First-line non–anthracycline-based chemotherapy for extranodal nasal-type NK/T-cell lymphoma: a retrospective analysis from the CLCG

Extranodal natural killer/T-cell lymphoma, nasal type (ENKTCL) is rare in Western populations but more frequent in East Asia.^1-3  The disease is unique among aggressive lymphomas in terms of its clinical features and treatment principle. ENKTCL is resistant to anthracycline (ANT)-based regimens but sensitive to radiation.^2,4-6  Radiotherapy plays an important role in both locoregional disease control and maintaining long-term survival in early-stage patients.^2-6  In a recent registry study from the National Cancer Database,²  the omission or inadequate dose (<50 Gy) of radiotherapy is negatively associated with overall survival (OS). Furthermore, in a previous study from the China Lymphoma Collaborative Group (CLCG),⁶  improved locoregional control using appropriate radiotherapy is associated with prolonged OS and progression-free survival (PFS). Primary radiotherapy achieves a favorable prognosis in patients with early-stage ENKTCL,^4-6  whereas adding ANT-based chemotherapy into radiotherapy significantly improves survival in high-risk early-stage patients.⁵  However, the prognosis of patients with localized or disseminated diseases is poor using ANT-based chemotherapy alone.^2,5 

Prospective phase 1/2 trials and retrospective studies demonstrated that non–ANT-based regimens provided survival benefits compared with ANT-based regimens in newly diagnosed or relapsed/refractory ENKTCL.^7-18  A variety of non–ANT-based regimens, mainly asparaginase (ASP)-based regimens, have been recommended as first-line treatments for ENKTCL. However, evidence supporting the clinical use of non–ANT-based regimens is limited to single-arm phase 1/2 trials or retrospective studies with small cohorts of patients. The restrictive recruitment criteria in prospective uncontrolled trials might select out suitable young patients with preserved organ function, leading to favorable treatment outcomes. Therefore, the beneficial effect of non–ANT-based regimens needs to be validated in a large comparative study.

Given the rarity of ENKTCL and the difficulty in conducting randomized controlled trials, we designed a large-scale retrospective study to compare the survival benefit of non–ANT-based over ANT-based regimens.

The renewed ENKTCL database from the CLCG included 3046 patients between 2000 and 2015 from 20 institutions. The eligibility criteria for this study included: (1) a newly diagnosed ENKTCL with the typical histological and immunophenotypic evaluations that included CD20/CD79α, CD3ε, CD3s, CD56, cytotoxic molecules (TIA-1, Gram-B, perforin), and Epstein-Barr virus–encoded RNA in situ hybridization, according to the World Health Organization classification of lymphomas; (2) patients received chemotherapy with or without radiotherapy; and (3) complete clinicopathologic and follow-up information. Patients receiving radiotherapy alone (n = 389) or unknown regimens (n = 97) were excluded. Finally, 2560 patients formed the study population. The institutional review boards approved the project and waived the need for informed consent because of the deidentification of patient data.

Clinical staging evaluation and the definition of primary tumor invasion (PTI) have been described previously.¹⁸  Briefly, clinical evaluation included history and physical examination; endoscopy of the upper-aerodigestive tract (UADT); blood biochemistry; computed tomography (CT) scans of the head and neck, chest, abdomen, and pelvis; magnetic resonance imaging of the head and neck; and bone marrow examination. Positron emission tomography (PET) CT has been recommended for all patients since 2010, particularly for those with locally advanced-stage or disseminated diseases. Patients were staged using the Ann Arbor staging system and stratified using 3 ENKTCL-specific models: the nomogram-revised risk index,^3,19  the Korea prognostic index,²⁰  and the prognostic index of natural killer lymphoma.²¹  Quantitative measurement of circulating Epstein-Barr virus DNA was not performed in this study.

