Relapse in stage I(E) diffuse large B‐cell lymphoma

Abstract Despite a general favourable outcome in limited stage diffuse large B‐cell lymphoma (DLBCL), relapses occur in about 10 to 20% of patients. Prognostic models only partially identify patients at risk for relapse. Moreover, it is not known whether the outcome after such a relapse is similar to the outcome after relapse in advanced stages. From January 2004 through December 2012, all newly diagnosed patients with stage I(E) DLBCL were retrospectively analysed from 2 clinical databases to investigate the relapse pattern and outcome in relation to initial treatment and clinical characteristics. In 126 patients (median age 64 years), histologically confirmed stage I(E) DLBCL was diagnosed. With a median follow‐up of 53 months (range 5‐132 months), 1 progressive disease and 18 relapses occurred. The 5‐year time to tumour progression and disease‐specific survival were 85% (95% CI 79‐91%) and 92% (95% CI 87%‐97%), respectively. We observed no significant difference in relapse localization, time to tumour progression, and disease‐specific survival between patients treated with abbreviated R‐CHOP plus involved field radiotherapy or with 6 to 8 cycles of R‐CHOP. Analysis of relapses showed relapse >5 years after initial treatment (late relapse) in 5 of 19 patients (26%). Six of 19 patients (32%) had central nervous system relapse. Three of 11 relapsed cases available for analysis (28%) showed an MYC translocation, suggesting an overrepresentation in the relapse group. Outcome of patients with a relapse was poor with a median survival after relapse of 8 months. Only 1 patient (5%) underwent successful autologous stem cell transplantation. To improve outcome in these patients, early identification of new biological factors such as a MYC translocation or a high risk for CNS dissemination might be helpful. Moreover, treatment of any relapse after stage I disease should be taken seriously. Salvage treatment should be similar to relapses after advanced DLBCL.


| INTRODUCTION
Diffuse large B-cell lymphoma (DLBCL) accounts for 25 to 30% of adult non-Hodgkin lymphomas. 1 Twenty-five to 40% of patients present with limited stage disease, defined as stages I and II according to the Ann Arbor classification.
Until the beginning of this century, optimal treatment for limited stage DLBCL used to consist of 3 cycles of cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) chemotherapy plus involved field radiotherapy (IFRT). This combined modality approach resulted in a significantly better overall survival (OS) than treatment with 8 cycles of CHOP alone. 2 The addition of rituximab to CHOP has increased OS with 10-15% in both limited stage and advanced stage DLBCL. [3][4][5] Apparently, this became for many haematologists a reason to refrain from consolidation IFRT for patients with stage I and II. 6 Although randomized controlled trials are lacking, a very large registry study covering >59,000 patients strongly suggested that combined modality therapy was associated with better OS, even in the rituximab era. 6 Despite the generally favourable outcome, relapses still occur in 10 to 20% of patients with limited stage DLBCL and 5-year OS ranges between 75% and 94%, which suggests that salvage of relapses is frequently unsuccessful. [6][7][8][9][10] Clinical prognostic models only partially identify patients at risk for relapse. 2,11 Biological tumour characteristics such as cell of origin and especially presence of MYC translocation have prognostic significance in DLBCL. 12,13 We decided to analyse in an observational cohort study the relapses of patients with stage I(E) DLBCL focusing on (1)

| Treatment and follow-up
Patients were categorized into 2 treatment regimens: combined modality treatment consisting of abbreviated R-CHOP (3-4 cycles) plus IFRT or R-CHOP only (6-8 cycles). The number of R-CHOP cycles administered was registered for all patients. Staging was performed by 18 FDG PET/CT scan in 59% of patients (n = 74) and by CT scan in 41% (n = 52). End of treatment response was assessed by 18 FDG PET/CT scan in 75% of patients (n = 91) and by CT scan in 25% (n = 31). Tumour responses were classified as complete remission, partial response, stable disease, or progressive disease (PD) according to the International Working Group. 15 In all patients, relapse was confirmed with 18 FDG PET scan, CT scan, and/or magnetic resonance imaging. Histological confirmation of relapse (or in case of CNS localization multicolor flow cytometry of spinal fluid) was available in 68% (n = 13) of patients. Reasons for not performing histological conformation were early relapse (n = 4) and CNS localization (n = 2). In case of PD or relapse, the subsequent salvage treatment and response was retrieved from the clinical records. Early treatment-related mortality was restricted to death during or ≤3 months after treatment. Relapses after 5 years were designated as late relapses. 16,17 Follow-up was completed until December 2015.

