Table 3.
Summary of statistics generated using the POD model and other models for the detection of Salmonella Infantis ATCC 51741 in 25 g dry dog food test portions by the LAMP method (with LB preenrichment) versus the BAM reference method in a paired study design.
| Parameter and comparisons | Combined POD and associated statistics (lower control limit, upper control limit) among dry dog food test portions | ||
|---|---|---|---|
| Uninoculated (0 MPN/25 g) | Low-level (0.65 MPN/25 g) | High-level (3.01 MPN/25 g) | |
| Statistics generated using the POD modela | |||
| LB-LAMP positive/total number | 1/96 | 58/98b | 94/94 |
| LPOD | 0.01 (0.00, 0.06) | 0.59 (0.46, 0.72) | 1.00 (0.96, 1.00) |
| sr | 0.10 (0.09, 0.19) | 0.47 (0.41, 0.53) | 0.00 (0.00, 0.19) |
| sL | 0.00 (0.00, 0.05) | 0.15 (0.00, 0.35) | 0.00 (0.00, 0.19) |
| sR | 0.10 (0.09, 0.12) | 0.50 (0.44, 0.53) | 0.00 (0.00, 0.27) |
| P-value | 0.4336 | 0.0726 | 1.0000 |
| BAM positive/total number | 0/96 | 58/98b | 94/94 |
| LPOD | 0.00 (0.00, 0.04) | 0.59 (0.47, 0.72) | 1.00 (0.96, 1.00) |
| sr | 0.00 (0.00, 0.19) | 0.48 (0.41, 0.53) | 0.00 (0.00, 0.19) |
| sL | 0.00 (0.00, 0.19) | 0.14 (0.00, 0.33) | 0.00 (0.00, 0.19) |
| sR | 0.00 (0.00, 0.27) | 0.50 (0.44, 0.53) | 0.00 (0.00, 0.27) |
| P-value | 1.0000 | 0.0974 | 1.0000 |
| Comparisons based on the POD model and other statistical modelsc | |||
| dLPOD (LB-LAMP vs. BAM) | 0.01 (-0.03, 0.06) | 0.00 (-0.18, 0.18) | 0.00 (-0.04, 0.04) |
| P-value (LB-LAMP vs. BAM) | 0.99 (f) | 1 (1) | N/A (N/A) |
a LPOD is a composite POD across collaborators and includes between-collaborator variation in addition to variation inherent in the binomial nature of the binary probabilities. sr is repeatability standard deviation, sL is among-collaborator standard deviation, sR is reproducibility standard deviation. P-value is homogeneity test of laboratory (collaborator) PODs. b Collaborator 13 had one positive sample by LB-LAMP only, while collaborator 8 had one positive sample by BAM only. c dLPOD is the difference in LPOD between two methods. The numbers in parenthesis are 95% confidence interval (lower control limit [LCL], upper control limit [UCL]) estimates on dLPOD. A confidence interval for dLPOD that does not contain 0 indicates a statistically significant difference between the two methods being compared. Obuchowski’s modified McNemar’s test and a conditional logistic regression model (numbers in parenthesis) were used for LB-LAMP vs. BAM comparisons. f indicates that model fitting failed to converge. N/A, no test was done because of complete match of the results.