Non Inferiority Trial
Feb 25, 2016
Non Inferiority Trial
0% 100%
experiment - drug control - placebo
event rate
superiority study
0% 100%event rate
0ARR=EER-CER
-∞ +∞
experiment better experiment worse
experiment - drug control - placebosuperiority study
EER CER
0% 100%
experiment - drug B control - drug Anon inferiority study
presumption that experiment may be worse than control! But does it matter?
is it clinically significant?
a positive non inferiority trial result supports the statement that
the experimental treatment is no worse than the standard
treatment✱
✱within a predetermined acceptable margin of equivalence
margin of equivalence“how much worse than the
active control you’re willing to accept”
0% 100%
ok if not worse than X%
experiment - drug B control - drug Anon inferiority study
0% 100%
X%
experiment - drug B control - drug Anon inferiority study
0-∞ +∞
experiment better experiment worse
if upper CI is less than X, non inferior!
non inferior inconclusive
the margin of equivalence is determined by
•what is deemed to be a clinically significant difference,
•the expected event rates,
•regulatory requirements,
•the known effect of the standard treatment over placebo,
•severity of the disease,
•toxicity,
•inconvenience, or cost of the standard treatment, adherence and
•the primary endpoint.
Superiority Trial
•type I error is falsely finding a treatment effect when there is none
•type II error is failing to detect a treatment effect when truly one exists.
Non inferiority Trial
• the null and alternative hypotheses are reversed;
•type I error is the erroneous acceptance of an inferior new treatment,
•type II error is the erroneous ejection of a truly non inferior treatment
0% 100%3.3%
7%
experiment - No Abxcontrol - Abx
0-∞ +∞
experiment better experiment worsenon inferior inconclusi
ve
treatment failure at day 10
0% 100%
7%
experiment - No Abxcontrol - Abx
0-∞ +∞
experiment better experiment worse
4.1% 5.3%
1.2% 6.8%
non inferior inconclusive
treatment failure at day 10
0% 100%
7%
experiment - No Abxcontrol - Abx
0-∞ +∞
experiment better experiment worse
12.9% 26.4%
13.5%24.3%
non inferior inconclusive
new lesions at day 10
want to prove
equivalence
thresholdnull
hypothesistype I error
type II error
ITT analysis
commonality
superiority trial
experiment is superior to placebo or other
treatment
not applicable
experiment and placebo are equally effective
falsely finding
treatment effect when
there is none
failing to detect a
treatment effect when there is one
good very common
non inferiority
trial
experiment is non
inferior to active
control (no placebo)
acceptable “worst”
treatment effect
experiment is inferior to
active control
falsely finding non inferiority when the
experiment is inferior
falsely rejecting
the experiment when it is
non inferior
may increase
type I errorincreasing
equivalence trial
experiment is equally
effective to active
control (no placebo)
acceptable “worst” and
“best” treatment
effect
experiment and active control are not equally effective
falsely finding
equivalence
falsely rejecting
equivalence
may increase
type I errorrare