1 Fred Paccaud, IUMSP, LausanneScreening as a public health strategy 9th International Seminar on the Public Health Aspects of Noncommunicable Diseases Geneva and Lausanne, Switzerland 4-8 MAY 2015 Fred Paccaud, IUMSP, Lausanne

2 Strategies of public health“Comprehensive care of NCDs encompasses primary prevention, early detection/screening, treatment, secondary prevention, rehabilitation, palliative care and attention to improving mental health as a priority for social development and investment in people” [GLOBAL ACTION PLAN FOR THE PREVENTION AND CONTROL OF NCDs ]

3 Best buys in LMICs: harmful use of alcoholIncreasing capacity of health-care services to deliver prevention and treatment interventions for hazardous drinking and alcohol use disorders, including screening and brief interventions in all settings providing treatment and care for NCDs [GLOBAL ACTION PLAN FOR THE PREVENTION AND CONTROL OF NCDs ] Screening includes a short questionnaire to detect heavy drinkers (e.g., Michigan Alcoholic Screening Test) and a short psychotherapeutic-like intervention

4 Definitions and characteristicsNCDs and screening Conditions for screening (i): frequent disease Conditions for screening (ii): preclinical marker Conditions for screening (iii): effectiveness of early treatment Conclusions

5 Bibliography Rose G. The strategy of preventive medicine. Oxford: OUP, 1992 Raffle A, Muir Gray JA. Screening: Evidence and Practice. Oxford: OUP, 2007 Chiolero A, et al. Swiss Med Wkly doi: /smw

6 Definitions (i) End of Middle Age in Europe: Screening of sex workers to exclude those with sexually transmitted diseases Aiming the protection of the consumers, not the sex workers 1900: AMA proposes the institution of comprehensive annual testing (check up) of healthy adults in the USA RCTs: no benefit of systematic checkups

7 Definitions (ii) Breast cancer screening Cervical cancer screeningBreast palpation introduced in Germany in the 1930s Cervical cancer screening Implemented after the 2nd World War in the UK as an organized screening (1960), reorganized in the 1990s Screening for cardiovascular disease (high blood pressure, dyslipidemia, …) Pre-, peri- and post-natal screening (incl. screening to consider pregnancy termination) Public health genomic

8 Definitions (iii) Pre-clinical disease Clinical diseaseDeterminants Pre-clinical disease Clinical disease Consequence of disease Health promotion Primary prevention Secondary prevention (=early diagnosis, detection, screening) Clinical care

9 A PYRAMID OF HEALTH AND DISEASEDiseased, diagnosed & controlled Diagnosed & uncontrolled Undiagnosed or wrongly diagnosed Risk factors for disease Free of risk factors

10 Definitions (iv) 20 30 40 50 60 70 80 Birth Exposure Neoplasia Exfoliation Screened diagnosis Symptom diagnosis Death Total Pre-Clinical Phase (TPCP) Detectable Pre-Clinical Phase (DPCP) Begins at the initiation of disease ends when the disease is clinically manifested Begins when screening test is able to detect disease Ends when disease is clinically evident

11 Definitions NCDs: big opportunity for screening strategies, mainly because the long or very long incubation periods of most NCDs Increasing life expectancy for people aged more than 60 years increases the probability of developing NCDs before death

12 Rectangularization of the survival curve Women, France, 1876-2001Proportion de survivants Age

13 Definitions Development of biomedical and medical technologies, allowing new interventions Genomic of complex diseases (e.g., cystic fibrosis, breast cancer, …) Medical imaging (e.g., lung cancer, abdominal aortic aneurysm)

14 Definitions Secondary prevention (=early diagnosis, detection, screening) of NCDs is aiming a better prognosis through an earlier treatment NB: Primary prevention is aiming a decline of incidence Provided to asymptomatic persons Usually provided as a two-steps phase, with (i) a safe, unexepensive test and (ii) a full diagnostic procedure Those tested positive have an higher probability to be diseased, deserving a full diagnostic investigation Those tested negative have a low probability of disease, low enough to skip diagnostic investigation

15 Screening for NCDs Cardiometabolic diseasesAtheromatous disease (ischaemic heart disease, stroke, aortic aneurysm, …) Diabetes Chronic renal diseases Cancer (cervix, breast, colorectum, prostate, lung) Neuropsychiatric diseases (major depression, Alzheimer, Parkinson) Musculo-squelettal diseases

