11

Study Design in

Clinical Research

Study Design in

Clinical Research

Jarir At Jarir At ThobariThobari, MD, DPharm, PhD, MD, DPharm, PhDFacultyFaculty of Medicine, of Medicine, UGMUGM CrossCrossCrossCross----

sectionalsectionalsectionalsectional

Case Control StudyCase Control StudyCase Control StudyCase Control Study

CauseCauseCauseCause

Cohort StudyCohort StudyCohort StudyCohort Study

DiseaseDiseaseDiseaseDisease

Case Case Case Case reportreportreportreportCase Case Case Case seriesseriesseriesseriesSurvey Survey Survey Survey

Descriptive

OutcomeOutcomeOutcomeOutcome ExposureExposureExposureExposure

RCTRCT QuasiQuasiAnalytic

OBSERVATIONALOBSERVATIONAL

RESEARCH DESIGNRESEARCH DESIGNRESEARCH DESIGNRESEARCH DESIGNRESEARCH DESIGNRESEARCH DESIGNRESEARCH DESIGNRESEARCH DESIGN

EXPERIMENTALEXPERIMENTAL

• DescriptiveDescriptiveDescriptiveDescriptive: Most basic : Most basic : Most basic : Most basic demographic demographic demographic demographic studies studies studies studies

(without comparison group)(without comparison group)(without comparison group)(without comparison group)

• AnalyticalAnalyticalAnalyticalAnalytical: Comparative studies testing an : Comparative studies testing an : Comparative studies testing an : Comparative studies testing an

hypothesis (with comparison group)hypothesis (with comparison group)hypothesis (with comparison group)hypothesis (with comparison group)

• crosscrosscrosscross----sectionalsectionalsectionalsectionala snapshot; no idea on causea snapshot; no idea on causea snapshot; no idea on causea snapshot; no idea on cause----andandandand----effect relationshipeffect relationshipeffect relationshipeffect relationship

• casecasecasecase----controlcontrolcontrolcontrolretrospective; causeretrospective; causeretrospective; causeretrospective; cause----andandandand----effect relationship can be inferredeffect relationship can be inferredeffect relationship can be inferredeffect relationship can be inferred

• cohortcohortcohortcohortprospective; causeprospective; causeprospective; causeprospective; cause----andandandand----effect relationship can be inferredeffect relationship can be inferredeffect relationship can be inferredeffect relationship can be inferred

Observational Study Hierarchy of Epidemiologic Study DesignHierarchy of Epidemiologic Study Design

�Tower & Spector, 2007

22

Grimes & Schulz, 2002Grimes & Schulz, 2002

Observational studyObservational study Cause and effect relationshipCause and effect relationship

1. Strength of association

2. Biologic plausibility

3. Dose-response relationship

4. Temporality

5. Consistency

6. Reversibility

CROSS SECTIONAL

STUDY

CROSS SECTIONAL

STUDY �� involvesinvolves

�� identifying a defined population at a identifying a defined population at a

particular point in timeparticular point in time

�� measuring a range of variables on an measuring a range of variables on an

individual basis individual basis

DefinitionDefinitiona type of observational or descriptive study, a type of observational or descriptive study, the research has no control over the exposure the research has no control over the exposure of interest (ex. diet).of interest (ex. diet).

33

A “Snapshot”

Cross Sectional StudyCross Sectional Study

Data collected at a single point in timeData collected at a single point in time

Describes associationsDescribes associations

PrevalencePrevalence

Uses of cross-sectional studies

Prevalence surveyPrevalence surveyThe studies are commonly used to The studies are commonly used to describe the prevalence of the problems describe the prevalence of the problems and its distributionand its distribution

Uses of cross-sectional studies

DDescribe population characteristicsescribe population characteristics�� They are also commonly used to They are also commonly used to

describe population characteristics, often describe population characteristics, often in terms of person (who?) and place in terms of person (who?) and place (where?)(where?)

�� To describe various age groups in the To describe various age groups in the population in terms of food and nutrient population in terms of food and nutrient intake and range of other personal and intake and range of other personal and lifestyle characteristics.lifestyle characteristics.

