第一部分 含不良预后因素的早期结外NK/T细胞淋巴瘤综合治疗模式的随机对照研究及预后分析

【目的】目前，含不良预后因素的早期结外NK/T细胞淋巴瘤（extranodal natural killer/T cell lymphoma，NKTCL）的最佳治疗策略尚未完全明确。我们探索了调强放射治疗（intensity modulated radiation，IMRT）序贯含吉西他滨、地塞米松、顺铂（GDP）药物的化疗方案联合或不联合西达本胺一线治疗含不良预后因素的早期上呼吸消化道（upper aerodigestive tract，UADT）-NKTCL的疗效和安全性。

【方法】这是一项在中国开展的双中心、开放标签、随机对照II期临床研究。本研究计划共入组98例患者。主要纳入条件为：（1）新诊断的早期NKTCL患者；（2）根据列线图简化风险指数，患者至少含有下列一个危险因素：＞60岁、血清乳酸脱氢酶升高、原发肿瘤侵犯、区域淋巴结受累（分期为Ⅱ期）或东部肿瘤协作组体力状态（Eastern Cooperative Oncology Group Performance Status，ECOG PS）评分＞1。患者接受IMRT序贯GDP化疗方案一线治疗，联合或不联合西达本胺。符合条件的患者以1:1的比例随机分配到西达本胺组或对照组。主要研究终点为2年无进展生存（progression-free survival，PFS）率。次要研究终点包括客观缓解率（objective response rate，ORR）、2年总生存（overall survival，OS）率和治疗相关的毒副作用。

【结果】从2015年5月至2019年12月，最终共有74例符合条件的患者纳入本研究。研究进行过半时，我们对已经入组的58例患者进行了中期分析，但未观察到两组患者PFS和OS的显著差异。鉴于随访时间较短，试验进行到三分之二时，我们再次进行分析，仍未观察到研究主要终点和次要终点的显著获益趋势。因此，考虑到西达本胺的使用可能导致不良事件发生率增加，我们提前终止了试验招募。74例患者中，37例患者接受了IMRT+GDP+西达本胺方案治疗（西达本胺组），另37例患者接受了IMRT+GDP方案治疗（对照组）。中位随访时间为43.4个月（范围1.0-74.6）。西达本胺组ORR为86.5%，对照组为78.4%（p = 0.359）。西达本胺组的 2 年 OS 率和 PFS 率分别为 89.2% 和 75.2%，而对照组分别为 83.8% （p = 0.388）和 70.2%（p = 0.821）。条件生存分析显示，存活超过7个月的患者中，西达本胺组的生存率较对照组更优（2年OS率，100% vs. 88.6%，p = 0.028）。在两组中，主要的不良反应为血液学毒性和口腔黏膜炎，且在两组间无显著统计学差异（p＞0.05）。多因素分析表明血清LDH升高和ECOG 评分＞1是PFS的独立预后因素（HR：2.653，95%CI：1.119-6.291，p = 0.027；HR：3.618，95%CI：1.515-8.645，p = 0.004）。

【结论】在含不良预后因素的早期NKTCL患者中，IMRT序贯GDP化疗方案联合西达本胺相较于不联合西达本胺没有显著生存获益，两种治疗方案的毒性没有显著差异。血清LDH和ECOG 评分对评估NKTCL患者预后有一定的预测价值。

第二部分  利用全外显子测序技术探索结外NK/T细胞淋巴瘤治疗及预后生物标志物

【目的】目前，对于结外NK/T细胞淋巴瘤（extranodal natural killer/T cell lymphoma，NKTCL），缺乏广泛适用于临床实践的预后分子标志物。本研究通过全外显子测序技术分析NKTCL患者的基因组数据和突变特点，旨在探讨肿瘤遗传背景对临床疗效的影响，并寻找可能影响NKTCL治疗及预后的生物标志物。

