Abstract
Misuse of statistical methods in biomedical research remains widespread, undermining scientific integrity and public health.Flawed analyses can lead to misleading clinical guidelines, unnecessary treatments, or concealment of true treatment effects.Statistical errors often result from a structural training gap that affects the research process from study design to reporting, rather than from individual incompetence.Authors should be able to select appropriate measures of central tendency based on the data distribution, apply parametric and nonparametric tests correctly, use regression analyses with attention to statistical power and collinearity, and interpret P values in the context of confidence intervals.The growing number of retractions in statistics and the continued citation of retracted articles highlight the limitations of current peer-review processes.Statistical Package for the Social Sciences is the most widely used statistical software in health sciences research, while R/RStudio is gaining increasing adoption due to its open-source nature, analytical flexibility, and capacity for fully reproducible and transparent analyses.Software selection can significantly influence analytical outcomes due to differences in default algorithms.Therefore, full disclosure of software versions and procedures is essential for reproducibility.Integrating artificial intelligence-based tools into statistical workflows introduces risks, including output hallucination, a lack of algorithmic transparency, and unresolved accountability.These tools should not be used without internationally accepted validation guidelines.At the publisher level, employing statistical editors, enforcing open data-sharing policies, and adopting contributor role taxonomies such as Contributor Roles Taxonomy are the most effective structural interventions.Enhancing the statistical quality of biomedical literature requires coordinated, sustained efforts from all stakeholders in scientific communication.