Follicular lymphoma (FL) frequently transforms into diffuse large B-cell lymphoma (DLBCL). To clarify the associated clinicopathological prognostic parameters, we examined the correlation of 11 histopathological parameters with progression-free survival (PFS) and overall survival (OS) in 107 consecutive patients who had DLBCL with pre-existing (asynchronous) or synchronous FL. The patients comprised 58 men and 49 women with a median age of 56 years. For DLBCL, the complete response rate was 81%, overall response rate was 88%, and 5-year PFS and OS rates were 55% and 79%, respectively. Immunohistochemical analysis of the DLBCL component revealed the following positivity rates: CD10, 64%; Bcl-2, 83%; Bcl-6, 88%; MUM1, 42%; GCB, 82%; cMyc index ≥80%, 17%; and Ki-67 index ≥90%, 19%. IGH/BCL2 fusion was positive in 57% of DLBCL cases. In univariate analyses, asynchronous FL and DLBCL (24%, P = 0.021), 100% proportion of DLBCL (29%, P = 0.004), Bcl-2 positivity (P = 0.04), and high Ki-67 index (P = 0.003) were significantly correlated with shorter PFS. Asynchronous FL and DLBCL (P = 0.003), 100% proportion of DLBCL (P = 0.001), and high Ki-67 index (P = 0.004) were significantly correlated with shorter OS. In a multivariate analysis, asynchronous FL and DLBCL (P = 0.035) and 100% proportion of DLBCL (P = 0.016) were significantly correlated with shorter OS. Thus, asynchronism and 100% proportion of DLBCL, that is, FL relapsed as pure DLBCL, or FL and DLBCL at different sites, were significant predictors of unfavorable outcome of patients with DLBCL transformed from FL.
Follicular lymphoma (FL) is one of the most common subtypes of non-Hodgkin lymphoma in the Western world, accounting for 22% of all cases worldwide. The risk of FL transformation was reported to be approximately 20% at 8 years.[2, 3] Transformation to diffuse large B-cell lymphoma (DLBCL) is observed frequently, with cells most commonly resembling centroblasts, but occasionally resembling anaplastic large cells with CD30 expression. In rare cases, transformation to Burkitt or Burkitt-like lymphoma or precursor B-lymphoblastic lymphoma/acute lymphoid leukemia has been noted. Composite FL and Hodgkin lymphoma have been suggested to represent two morphologic manifestations of the same tumor clones.[8, 9]
Transformation to DLBCL is frequently associated with a rapidly progressive clinical course and death from a tumor that is refractory to treatment. Ghesquières et al. analyzed 60 DLBCLs transformed from low-grade B-cell lymphoma and reported that complete remission with multidrug chemotherapy regimens was achieved in 60% of the patients, but 48% of them relapsed. Overall survival (OS) and freedom-from-progression rates at 5 years were 57% and 33%, respectively.
In recent years, several analyses of genetic alterations that appear to affect the risk for FL transformation have been reported, including activation of cMYC,[6, 11] inactivation of TP53,[12, 13] and inactivation of p16INK4a.[14, 15] However, few histopathological and immunohistochemical analyses of transformed FL have been performed.[12, 16, 17] Hans et al. reported that FL grade 3 cases with a predominant diffuse component (>50% diffuse) had a significantly worse overall survival and event-free survival than the remaining FL grade 3 cases.
Since 2000, DLBCL has been subdivided into germinal center B-cell (GCB) and non-GCB subgroups (including the activated B-cell phenotype [ABC] and type 3 phenotype) by using the cDNA microarray technique.[18, 19] The GCB subgroup shows a better outcome and includes cases with a translocation (14;18)(q32;q21). In clinical practice, Hans et al. showed that a panel of immunohistochemical markers comprising CD10, Bcl-6, and MUM1 could be used on paraffin-embedded tissues to classify DLBCL as tumors with a GCB or non-GCB phenotype. Davies et al. examined 35 cases of transformed FL and found that 89% of them had a GCB phenotype and 9% had a non-GCB phenotype. We previously reported that 84% of transformed FL cases had a GCB phenotype and 16% had a non-GCB phenotype. However, the prognostic implications were not determined.
