Title: Joel Singer, Programme Head, Methodology and Statistics, CIHR Canadian HIV Trials Network
1Joel Singer, Programme Head, Methodology and
Statistics,CIHR Canadian HIV Trials Network
2What is a Non-inferiority Trial?
- A trial to show that a particular therapeutic
agent or policy is no worse than some other
standard.
3What is a Non-inferiority Trial?
- Type 1 Both agents have been shown superior to
placebo/no treatment - Type 2 Only one agent has been shown superior to
placebo/no treatment
4What is a Non-inferiority Trial?
- The standard (STD) should have clinical efficacy
(i) that is of substantial magnitude (ii) is
precisely estimated (iii) estimates that are
relevant to the setting in which the trial is
being conducted.
5ICH Guidelines
- A suitable active comparator could be a widely
used therapy whose efficacy in the relevant
indication has been clearly established - and
6ICH Constancy Assumption
- quantified in well-designed and well documented
superiority trials and which can be reliably
expected to have similar efficacy in the
contemplated active control trial.
7Example of non-Constancy
- Suppose vancomycin, has been shown to provide a
large improvement, compared to no treatment, in
cure rate in patients with various infections.
8Example of non-Constancy
- Suppose a high prevalence of vancomycin-resistant
enterococci (VRE) has developed. NI trial
comparing an experimental antibiotic (EXP) with
vancomycin is conducted where VRE is prevalent
9Example of non-Constancy
- A biased estimate of the efficacy of EXP would be
obtained if the constancy assumption is falsely
presumed to hold in the NI study.
10What is a Non-inferiority Trial?
- Because proof of exact equality is impossible, a
prestated margin of noninferiority for the
treatment effect in a primary patient outcome is
defined.
11What is a Non-inferiority Trial?
- Non-inferiority trials are intended to
- show whether a new treatment has at least as
much efficacy as the standard or is worse by an
amount less than ?,often on the premise that it
has some other advantage
12What is a Non-inferiority Trial?
- Both participants and outcome measures in a
noninferiority or equivalence trial should be
similar to those in trial(s) that established the
efficacy of the reference treatment.
13What is a Non-inferiority Trial?
- The required sample size is calculated using the
confidence interval (CI) approach, considering
where the CI for the treatment effect lies with
respect to both the margin of noninferiority and
a null effect.
14What is a Non-inferiority Trial?
- Sample size depends on the level of confidence
chosen, the risk of type II error (or desired
power), and ?.
15What is a Non-inferiority Trial?
- A prestated margin of noninferiority can be
specified as a difference in means or proportions
or the logarithm of an odds ratio, risk ratio, or
hazard ratio.
16What is a Non-inferiority Trial?
- The meaning and clinical importance of measures
such as relative risk, odds ratio and relative
hazard will be dependent on the base rate of
outcome
17What is a Non-inferiority Trial?
- A prestated margin of noninferiority is often
chosen as the smallest value that would be a
clinically important effect.
18What is a Non-inferiority Trial?
- The required size of non-inferiority trials is
therefore usually larger than that for
superiority trials, - But
- The actual calculation is the same as in
superiority trials
19What is a Non-inferiority Trial?
- One should avoid features that might dilute true
differences between treatments, thereby enhancing
the risk of erroneously concluding non-inferiority
20What is a Non-inferiority Trial?
- For superiority trials, intention-to-treat (ITT)
analysis (analyzing all patients within their
randomized groups, regardless of whether they
completed allocated treatment) is recommended
21What is a Non-inferiority Trial?
- In non-inferiority trials, ITT analysis will
often increase the risk of falsely claiming
non-inferiority (type I error), although not
always
22What is a Non-inferiority Trial?
- In non-inferiority and equivalence trials,
non-ITT analyses might be desirable as a
protection from ITTs increase of type I error
risk (falsely concluding non-inferiority).
23What is a Non-inferiority Trial?
- Interpreting a non-inferiority trials results
depends on where the CI for the treatment effect
lies relative to both the margin of
non-inferiority and a null effect. The observed
treatment effect is not sufficiently informative.
24What is a Non-inferiority Trial?
- For superiority trials, intention-to-treat (ITT)
analysis (analyzing all patients within their
randomized groups, regardless of whether they
completed allocated treatment) is recommended
25Major Methodological Issues for
Equivalence/Non-inferiority Trials
- 1. Difficult to get consensus on difference
considered equivalence or non-inferior. - 2. In superiority trials, lack of rigor in the
conduct of a trial will tend to lead to
equivalence between the comparators and will
cause a conservative bias.
