Cerhan, James R.

Abstract Background: Follicular lymphoma (FL) is a germinal center–derived B cell malignancy characterized by the clonal expansion of a dominant malignant B cell population. While prior studies have identified transcriptional subtypes across patients, most analyses have treated B cells within individual tumors as a transcriptionally uniform population. However, this assumption has not been systematically tested using high-resolution, clonotype-resolved transcriptomic data. The broader landscape of intratumoral B cells—including those not belonging to the dominant clone—remains poorly defined, and their roles in disease biology are not well understood. These unexpanded, intratumoral B (ITB) cells may include residual non-malignant populations, early-stage malignant cells, or immune-modulating cells with distinct roles in the tumor microenvironment. It is also unknown whether their presence varies across patients or correlates with clinical features or immune microenvironmental interactions. Methods: We performed integrated single-cell 5′ RNA sequencing and B cell receptor (BCR) repertoire sequencing on 7 newly diagnosed FL 1-3A fresh-frozen lymph node biopsies (Dasari et al., Blood Cancer Discov 2025). Using a high-confidence approach that incorporates heavy and light chain pairing, V and J gene usage, and CDR3 nucleotide junction identity, we classified B cells into two groups: (1) cells belonging to the dominant expanded clone (FL B cells) and (2) all other B cells with detectable and non-expanded BCRs (ITB). These clonal identities were then integrated with single-cell transcriptomic profiles using Seurat and Immcantation pipelines. Differential expression analysis was conducted using edgeR to compare FL B cells to ITB cells across all FL biopsies. Results: After quality filtering, a total of 44,637 single cells were retained across all 7 samples (minimum per sample: 3,847 cells), including malignant and non-malignant B cells as well as diverse tumor microenvironment populations. The proportion of B cells assigned to the dominant FL clone varied substantially across tumors, ranging from 22% to 89%, indicating that some tumors contain a substantial population of ITB cells. In one tumor, we identified two expanded malignant B cell clones comprising 40.6% and 34.4% of all cells. These clones exhibited non-overlapping V and J gene usage (IGHV3-11/IGHJ5 vs. IGHV3-72/IGHJ3), distinct CDR3 junctions, and different immunoglobulin isotypes (IGHG1 vs. IGM). BCL2 was expressed in most cells within both clones (Clone 1: 87.8%, Clone 2: 67.0%), with mean log-normalized expression levels of 1.79 and 0.95, respectively. Both clones formed distinct UMAP clusters and showed transcriptional divergence from ITBs, with 453 differentially expressed genes (DEGs) identified for Clone 1 and 98 for Clone 2 (adjusted p<0.01, |log₂FC|>1, min.pct≥0.25). Additionally, 748 DEGs were identified between the two clones. This evidence of clonal divergence within a single FL tumor reinforces the need to reconsider intratumoral homogeneity in FL and suggests that multiple malignant clones may sometimes coexist. When comparing FL clones to ITB cells across tumors we found that FL clones exhibited a transcriptional program enriched for proliferation and cell cycle regulators, including MKI67 (p<0.005), E2F1 (p<0.0001), CDKN2A (p<0.0001), and CDKN2B (p<0.01), as well as elevated expression of AICDA (p<0.01), EZH2 (p<0.0001), and PD-L1 (p<0.05). As expected, the malignant cells also had higher expression of BCL2 (p<0.01) and CD10/MME (p<0.0001). In contrast, ITB cells demonstrated relatively higher expression of SELL (p<0.0001) and TNFRSF14 (p<0.05), which were significantly downregulated in the FL clone. Conclusions: Together, these findings challenge the prevailing notion that FL tumors are defined by a single homogeneous malignant B cell population. Instead, our data support a model in which multiple transcriptionally and functionally distinct B cell populations coexist within tumors, including dominant clones and a diverse population of ITB cells. This is, to our knowledge, the first study to comprehensively characterize these populations using integrated, lineage-aware single-cell multiomic profiling across FL patients. These insights may ultimately inform new strategies for understanding tumor evolution, immune evasion, and therapeutic targeting in follicular lymphoma.

