CEOP/IVE/GDP alternating regimen compared with CEOP as the first-line therapy for newly diagnosed patients with peripheral T cell lymphoma: results from a phase 2, multicenter, randomized, controlled clinical trial

Cyclophosphamide, doxorubicin, vincristine, and prednisolone (CHOP)/CHOP-like chemotherapy is widely used in peripheral T cell lymphoma (PTCL). Here we conducted a phase 2, multicenter, randomized, controlled trial, comparing the efficacy and safety of CEOP/IVE/GDP alternating regimen with CEOP in newly diagnosed PTCL. PTCL patients, except for anaplastic large cell lymphoma-anaplastic lymphoma kinase positive, were 1:1 randomly assigned to receive CEOP/IVE/GDP (CEOP, cyclophosphamide 750 mg/m2, epirubicin 70 mg/m2, vincristine 1.4 mg/m2 [maximum 2 mg] on day 1, and prednisone 60 mg/m2 [maximum 100 mg] on days 1–5 every 21 days, at the first and fourth cycle; IVE, ifosfamide 2000 mg/m2 on days 1–3, epirubicin 70 mg/m2 on day 1, and etoposide 100 mg/m2 on days 1–3 every 21 days, at the second and fifth cycle; and GDP, gemcitabine 1000 mg/m2 on days 1 and 8, cisplatin 25 mg/m2 on days 1–3, and dexamethasone 40 mg on days 1–4 every 21 days, at the third and sixth cycle) and CEOP (every 21 days for 6 cycles). Analysis of efficacy and safety was of the intent-to-treatment population. The primary endpoint was a complete response rate at the end of treatment. Meanwhile, whole exome sequencing and targeted sequencing were performed in 62 patients with available tumor samples to explore prognostic biomarkers in this cohort as an exploratory post hoc analysis. Among 106 patients, 53 each were enrolled to CEOP/IVE/GDP and CEOP. With 51 evaluable patients each in two groups, a complete response rate of the CEOP/IVE/GDP group was similar to that of the CEOP group (37.3% vs. 31.4%, p = 0.532). There was no difference in median progression-free survival (PFS; 15.4 months vs. 9.2 months, p = 0.122) or overall survival (OS; 24.3 months vs. 21.9 months, p = 0.178). Grade 3–4 hematological and non-hematological adverse events were comparable. Histone modification genes were most frequently mutated (25/62, 40.3%), namely KMT2D, KMT2A, SETD2, EP300, and CREBBP. Multivariate analysis indicated that CREBBP and IDH2 mutations were independent factors predicting poor PFS and OS (all p < 0.001), while KMT2D predicting poor PFS (p = 0.002). CEOP/IVE/GDP alternating regimen showed no remission or survival advantage to standard chemotherapy. Future clinical trials should aim to develop alternative regimen targeting disease biology as demonstrated by recurrent mutations in epigenetic factors. The study was registered on ClinicalTrial.gov (NCT02533700) on August 27, 2015.

(Continued from previous page) sequencing and targeted sequencing were performed in 62 patients with available tumor samples to explore prognostic biomarkers in this cohort as an exploratory post hoc analysis. Results: Among 106 patients, 53 each were enrolled to CEOP/IVE/GDP and CEOP. With 51 evaluable patients each in two groups, a complete response rate of the CEOP/IVE/GDP group was similar to that of the CEOP group (37.3% vs. 31.4%, p = 0.532). There was no difference in median progression-free survival (PFS; 15.4 months vs. 9.2 months, p = 0.122) or overall survival (OS; 24.3 months vs. 21.9 months, p = 0.178). Grade 3-4 hematological and nonhematological adverse events were comparable. Histone modification genes were most frequently mutated (25/62, 40.3%), namely KMT2D, KMT2A, SETD2, EP300, and CREBBP. Multivariate analysis indicated that CREBBP and IDH2 mutations were independent factors predicting poor PFS and OS (all p < 0.001), while KMT2D predicting poor PFS (p = 0.002). Conclusions: CEOP/IVE/GDP alternating regimen showed no remission or survival advantage to standard chemotherapy. Future clinical trials should aim to develop alternative regimen targeting disease biology as demonstrated by recurrent mutations in epigenetic factors. Trial registration: The study was registered on ClinicalTrial.gov (NCT02533700) on August 27, 2015.
Dysregulated epigenetic mechanisms play a central role in the pathogenesis of PTCL. Main genes involved in epigenetic regulation are frequently mutated in PTCL, including histone modification (KMT2D, KMT2A, SETD2, KDM6A, EP300, and CREBBP), chromatin remodeler (ARID1B and ARID2) [12], and DNA methylation and demethylation genes (TET2, DNMT3A, and IDH2) [13]. We previously reported that histone modification gene mutations are associated with inferior progression-free survival (PFS) time of the patients [12]. However, the prognostic value of individual epigenetic regulator remains to be further investigated in PTCL, particularly for identifying patients with poor prognosis who may benefit from targeted agents.
Given the poor response to CHOP-based regimens and the potential anti-lymphoma activity by alternating chemotherapy in PTCL, we conducted a phase 2, multicenter, randomized, controlled trial, comparing the efficacy and safety of CEOP/IVE/GDP alternating regimen with CEOP in a Chinese cohort of newly diagnosed patients with PTCL. Meanwhile, we performed whole exome sequencing (WES) and targeted sequencing in 62 patients with available tumor samples to explore prognostic biomarkers in this prospective cohort. Results from this study were previously presented as an abstract [14].

