Analytical Epidemiologic Study Panithee Thammawijaya Bureau of Epidemiology 1
Jan 01, 2016
Analytical Epidemiologic Study
Panithee Thammawijaya
Bureau of Epidemiology1
เป้�าหมายของการศึ�กษาทางระบาดวิ�ทยา
• DESCRIBE• ม�ผู้��ป้�วิยด�วิยโรคห�วิใจหลอดเล!อดมากน้�อยเพี�ยงใดใน้จ�งหวิ�ด
แห%งหน้�&ง• ผู้��ป้�วิยด�วิยโรคห�วิใจหลอดเล!อดเป้'น้สั�ดสั%วิน้เท%าไรใน้ผู้��หญิ�งและ
ใน้ผู้��ชาย• EXPLAIN
• ท,าไมผู้��ชายจ�งป้�วิยด�วิยโรคห�วิใจหลอดเล!อดมากกวิ%าผู้��หญิ�ง• การสั�บบ-หร�&เพี�&มควิามเสั�&ยงใน้การเป้'น้โรคห�วิใจหลอดเล!อดหร!อ
ไม%• PREDICT• ถ้�าสัามารถ้รณรงค0ให�คน้ใน้ช-มชน้เล�กสั�บบ-หร�&ได�เป้'น้ผู้ลสั,าเร1จ
จ,าน้วิน้ผู้��ป้�วิยโรคห�วิใจหลอดเล!อดรายใหม%ใน้ป้3หน้�าจะลดลงเป้'น้จ,าน้วิน้เท%าไร
Source: Morgenstern, 2001 (modified)
Measure of Frequency
Measure of Association
Measure of Impact
2
• CONTROL• มาตรการท�&เหมาะสัมสั,าหร�บช-มชน้ (ภายใต�ข�อจ,าก�ดต%างๆ) ค!อ
อะไร
3
การวิ�ดทางระบาดวิ�ทยา•Measure of Frequency: ขน้าดป้7ญิหา
– ควิามช-ก Prevalence – อ-บ�ต�การ Incidence
•Measure of Association: วิ�ดขน้าดควิามสั�มพี�น้ธ์0ระหวิ%างป้7จจ�ยก�บโรค–Risk Ratio, (Incidence) Rate Ratio
===> Cohort Study–Odds Ratio ===> Case – Control
Study–Prevalence Ratio, Prevalence Odds
Ratio ===> Cross-sectional Study
From Last Time…(1)If you want to count…
“State”Existing of… at a point of time
Prevalence(=New + Old cases)
“Event”Occurring of… during a period of time
Incidence(=New cases)
E.g. •Number of all DM cases a village in Jan 2009 = 120 •Proportion of current smokers in company on Jan 1st, 2010 = 15% of total employees
E.g. •Number of flu cases occurred in a village 2009 = 150•Proportion of new smokers in a company during Jan to May 2010 = 2% of non-smoker on Dec 31st, 2009
From Last Time…(2)
Point Prevalence
Prevalence
Period Prevalence
=
=
At time t1
During time t1-t2
= sick (new) = not sick= sick (old)
From Last Time…(3)
Incidence Proportion(Risk; Cumulative
Incidence )
Incidence
=
= During time t1-t2
Incidence Rate(Rate; Incidence Density)
= sick (new) = not sick= sick (old)
During time t1-t2
= person-time
7
2x2 Table and Measure of Association(Count Data)
Disease Non disease Total
Exposed A B A+B
Unexposed C D C+D
Total A+C B+D A+B+C+D
Risk Ratio (RR) = [A/(A+B)] / [C/(C+D)] Odds Ratio (OR) = [A/C] / [B/D] = AD/BC Prevalence Ratio (PR) =
[A/(A+B)] / [C/(C+D)]
8
2x2 Table and Measure of Association(Person-Time Data)
No. of Cases Person-Time
Exposed A TE
Unexposed B TU
Total A+B TE+TU
Incidence Rate Ratio (IRR) = [A/TE] / [B/TU ]
Ratio Scale Measures and Theirs Relationships
OR OR
0 ∞1IRR IRRRR RR
The null value(no association)
weaker stronger
Causative Effect
weakerstronger
Protective Effect
9
Causative FactorProtective Factor
Risk Factor
How epidemiologists work?1. Counting:
