It should be pointed out that looking at the average cost of developing a drug is misleading, since one has to include the cost of all the drugs that failed to make it to market. One also has to include the money spent by small companies that failed and were not bought out, not just the money the big companies spend buying the successful ones.

This is critical, and exactly the sort of thing someone will gloss over, intentionally or not. The numerator is the total cost of developing ALL the drugs, even (especially) the failed ones, but the denominator is only those drugs that are successful.

On the contrary, often pharma proponents will gloss over the fact that most “failures” are discovered and canned at a small fraction of the investment of getting a drug to market (when it’s obvious it won’t do what you need or has other challenges).

Actually not in the contrary - you are right. It’s just that the failures often cost a small fraction. They don’t get to 95% testing and approval before “nope, not even close”.

I understand 60% of drug candidates fail in phase 3, the last and most expensive phase of testing.

There are of course lots of chemicals ruled out early, but that's before you have something you'd call a drug.

No, that's not quite accurate - the phases are effectively additive.

37% fail in Phase 1. Of those that make it through, 69% of those fail in Phase 2, and of those, 42% fail in Phase 3.

So, of 1,000 possibles, you have 630 make it through Phase 1, and 195 that make it through phase 2, i.e. 80.5% of your drug candidates didn't even make it to Phase 3, that "most expensive phase".

A more accurate phrasing would be that 42% of the drugs that made it through Phase 2 fail to make it through Phase 3.

Supposing the cost ramps up exponentially at each phase, e.g. it costs $10 million to get to Phase 1, $100 million to get to Phase 2, and $1 billion to get to Phase 3, then we see the total expenditure for 1000 possible drugs as:

    $  3.7 billion for Phase 1 failures  (370 drugs)
    $ 43.5 billion for Phase 2 failures  (435 drugs)
    $ 82   billion for Phase 3 failures  ( 82 drugs)
    $113   billion for Phase 3 successes (113 drugs)

This sums to a little over $242 billion spent against 113 successful drugs, or about $2.14 billion per successful drug, or more generally, accounting for failed drugs, the full cost of a successful drug is a little more than twice what was directly spent on its development.

We do have some shaky and hard to interpret data.

Published estimates of trial costs from a 2011 systematic review ranged over an OOM.[1]

A 2017 report focused on 7 top-20 companies and 726 studies from 2010-2015 found " median cost of conducting a study from protocol approval to final clinical trial report was US$3.4 million for phase I trials involving patients, $8.6 million for phase II trials and $21.4 million for phase III trials". [2] These are not all that far off from another 2016 study on cost drivers of pharma clinical trials in the US using means and breaking down costs by therapeutic area. [3]

Plugging the first study's numbers into your 1000-drug profile, we'd have:

$ 1.3 billion for Phase 1 failures (370 drugs)

$ 3.7 billion for Phase 2 failures (435 drugs)

$ 1.8 billion for Phase 3 failures ( 82 drugs)

$ 2.4 billion for Phase 3 successes (113 drugs)

That sums to $9.2 billion spent against 113 successful drugs, or about $80 million per successful drug. This implies the full cost of a successful drug is almost 4x what was spend on its development.

One limit here is we're working with medians, not means, and I wouldn't be surprised if this is an underestimate of clinical trial costs.

Roche has had pretty stable net income (profit) of $9.2-$15.2B/year from 2011-2023 against revenue from $49.9-$72B/year in the same time period. Using this estimate, ignoring inflation, if they ran 1,000 clinical trials per year it would account for a maximum of about 18% of their costs and they'd get 113 new drugs out of it annually.

Obviously that is not what's happening: there are an average of 53 FDA new drug approvals per year across the entire industry. If Roche was the only pharma company running clinical trials, the total cost of those trials would be more like $4B, so a max of about 10% of their annual costs. In reality this estimate makes it seem like it must be substantially lower.

The Congressional Budget Office[4] says total pharma R&D spending in 2019 was $83 billion. With 53 new drug approvals per year on average, that implies about an average cost of about $600 million per drug in R&D spending, compared with the $80 million estimate obtained above.

So this makes it sound like running clinical trials account for only about 10% of total R&D spending. Given that Roche's costs alone look to be in the tens of billions per year, as compared to $83B or so annually for R&D across the industry, it also looks like R&D is only a part of the story on cost drivers for pharma companies. Google is not being helpful on this question (almost all the conversation is on R&D cost, it seems), but my guess is it's costs of manufacture, legal, sales, etc.

