Patient and CAR T product disposition
In total, 981 LBCL patients were approved for CAR T-cell therapy (axi-cel 805, tisa-cel 176) between January 2019 and January 2023. We identified 38 patients who had at least 1 CAR T-cell MF. The intended CAR T-cell product was axi-cel in 28 and tisa-cel in 10 patients. Overall MF frequency was 3.9% (3.5% for axi-cel and 5.7% for tisa-cel). To analyse MF risk across different time periods, we compared MF frequency in era 1 (January 2019–January 2021) with era 2 (February 2021–January 2023). There was no difference by time period with MF frequency of 17/462 (3.6%) in era 1 and 21/519 (4.0%) in era 2 (P = 0.74).
The number of manufacturing attempts and resulting product outcomes are shown in the flow diagram (Fig. 1). Details of the reasons for MF per patient and per manufacturing attempt are shown in Supplementary Table 1. All 38 patients had an MF after the 1st manufacturing attempt (OOS in 18 and no product available in 20). Some of these went on to have further manufacturing attempts from either the same apheresis material or by repeating apheresis. In the end, a total of 59 manufacturing attempts were made for the 38 patients; 20 resulted in an OOS product (18 after 1st and 2 after 2nd attempt) 13 of which were infused, 13 produced a product in-specification after one or more remanufacturing attempts (11 after 2nd, 1 after 3rd and another after 4th attempt) 11 of which were infused and 26 resulted in no product being available (20 after 1st, 5 after 2nd and 1 after 3rd attempt). Analysing the 46 manufacturing attempts resulting in either an OOS product or no product, the most frequent reasons for MF were low cell viability (n = 8), low T-cell purity (n = 8), poor growth in culture (n = 6), low interferon-gamma (n = 6) and low CAR T-cell dose (n = 4). Most frequent reasons for MF with an available OOS product were low cell viability (n = 5), low interferon-gamma (n = 5) and low T-cell purity (n = 4) (Supplementary Tables 1 and 2). Although numbers were limited, analysing reasons for MF over time did not reveal any obvious differences between era 1 and era 2 (Supplementary Table 2).
Of the 38 patients with MF, OOS product was infused in 13 patients (OOS-infused cohort) following approval from the NHSE OOS CAR T panel; 11 patients went on to receive a delayed infusion with an in-specification CAR T-cell product (delayed-infused cohort) and 14 patients did not proceed to CAR T-cell infusion (MF-no-infusion cohort) (Fig. 1). For comparison with the 38 MF patients, we included 38 randomly selected LBCL controls matched for CAR T-cell product (28 axi-cel and 10 tisa-cel) without MF, 29 of whom received infusion (controls-infused cohort).
Risk factors for MF
An assessment of baseline variables and their impact on risk of MF is shown in Table 1. There was no difference in age, sex, BMI, number of prior lines of therapy, prior high-dose cytarabine, prior stem cell transplant or the need for holding therapy between MF patients and controls. Bendamustine therapy was the only baseline variable significantly associated with risk of MF with prior exposure in only 3 out of 38 (7.9%) controls compared with 11 out of 38 (28.9%) MF patients (P = 0.026). In all but 2 patients, bendamustine was administered as part of the rituximab, polatuzumab and bendamustine regimen. Two patients had prior bendamustine (one each with rituximab and obinutuzumab) for follicular lymphoma. The increased risk with bendamustine appeared to be largely due to therapy within 6 months; with 0% of controls vs 9 of 38 (23.7%) of MF patients (P = 0.0029) receiving it within 6 months of apheresis (Table 1 and Supplementary Fig. 1). We went on to analyse the impact of the number of bendamustine cycles received on the risk of MF. We found a significant association between bendamustine timing and the number of cycles received (P = 0.021), with patients given bendamustine closer to apheresis having received fewer cycles in total; of the 11 MF patients with prior bendamustine, 7 received it within 3 months of apheresis; 4 received only 1 cycle (all as holding therapy) and the others received 3 cycles (Fig. 2). Therefore, we were not able to evaluate the effect of number of cycles of bendamustine on risk of MF as it was confounded by the timing of delivery.
Association between timing of bendamustine in relation to apheresis and number of cycles received.
Assessment of variables at apheresis, including biochemical and haematological parameters, and their impact on risk of MF is shown in Table 2. None of the variables analysed, including total white cell count, absolute neutrophil, absolute lymphocyte count, platelet count, CD3 count, LDH, CRP and volume of blood apheresed were significantly associated with a risk of MF. As bendamustine was the only baseline variable associated with a risk of MF, we further analysed the impact of recent bendamustine (<6 months) on blood counts at apheresis (Table 3). Apart from a lower platelet count of borderline statistical significance (P = 0.051), there was no difference in any other parameter between those exposed to recent bendamustine vs those not.
