Design of Observational Design of Observational and Experimental Clinical and Experimental Clinical

StudiesStudies

Kenneth F. Schulz, PhDKenneth F. Schulz, PhD

EpidemiologyEpidemiology

More than statistics; biostatistics; More than statistics; biostatistics; medical science - A specific research medical science - A specific research methodologymethodology

““Methods” conjures up what images?Methods” conjures up what images? Connotation? Word association? Connotation? Word association?

Presentation?Presentation? NEJM Methods SectionNEJM Methods Section

Bad Rap - really fun and exhilaratingBad Rap - really fun and exhilarating

Epidemiology of AIDS Before Epidemiology of AIDS Before HIVHIV

Risk groups identifiedRisk groups identified MSMMSM Blood recipientsBlood recipients IVDUIVDU

>> 100 partners/year 100 partners/year (implying that 99 are safe…)(implying that 99 are safe…)

Anal receptive Anal receptive intercourse/fistintercourse/fist

Not using condomsNot using condoms Nash Herndon/FHI

AIDS Prevention Before AIDS Prevention Before HIVHIV

Exclusion of risk groups from blood Exclusion of risk groups from blood donationdonation

Education/recommendations on Education/recommendations on the risk factorsthe risk factors

AIDS Prevention Before HIV: AIDS Prevention Before HIV: EffectsEffects

Transfusion associated, Transfusion associated, transmission droppedtransmission dropped

In certain areas, risk factor In certain areas, risk factor transmission dropped, witness the transmission dropped, witness the 90% reduction in syphilis in San 90% reduction in syphilis in San FranciscoFrancisco

Supported by subsequent HIV Supported by subsequent HIV analysis of banked bloodanalysis of banked blood

Bacterial Vaginosis and Bacterial Vaginosis and Prematurity in Indonesia*Prematurity in Indonesia*

Association between bacterial Association between bacterial vaginosis diagnosed at 16 to 20 vaginosis diagnosed at 16 to 20 weeks and preterm deliveryweeks and preterm delivery (gestational age <37 weeks)(gestational age <37 weeks)

OR = 2.0; OR = 2.0; 95% CI95% CI 1.0 - 3.91.0 - 3.9

Above OR adjusted for mother’s age, Above OR adjusted for mother’s age, education, smoking and education, smoking and TTrichomonasrichomonas

*Joesoef et al. Am J Obstet Gynecol 1993;169:175-8.*Joesoef et al. Am J Obstet Gynecol 1993;169:175-8.

Bacterial Vaginosis and Bacterial Vaginosis and Prematurity in IndonesiaPrematurity in Indonesia* *

(Cont.)(Cont.)

““Only bacterial vaginosis Only bacterial vaginosis diagnosed early in the second diagnosed early in the second trimester of pregnancy plays trimester of pregnancy plays a major role as a risk factor a major role as a risk factor for preterm delivery”for preterm delivery”

Clindamycin vaginal cream - Clindamycin vaginal cream - an effective treatmentan effective treatment

*Joesoef et al. Am J Obstet Gynecol 1993;169:175-*Joesoef et al. Am J Obstet Gynecol 1993;169:175-8.8.

Nash Herndon/FHI

Begin Program in Indonesia?Begin Program in Indonesia?

PPEPPE

Corollary to PPE: SECorollary to PPE: SE

Conduct an RCTConduct an RCT

Outcome Clindamycin Placebo OR (95% CI)

Preterm (<37) 15.0% 13.5% 1.1 (0.7 – 1.7)

Preterm (<32) 4.7% 2.6% 1.8 (0.8 – 4.2)

LBW (<2500gm) 9.0% 6.8% 1.3 (0.8 – 2.4)

Intravaginal Clindamycin for BV: Intravaginal Clindamycin for BV: Effects on Preterm Delivery and Effects on Preterm Delivery and

LBW*LBW*

**Joesoef et al. Am J Obstet Gynecol 1995; 173:1527-Joesoef et al. Am J Obstet Gynecol 1995; 173:1527-31.31.

Study Design is CriticalStudy Design is Critical

Reading the literatureReading the literature

No study is perfect – the limitations No study is perfect – the limitations MUSTMUST be be understoodunderstood

Conducting researchConducting research

Serious mistakes in the study design stage Serious mistakes in the study design stage cannot be corrected in subsequent stagescannot be corrected in subsequent stages

