Cohort Study - interfetpthailand · of cohort study design. • Students are able to identify types of cohort studies • Students are able to design a simple cohort study • Student

Cohort StudyCohort Study

Ram Rangsin, MD MPH Ram Rangsin, MD MPH DrPHDrPHDepartment of Military and Community MedicineDepartment of Military and Community MedicinePhramongkutklao College of Medicine, THAILANDPhramongkutklao College of Medicine, THAILAND

1

Learning objectivesLearning objectives

•• Students are able to identify the fundamental concepts Students are able to identify the fundamental concepts

of cohort study design.of cohort study design.

•• Students are able to identify types of cohort studiesStudents are able to identify types of cohort studies

•• Students are able to design a simple cohort studyStudents are able to design a simple cohort study

•• Student are able to understand the concept of Poisson Student are able to understand the concept of Poisson

Regression Model using in data analysis of the Cohort Regression Model using in data analysis of the Cohort

Study Study

2

Epidemiological StudiesEpidemiological Studies

•• Describe the problemDescribe the problem

–– How large is the problem?How large is the problem?

–– How dose the problem distribute in the How dose the problem distribute in the population?population?population?population?

–– Descriptive StudyDescriptive Study

•• TimeTime

•• PlacePlace

•• PersonPerson

3

Classification epidemiological studyClassification epidemiological study

ObservationObservation(natural exposure)(natural exposure)

ExperimentExperiment(exposure given by researcher)(exposure given by researcher)

Descriptive Descriptive (no comparison group)(no comparison group)

Analytic Analytic (with comparison group)(with comparison group)

CrossCross--sectionalsectional Case controlCase control CohortCohort

4

Hierarchy of Hierarchy of Epidemiological studiesEpidemiological studies

Analytic

Experiment•• Clinical TrialClinical Trial

•• CohortCohort

•• CrossCross--sectional / Case controlsectional / Case control

Descriptive

Analytic•• CrossCross--sectional / Case controlsectional / Case control

•• CrossCross--sectionalsectional

•• Case seriesCase series

•• Case reportCase report

5


•• The most powerful observational study for The most powerful observational study for

identifying an association between risk factors identifying an association between risk factors

and a diseaseand a diseaseand a diseaseand a disease

•• The most time consuming The most time consuming

•• The most expensive The most expensive

6

“COHORT”“COHORT”“COHORT”“COHORT”

A unit of A unit of 300300--600 600 men men in the ancient Roman armyin the ancient Roman army

7

A Roman CohortA Roman Cohort

Two centuries made one maniple and three maniples made up one cohort. Two centuries made one maniple and three maniples made up one cohort.

8

“COHORT”“COHORT”

in Epidemiologyin Epidemiology

A group of persons who are followed A group of persons who are followed over timeover time

9

•• Start with a group of people without the diseaseStart with a group of people without the disease

•• Then divide people based on the basis of the Then divide people based on the basis of the

exposure to a suspected risk factorexposure to a suspected risk factor


•• Follow the Follow the “whole group”“whole group” for a period of timefor a period of time

•• Then asses the disease occurrence outcomeThen asses the disease occurrence outcome

10

EffectEffectCauseCause

Using epidemiology to identify Using epidemiology to identify the cause of a diseasethe cause of a disease


••RISK FACTORRISK FACTOR ••CigaretteCigarette

••DISEASE DISEASE ••Lung CancerLung Cancer

11


FactorsFactors DiseaseDisease

CaseCase--ControlControl

FactorsFactors DiseaseDisease

Cohort12

FactorsFactors

DiseaseDisease

ExposedExposed Not ExposedNot Exposed

DevelopDevelopDiseaseDisease

Do NotDevelopDisease



13

When we are conducting a cohort When we are conducting a cohort study,study,

we are dealing with we are dealing with “INCIDENCE”.“INCIDENCE”.

