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.