The translocation (14;18)(q32;q21) is present in 80–100% of FLs in Western countries,[21, 22] whereas in South-east Asia, including Japan, the incidence of translocation has been reported to be considerably lower at approximately 60%. The outcome of the patients with DLBCL transformed from FL with t(14;18) is unclear, although we recently reported that this fusion does not affect clinical outcome.
The aim of this study is to clarify the histopathological prognostic parameters of DLBCL transformed from FL. Eleven histopathological parameters influencing progression-free survival (PFS) and/or OS were evaluated using univariate and multivariate analyses, including the presence of synchronous/asynchronous FL and DLBCL, the proportion of DLBCL, the highest grade of FL, expressions of CD10, Bcl-2, Bcl-6, MUM1, cMyc, and Ki-67, classification as GCB/non-GCB subgroups, and detection of IGH/BCL2 fusion using fluorescence in situ hybridization (FISH) analysis.
Materials and Methods
The study subjects were 107 consecutive patients with DLBCL transformed from FL, treated at the National Cancer Center Hospital, Tokyo, Japan, between 2003 and 2010. Clinical information was extracted from medical records. The Ann Arbor system was used for staging. The staging procedures included bone marrow aspiration or biopsy, endoscopy of the upper gastrointestinal tract, computed tomography, and optionally, positron emission tomography. Following DLBCL diagnosis, most patients received rituximab-containing chemotherapy that consisted of rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) or R-CHOP-like regimens with or without irradiation (86 patients, 80%), and the remainder received chemotherapy and/or irradiation without rituximab (21 patients, 20%). The median follow-up time after DLBCL diagnosis was 70 months (range, 2–162 months), and that after the initial FL diagnosis was 93 months (3–411 months). Informed consent was obtained from each patient. The study was approved by the institutional review board of the National Cancer Center.
Biopsy materials were fixed in 10% neutral-buffered formalin, embedded in paraffin, cut into 4-μm-thick sections, and stained with H&E for histopathological evaluation. All specimens were evaluated by two pathologists (AMM and HTa) according to the “WHO Classification of Tumours of Hematopoietic and Lymphoid Tissues”, published in 2008. Diffuse large B-cell lymphoma was subclassified as centroblastic, anaplastic, immunoblastic, or T cell/histiocyte-rich variants. The presence of synchronous or asynchronous FL and DLBCL, the proportion of the DLBCL component, and the highest grade of FL were evaluated. If FL and pure DLBCL were detected at different sites, the proportion of the DLBCL component in the DLBCL specimen was indicated as 100%. In such cases, a FL component was usually found in bone marrow, the stomach, or the duodenum during the staging procedure.
Immunohistochemistry on formalin-fixed paraffin-embedded tissues was performed using a panel of monoclonal antibodies. The number of available specimens differed for each stain. Sections, 4 μm thick, were cut from each paraffin block, deparaffinized, and incubated at 121°C in citrate buffer, pH 6.0, for 10 min for antigen retrieval. The antibodies used included those against the following antigens: CD3 (PS1, ×25; Novocastra, Newcastle-upon-Tyne, UK), CD10 (56C6, ×100; Novocastra), CD20 (L26, ×200; Dako, Glostrup, Denmark), Bcl-2 (124, ×100; Dako), Bcl-6 (PG-B6p, ×20; Dako), cMyc (Y69, ×50; Epitomics, Burlingame, CA), Ki-67 (MIB-1, ×100; Dako), and MUM1 (MUM1p, ×200; Dako). An autostainer was used with the standard polymer method (Dako Autostainer Plus). All cases were positive for CD20 and negative for CD3. Immunoreactivity for CD10, Bcl-2, Bcl-6, and MUM1 was judged positive if more than 20% of the tumor cells were stained. The labeling index was counted for Ki-67 and cMyc: a high labeling index was defined as 90% or more for Ki-67 and 80% or more for cMyc. The area with the greatest staining was selected, 500 nuclei were counted manually, and the proportion of positive cells was calculated. To classify each case as having either a GCB phenotype or a non-GCB phenotype, a panel of three antigens (CD10, Bcl-6, and MUM1) was used according to the protocol described by Hans et al.
