Introduction

Diffuse large B-cell lymphoma (DLBCL) is the most common non-Hodgkin lymphoma, with 25–30% of patients presenting as a limited-stage disease (stage I and II) [1, 2]. Although most DLBCL arises in lymph nodes, approximately 40% of limited-stage DLBCL (LS-DLBCL) arises in a variety of organs, such as the gastrointestinal tract, sinus/nose, breast, testis, thyroid, and many others [3, 4]. Extranodal DLBCL exhibits heterogeneous clinical and molecular features, with a subset of patients showing a more aggressive course with relapses and inferior outcomes compared to nodal disease [3, 5,6,7,8]. Specifically, extranodal involvement in sites such as the breast and uterus demonstrate a high prevalence of the ABC phenotype and the MCD (MYD88/CD79b mutated) genomic subtype, which might be associated with an increased risk of central nervous system relapse [9,10,11].

Patients with LS-DLBCL generally have an excellent prognosis treated with rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) [12,13,14,15,16], which has led to further question the role of consolidative radiotherapy (RT) in the rituximab era. The single-arm phase II study SWOG S0014 evaluated 3 cycles of R-CHOP with RT in LS-DLBCL patients with at least one stage-modified International Prognosis Index (SM-IPI) adverse risk factor. The 4-year progression-free survival (PFS) and 4-year overall survival (OS) rates were 92% and 88%, respectively [17]. Despite the inclusion of patients with more adverse risk factors, S0014 study appeared to improve outcomes compared to the historical SWOG cohort [18]. In addition, several retrospective studies have investigated the impact of consolidative RT in LS-DLBCL and observed continued benefit even in the rituximab era [19, 20]. However, the prospective LYSA/GOELAMS trial showed that R-CHOP alone is not inferior to R-CHOP followed by RT in patients with non-bulky LS-DLBCL [14]. The S1001 study prospectively evaluated a PET-CT adapted approach in patients with non-bulky LS-DLBCL. Interim PET (iPET) scanning was performed after three cycles of R-CHOP treatment. Patients with a negative iPET received an additional cycle of R-CHOP, while those with a positive iPET received radiotherapy (RT) plus ibritumomab tiuxetan radioimmunotherapy. The results indicated no statistical difference between the iPET-negative and iPET-positive groups [16]. It is important to note that in most studies, only 30–40% of patients included had extranodal involvement, or extranodal disease was not analyzed separately. Thus, extrapolating these findings to extranodal LS-DLBCL may not have the same outcomes.

Given recent studies have attempted to omit consolidative RT in LS-DLBCL and the benefit of consolidative RT in extranodal LS-DLBCL remains controversial, we undertook this multicenter retrospective study and aimed to clarify the role of consolidative RT in extranodal LS-DLBCL in the rituximab era.

Materials and methods

Patients

In this multicenter retrospective study, all patients with newly diagnosed extranodal LS-DLBCL treated with rituximab-based chemoimmunotherapy with or without consolidative RT were eligible for inclusion in the analysis. Data were collected from 5 institutes in China, with patients treated between January 2008 and December 2020. LS-DLBCL was defined as stage I or non-bulky disease stage II disease according to Ann Arbor classification [18]. The initial staging was assessed using positron emission tomography (PET-CT), contrast-enhance computed tomography (CT), or magnetic resonance imaging (MRI). All patients underwent a bone marrow biopsy. Lymphoma limited to lymph nodes, Waldeyer ring, thymus, or spleen was considered as a nodal disease and excluded. Patients with primary mediastinal lymphoma, primary central nervous system (CNS) lymphoma, and transformation of a previous indolent lymphoma were also excluded. Bulky disease was defined as any mass > 10 cm in maximum dimension. SM-IPI is based on age > 60 years, elevated lactated dehydrogenase (LDH), performance status ≥ 2, and stage II or IIE [18]. Cell of origin (COO) was classified as eighter germinal B-cell like (GCB) phenotype or non-GCB according to the Hans algorithm [21]. Staging workup and initial treatment for patients were performed according to local clinician discretion. Independent or centralized pathologic verification of diagnosis was not performed since all specimens had already undergone review by an expert hematopathologist. This study was approved by the institutional review board and was conducted in accordance with the Declaration of Helsinki.

Statistical methods

Patient characteristics were compared using the Pearson’s Χ2 test or Fisher’s exact test for categorical variables, and the Mann-Whitney U-test for continuous variables. OS was defined as the time from the initial diagnosis to death from any cause or to the last follow-up. PFS was defined as the time from the initial diagnosis to disease progression or death from any cause. Time to relapse was calculated from the date of diagnosis to the date of relapse, progression, or last follow-up. The median follow-up time was estimated by the reverse Kaplan-Meier method [22]. OS and PFS were calculated by the Kaplan–Meier method. Survival outcomes were compared using the log-rank test. Univariate and multivariate analyses for PFS and OS were performed using the Cox proportional hazard model. Competing risk analyses were performed using the Fine & Gray method, with death without relapse as competing events [23]. Significant variables (P < 0.1) in univariate analysis were included in multivariate analysis. Propensity score-matching with the nearest neighbor method with a caliper width of 0.2 was used to match patients treated with consolidative RT to those treated without. A two-tailed P value < 0.05 was considered statistically significant. All statistical analyses were performed using SPSS statistics (version 26.0) and R version 4.3.1.

