Circulating NK-cell counts improve prognostic stratification of the PINKE model in extranodal NK/T-cell lymphoma: a retrospective cohort study.

PURPOSE: The prognostic index of natural killer lymphoma-Epstein-Barr virus (PINKE) is a standard model for extranodal NK/T-cell lymphoma (ENKTL) but relies solely on clinical and virological parameters. This study aimed to integrate peripheral blood immune biomarkers into PINKE to enhance prognostic accuracy. METHODS: In this retrospective single-center cohort of 203 patients with newly diagnosed ENKTL, we developed a refined prognostic model. Candidate predictors were screened using LASSO regression, and an independent immune risk factor was identified through multivariable Cox analysis. This factor was then incorporated into the conventional PINKE model. The incremental value of this enhanced model was assessed through Harrell's C-index, time-dependent ROC analysis, internal validation via bootstrapping with 1000 resamples, and decision curve analysis. RESULTS: A low absolute NK cell count (ANKC < 0.13 ×10⁹/L) was established as an independent adverse prognostic factor for overall survival (hazard ratio, 3.17; 95% confidence interval, 1.40-7.17; P = 0.006). Lower ANKC correlated with aggressive disease features, including advanced stage and unfavorable immune profiles. Integration of ANKC into the PINKE framework led to the development of the refined NKPINKE model, which demonstrated superior and more stable predictive accuracy compared to the original PINKE model, with an improved C-index (0.786 vs 0.756) and higher AUC values at 1, 3, and 5 years (0.864, 0.844, and 0.841 vs 0.834, 0.803, and 0.792, respectively). Decision curve analysis confirmed a superior net benefit, and calibration analysis indicated acceptable model performance. CONCLUSIONS: The NKPINKE model, incorporating peripheral immune biomarkers, improves prognostic discrimination over the original PINKE score. External validation is required to confirm its generalizability prior to clinical application.