I suppose that one might argue that drug companies will be less likely to allow "non-trusted" researchers to participate in studies if their contract terms are either held unenforceable or deleted for the future. That's always a risk, of course; the problem is that the FDA's licensing scheme already calls into question whether the drug companies have any "legitimate expectation of privacy" in efficacy and safety testing of their drugs. I don't even have to rely on "public interest" issues.
This resembles practices in the entertainment/publishing industry regarding reporting of sales and royalties; but explaining how would take too much space and effort to be appropriate in a blawg, not to mention crawl awfully close to the line drawn by certain protective orders. In any event, that's far less serious than just saying "Pinto gas tank."
For whatever reason, the drug companies as a group appear to believe that hiding data is economically in their best intereststhat the potential rewards outweigh the potential risks. That alone indicates that we have a serious risk of market failure… and that, in turn, assumes that drug choice is even the result of fully (or even reasonably) informed choice in the first place.