Among the 71 patients who were excluded from the study during screening, 55 did not meet all clinical eligibility criteria, 8 decided not to participate, 4 did not complete the screening process, 2 had an adverse event, 1 died, and 1 had the treating physician decide against participation. Among the 38 enrolled patients who had other reasons for discontinuing participation in the study, 16 died before infusion, 16 had their treating physician decide against further participation, 3 had an adverse event, 2 decided against further participation, and 1 had a protocol deviation. Reasons for discontinuation such as death, physician’s decision, and patient’s decision were mainly related to disease progression as reported by the investigator. The full analysis set and safety set were made up of all the patients who received an infusion, including those treated with tisagenlecleucel manufactured in the United States (main cohort) and those treated with tisagenlecleucel manufactured in the European Union (cohort A).
Between July 2015 and the data cutoff date, December 8, 2017, a total of 238 patients were screened and 165 were enrolled (Figure 1). Of the enrolled patients, 111 (67%) received an infusion: 95 in the main cohort and 16 in cohort A (Figure 1, and the Methods section in the Supplementary Appendix); 4 patients (2%) were awaiting infusion at the time of analysis. Patients received infusions in either inpatient or outpatient settings. The median time from enrollment to infusion was 54 days (90% of patients received infusions between 30 days and 92 days after enrollment). The median time from infusion to data cutoff was 14 months (range, 0.1 to 26). The baseline characteristics of the enrolled patients and the patients who received an infusion were similar (Table 1, and Table S3 in the Supplementary Appendix); however, the patients who did not receive an infusion tended to have a lower performance status than those who did receive an infusion, and a greater proportion of the patients who did not receive an infusion had DLBCL that was refractory to the last therapy they received before enrollment.
Before infusion, 92% of the patients received bridging therapy, including combinations of rituximab (54%), gemcitabine (40%), etoposide (26%), dexamethasone (25%), cisplatin (19%), and cytarabine (19%), as well as newer agents such as ibrutinib (9%) and lenalidomide (7%). A total of 103 patients (93%) received lymphodepleting chemotherapy (73% received combination fludarabine–cyclophosphamide, and 20% received bendamustine). All 111 patients received a single infusion of tisagenlecleucel (median dose, 3.0×108 CAR-positive viable T cells; range, 0.1×108 to 6.0×108) (Table S4 in the Supplementary Appendix).
The best overall response rate was the combined percentage of patients who had a complete or partial response. The dashed vertical line indicates a rate of 20% (the null hypothesis was that the best overall response rate would be 20% or less). IPI denotes International Prognostic Index; an IPI score of less than 2 (i.e., fewer than two risk factors) indicates a low risk, a score of 2 a low–intermediate risk, a score of 3 a high–intermediate risk, and a score of 4 or 5 a high risk of death within 5 years.
The null hypothesis with regard to the primary end point (i.e., that the best overall response rate would be ≤20%) was rejected in the interim analysis (P<0.001).22,23 Among the 93 patients in the efficacy analysis set who had 3 months or more of follow-up or had discontinued participation in the study before 3 months, the best overall response rate was 52% (95% confidence interval [CI], 41 to 62): 40% of patients had a complete response, and 12% of patients had a partial response (Table S5 in the Supplementary Appendix). The rates of overall and complete response were 38% and 32%, respectively, at month 3 and were 33% and 29% at month 6. A high concordance (85%) was found between local and central assessments of response. Response rates did not differ substantially according to the type of lymphodepleting therapy received (Table S6 in the Supplementary Appendix), and univariate analyses showed a homogeneous and consistent treatment effect across major demographic and prognostic subgroups, including the subgroup based on disease response to previous therapy (Figure 2, and Fig. S1 in the Supplementary Appendix).
