1. Diagnosis, screening and prevention

2. Puzzle Virus present randomly in 1 in 1000 population Test for virus 99% reliable –i.e. misses 1% of infected individuals and falsely labels 1% of non infected individuals Select 1 individual at random. If the test result is positive, what is the probability that that individual is infected?

3. 1000 1 999 ~ 1 ~ 0 10 989 Probability that that individual is infected = 1 / 10+1 = 9% (= PPV) TP FP FN TN

4. Diagnostic tests Objectives Student should be able to: describe how tests are evaluated define and calculate se\sp\pv and describe their inter-relationships demonstrate the ability to correctly interpret test results utilise prior probability in the judicious selection of diagnostic tests

5. Objectives contd. describe how prior probablilities are derived sources of health information in Ireland / other countries

6. Key points 1. Purpose of testing –to move probability of disease towards 0% or 100% 2. Tests vary in performance (validity) –sensitivity and specificity 3. + ve test result  person has disease –PPV, NPV 4. Same test performs differently in different situation –Diagnostic vs screening contexts

7. 1. Purpose of testing To move probability of disease towards 0% or 100%

9. 2. Test validity Validity measured by se and sp. Sensitivity is the probability of a positive test result in the presence of disease. Specificity is the probability of a negative test result given the absence of disease. Sensitivity and specificity are inversely related.

10. Test validity

11. Sensitivity –probability of a positive test result in the presence of disease = a / a + c Specificity –probability of a negative test result given the absence of disease = d / d + b

12. Se = 99% Sp = 99% Puzzle: Pre-test probability (prevalence) = 1/1000 Test for virus 99% reliable i.e. misses 1% of infected individuals, and falsely labels 1% of non infected individuals

13. Comparison of fine needle aspiration test results with findings from surgical excisional biopsies in women without and with palpable breast masses (from C Smith et al. Surgery 1988 103:178))

14. Comparison of fine needle aspiration test results with findings from surgical excisional biopsies in women without and with palpable breast masses (from C Smith et al. Surgery 1988 103:178))

15. Se = 113/113 + 8 Sp = 181/181+15 = 93%= 92%

16. 3. Positive test result  person has disease PPV: probability of disease, given positive test result= a /a + b NPV: prob. of no disease, given negative test result= d / d + c PPV rises: as specificity increases as prior probability/ prevalence increases

17. PPV NPV Se = 99% Sp = 99% Puzzle: Virus present randomly in 1 in 1000 population. Test 99% reliable Select 1 individual at random: If the test result is positive, what is the probability that that individual is infected?

18. PPV = a/a+b  1/11 = 9.1% NPV = d/d+c = 989/989  100.0% ( Se = 99% Sp = 99% Puzzle: Virus present randomly in 1 in 1000 population. Test 99% reliable Select 1 individual at random: If the test result is positive, what is the probability that that individual is infected?

19. Comparison of fine needle aspiration test results with findings from surgical excisional biopsies in women without and with palpable breast masses (from C Smith et al. Surgery 1988 103:178))

23. Positive predictive values (PPV) are highly dependent on PREVALENCE For a screening test with: 95% sensitivity and 95% specificity: if prevalence is 20% if prevalence is 10% if prevalence is 1% if prevalence is 0.1% PPV is83% PPV is68% PPV is16% PPV is 2%

24. Positive predictive values (PPV) also vary greatly with SPECIFICITY For a disorder with prevalence (pre-test probability) of 1%, & a test with 60% sensitivity: if specificity is 60% if specificity is 80% if specificity is 90% if specificity is 95% if specificity is 99% PPV is1.5% PPV is2.9% PPV is5.7% PPV is 10.8% PPV is 37.7%

25. 4. Same test performs differently in different situations Usefulness varies according to pretest probability (prevalence) of disease. Implications of link between prior probability and PPV: –Clinical if prior probability is very low - or very high - test contributes little information –Screening PPV is usually very low, especially for disease of low frequency

26. CASS: % with coronary artery disease according to symptom history (DA Weiner et al. NEJM 301:230-5,1979)

28. Prevention and screening Objectives Student should be able to: describe the levels of prevention list and apply the criteria for screening describe the impact of prevalence on predictive values explain why there are difficulties with screening

29. PRIMARY PREVENTION –Prevention of disease by controlling risk factors, removing causes e.g. non-smoking promotion SECONDARY PREVENTION –Reduction in consequences of disease by early diagnosis and treatment e.g. cervical cancer screening TERTIARY PREVENTION –Reduction of complications of disease e.g. MV crashes and ICU Levels of prevention

