A new Science analysis argues that key randomized trials on UPFs were not controlled enough to show whether ultra-processing itself is harmful.