Correspondence of linearity evaluations to the performance of clinical chemistry instrument and reagent systems.

Abstract

The importance of demonstrating linearity for a clinical chemistry analytical system has caused considerable debate. Disagreement about the way linearity is demonstrated has led to some misunderstanding about just what linearity means to the analytical goals. In many cases where linearity is evaluated by statistical means, the outcome will indicate a nonlinear condition. A subsequent graph plot of the same data usually produces a line that is virtually straight. Confusion over which linearity evaluation outcome to accept creates a lack of confidence in either approach owing to the sometimes obvious disagreement between the two outcomes and the ill-defined impact linearity has on analytical error or clinical usefulness. Models of analytical performance can be constructed in such a way that the paradox of conflicting linearity evaluations can be examined through their effects on the analytical goals.