Hacker News Reader

Comment by zepto | Hacker News Reader

Not really. You’d expect an organization like the FDA to be policy based.

You expect consistency from the IRS, because that domain is easy to fully articulate and constrain inside policy.

You don't expect it in the FDA because the trade-offs between safety, efficacy, urgency, clinical/statistical significance, and other dimensions are going to be rather specific to each silo of expertise/each product, and rather difficult to fully prescribe and encode a priori. So out of necessity, some amount of discretion and autonomy are going to be provided to each group of specialists, and you are bound to get inconsistency.

That doesn’t mean there can’t be a policy. Drug safety is based on science, after all.

There can be policy, it just can't be as prescriptive as you're asking for.

Consider the review process for paper submissions to scientific journals. You can't prescribe an exhaustive checklist to Nature reviewers and ask them to mechanically follow it. You need to give the reviewers some autonomy. And with that autonomy will come inconsistency.

> can't be as prescriptive as you're asking for.

What do you think I am asking for?

It sounds like you want a policy that sets a standard and for discretion to be explicitly mentioned in the policy, and that if one drug that's less effective is approved over another then policy violation is a candidate explanation.

Which is all probably already the case. The standard is probably rather loosely defined out of necessity, due to the reasons I suggested previously, and it follows that "inconsistency" will necessarily come from the part of the policy that allows for discretion. It doesn't follow that this inconsistency is bad, it could mean that in one domain or one product, certain considerations should get more weight. That's the point of the discretion, to allow the specialists to upweight and downweight those considerations that matter more or less for a given product.

So my point is that some inconsistency is to be expected, a subset of it is good, and no inconsistency would actually be a cause for concern.

What are we disagreeing about?

> Which is all probably already the case.

Remember - my comments about policy were a response to this:

> It probably makes more sense if we view the FDA not as a monolith but a collection of lots of individuals, some of which don't communicate or work together directly.

The comment I was replying to implied that the discrepancy was simply due to disparate groups of people not in communication.

Your view now seems to be the opposite - it is that they are in communication, and have a policy but that subject matter is complex and will legitimately generate the range of responses we see.

They're almost certainly not in communication but share an overarching policy which is intentionally vague enough to allow for some autonomy and flexibility. Hence expected inconsistency.

Sure, you'd expect and demand that, but that's the goal to strive towards not the reality of large organizations. Which is why permanent pressure and scrutiny is required on any government or corporate entity. That's democracy.

But in terms of analyzing these organizations it's an excellent point that we can't assume consistency. They are complex and contradictory.

True but how is that relevant? The point of policy is to manage inconsistency. In this case we would expect there to be a policy that sets the required standard, and to be able to see whether these decisions match the policy or not.

All organizations, even ones which are policy based, function like an artist's collective at some level - the human level. They all employ humans, who have to interpret those policies, and can interpret them in favor of or against something they personally have an opinion about.

If the difference is a matter of interpretation, we should be able to tell by comparing against the policy.

Policy has to be interpreted and enforced by people and that isn't always going to be consistent even with the same person let alone the fact that's its never the same person or group of people overseeing every decision.

True, but that doesn’t change the calculus. You’d expect a policy to set the required standard, and you’d expect a review process to identify major deviations.

Policy also has to be flexible when it comes to things with as many complexities as medication. Coming up with a universal standard for approval is extremely difficult. There can be issues both if it's too rigid or too lax. The role of policy in most cases such as this is supposed to be a guide, not a strict pass or fail for every possible criteria. Discretion from the regulators is needed to accommodate for cases that don't fit neatly into existing policies.

As a comparison, look at how nearly all of the big social media platforms all struggle with policy interpretation and enforcement around topics such as misinformation or hate speech.

> The role of policy in most cases such as this is supposed to be a guide

Sure, so we should be able to see what guidance was followed.

> Discretion from the regulators is needed to accommodate for cases that don't fit neatly into existing policies.

Discretion doesn’t mean policy can be ignored. Discretion would be part of the policy, and we should be able to see what discretion was used to come to these decisions.

I didn't say discretion means policy can be ignored, it's part of the interpretation of policy. Whether or not a policy is applicable for specific circumstances is part of discretion. It explains why policy is not universally applied and appears inconsistent.

Nothing about one drug being denied approval despite having more successful trials than another drug alone demonstrates that policy was ignored or that it wasn't considered.

> Nothing about one drug being denied approval despite having more successful trials than another drug alone demonstrates that policy was ignored or that it wasn't considered.

No, it doesn’t ‘demonstrate’ it, but it certainly makes it into a candidate explanation that can’t be dismissed.

Reply on news.ycombinator.com