Transcript Diapositiva 1

HISTORY OF "CLINICAL TRIALS“ -
1
Treatment of yellow fever by bleeding, RUSH, 1794
“I began to extract a small amount of blood each time.
The appearance of the blood and the effect of the
intervention made me satisfied on the safety and
efficacy of treatment ....... I have never experienced a
similar feeling of sublime joy when I was able to
contemplate the success of my remedy ..... Thank
God, the more than 100 patients that I treated
personally or to which I have commanded my
attention that day, I have not lost even one "
Rush B. An account of the bilious remitting yellow fever as it appeared in the
city of Philadelphia in 1793. Philadelphia, Dobson, 1794
HISTORY OF "CLINICAL TRIALS“ - 2
Louis P.C.A. (1835)
Analysis of treatment with "bleeding" in various
diseases:
• Pneumonia (78 cases)
• Erysipelas (33 cases)
• Inflammation of the chest (23 cases)
• Division into two groups, treated and untreated
• Result = no average number of better results of
treatment compared with no treatment
Louis P.C.A.: Recherches sur les Effects de la Saignée,
Paris, De Mignaret Ed. 1835
HISTORY OF "CLINICAL TRIALS“ - 3
Lister J. (1870)
Surgical treatment of gangrene according to the traditional
method in comparison with the method that involved using
antisepsis of the operative field by means of carbolic acid
2 groups of patients 35 and 40 cases:
a)Traditional technique mortality = 43%
b) New method mortality = 15%
Lister J.: On the effects of the antiseptic system upon the
salubrity of a surgical hospital. Lancet, i, 4, 1870
CRITERIA FOR EVALUATION OF
A HEALTH INTERVENTION
• EFFICACY
• EFFICIENCY
• SAFETY
TYPES OF CLINICAL STUDIES
ON TREATMENT
•
•
•
OBSERVATIONAL NOT CONTROLLED
STUDIES
a) case series studies
b) before-after studies or with historical
controls
QUASI-EXPERIMENTAL STUDIES
a) controlled studies with contemporary
controls, but not randomized
EXPERIMENTAL STUDIES
a) randomized controlled
b) quasi-randomized controlled
HCT: Historical control trial
RCT: Randomized control trial
Conclusions of "HCT" and "RCT" on 6 different
therapeutic problems
HCT
Result
(outcome)
+
44
12
HCT after
RCT
adjustment for
prognostic
factors
Result
Result
+
+
21
3
10
40
TREATMENTS MUST BE ASSESSED
COMPARATIVELY
• Against another treatment, already
proven effective, in terms of
increased effectiveness, efficiency
or safety.
• Against an inert substance (placebo)
when there is no proven effective
treatment.
CONTEMPORARY ASSESSMENT OF
THE EFFICACY
The judgment of efficacy must be
based on a comparison between the
natural history of the disease in
relation to two different
treatments:
• experimental treatment
• comparator treatment
NEED OF STATISTICS
The evaluation of efficacy must
depend on comparison of the
frequency of significant events in a
group of treated subjects and in a
control group.
The comparison can not disregard a
probabilistic assessment and, then,
from a statistical analysis of the
data.
THE SAMPLE SIZE
THE PROBABILISTIC APPROACH AND
SUBSEQUENT USE OF STATISTICS
IMPLIES PRE-DETERMINATION OF
THE SAMPLE SIZE REQUIRED TO
EVALUATE THE EFFECT OF A HEALTH
INTERVENTION.
SEVERAL ASPECTS MUST BE TAKEN
INTO ACCOUNT.
The probabilistic approach
• Concept of statistical inference
• 95% confidence intervals
• Significant difference
Whole sample
23 red/64 =35%
Two samples
7 reds/27=25%
Four samples
7 reds/18=22%
10 reds/20=50%
1 red/10=10%
5 reds/15=30%
0
35
100
0
16 reds/37=43%
35
100
Whole sample
23 reds/64 =35%
26 samples
from
1 red/1= 100%
to
0 red/3= 0%
0
35
100
95% Confidence intervals of the Mean
• The minimum and maximum values that can
be found in 95% of our samples
• The larger the sample the more the value
found is close to reality
• The larger the sample the more narrow the
confidence interval
DIFFERENT STAGES OF THE
CLINICAL DRUG "TRIAL"
• PHASE I: clinical pharmacology and
toxicology
• PHASE II: preliminary clinical efficacy
and safety of treatment, "dose
ranging", pharmacokinetics and
pharmacodynamics in human.
• PHASE III: definitive study of the
effect of treatment-RCT
• PHASE IV: "post-marketing” monitoring
Phase III: Randomized
Controlled Trial
• Some preliminary concept:
internal /external validity
randomization
masking
power of the study
Phase III: Randomized
Controlled Trial
Randomization:
to avoid the "selection bias”
Must be "concealed"
Phase III studies: double
blind RCT
Masking:
to avoid "bias" in the evaluation
Expectation bias
Co-interventions
Single, double, triple ... better describe
who is blinded to treatment
Phase III studies: Randomized
Controlled Trial
• Power of the study
No. of patients sufficient to show any
effect of the treatment with a good
probability and with few probability of
being wrong.
ISSUES RELATED TO THE SAMPLE SIZE
• WHICH AIMS HAS THE STUDY?
• WHAT MEASURE IS USED FOR EVALUATING THE
OUTCOME OF THE PATIENT?
