You are on page 1of 41

Study Designs in Clinical Research:

An Overview
Kuntjoro Harimurti
Department of Internal Medicine Center for Clinical Epidemiology and EBM (CE-EBM) Cipto Mangunkusumo Hospital / Faculty of Medicine UI, Jakarta

...a poor design cannot be salvaged by a good statistics

Research Design
A specific plan or protocol for conducting the study, which allows the investigator to translate the conceptual hypothesis into an operational one All procedures for selecting and recruiting individuals in the study sample

Consideration when choosing a study design

No one approach is always better than the others Each research question requires a judgment about which design is the most efficient way to get a satisfactory answer Its choose depends on the study objectives, but may also rely on practical issues such as costs or data availability

Clinical Study Types

Observational Studies
Case report Case series Cross-sectional Case-control Cohort

Experimental Studies
Uncontrolled Controlled

Observational Studies
A study in which the investigator monitors, but does not influence, the exposure status of individual subjects and their subsequent disease status

Case-report and case series

Not considered to be true investigation Only describe clinical / laboratory characteristics No hypothesis, no statistical analysis, no sample size estimation Involve new disease, rare disease, or rare manifestations of common diseases Sometimes useful to identify research problem and generating hypothesis

Cross-sectional study
A study in which determinant (risk factors) and outcome (disease) are collected at the same point in time for each participant Characteristics of cross-sectional study:
Observational = non-experimental
No time-axis Individual observed only once Could be descriptive or analytic

Use of cross-sectional studies

Prevalence studies (descriptive):
survey, census

Etiologic studies

Determinant characteristics that do not change (sex, gene expression) Estimate probability of disease presence on basis of diagnostic determinants

Diagnostic studies
Reference range studies Repeated cross-sectional studies
Measure change / evaluate intervention

Measures and analyses in cross sectional studies

Categorical variables
Prevalence, percentage Prevalence ratio Odds ratio

Numerical variables
Mean Means difference Correlation coefficient

Case-control study
I would trust only six people in the world to do a proper case-control study (David L. Sackett) A study in which outcomes (disease/cases and no-disease/controls) identified first and risk factors accounted in and compared between cases and controls Characteristics of case-control study:
Observational Retrospective Analytic

Essence of case-control study: sampling!

Analysis in case-control studies

Objective: to measure association, estimating relative risk/risk ratio (RR) Risk could not directly calculated from a casecontrol data alternative measure: odds
Odds: probability of event / probability of no-event (p/1-p)

Odds ratio (OR): ratio of two odds (with CI and pvalue)

Interprete cautiously as RR (risk ratio or relative risk)

Cohort study
Cohort: the tenth division of the Roman army. Cohort of war fighters In research: a cohort is a group of subjects from which data are collected

Cohort studies: disease-free subjects selected first according to the risk factor and further followed for the disease Prospective and restrospective cohort

Analysis in cohort studies

Measuring incidence:
Cumulative incidence (risk) Incidence density: person-time

Measuring (strength of) association:

Risk ratio/relative risk (RR) Sometimes expressed as OR With p-value and CI

Special analysis in cohort studies survival analysis:

Time to event as outcome Calculating prob. of survival in a specific period Measure of association: hazard ratio (HR)

Observational Designs
E(-) Controls

Prospective Cohort Today

Participants, Patients, Subjects
No Expo.
Control Control Case Case






Retrospective Cohort





Cross-sectional Time



Comparisons of observational study designs

Crosssectional Casecontrol Retrospective Cohort Prospective Cohort

Cost Loss to follow up Estimating risk / RR Bias Cause-effect relationship

+ + + +

+ ++ +++ ++

+ + +++ ++ +++

+++ ++ +++ + +++

Experimental Study
A study in which the investigator influences the exposure status of individual subjects and then monitors the subjects outcome

Types of experimental studies (trials)



Not controlled


Not randomised


Not blinded

A randomized double-blind controlled clinical trial (RCT)

Clinical Trial
A randomized double-blind controlled clinical trial (RCT) Gold standard of research design which provide the most convincing evidence of relationship between exposure (intervention) and outcome Use human subject Always prospective Comparing two or more intervention strategies

Clinical Trial
Key elements of RCT:
Randomization Blinding Control
Objective: To achieve comparabilities

Comparabilities in RCT:
Comparability of prognostic factors Comparability of extraneous factors Comparability of outcome measures

