Bridging Radiation Rapidly and Effectively Cytoreduces High Risk Relapsed/Refractory Aggressive B-Cell Lymphomas Prior to Chimeric Antigen Receptor T-Cell Therapy

Structured Abstract

Background:

Greater tumor burden before CD19-targeted chimeric antigen receptor T-cell (CAR T) therapy predicts lower complete response rate and shorter overall survival (OS) in aggressive non-Hodgkin lymphoma (NHL). Recent patterns of failure studies identified lesion characteristics including size, SUV, and extranodal location as associated with post-CAR T failure.

Objectives:

Here we analyze the effect of bridging radiation-containing regimens (BRT) on pre-CAR T lesion-and patient-level characteristics and post-CAR T outcomes including patterns of failure.

Study Design:

Consecutive NHL patients irradiated between 30d pre-leukapheresis until CAR T infusion were reviewed. Metabolic tumor volume (MTV) was contoured with threshold SUV 4. First post-CAR T failures were categorized as pre-existing/new/mixed with respect to pre-CAR T disease, and in-field/marginal/distant with respect to BRT.

Results:

Forty-one patients with diffuse large B cell lymphoma (DLBCL; n=33), mantle cell (n=7) and Burkitt’s lymphoma (n=1) were identified. BRT significantly improved established high-risk parameters of post-CAR T progression including in-field median MTV (45.5 to 0.2cc; p<0.001) maximum SUV (18.1 to 4.4; p<0.001), diameter (5.5 to 3.2cm; p<0.001), and LDH (312 to 232; p=0.025). DLBCL patients with lower LDH post-BRT had improved progression-free survival (PFS; p=0.001). In DLBCL, first failures were new in 7/19, pre-existing in 5/19, and mixed in 7/19; with respect to BRT, 4/19 were in-field and 4/19 marginal. Post-CAR T survival was similar in patients with initially low vs. newly low MTV post-BRT using a statistically determined threshold of 16cc (PFS: 26 vs. 31mo; OS unreached for both).

Conclusions:

BRT produced significant cytoreduction in diameter, SUV, MTV, and LDH, all predictors of poor post-CAR T outcomes. Similar progression-free and overall survival in patients with initially low MTV and those who newly achieved low MTV after BRT suggest BRT may “convert” poor-risk patients to better risk. In the future, response to BRT may allow for risk stratification and individualization of bridging strategies.

Graphical Abstract

Introduction

CD19-targeted chimeric antigen receptor T-cell (CAR T) therapies have transformed the treatment landscape for patients with relapsed/refractory B cell lymphoma.^1,2,3 Yet, up to 60% of patients will not achieve a complete response (CR) following CAR T and unfortunately, for such patients, median response duration is as low as 2 months.³ Thus, approximately two-thirds of CAR T patients ultimately develop progressive disease (PD) with dismal outcomes.^4,5 Given the high cost of CAR T and high baseline toxicity rates, there is strong incentive to identify strategies that may increase durable disease control after CAR T. Since many patients require some form of “bridging” therapy to treat active disease between leukapheresis and infusion,⁶ one possible approach to improve outcomes is to optimize bridging strategies. Bridging therapy options include radiotherapy, chemotherapy, immunotherapy, and steroids or combinations of these approaches. There is currently little data to guide the choice between these modalities and clinical practice is largely non-standardized.

Greater baseline tumor volume is associated with poorer CAR T outcomes. High tumor burden as evaluated by various metrics including lactate dehydrogenase (LDH), sum of the product diameters of up to 6 index lesions, and metabolic tumor volume (MTV) have been associated with inferior outcomes including lower overall response rate (ORR) and lower CR rate, shorter progression free survival (PFS), poorer overall survival (OS), and higher incidence of treatment-associated high-grade cytokine release syndrome (CRS) toxicity.^{2,7,8,9,10,11,12,13,14} Building on this, a recent pattern of failure study has associated certain baseline lesion characteristics including size, SUV, and extra-nodal status with greater risk of local post-CAR T failure.¹⁵ Taken together, these findings suggest it may be feasible to better identify high-risk patients – and specific high-risk lesions – that might benefit from focal therapeutic intervention prior to CAR T.

