Over the last ten years the quality of published clinical trials has improved considerably. Randomised controlled design, concealed allocation, blinding, clinically relevant outcomes, large sample sizes and almost complete follow-up are standard features of most trials. As a result, standard evidence-based medicine criteria can no longer be used to separate the "wheat from the chaff". Here, we discuss recent studies which reveal only marginal superiority of new treatments over well-established and effective therapies. Nowadays, readers of reports should be particularly aware of the definition of control therapies, of clinical relevance when results are statistically significant, of the external validity and whether a trial serves a pragmatic or explanatory purpose.