Study design and patients
This phase 2, multicenter, randomized, controlled trial was conducted at four centers (Additional file 1: Table  S1) in China within cooperative network of the Multicenter Hematology-Oncology Programs Evaluation System (M-HOPES). Eligible patients were ≥ 16 years of age with newly diagnosed, histologically confirmed PTCL according to 2008 WHO classifications [15]: peripheral T cell lymphoma-not otherwise specified (PTCL-NOS), angioimmunoblastic T cell lymphoma (AITL), anaplastic large cell lymphoma (ALCL)-anaplastic lymphoma kinase (ALK) negative, EATL, subcutaneous panniculitis like T cell lymphoma, and hepatosplenic T cell lymphoma (HSTL). Patients were required to have WHO performance status of ≤ 2, no previous history of malignancy, radiologically measurable disease, and a life expectancy of ≥ 6 months. Patients were not eligible once meeting any of the following criteria: NK/T cell lymphoma, ALCL-ALK positive, or primary central nervous system (CNS) lymphoma, previous systemic chemotherapy or local therapy, previous hematopoietic stem cell transplantation (HSCT), antifungal or antiviral therapy, uncontrollable cardiocerebrovascular, coagulative, autoimmune, or serious infectious disease, left ventricular ejection fraction (LVEF) ≤ 50%, other uncontrollable medical condition that may interfere with their participation in the study. Patients were also not enrolled if they had inadequate renal, hepatic, or bone marrow function, were not able to comply with the protocol for any reasons, were pregnant or breastfeeding, or had active liver or biliary disease, or human immunodeficiency virus (HIV) infection.
The study was approved by the Institutional Review Boards of all centers. Informed consent was obtained from all patients in accordance with the Declaration of Helsinki. The study was registered on ClinicalTrial.gov (NCT02533700) (https://www.clinicaltrials.gov/ct2/ show/NCT02533700?term=NCT02533700&rank=1) on August 27, 2015. The trial was overseen by trial management and trial steering committees. The CONSORT checklist for reporting a randomized trial can be found in Additional file 2.

Whole exome sequencing
Genomic DNA was extracted from frozen tumor samples of 27 patients [16,17] and peripheral blood of 21 patients using Wizard® Genomic DNA Purification Kit (Promega, Wisconsin-Madison, USA) according to the manufacturer's instructions. Exome regions of all samples were captured using a SeqCap EZ Human Exome kit (version 3.0) and sequenced on HiSeq 4000 platform with 150 bp paired-end strategy. Read pairs were then aligned to Refseq hg19 by Burrows-Wheeler Aligner (BWA) version 0.7.13-r1126, with PCR duplications removed by Samtools version 1.3 to generate chromosomal coordinate-sorted bam files. The mean depth of each sample was 90.1× (range 68.7-107.2×), with an average 98.2% (range 96.9~97.9%) of the target sequence being covered sufficiently deep for variant calling (≥ 10× coverage). A total of 33 genes were selected based on two criteria: (i) genes with recurrent mutations (> 3%) and/or (ii) genes with relevance to oncogenesis of PTCL (Additional file 1: Table S2). Population-related variants were eliminated using the variants in the 1000 Genomes Project, and SNP mutations were identified by COSMIC (the Catalogue Of Somatic Mutations In Cancer version 77).

