148 patients were screened (Fig. 1), 6 were deemed ineligible for tisagenlecleucel due to active graft-versus-host disease (n = 3), CD19 negative disease (2), and lymphopenia precluding leukapheresis (1). The intention-to-treat cohort was therefore formed by 142 eligible patients. Overall, 17 eligible patients (12%) were not infused. Eight patients were not harvested: 5 patients underwent alternative therapy or palliation, and 3 had progressive disease. Nine harvested patients were not infused—in 6 this was due to progressive leukaemia, 1 patient succumbed to infection after lymphodepletion (Fig. 1).

CR/CRi complete remission with/without haematological recovery, DoD died of disease, PD progressive disease, HSCT allogeneic haematopoietic stem cell transplantation, MRD minimal residual disease. *Patients died from procedural mortality (n = 3).

Table 1 shows patient baseline characteristics. The median age at CAR T-cell screening was 11 years (IQR 6.5–15.9, range 0.7–25.7), 79/142 (55.6%) were male. Fifteen patients (10.6%) were infants at diagnosis, and 14 of these had KMT2A rearrangement. 44/142 (31.0%) presented with high-risk features and 2 presented as B-cell lymphoblastic lymphoma. Half of the population (74/142, 52.1%) were deemed refractory at some point prior to infusion with tisagenlecleucel. This was a heavily pre-treated cohort with a median of 3 (IQR 2–3) lines prior to CAR T (excluding HSCT) and 58 (40.8%) having received HSCT (one patient had received 2 allografts). 43/142 (30.3%) had received blinatumomab (12/43, 27.9% were non-responders) and 18/142 (12.7%) had received inotuzumab prior to tisagenlecleucel. Six patients (4.2%) had received experimental anti-CD19 and/or CD22 directed CAR T-cell products prior to consideration for tisagenlecleucel. Sixty-four patients (45.1%) had central nervous system (CNS) involvement prior to CAR T referral and 17 (12%) non-CNS extramedullary disease (EMD), with testicular disease being the single most common site (Supplementary Table 1).

Disease status was assessed prior to lymphodepletion (Table 2). Patients had a median BM disease burden of 2% (IQR 0.09–12%) blasts. 57/125 of infused patients (45.6%) had low (<5% blasts) BM disease burden, 48/125 (38.4%) proceeded with high (≥5%) disease burden and 15/125 (12%) had undetectable BM disease. There were no patients with CNS3 disease. Three patients with EMD proceeded with either stable (n = 2) or reduced (n = 1) disease burden compared with previous MRI/US imaging.

All patients received lymphodepletion comprising intravenous fludarabine (4 days at 30 mg/m2/day) and cyclophosphamide (2 days at 500 mg/m2/day). In 3 patients fludarabine doses were adjusted to renal function.

134 patients underwent harvest with a median of 1 day needed for leukapheresis (IQR 1–1; range 1–4) (Table 3). In one case, the leukapheresis product did not meet manufacturer’s guidance (low CD3 count) but manufacture was ultimately successful. In 5 cases (3.7%) the manufactured product did not meet manufacturer’s release criteria: 1 was infused as an out-of-specification product (viability <70%), 3 underwent successful repeat harvest and subsequent manufacture. One patient progressed before another harvest could be re-attempted (Supplementary Table 3).

Outcomes for the 17 non-infused patients are shown in Fig. 1. Median survival for the 17 non-infused patients was 4.4 months (95%CI 3, - not reached [NR]), 4/17 were alive in remission at last follow-up.

The CR/CRi rate at day 30 for the 125 infused patients was 92% (115/125). Of 115 responders, 104 (90.4%) were negative for BM MRD by IgH/TCR PCR. Two patients died prior to day 30 and were not evaluable for response: one died of refractory disease five days after infusion, the other died of fungal pneumonia whilst in remission on day 24. Of 123 evaluable patients at day 30, 8 did not achieve CR/CRi: 7 died of disease (all within 16 months) and 1 is alive in remission 2 years after HSCT.