Chemotherapy was dichotomized into non–ANT-based (n = 1351, 52.8%) and ANT-based regimens (n = 1209, 47.2%). ANT-based regimens included CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisolone; n = 919) or CHOP-like (CHOP plus etoposide; n = 290), whereas the most commonly used non–ANT-based regimens were ASP (l-ASP or pegaspargase)–containing regimens (n = 1054), followed by platinum-containing (n = 166) and gemcitabine (GEM)–containing (n = 131) regimens. Based on previous systematic review and meta-analysis studies on efficacy and toxicity,^22,23  we further subclassified non–ANT-based regimens into 5 categories (supplemental Table 1): ASP/ANT based (30.3%), ASP/GEM based (25.5%), ASP/methotrexate based (8.8%), ASP/not otherwise specified based (13.4%), and platinum/other regimens (18.9%, usually with GEM). The median number of chemotherapy cycles was 4. Extended involved-site radiotherapy was delivered, with a median dose of 50 Gy.

Primary end points were OS and PFS. OS was calculated from the date of first treatment until the time of death or last follow-up and PFS from the date of first treatment until the date of disease progression, relapse, death, or time of last follow-up. Survival was estimated using the Kaplan-Meier method and compared using a log-rank test. Propensity score matching (PSM) analysis was applied to adjust confounding variables and generate comparable study arms; 1:1 patient matching without replacement was used to pair each patient receiving non–ANT-based regimens with another patient receiving ANT-based regimens whose propensity score was within the designated caliper size. After PSM, baseline covariates and survival rates were compared between chemotherapy groups. Standardized mean difference is used to examine the balance of covariate distribution between treatment groups. Covariates were considered well balanced when the standardized mean difference was <0.10. The measured covariate balances were assessed both graphically and analytically. Cox proportional hazards regression was performed for multivariable analysis in entire group and prespecified subgroups based on ENKTCL-specific models. The interactions between covariates and treatment in Cox model were checked.

The patient characteristics are summarized (Table 1). The median age was 43 years (range, 6-84 years), and the male-to-female ratio was 2.42:1. Most patients had good performance status and primary UADT site (93.8%). Elevated LDH was present in 31.5% of patients, PTI was present in 57.3%, and the majority had early-stage disease (87.0%).

We investigated the survival benefit of non–ANT-based regimens in the entire cohort. With a median follow-up time of 48 months for surviving patients, the 5-year OS and PFS rates were 68.9% (95% confidence interval [CI], 65.9-72.1) and 59.5% (95% CI, 56.3-62.9) for non–ANT-based regimens compared with 57.5% (95% CI, 54.5-60.6; hazard ratio [HR], 0.69; 95% CI, 0.60–0.78; P < .001; Figure 1A) and 44.5% (95% CI, 41.6-47.6; HR, 0.65; 95% CI, 0.57-0.73; P < .001; Figure 1B) for ANT-based regimens.

After adjustment using PSM, the clinical variables were comparable between the 2 groups (Table 1; supplemental Table 2). Non–ANT-based regimens resulted in significantly better OS (HR, 0.64; 95% CI, 0.55-0.75; P < .001; Figure 1C) and PFS (HR, 0.62; 95% CI, 0.54-0.71; P < .001; Figure 1D) than ANT-based regimens. After adjusting for confounding variables, treatments, and time periods via multivariable analysis, in addition to other clinical factors (age, performance status, stage, LDH, regional and distant lymph node involvement, and PTI), both non–ANT-based regimens and radiotherapy were independent prognostic factors for survival (Table 2). The HRs for OS and PFS of radiotherapy vs no radiotherapy were 0.42 (95% CI, 0.35-0.49; P < .001) and 0.38 (95% CI, 0.33-0.44; P < .001). The HRs for OS and PFS of non–ANT-based regimens vs ANT-based regimens were 0.72 (95% CI, 0.60-0.85; P < .001) and 0.63 (95% CI, 0.54-0.73; P < .001). Thus, non–ANT-based regimens are associated with ∼30% improvement in OS and PFS in the entire cohort. Since 26.5% of the patients in the non–ANT-based regimens were treated with CHOP plus ASP (CHOP-ASP), a crude comparison with CHOP and CHOP-ASP was made to address the role of ASP in ENKTCL. The addition of ASP to CHOP was associated with significantly better OS and PFS in the entire cohort (supplemental Figure 1). Explorative analysis of survival outcomes with other non–ANT-based regimens was not performed.

