Ketone Bodies Are Potential Prognostic Biomarkers in Relapsed/Refractory Diffuse Large B-Cell Lymphoma: Results from the R2-GDP-GOTEL Trial

Academic Editor: Alexandar Tzankov Received: 3 December 2024 Revised: 10 January 2025 Accepted: 16 January 2025 Published: 5 February 2025 Citation: Fernández-Castillejo, S.; Badia, J.; de la Cruz-Merino, L.; Martín Garcia-Sáncho, A.; CarniceroGonzález, F.; Palazón-Carrión, N.; Ríos-Herranz, E.; de la Cruz-Vicente, F.; Rueda-Domínguez, A.; MartínezBanaclocha, N.; et al. Ketone Bodies Are Potential Prognostic Biomarkers in Relapsed/Refractory Diffuse Large B-Cell Lymphoma: Results from the R2-GDP-GOTEL Trial. Cancers 2025, 17, 532. https://doi.org/10.3390/ cancers17030532 Copyright: © 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/ licenses/by/4.0/). Article Ketone Bodies Are Potential Prognostic Biomarkers in Relapsed/ Refractory Diffuse Large B-Cell Lymphoma: Results from the R2-GDP-GOTEL Trial Sara Fernández-Castillejo 1,2,† , Joan Badia 1,2,† , Luís de la Cruz-Merino 3,4 , Alejandro Martín Garcia-Sáncho 5,6 , Fernando Carnicero-González 7 , Natalia Palazón-Carrión 3,4 , Eduardo Ríos-Herranz 8 , Fátima de la Cruz-Vicente 9 , Antonio Rueda-Domínguez 10 , Natividad Martínez-Banaclocha 11 , José Gómez-Codina 12, Jorge Labrador 13 , Francisca Martínez-Madueño 1,2,14, Núria Amigó 14,15 , Antonio Salar-Silvestre 16, Delvys Rodríguez-Abreu 17, Laura Gálvez-Carvajal 10 , Margarita Sánchez-Beato 18 , Mariano Provencio-Pulla 19 , Maria Guirado-Risueño 20 , Esteban Nogales 3,4, Víctor Sánchez-Margalet 21 , Carlos Jiménez-Cortegana 21 , Guillermo Rodríguez-García 9, Raquel Cumeras 1,22,* and Josep Gumà1,2,14 on behalf of the Spanish Lymphoma Oncology Group (GOTEL) 1 Translational, Epidemiological and Clinical Oncological Research Group (GIOTEC), Department of Oncology, Institut d’Investigació Sanitària Pere Virgili (IISPV), 43204 Reus, Tarragona, Spain; [email protected] (S.F.-C.); [email protected] (J.B.); [email protected] (F.M.-M.); [email protected] (J.G.) 2Institut d’Oncologia de la Catalunya Sud (IOCS), Hospital Universitari Sant Joan de Reus, 43204 Reus, Tarragona, Spain 3Cancer Immunotherapy Group, Oncohematology and Genetics Department, Biomedicine Institute of Seville (IBIS)/CSIC, 41013 Seville, Spain; [email protected] (L.d.l.C.-M.); [email protected] (N.P.-C.); [email protected] (E.N.) 4 Department of Clinical Oncology, University Hospital Virgen Macarena and School of Medicine, University of Sevilla, 41013 Sevilla, Spain 5Department of Hematology, Hospital Universitario de Salamanca, Instituto de Investigación Biomédica de Salamanca (IBSAL), Universidad de Salamanca, 37007 Salamanca, Spain; [email protected] 6CIBER de Cáncer (CIBERONC), Institute of Health Carlos III, 28029 Madrid, Spain 7Department of Hematology, San Pedro de Alcántara Hospital, 10003 Cáceres, Spain; fcarnicer[email protected] 8Department of Hematology, Hospital Universitario Virgen de Valme, 41014 Sevilla, Spain; [email protected] 9Department of Hematology, Hospital Universitario Virgen del Rocío, 41013 Sevilla, Spain; [email protected] (F.d.l.C.-V.); [email protected] (G.R.-G.) 10 Department of Clinical Oncology. Hospital Universitario Virgen de la Victoria, 29010 Málaga, Spain; [email protected] (A.R.-D.); [email protected] (L.G.-C.) 11 Department of Oncology, Dr. Balmis General University Hospital, Alicante Institute for Health and Biomedical Research (ISABIAL), 03010 Alicante, Spain; [email protected] 12 Department of Clinical Oncology, Hospital Universitario y Politécnico La Fe, 46026 Valencia, Spain; [email protected] 13 Department of Hematology, Hospital Universitario de Burgos, 09006 Burgos, Spain; [email protected] 14 Faculty of Medicine and Health Sciences, Universitat Rovira i Virgili (URV), 43201 Reus, Tarragona, Spain; [email protected] 15 Biosfer Teslab, 43206 Reus, Tarragona, Spain 16 Department of Hematology, Hospital del Mar, 08003 Barcelona, Spain; [email protected] 17 Department of Clinical Oncology, Hospital Universitario Insular de Gran Canaria, 35016 Las Palmas de Gran Canaria, Las Palmas, Spain; [email protected]g 18 Lymphoma Research Group, Department of Medical Oncology, Hospital Universitario Puerta de Hierro-Majadahonda, IDIPHISA, 28222 Majadahonda, Madrid, Spain; [email protected]g 19 Department of Clinical Oncology, Hospital Universitario Puerta De Hierro-Majadahonda, IDIPHISA, 28222 Majadahonda, Madrid, Spain; [email protected] 20 Department of Clinical Oncology, Hospital Universitario de Elche, 03203 Elche, Alicante, Spain; [email protected] 21 Medical Biochemistry and Molecular Biology and Immunology, Hospital Universitario Virgen de la Macarena, 41009 Sevilla, Spain; [email protected] (V.S.