Study Designs and Randomized Trials

Basic study designs in Epidemiology

 Types of Studies
• Descriptive Studies
• Observational Analytic Studies
• Experimental Studies
Types of Studies 1: Descriptive

•Document what is going on or what exists

•Provide information on something in a given
context
 Types of Studies 2: Analytic
• Examine the relationship between two or more variables

Effect of other variables in the relationship

B
Did the investigator assign
exposure?
Did the investigator assign
exposure? NO

Non-experimental/
observational
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random
allocation?
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random
allocation?
NO
Non-randomized
(qasi-experimental
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random
allocation?
NO
Non-randomized YES
(quasi-experimental Randomized
trials
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random
allocation?
NO
Non-randomized YES
(quasi-experimental Randomized
trials

Blinding?
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random
allocation?
NO
Non-randomized YES
(quasi-experimental Randomized
trials

Blinding?

YES

Blinded
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random
allocation?
NO
Non-randomized YES
(quasi-experimental Randomized
trials

Blinding?

YES NO

Blinded Not-Blinded
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random Comparing groups?

allocation?
NO
Non-randomized YES
(quasi-experimental Randomized
trials

Blinding?

YES NO

Blinded Not-Blinded
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random Comparing groups?

allocation?
NO NO
Non-randomized YES Descriptive
(quasi-experimental Randomized
trials

Blinding?

YES NO

Blinded Not-Blinded
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random Comparing groups?

allocation? YES
NO NO
Non-randomized YES Analytical Descriptive
(quasi-experimental Randomized
trials

Blinding?

YES NO

Blinded Not-Blinded
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random Comparing groups?

allocation? YES
NO NO
Non-randomized YES Analytical Descriptive
(quasi-experimental Randomized
trials

Blinding?

YES NO

Blinded Not-Blinded
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random Comparing groups?

allocation? YES
NO NO
Non-randomized YES Analytical Descriptive
(quasi-experimental Randomized
trials
Direc
Blinding?
tion?
YES NO

Blinded Not-Blinded
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random Comparing groups?

allocation? YES
NO NO
Non-randomized YES Analytical Descriptive
(quasi-experimental Randomized
trials
Direc
Blinding?
tion?
YES NO

Blinded Not-Blinded

Cross-
section
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random Comparing groups?

allocation? YES
NO NO
Non-randomized YES Analytical Descriptive
(quasi-experimental Randomized
trials
Direc
Blinding?
tion?
YES NO

Blinded Not-Blinded

Case- Cross-
control section
 Did the investigator assign
YES exposure? NO

Non-experimental/
Experimental
observational

Random Comparing groups?

allocation? YES
NO NO
Non-randomized YES Analytical Descriptive
(quasi-experimental Randomized
trials
Direc
Blinding?
tion?
YES NO

Blinded Not-Blinded

Case- Cross-
Cohort
control section
Randomized Trials
 Some history of trials
• In 1537, Paré was responsible for the treatment of the wounded
after the capture of the castle of Villaine- when the standard
treatment for war wounds was the application of boiling oil.
“At length my oil lacked and I was constrained to apply in its place a
digestive made of yolks of eggs, oil of roses and turpentine”

“I found those to whom I had applied the digestive medicament

feeling but little pain, their wounds neither swollen nor inflamed,
and having slept through the night. The others to whom I had
applied the boiling oil were feverish with much pain and swelling
about their wounds. Then I determined never again to burn thus so
cruelly the poor wounded”

What was the design?

this was not a randomized trial , it was a form of unplanned trial
 Some history of trials
A planned trial was described by the Scottish surgeon James Lind in 1747 on scurvy
 Some history of trials
I took 12 patients in the scurvy on board the Salisbury at sea. The cases were
as similar as I could have them…they lay together in one place and had one
diet common to them all.
May 20th 1747, on the HMS Salisbury at sea,6 pairs of seamen
Treatment Effect

Electuary (garlic, mustard seed, etc) 0

(Reference group)

Quarter of cider a day Improved after 20 days

Elixir vitriol (3x 25 gttes/d) Slight improvement

Vinegar (3x2 spoonfuls/d) 0

Sea water (1/2 pint/d) 0

2 Oranges and 1 lemon /d Cured after 6 days

Oranges & lemon can cure scurvy

 Allocation of treatment without
randomization
• Studies without Comparison
– case study or case series. no comparison is made with an untreated group
– The issue of comparison is important because we want to be able to derive
a causal inference
• Studies with Comparison
– Historical Controls- a comparison group from the past for comparison, we
will go back to the records of patients with the same disease who were
treated before the new therapy became available
• Inference could be baised due to problems in records
• not be sure that the difference is due to the therapy, because many
things other than the therapy change over calendar time
• Important for disease is uniformly fatal and a new drug becomes
available
– Simultaneous Nonrandomized Controls
– Quasi-experimental
 Randomized controlled trial
• The randomized trial is generally considered the “gold standard” of study
designs.
• Investigator controls the predictor variable (intervention or treatment)
• Controls concurrent with experimental group
• Placebos (inactive “treatment”)
• Advantage over other type of studies-ability to demonstrate causality
• Randomization controls unmeasured confounding
• Aims to find an identical comparison, does not necessarily ensure
• Only for mature research questions
• Comparative studies with an intervention group and a control group
• The assignment of the participants to a group is determined by the
formal procedure of randomization
• Limited mostly to trials at efficacy level
• Ethical drawbacks
Basic Design of RCT
Basic Design of RCT
Basic Design of RCT

