3650 - Predictive Value of Tumor Shrinkage in Relapsed/Refractory DLBCL Patients Treated with Radiation

Purpose/Objective(s): To determine whether percent change in planning target volume (PTV%) would predict for future disease relapse in patients with relapsed/refractory DLBCL.

Materials/Methods: We analyzed patients with relapsed/refractory DLBCL who received radiation therapy (RT) using online adaptive radiation treatment (oART) on the Ethos platform of a technology company at a single institution. Patients were included if they had active disease that required RT for either bridging to CAR T therapy or as salvage for post-chemotherapy or post-CAR T failure. PTV% change was calculated as the percent change in tumor volume from the time of the first to the last treatment. Disease recurrence was documented as a binary outcome, with further classification into local, regional, and distant failure. Correlation analyses, logistic regression, and receiver operating characteristics (ROC) were performed to determine the predictive value of PTV% change.

Results: A total of 24 patients were included in this analysis with a median follow up time of 7.5 months (range 0-31 months). Most patients received RT for bridging prior to CAR T therapy (n=13), followed by salvage RT for post-CAR T failure (n=8). All measurable disease was irradiated in all but 3 patients. The median RT dose was 30 Gy (range 20-44 Gy) in 5 fractions (range 5-20). The median PTV% change was -14% and ranged from +143% to -67%. An independent t-test showed that patients with recurrence had significantly different tumor size changes compared to those without (t = 2.40, p = 0.027, Cohen’s d = 0.98). Logistic regression indicated an increased recurrence risk with less tumor shrinkage (p = 0.050). An optimal cutoff of -16% tumor shrinkage was identified using the Youden Index, maximizing sensitivity (100%) and specificity (64%). Patients with less than 16% shrinkage had a 69% probability of recurrence, while those exceeding this threshold had 0% recurrence risk.

Conclusion: In our dataset we found that tumor shrinkage of -16% is a strong predictor of recurrence despite our modest patient numbers. The high NPV (100%) suggests that tumors that shrink sufficiently have almost no recurrence risk. In our dataset we found that tumor shrinkage of -16% is a strong predictor of recurrence despite our modest patient numbers. The high NPV (100%) suggests that tumors that shrink sufficiently have almost no recurrence risk. These findings highlight the potential prognostic value of real-time tumor response assessment and indicate a need for closer monitoring and potential treatment adjustments for high-risk patients such as radiation dose escalation or prompt enrollment onto a clinical trial.