==> picture [215 x 22] intentionally omitted <==

Single Technology Appraisal

Axicabtagene ciloleucel for treating relapsed or refractory diffuse large B-cell lymphoma after 1 systemic treatment [ID1684]

Committee Papers

==> picture [215 x 22] intentionally omitted <==

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

SINGLE TECHNOLOGY APPRAISAL

Axicabtagene ciloleucel for treating relapsed or refractory diffuse large B-cell lymphoma after 1 systemic treatment [ID1684]

Contents:

The following documents are made available to consultees and commentators:

Access the final scope and final stakeholder list on the NICE website.

Pre-technical engagement documents

1. Company submission from Kite, a Gilead company

2. Clarification questions and company responses

3. Patient group, professional group and NHS organisation submissions from:

• a. Anthony Nolan

• b. Blood Cancer UK

• c. Lymphoma Action

• d. Royal College of Pathologists

• e. NHSE CAR-T tariff summary

• f. Gilead response to NHSE CAR-T tariff documents

4. Evidence Review Group report prepared by Aberdeen HTA Group

5. Evidence Review Group report – factual accuracy check

Post-technical engagement documents

6. Technical engagement response from company

7. Technical engagement responses and statements from experts:

• a. Dr Andrew McMillan – clinical expert, nominated by NHS England

• b. Dr Sridhar Chaganti – clinical expert, nominated by The Royal College of Pathologists

• c. Robert Cross – patient expert, nominated by Anthony Nolan

• d. Rebecca Hallam – patient expert, nominated by Anthony Nolan

8. Evidence Review Group critique of company response to technical engagement prepared by Aberdeen HTA Group

• a. ERG critique of company TE response

• b. ERG critique of CAR-T tariff

==> picture [215 x 22] intentionally omitted <==

Any information supplied to NICE which has been marked as confidential, has been redacted. All personal information has also been redacted.

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

Single technology appraisal

Axicabtagene ciloleucel for treating relapsed or refractory diffuse large B-cell lymphoma after 1 systemic therapy [ID1684]

Document B

Company evidence submission

March 2022

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 1 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 2 of 162

Contents

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 3 of 162

Tables and figures

Table 1: The decision problem ................................................................................... 8 Table 2: Technology being appraised ...................................................................... 10 Table 3: Studies providing insight into the prognosis of primary refractory or early relapse DLBCL patients intended for transplant ....................................................... 18 Table 4: Clinical effectiveness evidence ................................................................... 25 Table 5: Summary of trial methodology for ZUMA-7 ................................................ 28 Table 6: Baseline characteristics of patients in ZUMA-7 .......................................... 33 Table 7: Summary of statistical analyses for ZUMA-7 .............................................. 36 Table 8: Summary of ORR and best overall response per central assessment, FAS 42 Table 9: Overall summary of treatment-emergent adverse events, SAS .................. 56 Table 10: Incidence of TEAEs occurring in ≥ 10% of patients in either treatment group, SAS ............................................................................................................... 58 Table 11: Incidence of treatment-related TEAEs occurring in ≥ 10% of patients in either treatment arm, SAS ........................................................................................ 61 Table 12: Summary of treatment-emergent neurological events occurring in ≥ 5% of patients in either treatment group, SAS .................................................................... 63 Table 13: Summary of treatment-emergent CRS and CRS symptoms occurring in ≥ 5% of patients in the axi-cel group, SAS .................................................................. 64 Table 14: Summary of treatment-emergent cytopenia events in either treatment group, SAS ............................................................................................................... 65 Table 15: TEAEs observed in patients receiving CAR T-cell therapy ....................... 68 Table 16: End-of-life criteria ..................................................................................... 77 Table 17: Summary list of published cost-effectiveness studies .............................. 78 Table 18: Summary of previous TAs of CAR T-cell therapies in DLBCL .................. 79 Table 19: General model settings ............................................................................ 83 Table 20: Features of the economic analysis ........................................................... 84 Table 21: EFS, mixture cure model, implied cure fractions ...................................... 93 Table 22: EFS, mixture cure model AIC and BIC statistics and landmark survival estimates .................................................................................................................. 94 Table 23 Summary of OS results from ITT and standard RPSFTM analyses (stratified) ................................................................................................................. 97 Table 24 : Axi-cel OS, implied cure fractions .......................................................... 101 Table 25: OS mixture cure model AIC and BIC statistics and landmark survival estimates, crossover .............................................................................................. 103 Table 26: Axi-cel TTNT, implied cure fractions ....................................................... 105 Table 27: TTNT mixture cure model AIC and BIC statistics and landmark survival estimates ................................................................................................................ 107 Table 28: Health state utility values from prior NICE appraisals ............................. 111 Table 29: Adverse event data (Grade 3+, ZUMA-7) ............................................... 112 Table 30: Adverse event disutilities ........................................................................ 113 Table 31: UK general population utility[94] ................................................................ 115 Table 32: Summary of utility values for cost-effectiveness analysis ....................... 116 Table 33: Unit costs of leukapheresis ..................................................................... 119 Table 34: Bridging therapy cost calculations .......................................................... 120 Table 35: Unit costs conditioning chemotherapy .................................................... 120 Table 36: Conditioning chemotherapy cost calculations ......................................... 121 Table 37. Standard of care chemotherapy acquisition costs .................................. 124

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 4 of 162

Table 38: Chemotherapy administration costs ....................................................... 127 Table 39: Total costs for administration of each regimen ....................................... 128 Table 40: Stem cell harvest and reinfusion costs ................................................... 129 Table 41: Unit costs, BEAM (high dose chemotherapy) regimen ........................... 130 Table 42: Cost per cycle of BEAM ......................................................................... 130 Table 43: Pre- and post-event healthcare resource use unit costs and frequencies per model cycle ...................................................................................................... 131 Table 44: Grade 3+ adverse event rates, ZUMA-7 ................................................. 132 Table 45: Adverse event costs ............................................................................... 133 Table 46: Adverse event costs applied in the model .............................................. 135 Table 47: Subsequent treatments received in ZUMA-7 .......................................... 136 Table 48: Subsequent treatments applied in the base case (crossover adjusted) analysis .................................................................................................................. 137 Table 49: Subsequent therapy unit costs ............................................................... 137 Table 50: Subsequent therapy drug and administration costs ................................ 138 Table 51: Key model assumptions ......................................................................... 139 Table 52: Base-case results ................................................................................... 142 Table 53: Scenario analyses results....................................................................... 148 Table 54: Modelled median EFS and ZUMA-7 median EFS (central assessment, investigator-assessed) estimates for axi-cel and SoC ............................................ 151 Table 55: Modelled median OS and ZUMA-7 median OS estimates for axi-cel and SoC 151

Figure 1: Axi-cel anti-CD19 CAR construct and mode of action ............................... 13 Figure 2: Process of manufacturing and administering axi-cel ................................. 13 Figure 3: Estimated cure rates with current treatment for DLBCL ............................ 16 Figure 4: Clinical pathway of care for DLBCL and proposed axi-cel positioning ...... 22 Figure 5: Study scheme for ZUMA-7 ........................................................................ 27 Figure 6: Kaplan–Meier plot for EFS as per central assessment, FAS..................... 41 Figure 7: Kaplan–Meier plot for OS, FAS ................................................................. 44 Figure 8: Kaplan–Meier plot of OS – sensitivity analysis using RPSFT model, FAS 45 Figure 9: Kaplan–Meier plot for EFS per investigator assessment, FAS .................. 46 Figure 10: Kaplan–Meier plot for PFS per investigator assessment, FAS ................ 47 Figure 11: Kaplan–Meier plot for DOR per central assessment, FAS ...................... 48 Figure 12: Kaplan–Meier plot for mEFS per central assessment, FAS .................... 50 Figure 13: Kaplan–Meier plot of TTNT, FAS ............................................................ 51 Figure 14: Mean (95% CI) EORTC QLQ-C30 Global Health Status scores over time by treatment group, QoL analysis set ....................................................................... 52 Figure 15: Mean (95% CI) EORTC QLQ-C30 Physical Functioning scores over time by treatment group, QoL analysis set ....................................................................... 53 Figure 16: Mean (95% CI) EQ-5D-5L VAS scores over time by treatment group, QoL analysis set .............................................................................................................. 54 Figure 17: Mean (95% CI) EQ-5D-5L index scores over time by treatment group, QoL analysis set ....................................................................................................... 55 Figure 18: Model structure ....................................................................................... 80 Figure 19: Kaplan–Meier plot for EFS as per central assessment, FAS................... 87 Figure 20: Kaplan–Meier plot for OS, FAS ............................................................... 88 Figure 21: Kaplan–Meier plot of OS – analysis using RPSFT model, FAS .............. 88 Figure 22: Kaplan–Meier plot for TTNT, FAS ........................................................... 89

