Transcript Statistical Evaluation of Suicidality in Adults Treated with

Acceptability of Foreign Data by
the US
Russell Katz, M.D.
Director
Division of Neurology Products
Center for Drug Evaluation and Research
Version: 27 May 2007
Outline
1. Regulations and Guidance
2. What are we worried about?
3. How often do they really differ?
4. Current status
2
History of Interest
1.
2.
3.
4.
5.
Regulations have permitted reliance on foreign data
since 1985 (in part or completely)
Most of our experience (prior to last 10-15 years) has
been with Canada, Western Europe, Australia
Now, many more sites in Latin America, Eastern
Europe, Asia
We are gaining experience with inspections in these
new areas, but are still grappling with differences in
cultures and medical practice
Let’s remember that the US is anything but
homogeneous!
3
Regulations
FDA clearly can legally rely on non-US data as a basis for approval.
21 CFR 314.106 (since 1985): Foreign Data
When will FDA rely solely on foreign data to support marketing?
• When the data are applicable to US
• Studies by investigators of recognized competence
• We can inspect or don’t need to
Again, in the past these situations mostly involved data from
familiar places, mainly Western Europe.
4
Regulations (cont)
21 CFR 312.120 (revised 2008)
FDA will accept a foreign study not conducted under an IND as
support for an IND or NDA if it was “well-designed and conducted,”
and
• Study conducted in accordance with GCP, including initial
and continuing review by an IEC, proper informed consent.
• FDA can validate with onsite inspection if felt necessary
• Supporting information provided to show how GCP was
assured
• In addition, we’d expect all the very complete documentation
we usually get (data listings, CRF’s, etc)
5
ICH E-5 (1998)
Ethnic Factors in the Acceptability
of Foreign Clinical Data
The guidance reflected awareness of what could lead results to differ
among regions, but we had little real evidence that they did .
There are potential differences to worry about:
− Genetic differences (metabolism, risk of AEs)
− Differences in disease severity, frequency and outcome (heart
attack vs stroke, rate of heart failure) in different regions
− Differences in concomitant illness (growing diabetes and obesity
in US; smoking rates), but all regions have all at least to some
extent
− Different clinical practices (rate of post-infarction percutaneous
intervention; rate of CABG, different use of coated and uncoated
stents)
− Differences in study conduct and follow up
But not so many illustrations of documented consequences (wellestablished different conclusions about a drug by region).
6
ICH E-5 (cont)
Ultimately, what could be described as collective “freefloating anxiety” was resolved as follows:
A country faced with an entirely acceptable (study design,
population, etc) database derived from extra-regional
studies could resolve its concern about possible ethnic
differences by asking for:
• One adequate and well-controlled “bridging” study in
its region (often a D/R study); could be anything
needed: clinical trial; clinical pharmacologic (PD)
study showing effect on pertinent biomarker, e.g.,
physiologic effect. What was needed could depend
on, e.g., prior experience with the drug class.
• Additional local safety data (reflecting concern that
attitudes toward ADR’s and reporting practices could
differ by region).
7
ICH E-5 (cont)
A Q & A addition to E-5 (2006) described use of a “global”
development approach that would assure substantial
representation of all regions in multi-regional trials as distinct
from doing a separate study in the new region, thus anticipating
(as opposed to adding after the fact) need for regional data (now
very common). It emphasizes careful attention to:
•
•
•
•
•
•
Definition of the condition and description of patients
Choice of control
Efficacy variables
Safety assessments
Medical practice, concomitant meds
Dose, regimens
The Q & A, like E-5, expresses the hope/expectation that as
experience grows, need for bridging studies will decrease.
Certainly, we are historically less concerned about trials in
Canada, Western Europe, Australia, but with more experience
elsewhere, we should get more comfortable.
8
Future
Experience matters
ICH E-5 suggests, for a global (multi-regional) development
effort, that the most persuasive data is a “stand alone”
regional result together with an overall study result. This, of
course, would require huge regional sub-populations, and
consistent trends, similar D/R, history with drugs in the
same class could provide support based on a less strong
result.
Progress will, however, depend on experience and
understanding of what could lead to different results in
regions or subsets of the overall population. It will also be
critical to analyze our inspection experience, growing
rapidly but still a considerable challenge.
