Gascoyne, Randy D.

Abstract Abstract 1545 Introduction: The diagnosis of peripheral T-cell lymphoma (PTCL) is difficult and accurate diagnosis and subclassification relies on correlating histologic, immunophenotypic, molecular genetic, and clinical data. Evidence-based guidelines for the appropriate diagnostic work-up of PTCL are lacking. The objective of this study was to evaluate a large series of PTCLs by experienced hematopathologists with a tiered approach to immunohistochemistry (IHC) and molecular genetic characterization to document overall diagnostic accuracy and clinical relevance using this approach. Methods: 7 experienced hematopathologists from 5 institutions reviewed 374 cases of peripheral T and NK cell lymphomas (referred to collectively as PTCL). 6 cases of cutaneous T-cell lymphoma were excluded after final review in addition to 29 non-PTCL cases submitted as control cases to mimic diagnostic practice during the review. Cases received tier 0, 1 and 2 diagnoses by 3 independent pathologists, based on review of hematoxylin and eosin (HE) stain with basic demographic data, panel 1 IHC (CD3, CD5, CD10, CD20, CD21, CD30, CD45, PAX5), and panel 2 IHC (CD2, CD4, CD7, CD8, CD23, PD1, CD56, EBER, ALK, TIA1, TCRg, TCRbF1), respectively. A tier 2b diagnosis was then rendered after gene rearrangement data were available. A final consensus diagnosis was rendered after discussion of each case by the 3 reviewers with all available clinical data. Overall survival (OS) was assessed using Kaplan-Meier (KM) curves and Cox proportional hazards models. Results: 1122 individual diagnoses leading to 339 final consensus PTCL diagnoses were rendered. 341 (91%) cases had complete phenotypic data and 241 had gene rearrangement data. There was no bias of missing data according to final diagnosis subtype. Reviewer diagnoses using specific WHO subclassification were 16.3%, 36.9%, 82.7%, and 85.9% for tier 0, 1, 2, and 2b, demonstrating a significant increase in diagnostic certainty after a complete IHC panel. Gene rearrangement only contributed to a change in diagnosis in 51/650 (8%) individual reviews. Across all 374 cases, a small number of cases (n=28, 7.5%) showed no agreement among the 3 independent reviewers after tier 2b and required debate. These generally represented refinement of the subclassification of a PTCL. The most common disagreements were between PTCL, not otherwise specified (nos) vs. unclassifiable T-cell lymphoma, and PTCL, nos vs. angioimmunoblastic T-cell lymphoma (AITL). Currently, OS data was available in 198 cases and 52% have died (median age follow-up of 15 months for those still alive, range 0–160 mo). Figure 1 shows the KM survival curves for types with more than 10 cases. Of note, unclassifiable PTCL cases had poor OS, comparable to PTCL nos; a trend was seen for CD30+ ALK- ALCL to have a longer OS compared to PTCL nos (HR=0.5, 95% CI: 0.21–1.19, P=.09). EBER positivity in tumor cells was associated with poor OS in the cohort of all PTCL patients (HR=2.32, 95% CI: 1.28–4.23, p=0.006, Figure 2); the association was similar when nasal NK/T-cell lymphoma cases were excluded (HR=2.50, 95% CI: 1.00–6.26, p=0.05). Trends for poor prognosis were also seen for TIA-1 expression in PTCL, nos (HR=1.9, 95% CI 0.91–3.96, P=.09) and PD1 expression in AITL (HR=6.25, 95% CI 0.85–46.09, P=.07). Clinical data collection is still ongoing and will be updated. Conclusions: We demonstrate the diagnostic accuracy among experienced hematopathologists of a defined IHC panel, showing an overall ability to reach consensus diagnosis of 93% in PTCL cases. The resulting consensus diagnostic subtypes showed expected outcomes relative to other large series of PTCLs, and EBER positivity is a poor prognostic marker among T and NK cell lymphomas. A tiered approach to IHC is recommended when a PTCL is under differential diagnostic consideration since the first tier can often resolve the main question of whether a lesion is reactive or lymphoma. This can lead to more efficient use of the expanded tier 2 panel that will enable diagnosis and specific subclassification of PTCL. Gene rearrangement studies are not required in the great majority of cases. This evidence-based approach to the diagnosis of PTCL should inform practicing pathologists, clinical trial design, and policy makers regarding required ancillary studies in this group of diseases. Disclosures: Hsi: Allos: Research Funding. Said:Allos: Research Funding. Macon:Allos: Research Funding. Rodig:Allos: Research Funding. Gascoyne:Seattle Genetics: Research Funding. Ondrejka:Allos: Research Funding. Dorfman:Allos: Research Funding. Maurer:Allos: Research Funding. Dogan:Allos: Research Funding.

