Back to the Top
Validation issue to bioanalytical methods (specially in pharmacokinetics
and bioequivalence studies) describes the necessity of evaluation of the
parameter linearity. So, calibration curves are generally evaluated by
ordinary linear regression (unweighted or weighted). So, I have a
question: Should always the relationship between concentration of
standard solution and chromatography signal be described by a linear
regression or any regression technique may be applied, as polynomial
Any comments will be really appreciated.
Daniel Rossi de Campos, M.Sc
Back to the Top
There are a few times when you quantitate using standards that do not
linear response (fluorescence is one example), but I am not sure of a
justification for doing this if you are using standard UV or MS
systems. Try to decrease the range of your standard curve. This
helps with linearity problems. You can dilute your samples differently
(doing the appropriate dilution QCs) to have all of your samples fit
within your standard curve range.
Best of Luck
Mr. Campos,Back to the Top
I am not a pharmacokineticist so I am venturing onto unfamiliar turf.
But you are describing chromatographic analysis withan analyte
detector, say UV or refractive index, for example; and there is no
reason to always believe that the detector response is linearlyrelated
to concentration. Many times it is not and the empirical calibration
relationship adequately describes the analytical system. Performing
analyses within the observed range ofan observed
empiricalrelationship is reliable. Of course, we expect this
relationship to hold every time we perform the method.
Stanley L. Alekman PhD
S.L. Alekman Associates, Inc.
Dear Mr. Rossi,Back to the Top
The problem of linearity of the calibration curves and how to evaluate
discussed in detail in this paper.
Non-linear heteroscedastic regression model for determination of
methotrexate in human plasma by high-performance liquid chromatography,
Sadray, S. Rezaee and S. Rezakhah, Journal of Chromatography B, Vol 787,
2003, pp 293-302
Dr. Sima Sadray
I neglected to make clear in my first answer that the only statisticalBack to the Top
test to support linearity is the Lack-of-Fit testwhich requires
independent replicates for each calibrator. Correlation and regression
coefficients, and the value of the standard error,are not evidence for
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (email@example.com)