Transcript 500

Critical appraisal of RCTs
• Internal validity ( to control bias)
inclusion and exclusion criteria
randomization
blinding or masking
outcome definition and measure
primary and secondary end-point
study power
co-treatments definition and standardization
drop-out and lost to follow-up
statististical analysis
Critical appraisal of RCTs
• External validity
results generalization
clinical features of patients
availability and applicability
economic sustainability
Internal validity
• Inclusion and exclusion critera
well defined and described
more strict less external validity
Internal validity
• Randomization (selection bias)
concealed
simple
stratified
with blocks (multicenter studies)
To balance prognostic variables
Internal validity
• Blinding (information bias, performance
bias,detection bias)
drug appearence (colour, taste , size)
side effects
administration route( “double dummy”)
Who is unaware ( blind) of treatments:
patients
investigators
outcome assessor
statisticians
investigators writing the reports
Internal validity
• Definition and measure of outcomes
which outcomes (e.g. disability, pain)
measurement tools reliability (e.g.
validate scale)
appropriateness
reproducibilty
Internal validity
• Primary and secondary end-points definition
Internal validity
• co-treatments definition and standardization
necessary for open label studies
always relevant to avoid treatment bias
Internal validity
Study power calculation (type I and II errors)
on the basis of primary end point
background rates estimation ( phase II
studies, epidemiological data)
Internal validity
• Drop-out and lost to follow-up*
drop-out: patients who stopped the treatment but
were followed and with outcome available
lost at follow-up: patients with outcome
unavailable
(* If more than 20%, doubts on results reliability)
Internal validity
• Statistical analysis
per protocol
intention to treat (attrition bias)
sensitivity analysis
LOCF
• Statistical significance and clinical relevance
Measures to describe results
Control group
Experimental group
End point +
40
20
End point -
460
480
500
500
40
20
Absolute frequecy =
Relative frequency = 40/500 = 0,08
Odds =
40/460 =
0,09
20/500 = 0,04
20/480 =
0,04
Measure to compare results
Control grpup
Experimental group
End point +
40
20
End point-
460
480
500
500
8%
= 0,09
Relative frequency =
Odds
:
:
4%
0,04
RR = 2,0
OR = 2,3
Measure to quantify results
Control grpup
Experimental group
End point+
40
20
End point -
460
480
500
500
Relative frequency =
_
4%
________________
8%
8%
ARR = 4%
RRR = 50%
Measure to compare results
• NNT = number needed to treat
to obtain one unfavourable
outcome less
1/ARR
1/0.04 = 25
Effect of end point frequency on
ARR and RRR
Control group
Experimental group
End point +
400
200
End point -
100
300
500
100
Relative frequency=
_
40%
________________
80%
80%
ARR = 40%
RRR = 50%
NNT
1/0.40=2.5
Effect of end point frequency on
ARR and RRR
Control group
End point +
End point -
Relative frequency =
Experimental group
4
2
496
498
500
500
_
0,4%
________________
0,8%
0,8%
ARR = 0,4%
RRR = 50%
NNT
1/0.004=250
External validity
•
•
•
•
•
Patients features
Setting
Treatment avalaibility
Treatment applicabilty
Cost benefit ratio