Well first of all, hiding negative date is not the same as hiring corrupt or incompetent researchers to generate fraudulent data, which was HungryHobo's claim. Second, drug developers can and do manipulate which data they publish, but it's much harder for them to conceal negative data from the regulators. That's even more true now that regulators require up-front registration of any trials that a developer hopes to use to gain approval. You can't start 20 trials, hoping to get approval based only on the 1 that randomly meets p < 0.05, because you'd have to register all of them. And the regulators will ask you for results on them all.
In fact, HungryHobo's link about Tamiflu makes that point pretty clearly:
From the Plain Language Summary (see http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD008965.pub4/abstract;jsessionid=D0952B85723A29FF11B3087DB5F71938.f03t02
In previous versions of this review we identified unresolved discrepancies in the data presented in published trial reports and substantial publication bias. As a consequence, we elected not to use data from journal articles but included the documents generated during licensing processes. We have accessed such data from the UK, USA, European Medicines Agency (EMA), Japanese regulators and clinical study reports from the manufacturers (after a protracted media campaign). This has enabled us to verify information from the randomised, placebo-controlled trials on adults and children with confirmed or suspected exposure to naturally occurring influenza.
Based on our assessments of the regulatory documents (in excess of 160,000 pages), we came to the conclusion that there were substantial problems with the design, conduct, reporting and availability of information from many of the trials.
The lack of good evidence demonstrating an effect on complications agrees with the conservative conclusions on both drugs drawn by the US Food and Drug Administration (FDA). The FDA only allowed claims of effectiveness of both drugs for the prevention and treatment of symptoms of influenza and not for other effects (including the interruption of person-to-person spread of the influenza virus or prevention of pneumonia). The FDA described the overall performance of both drugs as 'modest'.
The point being that even if the neuraminidase inhibitor developers were trying to conceal much of that negative data from the general public & from most MDs & researchers, FDA still had access to all of it. That was where Cochran got the data to draw their conclusions in the first case. And FDA did not allow them to make any claims about preventing complications, precisely because of all that negative data that they did review.
Of course, that doesn't mean the system actually works well. It would be far better if the results of those trials were made available to the MDs, researchers, and the public, not just to the FDA. Absent that, drug companies (& their employees) will absolutely try to get their drugs used as much as possible, even under circumstances that may not be appropriate.
In fact, there are efforts to try to 'force' publication of all clinical trials, whether they're run by drug companies or by academics. One interesting irony is that the drug companies currently seem to be better at this than academics. I don't have the citation to hand right now, but I believe I read about it over at Derek Lowe's In The Pipeline blog, if anyone wants to try to find it.