Results

Patient characteristics

We identified 328 patients with newly diagnosed extranodal LS-DLBCL. The baseline characteristics and a comparison of characteristics between patients who received consolidative RT and those who did not are summarized in Table 1. The median age at diagnosis was 54 years (range, 18–88 years). By Ann Arbor staging system, 129 patients (39.3%) were diagnosed with stage IE disease, and 199 patients (60.7%) with stage IIE disease. According to SM-IPI, 36.3% of patients were classified as high-risk [2,3,4]. The most common sites of extranodal disease at diagnosis were as follows: stomach (21.0%), intestine (18.6%), sinus/nose (17.7%), and breast (15.5%) (Supplementary Table S1). With the exception of patients who underwent complete resection, baseline characteristics in patients who received consolidative RT or those who did not were not significantly different.

Table 1 Baseline patient characteristics
Full size table

Treatment and response

A total of 84 patients (25.6%) underwent complete surgical resection. In terms of the first-line chemotherapy regimen, 310 (94.5%) received R-CHOP, 12 (3.7%) received R-EPOCH (rituximab, etoposide, prednisone, vincristine, cyclophosphamide, and doxorubicin), and 6 (1.8%) received R-CHOPE (rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone, and etoposide). The median number of chemoimmunotherapy cycles overall was 6, with 74.4% (n = 244) of patients receiving 6–8 cycles. At the end of chemoimmunotherapy, 296 patients (90.2%) achieved complete response (CR), 20 patients (6.1%) achieved partial response (PR) and 12 patients (3.7%) experienced progressive disease (PD) during the initial treatment. Consolidative RT was given to 114 of the 296 patients who achieved CR, and to 15 patients who achieved PR. The median RT dose for patients who achieved CR was 36.0 Gy (range, 30.0–40.0 Gy). Patients were most likely to receive consolidative RT with primary thyroid (77.3% of 22), sinus/nose (46.6% of 58), breast (45.1% of 51), and stomach (36.2% of 69) involvement.

Survival outcome

With a median follow-up time of 5.1 years (range, 0.3–14.8 years), the 5-year PFS rate among all patients was 75.4% (95% CI 70.4%-80.8%) and the OS rate was 83.9% (95% CI 79.6%-88.4%), respectively. Fifty-five patients died during the follow-up period. The cause of death was lymphoma related in 46 patients, and non-lymphoma related in 9 patients.

On univariate analysis of all patients, elevated LDH, SM-IPI score, and consolidative RT significantly affected both PFS and OS. Stage II disease is also associated with inferior PFS (Supplemental Table 2).

Consolidative RT significantly improved the 5-year OS (90.0% vs. 80.0%, P = 0.002) and PFS rates (84.6% vs. 68.7, P = 0.003) compared with no RT (Fig. 1A and B). In the subgroup of patients who achieved a CR, consolidative RT was still associated with better 5-year OS (93.7% vs. 87.6%, P = 0.011) and PFS (89.9% vs. 74.7%, P = 0.001) compared with no RT.

Fig. 1
figure 1

Survival outcomes for patients with extranodal LS-DLBCL. (A) Overall survival by the administration of consolidative RT or not. (B) Progression-free survival by the administration of consolidative RT or not. (C) Overall survival by SM-IPI risk groups. (D) Progression-free survival by SM-IPI risk groups

Full size image

Patients with high SM-IPI risk had significantly worse OS (5-year, 77.3% vs. 88.1%, P = 0.005) and PFS (5-year, 68.9% vs. 79.8%, P = 0.005) compared to those with low SM-IPI risk (Fig. 1C and D).

In multivariate analyses (Table 2), patients who received consolidative RT had a superior OS (HR 0.399, 95% CI 0.213–0.747, P = 0.004) and PFS (HR 0.477, 95% CI 0.285–0.798, P = 0.005) compared to those who did not receive RT. High SM-IPI risk predicted inferior OS (HR 2.389, 95% CI 1.397–4.086, P = 0.001) and PFS (HR 1.929, 95% CI 1.224–3.039, P = 0.005).