Radiotherapy provides rapid cytoreduction in hematologic malignancies and has retained efficacy for chemorefractory disease, making it an attractive bridging tool.¹⁶ Furthermore, RT has known immunomodulatory properties, raising the possibility of immune synergy with CAR T.¹⁷ Clinical incorporation of BRT is increasing and with many additional cellular therapies currently under development, this scenario will only become more frequent; what is learned about optimal bridging strategies likely will have relevance outside of lymphoma.¹⁸ Additionally, considering the recent positive results from ZUMA-7¹⁹ and TRANSFORM,²⁰ CAR T is now being considered earlier in the relapsed/refractory treatment paradigm which could expand the relevance of optimized BRT to the 40% of large B cell lymphoma patients who require second line therapy.²¹

At present, there are limited data to guide the rational integration of RT with CAR T. Published retrospective studies have not identified signals of excess toxicity or suggestions of markedly inferior clinical outcomes to imply reduced activity of CAR T after bridging radiotherapy (BRT), however, conclusions are limited by small cohort size (n = 3-17) and variable intent and applications of BRT.^{22,23,24,25,26} Many open questions remain with regard to the optimal application of BRT including which patients benefit most from this approach and whether BRT can alter the outcomes of patients who would otherwise not obtain a durable response to CAR T. Furthermore, particularly in the setting of multifocal disease, it is not yet clear which lesions should be targeted with BRT and to what dose.

Here, we report the largest series of NHL patients treated with BRT before CD19-targeted CAR T, and the first with quantitative radiomic correlatives. We hypothesized that BRT can significantly improve pre-CAR T risk status, including tumor burden and high-risk lesion features recently associated with post-CAR T failure. We hypothesized that in effectively cytoreduced patients, BRT would improve patients’ overall risk status and post-CAR T outcomes, and that BRT may alter subsequent patterns of failure by reducing the recurrence risk of treated lesions. Finally, we sought to characterize which clinical and treatment factors are associated with greatest benefit from BRT.

Materials and Methods

Patient Population

We included all consecutively treated patients with pathologically confirmed NHL who received radiotherapy in the period between 30d pre-leukapheresis to commercial CD19 CAR T infusion at Memorial Sloan Kettering Cancer Center (MSKCC) between 2017 and 2021; patients were excluded from cytoreduction analysis if they did not undergo interim PET between BRT and CAR T infusion. Medical records were reviewed to extract clinicopathologic features, treatment histories, and outcomes. Sex as a biological variable and age at BRT were collected. This research was approved by the MSKCC Institutional Review Board (protocol 16-1512). All patients signed informed consent for treatment with immune effector cells.

Bridging Radiation

Photon-based BRT was delivered using intensity modulated radiotherapy (IMRT) or conventional techniques. BRT fields were evaluated at the lesion level and categorized as ‘comprehensive’ (defined as treating 100% of the metabolic tumor volume [MTV; see radiographic and radiomic parameters below] or ‘partial.’ BRT targets were further classified by anatomic site as CNS vs. non-CNS and nodal vs. extranodal. Patients who received concurrent systemic therapy during the bridging period including chemotherapy, targeted agents, and steroids were included.

Radiographic and Radiomic Parameters

All radiographic and radiomic measurements were performed by an oncologic nuclear medicine physician who was blinded to clinical outcomes. Radiomic analysis was performed on the most recent PET prior to BRT start and a PET obtained between BRT completion and lymphodepleting chemotherapy (the ‘interim PET’). Maximum disease diameter was measured in any dimension; bulky disease was defined as a max diameter of ≥7.5cm per National Comprehensive Cancer Network Guidelines. To increase rigor and applicability, MTV was replicated and compared using two approaches with fully manual and semi-automated methods. First, disease was outlined manually in MIM PACS version 6.9.7 with a threshold of SUV 4. Separately, automated MTVs were acquired using the Beth Israel plugin for FIJI software (RRID: SCR_002285) also with a threshold of SUV 4.²⁷ The fixed absolute threshold of SUV 4 was chosen over a fixed relative approach such as 41% of SUV Max for greater ability to discriminate between active malignancy and post-radiation change. For both manual and automated methods, physiologic and inflammatory avidity were subtracted. Bone marrow avidity was included only if focal uptake was present. Agreement between manual and automated MTV values was evaluated using a concordance correlation coefficient (Supplementary Table S1). Given the excellent concordance (correlation coefficients all >0.96), automated values were used for analysis due to superior reproducibility.²⁷

Laboratory Parameters

“Pre-BRT” and “Post-BRT” laboratory parameters were collected including white blood cell count, lymphocytes, absolute lymphocytes, platelets, and LDH. The most recent peripheral blood collection prior to BRT start (“Pre”) and the most recent peripheral blood collection prior to lymphodepleting chemotherapy (“Post”) were used. Simplified Endothelial Activation and Stress Index (sEASIX) scores were then calculated [LDH divided by platelets]. sEASIX was chosen due to lack of baseline C Reactive Protein levels for most patients precluding calculation of modified EASIX scores.²⁸

Outcome Assessment

Interim PET imaging and all post-CAR T scans were evaluated per Lugano criteria.²⁹ Per institutional practice, patients underwent PET at 30-, 90-, and 180-days post-CAR T. We identified date(s) of first progressive disease (PD) overall and specifically within the BRT treatment area (i.e., local failure) post-CAR T. Pattern of failure analysis was performed to categorize PD as pre-existing (present pre-CAR T), new, or mixed and as in-field, marginal, or distant with respect to BRT. Scans demonstrating PD were cross-referenced with BRT plans and “in-field” PD was defined as PD within the 90% isodose line (area receiving 90% of the prescribed radiation dose). Marginal was defined as failure outside of the 90% isodose line but within 1cm of BRT field edge in any direction. Cytokine release syndrome (CRS) and Immune Effector Cell-Associated Neurotoxicity Syndrome (ICANS) were graded according to ASTCT Consensus criteria.³⁰ Radiation-related adverse events were graded using Common Terminology Criteria for Adverse Events Version 5.0.

