IntroductionDiffuse large B-cell lymphoma (DLBCL) is the most common subtypes of lymphoma. Clinical biomarkers are still required for DLBCL patients to identify high-risk patients. Therefore, we developed and validated the platelet-to-albumin (PTA) ratio as a predictor for DLBCL patients.MethodsA group of 749 patients was randomly divided into a training set (600 patients) and an internal validation set (149 cases). The independent cohort of 110 patients was enrolled from the other hospital as an external validation set. Penalized smoothing spline (PS) Cox regression models were used to explore the non-linear relationship between the PTA ratio and overall survival (OS) as well as progression-free survival (PFS), respectively.ResultsA U-shaped relation between the PTA ratio and PFS was identified in the training set. The PTA ratio less than 2.7 or greater than 8.6 was associated with the shorter PFS. Additionally, the PTA ratio had an additional prognostic value to the well-established predictors. What’s more, the U-shaped pattern of the PTA ratio and PFS was respectively validated in the two validation sets.DiscussionA U-shaped association between the PTA ratio and PFS was found in patients with DLBCLs. The PTA ratio can be used as a biomarker, and may suggest abnormalities of both host nutritional aspect and systemic inflammation in DLBCL.