[The level of use of statistical data in articles in the Orvosi Hetilap]

The objective was to assess the level of use of statistics in the articles of the Orvosi Hetilap (OH). Eight clinical and 25 randomly selected general medical papers (published in 1992) were evaluated with the help of checklists used in the British Medical Journal (BMJ). The result was compared with the survey conducted in the BMJ (Gardner and Bond, JAMA, 1990, 263, 1355-1357). Two clinical studies were found suitable for publication without further revision. All the general papers needed revision and in 8 studies no changes appeared to be sufficient for publication. For the clinical and general articles, respectively, loglinear modelling showed a 4 and 6 times higher likelihood for satisfactory qualification in the BMJ than in the OH. In the OH, the quality of the analysis and presentation was rated the least satisfactory while the design was assessed the best among all checklist items. The main source of methodological flaws were either the lack of any statistical method or the inappropriate selection of them. The validity of the results published in the papers evaluated was often diminished by multiple testing or neglecting potential confounders. Contribution of statisticians appears desirable in planning and evaluating the investigation. Following the practice in the major medical journals, introduction of statistical review in the OH was suggested. In the view of the authors, reforms in medical education and research seem to be unavoidable for responding to the challenge of today's medical methodology.