ALBERT

Description: Introduction: Primary CNS lymphomas (PCNSL) are heterogeneous, aggressive, extra-nodal non-Hodgkin lymphomas limited to the neuraxis. Published response rates to high-dose methotrexate (MTX) based induction regimens for PCNSL range from 35-78%. However, 〉50% of patients relapse and have a median survival of 2 months without additional treatment. Our ability to prognosticate outcomes is limited to clinical models like the International Extranodal Lymphoma Study Group (IELSG) score and Memorial Sloan-Kettering Cancer Center (MSKCC) classifier. There is an urgent need to develop improved biologic and radiologic predictive models for PCNSL to facilitate therapeutic advances. We hypothesize that a machine learning model using advanced magnetic resonance imaging (MRI) tumor characteristics will improve the accuracy of clinical models to predict response to MTX and survival outcomes. Methods: Data from patients with PCNSL treated at UT Southwestern and Parkland Health and Hospital System hospitals from 2008-2020 (n=95) were collected. An analytical dataset of 61 patients was selected based on the availability of T1 postcontrast (T1c) and T2w FLAIR MR images. A subset of 47 patients was used to evaluate MTX treatment response. Expert neuroradiologists drew regions of interest (ROIs) on the multiparametric MR images including whole tumor (consisting of edema + enhancing tumor + necrosis), enhancing tumor and necrosis (Figure 1). Response to methotrexate-based induction was defined per the International Primary CNS Lymphoma Collaborative Group (IPCG) criteria. For overall- and progression-free survival (OS and PFS) analysis, short (≤1 year) and long-term (〉1 year) survivor groups were defined. A support vector machine (SVM) network was used for predicting treatment response to MTX and for predicting the OS groups. A Multinomial Naive Bayes (MNB) network was used for predicting the PFS groups. PyRadiomics package was used to extract 106 texture-based features from the combination of each MR image and tumor ROI. A total of 642 features were extracted from the imaging parameters. Clinical features including age, race, performance status, MSKCC class, IELSG score, histology, delay from 1st MRI to start of treatment, induction and consolidation treatments used were included in the analysis. Feature reduction methodology based on the feature importance derived from the gradient boost model was applied to reduce the number of features. 17 features (imaging = 14, clinical = 3) were used for predicting OS/PFS and 7 features (imaging = 5, clinical = 2) were used for predicting treatment response to MTX. Networks utilizing only clinical features were analyzed for comparison. The sklearn package in python was used for the machine learning analysis. 5-Fold cross validation was performed to generalize the network performance. Results: Baseline wclinical characteristics of the study population is shown in Table 1. Table 2 lists the accuracy, F1 score, sensitivity, specificity, positive predictive value, negative predictive value and area under the curve (AUC) values averaged for the 5-fold cross validation. The SVM network achieved a mean testing accuracy of 81.1 ± 12.3% for predicting the treatment response to MTX-based induction. Sensitivity, specificity and AUC values were 90.5 ± 13.1%, 63.3 ± 22.1% and 0.81 ± 0.14 respectively. The SVM and the MNB network achieved mean testing accuracies of 80.3 ± 11.4% and 83.3 ± 11.8% for predicting the long and short survival groups in OS and PFS respectively. Sensitivity, specificity and AUC values for the SVM and MNB networks were 79.3 ± 6.5%, 80.5 ± 16.5% and 0.86 ± 0.12 and 85.3 ± 12.9%, 81.9 ± 11.8% and 0.86 ± 0.13 respectively. The accuracy values for predicting treatment response to MTX, OS and PFS using only the clinical features were 61.6 ± 9.2%, 59.1 ± 16.4% and 62.1 ± 17.5% respectively. Conclusion: This machine learning model boosted the accuracy (≥20%) over currently validated clinical models alone in predicting response to methotrexate-based therapies and survival outcomes in PCNSL. The current analysis is limited by the small sample size, and we plan to statistically test this model across a larger dataset and report results at the meeting. Our preliminary results suggest that machine learning based radiomic analysis may predict biologic aggressiveness in PCNSL and has the potential to be integrated in clinical predictive tools and design of clinical trials. Disclosures Awan: Blueprint medicines: Consultancy; Celgene: Consultancy; Sunesis: Consultancy; Karyopharm: Consultancy; MEI Pharma: Consultancy; Astrazeneca: Consultancy; Genentech: Consultancy; Dava Oncology: Consultancy; Kite Pharma: Consultancy; Gilead Sciences: Consultancy; Pharmacyclics: Consultancy; Janssen: Consultancy; Abbvie: Consultancy. Desai:Boston Scientific: Consultancy, Other: Trial Finding.