Interphase FISH analysis
Sections, 4-μm thick, were cut from each paraffin block and used for FISH analysis. Fluorescence in situ hybridization and judgment of the fusion gene was performed as described previously. Dual–color LSI IGH Spectrum Green/LSI BCL2 Spectrum Orange Dual Fusion Translocation Probes (Abbott, Downers Grove, IL, USA) were used to detect IGH/BCL2 fusion.
All survival curves were estimated using the Kaplan–Meier method. Univariate and multivariate analyses were performed using Cox's proportional hazard model. Multivariate analysis was performed for significant parameters detected by univariate analyses. Differences were considered significant when the P-value was <0.05. Overall survival for DLBCL was calculated from the date of DLBCL diagnosis to the date of death from any cause. Overall survival for FL was calculated from the date of FL diagnosis to the date of death from any cause. Progression-free survival for DLBCL was calculated from the initial date of treatment to the date of disease progression, relapse, or death from any cause.
Clinical information is summarized in Table 1. The patients comprised 58 men and 49 women, ranging in age from 19 to 83 years with a median age of 56 years at the time of DLBCL diagnosis. Synchronous FL and DLBCL were seen in 81 patients (76%) and asynchronous FL and DLBCL (FL followed by DLBCL) were seen in 26 patients (24%). Among the patients with asynchronous FL and DLBCL, the interval between the diagnosis of FL and DLBCL ranged from 11 to 263 months with a median of 71 months. In a total of 107 patients, rituximab-containing chemotherapy for DLBCL was administered to 86 patients (80%) and chemotherapy without rituximab was administered to 21 patients (20%); PFS and OS did not differ significantly between the two groups. For DLBCL, the complete response (CR) rate was 81%, and the overall response rate was 88%. After DLBCL diagnosis, the 5-year PFS rate was 55% and the 5-year OS rate was 79%. After the initial FL diagnosis, the 5-year OS rate was 88% and the 10-year OS rate was 80%. The number of regimens before the diagnosis of DLBCL was 0 in 81 patients (76%), 1–2 in 16 patients (15%), and ≥3 in 10 patients (9%); PFS and OS differed significantly between patients receiving 0–2 regimens and those receiving ≥3 regimens (P = 0.009 and P < 0.0001, respectively).
Table 1. Characteristics of 107 patients with DLBCL transformed from FL
Histopathological, immunohistochemical, and FISH results
Histopathological results are summarized in Table 2. Follicular lymphoma and DLBCL components were synchronous in 76% (81/107) of the patients and asynchronous in 24% (26/107) of the patients. The proportion of DLBCL was <50% in 33% (35/107) of the patients, 50% to <100% in 38% (41/107) of the patients, and 100% in 29% (31/107) of the patients. The highest grade of FL was 1–2, 3a, and 3b in 31% (33/107), 47% (50/107), and 22% (24/107) of the patients, respectively. The 104 DLBCLs were centroblastic subtype and the remaining three DLBCLs were anaplastic subtype.
Table 2. Histopathological parameters and Cox's proportional univariate analyses
No. patients (%)
5-year PFS rate(%), DLBCL
5-yr OS rate (%), DLBCL
DLBCL, diffuse large B-cell lymphoma; GCB, germinal center B-cell phenotype; FISH, fluorescence in situ hybridization; FL, follicular lymphoma; NS, not significant; OS, overall survival; PFS, progression-free survival.
P-value, hazard ratio (HR), and confidence interval (CI) of univariate analyses were calculated according to Cox's proportional hazard model.