Abstract Background The international prognostic score (IPS) (Hasenclever et al., NEJM 1988) uses 7 factors (age> 45, male sex, hemoglobin<10.5, stage IV, leukocytosis > 15,000, and lymphopenia< 600) to predict a 5 year freedom from progression (FFP) of 42%-84% and overall survival (OS) of 56%-89% for patients with advanced HL. Constructed from a retrospective analysis of patients treated before 1992, the IPS continues to be the most commonly used risk stratification index for advanced HL. Recent studies suggest that the predictive range of the IPS has narrowed due to improved outcomes of patients treated with current therapy (Moccia et al. JCO 2012). In this report we prospectively evaluated the ability of the individual components of the IPS to predict outcome in patients enrolled on the US Intergroup trial E2496. Methods All seven IPS (IPS-7) variables were recorded for all patients at study entry. FFP was defined as the time from study entry to disease progression or relapse; deaths that occurred during remission that were not preceded by disease progression/relapse were censored. OS was defined as the time from study entry to death from any cause. Kaplan-Meier methodology was used to construct survival curves. Univariate and multivariate analysis (MVA) was performed using Cox proportional-hazards models. We subsequently constructed an alternative prognostic score utilizing the factors which were significant on MVA (PS-3). Results From 1996-2006, 854 patients with advanced HL, were randomized to treatment with either ABVD or Stanford V, with no significant differences in outcome (Gordon et al, JCO 2013). While the IPS-7 remained prognostic it did not stratify the lowest risk patients (0-1 risk factor) or patients with 3-5 risk factors, as its predictive range was narrowed due to improved clinical outcomes (Fig 1a and 2a). Table 1 shows the univariate and multivariate analysis for IPS-7 and outcomes. In contrast to the original IPS-7, on MVA, only two factors, hemoglobin and stage were significant for FFP, and three factors for OS: hemoglobin, stage, and age. We then evaluated a new 3 factor score (PS-3) utilizing variables significant on the MVA. The PS-3 was significant for both FFP (p=0.0001) and OS (p<.0001) and clearly separated patients into 4 distinct risk groups with either: 0, 1, 2 or 3 risk factors. The five year FFP was 83%, 74%, 68%, and 63% and OS was 95%, 85%, 75%, and 52% for patients with 0, 1, 2, and 3 risk factors, respectively (Table 2, Fig 1b and 2b). Conclusion In E2496, for patients with advanced HL, the PS-3 may stratify patients more clearly compared to IPS-7 and provide a simpler model to assess risk. Studies comparing PS-3 with IPS-7, and are ongoing. The incorporation of novel biomarkers and biologic factors into PS-3, as well as validation with other data sets is warranted. Disclosures: No relevant conflicts of interest to declare.

Abstract Background: Follicular lymphoma (FL) is the second most common nodal lymphoma worldwide and remains incurable for most patients (pts). FL is recognized to be a highly heterogeneous disease and a subset of pts experience remarkable poor outcome. In a recent study, 19-26% of pts receiving first-line R-CHOP experienced progression of disease (POD) within 24 months after diagnosis and had a 5-year (yr) overall survival (OS) of only 34-50%, as compared to a 5-yr OS of 90-94% for pts without POD within 24 months (Casulo et al., J Clin Oncol 2015). Consequently, event-free survival at 12 (EFS12) and 24 months (EFS24) have been suggested as novel surrogate endpoints for poor overall survival in clinical trials and useful for patient counseling (Maurer et al., ASH 2014). We have previously shown that a clinicogenetic risk model (m7-FLIPI) that includes the mutation status of seven genes (EZH2, ARID1A, MEF2B, EP300, FOXO1, CREBBP, and CARD11), the FL International Prognostic Index (FLIPI), and the Eastern Cooperative Oncology Group (ECOG) performance status, improves risk stratification for failure-free survival (FFS) and OS in pts with FL receiving first-line immunochemotherapy (Pastore et al., Lancet Oncology 2015). An online tool is available at: http://www.glsg.de/m7-flipi. It would be advantageous to determine prognosis prior to front-line therapy rather than after POD. Thus, we aimed to investigate the predictive utility of the m7-FLIPI for early POD. Patients and Methods: Clinical and mutation data were available from two independent cohorts: 151 pts who received R-CHOP and IFN maintenance as part of a randomized trial of the German Low-Grade Lymphoma Study Group (GLSG), and 107 pts from a population-based registry of the British Columbia Cancer Agency (BCCA) who received R-CVP. Among the latter, 93 (87%) received R-maintenance by intention to treat. All pts had symptomatic, advanced stage or bulky disease considered ineligible for curative radiotherapy, and a