Abstract Background: Prognostication for patients with non-transformed relapsed/refractory (R/R) follicular lymphoma (FL) is not well defined. Progression of disease within 24 months (POD24) and number of lines of prior therapy are commonly used to benchmark outcomes and for trial eligibility in this space. However, a comprehensive model for risk prediction in R/R FL is lacking. Here we utilize patients from the LEO Consortium for Real World Evidence (CReWE) and Iowa/Mayo Clinic Lymphoma SPORE MER to develop a clinical prediction model for R/R FL. Methods: Patients with grade 1-3A FL were prospectively enrolled at diagnosis in the MER and/or initiated second line or later therapy (2L+) for non-transformed FL at one of the 8 LEO Centers from a previous LEO CReWE study (Casulo et al, Lancet Haem 2022). All available lines of therapy were abstracted; lines of therapy after histologic transformation were not used in the analyses. A management plan of observation was considered as a type of therapy in the 1L setting but not in the 2L+ setting. Clinical variables and outcomes were abstracted at each line of therapy. Variables considered for modeling included demographics, clinical labs, treatment history, and pertinent clinical variables (including FLIPI and FLIPI24 components). Multiple imputation was used to address missing data in modeling; beta-2 microglobulin (B2M) was unable to be imputed due to the amount of missing data. Index therapy was defined as the date of initiation of a line of therapy (2L, 3L or 4L) for non-transformed FL. The primary endpoint for model development was the time from index therapy until the histologic transformation or lymphoma-related death (HT/LD). Cox proportional hazards models were used for model development. Modeling was first performed separately for each line of therapy and then evaluated across lines. Bootstrapping on imputed datasets was used to assess optimism in model performance. Results: N=1001 patients initiated 2L+ therapy and were utilized for modeling. Of these patients, N=654 initiated 3L and N=344 initiated 4L therapy. Median age at index therapy was (60, 61, and 62) for lines 2-4, respectively. Median follow-up by line of therapy was 8.2, 6.1, and 6,1 years and number of HT/LD events for modeling was 272, 154, and 88, respectively. Observed outcomes decreased modestly as the line therapy increased: median event-free (1.6, 1.5, and 1.2 years) and overall survival (17.6 years, 14.5 years, and 12.3 years), for lines 2-4, respectively. 5-year estimates of HT/LD by line of therapy were 21.6% (95% CI: 18.9-24.3), 23.7% (95% CI: 19.9-27.4) and 29.3% (95% CI: 23.3-34.7), respectively. A clinical prediction model was independently developed for each line of therapy; the final model for each line consisted of the same variables: increasing age, male sex, higher LDH, lower hemoglobin, higher white blood cell count, prior immunochemotherapy exposure, and shorter time since initial diagnosis. The parameter estimates for these variables were broadly similar across the different lines of therapy. The model output is an individual patient's risk of HT/LD. Patients with high-risk FLP-R, as defined using the upper quartile of FLP-R risk scores for each line, had observed 5 year HT/LD rates of 43%, 44% and 51% in lines 2-4, respectively. In contrast, patients with the lower quartile of FLP-R risk scores had observed 5 year HT/LD rates of 10%, 6% and 8% in lines 2-4, respectively. The optimism-corrected c-statistic for the FLP-R model across all lines of therapy was 0.685. This added significant information beyond FLIPI diagnosis (c=0.533) FLIPI at index line (c=0.563), the line of therapy (c=0.525), POD24 status (c=0.578), or all 4 variables combined (c=0.614). Conclusions: The FLP-R model identifies a group of patients with non-transformed FL in the R/R setting who are at very high risk of transformation and/or lymphoma-related death across lines of therapy. FLP-R features 4 of the 5 components of the 1L FLIPI24 prediction model (B2M was unavailable for modeling) with the addition of sex, treatment history and time since initial diagnosis. The number of prior therapies is not sufficient to identify high-risk populations for clinical trial enrichment in the R/R FL setting and model discrimination for FLP-R was greatly improved over FLIPI and/or POD24 across lines of therapy. A Shiny app will be provided for model implementation; external validation is planned.

Abstract Introduction Short diagnosis-to-treatment interval (DTI) is associated with aggressive disease biology and inferior outcomes in diffuse large B-cell lymphoma (DLBCL), and is recognized as a potential source of selection bias in clinical trials if not properly addressed. However, its prognostic significance in follicular lymphoma (FL) remains poorly characterized. We aimed to evaluate the prognostic significance of DTI in newly diagnosed FL patients who received frontline (1L) immunochemotherapy (IC), using a large international dataset and an independent external validation cohort. Methods For the discovery dataset, we evaluated patients diagnosed between 2002 and 2018 from ten observational cohorts as part of the FLIPI24 Consortium. External validation was conducted using data from patients enrolled in the Lymphoma Epidemiology of Outcomes (LEO) Cohort between 2010 and 2015. All patients had FL grade 1–3A and received 1L IC within 100 days of diagnosis. Patients with a concurrent DLBCL were excluded. Missing baseline laboratory values were addressed via multiple imputation. The primary endpoint was event-free survival at 24 months (EFS24); secondary endpoint was 5-year overall survival (OS), both calculated from treatment initiation. DTI was analyzed as continuous (via splines) and dichotomized variable (≤14 vs >14 days; as defined in DLBCL). Results In the discovery cohort (n=3,907), median age was 61 years (IQR 53–69), and 50% were male; 49% of patients were high-risk according to FLIPI, 26% according to PRIMA-PI, and 38% according to FLIPI24. Included patients received R-CHOP (54%), R-CVP (26%), or B-R (20%) as 1L treatment, followed by anti-CD20 maintenance in 60%. The median DTI was 30 days (IQR 17–48). Using the 14-day cut-off, 851 patients (22%) had short DTI. These patients had more adverse baseline characteristics, including B symptoms (42% vs 32%, P<0.01), ECOG performance status ≥2 (15% vs 8%, P<0.01), elevated LDH (51% vs 40%, P<0.01), hemoglobin <12 g/dL (30% vs 17%, P<0.01), and elevated β2-microglobulin (56% vs 44%, P<0.01). Prognostic indices (PIs) were also more frequently high-risk in the short-DTI group: FLIPI (56% vs 47%), PRIMA-PI (36% vs 24%), and FLIPI24 (51% vs 34%, all