This study was a phase II, multicenter, open-label randomized trial at 21 hospitals that belonged to the CISL in Korea (CISL-1504/ROSE study). Eligible participants were patients aged 20–65 years with previously untreated histologically confirmed PTCLs based on the World Health Organization classification 2008 including PTCL-NOS, AITL, ALK-negative ALCL, enteropathy-associated T-cell lymphoma (EATL), and hepatosplenic T-cell lymphoma (HSTL). Patients with ALK-positive ALCL, extranodal NK/T-cell lymphoma, and mycosis fungoides/Sezary syndrome were not included in the study. For participation, patients were required to have stage I–IV disease and an Eastern Cooperative Oncology Group performance status ≤ 2. Patients were not eligible if they did not have adequate cardiac, renal, hepatic, and bone marrow function. Thus, patients with an absolute neutrophil count < 1,500 cells/mm³ and platelet count < 100,000/mm³ (a platelet count < 75,000/mm³ in the case of bone marrow involvement) were excluded from the study. Patients with central nervous system or leptomeningeal involvement, positive serology for HIV-1, or active hepatitis B or C were also not eligible for the study. Consolidation treatment with ASCT was mandatory for all participants who responded to induction treatment. Patients were required to provide written informed consent before registration. The protocol was approved by the Institutional Review Boards of each participating institute, and the study was undertaken in accordance with the Declaration of Helsinki (ClinicalTrials.gov: NCT02445404).

After enrollment, patients underwent baseline assessments within 21 days before the first dose of the study drugs including blood tests, echocardiography, bone marrow aspiration/biopsy contrast-enhanced computed tomography (CT) scans of the neck, thorax, abdomen and pelvis, and Flourine-18 fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG-PET-CT) scans. Patients were assigned in a 1:1 ratio of CHOP to fractionated ICED by the CISL office according to two stratification variables: 1) histological subtypes (PTCL-NOS versus AITL) and 2) International Prognostic Index (IPI) risk (low/low-intermediate vs. high-intermediate/high). Treatment started within 4 days after randomization. Patients who were assigned to the CHOP group received cyclophosphamide 750 mg/m², doxorubicin 50 mg/m², and vincristine 1.4 mg/m² (maximum 2 mg) intravenously on day 1, and oral prednisolone 100 mg on days 1–5 every three weeks, for six cycles. Patients in the ICED group received intravenous ifosfamide 1,670 mg/m² on days 1–3, carboplatin 5 x area under the curve (AUC) on day 1, intravenous etoposide 100 mg/m² on days 1–3, and oral or intravenous dexamethasone 40 mg on days 1–4 every three weeks, for six cycles. For both treatment groups, dose modifications were required for treatment-related toxicities in accordance with the study protocol. Pegylated granulocyte colony stimulating factor (G-CSF) was administered on day 2 after CHOP and day 4 after ICED. Administration of supportive medication, including antiemetic therapy was given in accordance with local practice. Prophylaxis for Pneumocystis jirovecii pneumonia was mandatory for all patients at each cycle whatever the treatment arm. The response evaluation was done by contrast-enhanced CT and ¹⁸F-FDG-PET-CT scans prior to the fourth cycle and after the sixth cycle of each treatment arm. Patients who achieved a complete or partial response after the completion of six cycles received ASCT as consolidation therapy within three months from the start of the sixth cycle of chemotherapy for both treatment arms. Peripheral blood stem cells for ASCT were collected during the fourth–sixth cycles of the treatment period or G-CSF mobilization after the completion of six cycles according to the investigator’s decision. The minimum CD34+ cell count was over 3.0 × 10⁶/kg of the recipient’s body weight. Intravenous administration of busulfan 3.2 mg/kg on day –7 to –5, etoposide 200 mg/m², twice a day on day –5 to –4, and cyclophosphamide 50 mg/kg on day –3 to –2 was done as conditioning chemotherapy prior to ASCT. After completion of the study treatment, patients were monitored by CT scan every three months for the first year, every four months for the next year, and every six months thereafter. If a patient showed any signs of clinically suspicious relapse or progression, investigators performed CT and ¹⁸F-FDG-PET-CT scans. Survival status was monitored after disease relapse or progression was confirmed.

