Diagnostic Tests: Types and Importance

1. Diagnostic Tests Dr. Fortune Sithole

2. Types of Diagnostic Tests

3. Definition of a Diagnostic Test • Broadly speaking: diagnostic tests provide information thataids in making a specific diagnosis. • Diagnosis = classify animal(s) as having a disease (infection, condition, etc.) or not • Many uses: • Clinical medicine: used for Txand prognosis • Surveillance: ID changes in dz status in a population • Border security: keep dzed animals out • Research: identify research subjects

4. Diagnostic Test Results Can Be Dichotomous or Continuous • Dichotomous = only two possible answers • Positive or negative • Image on a radiograph, MRI • Presence of an organism, lesion, fracture, etc. • Continuous = a continuum of possible answers • Serum chemistry • Blood glucose test results • Antibody titres • CBC, differential

5. Type of Diagnostic Tests • Clinical examination • Post mortem examination • Radiography, ultrasound, MRI • EKG • Hematology tests • Biochemical tests • Immunological tests • Pathogen-detection tests

6. HematologyDiagnostic Tests • Measures: amount of different cell types in a venous blood sample • Often, any given test is not specific to any one disease. • Suggestive—presumptive diagnoses, limit Df Dx list • Consortia of test results can lead to diagnosis • Examples • Complete blood count • Differential blood count • PCV

7. Biochemical Diagnostic Tests • Measures: enzymes, metabolites, proteins, usually in venous blood. • Mostly = clinical pathology • Often, any given test is not specific to any one disease • Suggestive—presumptive diagnoses, limit Df Dx list • Consortia of test results can lead to diagnosis • Examples of biochemical tests • Liver enzymes, BUN, creatinine, amylase, lipase, etc. • Acute phase proteins, hormone assays

8. Immunological Diagnostic Tests • MEASURES: antibodies • ELISA: Enzyme linked immunosorbent assay • IFA (aka IFAT, MIF) immunoflourescent antibody • Western blot • SNAP tests for FIV and E. canis • Using antibodies to detect antigen • SNAP tests for FeLV and heartworm • Immunohistochemistry: BVD virus detection using ELISA • Antibodies are produced in response to infection • May not be present in the initial phase of the disease • Antibodies usually persist after the animal has recovered

9. Pathogen Detection Tests • Detect the pathogen itself (not antibodies against it) • Present during the infection, with no time lag • Some examples: • Direct visualization • Virus isolation • Bacteria culture • PCR • Some tests used for epidemiologic investigations fingerprintpathogens or discriminate infected from vaccinated animals (DIVA) PCR is a method for detecting pathogen DNA in a clinical sample

10. How diagnostic tests work:How a positive or negative decision is made

11. Most Diagnostic Tests (even Dichotomous ones) Measure Something… • Usually can’tmeasure the disease itself • What is measured: • Something (chemical, antigen, etc.) that is present when an animal has a certain dz • What is measured = Test value • Hematology—amount of different blood cells • Biochemical—amount of enzymes, metabolites, proteins • Immunological—amount of antigens or antibodies

12. How it Works How it Works Sample A sample is taken from an animal Test Value: the test is run on the sample and a measurement (test value) is produced Test Result: a decision (positive or negative) is made by the test or the diagnostician Test Value Test makes decision if result is positive or negative Diagnostician decides if result is positive or negative

13. How is the decision made? • A cut-off value is used • Normal versus abnormal blood glucose • Positive versus negative antibody titers • The cut-off is determined experimentally, as the test value that minimizes false positive/negative results NEGATIVE POSITIVE

14. How do we determine when a test value should be positive or negative? • Test a bunch of individuals that are provenbeyond a doubt to have the condition or disease or infection. • Apply the test and record the test values. • Test a bunch of individuals that are provenbeyond a doubt to not have the condition or disease or infection. • Apply the test and record the test value. • Define a cut-off value that bestseparates the two groups.

