Science and Technology CommitteeWritten evidence submitted by Dr Elizabeth Wager
Statement of Competing Interests
I am a freelance writer, editor, trainer and publications consultant. I work with pharmaceutical companies, publishers, academic institutions and individual researchers. In 2012 about 40% of my income came from pharmaceutical companies. I am a former employee of the pharmaceutical industry (I was UK Medical Writer for Janssen Cilag 2002–09 and UK Head, International Medical Publications, Glaxo Wellcome/GlaxoSmithKline 2009–11) and will be eligible for some pension from each of these companies. I am a member of the European Medical Writers Association (EMWA) and the International Society for Medical Publications Professionals. I have received travel expenses and run workshops for both these organizations. I am the author of various guidelines related to this topic, including Good Publication Practice for Pharmaceutical Companies and the EMWA guidelines on the role of medical writers in developing peer-reviewed publications.
Scope of Submission
I wish to comment on two questions:
(Q3)
(Q4)
What evidence is there that pharmaceutical companies withhold clinical trial data and what impact does this have on public health?
3.1 There is considerable evidence that pharmaceutical companies withhold clinical trial data. This may take the form of failing to publish all results from studies (ie selectively publishing only some of the findings) or failing to publish entire trials.
3.2 Recent evidence of the possible effects of incomplete reporting comes from Egan et al (Can J Hosp Pharm 2012;65:387–93). They examined different meta-analyses (which combine data from several trials) about drugs commonly used to treat high blood pressure. Some of these meta-analyses suggested the drugs were associated with an increase in the incidence of cancer, while others did not. Egan et al concluded that the reason the studies reached different conclusions was outcome reporting bias (ie selective reporting of study findings). An earlier study of anti-depressants comparing information submitted to regulatory authorities with results published in journals also provided clear evidence of publication bias by drug manufacturers (Melander et al, BMJ 2003;326:1171–3). The effect of this bias (in which positive studies were published more than once and studies with negative findings were not published) was to make the drugs appear more effective than they really were.
3.3 More evidence of reporting bias comes from a study published in the BMJ last year (2012;344:d7202). Hart et al, took 42 meta-analyses of nine drugs approved by the FDA (the US regulator) in 2001–02. They re-analysed the published meta-analyses (which had included only published studies) and included unpublished data obtained from the FDA (ie supplied by the manufacturers to the US regulator but never published). Including the unpublished data caused the drug to appear less effective in 46% of cases and more effective in 46%.
3.4 From my personal experience of working within the pharmaceutical industry (2002–11) I was aware of under-reporting. This had a variety of causes, the most common being:
transfer of resources from drugs that were no longer being developed;
lack of interest of clinical investigators (who did not perceive findings to be particularly interesting or did not have time to write up the results of their research);
journal space constraints (especially before the availability of electronic supplementary files);
rejection by journals (especially before the creation of less selective journals, such as PLoS One, and journals specifically focused on negative findings); and
omission of unfavourable or inexplicable outcomes.
In my experience, these reasons were much more common than deliberate policies to suppress findings or studies, although I am aware that there is evidence that companies have engaged in such behaviour.
3.5 While this enquiry may focus on problems with pharmaceutical companies, it is important to note that non-publication and selective reporting are also well documented among academic research. For example, Chan et al, examined studies funded by the Canadian Institutes of Health Research and found that 59% of outcomes related to treatment adverse effects were incompletely reported (Chan et al, Canadian Medical Association Journal 2004;171:735–40).
How could the occurrence and results of clinical trials be made more open to scrutiny? Who should be responsible?
4.1 The most obvious way to ensure that the occurrence of trials is transparent is to require that all trials must be registered on a public register such as the ISRCTN or ClinicalTrials.gov. Registration of study design details before the study begins can also help to reduce, or at least identify, the selective reporting of outcomes, or changes in study design occurring between initiation and publication. Since 2005, many of the major general medical journals (including The Lancet and BMJ) have refused to publish trials unless they have been registered. This policy led to a sharp increase in trial registrations. US legislation (specifically the FDA Amendments Act of 2007–08) strengthened this trend by making registration compulsory for many trials of new medicines. Most pharmaceutical companies therefore comply with this legislation and the journal editors’ requirements and register all Phase II to IV studies of new drugs.
4.2 The FDAAA also required the posting of summary tables of study findings on ClinicalTrials.gov within 12 months of the end of most studies of new drugs. Although studies have revealed shortcomings in these postings, and the occurrence of late or incomplete disclosure, this legislation has greatly increased the availability of summary trial findings.
4.3 However, while the summary results tables on ClinicalTrials.gov are useful, they are not easy to understand without some knowledge of trial design. Therefore clinicians and patients continue to rely on other sources of information that present results in context and provide more explanation and interpretation. The conventional method of communicating study results to doctors is via articles in peer-reviewed medical journals and this continues to be regarded as the best method of publication although it is by no means perfect (see Wager PLoS Clinical Trials 2006;1(6):e31).
4.4 Given the reliance of doctors on journal articles and the benefits that this type of publication carries (such as permanence, the possibility for corrections or retractions, some measure of quality control via peer review, the opportunity for post-publication commentary and discussion) companies should be encouraged to submit reports for publication. One mechanism would be to require companies (and associations such as the ABPI) to endorse Good Publication Practice guidelines (see Graf et al, BMJ 2009;339:b4330) which call on companies to endeavour to publish all results of clinical trials on marketed products.
4.5 Preparing articles for medical journals requires time and expertise. In my experience of working for pharmaceutical companies, external investigators are not always equipped or prepared to do such work, especially for studies they consider routine or relatively uninteresting. Most journal editors therefore recognise that professional writers can have a legitimate role in helping to develop such publications, so long as their involvement and financing are fully disclosed. Professional writers should only be regarded as “ghost writers” if their contribution or link to the funder is not properly acknowledged. There are several guidelines on the role of medical writers in peer-reviewed publications, such as those from the European Medical Writers Association (Jacobs & Wager, Current Medical Research & Opinion 2005;21:317–21). The case for greater involvement of professional writers has recently been put forward by Woolley et al (Current Medical Research & Opinion 2012;28:1857–60). They conclude that “professional medical writers could help ensure results are reported in a complete, timely, and ethical manner”.
4.6 While journal articles will continue to be widely used by doctors and patients, alternative methods for disseminating research results should be investigated. Structured summaries or tabular formats (such as those on ClinicalTrials.gov) should be developed. Regulatory authorities (such as the EMA) require companies to produce detailed clinical trial reports but these remain confidential after submission to the authorities. Greater transparency could be achieved if regulators opened up their archives to the public or if companies posted clinical trial reports (or their summaries) on their corporate websites.
4.7 One barrier currently preventing companies from making trial reports available is that medical journals will only consider findings that have not been published elsewhere. Therefore if a company wishes to publish an article in a medical journal it will be deterred from posting a study report or extended summary on a website. The journal editors do allow the very short summary postings required by FDAAA, but should be encouraged to relax their requirements about prior publication to encourage the wider dissemination of full clinical trial reports or extended summaries (see Wager & Abbasi, Journal of the Royal Society of Medicine 2009;101:1–2). Regulators (such as the EMA) should also become more transparent and disclose the information submitted to them that forms the basis of product licensing. The EMA has recently indicated that it is considering greater transparency and this is to be welcomed, but it is not yet clear either when this will happen or how.
February 2013