Prevelence use of herbal and its predictors

44

Case fatality dengue Fever

�� KAP (KAP (knowledgesknowledges, attitudes, and , attitudes, and practices) study:practices) study:

�� KAP studies are purely descriptive and help KAP studies are purely descriptive and help to build up a better understanding of the to build up a better understanding of the behavior of the population, without behavior of the population, without necessarily relating this to any disease or necessarily relating this to any disease or health outcome.health outcome.

Uses of cross-sectional studies

Management toolManagement tool�� health service managers and planners health service managers and planners

may make use of crossmay make use of cross--sectional survey sectional survey to assess utilization and effectiveness of to assess utilization and effectiveness of serviceservice

Uses of cross-sectional studies

Development of hypothesis:

� Hypotheses on the causes of disease may be developed using data from cross-sectional study survey.

Uses of cross-sectional studies

55

CASE CONTROL STUDYCASE CONTROL STUDY � Retrospective

� Two groups, selected by disease/event status (diseased: cases, non-diseased: control)

� Cases and controls are compared for differences in the prevalence of exposures/factors of interest

Case Control Study: The Basics

Exposure +Exposure +

Exposure Exposure --CasesCases

Exposure +Exposure +

Exposure Exposure --ControlControl

Case control: Design

Exposure rare (no!)Exposure rare (no!) smokingsmoking

Rare outcome/casesRare outcome/cases Ca pulmo (cases)Ca pulmo (cases)

When case control study is used?

Extended time lapse between exposure and outcome

66

Prevalent case-controlIncident case-controlNested case-control

Case-cohortMatched case-control

Case control: variations

Usually based on Usually based on outcomeoutcome

�� Incident outcomesIncident outcomes�� Prevalent outcomesPrevalent outcomes

Case control: cases selection

�� Definition: A sample of the source Definition: A sample of the source population that gave rise to the cases.population that gave rise to the cases.

�� Purpose: To estimate the exposure Purpose: To estimate the exposure distribution in the source population that distribution in the source population that produced the cases.produced the cases.

Case control: control

Selection criteria determined before study started.

Subjects are chosen from the same population and the same survey and having similar factors to get distortion

To detect the outcome

To reduce the bias

Case control: control selection

77

�� Pairing of one or more Pairing of one or more controls to each case controls to each case

on the basis of their on the basis of their similarity with respect to similarity with respect to

selected variablesselected variables

�� Any characteristic or Any characteristic or attribute of an individual attribute of an individual may serve as a basis of may serve as a basis of matchingmatching

�� age, gender, race, age, gender, race, blood group, marital blood group, marital status, weight, hospital status, weight, hospital of admission, parity, inof admission, parity, in--come, neighborhood, come, neighborhood, occupation, etcoccupation, etc..

Matching

� Measure of association: odds ratio

� Analysis methods: odds ratio, logistic regression

� Cases-control studies do not provide an estimate of the general risk of the outcome

Case control: analysis

Cases ControlCases Control TotalTotal

ExposurExposuree

++ aa bb a+ba+b

-- cc dd c+dc+d

a+ca+c b+db+d

Exposure odds ratio

Odds 1 =a ( a + c )c ( a + c )

ac

=

Odds 0 =b ( b + d )d ( b + d )

bd

=

OR = a/cb/d = a.d

b.c

Case control: Design

BMJ 2010;341:4444

88

Case control: Design

�� Efficient for rare diseases and diseases with Efficient for rare diseases and diseases with long induction and latent period.long induction and latent period.

�� Can evaluate many risk factors for the same Can evaluate many risk factors for the same disease. So, good for diseases about which disease. So, good for diseases about which little is known.little is known.

Case control: strength

�� Inefficient for rare exposuresInefficient for rare exposures�� Vulnerable to bias because of retrospective Vulnerable to bias because of retrospective

nature of studynature of study�� May have poor information on exposure because May have poor information on exposure because

retrospectiveretrospective�� Difficult to infer temporal relationship between Difficult to infer temporal relationship between

exposure and disease exposure and disease

Case control: weakness

Cohort studyCohort study

99

Study group

Exposure status before outcome coming

Control group: without exposure

Follow up until:

• Outcome coming• Subyek death• The end of the study• Subjects could not follow anymore

2 type of cohort study2 type of cohort study

ProspectiveProspectiveRetrospective/

historical:

Exposure determined during study started

Exposure determined before study started

Subject

Exposure +

Exposure -

Present Future

Prospective c.s.

