The influence of study characteristics on reporting of subgroup analyses in randomised controlled trials: systematic reviewBMJ 2011; 342 doi: https://doi.org/10.1136/bmj.d1569 (Published 28 March 2011) Cite this as: BMJ 2011;342:d1569
All rapid responses
Sun et al. caution us about subgroup analyses in industry funded
trials1. We should also take a wider perpective and pay attention to the
relevance of sponsorship and conflicts of interest for authors and for the
results of clinical trials. These have been extensively discussed by the
scientific community and the general public, and the evidence is that the
results and publication of the latter can be influenced by the former2. As
a consequence, numerous organizations have taken steps to minimize such
influence as shown, for instance, by the trial registration requirement3.
With scientific, peer reviewed publications, the most noticeable
consequence has been the disclosure of potential conflicts of interest by
the authors usually informed at the end of the paper, which can sometimes
be a lenghty list, unsuitable for an abstract.
Furthermore, awareness of sponsorship is essential for critical
appraisal; it is a source of bias which can influence significantly any
published paper, either one containing original information, or one
reviewing information already available.
Given the time constraints involved in clinical practice, most
physicians have little time to gather information from the literature. As
a consequence, they most likely follow the steps of reading the title
before deciding whether to proceed to the abstract, then examining the
abstract before going through the full paper or parts of it. This stepped
approach is taken, for example, on systematic reviews4. Consequently,
abstracts are read more frequently than entire papers. Moreover, abstracts
are usually available for free, whereas access to the full articles may
have a cost.
Currently, there is no information about sponsorship available in the
abstracts of major scientific journals.
We suggest the inclusion of the information about sponsorship in the
abstract of all published articles. By sponsorship we mean all sources of
funding directed specifically towards the work reported by the article.
Such information would be restricted to the sponsorship related to the
work described in the publication, for the sake of brevity, without
including other sources of bias usually disclosed in the declaration of
potential conflicts of interest by the authors.
A problem for authors is that they typically struggle to adequately
describe their research within the 250 words (or less) allowed for an
abstract. Do we suggest subtracting the 20 or so words needed for a
sponsorship statement? No, instead, we advocate allowing authors a "free"
20 words at the end of the abstract to describe their sponsorship.
Sponsorship information in the abstract will provide the reader with
a crucial piece of information to critically appraise a scientific paper
without requiring the extra steps of accessing and reading the full paper.
1. Sun X, Briel M, Busse JW, et al. The influence of study characteristics
on reporting of subgroup analyses in randomized controlled trials:
systematic review. BMJ 2011;342:d1569
2. Lexchin J, Bero LA , Djulbegovic B and Clark O. Pharmaceutical industry
sponsorship and research outcome and quality: systematic review. BMJ
3. De Angelis C, Drazen JM, Frizelle FA, et al. Clinical trial
registration: a statement from the International Committee of Medical
Journal Editors. N Engl J Med 2004;351:1250-1.
4. Sanders S, Glasziou PP, Del Mar C and Rovers M. Antibiotics for acute
otitis media in children (Review). Cochrane Database of Systemic Reviews.
Saulo Castel, MD, PhD, FRCPC
Director, Inpatient Services
Department of Psychiatry,
Sunnybrook Health Sciences Centre,
Assistant Professor, University of Toronto
2075 Bayview Avenue, FG21a,
Toronto, ON, Canada, M4N 3M5
Helena M. Calil, MD, PhD
Professor of Psychopharmacology
Department of Psychobiology,
Federal University of Sao Paulo
Former editor of the Brazilian Journal of Psychiatry
Competing interests: No competing interests
The paper by Sun et al  is a useful reminder to the medical
profession about the need for caution when interpreting trials with
subgroup analyses, particularly in trials with non-significant primary
outcomes and when the subgroup analyses are not prespecified. However, in
view of the importance of prespecification, it is disappointing to find
that Sun et al's analysis appears to have deviated from their prespecified
analysis plan .
In the analysis plan, Sun et al state that they will analyse the
influence of study characteristics (industry funding, study size etc) on
whether reports of trials that included a subgroup analysis make a claim
of a subgroup effect on the basis of that analysis. That analysis does not
appear to be reported.
On the other hand, Sun et al do report an analysis of the factors
associated with prespecification of subgroup analysis, although that
analysis, ironically, does not appear to have been prespecified.
We would be grateful if Sun et al could explain these apparent
discrepancies between the results that they report and their prospectively
published analysis plan.
1. Sun X et al. The influence of study characteristics on reporting
of subgroup analyses in randomised controlled trials: systematic review
BMJ 2011; 342:d1569
2. Sun X et al. Subgroup Analysis of Trials Is Rarely Easy (SATIRE):
a study protocol for a systematic review to characterize the analysis,
reporting, and claim of subgroup effects in randomized trials. Trials
Competing interests: AJ & SH are the director and an employee of Dianthus Medical Limited, a company which provides professional medical writing services to both industry-funded and non-industry-funded researchers, but much more often to the former.
Too frequently, subgroup analyses breach the basic methodological
prerequisites: a) a priori definition on the basis of known mechanisms or
in response to previous findings; (b) declaration of the number of
analyses performed and adjustment for multiple testing. However, they are
an epidemic, being present in 44% of randomized trials.(1)
Sun et al. illuminated the hidden face of subgroup analyses.(1) They
rarely serve to generate important hypotheses (exploratory subgroup
analyses) which will be tested in subsequent randomised trials. In
contrast, they seem to be use as sleights to either overcome negative
results of the primary outcome or serve vested interests.(1)
Sleight warned us against this sleight.(2) He showed, from the ISIS-
2 trial (effect of one month of oral aspirin after fibrinolytic therapy
on survival after suspected acute myocardial infarction, a series of
17,187 patients), that aspirin therapy was significantly beneficial for
all patients, except those born under the astrological signs of Gemini or
Libra.(2) Peter's lesson was fun. Who can understand why so many are still
1 Sun X, Briel M, Busse JW, et al. The influence of study
characteristics on reporting of subgroup analyses in randomised controlled
trials: systematic review BMJ. 2011 Mar 28;342:d1569.
2 Sleight P. Subgroup analyses in clinical trials: fun to look at -
but don't believe them! Curr Control Trial Cardiovasc Med 2000;1:125-127.
Competing interests: No competing interests