Randomized Controlled

Trials

Madhukar Pai, MD, PhD

McGill University
Montreal
[email protected]

1
What is a randomized controlled trial?

Simplest definition: Individuals are allocated at

random to receive one of several interventions
(at least two total).

RCTs are experimentalthe intervention is

controlled by the investigator

RCTs are usually comparative studies

(controlled in the RCT)

2
Courtesy: J Colford
What is random allocation?
Random allocation means that all
participants have a defined probability of
assignment to a particular intervention
Allocation is not determined by the
investigator, clinicians, or participants
Allocation is not predictable based on a pattern

3
Courtesy: J Colford
Are these randomized designs if
based on
Date of birth (odd to group 1; even to group 2)
Hospital record number (last digit; odd to group
1, even to group 2)
Day of enrollment (Monday=Rx, Tues=Placebo,
etc)
Alternating (first person=Rx, second
person=placebo, etc)

No, these are called quasi randomized

4
Courtesy: J Colford
What purpose is served by random
allocation?
Covariates are distributed equally across the
groups at baseline
Not always (especially if N is small)!
Affects both measured and, more importantly,
unmeasured variables
The risk of imbalance remains even after properly
executed randomization
Table 1 in most RCTs will provide a comparison of
treatment and comparison groups, with p-values
If randomisation has been performed correctly, chance
is the only explanation for any observed difference
between groups, in which case statistical tests are
considered superfluous 5
Courtesy: J Colford
Schulz KF, Grimes DA. Generation of allocation
6
sequences in randomised trials: chance, not
choice. Lancet. 2002 Feb 9;359(9305):515-9
What elements of a trial can be
randomized?
Most common unit is individual patient
Sometimes groups are randomized=cluster
randomization
Examples: families, schools, towns, hospitals,
communities
Worry about contamination in cluster randomization
Special statistical techniques needed to cope with the
loss of independence of the individual units

7
Courtesy: J Colford
Example of a cluster randomized trial

8
How is randomization achieved?
Two steps involved:
Generation of allocation sequence
Implementation of allocation (concealment of allocation)
While both are important, there is evidence that
concealment of allocation is more critical

9
Generation of allocation sequence
Simple randomization
Analogous to a repeated fair coin tossing
Restricted randomization
Blocking
Done to ensure equal balance of arms throughout all
portions of the study
For example, blocks of six would have 3 active/3 control
Block size itself can/should vary
Stratified randomization
Individuals are identified based on important covariates
(sex, age, etc.) and then randomization occurs within
the strata
Dynamic or adaptive methods (not common)
10
Courtesy: J Colford
Concealment of allocation
If those making the decision about patient eligibility are
aware of the arm of the study to which the patient will be
allocated --if randomization is unconcealed-- they may
systematically enroll sicker-- or less sick-- patients to either
treatment or control groups.
This will defeat the purpose of randomization and the study
will yield a biased result.

Example: RCT of open vs laparoscopic appendectomy

(example from Users Guides):
trial ran smoothly during the day
at night, however, the attending surgeon's presence was required
for the laparoscopic procedure but not the open one; and the
limited operating room availability made the longer laparoscopic
procedure an annoyance.
reluctant to call in a consultant, and particularly reluctant with
specific senior colleagues, the residents sometimes adopted a
practical solution.
when an eligible patient appeared, the residents checked the
attending staff and the lineup for the operating room and, Schulz KF, Grimes DA. Allocation concealment
in randomised trials: defending against
depending on the personality of the attending surgeon and the deciphering. Lancet. 2002 Feb
length of the lineup, held the translucent envelopes containing 16;359(9306):614-8
orders up to the light.
as soon as they found one that dictated an open procedure, they
opened that envelope. The first eligible patient in the morning
would then be allocated to a laparoscopic appendectomy group
according to the passed-over envelope
If patients who presented at night were sicker than those who
presented during the day, the residents' behavior would bias the 11
results against the open procedure.
Schulz KF, Grimes DA.
Allocation concealment in
12
randomised trials: defending
against deciphering. Lancet.
2002 Feb 16;359(9306):614-8
Allocation concealment