Difference
between
randomization and
random selection?
Basic Design of RCT
 ALLOCATING SUBJECTS USING
RANDOMIZATION
Process by which all participants are equally likely to be
assigned to either the intervention group or control group
Advantage
– Tends to produce true comparable groups
Different types of randomization
– Simple randomization
– Blocked randomization
– Stratified randomization
 Type of randomization
Simple randomization
• Easy to implement
• To toss an unbiased coin
• Random number generator (Statistical software)
• Table of random numbers
• Potential for imbalance, especially for small trials
• Example with a coin……….

Block randomization
• Also called permuted block randomization
• Widely used randomization procedure
• Used to avoid imbalance in the number of participants assigned to each
group
 Block randomization: Example
• Example
• Two groups-A and B
• Block size of 4
• Possible order of treatment assignment of four subjects to
two treatments (A and B) to achieve equal numbers treated
in each group
• AABB ABAB ABBA BAAB BABA BBAA
• The size of the block can vary over the period of the study
 Stratified randomization
• Assigns patients to groups (strata) according
to a few baseline characteristics
– Example: study site, stage of disease
• Within each stratum, assign patients to
treatment groups using simple
randomization or blocked randomization
• Guarantees balance of stratification
factor
 Blinding
• Purpose: To avoid differences in information between groups
that result from awareness of the intervention
– subjects not to know which group they are assigned to
– mask (or blind) the observers or data collectors in regard to which
group a patient is in
– Analyst also will be masked
• Blinding types
Single blind: Participants are not aware of the treatment group
Double blind: Both participants and investigators are unaware
Triple blind: Various meaning
• Persons who perform the tests/assign the treatment
• Outcome adjudicators
• Data safety monitoring group
Other designs: cross-over (planned)
 Other designs: cross-over (planned)
• Participants serve as control for their own-selves
• Best utilization of limited number of samples
• variation between individuals in many characteristics that could
potentially affect a comparison of the effectiveness of two agents
are eliminated
• carryover: if a subject is changed from therapy A to therapy B and
observed under each therapy, the observations under therapy B
will be valid only if there is no residual carryover from therapy A.
There must be enough of a washout period to be sure none of
therapy A, or its effects, remains
• the order in which the therapies are given may elicit psychological
responses
• planned crossover design is clearly not possible if the new therapy
is surgical or if the new therapy cures the disease
Other designs: cross-over (unplanned)
 Challenge in unplanned cross-over
• Unplanned crossovers pose a serious challenge in analyzing the data.
• If we analyze according to the original assignment (called an intention
to treat analysis), we would compare the patients according to the
treatment to which they were originally randomized, regardless of
what treatment actually occurred
• we analyze according to the treatment that the patients actually
receive, we will have broken, and therefore lost the benefits of, the
randomization
• No perfect solution is available for this dilemma. Current practice is to
perform the primary analysis by intention to treat—according to the
original randomized assignment.
• We would hope that the results of other comparisons would be
consistent with this primary approach.
• The bottom line is that because there are no perfect solutions, the
number of unplanned crossovers must be kept to a minimum.
 Other designs: Factorial

Advantage: can estimate the individual and combined impact from the same study
 Non-compliance
• Patients may agree to be randomized, but following randomization
they may not comply with the assigned treatment.
• Noncompliance may be overt or covert: On the one hand, people may
overtly articulate their refusal to comply or may stop participating in
the study.
• These noncompliers are also called dropouts from the study.
• People may just stop taking the agent assigned without admitting
this
to the investigator or the study staff.
• checks on potential noncompliance are built into the study
• Another problem called drop-ins. Patients in one group may
inadvertently take the agent assigned to the other group.
• The net effect of noncompliance on the study results will be to reduce
any observed differences, because the treatment group will include
some who did not receive the therapy, and the no-treatment group may
include some who received the treatment. Thus, even if there is a
difference in the effects of the treatments, it will appear much smaller
 Ethical considerations
• There must be reasonable doubt of the efficacy of one treatment (to
give the other treatment to some patients) and reasonable belief in the
benefits of one treatment (to give it to some of the patients)
– Risks to the subjects
– Human subjects involvement and characteristics
– Sources of Materials
– Potential risks
• Adequacy of protection against risks
• Recruitment and Informed Consent
• Potential benefit to the subjects and others
• Importance of the knowledge to be gained
• Data and Safety monitoring
 Sample size
• How many subjects do we have to study?
– Sample size measures the number of individual samples
or
observations used in a survey or experiment.
• Purpose: sample statistics to estimate population parameter
• need to be determined before the study is started.
• Depends on the study objective and design
• Considers certain assumptions
• Two basic types
– Precision based
• mostly for estimation of indicators
– Power based
• to detect difference between two groups
• How to calculate it: will learn in biostatistics
 Example
•Is there a larger (or smaller)
proportion of blue beads in jar A
than in jar B?
•We have to draw sample
•We are looking at samples and
trying to draw a conclusion
regarding a whole universe—the jars
from which we have drawn the
samples
Sample 1 Sample 2

Every single time we draw sample it could be different!

