Beyond conflict of interest

Conflict of interest is being taken more seriously by doctors and by society at large. The New England Journal of Medicine has twice recently been heavily criticised for failing to declare authors’ conflicts of interest—despite its declared policy of doing so.¹,² Last week the BBC halted a £360 000, well reviewed television series because of a “potential conflict of interest”: the producer owned commercial property featured in the series.³ Despite the rising concern, medical journals have done an indifferent job in tackling the problem.⁴ Four years ago I wrote an editorial arguing that we had to do better,⁵ and we began then to require all authors to sign forms declaring conflicts of interest. Unfortunately authors often fail to declare conflicts of interest. This issue of the BMJ contains a collection of material on the subject, and we are proposing new policies.

A common problem

Conflict of interest has been defined as “a set of conditions in which professional judgment concerning a primary interest (such as patients’ welfare or the validity of research) tends to be unduly influenced by a secondary interest (such as financial gain).”⁶ It is a condition not a behaviour, and there is nothing wrong with having a conflict of interest. It is common.

Some people have taken the view that conflict of interest is a lot of fuss about nothing, or, worse, that identifying people’s conflicts of interest is a form of McCarthyism.⁷ Those who argue against concerns about conflict of interest say that science is science, methods are transparent, data either support the conclusions or do not, and it is neither here nor there whether researchers have, for example, shares in a company that manufactures a drug included in a trial.

This argument is becoming steadily less tenable as evidence accumulates on the influence of conflict of interest. Several studies have shown that financial benefit will make doctors more likely to refer patients for tests, operations, or hospital admission,^8–10 or to ask that drugs be stocked by a hospital pharmacy.¹¹ Now we are beginning to have data on the effects of conflict of interest on publications. Original papers published in journal supplements sponsored by pharmaceutical companies are inferior to those published in the parent journal.¹² Reviews that acknowledge sponsorship by the pharmaceutical or tobacco industry are more likely to draw conclusions that are favourable to the industry.^13–16

This year has seen the publication of two important studies that mean we must take conflict of interest more seriously. Stelfox et al showed in a paper published in the New England Journal of Medicine that authors were much more likely to be supportive of calcium channel antagonists for treating cardiovascular disorders if they had a financial relationship with manufacturers of the drugs.⁴ The safety of calcium channel antagonists was a good subject to investigate because it is intensely controversial and the market for the drugs is huge and lucrative. The authors looked at 70 articles (mostly reviews or letters) published in medical journals between March 1995 and September 1996 and classified them as critical of calcium channel antagonists (23), supportive (30), or neutral (17). They then contacted all the authors and inquired about financial relationships with manufacturers: financial support to attend a symposium, speak, organise education, or perform research, and employment and consultation. Two thirds of the authors had a financial relationship with manufacturers, but (and this may be the most important result of the study for journals) “only two of the 70 articles ... disclosed the authors’ potential conflicts of interest.” Almost all supportive authors (96%) had financial relationships with manufacturers, compared with 60% of neutral authors and 37% of critical authors. The study has been criticised for being more about the nature of evidence than about conflict of interest: many of the supportive authors were clinical researchers who are more likely than epidemiologists (most of the critical authors) both to receive funding from manufacturers and to give more weight to clinical judgment than to evidence from randomised controlled trials.¹⁷ Nevertheless, this remains an important study, not least for its demonstration of journals’ failure to disclose conflicts.

Building a convincing case

The second study, published in JAMA, looked at what characteristics determined the conclusions of review articles on passive smoking.¹⁸ The authors identified 106 reviews, with 37% concluding that passive smoking was not harmful and the rest that it was. A multiple regression analysis controlling for article quality, peer review status, article topic, and year of publication found that the only factor associated with the review’s conclusion was whether the author was affiliated with the tobacco industry. Three quarters of the articles concluding that passive smoking was not harmful were written by tobacco industry affiliates. The study authors suggest that “the tobacco industry may be attempting to influence scientific opinion by flooding the scientific literature with large numbers of review articles supporting its position that passive smoking is not harmful to health.” Again, only a minority of the articles (23%) disclosed the sources of funding for research. The authors had to use their own database of researchers linked with the tobacco industry to determine whether authors had such links.

These two papers and their predecessors begin to build a solid case that conflict of interest has an impact on the conclusions reached by papers in medical journals. They also show convincingly that medical journals are failing to get authors to declare conflicts of interest. Readers might want to bear these thoughts in mind as they try to unravel the accusations and counteraccusations in our large cluster of letters that feature conflict of interest (beginning on p 343), many of which are concerned with passive smoking. Look too at the three pairs of papers on whether researchers should take money from industry (starting on p 333): tobacco researchers generally don’t; alcohol researchers are moving towards not taking money; and those researching infant feeding remain divided over taking money from baby milk manufacturers.

What should the BMJ be doing?

The BMJ’s policy is disclosure of conflict of interest rather than prohibition.⁵ We simply don’t think prohibition is feasible, although we try to avoid having an editorial written by somebody with a major conflict of interest. We send authors of all original papers, editorials, and review articles and of selected letters a form in which we define what we mean by conflict of interest and ask them to sign to say whether they have one. We have gone for a broad definition that extends beyond financial interests to personal, political, academic, and religious ones. With original papers we give the source of funding and disclose what authors have told us about whether or not they have other interests. With the other articles we add a note only if authors tell us they do have a conflict of interest.

Our impression, supported by the two recent papers, is that many authors are willing to sign that they don’t have a conflict of interest when by our definition they do. We have two hypotheses to explain this. Firstly, authors think that an admission of a conflict of interest implies wickedness. We don’t think so. Secondly, authors are confident that they have not been influenced by a conflict of interest and so don’t tell us they have one. Our response is that bias works in subtle ways and that none of us is blessed with knowledge of our own motivations and mental mechanisms. We are thus proposing some changes to see if we can do better. They will be phased in from now.

• • We will replace the term “conflict of interest” with “competing interests.” This will, we hope, reduce the sense of wrongdoing and encourage people to disclose competing interests.

• • We will restrict ourselves to financial interests and modify our form accordingly. The authors of the New England Journal of Medicine article suggest that authors should be sent a questionnaire similar to the one they used in their study, and we have adopted this idea (see the form on our website (www.bmj. com/guides/advice.shtml)). Restricting ourselves to financial interests is a tactical move: narrowing the range may make it more likely that authors will declare competing interests. If authors want to disclose other competing interests then we will disclose them to readers.

• • Authors of all original papers, editorials, and review articles will be asked to complete our questionnaires. Competing interests will be disclosed, and if authors tell us they have none (the usual case) we will write “none declared” rather than “none.” With letters we will continue to encourage authors to disclose competing interests but will send them a questionnaire to complete only if we suspect that authors might have competing interests. Authors of letters about drugs will usually be sent a questionnaire.

• • If we learn after publication that authors had competing interests that they did not disclose then we will tell readers.

Some readers will regret such moves and remember a golden age when conflict of interest was not an issue. Times have changed however, and trans- parency and accountability are increasingly expected in all aspects of society. I doubt that the changes we are proposing will solve the problem, but they seem to us to be a step in the right direction. Authors and readers who disagree will no doubt tell us—and we will listen.

News p 301, Education and debate p 333, Filler p 318, Letters p 344