Outcome Exposure Exposure Exposure Exposure ----

PresentPresentPresentPresent

Historical cohortHistorical cohortHistorical cohortHistorical cohort

In the pastIn the pastIn the pastIn the past

SubjectSubjectSubjectSubject

Exposure +Exposure +Exposure +Exposure +OOOOuuuuttttccccoooommmmeeee

1010

Cough

No cough

Cough

No cough

Is there relation between cough and captopril?

Is there relation between cough and captopril?

Present

Follow-upHypertension

Captopril (+)

Captopril (-)

Relative Risk (RR)Relative Risk (RR)

outc+ outc-

Exposed a b

Unexposed c d

a------a + b

RR = ----------------c

-------c + d

Exposure +

Exposure -

Total

a=18

c=29

b=200

d=151

218

180

47 351

a/(a+b) = 0.08

c/(c+d) = 0.16

outcome(+)

Outcome (-)

Total Proportiondeceased

RR 0.08/0.16 = 0.5

Hypothesis testing

�� When the exposure groups in a cohort When the exposure groups in a cohort study are defined at the onset of the study are defined at the onset of the study without movement of individuals study without movement of individuals between exposure groups, the between exposure groups, the exposure groups are referred to as exposure groups are referred to as fixed cohorts. (occupational. war)fixed cohorts. (occupational. war)

FIXED COHORT

1111

STRUCTURE COHORT STUDIESCOHORT STUDIES

�� Fixed CohortFixed Cohort

Exposure

(+)(+)

((--))

x

x

x

X = outcomeX = outcome

Relative Relative risk risk

= (= (2/32/3)/()/(1/31/3) )

= 2.0= 2.0

�� The other type of prospective cohort The other type of prospective cohort study is the open or dynamic cohort study is the open or dynamic cohort study wherein individuals can be study wherein individuals can be unexposedunexposed at one time period and at one time period and unexposedunexposed at another time. The at another time. The personperson--time analysis can take this into time analysis can take this into account in calculating account in calculating incidence incidence densitiesdensities

OPEN OR DYNAMIC COHORT

OPEN OR DYNAMIC COHORT STRUCTURE

1212

COHORT STUDIESCOHORT STUDIES

�� Dynamic CohortDynamic Cohort

Exposure

(+)(+)

((--))

Rel. Risk

= 2/3/2/3 =1

or

2/5py/2/10py

= 2.0Years

XX

XX

� Linked Medicare data, state pharmacy assistance program data, and state cancer registry

� Cohort and Exposure� Patients enrolled in Medicare and state pharmacy

assistance program� New user design� Exposed group:

� Initiated statins (no use in the past 12 month)� Dispensed ≥ 3 prescriptions for statins during the first 180 days

after the initation (chronic use)

�

DATA SOURCES COHORT STUDIES

DATA SOURCES COHORT STUDIES

� Comparison� Initiated glaucoma drugs� Dispense ≥ 3 prescriptions for glaucoma drugs

during the first 180 days after the initation� Avoided non-drug users (active comparison

drug)� Similar patients (patients who used other

preventive medications)

DATA SOURCES COHORT STUDIES

DATA SOURCES COHORT STUDIES

1313

� Outcome� Identified using cancer registry

� Confounding adjustment� Multivariate and propensity score� Demographic, comorbidity, drug use, and

preventive service use

OUTCOMEOUTCOMEBASIC CHARACTERISTICSBASIC CHARACTERISTICS

Number of Cases, Person-Years, and Incidence Rate of Cancer

Number of Cases, Person-Years, and Incidence Rate of Cancer

Effects of Statin Use on Selected Cancers

Effects of Statin Use on Selected Cancers

1414

Advantages of Cohort StudiesAdvantages of Cohort Studies

- Can establish population-based incidence

- Accurate relative risk (risk ratio) estimation

- Can examine rare exposures (asbestos > lung cancer )

