Systemic Inflammatory Response Markers Improve the Discrimination for Prognostic Model in Hepatocellular Carcinoma

Background/purpose of the study: We aimed to evaluate the performance of neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR) and their combination (combined NLR-PLR, CNP) on overall survival (OS) and recurrence-free survival (RFS) in a large cohort of unselected hepatocellular carcinoma (HCC) patients. Methods: Training and validation cohort data were retrieved from the Italian Liver Cancer (ITA.LI.CA) database. The optimal cut-offs of NLR and PLR were calculated according to the multivariable fractional polynomial and the minimum p-value method. The continuous effect and best cut-off categories of NLR and PLR were analyzed using multivariable Cox regression analysis. A shrinkage procedure adjusted over-fitting HR estimates of best cut-off categories. C-statistic and integrated discrimination improvement (IDI) were calculated to evaluate the discrimination properties of the biomarkers when added to clinical survival models. Results: 2,286 patients were split into training (n=1,043) and validation (n=1,243) cohorts.The optimal cut-offs for NLR and PLR were 1.45 and 188, respectively. NLR (HR 1.58, 95%CI 1.11-2.28, p=0.014) and PLR (HR 1.79, 95%CI 1.11-2.90, p=0.018) were independent predictors of OS. When added to the clinical prognostic model, including age, alpha-fetoprotein (AFP), CHILD-Pugh score and Barcelona Clinic Liver Cancer (BCLC) staging system, CNP had a significant incremental value in predicting OS (IDI 1.3%, p=0.04). Data were confirmed in the validation cohort. NLR (p=0.027) and CNP (p=0.023) predicted RFS in the training cohort. Conclusions: NLR, PLR, and CNP independently predicted shorter OS in HCC patients. The addition of CNP into the survival prediction model significantly improved the model's predictive accuracy for OS.