14

IncidenceIncidence

•• Population Population 1010,,000 000 peoplepeople

•• Blood pressure measurementBlood pressure measurement

•• 1000 1000 people had BP >= people had BP >= 140140//90 90 mmHgmmHg

•• Please design a study to identify Please design a study to identify incidence of HT in this populationincidence of HT in this population

15

Type of Cohort StudyType of Cohort Study

I. Concurrent Cohort StudyI. Concurrent Cohort Study

(Prospective Cohort Study)(Prospective Cohort Study)

ExposedExposed Not ExposedNot Exposed20201919

16






20201919

20202929

Type of Cohort StudyType of Cohort Study

II. Retrospective Cohort StudyII. Retrospective Cohort Study

(Take advantage of records collected)(Take advantage of records collected)

ExposedExposed Not ExposedNot Exposed19701970

17






19701970

20201919


SmokeSmoke

##500 500 personspersons

Not SmokeNot Smoke


19701970 Persons without the disease!!!!Persons without the disease!!!!

18

DiseaseDisease No DiseaseNo Disease


Lung CancerLung Cancer

# # 4545

No Lung CancerNo Lung Cancer

# # 455455


DiseaseDisease No DiseaseNo Disease

Lung CancerLung Cancer

# # 11No Lung CancerNo Lung Cancer

# # 499499

20201919

CA LungCA Lung No CANo CA

SmokeSmoke

Not smokeNot smoke

4545 455455

11 499499

500500

500500

•• Incidence of Smoker who develop Lung Cancer = Incidence of Smoker who develop Lung Cancer = 45 45 //500500

•• Incidence of Non Incidence of Non --Smoker Smoker whodevelopwhodevelop Lung Cancer = Lung Cancer = 1 1 //500500

•• Relative Risk of smoking for Lung Cancer = = Relative Risk of smoking for Lung Cancer = = 4545

•• Those who smoked were Those who smoked were 45 45 times more likelytimes more likely to get to get

lung cancer.lung cancer.

4545//500500

11//500500

19

Relative RiskRelative Risk

CA LungCA Lung No CANo CA

SmokeSmoke AA BB A+BA+BA+BA+B

Not smokeNot smoke CC DD C+DC+DC+DC+D

Relative Risk Relative Risk = = A/A+BC/C+D

20

Interpretation of Relative Risk Interpretation of Relative Risk (RR)(RR)

Interpretation of Relative Risk Interpretation of Relative Risk (RR)(RR)

•• Relative Risk of smoking for CA Lung = Relative Risk of smoking for CA Lung = 4545

•• Those who smoked Those who smoked were were 45 45 times more times more likelylikely to develop lung cancer than those to develop lung cancer than those likelylikely to develop lung cancer than those to develop lung cancer than those who did not smoke.who did not smoke.

21

Cohort StudyCohort StudyCohort StudyCohort Study

Comparison between

“a group of persons “a group of persons with with

a factor a factor ---- Exposed”Exposed”a factor a factor ---- Exposed”Exposed”

VS

“a group of persons “a group of persons without without

the factor the factor ---- NonNon--exposed”exposed”

22

Measurement of AssociationsMeasurement of Associations

•• CrossCross--sectional sectional Prevalence Rate RatioPrevalence Rate Ratio

•• CaseCase--Control Control

•• Cohort Cohort

Odds RatioOdds Ratio

Relative RiskRelative Risk

23

Advantages Advantages of a cohort studyof a cohort study

•• Temporal sequence Temporal sequence (exposure occur prior to (exposure occur prior to

the disease)the disease) can be more clearly establishedcan be more clearly established

•• Well suited for assessing the effect of RARE Well suited for assessing the effect of RARE

EXPOSURE (e.g. Radiation,)EXPOSURE (e.g. Radiation,)EXPOSURE (e.g. Radiation,)EXPOSURE (e.g. Radiation,)

–– Persons are enrolled on the basis of exposurePersons are enrolled on the basis of exposure

