Diffuse Large B-cell Lymphoma Research Articles

Prognostic outcomes of diffuse large B-cell lymphoma patients with myelocytomatosis oncogene (MYC) and B-cell lymphoma 2 (BCL2) co-expression.

Double expressor lymphoma (DEL) refers to diffuse large B-cell lymphoma (DLBCL) cases characterized by the overexpression of both MYC and BCL2 proteins, as determined by immunohistochemistry (IHC), without requiring underlying genetic rearrangements. DEL is associated with more aggressive disease behavior and poorer prognosis. This study aimed to assess the impact of DEL on progression-free survival (PFS) and overall survival (OS) compared to non-DEL patients. We conducted a retrospective study at the Hospital, analyzing 177 patients diagnosed with DLBCL between March 2014 and March 2021. Patients were classified as DEL or non-DEL based on immunohistochemical analysis. Survival rates, clinical characteristics, and treatment responses were compared using Kaplan-Meier survival analysis, and multivariable Cox regression was performed to identify independent prognostic factors. Among 177 patients, 113 (63.8%) were DEL and 64 (36.2%) non-DEL. DEL patients had significantly worse outcomes, with a median follow-up of 39.4months. The 3-year PFS (44.2% vs. 68.8%) and OS (54.9% vs. 81.3%) were significantly lower in DEL (PFS: p < 0.001; OS: p = 0.001). Median PFS in DEL was 19months. Multivariable analysis confirmed DEL as an independent predictor of worse PFS (HR: 1.488, 95% CI: 1.091-2.03, p = 0.012) and OS (HR: 1.376, 95% CI: 1.011-1.873, p = 0.043). DEL status is strongly linked to poor survival in DLBCL, highlighting the need for targeted therapies beyond R-CHOP. Future research should explore personalized treatment strategies to improve outcomes in this high-risk group.

CtBP2 Regulates Wnt Signal Through EGR1 to Influence the Proliferation and Apoptosis of DLBCL Cells.

Diffuse large B-cell lymphoma (DLBCL) is the most prevalent form of lymphoma. The overexpression of CtBP2 in tissues may contribute to tumor occurrence and progression. The expression of EGR1 in DLBCL is elevated, suggesting its potential role as an oncogene that promotes the proliferation of DLBCL cells. Database predictions indicate that CtBP2 can bind to EGR1. The objective of the present study was to investigate whether CtBP2 can influence the proliferation and apoptosis of DLBCL cells by regulating the Wnt signaling pathway through EGR1. Western blot assay showed that CtBP2 expression was upregulated in DLBCL cells. Cell proliferation level was detected by CCK8 assay and EdU staining, and the apoptosis level and cycle distribution were analyzed through flow cytometry. Our data indicated that interference with CtBP2 and EGR1 can inhibit the proliferation and cell cycle progression of DLBCL cells while promoting apoptosis. The predictions from the HDOCK server, along with the results of Co-IP experiments, suggested that EGR1 and CtBP2 can effectively bind to each other, with EGR1 positioned downstream of CtBP2 and regulated by it. Furthermore, interference with CtBP2 could also inhibit the expression of the Wnt/β-catenin signaling pathway. Overexpression of EGR1 counteracted the effects of siRNA-CtBP2, promoting cell proliferation and cycle, inhibiting apoptosis and upregulating the expression of the Wnt/β-catenin signaling pathway. From the above experiments, we found that CtBP2 can regulate the Wnt/β-catenin signaling pathway through EGR1 to influence the proliferation and apoptosis of DLBCL cells. Therefore, EGR1 may be one of the key contributors involved in the regulation of Wnt/β-catenin signaling by CtBP2.

Discordant lymphoma characterized by the coexistence of diffuse large B-cell lymphoma in the brain and mantle cell lymphoma in the colon, rectum, and bone marrow.

