PUBLICATION BIAS IN TWO SPANISH MEDICAL JOURNALS
Carlos Campillo
Gerencia de Atención Primaria de Baleares, c/ Reina
Esclaramunda, 9, 07003 Palma de Mallorca, Spain
Objective: To assess the factors associated with publication status of papers submitted to 2 major Spanish medical journals (Atencion Primaria and Medicina Clinica)
between 1994 and 1996; and to investigate the existence of publication bias in both journals. The study hypothesis was that publication bias stems from authors, not from editors.
Design: A case-control study based on a consecutive and representative sample of 100 published papers, half taken from each journal (cases), and 100 rejected manuscripts,
half rejected by each journal (controls). Papers reporting observational and experimental studies were included. The following risk factors for publication bias were blindly
assessed with regard to the journal and to publication status (outcome variable: published vs rejected): type of study, objective of the study, region of Spain where the study
was conducted, sample size, complexity of the analyses, statistical significance, and methodologic quality. Quality scores were blindly assigned to each paper by means of a
previously validated questionnaire. The association of each independent variable with publication status was assessed using bivariate and multivariate analyses.
Results: Only 4 papers out of 200 were excluded because they did not report observational or experimental studies. Quality (high vs low) was the only risk factor
associated with publication status (Atencion Primaria: OR=17.3; 95% CI, 5.7-55.0); Medicina Clinica: OR=234; 95% CI, 0.6-241.6)). Out of 200 papers, 146 included hypothesis testing,
of which 143 were positive studies, and 3 were negative(2 were published and 1 was rejected).
Conclusions: Quality was the only factor associated with publication status in both journals. If publication bias exists, it stems from authors, not editors.
Return to Session Information