Statistical Analysis

Wilcoxon signed ranked tests and McNemar’s Chi squared tests were used to compare parameters before and after BRT for continuous and categorical variables, respectively. Agreement between manual and automated MTV methods was examined using concordance correlation coefficients. A cut point for post-BRT MTV values was then established using a maximally selected log rank statistic for overall survival via the maxstat package in R. Within the DLBCL-only subcohort (the largest with homogenous histology), Kaplan-Meier methods were utilized to calculate PFS and OS; reverse Kaplan-Meier was used for median follow up. Time zero for survival measures was CAR T infusion. PFS and OS were stratified by the aforementioned post-BRT MTV cut point value of 16.4cc and by the comprehensiveness of BRT fields. Univariable analysis was performed using Cox proportional hazards models and variables for which the models did not converge were removed. CRS was treated as a binary endpoint as all events occurred within 11 days, and univariable analysis was conducted using logistic regression. Variables significant on univariable analysis (p<0.05) were selected by clinical importance for inclusion in multivariable models. All confidence intervals (CI) reported are 95%. All analyses were conducted using R v4.0.5 (RRID: SCR_001905).

Results

Patient Characteristics

We identified 41 eligible aggressive NHL patients (33 DLBCL, 7 mantle cell lymphoma [MCL], and 1 Burkitt lymphoma). Thirty-two patients (78%; 25 DLBCL, 6 MCL, and 1 Burkitt) had both pre- and post-BRT PET imaging and were included in quantitative cytoreduction analyses (Supplementary Figure S1). Baseline clinicopathologic features are summarized in Table 1. The median age at BRT was 66 years (range: 22–85) with male predominance (73%). Most (90%) had baseline ECOG ≤1. Most (n=30, 73%) had advanced stage disease pre-BRT and 19 (46%) were primary refractory. Age-adjusted IPI scores were most commonly high-intermediate (49%) or low-intermediate (29%). The median number of prior lines of therapy was 3 (range: 2–8). Eight patients (20%) had previously undergone autologous hematopoietic cell transplantation. In DLBCL patients, 18 (55%) were GCB-type, 14 (42%) double expressor, and 3 (9%) double/triple hit. Fifteen DLBCL patients (45%) were secondarily transformed from a different histology (8 follicular lymphoma, 4 marginal zone lymphoma, 2 chronic lymphocytic leukemia, 1 unknown).

TABLE 1.

CLINICAL CHARACTERISTICS

Age at BRT (median, range)	66 (22 – 85)
Sex (% male)	30 (73)
Race (% white)	33 (80)
ECOG prior to BRT
0 – 1 (%)	37 (90)
2 – 3 (%)	4 (10)
Histology
Diffuse Large B Cell Lymphoma (%)	33 (80)
Germinal Center Type (%)	18 (55)
Double Expressor (%)	14 (42)
Double/Triple Hit (%)	3 (9)
EBV expressed (%)	2 (6)
Transformed (%)	15 (45)
Follicular Lymphoma (%)	8 (53)
Marginal Zone Lymphoma (%)	4 (27)
Chronic Lymphocytic Leukemia (%)	2 (13)
Unknown (%)	1 (6)
Mantle Cell Lymphoma (%)	7 (17)
Burkitt’s Lymphoma (%)	1 (2)
Primary refractory disease (%)	19 (46)
Stage at BRT
Localized (stage I / II, %)	11 (27)
Advanced (stage III / IV, %)	30 (73)
Age-Adjusted IPI Score at BRT
Low (%)	6 (15)
Low-Int (%)	12 (29)
Int-High (%)	20 (49)
High: (%)	3 (7)
Lines of therapy prior to CAR T (median, range)	3 (2 – 8)
Prior autologous stem cell transplant (%)	8 (20)
Time from diagnosis to CAR T infusion, months (median, range)	17 (2 – 209)

BRT = bridging radiotherapy; ECOG = European Cooperative Oncology Group; EBV = Epstein-Barr Virus; IPI = international prognostic index; CAR T = chimeric antigen receptor T cell

