Transcript Number Needed to Treat

Number Needed to Treat
Alex Djuricich, MD
Indiana University School of Medicine
Department of Medicine
Ambulatory Rotation 2006-2007
Objectives
• To learn basic information used in evaluating
articles about therapy
• To learn definitions of the following and be able to
apply them to appraising articles on therapy:
•
•
•
•
•
CER (Control event rate)
EER (Experimental event rate)
RRR (Relative risk reduction)
ARR (Absolute risk reduction)
NNT (Number Needed to Treat)
Objectives
• To learn the “correlaries” to each of these
definitions for harm
• RRI (Relative risk increase)
• ARI (Absolute risk increase)
• NNH (Number needed to harm)
Hypothetical trial
• Trial of 1000 patients to compare use of
MiGone (I made that one up? Like the
name? Get it?) vs. placebo to prevent MI,
given over 5 years. Patients for study are
all healthy gym rats, coming from Bally’s
Gym.
Trial stats
Total number of patients N=1000
Control group
N=400
MiGone group
N=600
# who developed an
MI
20
18
# who developed
myalgias
4
12
For MI
• CER
• 20/400 = 0.05, or 5%
• EER
• 18/600 = 0.03, or 3%
• ARR
• 0.05 – 0.03 = 0.02, or 2%
• RRR
• (0.05 – 0.03)/0.05 = 0.02/0.05 = 40%
For MI
• NNT
• 1/ARR
• 1/0.02 = 50
• 50 patients would need to be treated to
prevent one MI in 5 years in Bally’s Gym
patients
For myalgias
• CER
• 4/400 = 0.01, or 1%
• EER
• 12/600 = 0.02, or 2%
• ARI
• 0.02 – 0.01 = 0.01, or 1%
• RRI
• 0.02 – 0.01/0.01 = 0.01/0.01 = 1, or 100%
For myalgias
• NNH
• 1/ARI
• 1/0.01 = 100
• 100 patients would need to be treated with
MiGone over 5 years to cause one case
(harm) of myalgias
Case #1
• You just completed your GI rotation, and
are now doing Wishard wards.
• You admit a 50 year old alcoholic
gentleman for abdominal pain. After
imaging, you find that he has radiologic
“cirrhosis” with evidence of portal
hypertension. He has NEVER had an upper
GI bleed
Case #1 continued
• You discuss preventive pharmacologic
treatments in rounds with the attending
• In patient with cirrhosis and portal
hypertension, does a non-selective beta
blocker prevent gastroesophageal varices?
• Yes
• No
Article
• Beta-blockers to prevent gastroeophageal
varices in patients with cirrhosis
• Groszmann RJ, et al. New Engl J Med
2005;353:2254-61.
Focus on Table #2
Total number of patients N=213
Control group
N=105
Timolol group
N=108
# who developed
varices (%)
42 (40%)
42 (39%)
# who developed
“Serious event”
(buried on p.2258)
6 (5.7%)
20 (18.5%)
Control Event Rate
• Percentage of those controls who develop
the unwanted condition
• In our example, varices or complication from it
• 42/105
• 40%
Experimental Event Rate
• Percent of those in treatment group that
developed the condition (varices)
• 42/108
• 38.9%
Absolute Risk Reduction
• The true reduction in risk between control
patients and treated patients
• Difference between the CER and the EER
• CER – EER
• 0.4 – 0.389
• 0.011, or 1.1%
• That’s ok, but not very impressive!
Relative Risk Reduction
•
•
•
•
(CER – EER)/CER
(0.4 – 0.389)/0.4
0.011/0.4
0.0275, or 2.75%
• Remember, 2.75% sounds better than 1.1%,
but it still isn’t very “good”
Number Needed to Treat (NNT)
• Meaningful way of expressing benefit of an
active treatment over a control
• Defines treatment-specific event of an
intervention
NNT
• Expression of the # of patients one would need to
treat to prevent 1 additional adverse outcome (MI)
or attain one additional benefit
• 1/ARR
• 1/0.011
• 91 (that is, 91 patients would need to be treated
with timolol to prevent one additional case of
varices)
• Impressive? Not impressive?
NNT
• Function of 4 elements
•
•
•
•
Condition
Intervention
Events being prevented
Duration of follow-up
• Remember “DICE”
Cirrhosis pts
Timolol vs. placebo
Varices
55 months
NNT
• Tells us in more concrete terms how much
effort clinicians must expend to prevent one
event, thus allowing comparisons with the
amounts of efforts that must be expended to
prevent the same or other events in patients
with disorders
NNT
• Is a “good” NNT a low or a high number?
• The lower the NNT, the larger the
magnitude of treatment effect
What about the “serious adverse
events”?
• Need to remember that medications have
side effects, and that one can calculate how
often we “harm” someone by giving a
medication designed to help.
• This “harm” is the side effect
CER
• For harm, the CER is still the same as
before.
• However, it applies to the harm instead of
the “good” effect
• In the example, it refers to “serious adverse
events” (such as bradycardia)
EER
• This is really the same as before, except it
applies to the rate for harm instead of for
benefit
• In our example, the EER is for “serious adverse
events”
ARI for “Harm”
• Absolute Risk Increase
• EER – CER
• Note NOT “CER – EER”
RRI
• Relative Risk Increase
• (EER – CER)/CER
• Remember, its always the Control that goes
in the denominator
What about NNH?
• In general, a decision to use therapy
involves risk vs. benefit
• Most therapies have side effects
• Same concepts for NNT apply to side
effects. We call that number the NNH
(number needed to harm)
NNH
• The number of patients who, if they
received the experimental treatment, would
lead to one additional patient being
“harmed”, compared with patients who
received the comparison treatment.
