19.10.2023 Views

Health in a post-pandemic EU

As this magazine goes to press, the World Health Organisation has registered just shy of seven million global deaths due to COVID-19. It also has registered over 13.5 billion doses of vaccines administered. When it comes to health, it’s truly an extraordinary period to be alive. And to stay alive, if all works out as it seems to be doing.

As this magazine goes to press, the World Health Organisation has registered just shy of seven million global deaths due to COVID-19. It also has registered over 13.5 billion doses of vaccines administered. When it comes to health, it’s truly an extraordinary period to be alive. And to stay alive, if all works out as it seems to be doing.

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

KEEPING AN EYE ON AI-POWERED DRUGS<br />

If these tests are successful, AI could<br />

upend the drug discovery process. Researchers<br />

typically spend years sift<strong>in</strong>g<br />

through troves of data and test results to<br />

land on promis<strong>in</strong>g drug candidates <strong>in</strong> the<br />

lab – only for many to fail dur<strong>in</strong>g cl<strong>in</strong>ical<br />

trials.<br />

AI models could improve the odds by<br />

help<strong>in</strong>g researchers identify the right target<br />

<strong>in</strong> the body for a particular disease,<br />

then f<strong>in</strong>d or even create the right molecule<br />

to <strong>in</strong>teract with it, and lastly, predict<br />

which patients it could help. Pharma<br />

companies could then <strong>in</strong>vest <strong>in</strong> only<br />

the most favorable options, cutt<strong>in</strong>g out<br />

much of the early trial and error.<br />

It isn’t a surefire strategy: Exscientia’s<br />

OCD trial, for example, shuttered <strong>in</strong> 2021<br />

after fail<strong>in</strong>g to meet its targets. But ultimately,<br />

the goal is to br<strong>in</strong>g cheaper medic<strong>in</strong>es<br />

to patients faster, while br<strong>in</strong>g<strong>in</strong>g <strong>in</strong><br />

billions of dollars <strong>in</strong> profits.<br />

“Just from drug discovery to cl<strong>in</strong>ical<br />

development, that span is about five and<br />

a half years,” said Aarti Chitale, a senior<br />

<strong>in</strong>dustry analyst for health care and life<br />

sciences at the advisory firm Frost & Sullivan.<br />

“Some of the lead<strong>in</strong>g AI vendors<br />

are able to br<strong>in</strong>g that duration down to<br />

only about 18 months.”<br />

Money pours <strong>in</strong><br />

Investors have taken note of the opportunity,<br />

pour<strong>in</strong>g at least $10 billion <strong>in</strong>to<br />

startups target<strong>in</strong>g AI <strong>in</strong> early drug development<br />

s<strong>in</strong>ce 2019, while European<br />

pharma giants have announced major<br />

deals to expand their AI capabilities.<br />

France’s Sanofi, for example, <strong>in</strong>ked a<br />

$1.2 billion pact with Atomwise to sort<br />

through small molecules <strong>in</strong> 2022, while<br />

the British-Swedish AstraZeneca expanded<br />

its partnership with the United<br />

K<strong>in</strong>gdom’s BenevolentAI to hunt for<br />

treatments for systemic lupus erythematosus<br />

and heart failure, <strong>in</strong> addition<br />

We believe<br />

that there is<br />

huge promise<br />

from artificial<br />

<strong>in</strong>telligence<br />

<strong>in</strong> terms of<br />

medic<strong>in</strong>es<br />

development,”<br />

Peter Arlett<br />

to chronic kidney disease and idiopathic<br />

pulmonary fibrosis.<br />

As of 2022, there were nearly 270 companies<br />

work<strong>in</strong>g on AI-powered drug discovery<br />

around the world, with Western<br />

Europe serv<strong>in</strong>g as a grow<strong>in</strong>g hub, accord<strong>in</strong>g<br />

to consultancy firm McK<strong>in</strong>sey & Co.<br />

“We believe that there is huge promise<br />

from artificial <strong>in</strong>telligence <strong>in</strong> terms of<br />

medic<strong>in</strong>es development,” said Peter Arlett,<br />

head of data analytics and methods<br />

for European Medic<strong>in</strong>es Agency, which<br />

oversees pharmaceutical products for the<br />

European Union.<br />

Notably, the use of AI for drug discovery<br />

is generally considered low-risk because<br />

if a potential medic<strong>in</strong>e fails, it fails <strong>in</strong> a<br />

