engleski

OVERVIEW OF VALIDATION PRINCIPLES I: LINEARITY, WORKING RANGE, SENSITIVITY AND ACCURACY

Most measurement techniques make use of a calibration graph to estimate the analyte concentration in unknown samples. This implies that a decision concerning the nature of the relationship between the concentration and the response has to be taken. Very often a simple straight line relationship is preferred and many measurement techniques are designed to achieve proportionality between response and concentration. In practice, however, deviations from linearity are frequently observed. Therefore, it is essential for a method validation program to include a linearity test. Therefore, it is essential for a method validation program to include a linearity test. Such tests and closely related parameters (working range and sensitivity) are discussed in this session. Accuracy and bias are of course in close relation with linearity estimation. Systematic errors are characterized by terms such as trueness and bias related to term accuracy. Systematic errors may be constant (absolute) or proportional (relative). A constant error refers to a systematic error independent of the true concentration of the analyte and should be expressed in concentration units. A proportional error depends on the concentration of the analyte and should be expressed in relative units, such as percentage. The main sources of constant error are insufficient selectivity, which is caused by another component that also yields a response, and inadequate blank corrections. Proportional errors are caused by errors in the calibration, for instance by different slopes of the calibration lines of the standards and in the sample (matrix interference). The incorrect assumption of linearity over the range of analysis will also cause errors related to the concentration to be determined. The experimental design and statistical background necessary for estimation of linearity working range, sensitivity and accuracy will be discussed and presented. Moreover, practical examples will be calculated on sight using MS Excel offering easy transfer and implementation of presented know-how in participants own laboratory.

validation; linearity; working range; sensitivity; accuracy

nije evidentirano