Counts cases or health events, and describes them in terms of time, place, and person
2. Dividing:Divides the number of cases by an appropriate denominator to calculate “rates”
3. Comparing:Compares these “rates” over time or for different groups of people
* Rate, in this case, simply means division of one number by another
DescriptiveEpidemiology
AnalyticEpidemiology
11
Classification epidemiological study
Observational Study
(natural exposure)
Experimental Study(exposure given by researcher)
การศึ�กษาเช�งพีรรณน้าDescriptive Study
(ไม%ม�กล-%มเป้ร�ยบเท�ยบ)
การศึ�กษาเช�งวิ�เคราะห0Analytic Study(ม�กล-%มเป้ร�ยบเท�ยบ)
Cross – sectional Case control CohortFrom: Ram Rungsin, modified
Case report Case series
12
Case report: a hypertension
case in young adult
Case series: three
hypertension cases in
young adults
Cross – sectional study: a
hypertension survey
Cross – sectional study: HT
vs Salt consumption
Case – control study: HT vs
Salt consumption
Cohort study: HT vs Salt
consumption
Clinical trial: Beta blocker
vs Hypertension
Descriptive
AnalyticExperiment
ลำ��ดั�บชั้��นของก�รศึ�กษ�ท�งดั��นระบ�ดัวิ�ทย�
ลำ��ดั�บชั้��นของก�รศึ�กษ�ท�งดั��นระบ�ดัวิ�ทย�
From: Ram Rungsin
13
การศึ�กษาเช�งพีรรณน้า
ไม%ป้�วิย
ผู้��ป้�วิย เป้�าหมายMagnitude and
severityDistribution: Time,
Place, Person<<Hypothesis formulation>>
เป้�าหมายMagnitude and
severityDistribution: Time,
Place, Person<<Hypothesis formulation>>
สัน้ใจเฉพีาะกล-%มผู้��ป้�วิย
14
การศึ�กษาเช�งวิ�เคราะห0
ไม%ป้�วิยป้7จจ�ย A?
ป้�วิยป้7จจ�ย A?
สัน้ใจท�:งกล-%มผู้��ป้�วิย
และไม%ป้�วิย
เป้�าหมายAssociation between Disease and Factor A
<<Hypothesis testing>>
เป้�าหมายAssociation between Disease and Factor A
<<Hypothesis testing>>
What Is the “Cause” of a Disease? (1)
1990 2010
Mr. A• 20-yrs male, Thai, farmer, etc.
Mr. A• 20-yrs male, Thai, farmer, etc.
Did pumpkin have an effect on the disease in Mr. A? Yes, causative effect.
Event actually occurred-observed
Counterfactual-not observed
What Is the “Cause” of a Disease? (2)
1990 2010
Mr. A• 20-yrs male, Thai, farmer, etc.
Mr. A• 20-yrs male, Thai, farmer, etc.
Did pumpkin have an effect on the disease in Mr. A? No. He is doomed.
Event actually occurred-observed
Counterfactual-not observed
What Is the “Cause” of a Disease? (3)
1990 2010
Mr. A• 20-yrs male, Thai, farmer, etc.
Mr. A• 20-yrs male, Thai, farmer, etc.
Did pumpkin have an effect on the disease in Mr. A? No. He is immune.
Event actually occurred-observed
Counterfactual-not observed
What Is the “Cause” of a Disease? (4)
1990 2010
Mr. A• 20-yrs male, Thai, farmer, etc.
Mr. A• 20-yrs male, Thai, farmer, etc.
Did pumpkin have an effect on the disease in Mr. A? Yes, protective effect.
