Using Circulating Tumor DNA to Spare Patients with DLBCL Unnecessary Treatment and Cost: A Model-Based Analysis

Introduction Despite its advantages, 18F FDG positron emission tomography (PET) can yield false positives in diffuse large B cell lymphoma (DLBCL), potentially putting patients at risk of overtreatment. This was seen in long-term follow-up of the GOYA trial, where >50% of patients who did not achieve complete remission on end of treatment (EOT) PET following first line chemoimmunotherapy remained alive and progression-free many years later. Similar outcomes were seen in CALGB 50303. Plasma derived circulating tumor DNA (ctDNA) is an emerging biomarker in DLBCL. PhasED-Seq is a ctDNA assay that requires recovery of multiple variants on the same single strand of DNA, and thereby offers improved specificity for determining true disease activity. Chimeric antigen receptor T-cell (CAR-T) therapy has improved survival in primary-refractory and early relapsed DLBCL compared to conventional salvage chemotherapy and autograft, but has substantial therapeutic and financial toxicities that warrant judicious utilization. We therefore modeled an EOT ctDNA-guided strategy for patients with DLBCL completing frontline therapy, focused on those who are EOT PET positive to reduce potential CAR-T overtreatment. Methods We used a decision tree model to compare current practice with potential future treatment algorithms that incorporate ctDNA testing to guide treatment decision making for DLBCL following completion of frontline R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone). The model was built using TreeAge Pro Healthcare 2024. The standard care paradigm includes an EOT PET, where PET-positive cases receive either empiric second line therapy, biopsy, or repeat PET scan, with biopsy positive or repeat PET positive cases receiving second line CAR-T with axicabtagene ciloleucel (axi-cel). In the intervention strategy, only patients with positive EOT PET results received ctDNA testing, using the PhasED-Seq assay based on published performance metrics. Patients with positive ctDNA results were deemed to have active DLBCL, and thus received CAR-T. PET-positive, ctDNA-negative patients were not treated unless their disease progressed. To evaluate secondary cost outcomes, we performed a cost-minimization analysis from a health system perspective to evaluate direct medical costs. The model included cost of PET scans, biopsies, CAR-T (including supportive care, hospitalizations, and associated clinic visits), and ctDNA testing. Event probabilities and cost data were derived from a combination of recent publications, conference presentations, and Medicare procedure pricing. Since pricing for ctDNA testing by PhasED-Seq is not yet available, we estimated a base case price of $5,000 per test. Results In our model, an EOT ctDNA-guided treatment strategy significantly decreased CAR-T utilization in patients with PET-positive DLBCL (from 83% to 57% of patients) and reduced overall CAR-T use (from 30% to 22% of patients). Additionally, the ctDNA strategy lowered average healthcare costs (from $139,820 to $106,273 per patient). Sensitivity analyses showed that ctDNA guided treatment reduced CAR-T utilization in almost all circumstances, except if: 1) the probability of a patient with a positive EOT PET having negative ctDNA testing was <18% (compared to 44% in clinical data); or 2) the probability of DLBCL progression within one year with PET positivity and ctDNA negativity was >61% (compared to 4% in early clinical data). Conclusions Our model suggests ctDNA testing can refine diagnostic precision in DLBCL after first-line treatment, reducing CAR-T exposure for patients unlikely to benefit based on likely false positive EOT PET, thus mitigating therapeutic harm and costs. However, the lack of large-cohort data of standardized-methodology ctDNA highlights the urgent need for clinical trials investigating measurable residual disease-adapted approaches in DLBCL to both optimize clinical outcomes and ensure efficient healthcare spending.