HAD 5303H

CONTROLLED CLINICAL TRIALS

Bruno R da Costa, BScPT, MScPT, MScMedStat, PhD
Associate Director, Applied Health Research Centre (AHRC)
Li Ka Shing Knowledge Institute of St. Michael's Hospital

Assistant Professor, Institute of Health Policy, Management and Evaluation

University of Toronto

Acknowledgment to Niall Ferguson

HAD 5303 Controlled Clinical Trials

Course Coordinators: Teaching Assistant:

Bruno da Costa Christina Belza
bruno.dacosta@utoronto.ca christina.belza@sickkids.ca

Niall Ferguson
niall.ferguson@uhn.ca
Course Objective

Train health care providers & graduate

students in the design, conduct and analysis
of randomized controlled trials (RCTs)
• Basic understanding of all aspects of RCTs
• Write an RCT protocol that could be submitted to a grant
application
Course Format

10 weekly one-hour didactic sessions

Followed by 1 ½ hour small-group tutorials where students

will work on their own protocols
• 4 tutorial groups (5-6 students per group)
• 2 consistent tutors throughout whole course, 2 weeks of statistical
tutors

Expectation: students will prepare a protocol for an RCT

(suitable for a grant submission)
Course Format – Didactic Sessions
Course schedule has been provided with didactic session
titles and speakers
By the prior Friday of each week, a detailed outline of the
next session will be posted on Quercus
Read appropriate section in the textbook, Trial Protocol
Tool, posted articles
Prepare relevant section for your own research study
The lecture slides will be posted on Quercus by the
Tuesday of that week
Readings

Readings from Clinical Epidemiology: How to do

clinical practice research. Haynes.
Pragmatic Trials in Health Care (PRACTIHC) Trial
Protocol Tool (www.practihc.org/toolindex) contains
information needed to prepare protocol
Weekly articles posted on Quercus
Optional textbook
Course Format - Tutorials
Bring ~ 2-4 copies of your section for distribution to your tutorial
group (1 tutor, yourself, other students); bring your computer

Discussion of each student’s progress ~ 10-15 minutes per

student

An attempt will be made to review all 6 protocols for detailed

discussion with the entire group, with facilitation by the tutor.
Due to time constraints, some protocols may be left for the next
tutorial and the next tutorial will begin with those protocols
Assignments and Evaluation
•Assignment #1 in Oct 10%
– 1-page summary of proposal
(PICOT format + study diagram)

•Final written assignment in Dec 50%

– Draft due in Nov
– Full research proposal, 10-13 pages

•Oral presentation, 2 last days 10%

– 15 minutes

•Student review of fellow student protocol, 2 last days 10%

– 1-2 pages

•Participation (1/2 lectures, 1//2 tutorials) 20%

Assignment Guidelines

Detailed information on assignment content, format,

length, etc. is available on Quercus in the document
entitled, “HAD5303H Assignment Guidelines”

Examples of assignments from previous years will

also be posted
Course Administration

Lecture and tutorial participation

Complete weekly lecture evaluation forms
Questions & Special requests - Send email to all –
Bruno da Costa (bruno.dacosta@utoronto.ca), Niall
Ferguson (niall.ferguson@uhn.ca) and Christina Belza
(christina.belza@sickkids.ca)
OVERVIEW OF
RANDOMIZED
CONTROLLED TRIALS
Let’s get started …
The first recorded “Clinical Trial”:
King Nebuchadnezzar
502-562 BC – ruler of
Babylon
Ordered people to eat only
meat and drink only wine
versus
Permitted dissenters to follow
a diet of legumes and H2O for
10 days
At the end of experiment
bean-eaters appeared better
nourished Source: CMAJ 2009, quoting the Book of Daniel
The first parallel-arm experiment (“The
Scurvy Trial”, 18th century):

Dr. James Lind divided 12 sailors into groups of 2:

Cider, few drops of acid, vinegar, seawater, nutmeg
and barley water, or oranges or lemons

Findings:
“all in general had putrid gums, the spots and
lassitude, with weakness of knees”
 The first parallel-arm experiment:
Dr. James Lind

After 6 days:

All remained sick

except the 2 allocated
to citrus treatment
Problems with these early trials

Potential for biases, gaming, chance differences

and developments to influence outcomes

“imperative to demonstrate medical outcomes

uncorrupted by poor judgment, illusion, over-
enthusiasm, imagination, and fraud”
First Medical RCT

Austin Bradford Hill

First Medical RCT
First Medical RCT
Why Randomise?

