If even in those circumstances they're not getting strong signals that their drug is effective then changing the experiments until they show what you're hoping

If even in those circumstances they're not getting strong signals that their drug is effective then changing the experiments until they show what you're hoping for is not the right approach. Inconclusive results are not conclusive negative results. Granted, the drugs here have small effects which can be masked by noise -- they cure only a small fraction of people who take them. Drugs are far from curing everyone all the time, and that's still valid. A couple percentage points translates to thousands of lives, which is better than nothing (if costs are taken into account too).