INTRODUCTION

According to the international agency for research on cancer, in 2015, approximately 414,772 new cases of non-Hodgkin lymphoma (NHL) were diagnosed worldwide¹. Diffuse large B-cell lymphoma (DLBCL) is the most frequent type of NHL. Although this disease is potentially curable², the outcome of patients may vary according to the primary site involved. Before the WHO classification of lymphoproliferative disorders³, involvement of the central nervous system (CNS) was the most frequent extranodal manifestation; however, this site constitutes an independent entity⁴. Recently, some authors have reported that primary extranodal (PE) involvement may be associated with a longer survival in comparison with patients with only nodal involvement in a study of patients with DLBCL of the head and neck⁵. In fact, primary breast, primary bone, and primary craniofacial DLBCL are extranodal manifestations of DLBCL, and most studies agree with the low incidence in these sites. As a primary manifestation, primary breast lymphoma represents up to 0.5% of all NHL and approximately 2% of extranodal presentations⁶,⁷; primary bone DLBCL constitutes approximately 4-5% of extranodal lymphomas and approximately 3% of all malignant bone tumors⁸,⁹; and craniofacial DLBCL accounts for approximately 1% of all lymphomas and 2-12% of extranodal lymphomas¹⁰,¹¹.

Because of the rarity of these manifestations, most of the studies are limited and, as far as we know, there is no available information regarding extranodal DLBCL (N-DLBCL) (extranodal-DLBCL) in Latin American countries. The objective of this work was to study at a single academic cancer center, the relative frequency of the location, clinical characteristics, clinical response, and survival of patients with these primary sites and compares them with those of the classical primary N-DLBCL.

METHODS

RESULTS

Treatment, Response, and Survival

Patients were treated with either chemotherapy only (n = 361, 56.6%) or chemotherapy plus radiotherapy (n = 275, 43.1%), and one patient (0.15%) was treated with surgery and chemotherapy. Table 2 details the treatment modality, scheme and response in all cohorts. A higher complete response rate was documented in cohorts E and C (100% and 91%, respectively), compared with other sites of extranodal disease (cohort A: 83.3%, cohort B: 80%, or cohort D: 73.3%). It is important to emphasize that none of the patients within cohort A had bulky disease; therefore, they were treated only with chemotherapy.

Table 2 Treatment and responses according to nodal/extranodal group 

Variable	Extranodal groups					Nodal F	p-value
	A Breast	B Bone	C Craniofacial	D Gastrointestinal	E Testicular
Treatment n (%)
Chemotherapy	6 (100)	1 (20)	5 (41.67)	15	-	332 (56.65)	<0.0001
Chemotherapy and radiotherapy	-	4 (80)	7 (58.33)	9	1 (33)	254 (43.35)
Surgery and chemotherapy				1	2 (66)
Chemotherapy scheme n (%)
R CHOP	6 (100)	3 (60)	10 (83.33)	22 (88)	3 (100)	513 (87.54)	<0.0001
R COP	-	1 (20)	2 (16.67)	2 (8)	-	36 (6.14)
CHOP	-	1 (20)	-	1 (4)	-	31 (5.29)
DA EPOCH R	-	-	-	-	-	6 (1.03)
Response n (%)
CR	5 (83.33)	4 (80)	11 (91.67)	20 (80)	3 (100)	430 (73.38)	0.958
PR	1 (16.67)	-	1 (8.33)	1 (4)	-	41 (6.99)
SD	-	-	-	1 (4)	-	5 (0.86)
PD	-	1 (20)	-	3 (12)	-	110 (18.77)
Relapse	2 (33.33)	0	0	6 (24)	1 (33)	44 (7.50)	0.078
Median follow-up (months)	26	21	25	32	16	27	0.658
DFS 5-y	42	100	100	76	53	60	0.144
OS 5-y	100	80	58			62	0.268

R CHOP: rituximab, cyclophosphamide, doxorubicin, vincristine and prednisone; R COP: rituximab, cyclophosphamide, vincristine and prednisone; CHOP: cyclophosphamide, doxorubicin, vincristine and prednisone; DA EPOCH R: dose-adjusted etoposide, prednisone, vincristine, cyclophosphamide, rituximab, CR: complete response, PR: partial response, SD: stable disease, PD: progressive disease, DFS: disease-free survival, OS: overall survival, 5y: 5-year.