Survival for both the ITT and infused cohorts are shown in Fig. 2a, b. With a median follow-up of 26.3 months (IQR 19.7–36.7), the 2-year OS in the ITT cohort was 65.2% (95%CI 57.2–74.2%) and in the infused cohort was 70% (95%CI 61.7–79.4%). Two-year EFS as per ELIANA criteria was 46.5% (95%CI 37.6–57.6%) and 51.7% (95%CI 42.1–63.5%) for the ITT and infused cohorts respectively. Two-year stringent EFS was 35.6% (95%CI 28.1–44.9%) for the intention-to-treat cohort and 40.4% (95%CI 32.2–50.7%) for the infused cohort. The median OS for the ITT cohort was not reached, median EFS as per ELIANA criteria was 22.5 months (95%CI 17.3—NR), median stringent EFS was 8.7 months (95%CI 6.3–16.3). In the infused population the median OS was not reached, the median EFS was 25.7 (19.2-NR) months and the median stringent EFS was 8.6 months (5.5–25).

Survival in the intention to treat (a) and infused (b) cohorts. OS overall survival, EFS event-free survival as per ELIANA definition, stringent EFS stringent event-free survival.

Using the standard definition we noted a worse EFS in association with greater disease burden (≥5%, HR 1.95, p = 0.02), prior inotuzumab exposure (HR 2.21, p = 0.03), refractory disease, particularly refractory relapse (HR 3.05, p < 0.0001), blinatumomab non-response (HR 2.90, p = 0.02) (Table 4) and number of therapy lines by univariable regression analysis (Supplementary Fig. 1). However, follow-up for the 18 patients who had received inotuzumab (Supplementary Fig. 1) was short and as such, its longer-term effect could not be reliably estimated. We tested these and other clinically relevant variables in multivariable models. In the final adjusted model only refractory disease (HR 2.33, 95%CI 1.1–4.9), (p = 0.02) was noted to be significantly associated with worse EFS after tisagenlecleucel. In contrast to frontline chemotherapy [19], age at tisagenlecleucel screening did not impact survival (Table 4).

Toxicity data were available for 123 of 125 infused patients. Tisagenlecleucel was generally well tolerated, the procedural mortality was 2.4% (accounting for 3/126 patients who proceeded with lymphodepletion) (Fig. 1). Although 106/123 (86.2%) patients suffered cytokine release syndrome (CRS), this was severe (grade 3–4) in only 16/123 (13%), with a median duration of 3 days (Table 5a). Immune effector cell associated neurotoxicity syndrome (ICANS) affected 26/123 (21.1%) patients and was severe in 10 (8.1%), the median duration was 3 days. One patient died of encephalopathy on day 51 post-infusion, an extensive infectious screen was negative.

Cytopenias persisting or occurring beyond day 30 post-infusion were noted in 56 of 95 (58.9%) evaluable patients (Table 5b), median duration was 57.5 days (IQR 20–90) after day 30 and 33/95 (34.7%) were grade 4. Neutropenia was the commonest cytopenia, with 23/95 (24.2%) having isolated neutropenia and 28/95 patients (29.5%) having neutropenia in combination with anaemia/thrombocytopenia. Supportive interventions for cytopenia are shown in Table 5b.

There were 45 episodes of infection in 35/123 (28.5%) evaluable patients, 26/123 (21.1%) had 1, 8 (6.5%) had 2 and 1 (0.8%) had 3 episodes of infection; 27/45 (60%) of infections were severe, i.e., grade 3 or higher (Table 5c). One patient died of a combined JC/HHV6 CNS infection and invasive fungal chest infection (see procedural mortality above). The most frequent infections were viral infections (11/123, 8.9%, Supplementary Table 4) which were generally mild (9/11, 81.8% grade 1–2) and sepsis/bacteriaemia (11/123, 8.9%), all of which were severe (10/11 grade 3 and 1/11 grade 4). Other type of infections (10/123, 8.1%) included skin/deep tissue infections (3), respiratory infections (2), urinary tract infections (2), otitis media (1), conjunctivitis (1) and pancreatitis (1). Febrile neutropenia (grade 3) affected 9/123 (7.3%); all concomitantly affected by CRS, which may have been a confounder.