Event actually occurred-observed
Counterfactual-not observed
• วิ�ดการเก�ดโรคใน้กล-%มต�วิอย%างกล-%มหน้�&งท�& “exposed” เป้ร�ยบเท�ยบก�บการเก�ดโรคใน้กล-%มเด�ยวิก�น้(คน้เด�ม)น้�:น้หากวิ%าพีวิกเขาไม%ได� exposed , หร!อ
• วิ�ดการเก�ดโรคใน้กล-%มต�วิอย%างกล-%มหน้�&งท�& “unexposed” เป้ร�ยบเท�ยบก�บการเก�ดโรคใน้ป้ระชากรกล-%มเด�ยวิก�น้(คน้เด�ม)น้�:น้หากวิ%าพีวิกเขาได� exposed
• สร�ป โดยหล�กการ การศึ�กษาเพี!&อค�น้หาสัาเหต- จะต�องเป้ร�ยบเท�ยบ
“actual outcome” vs. “potential outcome”
Causal Inference in Modern Epidemiology
19
Causal Inference in Modern Epidemiology• ใน้ทางป้ฏิ�บ�ต� เราไม%สัามารถ้สั�งเกตการเก�ดโรคใน้ภาวิะท�&เป้'น้
“counterfactual” หร!อ “the potential outcome” ได�• เราจ�งต�องท,าการเป้ร�ยบเท�ยบกล-%มต�วิอย%างท�&“exposed”
ก�บกล-%มต�วิอย%างอ!&น้แทน้ (Substitute population)
• กล-%มต�วิอย%างอ!&น้ท�&ใช�แทน้ได� จะต�องเป้'น้กล-%มต�วิอย%างท�&ม�ล�กษณะท�&เป้'น้ต�วิแทน้(represent) ของ กล-%มต�วิอย%างท�&“exposed”น้�:น้หากวิ%าไม%ได� “exposed”
• Validity of inference ข�:น้อย�%วิ%ากล-%มต�วิอย%างท�&น้,ามาเป้ร�ยบเท�ยบก�น้น้�:น้ (exposed and unexposed groups) สัามารถ้เป้ร�ยบเท�ยบก�น้ได� (comparability) มากน้�อยเพี�ยงใด
20
What Is the “Cause” of a Disease?
1990 2010
Did pumpkin have an effect on the disease in population? Validity & Precision?
?
?Exposed group• age 15-25 yrs
Unexposed group• age 15-25 yrs
Analytic Epidemiological Study
Exposure DiseaseEffect?
•เป้�าหมายสั,าค�ญิของ Analytic study ค!อ การวิ�ด effect ของ exposure ท�&ม�ต%อโรคหน้�&งๆ•ใน้ Observational study ไม%สัามารถ้วิ�ด effect ได�โดยตรงเน้!&องจากกล-%มเป้ร�ยบเท�ยบอาจจะไม% Comparable•ใน้ทางป้ฏิ�บ�ต�จ�งวิ�ดได�แต%เพี�ยง ควิามสั�มพี�น้ธ์0(Statistical association)
Association?
22
Cause? Risk factor?
•ใน้ทางป้ฏิ�บ�ต�ม�กจะไม%สัามารถ้ระบ-ได�แน้%ช�ดวิ%าสั�&งใดเป้'น้สัาเหต- (cause) ท�&แท�จร�งของโรคหน้�&งๆ เน้!&องจากข�อจ,าก�ดของควิามร� �(เช%น้ ด�าน้ช�วิวิ�ทยาหร!อกลไกการเก�ดโรค เทคโน้โลย�ใน้การวิ�ด ฯลฯ)•ใช�ค,าวิ%า Risk factor แทน้เพี!&อแสัดงถ้�งข�อจ,าก�ดด�งกล%าวิ
23
ก�รศึ�กษ�เชั้�งวิ�เคร�ะห์�• Cross – sectional Study
• Case – Control Study
• Cohort Study
Cross-sectional studyIn a cross-sectional study,
the measurementsmeasurements of exposure andeffect are made at the same time
24
25
ส��รวิจภ�วิะควิ�มดั�นโลำห์�ตส$งแลำะไขม�นในเลำ(อดั
• The Health Department of Hanoi City in 2000
• 1,000,000 Hanoi population
• ส�มภ�ษณ์�• เจ�ะเลำ(อดัวิ�ดั Cholesterol
• วิ�ดัควิ�มดั�นโลำห์�ต
26
• 60,000 = hypertension
• 200,000 = high blood
cholesterol
• Prevalence of HT = ? 600001000000 6, / , , = %
“ ควิามด�น้โลห�ตสั�งและไขม�น้ใน้เล!อด
ม�ควิามสั�มพี�น้ธ์0ก�น้หร!อไม% ”
ส��รวิจภ�วิะควิ�มดั�นโลำห์�ตส$งแลำะไขม�นในเลำ(อดั
27
ส��รวิจภ�วิะควิ�มดั�นโลำห์�ตส$งแลำะไขม�นในเลำ(อดั
ID Age Sex -Hypertension
HighChol
1 18 M No Yes
2 36 M No No
3 50 F Yes Yes
28
Defined PopulationDefined Population
Exposed:Have disease
Exposed:Have disease
Exposed:No diseaseExposed:
No diseaseNot Exposed:Have diseaseNot Exposed:Have disease
Not Exposed:No disease
Not Exposed:No disease
Gather Data on Exposure & Disease at the same timeGather Data on Exposure & Disease at the same time
ส��รวิจภ�วิะควิ�มดั�นโลำห์�ตส$งแลำะไขม�นในเลำ(อดั
29
a b
c d
DiseaseNo Disease
Exposed
Not Exposed
ส��รวิจภ�วิะควิ�มดั�นโลำห์�ตส$งแลำะไขม�นในเลำ(อดั
30
20000,
40000,
HT NoHT
High Chol.
