Avian Influenza Epidemiology

Training Workshop in Indonesia

Workshop Notes

Concepts & Applications of Clinical and Serological testing

Introduction
Clinical tests are used to serve several purposes in veterinary epidemiology. For example, the type of survey most commonly done in veterinary epidemiology, a prevalence survey, is employed to determine the frequency and distribution of some infectious agent within a population of animals. In a prevalence survey, the clinical test usually used, a serological test, measures the occurrence of antibody that is produced against the infectious agent of interest in the serum of the animal. A serological test is a laboratory test used to detect or to measure the amount of antibodies or antigens in blood sera. Serum antibodies indicate a humoral immune response to a specific infectious agent or foreign antigen.

Screening & Diagnostic Tests

Clinical tests can be applied for screening purposes. Screening is the presumptive identification of unrecognized disease to sort out apparently healthy animals, which may have the disease, from those that probably do not have the disease. Therefore, screening tests, whether they are serological tests, metabolic profiles, cell mediated immuneresponse assays, or physical measurements, are applied to apparently healthy animals in search of disease. Screening tests are not meant to be diagnostic. As such, a screening test is designed to be innocuous, rapid, and inexpensive. Screening tests are usually done with the objective of early disease detection unlike prevalence surveys, which are done to measure the amount of disease in a population. As a general rule, screening tests are applied to a large number of animals and often are followed by additional diagnostic tests on those animals that are found to be positive. The additional diagnostic tests serve as confirmatory tests. When a screening test is applied to a high risk group of animals, it is thought of as case finding. A high risk group includes animals that are suspected or known to have a higher prevalence of the disease compared to the total population. These animals may be at a higher risk due to factors such as age, location, use, or an increased exposure to disease agents. In conclusion, the aim of a screening test is usually early detection, which leads to prevention, early treatment, and/or control of the disease of interest. The aim of a screening test is usually for early detection, which leads to prevention, early treatment, and/or control.

Sources of Uncertainty
Often, the result of a clinical test is used to confirm or diagnose a disease in an animal or to screen a population of seemingly healthy animals to determine if an animal is diseased
Page 1 of 8

Serological Testing

or not diseased. Ideally, a clinical test should be accurate, precise, and be able to determine the disease status of an animal without error. Unfortunately, all tests are not equal in accomplishing these tasks. Some tests have inherent biases in them. Some tests are not very repeatable. Moreover, not all tests are equal in differentiating normal (not diseased) from abnormal (diseased). As a result, knowing as much as possible about a specific screening or diagnostic test is of paramount importance for proper interpretation of results. If a given test's weak points are understood, classification of the animal as diseased or not diseased can be enhanced or qualified. The result of a serological test is classified as either positive or negative; however, a positive result does not mean necessarily that the animal has been recently infected with the agent in question and is currently diseased. A positive result also can indicate an incubating or a recovering animal. Moreover, a positive result may indicate a prior vaccination to the agent of interest or passive antibody transfer, such as through the colostrum. A serological test simply cannot differentiate among them. Another bias of the test is seen in that occasionally, antibodies to another agent that has infected the animal will cross-react in tests used to determine exposure to a different agent. On the other hand, a test may be negative when the animal is actually infected. For example, an animal may have been infected recently at the time of the test, so sufficient time to develop an antibody response is not present. Also, a test may not be finely tuned and able to detect small quantities of antibody to an agent. In addition to the above factors, laboratory instrumentation and sample-handling errors can impact the test results. As can be seen, a variety of factors can influence test results and the ability to correctly classify an animal as diseased or not diseased.

