1、Predictive ValuesBob L.Larson,DVM,PhD University of MissouriCommercial Agriculture,Beef Focus Team1Use and Interpretation of Diagnostic Tests Help to make a diagnosis(does not stand alone)Judge severity of disease Predict subsequent clinical course of disease Estimate likely responsiveness to therap
2、y Determine response to therapy(post therapy)1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests2Diagnostic Tests Must be tested against a“gold standard”Although no perfect test exists,a gold standard is an accepted,reliable test Properties of diagnostic t
3、ests Sensitivity=proportion of true positives that test positiveHow effective is the test at identifying animals with the condition?Specificity=proportion of true negatives that test negativeHow effective is the test at identifying animals without the condition?1.Measurement of disease events2.Infec
4、tious disease epidemiology3.Interpreting diagnostic tests3Diagnostic TestsDisease StatusPresentAbsentTest PositiveTest Negative1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests4Diagnostic TestsDisease StatusPresentAbsentTest PositiveTrue PositiveTest Neg
5、ative1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests5Diagnostic TestsDisease StatusPresentAbsentTest PositiveTest NegativeTrue Negative1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests6Diagnostic TestsDisea
6、se StatusPresentAbsentTest PositiveFalse PositiveTest Negative1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests7Diagnostic TestsDisease StatusPresentAbsentTest PositiveTest NegativeFalse Negative1.Measurement of disease events2.Infectious disease epidemi
7、ology3.Interpreting diagnostic tests8Diagnostic TestsDisease StatusPresentAbsentTest Positiveaba+bTest Negativecdc+da+cb+da+b+c+dSensitivity =a/(a+c)1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests9Diagnostic TestsDisease StatusPresentAbsentTest Positiv
8、eaba+bTest Negativecdc+da+cb+da+b+c+dSensitivity =a/(a+c)Specificity=d/(b+d)1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests10Diagnostic TestsDisease StatusPresentAbsentTest Positiveaba+bTest Negativecdc+da+cb+da+b+c+dSensitivity =a/(a+c)Specificity=d/(
9、b+d)False negative=c/(a+c)1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests11Diagnostic TestsDisease StatusPresentAbsentTest Positiveaba+bTest Negativecdc+da+cb+da+b+c+dSensitivity =a/(a+c)Specificity=d/(b+d)False negative=c/(a+c)False positive=b/(b+d)1.
10、Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests12Number of animalsMeasurementDiagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)Population APopulation B1.Measurement of disease events2.Infectious disease epidemiology
11、3.Interpreting diagnostic tests13Number of animalsMeasurementDiagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)Population APopulation B1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests14Number of animalsMeasurementD
12、iagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)Population APopulation B1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests15Number of animalsMeasurementDiagnostic tests use an arbitrary cut-off to separate two popul
13、ations(normal vs.abnormal)Population APopulation B1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests16Number of animalsMeasurementDiagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)Population APopulation B1.Measuremen
14、t of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests17Diagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)NumberMeasurementWhere cut-off is placed determines sensitivity and specificity1.Measurement of disease events2.Infectious dis
15、ease epidemiology3.Interpreting diagnostic tests18Diagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)NumberMeasurementWhere cut-off is placed determines sensitivity and specificityNormalAbnormalFew False NegativesHigh Sensitivity-result+result1.Measurement of di
16、sease events2.Infectious disease epidemiology3.Interpreting diagnostic tests19Diagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)NumberMeasurementWhere cut-off is placed determines sensitivity and specificityNormalAbnormalFew False PositivesHigh Specificity-resu
17、lt+result1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests20Diagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)NumberMeasurementWhere cut-off is placed determines sensitivity and specificityNormalAbnormalTrade off be
18、tween False and False+-result+result1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests21Diagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)NumberMeasurementTest with High SensitivityNormalAbnormalA negative result all
19、ows you to rule-out whatever you are testing-result+resultSnNout1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests22Diagnostic tests use an arbitrary cut-off to separate two populations(normal vs.abnormal)NumberMeasurementTest with High SpecificityNormalA
20、bnormalA positive result allows you to rule-in whatever you are testing-result+resultSpPin1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests23Diagnostic Tests Sensitivity and specificity are inversely related Sensitivity and specificity are properties of
21、the diagnostic test and tell us if the test is valid Test sensitivity and specificity have little effect on the confidence a veterinarian has in an individual result.1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests24Because NumberMeasurementMy illustrat
22、ion is not consistent with true disease dynamics in a population(the real world)1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests25NumberMeasurementBecause My illustration is not consistent with true disease dynamics in a population(the real world)1.Meas
23、urement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests26Predictive ValueWhat we are really interested in Predictive values tell us if a test is useful Positive Predictive Value(PPV)The probability that an animal with a positive test result is truly positive Negativ
24、e Predictive Value(NPV)The probability that an animal with a negative test result is truly negative Dependent on disease prevalence and test sensitivity and specificity1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests27PPV and NPV Assume a test for bovin
25、e Johnes is 60%sensitive and 90%specific On examination of a 5 year old dairy cow,with typical history,clinical signs and positive herd-matesWe assume 90%of cows with this description are positive(prevalence=90%)Dz+Dz-Test Positive54010550Test Negative360904509001001,000Prevalence is 90%(900/1000=90
26、%)1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests28PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%specific On examination of a 5 year old dairy cow,with typical history,clinical signs and positive herd-matesWe assume 90%of cows with