30. Screening: process of identifying unrecognised diseases/defects using tests that can be applied rapidly on a large scale Tests sort apparently healthy from those with (subclinical) disease Not diagnostic –investigative follow-up and treatment required Safety paramount –initiative from health service, not client

31. 1. Purpose of testing To move probability of disease towards 0% or 100%

32. NATURAL HISTORY OF DISEASE

33. Screening Criteria Important health problem –high prevalence \ radical consequences Natural history known –long pre-clinical phase Tests: valid, acceptable, ‘cheap’ Effective treatment –adequate facilities for dx & Rx Economically justifiable

34. We screen for many different types of disease : metabolic disorders genetic disorders congenital defects developmental problems infectious diseases cancers and other chronic diseases

35. Types of screening Mass screening Opportunistic screening (case finding) Targeted screening Multiphasic screening

36. SCREENING BIASES Lead time bias: –date of diagnosis automatically advanced for those cancers detected by screening Length bias: –tendency of screening to detect cancers which spend longer in the asymptomatic state (slower growing) Selection bias: –tendency for non-acceptors of screening to be at higher (or lower) risk of developing and/or dying from this disease than the general population

37. NATURAL HISTORY OF DISEASE

39. SCREENING BIASES Lead time bias: –date of diagnosis automatically advanced for those cancers detected by screening Length bias: –tendency of screening to detect cancers which spend longer in the asymptomatic state (slower growing) Selection bias: –tendency for non-acceptors of screening to be at higher (or lower) risk of developing and/or dying from this disease than the general population.

40. Evaluation of screening programmes RCT Apparent ‘benefit’ of screening enhanced if: –‘poor’ quality RCT –Non RCT e.g. case control studies

41. Target groups for screening: NEONATES "FOETUSES” PREGNANT WOMEN/PARENTS -TO-BE ADULTS ELDERLY PKU/CHT/GAL/HCU/MSUD, CDH, NTD, Down syndrome, thalassaemia UTI, STD, AIDS, rubella, blood sugar, Rhesus factor Rubella, AIDS, sickle cell anaemia, thalassaemia, Tay Sachs cancers (breast, cervix, rectal), hypertension, cholesterol glaucoma, cataract, hearing, mobility.

42. The treatment aims of screening differ Target group Neonates "Foetuses" Pregnant women Parents-to-be Children Adults Elderly Treatment Prevention of serious morbidity Termination Prevention of foetal damage (also maternal welfare) Prevention of conception Alleviation / prevention of morbidity Reduction in morbidity or mortality / prevention of disease onset Alleviation

43. Key points 1 - 2 1. Purpose of testing –to move probability of disease towards 0% or 100%. 2. Tests vary in performance (validity) –Sensitivity probability of pos. test result when disease present –Specificity probability of neg. test result when disease absent –Trade off between sensitivity and specificity

44. Key points 3 - 4 3. +ve test result  person has disease –PPV : probability of disease, given a pos. test result –NPV: prob. of no disease, given neg.test result –PPV rises: as specificity increases as prior probability/ prev. increases 4. Implications of link between prior prob. and PPV: –Same test performs differently in different situations –Clinical context if prior probability is very low - or very high - test contributes little information –Screening for disease of low frequency PPV is very low

45. Task force on sudden death in sport (Nov 2005) Clinical screening for sudden death in all teenagers and adults who play sport –PPV = 1/1000 Better to use family history i.e.targeted screening

46. * Screening test: blood spot trypsinogen Nutritional benefits of neonatal screening P. Farrell et al. NEJM 1997; 337:963-9

47. Intervention Cost per QALY* in US dollars Opportunistic screening for type 2 diabetes$56,600 Intensive glycaemic control in type 2 diabetes$16,000 Plus tight blood pressure control additional $700 Detection of mild thyroid disease in women aged > 35 years, during a health check$9,000 Breast cancer screening with mammography in women aged 50 - 65 years $150,000 Colon cancer screening using FOBT in patients aged 50 – 75 years$16,000 Cervical cancer screening using pap smears every 4 years for women aged 20 – 75 years$16,000 * Quality adjusted life years

48. Cochrane review of breast cancer screening Relative risk of death from breast cancer in screened vs control groups Gotzsche & Olsen, Lancet 2000; 355: 129-34 Total mortality - no difference Breast cancer mortality - marginal difference

49. Relative risk of death from breast cancer in screened vs control groups Gotzsche & Olsen, Lancet 2000; 355: 129-34

50. Trials with adequate randomisation