Types of Epidemiologic Types of Epidemiologic StudiesStudies

Observational:Observational: “natural “natural experience”experience”

Descriptive:Descriptive: like newspaper reportinglike newspaper reporting has no control or comparison grouphas no control or comparison group

Analytic:Analytic: has a control or comparison grouphas a control or comparison group

Experimental:Experimental: investigator investigator controlscontrols

has an exposed or experimental group has an exposed or experimental group plus a plus a control or comparison group control or comparison group determined by determined by investigator investigator

com

ple

xit

ycom

ple

xit

y

Descriptive StudiesDescriptive Studies

Estimate frequencyEstimate frequency

Time trendTime trend

Generate more specific Generate more specific etiologic hypothesesetiologic hypotheses

Descriptive Study of WomenDescriptive Study of Womenwith PIDwith PID

Outcome (PID)Outcome (PID)

SES (Age, Marital Status, Etc.)SES (Age, Marital Status, Etc.)

Exposure (STI, IUD)Exposure (STI, IUD)

© Ruth C. Kennedy

© Harvey Nelson

Observational Analytic Observational Analytic StudiesStudies

CohortCohort ExposureExposure

Exposed Outcome?

Not Exposed Outcome?

Case-ControlCase-Control OutcomeOutcome

Exposed? Outcome

Exposed? Control

The FutureThe Future

Outcome

Outcome

No Outcome

No Outcome

Prospective Cohort DesignProspective Cohort Design

The PresentThe Present

Exposure present

Exposure absent

SampleSample

Population

The FutureThe Future

Death

Death

No Death

No Death

Example of a Prospective Cohort Example of a Prospective Cohort Design: Treatment of Severe Design: Treatment of Severe Malaria in ChildrenMalaria in Children

The PresentThe Present

SampleSample

PTX present

PTX absent

Population

Case-Control DesignCase-Control Design

The Past or PresentThe Past or Present The PresentThe Present

ExposurExposuree presentpresent

ExposurExposuree presentpresent

SampleSampleof casesof cases

PopulatiPopulation with on with disease disease (cases)(cases)

SampleSampleof of

controlscontrols

Much Much larger larger

populatiopopulation without n without disease disease

(controls)(controls)

NoNoOutcomOutcomee

OutcomOutcomee

ExposureExposure absentabsent

ExposurExposuree absentabsent

Example of Case-Control Example of Case-Control Design: OC Use and Breast Design: OC Use and Breast

CancerCancerThe Past or PresentThe Past or Present The PresentThe Present

OC UseOC Use

OC UseOC Use

SampleSampleof casesof cases

PopulatiPopulation with on with disease disease (cases)(cases)

SampleSampleof of

controlscontrols

Much Much larger larger populatiopopulation without n without disease disease (controls)(controls)

NoNoBreast Breast CancerCancer

Breast Breast CancerCancer

OC Use OC Use absentabsent

OC Use OC Use absentabsent

RCT ParadigmRCT Paradigm

Outcome AssessmentOutcome Assessment

UnexposedUnexposed ExposedExposed

RandomizeRandomize

Other allocation approach?

Population of Population of InterestInterest

Sample (Subset)

RCT ParadigmRCT Paradigm

Outcome AssessmentOutcome AssessmentRecovered within 7 daysRecovered within 7 days

New New AntibioticAntibiotic AmpicillinAmpicillin

RandomizeRandomize

Sample (Subset)Sample (Subset)

Women with

Postpartum Febrile

Morbidity

During the Emergence of Modern During the Emergence of Modern EpidemiologyEpidemiology (around 1950)(around 1950)

Sir Austin Bradford Hill, Sir Richard Doll and Sir Austin Bradford Hill, Sir Richard Doll and colleagues conducted:colleagues conducted: A case-control study of lung cancerA case-control study of lung cancer A cohort study of smoking and lung cancerA cohort study of smoking and lung cancer The first RCT in medicine, streptomycin for The first RCT in medicine, streptomycin for

treatment of TBtreatment of TB

ENDEND

Epidemiologic GoalsEpidemiologic Goals

Read the literature more criticallyRead the literature more critically

Ability to conceptualize EPI research Ability to conceptualize EPI research designsdesigns

Implement simple designsImplement simple designs

Seek assistance from the appropriate Seek assistance from the appropriate people and ask the right questionspeople and ask the right questions