24

Advantages Advantages of a cohort studyof a cohort study

•• Able to examine multiple diseases outcome of a Able to examine multiple diseases outcome of a

single exposuresingle exposure

–– The Nurse Health Study, USAThe Nurse Health Study, USA

–– 120120,,000 000 female nursesfemale nurses

–– Exposure: Exposure: Oral Contraceptive PillOral Contraceptive Pill

–– Outcomes: Outcomes:

•• Breast cancerBreast cancer

•• Ovarian CancerOvarian Cancer

•• Malignant melanomaMalignant melanoma

25

DisadvantagesDisadvantages

•• Insufficient for the evaluation of rare diseasesInsufficient for the evaluation of rare diseases

•• Extremely expensive and time consuming Extremely expensive and time consuming

(Prospective)(Prospective)

Required the availability of adequate records Required the availability of adequate records •• Required the availability of adequate records Required the availability of adequate records

(Retrospective)(Retrospective)

•• Loss to followLoss to follow--upup

26

Rate Rate

•• The central tool of Epidemiology is the The central tool of Epidemiology is the comparison of RATEScomparison of RATES

•• RATE = RATE = NumeratorNumerator•• RATE = RATE = NumeratorNumerator DenominatorDenominator

–– Mortality RateMortality Rate–– PrevalencePrevalence–– IncidenceIncidence

27

There are two ways of measuring

11) ) Cumulative incidence Cumulative incidence

= number of new case in specified time = number of new case in specified time population at risk in specified timepopulation at risk in specified time

X X 1010(n)(n)

Measuring the incidenceMeasuring the incidence

population at risk in specified timepopulation at risk in specified time

= = 4040 = = 11..2525 //11,,000000 3232,,000000

28

22) ) Incidence density or Incidence rateIncidence density or Incidence rate

• Adding “TIME Dimension” into the denominator


• ppersonerson--month, month, ppersonerson--yearyear

•• 1 1 ppersonerson--year year = Following = Following 1 1 person for person for 1 1 year periodyear period

•• 10 10 ppersonerson--year year = Following = Following 1 1 person for person for 10 10 year periodyear period

= Following = Following 10 10 persons for persons for 1 1 year periodyear period

“Person“Person--time”time”

1 1 pp--yy 1 1 pp--yy

1 1 pp--yy

29

•• 22) Incidence density or Incidence rate) Incidence density or Incidence rate

= = Number of new case in specified time Number of new case in specified time

PersonPerson--years of observation which is disease freeyears of observation which is disease free


X 10(n)

PersonPerson--years of observation which is disease freeyears of observation which is disease free

•• If If 100 100 subjects are followed for subjects are followed for 1 1 yearyear and and 20 20 develop diseasedevelop disease, the incidence density is, the incidence density is

•• 20 20 cases/cases/100 100 personperson--years of observationyears of observation

•• 20 20 = = 20 20 / / 100 100 personperson--yearsyears

100 100 personperson--yearsyears1 1 pp--yy 1 1 pp--yy

1 1 pp--yy

30

Relationship between cigarette smoking and incidence rate of stroke

in a study of 118,539 population in over 8 years period

SmokingSmoking No. of stroke PersonNo. of stroke Person--yearsyears Incidence rateIncidence rate

of observationof observation //100100,,000 000 personperson--yearsyearsof observationof observation //100100,,000 000 personperson--yearsyears

Never 70 395,594 17.7

Ex-smoker 65 232,712 27.9

Smoker 139 280,141 49.6

TotalTotal 274274 908908,,477477 3030..22

31

Database of Database of 118118,,539 539 subjectssubjects

ID Age smoking Stroke Enter Last

Contact Person-Year

1 18 No No 1990 1998 8

2 36 No No 1990 1992 2

3 50 Yes Yes 1991 1998 7

4 42 Ex No 1993 1995 2 4 42 Ex No 1993 1995 2

. . . . . . .