We evaluated the survival benefit of non–ANT-based regimens in different stages, because treatment strategies and prognosis varied between early-stage and advanced-stage diseases. In early-stage disease, the 5-year OS and PFS rates were 73.3% (95% CI, 70.2-76.6) and 64.0% (95% CI, 60.6-67.5) for non–ANT-based regimens compared with 61.0% (95% CI, 57.8-64.2; HR 0.65, 95% CI, 0.55–0.75; P < .001; Figure 2A) and 47.8% (95% CI, 44.7-51.1; HR 0.62, 95% CI, 0.54-0.70; P < .001; Figure 2B) for ANT-based regimens. In advanced-stage disease, the 5-year OS and PFS rates were 39.8% (95% CI, 31.7-49.9) and 30.1% (95% CI, 22.8-39.9) for non–ANT-based regimens compared with 29.9% (95% CI, 22.5-39.8; HR 0.70, 95% CI, 0.51-0.92; P = .013; Figure 2C) and 18.8% (95% CI, 12.8-27.7; HR 0.67; 95% CI, 0.50-0.86; P = .003; Figure 2D) for ANT-based regimens.

PSM adequately balanced the clinical variables between the chemotherapy groups (Table 1; supplemental Table 2). The adjusted HRs for OS and PFS of non–ANT-based regimens vs ANT-based regimens were 0.65 (95% CI, 0.55-0.77; P < .001) and 0.61 (95% CI, 0.53-0.70; P < .001) for early-stage patients and 0.63 (95% CI, 0.44-0.89; P = .009) and 0.58 (95% CI, 0.42-0.80; P = .001) for advanced-stage patients. In multivariable analysis (Table 2), the HRs for OS and PFS were 0.79 (95% CI, 0.64-0.97; P = .021) and 0.67 (95% CI, 0.56-0.79; P < .001) for early-stage patients and 0.55 (95% CI, 0.38-0.78; P = .001) and 0.54 (95% CI, 0.39-0.75; P < .001) for advanced-stage patients. Thus, non–ANT-based regimens provided significantly better survival than ANT-based regimens, independent of stages.

We determined whether the benefit of non–ANT-based regimens existed in early-stage patients receiving radiotherapy. For patients receiving combined modality therapy (CMT), the 5-year OS and PFS rates were 77.0% (95% CI, 73.6-80.2) and 67.7% (95% CI, 64.0-71.5) for the non–ANT-based regimens compared with 65.3% (95% CI, 61.9-68.7; HR; 0.64, 95% CI, 0.54-0.77; P < .001; Figure 3A) and 54.4% (95% CI, 51.0-58.0; HR, 0.65; 95% CI, 0.56-0.76; P < .001; Figure 3B) for ANT-based regimens. For patients receiving chemotherapy only, the 5-year OS and PFS rates were 50.1% (95% CI, 42.0-62.1) and 39.7% (95% CI, 33.5-51.3) for non–ANT-based regimens compared with 36.7% (95% CI, 30.2-47.2; HR, 0.75; 95% CI, 0.55-1.03; P = .078; Figure 3C) and 14.1% (95% CI, 9.7-21.8; HR, 0.55; 95% CI, 0.42-0.71; P < .001; Figure 3D) for ANT-based regimens.