Bridging Therapy Characteristics

Most frequently BRT-treated anatomic sites were head and neck (n=14), extremity (n=7), pelvis (n=7), and retroperitoneum (n=6). For 54% (n=22), the BRT target included extranodal disease and in 45% (n=19), targeted disease was bulky. BRT fields were comprehensive in 16 patients (39%). Median total dose was 30Gy (range: 4–54) delivered over a median of 10 fractions (range: 2 – 30); 23 patients (56%) received ≥30Gy. The median biologically effective dose using an alpha/beta ratio of 10 was 39Gy (range: 5–65Gy). Most frequently utilized BRT regimens were 4Gy x 5 (n=11) and 3Gy x 10 (n=10). The majority of BRT was intensity modulated radiation therapy (71%) with 24% conventional and 5% 3D-CRT. Median time between BRT and lymphodepletion was 14 days (range 1–75). Seven patients (17%) received all BRT prior to leukapheresis. See Table 2 for complete radiation characteristics.

TABLE 2.

BRIDGING THERAPY CHARACTERISTICS

	N = 41
Irradiated sitea
Head and Neck (%)	14 (33)
Extremity (%)	7 (17)
Pelvis (%)	7 (17)
Retroperitoneum (%)	6 (14)
Spine (%)	2 (5)
Spleen (%)	2 (5)
Axilla (%)	1 (2)
Sternum (%)	1 (2)
Lung (%)	1 (2)
Brain (%)	1 (2)
Extranodal irradiated site (%)	22 (54)
Bulky irradiated site: >7.5 cm (%)	19 (46)
Largest lesion irradiated (%)	33 (80)
Most avid lesion irradiated (%)	30 (73)
Comprehensive radiation fields (%)	16 (39)
Planning target volume, cc (median, range)	634 (18 - 5846)
Total BRT Dose, Gy (median, range)	30 (4 – 54)
≥30 Gy (%)	23 (56)
Fractions per course (median, range)	10 (2 – 30)
Biologically effective dose, Gy (median, range)	39 (5 – 65)
Most common BRT regimens
4 Gy × 5 (%)	11 (27)
3 Gy × 10 (%)	10 (24)
1.5 Gy × 20 (%)	2 (5)
3 Gy × 12 (%)	2 (5)
2 Gy × 2 (%)	2 (5)
BRT modality
IMRT (%)	29 (71)
Conventional (%)	10 (24)
3D-CRT (%)	2 (5)
All radiation prior to apheresis (%)	7 (17)
Systemic therapy between apheresis and CAR T infusion (%)	17 (41)
Regimen includes chemotherapy (%)	10 (59)
Regimen includes targeted therapy (%)	7 (41)
Regimen includes steroids (%)	4 (24)
Systemic therapy before BRT (%)	4 (24)
Rituximab + Etoposide + Prednisone + Vincristine + Cyclophosphamide + Doxorubicin	1
Rituximab + Dexamethasone + Cytarabine + Oxaliplatin	1
Rituximab + Polatuzumab Vedotin	1
Polatuzumab Vedotin	1
Systemic therapy concurrent with BRT (%)	11 (59)
Dexamethasone	2
Zanubrutinib	1
Ibrutinib	1
Rituximab + Prednisone	1
Ibrutinib + Lenalidomide	1
Gemcitabine + Oxaliplatin	1
Intrathecal Cytarabine + Venetoclax	1
Methotrexate + Leucovorin + Cytarabine	1
Dexamethasone + Cytarabine + Oxaliplatin	1
Rituximab + Doxorubicin + Cyclophosphamide + Bortezomib	1
Systemic therapy after BRT (%)	2 (12)
Etoposide + Ifosfamide + Cytarabine + Methotrexate	1
Topotecan + Vinorelbine + Thiotepa + Clofarabine	1

^aN = 42 due to one patient irradiated to head and neck and pelvis concurrently

BRT = bridging radiotherapy; IMRT = intensity modulated radiotherapy; 3D-CRT = Three-Dimensional Conformal Radiation Therapy; CAR T = chimeric antigen receptor T-cell therapy;

Seventeen patients (41%) also received systemic therapy during the bridging period. Systemic therapy was concurrent with BRT in 11 patients, preceded BRT in 4 patients, and followed BRT in 2 patients. Systemic regimens were highly variable, containing chemotherapy in 59% (n=10), targeted inhibitors in 41% (n=7), and steroids in 24% (n=4). Systemic regimens are detailed in Table 2.

All patients received fludarabine and cyclophosphamide lymphodepletion prior to CAR T infusion. Infused CAR T products were: axicabtagene ciloleucel (n=19, 46%), tisagenlecleucel (n=10, 24%), lisocabtagene maraleucel (n=9, 22%), and brexucabagene autoleucel (n=3, 7%). Four patients with mantle cell lymphoma were treated with lisocabtagene maraleucel on clinical trial NCT02631044.

