The need for evidence-based medicine as a foundation for optimal patient care requires application of the best scientific methods. Various methods used in the search for clinical truth are discussed. The most powerful method to test a clinical hypothesis is the randomized, controlled clinical trial. By contrast, epidemiology/observation studies (EOS) have inherent weaknesses that can lead to erroneous conclusions. Five such examples, two older and three very recent, are discussed to provide historical perspective and demonstrate potential problems with EOS. Each incident has produced widespread consternation or confusion among physicians. Types of bias leading to errors in EOS are discussed, and the need is emphasized for more rigorous evaluation of EOS by investigators, as well as editors, to avoid repetition of past mistakes and to ensure publication of correct medical information.