P<0.01). Spline modeling showed a decreasing EFS24 rate with increasing DTI, reaching a plateau at 30–40 days. In logistic regression model, each additional week of DTI improved EFS24 (OR 0.92, P<0.01). DTI attenuated but remained prognostic after adjusting for PIs (OR 0.93 for FLIPI, 0.95 for PRIMA-PI, 0.96 for FLIPI24, all P<0.01). When analyzed as a dichotomized variable, short DTI was associated with inferior EFS24: OR 1.74 adjusted for FLIPI, 1.63 for PRIMA-PI, and 1.53 for FLIPI24 (all P<0.01). The 2-year EFS of short vs long DTI groups was 71% vs 82% (HR 1.50, 95% CI 1.34–1.67, P<0.01), 5-year OS was 80% vs 86% (HR 1.38, 95% CI 1.21–1.56, P<0.01), and the 5-year cumulative risk of histologic transformation was 7.7% vs. 5.4% (P<0.01). In the validation cohort (n=516), median age was 61 years (IQR 52–69 years), 55% were male; 36% of patients were high-risk according to FLIPI; 29% according to PRIMA-PI, and 33% according to FLIPI24. Patients received B-R (59%) or R-CHOP (41%) as 1L treatment. The median DTI was 31 days (IQR 19–48), 84 patients (16%) had DTI ≤14 days. Similar patterns were observed; patients with short DTI had adverse baseline characteristics, PIs, and shorter 2-year EFS (72% vs 84%, HR 1.69, 95% CI 1.13–2.55, P=0.01). DTI remained associated with EFS24 when analyzed as a continuous variable (per week: OR 0.88 unadjusted, 0.90 with FLIPI, 0.89 with PRIMA-PI, and 0.89 with FLIPI24), and dichotomized (≤14 vs >14 days: OR 1.83 adjusted for FLIPI, 1.88 for PRIMA-PI, and 1.73 for FLIPI24). Conclusion Short DTI, defined as ≤14 days (observed in 22% of FL patients vs 46% in DLBCL), is a strong and independent predictor of inferior outcomes in newly diagnosed FL patients treated with IC. While DTI correlates with adverse baseline features and established PIs, it retains independent prognostic value. Although a 14-day threshold is reasonable for risk modeling, DTI contains greater informational value when modeled as a continuous variable. Among PIs, FLIPI24 most strongly attenuated the DTI effect, underscoring its superior discriminative power over FLIPI and PRIMA-PI. Clinical trials designs should mitigate any potential barriers to enrollment of FL patients requiring urgent therapy.

Abstract Introduction There is currently no standard-of-care for patients (pts) who cannot tolerate full dose anthracycline regimens in frontline large B-cell lymphoma (LBCL). R-miniCHOP is a commonly used regimen that provides curative intent treatment with manageable toxicity. Despite this, lymphoma and treatment-related mortality are the leading causes of death in these pts. Additionally, many pts may not be eligible to receive R-miniCHOP due to comorbidity or frailty. There are limited clinical trial and real-world data focused on this population. Methods Pts with LBCL age 18 years (yrs) and older were identified from the Lymphoma Epidemiology of Outcomes (LEO) cohort, a prospective study of newly diagnosed pts with non-Hodgkin lymphoma enrolled from 2015-2020 at 8 academic medical centers in the US. Pts with LBCL were included in this analysis if they had received R-miniCHOP or a regimen that did not include an anthracycline (nonAC). NonAC regimens included R-CVP, BR, rituximab, radiation and lenalidomide and rituximab. Pts completed the Vulnerable Elders Survey (VES-13) and were considered vulnerable if they had a score >3. Event-free survival (EFS) was defined as the time from start of treatment until progression/relapse, retreatment, or death due to any cause; EFS at 24 months (EFS24) was a secondary endpoint. Overall survival (OS) was defined as the time from start of treatment until death due to any cause. EFS and OS for R-miniCHOP and non-IC groups, as well as vulnerable and not vulnerable by VES-13 were estimated using Cox models. Results A total of 190 pts received R-miniCHOP (N=139) or nonAC (N=51). R-miniCHOP-treated pts were older (median 82 vs 72 yrs) and more commonly female (55% vs 35%). The majority in both groups (86% vs 78%) received their treatment at a LEO center and did not receive therapy on a clinical trial (89% vs 100%). Most pts receiving R-miniCHOP had high-intermediate (32%) or high (30%) international prognostic index (IPI) score, while both high-intermediate (22%) and high (20%) IPI were lower in the nonAC-treated pts. Cell of origin was determined by immunohistochemistry as available clinically, with GCB 42% vs 31%, non-GCB 32% vs 29%, and not done/unknown 26% vs 39% in R-miniCHOP vs nonAC pts. Double-hit status by FISH was known in 115 pts and was positive in 13% of R-miniCHOP pts and 10% of nonAC pts. The diagnosis-to-treatment interval (DTI) was ≤14 days in 30% of R-miniCHOP pts and 24% of nonAC pts. Of pts who had a VES-13 score (N=119), vulnerable status trended higher in the R-miniCHOP (66%) vs the non-IC group (50%; p=0.15). With a median follow-up of 50.5 months, the median EFS was 37.1 months and median OS was 50.0 months. R-miniCHOP, compared with non-IC, was associated with improved EFS (unadjusted HR=0.54, 95% CI 0.36-0.81; p=0.003) and OS (unadjusted HR=0.58, 95% CI 0.38-0.88; p=0.011). When adjusted for age, IPI, LBCL subtype (diffuse large B-cell lymphoma, not otherwise specified vs other), DTI and VES-13, R-miniCHOP was strongly associated with improved EFS (HR=0.26, 95% CI 0.14-0.50; p<0.001) and OS (adjusted HR=0.29, 95% CI 0.15-0.57; p<0.001). EFS24 rate was 56% in all pts and was superior in pts treated with R-miniCHOP vs non-IC (62.7% vs 34.1%; p=0.001). In the R-miniCHOP group, vulnerable status (adjusted for IPI) was associated with inferior OS (HR=2.22, 95% CI 1.09-4.50; p=0.028) and trended towards inferior EFS (HR=1.76, 95% CI 0.93-3.32; p=0.083). In a competing-risk cumulative-incidence analysis, lymphoma was the leading cause of death for pts receiving R-miniCHOP at 2 (0.16, 95% CI 0.11-0.24) and 5 (0.20, 95% CI 0.14-0.28) yrs, as well as for those receiving non-IC at 2 (0.35, 95% CI 0.24-0.53) and 5 (0.38, 95% CI 0.26-0.56) yrs. Treatment-related mortality was similar in both groups. In the R-miniCHOP group, the 2-yr (0.20 vs 0.03) and 5-yr (0.20 vs 0.11) lymphoma-related mortality was higher in pts who were vulnerable vs not by VES-13, while there were no differences for mortality due to therapy or other causes by vulnerable status. Conclusions These data support R-miniCHOP as a curative regimen in older and frail pts with the majority achieving EFS24. R-miniCHOP compared to non-IC use is strongly associated with improved EFS and OS when adjusted for disease and patient related variables. Within the R-miniCHOP group, vulnerable status is associated with inferior OS. Additional research efforts are needed to optimize outcomes for this group of pts.