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 5 of 162

Figure 23: Axi-cel EFS, mixture cure models ........................................................... 93 Figure 24: SOC EFS, mixture cure models .............................................................. 94 Figure 25: Modelled base case EFS curves ............................................................. 95 Figure 26: OS estimates adjusted for crossover, compared to external datasets .... 98 Figure 27 : Axi-cel OS, mixture cure models .......................................................... 101 Figure 28 : SOC OS, mixture cure models, crossover............................................ 102 Figure 29: Modelled base case overall survival curves .......................................... 104 Figure 30 : Axi-cel TTNT, mixture cure models ...................................................... 106 Figure 31 : SOC TTNT, mixture cure models ......................................................... 106 Figure 32: Modelled base case TTNT curves ......................................................... 108 Figure 33: Lifetime Markov trace for axi-cel ........................................................... 143 Figure 34: Lifetime Markov trace for SOC .............................................................. 143 Figure 35: PSA scatter plot at £50,000 threshold ................................................... 144 Figure 36: Cost-effectiveness acceptability curve .................................................. 145 Figure 37: PSA convergence plot ........................................................................... 146 Figure 38: One-way sensitivity analysis, Tornado diagram .................................... 147

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 6 of 162

Decision problem, description of the technology and clinical care pathway

B.1.1. Decision problem

This submission focuses on part of the expected technology’s marketing authorisation, aligning the proposed population to the pivotal evidence base.

The decision problem addressed is therefore the potential value of axicabtagene ciloleucel (axi-cel; Yescarta[®] ) for the treatment of adults with primary refractory or early relapse (≤ 12 months) diffuse large B-cell lymphoma (DLBCL) who are intended for transplant. In this position, axi-cel would displace current second-line standard of care (SOC) of re-induction therapy followed by high-dose therapy (HDT) plus autologous stem-cell transplant (auto-SCT) consolidation in responders.

ZUMA-7 provides direct data of relevance to this decision problem and shows that in this poor prognosis patient group with high unmet need, axi-cel offers a three-fold increase in the number of patients receiving definitive therapy and a 2.5-fold increase in the number of patients living event-free for at least 2 years compared with current second-line SOC.[1]

Full details of the decision problem that the submission addresses are summarised in Table 1. Full details of the technology and health condition are provided in Sections B.1.2 and B.1.3; full details of the clinical effectiveness evidence are provided in Section B.2.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 7 of 162

Table 1: The decision problem

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 8 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 9 of 162

B.1.2. Description of the technology being appraised

A description of axicabtagene ciloleucel (axi-cel; Yescarta[®] ) is presented in Table 2.

The draft summary of product characteristics (SmPC) for the licence extension to the second-line setting is presented in Appendix C. The European Public Assessment Report (EPAR) can be provided on receipt.

Axi-cel was the first in a breakthrough class of chimeric antigen receptor (CAR) T- cell therapies that are manufactured from patients’ own T-cells and is engineered ex vivo to express antigen-specific CARs, enabling them to target and kill antigenexpressing tumour cells on return to the patient. The CAR construct used in axi-cel is a single-chain antibody fragment directed against CD19 and linked to CD3ζ and CD28 T-cell activating domains; CD19 is a B-cell-specific cell surface antigen ubiquitously expressed in B-cell malignancies.[3]

Axi-cel is given as a single infusion treatment. The median target timescale from collection of the patient’s T-cells by leukapheresis, through transportation to the manufacturing facility, product manufacture, and qualified person (QP) release in Europe is '''''' ''''''''''''.[4]

The axi-cel construct and mode of action is depicted in Figure 1. The manufacturing and administration process for axi-cel is depicted in Figure 2.

Table 2: Technology being appraised

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 10 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 11 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 12 of 162

Figure 1: Axi-cel anti-CD19 CAR construct and mode of action

==> picture [430 x 314] intentionally omitted <==

Key: CAR, chimeric antigen receptor; LTR, long terminal repeat; scFv, single-chain variable region fragment.

Figure 2: Process of manufacturing and administering axi-cel

==> picture [452 x 113] intentionally omitted <==

Key: CAR, chimeric antigen receptor.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 13 of 162

B.1.3. Health condition and position of the technology in the treatment pathway

B.1.3.1. Disease overview

Non-Hodgkin’s lymphoma (NHL) comprises a diverse group of cancers of the lymphatic system.[8] DLBCL is the most common type of NHL, accounting for approximately 40% of all NHL cases.[8] DLBCL is an aggressive, high-grade form of NHL, characterised by abnormal and enlarged B-cells that quickly grow and spread if left untreated.[8] An estimated 5,180 people are diagnosed with DLBCL each year in the UK.[9]

There are several different subtypes of DLBCL, which demonstrates the heterogeneity of the clinical and pathological features of this disease beyond B-cell abnormality. DLBCL, not otherwise specified (NOS) is defined by excluding unique features and is the most common subtype, estimated to account for over 80% of large B-cell lymphomas.[10] Rarer subtypes recognised by the World Health Organization (WHO) include: T-cell/histiocyte-rich large B-cell lymphoma; primary DLBCL of the central nervous system (CNS); primary cutaneous DLBCL; and Epstein-Barr virus (EBV)-positive DLBCL.[11] Double-hit or triple-hit lymphoma (i.e. lymphoma with MYC , BCL2 and/or BCL6 genetic aberrations) was also traditionally considered a DLBCL subtype, but it is now included in the new category of highgrade B-cell lymphoma (HGBL) in the most recent WHO classification of lymphomas.[11] There are no therapies specifically indicated for HGBL and there is no consensus on whether a different management approach is needed for these lymphoma types, although patients with double-hit or triple-hit lymphoma typically have a poor prognosis.[12-14] From this point in the document, DLBCL is used to describe patients with any DLBCL/HGBL subtype that aligns to the eligibility criteria of the pivotal trial supporting the use of axi-cel (see Section B.2).

DLBCL has a complex and multifactorial aetiology with several risk factors identified. These include: demographic characteristics such as body mass index; clinical characteristics such as weakened immune function; environmental factors such as carcinogen exposure; and genetic susceptibility.[15, 16] Most patients are at least 60 years old at diagnosis (median age at diagnosis estimates range from 61 to 70 years

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 14 of 162

across datasets) and almost all patients (~90%) present with advanced-stage disease (Ann Arbor III/IV).[9, 17-19] There is a slight male dominance in cases.[9, 17]

Following diagnosis of DLBCL, patients will undergo prognostic assessment via the International Prognostic Index (IPI) which considers: age (> 60 years = 1 risk factor); lactate dehydrogenase (LDH) levels (> upper limit of normal [ULN] = 1 risk factor); Ann Arbor disease staging (III/IV = 1 risk factor); performance status (> 1 = 1 risk factor); and spread of disease (extranodal sites of disease > 1 = 1 risk factor) to estimate a prognostic risk (low = 0–1 risk factors; high = 4–5 risk factors).[20] Additional poor prognostic factors include genetic factors (see previous note on double-hit or triple-hit lymphoma) and bulky disease (tumour diameter > 7.5 cm).[21] Further factors predicting prognosis at relapse are captured in the secondary ageadjusted IPI (sAAIPI), which considers three of the IPI risk factors (LDH, disease staging and performance status) to estimate a prognostic risk (low = 0 risk factors; high = 3 risk factors).[22]

B.1.3.2. Clinical outcomes

DLBCL is a curable disease with 80% of patients receiving frontline therapy of rituximab with cyclophosphamide, doxorubicin hydrochloride, vincristine and prednisolone (R-CHOP) with curative intent.[23] Despite this, frontline R-CHOP does not cure all patients; approximately 10–15% of patients develop primary refractory disease (i.e. an inadequate response to frontline treatment) and a further 20–25% patients relapse following treatment.[10] Outcomes remain poor for these patients in whom frontline treatment fails, particularly for those with primary refractory or early relapse disease.[24-27]

