Transcript Why Do Clinical Research?

Why Do Clinical Research?
• Satisfaction of answering important
questions which will improve the
health of our patients
•
•
•
•
Status of researchers
Skill advancement
Professional advancement
Salary and Job Security
What is Research?
• Research is the endeavor to
discover new facts, procedures,
methods, and techniques by the
scientific study of a course of
critical investigation
Clinical Research
• Clinical research involves working
with human subjects to answer
questions relevant to their wellbeing
• Patient oriented research is where
the ‘rubber meets the road’!
‘How To Do’ Research
• Start with defining the question
• Write down a clear aim
• Divide the problem into smaller,
answerable questions
‘How To Do’ Research
• Develop hypotheses
• Decide what data is needed to test
the hypotheses
• Refine the above and check the line
of thought
Good Research
• CLEAR
– Essential for both the problem and the
answer
• ACCURATE
– Exactness and precision come from
hard work and responsible effort
• RELIABLE
– If repeated will the answer be the
same?
Good Research
• OBJECTIVE
– The researcher exposes all possible
prejudices at the onset of the study
design and strives to overcome them
– Will the research be untarnished by
personal gain, biases, vested interests,
etc?
Researcher Qualities
• Knowledgeable
• Motivated
• Observant
• Independent
• Logical
• Flexible
• Open-minded
• Careful
• Honest
Researcher Qualities
• Curious
• Persistent
• Inquisitive
• Patient
• Eager to learn
• Original
• Skeptical
• Creative
• Perceptive
Getting Started
•
•
•
•
Learn your subject
Read, Read, Read
Start general and then focus
Begin with the problem
Getting Started
• Formulate the problem as a research
question
• Reduce the question to a single
unambiguous question that is welldefined and answerable
Stages in Creativity
• SENSE
– Realize the need for a study
• PREPARE
– Gather relevant information
• INCUBATE
– Think through the problem
• ILLUMINATE
– Imagine possible solutions
• VERIFY
– Evaluate the solutions you have generated
Hypothesis
• Thesis is the position that you
believe represents truth
• Hypothesis is the foundation on top
of which you build your thesis
Hypothesis
• Hypothesis is a tentative construct
to be proved or disproved according
to the evidence
• The hypothesis is sometimes
expressed as a null hypothesis
A Good Hypothesis Should:
• Be testable
• Convey the nature of the
relationship being tested
• State exactly what variables form
this relationship
• Reflect all variables of interest
• Be formulated early on in the
planning stage
Study Types
• Will you test a hypothesis or
describe a phenomenon?
• Observational
– Longitudinal
– Cross-sectional
• Randomized, double-blind, parallel
group, placebo controlled trial
Epidemiology vs RCT
• Epidemiology allows the study of the
real world and the development of
hypothesis regarding disease states
• Randomized, controlled trials allow
the rigorous testing of hypothesis in
a well characterized manner that is
less real world in nature
Study Design
• Study Population
– Age
– Gender
– Ethnicity/Race
– Disease characteristics
– Exclusions
– Number
– Stratification
– Randomization
Human Subjects
• The safety and rights of human
subjects must be protected
– Study Design
– Institutional Review Board
– Informed consent
– Data Safety Monitoring/Medical
Monitors
Key Questions
• What is the main purpose of the
trial?
• What treatments will be used and
how?
• What is the participant risk?
• What are the possible benefits?
• How will patient safety be
monitored?
Key Questions
• Are there alternative treatments?
• Who is sponsoring the trial?
• What is the participant burden?
– How long and where?
– What do the participants have to do?
– Will there be any discomfort even if
there is no risk?
Methods
•
•
•
•
Define methods carefully
Decrease variability
Check reliability/reproducibility
Are you testing what you think you
are testing?
Methods
• Try to ‘walk through’ the study and
consider as many likely scenarios as
possible.
• Try to design in any variations in
treatment or data collection that you
think will occur before the study
starts
Operationalize Concepts
• Specify how you will repeatably and
reliably measure the variables you
are using to answer the question
• An operational definition specifies
how your concepts will be observed
and measured
• This should allow your research to
be reproduced
Data
• Data are the facts you measure
• They should be carefully recorded in
an unbiased manner
• They should be measured in a
manner that minimizes random
variation
• They should be derived from the
operational definitions you have
developed
Data Validation
• Do the data make sense?