118,539 24 Yes No 1993 1998 5

Total 908,477

32

Relationship between cigarette smoking Relationship between cigarette smoking and incidence rate of strokeand incidence rate of stroke

in a study of in a study of 118118,,539 539 population population in over in over 8 8 years periodyears period

SmokingNo.

of stroke

Person-years

of observation

Incidence rate

/100,000

person-years

Relative Risk

33

Never 70 395,594 17.7

Ex-smoker 65 232,712 27.9

Smoker 139 280,141 49.6

Total 274 908,477 30.2

Data Analysis Data Analysis in Cohort Studyin Cohort Study

• Incidence – Cumulative incidence = … / 100 persons

– Incidence Rate (Density) = … / 100 person-years

• Relative Risk: Univariate Analysis• Relative Risk: Univariate Analysis– Cumulative incidence = Relative Risk (… / 100 persons)

– Incidence Rate (Density) = Relative Risk (… / 100 person-years)

• Relative Risk: Multivariate Analysis– Cumulative incidence = Adjusted Relative Risk

Multiple Logistic Regression Model

– Incidence Rate (Density) = Adjusted Relative Risk

Poisson Regression Model

34

Poisson DistributionPoisson Distribution

18381838: : Recherches sur la probabilité Recherches sur la probabilité des jugements en matières des jugements en matières criminelles et matière civilecriminelles et matière civile

35

“Research on the Probability of Judgments in Criminal “Research on the Probability of Judgments in Criminal and Civil Matters”and Civil Matters”


40

50

60

70

CountsCounts

Number of occurrences of an event in a give time span Number of occurrences of an event in a give time span

36

0

10

20

30

0 1 2 3 4 5 6 7 8 9

Full Moon

Number of crimes per day (N=Number of crimes per day (N=183 183 full moon days )full moon days )

CountsCounts

Mean=1.4


60

80

100

120

CountsCounts


37

0

20

40

60

0 1 2 3 4 5 6 7 8 9

New Moon

Number of crimes per day (N=Number of crimes per day (N=186 186 new moon days )new moon days )

CountsCounts

Mean=0.5


60

80

100

120

CountsCounts


38

0

20

40

60

0 1 2 3 4 5 6 7 8 9

Full Moon New Moon

Number of crimes per day (N=Number of crimes per day (N=183 183 full moon days & N=full moon days & N=186 186 new moon days )new moon days )

CountsCounts

Mean=1.4 Mean=0.5

Poisson ModelsPoisson Models

•• Incidence rates can be calculated Incidence rates can be calculated from the results as follows:from the results as follows:

•• IR = e IR = e bb00++ bb

11xx

11+ b+ b

22xx

22+….++….+bb

kkxx

kk•• IR = e IR = e bb00++ bb

11xx

11+ b+ b

22xx

22+….++….+bb

kkxx

kk

•• Relative Risk of XRelative Risk of X11= = eeBB11

Framingham studyFramingham study

•• Framingham study of Framingham study of

cardiovascular diseasecardiovascular disease

–– Individuals Individuals 30 30 –– 62 62 years old in years old in –– Individuals Individuals 30 30 –– 62 62 years old in years old in

community at risk for diseasecommunity at risk for disease

–– Framingham, MA, Framingham, MA, 1948 1948 to presentto present

44,,46946922,,44544522,,024024RespondentsRespondents

66,,50750733,,4334333,0743,074Random SampleRandom Sample

Total Total No. No. WomenWomen

No. No. MenMen

Framingham study

55,,12712722,,84584522,,282282Total free of CHDTotal free of CHD

734734427427307307Volunteers free of CHD Volunteers free of CHD

44,,39339322,,41841819751975Respondents free of CHDRespondents free of CHD

740740428428312312VolunteersVolunteers

Framingham studyFramingham study

Cholesterol level CHD No CHD Total

`High'`High' 5757 305305 362362

`Low' <250`Low' <250 7171 10981098 11691169

Exposure assessmentExposure assessment

•• Exposed VS NonExposed VS Non--exposedexposed

•• Fixed ExposureFixed Exposure

•• TimeTime--dependent Exposure(Exposure dependent Exposure(Exposure level changes through time)level changes through time)level changes through time)level changes through time)