We describe a rare case of discordant lymphoma characterized by the coexistence of diffuse large B-cell lymphoma (DLBCL) in the brain and mantle cell lymphoma (MCL) in the colon, rectum, and bone marrow. A 63-year-old male patient with consciousness impairment and gait disturbance was admitted to our institution. Head computed tomography scan and contrast-enhanced magnetic resonance imaging showed a mass in the right temporal lobe and rectal wall thickening. Brain biopsy revealed DLBCL, and bone marrow and rectum biopsy showed MCL. According to a polymerase chain reaction analysis of immunoglobulin heavy-chain gene rearrangements using brain and bone marrow specimens, the two lesions were clonally unrelated lymphomas. After five cycles of R-MPV (rituximab, methotrexate, procarbazine, vincristine) therapy and three cycles of R-ESHAP (rituximab, etoposide, cytarabine, cisplatin, methylprednisolone) therapy, the patient received autologous hematopoietic stem cell transplantation using R-MEAM (rituximab, ranimustine, etoposide, cytarabine, melphalan) regimen after bridging therapy with ibrutinib. In addition, he received whole-brain irradiation at a dose of 40Gy in 20 fractions as consolidation therapy. He did not relapse within 3years of transplantation. To the best of our knowledge, this is the first case report of DLBCL and MCL coexistence.

Temporal trends in time toxicity of R-CHOP: a nationwide hospital-based database analysis in Japan

PurposeWhile the prognosis of patients with cancer has improved, the time burden of treatment has recently been recognized as time toxicity; although, the actual clinical situation remains largely unexplored. This retrospective study aimed to elucidate the time toxicity of rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) in patients with B-cell lymphoma and the factors influencing it.MethodsWe used a nationwide hospital-based database between January 2010 and November 2021 in Japan. We extracted the claims data of patients with diffuse large B-cell lymphoma and follicular lymphoma who were hospitalized and/or visited hospitals for chemotherapy.ResultsAmong the 7760 R-CHOP administered to 2006 patients, the rate of outpatient therapy increased over time (2010–2015: 17.9%; 2016–2021: 31.8%). In 2016, the median length of hospitalization was the shortest at 13 days (IQR 8–19), which coincided with the peak use of pegylated granulocyte colony-stimulating factor (Peg-G-CSF) during hospitalization in 2015–2016, likely driven by changes in the insurance system. In multivariate analysis, the factors associated with longer hospital stays were older age and poor activities of daily living, whereas the use of Peg-G-CSF, a reduced-dose regimen, and treatment at cancer-designated hospitals were associated with shorter stays.ConclusionThe time toxicity of R-CHOP has improved and may be influenced by the patient’s condition, adequate supportive care, changes in the insurance system, and center-specific treatment proficiency.

Expression and Clinical Significance of CXCR5 and LAG-3 on Peripheral Blood CD8+ T Cells in Patients With Diffuse Large B-Cell Lymphoma.

Diffuse large B-cell lymphoma (DLBCL) exhibits substantial biological and clinical heterogeneity. This study investigated the expression and prognostic implications of C-X-C chemokine receptor type 5 (CXCR5) and lymphocyte activation gene-3 (LAG-3) on peripheral blood CD8+ T cells in patients with DLBCL. A total of 71 DLBCL patients and 71 healthy controls were enrolled. The expression levels of CXCR5 and LAG-3 on peripheral blood CD8+ T cells were assessed and analyzed for their impact on 5-year progression-free survival (PFS) and overall survival (OS). Results revealed significantly elevated CXCR5 and LAG-3 expression levels in DLBCL patients compared to controls. CXCR5 expression correlated with lactate dehydrogenase (LDH) levels, extranodal involvement, Ann Arbor stage, and International Prognostic Index (IPI) scores, while LAG-3 expression was associated with Eastern Cooperative Oncology Group (ECOG) scores, number of extranodal sites, bone marrow involvement, Ann Arbor stage, and IPI scores. Multivariate analysis identified advanced age, Ann Arbor stage III-IV, and elevated CXCR5 and LAG-3 expression as independent risk factors for poorer 5-year PFS and OS. Furthermore, patients with higher CXCR5 and LAG-3 expression levels demonstrated significantly reduced 5-year PFS and OS rates. In conclusion, elevated CXCR5 and LAG-3 expression on peripheral blood CD8+ T cells plays a pivotal role in DLBCL progression and prognosis, making these markers potential therapeutic targets or prognostic indicators.