27、 this description are positive(prevalence=90%)Dz+Dz-Test Positive54010550Test Negative360904509001001,000Test Sensitivity=540/900=60%1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests29PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%spe
28、cific On examination of a 5 year old dairy cow,with typical history,clinical signs and positive herd-matesWe assume 90%of cows with this description are positive(prevalence=90%)Dz+Dz-Test Positive54010550Test Negative360904509001001,000Test Sensitivity=60%NPV=90/450=20%1.Measurement of disease event
29、s2.Infectious disease epidemiology3.Interpreting diagnostic tests30PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%specific On examination of a 5 year old dairy cow,with typical history,clinical signs and positive herd-matesWe assume 90%of cows with this description are positive(p
30、revalence=90%)Dz+Dz-Test Positive54010550Test Negative360904509001001,000Test Sensitivity=60%NPV=90/450=20%Test Specificity=90/100=90%1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests31PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%sp
31、ecific On examination of a 5 year old dairy cow,with typical history,clinical signs and positive herd-matesWe assume 90%of cows with this description are positive(prevalence=90%)Dz+Dz-Test Positive54010550Test Negative360904509001001,000Test Sensitivity=60%NPV=90/450=20%Test Specificity=90%PPV=540/5
32、50=98%1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests32PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%specific On examination of a 5 year old dairy cow,with typical history,clinical signs and positive herd-matesWe assume 90%of cows
33、with this description are positive(prevalence=90%)Dz+Dz-Test Positive54010550Test Negative360904509001001,000PPV=540/550=98%NPV=90/450=20%With positive test we are 98%sure she is a true positiveWith negative test we are 20%sure she is a true negative1.Measurement of disease events2.Infectious diseas
34、e epidemiology3.Interpreting diagnostic tests33PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%specific On examination of a 5 year old beef cow,with typical history,clinical signs and unknown herd statusWe assume 50%of cows with this description are positive(prevalence=50%)Dz+Dz-T
35、est Positive30050350Test Negative2004506505005001,000Prevalence is 50%(500/1000=50%)1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests34PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%specific On examination of a 5 year old beef cow,wit
36、h typical history,clinical signs and unknown herd statusWe assume 50%of cows with this description are positive(prevalence=50%)Dz+Dz-Test Positive30050350Test Negative2004506505005001,000Test Sensitivity=300/500=60%1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagno
37、stic tests35PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%specific On examination of a 5 year old beef cow,with typical history,clinical signs and unknown herd statusWe assume 50%of cows with this description are positive(prevalence=50%)Dz+Dz-Test Positive30050350Test Negative20
38、04506505005001,000Test Sensitivity=60%NPV=450/650=69%1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests36PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%specific On examination of a 5 year old beef cow,with typical history,clinical sign
39、s and unknown herd statusWe assume 50%of cows with this description are positive(prevalence=50%)Dz+Dz-Test Positive30050350Test Negative2004506505005001,000Test Sensitivity=60%NPV=450/650=69%Test Specificity=450/500=90%1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting di
40、agnostic tests37PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%specific On examination of a 5 year old beef cow,with typical history,clinical signs and unknown herd statusWe assume 50%of cows with this description are positive(prevalence=50%)Dz+Dz-Test Positive30050350Test Negati
41、ve2004506505005001,000Test Sensitivity=60%NPV=450/650=69%Test Specificity=90%PPV=300/350=86%1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests38PPV and NPV Assume a test for bovine Johnes is 60%sensitive and 90%specific On examination of a 5 year old beef
42、 cow,with typical history,clinical signs and unknown herd statusWe assume 50%of cows with this description are positive(prevalence=50%)Dz+Dz-Test Positive30050350Test Negative2004506505005001,000PPV=300/350=86%NPV=450/650=69%With positive test we are 86%sure she is a true positiveWith negative test
43、we are 69%sure she is a true negative1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests39Predictive Value of a Diagnostic TestPPV and NPV PPV and NPV are greatly influenced by pre-test likelihood of diseasePublished prevalence dataYour clinical skills and
44、 judgmentPositive%Negative%PrePostPrePostExample 190981020Example 2508650691.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests40PPV and NPV The immunohistochemistry test for persistently infected BVD animals is very good:96%sensitive and 99%specific Howeve
45、r,the prevalence of PI animals in the general calf population is low(0.3%)So,how does a very good test perform in a population with low prevalence?1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests41PPV and NPV The immunohistochemistry test for persistent
46、ly infected BVD animals is very good:96%sensitive and 99%specific However,the prevalence of PI animals in the general calf population is low(0.3%)So,how does a very good test perform in a population with low prevalence?BVD PIDz+Dz-Test Positive2889971,285Test Negative1298,70398,71530099,700100,000Te
47、st Sensitivity=96%Test Specificity=99%1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests42PPV and NPV The immunohistochemistry test for persistently infected BVD animals is very good:96%sensitive and 99%specific However,the prevalence of PI animals in the
48、 general calf population is low(0.3%)So,how does a very good test perform in a population with low prevalence?BVD PIDz+Dz-Test Positive2889971,285Test Negative1298,70398,71530099,700100,000Test Sensitivity=96%NPV=98,703/98,715=99.9%Test Specificity=99%1.Measurement of disease events2.Infectious dise
49、ase epidemiology3.Interpreting diagnostic tests43PPV and NPV The immunohistochemistry test for persistently infected BVD animals is very good:96%sensitive and 99%specific However,the prevalence of PI animals in the general calf population is low(0.3%)So,how does a very good test perform in a populat
50、ion with low prevalence?BVD PIDz+Dz-Test Positive288True+997False+1,285Test Negative1298,70398,71530099,700100,000Test Sensitivity=96%NPV=98,703/98,715=99.9%Test Specificity=99%PPV=288/1,285=22%1.Measurement of disease events2.Infectious disease epidemiology3.Interpreting diagnostic tests44PPV and N
