Can the randomized controlled trial literature generalize to nonrandomized patients?