| Comment Record |
|
Commentor |
Dr. Janet Wittes |
Date/Time |
2002-01-13 20:34:16 |
|
Organization |
Statistics Collaborative |
|
Category |
Company |
| Comments for FDA General |
| Questions |
|
1. General Comments
|
I think in general the guidance is excellent. It will go a long way to improve the way things run.In fact, I have already seen a difference in the way sponsors are thinking of DMCs. Congratulations!
I do have a few comments.
a) Section 3.2, line 2. I have never seen a Steering Committee write the
final study report. They often write the paper describing the results, but
in my experience the sponsor writes the study report for the FDA. And that
is what most companies refer to as the final study report or the clinical
study report.
b) Section 4.1, last line. I think it is important that everyone make a firm
commitment to participate until the end of the trial. I wouldn't limit this
to the Chair.
c) Section 4.2.2. I think you should mention the executive session here. You
refer to it in section 4.3.1.2, but it is sort of lost.
d) Section 4.3.1.1, par 1. I would delete the parenthetical (perhaps
most).
e) Section 4.3.1.1, par 2. I find the section too strongly in favor of
face-to-face meetings. This guidance is going to increase the number of DMCs
considerably and travel is difficult. I would soften this recommendation. In
many settings telephone or video meetings work very well. For example, in a
small study looking at very specific safety data, a telephone conference
call is fine. I think you should leave more room for committees to decide
how they are going to operate.
This is one area in which I have already seen a difference. In a Phase I trial that is setting up a DMC, the sponsor wanted biweekly face-to-face meetings of the committee because of this section. The committee said that was impossible.
f) Section 4.3.1.4. I would add and on the final analytic plan to the end
of the paragraph. I have seen DMC's influence the final analysis plan.
(Think of HA1A).
g) Section 4.4.2, last sentence. I don't know what you are referring to
here. What situations will be infrequent? Is this sentence necessary?
i) Section 6, line 1. independent defines independent. I have a question
about the restriction against having been involved in the design of the
trial. I could imagine a situation, especially for a small company or for an
early phase trial, where someone has served as a consultant designing the
trial. Then, when the various roles are divided, that person sits on the
DMC. Why is that necessarily a problem? (I actually don't know of any
specific case, but it doesn't seem problematic to me.)
Other issues--
I would like to see the guidance unambiguous about the desirability of the
DMC's having access to unblinded data. The language of section
4.3.1.4 pretty clearly argues for unblinded data. In the public meeting, however, some FDA staff seemed to argue for blinded data. I believe
that would be a mistake. Again, many members of the FDA tell sponsors that the DMC should see blinded data unless something requires unblinding. I believe this is a mistake and the guidance should be unambiguous.
The big issue that is bothering me is the one that Joe Costantino brought up
at the session. He and I did a webcast for FDANews discussing the
guidance. The session was very lively. The main issue that we struggled with was how
to ensure that the reporting statistician understands the data well enough
to present an intelligent report to the DMC. Joe's position is that only the
statistician working day-to-day with the data can understand the study well
enough to present the data to the committee. My view, on the other hand, is that if the
charge of the committee is clearly focussed, a sensible, responsible
outsider can do the job very effectively.
Nonetheless, Joe is correct that many people present data to the DMC without
understanding much of what is going on. I have been appalled by the kinds of
reports I have seen when I have served on DMCs for industry. I no longer
will sit on them because I find the general quality of the reports shoddy
and incomplete. I have been in situations where the DMC does
not have sufficient access to data necessary for intelligent
decision-making. So, there is a trade-off. It may be that you want to stress
the importance not only of protecting the integrity of the study (i.e.,
ensuring that knowledge of ongoing data does not influence the conduct of
the study) but also of ensuring high-quality presentations to the DMC. I
think the current draft emphasizes the former but gives short shrift to the
latter.
Also, I waffle about the difference between industry and government
sponsored trials. Sometimes, I agree with the spirit of the guidance and
feel that both should run under the same model, but sometimes I think that
government sponsors, as custodians of the public monies, cannot be excluded
from the discussions at the DMC. I have been a member of DMCs for many of
the institutes (NHLBI, NIDDK, NCI, NIAID, NEI) and they all run slightly
differently. In all cases except the NCI trials, the NIH project officers
have sat through the executive sessions. I don't think that is wrong or that
it jeopardizes the integrity of the trial. In principle, yes; in practice,
generally, no. Perhaps you should allow more latitude for NIH-sponsored
trials, especially the very large, multi-center trials that are asking big
public-health questions.
As I write this, I suddenly think I understand the problem. Industry DMCs
tend to have quite limited scope. They are looking at very specific
questions and are not interested in examining all the data from an ongoing
trial. Therefore, an independent statistician can get to know the relevant
data well enough to produce a useful report that allows a DMC to make
intelligent decisions.
NIH DMCs, on the other hand, tend to look at huge books of data. They will
review every variable collected because they have a very broad mandate for
the trial. They serve not only as safety monitors, but as general advisors
to the trial. There is no way that an independent statistician, not involved
in the trial, can produce the kind of monitoring report that a typical NIH
DMC sees.
The guidance does not appear to distinguish these two types of beasts.
|
|