Cross Validation

Cross validation of analytical methods is a special case. Cross validations are a comparison of the validation parameters of two or more bioanalytical methods. Generally, most bioanalysts develop and validate an analytical method prior to the start of a clinical study. However, there are two situations that can arise where cross validations should be conducted: when two or more analytical methods are used to generate data within a single study (including situations where one method was significantly changed during the study), or when two or more analytical laboratories are used to generate data within a single study. In addition, the analyst should consider cross validation in cases where significantly different analytical methods were used to generate data in different studies, if both studies produced data of pivotal importance to the NDA. Unfortunately, there is no uniformly accepted format for conducting cross validations. However, there are two general approaches, which are quite similar. First, spiked samples of low, medium, and high concentrations are simply analyzed by both methods and compared.

Alternatively, clinical samples are analyzed by the different methodologies and plotted against each other (see Fig. 7). Both methods should provide the same value, and the slope of the line should equal unity. This approach also allows certain statistical comparisons to be made [13]. Generally, the FDA recommends that both spiked samples and patients samples should be compared between methods. However, it is also unlikely that both methods will be exactly equal. The question then is how much difference is acceptable. This issue has not been fully addressed, but usually the ± 15/20% criteria used for accuracy and precision has been applied. It is

FIGURE 7 Cross validation. A set of patient samples were analyzed with two different methods, A and B. The concentrations determined by each method are plotted against one another. Ideally, if both methods were equal, they would produce the same concentrations and a slope equal to one. In this case the slope is 0.66, which indicates that Method A reports higher concentrations than Method B.

0 1000 2000 3000 4000 5000 Method A (ug/L)

FIGURE 7 Cross validation. A set of patient samples were analyzed with two different methods, A and B. The concentrations determined by each method are plotted against one another. Ideally, if both methods were equal, they would produce the same concentrations and a slope equal to one. In this case the slope is 0.66, which indicates that Method A reports higher concentrations than Method B.

advisable that the bioanalyst assess the objectives of the clinical study, and set the requirements for cross validation appropriately.

Was this article helpful?

0 0

Post a comment