• 1/ARI
Table #2 Again
Total number of patients N=213
Control group
N=105
Timolol group
N=108
# who developed
varices (%)
42 (40%)
42 (39%)
# who developed
“Serious event”
(buried on p.2258)
6 (5.7%)
20 (18.5%)
Let’s calculate for “serious adverse
events”
• CER
• 6/105 (5.7%)
• EER
• 20/108 (18.5%)
For “adverse events”
• ARI
• 18.5% - 5.7% = 12.8%, or 0.128
• RRI
• (18.5% - 5.7%) / 5.7%= 224.5%, or 2.245
• NNH
• 1/ARI
• 1/0.128
• 7.8 (rounds to 8) patients need to be treated with
timolol to cause one case of “serious adverse events”
• This is unfortunately not a very good number; we want
a very high NNH (as opposed to NNT)
Case Take Home
• According to this one study, the numbers
are not impressive enough to recommend
timolol to PREVENT varices in patients
with diagnosed cirrhosis.
• Remember that this is different from
patients who have ALREADY had an upper
GI bleed from varices secondary to
cirrhosis. Those patients should get a beta
blocker.
Questions?
Case #2
• You admit a patient on University Wards at night
for bacterial meningitis. The ER doctor calls you
with results of the CSF showing a WBC count of
1024, after noting cloudy CSF from the tap. She
wants to know which antibiotic you would like to
use; need an answer in 1 minute; meningitis is a
true medical emergency!!
• You recall from your pediatrics rotation about
giving steroids in addition to antibiotics.
• In patients with bacterial meningitis, does the use
of concomitant dexamethasone improve
neurological outcome?
Example NEJM paper
• Dexamethasone in adults with bacterial
meningitis
• deGans J, vandeBeek D, et al. New Engl J
Med 2002;347:1549-56.
• Total patients: 301 patients
• 157 in dexamethasone (experimental) group
• 144 in placebo (control) group
NNT: Function of 4 elements
• 1. Condition
• meningitis
• 2. Intervention
• dexamethasone vs. placebo. Keep in mind that other
treatments (antibiotics) were given to both groups, so
placebo isn’t really “nothing”
• 3. Events being prevented
• unfavorable outcome (they define it), death
• 4. Duration of follow-up
• 8 weeks
Dexamethasone paper
Total # of patients N=301
Control group
N=144
Dexamethasone
group N=157
Unfavorable
outcome
36
23
Death
21
11
Dexamethasone paper
• CER for unfavorable outcome
• 36/144
• 0.25, or 25%
• CER for death
• 21/144
• 0.25, or 15%
Dexamethasone paper
• EER for unfavorable outcome
• 23/157
• 0.15, or 15%
• EER for death
• 11/157
• 0.07, or 7%
Dexamethasone paper
• ARR for unfavorable outcome
• CER – EER
• 0.25 – 0.15
• 0.10, or 10%
• ARR for death
• CER – EER
• 0.15 – 0.07
• 0.08, or 8%
Dexamethasone paper
• NNT for unfavorable outcome
• 1/ARR
• 1/0.10
• 10. You would need to treat 10 meningitis patients with
dexamethasone to prevent one patient getting an
“unfavorable outcome” after 8 weeks
• NNT for death
• 1/ARR
• 1/0.08
• 12.5 (rounds up to 13). You would need to treat 13
meningitis patients with dexamethasone to prevent one
death after 8 weeks
Dexamethasone
• How about NNH?
• Calculate it for each of the “side effects”
you are interested in.
Dexamethasone paper
Total # of patients
Control group
N=144
Dexamethasone
group N=157
Hyperglycemia
37
50
Fungal infection
4
8
Dexamethasone paper
• CER for hyperglycemia
• 37/144
• 0.26, or 26%
• CER for fungal infection
• 4/144
• 0.03, or 3%
Dexamethasone paper
• EER for hyperglycemia
• 50/157
• 0.32, or 32%
• EER for fungal infection
• 8/157
• 0.05, or 5%
Dexamethasone paper
• ARI for hyperglycemia
• EER – CER
• 0.32 – 0.26
• 0.06, or 6%
• ARI for fungal infection
• EER – CER
• 0.05 – 0.03
• 0.02, or 2%
Dexamethasone paper
• NNH for hyperglycemia
• 1/ARI for hyperglycemia
• 1/0.06
• 16.7 (rounds to 17)
• You would need to treat 17 patients with
bacterial meningitis to cause one case of
hyperglycemia over 8 weeks
Dexamethasone paper
• NNH for fungal infection
• 1/ARI for fungal infection
• 1/0.02
• 50
• You would need to treat 50 patients with
meningitis with dexamethasone to cause
one case of fungal infection over 8 weeks
Case Take Home
• Consider steroids at the time of diagnosis of
suspected meningitis.
• They clearly have a role in improving
outcomes, but this comes at a cost of
increased fungal infection
• Is it worth the risk? That is the “art” of
medicine.
Summary
•
•
•
•
•
•
CER (control event rate)
EER (experimental event rate)
ARR (absolute risk reduction) and ARI
RRR (relative risk reduction) and RRI
NNT (number needed to treat)
NNH (number needed to harm)
Summary
• NNT: Expression of # of patients one would need
to treat to prevent one additional adverse outcome
or attain one additional benefit
• NNT is a function of:
•
•
•
•
•
Condition
Intervention
Events being prevented
Duration of follow-up
Remember, “DICE”