The only potentially curative treatment option available at first relapse is currently HDT-auto-SCT, which can only follow a response to re-induction therapy (Figure 4). Due to advanced age and coexisting medical conditions, only half of relapsed or refractory (r/r) DLBCL patients are fit enough to be considered for such high-intensity treatment, and only half again go on to receive auto-SCT.[10, 26, 28] Reasons why r/r DLBCL patients intended for transplant may not receive auto-SCT include: insufficient response or intolerance to re-induction therapy; intolerance to HDT; progressive disease during re-induction therapy or HDT; and stem cell mobilisation failure. In the primary refractory or early relapse patient group intended for

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 15 of 162

transplant, there is a higher risk of one or more of these factors preventing auto-SCT receipt, and closer to two-thirds of patients in this group will not receive auto-SCT despite intent.[1, 27, 29] Primarily, patients refractory to or relapsing quickly after frontline R-CHOP have a lower chance of sufficient response to chemotherapy-based reinduction therapy to accommodate HDT-auto-SCT and a higher chance of platinumsalvage toxicity.[30] Even for patients who do receive auto-SCT there is no guarantee of cure, with approximately half of r/r DLBCL patients treated with auto-SCT experiencing further relapse.[26, 28] It has previously been estimated that out of 100 r/r DLBCL patients, only 10 will be cured with current second-line care, as depicted in Figure 3.

Figure 3: Estimated cure rates with current treatment for DLBCL

==> picture [312 x 396] intentionally omitted <==

Key: ASCT, autologous stem cell transplant; DLBCL, diffuse large B-cell lymphoma; R-CHOP, rituximab with cyclophosphamide, doxorubicin hydrochloride, vincristine and prednisolone. Source: Friedberg 2011.[31]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 16 of 162

Historical trial data that are specific to the primary refractory or early relapse DLBCL group intended for transplant and reflective of current pathways of care are limited. However, some studies provide insight into the poor prognosis of this population, as summarised in Table 3. Further data from historical randomised controlled trials (RCT) are provided in Appendix N. Event-free survival (EFS) rates were approximately 16% at 2 years and 13% at 3 years in the subgroup of patients with primary refractory or early relapse DLBCL who had received frontline rituximabbased therapy in the CORAL trial. Similarly, this was 17% at 2 years in all patients enrolled in the ORCHARRD trial (r/r DLBCL despite frontline rituximab-based therapy) which included a majority (71%) of patients with either primary refractory or early-relapse disease.[25, 27] Median overall survival (OS) in this primary refractory or early relapse DLBCL group of the ORCHARRD trial (all of whom received frontline rituximab-based therapy) was less than 1 year (estimated at approximately 9 months from the Kaplan–Meier curve) and the 2-year OS rate was 31%.[27] Median OS in the primary refractory DLBCL group of the SCHOLAR-1 study was 7.1 months and the 2-year OS rate was 24%; in the overall population (refractory to frontline or later-line therapy or relapsed ≤ 12 months from auto-SCT), median OS was 6.3 months, and the 2-year OS rate was 20%.[24]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 17 of 162

Table 3: Studies providing insight into the prognosis of primary refractory or early relapse DLBCL patients intended for transplant

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 18 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 19 of 162

B.1.3.3. Burden of disease

The most common symptoms of DLBCL include enlarged lymph nodes and general ‘B symptoms’ that include night sweats, fever, involuntary weight loss and unexplained itching.[33] Other physical symptoms may depend on where DLBCL appears and spreads. For example, patients may experience breathlessness if lymphoma is affecting nodes in their chest.[33]

There is also an emotional burden associated with a diagnosis of DLBCL, and this is exacerbated for patients who experience treatment inefficacy[34] , such as primary refractory or early relapse disease. Ineligibility to receive effective treatment can also impact patients’ emotional status. For example, patients who go through the process of assessment and preparation for auto-SCT, but who then do not receive auto-SCT treatment, may experience a range of negative emotions. The emotional burden extends to carers of patients with r/r DLBCL who are often trying to support the patient with their feelings while coping with their own, which can lead to high levels of anxiety and stress.[35]

Patients undergoing treatment can experience additional physical and emotional symptoms relating to treatment side effects, and these are shown to negatively impact health-related quality of life (HRQL).[36] Patients undergoing stem cell transplant are at particular risk of treatment side effects that can adversely affect HRQL over the long term. A study investigating the HRQL of long-term survivors after auto-SCT using the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ)-C30 questionnaire showed that global health status did not return to general population levels until 4 years posttransplant.[37] Emotional, physical, role, social and cognitive functions were also all shown to be negatively impacted over the long term.

In addition to the long-term impact on HRQL, auto-SCT survivors are also at risk of late effects, such as secondary malignancies and cardiac or pulmonary toxicity that can be fatal, with late effects reported in around 10% of patients.[38, 39] In a retrospective long-term follow-up of r/r DLBCL patients undergoing auto-SCT in a US haematology clinic (n = 309), while relapse was initially the more likely cause of death, non-relapse mortality became the major cause of death after 8 years.[40]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 20 of 162

B.1.3.4. Clinical pathway of care

The clinical pathway of care for DLBCL is depicted in Figure 4.

Frontline treatment consists of R-CHOP for nearly all newly diagnosed patients treated with curative intent, with the number of cycles determined according to baseline prognosis.[10, 21] Consolidation radiotherapy and CNS prophylaxis may also be considered alongside R-CHOP for patients with bulky disease or who are at risk of CNS lymphoma, respectively.[10, 21, 41]

At first relapse, patients who are who are fit enough to tolerate intensive therapy are offered further multi-agent immunochemotherapy (re-induction therapy) to try to obtain sufficient response for HDT-auto-SCT consolidation.[10, 21, 41] The most common re-induction therapy regimens used at first relapse are: rituximab with dexamethasone, high-dose cytarabine, and cisplatin (R-DHAP); rituximab with ifosfamide, carboplatin, and etoposide (R-ICE); rituximab with etoposide, methylprednisolone, cytarabine and cisplatin (R-ESHAP); and rituximab with gemcitabine, dexamethasone, and cisplatin (R-GDP).[10, 17, 29, 41]

Allogeneic stem cell transplant (allo-SCT) can be considered instead of auto-SCT where stem cell harvesting is not possible, or for people with chemo-sensitive DLBCL that relapses after auto-SCT. Additional treatment options at second relapse (i.e. for people who have received two prior lines of therapy) include the CAR T-cell therapies axi-cel or tisagenlecleucel (tisa-cel), which are currently available through the Cancer Drugs Fund (CDF).[41] Patients who are not considered eligible for CAR T- cell therapy may be treated with further chemotherapy, enrolled to a clinical trial (if available) or managed with palliative or best supportive care.[17]

Axi-cel offers an alternative second-line treatment option to re-induction therapy plus HDT-auto-SCT in responders for patients with primary refractory or early relapse DLBCL who are intended for transplant (aligning with the ZUMA-7 trial population), as depicted in Figure 4.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 21 of 162

Figure 4: Clinical pathway of care for DLBCL and proposed axi-cel positioning

==> picture [452 x 274] intentionally omitted <==

Key: auto-SCT, autologous stem cell transplant; BSC, best supportive care; DLBCL, diffuse large B- cell lymphoma; HDT, high dose therapy; R-CHOP, rituximab with cyclophosphamide, doxorubicin, vincristine and prednisolone.

Notes: * Pixantrone is rarely used in clinical practice but is included here for completeness. ^ An allogeneic transplant can also be considered instead of auto-SCT where stem cell harvesting is not possible.

Green refers to the target population for axi-cel.

Blue refers to the proposed positioning of axi-cel at second-line.

