Jonathan Latham and Allison Wilson
Failure to declare a conflict of interest, as Lester Crawford has been reminded (see news item), is a federal offence for United States Government employees, punishable by a prison term. To many scientists however, conflicts of interest are a fact of life. Members of hundreds of government advisory panels hold shares in, consult for, or are employed by, the companies about whose products they are supposed to provide ‘independent’ guidance (Krimsky, 2003). Similarly, many public interest organisations, notably patient groups and charities, are in the similar position of receiving money from corporations affected by their policies, conduct and advice. The prevailing attitude in science is that these conflicts either are unavoidable, because most successful scientists have them, or that they do not matter because scientists are sufficiently objective to discount them. In the words of numerous editorials and official guidelines, these conflicts are ‘apparent’ rather than real. None of these arguments should carry much weight.
Firstly, there is no such thing as an ‘apparent conflict of interest’. All conflicts of interest are real. Either an individual or institution serves more than one master, or they do not. This is true regardless of whether the holder of that conflict (or anyone else) acknowledges its existence, or believes that their opinions or actions are altered by the fact of their conflict. Secondly, it is probably not true that such conflicts are unavoidable. According to Sheldon Krimsky of Tufts University (Krimsky 2003), companies go out of their way to manufacture conflicts by offering consultancies and other employment to as many scientists as they can. However, they would be unlikely to do so if their favoured scientists were consequently debarred from committee service. Thus the unavoidability argument is seemingly a circular one.
In medical research it has proved impossible to demonstrate in individual cases, whether scientists are, or are not, swayed by their conflicts of interest. Nevertheless, in most of public life it is accepted that conflicted individuals make bad decision makers. This judgement has been recently supported by multiple studies of medical research showing that even in science results depend very much on who funds them (e.g. Kjaergard and Als-Nielsen 2002; Bekelman et al., 2003). Since there is no reason at all to believe that scientists are more blinkered or indifferent to financial gain than individuals from other professions and since a large proportion of safety and efficacy studies (and not just in medicine) are funded by interested parties, it is reasonable to assume that the peer-reviewed literature in many disciplines has become dangerously compromised. It is time for science to take conflicts of interest seriously.
Until now, conflicts of interest in agriculture and food science had not been formally studied (Lesser et al., 2007). Nevertheless, significant portions of agricultural research are conducted by industry-affiliated scientists and many of these papers are grossly unsound in their methods: and specifically in ways likely to generate results that are favourable to their employer. In our experience many examples pertain to the biosafety properties of commercial transgenic crops, but there is evidence that nutrition and pesticide research are similarly affected (Lesser et al., 2007). These papers typically follow a simple logic. If a method can be devised that is sufficiently insensitive, then results that show no statistically significant difference (e.g. for a safety character) between a new product and a control, can be almost guaranteed. The simplest way to achieve this is to use sample sizes that are so small that the experiment lacks the statistical power necessary to reveal potentially hazardous differences. But other methods can also achieve the aim of obscuring potentially damaging findings. These include combining data points inappropriately, using inappropriate controls or failing to apply appropriate (or sometimes any) statistical methods. All of these defects are found for example in Ridley et al., (2002) a paper whose purpose is to demonstrate the substantial equivalence of NK603 (glyphosate-tolerant) transgenic maize. The authors of this paper, for example, not only pooled data from field trials of different sites but also from field trials conducted in different countries (Ridley et al., 2002). The authors of another maize compositional analysis paper went one better and pooled data from different continents (Sidhu et al., 2000).
Compositional and feeding studies of GM crops are not the only subject areas affected. Others have noted such tendencies in studies of ecological effects of transgenic plants (BTO, 2003; Marvier, 2001) and nutrition (Lesser et al., 2007). Examples can also be found in studies of the implications of virus-resistant transgenic plants for virus evolution. The latter includes a remarkable study ‘looking’ for novel viruses arising from recombination with virus-derived transgenes (Thomas et al., 1998). Novel recombinant viruses were not found but this may have been related to the fact that the authors searched for recombinant viruses in samples that were highly unlikely to contain them, using methods that were highly unlikely to detect them and in any case never ascertained that the host plants were transgenic for the viral gene with which recombination was supposed to be occurring (Thomas et al., 1998).
These papers, though astonishing in their egregious breaches of scientific norms and expectations, are far from unusual, nor do they come only from one company. The authors of course should bear ultimate responsibility but how much should be shared by the scientific community? Part of the answer is that these are peer-reviewed publications that have been accepted by editors and, so far at least, tolerated by the scientific community at large. Some of these papers have been more than tolerated, they appear to have had significant influence. The negative results of Thomas et al., (1998) were prominent in the conclusions of the British Government commissioned GMO Science Review Panel. But of course if such papers had no influence they would never have been produced in the first place.
Solutions to the problem of conflicts of interest must be systemic if they are to work. Declarations of conflicts of interest are not enough, especially when these are not enforced, or even required by most journals. Nor is it enough to disbar from committees individuals whose institutions have conflicting interests. What is needed is a significant and diverse body of scientists who do not have conflicts-of either a personal or an institutional nature. Much could be achieved by two simple changes-forbidding financial arrangements between public universities (and their employees) and outside interests and requiring data submitted to public committees to be generated by this public research system. Even this arrangement would not be enough, since there remains the problematic issue of patents which introduce a somewhat distinct pattern of institutional conflicts for public universities. But it would certainly be a fine start.
Bekelman J. et al. (2003) JAMA 289: 454-465
British Trust for Ornithology comments on the GM Science Review Panel (2003)
Kjaergard L. and Als-Nielsen B. (2002) BMJ 325: 249-252
Krimsky S. (2003) Science in the private interest (Rowman and Littlefield)
Lesser L. et al. (2007) PloS Medicine 4(1) e5
Marvier M (2001) American Scientist 89: 160
Ridley W. (2002) J. Agric. Food Chemistry 50: 7235-43
Sidhu R et al. (2000) J. Agric. Food Chemistry 48: 2305-12
Thomas P et al. (1998) Mol. Breeding 4: 407-417
If this article was useful to you please consider sharing it with your networks.