- Temporal relationship can be inferred (prospective design)

- Time-to-event analysis is possible

- Can be used where randomization is not possible

- Magnitude of a risk factor’s effect can be quantif ied

- Selection and information biases are decreased

- Multiple outcomes can be studied (smoking > lung cancer, COPD, larynx cancer)

Disadvantages of Cohort StudiesDisadvantages of Cohort Studies

- Lengthy and expensive

- May require very large samples

- Not suitable for rare diseases

- Not suitable for diseases with long-latency

- Unexpected environmental changes may influence the association

- Non-response, migration and loss-to-follow-up bias es

- Sampling, ascertainment and observer biases are st ill possible

1515

EXPERIMENTAL DESIGNRandomized Controlled Trial

Definitions

�Experimental group (EG)

The group that is exposed to some sort of a change or manipulation in the IV

Control group (CG)

A group not exposed to changes or manipulations that serves as a baseline comparison to the experimental group

�KEY Features of Experimental Designs

�at least two groups of individuals / participants

�random assignment to groups

�an independent variable manipulated by the experimenter

�dependent variable being measured

�1

�2

�3

�4

RCT (randomized controlled clinical trial)

Randomized Randomisation

Controlled

Inclusion

Procedure

Outcome

Clinical TrialIntervention

Vs. control group

1616

RCT

Investigator Participants

Clinical Manoeuvre

RCT

Quantitative Comparative Control Experiment

Measuring outcomequantitatively

Comparing 2 or more

intervention

All variables areClosely controlled

Objectives of RCT

RCT

Intervention Sample population

Efficacy Safety

(RCT-parallel design)

(RCT cross(RCT cross--over over design)design)

(RCT factorial (RCT factorial design)design)

DESIGN

RCTs according to how the participants are exposed to the interventions

1717

RCTRCT--parallel designparallel design

PatientPatient

eligibleeligible RandomRandom

Treatment BTreatment B

OUTCOME

treatment Atreatment A

RCT crossRCT cross--over designover design

Treatment BTreatment B

Treatment ATreatment A

washedwashedoutout

eligibleeligible

PatientPatient

Treatment ATreatment A

RandomRandom

Treatment BTreatment B

OUTCOME

RCTRCT--factorial designfactorial design

PatientPatient

RandomRandomeligibleeligible

OUTCOME

Tx BTx B

tx Atx A

tx A + tx Btx A + tx B

RCTs according to whether the investigators and participants know

which intervention is being assessed

Open trials

Triple and quadruple-blind trials

Double blind trials

Single blind trials

1818

Review of Steps for RCT Design

1. Select sample from population

2. Measure baseline variables

3. Randomize

4. Apply interventions & placebo

5. Follow-up cohorts

6. Measure outcomes

EXPERIMENTAL DESIGNQuasi

Notasi

�� XX = treatment= treatment

�� O O = observasi= observasi�� ---------------- = non randomised= non randomised

Design quasi experimental yang tidak memungkinkan causal inferences

1. One group posttest only design

3. One group pretest- posttest design

2. Posttest only designwith non-equivalent group

1919

Tanpa pretest observationTanpa pretest observationTanpa kelompok kontrolTanpa kelompok kontrol

1. One group posttest only design

Design: X O

Pemberian pengobatanEvaluasi hasil pendidikanEnvironmental toxicology

Kapan dilakukan?

Tanpa pretest observationTanpa pretest observationEfek treatment vs. selection effectEfek treatment vs. selection effectnon ekuivalennon ekuivalen

Design:

progress suatu treatment tidak bersamaan

Kapan dilakukan?

X O

O

2. Posttest only design with non-equivalent group

Contoh:Contoh:

X 12 O

X 9 O

X 6 O

X 0 O

Mengamati hasil counseling 12 bulan, 9 bulan,6 bulan, dan tanpa counseling

Pada satu kelompokPada satu kelompokDisertai pretest observationDisertai pretest observation

Design:

Apa kelemahannya ?