43

Fixed ExposureFixed Exposure

•• Exposure do not change over timeExposure do not change over time

–– Sex (Male / Female)Sex (Male / Female)

–– Blood group (A / B / O / AB)Blood group (A / B / O / AB)–– Blood group (A / B / O / AB)Blood group (A / B / O / AB)

–– Race (White / Black / Asian)Race (White / Black / Asian)

–– Expose to radiation from the power plant Expose to radiation from the power plant

explosionexplosion

–– Adult heightAdult height

44

TimeTime--dependentdependent

• Exposure level changes over time

– Body weight

– Alcohol consumption– Alcohol consumption

– Blood pressure level

45

Outcome assessmentOutcome assessment

•• DiseaseDisease

•• Specify clearly what is your Specify clearly what is your final outcomefinal outcome

–– DiseaseDisease–– DiseaseDisease

–– DeathDeath

–– Intermediate outcomesIntermediate outcomes

•• CDCD44+ count+ count

•• Increased Increased CreatinineCreatinine

46

Conducting a Cohort StudyConducting a Cohort Study

• Selecting a group of people without

the disease

•• Defining the Exposed groupDefining the Exposed group•• Defining the Exposed groupDefining the Exposed group

• Defining the Non-exposed group

•• Evaluate the disease outcome Evaluate the disease outcome

among both Exposed and Nonamong both Exposed and Non--

exposedexposed

• Calculating Relative Risk

47

•• You are interested in the association You are interested in the association

between blood cholesterol level and between blood cholesterol level and


between blood cholesterol level and between blood cholesterol level and

coronary artery heart diseasecoronary artery heart disease

•• Please conduct a cohort study to verify Please conduct a cohort study to verify

the association the association

48

•• What population would you like to start What population would you like to start with?with?

Conducting a Cohort StudyConducting a Cohort StudyConducting a Cohort StudyConducting a Cohort Study

with?with?

49

•• How can you identify those who will How can you identify those who will be the “study population”?be the “study population”?


be the “study population”?be the “study population”?

50

•• How can you identify exposed and How can you identify exposed and nonnon--exposed groups?exposed groups?


nonnon--exposed groups?exposed groups?

51

•• What is you followWhat is you follow--up plan?up plan?

–– What is you outcome of interest?What is you outcome of interest?


–– What is you outcome of interest?What is you outcome of interest?

–– How often you would like to asses the How often you would like to asses the

outcome?outcome?

–– How long will you follow the population?How long will you follow the population?

52

•• What is you plan for the analysis?What is you plan for the analysis?


–– What will be the measurement of What will be the measurement of

association from your study?association from your study?

–– What would you like to compare? What would you like to compare?

53

ReferencesReferences

•• GordisGordis, L, L. . 20002000. . Epidemiology, Epidemiology, 22nd nd eded.., Philadelphia, Philadelphia: : WW..BB. . SaundersSaunders..

•• Last, MLast, M. . etet. . alal. . 20012001. . Dictionary of Dictionary of Epidemiology, Epidemiology, 44th th eded.., New York; Oxford , New York; Oxford University PressUniversity Press..University PressUniversity Press..

•• HennekensHennekens, H, H. . HH.., et, et. . alal. . 19871987, Epidemiology in , Epidemiology in Medicine, BostonMedicine, Boston: : Little, BrownLittle, Brown..

•• [email protected]@pmk.ac.th

54

Documents

Cohort Study - interfetpthailand · of cohort study design. • Students are able to identify types of cohort studies • Students are able to design a simple cohort study • Student