Issues in reporting epidemiological studies
http://www.100md.com
《英国医生杂志》
EDITOR—Pocock et al surveyed some recent epidemiological publications.1 We agree with the general idea that the practice of reporting in epidemiologic publications needs improvement, but we are concerned about their methodological approach and conclusions.
Firstly, they say that few studies gave any power calculations to justify the study sizes. Ironically, Pocock et al's survey does not give any justification for the chosen study size of 73 epidemiological papers.
Secondly, Pocock et al's opinion on studies too small for meaningful analysis can be seen from a different point of view. Studies that are statistically underpowered are obviously less well planned and are often disregarded because they more often suffer from type I and type II errors compared with larger studies. However, from a public health perspective small studies may indicate true increased risks for exposures of interest that may have a substantial population-wide effect if the exposure is prevalent. These studies could be regarded as early warnings.
Thirdly, we are generally concerned about the discussion section. Pocock et al nearly completely omitted any critical discussion of their own results. It is not our purpose to persuade the reader that every paper published in 2001 eligible with respect to the study by Pocock et al is appropriate from a methodological point of view. However, we are convinced that this study adds little to our (those working in the field) knowledge and to science in general. The study lacks a clear hypothesis and design, has an insufficient study size in view of the high "bibliometric" variability, and is prone to misinterpretation for those with little knowledge of epidemiology.
Karl-Heinz J?ckel, professor of medical informatics, biometry and epidemiology
Institute of Medical Informatics, Biometry and Epidemiology, University Hospital, University of Duisburg-Essen, Hufelandstrasse 55, D-45147 Essen, Gemany k-h.joeckel@uni-essen.de
Andreas Stang, professor of clinical epidemiology
Institute of Medical Epidemiology, Biometry and Informatics, Medical Faculty, Martin-Luther-University of Halle-Wittenberg, Magdeburger Strasse 27, D-06097 Halle, Germany andreas.stang@medizin.uni-halle.de
Competing interests: None declared.
References
Pocock SJ, Collier TJ, Dandreo KJ, de Stavola BL, Goldman MB, Kalish LA, et al. Issues in the reporting of epidemiological studies: a survey of recent practice. BMJ 2004;329: 883-98. (16 October.)
Firstly, they say that few studies gave any power calculations to justify the study sizes. Ironically, Pocock et al's survey does not give any justification for the chosen study size of 73 epidemiological papers.
Secondly, Pocock et al's opinion on studies too small for meaningful analysis can be seen from a different point of view. Studies that are statistically underpowered are obviously less well planned and are often disregarded because they more often suffer from type I and type II errors compared with larger studies. However, from a public health perspective small studies may indicate true increased risks for exposures of interest that may have a substantial population-wide effect if the exposure is prevalent. These studies could be regarded as early warnings.
Thirdly, we are generally concerned about the discussion section. Pocock et al nearly completely omitted any critical discussion of their own results. It is not our purpose to persuade the reader that every paper published in 2001 eligible with respect to the study by Pocock et al is appropriate from a methodological point of view. However, we are convinced that this study adds little to our (those working in the field) knowledge and to science in general. The study lacks a clear hypothesis and design, has an insufficient study size in view of the high "bibliometric" variability, and is prone to misinterpretation for those with little knowledge of epidemiology.
Karl-Heinz J?ckel, professor of medical informatics, biometry and epidemiology
Institute of Medical Informatics, Biometry and Epidemiology, University Hospital, University of Duisburg-Essen, Hufelandstrasse 55, D-45147 Essen, Gemany k-h.joeckel@uni-essen.de
Andreas Stang, professor of clinical epidemiology
Institute of Medical Epidemiology, Biometry and Informatics, Medical Faculty, Martin-Luther-University of Halle-Wittenberg, Magdeburger Strasse 27, D-06097 Halle, Germany andreas.stang@medizin.uni-halle.de
Competing interests: None declared.
References
Pocock SJ, Collier TJ, Dandreo KJ, de Stavola BL, Goldman MB, Kalish LA, et al. Issues in the reporting of epidemiological studies: a survey of recent practice. BMJ 2004;329: 883-98. (16 October.)