-M.); [email protected] (C.J.-C.) 22 Department of Electrical and Automatic Electronic Engineering, Universitat Rovira i Virgili (URV), 43002 Tarragona, Spain *Correspondence: [email protected] †These authors contributed equally to this work. Cancers 2025,17, 532 https://doi.org/10.3390/cancers17030532 Cancers 2025,17, 532 2 of 16 Simple Summary: Patients with relapsed or refractory diffuse large B-cell lymphoma (DLBCL) have poor outcomes and limited treatment options. A phase II trial by GOTEL evaluated the R2-GDP regimen (combination of lenalidomide, rituximab, gemcitabine, dexamethasone, and cisplatin), demonstrating feasibility and effectiveness. Baseline serum metabolomic analysis of 69 patients enrolled in the trial identified two independent metabolites, 3-hydroxybutyrate (3OHB) and acetone, as being significantly associated with overall survival and progression-free survival. Elevated 3OHB levels (>141 µ M) were specific to the ABC subtype of DLBCL, while acetone levels were elevated in both types of DLCBL but more pronounced in ABC cases. These biomarkers, irrespective of sex, age, and BMI, could help predict outcomes and guide treatment strategies in relapsed/refractory DLBCL. Abstract: Background: Patients with relapsed or refractory (R/R) diffuse large B-cell lymphoma (DLBCL) who are ineligible for high-dose chemotherapy have limited treatment options and poor life expectancy. The purpose of this study is to identify a serum metabolomic profile that may be predictive of outcome in patients with R/R-DLBCL. Methods: This study included 69 R/R DLBCL patients from the R2-GDP-GOTEL trial (EudraCT 2014-001620-299). Serum samples were collected at baseline, and the mean length of follow-up was 41 months. Serum metabolites were analyzed by nuclear magnetic resonance (NMR). Metabolites were correlated with treatment response, progression-free survival (PFS), and overall survival (OS). Results: Serum levels of 3-hydroxybutyrate (3OHB) and acetone were significantly (p< 0.001) associated with PFS (3OHB: hazard ratio [HR] 7.7, 95% confidence interval [CI] 2.5–24.1; acetone: HR 9.32, 95% CI 2.75–31.6) and OS (3OHB: HR 9.32, 95% CI 2.75–31.6; acetone: HR 1.92, 95% CI 1.36–2.69). Serum values of 141 µ M for 3OHB and 40 µ M for acetone were the optimal cutoffs associated with the survival outcomes. Elevated 3OHB levels (>141 µ M) were specific to the ABC subtype of DLBCL, while acetone levels were elevated in both types of DLCBL but more pronounced in ABC cases. In a multivariate survival analysis, including the International Prognostic Index (IPI) score and refractoriness status (R/R), 3OHB and acetone remained significant. To aid oncologists employing the R2-GDP regime, we constructed PFS and OS nomograms for R/R-DLBCL risk stratification, incorporating 3OHB levels or acetone levels, IPI score, and refractoriness status. The nomogram with 3OHB and refractoriness status showed a time-dependent AUC of 0.86 for 6-month PFS and 0.84 for 12-month OS. These nomograms provide a comprehensive tool for individualized risk assessment and treatment optimization. Conclusions: The ketone bodies 3OHB and acetone are potential prognostic biomarkers of poor outcome in R/R DLBCL patients treated with the R2-GDP regimen, independently of IPI score and chemorefractoriness status. Keywords: ketone bodies; diffuse large B-cell lymphoma (DLBCL); relapsed/refractory lymphoma; 3-hydroxybutyrate; 3OHB; acetone; prognostic biomarkers; metabolomics 1. Introduction Approximately 60% of patients with diffuse large B-cell lymphomas (DLBCL) are cured using upfront therapy with the CHOP-R regimen or other anthracycline and rituximabbased chemotherapies, whereas the remaining 40% are refractory or relapsed (R/R) following first-line chemotherapy. At the time this clinical trial was conducted, the standard treatment for R/R DLBCL patients was second-line conventional chemotherapy followed by consolidation with high-dose chemotherapy in chemosensitive patients. Palliative chemotherapy is an option for patients who are unable to receive high-dose chemotherapy Cancers 2025,17, 532 3 of 16 or CAR-T therapy; nevertheless, the best therapeutic option may involve enrolling the patient in investigational clinical trials. The identification of biomarkers capable of predicting the outcome of DLBCL patients has been the focus of