Patients in different treatment groups A and B are

similar for confounders such as age, gender, social
class, disease severity,…
And also for unknown confounders!
Randomisation sequence must be unpredictable
Generation of a random allocation sequence
Concealment of allocation sequence from personell
assigning patients to treatments.
Minimise allocation bias, selection bias
Subversion of Randomization
Not Necessarily Absolute
Poor Quality RCTs

Trials with high risk of bias

Inadequate concealment
Unbalanced cointerventions
Lack of blinding
“Significant” loss to follow-up
Lack of intention-to-treat analysis (?)
Lack of trial registration / published protocol
…
Trials with low statistical power (“small trials”)
Chance findings
A 2-arm randomised controlled trial

Sample

Concealment of
Randomisation R
allocation

Blinding
Treatment A B
(patients/doctors)

Outcome assessment ? ? Blinding (assessors)

Analysis Intention-to-treat
analysis
Summary of potential biases in clinical trials

Sample

Concealment of
Selection bias R
allocation

Blinding
Performance bias A B
(patients/doctors)

Detection bias ? ? Blinding (assessors)

Attrition bias Intention-to-treat

analysis
Main types of trials
Basic trial design – expectation and content of this course
Conventional 2-armed parallel superiority trial

Advanced trial design

Cross-Over
Multi-armed
Equivalence, non-inferiority
Factorial
Assignment by groups (cluster)
Sequential / Platform
Individual (N = 1)
 Summary of phases of clinical trials
Phase Goal Dose Approx. No. Comments
patients
Pre-clinical Non-human efficacy, Unrestricted In vitro and in
toxicity vivo animals
Phase 0 Pharmokinetics Very small, sub- 10 people Often skipped and
therapeutic joined to phase I
Phase I Pharamcovigilence, Ascending doses 20-100 people Determine I drug safe
tolerability, toxicity, some to check for efficacy
dose finding
Phase II Initial efficacy, toxicity, Therapeutic dose 100-300 people Determines final dose &
dose finding if drug has true efficacy
Phase III Testing for intended Therapeutic dose 300-3000 Determines drug’s
purpose in clinical practice people efficacy
Phase IV Post-marketing Therapeutic dose Anyone Watches for drug long
surveillance term AEs
Phase V Translational research No dosing All reported use
 Phases of clinical trials

The phases of clinical trials are the steps in which experiments with
a health intervention are conducted in order to find enough evidence
to assess if it is useful as a medical treatment.
• Pre-clinical
• Phase 0
• Phase I
• Phase II
• Phase III
For the purpose of this course, only Phase
• Phase IV III and IV clinical trials are to be written
• Phase V
THE RESEARCH
QUESTION
OBJECTIVES
Describe:
The role of a research question in a clinical trial
The source of the question
How to establish the importance of the question
The elements of the question
The types of questions
TYPES OF CLINICAL QUESTIONS

1. Frequency
2. Etiology – Causality
3. Diagnostic tests
4. Effectiveness of interventions
5. Clinical Course and Prognosis
6. Efficiency ( Economic evaluation)
WHY IS THE RESEARCH QUESTION
IMPORTANT?

1. It guides the selection of the most

appropriate study design
2. It improves the credibility of trial
outcomes
Origins of the Research Question

Master the literature

• Systematic Review, General review; Attend meetings;
Network…
Being alert to new ideas & techniques
• Networking; Healthy skepticism; New technologies
Imagination
• Questions from clinical rounds; Creativity; Tenacity
Mentoring
• Need a mentoring team
SELECTION OF THE QUESTION

1. The number of possible questions about

a subject can be very big
2. There are limited resources and time to
answer them
3. The relative importance of a question
can change depending on the
perspective of whomever poses it
CHARACTERISTICS OF A GOOD
RESEARCH QUESTION

Innovative:
• Provides new findings
• Confirms, refutes or widens previous
findings
Approachable, answerable:
• Adequate technical expertise
• Feasible (time and money)
• Adequate size of the sample
What Makes a Good Research Question?

