Association of Improved Locoregional Control With Prolonged Survival in Early-Stage Extranodal Nasal-Type Natural Killer/T-Cell Lymphoma

Key Points

Question  Does improved locoregional control translate into a survival benefit in patients with early-stage extranodal natural killer/T-cell lymphoma (NKTCL)?

Findings  In a cohort study of 1332 patients in China with early-stage NKTCL, radiotherapy had a dose-dependent association with locoregional control, progression-free survival, and overall survival. Improved locoregional control was associated with prolonged progression-free survival and overall survival.

Meaning  Clinicians and patients should be aware of the essential role of radiotherapy in both locoregional disease control and maintaining long-term survival.

Importance  The long-term survival benefit for radiotherapy (RT) in early-stage extranodal natural killer/T-cell lymphoma (NKTCL) is not known, and it is unclear whether improved locoregional control (LRC) translates into a survival benefit.

Objective  To investigate the dose-dependent effect and potential survival benefits of RT on the basis of LRC improvements.

Design, Setting, and Participants  Review of clinical data of patients with early-stage NKTCL at 10 institutions in China between 2000 and 2014. Radiotherapy dose as a continuous variable was entered into the Cox regression model by using penalized spline regression to allow for a nonlinear relationship between RT dose and events. Regression analysis was used to assess whether a linear correlation exists between LRC and progression-free survival (PFS) or overall survival (OS). Patients received chemotherapy (CT) alone, RT alone, or a combination. Chemotherapy alone was defined as 0 Gy.

Main Outcomes and Measures  The association between LRC and OS or PFS.

Results  A total of 1332 patients (923 [69%] male; median age, 43 years [range, 2-87 years]) were reviewed. For patients treated with RT, median dose was 50 Gy (range, 10-70 Gy); 996 (86%) received at least 50 Gy, and 164 (14%) received 10 to 49 Gy. The risk of locoregional recurrence, disease progression, and mortality decreased sharply until 50 to 52 Gy. For patients receiving RT, high-dose RT (≥50 Gy) was associated with significantly better 5-year LRC (85% vs 73%; P < .001), PFS (61% vs 50%; P = .004), and OS (70% vs 58%; P = .04) than low-dose RT (<50 Gy). Improved LRC with high-dose RT was independent of the RT/CT sequence or initial response to CT. Radiotherapy yielded a dose-dependent effect on LRC (range, 41%-87%), PFS (18%-63%), and OS (33%-71%). Dose-response regression analysis revealed a linear correlation between 5-year LRC and 5-year PFS (correlation coefficient, r = 0.994, P < .001; determination coefficient, R² = 0.988) or 5-year OS (r = 0.985, P = .002; R² = 0.97), which was externally validated using published data.

Conclusions and Relevance  The optimal dose was 50 Gy for patients with early-stage disease. The improved LRC was associated with prolonged survival. These findings emphasize the importance of RT in optimizing first-line therapy, and provide evidence for making treatment decisions and designing clinical trials.

Extranodal nasal-type natural killer/T-cell lymphoma (NKTCL) is rare in Western countries but relatively common in East Asia and South America.^1,2 It has been recognized as a new entity in the World Health Organization classification.³ In contrast to other aggressive lymphomas, NKTCL is associated with the Epstein-Barr virus, frequently involves the upper aerodigestive tract, and has an aggressive natural locoregional history (most patients present with localized disease) and propensity for extranodal spread.^4-10 Reported 5-year overall survival (OS) rates vary from 40% to 90% for early-stage disease and 10% to 30% for advanced-stage disease, reflecting varied stage selections, treatment options, radiation doses, and chemotherapy regimens.^1,5,8,11-19

The standard of care for the most common type of aggressive lymphoma, diffuse large B-cell lymphoma (DLBCL), is chemotherapy (CT) with optional radiotherapy (RT). Given that DLBCL is chemosensitive, CT alone achieves a high complete response (CR) rate and cures most cases.^20-22 The CR rate after CT is a valid end point for predicting survival; RT is only considered as consolidation therapy following CT in patients achieving complete or partial response.^21,22 In contrast, NKTCL is radiosensitive but chemoresistant. Radiotherapy is the most effective single modality in terms of locoregional control (LRC)^23,24; however, the long-term survival benefit of RT in early-stage NKTCL remains a matter of debate due to the frequency of systemic relapse.²⁵ As the primary goal of RT is to achieve uncomplicated LRC, it is unclear whether improved LRC translates into a survival benefit by preventing systemic dissemination or postponing death.