increasing interest due to the marked genetic and molecular heterogeneity that underlies disease aggressiveness and tumor progression. Metabolomics is a powerful tool that can identify cancer biomarkers and drivers of tumorigenesis. In the field of lymphomas, different studies have evaluated untargeted metabolomics using gas (GC) or liquid chromatography (LC) coupled with mass spectrometry (MS) in patients with lymphoid neoplasms and healthy populations [ 1 – 5 ]. Although most studies showed metabolomic differences between patients and healthy subjects, the identification of differential metabolites has been inconsistent. This is likely due to differences in laboratory techniques (GC-MS and LC-MS) and/or biological samples (blood, urine, and feces). However, the metabolomic profile in patients with DLBCL as a prognostic factor for survival has been evaluated in only two studies. Stenson et al. [ 6 ] used nuclear magnetic resonance (NMR) spectroscopy in 87 patients with DLBCL prior to starting first-line treatment with chemoimmunotherapy. Statistically significant differences were found in the metabolomic profile between patients who achieved a complete response and long survival and those who were refractory to treatment or relapsed. Patients who had been cured showed higher levels of aspartate, valine, ornithine, and pyroglutamate, whereas R/R patients had higher concentrations of lysine, arginine, cadaverine, and 2-hydroxybutyrate. In a second study, Mi et al. [ 7 ] used GC-MS to assess pre-treatment serum samples from 80 DLBCL patients and reported that higher levels of pyroglutamic and hexadecenoic acids and lower levels of valine were associated with significantly higher overall survival. The Spanish Group for the Treatment and Study of Lymphomas (GOTEL) conducted a phase II clinical trial to evaluate the combination of lenalidomide with R-GDP (rituximab, gemcitabine, dexamethasone, and cisplatin) in patients with R/R DLBCL who were either unsuitable for high-dose chemotherapy or whose treatment had not worked. A total of 78 patients were included in the R2-GDP-GOTEL study, and after a median follow-up of 37 months, 7.9% of patients were still alive without progression at 24 months [ 8 ]. Taking advantage of baseline data from this clinical trial, the present study was designed to identify a serum metabolomic profile that may be predictive of outcomes in patients with R/R DLBCL treated with the R2-GDP combination. 2. Materials and Methods 2.1. Study Design and Patients The R2-GDP-GOTEL clinical trial was a phase II, multicenter, open-label, and singlearm study carried out in 78 R/R DLBCL patients who were treated in 18 Spanish hospitals between April 2015 and September 2018. Briefly, the R2-GDP regimen included an induction treatment with a combination of lenalidomide, rituximab, gemcitabine, dexamethasone, and cisplatin (R2-GDP), for up to 6 cycles (every 3 weeks), followed by maintenance with lenalidomide for up to 24 months, unless there was a progression, unacceptable toxicity, or voluntary withdrawal [ 8 ]. Additional information about the dropout rate (attrition rate) and a power analysis of the R2-GDP trial have been previously reported [ 8 ]. Since this trial has a single-arm design, randomization and blinding do not apply. Eligible participants were patients with chemorefractory or relapsed DLBCL unsuitable for high-dose chemotherapy, with an Eastern Cooperative Oncology Group (ECOG) performance status of 0–1, and who had previously received at least one line of immunochemotherapy, including rituximab. Patients with leptomeningeal or central nervous system (CNS) involvement, and those with hematological, renal, or liver dysfunction, were excluded from the initial R2-GDP-GOTEL clinical trial. Since all the lymphomas included in this study were DLBCL, all the included Cancers 2025,17, 532 4 of 16 patients expressed CD45 and the pan B-cell markers CD19, CD20, CD22, and CD79a in the initial diagnostic immunohistochemical study. The objective of the present substudy was to assess the baseline metabolomic profile of R/R DLBCL patients and to identify serum metabolites predictive of outcome. For that purpose, 69 patients from the R2-GDP-GOTEL clinical trial for whom a sufficient blood sample was available for metabolomic profiling were included in the metabolomic profiling substudy. 