Intro Clinical Trails
This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA
Overview
Download & View Intro Clinical Trails as PDF for free.
More details
- Words: 1,169
- Pages: 28
© 2003, Greenfield Research
1
introduction to Statistical aspects of
clinical trials
© 2003, Greenfield Research
2
© 2003, Greenfield Research
3
New Clinical Entity (NCE) In-vitro: combinatorial chemistry; pharmacogenetics; effects First Often studies first studies inregistration man, in patients. healthy Attempts volunteers. to prove efficacy; dose After Large or scale during studies tousually prove to discover drug suitable more for about registration. safety, efficacy including, metabolites, toxicity, drug-drug interactions; Pharmacokinetics, finding; tolerability pharmacodynamics, basic tolerabilityidentify in Control different group: populations. placebo or Many proven lacktherapy. control group. surrogate markers.
10 to 20 years
Phase two
Pre-clinical Phase three phase Phase one concept Phase four © 2003, Greenfield Research
profit 4
Clinical trials To determine whether or not there are differences between the effects of treatments Treatments A and B purpose
A
design
trial
data
difficulties
B
? A
=
B
Analysis method knowledge
•Previous studies •publications •data bases •theory
© 2003, Greenfield Research
5
Difficulties and ethics: •
Patients • • • • • •
•
Data • • • • •
• © 2003, Greenfield Research
availability inclusion and exclusion criteria willingness to participate presentation rates compliance how many? measures of effects adverse effects influence of other factors random variation of effects between people bias • allocation bias • assessment bias • analysis
Cost 6
Assume everything else controlled except for random variation of effects between people
Consider: • a single continuous quantitative variable and its presentation distribution • random allocation of patients to two treatments • posterior (after treatment) distributions: are they they same or different for the two treatments, as measured by their means and standard deviations?
© 2003, Greenfield Research
7
FEV1: • prior (presenting) distribution:
© 2003, Greenfield Research
• posterior (after treatment) distributions:
• mean = 1.70
• treatment A mean = 1.75
• sd = 0.30
• treatment B mean = 1.80
8
Two approaches to analysis: 1.
Compare means of the two posterior distributions.
2.
Compare the means of the changes for all the patients in the two treatment arms.
Statistical tests are needed for these comparisons Statistical tests help us to decide how many patients will be needed
© 2003, Greenfield Research
9
Form of a test statistic:
What you want to test Variability of what you want to test Need to understand: • Probability distributions, means and standard deviations • Means and standard deviations of sums of variables • Sampling distributions of means, standard errors of means • Central limit theorem © 2003, Greenfield Research
10
Is treatment A better than treatment B?
© 2003, Greenfield Research
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
11
If treatment A is better than treatment B . . .
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Why did these eight recover? Were they younger, stronger, or in a better condition than those who did not recover? © 2003, Greenfield Research
12
If treatment A is better than treatment B . . .
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Why didn’t these three recover? Were they older, more feeble, or in a worse condition than those who did? © 2003, Greenfield Research
13
Perhaps treatment A is not better than treatment B? So . . .
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Could there have been an allocation bias?
© 2003, Greenfield Research
14
Perhaps treatment A is not better than treatment B? So . . .
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Could there have been an assessment bias?
© 2003, Greenfield Research
15
Perhaps treatment A is not better than treatment B? So . . . Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Could the results have occurred by chance?
