That's not true. Drug companies can't just cherry pick trials to report to the FDA. The FDA is involved from very early on in the process. Furthermore, if the drug company is doing suspicious things like shutting down trials, the FDA will be even more strict in their requirements for approval.
The drug companies don't need to shut down trials, they just need to spread their bets sufficiently to ensure there are some winners. Sort of a VC model, except they have a clearer victory condition in FDA approval. Once a drug is approved, the pharmaceutical equivalent of retail shelf space is assured and the marketing machine can kick in to guarantee profitability. And the FDA has little incentive to prevent marginally effective drugs from reaching the market as long as no better alternative treatment exists. Which it won't, if big drug companies can keep churning out pharmaceutical 'hits' without much regard for efficacy. The system is a self-sustaining money machine for them.
If what you claim is true, and drug companies only get things approved by brute force, then we'd expect just 5% (for p-value of 0.05, as is commonly used) of trials to succeed at every level. But that's not the case.