2.2. Sample Preparation and Metabolomic Profiling A comprehensive methodological approach for high-throughput screening by NMR spectroscopy using a Bruker Avance 600 MHz NMR spectrometer (Bruker BioSpin, Ettlingen, Germany ) was used to analyze a broad spectrum of metabolites in serum samples [ 9 ]. The analysis included the lipoprotein, glycoprotein, and metabolite profiles from intact serum, in addition to the lipid profile from lipid serum extracts obtained by a biphasic extraction with methyl tert-butyl ether (MTBE). Lipid extracts were dried and reconstituted in 0.01% tetramethylsilane (TMS) solution (0.067 mM) and deuterated solvents. All analyses were carried out at Biosfer Teslab (Reus, Tarragona, Spain). In addition, 1D Nuclear Overhauser Effect Spectroscopy (NOESY) was used to characterize small molecules such as amino acids and small carbohydrates, while larger molecules like lipoproteins and glycoproteins were detected using LED Diffusion (Diff) experiments. All the sequences ran at 37 ◦C in quantitative conditions. Samples were coded to maintain the subjects’ anonymity. The Liposcale ® test (IVD-CE) was used to determine the lipid composition, particle size, and concentration of major lipoprotein classes as well as the particle concentration of nine subclasses [ 10 ]. Circulating glycoproteins were obtained by deconvoluting the specific NMR spectral regions and quantifying areas correlating to the concentration of the acetyl groups of N-acetylglucosamine and N-acetyl galactosamine (GlycA) and acetyl groups of N-acetylneuraminic acid (GlycB). A Carr–Purcell–Meiboom–Gill (CPMG) filter on the 1H-NMR spectra was used to profile and absolutely quantify the metabolomics profile. The BUME protocol [ 11 ] was used for lipid quantification, based on the Lipspin software [ 12 ]. Succinctly, we used lineshape fitting analysis of spectral regions to quantify the lipids. 2.3. Endpoints The primary endpoints were the characterization of the baseline metabolomic profile in R/R DLBCL patients according to response to treatment and outcome. Key secondary endpoints were to determine the predictive performance of nomograms for R/R DLBCL risk stratification based on the metabolites identified and their optimal cutoffs associated with the outcome. Tumor response was evaluated according to the International Working Group Criteria [ 13 ] using computed tomography (CT) after the third induction cycle and positron emission tomography (PET) in the following 4 weeks after the last cycle of the induction phase. Outcome included progression-free survival (PFS) and overall survival (OS). PFS was defined as the time between the first dose of the R2-GDP schedule to the progression of disease or death, and OS was defined as the period from the first dose of the R2-GDP schedule to death from any cause. 2.4. Statistical Analysis Statistical analyses were performed with R software using the R Stats Package (v.4.3.2.), survival (v.3.5-7), survminer (v.0.4.9), maxstat (v.0.7-25), caret (v.6.0-94), rms (v.6.7-1), and survivalROC (v.1.0.3.1). The analyses were restricted to metabolites identified in >90% of patients. Metabolite missing values were not imputed, and metabolite concentrations were only scaled in multivariate analyses. For reproducibility, the random seed was set at 123 via the set.seed function (R base). Univariate statistics was performed using the non-parametric Cancers 2025,17, 532 5 of 16 Mann–Whitney–Wilcoxon statistical test. Survival analysis was carried out using the nonparametric Kaplan–Meier method and the log-rank test for the comparison of survival curves. Cutoffs for numerical variables were calculated using the maximally selected rank statistics, with a 10% minimum proportion of observations, and missing values were assigned to the lowest group. Cox regression analyses were run twice: once to search for statistically significant individual prognostic metabolites, and again to confirm whether they remained statistically significant in a multivariate analysis with other statistically significant clinical prognostic factors, such as the International Prognostic Index (IPI) [ 14 ] and refractoriness status. The hazard ratio (HR) and the 95% confidence interval (CI) were calculated. Statistical significance was set at p< 0.05, and a false discovery rate (FDR) pvalue adjustment was applied when necessary. Fold change (FC) analyses were used to compare the absolute value of change in the means of each metabolite between responders and non-responders. The nomogram for R/R DLBCL risk stratification incorporated the significant metabolites identified in the univariate Cox analysis and the significant clinical risk factors within a Cox proportional hazards framework. A 70% train split was used for model fitting via bootstrap calibration (B = 1000), while the remaining 30% test split was used for predictive performance evaluation with a time-dependent receiver operating characteristic curve (ROC) and area under the ROC (AUC). The time points selected were 6 months for PFS and 12 months for OS. Pearson’s product-moment correlation coefficient (r) was used to assess the relationship between the metabolites identified. 