biopsy specimen obtained </=12 months prior to the initiation of first-line therapy. POD was defined as progression, relapse, or death due to any cause. POD12 and POD24 were defined based on POD status at 12 and 24 months after therapy initiation, respectively. Logistic regression analysis was performed to assess if m7-FLIPI was predictive of early POD using the statistical software R (version 3.1.2). Results: In the GLSG cohort, median age was 57 yrs (range 27-77), 51% had high-risk FLIPI, and 5-yr failure-free survival (FFS) and OS rates were 66% and 83%, respectively (median follow-up for OS 7.7 yrs). In the BCCA cohort, median age was higher (62 yrs, 37-83), but high-risk FLIPI was similarly frequent (50%); 5-yr FFS and OS rates were 58% and 74%, respectively (median follow-up for OS 6.7 yrs). A total of 43 GLSG (28%) and 24 BCCA pts (22%) were classified as high-risk by m7-FLIPI, with a 5-yr OS of 65% and 42%, respectively. In the GLSG cohort, 8 (5%) and 29 pts (19%) had POD within 12 and 24 months, respectively, 4 and 10 pts were censored before POD12 and POD24 (table). High-risk m7-FLIPI pts were significantly more likely to experience early POD with an Odds Ratio (OR) of 20.80 (95% confidence interval (CI) 3.53-395.96; p=0.0052) for POD12 and 6.33 (95% CI 2.67-15.69; p<0.0001) for POD24. Likewise, in the BCCA cohort, 16 (15%) and 28 pts (26%) experienced POD12 and POD24, respectively (table). The OR for high-risk m7-FLIPI pts was 4.69 (95% CI 1.52-14.65; p=0.0068) for POD12, and 5.36 (95% CI 2.03-14.60; p=0.0008) for POD24. In the GLSG and BCCA cohorts, the m7-FLIPI had a sensitivity of 88% and 50%, a specificity of 75% and 82%, a positive predictive value (PPV) of 16 and 33%, and a negative predictive value (NPV) of 96% and 90% to predict POD12. For POD24, sensitivity was 62% and 46%, specificity 79% and 86%, PPV 42% and 54%, and NPV 82% and 82%, respectively. Conclusion: We conclude that patients classified as high-risk m7-FLIPI are highly enriched among those with early progression of disease. However, approximately one-half of patients with progression by 24 months after therapy initiation are classified as low-risk m7-FLIPI. Thus, further efforts are needed to refine the m7-FLIPI and thereby capture additional cases that may benefit from innovative first-line approaches. Table. Population m7-FLIPI N POD12 POD24 GLSG Total 151 8/147* 29/141* High-risk 43 7 18 Low-risk 108 1 11 BCCA Total 107 16/107 28/107 High-risk 24 8 13 Low-risk 83 8 15 *Pts censored were removed Disclosures Ansell: Bristol-Myers Squibb: Research Funding; Celldex: Research Funding.

Abstract Abstract 3705 Improved outcomes were reported in USA studies in DLBCL (Habermann et al. J Clin Oncol 2006) in a two-staged randomized study of R-CHOP (rituximab, cyclophosphamide, Adriamycin, vincristine, and prednisone) versus CHOP with a second randomization assignment to maintenance rituximab (MR) or observation (OBS) (E4494) and in follicular lymphoma (FL) in patients in a trial with CVP (cyclophosphamide, vincristine, and prednisone) followed by randomization to MR versus OBS in advanced-stage indolent lymphoma (E1496) (Hochster et al. J Clin Oncol 2009). The details of the treatment regimens have been previously described. The German High Grade Lymphoma Study Group has reported differences between men and women in outcomes in DLBCL with improved outcomes in women treated with rituximab (Müller et al. Blood 2012). Methods: Survival outcomes by sex were analyzed using a stratified weighted cox regression, removing the effect of maintenance rituximab in DLBCL (E4494), and a log-rank test in patients with follicular histology (E1496). A Cox regression model was used to evaluate the differences between males and females together with weight (below or above the median). Wilcoxon test and Fisher's exact test were used to compare medians and proportions, respectively. Results: In DLBCL, there were 273 eligible males and 273 females. The median age was 69 in males and 70 in females. There were no differences in baseline patient characteristics among the male and female populations. 138 males and 129 females were treated with R-CHOP. There was no difference in treatment received, with 83% females and 75.3% males receiving 6 cycles or more of R-CHOP treatment (p=0.14). The complete remission (CR) and partial remission (PR) rates after RCHOP were higher in females (82.6%) versus males (71.5%) (P = 0.04). With a median follow-up of 9.45 years, the failure free survival (FFS) and overall survival (OS) were improved in females (P = 0.02, 0.002; HR=0.63, 0.54) compared with males in the R-CHOP group. The outcomes were not different in the CHOP group for FFS (P = 0.81) or OS (P= 0.57) between males and females. Within the R-CHOP group, the failure free survival (FFS) was significantly different between women and men (P = 0.002) and body weight (P = 0.03), but only female sex (P = 0.001) was significant and not weight (P = 0.26) for OS. Of 282 evaluable patients with advanced-stage follicular lymphoma who received initial treatment with CVP, 115 patients were randomly assigned to MR and 113 to OBS. 120 patients were male, and 108 were female. The median age was 58 years in the MR arm and 54 years in the OBS arm. At a median follow-up of 8.04 years, MR markedly improved the PFS (P=0.003) compared with OBS. There were no differences in PFS (P= 0.17) between males and females treated with MR or OS (P = 0.27). In conclusion, in induction therapy with rituximab in DLBCL there is a sex dependent effect with rituximab with males benefiting less than females. In contrast, in patients with follicular lymphoma treated with chemotherapy followed by maintenance rituximab, there were no differences. The differences in outcomes in patients of different sex with DLBCL treated with immunochemotherapy and FL treated with maintenance rituximab will require further analysis of multiple clinical and biologic variables. Disclosures: Fisher: Roche: Advisory Board Other. Cheson:Genentech: Consultancy. Kahl:Genentech, Roche: Consultancy, Research Funding. Horning:Genetech, Roche: Employment, stock (Roche) Other.

Abstract Abstract 687 Background: A persistently positive PET scan after just a few cycles of therapy is predictive of a poor clinical outcome in diffuse large B-cell lymphoma (DLBCL). Prior data suggested that about one third of patients remain PET positive after 2–4 cycles, with ≤ 20% 2 year progression-free survival (PFS) for such patients. A response-adapted strategy was studied in E3404 to test the hypothesis that an early treatment change to four cycles of R-ICE might reduce the treatment failure rate of patients whose PET scan remained positive after an initial 3 cycles of R-CHOP. Study Design: Previously untreated patients with DLBCL stage III, IV, or bulky II, with measurable disease, HIV negative, and with adequate organ reserve, were eligible. PET/CT scan was performed before treatment and again after 3 cycles of R-CHOP. A fourth cycle of R-CHOP was given while the scan was centrally reviewed and scored as either positive or negative for FDG-avid tumor by a single reviewer using criteria based on modifications of the Harmonization Criteria. Persistently PET-positive patients received 4 cycles of R-ICE, while PET-negative patients received 2 more cycles of R-CHOP, for a total of 6 cycles. PET/CT was performed again at the end of treatment. A ≥45% 2 year PFS for mid-treatment PET- positive patients was viewed as a promising result, with 88% power if 33 such patients were accrued. Total accrual of 99 patients was therefore planned. Results: Of 100 patients accrued, 78 were eligible; all but 1 ineligibility was based on central pathology review. Fifty-eight were male; median age was 62 years (20–74); 13% IPI 0–1, 31% IPI 2, 37% IPI 3, and 19% IPI 4–5. Seventy-four of 78 (95%) patients completed the first 3 cycles of R-CHOP. Of 72 patients continuing treatment, 67 (93%) completed protocol treatment: 83% of mid-treatment PET-positive and 93% of PET-negative patients. Of 74 patients undergoing mid-treatment PET scan, 12 (16%) were scored as positive, and 62 (84%) as negative. At the end of treatment, 13% were scored as positive, and 87% as negative. Two-year PFS, from the time of study entry, was 72% among all eligible patients. Two year PFS from the time of mid-treatment PET scan was 45% (90% CI, 21–67%) for patients scored as mid-treatment PET-positive, and 77% (90% CI 67–85%) for patients scored as mid-treatment PET-negative. The 80% confidence interval (corresponding to a test with a one-sided type I error of 10%) did not include the null hypothesis of 25%, implying that the results met the pre-specified threshold. Among all 78 eligible patients, 10 have died (13%); 5-year overall survival was 87% (90% CI 78–92%) Three-year overall survival was 67% (90% CI 40–84%) for mid-treatment PET-positive patients, and 93% (90% CI 85–97%) for mid-treatment PET-negative patients. Conclusions: The 2-year PFS for mid-treatment PET-positive patients met the threshold to be considered promising, but the confidence interval was wider than expected due to the small patient number. In addition, the inter-observer variability in the interpretation of mid-treatment PET scans as a binary variable in this study (studied and reported separately in Horning SJ et al. Blood 2010) implies that treatment modification based on early PET scanning should remain confined to clinical trials. Disclosures: Swinnen: Celgene: Consultancy; Genentech: Research Funding. Quon:Genentech: Research Funding. Advani:Pharmacyclics: Research Funding. Kahl:Roche/Genentech: Consultancy; Celgene: Consultancy; Cell Therapeutics: Consultancy; Janssen: Consultancy; GSK: Consultancy; Gilead: Consultancy. Horning:Genentech: Employment; Roche: stock, stock Other.