Upfront RT, risk-adapted therapies, and more effective CT have improved outcomes in early-stage NKTCL.^{6-10,18,26,27} High LRC (approximately 90%) is associated with favorable survival (5-year OS, approximately 70%) when modern RT techniques are applied using the proper field and dose.^{6-10,12,13,24,26} However, low-dose, small-field RT is not sufficient for disease control, with up to 50% of patients experiencing local recurrence and only 40% surviving at 5 years.^14,28-31 Current recommendations for multidisciplinary care including RT are largely based on small cohort studies, extrapolated from patients with DLBCL, or limited by patient selection criteria and different RT fields and arbitrary RT dose cutoff points.^14,15,30-34

Based on the uncertainties regarding optimal therapy, we hypothesized that improved LRC is associated with survival and disease control benefits in early-stage NKTCL. To investigate the association between improved LRC and survival, we evaluated the dose effects of RT on LRC and survival outcomes using modern statistical methods in a large cohort of patients, and validated this relationship using external data.

The cohort comprised 1332 patients with early-stage NKTCL at 10 institutions in China between 2000 and 2014. The project was approved by institutional review board with written informed consent in accordance with the Declaration. Routine staging methods were described previously.¹⁰ Positron emission tomography was performed in 270 patients. Primary tumor invasion (PTI) was defined as the presence of primary disease that extended into adjacent structures or organs, or the involvement of multiple, contiguous primary sites.⁹

Patient characteristics (stratified by RT dose) are summarized in eTable 1 in the Supplement. Median age was 43 years (range, 2-87 years). Elevated lactate dehydrogenase (LDH) levels were present in 423 (32%) patients; PTI, in 734 (55%); and most (987 [74%]) had stage I disease.

Patients received CT followed by RT (CT + RT, n = 653), RT followed by CT (RT + CT, n = 215), RT alone (n = 292), or CT alone (n = 172). Radiotherapy included an extended involved field encompassing primary tumor and adjacent sites. Median dose was 50 Gy (range, 10-70 Gy). Most (996 [86%]) received at least 50 Gy; 164 (14%) received 10 to 49 Gy; and 53 (4%) received less than 40 Gy (eTable 1 in the Supplement). Chemotherapy alone was defined as 0 Gy for dose-effect analysis. Of the 1040 patients receiving CT, 832 (80%) received doxorubicin-based regimens; 208 (20%) received L-asparaginase–based or gemcitabine-based regimens. The median number of CT cycles was 4 (range, 1-14).

Based on the international response criteria for non-Hodgkin lymphoma,³⁵ CR was defined as complete disappearance of all detectable clinical and radiographic evidence of disease. Locoregional control was measured from first therapy to locoregional recurrence (LRR), regardless of distant lymph node recurrence and/or extranodal recurrence; OS, to death, regardless of cause; and progression-free survival (PFS), survival without evidence of relapse, progression, or death. Locoregional control and survival were calculated using Kaplan-Meier analysis; subgroups were compared using the log-rank test. The clinical features of the different subgroups were compared using χ² analysis. Cox proportional hazards regression was performed to identify independent risk factors for LRC and survival. The selection of the final model was performed using a backward step down-selection process. A 2-sided P value of <.05 was considered significant.

The penalized spline (P-spline) fit in the Cox model allows a nonlinear relationship of RT dose with the logarithm (ln hazard ratio [HR]) of LRR, disease progression, or mortality, estimated from the full Cox regression model adjusted for all covariates.³⁶ We used the dfmacox (degrees of freedom in multivariate additive Cox models) function in smoothHR to obtain the optimal number of degrees of freedom in the extended Cox-type additive multivariate analysis.³⁶ Regression analysis was used to assess whether a linear relationship exists between LRC and OS or PFS.³⁷ Cox proportional hazards regression was performed using the rms package and P-spline using the smoothHR package in R, version 3.2.3 (http://www.r-project.org/).

To quantify the dose-dependent effect, we entered RT dose as a continuous variable into the Cox regression using P-splines in smoothHR to allow for nonlinear relationships between RT dose and end points. This model confirmed that the risk (ln HR) of LRR, disease progression, and death decreased sharply until 50 to 52 Gy, then increased slightly (Figure 1). The dose-dependent effect of RT was more constant for PFS and OS than LRC. After adjustment for all covariates (age, sex, Eastern Cooperative Oncology Group performance status [ECOG PS], B symptoms, LDH, stage, PTI) in multivariable analysis, the HR curves revealed similar trends for LRC, PFS, and OS (eFigure in the Supplement). These findings indicate that 50 to 52 Gy is an optimal dose and confirm the dose-dependent effect of RT on LRC, PFS, and OS.

