Although randomised trials are important for evidence-based medicine, little is known about their overall characteristics. We assessed the epidemiology and reporting of methodological details for all 519 PubMed-indexed randomised trials published in December, 2000 (383 [74%] parallel-group, 116 [22%] crossover). 482 (93%) were published in specialty journals. A median of 80 participants (10th-90th percentile 25-369) were recruited for parallel-group trials. 309 (60%) were blinded. Power calculation, primary outcomes, random sequence generation, allocation concealment, and handling of attrition were each adequately described in less than half of publications. The small sample sizes are worrying, and poor reporting of methodological characteristics will prevent reliable quality assessment of many published trials.