Grey refers to treatments currently recommended within the Cancer Drugs Fund. Source: Adapted from the NICE pathway for treating DLBCL[41] and the British Society for Haematology guidelines for the management of DLBCL.[21]

B.1.3.5. Unmet need

DLBCL is a curable disease, but not all patients achieve cure within the current management pathway. Outcomes remain poor for patients for whom frontline treatment fails, particularly patients with primary refractory or early relapse disease. In this difficult-to-treat group, only a third of patients intended for transplant receive auto-SCT at first relapse, and the overall cure rate is expected to fall between 13‒ 17% based on EFS rates reported in historical trials of current second-line care (reinduction therapy plus HDT-auto-SCT in responders) (Table 3). Those who do receive auto-SCT are also at risk of persistent and late side effects that can negatively impact long-term quality of life.[37-40]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 22 of 162

During scoping consultation, the urgency of this appraisal in consideration of the significant unmet need in second-line therapy for r/r DLBCL was highlighted by commentators, where the specific target population (primary refractory or early relapse patients) was described as having a ‘ dismal outcome’.[42] In Kite-sponsored consultation settings, one clinical expert in the UK described the current treatment option of auto-SCT as ‘unpleasant’ and with ‘modest expectation of success’ in all second-line patients, but particularly in those with primary refractory or early relapse disease; another described auto-SCT as ‘cruel punishment’, adding that current second-line care ‘fails the majority’ of primary refractory or early relapse patients.[29, 43]

CAR T-cell therapy is an alternative, potentially curative, treatment option to HDTauto-SCT for r/r DLBCL, but it is currently only available at the third- or later-line setting. By the time patients reach this setting, they have already received two intensive lines of treatment with suboptimal response and may not be fit enough (or willing) to receive another.[29] Generally we would expect decreased tumour burden and comorbidities in second-line versus third-line patients, and higher general and T- cell fitness.[30] This was also acknowledged by commentators during the scoping consultation, who noted that ‘although patients can potentially access CAR T-cell therapy, third-line disease progression may result in poor performance score (i.e. not 0‒1) and hence be ineligible for this treatment modality’ and that ‘second-line rather than third-line could result in improved access, with improvement in the outcomes measured’.[42]

ZUMA-7 directly investigates the potential benefit of treating primary refractory or early relapse DLBCL patients with axi-cel versus HDT-auto-SCT in the second-line setting and shows that patients intended for axi-cel treatment are three times more likely to receive definitive therapy than patients intended for transplant, and are 2-3 times more likely to live event-free for at least 2 years (see Section B.2).[1] The availability of axi-cel for primary refractory or early relapse DLBCL patients intended for transplant would not only increase the definitive therapy receipt and associated cure rates, but could also reduce the negative long-term physiological and psychological impacts of current second-line care.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 23 of 162

B.1.4. Equality considerations

There is an age inequality issue with current second-line SOC in that auto-SCT is not considered a treatment option for older patients, with a typical ‘cut-off’ age between 65 and 70 years.[29, 43] Such an age restriction would not be applied to axi-cel and therefore its introduction could help to reduce this current age inequality.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 24 of 162

==> picture [29 x 13] intentionally omitted <==

Clinical effectiveness

B.2.1. Identification and selection of relevant studies

See Appendix D for full details of the process and methods used to identify and select the clinical evidence relevant to the technology being appraised.

B.2.2. List of relevant clinical effectiveness evidence

Table 4 summarises the evidence that supports axi-cel for the treatment of adults with primary refractory or early relapse DLBCL who are intended for transplant.

Table 4: Clinical effectiveness evidence

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 25 of 162

B.2.3. Summary of methodology of the relevant clinical effectiveness evidence

Table 5 provides a summary of the trial methodology for ZUMA-7.

ZUMA-7 is a Phase III, randomised, open-label, parallel assignment trial that evaluates the efficacy of axi-cel versus SOC therapy in adults with relapsed/refractory DLBCL. Eligible patients were at least 18 years of age. They had histologically confirmed DLBCL that was refractory to frontline treatment (no complete response [CR]), or that had relapsed from CR ≤ 12 months after the completion of frontline chemoimmunotherapy. This included an anti-CD20 monoclonal antibody and anthracycline-containing regimen. The patients also intended to proceed to HDT and auto-SCT.[1]

Patients were randomised in a 1:1 ratio to receive axi-cel or SOC.[1] . Randomisation was stratified according to response to frontline therapy (refractory versus relapsed disease) and the sAAIPI (0 or 1 risk factor versus 2 or 3 risk factors). Although crossover between the treatment groups was not permitted within the trial, patients who had no response to SOC could receive subsequent cellular immunotherapy outside of the trial protocol (reflecting ‘treatment switching’).OS outcomes in the SOC arm are therefore augmented and reflect a treatment sequence that includes CAR T- cell therapy at third- or later-line settings Each patient was to proceed through the study periods depicted in Figure 5.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 26 of 162

Figure 5: Study scheme for ZUMA-7

==> picture [445 x 209] intentionally omitted <==

Key: auto-SCT, autologous stem cell transplant; CAR, chimeric antigen receptor; HDT, high-dose therapy; R-DHAP, rituximab + dexamethasone, high-dose cytarabine and cisplatin; R-ESHAP, rituximab + etoposide, methylprednisolone, cytarabine, cisplatin; R-GDP, rituximab + gemcitabine, dexamethasone and cisplatin/carboplatin; R-ICE, rituximab + ifosfamide, carboplatin, and etoposide; SCT, stem cell transplant; SOCT, standard of care therapy.

Notes:[a] At the discretion of the investigator, corticosteroid bridging therapy could have been considered for patients with high disease burden at screening.[b] Minimum observation period of 7 days unless otherwise required by country regulatory agencies (e.g. 10 days for patients treated in Germany, Switzerland, and France).[c] Disease assessments were to be calculated from the date of randomisation and not the date of dosing with axi-cel or SOCT. Independent of the treatment arm, study procedures and disease assessments were to occur at the same protocol-defined timepoints. Source: ZUMA-7 CSR.[44]

The primary endpoint of the ZUMA-7 trial was EFS, defined as the time from randomisation to the earliest date of disease progression per the Lugano Classification[45] as determined by blinded central assessment, commencement of new lymphoma therapy, death from any cause, or a best response of stable disease (SD) up to and including the response on the Day 150 assessment after randomisation.[1] Secondary endpoints included: objective response rate (ORR); OS; progression-free survival (PFS); duration of response (DOR); modified EFS (mEFS); safety; and patient-reported outcome (PRO) endpoints (Table 5).

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 27 of 162

Table 5: Summary of trial methodology for ZUMA-7

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 28 of 162

• No known history or suspicion of CNS involvement by lymphoma

• ECOG performance status of 0 or 1

• Adequate bone marrow, renal, hepatic, pulmonary and cardiac function defined as:

  • Absolute neutrophil count ≥ 1000/μL

  • Platelet count ≥ 75,000/μL

  • Absolute lymphocyte count ≥ 100/μL

  • Creatinine clearance (as estimated by Cockcroft Gault) ≥ 60 mL/min

  • Serum alanine aminotransferase/aspartate aminotransferase (ALT/AST) ≤ 2.5 ULN

  • Total bilirubin ≤ 1.5 mg/dL, except in patients with Gilbert's syndrome

  • Cardiac ejection fraction ≥ 50%, no evidence of pericardial effusion as determined by an ECHO, and no clinically significant ECG findings

  • No clinically significant pleural effusion

  • Baseline oxygen saturation > 92% on room air

Key exclusion criteria:

• History of malignancy other than non-melanoma skin cancer or carcinoma in situ (e.g. cervix, bladder, breast) unless diseasefree for at least 3 years

• Received more than one line of therapy for DLBCL

• History of auto-SCT or allo-SCT

• Presence of fungal, bacterial, viral, or other infection that was uncontrolled or requiring IV antimicrobials for management

• Known history of infection with HIV, hepatitis B or hepatitis C. If there was a positive history of treated hepatitis B or hepatitis C, the viral load must have been undetectable per quantitative polymerase chain reaction and/or nucleic acid testing

• Patients with detectable cerebrospinal fluid malignant cells or known brain metastases, or with a history of cerebrospinal fluid malignant cells or brain metastases

• History or presence of non-malignant CNS disorder, such as seizure disorder, cerebrovascular ischaemia/haemorrhage, dementia, cerebellar disease, or any autoimmune disease with CNS involvement

• Presence of any indwelling line or drain. Dedicated central venous access catheters, such as a Port-a-Cath or Hickman catheter, were permitted

• History of myocardial infarction, cardiac angioplasty or stenting, unstable angina, New York Heart Association Class II or greater congestive heart failure, or other clinically significant cardiac disease within 12 months before enrolment

• History of symptomatic deep vein thrombosis or pulmonary embolism within 6 months before enrolment

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 29 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 30 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 31 of 162

B.2.3.1. Baseline characteristics

Table 6 provides a summary of baseline characteristics, including demographic and clinical characteristics.