6511 Background: Conventional adverse event (AE) analysis (ToxC) focuses on incidence of grade (gr) 3+ toxicities, and fails to capture AE time profile. Novel metrics that reflect chronic low gr and overall AE burden are needed. We applied the Toxicity over Time (ToxT) approach to ECOG-ACRIN 2408 to depict time-dependent toxicity of lenalidomide (L) with rituximab maintenance (MR) in follicular lymphoma (FL), and we developed a novel summary metric of symptomatic AE burden, the maximum gr over time (MGOT). Methods: In E2408, high risk FL patients (pts) were randomized (1:2:2) to: A) bendamustine-rituxumab (BR) x 6 then MR x 2 years (yrs) vs B) BR-bortezomib x 6 then MR x 2 yrs vs C) BR x 6 then MR x 2 yrs + L x 1 yr (MRL). Analysis included 3 laboratory and 5 symptomatic AEs of highest incidence during maintenance on arms A and C. Treatment-related AEs of any gr were analyzed by ToxC and ToxT. Repeated measures, time-to-event (TTE) and area under the curve (AUC) analyses capture trends over time in ToxT. MGOT combines the 5 symptomatic AEs. Results: 104 randomized pts (30 MR, 74 MRL) were included. For the laboratory AEs, by ToxC, neutropenia incidence was significantly higher in MRL (84%) than MR (47%, p < .001). ToxT additionally shows neutropenia does not worsen over time (10/14/20% gr 1/2/3+ at c1, 6/21/12% gr 1/2/3+ at c12). For the symptomatic AEs, ToxC indicates 2% gr 3+ GI AEs. However, gr 1-2 GI AEs are more common on MRL (59%) than MR (26%, p < .001). ToxT AUC captures a higher burden of GI AEs over time on MRL(2.8) vs MR(1.4, p = .002). TTE depicts sooner GI AE onset in MRL (10% vs 0% gr 2+ GI by day 50, p = 0.03). Bar charts of incidence and grade by cycle illustrate this improves over time (34/7/4% gr 1/2/3+ at c1, 13/0/0% gr 1/2/3+ at c12). ToxT MGOT analyses demonstrate earlier time to gr 2+ symptomatic AE on MRL vs MR (63% vs 31% by day 50, p < .001) and suggest that overall AE burden over time is higher for patients on MRL(AUC 18.2) than MR(11.8, p < .001). Conclusions: ToxT depicts AE time profile and can guide AE interventions. Summary metrics suggest that symptomatic AEs occur earlier and their burden over time is higher on MRL. We are implementing ToxT in patient-reported AE data to better characterize pt tolerability.

Abstract Primary mediastinal large B-cell lymphoma (PMBCL) is recognized as a distinct entity in the World Health Organization classification. Currently, the diagnosis relies on consensus of histopathology, clinical variables, and presentation, giving rise to diagnostic inaccuracy in routine practice. Previous studies have demonstrated that PMBCL can be distinguished from subtypes of diffuse large B-cell lymphoma (DLBCL) based on gene expression signatures. However, requirement of fresh-frozen biopsy material has precluded the transfer of gene expression–based assays to the clinic. Here, we developed a robust and accurate molecular classification assay (Lymph3Cx) for the distinction of PMBCL from DLBCL subtypes based on gene expression measurements in formalin-fixed, paraffin-embedded tissue. A probabilistic model accounting for classification error, comprising 58 gene features, was trained on 68 cases of PMBCL and DLBCL. Performance of the model was subsequently evaluated in an independent validation cohort of 158 cases and showed high agreement of the Lymph3Cx molecular classification with the clinicopathological diagnosis of an expert panel (frank misclassification rate, 3.8%). Furthermore, we demonstrate reproducibility of the assay with 100% concordance of subtype assignments at 2 independent laboratories. Future studies will determine Lymph3Cx’s utility for routine diagnostic purposes and therapeutic decision making.