【方法】本研究纳入了中国医学科学院肿瘤医院和北京大学肿瘤医院提供的33例NKTCL患者样本，对样本进行了全外显子测序并对数据进行了生物信息学分析。根据患者是否在治疗/免疫治疗后获得完全缓解，将患者分成治疗敏感组和非治疗敏感组/免疫治疗敏感组和非免疫治疗敏感组。通过Fisher精确检验筛选差异基因，并研究它们与临床特征和预后的相关性。然后，运用R软件包“musicatk”、“deconstructSigs”和“decompTumor2Sig”进行突变频谱印记分析，探索突变特征与临床特征、治疗效果及预后之间的关联。其次，评估NKTCL样本的等位基因突变的肿瘤异质性（mutant allele tumor heterogeneity，MATH）评分，量化NKTCL样本的肿瘤内异质性，并研究MATH评分与临床特征、治疗效果和预后的相关性。

【结果】33例NKTCL患者中位肿瘤突变负荷（tumor mutation burden，TMB）为4.66/Mb（范围0.22-8.76）。患者根据治疗后的疗效评价被分为治疗敏感组（CR组，n = 23）和非治疗敏感组（non-CR组，n = 10），或免疫治疗敏感组（iCR组，n = 5）和非免疫治疗敏感组（non-iCR组， n = 4）。在non-CR组中，ZXDA（p = 0.036）、HAPLN2（p = 0.036）、COL4A3（p = 0.036）、MGA（p = 0.036）、LAMA1（p = 0.036）基因突变率显著高于CR组，non-iCR组中ANK2（p = 0.047）基因突变率显著高于iCR组。预后分析显示，MGA（p = 0.005）和LAMA1（p = 0.031）突变与NKTCL较短的总生存期（overall survival，OS）显著相关。此外，本研究揭示了C>T/G>A变异为NKTCL最常见的体细胞单核苷酸变异，占比41.8%，且94%的患者具有COSMIC signature 1特征。45.4%的患者中具有COSMIC signature 6特征，该群体的OS（p = 0.011）和无进展生存期（progression-free survival，PFS）（p = 0.024）显著短于不具有该特征的患者。33例患者的中位MATH评分为49.4，基于此评分，患者被分为高MATH组与低MATH组，其中低MATH组患者的PFS及OS均显著优于高MATH组患者（p = 0.023；p = 0.026）。

【结论】本研究识别了与NKTCL患者治疗反应和预后密切相关的关键基因和遗传特征，包括MGA和LAMA1基因突变，以及COSMIC signature 6突变特征，这些因素均有可能作为NKTCL不良预后的潜在标志。此外，通过MATH评分进行NKTCL肿瘤内异质性的量化分析，揭示了高MATH评分与不良预后之间的关联。这些发现为NKTCL的个体化治疗提供了重要的分子依据，并发现了NKTCL潜在的生物标志物，未来可针对本研究发现进行进一步探索。

第三部分  肿瘤知情和肿瘤未知液体活检方法在结外NK/T细胞淋巴瘤患者预后评估中的应用：三个病例报告

【目的】比较肿瘤知情ctDNA（tumor-informed ctDNA-MRD）和肿瘤未知ctDNA（tumor-naïve ctDNA）两种检测方法在结外NK/T细胞淋巴瘤（extranodal natural-killer/T-cell lymphoma，NKTCL）患者中的应用价值。

【方法】本研究纳入3例2018年至2020年在中国医学科学院肿瘤医院治疗的NKTCL患者。本研究对这3例患者进行了前瞻性、连续、动态的血液样本收集，并在2018年至2020年期间进行了tumor-naïve ctDNA检测。2023年，本研究对这3例患者外周血样本及治疗前组织样本进行了tumor-informed ctDNA-MRD检测。Tumor-naïve ctDNA检测采用了包含112个淋巴瘤相关基因面板进行靶向测序。Tumor-informed ctDNA-MRD检测采用brPROPHET^TM技术对ctDNA进行检测与定量，用于评估MRD。

