Industry funds a large share of the research that evaluates its own products.
The arrangement is not avoidable in any straightforward way, since the alternative is less research.
The measurable consequence is well established across several fields.
Industry-funded trials are more likely to report results favourable to the , and the difference is not explained by study quality.
Several mechanisms produce that pattern without anybody anything.
The comparison drug may be given at a that is too low or too high.
The outcome measured may be the one on which the product performs best.
The trial may end before an would have appeared.
Publication is the strongest mechanism of all and operates after the science is finished.
A result that disappoints a is less likely to be submitted, and a contract may permit a delay.
Trial was introduced to close that route and has substantially done so.
Registering the outcome measure in advance makes a later change visible to anybody who checks.
Checking, however, is unpaid work that competes with a researcher's own publications.
is the standard remedy and is weaker than it appears.
Experiments show that a declared conflict can increase the influence of advice rather than reduce it.
A reader treats disclosure as a sign of honesty, and an adviser who has disclosed feels licensed to .
The structural answer is to separate funding from design through an body that allocates the money.
Industry pays into a pool and does not choose the question, which is the arrangement it has resisted in every sector where it has been proposed.