Properties of Tests
Accuracy of a Test Ideally, a clinical test should be accurate, precise, and be able to determine the disease status of an animal without error. Unfortunately, all tests are not equal in accomplishing these tasks. Accuracy is the ability of a test to correctly identify an animal as diseased. If a test does not correctly identify an animal as diseased or not diseased, a problem of misclassification exists. The accuracy of a test can be measured and expressed in terms of test sensitivity (Se) and specificity (Sp). Sensitivity is defined as the probability of a test to identify correctly those animals that are infected. Specificity is defined as the probability of a test to identify correctly those animals that are not infected. Example Calculations Knowing as much as possible about a specific screening or diagnostic test is of paramount importance for proper interpretation of results. Thus, it is important to know the accuracy of a test, the amount of misclassification, and its sensitivity and specificity. Sensitivity and specificity can be calculated for a given test by comparing the results obtained with the true disease status of an animal, which is determined by a Gold

Page 2 of 8

Serological Testing

Standard test or by other means. The results can be tabulated in a 2-by-2 table from which sensitivity and specificity can be calculated. For example, below is a two-by-two table for a generic disease: Status Positive Test result (T+) Negative Test result (T-) Total Diseased (D+) 18 TP 38 FN 56 Not Diseased (D-) 15 FP 78 TN 93 Total 33 116 149 N The definitive diagnosis of the disease as determined by a standard method (i.e. viral/bacterial isolation), named the Gold Standard, and is designated by D+ & D- in the above table. The positive and negative results listed above are those determined by a screening test. The following information can be obtained from the table: 1. The total number positive (T+) according to the screening test is 33. 2. The total number negative (T-) according to the screening test is 116. 3. The total number of diseased animals (D+) as determined by the Gold Standard method is 56. 4. The total number of non-diseased animals (D-) as determined by the Gold Standard method is 93. Those animals that are misclassified by the screening test in relation to the Gold Standard method are: 1. False positive (FP): those animals that are not diseased but are positive to the screening test. (15) 2. False negative (FN): those animals that are diseased but are negative to the screening test. (38) Those animals correctly identified are: 1. True Positive (TP): those animals that have the disease and are positive on the screening test. (18) 2. True Negative (TN): those animals that do not have the disease and are negative on the screening test. (78) In the application of most screening tests, true positives and true negatives are not known and the consequences of misclassification must be understood. Interpretation of misclassified results depends upon the purpose of the test and upon the person who does the interpretation. To understand test sensitivity, specificity and the consequences of misclassification, several calculations, graphs and explanations follow.

Page 3 of 8

Serological Testing

Calculating the Properties of Screening Tests

True vs. Apparent Prevalence:

The true prevalence of disease is the proportion of animals tested that have the disease condition of interest as determined by the Gold Standard. true prevalence = D+ N (total # of animals tested)

For our generic disease, the true prevalence of disease would be 56/149. The apparent prevalence of disease is the proportion of animals that test positive according to the screening test. This is also known as the test positive rate. apparent prevalence = T+ N

For our generic disease, the apparent prevalence would be 33/149.

Accuracy vs. Misclassification:
The calculations for the generic disease will be shown beside the equation in parentheses for the following calculations.

The accuracy of the test is the proportion of those animals whose disease status was correctly identified by the test. accuracy = TP + TN (96/149) N (total # of animals tested) Misclassification is the proportion of those animals whose disease status was not correctly identified by the test. misclassification = FP + FN N
Sensitivity (Se) vs. Specificity (Sp):

(53/149)

Se: the proportion of truly diseased animals (D+) that the test correctly identifies (those among the diseased population that tested positive). Se = TP = TP D+ TP + FN (18/56 or 32%)

Thus, in our generic disease example, 32 out of 100 infected animals will be positive on the screening test. Sp: the proportion of the truly non-diseased animals (D-) that the test correctly identifies (those among the non-diseased population that tested negative). Sp = TN = TN DTN+FP (78/93 or 84%)

Page 4 of 8

Serological Testing

Thus, in our generic disease example 84 out of 100 non-diseased animals will be negative on the screening test. If the Se and Sp are known, the true prevalence of disease can be calculated: true prevalence = apparent prevalence + Sp - 1 Sp + Se - 1
Proportions of False Positive and False Negative:

The proportion of false positives is equal to the proportion of truly non-diseased animals (D-) that the test misidentifies as positive. Proportion of false positive= FP TN + FP (those animals without disease = D-) The proportion of false positives for the generic disease is 15/93 or 16%. Thus, on the average 16 out of 100 truly non-diseased animals will have a positive test. The proportion of false negatives is equal to the proportion of truly diseased animals (D+) that the test misidentifies as negative. Proportion of false negative= FN TP + FN (those with the disease = D+) The proportion of false negative for the generic disease is 38/56 or 68%. Thus, on the average 68 out of 100 truly infected animals (D+) will have a negative test (T-). Predictive Values: The positive predictive value (PV+) is an important property of a screening test. The positive predictive value indicates what proportion of the test positive animals are really infected (D+). It is the probability that a positive test result is correct. PV+= TP T+ = TP TP + FP (18/33 or 54%)

