Time filter

Source Type

Dorset, United Kingdom

Background: Calibration curves for use in chromatographic bioassays are normally constructed using least squares regression analysis. However, the effects of changes in the number and distribution of calibrators, together with the choice between linear and nonlinear regressions and their associated weighting factors, are difficult to quantify. Results/discussion: A Monte Carlo simulation software package is under development that uses the assay range, concentration versus response relationship and intra-assay measurement precision profile to quantify the errors resulting from different calibration procedures. Criteria for assay batch acceptance, in terms of calibrator and QC sample estimates being within specified limits, are included in the design. Conclusion: Monte Carlo software provides a means to evaluate calibration strategies and maximize assay batches meeting acceptance criteria. © 2013 Future Science Ltd. Source

Background: The acceptance criteria for chromatographic bioassays are generally based upon the specifications found in the US FDA Guidance for Industry document, whereas regression analysis is a statistical process and, thus, most suited to large data populations, rather than the small populations used in assay calibration. Data that form part of the statistical population are often excluded in order that calibrations meet these acceptance criteria. Results/discussion: Monte Carlo simulations of various linear regression calibration strategies, which take into account guidance acceptance criteria, have been used to quantify calibration errors. Conclusion: It is concluded that 1/x2 weighting generally produces results that meet both theoretical and regulatory requirements, and that use of single, rather than duplicate, calibrators can result in misleading statistics at the LLOQ. © 2013 Future Science Ltd. Source

An introduction to the use of the mathematical technique of Monte Carlo simulations to evaluate least squares regression calibration is described. Monte Carlo techniques involve the repeated sampling of data from a population that may be derived from real (experimental) data, but is more conveniently generated by a computer using a model of the analytical system and a randomization process to produce a large database. Datasets are selected from this population and fed into the calibration algorithms under test, thus providing a facile way of producing a sufficiently large number of assessments of the algorithm to enable a statically valid appraisal of the calibration process to be made. This communication provides a description of the technique that forms the basis of the results presented in Parts II and III of this series, which follow in this issue, and also highlights the issues arising from the use of small data populations in bioanalysis. © 2013 Future Science Ltd. Source

Burrows J.,Anastats | Watson K.,Anastats
Bioanalysis | Year: 2015

Chromatography-based drug assays are generally considered to be linear, provided that the back calculated concentrations of 75% of the calibrators are within ±15% of their target values. Data meeting this criterion are not usually subject to further examination in order to evaluate the extent of any nonlinearity and whether use of a nonlinear calibration function would improve the accuracy. Examples of nonlinear behavior are presented for several chromatographic systems and most are best described by the nonlinear equation y = a + bx + cxln(x). A more critical evaluation of linearity, as presented herein, can lead to the identification of nonlinear behavior and an improvement in accuracy by use of a nonlinear calibration regression. © 2015 Future Science Ltd. Source

Traditional approaches to the assessment of the linearity of chromatographic assays initially assume that the concentration-response relationship is linear and only deemed to be nonlinear if certain parameters lie outside pre-set acceptance criteria. A different approach is described in that the a priori assumption is made that the assay does have some curvature that can be described by fitting a nonlinear regression curve. Any nonlinearity can be quantified from the curvature terms of the resultant regression coefficients. The success of this alternative approach is dependent upon the appropriate choice of nonlinear regression to accurately describe concentration versus response relationship of the assay. © 2015 Future Science Ltd. Source

Discover hidden collaborations