For many medicines, you really can't separate efficacy from safety. As an extreme case, take for example cancer drugs. They typically have high side effects and every year, a certain proportion of the people taking them, die from the side effects. That is an acceptable risk because there is evidence that they are efficacious in treating cancer. If they are not efficacious, that is a completely unacceptable risk.

Less extreme are antibiotics which can also have bad side effects, but are useful drugs because they actually kill bacteria. If they did not, the side effects would be unacceptable.

Should a doctor be making that decision, or should the government?

Or should the government make that decision for on-label uses and doctors make that decision for off-label uses? That's the status quo. Whatever argument you have for one half of that is an argument against the other half.

Are you suggesting when a new drug hits the market doctors should just try it out on a few patients to see if actually does what it says it does? If only there was some way this sort of trial and error could be performed before a drug was widely available, in a controlled environment...

" If only there was some way this sort of trial and error could be performed before a drug was widely available, in a controlled environment..."

This is exactly how doctors determine what works, with or without the government. See the off-label discussion in the link above.

I'm familiar with the off-label discussion, but what number of these off-label uses were discovered by treating coinciding conditions in patients? How many were last ditch attempts when other treatment options either don't exist or failed?

I'm not suggesting that current regulation by the FDA isn't flawed, or that it doesn't need an overhaul, but I think it's clear that there should be some proof that efficacy studies were done before a drug becomes widely available, especially when the potential damage done by a failed treatment is so high. The current opioid crisis has completely eroded any faith that the pharmaceutical industry is capable or willing to regulate itself.

so, in effect, subsidize pharma via selling placebo effect?

No.