© 2003, Greenfield Research
16
Pharmacogenetics:
How differences in genes can influence responses to drugs
© 2003, Greenfield Research
17
Example of pharmacogenetics
codeine
placebo
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Perhaps these 3 cannot respond to codeine © 2003, Greenfield Research
18
codeine
loss of
morphine
methyl group enzyme:
cyp2d6
Gene encoding cyp2d6 is on chromosome 22
It is altered in ten percent of people
© 2003, Greenfield Research
19
The efficiency and effectiveness and cost of a clinical trial depend on:
• • • • • • • • © 2003, Greenfield Research
response to each treatment influence of other factors such as age, gender or life style number of patients how patients are selected for the trial how patients are allocated to treatments type of trial:
parallel or crossover
compliance of patients to treatments how data are recorded, analysed and interpreted 20
Type of trial: • Parallel • Cross-over
© 2003, Greenfield Research
•
Sequential
•
Group sequential
•
Factorial 21
Parallel Designs POPULATION Assessment Eligible and willing subjects Allocation Treatment A
Treatment B
Assessment Data analysis
© 2003, Greenfield Research
arm 1
arm 2
22
Two-period Cross-over designs POPULATION Assessment Eligible and willing subjects Allocation
First period
Treatment A
Treatment B
Assessment
Second period
Treatment B
Treatment A
Assessment Data analysis
© 2003, Greenfield Research
arm 1
arm 2
23
Allocation Random • Simple randomisation •
Weighted randomisation
• Block randomisation • Sequential randomisation Semi-random • Minimisation Non-random • Open © 2003, Greenfield Research
24
Sample size
number of patients in a trial
based either on independent samples t-test or on comparison of proportions
based either on a
paired t-test
or on comparison of paired proportions
© 2003, Greenfield Research
25
Sample size
number of patients in a trial
alpha value:
probability of a type one error : of rejecting the null hypothesis when it is actually true 0.05 or 0.01
power:
probability (80% or 90%) of getting a statistically significant difference if the true difference between treatments is of a given size (clinically significant difference) adjust for expected dropouts
© 2003, Greenfield Research
26
Exclusions: patients presenting with the condition being studied must be assessed for suitable inclusion: they may be excluded if, for example: • • • • •
over-weight over or under age pregnant exhibiting what may be an adverse response on interacting medication
Exclusion must be decided before allocation to treatment © 2003, Greenfield Research
27
Intention to treat After a patient has been entered into a trial, after allocation to treatment, that patient’s data must be included in the analysis even if the patient has dropped out of the trial before completion of treatment or if there are any missing values. If ‘intention to treat’ analysis is being applied, missing values of a variable after drop out will be replaced by the last values recorded. © 2003, Greenfield Research
28
1
introduction to Statistical aspects of
clinical trials
© 2003, Greenfield Research
2
© 2003, Greenfield Research
3
New Clinical Entity (NCE) In-vitro: combinatorial chemistry; pharmacogenetics; effects First Often studies first studies inregistration man, in patients. healthy Attempts volunteers. to prove efficacy; dose After Large or scale during studies tousually prove to discover drug suitable more for about registration. safety, efficacy including, metabolites, toxicity, drug-drug interactions; Pharmacokinetics, finding; tolerability pharmacodynamics, basic tolerabilityidentify in Control different group: populations. placebo or Many proven lacktherapy. control group. surrogate markers.
10 to 20 years
Phase two
Pre-clinical Phase three phase Phase one concept Phase four © 2003, Greenfield Research
profit 4
Clinical trials To determine whether or not there are differences between the effects of treatments Treatments A and B purpose
A
design
trial
data
difficulties
B
? A
=
B
Analysis method knowledge
•Previous studies •publications •data bases •theory
© 2003, Greenfield Research
5
Difficulties and ethics: •
Patients • • • • • •
•
Data • • • • •
• © 2003, Greenfield Research
availability inclusion and exclusion criteria willingness to participate presentation rates compliance how many? measures of effects adverse effects influence of other factors random variation of effects between people bias • allocation bias • assessment bias • analysis
Cost 6
Assume everything else controlled except for random variation of effects between people
Consider: • a single continuous quantitative variable and its presentation distribution • random allocation of patients to two treatments • posterior (after treatment) distributions: are they they same or different for the two treatments, as measured by their means and standard deviations?
© 2003, Greenfield Research
7
FEV1: • prior (presenting) distribution:
© 2003, Greenfield Research
• posterior (after treatment) distributions:
• mean = 1.70
• treatment A mean = 1.75
• sd = 0.30
• treatment B mean = 1.80
8
Two approaches to analysis: 1.
Compare means of the two posterior distributions.
2.
Compare the means of the changes for all the patients in the two treatment arms.
