Randomized and nonrandomized studies: complementary or competing?