Study design, statistical method, and level of evidence in Japanese and American clinical journals

Clinical articles published in Japanese journals are said to be characterized by poor study design, less sophisticated statistics, and producing few high-grade clinical evidences. Two American and two Japanese medical journals, published in 1990, 1993, 1996, and 1999 were compared to find out the differences regarding study design, statistical methods, and level of clinical evidence of original articles and synthetic studies. There were 1689 original articles in American and 308 in Japanese joumals. Regarding study design, American articles contributed much more to randomized controlled trials/controlled trials/clinical trials (27.9% vs. 14.3%, p=0.001), cohort studies (21.6% vs. 6.2%, p=0.001), and case-control studies (6.5% vs.0.3 %, p=0.000). Among original articles in American and Japanese journals, mean number of statistical methods used were 2.4 and 1.7 per article (p=0.000), respectively. Articles providing high grade clinical evidence (grade Ia, Ib & IIa) were much greater in proportion in American journals than Japanese journals (31.1% vs. 12.7%, p=0.001). The overall picture of Japanese medical articles seems to be improving recently, at least in terms of statistical methods toward more diversified and sophisticated way of use, compared to the previous data.