After adjustment with PSM, matched cohorts were well balanced in early-stage disease (supplemental Tables 2 and 3). The adjusted HRs for OS and PFS (non–ANT-based vs ANT-based regimens) were 0.60 (95% CI, 0.49-0.74; P < .001) and 0.59 (95% CI, 0.50-0.71; P < .001) for CMT, and 0.78 (95% CI, 0.55-1.10; P = 0.161) and 0.57 (95% CI, 0.43-0.76; P < .001) for chemotherapy only. In multivariable analysis (supplemental Table 4), the HRs for OS and PFS were 0.64 (95% CI, 0.50-0.83; P = .001) and 0.60 (95% CI, 0.48-0.74; P < .001) for CMT, and 1.11 (95% CI, 0.73-1.69; P = 0.633) and 0.69 (95% CI, 0.48-0.98; P = .037) for chemotherapy only. Thus, non–ANT-based regimens significantly improved the PFS for early-stage ENKTCL, independent of radiotherapy.

In early-stage patients who received either non–ANT-based regimens or ANT-based regimens, chemotherapy alone was associated with significantly inferior survivals compared with CMT (Figure 4). The HRs of CMT vs chemotherapy alone were 0.36 (95% CI, 0.24-0.54; P < .001; Figure 4A) for OS and 0.38 (95% CI, 0.27-0.54; P < .001; Figure 4B) for PFS in the non–ANT-based regimens group; and 0.43 (95% CI, 0.32-0.59; P < .001; Figure 4C) for OS and 0.33 (95% CI, 0.25-0.43; P < .001; Figure 4D) for PFS in the ANT-based regimens group.

The sequences of treatment with radiotherapy and chemotherapy were evaluated in early-stage patients who received either non–ANT-based or ANT-based regimens. The baseline characteristics in each group are listed in supplemental Table 5. In either group, radiotherapy followed by chemotherapy was associated with a better survival than chemotherapy followed by radiotherapy (supplemental Figure 2).

We evaluated the benefit of non–ANT-based regimens in the risk-stratified patients according to 3 ENKTCL-specific models. Figure 5 shows the 5-year OS and PFS rates and HRs of the non–ANT-based regimens vs the ANT-based regimens in each risk subgroup. Non–ANT-based regimens were associated with a significant PFS benefit across all risk subgroups within each model. The HRs for PFS ranged from 0.47 to 0.75 (all P < .05), whereas the HRs for OS varied from 0.47 to 0.80 across these risk subgroups. Thus, non–ANT-based regimens appear to be beneficial in risk-stratified patients.

This is a large-scale, real-world multicenter comparison of non–ANT-based regimens with ANT-based regimens to treat ENKTCL. Non–ANT-based regimens provided OS and PFS benefits over ANT-based regimens, even after adjusting for prognostic confounders and time periods via PSM and multivariable analyses. The clinical advantage of non–ANT-based regimens was consistent across all subgroups examined, regardless of stage and risk subgroup, and was retained for early-stage patients receiving radiotherapy.

Treatment of ENKTCL has evolved, with the introduction of up-front modern radiotherapy and non–ANT-based chemotherapy that have improved survival outcomes.^2,4-18,24-28  However, there are no published data describing the benefit of non–ANT-based regimens in a randomized controlled trial or a large multicenter comparative study. The present comprehensive study from real-world data represents a critical step toward understanding the effect of non–ANT-based regimens on long-term survival. Our study provided direct comparison information with abundant sample size, which allowed statistical adjustments for potentially confounding factors, and showed a decreased mortality risk of ∼30% using non–ANT-based regimens in the entire cohort. Non–ANT-based regimens significantly improved PFS and OS for ENKTCL. The beneficial effect of non–ANT-based regimens remained consistent after adjustments using PSM and multivariable analyses and persisted within different stages and individual risk groups. The survival benefit profiles of non–ANT-based regimens after risk stratification could be used for risk-adapted therapy deintensification or intensification in ENKTCL. The survival gain in this large cohort of patients across a substantial number of institutions demonstrated the efficacy and feasibility of non–ANT-based chemotherapy for ENKTCL.

