2. 2
Research is to see what
everybody else has seen, and to
think what nobody else has
thought.
Albert Szent-Gyorgyi
Definitions of Research
3. Research Designs
Observational Studies
Where the exposure or assignment of subjects into
a treated or control groups is beyond the scope of
the investigator.
Interventional Study
Where exposure and allotment of subjects into
treated and control groups is controlled by the
investigator.
2/11/2020 3
4. 4
Where to Start?
A good clinical study starts with
a good question based on good hypothesis that is based on
good and comprehensive review of the available evidence
from pre-clinical and clinical data
Type of design depends on the question to be
answered
5. 5
Formulating a Research Question
Focused and specific
What is the prevalence of Hepatitis B surface
Antigen in India?
Cross-sectional study
What are the risk factors for hepatitis B infection?
Prospective cohort or case-control
Is interferon a useful therapy for hepatitis B
infection?
Therapeutic clinical trial
6. 6
Objectives
Specific aims
Clear and detailed
End point(s)
Primary
The main answer to the research question
Secondary
Answer other related questions
7. 7
Observational study Clinical trial
exposed
non exposed
outcome
Clinical
Trial
observational
studydescribe as
occurring in nature
allocate
randomly
Ethics!
8. 8
Important issues in Study Design
Validity: Truth
External Validity:
Can the study be generalized to the population
Internal Validity:
Results will not be due to chance, bias or
confounding factors
Symmetry Principle: Groups are similar
9. 9
Confounding: distortion of the effect of one risk factor by the
presence of another
Bias: Any effect from design, execution, & interpretation that
shifts or influences results
Confounding bias: failure to account for the effect of
one or more variables that are not distributed equally
Measurement bias: measurement methods differ
between groups
Sampling (selection) bias: design and execution
errors in sampling
Important issues in Study Design
10. 10
Introduction
Why this study is needed ?
What is the purpose of this study?
Was purpose known before the study?
What has been done before and how does this
study differ?
inadequacies of earlier work or next step in an
overall research project
Does the location of the study have relevance?
11. 11
Why doing a study?
Alternative:
census: test every individual in the population
use available data, e.g. hospitals
But:
- data availability
- data quality
- cost
- questions require specific type of data and
circumstances
12. 12
Clinical Study Types
Observational Studies
Cohort (Incidence, Longitudinal)
Case-Control
Cross-Sectional (Prevalence)
Case Series
Case Report
Experimental Studies
Uncontrolled Trials
Controlled Trials
13. 13
Types of observational studies
Cross - sectional study
Cohort study
Case control study
Case series/case reports
14. 14
Characteristics of observational studies
No control over study units
need to clearly describe study individuals
Can study risk factors that have serious consequences
Study individuals in their natural environment (>>
extrapolation)
Possibility of confounding
15. 15
Aims of observational studies
Evaluate the effect of a suspected
risk factor (exposure) on an outcome
(e.g. disease)
define ‘exposure’ and ‘disease’
Describe the impact of the risk factor
on the frequency of disease in a
population
20. 20
Cohort studies
Follow-up studies; subjects selected on presence or
absence of exposure & absence of disease at one
point in time. Disease is then assessed for all subjects
at another point in time.
Typically prospective but can be retrospective,
depending on temporal relationship between study
initiation & occurrence of disease.
21. 21
Cohort Study (1)
Individuals selected by exposure status and future
occurrence of disease measured
present futurepast
n
Exposed yes
no
disease ?
disease ?
22. 22
Cohort studies (2)
More clearly establish temporal sequence
between exposure & disease
Allows direct measurement of incidence
Examines multiple effects of a single exposure
(nurses’ health study, OC and breast, ovarian
cancers)
23. 23
Cohort studies (3)
Limitations:
time consuming and expensive
loss to follow-up & unavailability of data
potential confounding factors
inefficient for rare diseases
26. 26
Case-Control Study (1)
Retrospective
Can use hospital or health register data
First identify cases
Then identify suitable controls
Hardest part: who is suitable ??
Then inquire or retrieve previous exposure
By interview
By databases (e.g. hospital, health insurance)
27. 27
Case-Control Study (2)
Diseased and non-diseased individuals are selected
first
Then past exposure status is retrieved
present futurepast
n
yes
no
diseaseexposed ?
exposed ?
28. 28
Case-Control Study (3)
Good for rare disease (e.g. cancer)
Can study many risk factors at the same time
Usually low cost
Confounding likely
OR (not RR !!)