3. Results 3.1. Baseline Characteristics of Patients and Survival Sixty-nine participants in the R2-GDP-GOTEL clinical trial (35 men and 34 women) with a median age of 70 years were included in the metabolomic profiling study. Key baseline data are displayed in Table 1. Activated B-cell-like (ABC) lymphomas, relapsed DLBCL, and IPI low/high-intermediate risk category (0–3) were the most common characteristics. Table 1. Key characteristics of the study population. Variables 1Total Patients (n= 69) Sex, n(%) Men 35 (50.7%) Women 34 (49.3%) Age, years, median (IQR) 70.1 (61.7–74.5) Body mass index (BMI), kg/m2, median (IQR) 27.0 (23.9–30.2) Cell-of-origin (CoO), n(%) [n= 65] Germinal center B-cell-like 27 (41%) Activated B-cell-like 39 (59%) Refractoriness, n(%) Relapsed 40 (58.0%) Chemorefractory 29 (42.0%) International Prognostic Index (IPI) score, n(%) Low risk (0–1) 12 (17.4%) Low/High-intermediate risk (2–3) 37 (53.6%) High risk (4–5) 20 (29.0%) Response, n(%) Responders 41 (59.4%) Non-responders 28 (40.6%) 1IQR: interquartile range (25th–75th percentile). In this subset of patients, the overall response to R2-GPD treatment was 59.4% ( n= 41 ) (complete response 37.7%, partial response 21.7%), similar to those previously reported in the whole R2-GDP population [ 8 ]. The individual clinical evolution of the 69 patients Cancers 2025,17, 532 6 of 16 is presented in Supplementary Figure S1. After a median follow-up of 41 months, the median PFS was 5 months (36%, 16%, and 7.9% at 6, 12, and 24 months, respectively) and the median OS was 12 months (66%, 47%, and 36% at 6, 12, and 24 months, respectively) (Supplementary Figure S2A,B). Survival analyses stratified by cell-of-origin (CoO) showed a median PFS of 6.0 months (95% CI 3.0–11) for the germinal center B-cell-like (GBC) subtype and 5.0 months (95% CI 2–6) for ABC (p= 0.099) (Figure S2C), similar to those previously reported in the whole R2-GDP population [ 8 ]. The median OS was 16 months (95% CI 6.2—not reached) for GBC and 9.3 months (95% CI 5.3–24) for ABC (p= 0.59) (Figure S2D). Concerning DLBCL status, patients with chemorefractory disease showed a median PFS of 3 months (95% CI 2–6), whereas those with relapsed disease had a median PFS of 6 months (95% CI 4–9) (p= 0.083) (Figure S2E). The OS was significantly longer for the relapsed group (median 24 months, 95% CI 12—not reached) than for the chemorefractory group (median 6.2 months, 95% CI 4.5–13) (p= 0.0034) (Figure S2F). Patients in the low/high-intermediaterisk IPI category showed a median PFS of 6 months (95% CI 4–9), whilst patients in the high-risk category had a median PFS of 2 months (95% CI 2–7) (p= 0.0044) (Figure S2G). The median OS was also longer for the low/high-intermediate-risk IPI category (median 15 months, 95% CI 9.3–33) compared with the high-risk IPI category (median 3.5 months, 95% CI 1.8—not reached) (p= 0.074) (Figure S2H). 3.2. Treatment Response Metabolomic Profile A total of 66 metabolites were identified, 7 of which were excluded as they were detected in less than 90% of the patients. In Supplementary Table S1, the mean values for the treatment response of the 59 included metabolites are presented. Twelve metabolites were significantly different (p< 0.05) in responders vs. non-responders, including creatinine, lactate, glycoprotein A, free cholesterol, esterified cholesterol, intermediate-density lipoprotein (IDL) cholesterol, IDL triglycerides, high-density lipoprotein (HDL) triglycerides, medium and large low-density lipoproteins (LDLs), and medium and large HDLs. All metabolites except for creatinine showed significantly lower mean values in responders to R2-GPD treatment than in non-responders. By contrast, creatinine was significantly higher in responders than in non-responders (Table 2). The box plots for significant metabolite among responders and non-responders are shown in Supplementary Figure S3. Table 2. Significant serum metabolites in responders and non-responders. Metabolites 1Responders (n= 41) Non-Responders (n= 28) pValue Fold Change 2 Low