Normal Chol.
1000000, ,
200000,
800000,
60,000,940000,
ส��รวิจภ�วิะควิ�มดั�นโลำห์�ตส$งแลำะไขม�นในเลำ(อดั
180000,
760000,
31
• High Cholesterol : HT Prevalence Rate = 20,000 / 200,000 = 10%
• Normal Cholesterol : HT Prevalence Rate = 40,000 / 800,000 = 5%
• Prevalence Ratio (PR) = 10% / 5% = 2
ส��รวิจภ�วิะควิ�มดั�นโลำห์�ตส$งแลำะไขม�นในเลำ(อดั
32
Prevalence Ratio (PR)• Prevalence Ratio = 10% / 5% = 2
• แป้ลวิ%า “ ผู้��ท�&ม�ภาวิะ high cholesterol ม�โอกาสัท�&จะ พีบวิ%าม�โรคควิามด�น้โลห�ตสั�งอย�%ด�วิยเป้'น้ 2 เท%าของผู้��
ท�&ไม%ม� high cholesterol”
PR จากการศึ�กษาแบบต�ดขวิาง สัามารถ้ใช�ป้ระมาณค%า “Risk Ratio” ถ้�าหาก
•ม�&น้ใจวิ%า Exposure เก�ดก%อน้ Disease (No temporal ambiguity)•Cases ท�&อย�%ใน้การศึ�กษาเป้'น้ต�วิแทน้ของ Incidence cases ท�:งหมด (ไม%ม� selective survival or prevalence-incidence bias)
33
Cross-sectional Studies
• Advantages:– quick, inexpensive– Useful for health administration and hypothesis formulation
• Disadvantages:– low prevalence due to
• Low incidence (rare disease)• short duration
– Uncertain temporal relationships– Selection Bias (Selective survival)– Information Bias (Recall bias)
Cohort study A study in which the
incidence proportion/rate of disease in 2 or more cohorts
is compared
34
35
A Roman CohortA Roman Cohort
Two centuries made one maniple and three maniples made up one cohort.