【结果】病例1在入组后第1-3次随访期间，tumor-naïve ctDNA和tumor-informed ctDNA-MRD均呈阳性，后续患者在西达本胺治疗9个月后出现系统性复发，并因病情进展死亡。这表明这两种检测方法结果具有一致性，也与临床进程相符，持续的ctDNA/ctDNA-MRD阳性可能预示不良预后。病例2自化疗第4周期结束时起，tumor-naïve ctDNA和tumor-informed ctDNA-MRD 均呈阴性，截止2024年3月，患者仍存活。再次提示这两种检测方法结果具有一致性，持续的ctDNA阴性可能与较长生存时间有关。病例3的第2次和第3次外周血ctDNA监测显示，tumor-naïve ctDNA结果均为阴性，而tumor-informed ctDNA-MRD结果均为阳性（ctDNA-MRD分数分别为2.1%和5.0%），影像学检查未显示疾病进展或复发迹象。后续患者出现系统性复发，并因病情进展死亡。这表明与tumor-naïve ctDNA相比，tumor-informed ctDNA-MRD在预测NKTCL早期复发和预后方面更敏感。

【结论】本研究表明，血浆ctDNA/ctDNA-MRD的动态监测在预测NKTCL疾病进展和复发方面具有重要价值，为制定NKTCL患者个体化治疗策略提供了依据。其次，tumor-informed ctDNA-MRD检测方法在评估NKTCL患者预后和预测肿瘤复发方面，相较于tumor-naïve ctDNA检测方法具有更高的敏感性。血浆ctDNA/ctDNA-MRD监测和影像学评估相结合可以对患者病情进行更全面的监测，以及帮助临床医生更早识别复发迹象。未来，需要更大队列和更长随访时间的临床研究来验证本研究发现。

Part 1：

Objective: Optimal treatment strategies for early-stage extranodal natural killer/T cell lymphoma (NKTCL) patients with adverse prognostic factors have not been fully defined. We investigated the safety and efficacy profile of first-line intensity modulated radiation therapy (IMRT) followed by gemcitabine, dexamethasone, cisplatin (GDP) plus chidamide for early-stage upper aerodigestive tract (UADT)-NKTCL patients with adverse prognostic factors.

Methods: This is an open-label, randomized phase 2 trial performed at 2 centers in China. The study planned to enroll a total of 98 patients. Patients were eligible if they were newly-diagnosed with early-stage NKTCL with at least one risk factor based on a nomogram-revised risk index: >60 years old, elevated serum lactate dehydrogenase (LDH), primary tumor invasion, stage II or Eastern Cooperative Oncology Group performance status (ECOG PS) >1. Patients were treated with IMRT followed by GDP, with or without chidamide, in the first-line setting. Eligible patients were randomly assigned in a 1:1 ratio to either the chidamide group or the control group. Two-year progression-free survival (PFS) rate comprised the primary endpoint. Objective response rate (ORR), 2-year overall survival (OS) rate, and toxicities comprised the secondary endpoints.

Results: From May 2015 to December 2019, a total of 74 eligible patients were ultimately enrolled in this study. Midway through the study, a midterm analysis of the 58 patients enrolled up to that point showed no significant difference in PFS and OS between the two groups. Given the short follow-up period, two-thirds of the way through the trial, the analysis was performed and again, no significant benefit was observed for the primary and secondary endpoints. Due to concerns over a potentially higher frequency of adverse events associated with chidamide, recruitment for the trial was terminated early. Among the 74 patients, 37 patients were treated with IMRT + GDP + chidamide (chidamide group), whereas 37 cases were treated with IMRT + GDP (control group). Follow-up comprised a median of 43.4 months (range, 1.0-74.6 months). ORR was 86.5% in the chidamide group and 78.4% in the control group (p = 0.359) at the end of treatment completion. The 2-year OS and PFS rates were 89.2% and 75.2% in the chidamide group versus 83.8% (p = 0.388) and 70.2% (p = 0.821) in the control group, respectively. Conditional survival analysis showed that among patients who survived more than 7 months, those in the chidamide group had superior survivals compared with those in the control group (2-year OS, 100% vs. 88.6%, p = 0.028). Hematological toxicities and mucositis were the main adverse events with no significant difference between the two groups (p > 0.05). Multivariate analysis showed that elevated serum LDH levels and the ECOG score >1 were independent prognostic factors for PFS of NKTCL (HR：2.653，95%CI：1.119-6.291，p = 0.027；HR：3.618，95%CI：1.515-8.645，p = 0.004).