Outcomes

Imaging Response Post-BRT

We first focused on the 32 evaluable patients with pre- and post-BRT PET imaging. For this group, median time to interim PET evaluation from completion of BRT was 13 days (range: 0–50). At this early timepoint, most achieved an in-field CR/PR (84%; n=27). However, outside of the BRT field, many (56%; n=18) progressed during the bridging period. Of the patients with out-of-field progression, 8/18 (44%) had also received systemic therapy during the bridging period. Marginal failures post-BRT but pre-infusion occurred in 4 patients (13%). Best responses on interim PET are described in Supplementary Figure S2a.

Quantifying Cytoreduction Due to BRT

Changes in radiographic/radiomic parameters following BRT were next analyzed (Figure 1 and Supplementary Table S2). Significant decreases in several adverse disease parameters were noted including reductions in overall median maximum SUV (18.7 to 12.6; p=0.03) and maximum disease diameter (5.7 to 5.4cm; p=0.006). Three patients (9%) were metabolically negative following BRT (no SUV above background). The number of patients with initially bulky disease decreased from 44% to 16% pre-infusion, p=0.008.

Figure 1.

Quantification of cytoreduction at the overall patient level including irradiated and non-irradiated disease (All disease) and only within the BRT field (Irradiated disease) over the bridging period. Values before and after BRT for maximum diameter in centimeters (a), maximum SUV (b), metabolic tumor volume (c), and LDH (d) in all and irradiated disease. Crossbars indicate median value. Patients of any histology with post-BRT interim PET included for 1a/b/c (n = 32), all patients included for LDH analysis in 1d (n = 41).

These improvements were more prominent when considering just the disease treated with BRT. Specifically, we report highly significant reductions in in-field maximum disease diameter (5.5 to 3.2cm; p<0.001), bulky disease (44 to 13%; p=0.004), maximum SUV (18.1 to 4.4; p<0.001), and presence of any lesion with SUV>10 (72 to 25%; p=0.001). Ten patients (31%) had metabolically negative disease in-field after BRT.

Large, significant reductions in in-field median MTV occurred following BRT (45.5 to 0.2cc; p<0.001) (Figure 1 and Supplementary Table S2). While change in overall (irradiated and non-irradiated) MTV (median 50.3 to 21.5cc; p=0.073) was not significant, LDH, a serologic correlate for overall disease burden, was significantly reduced after BRT (median 312 to 232U/L; p=0.025). Patients with metabolically active extranodal disease decreased from 81% to 53% (p=0.016). Figure 2 shows the percent and absolute reduction in MTV after BRT by histology both in and out-of-field.

Figure 2.

Change in MTV over the bridging period. Percent change in MTV in only irradiated (a) and all disease (b) as well as absolute change in MTV (c) over the bridging period for patients of any histology with a post-bridging radiation PET; n = 32

BRT Impact on Hematologic Parameters

Hematologic parameters were measured at a median of 3 days (range: 1–28) pre-BRT start and post-BRT at a median of 2 days (range: 1–35) prior to lymphodepletion. White blood cell count was significantly decreased following BRT (median 6 to 5k/mcL; p=0.014) as was absolute lymphocyte count (median 0.8 to 0.5k/mcL; p=0.01). However, simplified EASIX score (median 1.9 to 1.4; p>0.9) and platelet count (154 to 156k/mcL; p=0.1) were stable following BRT (Supplementary Figure S3).

Of the 7 patients who received BRT prior to apheresis, 1 experienced new leukopenia and 2 experienced new lymphopenia during BRT; in all cases counts recovered to normal levels by the time of apheresis. Additionally, the patients with new leukopenia/lymphopenia during BRT did not receive BRT to significant bone marrow volume, and two received cytotoxic chemotherapy during the bridging period, raising the possibility that BRT was not the principal driver of count reduction.

Survival Outcomes

The median follow-up from CAR T infusion was 20.3 months (interquartile range: 11.5–29.5). A swimmer’s plot describing outcomes following CAR T infusion for the overall cohort is illustrated in Figure 3a. Overall, 76% of patients had a CR/PR post-CAR T: 61% CR and 15% PR (Supplementary Figure S2b). In the DLBCL sub-cohort (n=33), ORR was 79% (64% CR), median PFS was 20 months (CI: 2.7–unreached) and median OS was not reached (CI: 27–unreached) (Figure 3b & 3c).

Figure 3.

