Recurrent implantation failure in assisted reproduction: redefining diagnostic boundaries and shifting from empirical intervention to evidence-based selection

Recurrent implantation failure (RIF) is defined as the failure to achieve clinical pregnancy despite multiple transfers of high-quality embryos during assisted reproductive technology (ART) treatment. The definition of RIF has not been unified, with discrepancies primarily centering on maternal age, number of embryo transfers and total embryos transferred, criteria for high-quality embryo assessment, and estimation of cumulative pregnancy probability. In this article, we systematically review the evolution of RIF definitions and the impact of aneuploid embryo transfer on pregnancy outcomes; evaluate the diagnostic and therapeutic limitations of preimplantation genetic testing for aneuploidy (PGT-A) and endometrial receptivity analysis (ERA); re-examine the evidence-based rationale for screening chronic endometritis, endometrial microbiome, and immunological and thrombophilic factors; and highlight recent major advances in single-cell sequencing of the endometrium. We advocate for active investigation and management of identifiable pathological conditions that affect pregnancy outcomes, caution against “empirical multiple interventions,” and call for a return to the therapeutic principle of “first, do no harm.” It is anticipated that artificial intelligence-driven integrative multimodal modeling and real-world data registry studies will serve as breakthroughs for the diagnosis and treatment of RIF.