The primary end point was progression-free survival (PFS) at three years, and the PFS was defined from the date of the first day of the first cycle treatment until the date of the first observation of documented disease relapse or progression or any kind of death. Patients who received at least one cycle of treatment were defined as the assessable population, and the primary end point was assessed in the assessable population. Secondary end points were overall survival (OS) between the date of randomization and any kinds of death, an objective response rate (ORR) consisting of complete response (CR) and partial response (PR), and the safety outcomes of CHOP and ICED. Response evaluation and the assessment of stable disease (SD) or relapse/progressive disease (PD) were conducted by the participating investigators according to the Revised Response Criteria for Malignant Lymphoma (17). The interim and final response evaluation were performed after the third and sixth cycle of each treatment. Patients undergoing upfront ASCT after the completion of induction treatment were considered as the maintenance of response without any event for the estimation of PFS. Safety outcomes were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (version 4.03). A subgroup analysis was done to assess factors associated with the primary end point based on subtypes of PTCLs and clinical and laboratory parameters. As an exploratory analysis, we compared the mutation profiles of tumor tissue at diagnosis with the treatment outcomes in patients who had archived targeted sequencing data available for analysis as previously described (18, 19). Mean sequencing coverage was greater than 700 x, and somatic alterations were called by a previously described pipeline: MuTect version 1.1.6, Lowfreq version 0.6.1, Pindel version 0.2.5a4 software, and a custom-built in-house algorithm (19–21).

Sample size determination was based on the PFS. This study aimed to test the hypothesis that three-year PFS of ICED would be improved up to a rate of 60% compared with 40%, the rate expected for three-year PFS in the CHOP arm. Interim analysis was performed using group sequential design and, to correct the errors caused by multiple testing, the O’Brien–Fleming method was used. Interim analysis of PFS between the two groups was performed at 41 cumulative progression/death events with the two-sample log-rank test. If the standardized p-value of the log-rank test was greater than 0.686, the effectiveness of the ICED group was concluded as no better compared to CHOP, so the study was stopped. However, if the interim analysis result showed a p-value less than 0.017, it would be declared that the effectiveness of ICED therapy was statistically highly significantly better than CHOP, and the study would be terminated. If the p-value was between 0.017 and 0.686, the study would continue to proceed. The log-rank for one-sided alpha = 10% for 90% power required 134 patients (67 patients in each arm) and under the 2% drop-out assumption, a total 138 patients were enrolled. The final analysis was planned when at least 82 cumulative progression/death events occurred. Fisher’s exact test was used for associations between categorical variables. The Kaplan–Meier method was used to estimate OS and PFS, and the results were compared using the log-rank test. Data were analyzed using Statistical Package for Social Sciences software (version 24.0; IBM Corp, Armonk, NY, USA).

Between September 2015 and March 2021, 145 patients were screened, and 138 patients were enrolled ( Figure 1 ). The interim analysis on PFS between the two groups in 2019 showed that the p-value for PFS difference was 0.640, and thus the study was continued until the final enrollment. After randomization, two patients withdrew informed consent, and one patient’s diagnosis was changed in the ICED group. Accordingly, 66 patients received the first cycle of ICED, whereas 69 patients received the first cycle of CHOP after assignment ( Figure 1 ). The median age of 135 assessable patients was 57 years (range, 27–65 years), and the most common subtypes were PTCL-NOS (n = 61) and AITL (n = 53, Table 1 ). Although patients were randomized based on the IPI and subtype, a greater number of patients with ALK-negative ALCL and stage III/IV were assigned to the ICED group. However, there was no significant difference in baseline characteristics between the two arms ( Table 1 ). In the CHOP arm, response evaluation was not done in three patients who dropped out after the first cycle: two deaths (one patient with EATL died due to bowel perforation, and the other patient died due to central catheter-related air embolism) and one patient refused to participate in the study. In the ICED arm, three patients’ responses were not evaluated because two patients died due to acute myocardial infarction and sepsis, respectively, after the first cycle, and the other patient died due to unknown causes after the third cycle ( Figure 1 ). The interim response evaluation showed 31 CR, 25 PR, and 10 PD in the CHOP arm. Out of 25 patients with interim PR, only seven patients achieved CR, whereas 14 patients showed PD at the end of treatment. Finally, the ORR of CHOP was 59.4% (41/69), consisting of 36 CR and five PR ( Figure 2A ). The interim response to ICED showed 21 CR, 27 PR, four SD, and 11 PD. Out of 27 with interim PR, eight patients achieved CR, whereas nine patients progressed at the final response evaluation. Thus, the number of PR patients was 10 at final response, whereas 27 patients achieved CR at final response because three patients with interim CR progressed at final response. As a result, the ORR of ICED was 56.1% (37/66), consisting of 27 CR and 10 PR ( Figure 2A ). Although the number of patients with CR was different (CHOP: n = 36 vs. ICED: n = 27), the ORR was not different between CHOP (41/69, 59.4%) and ICED (37/66, 56.1%) arms (p = 0.421).

