Curie Brief
Turn on cookies to sign in
Signing in saves your progress to your Curie account. We can only do that with cookies on — turn them on to continue.
News
AI can't replace the MDT for severe asthma treatment. A new study found that five major AI models — ChatGPT, Copilot, DeepSeek, Gemini, and PerplexityAI — all showed poor agreement with multidisciplinary team (MDT) biologic treatment decisions for uncontrolled severe asthma. Researchers flagged erroneous outputs and non-evidence-based reasoning as serious safety concerns.
AI stumbles on severe asthma treatment decisions
Five leading AI models couldn't keep up with real-world clinical expertise when it came to choosing biologic therapies for severe asthma. A retrospective study of 114 biologic-naïve adults with uncontrolled severe asthma found that ChatGPT, Copilot, DeepSeek, Gemini, and PerplexityAI all showed poor agreement with decisions made by multidisciplinary team (MDT) meetings — the current gold standard for this patient population.
Researchers were particularly concerned by the quality of AI outputs, which included non-evidence-based reasoning, logical inconsistencies, and nonsensical recommendations — even when models were explicitly instructed to follow Global Initiative for Asthma (GINA) guidelines.
By the Numbers
Why it matters: As AI tools become more common in clinical settings, this study is a clear reminder that AI should support — not replace — human clinical judgment, especially in complex, high-stakes decisions like biologic therapy selection for severe asthma.