The characteristics of the patients at baseline were generally balanced between the two treatment groups.[1] The median age was 59 years and 30% of the patients were 65 years of age or older. In total, 74% of patients had primary refractory disease, with 26% experiencing relapse ≤ 12 months after the initiation or completion of frontline therapy. Almost half of patients (45%) had a high sAAIPI with two or three risk factors and the majority (79%) had stage III or IV disease. Differences of ≥ 10%

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 32 of 162

were observed between the axi-cel and SOC arms for sex (male: 61% versus 71%, respectively) and extranodal disease (''''''''''' versus ''''''''''', respectively).[44]

Table 6: Baseline characteristics of patients in ZUMA-7

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 33 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 34 of 162

B.2.4. Statistical analysis and definition of study groups in the relevant clinical effectiveness evidence

Table 7 provides a summary of the statistical analysis for ZUMA-7.

The study was primarily designed to investigate the EFS in patients with r/r DLBCL treated with axi-cel or SOC, with a hypothesised target of 50% improvement in the median EFS time for axi-cel compared with SOC.[44] Approximately 350 patients were to be randomised (175 patients per treatment group) to achieve approximately 90% power at the 1-sided 2.5% significance level to detect a 50% improvement in EFS. To preserve the overall significance level, statistical testing of the primary and key secondary efficacy endpoints followed a hierarchical scheme:[44]

• EFS was to be tested at the primary analysis using a log-rank test stratified by randomisation factors to test the null hypothesis of no difference in EFS

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 35 of 162

• Conditional on a statistically significant improvement in EFS, ORR was to be tested at the time of the primary EFS analysis. ORR was to be tested with a stratified Cochran-Mantel-Haenszel (CMH) test using randomisation factors

• Conditional on a statistically significant improvement in EFS and ORR, OS was to be tested up to three times:

  • The first interim analysis of OS was to be tested at the time of the primary EFS analysis

  • The second interim analysis of OS was to be tested when approximately 160 deaths had been observed, or no later than four years after the first patient was randomised

  • The primary analysis of OS was to be tested when approximately 210 deaths had been observed, or no later than five years after the first patient was randomised

The primary analysis was planned to occur when all patients had the opportunity to be followed for the Month 9 disease assessment (i.e. the Month 9 timepoint had passed for all patients) and 250 EFS events had been observed by blinded central assessment.[44] This submission presents data from the primary analysis of EFS with a data cut-off date of 18 March 2021. The median potential follow-up time was 24.9 months, and the median actual follow-up time was '''''''''' months.[1, 44] The full analysis set (FAS) was used for the primary efficacy analysis. An algorithm included in the statistical analysis plan (SAP) was to be used to impute partial or missing event dates.

Table 7: Summary of statistical analyses for ZUMA-7

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 36 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 37 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 38 of 162

B.2.4.1. Patient disposition data

At the data cut-off date (18 March 2021), 359 patients with primary refractory or early relapse DLBCL intended for transplant were enrolled, of which 180 patients were randomised to the axi-cel group and 179 patients were randomised to the SOC group.[1] Among the patients in the axi-cel group, 178 (99%) underwent leukapheresis and 170 (94%) received axi-cel. Six patients received neither lymphodepleting chemotherapy nor axi-cel, and two patients received lymphodepleting chemotherapy but not axi-cel for the following reported reasons:

• Adverse events (AEs; n = 4)

• Death (n = 2)

• Disease progression (n = 1)

• Other reason (n = 1)

Axi-cel was successfully manufactured for all the patients who underwent leukapheresis, and 65 patients (36%) received bridging therapy with glucocorticoids while awaiting axi-cel.[1] Among the 170 patients who received axi-cel, the median time from randomisation to leukapheresis was ''' days (range: ''''''''''''), the median time from leukapheresis to delivery of axi-cel to the study site was ''''' days (range: ''''''''''''''''') and the median time from leukapheresis to axi-cel administration was '''''' days (range: ''''''''''''''''').[44] Overall, the median time from randomisation to axi-cel infusion was 29 days (IQR: 27–34).[1] After axi-cel treatment, '''''''''' patients who had a response and later progressed were retreated with axi-cel.[44]

Among the patients in the SOC group, 168 (94%) received platinum-based chemotherapy (R-ICE, 84 [50%]; R-ESHAP, 5 [3%]; R-GDP, 42 [25%]; R-DHAP/RDHAX, 37 [22%]), and 64 (36%) received high-dose chemotherapy and underwent auto-SCT (including two patients who underwent auto-SCT outside the protocol).[1]

The CONSORT diagrams for the ZUMA-7 study are presented in Appendix D.

B.2.5. Quality assessment of the relevant clinical effectiveness evidence

A quality assessment of the ZUMA-7 study was conducted using the NICE checklist; the full details of this checklist are provided in Appendix D.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 39 of 162

The study was approved by the institutional review board and independent ethics committee and was conducted according to Good Clinical Practice. Overall, the study is considered to be a methodologically robust and high-quality study with a comprehensive approach to patient allocation, control of confounding factors, and an overall low risk of bias.

The ZUMA-7 study required open-label treatment due to the autologous cellular therapy nature of axi-cel. Although the primary analysis included a blinded central assessment to minimise bias, this open-label design did result in a small proportion (< 5%) of patients who were randomised to the SOC arm withdrawing consent before receiving treatment. They were therefore immediately censored in the time-to-event analyses that were conducted according to the intention-to-treat principle.

Disease assessments were conducted in line with recommended and accepted classification systems. The outcomes that were measured reflected established trial outcomes within the DLBCL setting and those relevant to patients and healthcare providers. Importantly, the ZUMA-7 study provides applicable data to the intended use of axi-cel in clinical practice and the decision problem under appraisal. This is further discussed in Section B.2.13.

B.2.6. Clinical effectiveness results of the relevant trials

B.2.6.1. Primary efficacy endpoint

B.2.6.1.1. EFS per central assessment

At the time of the data cut-off (18 March 2021), ''''''''' EFS events by blinded central assessment occurred for ''''''''' patients (''''''%) in the axi-cel group and '''''''''' patients (''''''%) in the SOC group.[44] Axi-cel treatment was superior to SOC, with a stratified hazard ratio (HR) of 0.40 (95% confidence interval [CI]: 0.31, 0.51; p < 0.001; see Figure 6).[1]

The median EFS was significantly longer in the axi-cel group (8.3 months; 95% CI: 4.5, 15.8) than in the SOC group (2.0 months; 95% CI: 1.6, 2.8).[1] The estimated EFS at 24 months was 41% (95% CI: 33, 48) in the axi-cel group compared with 16% (95% CI: 11, 22) in the SOC group (Appendix L).[1] The median follow-up time for EFS

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 40 of 162

using the reverse Kaplan–Meier method was ''''''''''' months in the axi-cel group and '''''''''' months in the SOC group.[44]

The most common EFS events in either the axi-cel or SOC groups were disease progression ('''''% and ''''''%, respectively), new lymphoma therapy ('''% and ''''''%, respectively) and death from any cause ('''% and '''%, respectively).[44]

Findings of the EFS sensitivity analyses were supportive of and consistent with results for the primary analysis of EFS (Appendix L).

