Biomedical Engineering Reference
In-Depth Information
Due to the intrinsic variability of biological matrices, guidelines require to prove
that in several independent sources of the same matrix: FDA requires a minimum of
six, EC requires a minimum of 20 representative blank samples.
Another approach to verify selectivity is described (also reported in EC guide-
lines) as follows: once selected a range of chemically related compounds (metabo-
lites, derivatives, etc.) or other substances likely to be encountered with the
compound of interest that may be present in the sample, investigate after analysis
whether the presence may lead to a false identification, the identification of the tar-
get analyte is hindered by the presence of one or more of the interferences, or the
quanti fi cation is notably in fl uenced.
7.2.5
Stability
Insufficient stability of the analyte or matrix constituents in the sample during stor-
age or analysis may give rise to significant deviations in the outcome of the result of
analysis. Furthermore, the stability of the calibration standard in solution should be
checked. Usually analyte stability is well characterized under various storage condi-
tions. Drug stability in a biological fluid is a function of the storage conditions, the
chemical properties of the drug, the matrix, and the containing system. Stability
procedures should evaluate the stability of the analytes during sample collection and
handling, after long-term (frozen at the intended storage temperature) and short-
term (bench top, room temperature) storage, and after going through freeze and
thaw cycles. The procedure should also include an evaluation of analyte stability in
stock solution.
7.2.6
Calibration
The relationship between the concentration of analyte in the sample and the corre-
sponding response (in bioanalytical methods mostly the area ratio of analyte versus
internal standard (IS)) must be investigated. There is general agreement that for
bioanalytical methods, calibrators should be matrix-based, i.e., prepared by spiking
the blank matrix with the standard. Calibrator concentrations must cover the whole
calibration range and should be evenly spaced across the range [
60
] . An acceptable
correlation coefficient is 0.99; however, in some cases 0.98 is minimally acceptable.
In addition, it is good practice to evaluate the range of the calibration by calculating
the value of each calibrator against the curve. Values of ±20 % are generally accept-
able for most applications, although ±10 % is preferred. Single point calibrations
are discouraged unless controls are used at or close to the upper and lower quantita-
tive reporting limits [
56
] .
Usually, linear models are preferable (linear ordinary, i.e., unweighted, least
squares regression model is not appropriate in many cases, in which weighted least
squares model should be applied), but, if necessary, nonlinear (e.g., second order)
models can be used [
60
] .
