Monday, June 24, 2024

Billionaire Jim Breyer Is Backing This Health AI Startup With A Radical Approach To Data

Must read

Doctors usually counsel people with diabetes not to abruptly switch or stop taking their regular medications before a surgery. That’s why David Klonoff, an endocrinologist who treats people with diabetes, was surprised when the American Society of Anesthesiologists issued guidance last summer saying patients should stop taking GLP-1 agonists, the new diabetes and weight loss drugs like Wegovy and Ozempic, before any scheduled surgery. “I was concerned that the downside of that decision might be worse than the upside,” Klonoff, the medical director of the Diabetes Research Institute at Mills Peninsula Medical Center in San Mateo, California, told Forbes.

The drugs slow stomach emptying and the anesthesiologists were worried that patients on GLP-1 agonists might vomit during surgery, swallow some of it and end up with pneumonia, among other issues. The national anesthesiologist group decided the risks of potential complications outweighed the benefit of staying on the drugs. But what irked Klonoff was that there was little evidence to support the position, aside from a handful of case reports. “There was no data,” he said.

In medical school, doctors learn that randomized controlled trials are the gold standard to inform decision-making around how they should treat patients. But every day doctors have to make life-or-death decisions in situations that can’t be found in a clinical trial. They instead have to rely on training, intuition and experience to triangulate what’s best for an individual patient. “There is not enough time and effort and money to do a randomized control trial for everything,” said Klonoff.

That’s why he joined the medical advisory board of Atropos Health, a startup Klonoff believes has come up with the next best alternative. Atropos is tapping into 200 million patient records to help doctors make decisions based on what’s happening in the real-world, outside of these carefully designed clinical trials.

Say there’s a 45-year-old woman with bladder cancer – would she have fewer complications and better outcomes with radiation treatment or bladder removal? Atropos’ software combs through millions of those records to find similar patients and how they’ve fared and generates a statistical analysis in a matter of days – and, now with recent advances in generative AI – minutes.

On Thursday, Atropos announced a $33 million Series B round led by Valtruis, the venture growth arm of private equity firm Welsh, Carson, Anderson & Stowe. Forbes estimates the company’s valuation to be around $250 million; it has raised $54 million to date. Existing investors Breyer Capital, Emerson Collective and Presidio Ventures participated in the round, along with new strategic investors Cencora Ventures (formerly AmerisourceBergen), McKesson Ventures and Merck Global Health Innovation Fund.

Atropos has a radically different approach from most other health data companies. It doesn’t move healthcare data around to create one massive dataset. Instead, it runs separate queries on each data pile, or node, individually, which is known as a federated data model.

The uniqueness of the approach appealed to billionaire venture capital investor Jim Breyer, an early investor in Facebook, who’d realized how hard it is for doctors to share data across different hospitals when navigating both of his parents’ cancer diagnoses. “There are huge challenges in allowing medical professionals, be they great doctors or nurses, to have access to a full suite of data, which currently is as siloed as any industry that I know,” he told Forbes. Breyer joined Atropos’ seed round and board in 2020 and has participated in every financing since.

The startup is named for one of the three fates in Greek mythology whose scissors have the power to end or extend the lives of mere mortals. The hope is that by generating evidence to help clinicians make better decisions, the company can also help improve outcomes and extend people’s lives.

“Every time we try and aggregate all the world’s data into one place … the largest dataset always wins.”

Brigham Hyde, cofounder and CEO, Atropos Health

Before starting Atropos, its CEO Brigham Hyde had spent a decade building and running companies that worked to aggregate healthcare data to sell to pharma companies for research. Companies pull this data together with the goal of creating a complete patient record, which is crucial to helping develop new drugs. For example, if a patient got a drug from one doctor’s office and surgery at a different hospital and a cancer diagnosis at another hospital, the data is only valuable if all the pieces of the puzzle are together.

But as Hyde discovered, the status quo, where companies simply buy up more and more data, is expensive and time-consuming. There’s risk to patients’ privacy, too — the more healthcare data that’s shipped around, the more risk there is for all of the parties involved under the federal privacy law known as HIPAA. But crucially, Hyde realized there was one major flaw. “Every time we try and aggregate all the world’s data into one place,” he told Forbes, “the largest dataset always wins.” Essentially, the dataset will be biased towards the entity that supplied the most data.

Hyde cofounded Atropos with two experts in healthcare informatics: Nigam Shah, chief data scientist at Stanford Health Care, and pathologist Saurabh Gombar. The trio built software that runs its analyses on each individual data source, in this case, each healthcare institution, and then combines the results in what’s known as a meta-analysis. Using this technique, the company can provide doctors and researchers with retrospective studies based on how real people responded to different therapies. And because there’s no moving data around, there are fewer privacy concerns. Now, a researcher or company who wants to conduct a retrospective analysis doesn’t need to buy the data, said Hyde: “They can just buy the answer.”

So far Atropos has generated more than 10,000 of these retrospective studies. This includes working with Klonoff, the endocrinologist, to figure out if there were increased risk of complications for people on GLP-1 agonists who had nine common surgeries. An analysis of over 13,000 patients with diabetes – around 2,200 of them on GLP-1s – published this month in Diabetes, Obesity and Metabolism journal, found no evidence to support stopping the drugs before surgery.

Klonoff said he was swayed by the results, but he’s not the one in the operating room ultimately responsible for whether the patient lives or dies. And Girish Joshi, an anesthesiologist and vice chair of the committee that came up with the recommendation, said this particular study is flawed. There are too few patients, especially patients who died, to suggest meaningful correlations. Around 1 in 3,000 patients under anesthesia will aspirate, or inhale their own vomit. And Joshi said the study used medical diagnosis codes that were too broad, so it couldn’t correctly drill down on this specific phenomenon. Joshi said he’s not opposed to the general idea of using retrospective analysis to inform guidance, but the study needs to be better designed and include many more patients.

Ultimately, that’s what Hyde hopes he’ll be able to provide. Even if there may not be enough data to convince Joshi now, Atropos will keep adding new patients and new medications. “What you need is new evidence,” said Hyde. “And that’s what we generate.”


ForbesThis Health AI Startup Aims To Keep Doctors Up To Date On The Latest ScienceForbesWhy $4.6 Billion Health Records Giant Epic Is Betting Big On Generative AIForbesMicrosoft And Federal Agencies Launch Nonprofit Supergroup To Wrangle Health AI’s Wild WestForbesThis $2.2 Billion Startup’s GPT-4 Powered AI Bot Demystifies Health Insurance

Latest article