Blunted CD40-responsive enhancer activation in CREBBP-mutant lymphomas can be restored by enforced CD4 T-cell engagement.

Blunted CD40-responsive enhancer activation in CREBBP-mutant lymphomas can be restored by enforced CD4 T-cell engagement.

Biology as vulnerability in follicular lymphoma: genetics, epigenetics, and immunogenetics.

Biology as vulnerability in follicular lymphoma: genetics, epigenetics, and immunogenetics.

A predictive diagnostic model for refractory diffuse large B-cell lymphoma: a single-center retrospective cohort study.

To develop a baseline predictive model for refractory diffuse large B-cell lymphoma (DLBCL) utilizing imaging data including ultrasound findings and PET-CT in conjunction with clinical parameters. We retrospectively analyzed data from 140 patients with newly diagnosed DLBCL treated at Peking University Third Hospital between January 2018 and January 2023. All patients underwent ultrasound, histopathological examinations and PET-CT examinations. After completing 6-8 cycles of standardized chemotherapy, patients were categorized into refractory and non-refractory groups according to the Lugano International Response Assessment Criteria. Univariate analyses were performed using T-tests and Chi-Squared Tests, and independent risk factors for refractory DLBCL were identified through logistic regression. A nomogram predictive model was constructed using the R package "rms," and its predictive performance was subsequently validated. Univariate analysis and logistic regression identified that blurred margins of the affected lymph nodes in ultrasound images (P < 0.001, OR = 18.238) and IPI score(P = 0.051, OR = 3.131) were significant risk factors for disease progression. The predictive nomogram established for refractory diffuse large B-cell lymphoma demonstrated an area under the receiver operating characteristic curve (AUC) of 0.835, with a sensitivity of 85.5% and specificity of 79.5%. Following internal validation, the predictive model exhibited a high degree of alignment between the estimated risk of refractory diffuse large B-cell lymphoma and the actual observed progression events. The prediction model of the R-DLBCL prediction model, amalgamating ultrasonic characterizations and clinical indicators, proves instrumental in identifying high-risk DLBCL groups.

Lymphocyte/monocyte to lactate dehydrogenase ratio prior to lymphodepletion impact the outcomes of patients with diffused large B cell lymphoma undergoing CAR-T cell therapy

Factors associated with outcomes of chimeric antigen receptor (CAR)-T cell therapy in patients with relapsed or refractory (R/R) diffuse large B-cell lymphoma (DLBCL) have not been fully elucidated. We explored the impact of the prelymphodepletion (pre-LD) lymphocyte to monocyte ratio (LMR) and its ratio to lactate dehydrogenase (LDH) (LMR/LDH) on the efficacy and prognosis of 60 patients with R/R DLBCL undergoing CAR-T cell therapy. The optimal cutoff values for pre-LD LMR and LMR/LDH were 3.583 and 0.0103, respectively. The overall response rate (ORR)s were higher in patients with high pre-LD LMR or LMR/LDH than those with low pre-LD LMR or LMR/LDH (ORR, 100% vs. 65.79%, P = 0.006 and 96.15% vs. 38.24%, P < 0.0001, respectively). Pre-LD LMR/LDH was an independent factor associated with ORR (P = 0.010, odds ratio = 18.757; 95% confidence interval [CI] 2.046–171.975) by multivariate logistic regression analysis. Patients with high pre-LD LMR/LDH had significantly longer progression-free survival (PFS) (median PFS, 29.73 vs. 2.47 months, P < 0.0001) and overall survival (OS) (median OS, not reached vs. 7.4 months, P = 0.0002) than those with low pre-LD LMR/LDH. Multivariate Cox regression analysis showed that pre-LD LMR/LDH and ORR were independent factors affecting PFS (P = 0.030, hazard ratio [HR] = 2.561; 95% CI 1.093–5.999 and P = 0.024, HR = 2.202; 95% CI 1.22–4.369, respectively); pre-LD LMR/LDH was an independent factor affecting OS (P = 0.029, HR = 3.331; 95% CI 1.131–9.807). In conclusion, the pre-LD LMR/LDH was an independent factor associated with ORR and an independent prognostic factor in patients with R/R DLBCL undergoing CAR-T cell therapy.