9
Foreign Clinical Investigator
Inspections
120
100
80
60
40
20
0
(CDER, FY 1997 – 2009; All
Indications)
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
10
Clinical Investigator Deficiencies
CDER Inspections - FY 2008
60%
50%
40%
30%
20%
10%
Foreign
Domestic
0%
NAI
Protocol Records
*Based on Letter Issued Date
Drug
Acct
Consent
AEs
Foreign n = 72*
Domestic n = 345*
11
Examples – Why We Worry
1. Antidepressant
We saw an antidepressant with clearly successful
results in Latin America and Eastern Europe. But trials
in Western Europe and US failed and in two cases a
positive control succeeded while the drug failed, very
unusual for an effective drug.
So far no explanation, but plainly a source of concern.
12
Examples – Why We Worry
2. Toprol XL
MERIT-HF (4000 patient, class II-IV CHF) trial, 25% US, showed clear
overall effect on death plus hospitalization and CV mortality.
Toprol Pla
N
1990
2001
Death & Hosp
641
767
RR
p
0.81
0.0001
Death
145
217
0.66
0.0001
Death & CHF
Hosp
311
439
0.69
<0.001
In US, about 25% of population, there was a similar effect on hospitalization,
but no effect at all (not even a “lean”) on mortality or CV mortality, despite
about 100 events. Remains unexplained.
Drug was approved: labeling describes the disparate results but notes
uncertainty as to meaning.
13
Examples – Why We Worry
3. BiDil
BiDil, combination of isosorbide dinitrate and hydralazine was
dramatically effective on hospitalization and survival in selfidentified blacks (a > 40% mortality reduction) but had little or no
effect in others. We have no idea of the reason for this. Could it
signal potential ethnic differences in treatment of CHF (genetic,
diet, other therapy)?
4. Losartan
LIFE (losartan vs atenolol) showed, overall, reduction in stroke on
losartan in 9193 patients, but no effect (significantly adverse) in
blacks, a small subset.
Losartan
Atenolol
RR
P
Stroke (F/NF)
232 (5%)
309 (7%)
25%
0.001
Race
Black (n=533)
White (n=8503)
9%
5%
5%
7%
Significant
Significant
14
Examples-Why We Worry
Labeling notes this result but also notes
problems with subset analyses
Labeling then states:
However, the LIFE study provides no evidence
that the benefits of Cozaar on reducing the risk
of cardiovascular events in hypertensive
patients with left ventricular hypertrophy apply
to Black patients.
15
Examples – Why We Worry
5. Losartan (RENAAL)
Protection against nephropathy in type II diabetes. Primary endpoint first
occurrence creatinine doubling, ESRD, or death; but almost all first events
were creatinine doubling. Also collected longer term data on ESRD. Initial
showed all effect in “Asian” subset but “Asian” included Israel, Australia.
Even defined ethnically, early result (mostly creatinine doubling) was “Asian,”
but ESRD showed much less disparity, suggesting differences were random.
Group
N
Overall Primary
ESRD
HR (CI)
HR (CI)
Overall
1513
0.839 (0.721, 0.977)
0.714 (0.576-0.885)
Race
Asian
Black
Hispanic
White
252
230
277
735
0.655 (0.453-0.947)
0.983 (0.647-1.495)
1.003 (0.728-1.380)
0.809 (0.645-1/013)
0.625 (0.367-1.066)
0.831 (0.456-1.516)
1.024 (0.661-1.586)
0.596 (0.427-0.831)
Similar result in irbesartan study (IDNT). Hard to explain, but the
differences were almost certainly spurious.