FL and DLBCL
Proportion of DLBCL
100% (different sites)
Highest grade of FL
Immunohistochemistry for DLBCL
Paraffin sections were available for the all 107 DLBCL specimens, but in a few cases, sections were not available for some of the markers. All DLBCL cases were positive for CD20 and negative for CD3. Immunohistochemical analysis of the DLBCL component revealed the following positivity rates: CD10, 64% (69/107); Bcl-2, 83% (87/105); Bcl-6, 88% (64/73); MUM1, 42% (30/72); GCB, 82% (77/94); cMyc index ≥80%, 17% (12/72), and Ki-67 index ≥90%, 19% (15/81). The IGH/BCL2 fusion was positive in 57% (29/51) of the patients examined.
Univariate analysis revealed that asynchronous FL and DLBCL (P = 0.021), 100% proportion of DLBCL (P = 0.004), Bcl-2 positivity (P = 0.04), and high Ki-67 index (≥90%) (P = 0.003) were significant parameters for shorter PFS (Table 2). Similarly, asynchronous FL and DLBCL (P =0.003), 100% proportion of DLBCL (P = 0.001), and high Ki-67 index (P = 0.004) were significant factors contributing to shorter OS (Table 2).
In the multivariate analysis for the significant parameters detected in univariate analyses, asynchronous FL and DLBCL (P = 0.035) and 100% proportion of DLBCL (P = 0.016) were independently correlated with shorter OS (Table 3). No parameters were significantly correlated with shorter PFS.
Table 3. Results of Cox's multivariate proportional hazards analysis for overall survival
We performed subgroup analyses to examine the predictive value of asynchronous/synchronous and proportion of DLBCL for PFS and OS. Progression-free survival and OS curves were compared among the following four groups (Fig. 1): asynchronous/100% proportion of DLBCL (FL relapsed as DLBCL or FL and DLBCL at different sites) (16 patients), asynchronous/<100% proportion of DLBCL (FL relapsed as mixed FL and DLBCL) (10 patients), synchronous/100% proportion of DLBCL (FL and DLBCL at different sites) (15 patients, Fig. 2a,b), and synchronous/<100% proportion of DLBCL (synchronous mixed FL and DLBCL, Fig. 2c–e). The asynchronous/100% proportion of the DLBCL group had significantly shorter PFS (P = 0.001; 5-year PFS rate, 24%) and OS (P < 0.0001; 5-year OS rate, 30%) than the remaining three groups, but significant differences were not found among the remaining three groups (Fig. 3a,b). The latter three groups, synchronous/mixed FL and DLBCL, synchronous/FL and DLBCL at different sites, and asynchronous/FL relapsed as mixed FL and DLBCL groups, showed a tendency of better OS: 5-year OS rates were 87%, 79%, and 90%, respectively.
Follicular lymphoma transforms primarily to DLBCL followed by intermediate DLBCL/Burkitt lymphoma, classical Hodgkin lymphoma, and B acute lymphoblastic leukemia. Transformation of FL to DLBCL is currently the focus of widespread clinical and pathological interest. We therefore studied the histopathological prognostic parameters of DLBCL transformed from FL using morphological, immunohistochemical, and FISH analyses.
To evaluate the outcome of patients with DLBCL transformed from low-grade B-cell lymphoma, Ghesquières et al. analyzed 60 such cases, and reported that OS and freedom-from-progression rates at 5 years were 57% and 33%, respectively. The better outcomes for patients in our present study (5-year PFS rate, 55% and 5-year OS rate, 79%) might be partly associated with the younger median age (56 years) and relative frequency of cases with a low/low-intermediate International Prognostic Index. In addition, most of these patients received rituximab-containing treatment, although the benefit of rituximab was not proved in the current study.
