If you ask someone in my fields who has just used the term “conflict of interest” what it means, I would estimate that you would get a meaningful answer less than 5% of the time. Most answers that were other than a blank stare would be some mechanistic and legalistic statement along the lines of “you are supposed to declare who has funded your work when you publish in a journal”, which is not actually a response to the question.
This is especially sad since the definition is built into the term: all you need to do is figure out what interest someone has and what other interest it conflicts with. Even with that hint, though, my experience is that those who are fond of exclaiming that someone is guilty of COI has no idea what it means.
Here’s what it means in the presently relevant context: The interest that may be threatened is the mythical scientific or journalistic “objectivity” – roughly the idea that someone is completely indifferent about how their analysis turns out and are just following the data. This is pretty much never the case in the health sciences (and probably most areas of inquiry) – very seldom does someone start an analysis without some notion of how it will come out, and some preference about whether that outcome is realized. This is why a huge portion of the epidemiology literature suffers from publication bias in situ.
That said, most of the time most analysts working on most subjects do the best they can to analyze and report in a way that is close to the theoretical ideal. The interests that conflict fiercely with that goal are strong preferences about the world that would be aided by a particular result of the analysis. The most important of these are preferences about how the world should be that might be supported by the results of the research. Thus, someone who believes that all tobacco products should be eliminated and who reports research on harms from tobacco products has an obvious enormous conflict of interest: They are looking for evidence that might encourage consumers or regulators to conform to their wishes. An equally compelling conflict of interest is when someone is employed by an organization that has a similar worldly preference. Of course, if the employee already shares that preference (as they likely would, having taken that job) the employment relationship probably adds little to the motivation.
MSNBC suspended the anchor/commentator who is the only apparent reason for watching MSNBC (or, I am informed by Maddow fans, one of two), Keith Olbermann, because he donated money to some Democratic candidates in the last election, violating company rules about how their journalists must avoid COI or remain objective. (Those are my characterizations of the motivation for the policy; I have not noticed them being explicitly stated in what has been written about the matter.) Of course, they quickly reinstated him after the public outcry about his absence and the legality of the rule, and the publicity was so valuable that the truly cynical can suggest it was all planned. But there was nothing in the retreat that suggested that anyone realized how naive the action was in the first place.
For those who do not know, Olbermann is probably the most interesting and insightful American liberal television commentator (sorry, Jon Stewart – that used to be you). As such, there is no question that his role is to advocate for a particular worldview. If you interpret his job as being that of “objective journalist” (whatever that mythical concept means), then his strong beliefs present a COI (as do most “objective” journalist’s worldly beliefs). If, however, you consider his job to be the spokesman for his causes (and, even more so, brilliant critic of the dishonest and/or nutcase opponents of those causes) then there is no conflict. But in either case, his donations to the cause cannot possibly create a COI. He makes the donations because he has particular beliefs, and those beliefs matter. But to anyone who thinks that the changing hands of money matters in this case apparently has some difficulty with the ordering of cause and effect.
To draw an analogy back to the tobacco researcher, if that researcher were doing a study of whether a particular intervention were effective at discouraging tobacco product use, there would be no COI with her worldly preference for eliminating use. Indeed, the motivation to figure out exactly what works would tend to favor doing the best possible analysis. (So why does it seem that researchers evaluating anti-tobacco interventions always seem to twist their results to make the intervention appear more effective than it really was? Perhaps they are not very good at what they do. Also, it is often because they are employed by the organization that implemented the intervention, which creates an obvious COI that might overwhelm the motivation to get the right answer.) Just as Olbermann’s personal beliefs, and even ties to the Democratic party, do not conflict with him being a good spokesman for his causes, a tobacco researcher’s goal of eliminating use would not create a conflict with research designed to figure out how to eliminate use.
So if Olbermann pretended he was not doing what he does – trying to make the best possible arguments for a particular viewpoint – then his beliefs would represent a COI (even still, the exchange of money that caused the trouble would still be irrelevant, serving only as evidence of the real COI). Similarly, if the anti-tobacco researcher reported study results that tended to support the anti-tobacco cause while pretending to be an objective researcher, then there would be a COI. So, naturally, we would expect her to declare her COI when publishing the results. (Bazinga!)
Contrast this with the recent hand wringing by the NYT over analysis by their excellent business reporter/columnist Joe Nocera, who condemned an executive’s ethics in an article only to find out (after it was published, unless someone is lying) that the law firm where his fiancee works is representing a company that is suing over that executive’s actions. Nocera reported this when he learned it and the NYT published what was basically a COI disclosure which appears with the original column (ok, it could not be added to the print version, but does anyone actually read printed periodicals anymore?). As it played out, there was actually no COI because the author did not know his family had an interest, via employment, in making the arguments he made. But if he had known, this would have been a clear COI that should be reported since the appearance of disinterested journalism would have been misleading. (As and aside, the NYT never once flagged either the original column or their subsequent discussion with the topic label “ethics”, a label they usually inappropriately reserve for sex scandals and insider activities that are clearly criminal rather than interesting ethical questions.)
Nocera declared that he would not have written the story if he had known about the chain of relationships. This is a reasonable strategy in this case. No matter how heartfelt, it is difficult to make worldly declarations while forgetting that they could affect the success of an employer (of oneself or an immediate family member). And the downside of not doing it is not so great since for newspaper reporting, anyone else on the beat could pick up the story, or for that matter the world would not notice if no one did. The latter contrasts with scientific situations where the people most invested in a topic are likely to have worldly preferences, but are also the only ones who are expert enough to do the analysis, so simply avoiding it is not possible. So long as someone is well-known to be or discloses that they are an advocate for a particular worldly goal (and similarly makes clear the preferences of their employer), there is no reason to avoid making the best case for that goal with the Olbermann approach: choosing particular topics to analyze, disputing opposing claims, and making opinionated declarations. This does not extend to making factually misleading statements or analyzing data in ways to skew the result, but that is a topic for another time.