Conclusions: In early-stage NKTCL patients with adverse prognostic factors, the IMRT followed by GDP chemotherapy regimen combined with chidamide did not show a significant survival benefit compared to the regimen without chidamide. There was no significant difference in toxicity between the two treatment approaches. Serum LDH levels and ECOG score have predictive value in assessing the prognosis of patients with NKTCL.

Part 2：Utilizing Whole-exome Sequencing to Explore Biomarkers for Therapeutic Outcomes and Prognosis in extranodal natural killer/T cell lymphoma

Objective: Currently, there is a lack of broadly applicable prognostic molecular markers for extranodal natural killer/T-cell lymphoma (NKTCL) in clinical practice. In this study, we analyzed the genomic data and mutational characteristics of patients with NKTCL by whole-exome sequencing (WES), aiming to investigate the impact of tumor genetic background on clinical efficacy and to identify potential biomarkers that could significantly impact the management and prognostication of NKTCL.

Methods: 33 NKTCL specimens from the Cancer Hospital of the Chinese Academy of Medical Sciences and Beijing Cancer Hospital were analyzed using WES complemented by in-depth bioinformatics analysis. Patients were categorized into two main groups based on their response to treatment/immunotherapy: sensitive and non-sensitive. This classification was determined by whether they achieved complete remission (CR) undergoing treatment or immunotherapy, thereby creating subsets of treatment-sensitive and immunotherapy-sensitive groups, alongside their respective non-sensitive counterparts. Statistical analysis, including Fisher's exact test, was employed to pinpoint genes with differential mutation rates between groups. Furthermore, mutation signature analysis, employing R packages such as "musicatk", "deconstructSigs", and "decompTumor2Sig" aimed to delineate the association between mutation profiles and clinical outcomes, therapeutic efficacy, and prognosis. Moreover, the mutant allele tumor heterogeneity (MATH) scores were calculated to quantify intra-tumor heterogeneity within NKTCL samples, exploring its correlation with clinical manifestations, therapeutic response, and survival outcomes of patients with NKTCL.

Results: The median tumor mutation burden (TMB) of 33 patients with NKTCL was 4.66/Mb (range 0.22-8.76). Patients were categorized based on post-treatment/post-immunotherapy efficacy evaluation into treatment-sensitive (CR group, n = 23) and non-treatment-sensitive groups (non-CR group, n = 10), or into immunotherapy-sensitive (iCR group, n = 5) and non-immunotherapy-sensitive groups (non-iCR group, n = 4). In the non-CR group, mutation frequencies of ZXDA (p = 0.036), HAPLN2 (p = 0.036), COL4A3 (p = 0.036), MGA (p = 0.036), and LAMA1 (p = 0.036) were significantly higher compared to the CR group, and in the non-iCR group, the mutation rate of the ANK2 (p = 0.047) was significantly higher compared to the iCR group. Prognostic analysis showed that MGA (p = 0.005) and LAMA1 (p = 0.031) mutations were significantly associated with shorter overall survival (OS) of patients with NKTCL. Furthermore, the study revealed that C>T/G>A mutations were the most common somatic single nucleotide variations, accounting for 41.8%, and 94% of patients with NKTCL had COSMIC signature 1. Patients with COSMIC signature 6 (45.4%) showed significantly shorter OS (p = 0.011) and progression-free survival (PFS) (p = 0.024) compared to those without this feature. The median MATH score among the 33 patients was 49.4, and based on this score, patients were divided into High-MATH and Low-MATH groups, with the Low-MATH group showing significantly longer PFS and OS than the High-MATH group (p = 0.023; p = 0.026).