16 Prevention of cardiovascular disease: a simple frameworkPatho-physiological markers Hypertension High cholesterol Diabetes Obesity Cardiovascular phenotypes IHD Stroke Peripheral artery dis. Heart failure (some cancers) Modifiable risk factors Tobacco Physical inactivity Unhealthy diet (salt, fats& sugar, fruits/veg) Population strategy: Environment Policy Education High-risk strategy: Screening and treatment of high-risk individuals: Often medication ($$) + Surveillance guide & evaluate intervention

17 Three main arguments Frequent diseaseAvailability of an inexpensive, safe detection test Evidence of effectiveness of the early treatment

18 N diseased with positive test N non diseased with positive testN with positive test N diseased N non diseased N target population

19 Diseased Non diseased 100’000 N target population: 100’000

20 Estimated disease true prevalence: 5%Diseased Non diseased 5’000 95’000 100’000 N target population: 100’000 Estimated disease true prevalence: 5% N diseased: 5’000

21 Estimated disease true prevalence: 5% Diseased Non diseased Positive test 4’000 5’000 95’000 100’000 N target population: 100’000 Estimated disease true prevalence: 5% N diseased: 5’000 Persons with a positive test, i.e., % of true positive (= sensitivity): 80% N true positive: 4’000

22 Estimated disease true prevalence: 5% Diseased Non diseased Positive test 4’000 4’750 8’750 5’000 95’000 100’000 N target population: 100’000 Estimated disease true prevalence: 5% N diseased: 5’000 Persons with a positive test, i.e., % of true positive (= sensitivity): 80% N true positive: 4’000 % non diseased persons with a positive test, i.e., % false positive (= 1- specificity): 5% N false positive: 4’750 N positive (true+false): 8’750 Positive predictive value of the test: 54% (4’750/8’750)

23 Treatment Re-screen Clinical diagnosis Target population N=100’000To understand the process of screening , it should be clear that screening is not usually diagnostic and in most cases is followed by appropriate investigative procedures to confirm the results. The process of screening starts by selecting the population, followed by performing a screening procedure to classify individuals into either apparently healthy or affected. For the apparently healthy individuals, regular screening at fixed intervals is recommended. The apparently affected group is sent for further evaluation by a subsequent diagnostic procedure. Appropriate intervention is started among those with confirmed diagnosis while individuals who prove to be unaffected are asked to repeat the screening test at fixed intervals. Screening test Test – N=91’250 Test + N=8750 Re-screen Clinical diagnosis Diagnostic – N=4’750 Diagnostic + N=4’000 Treatment

24 Estimated disease true prevalence: 5% Diseased Non diseased Positive test 4’000 4’750 8’750 5’000 95’000 100’000 N target population: 100’000 Estimated disease true prevalence: 5% N diseased: 5’000 Persons with a positive test, i.e., % of true positive (= sensitivity): 80% N true positive: 4’000 % non diseased persons with a positive test, i.e., % false positive (= 1- specificity): 5% N false positive: 4’750 N positive (true+false): 8’750 Positive predictive value of the test: 54% (4’750/8’750)

25 Diseased Non diseased Positive test 4’000 4’750 8’750 5’000 95’000 100’000 N target population: 100’000 Estimated disease true prevalence: 5% N diseased: 5’000 Persons with a positive test, i.e., % of true positive (= sensitivity): 80% N true positive: 4’000 % non diseased persons with a positive test, i.e., % false positive (= 1- specificity): 5% N false positive: 4’750 N positive (true+false): 8’750 Positive predictive value of the test: 54% (4’750/8’750) What if we decrease or increase the prevalence of the disease (with the same characteristics of the test) ? 1% vs. 5% 10% vs. 5%

26 Estimated disease true prevalence: 1% % diseased Non diseased Positive test 800 4’950 5’750 1’000 99’000 100’000 N target population: 100’000 Estimated disease true prevalence: 1% % diseased N diseased: 1’000 Persons with a positive test, i.e., % of true positive (= sensitivity): 80% N true positive: 800 % non diseased persons with a positive test, i.e., % false positive (= 1- specificity): 5% N false positive: 4’950 N positive (true+false): 5’750 Positive predictive value of the test: 14% (800/5’750)

