Role of radiation therapy in primary breast diffuse large B‐cell lymphoma in the Rituximab era: a SEER database analysis

Abstract Primary breast diffuse large B‐cell lymphoma (PB‐DLBCL) is an uncommon extranodal non‐Hodgkin's lymphoma (NHL), which was traditionally treated with anthracycline‐containing regimens followed by consolidative radiation therapy (RT) to add therapeutic benefits. The introduction of anti‐CD20 antibody rituximab for the treatment of B‐cell NHLs has significantly improved the clinical outcome of these malignant diseases. It is unclear, however, whether consolidative RT could still add therapeutic benefits for PB‐DLBCL patients treated with rituximab. To answer this important question, we used the Surveillance, Epidemiology, and End Results (SEER) database to evaluate the impact of RT on the clinical outcomes of PB‐DLBCL patients in the rituximab era. Information on patient age, year of diagnosis, stage, race, laterality, and RT status for PB‐DLBCL patients diagnosed between 2001 and 2014 were extracted. Kaplan–Meier survival curves were plotted, and log‐rank test was used to compare the potential survival difference. Multivariate analysis using Cox proportional hazards model was employed to determine the impact of RT and other factors such as age, race, tumor laterality, stage, and year of diagnosis on survival. Among the 386 patients identified, the median follow‐up time was 45 months (range, 0–167 months); the median age was 64 years (range, 19–93 years); 33.9% of the patients were younger than 60 years of age; 69.9% of the patients were stage I; 79.0% were white; 51.8% received RT. The 5‐year OS and cause‐specific survival (CSS) for the whole cohort were 72.3% and 82.5%, respectively. The 5‐year OS was significantly superior for patients who received RT compared to those who did not receive RT (78.1% vs. 66.0%, P = 0.031). In multivariable analysis, RT remained significantly associated with improved OS (P = 0.026). In summary, our study suggests that RT still adds significant therapeutic benefits for patients with PB‐DLCBL in the rituximab era.

Introduction era. This type of extranodal lymphoma has a high risk of contralateral breast and central nervous system (CNS) relapse.
Due largely in part to the low incidence, there is a lack of prospective studies to guide the optimal treatment. It seems that radical resection should be avoided, as retrospective studies have demonstrated that mastectomy offered no benefit over nonmastectomy treatment in terms of the overall survival (OS) [3,5]. Combined treatment consisting of chemotherapy and RT seems to be the treatment of choice in the prerituximab era. For instance, Alives conducted a controlled study and found that combination therapy using six cycles of CHOP plus RT of 30 Gy could induce a higher rate of complete remission (CR) and a lower rate of relapse compared with RT alone or chemotherapy alone [9]. In agreement with this result, a study by the International Extranodal Lymphoma Study Group (IELSG-15) also recommended consolidative RT followed by chemotherapy based on the observation that RT was associated with better OS and a trend to improve progression-free survival (PFS) and cause-specific survival (CSS) and reduced the risk of ipsilateral progression [3]. Further studies also suggested that PB-DLBCL would benefit from such combined therapy [2].
The anti-CD20 antibody rituximab was tested in the management of B-cell NHLs in late 1990s and has since been confirmed to improve OS by 10-30% for most of B-cell NHLs. Several studies have assessed the value of rituximab in PB-DLBCL [7,8,[10][11][12][13]. The use of rituximab in chemotherapy tends to prolong PB-DLBCL patients' PFS and OS [13], suggesting that rituximab might change the biological behaviors of this disease. As the successful application of rituximab has been shown to improve the outcome of DLBCL, it is important to evaluate whether RT could still add therapeutic benefits for PB-DLBCL patients treated with rituximab or whether RT might only increase the risk of toxic side effect without additional therapeutic benefits. In this study, we used the Surveillance, Epidemiology, and End Results (SEER) database to evaluate the impact of RT on the clinical outcome of PB-DLBCL patients in the rituximab era.

Data source
The source of data for our study was from 18 SEER databases of the National Cancer Institute in the United States. SEER is a program that collects and publishes cancer incidence, treatment, and survival data from population-based cancer registries, representing For this study, we included patients with PB-DLBCL diagnosed between 2001 and 2014, a period expected to reflect frequent applications of rituximab in the treatment of lymphomas. Patients with unknown stage, no exact information on disease laterality, stage III, stage IV and bilateral disease diagnosed as stages I and II, with unknown race/ethnicity, younger than 18 years, or received no chemotherapy were excluded (Fig. 1). The variables of age, race, laterality, stage, RT, and year of diagnosis were subjected to both univariate and multivariate analyses to assess their prognostic value on survival.

