To describe the biomedical journal characteristics that are associated with the retraction of papers.A descriptive cross-sectional study was conducted. All papers retracted and indexed in PubMed from January 1 2013 to December 31, 2016 were included. We used nine main categories to classify retractions: aspects related with data, authors issues, plagiarism, unethical research, journal issues, review process, conflict of interest, other, and unknown. These categories were further classified as: misconduct, suspicion of misconduct, or no misconduct.The proportion of retraction was 2.5 per 10,000 publications. Retractions appeared in 611 journals. During the study period, retraction due to misconduct was more frequent among journals with low-impact factor. Within these retracted publications, among low-impact journals the presence of misconduct was higher with a 73% compared to 61% for the high-impact journals (p=0.001). There were differences in the percentage of retractions due to misconduct regarding the journal classification category (p<0.001).Retraction of publications is present in both high- and low-impact factor biomedical journals, but misconduct is more frequent among the papers retracted from lower impact journals. Measures before and after publication should be taken to limit misconduct.