Inclusion of baseline FDG PET/CT and clinical data for an improved predictive risk model in non-Hodgkin lymphoma patients treated using CAR T-cell therapy

Introduction: Chimeric antigen receptor T-cell therapy (CAR-T) has expanded treatment options for relapsed/refractory non-Hodgkin lymphoma (NHL) patients with many patients experiencing improved outcomes. The identification of biomarkers which can predict patient outcomes following CAR-T treatment is essential for appropriately guiding therapeutic strategies in these patients. The primary aim of this study was to assess the value of baseline clinical data as well as imaging data extracted from ¹⁸F-Fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) images in predicting overall survival (OS) and progression-free survival (PFS) in NHL patients who underwent CAR-T therapy.<w:sdt id="-252980971" sdttag="goog_rdk_0"></w:sdt>

Methods: Patients with relapsed/refractory NHL who received CAR-T treatment and underwent baseline FDG PET/CT imaging within 4 months prior to the start of therapy were eligible for this IRB-approved retrospective study. Baseline clinical data included age, sex, Eastern Cooperative Oncology Group performance status (ECOG), lactate dehydrogenase (LDH) levels, and tumor Ki-67 proliferation index results. The tumors with the highest maximum standardized uptake value (SUVmax) were semi-automatically segmented using MIM version 6.9. A fixed threshold value of 20% of the SUVmax was applied to the volumes of interest (VOIs). SUVmax along with mean and peak standardized uptake value (SUVmean and SUVpeak, respectively), total lesion glycolysis (TLG), kurtosis, and skewness were extracted from the VOIs using LIFEx version 7.1.11. Univariate analyses were used to evaluate the predictive ability of each baseline clinical and imaging metric. Those metrics with significant predictive power were then included in a multivariate model. The factors included in the multivariate model were divided into high risk and low risk groups. Risk cutoffs for clinical metrics were as follows: ECOG > 0, LDH > 250, and Ki-67 > 0.60. Receiver operator characteristic (ROC) analysis was used to determine cutoffs for PET metrics. Overall and progression-free survival (OS and PFS, respectively) were assessed using Kaplan-Meier methods. Hazard ratios (HR) were calculated using Cox proportional hazards models. Data analyses were performed using R version 4.1.2. A P-value less than 0.05 was considered significant.

Results: 109 NHL patients (86 male, 23 female; mean age = 65.2 ± 10.5 years) who received CAR-T therapy from December 2015 to September 2021 were included in this retrospective study. Median OS and PFS were 12.0 (IQR: 5.1-22.6) months and 4.8 (IQR: 2.8-18.4) months, respectively. Baseline age, sex, SUVmean, skewness, and kurtosis did not significantly predict OS or PFS. Baseline ECOG, LDH, and Ki-67 were all significant OS predictors (HR=2.1; P=0.0027, HR=1.8; P=0.0458, HR=2.4; P=0.0304, respectively). Baseline SUVmax, SUVpeak, and TLG were also significant OS predictors (HR=3.6; P=0.0006, HR=3.2; P=0.0001, HR=4.0; P<0.0001, respectively). Baseline ECOG and LDH were significant PFS predictors (HR=2.3; P=0.0210 and HR=2.0; P=0.0090, respectively). Baseline SUVmax, SUVpeak, and TLG were also significant PFS predictors (HR=3.5; P=0.0003, HR=3.4; P=0.0002, HR=3.2; P=0.00002, respectively). A multivariate model combining baseline ECOG, LDH, Ki-67, SUVmax, SUVpeak, and TLG showed the highest OS predictive ability (Fig. 1A; HR=7.0, P<0.0001). A model combining baseline ECOG, LDH, SUVmax, SUVpeak, and TLG showed the highest PFS predictive ability (Fig. 1B; HR=4.9, P<0.0001).

Conclusions: Several baseline clinical factors and semi-quantitative FDG PET/CT metrics were independently predictive of OS and PFS. Models combining baseline imaging and clinical data showed stronger predictive ability than any individual baseline metric considered here. Adding FDG PET/CT metrics to predictive models may improve identification of high risk patients at baseline as compared to clinical factors alone.

• Download figure
• Open in new tab
• Download powerpoint