Table 2.
Optimisation of results from lipid assay to predict status all 80 PEI PCR-positive samples compared to 80 F5 and F6 negatives, with single antigens and combinations of antigens in R.
| Antigen | Antigen Combination | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MOD 171 | JRRR 121 | RT 237 F2 | SMP 70 | ZAM 295 | ST 123 | ST123 + ZAM295 | ST123 + JRRR121 |
SMP70 + JRRR121 + ZAM295 | ST123 + JRRR121 + ZAM295 | All Six Antigens | MAP IDEXX |
||
| % Pooled Pos | Samples Meeting Cut-Off for Positive Diagnosis | ||||||||||||
| PCR+ n = 80 |
Sens % | 10.0 | 20.0 | 8.8 | 12.5 | 22.5 | 26.3 | 22.5 | 37.5 | 38.8 | 38.8 | 36.3 | 20 |
| Number matching PCR | 8 | 16 | 7 | 10 | 18 | 21 | 18 | 30 | 31 | 31 | 29 | 16 | |
| PEI ‘no-history’ negatives | |||||||||||||
| F5 + F6 n = 80 |
Spec % | 92.5 | 92.5 | 96.3 | 91.3 | 95.0 | 91.3 | 95.0 | 97.5 | 95.0 | 95.0 | 95.0 | 100 |
| Compared to PCR+ | ns | ns | ns | ns | ns | ns | ns | ns | ns | ns | ns | ns | |
| F5 n = 40 |
Spec % | 95 | 97.5 | 95 | 95 | 92.5 | 87.5 | 92.5 | 97.5 | 92.5 | 92.5 | 95 | 100 |
| p (compared to PEI+) | ns | <0.05 | ns | <0.0001 | <0.001 | <0.05 | <0.001 | <0.001 | <0.0001 | 0.0001 | <0.0001 | 0.05 | |
| F6 n = 40 |
Spec % | 90 | 87.5 | 97.5 | 87.5 | 97.5 | 95 | 97.5 | 97.5 | 95 | 97.5 | 95 | 100 |
| p (compared to PEI+) | ns | <0.05 | <0.05 | ns | <0.001 | ns | <0.05 | <0.05 | <0.001 | <0.05 | <0.05 | <0.05 | |
| Using a fixed specificity of 97.5% | |||||||||||||
| PCR+ sens. | Sens % |
5 | 17.5 | 7.5 | 7.5 | 11.3 | 17.5 | 17.5 | 37.5 | 33.8 | 18.9 | 30 | 21.3 |
| Number matching PCR PEI+ | 4 | 14 | 6 | 6 | 9 | 14 | 14 | 30 | 27 | 15 | 24 | 17 | |
The optimal sensitivity and specificity, assuming the 80 PCR positive samples are true positives, for the two sets of ‘no-history’ negatives from farms F5 and F6, and the 80 combined F5 plus F6 PEI samples. Data for sets of antigens are derived by optimising the combinations of antigen responses to provide the best distinction between 80 PCR positive and PCR negative samples using the R statistics package. Significance values were calculated using the Tukey test; ns = not significant. The final column is the result with a commercial IDEXX assay; the number of predicted positives in IDEXX, in comparison to the lipid assay, at 97.5% specificity was calculated by reducing the cut-off to reach that specificity with the 80 F5 and F6 negatives using the data in Table S1. An analysis of the data for other sample cohorts using these cut-offs and antigen combinations optimised for all 80 PCR positives is shown in Table S5.