AUTHORS
Yidan A Wang, Lily Kollitz2, Qing Cao, MS3, Marie Hu, MD4, Daniel O'Leary, MD5, Christopher D Graham, MD6, Binoy Yohannan, MD7, Mark B. Juckett, MD, MHCM8, Zuzan Cayci, MD9, Veronika Bachanova, MD, PhD10, Supriya Gupta, MD11
AFFILIATIONS
1 University of Minnesota Medical School, Minneapolis, United States
2 University of Minnesota Medical School, Minneapolis, United States
3 University of Minnesota Clinical and Translational Science Institute, Minneapolis, United States
4 University of Minnesota, Minneapolis, United States
5 University of Minnesota, Minneapolis, United States
6 University of Minnesota, Minneapolis, United States
7 University of Minnesota, Minneapolis, United States
8 University of Minnesota, Minneapolis, United States
9 University of Minnesota, Minneapolis, United States
10 University of Minnesota, Minneapolis, United States
11 University of Minnesota, Minneapolis, United States
ABSTRACT
Background: CAR-T is capable of inducing prompt complete remissions in R/R LBCL as early as day (D) 30 post CAR-T. However, many patients (pts) have persistent hypermetabolism due to incomplete response or inflammation, which makes response assessment challenging. Tools to predict long-term outcomes in pts with equivocal D30 responses are needed. Metabolic tumor volume (MTV) is a method of quantifying whole-body tumor burden by positron emission tomography (PET) scans which may provide prognostic value beyond standardized uptake value (SUV) alone.
Methods: We conducted a single-center retrospective study of 122 LBCL pts treated with commercial CD19 CAR-T therapy (2016 – 02/2026). Pts were categorized by disease response as D30 complete response (CR), D30 progressive disease (PD), converters (PR/SD at D30 & CR by D90), or late progressors (PR/SD at D30 & PD by D90). Central nervous system lymphoma was excluded. Clinical variables & PET metrics at pre-lymphodepletion (LD), D30, & D90 timepoints were compared. Overall survival (OS) & progression-free survival (PFS) were estimated with Kaplan-Meier curves, adjusting for pre-specified confounders in a Cox regression model. Recursive partitioning was used to find the optimal MTV & SUV cutoff based on 2-yr OS & PFS.
Results: At D30, 67 (55%) pts achieved a CR, 41 (34%) had PR/SD, & 14 (11%) had PD. Among pts with D30 PR/SD (primary analysis cohort), 24 (59%) converted to CR by D90 & 17 (41%) progressed. Median age for the entire cohort was 64.8 yrs (range 26.7-86.6). Diffuse large B-cell lymphoma was the most common subtype (66%). CAR-T products were axi-cel (41%), liso-cel (34%), & tisa-cel (25%). Age, race/ethnicity, sex, histology, stage, double/triple-hit status, Ki-67, bridging therapy, CAR-T product & rates of cytokine release syndrome, neurotoxicity & cytopenias were similar across groups.
Pre-LD ECOG performance status differed across groups (ECOG 0-1 in 97% of D30 CR, 88% of converters, 71% of D30 PD, & 65% of late progressors; p<0.01), as did CRP (median 0.3, 0.4, 2.9, 1.3 mg/dL, respectively; p<0.01), ferritin (205, 265, 618, 471 ng/mL; p=0.02), MTV (8.5, 40, 52.2, 62.3 mL; p<0.01), & SUVmax (10.7, 16.8, 22.7, 18.1; p=0.03).
The 2-yr OS differed by response: 78% for converters, 60% for D30 CR, 28% for late progressors, & 7% for D30 PD (p<0.01). The optimal MTV cutoff for OS & PFS was identified as 117.6. Inferior 2-yr OS was noted in pts with pre-LD MTV ≥117.6 (27% [n=20] vs 62% [n=86]; p<0.01), pre-LD SUVmax ≥43.2 (13% vs 60%; p<0.01) & ECOG ≥2 (23% vs 58%; p<0.01). In the primary analysis cohort, 2-yr OS was 78% for converters vs 28% for late progressors (p<0.01). Pre-LD CRP (0.4 vs 1.3; p=0.05) & D30 SUVmax (median 5.2 vs 13.5; p<0.01) were lower in converters. Inferior 2-yr OS was noted with pre-LD LDH ≥ 120.5 (41% vs 74%; p=0.04), pre-LD MTV ≥117.6 (30% vs 73%; p=0.02) & D30 SUVmax ≥5.1 (46% vs 100%; p<0.01), but D30 MTV did not differentiate convertors from progressors (9.1 vs 17.5; p=0.54). Inferior 2-yr PFS was noted with ECOG ≥2 (13% vs 56%; p=0.03) & D30 SUVmax ≥8.1 (21% vs 76%; p<0.01). On multivariate analysis adjusting for ECOG & pre-LD LDH, D30 SUVmax ≥ 5.1 remained independently associated with inferior PFS (HR 8.62, 95% confidence interval 1.10-67.60; p=0.04). More converters fell below the D30 SUVmax OS-optimal cutoff of 5.1 (38% vs 6%) & the PFS-optimal cutoff of 8.1 (58% vs 18%; both p<0.01). Among 8 pts with ClonoSEQ measurable residual disease (MRD) assessments at D30 & D90, MRD status was concordant in 6 (75%) pts; all 4 pts with persistent MRD negativity subsequently converted to CR.
Conclusion: LBCL pts yielding PR/SD 30 days after CAR-T may experience ongoing disease response with 60% achieving subsequent CR. PET metrics can provide meaningful prognostic information. Baseline MTV identifies pts at risk for inferior survival in the overall cohort but did not distinguish converters from late progressors. In pts with D30 PR/SD, the D30 SUVmax was a strong predictor of subsequent conversion to CR. Other factors included baseline ECOG & CRP. Our data supports continued observation rather than immediate intervention for pts with low D30 SUVmax, whereas high pre-LD MTV or D30 SUVmax may require closer monitoring including MRD testing & early intervention. Systematic incorporation of D30 PET metrics into post–CAR-T assessment may help refine risk stratification & inform surveillance & consolidation strategies for pts with LBCL with D30 PR/SD.
DISCLOSURES
Yidan A Wang: No relevant conflicts of interest to declare.