Conclusions: This study identified key genes and genetic features closely related to the treatment response and prognosis of patients with NKTCL, including MGA and LAMA1 mutations, as well as the COSMIC signature 6, which could serve as potential markers for poor prognosis in NKTCL. Additionally, the quantification of intra-tumor heterogeneity through MATH scoring revealed an association between high MATH scores and poor prognosis of NKTCL, suggesting that intra-tumor heterogeneity may be related to treatment resistance. These findings provide important molecular insights for the personalized treatment of NKTCL and identify potential biomarkers for further exploration in the future.

Part 3：

Objective: To compare the potential application value of tumor-informed ctDNA-MRD and tumor-naïve ctDNA detection methods in patients with extranodal natural-killer/T-cell lymphoma (NKTCL).

Methods: This study included 3 patients with NKTCL treated at the Cancer Hospital of the Chinese Academy of Medical Sciences between 2018 and 2020. In this study, prospective, continuous, and dynamic blood sample collection was conducted on these 3 patients, and tumor-naïve ctDNA detection was performed between 2018 and 2020. In 2023, tumor-informed ctDNA-MRD detection was performed on peripheral blood samples and pre-treatment tissue samples of these 3 patients. Tumor-naïve ctDNA detection used targeted sequencing of a panel containing 112 lymphoma-related genes. Tumor-informed ctDNA-MRD detection utilized the brPROPHET™ technology for ctDNA detection and quantification, aimed at assessing MRD.

Results: Both tumor-naïve ctDNA and tumor-informed ctDNA-MRD were positive for case 1 during the first to third follow-up periods after enrollment. Then case 1 experienced systemic relapse 9 months following the chidamide therapy and later died due to disease progression. This indicated that the results of the two detection methods were consistent and aligned with the clinical course. Persistent ctDNA/ctDNA-MRD positivity may indicate a poor prognosis of NKTCL. In case 2, both tumor-naïve ctDNA and tumor-informed ctDNA-MRD have been negative since the end of the fourth cycle of chemotherapy, and the patient was still alive as of March 2024. This again indicated consistency between the two detection methods, suggesting that persistent ctDNA negativity may be associated with a longer survival time. For Case 3, the second and third peripheral blood ctDNA monitoring showed tumor-naïve ctDNA was negative, while tumor-informed ctDNA-MRD was positive (ctDNA-MRD fraction of 2.1% and 5.0%, respectively), with no signs of disease progression or relapse on radiological assessment. The patient subsequently experienced systemic relapse and died due to disease progression. This suggested that compared to tumor-naïve ctDNA, tumor-informed ctDNA-MRD was more sensitive in predicting early relapse and prognosis in patients with NKTCL.

Conclusions: This study demonstrated that dynamic monitoring of plasma ctDNA/ctDNA-MRD held significant value in predicting disease progression and relapse in NKTCL, providing a basis for formulating personalized treatment strategies for patients with NKTCL. Moreover, the tumor-informed ctDNA-MRD detection method exhibited higher sensitivity in assessing prognosis and predicting tumor relapse in patients with NKTCL compared to the tumor-naïve ctDNA detection method. Combining plasma ctDNA/ctDNA-MRD monitoring with radiological assessment provided a more comprehensive monitoring of the patient's condition and helped clinicians identify signs of relapse earlier. In the future, clinical studies with larger cohorts and longer follow-up times are needed to verify the findings of this study.