89 / 1195
DuPont™ BAX
®
System Real-Time PCR Assay for
Salmonella
: Collaborative Study
DuPont Nutrition & Health
Page 11
POD analysis (the 95% confidence interval of the dLPOD included 0 in all cases). Two orange juice
1
samples (one from each of two collaborator sites) returned a presumptive positive result with the test
2
method but could not be culture confirmed. One sample indicated a very weak positive result,
3
suggesting either a cross-contamination event (most likely during a sample transfer step) or a very low
4
target cell density in the sample, which could be detected with the PCR method but was difficult to
5
detect by culture. The second sample returned a strong positive result with the test method, so it is
6
unclear what caused the discordant results between the test and reference methods. The remaining 502
7
orange juice samples tested from the alternative enrichment were in agreement with culture
8
confirmation from the alternative enrichment broths.
9
The results for frankfurters are summarized in Tables 4 – 6. At each inoculation level, the BAX® System
10
method and the reference method demonstrated no significant statistical difference as indicated by
11
POD analysis (the 95% confidence interval of the dLPOD included 0 in all cases). Two frankfurter
12
samples, both from the same collaborator site, returned a presumptive positive result with the test
13
method but could not be culture confirmed. Both samples indicated a very weak positive result,
14
suggesting either a cross-contamination event or a very low target cell density in the sample, which
15
could be detected with the PCR method but was difficult to detect by culture. The remaining 502
16
frankfurter samples analyzed with the alternative method were in agreement with culture confirmation
17
results. One sample initially returned an indeterminate result with the test method and was re-tested
18
according to the manufacturer’s instructions. Upon retest, this sample returned a negative result, which
19
was in agreement with culture confirmation results.
20
A POD summary of all test method results is shown in Table 7. Across all 3 inoculation levels for both
21
matrices, statistical analysis indicates that the test method presented demonstrates no significant
22
differences from the reference methods. The within laboratory component (S
r
) of the reproducibility S
R
23
value represents the sampling variability at very low spiking levels. It accounted for all of the S
R
value
24
observed for each matrix collaboratively studied, the S
L
value (between laboratory effect components of
25
S
R
) being zero in both data sets at each partial response spike level. This acceptable inter-laboratory
26
reproducibility is supported by the insignificant homogeneity test P
T
values (>0.1), which suggest that
27
the laboratory POD values are not significantly different when allowance is made for the sampling
28
variability. While interpretation of this latter test is subject to the study design, 10 or more laboratories
29
with 12 replicate sample portions per level for each of three levels (high, low and unspiked) per
30
laboratory is deemed adequate for such studies.
31
The graphical representation of the data (Fig. 1) demonstrates that the dose response curve for each
32
matrix encompasses the partial response region required for qualitative detection method analysis. The
33
95% confidence interval of each dPOD value determined at each concentration contains zero, which is
34
indicative of no significant difference between the candidate and reference methods and between the
35
candidate presumptive result and candidate confirmed result.
36
37
38
FINAL (Version 4)