= A unit of 300-600 men in the ancient Roman army
36
= A group of persons who are followed over time
= A group of persons who are followed over time
“COHORT”in Epidemiology
“COHORT”in Epidemiology
37
Cohort Study• โดัยท�+วิไป ถื(อวิ-�เป.นก�รศึ�กษ�แบบ
ส�งเกตท/+ม/ควิ�มถื$กต�องส$งส�ดัในแง-ก�รห์�ควิ�มส�มพั�นธ์�ระห์วิ-�งป2จจ�ยก�บก�รเก�ดัโรค
• ใชั้�เวิลำ�ในก�รศึ�กษ�น�นท/+ส�ดั
• ใชั้�งบประม�ณ์ในก�รศึ�กษ�ม�กท/+ส�ดั
38
การใช�ระบาดวิ�ทยาใน้การค�น้หาสัาเหต-ของ
การเก�ดโรค
RISK FACTOR (ป้7จจ�ยเสั�&ยง)•Cigarette
DISEASE (การเก�ดโรค)•Lung Cancer
Causeส�เห์ต�
Effectผลำ
39
Person at-risk (without disease) at startPerson at-risk (without disease) at start
เก�ดโรคเก�ดโรค ไม%เก�ดโรคไม%เก�ดโรค เก�ดโรค เก�ดโรค ไม%เก�ดโรคไม%เก�ดโรค
ExposedExposed Not ExposedNot Exposed
40
Persons without the disease !!!!Not ExposedExpose
d Not Smoke#500
persons
Smoke#500
personsDise
aseNo
Disease
Disease
No Disea
seNo Lung Cancer
# 455
Lung Cancer
# 45
No Lung Cancer
# 499
Lung Cancer
# 1
1970
2001
41
CA LungNo CA
Smoke
Not smoke
45455
1499
500
500
• Incidence of Smoker who develop Lung Cancer = 45/500• Incidence of Non -Smoker whodevelop Lung Cancer = 1/500• Risk Ratio of smoking for Lung Cancer = 45 • ผู้��ท�&สั�บบ-หร�&ม�โอกาสัเก�ดโรคมะเร1งป้อดมากกวิ%าผู้��ท�&ไม%สั�บ 45 เท%า
42
Risk RatioCA LungNo CA
Smoke
Not smoke
A B
C D
Risk Ratio = A/A+B C/C+D
A+B
C+D
43
•เล!อกป้ระชากรกล-%มท�&ย�งไม%เก�ดโรคแต%ม�โอกาสั
•ค�น้หา Exposed group และNon-exposed group
•ต�ดตามและวิ�ด incidence of disease outcome ท�:งใน้กล-%ม Exposed และ Non – exposed ระหวิ%าง ช%วิงเวิลาท�&ท,าการศึ�กษา
•ค,าน้วิณหา Risk Ratio หร!อ Incidence Rate Ratio
Conducting a Cohort Study
44
Design 1: Prospective Cohort Study
PopulationPopulation
People withou
t diseas
e
People withou
t diseas
eUnexposedUnexposed
ExposedExposed
No diseaseNo disease
DiseaseDisease
No diseaseNo disease
DiseaseDisease
Time of Study BeginTime of Study Begin
Direction of inquiryDirection of inquiry
Sampling?
•If the study started before the disease occurred…“Prospective cohort study”
Cause Effect
Ex: A Study of Smoking and Lung Cancer(Prospective cohort study with person-time data)
No. of Case F/U time (person-year)
Smoking 90 30,526Non smoking 10 28,364
Total 100 58,890
Incidence rate in smokers = 90 / 30,526 = 2.9 per 1000 person-years
Incidence rate in non-smokers = 10 / 28,364 = 0.5 per 1000 person-yearsRate ratio = 2.9/0.5
= 5.8
Rate of developing the disease in smokers is 5.8 times of that in non-smokers 45
46
Design 2: Retrospective Cohort Study
PopulationPopulation
People withou
t diseas
e
People withou
t diseas
eUnexposedUnexposed
ExposedExposed
No diseaseNo disease
DiseaseDisease
No diseaseNo disease
DiseaseDisease
Time of Study BeginTime of Study Begin
Direction of inquiryDirection of inquiry
Sampling?