Improvements in long-term survival have been reported in the first-line treatment of early-stage patients over the last decade, mainly as a result of adding radiotherapy into the less effective ANT-based regimens.^2-5  In the modern chemotherapy era, radiotherapy remained an essential component of first-line therapy for early-stage ENKTCL,^24-27  even after a complete response to ASP-based regimens.²⁷  Here, we further demonstrated that non–ANT-based chemotherapy (mostly with radiotherapy) significantly improved survival outcomes for stage I–II patients. The 5-year OS rate was 73.3% in this study, comparable to the recent large multicenter studies from Japan (72%)²⁹  and Asian joint data (∼74% to 79%),³⁰  but higher than the International T-cell Project registry data (median OS, 59 months),³¹  probably because of more heterogeneous treatments in the latter study. Interestingly, survival outcomes in these large series were similar to that of prospective phase 1/2 trials (3-year OS, 66%-87.5%; 5-year OS, 60%-82.1%) and retrospective studies (∼75%; Table 3).^{7-10,12-14,26,29-40}  Furthermore, the survival benefits of non–ANT-based regimens in this study were prominent in early-stage patients who underwent CMT vs chemotherapy only. In our previous study,²⁴  non–ANT-based regimens and radiotherapy were associated with higher conditional survival and lower annual failure hazard in early-stage patients. Consistent with the International T-cell Project registry study,³¹  even with more effective non–ANT-based regimens, chemotherapy alone still resulted in inferior outcomes in early-stage patients, with both 5-year OS and PFS rates of only 50.1% and 39.7% in the present CLCG study. In our previous study, up to 46.9% of early-stage patients who achieved a complete response with ASP-based chemotherapy developed disease relapse.²⁷  The current study also demonstrated that radiotherapy followed by chemotherapy provided survival benefit over chemotherapy followed by radiotherapy for early-stage ENKTCL in the CMT setting with ANT-based or non–ANT-based regimens. These results emphasize the important role of radiotherapy for early-stage patients in the modern chemotherapy era. On the other hand, patients with advanced-stage disease had extremely poor prognoses after non–ANT-based regimen chemotherapy (Table 3), with a median OS of 5.4 to 36.6 months and a 5-year OS of <40% in these prospective and retrospective studies.^29,31,41-46  Furthermore, the risk of death or progression is variable due to the interactions between clinical stages and treatments in patients with relapsed/refractory ENKTCL.^{7,15,16,47-49}  As indicated in Table 3, the favorable outcomes of non–ANT-based regimen chemotherapy (5-year OS, 55%-66.9%) reported in these patients are partly attributable to the inclusion of those with localized-relapsed/refractory^{7,15,16,47-49}  or newly diagnosed early-stage ENKTCL^{21,29,31,50-52}  treated with effective radiotherapy. These findings support further consideration of more effective systematic therapy or innovative treatment strategy for patients with advanced-stage or disseminated diseases and the use of risk-adapted therapy involving radiotherapy and better-tolerated non–ANT-based chemotherapy for patients with early-stage or localized diseases.

The rarity of the disease and lack of randomized controlled trial data have resulted in a variety of non–ANT-based regimens for ENKTCL being developed and recommended by guidelines and adopted in clinical practice among the CLCG institutions. Although treatment options and non–ANT-based regimens varied across countries and even within institutions, consistent with other studies,^29-31  ASP-based regimens have formed the mainstay of chemotherapy across China over the past decade. The wide utility of ASP-based regimens in the world indicated its feasibility in clinical practice and prompted their further validation and optimization. Interestingly, the addition of ASP to CHOP was associated with significantly improved survival compared with CHOP, indicating the important role of ASP in ENKTCL. However, only a small proportion of patients received first-line ASP/methotrexate–based regimens such as SMILE (l-ASP, ifosfamide, methotrexate, etoposide, and dexamethasone) in this study, probably because of its highly severe toxicities.^7,15,16  The heterogeneity of ASP-based regimens reflects the evolving practice pattern of chemotherapy regimens, associated with a lack of a general consensus on the standard regimens for ENKTCL. Further work is needed to determine the optimal non–ANT-based regimens.