Patient outcomes following CAR T infusion. Swimmer’s plot depicting patient courses from CAR T infusion for full study cohort; N = 41 (a). Progression free survival (b) and overall survival (c) in DLBCL patients (n = 33). Median PFS was 20 months (CI: 2.7–not reached) and median overall survival was not reached (CI: 27–not reached). Patterns of first failure in DLBCL patients with any post-CAR T progression (d); n = 19

Failure Patterns with Respect to BRT

Within study follow up, 19 DLBCL patients had PD after CAR T (Supplementary Table S3). First failures were relatively balanced between pre-existing and new sites (n=5 pre-existing, n=7 new, and n=7 mixed; Figure 3d). The majority of first failures were distant to the BRT field (n=11/19) with rates of in-field and marginal failure both 4/33 in DLBCL patients. Seven first failures were extranodal while 5 were nodal and 7 were mixed nodal/extranodal. Fourteen failures were pathologically confirmed of which 3 were CD19 negative.

Clinical Associations with Post-CAR T Failure

On multivariate analysis, higher post-BRT LDH was associated with shorter PFS (HR 1.03; p=0.001) (Supplementary Table S4). Radiation dose ≥30Gy was not significantly associated with PFS. Supplementary Table S5 shows the results of the univariate analysis from which select factors were included for multivariate analysis. Factors significant for PFS which were not able to be included in the multivariate model due to few events included age (HR 1.04; p=0.041), ECOG ≥2 (HR 4.97; p=0.041), CNS involvement (HR 7.36; p=0.006), and maximum disease diameter pre-BRT (HR 1.07; p=0.037). A multivariate model was not feasible for OS given few events.

We then stratified patients by MTV-associated risk using the post-BRT MTV cut point of 16.4cc established using a maximally selected log rank statistic for OS. There were non-significant trends toward improved PFS and OS in patients with a post-BRT MTV of <16.4cc (p=0.2 & p=0.079; Figure 4, top row). To explore the hypothesis that BRT can improve a patient’s risk of relapse through effective cytoreduction, three subgroups were created. Patients were considered ‘Poor Risk’ if they had an MTV >16.4cc post-BRT, ‘Good Risk’ if their pre- and post-BRT MTV were <16.4cc, and ‘Converted Risk’ if the pre-BRT MTV was greater than 16.4cc and post-BRT MTV improved to <16.4cc. Due to the small number of patients, log rank could not be reported. However, Converted Risk patients appeared to have PFS and OS similar to the Good Risk patients and improved compared to the Poor Risk group (Figure 4, middle row). Case examples for ‘Good’, ‘Poor’, and ‘Converted’ risk patients with radiographic and radiomic correlatives are shown in Figure 4c.

Figure 4.

Outcomes by MTV and BRT comprehensiveness in DLBCL. Progression free (a) and overall survival (b) in DLBCL patients with interim PET post radiation (n = 25) stratified by post radiation MTV cut point of 16.4cc (top row). In the middle row patients were stratified as ‘Poor Risk’ (>16.4cc MTV on post-radiation PET), ‘Good Risk’ (≤16.4cc MTV on both pre- and post-radiation PETs) or ‘Converted Risk’ (>16.4cc MTV on pre-radiation PET and ≤16.4cc MTV on post-radiation PET). In the bottom row patients were stratified by extent of radiation: ‘Partial’ (<100% of total MTV irradiated) or Comprehensive (100% of total MTV irradiated). P values were not evaluable for three group Kaplan Meier or for overall survival stratified by radiation extent due to low number of events. (c) Three example patient cases illustrating Good Risk (top row), Converted Risk (middle row), and Poor Risk (bottom row). Of note, the Good Risk patient received comprehensive radiation while the Converted and Poor Risk patients received partial radiation. PD = progressive disease; PR = partial response, CMR = complete metabolic response, QD = daily, BID = twice a day; fx = fractions

Finally, we stratified patients by extent of radiotherapy fields (Figure 4, bottom row). We found non-significant trends toward improved PFS and OS for patients receiving Comprehensive BRT compared with Partial BRT (p=0.074 & not reportable).

Toxicities

Seventy-six percent (n=31) experienced a toxicity at least possibly attributable to BRT, the majority of which (87%; n=27) were low grade (Supplementary Table S6). Grade 1 adverse events with >10% incidence were fatigue (n=23), dermatitis (n=7) and xerostomia (n=6). Grade 3 toxicities were anemia (n=2), mucositis (n=2), and trismus (n=1). The two patients with grade 3 anemia following BRT both had pre-existing grade 3 anemia prior to BRT start (in one case in the setting of known marrow disease involvement and in the other with history of recent cytotoxic chemotherapy). There were no grade 4-5 toxicities attributable to BRT. Following CAR T infusion, CRS was observed in 29 patients (71%) with grade 3 CRS in 2 patients. On univariate analysis, non-axicabtagene CAR T products were associated with a reduced risk of CRS (HR 0.09; p=0.003). ICANS was observed in 8 patients (20%); 4 patients had Grade 3 ICANS and 1 patient had grade 4. Four of twenty-eight evaluable patients (14%) had a grade ≥3 hematologic toxicity at 90 days post-CAR T infusion (G3 anemia: n=3 (11%); G4 thrombocytopenia: n=1 (4%); G3-4 neutropenia: n=0 (0%).