simulation, not a patient. Instead, AI<br />

likely poses a greater risk <strong>in</strong> later stages of<br />

drug development given the potential for<br />

ethical issues, risks of human biases to<br />

work their way <strong>in</strong>to algorithms or flawed<br />

data analyses that are used <strong>in</strong> a drug’s application<br />

for regulatory approval.<br />

Regulat<strong>in</strong>g pharma AI<br />

As pharmaceutical companies lean more<br />

heavily on AI across the therapeutic<br />

pipel<strong>in</strong>e, regulators are catch<strong>in</strong>g up to<br />

ensure these tools are used safely. The<br />

EMA published a draft paper this summer<br />

on the path forward for AI <strong>in</strong> drug<br />

development, and will hold a workshop<br />

<strong>in</strong> November to solicit feedback from the<br />

pharma sector and other stakeholders.<br />

“We see it as the start, the very start,<br />

of [AI] guidance and regulation <strong>in</strong> the<br />

pharmaceutical sector,” said Arlett, who<br />

is also co-chair of the EMA’s Big Data<br />

Steer<strong>in</strong>g Group.<br />

The reflection paper is set to be f<strong>in</strong>alized<br />

by late 2024, but it will likely “change<br />

significantly” before then based on external<br />

feedback, Arlett said. While the<br />

document won’t be b<strong>in</strong>d<strong>in</strong>g, it will offer<br />

a more concrete picture of the regulatory<br />

guidance to come <strong>in</strong> 2025 or 2026, which<br />

pharma companies will be expected to<br />

follow.<br />

Head<strong>in</strong>g <strong>in</strong>to the November workshop,<br />

Arlett said regulators broadly agree that<br />

they should categorize the risks of AI<br />

for different purposes so that “we don’t<br />

over-regulate where the use of AI may be<br />

just a background process, and not impact<br />

the benefit-risk balance for a medic<strong>in</strong>e.”<br />

Even so, he said regulators should have<br />

at least some access to drugmakers’ algorithms<br />

and the data used to tra<strong>in</strong> their<br />

models dur<strong>in</strong>g the discovery process, as<br />

well as <strong>in</strong>sight <strong>in</strong>to how algorithms are<br />

used to manage medic<strong>in</strong>es after they’ve<br />

been authorized – for example, if an algorithm<br />

helps to determ<strong>in</strong>e a patient’s<br />

<strong>in</strong>sul<strong>in</strong> dosages, the EMA wants to know<br />

how it works. The level of transparency<br />

that will be required is still up for debate.<br />

“Because the algorithm is learn<strong>in</strong>g, we<br />

will need probably to th<strong>in</strong>k <strong>in</strong>novatively<br />

as to how we oversee that,” Arlett said.<br />

“The exist<strong>in</strong>g framework, which is rather<br />

strict and very structured … may not be<br />

optimal for someth<strong>in</strong>g that’s as fast-mov<strong>in</strong>g<br />

as a learn<strong>in</strong>g algorithm.”<br />

The <strong>in</strong>dustry responds<br />

The pharma <strong>in</strong>dustry is keep<strong>in</strong>g tightlipped<br />

ahead of the November workshop,<br />

though executives from some<br />

lead<strong>in</strong>g firms, <strong>in</strong>clud<strong>in</strong>g Exscientia, have<br />

pushed back aga<strong>in</strong>st proposals to establish<br />

AI-specific drug discovery regulations.<br />

In a statement, the Brussels-based trade<br />

group European Federation of Pharmaceutical<br />

Industries and Associations<br />

said that new AI policies should “balance<br />

benefits and risks of AI while support<strong>in</strong>g<br />

and foster<strong>in</strong>g <strong>in</strong>novation,” and that<br />

“we already have a robust framework for<br />

handl<strong>in</strong>g statistical and predictive models<br />

and software that will apply to many<br />

uses of AI <strong>in</strong> medic<strong>in</strong>es development.”<br />

Regardless of loom<strong>in</strong>g changes to the<br />

regulatory landscape, drugmakers still<br />

need to figure out how to br<strong>in</strong>g AI-powered<br />

medic<strong>in</strong>es to market – and prove<br />

that they’re more beneficial than exist<strong>in</strong>g<br />

treatments. Ultimately, cl<strong>in</strong>ical success<br />

will be the key determ<strong>in</strong>ant for how<br />

widely AI is used for drug discovery, rather<br />

than time or cost sav<strong>in</strong>gs, as noted by<br />

the Boston Consult<strong>in</strong>g Group.<br />

Transform<strong>in</strong>g pharma<br />

The <strong>in</strong>dustry faces some other challenges,<br />

too. AI and mach<strong>in</strong>e learn<strong>in</strong>g models<br />

need robust, high-quality datasets<br />

to work well, and a central repository<br />

for drugmakers doesn’t yet exist <strong>in</strong> Europe.<br />

Further, most <strong>in</strong>vestment <strong>in</strong> the<br />

past five years has been <strong>in</strong> high-<strong>in</strong>come<br />

countries and focused on the profitable<br />

fields of oncology and neurology, leav<strong>in</strong>g<br />

<strong>in</strong>fectious diseases – which carry a much<br />

greater health burden globally – under<strong>in</strong>vested<br />

<strong>in</strong>, exclud<strong>in</strong>g COVID-19.<br />

Global economic uncerta<strong>in</strong>ty could also<br />

slow progress for smaller firms and startups,<br />

Chitale said. While venture capital<br />

fund<strong>in</strong>g for AI-powered drug discovery<br />

startups soared <strong>in</strong> 2021, reach<strong>in</strong>g $4.7 billion,<br />

that level was much lower <strong>in</strong> 2022<br />

and 2023, <strong>in</strong> l<strong>in</strong>e with a broader <strong>in</strong>vestment<br />

slowdown, accord<strong>in</strong>g to analytics<br />

firm CB Insights.<br />

Even so, <strong>in</strong>dustry players, academics and<br />

funders believe AI is poised to transform<br />

the pharma sector. In a recent survey, 84%<br />

of those currently us<strong>in</strong>g AI said they expect<br />

it to play a significant role <strong>in</strong> drug discovery<br />

over the next five years, compared with<br />

70% among those not us<strong>in</strong>g AI.<br />

In Europe, the use of AI isn’t limited to<br />

the early stages of research <strong>in</strong>to potential<br />

blockbuster medic<strong>in</strong>es. EFPIA, the drug<br />

<strong>in</strong>dustry trade group, said major pharma<br />

companies are “employ<strong>in</strong>g AI and ML<br />

approaches across the entire lifecycle of<br />

medic<strong>in</strong>es development” – from drug<br />

discovery and manufactur<strong>in</strong>g to safety<br />

monitor<strong>in</strong>g and beyond. •<br />

About<br />

Gaby Galv<strong>in</strong><br />

Gaby Galv<strong>in</strong> is a freelance journalist<br />

based <strong>in</strong> Amsterdam. Her work<br />

has appeared <strong>in</strong> Euronews, U.S.<br />

News & World Report and <strong>Health</strong>care<br />

Brew, among others.<br />

35

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!