Diffuse large B-cell lymphoma

hepatocellular carcinoma patients: evaluation of prognostic factors. Eur J Gastroenterol Hepatol 2011; 23: 1233–1238. 9. Demetri GD, van Oosterom AT, Garrett CR et al. Efficacy and safety of sunitinib in patients with advanced gastrointestinal stromal tumour after failure of imatinib: a randomised controlled trial. Lancet 2006; 368: 1329–1338. 10. Lipworth AD, Robert C, Zhu Ax. Hand-foot syndrome (Hand-foot skin reaction, palmar-plantar erythrodysesthesia): focus on sorafenib and sunitinib. Oncology 2009; 77: 257–271. 11. Van der Veldt AA, Boven E, Helgason HH et al. Predictive factors for severe toxicity of sunitinib in unselected patients with advanced renal cell cancer. Br J Cancer 2008; 99: 259–265. 12. Chu D, Lacouture ME, Weiner E, Wu S. Risk of hand-foot skin reaction with the multitargeted kinase inhibitor sunitinib in patients with renal cell and nonrenalcell carcinoma: a meta analysis. Clin Genitourin Cancer 2009; 7: 11–19. 13. George S, Blay JY, Casali PG et al. Clinical evulation of continuous daily dosing sunitinib malate in patients with advanced gastrointestinal stromal tumour after imatinib failure. Eur J Cancer 2009; 45: 1959–1968. 14. Lee WJ, Lee JL, Chang SE et al. Cutaneous adverse effects in patients treated with the multitargeted kinase inhibitors sorafenib and sunitinib. Br J Dermatol 2009; 161: 1045–1051. 15. Lacouture ME, Wu S, Robert C et al. Evolving strategies for the management of hand-foot skin reaction associated with the multitargeted kinase inhibitors sorafenib and sunitinib. Oncologist 2008; 13: 1001–1011. 16. Chu D, Lacouture ME, Fillos T, Wu S. Risk of hand-foot skin reaction with sorafenib: a systematic review and meta-analysis. Acta Oncol 2008; 47: 176–186. 17. Llovet JM, Ricci S, Mazzaferro V et al. Sorafenib in advanced hepatocellular carcinoma. N Engl J Med 2008; 359: 378–390. 18. Cheng AL, Kang YK, Chen Z et al. Efficacy and safety of sorafenib in patients in the Asia-Pacific region with advanced hepatocellular carcinoma: a phase III randomised, double-blind, placebo-controlled trial. Lancet Oncol 2009; 10: 25–34. 19. Azad NS, Aragon-Ching JB, Dahut WL et al. Hand-foot skin reaction increases with cumulative sorafenib dose and with combination anti-vascular endothelial growth factor therapy. Clin Cancer Res 2009; 15: 1411–1416. 20. Erber R, Thurnher A, Katsen AD et al. Combined inhibition of VEGF and PDGF signaling enforces tumor vessel regression by interfering with pericyte-mediated endothelial cell survival mechanisms. FASEB J 2004; 18: 338–340. 21. Yang CH, Lin WC, Chuang CK et al. Hand-foot skin reaction in patients treated with sorafenib: a clinicopathological study of cutaneous manifestations due to multitargeted kinase inhibitor therapy. Br J Dermatol 2008; 158: 592–596. 22. Buchler T, Klapka R, Melichar B et al. Sunitinib followed by sorafenib or vice versa for metastatic renal cell carcinoma–data from the Czech registry. Ann Oncol2012; 23: 395–401. 23. Dušek L, Mužík J, Gelnarová E et al. Cancer incidence and mortality in the Czech Republic. Klin Onkol 2010; 23: 311–324.

