| | | | |  | | By Ruth Reader, Erin Schumaker, Carmen Paun and Daniel Payne | | | | | 
Johns Hopkins is among the big names in medicine involved with CHAI. | Getty Images | A coalition of leading health systems and tech firms is pushing forward with its plan to self-regulate artificial intelligence in health care. The Coalition for Health AI, whose members include the Mayo Clinic, Microsoft and Johns Hopkins, plans a network of quality assurance labs to vet AI tools. Today, it laid out what the labs will need to do. The group says the labs will evaluate whether AI tools: — Function effectively and can be used in research — Deliver biased results or perform differently across patient populations — Provide ongoing performance monitoring CHAI says it will accredit labs based on their capabilities. Some may conduct pre-deployment evaluations and audits, others post-deployment monitoring. Some labs may be able to do all three. The group will certify labs based on domain expertise, such as image-based AI or bias detection. It will also ensure labs are free of conflicts of interest and protect the confidentiality of data they review. All labs will have to adhere to standards for testing laboratories as defined by the International Organization for Standardization, a Switzerland-based nongovernmental group that seeks to ensure tech platforms work across borders. The organization also revealed standards for AI transparency in line with the Department of Health and Human Services’ requirements for firms seeking government certification. Why it matters: In September, HHS officials said the department was working on an AI strategy, which will be made public by January, that included a network of assurance labs and an AI safety program. Shortly after, HHS Assistant Secretary for Technology Policy Micky Tripathi told POLITICO that assurance labs would be part of the strategy and that he is waiting for AI developers and purchasers to reach consensus on what the vetting process should look like. He also said HHS is closely following a number of private sector-led efforts, including those of the Coalition for Health AI; VALID AI, launched at the University of California, Davis; and the Health AI Partnership based at Duke University, and that he wasn’t endorsing any of them yet. But both Tripathi and FDA Director of Digital Health Troy Tazbaz previously served as advisers to CHAI.
| | | | 
Prince Edward Island, Canada | Shawn Zeller/POLITICO | This is where we explore the ideas and innovators shaping health care. Your cat is likely faster than your toddler in associating images with words, according to a new study from Japan, Science reports. Share any thoughts, news, tips and feedback with Carmen Paun at cpaun@politico.com , Daniel Payne at dpayne@politico.com, Ruth Reader at rreader@politico.com, or Erin Schumaker at eschumaker@politico.com. Send tips securely through SecureDrop, Signal, Telegram or WhatsApp.
| | |  Human organ tissue on chips like this could speed drug development | J. Adam Fenster/University of Rochester | Copies of human organ tissue on microchips could expedite the testing of drugs to prevent and treat respiratory diseases. That’s the theory behind a contract worth up to $7.1 million that HHS’ Center for Biomedical Advanced Research and Development Authority has awarded the University of Rochester to develop technology that can model how respiratory illnesses, like the flu or long Covid, affect brain function. Step one: Make organ tissue or small-scale copies of human tissue or 3D networks of human cells on a chip. The researchers then use the model to accelerate what they know about human biology and, potentially, supercharge the drug development process, Benjamin Miller, principal investigator on the project and a dermatology professor at the University of Rochester, told Erin. Step two: Integrate existing tissue-on-a-chip research, combining a lung chip and a brain chip to model what happens when a human contracts a respiratory virus infection and how the infection impacts the brain. One possible application: Using the model to test drugs that could reduce inflammation when you get a respiratory disease like the flu. Big picture: The BARDA project builds on a previous $7.5 million National Institutes of Health grant, which funded tissue-on-a-chip research at Rochester’s Translational Center for Barrier Microphysiological Systems. Why it matters: Before studying drugs in humans, researchers typically test them on animals. But animals aren’t people, Miller said, and the difference between them is a big part of why 90 percent of drug candidates fail when transitioning from animal to human testing. Testing drugs on human cell cultures in a petri dish is another option, but since it involves just one cell type, it’s overly simple compared to animal or tissue-on-a-chip testing. Real-time tracking: Continuous, real-time sensors will allow the researchers to understand what’s happening while the testing is underway. Miller compares the old method of setting up an experiment and checking results afterward to going to a movie theater, sitting down to watch the first frame and suddenly finding the credits rolling. “What we do is we integrate these sensors with the tissue chips so we can watch the whole movie, watch things change in real time,” he said.
| | |  Methadone remains the most effective medication for treating opioid addiction, a study found. | Kevin D. Liles/AP | People with opioid use disorder, particularly those using illicit fentanyl, are more likely to stay in treatment if they take methadone, another opioid that reduces their cravings, a new study found. How so? Canadian, Austrian and U.S. researchers analyzed the data of some 31,000 people in British Columbia who started treatment for opioid use disorder between January 2010 and mid-March 2020, with either a combination of buprenorphine and naloxone — known under the brand name of Subaxone — or with methadone. People who received Subaxone were at higher risk of stopping treatment within two years, at 88.8 percent, while those on methadone showed a lower risk: 81.5 percent, according to the study published Thursday in JAMA Network. The mortality risk while receiving treatment was similar between the two medications, researchers said. Both drugs were available through doctors’ offices and specialized drug treatment centers in British Columbia, unlike in the U.S., where methadone can only be prescribed and dispensed in specialized clinics. Why it matters: While the data didn’t include information about why people on methadone were less likely to drop out of treatment than those on Subaxone, “the primary explanation is that people are using street opioids with much higher potency now and our treatments, and treatment dosing guidelines, were all developed and tested before the fentanyl era,” said Bohdan Nosyk, one of the study authors, who leads the Health Economic Research Unit at the Centre for Advancing Health Outcomes in Vancouver, Canada. The study results suggest methadone should be considered as a first-line treatment for people with opioid use disorder, particularly those who haven’t been able to stick with other treatment regimens, Nosyk and his co-authors wrote. Ed Markey (D-Mass.), who together with Rand Paul (R-Ky.) proposed a Senate bill to expand where methadone can be prescribed and dispensed in the U.S., said the study underscores why the legislation is necessary. However, providers running methadone clinics have opposed the bill, arguing that the treatment approach could lead to diversion into the illicit drug market and potentially cause more fatal overdoses. | | | | Follow us on Twitter | | | | Follow us | | | | |