Multivariate analysis revealed that asynchronism and 100% proportion of DLBCL were significantly correlated with shorter OS. In addition, a high number of regimens administered (≥3) prior to transformation to DLBCL was correlated with a poor outcome. Our results suggested that patients with asynchronous FL and DLBCL, that is, patients who received prior chemotherapy regimens before the diagnosis of DLBCL, might have developed chemoresistance before receiving treatment for DLBCL. Among patients with DLBCL transformed from FL, the proportion of large lymphoid tumor cells ranged from 16 cells/high-power field to 100%. The present study suggested that a high proportion of large tumor cells as 100% was a poor prognostic parameter. As a result, subgroup analysis according to the presence of synchronous and asynchronous FL and DLBCL, and the proportion of DLBCL cells revealed that asynchronous FL and DLBCL and a 100% proportion of DLBCL cells (FL relapsed as pure DLBCL or FL and DLBCL at different sites) had a significantly worse outcome than the remaining three groups. In contrast, the remaining three groups had a tendency of better outcome, suggesting that initially treated patients with synchronous FL and DLBCL and those with asynchronous FL and DLBCL in whom FL relapsed as mixed FL and DLBCL (histologically “transforming” FL) showed better PFS and OS.
The Ki-67 index is associated with the proliferation activity of tumors. A high Ki-67 index is reported to be a poor prognostic parameter in de novo DLBCL; however, the issue is controversial. In the present study, Ki-67 was a significant predictor of PFS and OS in univariate analyses, although this significance was not sustained in multivariate analysis. We found that a 100% proportion of DLBCL in DLBCL transformed from FL displayed a worse outcome than the other groups. In addition, our results suggested that transformation to pure DLBCL with a high Ki-67 index was a poor prognostic indicator for transformed FL.
The Bcl-2 protein, an anti-apoptotic molecule, is expressed on resting B and T cells, but not on normal germinal center cells. Bcl-2 positivity has been reported to be associated with unfavorable prognosis in de novo DLBCL.[26, 28, 29] In the present study, Bcl-2 positivity predicted significantly shorter PFS in cases of DLBCL transformed from FL, and patients negative for Bcl-2 showed 100% 5-year OS. Conversely, patients with a low FL grade who were positive for IGH/BCL2 fusion tended to have poorer PFS and OS, although these results were not significant. We found that 57% of the cases had IGH/BCL2 fusion in DLBCL transformed from FL, and this frequency was similar to that reported for Japanese FL cases. Because Bcl-2 positivity, low grade of FL, and IGH/BCL2 fusion positivity were usually correlated with each other, they might be markers indicating a poor prognosis for cases of DLBCL transformed from FL.
In the present study, CD10, Bcl-6, and MUM1 were expressed in 64%, 88%, and 42% of the cases of DLBCL transformed from FL, and 82% were of GCB phenotype, whereas 18% were of non-GCB phenotype. Several previous studies examining difference in prognosis between de novo DLBCL patients with GCB phenotype and those with non-GCB phenotype revealed that the former group had a more favorable prognosis.[30, 31] However, this has recently become a controversial issue.[26, 32] In the present study, neither GCB versus non-GCB nor the expression of CD10, Bcl-6, or MUM1 was a significant prognostic factor for cases of DLBCL transformed from FL.
In conclusion, asynchronism and 100% proportion of DLBCL, that is, FL relapsed as pure DLBCL or FL and DLBCL at different sites, were predictors of shorter PFS and OS in patients with DLBCL transformed from FL.
The study was supported in part by the Foundation for Promotion of Cancer Research, National Cancer Center Research and Development Fund (23-A-17 and 23-A-23), and the Grant-in-Aid for the Third-Term Comprehensive 10-Year Strategy for Cancer Control from the Ministry of Health, Labour and Welfare, Japan. The authors would like to thank Ms C. Kina and Ms S. Miura for their technical assistance with immunohistochemistry.
Kensei Tobinai received research grants from Zenyaku Kogyo and Chugai Pharmaceutical. All the other authors, including the corresponding author, have declared no conflicts of interest.