Discussion

Here we report the largest series of patients with aggressive NHL treated with BRT. We found that BRT rapidly and significantly reduces the overall maximum diameter, SUV, and irradiated MTV – all established predictors of poorer post-CAR T prognosis. Preserved sEASIX scores and low rates of post-CAR T toxicity reinforce the safety of radiating between leukapheresis and infusion. Patterns of failure in this cohort were similar to published failure patterns after CAR T without BRT which may suggest effective risk reduction of high-risk pre-existing lesions which otherwise would have enriched failure in pre-existing sites. Similar progression-free and overall survival in patients with initially low MTV and those who newly achieved low MTV after BRT suggests the ability to “convert” poor risk patients to better risk, though larger cohorts are needed to confirm this finding.

This is the first experience to our knowledge in which MTV has been assessed before and after BRT, allowing quantification of the extent of cytoreduction achieved with BRT. Importantly, we validated the use of semi-automatic MTV measurements in this cohort which are methodologically more reproducible across studies and institutions as well as notably more time-efficient, however if not validated against manually contoured MTVs can lead to erroneous representation of disease burden. Additional novel aspects of this work include it being the largest published sub-group of mantle cell lymphoma patients treated with BRT/CAR T and the largest report of failure patterns following BRT and CAR T.

High risk lesions characterized by maximum diameter, SUV, and extranodal disease, which have been recently associated with post-CAR T failure¹⁵ were mitigated by BRT. Importantly, these features were significantly reduced not only within the irradiated field but in the overall patient, demonstrating that despite the typically rapid disease proliferation rate seen in this heavily pre-treated relapsed/refractory population, appropriately applied local therapy can improve the overall risk of patients awaiting CAR T. It is important to note that 41% of this cohort received bridging systemic therapy, which also contributed to overall disease control. This cytoreduction is remarkable considering that post-BRT PET was typically performed at a very early timepoint (median 13 days post-BRT) when post-treatment inflammation may still contribute to SUV and MTV values. Thus, radiomic risk reduction from BRT reported here may even underestimate ultimate clinical risk reduction.

Regarding post-CAR T outcomes, the median OS in DLBCL patients was not reached with 70% survival at both 12 months and 24 months. This compares favorably to the 58% 12-month OS reported in TRANSCEND³ and the 50.5% 24-month OS reported in ZUMA-1¹ and is similar to the 80% 12-month OS in prior smaller, retrospective BRT series.²³ The median PFS of 20 months in this cohort also compares favorably to randomized trial data where reported median PFS was 5.9 – 6.8 months.^1,3 The observed 51% 12-month PFS is similar to 44% reported in a retrospective cohort of 11 BRT patients.²⁴

While the value of baseline MTV in predicting post-CAR T outcomes has been demonstrated repeatedly^{8,9,10,11,12,13,14}; here we report for the first time that MTV reduction with targeted radiotherapy over the bridging period may be able to alter risk and relatively quickly convert patients who were at high-risk of progression and death based on their MTV at leukapheresis to a better risk state by time of infusion. Patients for whom MTV crossed below a threshold of 16.4cc after BRT had PFS and OS similar to those with MTV <16.4cc before BRT. Of note, the MTV cutoff of 16.4cc was determined statistically to evaluate the hypothesis that MTV reduction with BRT could convert patients to better risk and requires validation in larger cohorts. Notably, the suggestion that BRT could alter post-CAR T outcomes via effective MTV reduction is consistent with what has been observed with systemic bridging therapy, where larger change in MTV over the bridging period has been associated with improved PFS.¹⁴ If validated, BRT response could become a meaningful risk metric for CAR T patients and could suggest a role for an adaptive approach to bridging therapy, wherein an interim PET scan after an initial course of BRT could be used to determine the need for further BRT or alternate therapy prior to CAR T infusion. We are interested in better understanding if there are differences between BRT-associated and systemic bridging-associated cytoreduction and are currently in the process of analyzing this in our institutional CAR T cohort.

Comprehensive BRT fields had previously been suggested to be beneficial²⁴; while our data appear consistent with this suggesting improved PFS/OS in comprehensively treated patients, our study did not have enough statistical power to properly evaluate a difference and we are unable to discern if selection bias and inherent disease related features may drive these superior outcomes. However, it is notable that on interim PET the majority of MTV progression occurred in unirradiated rather than irradiated sites, suggesting that a more comprehensive BRT approach, when feasible, may be able to keep disease burden low prior to CAR T infusion.