[1] https://www.sciencedirect.com/science/article/pii/S016885101... [2] https://www.nature.com/articles/nrd.2017.70.pdf [3] https://pubmed.ncbi.nlm.nih.gov/26908540/ [4] https://www.cbo.gov/publication/57126

> it also looks like R&D is only a part of the story on cost drivers for pharma companies ... but my guess is it's costs of manufacture, legal, sales, etc.

Marketing. At least 7 of the top 10 pharma companies globally have marketing as a multiple of R&D for their spending (sometimes up to 7x). IIRC, at the other 3, it still exceeds R&D, less egregiously.

The big issue with "Marketing" spend is that though these numbers are global, there are only two countries in the world where you can advertise prescription medicines to consumers: the US, and New Zealand (and the latter, if I recall, is trying to phase it out, and only allowed it after being bullied by the US on a Trade Agreement).

So you end up with "US Marketing spend at many pharmaceutical companies grossly outpaces their global R&D spend" (and while a not insignificant portion of R&D happens in the US, most of those companies also have a notable R&D investment in Europe).

Marketing wouldn't go to zero without that, of course, but it'd be a huge sea change.

Certainly I do not mean to imply that Phase 1 and Phase 2 failures are cost-free. But it is challenging to measure. As seen elsewhere in this thread, Gilead essentially included the acquisition of another company who had a whole retinue of drugs and product lines as "R&D" for Truvada, I believe. That is creative accounting that would not pass an audit or SEC filing, which is why Gilead only counts it as an R&D cost in their press releases...

It certainly should count; that company didn't get delivered for nothing by the Drug Discovery Fairy. And even more: the companies that didn't get bought by Gilead should also count, since the funders of all those small companies could not tell ahead of time which would succeed enough to be bought out.

No, it really shouldn't. If I acquire a drug company that has, say, 100 patents on a suite of drugs for $1B (just using nice round numbers), I don't get to say "I spent $1B on "R&D" for 1 drug" as a sunk cost.

R&D is a sunk cost. Presumably acquiring an active company with a portfolio is an investment.

> One also has to include the money spent by small companies that failed and were not bought out, not just the money the big companies spend buying the successful ones.

If you're looking at the total amount spent by "the economy" (drug development costs X% of GDP), sure. If you're looking at "why are drug prices so high", it probably doesn't make sense to to include costs funded from other places (which in this example I assume would be research grants ie taxes, and venture capital funds).

For the private parts of development, the costs are absolutely priced in. A large drug company needs to amortize the cost of all development attempts, not just the successful ones. Private investments into smaller firms price in a very large chance of failure, so the cost of capital is quite high.

That's not quite right. A drug company with a new product will charge whatever the market will bear. What the costs do is control the scope of the industry: if profits are high, the industry expands to try more kinds of drugs, stopping when the attempts on the margin are just profitable enough (on average). If profits are not expected to be adequate, the industry contracts.

That's close to what I tried to say.

Perhaps you prefer: A company must think it's likely that they'll have a good return on all development costs, not just the costs of drugs that happen to be successful, to continue to invest.

> if profits are high, the industry expands to try more kinds of drugs, stopping when the attempts on the margin are just profitable enough (on average).

Of course, something like pharmaceutical products, with exclusive sales of specific products, few sellers, strategic conduct relative to other industries (insurers), and heavy regulatory influence is not guaranteed to converge to normal profit.

I'm not going to invest a drug company with a 90% chance of failure unless I can expect to get a 10x return if it succeeds.

The problem with this argument is it assumes the cost of a failure is the same as the cost of success, which it cannot be: the successful drug has to go through more rounds of testing and approvals than a failure.

In reality many failures are early or first round failures. Not free but a small fraction of the price of getting to market.

So to you example a 90% failure rate may only require a 2x or 3x return on your successes to “break even”.

Clinical trial failure rates (or inversely success rates) have been analyzed before.

https://www.nature.com/articles/nrd.2016.136

"They found that the probability of success was 63% in Phase I trials, 31% in Phase II trials, 58% in Phase III trials and 85% during the regulatory review process"

42% failure rates in phase 3 is enormously high. By then you've pretty much spent 90%+ of all the cost of getting a drug approved.

But it's 42% of 19%. So out of 1,000 drugs, you're looking at 805 being ruled out before you even get to that "most expensive phase", which is my point. At Phase 3, you're looking at 113 succeeding, so you're "only" eating the really expensive[1] costs of Phase 3 for 82[2] of 1,000 attempts.