Delayed Methotrexate Elimination Following High-dose Methotrexate Prophylaxis in High-risk Diffuse Large B-cell Lymphoma.

Objective High-dose methotrexate (HD-MTX) is widely used as central nervous system (CNS) prophylaxis in patients with diffuse large B-cell lymphoma (DLBCL) who are at a high risk of CNS relapse. Ensuring safe prophylactic administration with minimal adverse events is a key concern; however, few studies have detailed the safety profile of HD-MTX prophylaxis in patients with high-risk DLBCL. We analyzed the adverse events associated with HD-MTX in this population, focusing on delayed MTX elimination. Methods This multicenter retrospective study included 98 patients with DLBCL at high risk of CNS relapse who received HD-MTX as part of frontline therapy between 2014 and 2020. CNS prophylaxis involved 2 cycles of HD-MTX (3.0 g/m2) at 2-week intervals. Results The median age at the diagnosis was 63 (34-84) years old, and 34 patients received a reduced methotrexate (MTX) dose. The overall incidence of delayed MTX elimination was 18.4%. No cases of delayed MTX elimination were observed in the group that received a 3-h MTX infusion (n=50). Toxicities were more frequent in patients with delayed MTX elimination than in those without (77.8% vs. 26.2%, p≤0.05), including higher incidences of grade ≥3 adverse events and grade ≤2 renal dysfunction. Conclusion Delayed MTX elimination is associated with increased complications. Shorter MTX infusion rates, particularly at 3 h, may reduce the risk of delayed MTX elimination.

Subgroup analysis of elderly patients (≥60 years) with diffuse large B-cell lymphoma in the phase 3 POLARIX study.

In the phase 3 POLARIX study, polatuzumab vedotin plus rituximab, cyclophosphamide, doxorubicin, and prednisone (Pola-R-CHP) improved progression-free survival (PFS) versus rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) in patients with previously untreated diffuse large B-cell lymphoma (DLBCL). This post hoc subgroup analysis of POLARIX evaluated the efficacy and safety of Pola-R-CHP versus R-CHOP in elderly patients ≥60, ≥65, ≥70, and ≥75 years. As of June 15, 2022 (median follow-up 40 months), 629 patients ≥60 years were included (Pola-R-CHP, n = 311; R-CHOP, n = 318). Clinically meaningful improvements in PFS with Pola-R-CHP versus R-CHOP were observed across all age groups, particularly in patients ≥70 years whereby the risk of disease progression, relapse or death was reduced by 37% (unstratified hazard ratio [HR] 0.63; 95% confidence interval [CI]: 0.41-0.96). In patients ≥ 60 years, overall survival was similar with Pola-R-CHP versus R-CHOP (unstratified HR 0. 99; 95% CI: 0.67 -1.47 ). Safety profiles were similar for Pola-R-CHP versus R-CHOP among patients ≥60 years, including rates of grade 3-4 adverse events (AEs; 62.7% vs 61.5%), grade 3-5 infections (15.0% vs 12.9%), and grade 5 AEs (3.6% vs 3.2%); no novel toxicities were reported. Incidence of grade 3-4 febrile neutropenia was higher with Pola-R-CHP than R-CHOP (16.3% vs 7.6%), highlighting the importance of G-CSF prophylaxis in elderly patients receiving Pola-R-CHP. The benefit-risk profile favored Pola-R-CHP versus R-CHOP in elderly patients with previously untreated DLBCL. This trial was registered at ClinicalTrials.gov as #NCT03274492.

Measuring Total Metabolic Tumor Volume from 18F-FDG PET: A Reality Check.

