Test Method Validation - Accuracy

[Consumer Products Guy]

Can anyone put into layman's terms the generally acceptable limits for "Accuracy" in a test method validations? In the past I have used 98-102% accuracy as my guidelines, but are wider limits acceptable? I have good precision data, linearity, range, DL, QL, etc., I was wondering what others use, and how to express in layman's terms. This would be for a GLP validation, not cGMP or ICH directly, but all are similar (and similarly vague).

[indium]

How accurate do you want your method to be? The desired accuracy will depend on what the result is used for. For pharmaceutical main component HPLC assays the accuracy will be driven by the assay specification. For example, if the specification is 99.0-101.0% then you cannot have a 5% bias in the method. For the quantification of an impurity that is at a level of about 0.01% then a bias of 10% will not make any difference to the result if it is reported to two decimal places. For the most part the accuracy and precision will be dictated by the specification acceptance criteria, which may or may not be influenced by outside agency, e.g. FDA or EPA, expectations.

[mriley]

The company I work for has always used an average recovery of 95-105% with no individuals outside of the 93-107%. Our limits for the C of A are 90-110% for actives and 85-115% for other analytes.

[ccyting]

The accuracy and precision are determined by the method per se. When we start a method development, we should pre-set the limit of the method according to the requirements and to see the developed method can fit the limit. By various optimization we can evaluate the resulting accuracy and precision. Then we do the validation to check the pre-set limit can be achieved. We all know for the assay method of active, we need to achieve 1 or 2%. For impurity or degradation product at 0.1% level, we can broaden to 10%. If we have exhausted our method development and validation, whatever the precision and accuracy is what we would like to set up. Another words, we need to have a method that we can reproduce and transfer anytime. Good luck.

[KC]

Accuracy and precision of a method are not defined in the GLP's, they are more a guide to doing things in the proper manner and not definitive in how good you do them. The accuracy and precision of a method is usually defined by the type of study and the guideline of the agency for which the study is to be submitted. I develope and validate methods on a regular basis the first thing I check is the first page of the protocol to see where it is being submitted (usually EPA or EU). I do try to look at both regardless and pick the most stringent of the two in case the sponsor later decides to try and submit the data to both. The EU does a good job of spelling out parameters of specificity, accuracy and precision as well as LOQ (or limit of determination) as they call it which is confusing as hell because limit of determination (LOD) is definately not the same thing as limit of detection (LOD)....go figure.