Allocation concealed: the authors were deemed to have taken

adequate measures to conceal allocation to study group
assignments from those responsible for assessing patients for
entry in the trial (eg, central randomisation; numbered, opaque,
sealed envelopes; sealed envelopes from a closed bag; numbered
or coded bottles or containers; drugs prepared by the pharmacy;
or other descriptions that contain elements convincing of
concealment).
Allocation not concealed: the authors were deemed not to have
taken adequate measures to conceal allocation to study group
assignments from those responsible for assessing patients for
entry in the trial (eg, no concealment procedure, sealed envelopes
that were not opaque, or other descriptions that contain elements
not convincing of concealment).
Unclear allocation concealment: the authors did not report or
provide us with a description of an allocation concealment
approach that allowed for classification as concealed or not
concealed.

13

Haynes BR. Incorporating allocation concealment and blinding in randomised controlled trials Evidence-Based Medicine 2000; 5:38
When is it ethical to randomize?
At least two answers to this question:
uncertainty principle
clinical equipoise
Which is the preferred moral basis of the
RCT?

14
The uncertainty principle
Richard Peto et al. (1976):
Physicians who are convinced
that one treatment is better
than another for a particular
patient of theirs cannot ethically
choose at random which
treatment to give: they must do
what they think best for the
particular patient. For this
reason, physicians who feel they
already know the answer cannot
enter their patients into a trial.

15
www.uab.edu/ethicscenter/weijer.ppt
Clinical equipoise
Benjamin Freedman (1987):
Clinical equipoise exists
when there is genuine
uncertainty within the
professional community as
to which of the two
treatment arms is superior

16
www.uab.edu/ethicscenter/weijer.ppt
Types of RCTsclassification
schemes
Based on the type of interventions being
evaluated
Based on how participants are exposed to
interventions
Based on the number of participants
Based on whether goal is evaluation of
superiority vs. equivalence
Based on whether investigators and/or
participants know which intervention is being
studied

17
Courtesy: J Colford
Types of RCTsclassification
schemes
Based on the aspects of interventions
being evaluated
Efficacy vs effectiveness trials
Superiority vs equivalence trials
Phase I, II, III trials

18
Courtesy: J Colford
Efficacy vs. effectiveness
Efficacydoes the intervention work in the
people who actually receive it?
These trials tend to be explanatory
Goal here is high compliance
Effectivenesshow does the intervention
work in those offered it
Tend to be pragmatic

19
Courtesy: J Colford
Superiority vs. equivalence trials
Superiority trials
Intended to determine if new treatment is different from
(better than) placebo or existing treatment (active control)
Null hypothesis is that there is no difference between treatments.
Alternative hypothesis is that the new treatment is no different
from (two-sided) or better than (one-sided) control.
Equivalence trials
Intended to determine that new treatment is no worse
than active control
Null hypothesis and alternative hypotheses are reversed.
Null hypothesis is that difference between treatments is
greater than X.
Alternative hypothesis is that difference between
treatments is less than X

20
Courtesy: J Colford
Example of equivalence trial

21
Why do an equivalence trial?
Existing effective treatment
Placebo-controlled trial unethical
Life-threatening illness.
New treatment not substantially better
than existing treatment.
May have fewer side effects, greater
convenience, lower cost, higher quality of life,
or provide an alternative or second line
therapy.

22
Courtesy: J Colford
Phase I, II, III, IV trials

23
Types of RCTsclassification
schemes
Based on how the participants are
exposed to the intervention
Parallel trials
Crossover trials
Trials with factorial design

24
Courtesy: J Colford
Simple, two-arm (parallel) RCT

25

Hulley et al. Designing Clinical Research. 2nd Edition. Lippincott Williams & Wilkins, 2001
Cross-over RCT design

26

Hulley et al. Designing Clinical Research. 2nd Edition. Lippincott Williams & Wilkins, 2001
Example: Crossover trial

27
Factorial RCT design

28

Hulley et al. Designing Clinical Research. 2nd Edition. Lippincott Williams & Wilkins, 2001
Example: factorial design

29
Types of RCTsclassification
schemes
Based on the number of participants
N-of-1 trials to mega-trials
Fixed size
Sequential trials

