53 / 63
1448
K
oshy
et al
.:
J
ournal of
AOAC I
nternational
V
ol
.
99, N
o
.
6, 2016
Results and Discussion
Specificity
A representative chromatogram of the mixed standard
with UV detection is shown in Figure 1. The peaks are well-
resolved and the retention times are very similar to injections
of the individual standards (data not shown). Figures 2 and 3
show chromatograms of a methanolic extract sample and a
raw material sample using a UV detector and a PDA detector.
Although more peaks are observed in the samples compared
with the mixed standard, the six withanolide peaks are still well-
resolved and the retention times and relative retention times of
the six withanolide peaks in the samples are very similar to those
in the mixed standard (Table 6). Furthermore, the PDAspectra of
the withanolide peaks in the samples are identical to the spectra
of the corresponding standards, and the peak purities based on
PDA spectral analysis are >0.99 in the standards and samples
(data not shown). These results indicate no matrix interference
in the LC method. The relative retention times of the analyte
peaks were 0.70 for withanoside IV, 0.89 for withanoside V,
0.92 for withaferin A, 0.96 for 12-deoxywithastramonolide, 1.0
for withanolide A, and 1.15 for withanolide B. These relative
retention times agree with the values from the
United
States
Pharmacopeia
LC method as reported in SMPR 2015.007 (1).
Linearity
The mean peak areas of the linearity standards were plotted
against concentration and are presented in Figures 4–9. Table 7
summarizes the data. After performing linear regression,
linearity was assessed by determining goodness-of-fit (square
of the correlation coefficient, r
2
), by examining residuals over
the concentration range, and by determining the response factor
(peak area/concentration) at each concentration examined. Over
the full range of concentrations examined, r
2
varied from 0.76
for withanolide B to 0.99 for withanoside IV and V. A plot of
the residuals by concentration revealed a convex pattern with
significant residuals as high as 78%. The pattern indicates a lack
of linearity and suggests that alternative regression analyses
(e.g., weighted regression or polynomial regression) should be
evaluated for a better fit to the data over the large concentration
range. Alternatively, the concentration range can be narrowed
to allow for a better approximation of linearity. Using the
latter approach, the data for each analyte were analyzed over a
narrower concentration range until linearity was more closely
approximated. As can be seen in the nonrandom residual plots
of Figures 4–9, even over a narrower concentration range, the
data are not strictly linear. The residuals, however, are much
smaller and tolerable at ≤6.25%. In the linear regression plots,
all r
2
are >0.999 and the
y
-intercept has been greatly reduced
in most cases. Over the narrower concentration ranges, the
response factors, defined as mean peak area divided by
concentration, vary by <5% from the mean response factor. The
ranges for each analyte that approximate linearity are as follows:
withanoside IV 20–330 μg/mL, withanoside V 19–300 μg/mL,
withaferin A 23–184 μg/mL, 12-deoxywithastramonolide
22–178 μg/mL, withanolide A 22–173 μg/mL, and withanolide B
21–169 μg/mL. For routine analyses, the method includes three
concentrations of standards within the linear range.
Precision
Peak areas and retention times.—
Repeatability precision of
standards in methanol based on the measurement of the peak
area is presented in Table 7 and retention time in Table 8.
The RSDs from five replicate injections at each concentration
were all <1%. When a sample, WS/06Lot10, was analyzed
Figure 1. Representative chromatogram of mixed standard solution.
53