Targeted sequencing
Genomic DNA was extracted from frozen or paraffin tumor samples of 35 PTCL patients using a QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. PCR primers were designed by Primer 5.0 software and listed in Additional file 1: Table S2. Multiplexed libraries of tagged amplicons from PTCL tumor samples were generated by Shanghai Yuanqi Bio-Pharmaceutical Multiplex-PCR Amplification System. Deep sequencing was performed using established Illumina protocols on HiSeq 4000 platform (Illumina).

Study treatment
Patients in CEOP/IVE/GDP group received intravenous cyclophosphamide 750 mg/m 2 , epirubicin 70 mg/m 2 , vincristine 1.4 mg/m 2 (maximum 2 mg) on day 1, and oral prednisone 60 mg/m 2 (maximum 100 mg) on days 1-5 every 21 days, at the first and fourth cycle with CEOP. Intravenous ifosfamide 2000 mg/m 2 on days 1-3, epirubicin 70 mg/m 2 on day 1, and etoposide 100 mg/m 2 on days 1-3 every 21 days, at the second and fifth cycle with IVE. Intravenous gemcitabine 1000 mg/m 2 on days 1 and 8, cisplatin 25 mg/m 2 on days 1-3, and dexamethasone 40 mg on days 1-4 every 21 days, at the third and sixth cycle with GDP, for a total of 6 cycles. Patients in the CEOP group received CEOP regimen every 21 days for 6 cycles.
Supportive care was provided according to institutional clinical practice. Granulocyte-colony stimulating factor (G-CSF) was administered if absolute neutrophil count was < 1.0 × 10 9 /L. Consolidation radiotherapy was permitted for patients with residual disease at EOT, if existed. Transplantation eligible patients were recommended to receive HSCT once complete response (CR) or partial response (PR) was achieved. Second-line treatment was considered if the patient had stable disease (SD) at interim evaluation. Prophylaxis for CNS relapse was given to patients with involvement of bone marrow, nasal/paranasal sinuses, orbit, breast, or testis.

Randomization and masking
Patients were randomly assigned to two study groups by computer-assisted permuted-block (block size of 4, allocation ratio 1:1), receiving either CEOP/IVE/GDP alternating regimen or CEOP regimen. After obtaining informed consent from the patients, the investigators went online at the data center for registration and central randomization to assign treatment. A statistician located centrally in Shanghai Rui Jin Hospital supervised the randomization procedure. Investigators and patients were not masked to treatment assignment due to different ways of administration by each treatment group.

Assessments
Positron emission tomography-computed tomography (PET-CT) or CT scans with contrast were evaluated at baseline, after 3 cycles for interim evaluation and at EOT for final evaluation. Treatment responses were assessed according to 2014 Lugano criteria for non-Hodgkin lymphoma [18]. Central response assessment of PET and CT images was performed by radiologists of Shanghai Ruijin Hospital, who was not informative of the treatment group. CT scans of the neck, thorax, abdomen, and pelvis were repeated every 3 months thereafter to monitor disease progression in the first year, then every 6 months in the following 2 years, and yearly thereafter. The severity of adverse events was assessed according to the Common Terminology Criteria for Adverse Events (CTCAE) version 4.0.
Baseline clinical laboratory tests and examinations included complete blood cell count (CBC), serum lactate dehydrogenase (LDH), hepatitis B virus DNA (HBV-DNA), HIV, coagulation function including APTT, PT and fibrinogen, bone marrow aspiration and trephine biopsy, electrocardiography, echocardiography, PET-CT, or CT scans with contrast. Vital signs should be recorded at each visit. International Prognostic Index (IPI), National Comprehensive Cancer Network IPI (NCCN-IPI), and Prognostic Index for T cell lymphoma (PIT) were calculated for each patient at baseline.