15. Each Bar the count of the number of individuals tested that had a specific test value… as indicated on the x-axis Ideal Diagnostic Test Dzed Non-dzed Number 0 High Test Value These two charts are Frequency Distributions on the same X axis

16. Ideal Diagnostic Test—Perfect Test Cut-off Value A test value that best separates healthy from dzed individuals Cut Off Test Negative Test Positive Dzed Non-dzed Number 0 High Test Value

17. Ideal Diagnostic Test—Perfect Test Cut Off Test Negative Test Positive Dzed Non-dzed Number 0 High Test Value

18. Diagnostic Test Reality Cut-Off Usually, there is not a clear separation in the test values between dzed and non-dzed animals. Some non-dzed animals may have higher test values than some of the dzed animals, and some dzed animals may have lower test values than some of the non-dzed animals. Cut Off Test Negative Test Positive Non-dzed Dzed Number 0 High Test Value

19. Diagnostic Test Reality Cut Off Test Negative Test Positive Non-dzed Frequency There may be false positives. Non-dzed animals that test positive 0 High Test Value

20. Diagnostic Test Reality Cut Off Test Negative Test Positive Dzed Number There may be false negatives. Dzed animals that test negative 0 High Test Value

21. Diagnostic Test Reality Cut Off Test Negative Test Positive There may be suspects Too close to the cut-off value to be classified Dzed Non-dzed Number 0 High Test Value

22. How Diagnostic Tests Are Evaluated

23. Definitions for diagnostic test lectures The true health status of the animal: • Non-diseased (Non-dzed): Does not have the dz or infection or condition of interest • Diseased (dzed): Has the dz or infection or condition of interest Test Results for animals: • True positive: Dzed animal that tests positive • False negative: Dzed animal that tests negative • True negative: Non-dzed animal that tests negative • False positive: Non-dzed animal that tests positive

24. Evaluating Diagnostic Tests Every test is evaluated by how accurately it classifies: • Dzed individuals as TEST POSITIVE • Non-dzed individuals as TEST NEGATIVE How to evaluate a test: • Test a bunch of dzed animals: individuals that have the condition, disease, or infection • Test a bunch of non-dzed: individuals that do not have the condition, disease, or infection • A gold standard test is often used to determine the true disease status of individuals used for the evaluation • Gold standard tests are very accurate, reliable tests • But, gold standard tests are often very labor intensive, impractical, slow, and expensive. • Establish a cut-off value

25. Evaluating Diagnostic Tests We use two parameters to evaluate diagnostic tests: • Sensitivity (Se) • Specificity (Sp) • Sensitivity: Estimated by testing dzed animals • Specificity: Estimated by testing non-dzed animal Very Important Concepts Know these very well!

26. Sensitivity (Se) • The proportion of Dzed animals that the test correctly classifies as positive • Calculate: # dzed that test positive/# dzed tested • If you test 100 dzed animals and 95 of them are positive by the test, sensitivity = ____________ • The 5 dzed animals that test negative are called false negatives • Sensitivity can be low when: • There are very few/small amounts of whatever the test measures—low antibody titers, for example • Samples degraded before the lab received them 95 / 100 = 95%

27. Specificity (Sp) • The proportion of non-dzed animals the test correctly classifies as negative • Calculate: # non-dzed that test negative/# non-dzed tested • If you test 100 non-dzed animals and 98 of them are negative by the test, specificity = __________ • The 2 non-dzed animals that test positive are called false positives • Specificity can be low when: • Samples are contaminated • One pathogen is identified mistakenly as another one—antibody cross-reactivity, parasite misidentification • Animals have antibodies due to vaccine, not infection 98 / 100 = 98%

28. Yet another way of saying it… • Sensitivity is how sensitive the test is at identifying a dzed animal • For a very sensitive test • Most dzed animals will test positive • Very few dzed animals will test negative • Produces very few false negative test results • Most negative results are from non-dzed animals • SnOUT = Highly sensitive test rules disease out • A negative test result from a highly sensitive test is most likely from a non-dzed animal: so it rules out disease

29. Yet another way of saying it… • Specificity is how often the test detects the thing it is specificallyintended to detect—not healthy animals! • For a very specific test • Most non-dzed animals will test negative • Very few non-dzed animals will test positive • Produces very few false positives test results • Most positive results are from dzed animals • SpIN: Highly specific test rules dz in • A positive test result from a highly specific test is most likely from a dzed animal, so it rules disease in

30. When to use Se and Sp? When selecting a diagnostic test • Always consider both the Se and Sp • Select a test with Se and Sp that meets your requirements

31. High Se: 100% and High Sp: 100% Cut Off Test Negative Test Positive Ideal Use for any purpose This rarely happens. But, some tests are very close Non-dzed Dzed Frequency 0 High Test Value