molecular weight Creatinine, µM, mean (SD) 73.0 (43.3) 48.7 (17.6) 0.030 1.50 Lactate, µM, mean (SD) 1174.2 (989.7) 2023.5 (1655.6) 0.034 0.58 Glycoproteins Glycoprotein A, µmol/L, mean (SD) 846.0 (165.3) 915.5 (161.0) 0.028 0.92 Glycoprotein A, H/W ratio, mean (SD) 24.7 (4.3) 27.04 (4.7) 0.024 0.92 Cholesterol Free cholesterol, mmol/L, mean (SD) 2.36 (0.71) 2.84 (0.83) 0.016 0.83 Esterified cholesterol, mmol/L, mean (SD) 4.76 (1.20) 5.36 (1.40) 0.043 0.89 Lipoproteins IDL cholesterol, mg/dL, mean (SD) 14.14 (5.84) 17.39 (5.61) 0.015 0.86 IDL triglycerides, mg/dL, mean (SD) 13.38 (4.59) 15.64 (4.07) 0.018 0.81 LDL triglycerides, mg/dL, mean (SD) 18.94 (6.40) 22.90 (6.08) 0.005 0.83 Medium LDL-P, nmol/L, mean (SD) 340.83 (134.98) 424.09 (157.02) 0.008 0.80 Large LDL-P, nmol/L, mean (SD) 180.15 (44.78) 201.32 (51.39) 0.043 0.89 Medium HDL-P, µmol/L, mean (SD) 10.18 (1.68) 11.27 (1.72) 0.006 0.90 Large HDL-P, µmol/L, mean (SD) 0.30 (0.05) 0.33 (0.05) 0.005 0.90 1 SD: standard deviation; H/W: height/width of the NMR peak; IDL: intermediate-density lipoprotein; LDL: low-density lipoprotein; HDL: high-density lipoprotein; P: particle number. 2 Fold change (FC) > 1 indicates an increase in metabolite concentration and those with FC < 1 indicate a decrease in responders vs. non-responders. Cancers 2025,17, 532 7 of 16 3.3. Survival Outcome and Metabolomic Profile Two metabolites, 3OHB and acetone, were significantly associated with survival outcomes in the Cox univariate regression analysis. Higher serum concentrations of 3OHB or acetone were statistically significant (p< 0.001) prognostic factors for a worse PFS (3OHB: hazard ratio [HR] = 7.7; acetone: HR = 1.83) (Figure 1A) and OS (3OHB: HR = 9.32; acetone: HR = 1.92) (Figure 1B). Both 3OHB and acetone were independent predictors that were significantly associated with PFS and OS in the multivariate Cox regression model, which included IPI risk categories and refractoriness status (R/R) as classical DLBCL clinical prognostic factors (Figure 1C–F). In the overall study population, the mean serum levels of 3OHB were 120.6 µ M (range 0–1455.9 µ M) (n= 67) and the mean levels of acetone 40.48 µM (range 0–534.2 µM) (n= 68). Cancers 2025, 17, x FOR PEER REVIEW 8 of 18 Figure 1. 3-Hydroxybutyrate and acetone as prognostic metabolites in the Cox univariate analysis (A,B) and multivariate regressions models (C–F) for progression-free survival (PFS) (A,C,E) and overall survival (OS) (B,D,F). The optimal cutoffs for serum 3OHB and acetone were identified with an outcomeoriented method (see Methods) and were set at 141 µM and 40 µM, respectively. Following the application of the cutoff, the PFS and OS Kaplan–Meier survival curves were significantly different. For serum 3OHB, the median PFS was 5 months (95% CI 3–8) vs. 2 months (95% CI 1.0—not reached) (p = 0.044) (Figure 2A). The corresponding values for OS were 13 months (95% CI 8.9–33) vs. 3.2 months (95% 2.0—not reached) (p = 0.0035), Figure 1. 3-Hydroxybutyrate and acetone as prognostic metabolites in the Cox univariate analysis (A,B) and multivariate regressions models (C–F) for progression-free survival (PFS) (A,C,E) and overall survival (OS) (B,D,F). Cancers 2025,17, 532 8 of 16 The optimal cutoffs for serum 3OHB and acetone were identified with an outcomeoriented method (see Methods) and were set at 141 µ M and 40 µ M, respectively. Following the application of the cutoff, the PFS and OS Kaplan–Meier survival curves were significantly different. For serum 3OHB, the median PFS was 5 months (95% CI 3–8) vs. 2 months (95% CI 1.0—not reached) (p= 0.044) (Figure 2A). The corresponding values for OS were 13 months (95% CI 8.9–33) vs. 3.2 months (95% 2.0—not reached) (p= 0.0035), respectively (Figure 2B). Differences in PFS for serum acetone were 5 months (95% CI 3–9) vs. 2 months (95% CI 1.0—not reached) (p= 0.0054) (Figure 2C) and 15 months (95% CI 9–33) vs. 2.5 months (95% CI 1.8—not reached) (p= 0.00014) for OS, respectively (Figure 2D). Significant differences were also observed in univariate Cox regressions following the application of the cutoff. For serum 3OHB, the HR for PFS was 2.21, while the HR for OS was 3.21. For serum acetone, the HR for PFS was 2.76 and the HR for OS was 3.71 (Supplementary Figure S4). Cancers 2025, 17, x FOR PEER REVIEW 9 of 18 respectively (Figure 2B). Differences in PFS for serum acetone were 5 months (95% CI 3– 9) vs. 2 months (95% CI 1.0—not reached) (p = 0.0054) (Figure 2C) and 15 months (95% CI 9–33) vs. 2.5 months (95% CI 1.8—not reached) (p = 0.00014) for OS, respectively (Figure 2D). Significant differences were also observed in univariate Cox regressions following the application of the cutoff. For serum 3OHB, the HR for PFS was 2.21, while the HR for OS was 3.21. For serum acetone, the HR for PFS was 2.76 and the HR for OS was 3.71 (Supplementary Figure S4). Figure 2. Kaplan–Meier survival curves. (A,B), progression-free survival (PFS) and overall survival (OS) for the cutoff of 3-hydroxybutyrate (3OHB); (C,D) PFS and OS for the cutoff of acetone. 