Abstract Background: Rural/nonmetropolitan (nonmetro) counties face known barriers to comprehensive cancer care compared with metropolitan (metro) counties, which may impact the outcomes for lymphoma patients. We evaluated the impact metro vs nonmetro status on outcomes after a diagnosis of non-Hodgkin lymphoma (NHL) in the context of clinical, demographic, and lifestyle factors. Methods: Of 7724 participants enrolled in the prospective LEO cohort study from 2015-2020, 7610 (98.5%) had their county of residence at diagnosis assigned a RUCC (2013) code. We compared outcomes of participants from metro (defined as RUCC 1-3, which are central counties with ≥1 urban areas of ≥50,000) vs nonmetro counties (RUCC 4-9, which are counties with smaller urban areas as well as completely rural). We estimated hazard ratios (HRs) and 95% confidence intervals (CI) from univariable and multivariable Cox models for associations of metro status with overall survival (OS). Cause of death (lymphoma [including treatment related], non-lymphoma, unknown) was assessed using a competing risk approach. Results: Compared with metro counties (N=6096), LEO participants from nonmetro counties (N=1514) were older (median age 64 vs 62 years; p<0.001), more likely to be non-Hispanic white (94% vs 75%; p<0.001) and married (84% vs 75%; p<0.001), and less likely to have completed graduate or professional school education (16% vs 27%; p<0.001). There were no notable differences (>5%) by metro status for sex, employment status, NHL subtype, stage, performance status, abnormal lactate dehydrogenase (LDH), international prognostic index (IPI), number of comorbidities, time from diagnosis to treatment interval (DTI) or participation in a clinical trial. Compared with LEO participants from metro counties, those from nonmetro counties had inferior OS at 5 years (74.5% vs 78.4%; HR=1.23, CI 1.10-1.37; p<0.001). The OS association only slightly attenuated after adjustment for age, sex, race/ethnicity and NHL subtype (HR=1.20, CI 1.07-1.34; p=0.002), and held in models that adjusted for clinical prognostic factors (stage, NHL subtype, IPI, DTI), socioeconomic status (SES; education, marital status), health status (comorbidity, frailty), or lifestyle factors (smoking, body mass index). Using a competing risk model, the cumulative incidence of death at 5 years due to lymphoma (16.9% vs 14.1%; p=0.002) and non-lymphoma (5.9% vs 3.9%; p=0.002) was higher for participants from nonmetro vs metro counties, respectively. For participants <65 years, nonmetro vs metro status had inferior OS (HR=1.40, CI 1.17-1.67; p<0.001). This association remained after adjustment for age, sex, race/ethnicity and NHL subtype (HR=1.36, CI 1.14-1.63; p<0.001), as well as after adjustment for clinical prognostic, SES, health status, or lifestyle factors. The cumulative incidence of death at 5 years due to lymphoma was higher in participants from nonmetro vs metro counties (15.1% vs 11.5%; p=0.002), while there was no difference by metro status for non-lymphoma related deaths (2.1% vs 1.7%; p=0.47). In contrast, for participants age 65+ years, there was no differences by metro status for OS (HR=1.04, CI 0.91-1.20; p=0.56). After adjustment for age, sex, race/ethnicity and NHL subtype, participants from nonmetro areas had inferior OS (HR=1.12, 0.97-1.29) although this was not statistically significant (p=0.13), and was attenuated in models adjusting for clinical, health status, or lifestyle factors. The cumulative incidence of death at 5 years due to lymphoma was no different in participants from nonmetro vs metro counties (18.9% vs 18.0%; p=0.37), while non-lymphoma related deaths were higher in nonmetro counties (9.9% vs 6.8%; p=0.017). Conclusion: LEO participants from nonmetro vs metro counties had inferior OS overall, although this was predominately observed for younger (<65 years) participants. These associations did not appear to be explained by demographic, NHL subtype, clinical, SES, health status, or lifestyle factors. Of particular note, the risk of death due to lymphoma was higher in nonmetro areas for participants <65 years but not those 65+ years, while the risk of death due to other causes was higher in nonmetro areas for participants 65+ years but not those <65 years. These findings require further study, particularly related to barriers to care, and suggest the potential need for age-specific strategies to optimize outcomes for NHL patients from nonmetro/rural areas.

