Objective To analyze the causes of retracted publications and the main characteristics of their authors. Method A descriptive cross-sectional study was designed including all retracted publications from January 1st, 2013-December 31st, 2016 indexed in PubMed. The causes of retraction were classified as: data management, authorship issues, plagiarism, unethical research, journal issues, review process, conflict of interest, other causes, and unknown reasons. Then, misbehaviour was classified as misconduct, suspicion of misconduct or no misconduct suspicion. Results 1,082 retracted publications were identified. The retraction rate for the period was 2.5 per 10,000 publications. The main cause of retraction was misconduct (65.3%), and the leading reasons were plagiarism, data management and compromise of the review process. The highest proportion of retracted publications corresponded to Iran (15.52 per 10,000), followed by Egypt and China (11.75 and 8.26 per 10,000). Conclusions Currently, misconduct is the main cause of retraction. Specific strategies to limit this phenomenon must be implemented. It would be useful to standardize reasons and procedures for retraction. The development of a standard retraction form to be permanently indexed in a database might be relevant.