Statistical tests are needed for these comparisons Statistical tests help us to decide how many patients will be needed
© 2003, Greenfield Research
9
Form of a test statistic:
What you want to test Variability of what you want to test Need to understand: • Probability distributions, means and standard deviations • Means and standard deviations of sums of variables • Sampling distributions of means, standard errors of means • Central limit theorem © 2003, Greenfield Research
10
Is treatment A better than treatment B?
© 2003, Greenfield Research
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
11
If treatment A is better than treatment B . . .
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Why did these eight recover? Were they younger, stronger, or in a better condition than those who did not recover? © 2003, Greenfield Research
12
If treatment A is better than treatment B . . .
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Why didn’t these three recover? Were they older, more feeble, or in a worse condition than those who did? © 2003, Greenfield Research
13
Perhaps treatment A is not better than treatment B? So . . .
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Could there have been an allocation bias?
© 2003, Greenfield Research
14
Perhaps treatment A is not better than treatment B? So . . .
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Could there have been an assessment bias?
© 2003, Greenfield Research
15
Perhaps treatment A is not better than treatment B? So . . . Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Could the results have occurred by chance?
© 2003, Greenfield Research
16
Pharmacogenetics:
How differences in genes can influence responses to drugs
© 2003, Greenfield Research
17
Example of pharmacogenetics
codeine
placebo
Treatment A
Treatment B
total
Recovered
17
8
25
No better
3
12
15
total
20
20
40
Perhaps these 3 cannot respond to codeine © 2003, Greenfield Research
18
codeine
loss of
morphine
methyl group enzyme:
cyp2d6
Gene encoding cyp2d6 is on chromosome 22
It is altered in ten percent of people
© 2003, Greenfield Research
19
The efficiency and effectiveness and cost of a clinical trial depend on:
• • • • • • • • © 2003, Greenfield Research
response to each treatment influence of other factors such as age, gender or life style number of patients how patients are selected for the trial how patients are allocated to treatments type of trial:
parallel or crossover
compliance of patients to treatments how data are recorded, analysed and interpreted 20
Type of trial: • Parallel • Cross-over
© 2003, Greenfield Research
•
Sequential
•
Group sequential
•
Factorial 21
Parallel Designs POPULATION Assessment Eligible and willing subjects Allocation Treatment A
Treatment B
Assessment Data analysis
© 2003, Greenfield Research
arm 1
arm 2
22
Two-period Cross-over designs POPULATION Assessment Eligible and willing subjects Allocation
First period
Treatment A
Treatment B
Assessment
Second period
Treatment B
Treatment A
Assessment Data analysis
© 2003, Greenfield Research
arm 1
arm 2
23
Allocation Random • Simple randomisation •
Weighted randomisation
• Block randomisation • Sequential randomisation Semi-random • Minimisation Non-random • Open © 2003, Greenfield Research
24
Sample size
number of patients in a trial
based either on independent samples t-test or on comparison of proportions
based either on a
paired t-test
or on comparison of paired proportions
© 2003, Greenfield Research
25
Sample size
number of patients in a trial
alpha value:
probability of a type one error : of rejecting the null hypothesis when it is actually true 0.05 or 0.01
power:
probability (80% or 90%) of getting a statistically significant difference if the true difference between treatments is of a given size (clinically significant difference) adjust for expected dropouts
© 2003, Greenfield Research
26
Exclusions: patients presenting with the condition being studied must be assessed for suitable inclusion: they may be excluded if, for example: • • • • •
over-weight over or under age pregnant exhibiting what may be an adverse response on interacting medication
Exclusion must be decided before allocation to treatment © 2003, Greenfield Research
27
Intention to treat After a patient has been entered into a trial, after allocation to treatment, that patient’s data must be included in the analysis even if the patient has dropped out of the trial before completion of treatment or if there are any missing values. If ‘intention to treat’ analysis is being applied, missing values of a variable after drop out will be replaced by the last values recorded. © 2003, Greenfield Research
28
Related Documents
Intro Clinical Trails
May 2020 4
Clinical Trails Glossary
November 2019 12
Multicenter Trails
November 2019 27
Innovative Trails
June 2020 13
Chem Trails
December 2019 30