The prognostic significance of clinical characteristics and RT dose were evaluated in univariate analysis (eTable 2 in the Supplement). Chemotherapy alone (0 Gy, n = 172) resulted in significantly inferior 5-year LRC (41% vs 83%; P < .001), PFS (18% vs 59%; P < .001), and OS (33% vs 68%; P < .001), compared with RT with or without CT (n = 1160). We evaluated whether RT can be omitted in patients who achieve CR after initial CT (n = 234). Compared with CT alone (n = 53), CT + RT (n = 181) resulted in significantly higher 5-year LRC (84% vs 47%; P < .001), PFS (67% vs 33%; P < .001), and OS (75% vs 55%; P = .02) after CR, indicating that RT is essential in the first-line management of early-stage NKTCL.

Based on an optimal dose of 50 Gy, we divided patients treated with RT into 2 groups: high dose (≥50 Gy, n = 996) and low dose (<50 Gy, n = 164). Radiotherapy dose remained independently associated with LRC and survival outcomes in multivariate analyses (Figure 2A). Low-dose RT resulted in significantly higher HRs than high-dose RT. Five-year LRC, PFS, and OS were 85%, 61%, and 70% for the high-dose group and 73% (P < .001), 50% (P = .004), and 58% (P = .04) for the low-dose group. Moreover, PS and PTI were independent risk factors for LRC; additional prognostic factors including age, ECOG PS, stage, LDH, and PTI significantly influenced PFS and OS. These findings indicate that RT dose is an independent factor for LRC, PFS, and OS in early-stage NKTCL after RT.

Subgroup analyses according to treatment revealed that reduced-dose RT (<50 Gy) resulted in significantly lower LRC than high-dose RT (≥50 Gy) after RT + CT or CT + RT (Figure 2B). For patients treated with RT + CT, 5-year LRC was 89% for 50 Gy or more, compared with 78% for less than 50 Gy (P = .003). For CT + RT, 5-year LRC was 82% for 50 Gy or more and 72% for less than 50 Gy (P = .02).

Among patients who achieved a CR after initial CT and went on to receive RT (n = 181), reduced RT dose yielded a significantly lower LRC rate, with 5-year LRC of 88% for 50 Gy or more vs 73% for less than 50 Gy (P = .04) (Figure 2C). Nonsignificantly improved LRC for high-dose RT was observed after RT alone (P = .052) and in patients who did not achieve CR after CT (P = .07). Most clinical characteristics were comparable between groups (eTable 3 in the Supplement). These findings indicate that RT dose is an important risk factor for LRC, irrespective of treatment or initial response to CT.

To determine the association of LRC with survival, RT doses were divided into 5 groups for all patients (eTable 1 in the Supplement). Five-year LRC, PFS, and OS ranged from 41% to 87%, 18% to 63%, and 33% to 71% (P < .001) (eTable 2 in the Supplement), respectively, according to RT dose (0-70 Gy). Dose-response linear regression analysis revealed a strong correlation between 5-year LRC and 5-year PFS (correlation coefficient, r = 0.994, P < .001; determination coefficient, R² = 0.988, PFS = 0.954 × LRC – 19.59) or 5-year OS (r = 0.985, P = .002; R² = 0.97, OS = 0.879 × LRC – 4.084) (Figure 3). As estimated from the dose-response predictive models, an absolute gain of 10% LRC provides a 9.54% improvement in PFS and 8.79% improvement in OS, indicating that improvement in LRC is associated with increased survival probabilities.

To validate our findings, we analyzed the relationship between LRC and survival outcomes using published data. After searching PubMed/MEDLINE using the combined terms “NK/T-cell lymphoma” or “nasal lymphoma,” 3 authors (Y.Y., S.N.Q., Y.X.L.) read all publications by title and abstract, then the full text. Eligible studies had to meet the following criteria: early-stage NKTCL and reports of crude or cumulative LRC/LRR or local control (LC)/local recurrence (LR) and OS. A total of 3438 patients derived from 31 studies were included (eTable 4 in the Supplement).^{12-15,23,24,26-31,33,38-54} Patients received RT, CT, or a combination of RT and CT, with a variety of RT doses and CT regimens. This analysis confirmed the correlations between LRC/LC and PFS (r = 0.875, P < .001, R² = 0.765) (Figure 4A) or OS (r = 0.891, P < .001, R² = 0.794) (Figure 4B).

