analytical sensitivity"), which also depends on the concentration of the determinand according to the variance function. We tested these hypotheses by a comparison of two methods for the determination of cyclosporin A (ciclosporin, INN). Our results demonstrate that the data of the lower limits of the measuring interval and of the limit of detection agree well with data from the literature obtained in extensive interlaboratory surveys.