Figure 6: Kaplan–Meier plot for EFS as per central assessment, FAS

==> picture [452 x 244] intentionally omitted <==

Key: EFS, event-free survival; FAS, full analysis set. Source: Locke et al. 2021.[1]

B.2.6.2. Key secondary efficacy endpoints

B.2.6.2.1. ORR per central assessment

The ORR for patients in the axi-cel group was 83% (n = 150/180) compared with 50% (n = 90/179) for patients in the SOC group, with a difference between treatment groups of 33% (see Table 8).[1] The odds ratio comparing axi-cel with the SOC group was significantly in favour of axi-cel (OR: '''''''''', 95% CI: '''''''''', '''''''''''; p ''' '''''''''''''''''').[44] CR rates in the axi-cel and SOC groups were 65% (n = 117/180) and 32% (n = 58/179), respectively, and partial response (PR) rates were 18% (n = 33/180) and 18% (n = 32/179), respectively.[1]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 41 of 162

Improvement of response after the first evaluable disease assessment per central assessment occurred in both the axi-cel and SOC groups: improvement from PR to CR occurred for '''''' patients (''''''%) and '''''' patients ('''%), respectively; improvement from SD to CR occurred for ''''''' patients ('''%) and ''''''''' patient ('''%), respectively; and improvement from SD to PR occurred for ''''''''' patients ('''%) and ''''''''' patient (''''%), respectively.[44]

Table 8: Summary of ORR and best overall response per central assessment, FAS

B.2.6.2.2. ORR per investigator assessment

ORR per investigator assessment had a high concordance with central assessment (overall '''''''%; κ = ''''''''''; 95% CI: '''''''''''', ''''''''''; see Appendix L).[44]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 42 of 162

A summary of ORR, best overall response and concordance with central assessment is provided in Appendix L.

B.2.6.2.3. OS

At the time of analysis, 72 deaths in the axi-cel group (40%) and 81 deaths in the SOC group (45%) were reported.[1] The median OS, evaluated as an interim analysis, was not reached (95% CI: 28.3 months, not estimable [NE]) in the axi-cel group and was 35.1 months (95% CI: 18.5, NE) in the SOC group (see Figure 7). No statistically significant difference between the treatment groups was observed (HR: 0.73; 95% CI: 0.53, 1.01; p = 0.054).

In the interim analysis, the estimated OS at 2 years was 61% in the axi-cel group and 52% in the SOC group.[1] The median follow-up time for OS using the reverse Kaplan–Meier method was '''''''''''' months in the axi-cel group and '''''''''''' months in the SOC group.[44]

A total of 56% of the patients in the SOC group received subsequent cellular immunotherapy off-protocol.[1] OS outcomes are therefore augmented and reflect a treatment sequence that includes CAR T-cell therapy at third- or later-line settings. To address the confounding effect of off-protocol treatment switching, sensitivity analyses of OS that adjusted for crossover were conducted. Results from the sensitivity analysis showed a difference in OS in favour of axi-cel (HR: 0.58; 95% CI: 0.42, 0.81; see Figure 8) with the rank-preserving structural failure time (RPSFT) method. An additional analysis, which was conducted using the inverse probability of censoring weights model, showed a stratified HR of 0.70 (95% CI: 0.46, 1.05).

A summary of the pre-planned sensitivity analyses is provided in Appendix L. Further post-hoc sensitivity analyses that consider NICE recommendations for adjusting survival estimates in the presence of treatment switching were conducted for the economic modelling and are presented in Section B.3.3.4.1.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 43 of 162

Figure 7: Kaplan–Meier plot for OS, FAS

==> picture [452 x 223] intentionally omitted <==

Key: FAS, full analysis set; OS, overall survival. Source: Locke et al. 2021[1]

Figure 8: Kaplan–Meier plot of OS – sensitivity analysis using RPSFT model,

FAS

==> picture [452 x 227] intentionally omitted <==

Key: CI, confidence interval; FAS, full analysis set; HR, hazard ratio; mo, months; NE, not estimable; NR, not reached; OS, overall survival; RPSFT, rank-preserving structural failure time; SOC, standard of care.

Source: Locke et al. 2021.[1]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 44 of 162

B.2.6.3. Additional secondary efficacy endpoints

B.2.6.3.1. EFS per investigator assessment

At the time of the data cut-off (18 March 2021), ''''''''' investigator-assessed EFS events occurred for ''''''''' patients (''''''%) in the axi-cel group and '''''''''' patients (''''''%) in the SOC group.[44] Axi-cel treatment was superior to SOC, with a stratified HR of '''''''''' (95% CI: '''''''''', ''''''''''; see Figure 9).

The median EFS was significantly longer in the axi-cel group ('''''''''''' months; 95% CI: ''''''', '''''''''') than the SOC group (''''''' months; 95% CI: ''''''''', '''''''').[44] The estimated EFS at 24 months was '''''''% (95% CI: ''''''', ''''''') in the axi-cel group compared with ''''''% (95% CI: ''''''', ''''''') in the SOC group (see Appendix L). The median follow-up time for EFS using the reverse Kaplan–Meier method was '''''''''' months in the axi-cel group and ''''''''''' months in the SOC group.

The most common EFS events in either the axi-cel or SOC group were disease progression (''''''% and '''''''%, respectively), new lymphoma therapy (''''% and ''''''%, respectively) and death from any cause (''''% and ''''%, respectively).[44]

EFS per investigator assessment had a high concordance with central assessment (overall '''''''%; κ = '''''''''''; 95% CI: ''''''''''', ''''''''''; see Appendix L).[44]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 45 of 162

Figure 9: Kaplan–Meier plot for EFS per investigator assessment, FAS

==> picture [453 x 104] intentionally omitted <==

==> picture [453 x 105] intentionally omitted <==

Key: CI, confidence interval; EFS, event-free survival; FAS, full analysis set; HR, hazard ratio; SOCT, standard of care therapy. Source: Figure 10. ZUMA-7 CSR.[44]

B.2.6.3.2. PFS per investigator assessment

The median PFS was 14.7 months (95% CI: 5.4, NE) in the axi-cel group and 3.7 months (95% CI: 2.9, 5.3) in the SOC group (HR: 0.49; 95% CI: 0.37, 0.65; see Figure 10).[1] The estimated PFS at 24 months was 46% (95% CI: 38, 53) in the axicel group and 27% (95% CI: 20, 35) in the SOC group (Appendix L).[1] The median follow-up time for PFS using the reverse Kaplan–Meier method was '''''''''' months (95% CI: '''''''''', ''''''''''') in the axi-cel group and '''''''''' months (95% CI: '''''''''', '''''''''') in the SOC group.[44]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 46 of 162

Figure 10: Kaplan–Meier plot for PFS per investigator assessment, FAS

==> picture [460 x 175] intentionally omitted <==

Key: CI, confidence interval; FAS, full analysis set; mo, months; NE, not estimable; PFS, progressionfree survival.

Source: Locke et al. 2021.[1]

B.2.6.3.3. PFS per central assessment

PFS results per central assessment were consistent with those per investigator assessment. A summary of PFS per central assessment is provided in Appendix L.

B.2.6.3.4. DOR per central assessment

The median time to first objective response for patients who achieved a CR or PR per central assessment in the axi-cel group (n = 150/180) or the SOC group (n = 90/179) was '''''''' months (range: ''''''''''''''''''') and '''''''''' months (range: '''''''''''''''''), respectively (see Appendix L).[44] The median DOR in all responders for the axi-cel group was '''''''''' months (95% CI: '''''''''''', ''''''''') compared with ''''''' months (95% CI: ''''''', ''''''') for the SOC group (stratified HR: ''''''''''''; 95% CI: '''''''''', ''''''''''; see Figure 11). The median follow-up time for DOR using the reverse Kaplan–Meier method was ''''''''''' months (95% CI: '''''''''', '''''''''') in the axi-cel group and '''''''''' months (95% CI: '''''''''', '''''''''') in the SOC group.

The proportion of responding patients with an ongoing response at the time of data cut-off (18 March 2021) was '''''''''' in the axi-cel group compared with '''''''''' in the SOC group.[44] The estimated percentage of responding patients who remained in response at 24 months was ''''''% (95% CI: '''''', '''''') in the axi-cel group compared with ''''''% (95% CI: ''''', '''''') in the SOC group (see Appendix L).

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 47 of 162

Among patients who had a best overall response of CR in the axi-cel group (n = 117/180) and SOC group (n = 58/179), the median DOR was '''''''''''' months (95% CI: ''''''''''', ''''''') and ''''''' ''''''''''''''''''''' (95% CI: '''''''' '''''''''''''''''', '''''''''), respectively, with median follow-up times using the reverse Kaplan–Meier method of '''''''''''' months (95%CI: ''''''''''', ''''''''''') and ''''''''''' months (95% CI: ''''''''''', ''''''''''), respectively.[44] The proportion of complete responders with an ongoing CR at the time of data cut-off (18 March 2021) was '''''''''''' in the axi-cel group compared with ''''''''''' in the SOC group.[44] The estimated percentage of complete responders who remained in CR at 24 months was ''''''''''' (95% CI: '''''''' ''''''') in the axi-cel group compared with '''''''''''' (95% CI: '''''''' ''''''') in the SOC group (Appendix L).

