[Dysseus]

Could you expand a bit on that?

I'm having a hard time making the connection between declining efficiency and there necessarily being something broken about the science.

To expand just a bit: Scannell seems to be primarily focused on the disease model as the root issue here. This is curious to me. My understanding is that drugs fail primarily in two ways:

1) It's ineffective - fails to treat the disease. Which is partially covered by model validity, but also impacted by pharmacokinetics and distribution. Essentially your molecule can work, it just can't get where it needs to go in a high enough concentration to make a difference.

2) It's unsafe - your molecule is toxic either acutely or long term.

The data[0] I'm aware of indicates that these issues occur with roughly equal frequency. (My assumptions being: failure in Phase 1 trials are an issue of safety, failure in phase 2 can be caused by safety or effectiveness). Which for me calls into question the focus solely on good disease models.

[0] https://www.nature.com/articles/nrd3078

[refoundable]

The FDA does not allow Pharma companies to create safe and slightly less effective drugs than the best-in-class, even if such drugs were 10-100x cheaper and would thus alleviate most people's complaints about outrageous drug prices.

Scannell's observations are focused on the problem of finding drugs that are more effective than the current best-in-class, the so-called Better than the Beatles problem.

This turns out to be a really hard search problem, given for instance that animal models make unreliable and poor substitutes for humans beings, or that you're generally not allowed to base drug approval on small but highly representative patient populations (with some exceptions). There are even more technical problems I won't get into here (see Question 9 in the interview).

In short, I see no reason why we couldn't be getting slightly less effective but way cheaper drugs today if the FDA was willing to slightly relax its use of the precautionary principle. But in order to keep pushing the frontier of drug efficacy, we'll need new technical breakthroughs to help us solve the search problem.

[Dysseus]

First, really appreciate the engagement here. This is a hugely important problem and this interview and your presentation of it is a great contribution.

I think one of the things that gets lost when we talk about Eroom's law is that the original data points were established before congress passed the Kefauver-Harris amendments in 1962 which set standards for clinical trials, iNDA process, and basically required that drugs show efficacy before they could be marketed.

An important part of those amendments is they made the drug companies go back and review the 4,000 drugs already on the market and provide evidence on their efficacy. It took FDA a long time to work through that backlog, but when they did:

"In January 1968, the Drug Efficacy Study panels finally reported their conclusions to the FDA. They had reviewed over 16,500 therapeutic claims for 4,000 pre-1962 drugs. Only 434, about 12 percent of those examined, delivered on all their promised claims. Seven hundred and sixty-nine were marked as 'ineffective'" [0].

I bring that up to say two things:

1) Our baseline in examining Eroom's law is a bit skewed because standards have been going up since the graph begins.

2) We should be careful in how we change those standards. Many of them were bought with patients lives.

I need to go now, but I do want to address your comment on pricing later.

Thank you again. Really great work.

[0] Pharma - Gerald Posner - Pg 224. https://www.amazon.com/Pharma-Greed-Lies-Poisoning-America-e...

[refoundable]

No disagreement on the skewed stats. And I won't dispute the need for a minimum standard of efficacy.

However, if anything, the problem that we've had over the last 60 years is setting the standard of efficacy too high. This caused the Better than the Beatles problem, and placed Pharma companies in an impossible situation where they have to run faster and faster just to stay in the same place.

Moving forward, if we're not going to lower the efficacy standard, then the questions that Scannell raises about model validity and the search problem writ large become even more pressing and important.

On the other hand, if we're not going to solve the search problem, then the pragmatic solution would be to lower the efficacy bar somewhat so Pharma companies can at least create slightly less effective but way cheaper drugs.

In theory, we should be able to do both. In practice, we're likely to get one way before the other. The current path we're on of fewer drugs and higher prices is simply unsustainable.

[Dysseus]

This is helpful. Let me ask a question.

Why do you think that cheaper development costs will ultimately lead to lower drug prices?

I think we would both agree that drugs are not priced based on cost. Even if R&D were 10X cheaper - all else equal - prices are not coming down.

So what is the mechanism you think will lead to lower prices given lower costs of R&D?

Another question: Maybe as a patient (or a doctor), I like the better than the beetles problem. I don't need a thousand drugs on the market to treat every indication. Especially ones that have not passed a high bar for efficacy. I need a few drugs, preferably outside of patent protection, with enough diversity in structure to avoid specific toxicity effects.

If lowering the efficacy standards doesn't get me treatment for new indications, why would I want it? It's not clear to me that solving better than the beetles gets me to new indications. Presumably if I'm in a regime where BTTB applies, I've got a drug for that.

[refoundable]

> Why do you think that cheaper development costs will ultimately lead to lower drug prices?

If the cost of drug development were to drop by a factor of 100-1000, say from the $3B it costs today to the $3-30M range, that would likely unleash a massive wave of decentralization in the industry that would make it very hard to justify outrageously high drug prices for long.

However, short of a large drop of that magnitude, I agree that we'll probably continue to see Pharma companies charging as much as they can get away with and the political climate allows.