Abstract Introduction Marginal zone lymphoma (MZL) is an indolent and heterogenous disease. Several prognostic models have been proposed to assess outcomes in MZL, including FLIPI, IPI (primarily designed for overall survival [OS]), and MZL-specific scores such as MALT-IPI (event-free survival [EFS]) and MZL-IPI (progression-free survival). FLIPI24 is a recently developed prognostic index for patients with follicular lymphoma (FL) treated with first-line (1L) immunochemotherapy (IC), identifying a high-risk population with inferior survival and increased risk of lymphoma-related death. FLIPI24 has also demonstrated robust performance in the entire FL population, including patients under observation. This study aimed to evaluate the prognostic utility of FLIPI24 in patients with MZL and to compare its performance with other established indices (FLIPI, MZL-IPI, IPI, MALT-IPI). Methods We pooled harmonized data from newly diagnosed MZL patients with complete available FLIPI24 variables (age, white blood cell count [WBC], hemoglobin [HGB], serum lactate dehydrogenase [LDH], and beta-2-microglobulin [B2M] levels) treated with 1L IC. Data were collected from three prospective international cohorts: NiHiL (Czech Republic; n=897; enrolled 2005–2023), LEO (USA; n=62; 2015–2020), and MER (USA; n=32; 2010–2015), forming a pooled cohort of 991 1L IC-treated patients. The analysis was then extended to include all newly diagnosed MZL patients (n=2,116), regardless of treatment. The primary endpoint in the 1L IC-treated cohort was EFS at 24 months (EFS24), and for all MZL patients, it was OS. Secondary endpoints included OS24 for 1L IC treated cohort, lymphoma-related mortality by FLIPI24 risk groups, and comparative prognostic performance of FLIPI, IPI, and MALT-IPI, by C-statistics; performance of MZL-IPI was assessed on a subgroup of NiHiL patients with available data (n=877 all MZL, n=485 1L IC-treated). Results Among the 991 1L IC-treated patients, the median age at diagnosis was 66 years (range 22–94); 47% were male; 11% had ECOG performance status ≥2, 68% clinical stage III–IV, 42% elevated LDH, 40% IPI 3–5, 42% FLIPI 3–5, and 62% MALT-IPI 2–3. The median B2M level was 2.8 mg/L, WBC 7.0 × 10⁹/L, and HGB 13.1 g/dL. MZL subtypes were: 64% extranodal (EMZL), 22% nodal (NMZL), and 14% splenic (SMZL). Most common 1L IC regimens were R-CVP (46%), R-CHOP-like (32%), R-chlorambucil (9%), and B-R (9%). FLIPI24 stratified patients into low- (18%), intermediate- (37%), and high-risk (45%) groups. High-risk patients were older (median 54 vs 65 vs 71 years for low-, intermediate-, and high-risk groups, respectively; P<0.01), had more frequent adverse baseline features, and were more likely to have the SMZL subtype (2% vs 7% vs 25% for low-, intermediate-, and high-risk groups, respectively; P<0.01). After a median follow-up of 7.2 years, EFS24 was 89%, 85%, and 65% (P<0.01) across low-, intermediate-, and high-risk groups; OS at 2 years was 98%, 97%, and 79%; and OS at 5 years was 96%, 91%, and 62%, resp. (P<0.01). The 5-year cumulative incidence of lymphoma-related death was 13% in high-risk vs 6% in low-/intermediate-risk groups. FLIPI24 showed better prognostic accuracy for EFS compared with FLIPI, MZL-IPI, IPI and MALT-IPI (C-index: 0.640 vs 0.630, 0.622, 0.620 and 0.603, resp.), as well as for OS (C-index: 0.710 vs 0.672, 0.641, 0.663, 0.652, resp.). Prognostic performance was consistent across MZL subtypes. In the extended cohort of 2,116 newly diagnosed MZL patients (47% receiving IC, 17% rituximab monotherapy, 11% radiotherapy, 8% surgery, 2% antibiotics; 15% observed), FLIPI24 stratified patients into low- (24%), intermediate- (39%), and high-risk (37%) groups. OS at 2 year was 99%, 97%, and 83%, at 5 years it was 96%, 91%, and 62%, resp. (P<0.01). The 5-year cumulative incidence of lymphoma-related death was 12% in high-risk vs 4% in low-/intermediate-risk patients. FLIPI24 outperformed other indices (OS C-index: 0.754 vs MZL-IPI 0.654, FLIPI 0.700, MALT-IPI 0.694, IPI 0.699). Conclusion FLIPI24 demonstrates strong and consistent prognostic performance across all MZL subtypes, identifying ~40% of patients as high-risk with significantly worse survival and increased lymphoma-related mortality. The FLIPI24 outperforms FLIPI, MZL-IPI, IPI, and MALT-IPI in predicting outcomes and may serve as a practical tool to guide risk-adapted management and selection of patients for novel therapies in MZL. Funding: NU21-03-00411, P50 CA97274, U01 CA195568.

