L Mihan1, J Windeler. 1. Abteilung Medizinische Biometrie, Ruprecht-Karls-Universität Heidelberg.
Abstract
BACKGROUND: Standards concerning design, performance and analysis of controlled clinical trials have improved during the last decades. A number of recommendations has been published. Their impact is hardly known. METHODS: Using a standardized procedure controlled clinical trials published between 1979 and 1996 in the "Medizinische Klinik" were identified and reviewed for their quality of design, analysis and presentation of results. Results were subdivided by 6 three-year-periods. RESULTS: 132 trial reports were identified. Randomisation was reported in 107 papers (81%) with further details of the procedure in only 20. Adequate concealment is recognizable in 21 papers. Results of baseline comparisons were presented in 52 publications (39%). 57 reports did not mention any blinding although no arguments are given. Information about drop-outs can be found in 90 publications. In most papers, however, drop-outs were omitted from the analyses. An adequate intention-to-treat analysis is provided in 7 reports. Due to missing or incomplete data many results and statistical tests can not be recalculated by the reader. Only 6 papers presented some kind of power calculation. Methodological quality has improved over the complete time period with a certain weakening in the three years before 1996. CONCLUSION: Quality criteria for a meaningful controlled clinical trial are not sufficiently met in a number of cases. Percentages are in accordance with the international literature. Major problems comprise sample size calculation and the application of the intention-to-treat principle.
BACKGROUND: Standards concerning design, performance and analysis of controlled clinical trials have improved during the last decades. A number of recommendations has been published. Their impact is hardly known. METHODS: Using a standardized procedure controlled clinical trials published between 1979 and 1996 in the "Medizinische Klinik" were identified and reviewed for their quality of design, analysis and presentation of results. Results were subdivided by 6 three-year-periods. RESULTS: 132 trial reports were identified. Randomisation was reported in 107 papers (81%) with further details of the procedure in only 20. Adequate concealment is recognizable in 21 papers. Results of baseline comparisons were presented in 52 publications (39%). 57 reports did not mention any blinding although no arguments are given. Information about drop-outs can be found in 90 publications. In most papers, however, drop-outs were omitted from the analyses. An adequate intention-to-treat analysis is provided in 7 reports. Due to missing or incomplete data many results and statistical tests can not be recalculated by the reader. Only 6 papers presented some kind of power calculation. Methodological quality has improved over the complete time period with a certain weakening in the three years before 1996. CONCLUSION: Quality criteria for a meaningful controlled clinical trial are not sufficiently met in a number of cases. Percentages are in accordance with the international literature. Major problems comprise sample size calculation and the application of the intention-to-treat principle.