Outcome research/screening 1 Outcome evaluation of screening Wei-Chu Chie.

outcome research/screening 1

Outcome Outcome evaluation of evaluation of

screeningscreeningWei-Chu ChieWei-Chu Chie

outcome research/screening 2

Screening Screening • Secondary prevention

– early detection and prompt treatment– screening people with early, often

preclinical, manifestations of diseases, to prevent its progression

– must followed by diagnosis and treatment procedures

outcome research/screening 3

VariantsVariants• Mass screening• opportunistic screening• clinical case finding

outcome research/screening 4

Endpoints and efficacy Endpoints and efficacy evaluationevaluation

• Compare the outcomes of screened and non-screened (symptomatic) groups

• mortality– case-fatality/survival– total (overall) mortality– specific mortality

• specific morbidity

outcome research/screening 5

Case-fatality/survivalCase-fatality/survival• Count from the time of diagnosis• take only causes of death that

related to the target disease(s) of screening – lead-time bias– length bias (prognosis selection bias)

outcome research/screening 6

Overall mortalityOverall mortality• Count from the time of participation• take all causes of death

– unaffected by the ways of diagnosis or detection

– overall impact of the program • positive & negative

– sometimes insensitive

outcome research/screening 7

Specific mortalitySpecific mortality• Count from the time of participation• take only causes of death that

related to the target disease(s) of screening– may be affected by the ways of

diagnosis or detection– more sensitive– sometimes difficult to identify

outcome research/screening 8

Specific morbiditySpecific morbidity• Count from the time of participation• take only some severity indices of

the target disease(s) of screening• stages, tumor size, certain

complications, ...– not the final outcome– only surrogate in nature

outcome research/screening 9

Study designStudy design• Randomized controlled trial• non-randomized

– cohort– case-control– other variants

• meta-analysis

outcome research/screening 10

實驗性方法的三要素實驗性方法的三要素• 實驗單位 (experimental unit)• 處置 (treatment)• 評估 (evaluation)

outcome research/screening 11

實驗單位實驗單位• 個人為單位

– 病人 •藥物臨床試驗 •教育介入 如病人衛生教育

– 健康人 •疫苗預防性試驗 篩檢工具試驗•教育介入 如健康行為的養成

• 群體為單位– 社區 班級 團體

outcome research/screening 12

處置處置•新藥•新的飲食•診斷或篩檢•手術•醫療器材•教育介入•無治療

outcome research/screening 13

評估評估• 有效性

•事件 / 事件史資料 (event/time to event)• well-being 指標 : 生活品質 (quality of life)•成本 : 直接與間接醫療成本•暫時性或替代性指標 : 生理 生化 病理資料•又稱終點 (end-points)

• 安全性– 不良反應及相關指標

• 經濟學 : 成本效益

outcome research/screening 14

Common problemsCommon problems• No blindness

– screener– subjects– evaluator or rater

• sometimes randomized by cluster for the convenience of intervention– statistical help

outcome research/screening 15

ExamplesExamples• Breast self-examination for breast

cancer– mass screening

• General health screening for cardiovascular risks– by general practitioners

• Case-finding for depression in primary care– by general practitioners

outcome research/screening 16

BSEBSE• Thomas DB., et al. Randomized trial

of breast self-examination in Shanghai: final results. JNCI 2002; 94:1445-1457.

outcome research/screening 17

BSE: BSE: background/goal/hypothesisbackground/goal/hypothesis

• Background: – BSE efficacy unproven

• Goal:– to determine whether ...

• Hypothesis: – an intensive program of BSE instruction

will reduce the mortality (number of death ) of breast cancer

outcome research/screening 18

BSE: study designBSE: study design• Randomized controlled trial

– instruction group– control group– randomized by cluster (factories)

stratified by hospital affiliation– no statistical adjustment: large number

and identical rates

• mass screening

outcome research/screening 19

BSE: subjects BSE: subjects • Shanghai Textile Industry Bureau

(STIB) employees• women born between 1925-1958

(31-64)• eligibility, inclusion/exclusion criteria• 289292 ….266064

– instruction: 146437 …. 132979– control: 142955 …. 133085

outcome research/screening 20

BSE: exposure/interventionBSE: exposure/intervention• BSE instruction

– small group (10)– video/practice/discussion/reinforcement– review/supervision– proficiency assessment

outcome research/screening 21

BSE: endpointBSE: endpoint• Primary: death from breast cancer

– definition– judgement: blind

• Secondary?– Intermediate variables: benign

tumors/biopsies

• Follow-up: 11 years

outcome research/screening 22

BSE:data analysis BSE:data analysis • Survival

– from time of entry– from time of diagnosis

• Kaplan-Meier/log-rank test• Cox proportional hazards model

• comparability , compliance, …• intermediate variables

outcome research/screening 23

BSE: major BSE: major results/discussionresults/discussion

• Mortality/survival– from entry– from diagnosis

• comparability, compliance, …• intermediate variables

– benign tumors … biopsy (unnecessary burden)

outcome research/screening 24

CRSCRS• Engberg M., et al. General health screeni

ngs to improve cardiovascular risk profiles: a randomized controlled trial in general practice with 5-year follow-up. J Family Practice 2002;51:546-552.