Using the dose-response predictive models (Figure 3), predicted PFS and OS were calculated for each study using actual LRC/LC rates (eTable 4 in the Supplement), then compared with actual PFS or OS using regression analysis. As it approaches the diagonal line, predicted PFS or OS approximated to actual PFS (r = 0.849, P < .001) (Figure 4C) or OS (r = 0.879, P < .001) (Figure 4D), validating our finding that LRC is predictive of PFS and OS.

This is the first study to quantify the increased risk of disease events after therapy for a specific RT dose (as a continuous covariate) or evaluate the impact of LRC on PFS and OS in early-stage NKTCL. We identified several important findings in a large, real-world cohort of well-characterized patients. Radiotherapy exerted a dose-dependent effect on LRC and survival. Radiotherapy dose was an important prognostic factor for LRC, PFS, and OS in univariate and multivariate analyses. Improved LRC due to high-dose RT (≥50 Gy) was independent of treatment or initial response to CT. Ultimately, we observed a significant association between improved LRC and survival benefits in early-stage NKTCL, which was externally validated by analyzing published data. These findings provide a strong argument for the frontline use of optimal RT to expedite the development of breakthrough therapies.

This study represents a critical step toward understanding the dose-dependent effect of RT on LRC and survival outcomes in early-stage NKTCL. The effectiveness of RT for improving LRC or LC is well established for DLBCL and NKTCL.^21-24 However, considering aggressive lymphoma as a systemic disease, the question of whether improved LRC leads to disease-specific survival benefits remains. No matter how effective RT is with respect to LRC and LC, primary RT in chemoresistant NKTCL or the addition of RT to CT in chemosensitive DLBCL will provide no survival benefit if disease is not well controlled systemically. Research into these topics is essential to optimize treatment strategies and individualize RT treatment plans for aggressive lymphomas with different chemosensitivities. In early-stage NKTCL, optimization of first-line therapy is critical because relapsed and disseminated NKTCL have a poor prognosis, with only 29% 3-year OS, even after more effective L-asparaginase–based CT.^19,55 We demonstrate that RT not only improved LRC but also prolonged survival in a dose-dependent manner in early-stage NKTCL. Below the threshold of 50 to 52 Gy, LRC and survival outcomes increased with increasing RT dose (Figure 1). The lack of a significant impact of RT dose on LRC and survival outcomes at a higher dose level (60-70 Gy) may be attributed to selection bias toward patients with residual or bulky tumors containing higher numbers of radioresistant or stem cells.⁵⁶ Given the difficulty of defining RT dose in clinical practice, our finding supports the notion that 50 Gy should be the prescribed dose for early-stage NKTCL, with a boost of 5 to 10 Gy for residual disease (range, 50-60 Gy).

For patients with early-stage NKTCL treated with RT, high-dose RT (≥50 Gy) resulted in better LRC, PFS, and OS than low-dose RT (<50 Gy). Multivariate analyses revealed that RT dose was an independent prognostic factor for LRC, PFS, and OS. Moreover, LRC was largely dependent on the RT dose and primary tumor burden (Figure 2A), but regardless of CT and/or RT sequence (Figure 2B) or initial response to CT (Figure 2C), although these systemic factors influenced survival outcomes in high-risk early-stage patients.^10,45 This raises the possibility that CT improves survival by acting on distant micrometastatic disease more so than via local effects. In view of the recognized high cure rate of RT and evidence for a dose-response relationship, adherence to RT plays a greater role in both LRC and survival. Clinical trials designed to define optimal therapies or novel CT regimens without inclusion of RT may not be acceptable or ethical in early-stage NKTCL. A combination of upfront, optimal RT and effective systemic therapy may provide the ideal spatial combination for first-line therapy in early-stage NKTCL.

This large cohort study and literature-based systematic review robustly confirm a significant correlation between improved LRC and survival benefits in early-stage NKTCL. Progression-free and overall survival can be deduced from the dose-response predictive models for LRC (Figure 3). A 5-year OS rate of at least 75% could be predicted if LRC was at least 90%. Based on these data, achieving at least 90% 5-year LRC is possible and acceptable for radiosensitive NKTCL when irradiated at the appropriate dose. Importantly, the positive relationships between LRC and survival outcomes were externally validated using published data including patients from different countries who had varied treatments, RT doses, CT regimens, and follow-up times (Figure 4, eTable 4 in the Supplement). Given the unsatisfactory LRC (<90%) rates of some previous studies,^{14,15,29-31,33,40} the proportional survival gains provided by RT may be greater for current patients with early-stage disease than in previous studies, as treatment strategies have changed substantially and patients currently receive upfront RT at the optimal dose with better target coverage, as well as benefiting from more effective CT and advanced imaging techniques.