Among 54 patients who completed six cycles of CHOP treatment, 13 patients with interim CR or PR showed PD after the completion of CHOP. Four patients with CR refused ASCT and stem cells were not collected from one patient. Another patient with CR failed to undergo ASCT because leptomeningeal involvement occurred before ASCT. Thus, 35 patients received upfront ASCT according to the protocol ( Figure 1 ). In the ICED arm, 28 patients underwent upfront ASCT out of 42 patients who completed six cycles of ICED because of disease progression, collection failure, and refusal to undergo ASCT. Three patients with PR at the final response showed disease progression while they were preparing for ASCT ( Figure 1 ). The number of patients who underwent the planned upfront ASCT was not significantly different between CHOP (35/69, 50.7%) and ICED (28/66, 42.4%) either (p = 0.389). The number of surviving patients was not different between the CHOP (37/69, 53.6%) and ICED groups (37/66, 56.1%, p = 0.863) at the time of analysis. The most common cause of death was disease relapse or progression in both arms, and most deaths occurred during the post-treatment period. The survival outcome was significantly better in patients who completed the treatment protocol including upfront ASCT in both arms ( Figure 2B ). When the survival status of 63 patients who underwent upfront ASCT was compared according to the final response to CHOP or ICED treatment, patients with PR showed poor survival ( Figure 2C ).

As of 31 December 2022, after a median follow-up of 44.8 months (95% Confidence Interval [CI]: 40.8–48.8 months), the occurrence of relapse or progression was not significantly different between the CHOP (28/69, 40.6%) and ICED groups (24/66, 36.4%, p = 0.724). Accordingly, the primary end point of PFS was not met, and the three-year PFS was not significantly different between the CHOP (36.7%) and ICED groups (33.1%, p = 0.709, Figure 3A ). Among the patients who had relapse or progression after CHOP or ICED, seven patients in the ICED arm underwent salvage ASCT after they responded to subsequent therapy, whereas three patients in the CHOP arm underwent salvage ASCT. Thus, the three-year OS was not significantly different between the CHOP (54.9%) and ICED groups (57.2%, p = 0.900, Figure 3B ). Patients who completed the CHOP or ICED treatment followed by upfront ASCT according to the protocol did not show a significant difference in PFS (p = 0.294, Figure 3C ) and OS (p = 0.938, Figure 3D ) either. The PFS analyses were conducted in subgroups with potential prognostic factors for 135 patients. There was no statistically significant difference in PFS between the CHOP and ICED arms for any subgroup analyzed except subtypes such as AITL ( Figure 4A ). Thus, CHOP was favored over ICED in AITL patients, whereas ICED was favored over CHOP in EATL/HSTL patients ( Figure 4A ). Accordingly, the comparison showed a better PFS of AITL in CHOP than ICED with a marginal significance ( Figure 4B ). On the other hand, ICED showed a trend toward better PFS in PTCL-NOS patients, although it was not statistically significant ( Figure 4C ). Exploratory analysis of PFS in a group of patients (n = 25) whose sequencing data were available showed worse survival outcomes in patients with TP53 mutations regardless of treatment, although it was not statistically significant ( Figure 5 ).

Treatment-emergent adverse events were assessed in 135 patients who received at least one dose of the study treatment. CHOP was associated with more peripheral neuropathy of all grades ( Table 2 ). ICED was associated with more anemia, neutropenia, and thrombocytopenia of all grades ( Table 2 ). Accordingly, the frequency of febrile neutropenia was higher in the ICED arm vs. the CHOP arm although it was not statistically significant (19.7% vs. 10.1%, p = 0.143). Grade 4 adverse events were higher in the ICED (n = 43) arm vs. CHOP (n = 8) arm (65.1% vs. 11.6%, p < 0.001). As a result, the relative dose intensity (calculated as the average of the four) drugs for ICED (92.5%) was lower than that for CHOP (98.8%) because of the dose reduction and delay in the ICED arm.