Full Disclosure = MORE Bias?

I've lagged in posting lately in order to take my recertification exam in psychiatry. I hope I passed. For some reason, there were a lot of questions about physostigmine, a medication the Board seems to believe is crucial for psychiatric practice. I guess they got bored writing questions about SSRIs.

Lacking any urgent CME news, I caught up on some reading and came across this interesting article on the perverse negative effects of full disclosure. Usually, we think that disclosure minimizes the biasing effects of conflict of interest (COI). Accordingly, the ACCME's standards for commercial support are based primarily on an elaborate series of guidelines about how to disclose COI in educational activities. But a funny thing happens on the way to CME programs: even with full disclosure, they continue to be commercially biased. Why is that? Why doesn't disclosure solve the problem?

Business school researchers at Carnegie Mellon University conducted two experiments to answer these questions. Undergraduates from the university were asked to estimate the number of coins in a glass jar. Some were assigned the "estimator" role and others, the "advisor" role. In the CME world, the advisors correspond to hired gun speakers, and the estimators are physicians in the audience.

Here are the results:

1. In the first experiment, advisors were paid to give erroneous advice about the number of coins in the jar. Specifically, they were asked to exaggerate the number of coins (bias high) or to mimimize (bias low). Crucially, the estimators were warned ahead of time that the advisors had these incentives to give bad advice. But even with these disclosures in mind, the estimators followed the biased advice.

2. In this variation, some advisors were paid to give accurate advice, and others were paid to give high advice. As expected, the advisors who had an incentive to give high advice succeeded in influencing the estimators. But here's the interesting twist. A third group of advisors was paid to give high advice, and was told that the estimators would be warned about their incentive to bias the answer upward. The result? Advisors who knew their bias was disclosed actually provided more erroneous advice than the others. In other words, disclosing the COI actually makes hired guns more likely to provide biased information.

The authors speculate that disclosure worsens bias for two reasons. First, disclosure gives the advisors "moral license" to exaggerate. To quote from the paper: "When their conflict of interest is disclosed, people feel less of an obligation to protect the advice recipient." The other issue is "proactive exaggeration." Since advisors who disclose their COI worry that their advice will be discounted by the audience, they will tend to exaggerate their advice to make up for it.

This is quite an intriguing paper, and deserves to be studied by regulating bodies in all advice-giving industries, including medicine. It may well be that disclosing COI has the unintended effect of encouraging people to give worse advice than they otherwise would.