Intended for healthcare professionals

Papers

Association between competing interests and authors' conclusions: epidemiological study of randomised clinical trials published in the BMJ

BMJ 2002; 325 doi: https://doi.org/10.1136/bmj.325.7358.249 (Published 03 August 2002) Cite this as: BMJ 2002;325:249

Industry sponsored trials: an alternative view

Dear Editor,

Two major problems seriously limit the conclusions that may be
validly drawn from the study by Kjaergard and Als-Nielsen [1].

The study fails to take on board important aspects of drug
development within the pharmaceutical industry. Of crucial importance is
the backdrop of an exceedingly high failure rate of drugs in early phases
of development. Industry-sponsored studies of beneficial drugs are
scrutinised by regulatory agencies (e.g. the US Foods and Drugs
Administration and European Agency for the Evaluation of Medicinal
Products), and hence positive results and conclusions of trials submitted
for regulatory approval and subsequently published in journals should be
expected. Industry trials published in scientific journals such as the
British Medical Journal are based upon extensive clinical studies
including pre-clinical work i.e. animal, laboratory, and pharmaco-kinetic
and pharmaco-dynamic studies of many types. The proportions of various
types of trials conducted by industry may also be different to those
sponsored by non-industry sources. It would be useful to see data that
compares the results of placebo-controlled trials (often a requirement of
regulatory agencies) and comparative trials by type of competing interest.
Comparative trials are more likely to be conducted after registration of
drugs and, hence, we believe, less likely to be dominated by funding from
industry. Comparative trials are also more likely to involve comparisons
of non-pharmacological interventions, which might be less likely to reach
valid positive results. Regulatory authorities also approve drugs with
benefits on important surrogate endpoints. Use of important surrogate
endpoints, generally requiring shorter studies, is also likely to produce
more favourable comparisons than those employing outcomes further down the
course of a disease. The use of surrogate endpoints often requires
training and quality control monitoring to minimise random and systematic
error and such quality control is likely to be better in industry trials.
As a consequence of the higher quality of trials conducted by industry,
conforming rigorously to good clinical practice (2), and extensive
screening for trial eligibility, drop-outs may be smaller in industry-only
funded trials resulting in more valid favourable results and conclusions.
The important issue of drop-out was a component of quality that was not
taken into account in the paper by Kjaergard and Als-Nielsen (1).

The second major issue concerns use of less than ideal methodologies.
We believe that a meaningful analysis requires a comparison of 'like with
like'. Grouping studies that are methodologically different and hoping
that adjustment will correct for imbalances is suboptimal. The
inadequacies of statistical tests of heterogeneity and traditional
epidemiological analysis to adequately adjust for differences in
covariates (especially for those that are not included or unknown) make
this mandatory. In particular, whether potential confounding variables are
statistically significant or not is irrelevant. Hence, the analysis should
compare 'industry-only' versus 'non-industry-only' funded studies for at
least two specific subtypes of trials: 1) unconfounded, randomised, double
-blind placebo-controlled trials of drug therapies and 2) unconfounded,
randomised, controlled trials of drugs against another intervention in
comparative trials, making allowance for differences in surrogate
endpoints and dropouts.

There is enormous scope for selection bias inherent in this kind of
research, documented in letters by others in response to the article.
Hence, we believe that to provide a convincing basis for drawing
inferences, the primary objectives, methodology and results of each
relevant trial, without identification of sources of funding and drug,
should be sent to 'blinded' 'independent' academics working in fields
different to those of the topic of the trial.

Vindication for our approach and scepticism of the paper by Kjaergard
and Als-Nielsen (1) comes from a recent independent study that contradicts
the findings. This study, published in a less 'prestigious' journal than
the BMJ, found no evidence of bias from industry-sponsored trials, using
more robust methodology closer to what we advocate (3).

References

1. Kjaergard LL, Als-Nielsen B. Association between competing
interests and authors' conclusions: epidemiological study of randomised
clinical trials published in the BMJ. British Medical Journal 2002; 325:
249.

2. Senn SJ. Statistical quality in analysing clinical trials. Good
Clinical Practice Journal 2000; 7: 22-26.

3. Clifford TJ, Barrowman NJ, Moher D. Funding source, trial
outcome and reporting quality: are they related? Results of a pilot study.
BMC Health Services Research 2002; 2:18.

Nawab Qizilbash
Director, Epidemiology, GSK;
and Visiting Professor of Geriatric Medicine, University Hospital GM-
Cantoblanco, Madrid, Spain
nawab.qizilbash-1@gsk.com

Frank Rockhold
Senior Vice-President, Biomedical and Data Sciences, GSK

N Sreeharan
Senior Vice-President and European Medical Director, GSK

GlaxoSmithKline
Third Avenue
Harlow CM19 5AW
UK

Competing interests:  
All authors are employees of GlaxoSmithKline

Competing interests: No competing interests

02 June 2003
Nawab Qizilbash
Director, Epidemiology
Frank Rockhold, N Sreeharan
GlaxoSmithKline, Third Avenue, Harlow CM19 5AW