and the overall survival (OS) of 153 WR DLBCL patients with survival information were 69.5% and 77.8%, respectively. The GCB immunophenotype correlated with a better OS (P = 0.0015), while BCL2 expression predicted a worse OS (P = 0.037), an effect overcome by the GCB/non-GCB classification. Compared with matched nodal DLBCLs, WR DLBCLs with no age-adjusted international prognostic index factor disclosed a better 5-year PFS rate (77.5% versus 70.7%; P = 0.03).
Conclusions: WR DLBCLs display distinct clinicopathologic features compared with conventional DLBCLs, with usual localized-stage disease, common follicular features and a high frequency of GCB immunophenotype contrasting with a low rate of BCL2 rearrangements. In addition, they seem to be associated with a better outcome than their nodal counterpart. Key words: clinicopathologic features, diffuse large B-cell lymphomas, outcome, Waldeyer's ring introduction About one-third of diffuse large B-cell lymphomas (DLBCLs) manifest primarily in non-nodal locations [1]. In the recent WHO classification, subsets of DLBCLs arising in peculiar extranodal sites have been individualized as distinct disease subgroups ( primary DLBCLs of the central nervous system, primary cutaneous DLBCLs, leg-type) or as distinct disease entities ( primary mediastinal large B-cell lymphoma), based on specific clinical and/or pathologic features [2][3][4]. Yet, the majority of extranodal DLBCLs remain categorized as DLBCLs, not otherwise specified (NOS).
Waldeyer's ring (WR) represents one of the most common extranodal sites for DLBCL development [5], comprising up to 20% of the cases [6]. Peculiar features have been suggested for WR DLBCLs. Clinically, a relationship to gastrointestinal tract involvement has been mentioned, either concurrent with diagnosis or at subsequent relapse [7][8][9]. Pathologically, the presence of focal follicular features has been reported in a subset of tonsillar DLBCLs, suggesting a morphologic subgroup distinct from de novo nodal DLBCLs, possibly corresponding to either transformed follicular lymphomas, or transformed marginal zone lymphomas with follicular colonization [10]. Genetically, translocations affecting the IRF4 locus in 6p25 have recently observed in a subtype of germinal center-derived B-cell lymphomas frequently manifesting in the cervical region including the WR and particularly affecting children and young adults [11]. Several clinical comparative studies of nodal versus WR DLBCLs have variably suggested differences in response to therapy and outcome, or similar features [9,[12][13][14]. However, published series are based on rather limited cohorts of patients, oftentimes heterogeneous with respect to histopathology, and rarely correlated clinical and pathologic features [9,[14][15][16][17][18][19][20].
The purpose of this work was to study the pathologic features of primary WR DLBCLs, and to compare their clinical outcome with that of nodal DLBCLs, based on the analysis of a retrospective series of well-characterized patients with a long follow-up treated in GELA protocols.

patient selection WR DLBCL patients
The GELA files were searched for patients diagnosed with primary DLBCLs of the WR (coded as palatine tonsil, lingual tonsil or nasopharynx). Of 208 with pathologic documentation of a WR DLBCL consecutively enrolled into the protocols LNH93, LNH98, LNH01 and LNH03 between 1993 and 2003, 187 qualified for primary WR DLBCLs. Outcome analysis and clinicopathologic correlations were restricted to a subset of 153 patients with available updated survival information, corresponding to the protocols LNH 93-1 to 93-5, LNH 98-2 and LNH 98-5 (summarized in supplementary Table S1, available at Annals of Oncology online) [6,[21][22][23][24][25]. All patients had received anthracyclin-based polychemotherapy without immunotherapy, except for two patients treated with R-CHOP in the LNH 98.5 trial.

matching with nodal DLBCL patients
For the comparison of outcome, WR DLBCL patients with clinical annotations were matched to DLBCL patients presenting with nodal involvement, on the basis of sex, age (<60 versus >60 years), age-adjusted international prognostic index (aaIPI) and randomization arm, leading to 135 pairs. Comparison of outcome was further restricted to a subset of 107 pairs of patients with no aaIPI factor.

pathologic studies
The archival slides that had been centrally reviewed at the time of enrollment into clinical trials, comprising 74 surgical specimens (large biopsies or excisions) (40%) and 113 sampling biopsies (60%), were reexamined by 2 hematopathologists (LdeL and PG) for cytologic characterization, according to the variants described in the WHO classification [2], and for the assessment of the pattern of growth. All cases by definition exhibited at least partially a diffuse pattern, and were categorized as: purely diffuse (D), predominantly diffuse (D > N) or predominantly nodular (N ≥ D). A tissue microarray (TMA) comprising two 0.6 mm cores representative of 59 cases with available tissue blocks was constructed in duplicate.

FISH analysis
A subset of cases were studied by interphase FISH on TMA sections, using split-signal DNA probes targeting MYC/8q24, BCL2/18q21 and BCL6/ 3q27genes ( probes Y5410, Y5407, Y5408; Dako, SA, Glostrup, Denmark) according to the manufacturer's recommendations (www.euro-fish.org) and analyzed and scored as recently reported [28]. For the detection of breakpoints affecting the IRF4 locus, two different break-apart assays using differently labeled BAC/PAC clones were carried out [11].

statistical analyses
Results were expressed as means ± standard deviations, medians and ranges for continuous variables and as proportions for categorical variables.
Associations between pathologic variables were tested using the chi-square test. End points of interest were complete response (CR), progression-free survival (PFS) (defined as the time interval between randomization to primary treatment failure, relapse and death from any cause or last followup) and overall survival (OS) (defined as the time interval between randomization to last follow-up or death from any cause). The OS and PFS were analyzed using a Cox regression model taking matching into account. The OS and PFS were displayed by Kaplan-Meier curves in each group. Within the WR DLBCL group, the OS and PFS were evaluated using univariate and multivariate Cox regression models. Results were considered significant at the 5% level (P < 0.05). Statistical analyses were carried out using the SAS (version 9.1 for Windows) statistical package.