The follow-up period was similar between the cohorts. Median PFS was 37 and 73 months for patients within cohorts A and F, respectively. In cohorts B, C, D, and E, the median PFS was not reached until after 5 years of follow-up (Fig. 1). During this time, the higher relapse rate was in cohort F, with 207 cases achieving a relapse rate of 35.5%. One patient (33%) within cohort E relapsed, two cases in cohort A relapsed (33.3%) at the CNS, and this relapse rate was higher and significant (p = 0.001) compared with patients from cohort F, where only 3 (0.51%) relapsed at this site. Similarly, six patients (24%) in cohort D relapsed at retroperitoneum, and no relapses were documented during follow-up in cohorts B or C. The PFS and OS are shown in Figures 1 and 2, respectively. As detailed in table 2, OS was higher in cohort A, followed by cohort B (Fig. 2). During follow-up, three non-lymphoma-related deaths were documented, as follows: two patients (cohorts B and C, one each) died because of cardiovascular related incidents, and one had an accidental death (cohort C).

Figure 1 Progression-free survival, showing that patients with primary centrofacial, and primary bone diffuse large B-cell lymphoma (DLBCL) had better survival, in comparison with the other groups of DLBCL. 

Figure 2 Overall survival was higher for patients with primary breast, primary bone-diffuse large B-cell lymphoma (DLBCL), and gastrointestinal DLBCL. 

By multivariate analysis employing the Cox model, from all the clinical, biochemical, and histopathological molecular classification parameters included, none were independently associated with PFS nor OS. The achievement of a complete response was the single factor associated with PFS and OS.

DISCUSSION

PE-DLBCL is a rare and diverse pathological entity. Several efforts have been made to evaluate the clinical characteristics and management approaches of these varied presentations and they are associated with excellent disease control by employing modern therapies⁵,⁸,¹¹,¹⁸-²².

Regarding primary breast DLBCL, the relative frequency of diagnosis was higher (0.9%) than reported by other authors (0.04-0.5%)⁶,¹⁸. Clinically, our findings were consistent with those reported on literature, where the median age was 53.3 years, according to the range of 40-69 years reported in most studies⁶,¹⁸-²⁰,²³. In contrast, with N-DLBCL, data from eight US academic institutions¹⁸, concluded that the molecular classification of DLBCL into a GC and non-GC types had no impact on survival, and other authors have confirmed these results²⁴. In this cohort study, the most frequent subtype was non-GC, and a low-frequency of relapses occurred regardless of the molecular type (one in GC type and one in non-GC type). Different authors have reported a relapse rate ranging from 12% to 44% in the ipsilateral or contralateral breast²³. Ryan reported that ipsilateral relapse occurred within 2.6 years from the beginning of treatment, whereas contralateral breast relapse occurred within up to 13.3 years. In the present study, one patient (16.6%) had bilateral involvement at diagnosis, a higher frequency than recently reported by Lalani et al.¹⁹. None of our patients had either ipsilateral or contralateral relapse after a 5-year follow-up; however, two patients relapsed at CNS (33.3%), at 8 and 36 months. This relapse rate is slightly higher than that reported on literature, of 8-29%⁶,¹⁸,²³. The international extranodal lymphoma study group²³ defined the following prognostic factors influencing OS: IPI, anthracycline-containing regimen, and radiotherapy. A negative association has also been reported between patients undergoing radical mastectomy and OS²³,²⁵,²⁶. In this cohort study, most patients (80%) were within the low-risk group of IPI scores; all were treated with anthracyclines, and none were submitted to mastectomy. Each of these factors could have had a positive impact on OS, as shown in figure 2. In fact, primary breast DLBCL remains a rare disease, and no clear consensus concerning therapy has emerged to date. However, the most recent reports agree with the routine use of rituximab-based chemotherapy added to radiotherapy and CNS prophylaxis¹⁸,²⁰.

Primary bone DLBCL is also a rare, primary extranodal lymphoma that occurs with a median age of onset ranging from 40-years old and is predominant in males⁸,¹¹,²⁷, as was the case in this study. The most common initial symptoms described are local pain followed by nerve compression and local mass⁸, as in our study: about 80% of patients had pain, 40% referred paraparesis, and 20% had a bone fracture. A study from Japan²⁷ reported that the pelvis was the primary site involved (54%) rather than the extremities; however, Matikas et al.²⁸ and Hayase et al.²¹ described the spine as the most frequent primary involved site. In our study, 40% of patients had an axial localization, 40% were appendicular, and 20% had multifocal (axial and appendicular) bone involvement. Although elevated levels of LDH have been reported in 45-70% of patients¹⁰,²⁹, we documented normal levels of LDH in 80% of our cases. In agreement with Liu et al.³⁰ and Hattori et al.²⁷, most of our patients (80%) did not have B symptoms and 60% were classified within high-intermediate or high-risk groups, according to IPI score. This finding was also similar to 61% reported by Hattori et al.²⁷. In addition, it has been reported that the presence of bone fractures is a negative prognostic factor for OS⁸. We had only one patient with a bone fracture, and this patient responded well to chemotherapy plus radiotherapy and remained disease-free for 3 years of follow-up. Regarding treatment, a large study from Taiwan analyzing 11 years of experience³⁰, demonstrated that the administration of ≥ 6 cycles of chemotherapy had a positive impact on OS. The addition of radiotherapy has also been associated with an improvement in DFS, from 63% to 88%¹¹, as well as an increase from 49% to 57.4% in OS⁸. All of our patients received rituximab-based chemotherapy combined with radiotherapy, and the DFS and OS to 5 years in our study were 100% and 80%, respectively.