Outcomes
The primary endpoint in this study was CRR at EOT, measured by PET-CT or CT scans with contrast, according to 2014 Lugano classification. The secondary endpoints included PFS (defined as the duration time between the date of randomization and the date of disease progression or death from any cause), OS (defined as the duration time between the date of randomization and the date of death from any cause), overall response rate (ORR) at EOT, and the toxicity, evaluated according to the National Cancer Institute Common Terminology Criteria of Adverse Events, version 4.0. The sequencing analysis was an exploratory post hoc analysis.

Sample size and statistical analysis
The objective of this study was to compare the efficacy and safety of CEOP/IVE/GDP alternating regimen with CEOP regimen in newly diagnosed patients with PTCL. For sample size, we estimated that 35% of patients in the CEOP group and 60% of patients in the CEOP/IVE/GDP group would achieve CR. Fifty patients per group were required to show this difference with 5% significance (two-sided) and 73% power, with no plan for interim analysis. The number of patients achieving CR at the end of study was reported by treatment group with 95% CI using normal approximation and compared between groups using logistic regression, with odds ratio (OR [95% CI]) for CEOP/IVE/GDP versus CEOP. Additionally, a logistic regression model was fitted to adjust for the stratification variables, with OR (95% CI) for CEOP/ IVE/GDP versus CEOP. For those whose disease could not be evaluated were considered as no responders from statistically conservative consideration.
We planned subgroup analysis to assess treatment response in the predefined subgroups by logistic regression, with results displayed as ORs in a Forest plot. Kaplan-Meier methods were used, and we compared survival between two study groups by a log-rank test. Additionally, a Cox proportional analysis in the intentto-treatment (ITT) population combining both study groups: age, gender, PTCL subtype, IPI risk group, treatment, Epstein-Barr virus DNA (EBV-DNA), and gene mutations. ITT population was defined in the protocol as all subjects who were randomized to either of the treatment group and accepted at least one dose of treatment. The data analysis was generated using SPSS Statistics version 23 and GraphPad Prism 7.

Patient characteristics
Between September 22, 2015, and December 30, 2018, 106 patients were randomly assigned to two study groups: 53 each to the CEOP/IVE/GDP group and the CEOP group. One patient was excluded because of the change of diagnosis, and 3 withdrew informed consent before study treatment in both groups. Fifty-one patients in each group were included into efficacy and safety analysis as ITT population. Thirty-two patients failed to complete full 6 cycles due to disease progression (18), stable disease (11), toxicity (2), and consent withdrawal (1). In the CEOP/IVE/GDP group, 9 patients discontinued treatment due to failure to response (n = 8, including 4 with disease progression and 4 with stable disease at interim evaluation) and toxicity (n = 1). Median time from randomization to treatment failure was 3.3 months in patients with disease progression (range from 1.2 to 5.4 months) and 3.0 months in patients with stable disease (range from 2.5 to 3.4 months), respectively. In the CEOP group, 23 discontinued treatment due to failure to response (n = 21, including 14 with disease progression and 7 with stable disease at interim evaluation), toxicity (n = 1), and consent withdrawal (n = 1). Median time from randomization to treatment failure was 3.5 months in patients with disease progression (range from 0.3 to 5.6 months) and 3.2 months in patients with stable disease (range from 2.6 to 3.7 months), respectively. Seventy patients finally completed full 6 cycles of treatment (Fig. 1).

Mutation pattern
As shown in Fig. 2a No mutation of EZH2, KDM6A, APC, JAK1, STAT5B, IL2RG DDX3X, BCOR, or FYN was detected. Most of the somatic mutations were missense mutations (n = 70), followed by frameshift mutations (n = 21) and nonsense mutations (n = 11) (Fig. 2b). We observed a preference for C>T alterations analogous to the somatic single-nucleotide variation spectrum in other cancers (Fig. 2b). Histone modification gene mutations were mutually exclusive of each other, suggesting that they might be involved in distinct biological processes (Fig. 2c).

