Method validation of modern analytical techniques

Validation of analytical methods of well-characterised systems, such as are found in the pharmaceutical industry, is based on a series of experimental procedures to establish: selectivity, sensitivity, repeatability, reproducibility, linearity of calibration, detection limit and limit of determination, and robustness. It is argued that these headings become more difficult to apply as the complexity of the analysis increases. Analysis of environmental samples is given as an example. Modern methods of analysis that use arrays of sensors challenge validation. The output may be a classification rather than a concentration of analyte, it may have been established by imprecise methods such as the responses of human taste panels, and the state space of possible responses is too large to cover in any experimental-design procedure. Moreover the process of data analysis may be done by non-linear methods such as neural networks. Validation of systems that rely on computer software is well established. The combination of software validation with validation of the analytical responses of the hardware is the challenge for the analytical chemist. As with validation of automated equipment such as programmable logic controllers in the synthesis of pharmaceuticals, method developers may need to concentrate on the process of validation, as well as the minutiae of what is done.