With regard to primary craniofacial DLBCL, the differences in biological characteristics and the prognosis of lymphomas of nodal and extranodal origin of the head and neck have demonstrated a better outcome in patients with the primary extranodal disease³¹. In the same direction, other authors³²,³³ have reported that most patients with primary craniofacial DLBCL are diagnosed when they are older than 60 years, with an adequate ECOG status (0-1 groups), and low- or low-intermediate-risk groups, according to IPI score; these findings were consistent with our study, where the mean age was 67.9 years old, 75% of cases were in ECOG 0-1, and 91% were considered as low-intermediate or low-risk groups in IPI score. We found that sinonasal cavities were affected in 58.3% of patients, followed by the oral cavity in 41.7%. A study including only patients with sinonasal cavity involvement described an association with HIV or EBV infection in 32% of cases, and a clear predominance for the maxillary sinus (50%), followed by the ethmoid sinus (23%), nasal cavities (18%), and sphenoid sinus (9%). In our study, all patients were HIV negative. EBV status was not evaluated in this series. Different studies report no consensus regarding treatment. Lombard et al.²², in a study including 22 patients, treated three with localized involvement with radiotherapy, 16 patients with CHOP chemotherapy only, and the rest with combined chemotherapy plus radiotherapy. However, eight patients with an incomplete response required additional treatment, including salvage chemotherapy and bone marrow autograft (4 cases), and seven cases required complementary craniofacial radiotherapy. The OS rate was 73% at 36 months. It is important to note that patients treated with chemotherapy did not receive rituximab, which is now a standard drug in the treatment of this entity. The role of radiotherapy, as well as CNS prophylaxis in these patients, has been retrospectively analyzed by Murawski et al.³⁴ in 11 consecutive trials of the German High-Grade NHL study group, to identify factors that affect the outcome, and they concluded that the addition of rituximab abolishes the increased risk of CNS relapse. Therefore, intrathecal prophylaxis is not recommended. In our study, 66.6% were treated with a combined modality (rituximab-based chemotherapy plus radiotherapy) and one-third with rituximab-based chemotherapy only. No relapse was documented, and our OS was 78%-5 years, which is similar to that reported in literature²²,³⁴.

With respect to primary gastrointestinal lymphoma, the GI tract is the most common site of extranodal NHL. A prospective study from South India recently published³⁵ found that gastrointestinal lymphoma constitutes about 10-15% of all NHL. In this series, we documented only 25 cases, and this low frequency may be explained by the fact that we did not include other histologies, such as all low-grade lymphomas. As has been described³⁵,³⁶, our patients had a mean age of 55 years, but in contrast with the same series³⁵,³⁶, we documented a predominance of females; the reason for this predominance is unknown. Pain and gastrointestinal hemorrhage were the most frequent symptoms, followed by decreased appetite and weight loss, as has been reported³⁵-³⁷. In the same direction, the stomach was the most common site, followed by colon and small intestine. Although other series³-³⁵ have reported DLBCL involving the esophagus, we did not find any case affected at this site.

Primary testicular lymphoma has been described as a rare, clinically aggressive form of extranodal lymphoma³⁸, with a higher frequency in the seventh decade of life, as was in our three patients. The typical presentation with a firm mass was documented in our patients. Although HIV infection has been considered a risk factor³⁹, all our patients were HIV-negative. The predominance of non-GC type was consistent in this cohort. The presence of B symptoms and elevated LDH levels has been defined as negative prognostic factors both were present in our 3 cases, but only relapsed. As has been described, all received chemotherapy including rituximab and one received additional radiotherapy.

These PE-DLBCLs constitute rare, primary sites of lymphoproliferative disorders in most cases, with localized disease and good prognosis. A relative lower frequency of extranodal primary DLBCL was found in this series, and it may be secondary to a referral bias since this is a national cancer center. However, they require combined chemoimmunotherapy with radiotherapy in most cases to improve the local and systemic disease.