Table 2 Multivariate Cox regression analysis for factors affecting OS and PFS
Full size table

Propensity score-matched analysis

Propensity score-matched analysis was performed to compare the outcome of patients treated with or without consolidative RT. A total of 127 patients in each group were matched based on sex, SM-IPI score, and COO. Among propensity score-matched patients, multivariate analysis showed the addition of consolidative RT still improved both OS (HR 0.228, 95% CI 0.111–0.467, P < 0.001) and PFS (HR 0.308, 95% CI 0.167–0.566, P < 0.001) than those who did not receive RT. The 5-year OS and PFS for patients who received consolidative RT and those who without were 90.0% vs. 78.9% and 84.6% vs. 67.2%, respectively. SM-IPI was an independent prognostic factor for OS (HR 4.738, 95% CI 2.380–9.430, P < 0.001) and PFS (HR 3.185, 95%CI 1.721–5.894, P < 0.001).

Pattern of relapses

During the follow-up period, 49 of 296 (16.6%) patients who achieved a CR developed disease relapse. Fifteen patients (30.6%) had relapsed at local sites, 31 patients (63.3%) at distant sites, and 3 patients (6.1%) at both local and distant sites. The majority of relapse (n = 45, 91.8%) occurred within 5 years, and 55.1% of patients (n = 27) developed relapse within the first 2 years. Consolidative RT significantly reduced the cumulative risk of relapse (P = 0.002). The 2- and 5-year cumulative risk of relapse in patients treated with consolidative RT and without RT were 3.6% vs. 12.8% and 10.0% vs. 24.6%, respectively (Fig. 2). Among the patients who received consolidative RT, no patients experienced failures within the radiation field.

Fig. 2
figure 2

Cumulative incidence of relapse by the administration of consolidative RT or not

Full size image

Discussion

In present multicenter retrospective study, we specially evaluated the survival benefit of consolidative in patients with extranodal LS-DLBCL. Our results indicate that consolidative RT significantly improved survival outcomes and reduced relapse risk for patients with extranodal LS-DLBCL even in the rituximab era. We also confirmed that SM-IPI had robust prognostic utility for patients with extranodal LS-DLBCL.

Although R-CHOP with or without consolidative RT remains common practice for LS-DLBCL, no randomized trial has clearly demonstrated the survival benefit for extranodal LS-DLBCL in the rituximab era. The role of consolidative RT in this special subgroup is contentious. The UNFOLDER trial randomized patients to R-CHOP with or without RT to bulky and/or extranodal disease. However, the trial was stopped early due to a higher rate of treatment failure in the arm without RT, which indicated additional treatment were required in those patients [24]. Some retrospective studies have evaluated the role of consolidative RT on outcome in extranodal LS-DLBCL and the results were inconsistent [6, 19, 25, 26]. In a recent retrospective study restricted to patients with stage I DLBCL, a subgroup analysis of patients with extranodal disease suggested a benefit of adding RT to chemoimmunotherapy [6]. These discrepancies might be due to different patient populations including stage and extranodal disease sites, small sample size, and radiation dose. In the current study, we demonstrated that even in the rituximab era, the use of consolidative RT significantly improved both OS and PFS for patients with extranodal LS-DLBCL. After performing propensity score-matched analysis, consolidative RT was still associated with a significant improvement in outcome. In addition, consolidative RT showed better local control at the original extranodal site and reduced relapse risk compared to chemoimmunotherapy alone. Considering there is a tendency to forego RT in LS-DLBCL patients, our results suggest caution against the omission of consolidative RT in patients with extranodal LS-DLBCL.

Patients with LS-DLBCL generally have excellent survival and those at risk for inferior prognosis were difficult to identify, especially for extranodal LS-DLBCL. Only a few retrospective studies identified several prognostic factors with a small sample size and heterogeneous composition and treatment [6, 27, 28]. We confirmed SM-IPI was an independent factor for extranodal LS-DLBCL in the rituximab era, which was consistent with the previous study [6]. According to SM-IPI, we identified two main risk groups: patients with low SM-IPI risk (0–1) have excellent outcomes and 5-year OS and PFS rates of 88.1% and 79.8%, while patients with high SM-IPI risk [2,3,4] have worse 5-year PFS and OS rates of 77.3% and 68.9%.

In recent years, several studies reported a continuous pattern of relapse beyond 5 years in LS-DLBCL, which is in contrast to advanced disease with most relapses occurring within the first 2 years after frontline treatment [12, 15, 29]. However, in this study, we observed most relapses (91.8%) occurred within 5 years, and 55.1% of relapses occurred within the first 2 years. These distinct relapse patterns may indicate biological differences between nodal and extranodal LS-DLBCL, with the extranodal disease having more aggressive behavior. However, these observations need to be confirmed with longer follow-up and further generic study.

This study has several limitations. As its retrospective nature, selection bias may exist. The small number of each extranodal disease site makes it difficult to analyze the prognostic value of extranodal location. However, we performed propensity score-matched analysis to overcome potential bias between the groups. Given the lack of a prospective study, this real-world experience may support clinical management decisions and help to guide further prospective trials.

Our study showed that the addition of consolidative RT improved outcomes of patients with extranodal LS-DLBCL even in the rituximab era. A further prospective trial is necessary to confirm the role of consolidative RT in extranodal LS-DLBCL and better define the group of patients who will benefit from consolidative RT.