Measuring total metabolic tumor volume (TMTV) on 18F-FDG PET/CT images in clinical practice requires a fast, reliable, and easy-to-perform multilesional segmentation workflow. We conducted a field test to derive total metabolic volumes using 5 representative baseline 18F-FDG PET/CT scans from patients with diffuse large B-cell lymphoma. The scans were transferred to 10 different sites or readers who used different commercially available software platforms to derive TMTV after a recently proposed benchmark workflow. Observed TMTVs were compared with reference values, and overall analysis times were reported. Our results show that TMTVs can be obtained with reasonable accuracy across readers and platforms (within 10% compared with reference benchmark values for most TMTVs) but that processing times can vary considerably depending on reader experience and the software platform. Our study showed that there is an urgent need to improve TMTV segmentation workflows in clinical practice, requiring closer collaboration between users and software vendors.

Shorting of existing conditioning regimen for relapsed/refractory gastric diffuse large B-cell lymphoma transplant and studying its related outcomes: A first of its kind case report

Shorting of existing conditioning regimen for relapsed/refractory gastric diffuse large B-cell lymphoma transplant and studying its related outcomes: A first of its kind case report

Genetic subtyping by Whole Exome Sequencing across Diffuse Large B Cell Lymphoma and Plasmablastic Lymphoma.

Diffuse Large B-Cell Lymphoma (DLBCL) is a heterogeneous disease characterized by a limited number of molecularly defined subtypes. Recently, genomic-based algorithms have been proposed for the classification of this disease. The whole exome sequencing was conducted on 108 diagnostic samples of diffuse large B-cell lymphoma (DLBCL). Somatic variants, predicted copy number alterations (CNAs), and available fusion data were utilized to classify the cases. Additionally, the enrichment of mutations in the TP53, MYC, and MAPK/ERK pathways was analyzed. Genetic subtypes were identified in approximately 55% of the cases. Cases with a specific genetic subtype exhibited a significantly higher Tumor Mutation Burden compared to molecularly unclassified cases (Mann-Whitney U test, p = 0.024). The prevalence of subtypes varied according to the cell of origin phenotypes. GC-B type DLBCL NOS were classified as EZB (5 cases, 16%), ST2 (5 cases, 16%), and BN2 (1 case, 3%). Four cases (13%) were genetically composite. Three cases of HGBCL/DLBCL double-hit (MYC & BCL2) were classified as EZB-MYC. Forty-three non-GC-B type DLBCL cases were classified as ST2 (5 cases, 11%), BN2 (6 cases, 14%), and MCD (3 cases, 7%). Nine cases were genetically composite (20%). MYC pathway mutations were enriched in cases with EZB and ST2 genetic features, while they were absent in the MCD subtype. TP53 mutations were identified in 11% of the cases. Plasmablastic lymphomas exhibit genetic diversity, with 27% of tumors classified as ST2. Recurrent somatic mutations indicate dysregulation of the JAK/STAT, MAPK/ERK, and tyrosine kinase signaling pathways.

MYC networks associate with decreased CD8 T-cell presence in diffuse large B-cell lymphoma and may be addressed by the synergistic combination of AZD4573 and Selinexor – a preliminary analysis

MYC networks associate with decreased CD8 T-cell presence in diffuse large B-cell lymphoma and may be addressed by the synergistic combination of AZD4573 and Selinexor – a preliminary analysis

Selective Enhancer Dependencies in MYC-Intact and MYC-Rearranged Germinal Center B-cell Diffuse Large B-cell Lymphoma.

High expression of MYC and its target genes identify germinal center B-cell diffuse large B-cell lymphomas (GCB-DLBCL) associated with poor outcomes. We used CRISPR-interference profiling of human lymphoma cell lines to define essential enhancers in the MYC locus and non-immunoglobulin rearrangement partner loci, including a recurrent rearrangement between MYC and the BCL6 locus control region. GCB-DLBCL cell lines without MYC rearrangement are dependent on an evolutionarily-conserved enhancer we name "germinal center MYC enhancer 1" (GME-1), which is activated by the transcription factor complex of OCT2, OCA-B, and MEF2B, shows an active chromatin state in normal human and mouse germinal center B cells, and demonstrates selective acetylation and MYC promoter topological interactions in MYC-intact GCB-DLBCL biopsies. Whole-genome sequencing identified tandem copy gains of the GME-1 enhancer as a rare but recurrent event in DLBCL. Our findings shed new light on mechanisms that dysregulate MYC, a key driver of B cell malignancy.