3. The One-Group Pretest-Posttest design

Contoh: efek supervisi

O1 X O2

2020

Kelemahan

History (kenaikan gaji, perubahan policy)History (kenaikan gaji, perubahan policy)Regresi (produktivitas kerja)Regresi (produktivitas kerja)SpuriousSpuriousEfek maturasi (lebih skillful)Efek maturasi (lebih skillful)

The One-Group Pretest-Posttest design

O1 X O2

1. The untreated control group designwith pretest & posttest

O1 O2

Design:

Selection maturationInstrumentationdifferential statistical regressioninteraction of selection and history

Masalah:

O1 X O2

OUTCOME 1

Pretest Posttest

treatment

control

O1 O2

O1 X O2OUTCOME 2

Pretest Posttest

treatment

control

Terdapat pola selection-maturation interaction

Contoh: memberikan pendidikan ekstra pada anak yg. > pandai

O1 O2

O1 X O2

2121

OUTCOME 3

Pretest Posttest

treatment

control

O1 O2

O1 X O2

OUTCOME 4

Pretest Posttest

treatmentcontrol

Contoh: insentif extra untuk meningkatkan performan ce

O1 O2

O1 X O2

OUTCOME 5

Pretest Posttest

treatment

control

O1 O2

Contoh: intervensi pada paramedik vs. dokter

O1 X O2

Proxy Pretest Design

�� pretest based on pretest based on recollectionrecollection or or archivedarchived datadata

�� useful when you weren’t able to get a useful when you weren’t able to get a pretest but wanted to address pretest but wanted to address gaingain

N O1 X O2

N O1 O2

2222

Contoh: Tujuan studi: mengetahui efek suatu program terhadap peningkatan kepuasan pasien di UGD

�� Program sudah berjalanProgram sudah berjalan�� Kita bisa mendapatkan nilai post test, Kita bisa mendapatkan nilai post test,

tetapi tidak memiliki nilai pre test. tetapi tidak memiliki nilai pre test. �� Kita perlu mencari proxy variable yang Kita perlu mencari proxy variable yang

kirakira--kira dapat digunakan untuk estimasi kira dapat digunakan untuk estimasi pre test. pre test.

�� Sebagai contoh: gunakan nilai pretest Sebagai contoh: gunakan nilai pretest pasien di kamar bersalin sebagai proxy pasien di kamar bersalin sebagai proxy pretest.pretest.

Separate Pre-Post Samples

�� Groups with the same subscript come from the Groups with the same subscript come from the same contextsame context

�� here, Nhere, N11 might be people who were in the might be people who were in the program at Agency 1 program at Agency 1 last yearlast year, with those in N, with those in N22 at at Agency 2 last yearAgency 2 last year

�� This is like having a proxy pretest on a different This is like having a proxy pretest on a different groupgroup

N1 ON1 X ON2 ON2 O

Contoh: tujuan penelitian adalah meningkatkan kepuasan pasien di 2 rumah sakit

�� Pasien bergantiPasien berganti--ganti dari waktu ke waktu, kita ganti dari waktu ke waktu, kita tidak bisa mengukur nilai pretest untuk pasien tidak bisa mengukur nilai pretest untuk pasien yang sama. yang sama.

�� Lakukan pengukuran kepuasan pada satu saat, Lakukan pengukuran kepuasan pada satu saat, kemudian implementasikan program selama kemudian implementasikan program selama kurun waktu tertentu. kurun waktu tertentu.

�� Desain ini tidak cukup kuat, karena mengukur Desain ini tidak cukup kuat, karena mengukur orang yang berbedaorang yang berbeda

�� Oleh sebab itu yang dilihat hanya perubahan Oleh sebab itu yang dilihat hanya perubahan antara sebelum program dan setelah program. antara sebelum program dan setelah program.

Double Pretest Design

�� strong in internal validitystrong in internal validity

�� helps address selectionhelps address selection--maturationmaturation

�� how does this affect selectionhow does this affect selection--testing?testing?

�N O O X O�N O O O

2323

Switching Replications

�� strong design for both internal and strong design for both internal and external validityexternal validity

�� strong against social threats to internal strong against social threats to internal validityvalidity

�� strong ethicallystrong ethically

�N O X O O�N O O X O