Abstract Introduction The International Prognostic Index (IPI, NEJM, 1993) has guided risk stratification in large B-cell lymphoma (LBCL) for over three decades. Its simplicity and clinical utility have driven its widespread adoption, even in the era of precision medicine. However, the prognostic value of the IPI and its components may not be uniform across the age spectrum of LBCL patients. We evaluated the contemporary prognostic performance of the IPI in a large international cohort, focusing on age-stratified performance and the relative contributions of its individual components, grouped as patient- and disease-related factors. Methods We harmonized and pooled data from 6,941 patients with newly diagnosed systemic LBCL treated with R-CHOP-like regimens (R-CHOP, R-miniCHOP, DA-EPOCH-R, R-CHOEP, R-CHOP+X) with complete IPI data from three prospective cohorts: NiHiL (Czech Republic, n=4,590; 2010–2023), LEO (USA, n=1,789; 2015–2020), and MER (USA, n=562; 2010–2015). IPI components included age, ECOG performance status (PS) as patient-related, and Ann Arbor clinical stage, serum lactate dehydrogenase (LDH), and extranodal (EN) involvement as disease-related factors. Prognostic performance was evaluated using multivariable Cox regression models and C-statistics. Age-stratified analysis was performed for: ≤40 years (7%; n=492), 41–60 years (28%; n=1,909), 61–80 years (59%; n=4,080), and >80 years (7%; n=460). The primary endpoint was overall survival (OS). Results In the pooled cohort, the median age was 66 years (range 18–95). Compared to the original IPI cohort (n=3,273), our cohort was older (age >60 years: 66% vs 41%) but otherwise comparable: ECOG PS 2–4 (25% vs 24%), clinical stage III–IV (63% vs 66%), elevated LDH (60% vs 52%), and >1 EN involvement (32% vs 30%). In multivariable analysis, age (≤60 vs >60 years) was the strongest predictor of OS (HR=2.78; vs HR=1.96 in original IPI report), followed by ECOG PS (HR=2.07; vs 1.80), clinical stage (HR=1.50; vs 1.47), LDH (HR=1.41; vs 1.85, all P<0.01); EN involvement was not significant (P=0.50, HR=1.03 vs 1.48). Risk group distribution shifted from the original IPI: 28% were low-risk (0–1 factor; vs 35%), 23% low-intermediate (2; vs 27%), 24% high-intermediate (3; vs 22%), and 25% high-risk (4–5; vs 16%). Corresponding 2-year OS rates were 95%, 86%, 78%, and 63%, superior to those of the original IPI cohort (87%, 67%, 55%, 44%; pre-rituximab era). Age was associated with an increasing risk of mortality, with a more pronounced rise beyond 60 years. In a piecewise Cox model, the HR per year was 1.04 for ages ≤60 (linear), 1.05 for 61–69 (accelerated), and 1.06 for ≥70 years (exponential, all P<0.01). ECOG PS retained prognostic significance across age groups: HRs were 1.87 (P=0.06) for ≤40 years, 1.63 (P<0.01) for 41–60, 2.22 (P<0.01) for 61–80, and 1.43 (P<0.01) for >80 years. The strength of other IPI components declined with age. Clinical stage showed decreasing impact: HR 2.08 (P=0.09), 1.97 (P<0.01), 1.49 (P<0.01), and 1.31 (P=0.05). LDH was significant only up to age 80 years: HRs were 2.07 (P=0.04), 1.96 (P<0.01), 1.32 (P<0.01), and 1.21 (P=0.13). EN involvement was non-significant across all groups: HRs were 1.37 (P=0.34), 1.26 (P=0.05), 0.98 (P=0.74), and 1.06 (P=0.67). The IPI C-statistics was 0.676 overall and declined with age: ≤40 years (0.708), 41–60 years (0.683), 61–80 years (0.641), and >80 years (0.598). We next stratified the IPI factors into patient- (age, ECOG PS) and disease-related (clinical stage, LDH, EN involvement). In patients ≤40 years, disease-related factors outperformed patient-related (C-index 0.693 vs 0.602), as in the 41–60 group (C-index 0.677 vs 0.589). In patients aged 61–80, predictive values were comparable (0.613 vs 0.622). In >80-year-olds, both declined substantially (0.580 vs 0.568). Conclusion The IPI remains a valuable prognostic tool in LBCL, particularly among younger patients. However, its overall predictive accuracy declines with age. In patients ≤60 years old, prognosis is primarily driven by disease-related factors, suggesting a role for molecular classifiers to enhance risk stratification. In older patients, incorporation of additional patient-level factors, such as comorbidities and nutritional status could be considered. These findings underscore the need for more individualized prognostic tools for LBCL patients. Funding: NU21-03-00411, P50 CA97274, U01 CA195568, Charles University Haematology-Oncology Cooperatio Program.

