Quality control, calibration, standardisation and curve fitting
It is common practice to perform laboratory-based immunoassays as a batch or a single run that contains calibration and control material together with unknown samples. From the calibrators (or standards), a calibration curve can be derived and a concentration for the unknown sample can be determined. Inclusion of samples with known drug concentration (controls) at regular intervals allows an assessment of reproducibility within the batch. Duplication and control of reproducibility forms the basis of ‘within run’ precision assessment. Inclusion of the same quality-control material in consecutive runs gives an indication of run to-run reproducibility. This can be supplemented with the inclusion of previously analysed patient or test samples. Together with monitoring curve parameters, these components allow an opinion to be formed as to the acceptability of the batch-to-batch precision when compared with predefined parameters. Drug immunoassays can be run in a number of ways, depending on user requirements. Urine drug screening may only require a simple qualitative (positive or negative) result. Fully quantitative results are required for diagnostic purposes or TDM. In forensic applications, an assay run in a semi-quantitative mode (with a small number of calibrators) provides an estimated concentration. This estimated concentration is useful when progressing to a confirmation stage (e.g. GC-MS). For quantitative assays, the curve fit between data points is important and a number of curve fitting options are available. Some manufacturers supply ‘closed’ curve-fitting programs that determine the fit, and calibrator ranges are fixed on all instruments of that particular make. Open systems allow the user to select a suitable fit and transformation for the particular application. Care should be taken with this selection, as it is possible to mismatch the curve fit and immunoassay. Common problems include auto mated systems that attempt to fit a straight line to a curve or poorly fit the points of a curve. There is no substitute for the operator checking the raw data from an assay.
For toxicological analysis, it is often appropriate to use a point-to-point curve fit (while accepting that the curve when far from a point is less accurate). Other examples of curve fits include spline and four-parameter fits. The overriding consideration is to ensure, by validating the method, that the chosen fit connects the data points in the most suitable way. A simple example of how the same assay can produce different results because of data handling is shown in Figure 14.10. Both curves are plotted using the same results, but in Series 1 extra calibrator points are plotted. The curves plotted for Series 2 work perfectly satisfactorily as a ‘semi quantitative’ assay, but give inaccurate numerical results at points where the point-to-point curve fit is not suitable to provide a quantitative answer (illustrated best in Fig. 1 at the 100 lg/L region). The introduction of extra calibrators provides a more robust curve more suited to quantification of drug concentration. The addition of extra calibrators also ensures that a single calibrator point does not unduly bias the curve. In Figure 14.10, using Series 2 any small error or change to the 200 lg/L calibrator alters the curve significantly, but with Series 1 the curve is held by the 100 lg/L calibrator point.
Figure 1 Example to illustrate how different curve fitting programs can influence the results of an immunoassay.
