Analytical goals for rectilinear calibration functions.

ABSTRACT

Analytical goals for rectilinearity based on within-subject biological variation have not yet been advocated. On the other hand, the statistical tests to evaluate rectilinearity may be too restrictive for clinical purposes. If rectilinearity of the least-squares regression is rejected by the statistical test used, we propose to compare the systematic error introduced using such a regression line as the calibration function, with the allowable total error. Since total error ideally should be less than the within-subject biological coefficient of variation (C.V.Bw) the goal for rectilinearity we propose is that the maximum allowable systematic relative error produced by the calibration function (LoRi) when a lack of rectilinearity really occurs is LoRi < C.V.Bw -1.96 C.V.M, where C.V.M is the between-run metrological coefficient of variation of the measurement procedure, corresponding to the value of concentration under study.