A machine learning–derived sarcopenia index is associated with survival and nonrelapse mortality in DLBCL

Figures

• 

  Figure 1.. Kaplan-Meier survival curves. Survival curves depicting OS, PFS, and FFS according to SI at baseline (A-C) and SIrel at iPET (D-F). Groups are stratified by the lowest tertile vs the combined upper tertiles (prespecified stratification). Significance was determined using the log-rank test.

• 

  Figure 2.. Sarcopenia and the molecular landscape of DLBCL. (A) Oncoprint of recurrently mutated genes in patients (mutation frequency, ≥10%) with available WES data (n = 105). Columns represent individual patients, and rows represent genes; mutation classes are color-coded. Top annotations indicate baseline SI tertile, SIrel at iPET tertile, age group, and cell-of-origin subtype. (B) Box plots depicting the distribution of baseline SI (upper panel) and SIrel (lower panel) across genetic clusters, defined using the DFCI classifier and the LymphGen algorithm. Significance was determined using the Kruskal-Wallis test. COO, cell of origin; DFCI, Dana-Farber Cancer Institute; WES, whole-exome sequencing.

• 

  Figure 3.. Multivariable Cox proportional-hazards models for OS. Forest plot depicting HR and 95% CI from multivariable Cox proportional-hazards models with OS as the primary end point (63 events). The model used for the analysis shown in (A) only included covariates available at baseline, and survival was assessed at baseline. The model shown in (B) included covariates assessed at interim response assessment, and survival was assessed from interim response assessment.

• 

  Figure 4.. Cause-specific multivariable analysis. (A) Forest plot depicting cause-specific HR and 95% CI from multivariable Cox proportional-hazards models. HRs for lymphoma-specific mortality (24 events) are shown in red; HRs for NRM (34 events) are shown in blue. (B) Forest plot depicting subdistribution HR (sHR) and 95% CI from multivariable Fine-Gray competing risks regression models. sHR for lymphoma-specific mortality with NRM as the competing event is shown in red, sHR for NRM with lymphoma-specific mortality as the competing event is shown in blue.

• 

  Figure 5.. Logistic regression analysis for hematotoxicity. Forest plots depicting OR and 95% CI from multivariable logistic regression analyses assessing the association between baseline sarcopenia and grade ≥3 hematologic treatment toxicity (145 events; CTCAE version 3.0). BMI, body mass index; CTCAE, Common Terminology Criteria for Adverse Events; ECOG PS, Eastern Cooperative Oncology Group performance status; LDH, lactate dehydrogenase; ULN, upper limit of normal.