• HOW IS ANALYZED DATA TO IDENTIFY A POSSIBLE
DIFFERENCE BETWEEN TREATMENTS?
• WHAT IS THE EXPECTED RESULT OF THE "STANDARD"
TREATMENT OR THE PLACEBO?
• WHAT DIFFERENCE BETWEEN THE TWO TYPES OF
TREATMENTS COMPARED IS SUFFICIENT TO CONVINCE
US THAT ONE IS BETTER THAN THE OTHER?
• WHAT TYPE I ERROR AND TYPE II ERROR VALUES DO
WE DECIDE TO CHOOSE?
EXAMPLE OF THE SAMPLE SIZE CALCULATION
1) WHICH AIMS HAS THE STUDY?
To DETERMINE IF A NEW ANTIPLATELET
DRUG IS BETTER THAN “ASA” FOR
SECONDARY PREVENTION OF ISCHEMIC
STROKE
2) WHAT MEASURE OF "OUTCOME" WOULD
YOU LIKE TO USE?
MORTALITY + DISABLING STROKE WITHIN
2 YEARS AFTER A PREVIOUS TIA
EXAMPLE OF THE SAMPLE SIZE
CALCULATION
3) HOW DO WE ANALYZE DATA FOR
ASSESSING THE TREATMENT?
ASCERTAINMENT OF NEW STROKES AND
DEATHS THROUGH A SYSTEMATIC FOLLOWUP – COMPARISON OF THE FREQUENCY OF
EVENTS IN THE TWO GROUPS.
4) WHICH RESULT IS EXPECTED WITH ASA:
MORTALITY + disabling stroke EXPECTED
WITHIN 2 YEARS WITH ASA = 10%
EXAMPLE OF THE SAMPLE SIZE
CALCULATION
5) WHICH FREQUENCY OF EVENTS DO I
EXPECT WITH THE NEW DRUG TO DEFINE
IT MORE EFFECTIVE THAN THE
COMPARATOR?
EXPECTED RATE OF MORTALITY +
EXPECTED RATE OF DISABLING STROKE
WITHIN 2 YEARS WITH THE NEW
ANTIPLATELET = 5%
6) WHAT IS THE EXPECTED DIFFERENCE OF
THE FREQUENCY PERCENTAGE OF EVENTS?
10% (ASA) - 5% (NEW DRUG) = 5%
EXAMPLE OF THE SAMPLE SIZE
CALCULATION: ERROR TYPE I
6) WHICH VALUE OF
TO USE?

DO I INTEND
5 % (p < 0,05)
 = probability that the result of
the study may be positive due
to chance
EXAMPLE OF THE SAMPLE
CALCULATION: ERROR TIPE II
7) WHICH VALUE OF ß DO I CHOOSE?
10 %
ß = probability that is possibly
negative result of the study due to
chance
1- ß = power of the study (90%)
EXAMPLE OF THE SAMPLE SIZE
CALCULATION:
Size (N) of each group to be
studied = 578 (Total
subjects = 1156)
P1 = 10%
=
5%
P2 = 5%
ß = 10%
Classification of RCTs
• According to the intervention aspects
you want to check
• According to the manner in which the
participants are exposed to be treated
• According to the number of participants
• According to the participants and
investigators knowledge about the
treatment
• According to the objective
Classification of RCTs
• “explanatory”
have the purpose of assessing whether
a treatment works or not independently
in other variables
high internal validity
• “pragmatic”
have the purpose of assessing whether
the treatment works or not in
conditions similar to practice
high external validity
CLASSIFICATION OF RCTs
“explanatory” trial
efficacy
internal validity
“pragmatic trial”
effectiveness
external validity
Classification of RCTs
• Parallel (parallel groups)
• Cross-over
• Factorial
• No. of 1 trial
Clinical Trial
 Randomized and Controlled Clinical Trial with parallel groups
Randomization
Treatment
A
Analysis
of
Selected
sample
results
Treatment
B
Clinical Trial
 Randomized and Controlled Clinical Trial with Cross-over
Randomization
Treatment
A
Treatment
A
Analysis
of
Selected
sample
results
Treatment
B
Treatment
B
Clinical Trial
 Clinical Trial with Factorial Design
Treatment A
Treatment B
Selected
sample
Treatment A+B
Placebo
Analysis
of
results
Classification of RCTs
• open
• single-blind
• double-blind
• triple, quadruple blind
Classification of RCTs
• “mega-trial”
(very large simple pragmatic trial)
• sequential
• Pre-defined size
Classification of RCTs
• Superiority trials
• Equivalence trials
• Non-inferiority trials
95% confidence interval
P = 0.002
Strongly demonstrated superiority
P = 0.05
Demonstrated superiority
P = 0.2
Not demonstrated superiority
Control Better
0
Treatment Better
Difference between treatments
Equivalence margin (10%)
Equivalence
Demonstrated
Equivalence not
demonstrated
-
Control Better
+
0
Treatment Better
Difference between treatments
Non-inferiority margin (5%)
Non-inferiority
Demonstrated
Non-inferiority
Not demonstrated
-
Control Better
0
Treatment Better
Difference between treatments
Statistical Analysis
• For protocol
• Intention to treat
concept of “drop-out”
concept of lost to follow-up
Whole sample
100 points
35% reds
26 samples
from
2 reds/3=66%
to
1 red/6=16%