30
Courtesy: J Colford
N-of-1 trial
These can be thought of as a form of crossover trial
Each participant receives the experimental arm for a period
of time and then the control/comparison arm during a
different period of time
There can be many such periods of time in these studies
XCCCXXCCXX
The participant does not know which intervention is
occurring during each period

Eligible patient Drug Placebo Placebo Drug Drug Placebo

31
Courtesy: J Colford
Example: N-of-1 trial

32
Mega-trials (Large simple trials)
These studies are meant to be HUGE but to
collect only a limited amount of data (to make
them affordable and practical)
Are usually multi-center
Can pick up small effects

33
Courtesy: J Colford
Example: Mega-trial

34
Sequential trial
Contrast is with the more traditional fixed size trial in which
the number of participants is determined based on a priori
sample size calculations
Has a parallel design
Number of participants is NOT specified before the trial
begins
Participants are recruited until the question is answered (or
it becomes clear that there is no possibility to detect a
difference between the arms)
Usually the principal outcome occurs (or not) shortly after
the study begins

35
Courtesy: J Colford
 36
Meyhoff et al. Trials 2008 9:58
Types of RCTsclassification
schemes
Based on who knows what (about the
intervention that is being assessed)
Open trials
Single blind trials
Double blind trials
Triple and quadruple-blind trials

37
Courtesy: J Colford
Blinding
Relevant groups who may/may
not have knowledge of
treatment assignments
Participants
Investigators/clinicians
administering intervention
Investigators assessing outcomes
Data analyst(s)
Open trials Schulz & Grimes. Lancet 2002

All participants and investigators

know who is getting which
intervention
E.g. medical vs. surgical treatments

38
Courtesy: J Colford
Single, double, triple, and beyond
Single-blind
The participants (usually) or the
investigators assessing outcome
(alternately) do not the
assignments
Double-blind
Two groups do not know
usually it is the participants and
the outcome
assessors/investigators
Triple or quadruple blinding
Three or four of the relevant
groups (prior slide) are not Forder, MJA, 2005

aware of the treatment

assignment

39
Courtesy: J Colford
Blinding
Aspirin Myocardial Infarction Study (AMIS), 1982

Aspirin/Placebo survival for 3-4 years after

myocardial infarction

95 / 285 (33%) deliberately tested the capsule

Taste, smell, acid test or professional analysis

67% of testers guessed right (47% of non-testers)

40
Howard, J et al. 1982 Clin Pharmacol Ther 32(5), 543-53 Courtesy: J Colford
Please read B-File #5: will be discussed by Dr Stan
Shapiro on Monday, 2nd Nov

41
Blinding

Schulz KF, Grimes DA. Blinding in42

randomised trials: hiding who got what.
Lancet. 2002 Feb 23;359(9307):696-700
Blinding
Blinded: any or all of the clinicians, patients or
participants, outcome assessors, or statisticians were
unaware of who received which study intervention. If
"initially" is indicated (eg, blinded [patients and outcome
assessor initially]), the code was broken during the trial, for
instance, because of adverse effects.
Blinded (unclear): the authors did not report or provide
us with an indication of who, if anyone, was unaware of
who received which study intervention.
Unblinded: all participants in the trial (clinicians, patients
or participants, outcome assessors, and statisticians) were
aware of who received which study intervention.

43

Haynes BR. Incorporating allocation concealment and blinding in randomised controlled trials Evidence-Based Medicine 2000; 5:38
Concealment of allocation vs. blinding
Concealment of allocation:
Procedure to protect the randomization process before the subject
enters the trial
Failed concealment from the investigator or clinician
Failed concealment from the patient
Concealment of allocation is ALWAYS feasible
If not done, results in selection bias (randomization benefits are
lost, and treatment assignment is no longer truly random)
Blinding:
Masking of the treatments after randomization (once trial begins)
Failed masking of patients, investigators, outcome assessors, etc
Blinding is not always feasible
If not done, can result in patients biasing their responses because
of their knowledge of treatment; can also lead to biased outcome
assessment because investigators have knowledge of treatment
44
Bias in RCTs
Can occur at all phases:
Planning, selection of participants, administration of interventions,
measurement of outcomes, analysis of data, interpretation and reporting of
results, publication of reports, and even in the reading of the report!