Of the 37 patients who had a complete response, 16 had either stable disease (4 patients) or a partial response (12 patients) 1 month after infusion that improved to a complete response in a median of 2 months (range, 1 to 17). A conversion from a partial to a complete response occurred in 54% of the patients (13 of 24), including in 2 patients who were confirmed to have a complete response by positron-emission–tomography scanning performed 15 to 17 months after their initial response. Among the 35 patients who were in remission at month 3, the estimated probability of remaining in remission at month 12 was 81% (95% CI, 63 to 91). In an intention-to-treat analysis that included all 165 enrolled patients, including patients who discontinued participation before tisagenlecleucel infusion (mostly as a result of disease progression and death), the overall response rate was 34% (95% CI, 27 to 42).
Panel A shows the duration of response (time from the date of first documented disease response [complete response or partial response] to the date of first documented progression or death due to diffuse large B-cell lymphoma) among the 48 patients in the main cohort who had a complete or partial response. Data from 26 patients were censored because at data cutoff the patients continued to not have an event, with a duration of response between 181 and 527 days. Data from 7 other patients were censored, 5 because the patient received new cancer therapy as deemed necessary by the treating investigator (for disease progression [3 patients], stable disease , or partial response  by local assessment), 1 because of withdrawn consent, and 1 because adequate radiologic assessment was no longer available. Panel B shows progression-free survival (time from the date of tisagenlecleucel infusion to the date of disease progression or death from any cause) for all 111 patients who received an infusion. Panel C shows progression-free survival among patients with a response, according to response status at month 3. Panel D shows overall survival (time from the date of tisagenlecleucel infusion to the date of death from any cause) for the 111 patients in the full analysis set (gray line) and for patients who had a complete response (blue line).
The median response duration has not been reached (95% CI, 10 months to not reached); however, 79% (95% CI, 60 to 89) of patients who had a complete response and 65% (95% CI, 49 to 78) of all patients who had a response are projected to remain relapse-free at 12 months after having a response (Figure 3A, and Fig. S1A in the Supplementary Appendix). Durable responses were observed for up to 18.4 months after infusion. No patient proceeded to undergo transplantation while having a response. Six patients who did not have a response proceeded to undergo hematopoietic stem-cell transplantation (five underwent allogeneic transplantation, and one underwent autologous transplantation followed by allogeneic transplantation).
The median progression-free survival has not been reached for patients who had a complete response (Figure 3B); the estimated rate of progression-free survival at 12 months was 83% among patients who had a complete or partial response at 3 months (Figure 3C). The median overall survival among patients who received an infusion was 12 months (95% CI, 7 months to not reached) (Figure 3D, and Fig. S1B in the Supplementary Appendix). The estimated probability of survival at month 12 was 49% (95% CI, 39 to 59) among all patients and 90% (95% CI, 74 to 96) among patients with a complete response. In an intention-to-treat analysis that included all 165 enrolled patients, the median overall survival from the time of enrollment was 8.3 months (95% CI, 5.8 to 11.7) and the estimated probability of survival at month 12 was 40% (95% CI, 32 to 49) (Fig. S2 in the Supplementary Appendix).
Tisagenlecleucel Expansion and Persistence
Similar mean in vivo expansion and concentration–time profiles of tisagenlecleucel, measured as transgene level, median time to maximum transgene level, and mean area under the concentration–time curve from day 0 to day 28 (AUC0–28d), were observed in patients who had a response and those who did not (Table S7 in the Supplementary Appendix). Thus, no apparent effect of exposure on clinical outcome was observed. Persistent CAR transgene levels were observed for up to 2 years after infusion in patients with durable responses (Fig. S3 in the Supplementary Appendix). No relationship between dose and maximal in vivo expansion was apparent, and clinical responses were observed across a wide range of doses.24
The most common adverse events of any grade were cytokine release syndrome (58%), anemia (48%), pyrexia (35%), decreased neutrophil count (34%), decreased platelet count (33%), decreased white-cell count (33%), and diarrhea (32%) (Table S8 in the Supplementary Appendix). Grade 3 or 4 adverse events of special interest within the first 8 weeks after infusion (Table 2) included cytokine release syndrome (22% of the patients, according to the University of Pennsylvania grading scale19), cytopenias not resolved by day 28 (32%) (see Table S9 in the Supplementary Appendix for information on the different types of prolonged cytopenias), infections (20%), neurologic events (12%), and febrile neutropenia (15%).