Bell curve and probability
 Sample to conclusion
• In a study we are only looking at the sample of subjects, such as a
sample of patients with a certain illness who are being treated with
treatment A or with treatment B.
• Is treatment A more effective than treatment B in the total universe of
all patients with this disease who might be treated with treatment A or
treatment B?

Four possible situations

from sample based
study to conclusion
 Errors in conclusion

•The probability that we will make a type I error is designated as α, and the
probability that we will make a type II error is designated as β
•α is the P value, which is seen in any published papers and has been sanctified
by many years of use. “P < 0.05,” the reference is to α.
•P-value of .05 tells us the probability a difference could have arisen by chance
alone, and that the difference between our groups does not reflect any true
difference between therapies A and B, is only 0.05 (or 1 in 20)
 Errors in conclusion

The reality is that there is a difference between the therapies:

We might conclude, in error, that the therapies do not differ (type II error, β ).
Or we might conclude, correctly, that the therapies differ 1 − β , because the total
probability is 1.
This probability, 1 − β, is called the power of the study.
 Sample size assumptions

Assumptions/specifications needed to estimate the sample size:

•The difference in response rates to be detected
•An estimate of the response rate in one of the groups
•Level of statistical significance (α)
•The value of the power desired (1 − β)
•Whether the test should be one-sided or two-sided
 Analysis in RCT

• Primary: intention to treat

– Analyze according to original allocation
– Net effect of non-compliance is to reduce the
observed differences

• Secondary: actual treatment received

– Based on observed data
– No benefit of randomization
 Expressing randomized trial result
The risks of death or of developing a disease or complication in each
group can be calculated, and the reduction in risk (efficacy) can then be
calculated

The extent of the reduction in disease by use of the vaccine. Risks are
often calculated per person-years of observation

Another approach to reporting results from randomized trials is to

calculate the ratio of the risks in the two treatment groups (the relative
risk)

Difference between efficacy and effectiveness

Efficacy, or how well a treatment works under “ideal” conditions, may
be differentiated from effectiveness, or how well a treatment works
in “real-life” situations.
 Example
Vitamin C tonic was provided to seamen during voyage to explore its impact on
scurvy
Scurvy (D+) Disease free (D-) Total (N)

Vitamin C supplement (E+) 876 4265 5141

No supplement (E-) 1678 4221 5499

Diseased
Risk=

Total population at risk at baseline

876
Risk intervention = 1678
5141 = 0.17 5499 = 0.30
Risk control=

R(E+) 0.17
RR = = = 0.56

R(E-) 0.30

Interpretation:
The risk of scurvy infection among seamen who received vitamin C supplement was
 Example
Vitamin C tonic was provided to seamen during voyage to explore its impact on
scurvy
Scurvy (D+) Disease free Total (N) Person time
(D-)
Vitamin C supplement (E+) 876 4265 5141 4968
No supplement (E-) 1678 4221 5499 4894

New cases
Incidence rate=

Person time
876 1678
IR intervention = = 0.18pyo control=
= 0.34
IR pyo
4968 4894
IR(E+) 0.18
IRR = = = 0.51
IR(E-)
0.34
Interpretation:
The incidence of scurvy infection among seamen who received vitamin C supplement
was 0.51 times than that among seamen who received no supplement
 Interpretation of RCT result
Generalizability
•the ultimate objective is to generalize
the results beyond the study population
itself
•Ability to apply the results obtained in
study population to a broader
population is called the generalizability,
or external validity
•Depends on to what extent the patients
studied are representative of the
defined population, that is, of all
patients with the disease in question in
our community
•Depends on how select the sample.
 Interpretation of RCT result
Internal validity
• randomized trial is internally valid
if the randomization has been
properly done and the study is
free of other biases and is
without any of the major
methodologic problems.
• Randomized trials are considered
the gold standard of study
designs
– because randomization, if
correctly conducted,
prevents any biases on the
part of the study
investigators from
influencing the treatment
assignment for each patient.
 Advantages of RCT
Advantages
– One treatment is directly compared to another to establish superiority
– This study design can make causal inferences,
– Randomization minimizes allocation bias and selection bias
– Blinding minimizes performance bias
– Double-blinding minimizes assessment bias
– Allocation concealment minimizes both performance and assessment bias
– Prospective design minimizes recall error and selection bias
– Randomization makes groups comparable according both known and unknown
factors
– Statistical test of significance is readily interpretable when the study is randomized
Disadvantages
– Power calculation might demand vast samples size, more resources
– Results may not always mimic real life treatment situation (e.g. inclusion /
exclusion criteria; highly controlled setting)
– Randomization requires clinical equipoise: one cannot ethically randomize patients
unless both treatments have equal support in the clinical community
– Some research cannot be ethically performed as an RCT (classically, cigarette
smoking and lung cancer)