A sensitivity analysis was planned in which patients in the axi-cel group who underwent stem cell transplant (SCT) while in an axi-cel-induced response were imputed to have a PFS event at the time of SCT. ''''''''''''''''''''' '''''' '''''''''''''''''' '''''''''''' '''''''' ''''''''''''''''''.[44]

Figure 11: Kaplan–Meier plot for DOR per central assessment, FAS

==> picture [453 x 99] intentionally omitted <==

==> picture [453 x 99] intentionally omitted <==

Key: CI, confidence interval; DOR, duration of response; FAS, full analysis set; HR, hazard ratio; NE, not estimable; SOCT, standard of care therapy. Notes: One-sided p-value from log rank test is presented. Source: Figure 15. ZUMA-7 CSR.[44]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 48 of 162

B.2.6.3.5. DOR per investigator assessment

DOR results per investigator assessment were consistent with those per central assessment. A summary of DOR per investigator assessment is provided in Appendix L.

B.2.6.3.6. mEFS per central assessment

At the time of data cut-off (18 March 2021), ''''''''' mEFS events by central assessment occurred for '''''''''' patients ('''''''%) in the axi-cel group and '''''''''' patients (''''''%) in the SOC group.[44] Axi-cel treatment was superior to SOC, with a stratified HR of '''''''''' (95% CI: ''''''''''', ''''''''''''; p '''' '''''''''''''''''; see Figure 12).

The median mEFS was significantly longer in the axi-cel group (''''''''''' months; 95% CI: '''''''', ''''''''''') than the SOC group (''''''''' months; 95% CI: '''''''', '''''''').[44] The estimated EFS at 24 months was '''''''% (95% CI: '''''', ''''''') in the axi-cel group compared with '''''% (95% CI: ''''''', ''''''') in the SOC group (Appendix L). The median follow-up time for EFS using the reverse Kaplan–Meier method was ''''''''''' months in the axi-cel group and ''''''''''' months in the SOC group.

The most common EFS events in either the axi-cel or SOC group were disease progression (''''''% and ''''''%, respectively), new lymphoma therapy (''''% and '''''''%, respectively) and death from any cause (''''% and ''''%, respectively).[44]

Findings of the sensitivity analyses were supportive of and consistent with results for mEFS per central assessment (see Appendix L).

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 49 of 162

Figure 12: Kaplan–Meier plot for mEFS per central assessment, FAS

==> picture [453 x 101] intentionally omitted <==

==> picture [453 x 102] intentionally omitted <==

Key: CI, confidence interval; FAS, full analysis set; HR, hazard ratio; mEFS, modified event-free survival; NE, not estimable; SOCT, standard of care therapy. Source: Figure 18. ZUMA-7 CSR.[44]

B.2.6.3.7. mEFS per investigator assessment

mEFS results per investigator assessment were consistent with those per central assessment. A summary of mEFS per investigator assessment is provided in Appendix L.

B.2.6.4. Exploratory endpoint

B.2.6.4.1. Time to next therapy

Time to next therapy (TTNT) events occurred for ''''''' patients (''''''%) in the axi-cel group and '''''''''' patients (''''''%) in the SOC group.[44] Axi-cel treatment was superior to SOC, with a stratified HR of ''''''''''' (95% CI: ''''''''''', ''''''''''; p ''' ''''''''''''''''''; see Figure 13).

The median TTNT was significantly longer in the axi-cel group (''''''''''' months; 95% CI: '''''''', '''''''') than the SOC group (''''''''' months; 95% CI: '''''''', ''''''''').[44] At the time of data cut-off (18 March 2021), ''''''''''' of patients in the axi-cel group compared with ''''''''''' of the SOC group were alive and had not received subsequent therapy. The estimated number of patients who were event-free at 24 months was '''''''% (95% CI: '''''', '''''') in the axi-cel group compared with ''''''% (95% CI: ''''''', ''''''') in the SOC group

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 50 of 162

(Appendix L). The median follow-up time for EFS using the reverse Kaplan–Meier method was '''''''''''' months in the axi-cel group and '''''''''''' months in the SOC group.

Figure 13: Kaplan–Meier plot of TTNT, FAS

==> picture [453 x 97] intentionally omitted <==

==> picture [453 x 96] intentionally omitted <==

Key: CI, confidence interval; FAS, full analysis set; HR, hazard ratio; NE, not estimable; SOCT, standard of care therapy; TTNT, time to next treatment. Source: Figure 21. ZUMA-7 CSR.[44]

B.2.6.5. HRQL

B.2.6.5.1. EORTC QLQ-C30

At screening, the mean EORTC QLQ-C30 Global Health Status scores for evaluable patients in the quality of life (QoL) analysis set were comparable in the axi-cel (mean: '''''''''') and SOC group (mean: ''''''''''; Figure 14).[44] At Study Day 50, almost half of evaluable patients reported worsening scores in both the axi-cel (mean: '''''''''') and SOC groups (mean: ''''''''''). Scores in the axi-cel group rebounded at Study Day 100 (mean: '''''''''''), while those in the SOC group declined (mean: ''''''''''). At this point there was a statistically significant and clinically meaningful difference in the mean change of scores in favour of axi-cel (estimated difference: ''''''''''; 95% CI: '''''''''', '''''''''''; adjusted p ''' '''''''''''''''; see Appendix L). This difference was also statistically significant at Study Day 150 (estimated difference: '''''''''; 95% CI: ''''''', ''''''''''''; adjusted p = ''''''''''''''''''). Mean estimated scores for the axi-cel group had returned to or exceeded scores at screening by Study Day 100 versus at Month 9 for the SOC group.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 51 of 162

Figure 14: Mean (95% CI) EORTC QLQ-C30 Global Health Status scores over time by treatment group, QoL analysis set

==> picture [453 x 108] intentionally omitted <==

==> picture [453 x 107] intentionally omitted <==

Key: CI, confidence interval; EORTC QLQ-C30, European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Cancer-30; QoL, quality of life. Source: Figure 22. ZUMA-7 CSR.[44]

At screening, the mean EORTC QLQ-C30 physical functioning scores for evaluable patients in the QoL analysis set were comparable in the axi-cel (mean: '''''''''') and SOC groups (mean: ''''''''''''; Figure 15).[44] At Study Day 50, the majority of evaluable patients reported worsening scores in both treatment arms. Starting at Study Day 100, scores for both treatment groups rebounded. There was a statistically significant and clinically meaningful difference in the mean change of scores from screening to Study Day 100 in favour of axi-cel (estimated difference: ''''''''''''; 95% CI: '''''''', ''''''''''''; adjusted p '''' '''''''''''''''; Appendix L). Mean estimated scores for the axi-cel group had returned to or exceeded scores at screening by Study Day 150 versus at Month 12 for the SOC group.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 52 of 162

Figure 15: Mean (95% CI) EORTC QLQ-C30 Physical Functioning scores over time by treatment group, QoL analysis set

==> picture [453 x 115] intentionally omitted <==

==> picture [453 x 115] intentionally omitted <==

Key: CI, confidence interval; EORTC QLQ-C30, European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Cancer-30; QoL, quality of life. Source: Figure 24. ZUMA-7 CSR.[44]

Statistically significant differences in several EORTC QLQ-C30 measures were found between patients treated with axi-cel and those treated with SOC. Treatment with axi-cel resulted in more favourable outcomes in terms of: nausea and vomiting, diarrhoea, insomnia, and appetite loss measures at Day 100; role functioning at Day 100 and Day 150; and social functioning, fatigue, and dyspnoea measures at Day 100, Day 150, and Month 9 (see Appendix L).[44]

B.2.6.5.2. EQ-5D-5L

The mean visual analogue scale (VAS) score reported by evaluable patients in the axi-cel and SOC groups were comparable at screening ('''''''''' and ''''''''''', respectively; see Figure 16).[44] There was a statistically significant and clinically meaningful difference in the mean change of scores for the EQ-5D-5L VAS from screening in favour of axi-cel at Study Day 100 (estimated difference: ''''''''''''; 95% CI: '''''''', ''''''''''; adjusted p ''' '''''''''''''''''') and Study Day 150 (estimated difference: ''''''''''; 95% CI: ''''''', ''''''''''; adjusted p = '''''''''''''''''; see Appendix L).

