Chemistry Reference
In-Depth Information
medication, or other analytes of interest. The FDA guidelines recommend the analysis
of blank samples of the appropriate biological matrix from at least six different sources.
For example, six different sources of plasma should be spiked with known concen-
trations of analyte at the lower limit of quantification (LLOQ) to show that accurate
results can be obtained. Similarly, a blank extract of each matrix should be analyzed to
show the absence of interferences. In cases of a rare or difficult-to-obtain matrix (e.g.,
plasma from an exotic species or human tissue), the six-matrix requirement is relaxed.
7.5.2.2 Accuracy, precision, and recovery
The
accuracy
of a bioanalytical method is defined as the closeness of test results
to the true value as determined by replicate analyses of samples containing known
amounts of the analyte of interest; results are reported as deviations of the mean
from the true value. The FDA guidelines recommend the use of a minimum of five
determinations per concentration, and a minimum of three concentrations over the
expected range (a minimum of fifteen separately prepared samples). The guidelines
further recommend that the mean value be within ±15% of the actual value except at
the LLOQ, where ±20% is acceptable.
The
precision
of a bioanalytical method measures agreement among test results
when the method is applied repeatedly to multiple samplings of a homogeneous
sample. As in recent ICH guidelines, precision can be further divided into repeat-
ability (within-run or intrabatch) determinations, and intermediate (between-run or
interbatch) precision [2]. The FDA guidelines recommend the use of a minimum of
five determinations per concentration, and a minimum of three concentrations over
the expected range. The imprecision measured at each concentration level should not
exceed 15% RSD, except for the LLOQ, which should not exceed 20% RSD. Usually,
the same data is used to determine both precision and accuracy.
The assay recovery relates to the extraction efficiency and is determined by a
comparison of the response from a sample extracted from the matrix to the refer-
ence standard (with appropriate adjustments for dilution, etc.). The recovery of the
analyte does not necessarily need to be 100%, but it must be quantitative. That is, it
should be precise and reproducible. Recovery experiments should be carried out at
three concentrations (low, medium, and high), with a comparison of the results for
extracted samples versus unextracted samples (adjusted for dilution). Sometimes, it
is impractical to analyze unextracted samples (e.g., injection of unextracted plasma
will ruin most HPLC columns), so creative ways to show recovery may need to be
devised. For example, a liquid-liquid extraction of spiked matrix might be com-
pared to extraction of a matrix-free aqueous solution; or recovery from a solid-phase
extraction might be determined by calculation of volumetric recovery and compari-
son of the response from an extracted sample to a known concentration of reference
standard.
7.5.2.3 calibration/standard curve
A calibration curve or standard curve illustrates the relationship between the instru-
ment response and the known concentration of the analyte, within a given range
based on expected values. The simplest model that describes the proportionality
should be used (e.g., a linear fit is preferred over quadratic curve-fitting function).
