386
EQUINE VETERINARY EDUCATION / AE / july 2022
to be retested alongside the second serum sample, to provide a valid paired test result. This is due to the possibility of there being a between-assay (interassay) variation, which can occur with all serological assays. Therefore, only paired samples run on the same test run
can be reliably compared. The result obtained when the first sample was originally run will be compared to its rerun result as each assay will have an interassay variation value
that is deemed to be acceptable. Occasionally, when the first run and rerun results are compared, they may fall out of this allowable boundary and the test may need repeating to confirm them. Laboratories should have internal quality controls implemented to monitor and help prevent such scenarios.
Diagnostic test accuracy and result interpretation
When a new test is developed, irrespective of the laboratory accreditation, it should undergo stringent analysis to ensure the assay is suitable, optimal and standardised. Setting criteria for a good test involves establishing the following: intentional diagnostic use, standardisation, diagnostic sensitivity and specificity, analytical sensitivity and specificity and reliability (OIE 2013).
Intended use Tests are developed around their intended use and these include the following: confirmation of a disease in a clinical case, checking freedom from infection, use in disease eradication programmes and to establish immune response from vaccination.
Standardisation and optimisation Optimisation of a diagnostic test involves evaluating the assay’s parameters and establishing them at a level to optimise the test for its intended purpose. This process is achieved through assay validation (Saunders et al. 2015). When assays are developed, samples should be taken from the subject population to verify the test. For example, when designing an ELISA for Streptococcus equi (S. equi), samples were taken from known infected horses and known seronegative horses (Robinson et al. 2013). New tests are also compared to previous tests and if available, the gold standard test for validation. Table 2 demonstrates the results obtained from this comparison.
TABLE 2: Diagnostic testing definitions and their calculations New test result
True status of infection/disease
Diseased
Nondiseased Total
Sensitivity Specificity
Positive predictive value Negative predictive value
= =
= =
Positive TP FP
All test positives (TP + FP)
Negative FN TN
All test negatives (FN + TN) Total All diseased
All non-diseased TP + FN + FP + TN
TP/(TP+FN) Proportion of all diseased animals that test positive. A measure of a test’s ability to correctly identify disease cases
TN/(TN+FP) Proportion of nondiseased animals that test negative. A measure of a test’s ability to correctly identify non-disease cases
TP/(TP+FP) Indicates the chance of a case being diseased after a positive test result TN/(TN+FN) Indicates the chance of a case not being diseased after a negative test result
TP, True positive; FP, False positive; TN, True negative; FN, False negative. [Correction added on 02 May 2021 after first online publication: In Table 2, the division sign (/) was erroneously missed for each of the four entries under 'Diseased Nondiseased' column and these have now been added to this current version.]
© 2021 EVJ Ltd
Diagnostic sensitivity and specificity As a clinician, it is fundamental to be aware of a diagnostic test’s limitations when interpreting results. The ultimate test would correctly identify all cases with disease as infected and all disease-free animals as uninfected 100% of the time, but this is invariably an unrealisable expectation in reality. When setting cut-offs (the value to determine whether a result is classed as positive or negative), a compromise is made and determined by the tests intended use, with certain tests classifying those with disease better than others and vice versa (Gilbert et al. 2001). Quantifying the diagnostic accuracy of a test is done by calculating a test’s diagnostic sensitivity and specificity (Dohoo et al. 2010; Wong et al. 2011; Table 2). A test with a sensitivity of 90% will correctly detect 90% of
cases with the disease as positive (TP), but 10% with the disease will be falsely classified as negative (FN). The N in sensitivity can be used to remember that a test with a high sensitivity will create a low number of false negatives. A test with a specificity of 90% will correctly detect 90% of
cases without the disease as negative (TN), but 10% without the disease will be falsely classified as positive (FP). The P in specificity can be used to remember that a test with a high specificity will create a low number of false positives. A screening test should ideally have a high sensitivity, but
with this choice will come a compromise of lower specificity. Any positive results obtained from a screening test should then be clarified using a confirmatory test, which will have higher specificity, but lower sensitivity (Bourgeois and Oaks 2014). This is the case with EIA and EVA testing, with ELISA (high sensitivity but lower specificity) being the screening test for both and Coggins and VN testing (higher specificity), respectively, being the confirmatory tests. If test results are quantitative, a cut-off value for a positive
test result is usually needed and the sensitivity and specificity of a test can be calculated for the different cut-off options. A test’s selected cut-off value is often determined by receiver operating characteristic (ROC) analysis, with the most common criteria utilising a ROC curve. The ROC curve is a graph in which the true-positive rate (sensitivity) is plotted against the false-positive rate (1-specificity) for different values of cut-off points for a parameter. Each point on the ROC curve represents a sensitivity and specificity that corresponds to a particular decision threshold. The cut-off point is usually set at the point on the curve where the sensitivity and specificity of the test are each at their optimum in order to correctly
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88