• Look critically at the data
– Highest and lowest values
– Data entry errors
– Distribution: Normal or skewed
• Check selected data entries with
original data forms
Data Interpretation
• Do not interpret/analyze data until
after study is completed
• Do not ‘unblind’ subjects until the
study is completed other than for
safety reasons
• Do not interpret/analyze data until
after data has been validated and
the data set closed
Data Interpretation
• Use the research question and
hypotheses to guide analyses
• Use a priori definitions for any subset analyses
• Exploration of epidemiologic data
sets is OK, but need to avoid data
mining
Writing It Up
• If you don’t write it, then it didn’t happen
• Order of writing:
– Methods
– Results
– Introduction
– Discussion
– Abstract
– Title
Writing It Up
• After the first draft, new analyses
will usually be suggested by the
process of putting your ideas down
on paper
• Put the paper away for a few weeks
and then read it again
• Ask mentors and colleagues to read
the paper at the first draft stage
Sending It In
• When writing the paper, have the
journal you will submit to in mind
• Pick journals that will match your
paper’s topic and the quality and
importance of your work
• Aim high and, if needed, go low
• Persist, Persist, Persist
Clinical Research
Drug Development
Drug Development
• Preclinical/Laboratory Study
– Cell culture in animal and human cells
– Animal studies
– Looking both at toxicity/carcinogenicity
as well as effect, if relevant
• Develop Investigational New Drug
application with FDA (IND)
Phase I Studies
• Assess drug safety and tolerability
• Healthy volunteers, then those with target
disease
• Pharmacokinetics
– Absorption
– Metabolism
– Excretion
• Dose escalation
• 70% of new drugs pass this phase
Phase II Studies
• Assess drug efficacy
• Usually randomized, controlled trials
with smaller numbers up to several
hundred subjects
• Test different therapeutic strategies
• Use surrogate variables and are
usually short term
• Only 1/3 get past phase II
Phase III Studies
• Large scale RCT to assess efficacy and
safety of medication
• Several hundred to thousands of patients
enrolled
• Classic randomized, placebo-controlled
design
• Long-term study design with real world
outcome variables
• Define package insert content and allow
marketing
Study Size and Adverse Events
• The size of the treatment group
determines the likely frequency of
adverse events (side effects) that can be
detected
• A good rule of thumb is that you can
detect an adverse event rate that is one
event in the number of subjects divided
by three:
– A study with 100 patients will only detect AE’s
that occur at a rate of 1/33 = 3%
Phase IV Studies
• Compare drugs with other drugs on
the market
• Define broader target population
• Monitor long-term efficacy and
safety
• Conduct health economics
assessment and quality of life study
Reading Clinical Research
How to Approach RCT Reports
Reading Clinical Trials
• ‘All that glitters is not gold’ by Bengt
and Curt Furberg
• Just because a study is published in
a journal does not mean that it
represents truth
• ‘Throwaways’ and Drug company
sponsored newsletters have either
no or limited peer review
Was the question stated A Priori?
• Exploring data is acceptable to
define hypotheses, but cannot
definitively answer them
• Primary outcomes and limited
secondary outcomes should be
carefully defined before study
commences
Was the question stated A Priori?
• Multiple hypothesis testing can lead
to false association
• P <0.05 is subverted if there are 20
looks at the data
Is the question relevant?
• Does the answer clarify whether the
treatment will help patients to:
– Feel better
– Live longer
– Have less complications of illness
• Are the endpoints real world or
merely surrogates
• How can one generalize the
findings?
How is improvement quantified?
• Are the outcomes relevant?
• Do the measures used make sense?
• Is the magnitude of the difference
relevant to patient care?
• Is the study ‘over-powered’?
Are the outcomes relevant?
• Quality of life
• Mortality
• Health economics
• Surrogate markers of clinical
outcome
• Surrogate biologic markers
How are adverse events measured?
• Side effects are characterized as:
– Severe: Treatment must be stopped, or
patient hospitalized, or dies, or
develops cancer, or has congenital
anomaly in child
– Moderate: Dosage must be reduced,
usually leads to discomfort, temporary
disability, or reduction in functioning
– Mild: No change in treatment. Limited
discomfort or dysfunction
How are adverse events measured?
• AE’s are characterized as to
whether or not they are related to
the medication:
– Definitely
– Likely
– Probably
– Possibly
– Not associated
Are the patients representative?
• This is most problematic in
pediatrics where we often have to
extrapolate from adult studies
• Gender, age, and race can all alter
outcomes
• Disease classification and severity
can alter outcomes
• High risk patients are usually
excluded
Where the groups initially comparable?
• Even in studies of 150-200 subjects
substantive imbalance can occur
between treatment groups
• Was stratification used to ensure
balance?
• Did the treatment group start out
sicker so that they likely would
improve more than the placebo
group?
Excluded Subjects?
• Intent to treat analyses should be
reported
• Two unacceptable reasons to
exclude subjects are:
– After randomization where they do not
meet entry criteria
– Because they did not take the
medication
Do you need a statistician to read the
study?
• In clinical trials, design should allow
relatively straightforward
presentation of results
• Effect size and relevance are more
important than P values
Do you need a statistician to read the
study?
• Consider the number of patients
who would have to be treated to
avoid the outcome being prevented
• Subgroup analyses should be
avoided unless defined a priori
Economic Analysis
• “Of course our drug is more
expensive, but we need to convince
clinicians to use it more”
• Does the medication reduce direct
or indirect costs or both?
Economic Analysis
• Be sensitive to relationship between
the authors and the sponsor
• Be careful if soft assumptions are
used
• Beware of analyses based on the
clinical trial setting and not the real
world
• Beware indirect evidence with
surrogate markers