32. Med Se: 90%, Med Sp: 90 Cut Off Balanced Test Equal Se and SP Equally misclassifies dzed and non-dzed animals Test Negative Test Positive Non-dzed Dzed Frequency 0 High Test Value

33. Maximize sensitivity when… • You need to detect alldiseased or infected animals • You don’t care if there are a lot of false positives (you don’t care if test has poor specificity) • You don’t want any false negatives • A test that you would use to ensure that imported animals are not infected with a disease • Use a test with very high Se and only allow animals that test negative to enter the country • Don’t care about false positives, because no animals with positive test results will come into the country • A marginal test to enter Canada from the United States

34. High Se: 100%, Low Sp: 75% Highly Sensitive Test Correctly classifies ALL dzed animals as POSITIVE BUT May have a lot of False Positive test results if the Sp is low Cut Off Test Negative Test Positive Only allow test negatives into the country Non-dzed Dzed Frequency 0 High Test Value

35. Maximize specificity when… • The cost of a false positive is high • You don’t care if there are a lot of false negatives (i.e., you don’t care of the test has poor sensitivity) • Tests for very serious diseases with serious consequences and where informing the individual about having the diseases can cause serious stress • AIDS, Hepatitis C, cancer

36. Low Se: 75%, High Sp: 100% Cut Off Highly Specific Test Correctly classifies ALL non-dzed animals as negative NO Healthy animals are Test Positive BUT May have a lot of false negative test results if the Se is low Test Negative Test Positive Non-dzed Dzed Frequency 0 High Test Value

37. To accurately detect all dzed animals and eliminate the non-dzed animals, use two tests in series. 1ST test: High sensitivity • Detects most or all the dzed animals in the population…you won’t miss any! • Negatives are all true negatives! • But, you will have some false positive test results • Test positives are a mix of true and false positives 2nd Test: High specificity • Test all animals that test positive on the first test • High specificity means no false positive test results are produced…eliminates all the false positives from the first test • All test positives from the second test are true positives

38. How to interpret test results

39. Se and Sp don’t tell you the probability that an animal which tests positive in your clinic is truly diseased • Se and Sp help, because they tell you how good your test is. • BUT: • Se is the probability of a dzed animal testing positive • Sp is the probability of a non-dzed animal testing negative • You want to know: • The probability that a test positive animal is dzed, the opposite of Se!

40. Se and Sp don’t tell you the probability that an animal which tests positive in your clinic is truly diseased PPV and NPV do that. PPV: positive predictive value • The predictive value of a positive test result NPV: negative predictive value • The predictive value of a negative test result

41. Positive Predictive Value (PPV) • Answers the question: “My patient tested positive. What is the probability that it actually has the disease?” • The proportion of test positiveanimals that are dzed • You have 97 animals that test positive; 95 of those are diseased, and 2 are false positives. What is the PPV? ___________ • What if 84 have the disease and 13 are false positives? ___________ 95 / 97 = 97.9% 84 / 97 = 86.6%

42. Positive predictive value Cut Off Only concerned with positive test results Includes true positives and false positives Test Negative Test Positive Dzed Frequency True Positives PPV = Test Positives Non-dzed 0 High Test Value

43. Negative Predictive Value (NPV) • Answers the question: “My patient tested negative. What is the probability that my patient does not have the disease?” • The proportion of test negativeanimals that are non-dzed • If 103 animals test negative, but only 98 of those are non-dzed (5 are false negatives), what is the NPV? ___________ • What if only 34 of these are non-dzed (69 are false negatives)? ___________ 98 / 103 = 95.1% 34 / 103 = 33.0%

44. Negative predictive value Cut Off Test Negative Test Positive Only concerned with negative test results Includes true negative and false negatives Non-dzed Frequency True Negatives NPV = Test Negatives Dzed 0 High Test Value

45. A word of warning • Note that PPV and NPV change as the prevalence of the disease changes! • PPV depends upon the proportion of true positives and false positives. • If a disease is very rare, the number of false positives can easily outnumber the true positives (low PPV). • NPV depends upon the proportion of true negatives and false negatives. • If a disease is very common, the number of false negatives can quickly outnumber the true negatives (low NPV).

46. Important • Se and Sp are properties of the test and won’t change if the prevalence of disease in the population changes. • PPV and NPV are determined by the Se and Sp of the test and the prevalence of disease in the population. PPV and NPV will change as the prevalence of disease in the population changes.