clinical presentation of WR DLBCL patients
The 187 patients with primary WR DLBCLs comprised 169 patients with stage I-II disease, and 18 stage IV patients due to extensive local disease (n = 7) or dissemination limited to the bone marrow (n = 11). The population of patients with clinical follow-up comprised 106 men and 47 women at a median age of 57 years. Of these 153 patients, 141 had stage I or II disease and 12 patients (8%) had stage IV disease, due to locally invasive disease (n = 5) or bone marrow involvement (n = 7). The aaIPI, available for 144 patients, was 0 in 114 patients (79%), 1 in 27 patients (19%) and 2 in 3 patients (2%).

infrequent BCL2 breaks in WR DLBCLs
Evaluable FISH results of at least two loci were obtained for 42 cases sampled in the TMA (Table 2). Only three cases of 42 (7%) harbored a BCL2 rearrangement, one as the sole chromosomal break, and one each in association with BCL6 or MYC rearrangement. All three were positive for BCL2 by immunohistochemistry, two had a GCB immunophenotype and one was non-GCB. Nine cases of 36 assessable (25%) (5 GCB and 4 non-GCB) harbored a BCL6 rearrangement, of which 3 were positive for BCL6 expression. An MYC    The clinical features at the time of diagnosis were tested for their impact on outcome and response to treatment (Table 3). Age >60 years and higher aaIPI correlated with a shorter PFS and OS. The CR rate and the 5-year PFS were both adversely affected by a higher disease stage and a higher aaIPI.
Regarding associations with pathologic features (Table 1), no correlation was found with cytologic features, nor with a growth pattern. Expression of CD10 and a GCB-like immunophenotype strongly correlated with a better PFS (P = 0.0055 and P = 0.0003, respectively) and OS (P = 0.014 and P = 0.001, respectively) ( Figure 2). MUM-1 expression adversely affected PFS (P = 0.0046), and BCL6 expression was associated with a better OS (P = 0.048). BCL2 expression was predictive of shorter PFS (P = 0.017). These correlations remained significant irrespective of the aaIPI. However, BCL2 expression did not retain its adverse effect in a bivariate model, considering the GCB versus non-GCB profile together with BCL2 expression, while the good prognostic value of the GCB immunophenotype remained significant for both the PFS (P = 0.014) and the OS (P = 0. 047).

