Page 208 - Haematologica Vol. 110 - January 2025

LETTER TO THE EDITOR
A novel prognostic nomogram based on imaging and molecular parameters for newly diagnosed extranodal natural killer/T-cell lymphoma patients
Extranodal natural killer/T-cell lymphoma (ENKTL) is a highly aggressive non-Hodgkin lymphoma strongly as- sociated with Epstein-Barr virus (EBV) infection and is characterized by a high relapse rate.1 Recently, the 5-year overall survival (OS) of limited-stage ENKTL has increased to 72-74% as the result of the introduction of a novel strategy of concurrent chemoradiotherapy.2 However, the 5-year OS in advanced-stage disease remains around 15-25%.3 Since risk-adapted therapy plays a pivotal role in improving the survival of newly diagnosed patients,4 a better risk classification model could assist in precisely stratifying patients with ENKTL into risk groups and for- mulating appropriate individualized treatments to improve prognosis.
Several risk scoring systems for newly diagnosed ENKTL patients are currently available in clinical practice, in- cluding the International Prognostic Index (IPI), the Korean Prognostic Index (KPI), the prognostic index for natural killer lymphoma with or without EBV-DNA (PINK/PINK-E), and the nomogram-revised risk index (NRI).5-8 Among these systems, the PINK-E is the only scoring system that includes data regarding EBV-DNA and is widely used in clinical practice. Several studies have demonstrated the good prognostic value of EBV-DNA and maximum stan- dardized uptake value (SUVmax) in patients with ENKTL.9-12 However, no risk classification model has used SUVmax in 18F-fluorodeoxyglucose positron emission tomography/ computed tomography (18F-FDG PET/CT) and the quanti- tative value of EBV-DNA. In previous work, we integrated circulating tumor DNA (ctDNA) into the PINK-E to construct the PINK-EC model, which could overcome the poor dis- crimination efficiency of PINK-E for patients in low-risk and intermediate-risk groups.13 However, compared with the classic PINK-E system, our new PINK-EC model showed only slight improvement in terms of Harrell’s C-index for OS, which might not satisfy personalized clinical demand. Therefore, a more accurate and precise risk classification model is urgently needed. Here, we developed a nomo- gram model (we named it SEC, inspired by the initials of “SUVmax”, “EBV-DNA” and “ctDNA”), which included the semiquantitative radiomic parameter SUVmax and quanti- tative molecular parameters such as EBV-DNA and ctDNA to accurately stratify newly diagnosed ENKTL patients for optimal personalized treatment and management.
In this study, 91 patients newly diagnosed with ENKTL were enrolled at Xinqiao Hospital between February 2017 and No- vember 2023 (ClinicalTrials identifier: ChiCTR1800014813).
The data collected at diagnosis, as previously described,13,14 included age, gender, B-symptoms, Eastern Cooperative Oncology Group performance status (ECOG PS), primary site, regional lymph node involvement, distant lymph node involvement, numbers of extranodal sites, serum lactate dehydrogenase (LDH) level, whole blood EBV-DNA copy number, 18FDG PET/CT SUVmax value and ctDNA concen- tration. In this study, we extended the follow-up time to the date of this analysis. The methodology of ctDNA measurements was reported previously.13 This study was approved by the China Ethics Committee of Registering Clinical Trials (ChiECRCT-20,180,005).
Based on the conventional cutoff value, continuous vari- ables such as age and LDH concentration were divided into two categories. The ideal cutoff values for SUVmax, ctDNA, and EBV-DNA for survival prediction were deter- mined by the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. Progression-free survival (PFS) was calculated from diagnosis to disease progression, death from any cause, or the date of last follow-up. OS was measured from the date of diagnosis to the date of death due to any cause or the date of the last follow-up. Survival time was estimated using Kaplan-Meier survival curves and compared by log-rank tests. Univariate and multivariate analyses and calcula- tions of hazard ratios (HR) with 95% confidence intervals (95% CI) were performed using Cox regression models. A nomogram was generated based on the independent predictors of survival outcomes determined by univari- ate and multivariate analyses. A calibration curve (1,000 bootstrap resamples) was constructed to assess the consistency between the predicted and observed sur- vival. The discriminatory ability of the model was evalu- ated by Harrell’s C-index. A time-dependent ROC curve was used for the comparison of risk stratification for these prognostic models. All the statistical analyses were performed with IBM SPSS statistical software (version 25.0; IBM Inc., NY, USA) and R version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided P<0.05 was considered statistically significant.
The baseline characteristics of the whole cohort are listed in Online Supplementary Table S1. The optimal cutoff values determined by the ROC analysis for SUV- max, EBV-DNA, and ctDNA were 9.950, 1.4×104 copies/ mL, and 4.026 haploid genome equivalents per milliliter (hGE/mL), respectively (Figure 1).
According to the univariate and multivariate Cox re-
Haematologica | 110 January 2025
200