3.4. Nomogram for R/R DLBCL Risk Stratification Nomograms for risk stratification in patients with R/R DLBCL based on serum metabolites 3OHB and acetone showed a high predictive performance, slightly more favorable for 3OHB, with an AUC of 0.856 for 6-month PFS (Figure 3A,B) and 0.844 for 12-month OS (Figure 3C,D). Serum acetone showed a similar predictive performance, with an AUC of 0.840 for 6-month PFS (Figure 3D,E) and 0.830 for 12-month OS (Figure 3F,G). A nomogram that included both 3OHB and acetone was also explored; however, it performed worse (data not shown) since both metabolites showed a statistically significant correlation (r = 0.76, p < 0.001) (Supplementary Figure S5). Let us consider the OS nomogram of 3OHB for two patients: Patient 1 and Patient 2. Although both patients have an IPI score of 4–5 (3 points) and are both refractory (18 points), Patient 1 has a 3OHB concentration of 800 µM (50 points), whereas Patient 2 has a 3OHB concentration of 50 µM (3 points). Hence, for the 3OHB OS nomogram, Patient 1 scored 71 points (<10% survival probability at 12 months), while Patient 2 scored 24 points (30% survival probability at 12 months). Figure 2. Kaplan–Meier survival curves. (A,B), progression-free survival (PFS) and overall survival (OS) for the cutoff of 3-hydroxybutyrate (3OHB); (C,D) PFS and OS for the cutoff of acetone. 3.4. Nomogram for R/R DLBCL Risk Stratification Nomograms for risk stratification in patients with R/R DLBCL based on serum metabolites 3OHB and acetone showed a high predictive performance, slightly more favorable for 3OHB, with an AUC of 0.856 for 6-month PFS (Figure 3A,B) and 0.844 for 12-month OS (Figure 3C,D). Serum acetone showed a similar predictive performance, with an AUC of 0.840 for 6-month PFS (Figure 3D,E) and 0.830 for 12-month OS (Figure 3F,G). A nomogram that included both 3OHB and acetone was also explored; however, it performed worse (data not shown) since both metabolites showed a statistically significant correlation ( r= 0.76 , p< 0.001 ) (Supplementary Figure S5). Let us consider the OS nomogram of 3OHB for two patients : Patient 1 and Patient 2. Although both patients have an IPI score of 4–5 Cancers 2025,17, 532 9 of 16 ( 3 points ) and are both refractory (18 points), Patient 1 has a 3OHB concentration of 800 µM (50 points), whereas Patient 2 has a 3OHB concentration of 50 µ M (3 points). Hence, for the 3OHB OS nomogram, Patient 1 scored 71 points (<10% survival probability at 12 months), while Patient 2 scored 24 points (30% survival probability at 12 months). Cancers 2025, 17, x FOR PEER REVIEW 10 of 18 Figure 3. Nomograms for R/R DLBCL risk stratification based on single metabolites and their respective time-dependent ROCs. (A,B) 3-Hydroxybutyrate (3OHB) nomogram for progression-free survival (PFS) and its ROC; (C,D) 3OHB nomogram for overall survival (OS) and its ROC; (E,F) acetone nomogram for PFS and its ROC; (G,H) acetone nomogram for OS and its ROC. 4. Discussion The present data contribute to defining the metabolomic profile linked to R/RDLBCL patients’ responsiveness to treatment and survival. To the best of our knowledge, no previous metabolomics study has been reported in this setting. Treatment respondents Figure 3. Nomograms for R/R DLBCL risk stratification based on single metabolites and their respective time-dependent ROCs. (A,B) 3-Hydroxybutyrate (3OHB) nomogram for progressionfree survival (PFS) and its ROC; (C,D) 3OHB nomogram for overall survival (OS) and its ROC; (E,F) acetone nomogram for PFS and its ROC; (G,H) acetone nomogram for OS and its ROC. Cancers 2025,17, 532 16 of 16 28. Feng, S.; Wang, H.; Liu, J.; AA, J.; Zhou, F.; Wang, G. Multi-Dimensional Roles of Ketone Bodies in Cancer Biology: Opportunities for Cancer Therapy. Pharmacol. Res. 2019,150, 104500. [CrossRef] 29. Huang, D.; Li, T.; Wang, L.; Zhang, L.; Yan, R.; Li, K.; Xing, S.; Wu, G.; Hu, L.; Jia, W.; et al. Hepatocellular Carcinoma Redirects to Ketolysis for Progression under Nutrition Deprivation Stress. Cell