•If the study started after the disease occurred…“Retrospective (Historical) cohort study”
Cause Effect
47
Ex: An Diarrhea Outbreak in a Party(Retrospective Cohort study with count data)
Ill Not ill Total
Ate salad 150 50 200
Not eat 10 90 100
Total 160 140 300
Incidence proportion in exposed group = 150 / 200 = 75%
Incidence proportion in non-exposed group = 10 / 100 = 10%
Risk ratio = 75/10
= 7.5
Risk of developing the disease in exposed group is 7.5 times of that in non-exposed group
48
Cohort Studies - Advantages
• Can measure disease incidence • Can study the natural history • Provides strong evidence of casual
association between E and D (time order is known)
• Multiple diseases can be examined• Good choice if exposure is rare (assemble
special exposure cohort) • Generally less susceptible to bias
49
Cohort Studies - Disadvantages• Takes time, need large samples, expensive• Not useful for rare diseases/outcomes • With prolonged time period:
– Exposures change during follow-up period
• Selection Bias (loss-to-follow up in pros. cohort or selective survival in retro. cohort)
• Information Bias (recall bias in retro. Cohort)
Case-control studyKey: it begins with people with the
disease (cases) and compares them to people without the disease (controls)
50
51
เป้'น้การศึ�กษาระบาดวิ�ทยาเช�งวิ�เคราะห0ชน้�ดหน้�&ง เป้ร�ยบเท�ยบระวิ%างผู้��ท�&เป้'น้โรค (Case) ก�บกล-%มต�วิอย%างผู้��ท�&ไม%เป้'น้โรค (Control) โดยท,าการเป้ร�ยบเท�ยบป้ระวิ�ต�ของล�กษณะการม�ป้7จจ�ยเสั�&ยงท�&ก,าล�งศึ�กษา ระหวิ%าง 2 กล-%ม
Case – Control Study
52
Non CasesFactor A
CasesFactor A
CasesFactor ACases
Factor ANon CasesFactor A
Non CasesFactor A
Case – control Study
53
DiseaseDisease No DiseaseNo Disease
ExposedExposed NotExposed
NotExposedExposedExposed Not
ExposedNot
Exposed
Design for a case – control
Study
54
Design of a case-control study
Case Population
Case Population
Controls(People without disease)
Controls(People without disease)
Case(People with disease)
Case(People with disease)
Not exposedNot exposed
ExposedExposed
Not exposedNot exposed
ExposedExposed
Non-casePopulationNon-case
Population
•Identify true case, and true non-case populations•Sampling fractions from case<>non-case•Determine exposure status by history
Time of Study BeginTime of Study Begin
Direction of inquiryDirection of inquiryCause Effect
What is “Odds”?Odds of an event with an occurrence probability of p is the ratio of p to (1-p)Odds = Probability of event
Probability of non-event= p/(1-p)
Probability = odds/(1+odds)
Odds of Exposure among cases = a/(a+c) = a/c c/(a+c)
a bc d
D
+ -
+E - Odds of Exposure among noncases = b/(b+d) = b/d
d/(b+d)
Head-to-Head = 9 : 18
For case-control study:
What is “Odds Ratio”?Odds Ratio (OR) = Ratio of two odds
In case-control study, Exposure OR = Odds of exposure among cases Oddsof exposure among noncases
= a/c = ad/bc b/d
Odds of disease among the exposed = a/(a+b) = a/b b/(a+b)
Odds of disease among the unexposed = c/(c+d) = c/d d/(c+d)
a bc d
D
+ -
+E -
For cohort study:
In cohort study, Disease OR = Odds of disease among the exposed Odds of disease among the unexposed
= a/b = ad/bc c/d
57
2x2 Table and Measure of Association(Count Data)
Disease Non disease Total
Exposed A B A+B
Unexposed C D C+D
Total A+C B+D A+B+C+D
Risk ratio (RR) = [A/(A+B)] / [C/(C+D)] Odds ratio (OR) = [A/C] / [B/D] = [A/B] / [C/D]If disease is rare, then OR ~ RR
58
Factors DiseaseCase – Control
CohortFactors Disease
Cause Effect
Case-Control V.S. Cohort
59
Conducting a Case-control
Study• ค�น้หา “Cases”
• ท,าการค�ดเล!อก “Controls” โดยเล!อกจากกล-%มป้ระชากรท�&เป้'น้แหล%งก,าเน้�ดเด�ยวิก�น้ก�บ Cases ใน้การศึ�กษา (study base)
• วิ�ดล�กษณะการม�ป้7จจ�ยเสั�&ยง “ exposure ” ท�&สัน้ใจใน้กล-%ม cases และ controls
• เป้ร�ยบเท�ยบ exposure status ระหวิ%าง 2 กล-%ม
• ค,าน้วิณหา Odds Ratio
60
Sources of Cases• Population-based (ผู้��ป้�วิยใน้ช-มชน้)
• identify and enroll all incident cases from a defined population• e.g., disease registry, defined geographical area, vital records
• Hospital-based (ผู้��ป้�วิยท�&มาร�กษา)– identify cases where you can find them
• e.g., hospitals, clinics.
– But……• issue of representativeness?• prevalent vs incident cases?