[1] Which isn't to say there's zero cost for Phase 1 or Phase 2, but it's a lot lot less than Phase 3 trials.

[2] 1,000 drugs, 63%, 630 of which make it through Phase 1. In Phase 2, 195 drugs, 31% of 630 succeed and make it through to Phase 3, and then 82 drugs (58% of 195) make it to regulatory approval.

Right, but with the cost distribution for clinical trials, costs increase by 4x in phase 2, then 8x in phase 3 (relative to phase 1).

So despite attrition that reduces candidates by 9x by phase 3, costs have increased by 8x.

[1]https://www.sofpromed.com/how-much-does-a-clinical-trial-cos...

> "A new study in 2020 estimated that the median cost of getting a new drug into the market was $985 million, and the average cost was $1.3 billion, which was much lower compared to previous studies, which have placed the average cost of drug development as $2.8 billion.[4]"

PrEP repurposed Truvada, an existing blockbuster drug that had already reaped immense profit for Gilead for use in HIV treatment by the time the trials for PrEP began. The trials for PrEP were funded by the government, not Gilead. Gilead, however, got to retain all profits earned from PrEP.

A significant portion of the cost is the drug trials. Excedrin Extra Strength and Excedrin Migraine have identical formulations, but IIRC Bayer spent $300m on FDA approval for migraine treatment, which is why the migraine variant continues to be more expensive.

> A significant portion of the cost is the drug trials. Excedrin Extra Strength and Excedrin Migraine have identical formulations, but IIRC Bayer spent $300m on FDA approval for migraine treatment, which is why the migraine variant continues to be more expensive.

Your analogy would only be relevant if the US government paid $300M for the FDA approval and Bayer got to pocket 100% of the markup.

Did Gilead fund the R&D? There's a lot more to developing a new drug than just trials (though I think Gilead should have foot the bill for the trials too).

I don’t know if it’s the case here, but very, very, very often in biotech you’ve got the primary foundational research happening at university labs funded by grants, and it’s the productionization of the research (and then clinical trials, etc) that are what the biotech companies are doing. I’m not sure where that shifts the “who deserves what” conversation, but without university research labs, there’s no pharma industry.

If the university owned the IP, then its value should have been reflected in what was bid for it.

If the knowledge was not restricted by IP law, then any drug company could use it, and compete for new drugs based on it. As such, it would not provide any of them with a competitive advantage, and so would not be reflected in what they could charge.

What universities typically produce is not a chemical that can serve as an actual drug, but is only a starting point for a long and expensive process of producing such a chemical. And then, it's often found that the target of the class of potential drugs isn't actually a good one. One can't determine that until drug candidates are available to test on real patients.

> What universities typically produce is not a chemical that can serve as an actual drug, but is only a starting point for a long and expensive process of producing such a chemical

Remember you need to include all the failed attempts at finding useful things at university labs to see how much governments spend on research (just like you did failed pharma attempts), and if you add that up you see governments actually contribute a massive part of the cost to bring medicines to market.

What they produce is necessary to even begin the work pharma does, currently it is basically a gift from the people to the pharma industry.

"without university research labs, there’s no pharma industry." - I think you have it exactly backwards: Without the pharma industry, there's no medicine. Good research goes nowhere if you can't bring it to market.

The pharma industry COULD do their own foundational research, but the university system cannot bring a drug to market.

> The pharma industry COULD do their own foundational research, but the university system cannot bring a drug to market.

You can't use an "in theory" argument for one side but not the other.

In theory governments could bring medicine to market, in practice they don't/can't.

In theory pharma industry could do foundational research, but in practice they don't/can't.

If there’s no pharma industry?

You act like the solution would be “oh well, no meds for anyone then!” and not “let’s expand university programs to meet that need”.

> The pharma industry COULD do their own foundational research

Citation neeeded - have they ever done so? Would the shareholders accept it? Would they be able to manage borderline autistic PHD types detached from reality, and would these scientists want to work there?

Pharma companies are chock full of PhD types, as are the tech companies and Wall Street.

What companies don’t have is PhD students. They are numerous, smart, very cheap, and work very hard.

PhD types doesn't mean they make foundational research.

> They are numerous, smart, very cheap, and work very hard.

Yeah, this is a good reason to hire such people, but they generally don't do foundational research work at companies, and if they do it is extremely narrow.

