A recent study published in JAMA Health Forum revealed that fewer than 2% of FDA-cleared artificial intelligence (AI) medical devices reported randomized clinical trial data. Researchers analyzed 691 AI/ML devices cleared from September 1995 to July 2023, mostly classified as class II and cleared via the 510(k) pathway. Radiology dominated with 76.9% of devices. Premarket reporting was found incomplete: 46.7% lacked study design, and 53.3% didn’t disclose sample sizes. Only 1.6% cited randomized trials, and clinical performance metrics were rarely reported. Safety assessments were documented for 28.2% of devices, while only 6.1% reported potential health risks. Notably, 5.2% of devices faced 489 adverse events, leading to 40 recalls. These findings highlight the urgent need for standardized efficacy and safety evaluations in AI medical devices and emphasize the importance of enhanced regulatory pathways and postmarket surveillance for ensuring patient safety.
Source link