61
Sources of Controls• Population-based Controls
• ideal, represents exposure distribution in the general population, e.g.,
– driver’s license lists (16+)– Medicare recipients (65+)– Tax lists– Voting lists– Telephone RDD survey
62
Sources of Controls• Hospital-based Controls
– Hospital-based case control studies used when population-based studies not feasible
– More susceptible to bias
– Advantages• similar to cases? (hospital use means similar SES, location)• more likely to participate (they are sick)• efficient (interview in hospital)
– Disadvantages• they have disease?
– Don’t select if risk factor for their disease is similar to the disease under study e.g., COPD and Lung CA
• are they representative of the study base?
63
Other Sources of Controls• Relatives, Neighbors, Friends of Cases
– Advantages• similar to cases wrt SES/ education/ neighborhood• more willing to co-operate
– Disadvantages• more time consuming• cases may not be willing to give information?• may have similar risk factors (e.g., smoke, alcohol, golf)
64
Case : Control Ratio • อ�ตราสั%วิน้ของ case : control โดยท�&วิไป้อย�%ระหวิ%าง 1:1 ถ้�ง 1:4
• ถ้�าจ,าน้วิน้ case เท%าเด�ม
– การเพี�&มจ,าน้วิน้ control จะช%วิยเพี�&ม precision ของ Odds ratio
– แต%การเพี�&มจ,าน้วิน้ control ให�มากกวิ%า 4 ต%อ 1 case พีบวิ%าไม%ได�เพี�&ม precision
มากเท%าไรและอาจไม%ค-�มก�บต�น้ท-น้ท�&เพี�&มข�:น้
• ถ้�าจ,าน้วิน้รวิมของ Case ก�บ Control คงท�&
– อ�ตราสั%วิน้ 1:1 จะท,าให�ได� precision ของ Odds ratio มากท�&สั-ด
65
Cases
Controls
Lung
Cancer
#50 cases
Lung
Cancer
#50 casesSmoke
# 45
Not
Smoke
# 5
NO Lung
Cancer
#200
controls
NO Lung
Cancer
#200
controls
ExposedUnexposedSmo
ke
# 99
Not
Smoke
# 101
ExposedUnexposed
Ex: Smoking and Lung Cancer
66
Smoke
Not smoke
50 95
10 99
CA LungNo CA
Cohort Study
Case – Control Study
• Do not have incidence in exposed & incidence in non exposed • Cannot calculate the RR directly
CA LungNo CA
Smoke
Not smok
e
500 9,500
100 9,900
10,000
10,000
RR = (500/10000)/(100/9900) = 5
OR = (50/10)/(95/99) = 5.2
OR = (500/9500)/(100/9900) = 5.2
Ex: A Food Poisoning in a School (1)(Case-control study)
Case Control
Ate ice cream 40 17
Not eat 15 38
Total 55 55
Odds of eating ice cream in cases = (40/55) / (15/55) = 40/15
= 2.67
Odds of eating ice cream in control = (17/55) / (38/55) = 17/38
= 0.45
Odds ratio = 2.67 / 0.45 = 5.9 67
Ex: A Food Poisoning in a School (2)(Case-control study)
How to interpret odds ratio of 5.9 ???
In conventional case-control study: case vs. non-case
1. Study cases represent cases in population2. Study control represent non-case in population
>>> OR of 5.9 means “Odds of disease among the exposed is 5.9 times of that among the unexposed”
>>> OR ≈ RR
In population-based case-control study: • With case-cohort sampling
• With density sampling
>>> OR = RR
>>> OR = IRR
68
If 1.+2.+ 3. Rare disease
69
• Quick and cheap (relatively)– so ideal for outbreaks
• Can study rare diseases (or new)
• Can evaluate multiple exposures
Case-control Study - Advantages
70
Case-control Study - Disadvantages• uncertain of Exposure-Disease relationship (esp.
timing)• cannot estimate disease incidence• inefficient if exposures are rare• Selection Bias
– Much worry about representativeness of controls– selective survival if not using incidence cases
• Information Bias (recall bias)
71
Acknowledgement
• Dr. Chuleeporn Jiraphongsa• Dr. Ram Rungsin• Dr. Darin Areechokchai• Dr. Mathew J. Reeves
72