Abstract Background: Recent advances in computer vision have given rise to image foundation models (FMs), large-scale neural networks pre-trained on millions of unlabeled histopathology patches, which can learn generalizable representations of tissue morphology. In many cancers, these FMs have predicted genetic mutations and prognoses directly from histology, suggesting that morphology encodes “hidden” molecular information. In lymphoma, this field is nascent but growing.General-purpose histopathology FMs have limited representation of lymphoid morphology. We developed and evaluated LymphoVision, a lymphoma-specialized image foundation model pretrained with self-supervised learning, to extract clinically meaningful features from routine hematoxylin and eosin (H&E) histology. Methods: LymphoVision was pre-trained using knowledge distillation and consistency (DINOv2) on over 37 million multi-resolution patches (5x, 10x, 20x, 40x) sampled using a density-aware strategy from 31,211 H&E whole slide images (WSIs) from 9,155 archival lymphoma and reactive lymphoproliferative cases. To ensure diverse, representative samples, our patch sampling method uses an efficient graph-based clustering algorithm and prioritizes morphologically informative regions by clustering patch embeddings and performing density sampling within each cluster. The pretrained encoder was evaluated using a clustering-constrained attention multiple instance learning (MIL) framework on three diagnostic tasks: (1) cell-of-origin (COO) classification in diffuse large B-cell lymphoma (DLBCL), distinguishing germinal center B-cell (GCB) from non-GCB subtypes based on the Hans immunohistochemistry algorithm; (2) follicular lymphoma (FL) grading, differentiating low-grade (grades 1–2) from high-grade (grade 3); and (3) multiclass lymphoma subtype classification across six categories: DLBCL, FL, marginal zone lymphoma (MZL), mantle cell lymphoma (MCL), classical Hodgkin lymphoma (CHL), and benign/reactive lymphoid conditions. WSI-level labels were derived from pathology reports, and none of the evaluation cases were included in pretraining. Models used a ViT-Giant LymphoVision backbone to generate 20x patch embeddings, followed by MIL using 50% training data and 2-fold cross-validation (25% validation, 25% test). Performance was assessed using area under the receiver operating characteristic curve (AUC), accuracy, recall (sensitivity) and F1 score. Results: For DLBCL COO classification (total n=358 cases), LymphoVision achieved a mean test AUC of 0.93 (range 0.92–0.94), accuracy of 84.5% and F1 of 0.85 across two independent test folds (n=91 per fold, 50% GCB). These results outperformed general-purpose FMs on the same task and cases, including Virchow-v2 (AUC 0.85) and UNI-v2 (AUC 0.88), despite those models being pretrained on 10–100 times more images. For FL grading (total n=660 cases), the test set in each fold included 166 cases (18% grade 3), yielding a mean test AUC of 0.88 (range 0.87–0.90), accuracy of 89.2% and F1 of 0.80. In multiclass lymphoma subtype classification (total n=1,176 cases), the test set included 296 cases per fold. LymphoVision achieved a mean test AUC of 0.98 (range 0.97-0.99), mean test accuracy of 87.1% and mean F1 of 0.84. Class-wise recall was 92.0% for DLBCL (n=112), 88.2% for FL (n=72), 77.2% for MZL (n=29), 70.8% for MCL (n=35), 87.5% for CHL (n=24), and 95.8% for benign/reactive lymphoid conditions (n=24). Conclusion: LymphoVision is a lymphoma-specialized FM trained using self-supervision that achieves high diagnostic accuracy across multiple clinically meaningful tasks using weak supervision and standard H&E images, suggesting the encoder has learned fine-grained lymphoma morphology. Based on historical benchmarks from the 1997 International Lymphoma Study Group classification project, expert hematopathologists achieved diagnostic agreement rates of 55% to 84% using H&E morphology alone for the non-Hodgkin lymphoma (NHL) entities included in our multiclass task. LymphoVision approached or exceeded these benchmarks. These findings underscore the value of our approach using disease-specialized pretraining and intelligent patch sampling for data-efficient histopathology learning. LymphoVision has the potential to augment diagnostics, enable scalable AI-powered classification in real-world workflows, and serve as a foundation for prognostic and theragnostic applications. External validation and multimodal integration are ongoing.

Abstract Background: Although progression-free survival (PFS) in early-stage (ES) classic Hodgkin Lymphoma (cHL) is high, there remains a subset of patients (pts) at higher risk of worse outcomes, particularly in those with unfavorable risk disease. Leveraging pt-level data from within the HoLISTIC (Hodgkin Lymphoma International STudy for Individual Care) Consortium, the E-HIPI was developed in clinical trial pts and validated in real-world registry pts to predict 2-year (y) PFS (Rodday. NEJM Evidence 2025). Continuous, objective, and readily measurable pre-treatment variables were considered, with sex, maximum tumor diameter, albumin, and hemoglobin included as significant predictors. Using the E-HIPI, interactive calculators for pt risk, comparison, and stratification were created (https://rtools.mayo.edu/holistic_ehipi/). Although the E-HIPI was developed and validated in both unfavorable and favorable ES cHL pts, the clinical predictors are primarily markers of unfavorable disease. In the development and initial validation cohorts, 2y PFS was 93.7% and 90.3%, respectively, the C-statistic was 0.63 in both cohorts, calibration was strong with slight underprediction of events in the validation cohort, and the E-HIPI outperformed the historic binary classification of patients as favorable or unfavorable using EORTC criteria. However, since pts in the original E-HIPI development and validation were primarily treated with ABVD chemotherapy protocols, additional validation is needed, particularly among pts with unfavorable ES cHL who were treated initially with more intensive chemotherapeutic platforms (e.g., BEACOPP-based). Thus, we conducted an external validation of the E-HIPI in pts treated on recent unfavorable ES cHL GHSG trials. Methods: The validation cohort included 2367 adults (ages 18 to 60y) with newly-diagnosed unfavorable cHL from randomized phase 3 trials conducted by the GHSG (HD14 von Tresckow. JCO 2012; and HD17 Borchmann. Lancet 2021). We calculated the predicted probability of 2y PFS events based on the E-HIPI model equation (scale 1-100%). Model performance was assessed using discrimination and calibration. Discrimination, which is the model's ability to separate high and low risk pts, was assessed using Harrell's C-statistic. Calibration, which is the model's ability to accurately predict absolute risk, was assessed by comparing observed and predicted probabilities of 2y PFS events within quintiles of predicted probabilities and estimating calibration-in-the-large (predicted minus observed event rate) and the calibration slope. Results: For pts treated on unfavorable GHSG trials, mean age was 33y, 95% had stage II disease, mean baseline E-HIPI score (i.e., predicted 2y PFS event rate) was 6.0% (SD=2.3%), and the observed 2y PFS rate was 96.2% (95%CI: 95.4-97.0%). The C-statistic was 0.65 (95% CI: 0.59-0.70), calibration-in-the-large was 0.02, and the calibration slope was 1.08 (95% CI: 0.52, 1.64). Within the lowest risk E-HIPI quintile, the observed 2y PFS event rate was 1.6%, compared with 6.1% in the highest risk quintile. Conclusions: In external validation of the recently published E-HIPI in pts with cHL treated on unfavorable GHSG trials using more intensive BEACOPP-based protocols, discrimination was similar to the original development and validation cohorts. Collectively, the E-HIPI is a treatment-agnostic prediction tool across varied treatment regimens for unfavorable ES cHL pts. Better 2y PFS in the GHSG trials than in the E-HIPI development and initial validation cohorts may have contributed to mild misclassification, likely reflecting differences in treatment regimens. This may be resolved with recalibration or incorporating granular treatment variables into future models. Additional research is needed to understand E-HIPI performance in pts with favorable disease who have even lower event rates as well as to more fully predict individualized pt outcomes across varied chemotherapy platforms and radiation received for all ES cHL pts.