Res. 2016,26, 1112–1130. [CrossRef] 30. Zhang, J.; Chen, B.; Zhang, C.; Zhu, M.; Fan, Z.; Li, L.; Wang, J.; Jin, J. The Prognostic Value of Hydroxybutyrate Dehydrogenase in Diffuse Large B Cell Lymphoma. iScience 2024,27, 110905. [CrossRef] [PubMed] 31. Caro, P.; Kishan, A.U.; Norberg, E.; Stanley, I.; Chapuy, B.; Ficarro, S.B.; Polak, K.; Tondera, D.; Gounarides, J.; Yin, H.; et al. Metabolic signatures uncover distinct targets in molecular subsets of diffuse large B cell lymphoma. Cancer Cell 2012,22, 547–560. [CrossRef] [PubMed] 32. Penning, T.M.; Jonnalagadda, S.; Trippier, P.C.; Rižner, T.L.; Gottesman, M. Aldo-Keto Reductases and Cancer Drug Resistance. Pharmacol. Rev. 2021,73, 1150–1171. [CrossRef] [PubMed] 33. Dedkova, E.N.; Blatter, L.A. Role of β -Hydroxybutyrate, Its Polymer Polyβ -Hydroxybutyrate and Inorganic Polyphosphate in Mammalian Health and Disease. Front. Physiol. 2014,5, 101953. [CrossRef] 34. Huang, C.K.; Chang, P.H.; Kuo, W.H.; Chen, C.L.; Jeng, Y.M.; Chang, K.J.; Shew, J.Y.; Hu, C.M.; Lee, W.H. Adipocytes Promote Malignant Growth of Breast Tumours with Monocarboxylate Transporter 2 Expression via β -Hydroxybutyrate. Nat. Commun. 2017,8, 14706. [CrossRef] [PubMed] 35. Skorupa, A.; Po´nski, M.; Ciszek, M.; Cicho´n, B.; Klimek, M.; Witek, A.; Pakuło, S.; Boguszewicz, Ł.; Sokół, M. Grading of Endometrial Cancer Using 1H HR-MAS NMR-Based Metabolomics. Sci. Rep. 2021,11, 18160. [CrossRef] [PubMed] 36. Baranovicova, E.; Racay, P.; Zubor, P.; Smolar, M.; Kudelova, E.; Halasova, E.; Dvorska, D.; Dankova, Z. Circulating Metabolites in the Early Stage of Breast Cancer Were Not Related to Cancer Stage or Subtypes but Associated with Ki67 Level. Promising Statistical Discrimination from Controls. Mol. Cell Probes 2022,66, 101862. [CrossRef] [PubMed] 37. Gouirand, V.; Gicquel, T.; Lien, E.C.; Jaune-Pons, E.; Da Costa, Q.; Finetti, P.; Metay, E.; Duluc, C.; Mayers, J.R.; Audebert, S.; et al. Ketogenic HMG-CoA Lyase and Its Product B-hydroxybutyrate Promote Pancreatic Cancer Progression. EMBO J. 2022, 41, e110466. [CrossRef] 38. Ni, Y.; Xie, G.; Jia, W. Metabonomics of Human Colorectal Cancer: New Approaches for Early Diagnosis and Biomarker Discovery. J. Proteome Res. 2014,13, 3857–3870. [CrossRef] [PubMed] 39. Zhang, L.; Jin, H.; Guo, X.; Yang, Z.; Zhao, L.; Tang, S.; Mo, P.; Wu, K.; Nie, Y.; Pan, Y.; et al. Distinguishing Pancreatic Cancer from Chronic Pancreatitis and Healthy Individuals by 1H Nuclear Magnetic Resonance-Based Metabonomic Profiles. Clin. Biochem. 2012,45, 1064–1069. [CrossRef] 40. Yang, D.O.; Xu, J.; Huang, H.; Chen, Z. Metabolomic Profiling of Serum from Human Pancreatic Cancer Patients Using 1H NMR Spectroscopy and Principal Component Analysis. Appl. Biochem. Biotechnol. 2011,165, 148–154. [CrossRef] [PubMed] 41. Alfaifi, A.; Bahashwan, S.; Alsaadi, M.; Malhan, H.; Aqeel, A.; Al-Kahiry, W.; Almehdar, H.; Qadri, I. Metabolic Biomarkers in B-Cell Lymphomas for Early Diagnosis and Prediction, as Well as Their Influence on Prognosis and Treatment. Diagnostics 2022, 12, 394. [CrossRef] [PubMed] 42. Alfaifi, A.; Refai, M.Y.; Alsaadi, M.; Bahashwan, S.; Malhan, H.; Al-Kahiry, W.; Dammag, E.; Ageel, A.; Mahzary, A.; Albiheyri, R.; et al. Metabolomics: A New Era in the Diagnosis or Prognosis of B-Cell Non-Hodgkin’s Lymphoma. Diagnostics 2023,13, 861. [CrossRef] 43. Guaita-esteruelas, S.; Saavedra-Garcíagarcía, P.; Bosquet, A.; Borr, J.; Girona, J.; Amiliano, K.; Rodríguezrodríguez-Balada, M.; Heras, M.; Masana, L.; Gum, J. Adipose-Derived Fatty Acid-Binding Proteins Plasma Concentrations Are Increased in Breast Cancer Patients. Oncologist 2017,22, 1309–1315. [CrossRef] [PubMed] 44. Guaita-Esteruelas, S.; Gumà, J.; Masana, L.; Borràs, J. The Peritumoural Adipose Tissue Microenvironment and Cancer. The Roles of Fatty Acid Binding Protein 4 and Fatty Acid Binding Protein 5. Mol. Cell Endocrinol. 2018,462, 107–118. [CrossRef] 45. Guaita-Esteruelas, S.; Bosquet, A.; Saavedra, P.; Gumà, J.; Girona, J.; Lam, E.W.F.; Amillano, K.; Borràs, J.; Masana, L. Exogenous FABP4 Increases Breast Cancer Cell Proliferation and Activates the Expression of Fatty Acid Transport Proteins. Mol. Carcinog. 2017,56, 208–217. [CrossRef] [PubMed] 46. Bonuccelli, G.; Tsirigos, A.; Whitaker-Menezes, D.; Pavlides, S.; Pestell, R.G.; Chiavarina, B.; Frank, P.G.; Flomenberg, N.; Howell, A.; Martinez-Outschoorn, U.E.; et al. Ketones and Lactate “Fuel” Tumor Growth and Metastasis. Cell Cycle 2010,9, 3506–3514. [CrossRef] Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.