Prognostic relevance of immunoglobulin heavy chain rearrangement and immunoglobulin kappa light chain rearrangement in patients with diffuse large B cell lymphoma.

Evidence has demonstrated that monitoring of the variable, diversity, and joining gene segments (VDJ) rearrangement of immunoglobulin (Ig) gene in the circulating tumor DNA (ctDNA) is highly valuable in predicting the prognosis of patients with diffuse large B cell lymphoma (DLBCL). In this study, we investigated the role of both Ig heavy chain (IGH) and Ig kappa light chain (IGK) gene rearrangements detected in ctDNA samples in predicting DLBCL progression. Next-generation sequencing (NGS) was used to identify the dominant V(D)J clonotypic rearrangement in tissue samples of 33 DLBCL patients. Minimal residual disease (MRD) was monitored at the interim and end of the treatment, as well as the follow-up time by tracking the dominant V(D)J clonotypic rearrangement (defined as the "NGS MRD" method) in the peripheral blood (PB) ctDNA samples. The nomogram was established to predict the 12-month and 24-month progression-free survival (PFS) probability. Prior to treatment, the dominant clones identified in the tissue samples could be retrieved in tissue-matched PB of 26 (78.8%, 26/33) patients. The addition of IGK clones to IGH clones increased the MRD detection rate from 42.9% to 58.0% in the total series. NGS MRD and imaging scans showed poor concordance at the interim of treatment (Kappa = 0.24) and the follow-up time (Kappa = 0.28), and fair concordance at the end of treatment (Kappa = 0.46). However, we confirmed that the interim NGS MRD monitoring demonstrated improved prognostic performance compared to imaging scans, and both NGS MRD monitoring and imaging scans served as valuable prognostic factors for PFS at the end of treatment. Notably, NGS MRD monitoring predicted disease relapse in 3 patients prior to imaging scans. Furthermore, we found that both the faster IGH and IGK clone clearance rates were associated with favorable prognosis. The nomogram model identified IGH and IGK clone clearance rates, together with the interim NGS MRD result were the important predictors of 12-month and 24-month progression of DLBCL. MRD monitoring via NGS of Ig for both IGH and IGK is a promising noninvasive tool for prognosis prediction and early relapse prediction of DLBCL patients.

Prognostic Value of Very Early Interim FDG PET/CT After Single Cycle of Chemotherapy for 10-Year Survival in Diffuse Large B-Cell Lymphoma

Background/Objectives This study aimed to evaluate whether very early interim 18F-fluoro-2-deoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT) after a single cycle of first-line chemotherapy predicts long-term survival outcome in patients with diffuse large B-cell lymphoma (DLBCL). Methods A total of 51 patients (31 males and 20 females; mean age 55 years) had four FDG PET/CT studies, at baseline and after one, three, and six cycles of chemotherapy (PET0, PET1, PET3, and PET6). Visually and quantitatively assessed PET parameters were analyzed for associations with long-term survival. Results The estimated 10-year progression-free survival (PFS) and overall survival (OS) was 48% and 61%, respectively. During a median follow-up of 63 months (range 9–134), 17 patients (33%) exhibited disease progression and 15 (29%) died. On PET1, all but one showed decreased FDG uptake, and all showed decreased metabolic tumor volume. None of the PET1 or PET3 parameters were associated with survival. The PET6 parameters retained independent predictive value for OS after adjustment for the International Prognostic Index. Negative PET6 was associated with longer PFS (mean 99 vs. 50 mo, p = 0.04) and OS (mean 107 vs. 57 mo, p = 0.02). Con-clusions The FDG PET/CT parameters obtained after a single cycle of chemotherapy were not associated with long-term survival in DLBCL, while negative end-of-therapy FDG PET/CT was associated with longer PFS and OS. Tumor regression very early into first-line chemotherapy was not as clinically relevant as the presence of viable tumor on FDG PET/CT at the end of therapy for predicting long-term outcomes.

