The median age at diagnosis in Western countries is 60–64 years[4, 16], in East Asian countries the median age is lower (40–60 years)[5]. The median age was 57 years (range, 23–86 years) in our study, consistent with other reports. No specific or unusual manifestations and imaging features of breast lymphoma have been documented to distinguish PBL from breast carcinoma[17]. The most common signs and symptoms of breast lymphoma include a painless enlarging palpable mass (61% of cases), with a median diameter of 3-5cm and usually located in the external quadrant[6, 18]. The median diameter of the size of the primary tumor in the breasts was 2.9 cm (range, 0.4–5.3 cm).
Furthermore, this study investigated the association between pathological features and prognosis in patients with PBL. PB-DLBCL is the most common histologic subtype of PBL, comprising 40–80%[4, 13, 19, 20]. MZL (9–28%) and MALT lymphoma (12.2%) are the next most frequent[1, 4, 21, 22]. FL(10%-19%)[1, 21, 23] and Burkitt lymphoma (1%-5%)[24, 25] are also identified in the breast but have a lower prevalence[4]. In this study, PB-DLBCL accounts for a major portion (80.3%), followed by MZL/MALT (5.6%) and FL (4.2%). Among the 7 patients with recurrence, two of them were MZL/MALT subtypes, and the others were all DLBCL subtypes.
By the Han′s classification[15], DLBCL can be clinically divided into GCB types and non-GBC types. Previous studies showed that the majority of patients with PB-DLBCL are of the non-GCB, yet reported treatment outcomes were similar between the GCB and nonGCB patients. In line with the aforementioned studies, 63.8% of patients with PB-DLBCL were of the non-GCB type, accounting for a majority of these cases. Consistent with other studies, the 5-year OS and PFS in patients with GCB type were not significantly higher than were patients with the non-GCB type (P = 0.790; P = 0.779; Fig. 1a, 1b).
CD5 positivity was observed in 26.7% of our cases, higher than the prevalence of CD5 positivity reported in the general DLBCL population (5%-10%)[26]. The CD5 + DLBCL generally shows a poor prognosis and frequent CNS relapses. While the univariate Cox regression reached statistical significance, the multivariate Cox regression revealed that the positive of CD5 may not be an independent prognostic factor for PFS and OS. The c-Myc gene is a kind of proto-oncogene. Lymphoma with c-Myc gene translocations all showed aggressive clinical features, with poor response to treatment and poor prognosis. The c-Myc was observed as 32.4% positive and the EBER 9.9% positive, but no significant correlations were found for these two genes. BCL-2 and BCL-6 were 56.3% and 47.9% positive, respectively. Co-expression of c-Myc and BCL-2 was 32.4% observed, termed double-expresser lymphoma, has been suggested to be a prognostic factor of DLBCL[27]. But no significant correlations were found (Table 2).
For DLBCL, the correlation between IPI scores and survival outcomes in PBL has been controversial. Studies by Ryan et al.[6], Hu et al.[9], Zhang et al.[28], Shao et al.[12], Yhim et al.[29] demonstrated the IPI scores were independent risk factors for PFS and OS, while results from Zhao et al.[30], Qu et al.[31] and Zhang et al.[32] confirmed the IPI scores were not associated with PFS and OS in PBL. The result was presented in the forest graph format (Fig. 2). In this study, 45 (63.4%) patients had an IPI score of 0–1, and 26 (36.6%) patients had a high IPI score (≥ 2). The univariate and multivariate COX regression analysis both illustrated that IPI scores were independent prognostic factors for PFS and OS (P < 0.05, Table 2). The survival curves revealed that low IPI scores (0–1) were significantly associated with PFS and OS. (Fig. 1c, 1d).
Though surgical intervention beyond excisional biopsy has been shown to offer no improvement in survival or recurrence risk[2, 8, 9, 33, 34], in the present study, 60.6% of all patients diagnosed by surgery. Through the search of 27 studies, the ranges of rates of surgery were 4.4%[6]-100.0%[35]. The surgery rates in these three hospitals were 64.1%, 40.0%, and 100.0%, respectively. One reason could be that most FNAC results only suggested but not definitively diagnosed PBL and the second could be some patients with palpable masses tend to undergo surgery. The present study further corroborates that surgery offered no survival benefit and even associated with a trend toward worse OS[36]. We did not find this association, likely due to the comprehensive treatment.
By searching the reported literature, a total of 12 articles contained the comparisons of patient outcomes with/without rituximab (Supplementary files Table 2). Among these 12 studies, ten studies[2, 8, 12, 13, 31, 32, 37–39] found no significant improvement in PFS or OS from the addition of rituximab, whereas another three retrospective series[9, 40, 41] reported significantly better 5-year PFS or OS for PBL patients treated with RCHOP vs CHOP alone. Our findings demonstrated no statistically significant difference in PFS and OS with the addition of rituximab in chemotherapy of PBL (Fig. 1e, 1f).
The significant association between radiotherapy and prognosis of PBL has been verified in most studies, and the multivariate Cox regression analysis confirmed this conclusion (Table 2). By the subgroup analysis, survival analysis of IPI scores for the PFS and OS of patients using rituximab were not significantly different (P > 0.05). (Fig. 3a, 3b). But in patients who did not use rituximab, significant differences were found in the survival curves for the PFS (P = 0.034) (Fig. 3d). No significant differences were observed between survival curves for PFS in CHOP group (P = 0.083). (Fig. 3c). The results suggested that in the patients without rituximab, IPI proved to be valuable in predicting the prognosis of PBL; however, IPI seems to have lost some of its predictive value with the addition of the rituximab.
In this study, only six patients received prophylactic intrathecal chemotherapy, which may be due to the controversial clinical evidence on its benefit. The results of large-scale retrospective studies have not found that prophylactic intrathecal chemotherapy can prevent the recurrence of CNS for PBL patients. Hosein et al.[8] summarized the data of 76 PBL patients from 8 centers in the United States with the follow-up of 10 years and found that the incidence of CNS recurrence in PBL patients was high, and the incidence of CNS was not related to disease stage, LDH level, IPI score, and use of rituximab. Though none of these six patients had CNS infiltration, a correlation between prophylactic intrathecal chemotherapy and prognosis cannot be proven.
Several limitations need to be noted in this study. First, due to insufficient awareness in the early years, PET/CT was not widely used, especially in local hospitals. Not all patients are assessed for CNS involvement at diagnosis, which may lead to underestimation and inclusion of some patients with occult CNS disease. Due to the limited number of patients who received prophylactic intrathecal chemotherapy, we did not explore the preventive interventions of CNS. Further large-scale prospective clinical trials are needed to investigate how to prevent the CNS recurrence of PBL. Last, clinical data were often incomplete and difficult to unify, as this was a real-world multicenter study.
This study has several strengths, including the large sample size, long-term follow-up, and participating hospitals were from different parts of the country. The other two hospitals are both regional medical centers with a large number of patients. Medical institutions are widely distributed in China, with a great disparity in the medical levels of different regions, so it is necessary to consider differences in medical centers.