BACKGROUND Randomized controlled trials (RCTs) are frequently used to evaluate the effectiveness of health care interventions in family medicine. The value of the information obtained from RCTs depends largely on the quality of design and the way in which they are conducted and reported. Despite the increasing number of RCTs being conducted in family medicine, there is a scarcity of descriptive data on the methodological characteristics, including design features and quality of RCTs in this setting. METHODS All 55 RCTs published in four peer-reviewed US family medicine journals between 1987 and 1991 were identified and their methodological characteristics reviewed. Three potential sources of bias were assessed in each of the trial reports: (1) control of selection bias at entry, (2) control of selection bias after entry, and (3) control of bias in assessing outcome(s). RESULTS Fifty-five RCTs published between January 1987 and December 1991 were identified in the four journals. The absolute number of RCTs published over the 5 years increased steadily, and there was a 49% increase in the proportion of RCT articles. Measures used to control for selection bias before entry into the study were reported in 14 (25%) of the RCTs, the statistical power of the trial in 5 (9%), and whether the study had been reviewed by an institutional review board in 6 (11%). CONCLUSIONS The RCTs analyzed offered some imaginative solutions to the logistic difficulties of conducting RCTs in general practice. Nevertheless, the methodology and reporting of RCTs in the future should be improved.