This retrospective study had several limitations. First, the chemotherapy regimens were not randomly assigned; therefore, the results may be affected by selection bias. However, given the recognition of ENKTCL resistant to ANT-based regimens, a further prospective study designed as a direct randomized comparison of ANT-based and non–ANT-based regimens may raise ethical concerns. Second, non–ANT-based regimens were applied in more recent years, accompanied with more accurate imaging such as PET/CT or magnetic resonance imaging and advanced radiotherapy techniques.²⁸  As PET/CT scan was routinely used for staging of patients after 2010, the majority of patients recruited from 2000 to 2009 were therefore not staged properly according to the current standard. This would result in misleading conclusion on the effect of disease stage. We attempted to circumvent these limitations using PSM and multivariable analyses involving time period (2000-2004 vs 2005-2009 vs 2010-2015) and prognostic variables. However, even with PSM analysis, there remains a chance that the underlying confounding factors may have influenced the results. Some other factors, such as treatment toxicity and patient or physician preference, were not completely evaluated. Third, we only focused on effectiveness; however, the toxicity profiles of non–ANT-based regimens could be considered when dealing with different-stage patients or elderly patients. Finally, histologic results were not centrally reviewed in the CLCG; this may lead to misdiagnosis in few cases. However, with distinctive clinicopathological and immunophenotypic findings defined in the World Health Organization classification, ENKTCL is easily distinguished from other cytotoxic T-cell lymphomas.⁵³  We believe this is not likely to significantly affect our conclusions.

In summary, we demonstrated that non–ANT-based regimens significantly improved survival compared with ANT-based regimens, regardless of stage, risk subgroup, or radiotherapy. These findings provide additional evidence supporting non–ANT-based regimens as first-line treatments for ENKTCL.

The authors thank Ying Huang (Center of Cancer Research, National Cancer Institute, Bethesda, MD) for her critical review of the manuscript.

This work was supported by grants from the Chinese Academy of Medical Science Innovation Fund for Medical Sciences (CIFMS; 2016-I2M-1-001), the National Key Projects of Research and Development of China (2016YFC0904600), Capital’s Funds for Health Improvement and Research (grant 2018-1-2151), and the National Natural Science Foundation of China (grants 81670185 and 81870154).

Contribution: S.-N.Q., Y.Y., Y.-Q.S., J.Z., and Y.-X.L. designed the study, analyzed the data, and wrote the manuscript; S.-N.Q., Y.Y., Y.-Q.S., C.H., J.Z., and Y.-X.L. contributed to the study concept; Y.-X.L., J.Z., Y.-Q.S., S.-N.Q., and Y.Y. contributed to the study coordination; S.-N.Q., Y.Y., C.H., and Y.-X.L. performed the statistical analysis; and all authors contributed to data collection and interpretation, and finally approved the manuscript.

Conflict-of-interest disclosure: C.H. has received grants from the National Institutes of Health/National Cancer Institute and personal fees from Varian Medical Systems outside the submitted work. The remaining authors declare no competing financial interests. A complete list of the members of the China Lymphoma Collaborative Group appears in “Appendix.”

Correspondence: Ye-Xiong Li, State Key Laboratory of Molecular Oncology, Department of Radiation Oncology, National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, PanJiaYuan NanLi #17, ChaoYang District, Beijing 100021, People’s Republic of China; e-mail: yexiong12@163.com or yexiong@yahoo.com; and Jun Zhu, Key Laboratory of Carcinogenesis and Translational Research, Ministry of Education, Peking University Cancer Hospital & Institute, FuCheng Rd #52, HaiDian District, Beijing 100142, People’s Republic of China, e-mail: zhu-jun2017@outlook.com.

The members of the China Lymphoma Collaborative Group are: Y.-J.Z., Y.-X.L., J.Z., T.W., J.-Z.C., Y.Z., X.-Y.Q., M.S., H.S., Y.W., S.-Y.Z., Z.-Y.Y., X.H., G.W., H.W., B.-L.Q., J.-X.W., F.-Q.Z., L.-T.Q., and G.-F.L.