Because the relationship between LRC and survival in early-stage NKTCL is straightforward, the use of LRC as an end point not only incorporates survival, but also reduced disease relapse and progression. Therefore, LRC is an important end point of treatment efficacy in early-stage NKTCL. Compared with OS, assessment of LRC has a lower likelihood of confounding by subsequent or salvage therapies. Innovative treatment strategies with a large magnitude of effect on LRC may have a large effect on OS. In contrast, for chemosensitive DLBCL, the majority of patients achieve CR after CT and maintain long-term disease-free survival.^20-22 Thus, the therapeutic efficacy of CT in DLBCL mostly relies on initial response.^20-22 In contrast, most patients with early-stage NKTCL who achieve CR with CT alone eventually develop LRR and disease progression. The frequent tumor regrowth observed in complete responders indicates that NKTCL exhibits secondary resistance to CT, and CT alone is insufficient to provide long-term disease remission in most cases. Furthermore, patients who achieve CR still need to receive a high dose (≥50 Gy vs <50 Gy) to achieve better LRC, indicating that determination of prescribed RT dose by assessing CR to CT may have limited value. As testing of novel agents or systemic therapy for aggressive lymphomas is mainly based on CR or even overall response, but not LRC, this method carries the substantial risk that new CT regimens may be optimized in terms of their effect on CR, without significant improvement in curative potential for early-stage NKTCL. Therefore, in contrast to DLBCL, optimization of first-line therapy for early-stage NKTCL should aim to permanently improve LRC and, as outlined herein, eventually improve long-term survival.

This study has several limitations. First, as this study was limited to early-stage NKTCL, the value of RT in advanced-stage disease is not clear, although 1 study that added consolidation RT to primary CT reported better outcomes.¹⁷ However, it should be emphasized that in contrast to chemosensitive DLBCL (in which all stages are managed using primary CT), advanced-stage NKTCL should be managed differently from early-stage disease because RT is curative for early-stage but not disseminated NKTCL. Second, the effect of the RT field and technique on LRC and survival outcomes was not evaluated because of nonuniform data collection on target delineation from participating institutions, although the prognostic importance of RT field was established in earlier studies.^24,33 Other limitations included no information on positron emission tomography, some potential biomarkers such as Epstein-Barr virus DNA, toxicity, and potential benefit of concurrent chemoradiotherapy. Future research is required to explore LRC and survival models incorporating modern RT and imaging techniques, molecular predictors, and consolidation principles for advanced NKTCL.

Improvements in LRC provided by RT translate into significant PFS and OS benefits in early-stage NKTCL. Clinicians and patients should be aware of the essential role of RT in both locoregional disease control and maintaining long-term survival. These findings will help to define the standard of care and assist decision making and prospective clinical trial design.

Accepted for Publication: September 19, 2016.

Corresponding Author: Ye-Xiong Li, MD, Department of Radiation Oncology, National Cancer Center, Cancer Hospital and Institute, Peking Union Medical College (PUMC) and Chinese Academy of Medical Sciences (CAMS), Collaborative Innovation Center for Cancer Medicine, Beijing 100021, PR China (yexiong12@163.com).

Published Online: November 17, 2016. doi:10.1001/jamaoncol.2016.5094

Author Contributions: Dr Li had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Yang and Cao contributed equally to this work.

Study concept and design: Yang, Cao, Lan, Qian, Hou, F.-Q. Zhang, Y.-J. Zhang, Y. Zhu, Yuan, Li.

Acquisition, analysis, or interpretation of data: Yang, Cao, J.-X. Wu, T. Wu, S.-Y. Zhu, Hou, Y.-J. Zhang, Xu, Qi, Li.

Drafting of the manuscript: Yang, Cao, Hou, Li.

Critical revision of the manuscript for important intellectual content: Yang, Lan, J.-X. Wu, T. Wu, S.-Y. Zhu, Qian, Hou, F.-Q. Zhang, Y.-J. Zhang, Y. Zhu, Xu, Yuan, Qi, Li.

Statistical analysis: Yang, S.-Y. Zhu, Hou, Xu, Qi, Li.

Obtained funding: Xu, Li.

Administrative, technical, or material support: Yang, Cao, Lan, J.-X. Wu, T. Wu, S.-Y. Zhu, Qian, F.-Q. Zhang, Y.-J. Zhang, Xu, Li.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by the National Key Projects of Research and Development of China (2016YFC0904600).

Role of the Funder/Sponsor: The National Key Projects of Research and Development of China had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.