The "Bioanalytical Method Validation" document lays down some criteria that may be applied to the validation of the most critical assay parameters. It should be noted, however, that in some tests it may not be possible to reach the levels of accuracy and precision specified in this document, while in others it may be possible to set closer limits. The criteria should be test-specific.⁹

Accuracy is determined by replicate analysis of samples containing known amounts of the analyte. It is recommended that at least five determinations per concentration be done. A minimum of three different concentrations within the range of expected concentration should be tested so that data comprise at least 15 analyses. The mean value should be within 15% of the actual value, except at the lower limit of quantitation, where it should not deviate by more than 20%.

Precision should be measured using a minimum of five determinations on at least three different concentrations, testing multiple aliquots of a single homogenous sample. The precision determined at each concentration should not exceed 15% of the coefficient of variation (CV), except for the LLOQ, where it should not exceed 20%. The same criteria should be applied to each type of measurement of precision (intra-assay, inter-assay).

The simplest mathematical model that yields a useable relationship between concentration and response should be used.

In developing the standard curve, there must be <20% deviation of the LLOQ from the nominal concentration and <15% deviation of other standards. No fewer than six points should be used to define the curve. At least four of six non-zero standards should meet the above criteria, including the LLOQ and the calibration standard at the highest concentration. Considering the typical asymptotic curve generated by most biological quantitative assays, the guidance recommends that more than six points should be used to fit the curve. The concentration-response relationship is most often fitted to a four- or five-parameter logistic model. Computer programs can perform this automatically, and some of these are capable of dealing with non-parallelism as well.

Acceptance criteria should be set for bioanalytical test validation before the series is run. The guidance offers the following acceptance criteria: at least 67% of QC samples should be within 15% of their respective nominal value, and 33% of the QC samples (not all replicates at the same concentration) may be outside the 15% limit. Some tests may require wider limits.

Some small points to consider in performing the replicate runs required to validate a bioassay, using the ubiquitous microplate method, are as follows:

• Ensure that a certain number of microplates receive samples in reverse order ("front-to-back") or some other arrangement, to limit potential effect of time or position on the plate. Positional effects may be very important. It has been reported that multi-channel pipetting devices, which are often used for plate assays, may show variability between channels in the volume delivered.
• Perform identical validation runs using a sufficient number of lots of cells, media, standards, etc. to ensure that the test is sufficiently robust to withstand lot-to-lot variation.
• Ensure that the SOP describing the assay is sufficiently detailed and that it gives unequivocal instructions for procedures that may affect accuracy or precision of the test, e.g., mixing, washing steps.
• Be certain all reagents can withstand the exposure that occurs during plating out and further manipulation of the microplates during the assay.
• Check that the plate reader has been adequately qualified for this assay.
• Record all data directly after reading, and have mathematical manipulations checked by another analyst, or, if they are normally performed by the instrument's software, check this by manual re-calculation.
• If validation requires evaluation of a number of different test parameters, a "multi-dimensional" test plan can be used.