Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

EMC Harvard School of Public Health Vol 3 20 by FDADocs


									February 17, 2002

Docket Management Branch (HFA-305)
Food and Drug Administration
5630 Fishers Lane, Room 1061
Rockville, MD 20852

RE: FDA Draft Guidance for Clinical Trial Sponsors –
     “On the Establishment and Operation of Clinical Trial Data Monitoring
         Committees” (

     Docket No. 01D-0489, CBER 200139 (November 5, 2001)

Dear Sir or Madam,

We are writing to provide comment and express concern about the GCTS document, and
particularly with issues raised in the statement: “For all these reasons, the integrity of the
trial is best protected when the statistician preparing unblinded data for the DMC is
external to the sponsor, especially for critical studies intended to provide definitive
evidence of effectiveness. In any case, the statistician should have no responsibility for
the management of the trial and should have minimal contact with those who have such

One concern regards the definition of the word “sponsor.” In some cases, such as a trial
financed and conducted by industry, the sponsor is clearly defined. In other cases, such as
NIH-funded trials in which industry partners may provide drug or other forms of
ancillary support, it is not clear whether the sponsor refers to the companies involved, the
NIH itself, or all NIH-funded statistical and data management centers. Furthermore,
would academic centers or consulting companies that hold NIH grants and contracts or
contracts from the industry partners in question be considered “external” to the sponsor?
A clear definition of sponsor would be required for implementation of the proposed

More broadly, we have concerns about the impact of the policy on clinical research. The
goal of the proposed policy is one we wholeheartedly support, i.e. the need for
statisticians to provide monitoring boards with valid, confidential interim monitoring
reports, free of influence by proprietary or other interests. However, the solution
proposed by the document is not the only workable approach, and may do more harm
than good in many settings. In studies dealing with complex scientific issues and using

nonstandard designs, analyses, or endpoints, only statisticians intimately involved in the
study may have sufficient understanding to prepare a report that best reflects the data and
their limitations, and thus that ultimately serves the best interests of the patients. For
example, in AIDS research, a deep knowledge of the subject matter is often required to
compile toxicity reports. To fully characterize new toxicities that arise during the course
of the study, e.g. delayed hypersensitivity or peripheral renal tubular disease, statisticians
may need to use their understanding of the biology and the study data collection system
to compile information from many sources. This cannot always be done in a simple pre-
specified algorithm; informed judgment is required for interpretation of clinical signs and
symptoms, laboratory measures; in some instances, physician review of case reports
(blinded to therapy) might be required. Only a statistician who has been involved from
the earliest moment that a concern arises in a study is likely to understand how to prepare
a useful report. Similar concerns arise for efficacy analyses. While endpoints, stopping
rules and analysis plans should be prespecified, that does not preclude the need for
judgment that depends on a deep understanding of a study. This is particularly true when
ancillary analyses need to be identified and completed when there is any possibility of a
study being terminated early.

Choosing a statistician external to an organization to prepare a report may provide the
appearance, but not the certainty, of independence. As the GCTS correctly points out, the
use of monitoring boards for clinical trials is increasing rapidly. The need for external
statisticians would grow as well if the recommendations were implemented. A whole
industry of external statisticians used to produce DSMB reports would have to be
developed; these statisticians will thrive only by having close relationships to the drug
industry. While a good organization can assure the quality and integrity of its staff
through proper diligence, it may be harder to do so when employing external contractors.
Furthermore, the policy that encourages the use of such contractors presumes that an
adequate supply of highly qualified statisticians exists to staff them, and that that
contractors, which usually represent for-profit companies, are themselves free of any
conflicts of interest. This presumption may not be correct. In many settings, the goals of
the GCTS may be better served by nurturing truly independent and high quality statistical
groups within companies and other groups conducting studies that appear before the FDA
than by relying on external statisticians.

Although there is a great deal of concern raised in the document about the possibility of
pressure on statisticians to divulge confidential results, there is less focus on the
possibility of pressure that results in biased reporting. Both events are clearly detrimental
and should be avoided, and we believe that it would be valuable if the document
addressed this concern as well.

Finally, we urge the FDA to avoid specific proscription for preventing breaches of
confidentiality and biased reporting, but rather to emphasize the goal of eliminating such
problems. We say this for two reasons: First, there are several ways of addressing these
concerns and no single approach will work best for all settings; we know of no evidence
showing that the particular solution proposed in the GCTS document is optimal, and it
may not be a good choice in many settings. Second while we appreciate and value the

FDA’s role in the evaluation of drugs, devices, and biologics, we feel that it is not
appropriate for a regulatory agency to prescribe how studies should be conducted.
Guidelines from the FDA are very valuable, but because such guidelines can be perceived
as requirements for those conducting trials, we feel they should set goals and standards
rather than specify narrow solutions for how to achieve them.

On the issue of ensuring confidentiality and unbiasedness in interim analyses of studies,
we think that a better course of action for the FDA would be to strongly urge (or require)
sponsors to develop written policies that describe the steps they will take to ensure that
these goals are met. At the completion of a study, the FDA can request that both the
sponsor and the statistician who prepares the interim analyses attest in writing that these
policies have been followed. We believe that such an approach will reinforce the
importance of ensuring attainment of the stated goals, and will let organizations decide
for themselves how best to achieve and document them. Naturally, when the FDA and its
scientific advisors evaluate the results of studies, their perceptions on the effectiveness of
this policy will impact their interpretation, and thus there will be a strong incentive for
sponsors to ensure that they have an effective policy. Over time, we believe that such an
approach will increase the appreciation for the goals of confidentiality and unbiasedness
and lead to a variety of effective ways of attaining them, and, therefore, be in the best
interests of the participants in research studies and the public.


Janet Andersen, Executive Director
Center for Biostatistics in AIDS Research

Victor DeGruttola, Director
Statistical Center for the Adult AIDS Clinical Trials Group

Michael Hughes, Director
Statistical Center for the Pediatric AIDS Clinical Trials Group

Stephen Lagakos, Director
Center for Biostatistics in AIDS Research

L. J. Wei. Senior Statistician
Center for Biostatistics in AIDS Research

To top