Number of Sites of Involvement Predicts Central Nervous System Relapse in Patients with Localized Extranodal Natural Killer/T-Cell Lymphoma

Introduction: The risk of secondary central nervous system (CNS) involvement is approximately 7% in patients with extranodal natural killer/T-cell lymphoma (ENKL) who receive non-anthracycline-based chemotherapy. The prognosis of ENKL in patients who develop CNS relapse is extremely poor, and the median overall survival (OS) after CNS relapse is less than 4 months. The CNS-Prognostic Index of Natural Killer (CNS-PINK) model is widely used to predict CNS relapse (Kim H. Blood, 2020). The relationship between the number of sites of involvement and CNS relapse in ENKL remains unclear. First-line treatment approaches differ between localized and advanced ENKL. To investigate risk factors for CNS relapse in patients with ENKL, we analyzed combined data from two large retrospective studies in Japan.

Methods: We retrospectively analyzed the data of patients with newly diagnosed ENKL who were diagnosed at 54 hospitals in the NKEA (2000-2013; UMIN000015491; Yamaguchi M. JCO, 2017) and NKEA-next (2014-2021; UMIN0000463000) projects. Patients without CNS involvement at diagnosis and those who had received non-anthracycline-based chemotherapy were eligible in the present study. We analyzed the following sites of involvement: the orbit, nasal cavity, paranasal sinus, cheek, gingiva, palate, and salivary grand. CNS involvement was confirmed by radiological findings, cerebrospinal fluid test results, or autopsies. The time to CNS relapse was calculated from the date of initial diagnosis of ENKL to the date of CNS relapse. In the present study, the cumulative incidence of CNS relapse was analyzed via Gray’s test, with CNS relapse and death in patients without CNS involvement included as competing events. We identified risk factors for CNS relapse via Fine‒Gray regression analysis.

Results: Among 708 patients with newly diagnosed ENKL, the baseline characteristics of eligible 540 patients were as follows: median age, 57 years (range, 13-86); male sex, 67%; Eastern Cooperative Oncology Group performance status (ECOG PS) > 1, 14%; and elevated serum lactate dehydrogenase level, 38%. Seventy-four % (n = 401) had localized disease, and 26% (n =139) had advanced disease. High-risk CNS-PINK was 0% in patients with localized ENKL and 81% in those of advanced ENKL. Among patients with localized ENKL, the number of sites of involvement was 0 in 9% of the patients, 1 in 65%, 2 in 19%, 3 in 6%, 4 in 0.5%, and 5 in 0.5%. As the first-line treatment, 96% of the patients with localized ENKL received chemoradiotherapy with dexamethasone, etoposide, ifosfamide, and carboplatin (DeVIC), and 59% of the patients with advanced ENKL were treated with L-asparaginase-containing chemotherapy. For CNS prophylaxis, high-dose methotrexate (MTX) (≥ 2 g/m²) was used in 89 patients and intrathecal MTX or cytarabine was used in 44 patients.

At a median follow-up of 52 months, 25 patients (4.6%) experienced CNS relapse. The 2-year cumulative risk of CNS relapse was 4.3% (95% CI, 2.8–6.3), and the median time to CNS relapse was 6.5 months in all patients. Nineteen patients experienced isolated CNS relapse, and 6 patients experienced concurrent systemic relapse. The incidence of CNS relapse in the brain parenchyma and leptomeninges was 26% and 74%, respectively. The 2-year cumulative risk of CNS relapse was 3.4% (95% CI, 1.9-5.5) and 7.1% (95% CI, 3.5-12) in patients with localized and advanced ENKL, respectively. Univariate analysis of the 401 patients with localized ENKL revealed that > 1 sites of involvement (P = 0.009), ECOG PS > 1 (P = 0.01), and involvement of the orbit (P = 0.007), the paranasal sinus (P = 0.006), the gingiva (P = 0.004), and the palate (P = 0.02) were identified as risk factors for CNS relapse. Multivariate analysis of patients with localized ENKL identified > 1 sites of involvement (P = 0.01) and ECOG PS > 1 (P = 0.02) as independent risk factors for CNS relapse. For patients with advanced ENKL, univariate analysis revealed that involvement of the orbit (P = 0.01), the palate (P = 0.007), and the breast (P = 0.007) were risk factors for CNS relapse.

Conclusions: No patients with localized ENKL were classified as high-risk CNS-PINK in our cohort. Our present study highlights the importance of > 1 sites of involvement and ECOG PS > 1 for predicting CNS relapse in patients with localized ENKL, warranting further investigation.