Statistical analysis
Statistical analysis was performed using SEERstat 8.3.4 and Statistical Package for the Social Sciences (SPSS) version 20.0 software (IBM Corporation, Armonk, NY). Kaplan-Meier survival curves were plotted, and log-rank test was used to compare the survival difference. Multivariate analysis using Cox proportional hazards model was used to determine the impact of RT, race, age, sex, stage, and year of diagnosis on survival. A P value of <0.05 was considered statistically significant.

Results
From 2001 to 2014, a total of 386 patients with stage I and II PB-DLBCL were reported to the SEER registries. The median follow-up time was 45 months (range 0-167 months). The data showed that this disease often occurred in older patients, with a median age of 64 years (range 19-93 years). Of these patients, 79.0% were of white race and 69.9% presented with stage I disease. The disease involvement in the right breast was as often as in the left breast. Approximately half of the patients (51.8%) received RT. The use of RT remained relatively constant (48-54%) over time (Fig. 2). The basic characteristics between patients received RT and those without RT were compared as shown in Table 1. There were no statistical differences between RT group and non-RT group in terms of age, race, tumor laterality, disease stages, and year of diagnosis.

Clinical outcomes
The 5-year OS and CSS rates for the whole cohort were 72.3% and 82.5%, respectively (Fig. S1). Kaplan-Meier analysis showed that there was no significant difference in OS between the groups of right and left laterality, stages I and II, and year of diagnosis of the 2001-2005    S2). When Kaplan-Meier survival curves were plotted for the RT and non-RT groups, the data revealed that RT significantly improved OS (Fig. 3, P = 0.031). Univariate analysis showed that RT was associated with better OS [RT versus non-RT: HR = 0.680 (0.478-0.969)] ( Table 2).
In multivariate Cox regression analysis (Table 2), RT was significantly and independently associated with improved OS (HR 0.668, 95% CI 0.469-0.953). This was consistent with the results of univariate analysis. Patients younger than 60 years of age were also independently associated with better OS (Table 2).