27 Estimated disease true prevalence: 10% % diseased Non diseased Positive test 8’000 4’500 12’500 10’000 90’000 100’000 N target population: 100’000 Estimated disease true prevalence: 10% % diseased N diseased: 1’000 Persons with a positive test, i.e., % of true positive (= sensitivity): 80% N true positive: 800 % non diseased persons with a positive test, i.e., % false positive (= 1- specificity): 5% N false positive: 4’950 N positive (true+false): 5’750 Positive predictive value of the test: 64% (800/5’750)

28 Same detection test, different prevalences positive predictive values = 14% Prevalence 5%  positive predictive values = 54% Prevalence 10%  positive predictive values = 64% If the prevalence is too low, the effetiveness of the screening programm decreases

29 Same detection test, different prevalencesOne way to overcome the problem is to target high risk groups, i.e., with higher prevalence of the disease e.g., target population for breast cancer screening are women aged years Before the age 50, incidence of new cases and prevalence of subclinical cancer is too low after the 70, strong competition with causes of death other than breast cancer e.g., target population for colorectal cancer screening are people aged more than 50 years

30 Three main arguments Frequent diseaseAvailability of an inexpensive, safe detection test Evidence of effectiveness of the early treatment

31 Kundel, Polansky. Radiology 2003;228:303

32

33 Sensitivity 53.6% Specificity 98% PPV 10%faecal-occult blood screening Hardcastle et al. RCT of faecal-occult blood screening for colorectal cancer. Lancet 1996;348:1472 Sensitivity 53.6% Specificity 98% PPV %

34 Rapport sur la santé dans le monde, 2002. Genève: OMS, 2002:64

35 Incidence of myocardial infarction and total cholesterolCastelli. Atherosclerosis 124 (1996):S1

36 Framingham

37 PROCAM Incidence of coronary events in 8 years of follow-up (%)Prevalence (%) Total Cholesterol to HDL-Cholesterol Ratio

38 BP, CHD risk and number of deaths30 20 10 Prevalence (%) 20 40 Deaths (n) 60 40 20 Risk (%) BP distribution Risk Most CVD events occur in subjects with moderately elevated risk factors. Deaths 110 120 130 140 150 160 Systolic BP (mm Hg)

39 Three main arguments Frequent diseaseAvailability of an inexpensive, safe detection test Evidence of effectiveness of the early treatment

40 Effect of screening on mortality from colorectal cancer (J Med Screen 2002;9:54)Metaanalysis based on patients in 4 RCTs. Include Kewenter trial, based on patients aged yrs,2 screen RR= .84 Should be noted that analysis made according to the rule of intention to treat induces a dilution of the effectiveness of screening In Nottingham, there is a reduction among participants, with RR [acceptors vs. non acceptors]=.73 *Annual and biennial combined

41 Screening for abdominal aortic aneurysm

42 Mortality (per 100’000) by cervix cancer, all ageMortality (per 100’000) by cervix cancer, all age. Various countries, 1 2 3 4 5 6 7 8 9 1970 1975 1980 1985 1990 1995 2000 2005 France Germany Russia Sweden United Kingdom

43 Breast cancer: 5 years survival by staging. Norway 1992

44 Planning and implementationchoose the right things to do Choose the tests sequence: detection (screening test) and diagnosis (disease diagnosis) Ex: breast palpation and mammography vs. mammography alone Fecal occult blood test (FOBT) and coloscopy vs. colposcopy alone do them right blinded double reading to increase VPP

45 Planning and implementationMain Biases healthy screenee effect length time effect lead time effect Evaluation Methods randomised controlled trials time trend analyses case control studies  Additional Information : pilot or demonstration projects Modelling : Test performance (sensitivity, specificity, etc.)

46 Be sure that the downstream care is effective !

47 Best buys in LMICs: combined detectionInteraction of NCDs with infectious diseases increases the risk of infectious disease acquisition and susceptibility in people with pre-existing NCDs Maximize the opportunities to detect and to treat both NCDs and infectious diseases in health-care services, e.g.: smoking and diabetes, alcohol-use disorders, immunosuppression, etc. second-hand smoke and tuberculosis screening for diabetes and chronic respiratory disease in tuberculosis clinics and for tuberculosis in NCD clinics Integrate NCD programs or palliative care with HIV

48 Conclusions Screening is a complex task, which should be planned, implemented and monitored by public health specialists Strong cultural belief that all screening must be a good thing Screening is also an important commercial activity: these forces tend to oversimplify screening's benefits Screening needs to be delivered as a programme if it is to achieve risk reduction Different observers see some consequences more starkly than others depending on their viewpoint