Abstract Introduction The LEO Comorbidity Index (LCI), comprising ten equally weighted comorbidity categories (respiratory, cardiovascular, digestive, hepatic, renal, autoimmune, diabetes, cancer history, HIV, and stroke), has demonstrated prognostic utility in patients with non-Hodgkin lymphomas. However, its predictive accuracy may vary with increasing age, due to age-related rises in comorbidity burden and physiological vulnerability. To investigate these dynamics, we evaluated the age-stratified predictive performance of LCI in patients with large B-cell lymphoma (LBCL), with a focused analysis on individuals aged 80+ years. We also examined which specific comorbidities most strongly influenced overall survival (OS) and developed a simplified, clinically applicable index tailored to this population. Methods We first included adults ≥40 years with newly diagnosed LBCL and recorded comorbidities from the LEO registry (USA, 2015–2020; n=1,652) to assess age-stratified predictive performance of LCI using the C-index. Next, we focused on patients 80+ years (n=263), pooled from the LEO cohort (n=147) and NiHiL registry (restricted to General Hospital, Prague, Czech Republic, 2010–2023; n=116). Comorbidities were collected from patient self-reports (LEO) and/or medical records (LEO/NiHiL). Univariate and multivariate Cox models were used. The primary endpoint was OS. Results Among the total LBCL population (n=1,652), LCI predictive performance improved by age category with C-index for patients <65 of 0.584, 65–79 years 0.579, and ≥80 years 0.647, indicating stronger discrimination in the 80+ group. Subsequent analysis focused on patients aged 80+ years (n=263; median age 83). Within this cohort, 34% had ECOG performance status (PS) 2–4, 66% stage III–IV disease, 59% elevated lactate dehydrogenase, 35% >1 extranodal site involved, and 58% International Prognostic Index (IPI) score of 3–5. A total of 377 comorbidities were reported, averaging 1.43 per patient (1.22 in LEO; 1.61 in NiHiL). LCI distribution was: 0 points in 24% (n=62), 1 point in 33% (n=88), and 2–5 points in 43% (n=113). Higher LCI scores had worse ECOG PS 2–4: 19% (LCI=0), 33% (LCI=1), and 43% (LCI ≥2; P=0.01), and inferior OS (median 6.9 years LCI=0, 3.0 years LCI=1, and 2.8 years LCI≥2, resp.; P=0.02). Presence of ≥1 comorbidity was associated with shorter OS compared to no comorbidities (HR=1.67, P=0.01). The LCI was an independent predictor of OS after adjusting for age (HR=1.18, P=0.01, C-index 0.623), IPI (HR=1.16, P=0.03, C-index 0.628) as well as for recently proposed SENIOR-IPI (HR=1.14, P=0.049, C-index 0.665). When evaluating individual comorbidities, heart disease (HR=1.54, P=0.01), stroke (HR=1.89, P=0.03), and recent malignancy (HR=1.99, P=0.02) were independent predictors of OS. Based on these findings, we developed a simplified LCI (sLCI) for use in 80+ patients incorporating only the three comorbidities (1 point each). Patients were categorized as sLCI: 0 (49%, n=129), 1 (44%, n=115), and 2–3 points (7%, n=19). sLCI scores 0, 1, 2–3 were associated with ECOG PS 2–4 27%, 37%, and 63%, respectively (P=0.01) and inferior OS: median 5.3 years (0), 2.5 years (1), and 1.4 years (2–3; P<0.01). sLCI remained an independent and strong predictor of OS after adjusting for age (HR=1.71, P<0.01, C-index 0.636), IPI (HR=1.66, P<0.01, C-index 0.635), and SENIOR-IPI (HR=1.60, P<0.04, C-index 0.659). The results were consistent in both NiHiL and LEO cohorts and remained significant among patients treated with anthracyclines (sLCI adjusted for age: HR=1.58, P<0.01, C-index 0.607; for IPI: HR=1.53, P=0.01, C-index 0.623; for SENIOR-IPI: HR=1.51, P=0.01, C-index 0.633). Conclusion Comorbidity burden becomes increasingly prognostic with higher age in LBCL patients. In the 80+ individuals, comorbidities – particularly heart disease, recent cancer, and stroke – have an independent impact on OS. While the original LCI captures the impact of broadly defined comorbidities, its prognostic utility attenuates in the 80+ patients, likely reflecting limited relevance of indolent chronic conditions in this population's constrained survival horizon. The sLCI, focused on the most prognostically relevant comorbidities, offers practical risk stratification. Incorporation of sLCI into clinical decision-making – alongside other prognostic indices – could better identify high-risk individuals and guide treatment strategies. Funding: NU21-03-00411, U01 CA195568.