Description: Introduction:In R/R DLBCL patients receiving CAR T-cell therapy (CAR-T), bridging therapy (BT) with chemotherapy, targeted therapy, and/or radiation therapy (RT) is often administered during the manufacturing window after collection and prior to CAR-T infusion to aid in tumor debulking and/or control symptomatic disease. However, little is known about the optimal type of BT and specifically, the impact BT may have on patterns of failure. Thus, we sought to compare the patterns of failure in patients who received CAR-T for R/R NHL at a single-institution based on the type of BT received. We hypothesize that bridging RT decreases the risk of local recurrence in sites treated with RT prior to CAR-T therapy. Methods: An IRB-approved single-institution retrospective review was performed of all R/R DLBCL patients who underwent leukapheresis for planned CAR-T with axicabtagene ciloleucel (axi-cel). 20 patients were identified, of whom 1 died before CAR-T infusion and 3 died before day +30 post-CAR-T (D+30) PET/CT scans due to CAR-T related toxicities (n=2) or disease progression (n=1). Of the 16 patients who had D+30 PET/CT scans, demographic, disease, and treatment characteristics, as well as toxicity and survival outcomes, were collected and analyzed. PET/CT scans immediately before CAR-T, as well as D+30, D+90, 6 months, and 1-year post-CAR-T infusion were analyzed, with response assessment per the Lugano classification. FDG-avid (Lugano 4 or 5) lesions on pre-CAR-T scan were recorded as index lesions and compared to residual or new FDG-avid lesions on all available post-CAR-T scans. Results: Of the 16 patients who had D+30 PET/CT scans, 4 received no BT, 6 received bridging chemotherapy, 5 received bridging RT (median 30 Gy in 10 fxs), and 1 received a Bruton's tyrosine-kinase inhibitor (BTKi) on a clinical trial. At last follow up, 11/15 (73%) were alive (3/4 with no BT, 3/5 with bridging chemotherapy, 4/5 with RT, 1/1 with BTKi) and 8/15 (53%) were without disease progression (3/4 with no BT, 3/5 with chemotherapy, 2/5 with RT, 0/1 with BTKi); one patient was lost to follow up. Grade 3-4 cytokine release syndrome (CRS) occurred in 1/4 (25%) patients without BT, 2/6 (33%) patients with bridging chemotherapy, and 0/5 (0%) patients with bridging RT. Grade 3-4 immune effector cell-associated neurotoxicity syndrome (ICANS) occurred in 1/4 (25%) patients without BT, 2/6 (33%) patients treated with bridging chemotherapy, and 1/5 (20%) patients treated with bridging RT. Complete patient, disease, and treatment characteristics, as well as outcomes and toxicities, are summarized in Table 1. 6/16 (38%) patients had metabolic complete response (CR) at D+30 and 6/13 (46%) at D+90. In comparison among type of BT received, D+30 and D+90 CR rates, respectively, were 1/4 (25%) and 2/3 (67%) for no BT, 4/6 (67%) and 3/5 (60%) for bridging chemo, and 1/5 (20%) and 1/4 (25%) for bridging RT. In analysis of patterns of failure, there were 48 total index lesions identified on pre-CAR-T PET/CT scans. Of those, 20 received no BT, 15 were treated with bridging chemotherapy, 7 with bridging RT, 6 with bridging BTKi. On D+30 PET/CT, the rates of CR were 16/20 (80%) in lesions without BT, 8/15 (53%) in lesions treated with bridging chemo, 6/7 (86%) in lesions with bridging RT. By D+90, the rates of CR were 7/9 (78%) in lesions without BT, 8/11 (73%) in lesions treated with bridging chemo, and 3/3 (100%) of lesions treated with bridging RT. Of the 50 lesions noted on D+30 PET/CTs, 18 (36%) were index lesions on pre-CAR-T PET/CT, and only 4/28 (14%) lesions on D+90 PET/CT were initial index lesions. Conclusions: Patients who require BT before CAR-T have higher relapse rates, likely reflecting more aggressive disease biology. Bridging RT to CAR-T appears to be safe and effective in providing local control, even at palliative doses, but may not impact overall outcomes. Due to small sample size and retrospective biases, comparison among bridging treatments is limited, and the optimal bridging strategy remains unknown. However, these data suggest that bridging RT should be considered in sites where local control is a priority, such as symptomatic sites or sites where recurrence may cause significant morbidity. Radiating all sites of active disease pre-CAR-T may not improve outcomes though as the predominant pattern of failure appears to be distant. The optimal timing and combination strategies with RT and CAR-T for R/R DLBCL needs to be explored prospectively. Disclosures Awan: MEI Pharma:Consultancy;Karyopharm:Consultancy;Genentech:Consultancy;Astrazeneca:Consultancy;Abbvie:Consultancy;Janssen:Consultancy;Pharmacyclics:Consultancy;Sunesis:Consultancy;Gilead Sciences:Consultancy;Kite Pharma:Consultancy;Dava Oncology:Consultancy;Celgene:Consultancy;Blueprint medicines:Consultancy.Desai:Boston Scientific:Consultancy, Other: Trial Finding.

