Absolute and relative risk reduction

Download Report

Transcript Absolute and relative risk reduction

Reporting drugs and
treatments
Thomas Abraham
What we will learn today




The difference between absolute and relative
risk reduction
A basic way to interpret P values and CI
(confidence intervals)
Number needed to treat (NNT)
All of these are to help you evaluate and
report on studies
Nov 1995, results of the trial of a
drug announced



6,595 men in the west of Scotland with
elevated cholesterol level participated in trial
that lasted 5 years
One group (treatment group ) was given a
cholesterol lowering drug, pravastatin
Another group (control group) was given a
placebo
The results were impressive
For those who took the drug:
 31 % reduction of risk from heart attacks
 28 % reduction of risk of death from
coronary heart disease
 31 % reduction of risk from heart attack,
or death from heart disease
Combined risk from heart attacks and death from heart disease
7.9%
5.5%
Risk from heart attack and death from heart disease
Control group
A
7.9%
Treatment group
B
5.5%
How do you measure the difference between these two groups?
Absolute and relative risk
reduction


Absolute risk reduction
looks at the difference
between the treatments:
A-B =2.4%
Relative risk reduction
looks at how much better
one treatment is than the
other
(A-B/ A) x 100=30.37%
A
Control
7.9%
B
Treatment
5.5%
Absolute and relative risk



Absolute risk is your risk of getting a
particular disease, or condition
Relative risk measures how much your risk is
reduced with the new treatment, or
procedure, compared with the old.
Relative risk figures tend to look dramatic
Death from coronary heart disease
1.8
1.6
1.4
1.2
A
Control
1.7 %
B
Treatment
1.2%
1
0.8
0.6
0.4
0.2
0
Series1
Placebo
1.7
1.2
Series1


Absolute risk reduction=
Relative risk reduction=
Non fatal Myocardial Infarction
A
Control
6.5%
Absolute Risk= ?
B
Treatm
ent
4.6%
Relative risk= ?


What were the different outcomes the study
measured?
Which outcome was the treatment the most
effective in?
P values and Confidence
Intervals ( CI)
P value or probability value= The probability
that a result could have occurred by chance
P < .05 means a less than 5 percent probability
that this result could have occurred by chance
P< .01 means a less than 1 percent probability
that this could have occurred by chance
P <.001 means ?
(move decimal point two places to the right)


According to scientific convention, P<.05 is
significant, and P<.01 is highly significant.
Rank the following P values in
terms of significance
A.
B.
C.
D.
E.
F.
G.
H.
P<.1
P<.001
P<.01
P<.4
P<.04
P<.5
P<.0001
P<.05
Confidence intervals (CI)


The confidence level tells you the range
within which a true value will lie. It is
generally measured at the 95% level.
For example, 31 ( 17-43) 95% CI would
indicate that 95% probability that the real
value is between 17 and 43.
Number needed to Treat (NNT)


The number of persons who needed to be
treated with a drug, to prevent one person
from getting the disease.
Calculated as the inverse of the absolute risk
reduction percentage
Death from coronary heart disease
1.8
1.6
1.4
1.2
1
A
Control
1.7 %
0.8
0.6
0.4
B
Treatment
1.2%
0.2
0
Series1
Placebo
1.7
1.2
Absolute Risk Reduction = A-B= .5%
. 5% means out of 100 people, .5 death was prevented
To prevent one death: 200 people need to be treated ( 100/ARR%)
Time-5 years. Therefore 200 people over 5 years to prevent 1 case
Non fatal Myocardial Infarction
A
Control
6.5%
B
Treatm
ent
4.6%
Absolute Risk Reduction = A-B= 2.9%
Number needed to treat 100 people, 2.9 heart attacks prevented
To prevent one heart attack= 100/ARR%= 100/2.9
Over 5 years
Check list





When you hear about a new drug or
treatment, check both absolute and
relative risk reduction figures.
http://www.bbc.co.uk/news/health16961112
Ask ( or calculate) number needed to
treat
Ask about possible harms
Ask about costs

http://www.youtube.com/watch?v=lNFuEcy
5ekg&feature=related
From last week

-
-
When evaluating research findings
Published or not published?
Quality of study: numbers of people or cases
studied, methodology of study
Newsworthiness
Your course work






Three news stories from journal articles
Due Feb 28, March 28 and March 17(you can
turn them in earlier if you wish)
Length: 400-500 words
Go to Eureka alert, as well as Jama and BMJ
Links to press releases on our course website
Find an article that you think might be
newsworthy and interesting.




Do some background research on the topic.
Go to Pubmed for earlier studies
Contact authors of paper, as well as people in
Hong Kong who might have comments.
Give yourself one week for each
Contacting paper authors





Why?
Not to “get a quote”
To help bring out the significance, or
importance of study
To help you understand things you do not
understand
Read the paper and do research before
getting in touch, so you do not ask basic
questions
What I will look for




Is the story newsworthy?
Have you explained the research well?
Have you looked at the costs and benefits
(Look at the health news review website for
an idea of what I will be looking for)