Hypogammaglobulinemia is a predictable consequence of B-cell aplasia induced by CD19 CAR therapy in children, we documented this in 91/104 (87.5%) responders for whom this information was available. Intravenous immunoglobulin (Ig) replacement was started in 58/91 (63.7%) patients by the time of the data cut-off; at later time-points post CAR T infusion, subcutaneous formulations were used. The probability of remaining in B-cell aplasia after tisagenlecleucel infusion is shown in Fig. 3a. For the 55 patients who did have B-cell recovery or CD19 positive MRD re-emergence/relapse, the median time until these events was 4.1 months (IQR 2.4–8.7).

a Probability of remaining in B-cell aplasia. Main events are B-cell recovery or CD19 positive relapse, competing events are non-response, disease emergence other than CD19 positive and death. b Cumulative incidence of relapse. Death in remission and non-response are competing events. All documented relapses were considered, i.e. including those occurring after MRD emergence and/or other therapy following CAR T. This is because censoring further treatments received after CAR T would underestimate the true incidence of relapse after tisagenlecleucel infusion. c Cumulative incidence of relapse stratified by immunophenotype. Competing events non-response, death in remission, myeloid switch and relapse with unknown immunophenotype.

The cumulative incidence of frank relapse 1- and 2-years post-infusion was 21% (95%CI: 14–29%) and 33% (95%CI 24–43%) respectively (Fig. 3b). Relapses before further therapeutic intervention occurred between 1.3 and 43.6 months post-infusion. The cumulative incidence of CD19 negative relapse was 8.8% (95%CI 4.4–15) and of CD19 positive relapse 18% (95%CI 11–27) at 2 years post-infusion (Fig. 3c). Three patients experienced myeloid switch leukaemia after tisagenlecleucel: one non-responder (40 days post-infusion) and two responders (98 and 784 days post-infusion), two died of disease; a patient with late myeloid switch was alive with disease at last follow-up. A fourth patient evolved to a myeloid switch 1 month after starting blinatumomab for a CD19 positive MRD relapse following infusion and died of disease. Two of these patients with myeloid switch disease had KMT2A-rearranged disease, making the rate of lineage switch in this subgroup 2 of 18 infused cases (11.1%).

All 115 responding patients were evaluable for longer term outcomes (Fig. 1, Responding patients’ box). Two patients died in remission after day 30, one of neurotoxicity (see Toxicity above) and one of late infection. With a median follow-up of 23.3 months (IQR 14.9–32.8), 50 patients were alive and event-free by the stringent definition at the time of data cut-off, a single patient was lost to follow-up (3 months post-infusion). Of the remaining 62 patients (Fig. 1), in whom the median follow-up from infusion was 28.7 months (IQR 20.5–36.3), 20 patients first presented with a frank relapse, 20 patients had emergence of MRD, and 22 patients received further anti-leukaemic therapy for early LBCA with undetectable disease. Outcomes for these 62 patients are detailed in Fig. 4.

HSCT allogeneic haematopoietic stem cell transplantation, MRD minimal residual disease, TRM transplant related mortality, DoD died of disease.

Twenty patients relapsed at a median of 7.1 months (95%CI: 4.6–19.8) post tisagenlecleucel. Twelve (60%) had a CD19 positive relapse, 6 (30%) had a CD19 negative relapse, 1 (5%) patient relapsed with myeloid switch leukaemia and in 1, the immunophenotype was unknown. Thirteen (65%) had isolated BM relapse, 3 (15%) BM + non-CNS EMD (EMD sites: 2 orbital, 1 bilateral kidney involvement), 3 (15%) isolated CNS and 1 (5%) isolated EM (maxilla and bilateral kidney lesions) relapse. Six patients underwent HSCT: 4 are alive in molecular remission, 1 died of TRM and 1 died of disease after a further post-HSCT relapse. Of the 14 patients who did not undergo transplant, 5 are alive with disease (follow-up range 0–26.8 months) and 9 died of disease.

Re-emergence of MRD occurred at a median 3 months after infusion (95%CI 2–6). 3/20 (15%) patients never achieved a molecular remission and had rising levels of MRD. 9/20 (45%) had CD19 positive and 7/20 (35%) CD19 negative disease and in the rest (4/20, 20%) the immunophenotype of disease was unknown. 14/20 (70%) patients received HSCT (in 2 cases following a frank relapse), of which 3 died in remission of TRM and 2 died of disease. Within the 6 non-transplanted patients, 4 are alive after starting maintenance treatment: 3 are alive in molecular remission (2 have completed 2 years of maintenance) and 1 is alive with disease.