| MYC fluorescence in situ hybridization analysis
For evaluation of a MYC translocation, formalin-fixed paraffinembedded tissue blocks were collected of the relapsed DLBCL cases.
Interphase fluorescence in situ hybridization (FISH) was performed on 3-μm-thick whole tissue sections of the primary tumour as previously described by using Vysis break apart probes (Abbot Technologies). 18

| Statistical analysis
Duration of follow-up was calculated for all patients alive. The primary endpoints were OS, disease-specific survival (DSS), time to tumour progression (TTP), and survival after relapse. Overall survival was defined as time from diagnosis until death (from any cause), DSS as the time from diagnosis until death as a consequence of DLBCL, TTP as the time from diagnosis until relapse or progression, and survival after relapse as the time from relapse until death (from any cause). 15 Survival curves were estimated according to the Kaplan-Meier method. Between-group differences in DSS and TTP were evaluated by using the log-rank test.
All categorical variables were expressed as counts and percentages.
Where applicable, differences between groups were evaluated by chi-square for binary variables and independent t tests for continuous variables. Cox regression was used for univariate analysis. Given the low incidence of events no multivariate analysis was performed. A 2-tailed P value of less than.05 indicated statistical significance. All analyses were performed by using IBM SPSS Statistics version 22.

| Treatment
Of the 126 patients 97% (n = 122) completed, at least 3 cycles of R-CHOP were evaluable for comparison between treatment arms, e.g., combined modality (68%, n = 83) or R-CHOP alone (32%, n = 39); see Figure 1 and Table 1. In patients receiving R-CHOP, the number of cycles was reduced in 8% of patients (n = 3) because of previous tumour resection and treatment-related toxicity. The cumulative dosage of IFRT was 30 to 40 Gray. All patients with testicular localization received CNS prophylaxis with intrathecal methotrexate.
Patients with extranodal disease and those with elevated LDH more frequently received R-CHOP only (P < .01 and P = .01, Table 1).

| Clinical and biological characteristics of relapse
One patient had PD during the first 3 cycles of R-CHOP, and 18 patients experienced a relapse. Of these 18 patients, 28% of patients (n = 5) had a relapse more than 5 years after diagnosis: 3 of 14 patients treated with abbreviated R-CHOP plus IFRT and 2 of 4 patients treated with R-CHOP (p 0.52) (Figure 1). In 79% of cases (n = 15), the relapse/progression occurred at a site distant from the initial tumour localization.
In 32% of cases (n = 6), this involved the CNS, with either meningeal and/or parenchymal localization. Two of the CNS relapses occurred in patients with a testicular lymphoma, despite CNS prophylaxis. In the remaining cases, the primary tumour had nodal (n = 2) and nasopharyngeal localizations (n = 2) (Figure 2). The initial calculated CNS-IPI was low

| Patient outcome
The median duration of follow-up of the 126 patients was 53 months (range 5-132). Twenty-seven patients died (11 of relapse; 16 unrelated).

| DISCUSSION
In this population-based cohort study with a long median follow-up, we followed all consecutive patients with newly diagnosed DLBCL stage I(E) during an 8-year period. We observed no differences in relapse localization, TTP, and DSS between patients treated with  Patients with extranodal disease more frequently received R-CHOP, reflecting physicians' choice to avoid radiotherapy-induced toxicity.
By looking into the characteristics of relapsed patients, we made several observations that offer a reason for the occurrence of relapses in these good risk patients.
Regardless of initial therapy, one-third of relapses (26%) occurred more than 5 years after therapy. Although late relapses have been observed in the pre-rituximab era, 17 we and others observed these late relapses in patients with limited stage DLBCL treated with rituximab as well. 16 Although clonal relationship in late relapses is established, the biology underlying the long interval remains unclear. 19 Most relapses (73%, n = 15) arose at distant sites, indicative of good local tumour control with either abbreviated R-CHOP plus radiotherapy or R-CHOP. In nearly one-third of relapses, there was CNS involvement. Because rituximab and CHOP have only limited activity in the CNS, it is unlikely that either of the treatment regimens will prevent the CNS relapses. 20 Even when initial CNS prophylaxis with intrathecal MTX is provided, CNS relapses can occur as illustrated by 2 patients with lymphoma of the testes in our study. 21 Recently, the CNS-IPI as risk model for CNS relapse in patients with DLBCL was established to identify patients at highest risk for CNS relapse. 14 In patients with a low-risk CNS-IPI, less than 1% showed a CNS relapse.   It was recently reported that in contrast to advanced stage DLBCL, the cell-of-origin is not prognostic in limited stage disease. 9,13 To impact prognosis of these good risk patients, a biomarker, such as MYC breaks, might be helpful. 12 We found an MYC translocation in 3 of 11 evaluable relapsed cases. It is plausible that in the relapsed setting, 15 to 20% of DLBCL harbour an MYC translocation. 22 Although the number of analysed patients is low, we found no MYC positive DLBCL in late relapses. Despite an increased risk of CNS dissemination in MYC positive DLBCL, no translocation was detected in the 2 evaluable CNS relapses. 23 Combined analysis showed that two-third of relapses could be assigned to either a late relapse, CNS relapse, or MYC positive DLBCL.
In general, outcome of patients with relapsed or refractory DLBCL is very poor, with the exception of more favourable outcome in relapses more than 1 year after treatment. 24 We observed a similar poor out-

CONFLICT OF INTEREST
Not applicable.