outcome research/screening 25

CRS: CRS: background/goal/hypothesisbackground/goal/hypothesis

• Background: Danish trial– GP’s role in primary prevention for CVD

• Goal:– to investigate the impact of general

health screenings and health discussions on CV risk profiles

• Hypothesis: – … can reduce ...

outcome research/screening 26

CRS: study designCRS: study design• Randomized controlled trial

– 3 groups– by individual– stratified randomization based on

• GP, sex, age, cohabitation status, BMI

– too many cells• GP X sex X age X …

– stratified factors should be treated as covariates

outcome research/screening 27

Stratified randomizationStratified randomization– Individuals stratified by certain characteristi

cs– Independent random assignment to differe

nt groups in each cell• eg. Old-male, old-female, young-male, …• random code for each consecutive subject• can be generated by computer

– To avoid imbalance• may have empty cells if N is small• no need when N is large

outcome research/screening 28

CRS: subjects CRS: subjects • Ebeltoft, Aarhus County, Denmark• 9 GPs (general practitioners)• 3464 … random sample 2000 inhabitant

s• 30-49 years

outcome research/screening 29

CRS: exposure/interventionCRS: exposure/intervention• Health screening & follow-up

discussion– Group 1: health screening (502)– Group 2: health screening+follow-up

discussion (504)– Group 3: none (control) (501)– All three: questionnaire

outcome research/screening 30

CRS: endpointCRS: endpoint• Primary: score and categorical (high/low)

– CRS: gender, family inheritance (CAD <55 y), tobacco use, BP, total serum cholesterol, BMI

– BMI, SBP, DBP, serum cholesterol• by smoking and overweight subgroup

analysis

– No blindness• Follow-up: 5 years

outcome research/screening 31

CRS:data analysis CRS:data analysis • T-test, chi-square, …• intention-to-treat• compare values or high score rates

instead of changes from baseline– endpoints– baseline data– follow-up rates

outcome research/screening 32

CRS:major CRS:major results/discussionresults/discussion

• CRS, BMI, and serum cholesterol levels were lower in the Intervention group

• subgroup analysis: high vs. low risk ones

• no differences between two intervention groups

• not followed to the final endpoint

outcome research/screening 33

DepressionDepression• Williams JW. , et al. Case-finding for

depression in primary care: a randomized trial. Am J Med 1999; 106: 36-43.

outcome research/screening 34

Dep: background/goal/hypothesisDep: background/goal/hypothesis• Background:

– impact of depression/role of primary care

• Goal:– evaluate the efficacy of two case-finding

instruments vs. usual care

• Hypothesis: – … can enhance symptom reduction/recovery

for depression patients

outcome research/screening 35

Dep: study designDep: study design• Randomized controlled trial

– 3 groups– by individual– random assignment stratified by site

• computer-generated randomization blocked log

outcome research/screening 36

Dep: subjects Dep: subjects – GP and internal medicine clinics in San

Antonio, Texas and Washington, DC.– Consecutive adult patients– Only those diagnosed as with depression

were analyzed• screened with case-finding: positive---

additional questioning for clinical depression• all patients: DSM-III-R, CAGE, SF-36,

satisfaction• rater (evaluator) blindness

outcome research/screening 37

Dep: exposure/interventionDep: exposure/intervention• Case-finding instruments screening

– Group 1: single question (330)– Group 2: 20 item questionnaire (323)– Group 3: usual care (control) (316)

outcome research/screening 38

Dep: endpointDep: endpoint• Primary: only depressive patients

– recovery rates (<=1 DSM-III-R symptoms)

– symptoms score changes from baseline

• Secondary:– satisfaction

• Follow-up: 3 months

outcome research/screening 39

Dep:data analysis Dep:data analysis • Proportion and rates• Symptom numbers

– only for patients diagnosed as with depressions

– not severity indices– by medical record: counseling, referral,

medicine

• baseline data• follow-up rates

outcome research/screening 40

Dep:major results/discussionDep:major results/discussion• Recovery rate:

– case-finding better than usual care

• symptoms numbers– not different

• Only diagnosed patients were analyzed

• not followed to the final endpoint