Prognostic Value of Platelet-Associated Biomarkers in Nasopharyngeal Carcinoma Patients Received IMRT: A Retrospective Study.

PURPOSE: This study examines the prognostic significance of platelet-related biomarkers for overall survival (OS) in nasopharyngeal carcinoma (NPC) using Cox proportional hazards and random survival forest (RSF) models, and compares the two modes. METHODS: In total, 825 patients with confirmed nasopharyngeal carcinoma were included in this study, and a systematic analysis of their prognostic risk factors was conducted. The patient dataset was divided into training and validation datasets at a ratio of 7:3. A Cox model and RSF model were constructed using clinical data, platelet distribution width (PDW), the ratio of platelet distribution width to platelet count (PDW/P), mean platelet volume (MPV), mean platelet volume‒platelet count ratio (MPV/P) and follow-up survival data. The performances of the two prediction models were evaluated. RESULTS: Multivariate analysis revealed that age (P<0.001, hazard ratios (HR) =1.76, 95% confidence intervals (CI):1.28-2.42), T stage (T3: HR 2.81, 95% CI:1.62-4.83, p<0.001; T4: HR 4.10, 95% CI:1.97-8.52, p<0.001), PDW/P (P= 0.027, HR=1.82, 95% CI:1.07-3.22), and EBV-DNA level (P=0.016, HR=1.51, 95% CI:1.08-2.11) were independent prognostic factors for OS. The RSF model revealed that T stage, stage, the PDW/P and PDW had a notable impact on the model. The Log rank test results showed that when the risk score threshold of the RSF model was 39.63, it had efficacy in differentiating patient groups with significant differences in prognosis. The C-index of the training cohort for the COX model and the RSF model were 0.70 (95% CI: 0.663-0.737) and 0.71 (95% CI, 0.69-0.73) respectively, and 0.62 (95% CI, 0.553-0.681), 0.68 (95% CI:0.62-0.74) in the validation data. Moreover, the decision curve analysis demonstrated that the RSF model provided greater net benefit than the Cox proportional hazards model did. CONCLUSION: The PDW/P ratio was an independent prognostic factor for OS. The RSF model demonstrated high reliability in predicting clinical outcomes for patients with NPC.