is used in clinical practice. The use of the ITT is
required by the FDA as one of the analyses to
always be presented to them. The problems we
highlighted earlier in this section present chal-
lenges to the data analyst that can be addressed at
the analysis phase only to a limited extent. It is
impossible to design a trial so that these problems
will be prevented entirely. However, a careful
design and diligent execution and monitoring of
the trial can minimize them.
25.13 The dissemination of
clinical trials resultsClinical trials are complex and expensive scientific
endeavors. For this reason, most clinical trials are
supported by either pharmaceutical companies or
government. Pharmaceutical companies conduct
clinical trials not only as part of their clinical
development of new therapies but also to discover
new indications or special features of their
approved drugs as part of their marketing activities.
The results of trials conducted as part of the drug
approval process are summarized and submitted to
the FDA. In addition, the FDA requires sponsors to
summarize and submit the results of all other stu-
dies of the drug conducted by the sponsor or that
were published in the scientific literature. This
information becomes part of the public knowledge
after the drug is approved. Studies conducted out-
side of the New Drug Application (NDA) process
are treated quite differently. As these studies are
conducted for the purpose of promoting the sales of
the drug or in exploration of additional indications,
there is no requirement to submit the results to any
governmental agency unless the company decides
to submit a supplementary NDA. Many such trials
result with negative outcomes and the sponsors as
well as the clinical investigators have little or no
incentive to expend the resources of analyzing data
from trials and publish the results. Contributing to
this is the fact that scientific publications must
make editorial decisions as to what they will and
will not accept for publication. Negative studies
are often found uninteresting scientifically and
are refused publication. The result of this is
that the information available to the medical
community is selective and incomplete. This
might have serious public health consequences,
as it could influence medical practice. Medical
practitioners may not be aware of certain adverse
effects of a drug they prescribe, the usefulness of
drugs in the treatment of certain conditions may
not be known, or worse, treatments proven ineffec-
tive may continue to be employed. This problem
has been recognized and a federal law was passed
in 1997 requiring the sponsors of clinical trials
to report the existence and result of their trials to
the FDA to be included in a trial registry and
available to the public. The law has been largely
ignored by the pharmaceutical industry, and the
FDA did little to enforce it. A registry was estab-
lished in 1998 but only a small percentage of
industry-sponsored trials were posted in it. This
issue came to the attention of the American public
recently after it became known that one of the
nation’s major pharmaceutical manufacturers has
been withholding information of serious adverse
events of one of their antidepressants in children
and adolescents. (Bloomberg Business News,
2004; The Washington Post, 2004a, 2004b). It
seems that the publicity given to this issue may
encourage sponsors to be more diligent in making
the results of their trials known and the FDA enfor-
cing this requirement.25.14 The statistician’s role
Information derived from data collected in a clin-
ical trial is the ultimate product of the trial. Every
aspect of the trial from its conception to its execu-
tion impacts the quality of the data and the infor-
mation they contain. The final step in the process,
the analysis, is nothing but the application of
statistical methods for organizing the data, sum-
marizing them and extracting relevant informa-
tion; that is, separating the signal from the noise.
The statistician’s ability to make up for design
deficiencies and for noisy data is limited, and the
same rule defining good practice of medicine
applies here as well: the best treatment of a disease
is to prevent it. The statistician’s greatest impact25.14 THE STATISTICIAN’S ROLE 341