Efficacy outcomes
Median follow-up from approval for CAR T-cell therapy was 25.2 months (IQR: 18–35.6) for patients with MF and 35.9 months (IQR: 28.7–49.5) for controls. Median (IQR) time from approval for CAR T-cell therapy to infusion was 63 (57–69), 85 (70–119) and 62 (48–71) days for OOS-infused, delayed-infused and controls-infused patients. Baseline characteristics of infused CAR T patients is shown in Table 4, there were no significant differences in baseline characteristics between the infused cohorts (OOS-infused, delayed-infused and controls-infused).
ORR and CR were assessed by PET-CT scans at 1- and 3 months post infusion. Best ORR (CR) rates were 53.9% (53.9%), 54.6% (45.5%) and 71.5% (42.9%) for OOS-infused, delayed-infused and controls-infused, respectively (Fig. 3). Corresponding best ORR and CR rates by CAR T-cell product are shown in Supplementary Fig. 2, although numbers are too small to test for any differences in response rates by product. ORR (CR) rates at 1 month post infusion were 53.9% (53.9%), 54.6% (45.5%) and 71.4% (32.1%) and at 3 months were 46.2% (46.2%), 27.3% (27.3%) and 57.1% (39.3%) for the OOS-infused, delayed-infused and controls-infused cohorts respectively. There were no significant differences in the response rates between the three infused cohorts (P = 0.47 and P = 0.26). There was no suggestion for a delay in time-to-CR in either the OOS-infused or delayed-infused cohorts. All patients achieving a CR in both cohorts did so by 1 month post infusion.
Best response post CAR T infusion.
The 1-year OS (95% CI) from approval for OOS-infused, delayed-infused and controls-infused patients was 52.8% (23.4–75.5), 46.8% (14.8–73.9) and 68.4% (48.0–82.1) respectively with no significant difference between OOS-infused and controls-infused patients (HR: 1.52 (95% CI 0.65–3.58), P = 0.34), between delayed-infused and controls-infused patients (HR: 1.13 (95% CI 0.41–3.11), P = 0.81) or between OOS-infused vs delayed-infused (HR: 0.74 (0.24–2.28), P = 0.61). As expected, OS for non-infused patients (both for MF-not-infused and controls-not-infused) was poor with 1-year OS (95% CI) rates of 28.6% (8.8–52.4) and 11.1% (0.6–38.8), respectively (Fig. 4A).
A Overall survival (OS) from approval for CAR T. B Progression-free survival (PFS) from infusion.
The 1-year PFS (95% CI) for OOS-infused, delayed-infused and controls-infused patients was 46.2% (19.2–69.6), 24.2% (4.4–52.5) and 41.4% (23.7–58.3) respectively with no significant difference for OOS-infused vs controls-infused (HR 1.41 (95% CI 0.64–3.13), P = 0.40), for delayed-infused vs controls-infused (HR 1.64 (95% CI 0.70–3.83), P = 0.25) or for OOS-infused vs delayed-infused (HR: 0.86 (95% CI: 0.333–2.25), P = 0.76) (Fig. 4B). PFS for the three infused cohorts by CAR T-cell product is shown in Supplementary Fig. 3, although patient numbers are too small for any statistical comparisons to be performed.
Safety outcomes
In OOS-infused, delayed-infused and controls-infused patients, CRS of any grade (and grades 3–4) was seen in 83.3% (15.4%), 90.9% (0%) and 93.10% (6.90%), respectively, with no significant differences between the three infused cohorts either for any grade CRS (P = 0.81) or grades 3–4 CRS (P = 0.50). Corresponding rates for ICANS of any grade (and grades 3–4) were 38.5% (7.7%), 30.0% (18.2%) and 34.50% (10.30%), respectively, once again with no significant differences between the three infused cohorts either for any grade ICANS (P > 0.99) or grades 3–4 ICANS (P = 0.72).
The incidence of grades 3–4 neutropenia at 1 month (and 3 months) post CAR T-cell infusion was 16.7% (20.0%), 44.4% (40.0%) and 29.20% (19.05%) for OOS-infused, delayed-infused and controls-infused patients, respectively (P = 0.62 at 1 month and P = 0.81 at 3 months). Corresponding incidence of grades 3–4 thrombocytopenia at 1 month (and 3 months) was 33.3% (20.0%), 40.0% (20.0%) and 25.00% (4.80%) (P = 0.63 at 1 month and P = 0.24 at 3 months). There were no significant differences in cytopenia rates at either timepoint, though numbers assessable (alive without progression) at each point were small.
There were 4 deaths in remission, 1 in the OOS group at 21 months post CAR T (cause unknown, but after diagnosis of MDS), 3 in the control group (2 COVID at 13.1 and 45.9 months and 1 unknown cause at 8.4 months).