Feasible
Interesting
Novel
Ethical
Relevant
FRAMING THE QUESTION

Well-built questions have 3-5 components (PICOT

format)
Careful framing of the question greatly improves the
quality and utility of the answer.
ELEMENTS OF A WELL-RAISED
QUESTION: PICOT FORMAT
P – Population
I – Intervention
C – Comparison
O – Outcome
T – Time
“BACKGROUND” QUESTION

What is the best treatment for children with

pneumonia?
QUESTION IN “PICOT” FORMAT

In children between three months and five years of

age, hospitalized for severe pneumonia, is the rate
of failure after 48 hours of treatment with oral
amoxicillin equivalent to that observed with
intravenous ampicillin?
P – POPULATION

In children between three months and five years of

age, hospitalized for severe pneumonia [could
define], is the rate of failure after 48 hours of
treatment with oral amoxicillin equivalent to that
observed with intravenous ampicillin?
I - INTERVENTION

In children between three months and five years of

age, hospitalized for severe pneumonia; is the rate
of failure after 48 hours of treatment with oral
amoxicillin [add dose] equivalent to that observed
with intravenous ampicillin?
C – CONTROL

In children between three months and five years of

age, hospitalized for severe pneumonia, is the rate
of failure after 48 hours of treatment with oral
amoxicillin equivalent to that observed with
intravenous ampicillin?
O - OUTCOME

In children between three months and five years of

age, hospitalized for severe pneumonia, is the rate
of failure [could define] after 48 hours of treatment
with oral amoxicillin equivalent to that observed with
intravenous ampicillin?
T - TIME

In children between three months and five years of

age, hospitalized for severe pneumonia, is the rate
of failure after 48 hours of treatment with oral
amoxicillin equivalent to that observed with
intravenous penicillin?
PRIMARY AND SECONDARY
RESEARCH QUESTIONS

Primary:
The one of most interest to the investigator
The one that can be best answered with the
trial
The one that determines the size of the sample
EXAMPLE OF A PRIMARY RESEARCH
QUESTION

In children between three months and five

years of age, hospitalized for severe
pneumonia, is the rate of failure after 48
hours of treatment with oral amoxicillin
equivalent to that observed with
intravenous ampicillin?
THE SAME QUESTION, POSED AS AN
OBJECTIVE

To determine whether the rate of failure

after 48 hours if treatment with oral
amoxicillin, in children between three
months and five years of age, hospitalized
for severe pneumonia, is equivalent to that
observed with intravenous ampicillin
RESEARCH QUESTIONS IN
RANDOMIZED CONTROLLED TRIALS
Secondary:
Other questions related to the primary
question
Data from the study can help to answer them
Two types:
• Evaluate secondary outcomes
• Evaluate primary outcome in subgroups of
subjects
EXAMPLES OF SECONDARY
QUESTIONS
In children between three months and five years of
age, hospitalized for severe pneumonia, is length of
stay reduced with treatment with oral amoxicillin vs.
treatment with intravenous ampicillin?
In the subgroup of children without underlying
pulmonary disease, between three months and five
years of age, hospitalized for severe pneumonia, is
the rate of failure after 48 hours of treatment with
oral amoxicillin equivalent to that observed with
intravenous ampicillin?
OTHER QUESTIONS IN RANDOMIZED
CONTROLLED TRIALS

Adverse effects:
Frequent events
Not always possible to predict
Clinical course – natural history
Follow-up of the control group without an
intervention or receiving placebo
Evaluation of prognostic factors
CAUTION WITH SECONDARY
QUESTIONS

They must be limited in number:

• Raise complexity and costs of the study
• Reduction in the quality of the information
• Multiple statistical tests
They must be specified before collecting the
information
They must be founded on a solid rationale/
scientific basis
CONCLUSIONS

Establishing the research question is crucial for

planning the study
The Trial should have a primary question and a
limited number of secondary questions
The question should specify the population,
the experimental intervention, the control
intervention and the outcomes
HAD 5303 Controlled Clinical Trials

Course Coordinators: Teaching Assistant:

Bruno da Costa Christina Belza
bruno.dacosta@utoronto.ca christina.belza@sickkids.ca

Niall Ferguson
niall.ferguson@uhn.ca