discussion
In this report, we have investigated the clinical and pathologic characteristics of primary WR DLBCLs, with the aim of delineating the intrinsic features of DLBCLs in this anatomic site. Our study based on a cohort of 187 patients with primary WR DLBCLs, including 153 cases with updated survival information, represents the largest series of primary WR DLBCLs reported so far [9,20,29].
Our analysis has shed light on peculiar pathologic features of WR DLBCLs, somewhat different from those usually reported for DLBCLs, NOS. First, we show that WR DLBCLs comprise a majority of tumors (60%) with a GCB-like phenotypic profile. Most unselected series of nodal and extranodal DLBCLs categorized according to the Hans algorithm report a substantially higher proportion of lymphomas with a non-GCB-like immunophenotype [27,28,[30][31][32]. In a way similar to us, although their series was rather small, Lopez-Guillermo et al. [13] also found a higher prevalence of tumors with a GCB phenotype among WR DLBCLs than in nodal cases [9 of 12 cases (75%) versus 36 of 77 cases (47%)]. Second, the frequency and distribution of genetic alterations explored by FISH differ from those reported in unselected series in respect of a low frequency of BCL2 rearrangements (7%). This finding is particularly paradoxical since this alteration, reported in 20%-30% of DLBCLs, is strongly associated with a GCB immunophenotype [28,33]. Conversely, the frequency of BCL6 and MYC rearrangements in 25% and 10% of cases, respectively, is similar to those usually reported [34]. Finally, DLBCLs of the WR frequently exhibit, to variable extent, a follicular growth pattern, a feature better seen in surgical specimens. The presence of follicular features in tonsillar DLBCLs has been reported by Ree et al. [10], although at a lower frequency. The significance of the follicular pattern, however, remains unclear. In about half of the cases, a minor nodular pattern with poorly delineated follicles was suggestive of follicular colonization by DLBCLs in a site naturally containing many hyperplastic follicles. In such cases, the possibility of transformed marginal zone lymphoma with follicular colonization was considered but could not be confirmed given the absence of a small cell component of these tumors, and the absence of a true marginal zone pattern. In the other half of cases, the presence of sharply delineated neoplastic follicles was rather suggestive of a truly follicular pattern. However, these cases appear to be heterogeneous at the genetic level with only a minority (two of eight cases with contributive FISH analysis) harboring a BCL2 rearrangement of conventional follicular lymphoma, while conversely a subset harbor a BCL6 break (two of eight cases), or an IRF4 rearrangement (one of eight cases). Interestingly, BCL6 rearrangements typically uncommon in grades 1-3A follicular lymphoma (FL) are detected in up to one-third of grade 3B cases, especially those associated with DLBCLs [35][36][37][38], and translocations activating IRF4 identify a subtype of germinal center-derived B-cell lymphoma affecting predominantly children and young adults [11]. Thus, at least a small proportion of WR DLBCLs likely arise in association with grade 3B FLs harboring a BCL6 or IRF4 rearrangement. It is also noteworthy that the palatine tonsils and nasopharynx are among the most common sites of pediatric FL [39]. These, in contrast to the common adult type of FL, usually lack BCL2 gene rearrangements, are grade 3B, usually present with localized disease and have an excellent prognosis. Thus, the possibility that a subset of transformed FL in the WR may be related to transformed FL is questionable. DLBCLs of the WR represent a group of extranodal DLBCLs associated with a favorable outcome with an estimated 5-year OS of 80% and a PFS of 72% in this retrospective cohort of 153 subjects treated with polychemotherapy alone before rituximab era (except for two patients). Among the clinical and pathologic factors tested for their association with survival, age >60 years, the non-GCB immunophenotype and positivity for BCL2 were identified as adverse prognostic factors.
Strikingly, the GC and non-GC immunophenotypic profile intended to reflect the cell-of-origin classification had a highly significant impact on both the PFS and the OS. Independent studies have established that, when defined by gene expression patterns, the cell-of-origin classification into GC and non-GC subgroups is a strong predictor of survival in patients treated by chemotherapy, as well as immunochemotherapy [40][41][42][43][44][45][46]. Conversely, the predictive value of cell-of-origin phenotyping on paraffin-embedded tissues by the combined immunostaining of CD10, BCL6 and MUM-1, using Hans algorithm [27], is more controversial [28,[30][31][32][47][48][49]. Our study shows the strong prognostic value of the GC/non-GC classification in a retrospective multicentric series composed exclusively of extranodal DLBCLs in a single anatomic region ( Figure 2). Although the actual correspondence of the phenotypic and transcriptomic categorizations cannot be assessed, the observed correlation between the GC phenotype and a CB morphology, on the one hand, and between a non-GC phenotype and CD44 and BCL2 expression, on the other hand, lends support to the reliability of the immunophenotypic classification.
Reports in the literature on BCL2 expression as a prognostic biomarker have been contradictory; although multiple largescale trials have established an association between BCL2 expression and a decreased PFS or OS in DLBCLs [26,30,50,51], recent studies have suggested that the prognostic value of BCL2 expression is overcome by the use of rituximab [52] and may be restricted to non-GCB tumors only [53]. In this clinically homogeneous cohort of patients treated before the era of rituximab, we confirm the results previously reported by our group on the prognostic value of BCL2expression on both the PFS and the OS, irrespective of the aaIPI [26]. However, in bivariate analysis, the adverse effect of BCL2 expression was overcome by the GCB/non-GCB immunophenotype.
In our study, WR DLBCLs appear to carry a better outcome compared with matched nodal DLBCLs, a difference that reached significance for the PFS of patients with localized disease and no adverse factor of the aaIPI in this retrospective series of patients treated with chemotherapy. Whether immunochemotherapy protocols would alleviate this difference and/or could still improve the outcome of patients with primary WR remains to be investigated. Nevertheless, the clinical characteristics of WR DLBCLs, together with distinctive pathologic features, reinforce the concept that the anatomic location is a major determinant of DLBCL heterogeneity, and might suggest that WR DLBCLs could constitute a biologically distinct DLBCL subgroup.
acknowledgements LdeL, CB, GF and PG designed research, carried out research, analyzed data and wrote the paper. CC-B and RS carried out research and analyzed data. LS analyzed data, MB carried out research, and JB, TJM, BF, TP, JB, CG, CH, HT collected and analyzed data. The authors wish to thank Marion Fournier from GELARC for data retrieval; Emmanuelle Come and Marie-Laure Prunet from GELA-P for realization of the immunohistochemical techniques; and Reina Zühlke-Jenisch and Magret Ratjen for expert technical assistance.