26- 3. Lack of rigor or anything which
causes background noise in an
equivalence/noninferiority trial
will bias in favour of the
hypothesis of interest
(no difference).
27Important Variables for Defining
Equivalence/non-inferiority
- 1. Outcome.
- 2. Prevalence of the disease.
- 3. Other cost, invasiveness, ease of use.
28Trial Design
Nucleoside backbone d4T and 3TC
Inclusion criteria pVL gt 5000 copies/mL any
CD4 cell count any stage of CDC-classification
29Outcome Measures
- of patients with treatment failure at week 48
- less than 1log10 decline in pVL in first 12 weeks
- 2 consecutive pVL gt 50 copies/mL from week 24
onwards - new CDC-C event or death
- change of allocated treatment
30Methods
- All analyses Intention-to-Treat (unless stated
otherwise) - all randomised patients
- pVL data missingfailure
- Four pre-defined pairwise comparisons
- NVP-bd vs EFV (primary)
- NVP-od vs NVP-bd
- NVP-od vs NVPEFV
- EFV vs NVPEFV
31- Given that the 2NN trial was conducted as a
non-inferiority trial, how does one go about
interpreting the results? - The goal of an inferiority trial is to be able
to rule out differences greater than the
minimally important clinical difference (MICD).
32- Translated into a statistical statement, this
means we can say with confidence (based on the
95 confidence interval) that the true
difference is unlikely to exceed the difference
deemed clinically important.
33Methodological Concerns With the 2NN Trial
- 1. Choice of definition of outcome.
- 2. Background noise.
34Concerns in 2NN Study re non-inferiority
- 1. patients who never received treatment (20 in
EFV, 10 in NVP-BD). - 2. early compliance failures.
- 3. Question inclusion of changes in d4T or 3TC as
evidence of failure.
35What Were the Primary Results of the 2NN Trial?
36Treatment Success and Failure
of patients
Success only significant difference EFV vs
NVPEFV, plt 0.001
37- There was no statistical difference for the
primary comparison (NVP vs EFV) so one cannot say
definitively that one treatment is better than
the other.
38- The 95 confidence interval (1 in favour of
nevirapine to 13 in favour of efavirenz) does
not allow one to rule out a clinically important
difference (at least as defined by the
investigators) on the primary efficacy variable.
39Conclusions
- NVP and EFV have comparable potency in
suppressing HIV-1 replication - NVP-od and NVP-bd show comparable efficacy
- Co-administration of NVP and EFV results in
higher treatment failure due to increased
toxicity
40Comments
- Conclusion slide does not address primary outcome
(combination of disease progression, virologic
failure and change of therapy) - Does not address issue of proof of
non-inferiority
41Some Technical Sample Size issues
- Typically, one wants the sample size to be large
enough so that the bounds of the confidence
interval will exclude values outside the
non-inferiority range 80-90 of the time
42Some Technical Sample Size issues
- If the outcome is continuous, and one assumes
homogeneity of variance, then the sample size
required for non-inferiority under the null
hypothesis is the same as the sample size to
detect the indicated difference under the
alternative hypothesis
43Some Technical Sample Size issues
- If the outcome is dichotomous, then the sample
size required for non-inferiority under the null
hypothesis is not the same as the sample size to
detect the indicated difference under the
alternative hypothesis because the variability is
a function of p1 and p2
44Some Technical Sample Size issues
- In a non-inferiority trial, the hypotheses are
inverted. - The null hypothesis is that there is a difference
gt m, and the alternative is the null hypothesis
45Some Technical Sample Size issues
- Generally, the clinical world is concerned about
absolute differences between treatments, but
depending on the statistical model being used,
the null and alternative may be stated in other
terms eg hazard ratios
46Some Technical Sample Size issues
- The choice of what is considered non-inferior in
terms of a hazard ratio is typically driven by
what one assumes the baseline rates are - Eg if the hazard rates are relatively low, a
somewhat large hazard rate may translate into a
relatively low absolute difference
47Summary of 2NN Study
48- 1. Demonstrated a non-significant 6 advantage in
the primary outcome in favour of efavirenz over
nevirapine bid on intent-to-treat analysis. - 2. Confidence interval around obtained difference
between efavirenz and nevirapine bid include
confidence bounds to 13 in favour of efavirenz.
Cannot exclude non-inferiority on primary outcome
set out by study criteria.