Discussion
Using the traditional Wiseman and Liao's definition [1], our study mainly focused on lymphoma patients with the disease localized to the breasts and the regional lymph nodes. We did not include systemic diseases with breast involvement, as in such cases the primary disease sites are unclear. Thus, patients with Ann Arbor stage III and IV were excluded. The rare incident of PB-DLBCL makes it difficult to conduct large randomized studies to determine the optimal treatment. Although prospective clinical trials were initiated with a purpose to assess the impact of rituximab in the de novo PB-DLBCL with or without RT, such attempts failed as the studies were stopped early due to slow enrollment (https://clinicaltrials. gov/ct2/show/NCT01279772; https://clinicaltrials.gov/ct2/ show/NCT01144754). Results from retrospective studies indicate that rituximab tends to prolong OS of PB-DLBCL. It is unknown whether the use of rituximab in PB-DLBCL might make it unnecessary to add RT, which could induce additional toxic side effect. In this study, we conducted a large population-based analysis and found that adding RT was associated with the improvement of overall survival in the rituximab era. Thus, consolidative RT might still be recommended for PB-DLBCL patients treated with rituximab.
Consistent with the published results, we also found PB-DLCBL mostly occurred in elderly patients, with a  Treatment strategies for PB-DLBCL have evolved during the past decades. Surgical intervention should be avoided, as surgery seems to be associated with high cause-specific mortality and inferior outcome. Combined modality approaches such as chemotherapy in combination with RT produce superior outcomes for PB-DLBCL. Chemotherapy is the backbone of lymphoma treatment, and regimen consisting of anthracycline was associated with improved outcome. RT in combination with chemotherapy might further consolidate chemotherapy response. A large retrospective IELSG-15 study underscores the importance of consolidative RT in the management of PB-DLBCL [3]. It was found that RT plus anthracyclinecontaining chemotherapy was associated with improved OS compared to RT or chemotherapy alone. A controlled study by Aviles et al. [9] also confirmed that RT followed by six cycles of CHOP regimen was associated with significantly improved outcomes. Multiple studies have demonstrated the benefit of adding rituximab to chemotherapy regimen in DLBCL of different sites [14][15][16][17]. Aviles et al. [9,10] conducted a prospective study and found that addition of rituximab to the intensive chemotherapy regimen (CEOP-14) offered no benefit in terms of complete response rate and OS. Several retrospective studies reported that addition of rituximab showed a tendency to prolong the progression-free survival (PFS) and OS of PB-DLBCL. A small-scale study showed that rituximab could significantly improve PB-DLBCL patients' PFS and OS, with the 5-year PFS and OS of 82.4% for 6 patients who received R-CHOP and 67.5% for 15 patients who received CHOP (P = 0.038) [8]. These results should be interpreted with caution due to the small number of patients in this study. In the prerituximab era, PB-DLBCL patients were considered to have inferior clinical outcome and high rates of local and CNS relapse compared with nodal DLBCL [3,18]. Yhim et al. [13] found that the difference in outcome disappeared when rituximab was integrated into systemic chemotherapy. They treated PB-DLBCL and nodal DLBCL patients with R-CHOP and showed a similar OS between the two groups (3-year OS rate, 82.2% vs. 90.7%, respectively; P = 0.345). The reason for such similar clinical outcome in the rituximab era might be that most PB-DLBCL exhibited a nongerminal center B-cell-like (non-GCB) phenotype, and multiple studies have confirmed that rituximab might overcome the adverse prognostic impact of the non-GCB phenotype [16,19,20]. Although there is a lack of prospective study to confirm the efficacy of rituximab in PB-DLBCL, it is still recommended to add rituximab to traditional chemotherapy regimen for PB-DLBCL treatment due to the high effectiveness of rituximab in eliminating B lymphoma cells and reduce the CNS relapse risk.
In line with the published results, our analysis of the SEER database also revealed the change in treatment modalities for PB-DLBCL over the past decades (Fig. S3). The use of surgery has declined to under 10%, and RT increased to about 50%. Although RT is commonly used in the management of this disease, it should be noted that not all patients received RT. From 2001 to 2014, the use of RT remained constant at about 50% of PB-DLBCL patients (Fig. 2). When giving treatment recommendations, many oncologists often consider lung and cardiovascular toxicity associated with RT. Potential risk of secondary malignancies is another concern. Nevertheless, it seems appropriate to recommend RT for PB-DLBCL patients with stage I disease, as these patients can benefit from consolidative RT. It is possible that involved site radiotherapy (ISRT) to PB-DLBCL patients without prophylactic RT to the uninvolved regional lymph nodes or contralateral breast could reduce the risk of radiationrelated toxicities to contralateral breast, lung, and thyroid gland.
Our study selected the year 2000 as a cutoff time, when rituximab began to be widely used in the clinic. The patients diagnosed with PB-DLBCL from 2001 to 2014 were included in this analysis. Survival analysis of the SEER database shows favorable outcomes, with a 5-year OS of 72.3% for the whole cohort, which is comparable with the previously published results. A summary of several large studies on PB-DLBCL is given in Table 3. The 5-year OS varied from 50% to 78.1%. The 5-year OS of 78.1% for PB-DLBCL patients in the RT group represents the best results reported.
When we analyze the impact of RT on clinical outcome, we observed a significant improvement in OS in patients received RT (HR 0.680, 95% CI 0.478-0.969, P = 0.017). The 5-year OS was 78.1% and 66.0% for patients treated with or without RT, respectively. Interestingly, factors such as age, race, tumor laterality and stage did not affect the difference between RT and no RT groups (Table 1), thus decreasing the potential bias affecting prognosis derived from disease characteristics. The survival rate at 5 years observed in the RT group is the best-reported outcome for patients with PB-DLBCL (Table 3). The 12.1% improvement of 5-year survival in the RT group was significant. We also for the first time evaluated the influence of tumor laterality of breast on survival and showed similar disease involvement of left and right breasts.
Previous study suggested that stage II might be a bad prognostic factor for PB-DLBCL. Our study showed that the 5-year OS for stage II was seemingly inferior to stage I, but such difference was not statistically significant. We also found that age older than 60 years was correlated with the poor outcome. Interestingly, patients of white race exhibited relatively worse OS compared to the nonwhite patients. A possible explanation is that the median age at diagnosis is generally older in white patients, as older age was associated with worse OS.
It should be noted that this population-based study based on SEER database has some limitations. First, information of specific treatment regimen and the time course of treatment are lacking. This makes it difficult to determine what proportions of PB-DLBCL patients actually received rituximab and anthracycline-containing chemotherapy. However, it is reasonable to assume that most patients received anthracycline-based chemotherapy, as our study excluded the patients with no or unknown chemotherapy. Another limitation is that the volume and dosing of RT for those patients were unavailable in SEER database. Also, the detailed pathological information was not available in SEER database. As DLBCL consists of many subtypes such as double-or triple-hit large B-cell lymphoma, the aggressive property of this kind of lymphoma could significantly affect the clinical outcome and thus could complicate data interpretation.
To our knowledge, this SEER-based analysis is the first large retrospective study to evaluate the role of RT in patients with PB-DLBCL in the rituximab era. Our study revealed that RT was associated with better OS, which was confirmed in both univariate and multivariate analyses. From 2001 to 2014, only about half of PB-DLBCL patients received RT. As RT could significantly improve treatment outcome regardless of age, stage, tumor laterality, and race, it seems that the other half of the PB-DLBCL patients could benefit from adding RT to the treatment protocols. Before the available of prospective randomized study to confirm the impact of rituximab and RT on PB-DLBCL, it seems beneficial to give consolidative RT in combination with immunochemotherapy in PB-DLBCL patients.