Clinical Trial: Parallel Group Design

Experimental Treatment
With Outcome Without Outcome With Outcome Without Outcome

Participants screened for entry criteria

Control Treatment

Screening Baseline Treatment Outcome measurement

Analysis in RCT
Numerical outcome:
Means or medians difference after intervention (Student -t test or Mann-Whittney test) Per-protocol analysis

Categorical outcome:
Relative risk (RR), hazard ratio (HR) Relative and absolute risk reduction (RRR, ARR) Number needed to treat (NNT Intention-to-treat analysis

Special Design in Clinical Research


Quantitative method of combining the results of independent research (primary) studies and synthesizing conclusions to evaluate the effectiveness of treatments or procedures Begins with systematic finding, evaluating, and presenting the results of primary studies Systematic Review No collecting data directly from the study subjects secondary research Considered as true investigation and has highest rank in level-of-evidence

Advantages of Meta-analysis
Quantitatively summarize estimate from previous studies resolve controversies Using protocol to choose the individual studies avoid bias

Increase power for statistical test and increase precision for confidence intervals
Conclusions often reflect broad spectrum of patient and characteristics results are more generalizable

Important issues in meta-analysis

Study selection protocol
Finding all relevant studies

Assesing study quality

Critically appraise for validity of the studies

Combining combinable studies

Publication bias
Unpublished negative and small studies

Meta-analysis of RCTs with nominal outcome

Study I Study II Study III Study IV Study V Study VI Study VII1 Study VIII Combined 1992 1994 1995 1995 1996 1997 1999 2000


Favor drug

OR = 1


Favor placebo

Meta-analysis of RCTs with numerical outcome

Study I Study II Study III Study IV Study V Study VI Study VII1 Study VIII Combined 1992 1994 1995 1995 1996 1997 1999 2000

Mean difference (X1-X2) = 0 Favor drug Favor placebo



Clinical Research
Clinical research involves working with human subjects to answer questions relevant to their well-being Clinical (epidemiology) research addresses questions regarding clinical practice challenges Patient oriented research is where the rubber meets the road!

Challenges in clinical practice

1. Explanation of clinical profile: given the patient profile, what the patients illness?
2. Explanation of illness: why did this illness occur in this patient at this time? 3. Prediction of course: given the patients illness, its etiology, the clinical and nonclinical profile, etc., what will be the future course of the illness, depending on (absense of) treatment 4. Treatment decision 5. Treatment execution Diagnostic knowledge Etiologic knowledge Prognostic knowledge

Prognostic and decision analysis Skill

Diagnostic research
Aim: to predict the probability of presence of target disease from clinical and non-clinical profile Relevance for patient and physician to establish diagnosis and guide management Results presentation: sensitivity, specificity, positive and negative predictive values, likelihood ratios, scoring system Example: Wells clinical prediction rule for DVT and PE

Wells clinical prediction rule for DVT

Etiologic research
Aim: to causally explain ocurrence of target disaese from determinant (risk factor) Research relevance may indicate means of prevention and causal intervention

Results presentation: odds ratio (OR), relative risk (RR), hazard ratio (HR)
Example: smoking and lung cancer study by Sir Bradford Hills

Prognostic Research
Aim: to predict of disease from clinical and non-clinical profiles Relevance for patient and physician to learn about future and guide management

Results presentation: RR, HR, predictive values, scoring system

Example: TIMI risk score for STEMI, Framingham Risk Score, CURB-65 score

Therapy/Interventional research
Aims: To casually explain and to predict the course of disease as influence by treatment Relevance for patient and physician to decide an optimal management and for drug research/development/registration Results presentation: absolute/relative risk reduction (ARR/RRR), number needed to treat (NNT)

The Scandinavian Simvastatin Survival Study (4S)

Primary Endpoint: Overall Survival

The Scandinavian Simvastatin Survival Study (4S)

Summary of Key Endpoint Results

Take home messages: Choosing study design

Depends on:
research questions research goals researcher beliefs and values researcher skills time and funds

It is also related to:

status of existing knowledge occurrence of disease duration of latent period nature and availability of information available resources

Take home messages

Choosing design in clinical research should follow the nature of clinical challenges :
Diagnostic cross-sectional Etiologic cohort > case-control > crosssectional Prognostic cohort > case-control Therapy RCT > cohort > case-control

Thank you...