Unfortunately, ALT is not a solution, because it's not accepted by FDA, except as a temporary measure.

Any drug submission that supports shelf life via ALT data will be expected to also say "and we're doing real time testing to support this", and FDA will expect the follow up real-time data.

Then perhaps the solution is for the FDA to allow HALT testing. Having said that, I imagine there is little incentive for pharma companies to willingly extend the shelf life of their products - perhaps the real solution is for the FDA to actually require HALT testing.

I think the real thing here is, "Better safe than sorry". It's fine to use things like HALT testing initially to say, "this drug should be effective for 8 years" so that folks can get it to market before 8 years have passed.

But if, in reality, that drug only lasts for a strong 5 years, after which time it loses effectiveness, especially in high humidity or with high UV exposure. We could catch this with real-time tests and adjust accordingly - before it actually affects the general public's health.

A good model really isn't good enough in medicine. Real data from clinical trials is needed, and nobody has an incentive to fund that.

The largest barrier to accelerated life testing right now is that we really don't know all the mechanisms that cause degradation (or understand their impact on stability). There are many lab instruments that already provide HALT metrics on how stable a drug sample is expected to be under different conditions. And the Arrhenius and Eyring equations are useful for empirical modeling. These tools are great for quickly identifying when formulations are unstable. But it's very difficult to say with certainty (at least in the pharma world) that something will be stable for X years unless you actually test it for X years.

For the vast majority of manufactured products, HALT makes sense because any uncertainty that remains after testing is not likely to harm users. But with some of these drugs, any uncertainty, even after HALT, could cause major problems for users. And it seems that the FDA is unwilling to accept that risk. But as far as I know, the FDA is actively seeking methods for performing HALT that produce accurate and repeatable results.