U Vogel1, J Windeler. 1. Abteilung Medizinische Biometrie, Universitätsklinikum Heidelberg.
Abstract
BACKGROUND AND OBJECTIVE: The problem of so-called "publication bias"--distortion of statements about an event by selective publication--is receiving growing attention. Assessment of the results of a given study can be affected by knowing the factors that favour or inhibit the publication of its results. METHODS: The data were based on an analysis of 371 medical doctoral dissertations on clinical research. Characteristic features of the dissertations were obtained by questionnaire, supplemented by MEDLINE search and, additionally, by personal enquiry about the publications. Bivariate and multivariate analysis of the characteristic features was used to evaluate the factors that influenced publication of a dissertation. RESULTS: There were mainly three factors that influenced the decision whether or not a dissertation was published: (1) a positive result of the study (odds ratio 2.5, 95% confidence interval 1.12-5.75); (2) the publishing activity of the dissertation's referee (odds ratio 3.8, 95% confidence interval 1.77-7.97; and (3) the quality of the statistical analysis of the dissertation's data (odds ratio 1.67, 95% confidence interval 0.99-2.80). CONCLUSION: A distinct publication bias can be demonstrated also in the case of doctoral dissertations in that positive results are clearly published more often than negative ones.
BACKGROUND AND OBJECTIVE: The problem of so-called "publication bias"--distortion of statements about an event by selective publication--is receiving growing attention. Assessment of the results of a given study can be affected by knowing the factors that favour or inhibit the publication of its results. METHODS: The data were based on an analysis of 371 medical doctoral dissertations on clinical research. Characteristic features of the dissertations were obtained by questionnaire, supplemented by MEDLINE search and, additionally, by personal enquiry about the publications. Bivariate and multivariate analysis of the characteristic features was used to evaluate the factors that influenced publication of a dissertation. RESULTS: There were mainly three factors that influenced the decision whether or not a dissertation was published: (1) a positive result of the study (odds ratio 2.5, 95% confidence interval 1.12-5.75); (2) the publishing activity of the dissertation's referee (odds ratio 3.8, 95% confidence interval 1.77-7.97; and (3) the quality of the statistical analysis of the dissertation's data (odds ratio 1.67, 95% confidence interval 0.99-2.80). CONCLUSION: A distinct publication bias can be demonstrated also in the case of doctoral dissertations in that positive results are clearly published more often than negative ones.