Response to treatment and prognosis
Treatment response rate was summarized in Table 2 Figure S1). Seventy-four (72.5%) patients achieved response at interim evaluation in the whole study population. Among 35 CR patients at interim evaluation, 32

Factors impacting response to treatment
Post hoc subgroup analysis of CRR at EOT showed no difference in CRR between two study groups regarding age, gender, sex, Ann Arbor stage, performance status, serum LDH, PTCL subtype, EBV-DNA, and IPI risk group (Fig. 4). We also investigated the predictive ability of IPI, NCCN-IPI, and PIT model and found similar cindex for PFS (IPI 0.61, NCCN-IPI 0.57, PIT 0.60) and OS (IPI 0.68, NCCN-IPI 0.63, PIT 0.66). Interestingly, 11 patients of low-risk NCCN-IPI showed excellent prognosis with 2-year OS > 80% without HSCT (only 1 out of  Figure S3). The relationship of the recurrent mutated genes with the response status at EOT was shown in Fig. 5a. Among all, CREBBP and IDH2 gene mutations were significantly associated with fewer response to treatment at EOT, as compared to those without mutation (ORR 0% vs. 72.9%, p = 0.031 and 0% vs. 72.9%, p = 0.008, respectively). Moreover, KMT2D mutations were observed in 14 patients, of which only 3 patients with CR at EOT, and 6 patients with PR and 5 with NR. Comparing with those without mutation, patients with KMT2D mutation had higher proportion of no response and early progression after PR at interim evaluation (57.1% vs. 25.0%, p = 0.048). Among 7 patients who had PR at interim evaluation but PD at EOT, 3 had KMT2D mutation, 2 had ARID1B mutation, and 1 had CREBBP mutation.

Factors impacting prognosis
Clinicopathological parameters and recurrent mutated genes were assessed by univariable analysis. Non-ALCL subtype, intermediate/high-risk IPI, and detectable EBV-DNA, as well as mutations in CREBBP, IDH2, and ARID1B, were associated with inferior PFS and OS, while mutations in KMT2D were associated only with inferior PFS (Additional file 1: Table S3). By multivariate analysis, mutations in CREBBP, IDH2, and KMT2D were independent factors predicting poor PFS, while CREBBP and IDH2 were independent factors predicting poor OS (Additional file 1: Table S4). Median   Fig. 5c). According to KMT2D mutation, the median PFS in patients with KMT2D mutation was significantly reduced, as compared to those without KMT2D mutation (6.9 months [6.5-7.3] vs. 12.8 months [95% CI 9.2-16.4], HR 2.00, 95% CI 0.99-4.01, p = 0.047, Fig. 5d). Among eight patients with CREBBP or EP300 mutation, 2 of them received and responded to histone deaceylase (HDAC) inhibitor chidamide combined with chemotherapy as the second-line treatment (1 achieved PR after 4 months of treatment and remained alive, the other achieved PR after 2 months and died of disease progression after 4 months). Among 14 patients with KMT2D mutation, 8 of them received chidamide as the second-line treatment, either alone (n = 3) or in combination with chemotherapy (n = 5). Two patients achieved CR, three patients achieved PR, one patient had stable disease, and two patients had disease progression.

Toxicities
Adverse events of both hematological and nonhematological toxicity were listed in Table 3. Safety was assessed in 102 patients who received at least one dose of study treatment. No statistical significance on hematological adverse events was observed between the two study groups. Neutropenia was the most common event in both groups (82% in the CEOP/IVE/GDP group and 86% in the CEOP group). One patient in the CEOP group died from severe infection after 1 cycle and one patient in the CEOP/ IVE/GDP group was unable to continue treatment EOT end of treatment *One patient in the CEOP/IVE/GDP group could not be evaluated due to toxicity (n = 1). Two patients in the CEOP group could not be evaluated due to toxicity (n = 1) and consent withdrawal (n = 1). All the three patients were considered as no responders from statistically conservative consideration due to pulmonary infection after 2 cycles. For nonhematological adverse events, CEOP/IVE/GDP was associated with more headaches of all grades (27% in the CEOP/IVE/GDP group and 12% in the CEOP group).