This means that there is a 54% chance that an animal is truly infected if it has a positive test result. The negative predictive value (PV-) is the probability that a negative test result is correct. PV- = TN = TN T- TN+FN (78/116 or 67%)

This means that there is a 67% chance that an animal is not infected if it has a negative test result.

Page 5 of 8

Serological Testing

Predictive Values & Sensitivity/Specificity

PV+ is closely related to specificity and PV- is closely related to sensitivity. The predictive values indicate the test accuracy, given that the test result is known. Both predictive values depend on the prevalence of the disease in the population and the Sp and Se of the test used. Let us consider another way to calculate predictive values: PV+ = Se x Prevalence of diseased (True Prevalence) (Se x Prevalence) + (FP x Prevalence of non-diseased) PV- = Sp x Prevalence of non-diseased (Sp x Prevalence of non-diseased) + (FN x Prevalence of diseased)

The predictive values of positive and negative test results vary directly with the prevalence of disease when the Sp and Se are held constant. The following table illustrates this:

The Importance of Disease Prevalence The predictive value of a positive test increases and the predictive value of a negative test decreases as the prevalence of the disease increases as shown in the diagram below. In other words, as more animals in a population are infected, there is a greater the probability of an animal being truly infected if it has a positive test result.

Page 6 of 8

Serological Testing

Relationships among Sensitivity, Specificity and Predictive Values Specificity is related to PV+ in the following way:

As the specificity increases, the PV+ increases, and therefore the probability of a positive result being true increases. Specificity allows more confidence in a positive test. Sensitivity is related to PV- in the following way:

As the sensitivity increases, the PV- increases and therefore the probability of a negative result being correct increases. Sensitivity allows more confidence in a negative test. Choosing Between Tests 1. Use a test with a high Se and high PV- when: a) it is advantageous to "rule out" a diagnosis in the early stages of a diagnostic workup to decrease the possible number of animals to treat. Since there is more confidence in a negative test, this will allow confidence that those not treated (because they tested negative) will not spread the disease. b) a FN is dangerous. For example, a FN of an animal entering this country with a Foreign Animal Disease would have serious consequences.
Page 7 of 8

Serological Testing

2. Use a test with high Sp and high PV+ when: a) it is advantageous to confirm a diagnosis. Since there is more confidence in a + test, this allows those that should be treated to be confidently determined. b) a FP is dangerous. For example, if test and slaughter is the measure taken following a positive result, the cost of too many FPs could be quite high. If screening tests are performed with the purpose of identifying cases for treatment, it is desirable for the test to have a high PV+. Otherwise a large proportion of animals would be treated or slaughtered unnecessarily. Since screening aims to find potential cases of one specified disease (i.e. in the form of a diagnosis), the PV of a positive test can be termed its diagnositability. However, it is also very desirable for screening tests used in the early stages of a control program to be highly sensitive (so that there are few FNs) and the test used in the latter stages to be highly specific (to decrease FPs). This is especially true when prevalence is low (2%), when most of the animals are free of the disease and the results of even a highly sensitive and specific test will include a large number of false positives. Testing in Series and in Parallel Testing in series: When screening tests are run in series, the results of every test run must be positive, otherwise the animal is considered negative for disease. Therefore, only positive samples from a preceding test will be tested with the next test. For testing in series, the desirable outcome is to increase the overall Sp and PV+ so that there is confidence in the final positive result. Testing in parallel: When screening tests are run in parallel, the results of every test must be negative, otherwise the animal is considered positive for the disease. Therefore, only negative samples from a preceding test will be tested with the next test. For testing in parallel, the desirable outcome is to increase the overall Se and PV- so that there is confidence in the final negative result. Test Batteries: A test battery means that all the available tests and panels are run to screen for a particular disease. The more tests that are administered, the greater the probability of a false positive.

Page 8 of 8