The evaluation of the analytical performance of a quantitative medical device includes estimation of the following analytical figures of merit:(1) accuracy/bias, (2) precision, (3) linearity, (4) total analytical error,(5) limit of detection, (6) limit of quantification and (7) interference effects. It may be necessary to include some or all of these parameters in the 510(k) application submitted to the Food and Drug Administration (FDA). The type of samples used to evaluate the analytical performance must be representative of the type of samples encountered during the intended use of the device. The number of samples used to evaluate the analytical performance must be large enough to justify the manufacturer's claims.
There are no universal protocols to specify how the various tests should be done. Also, there are no universal acceptance criteria for the above analytical figures of merit. Acceptable results vary with analytes and the areas of application. Guidelines from appropriate regulatory and scientific organizations should be consulted for experimental details such as the types of samples, the number of samples and the necessary number of replicates. In the U.S.A., the FDA and the Clinical and Laboratory Standards Institute (CLSI) provide guidelines and detailed information on these and similar tests.
In evaluating analytical performance data, it is very helpful to generate plots whenever appropriate and to examine the plots for trends, patterns and any potential human errors. Following is an overview of the assessment of accuracy/bias, precision, linearity, total analytical error, limit of detection, limit of quantification and interference effects.