In the US you need two trials showing that a drug is effective, but you can have an unlimited number showing that it's not effective. Also it doesn't need to be shown to be effective against a disease, it can just be effective at reducing some number that correlates with a disease. And you don't need to show that it's effective in the longterm, only for up to 6 weeks - it could make the disease dramatically worse in the longterm, and that's fine even if the drug is meant for longterm use.
There are some rules trying to get all new trials registered, but the rules aren't followed at all and the FDA doesn't enforce them.
Wow, there is a long that is incorrect in your comment.
1. You are right that typically you need two registrational trials for approval and yes you can have many trials that show it doesn't work.
2. You are incorrect that you don't need to show it works against a disease. Sometimes you can show that you are impacting a biomarker (which I assume is what you mean by "some number"), but there must be evidence that changing the biomarker impacts the disease. Without that evidence you will not get approval.
3. You are incorrect you only need to show its effective for up to six weeks. If that were true, why would drug companies be running multi-year trials for cholesterol drugs? You typically need to show its effective for however long the drug is taken, but a follow-up period is also required.
2) C.f. statins. There are any number of books that talk about this. Overdosed America comes to mind, but you'd be hard pressed to find any recent book about the pharma industry that doesn't talk about statins.
3) C.f. SSRIs, and Robert Whitaker's book anatomy of an epidemic.
2) You are correct that new drugs can be approved by showing a lowering of LDL (biomarker), but that's only because the link between LDL lowering and cardiac risk reduction has already been demonstrated.
The best example I can give you right now are the drugs in development to treat Duchenne's muscular dystrophy. These drugs work by increasing the levels of a biomarker called dystrophin. The FDA recently came back and said "nope, not enough evidence that increasing dystrophin will improve patient's conditions, therefore we will not approve your drug".
As for the length of depression trials, it has been established that 6-week trials provide enough evidence to support an improvement in patients. [1]
Therefore, a 12-week trial is not necessary for all older patients; rather, the degree of improvement in the first 4 to 6 weeks identifies patients who are highly likely to benefit from continuing antidepressant treatment as well as those who very probably should have their treatment regimen altered at that point.
However, the depression drugs also run trials for long-term use: "Cymbalta is indicated for the treatment of major depressive disorder (MDD). The efficacy of Cymbalta was
established in four short-term and one maintenance trial in adults [see Clinical Studies (14.1)]."
There are some rules trying to get all new trials registered, but the rules aren't followed at all and the FDA doesn't enforce them.