The median time from infusion to the onset of symptoms of cytokine release syndrome was 3 days (all patients except one had onset within 9 days), and the median duration was 7 days (range, 2 to 30). The median time to the onset of grade 3 or 4 cytokine release syndrome was 4 days (range, 2 to 8); 97% of cases had resolved by data cutoff. Overall, 14% of the patients received tocilizumab, and 10% received both tocilizumab and glucocorticoids. No patient received more than two doses of tocilizumab (5% received one dose and 9% received two doses). Patients with cytokine release syndrome received supportive care, including oxygen supplementation (24%), endotracheal intubation (7%), high-dose vasopressors19 (6%), and dialysis (5%); 24% were admitted to the intensive care unit. Infections concurrent with cytokine release syndrome occurred in 6% of the patients.
Neurologic events of any grade occurred in 21% of the patients within 8 weeks after infusion; the median time to onset was 6 days (range, 1 to 17), and the median duration was 14 days. Headaches (not classified as a nervous system disorder) occurred in 20% of the patients 8 weeks or less after infusion. A total of 13 patients (12%) had grade 3 or 4 events, the majority of which had resolved by data cutoff with supportive treatment in accordance with local guidelines (e.g., glucocorticoids). Nine patients with grade 3 or 4 neurologic events had concurrent cytokine release syndrome. No fatal cerebral edema was observed.
Only one patient had normal CD19+ B-cell counts in peripheral blood before tisagenlecleucel infusion (normal range, 80 to 616 per cubic millimeter); the majority had CD19+ B-cell counts below the lower limit of quantitation (0.2 per cubic millimeter) (Table S10 in the Supplementary Appendix). After infusion, six patients with ongoing complete response had CD19+ B-cell counts return to the normal range (five patients at >6 months after infusion and one patient at month 3). Intravenous immune globulin was administered at the local investigator’s discretion; 30% of patients who received a tisagenlecleucel infusion were treated with intravenous immune globulin after the infusion.
Three patients died within 30 days after infusion, all from lymphoma progression. No deaths after infusion were attributed to tisagenlecleucel by the investigators.
Quantitative immunofluorescence analysis was performed on preinfusion tumor tissues to measure the expression of CD19; CD3, PD-1, and PD-L1; and CD3, TIM3, and LAG3. Samples were obtained 1 month to 1 year (in 60 of 82 patients) or more than 1 year (in 22 of 82 patients) before tisagenlecleucel infusion (Table S11 in the Supplementary Appendix). Responses to tisagenlecleucel were observed in patients with tumor samples that had unequivocal CD19 expression (best overall response rate, 49%; 95% CI, 34 to 64) and low or negative CD19 expression (best overall response rate, 50%; 95% CI, 29 to 71) (Fig. S4A and Table S12 in the Supplementary Appendix).
We found no apparent differences between the best-overall-response groups in the median or mean PD-1–PD-L1 interaction scores (see the Methods section in the Supplementary Appendix) or in the percentage of cells expressing immune checkpoint–related proteins (percentage of total cells positive for PD-L1, PD-1, LAG3, or TIM3 and percentage of total CD3 T cells expressing PD-1, LAG3, or TIM3) at baseline (Fig. S4B through S4E in the Supplementary Appendix and data not shown). However, the 5 patients with the highest PD-1–PD-L1 interaction scores either did not have a response to tisagenlecleucel (4 patients) or had a relapse by month 3 (1 patient). Similarly, the 11 patients with the highest percentages of LAG3+ T cells (among total T cells) did not have a response to tisagenlecleucel (7 patients) or had a relapse within 3 to 6 months (4 patients, 2 after a complete response and 2 after a partial response) (see the Methods section and Table S13 in the Supplementary Appendix). A clear absence of response or an early relapse was not observed in patients with the highest percentages of PD-1+ T cells or TIM3+ T cells (among total T cells).