Clinician-driven automated data preprocessing in nuclear medicine AI environments.

Artificial Intelligence (AI) approaches in clinical science require extensive data preprocessing (DP) steps prior to building AI models. Establishing DP pipelines is a non-trivial task, mainly driven by purely mathematical rules and done by data scientists. Nevertheless, clinician presence shall be paramount at this step. The study proposes a data preprocessing approach driven by clinical domain knowledge, where clinician input, in form of explicit and non-explicit rules, directly impacts the algorithms' decision-making processes, thus, making the DP planning phase more inclusive for clinicians. The rule set table (RST) was introduced as interface which accepts clinician's input as formal rules (including four actions: exp-keep, exp-remove, pref-keep, pref-remove features or samples) in human-readable form and translates it to machine readable input for preprocessing algorithms. A collection of commonly used algorithms was incorporated for data preprocessing of various clinical cohorts in both single and multi-center scenarios. The impact of RST was evaluated by utilizing 100-fold Monte Carlo cross-validation scheme for prostate and glioma cohorts (single center) with 80 - 20% training-testing split. Furthermore, diffuse large B-cell lymphoma (DLBCL) cohort was evaluated by using Center 1 as training and Center 2 as testing cohort for clinical endpoint prediction. Both scenarios were investigated in manual and automated data preprocessing setups across all cohorts. The XGBoost algorithm was employed for classification tasks across all established models. Predictive performance was estimated by confusion matrix analysis in validation samples of all cohorts. The performance of RST across all actions as well as without RST were compared in both manual and automated settings for each respective cohort. Performance increase of ML models with manual preprocessing combined with RST was up-to 18% balanced accuracy (BACC) compared to models without RST. The ML models with "exp-keep" and "pref-keep" instructions showed highest performance increase of + 18% BACC (glioma), + 6% BACC (prostate) and + 3% BACC (DLBCL) compared to other models across all datasets. The study demonstrated the added value of RST in predictive performance of oncology-specific ML models, hence, serving as proof of concept of a more inclusive clinician-driven DP process in future studies.

Evaluation of Etiologic Heterogeneity for Risk of Diffuse Large B-Cell Lymphoma Subtype Defined by Cell-of-Origin.

Diffuse large B-cell lymphoma (DLBCL) is clinically heterogeneous, and gene expression profiling has identified at least two biologically distinct DLBCL subtypes defined by their cell-of-origin (COO): germinal center B-cell (GCB) and activate B-cell (ABC) or non-GCB. We evaluated a variety of putative DLBCL risk factors for etiologic heterogeneity by COO in a clinic-based study of newly diagnosed DLBCL cases (N=638) and frequency-matched controls (N=2253). COO was determined on formalin-fixed, paraffin-embedded tumor tissue, with DLBCL classified as GCB (N=283), non-GCB (N=188), or undetermined/missing (N=167; mainly due to lack of tissue). Multivariable logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs). We identified heterogeneity by COO for low socioeconomic status (SES), which was only associated with non-GCB DLBCL (OR=1.88 for low vs average SES, 95%CI 1.08-3.27); alcohol use, which was only associated with GCB DLBCL (OR=0.48 for former drinkers, 95%CI 0.29-0.80; OR=0.47 for current drinkers, 95%CI 0.32-0.71); and borderline heterogeneity for regular use of regular/extra-strength aspirin, which was only associated with non-GCB DLBCL (OR=0.36, 95%CI 0.16-0.85). In contrast, there was no significant heterogeneity by COO for family history, medical history, or other lifestyle factors. While requiring confirmation, most risk factors for DLBCL did not show etiologic heterogeneity by COO, with some notable exceptions including alcohol use, SES, and perhaps regular use of regular/extra-strength aspirin. Mechanistically, these findings suggest that most of the DLBCL risk factors evaluated here influence lymphomagenesis prior to differentiation into COO subtypes, with selected factors acting later.