Of the 22 patients who were treated for early LBCA, LBCA occurred at 0–3 months post-infusion in 11/22 (50%) and at 3–6 months in 11/22 (50%). Nine patients received upfront HSCT for early LBCA, two of these died of TRM. Eight patients received maintenance regime chemotherapy due to a contraindication to HSCT, unsuitable donors or family preference [20]. No non-relapse mortality (NRM) was observed in this group, 5 patients are alive in molecular remission (one after subsequent consolidative HSCT). Four patients received a repeat tisagenlecleucel infusion with lymphodepletion, none achieved long term B cell aplasia, 2 are alive in molecular remission (Fig. 4). We identified 4 patients that had early LBCA but did not receive any further treatment. All relapsed eventually 1, 1.5, 5.8 and 24.5 months after B-cell recovery; 1 died of disease, 1 is alive with disease awaiting HSCT, and 2 are alive after HSCT.

The OS and EFS 1 year after CAR T failure for these 62 patients were 61.2% (95%CI: 49.3–75.8) and 55.3% (95%CI 43.6–70.2), respectively. The median OS was not reached, the median EFS was 14.8 months (95%CI 8.5-NR) (Fig. 5a).

a OS and EFS for the 62 responding patients that subsequently failed CAR T. b OS and (c) EFS stratified according to type of CAR T failure. MRD measurable residual disease emergence, LBCA early loss of B-cell aplasia.

As expected, OS was significantly associated with the reason for tisagenlecleucel failure in this group (p = 0.0014). In other words, there was a worse outcome for patients who suffered frank relapse (1-year 31.7%, 95%CI 14.4–69.7%), compared to patients who had MRD emergence (1-year 60.2%, 95%CI 40.9–88.7%; HR 0.31, 95%CI 0.12–0.8) or who received further treatment for early LBCA (1-year 81.3%, 95%CI 66.4–99.7%; HR 0.20, 95%CI 0.07–0.56) (Fig. 5b).

Although the difference was not statistically significant for EFS (p = 0.06), EFS following treatment for early LBCA (1-year 76.8%, 95%CI 60.8–96.9; HR 0.35, 95%CI 0.14–0.86) and MRD emergence (1-year 48.5%, 95%CI 30.5–76.9; HR 0.58, 95%CI 0.24–1.39) was better than that following frank relapse (31.7%, 95%CI 14.4–69.7) (Fig. 5c). With a median follow-up of 21.5 months after CAR T failure (IQR 12.2–30.5), 28/62 (45.2%) maintained/attained a further molecular remission at the time of data cut-off. In all 49 patients with disease emergence after tisagenlecleucel, emergence of CD19 positive disease (n = 26) had similar outcomes when compared to CD19 negative disease (n = 15) (OS HR 0.68, 95%CI 0.27–1.72; EFS HR 0.84, 95%CI 0.38–1.88).

We analysed the effect of post-tisagenlecleucel HSCT on outcomes. 33/62 (53.2%) patients received HSCT at a median 14.8 months (95%CI 9.7–NR) after tisagenlecleucel infusion. For 6 of these 33 patients this was a second and for 1 it was the third transplant. 6/33 (18.2%) died of TRM (Supplementary Table 5), mainly due to infection. Only 1 of these 6 patients had had a prior HSCT. There was no NRM amongst the 29/62 patients who did not receive a transplant post-tisagenlecleucel failure.

A multivariable OS regression model including type of CAR T failure, prior HSCT, and HSCT after tisagenlecleucel (Table 6), confirmed the better outcome of patients who received further therapy for early LBCA compared to those with frank relapse (HR 0.24, 95%CI 0.1–0.7). Pre-CAR HSCT emerged as a risk factor for OS and EFS after CAR T failure (OS HR 2.53 95%CI 1.1–6.0, p = 0.03; EFS HR 2.3, 95%CI 1.1–4.9, p = 0.03) (Table 6a).