Method validation should focus on demonstrating that an assay is fit for its intended purpose. We have applied the β-expectation tolerance interval - a statistical approach that predicts the accuracy of assay measurements in the future - to the validation of two different cell death biomarker assays, the M30 and M65 ELISAs. A meta-analysis was conducted on a total of 57 different M30 and M65 assays run over a 2 year period. All code utilised in calculations was developed using MATLAB. The optimal fit to the calibration curve for the M30 assay was shown to be a quartic curve which yielded a β-expectation tolerance interval of +20.5% and -23.6% at β=95% over a wide range of QC standards (88-810 U/L). However, such a fit required at least 7 points to avoid problems with over fitting. A linear fit to the M65 calibration curve normally produced a tolerance interval of less than ±20%, however, marked inter-batch variations were evident. Amelioration of batch to batch variations was accomplished by fitting M65 calibration data preferably to a 4-parameter logistic function or a cubic spline. The minimum number of QC replicates and different assays required to produce reliable accuracy profiles was determined. The β-expectation tolerance interval approach has resulted in further optimisation of the M30 and M65 ELISAs as biomarker assays that should translate into greater accuracy in results generated from clinical trials samples.
Copyright © 2011 Elsevier B.V. All rights reserved.