Evaluation of the effectiveness of seat-belt laws in reducing motor vehiclerelated morbidity and mortality requires appropriate epidemiologic methods. Experimental and cohort study designs are inadequate because these laws apply to all car users in a particular state or country, thus precluding random assignment of the intervention and a control group for comparison. Pretestposttest evaluations have been suggested as an alternative, but this design also has limitations. In a regional study of the effect of the 1979 French seat-belt law, the nature and severity of a person's injuries were recorded for all car crashes during the same 3-month period in 1976, 1977, 1978 (before the law), and in 1979 (after the law). Compared with 1978, crashes in 1979 showed a marked decrease in the frequency and severity of certain injuries, including those of the head and face, but an increase in the frequency and severity of other injuries, especially those of the cervical spine. Although, these changes are compatible with the biomechanical effects of increased seat belt use, one cannot conclude that the apparent changes were the effect of the law. A simple trend analysis of data from all 4 years showed that some changes from 1978 to 1979 had also occurred from 1977 to 1978, before the law had been passed. The trend analysis suggested that the results of a pretest-posttest design might be inaccurate for assessing the impact of the French seat-belt law and should be interpreted cautiously. We believe that at least a simple trend analysis should be done to evaluate the effect of seat-belt laws and that this approach suggest a beneficial effect of the 1979 French law.