Just like government work the problem is the environment, Governments hires management and planning types just like companies do, but that doesn't mean they can scale up like a company can. Same with foundational research, private companies aren't a good place to do that.

Really? Outside of pharma businesses do all kinds of time consuming, speculative and expensive foundational research.

I'm not quite sure why pharma is the exception.

It's difficult to overstate just how menial, fiddly, difficult, risky, time consuming and unclearly profitable foundational biomedical research is. A research project could easily take 4-5 years, have a 5% likelihood of success, and have no clear monetizability, yet end up being a groundbreaking foundational result and necessary to investigate.

In other fields, either there's some tangible hope of profit down the road, or at least you can attract talent and prestige. Not so much here.

i've always viewed big pharma as like pre-internet record labels. they pick up talent (that often comes from bohemia aka government funded research), vet it, run the trials and put up the money, do the engineering to deliver it at scale and then market it.

That’s also like any endeavor with tech.

The direct role that university research plays in drug development is overstated. The majority of cost and difficulty in pharma is _drug development_ not _drug discovery_. Pharma can do the discovery and the development, academics can only do the development. Absent academia, we'd have less drugs. Absent pharma we'd have no drugs.

Academics focus on drug discovery because it's better aligned with academic incentives and timelines (see this commentary for a brief description [0]). Drug development costs (including clinical trials, extensive and repeated med chem, etc) are borne mostly by drug companies.

Fair data on this is hard to come by because the two main sources have clear conflicts of interest (academics and pharma industry publications). One study Derek covered before (data from 1995-2007) shows only 24% of drug scaffolds were first found at a university and transferred to a biotech or pharma for development [1]. You can break this down further to highlight any story you want to support ('university ID'd drugs more innovative' vs. 'pharma ID'd drugs help more people') but they key point is that combining all the US research leads to only 24% of drug scaffolds that make it to market.

I think everyone acknowledges that outside of finding the scaffolds and the basic biology, pharma is paying the vast majority of clinical trial costs. [2] gives a figure of total NIH funding of clinical trials at 10% of overall (e.g. pharma covers 90%).

I think an argument could be made that the NIH training grants (which pay grad students in the biomedical sciences) subsidize the work force substantially, and might have a higher impact than direct research grants. I couldn't find quantitative data on this with a quick search, but I think this is often overlooked in the discussion.

Finally, a less quantitative pieces make me think the impact of the NIH/government funding is overstated even given the above numbers. In my own field (microbiome), academic research has been almost inimical to the production of quality drugs. For every disease there exists a paper suggesting that a certain gut microbe changes the likelihood/severity/X about that disease. Academic labs have incentives to publish significant results fast, and in the microbiome this has led to a) abysmal signal to noise ratio with very high likelihood of failure to replicate, and b) an epistemic closure about what types of microbiome data matter and how they should be pursued as drugs that is totally divorced from the reality of how drugs are developed. Much of the knowledge base is polluted by low-quality research that has been done for the purpose of publishing. While the NIH spends ~40 billion a year on external research grants [3], I think you have to heavily discount this for the amount of just pure "grad student needs to graduate gotta publish" material that gets produced.

[0] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10812233/ [1] https://www.science.org/content/blog-post/where-drugs-come-n... [2] https://www.fiercebiotech.com/research/report-industry-not-n... [3] https://www.nih.gov/about-nih/what-we-do/budget

The usual story is that academia finds an interesting mechanism to produce the desired effect. Though occasionally this is done by industry instead.

Then industry turns that into a specific molecule that can enter the human body in a standard way and doesn't produce too many side effects.

Then industry figures out how to produce that molecule at scale reliably in a sufficiently pure form.

And at the same time industry is shepherding the drug through clinical trials.

Also the government, aka the public, subsidizes a lot of those costs

Talking of studies, from the same wiki

> A 2022 study invalidated the common argument as is for high medication costs that research and development investments are reflected in and necessitate the treatment costs, finding no correlation for investments in drugs (for cases where transparency was sufficient) and their costs.[20][21]

The Wikipedia editor was a bit naive to think such a basic study could invalidate that whole claim. They measured the correlation between the list price, adjusted for use amount, and development cost. As far as I can tell they didn't take into account number of customers each drug would have, how long the drug would stay on the market before profits are cannibalized by competitors (see e.g. Wegovy), and definitely not the cost of failed drug development.

These original $3B numbers are highly misleading, to the extent that I deem them to be bordering on a straight up lie.

See: https://news.ycombinator.com/item?id=18693177