later the same negative result. Chalmers (1990),
somewhat hyperbolically, has actually character-
ized underreporting of clinical trials data as scien-
tific misconduct.
If this underreporting is suboptimal, then those
who publish clinical trials must take their share of
the blame. Incongruously, it is the same journal
editors who have traditionally been least likely to
publish negative data that are making the most
noise about the unsatisfactory performance of the
pharmaceutical industry in failing to publish the
data (e.g. Horton and Smith, 1999; Tonks, 1999).
This author cannot agree with Dickersinet al.
(1992) who wrote: ‘Contrary to popular opinion,
publication bias originates primarily with investi-
gators, not journal editors:..’ because the busy clin-
ical trialist is unlikely to waste his or her time
writing a paper that he or she knows has little
chance of being published.
The establishment of clinical trials registries
may be one way to overcome the bias against
reporting of negative clinical trials. This is not a
newidea (e.g.Simes, 1986)and several worthwhile
attempts have been made to accomplish this. The
National Health Service in the United Kingdom
(Peckham, 1991), an amnesty for the publication of
clinical trials offered by some journals (Roberts,
1998), and specialized databases (especially in the
areas of malignant disease and AIDS) have been
partial responses to the many pleas for registration
of clinical trials. Two large pharmaceutical com-
panies have taken an initiative to register their own
clinical trials (e.g. Sykes, 1998), but have been
ungratefully criticized both for doing too much
and for doing too little: some think that the regis-
tered information is insufficient, whereas others
believe that this creates a commercial disadvantage
(Horton and Smith, 1999).
A further bias in clinical trials publishing is the
selective reporting of subsets of secondary end
points. This is usually associated with active-
comparator trials having a primary objective of
demonstrating the superiority of one treatment
over the other. All too often, the primary objective
of the trial is not achieved: the authors then selec-
tively publish a few of the many secondary end
points that did support their hypothesis. The ‘if you
have 100 end points anda¼ 0 :05, then, at random,
five end points will be statistically significant’
principle supervenes; fallacious treatment differ-
ences are claimed after reporting only those five
end points. Solutions to this problem could include
an independently prepared summary of the pro-
tocol, with its prospective objectives and complete
list of end points, perhaps in mini-type, at the end
of such papers, as well as sensitization of reviewers
to this potential problem. Journal editors some-
times approach this ideal by asking for protocols
to accompany the submitted manuscripts; some
companies view their protocols as confidential,
and one wonders whether this is one of the
reasons why.
Thus, there are multiple ways in which publica-
tion bias may be created by study sponsors, pub-
licists, medical writers and those who control
journal content. Clinical trial registries still do
not exist in any comprehensive fashion. Those
constructing meta-analyses from published studies
should beware.42.4 The classic components
of a clinical trial report
in a peer-reviewed journalThe publication of clinical trials in peer-reviewed
journalsnormallyfollowsthesameformatasforany
other paper: title, authors, sponsorship, abstract,
introduction, methods, results, discussion, conclud-
ing paragraph, acknowledgments, references, tables
and figure legends, with each figure attached on a
separate sheet labeled on the reverse. The overall
philosophy is also the same as for any other paper,
namely that there should be enough information for
the study to be replicated in independent hands,
should the need arise. It is beyond the scope of
this chapter to teach how to write a scientific
paper: there are many other books, manuals and
journals that can devote enough space for this pur-
pose (succinct examples include Skelton, 1994;
Bonk, 1997; Fromteret al., 1999).
All journals publish guidelines describing the
formats for the often diverse types of article that
will be considered. The corollary is that the writer
should identify the target journal before putting42.4 THE CLASSIC COMPONENTS OF A CLINICAL TRIAL REPORT IN A PEER-REVIEWED JOURNAL 567