Discussion
To our knowledge, this is the first randomized clinical trial to evaluate the efficacy and safety of alternating regimen in treating newly diagnosed patients with PTCL. However, CEOP/IVE/GDP alternating regimen failed to meet the primary endpoint and showed no remission or survival advantage when comparing with CEOP. ALCL-ALK positive that was generally associated with good response to CHOP-based regimen had been excluded according to the study protocol, which could explain the short OS tendency of this study. Multi-drug combinations with different mechanisms of action may alter the clinical outcomes in lymphoma. Gemcitabine is not a substrate for MDRmediated efflux and proves effective in PTCL [4,[19][20][21][22]. A retrospective study comparing CHOP, CHOP-E, and GDP as first-line treatment of PTCL in 102 patients showed that GDP improves CRR, ORR, and OS [4]. Another randomized controlled trial also found that GDP with thalidomide is superior to standard CHOP in 103 newly diagnosed patients with PTCL for response rate and survival time [21]. However, a recent randomized trial in 87 patients with PTCL indicated that GEM-P (gemcitabine, cisplatin, methylprednisolone) is similar to CHOP in the first-line treatment [23]. As for IVE, it shows promising anti-tumor activity in patients with refractory/relapsed lymphomas [5][6][7][8][9][10] and is introduced as an effective front-line treatment in combination with CHOP and MTX in EATL [11]. Based on the above clinical data, we combined these regimens and conducted a phase 2, multicenter, randomized, controlled trial, aiming to increase efficacy in PTCL. Nevertheless, remission and prognostic status of PTCL may not be altered by chemotherapy in an alternating manner. Meanwhile, it is worth pointing out that response at EOT was a strong prognostic factor of both PFS and OS. Consistent with a previous study showing that complete response is a predictor of favorable outcome in PTCL [24], intensive strategies such as HSCT consolidation once CR was achieved are helpful to sustain long-term response and survival.
In terms of prognostic index, IPI, NCCN-IPI, and PIT had comparable capability for risk stratification. For lowrisk patients, CRR was much higher under all three prognostic indexes. Of note, NCCN-IPI can identify very low-risk population, with 2-year OS > 80% without HSCT, in accordance with our previous study [25]. Interestingly, for high-risk patients, CEOP/IVE/GDP alternating regimen was found to prolong PFS and OS, as risk stratified by NCCN-IPI (4-6) and PIT (3)(4). Further randomized clinical trials are needed to carry out in these high-risk subgroups of PTCL using alternating chemotherapy followed by HSCT.
Importantly, using WES and targeted sequencing, we identified mutations in epigenetic regulators that were closely related to disease progression of PTCL. It is reported that histone methylation gene KMT2D mutations correlate with shorter OS in Epstein-Barr virus-associated natural killer-cell lymphoma [26] and are the most frequent relapse-specific events in diffuse large B cell lymphoma (DLBCL) [27]. Presence of histone acetylation gene CREBBP mutations is an independent prognostic factor in DLBCL by mutational analysis of the SAKK 38/07 prospective cohort [28]. IDH2 mutations induce histone and DNA hypermethylation and define a unique subgroup of PTCL with follicular T-helper-like phenotype [29]. ARID1B, a subunit of the SWI/SNF chromatin complex, is also frequently mutated in aggressive lymphomas, including hepatosplenic T cell lymphoma [30] and DLBCL [31]. Therefore, we further confirmed that epigenetic factors can function as predictive biomarkers of inferior prognosis in PTCL.
Epigenetic regulator gene mutations render PTCL resistant to chemotherapy, but sensitive to targeted agents. Experimentally, HDAC inhibitor chidamide significantly inhibited the growth of EP300-mutated T-lymphoma cells and KMT2D-mutated T-lymphoma cells when combined with the hypomethylating agent decitabine [12]. Clinically, a phase 1 study of HDAC inhibitor romidepsin and hypomethylating agent 5azacytidine exhibited marked activity in patients with PTCL, with ORR and CR of 73% and 55%, respectively [32]. Therapeutic efficacy of novel small molecules like IDH2 inhibitors targeting IDH2, or modulators of the SWI/SNF chromatin complex targeting ARID1B could also be explored in PTCL. In our study, patients bearing CREBBP, EP300, or KMT2D may respond to epigenetic agent chidamide, either alone or in combination with chemotherapy.
Although these results should be further investigated in a larger cohort of patients, detection of epigenetic gene mutations may be helpful for selection of patients to targeted therapeutic agents and eventually improve the clinical outcome of PTCL in the era of precision medicine.
Some limitations in our study should be acknowledged. Firstly, although the majority of the patients were evaluated by PET-CT, methods of CT response assessment were also included. Secondly, maintenance therapy was not pre-specified and the proportion of patients undergoing HSCT was limited.