Assay accuracy data, poor

Case Study 3 Poor Assay Accuracy Data, an Out-of-Specification Investigation 260... 

The next step in the process is to assess the predictive capability of the alternative method. This may be done by confirming that alternative method data input into the predefined prediction models provides outputs that predict in vivo toxicity at the level of both accuracy and precision defined at the beginning of the study. Once the data are available from the validation study, each result from the alternative method should be converted by the algorithm(s) in the prediction model into a prediction of in vivo toxicity. Then the predicted toxicity should be directly compared with the actual toxicity of each test substance. If the method predicts toxicity within the limits defined by the prediction model, then it would provide strong evidence supporting the predictive capability of the assay. If the results from the alternative method poorly predict in vivo toxicity, then there would be little evidence supporting the utility of the method. 

---