Selection bias:
E.g. due to lack of concealment of allocation
Due to attrition and differential losses
Information bias:
Participant response bias (due to lack of blinding)
Outcome ascertainment bias (due to lack of blinding)
Bias due to competing interests
Reporting biases
Publication bias
Time lag bias
Outcome reporting bias, etc
45
Courtesy: J Colford
Selection bias
Definition: Selection bias is when there are systematic
differences in the way participants are accepted or
rejected for a trial, or in how the intervention is assigned
to participants once they have been accepted
Dont get a false sense of security as a result of
randomization, easy to introduce selection bias in a RCT!
Example: bias due to lack of concealment of allocation

46
Gluud, AJE 2006
Information (reporting, ascertainment or
detection) bias
Definition: Ascertainment bias occurs when the results
are systematically distorted by knowledge of which
intervention each participant is receiving

Can be introduced by the person administering the

intervention, the participants, the investigator, the data
analyst, or even the manuscript authors

Result: Can exaggerate the effect

47
Courtesy: J Colford
How can ascertainment bias be
minimized?
During.

Randomization Blind the participant as to

which intervention
receiving
Delivery of
intervention Blind the individuals who
administer the
interventions
Assessment of
outcomes Blind the individuals who
record the outcomes

Data Blind the statisticians

analysis/manuscript
48
Courtesy: J Colford
Other biases
During the course of a trial - inappropriate handling
of withdrawals, drop outs, and protocol violations
Intention to treat analysis all study participants are
included in the analyses as part of the groups to which
they were randomized regardless of whether they
completed the study or not
vs. per protocol analysis

Worst case scenario sensitivity analysis assign the

worst possible outcome to the missing patients or time-
points in the group that shows the best results and the
best possible outcomes to the missing patients or time-
points in the group with the worst results

49
Courtesy: J Colford
Bias due to not using intention-to-treat analysis

50
Per protocol analysis introduces bias into the estimate of intervention efficacy

Montori, V. M. et al. CMAJ 2001;165:1339-1341

51

Copyright 2001 Canadian Medical Association or its licensors

Loss to follow up and attrition

52
Bias due to competing interests

53
Bekelman, JAMA 2003
 54

JAMA 2004
Publication and reporting biases

55

Montori et al. Mayo Clin Proc 2000

Do design flaws actually affect RCT results?

56

Juni, P. et al. BMJ 2001;323:42-46

Quality assessment of RCTs

Various approaches used:

Checklist approach
Quality scoring system approach

Quality scores are complicated and tend to vary

depending on the instrument used so, not
encouraged

57
Checklist approach

58
Quality score approach

Jadad AR, et al. Assessing the quality of reports on randomized clinical trials: Is blinding necessary? Controlled Clin59
Trials 1996;17:1-12. URL: http://www.bmjpg.com/rct/chapter4.html
Initiatives to improve quality of
reporting of studies
CONSORT: reporting of
RCTs
STARD: reporting of
diagnostic studies
STROBE: reporting of
observational studies
PRISMA: reporting of
meta-analyses of RCTs
MOOSE: reporting of
meta-analyses of
observational studies http://www.consort-statement.org/

Move towards
registration of RCTs>>
60
61
CONSORT: checklist and flow diagram

62
Bibliography and further reading
Schulz KF, Grimes DA. Blinding in randomised trials: hiding who got what. Lancet.
2002 Feb 23;359(9307):696-700.
Schulz KF, Grimes DA. Allocation concealment in randomised trials: defending
against deciphering. Lancet. 2002 Feb 16;359(9306):614-8.
Schulz KF, Grimes DA. Generation of allocation sequences in randomised trials:
chance, not choice. Lancet. 2002 Feb 9;359(9305):515-9.
Haynes BR. Incorporating allocation concealment and blinding in randomised
controlled trials Evidence-Based Medicine 2000; 5:38
Users guide to the medical literature: a manual for evidence-based clinical practice.
Gordon Guyatt and Drummond Rennie (editors). Chicago: AMA Press, 2002.

If you are interested in RCTs,

take EPIB-635 Clinical Trials
Dr. S. Shapiro
63
Textbook readings for this lecture
Gordis text:
Chapter 7 and 8

64