16
Other examples
• Dimebon
– A 6 month study in patients with Alzheimer’s
Disease done in Russia demonstrated
remarkable effects
– We recommended the sponsor repeat this
study ASAP
– Next 6 month multi-center (North and South
America, Europe) study was negative
17
Other examples
• Lamictal XR for partial seizures
– Overall clearly positive study, but no effect in
the US
• Extensively analyzed
– The difference was felt to be related to a
differential effect in the Titration Phase,
related to titration regimen and resulting
higher lamotrigine levels in the non-US sites
(more VPA use outside the US)
18
Adverse events in a purely foreign
study
• Adverse Event
• Nausea
• Dry Mouth
• Paresthesia
Low dose
High dose
Placebo
(n=527)
%
(n=1313)
%
(n=386)
%
1
1
1
2
1
1
1
0.5
0.5
• This is the entire controlled trial AE table
19
Carbamazepine, HLA-B*1502, and
SJS
• About a 10 fold increase in risk of SJS with
carbamazepine in Han Chinese compared to
Caucasians, and strong association between
this reaction and HLA-B*1502
• Prevalence of the allele varies across various
groups identified as Asian (>15% in Hong Kong,
Thailand; 10% in Taiwan; 2-4% in India; <1% in
Japan and Korea)
• Almost non-existent in Caucasians
• Not entirely clear how to describe who is at risk
20
Suicidality
• Although there was a clear signal of
increased suicidality in our meta-analysis
of AED controlled trials, there was a
marked difference in the OR between US
(<2) and foreign (>4) studies
• The presence of a a signal in the US,
though, made this less of a problem
21
Other Examples
The Division of Psychiatric Products,
particularly Dr. Ni Khin, has examined US
and non-US experience in depression and
schizophrenia, showing drug and placebo
changes from baseline and drug effect
(comparison of changes from baseline).
22
Results in Depression
As you can see, US and
OUS look quite similar in
effect size (mean about 2.5
HamD points).
Total 86 trials
71 US: 19,600 patients
15 Foreign: 4,100
patients
23
Results in Depression
Placebo responses and treatment effects were similar over time.
24
Results in Depression
Even study success rates were fairly similar, at least since
1995, in the well recognized neighborhood of 50%. Specifically
OUS was not “the place to go to win.”
25
Results in Schizophrenia
There were fewer trials (23 US, 10 mixed, 4
OUS), with 13,500 patients, but very similar
results, with similar mean PANSS total,
about a 5 point improvement vs placebo, an
increasing placebo response over time in
both, but with perhaps an increasing drug
effect over time in US but not OUS. Maybe
somewhat larger effects in Asia and E
Europe, but too soon to tell.
26
Concerns
The matters that, overall, probably worry us most are not,
except in rare cases, genetic differences, even metabolic
differences, as these are readily assessed, but differences in
practice and assessment, especially differences that are not
recognized.
1. Do observers report all ADEs, especially serious ones, even
if they think the drug probably did not cause them.
2. Might they use unmentioned drugs, such as those in the
same class being studied or that might otherwise interfere with
the study.
3. Do they have the same standards for hospitalization, surgical
procedures, stopping treatment, etc, all of which could affect
retention and even outcomes. Is the “desire to please” different?
27
Concerns
4. Could illnesses or illness severity be defined differently, making
populations different in unknown ways.
5. Could attitude toward blinding differ, a potentially very critical issue.
6. Note that most of these concerns (except for blind-breaking) would create
noise, decreasing the likelihood of showing a drug effect.
7. What do we know about the “translatability” of the measures.
8. In a historical control setting (quite unusual), what if the historical control
data and the study data were obtained in different regions.
28
Concerns
9. We are moving, to a degree at least, from foreign but
familiar (W Europe, Canada, Australia, Israel) to less familiar
(E Europe, Russia, Japan, Taiwan, Latin America, itself very
varied) to very unfamiliar (China, India).
A very common way sponsors are dealing with this is multinational trials, a matter addressed in 2006 ICH E-5 Q and
A’s, which emphasized:
Make the region component large enough to 1) show
effectiveness (recognizing the challenge of this) and 2)
allow comparisons among regions to show lack of
sensitivity to ethnic factors. This does not mean the
region needs a significant stand-alone effect (although
that wouldn’t hurt).
29
Summary
• We always examine the effects in the US
vs non-US
• In most cases (even in the “newer” areas)
there appear to be no systematic
differences
• In those cases where there are
differences, we examine the data to
explain the differences
30
Summary
• In those cases where (prospectively) we
think regional differences may matter, we
ask the sponsor to enroll an “appropriate”
number of patients from the US
• Even when differences are seen, we
usually cannot establish a clear reason
why, and we are likely to “live with” these
differences (but not always!)
31