30. 30
• Study subjects selected on basis of whether
they have (case) or do not have (control) a
disease
• Useful for disease with long latency period
• Efficient in terms of time & costs
• Particularly suited for rare diseases
• Examines multiple exposures to a single
disease
Case-Control study (4)
31. 31
Case-control study (5)
Limitations:
(1) susceptible to bias (particularly selection &
recall)
(2) difficulties in selection of controls
(3) ascertainment of disease & exposure status
(4) inefficient for rare exposures unless
attributable risk is high
32. 32
Case Selection
• Define source population
• Cases
– incident/prevalent
– diagnostic criteria (sensitivity + specificity)
• Controls
– selected from same population as cases
– select independent of exposure status
33. 33
Control Selection
• Random selection from source population
• Hospital based controls:
– convenient selection
– controls from variety of diagnostic groups other
than case diagnosis
– avoid selection of diagnoses related to
particular risk factors
– limit number of diagnoses in individuals
34. 34
Characteristic Cross -
Sectional
Case Control Cohort
Sampling Random sample:
population
Purposive sample:
diseased/non-
diseased
Purposive sample:
Exposed/non-
exposed
Time One point Retrospective Prospective
Causality Statistical
association
Screening for
many risk factors
Testing one (or
few) risk factors
Frequency
measure
Prevalence None Incidence
Risk
parameter
Prevalence (risk)
ratio, odds ratio
Odds ratio Relative risk, odds
ratio
Summary of Observational Studies
36. 36
Animal Tests Can:
Suggest which drugs are likely to be effective in
humans
Indicate which drugs may not be harmful in humans
Animal Tests Cannot:
Predict with absolute certainty what will happen in
humans
Clinical trials in drug development
(Any alternatives)
37. 37
Clinical trial vs. Cross-sectional
Clinical trial:
Individuals selected by
entry condition
Control over exposure
Exposure groups fully
comparable
Outcome measured after
allocating individuals to
exposure
Therefore: causal
association likely
Cross Sectional Study:
Individuals selected
randomly
Exposure observed as
occurring in nature (groups
not ‘identical’)
Exposure AND outcome
measured at one point in
time
No causal interpretation
41. 41
Non-randomized Trials
May Be Appropriate
• Early studies of new and untried therapies
• Uncontrolled early phase studies where the
standard is relatively ineffective
• Investigations which cannot be done within the
current climate of controversy
• Truly dramatic response
42. 42
Advantages of Randomized
Control Clinical Trial
1. Randomization "tends" to produce comparable groups
2. Assure causal relationship
3. Randomization produces valid statistical tests
43. 43
Disadvantages of
Randomized Control Clinical Trial
1. Generalizable Results?
Participants studied may not represent general
study population.
2. Recruitment
Hard
3. Acceptability of Randomization Process
Some physicians will refuse
Some participants will refuse
4. Administrative Complexity
44. 44
Study Population
Subset of the general population determined by the
eligibility criteria
General population
Eligibility criteria
Study population
Enrollment
Study sample
Observed
45. 45
Eligibility Criteria
(inclusion & exclusion)
State in advance
Consider
Potential for effect of intervention
Ability to detect that effect
Safety
Ability for informed consent
46. 46
Method Outlines (1)
The independent (predictor) and dependent (outcome)
variables in the study should be clearly identified, defined,
and Measured?
How to choose subjects?
Random or not
Are they going to be representative of the population?
Random selection is not random assignment
Types of Blinding (Masking) Single, Double, Triple.
Control group? How is it chosen?
How are patients followed up? Who are the dropouts?
How is the data quality insured? Reliability?
Consider independent review of data? Compliance?
47. 47
Methods outlines (2)
Reference any unusual methods?
Statistical methods specified in sufficient
details
Is there a statement about sample size issues or
statistical power?
? multicenter study. Quality assurance
measures should be employed to obtain
consistency across sites?
48. 48
Comparing Treatments
• Fundamental principle
• Groups must be alike in all important aspects and only differ in the
intervention each group receives
• In practical terms, “comparable treatment groups” means
“alike on the average”
• Randomization
• Each participant has the same chance of receiving any of the
interventions under study
• Allocation is carried out using a chance mechanism so that neither the
participant nor the investigator will know in advance which will be
assigned
• Blinding
• Avoidance of conscious or subconscious influence
• Fair evaluation of outcomes
50. When to Unblind
1. Following an SAE
2. When treatment of an SAE depends upon the
treatment received in trial
3. Actually very few situations require unblinding
4. Physicians feel very uncomfortable when blind
procedures are done
50
56. 56
Sample Size
The study is an experiment in people
Need enough participants to answer the
question
Should not enroll more than needed to answer
the question
Sample size is an estimate, using guidelines and
assumptions
57. 57
Contingency Plans
Patient management
Evaluation and reporting to all relevant persons
and groups
Data monitoring plans
Protocol amendment or study termination
58. 58
Human Subjects Protection
• Institutional Review Board
• Informed consent
• Different levels of risk
• Confidentiality as well as risk of new tx
• Patient can refuse to participate w/o effect
• Path to exit study known
• Compensation
59. 59
Summary
Selection of design should be made on the basis of the
particular hypothesis to be tested with consideration of
current state of knowledge
Consider available resources when deciding on a study
design
A clear and organized study design leads to successful
results
Observational studies are especially valuable in
epidemiology
Clinical trials carry the highest level of evidence and
should be pursued whenever feasible