> I don't need a thousand drugs on the market to treat every indication. [...] If lowering the efficacy standards doesn't get me treatment for new indications, why would I want it?

It's true that the Better than the Beatles problem rewards consumers with cheaper drugs, but on the flipside, it robs them of the benefit of newer drugs.

As a result, all the technical capabilities that the Pharma industry gained into developing drugs that are less toxic or better targeted do not get translated into consumer benefit when the BTTB problem is in the way.

Moreover, once an area of drug development stops seeing new commercial success in decades, hysteresis sets in and can drive Pharma companies out of such areas, which has largely been the case with anti-infectives since the 1980s. (See Question 11 in the interview)

[Dysseus]

Again want to say I really appreciate how thoughtful you are about this.

I just want to offer a few of my thoughts because you've been so generous with yours. So you know, I work in the industry. More on the biz side than the scientific. And I am intensely interested in this issue.

100-1000 times cheaper I don't think is achievable in the medium term, if ever. As long as we're running clinical trials I think costs will stay constrained to about a 5x decrease. A 100-1000x reduction, I think, implies a world where we can largely simulate the impact of drugs without resort to in vivo testing. A future I believe in, but one that has a lot big unknown technical obstacles to overcome - like entirely new fields of science and engineering. I'm thinking here on a 50-100 year horizon.

This would also entail a huge change for the regulatory environment, but I think we're so far in the future in this scenario that it's hard to predict what that will look like.

I think a 2x reduction is possible in the near-medium term. If we can change the probability of technical success for trials through better tox models - not testing - that creates an enormous amount of value. The stuff people are doing to design better molecules pales in comparison to doing this. I think the math and know-how exists to do it now - the challenges are more institutional - Who pays? Who provides data? Lot's of coordination problems between actors who are in economic knife fights with each other, though people are trying.

Data problems too. Pharma companies are terrible at databases. It's a mess.

Also, it's better to solve tox because it's not disease specific. The effort invested in building the model (or models) will be distributed over many development life-cycles and would remain valuable after a good drug for an indication is invented.

This is a huge win for the world because it makes diseases that are important but not economic easier to justify. The phrase 'important but not economic' churns my stomach, but there it is.

None of this will fix pricing. Not as long as the patent structure stays the same. If you can legally do it, people will do it. Doesn't matter what it costs. Coke is still charging you two bucks for something that costs them cents to produce, and they don't have a patent.

If you want a near term future with lower drug prices (as I do), I fear your route lies through congress. I think it can happen, but I'm an optimist.

Thanks again for your thoughts and your work. It's good to see other people pulling in the same direction.

[nn3]

As I understand most of the cost of a new drug is the testing and certification. It's unlikely that a new drug with the same testing/certification requirements is significantly cheaper than an old one.

[tehjoker]

The costs of drugs is usually profit seeking. Many drugs can already be produced and delivered cheaply if the economic system would allow it. Trying to patch the drug instead of the economics just gives people worse drugs.

[refoundable]

I see no reason why we can't do both. We obviously need better drugs for the diseases we haven't yet cured, and we need cheaper drugs for the diseases we've already cured. The two are related yet distinct problems.

Incidentally, Scannell wrote one of the best teardowns of how drug pricing actually works: https://www.forbes.com/sites/matthewherper/2015/10/13/four-r...

[light_hue_1]

Two things, one is that toxicity is overwhelmingly the failure mode and the other is that 1 (effectiveness) and 2 (toxicity) are actually closely related not independent.

Failures are overwhelmingly in Phase II, https://en.wikipedia.org/wiki/Phases_of_clinical_research And we know why they happen, we don't need to guess. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6609997/#:~:tex... Mostly it's that we discover previously unknown toxic effects. Between Phase I failures due to toxicity, Phase II failures which are half due to toxicity, and Phase III failures, and Phase III trails failing 17% of the time due to toxicity https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6092479/ it is overwhelmingly the biggest problem. Only half of Phase III trials fail due to a lack of efficacy.

At the same time effectiveness and safety are not independent variables. Safety often a matter of dose. The dose required to achieve a clinical effect may turn out to be unsafe.

The issues goes back largely to model validity. That our models don't allow us to accurately establish safety margins and dosages, in addition to often being misleading about the wider effects that drugs will have.

[Dysseus]

I think we're saying the same thing. The disease model itself is necessarily limited. Sure you have a molecule that hits the target, great. But the struggle comes from being able to get that molecule to target in vivo without causing toxic effects.

Maybe we're just missing on how we're using words. I don't see how having a better disease model necessarily gets you to a better place on this. Sure, you can get better SAR and can decrease the dose. But as you point out, dosage is not just a function of SAR.

Having better tox models seems like the highest value, albeit very difficult, route here. Which to me is a separate, more general, problem than a specific disease model.

[refoundable]

> Having better tox models seems like the highest value, albeit very difficult, route here. Which to me is a separate, more general, problem than a specific disease model.

Agreed. Better toxicity testing would go a long way towards solving some aspects of the search problem.