First failures after CAR T were relatively balanced between new sites and pre-existing sites. This finding is similar to recently published failure patterns following CAR T alone which describe a predominance of pre-existing site and mixed site failure.¹⁵ Given that any BRT cohort is likely enriched for high-risk pre-existing lesions, observing similar patterns of failure – if not trending toward new site failure – may indicate effective risk-reduction of high-risk lesions however requires further validation. There has been one prior study to our knowledge reporting patterns of failure in BRT patients.³¹ We report slightly lower pre-existing site failure, which might be explained by higher BRT doses (median 30Gy in our cohort vs. 20Gy in the Saifi et al. cohort) with the caveat that cross study comparison is difficult and particularly limited given the low number of failures in both studies.

The marginal and in-field failure rates in our cohort may suggest that broader fields and higher doses should be considered, when safely feasible, in this population with rapidly proliferating and often refractory disease. The decision to increase dose, however, must be weighed carefully against the detriment of increasing BRT toxicity as well as treatment time and prolonging time to CAR T infusion. Furthermore, immunologically, lower RT doses have been shown to enable local CAR T killing independent of antigen expression supporting a strategy of broader low dose coverage.³² Interestingly, though higher BRT doses ≥3000cGy appeared associated with longer PFS on univariate analysis, this was no longer significant on multivariate when controlled for post-BRT LDH, perhaps suggesting that clinicians may choose higher doses for patients with lower overall disease volume reflecting an implicit bias that this group should be more aggressively bridged.

Another novel finding in this cohort is the preservation after BRT of sEASIX scores which have been shown to predict severe CRS and neurotoxicity following CAR T.²⁸ This reinforces that BRT does not appear to add excess toxicity to CAR T, as has been suggested by previous small retrospective series.^22,23,24 In the overall cohort, we observed grade 3 or higher CRS in 5% and grade 3 or higher ICANS in 12%; these rates fall within published toxicity rates after CAR T alone (2 – 22% for G3+ CRS and 10 – 32% for G3 ICANS)^1,2,3,33 again suggesting no substantial added toxicity with BRT. Importantly, in light of the known profound cytopenias following CAR T, we did observe a small but significant decrease in WBC and absolute lymphocytes after BRT. However, rates of persistent ≥grade 3 cytopenia at 90 days post-CAR T infusion in this cohort (11% anemia, 4% thrombocytopenia, and 0% neutropenia) appear comparable to ZUMA-1 (3% anemia, 7% thrombocytopenia, 11% neutropenia; no BRT allowed)¹ and real world data with mixed second-generation CAR T constructs (7% anemia, 10% thrombocytopenia, and 20% neutropenia).³⁴

Given the specific concern regarding cytopenias with BRT prior to leukapheresis impairing cell yields it is reassuring that of the 7 patients in this cohort irradiated prior to leukapheresis there were no RT-related cytopenias persistent at the time of apheresis. Nevertheless, we think continued caution with BRT pre-leukapheresis, particularly when a significant bone marrow volume must be irradiated and if BRT is given immediately prior to leukapheresis, is warranted.

This study has several limitations. Most notably, a substantial minority (~40%) of the patients in this cohort also received systemic therapy within the bridging period which we acknowledge affects both outcomes and toxicity. Patients treated with both systemic therapy and radiotherapy were included to reflect real world practice patterns. While this data should not be interpreted to argue for radiation alone bridging, the disproportionately significant reductions in disease burden metrics within the radiation field suggest that the cytoreduction achieved may be in large part due to radiation. Second, our cohort was heterogeneous in terms of histology and radiation approach, including those treated with strictly palliative intent, with the aim of immune augmentation, or with intentionally comprehensive fields. Additionally, based on prior practice patterns and to evaluate the safety of BRT prior to leukapheresis, we included patients irradiated within 30d prior to leukapheresis outside of the now largely accepted standard definition of the ‘bridging period’ between leukapheresis and CAR T infusion. Third, despite being the largest cohort published to date the total sample size is modest, reflecting the relatively recent approval of commercial CAR T products and referring provider comfort levels with BRT. Finally, there is inherent selection bias in this non-randomized cohort as clinicians likely chose patients with perceived higher-risk and/or progressive lesions to refer for local intervention. Prospective investigation evaluating these questions is ongoing at our institution (NCT05574114).

In conclusion, we found that BRT had minimal toxicity and significantly reduced radiographic and radiomic features associated with post-CAR T failure.¹⁵ Further work with larger cohorts and prospective comparison is needed to investigate the hypothesis raised here that BRT may alter post-CAR T patterns of failure and post-CAR T patient outcomes.

Highlights.

• Bridging radiation significantly reduced diameter, SUV, and metabolic tumor volume

• Bridging radiation may convert poor-risk CAR T candidates to better risk

• Bridging radiation response could guide individualization of bridging strategies