Description: Introduction: Primary CNS lymphomas (PCNSL) are rare, aggressive, extra-nodal, non-Hodgkin lymphomas confined to the CNS. Other than high-dose methotrexate (MTX) based induction regimens, there are no established standards-of-care for managing PCNSL. Treatment practices across institutions vary regionally and globally. At safety-net hospitals, there are often access issues, delays in seeking medical care, and limited treatment options due to insurance barriers and other prohibitive costs. This study compares the demographics, treatment patterns, and survival outcomes among PCNSL patients treated at a safety-net hospital versus a tertiary academic institution within the same healthcare system. Methods: Medical records of 94 patients who were treated for PCNSL from 2007-2020, at either a public safety-net hospital (n=34) or an academic tertiary-care center (n=60), both serving the same metroplex, were analyzed. Demographics, Memorial Sloan-Kettering Cancer Center (MSKCC) prognostic class, International Extranodal Study Group (IELSG) score, induction and consolidation treatment modalities used, and survival outcomes were compared between the 2 groups. Categorical variables and response rates were compared using Fisher's exact test. Continuous variables were compared using Mann-Whitney U test. Survival functions for overall- and progression-free survival (OS and PFS) were estimated by the Kaplan-Meier method and compared using a log-rank test. Cox proportional hazards regression was used for multivariate analysis. Results: Baseline characteristics of the study population is shown in Table 1. Compared to the tertiary academic center, patients at the safety-net hospital were significantly younger, had better Karnofsky performance status (KPS) and MSKCC prognostic class, and were more commonly Black or Hispanic, male, and HIV positive. Median follow-up was 39.4 months for surviving patients and 13.1 months for all patients. Induction regimens used were either chemotherapy (n = 74) or palliative whole brain radiation therapy (WBRT) (n = 16). Chemotherapy regimens were predominantly MTX-based (71/74, 95.9%). Safety-net patients were significantly less likely to receive induction chemotherapy (67.6% vs 85%, p = 0.004). Overall response to chemo-based induction was 70.3%. There was no significant difference in response rate to chemo-based induction between safety-net and academic center patients (73.9% vs 68.6%, p = 0.786). Patients who had a good response to induction chemotherapy (n = 52) were either given consolidative WBRT (n =11), autologous stem cell transplant (ASCT) (n = 14), chemotherapy (n = 17), or no further treatment (n = 10). When comparing safety-net and academic center patients, hospital location did not significantly affect whether patients received consolidation therapy (82.4% vs. 80%, p = 0.735). Safety-net hospital patients were significantly less likely to receive ASCT (0% vs. 40%, p = 0.001) and had higher rates of consolidative WBRT (41.2% vs 11.4%, p = 0.03). OS (71.7% vs. 51.1%, p = 0.218) and PFS (54.9% vs. 57.4%, p = 0.967) were not significantly different between safety-net and tertiary academic hospitals when adjusted for age and KPS. In a subset analysis, excluding patients who only received palliative care, there also were no differences in OS (68.9% vs 55.2%, p = 0.309) and PFS (50.3% vs 57.4%, p = 0.719) between hospitals (Figure 1). Patients who received consolidation treatment had significantly higher OS (84.8% vs 34.0%) and PFS (77.0% vs 30.3%), after adjusting for age and KPS (p = 0.001 for both). There were no significant differences between type of consolidation strategy employed for OS (p = 0.801) or PFS (p = 0.899). Conclusions: Safety-net patients with PCNSL were less likely to receive induction chemotherapy and could not receive ASCT as consolidation due to insurance barriers. Despite these differences in patterns of treatment, they had similar outcomes in terms of OS and PFS, even after adjusting for age and performance status. Consolidation treatments improved outcomes and it is critical to ensure eligible patients receive them, though the optimal type of consolidation remains unknown. This study is limited by the small sample size and retrospective analysis. Research to unearth the biologic heterogeneity of PCNSL and develop predictive biomarkers will be critical to personalize and optimize management strategies for these patients. Disclosures Awan: Genentech: Consultancy; MEI Pharma: Consultancy; Janssen: Consultancy; Astrazeneca: Consultancy; Celgene: Consultancy; Blueprint medicines: Consultancy; Sunesis: Consultancy; Karyopharm: Consultancy; Pharmacyclics: Consultancy; Gilead Sciences: Consultancy; Kite Pharma: Consultancy; Dava Oncology: Consultancy; Abbvie: Consultancy. Desai:Boston Scientific: Consultancy, Other: Trial Finding.