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 53 of 162

Figure 16: Mean (95% CI) EQ-5D-5L VAS scores over time by treatment group, QoL analysis set

==> picture [453 x 117] intentionally omitted <==

==> picture [453 x 116] intentionally omitted <==

Key: CI, confidence interval; QoL, quality of life; VAS, visual analogue scale. Source: Figure 26. ZUMA-7 CSR.[44]

The mean EQ-5D-5L index score was '''''''''''''' at screening for patients who received axi-cel and '''''''''''' for patients who received SOC (Figure 17).[44] At Study Day 50, many evaluable patients in both the axi-cel and SOC groups reported worsening scores (''''''''''% and ''''''%, respectively). The estimated mean difference in scores changing from screening was statistically significant and clinically meaningful at Day 100 in favour of axi-cel ('''''''''''''; 95% CI: '''''''''''', ''''''''''''; adjusted p = ''''''''''''''''; see Appendix L).

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 54 of 162

Figure 17: Mean (95% CI) EQ-5D-5L index scores over time by treatment group, QoL analysis set

==> picture [453 x 117] intentionally omitted <==

==> picture [453 x 118] intentionally omitted <==

Key: CI, confidence interval, QoL, quality of life. Source: Figure 29. ZUMA-7 CSR.[44]

B.2.7. Subgroup analysis

The subgroup analysis demonstrated consistent survival benefits with axi-cel over SOC. EFS, ORR, OS, PFS and mEFS outcomes were generally comparable with those observed in the overall population.

Results of the covariate analysis of EFS consistently showed axi-cel superiority over

SOC in most subgroups, including patients with high-risk features such as HGBL (including double- or triple-hit lymphomas), relapsed or primary refractory disease and being ≥ 65 years of age.

A summary of results for the analysed subgroups is provided in Appendix E.

B.2.8. Meta-analysis

The main evidence for the use of axi-cel in the second-line treatment of DLBCL is from ZUMA-7. Therefore, no meta-analysis is required.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 55 of 162

B.2.9. Indirect and mixed treatment comparisons

ZUMA-7 provides head-to-head data for the relevant comparator to the decision problem being addressed. Therefore, no indirect or mixed treatment comparisons have been performed.

B.2.10. Adverse reactions

B.2.10.1. Safety summary

Table 9 presents an overview of the safety data up to the data cut-off date (18 March 2021).

All patients experienced at least one AE of any grade. AEs of Grade 3 or higher occurred in 155 patients (91%) who received axi-cel and 140 patients (83%) who received SOC. Serious AEs of any grade occurred in 85 patients (50%) who received axi-cel and in 77 patients (46%) who received SOC.[1]

Seven patients (4%) died due to AEs in the axi-cel group, only one of which was considered by the investigators to be related to axi-cel (the treatment caused the hepatitis B virus to be reactivated).[1] Of the two patients (1%) in the SOC group who died because of AEs, both deaths were considered by the investigators to be related to high-dose chemotherapy (cardiac arrest and acute respiratory disease).

Table 9: Overall summary of treatment-emergent adverse events, SAS

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 56 of 162

B.2.10.2. Common AEs

Table 10 presents the most common treatment-emergent adverse events (TEAEs) occurring in ≥ 10% of patients in either treatment group.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 57 of 162

The most common TEAEs of any grade in the axi-cel group were: pyrexia (158 patients; 93%); neutropenia and hypotension ('''''' patients each; ''''''%); anaemia, fatigue and diarrhoea (71 patients each; 42%); headache (70 patients; 41%); nausea (69 patients; 41%), sinus tachycardia (58 patients; 34%); and a decreased neutrophil count ('''''' patients; ''''''%).[1, 44] The most common TEAEs of any grade in the SOC group were: nausea (116 patients; 69%); anaemia (91 patients; 54%); fatigue (87 patients; 52%); diarrhoea (66 patients; 39%); a decreased platelet count ('''''' patients; ''''''%); constipation (58 patients; 35%); and vomiting (55 patients; 33%).

The most common Grade 3 or higher TEAEs in the axi-cel group were: neutropenia ('''''' patients; '''''''%), anaemia (51 patients; 30%) and a decreased neutrophil count ('''''' patients; '''''%).[1, 44] The most common Grade 3 or higher TEAEs in the SOC group were anaemia (65 patients; 39%), a decreased platelet count ('''''' patients; ''''''%) and a decreased neutrophil count (''''' patients; ''''''%). The most common nonhaematological worst Grade 3 or higher TEAEs were: hypophosphatemia (31 patients; 18%), encephalopathy (20 patients; 12%) and hypotension (19 patients, 11%) for the axi-cel group; and hypophosphatemia (21 patients, 13%) for the SOC group.

Table 10: Incidence of TEAEs occurring in ≥ 10% of patients in either treatment

group, SAS

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 58 of 162

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 59 of 162

B.2.10.3. Treatment-related AEs

Table 11 presents the most common treatment-related TEAEs occurring in ≥ 10% of patients in either treatment group.

In the axi-cel and SOC groups, '''''''' patients (''''''%) and ''''''''' patients (''''''%), respectively, had treatment-related TEAEs; ''''''''' patients ('''''''%) and ''''''''' patients ('''''%), respectively, experience Grade 3 or higher treatment-related AEs.[44] The most common worst Grade 3 or higher treatment-related AEs in the axi-cel group were: pyrexia (''''''''' patients; ''''''%); hypotension ('''''' patients; ''''''%); and headache and sinus tachycardia ('''''' patients each; '''''''%). The most common worst Grade 3 or higher treatment-related AEs in the SOC group were: nausea (''''''''' patients; '''''''%); anaemia ('''''' patients; '''''''%); and fatigue ('''''' patients; ''''''%).

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 60 of 162

Table 11: Incidence of treatment-related TEAEs occurring in ≥ 10% of patients in either treatment arm, SAS

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 61 of 162

B.2.10.4. AEs of special interest

B.2.10.4.1. Neurological events

Table 12 presents neurological events following treatment with axi-cel and SOC.

Neurological events occurred in 102 patients (60%) who received axi-cel and in 33 patients (20%) who received SOC. Neurological events of Grade 3 or higher occurred in 36 patients (21%) and one patient (1%), respectively.[1] No deaths related to neurological events occurred.

The most common worst Grade 3 or higher treatment-emergent neurological events in the axi-cel group were encephalopathy (20 patients; 12%), aphasia (12 patients; 7%) and confusional state (nine patients; 5%).[1] One patient (1%) in the SOC group had a Grade 3 or higher treatment-emergent neurological event of delirium. The most common serious treatment-emergent neurological events of any grade in the axi-cel group were encephalopathy (''''''' patients; ''''''%), aphasia (''''''''''' patients; '''%) and confusional state ('''''' patients; '''%), and the only serious neurological event in the SOC group was encephalopathy.[44]

The median time to the onset of neurological events was 7 days (range: ''''''''''''''') in the axi-cel group and 23 days (range: ''''''''''''''') in the SOC group, and the median duration was 9 days (range: 1–817) and 23 days (range: ''''''''''''''''), respectively.[1, 44] At

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 62 of 162

the data cut-off date, two patients had ongoing neurological events; one patient who received axi-cel had Grade 2 paraesthesia and Grade 1 memory impairment, and one patient who received SOC had Grade 1 paraesthesia.[1]

Table 12: Summary of treatment-emergent neurological events occurring in ≥

5% of patients in either treatment group, SAS

B.2.10.4.2. Cytokine release syndrome

Cytokine release syndrome (CRS) is an AE induced by the activated T-cells upon engagement with the CD19 target, so it is considered to be related to treatment with CAR T-cell therapy. In ZUMA-7, the severity of CRS was graded according to a modification of the grading system proposed by Lee et al.[46]

Table 13 presents CRS events and the most common symptoms of CRS (occurring in ≥ 5% of patients) following treatment with axi-cel. CRS occurred in 157 patients (92%) who received axi-cel, of whom 11 (6%) had worst Grade 3 or higher CRS. No

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 63 of 162

deaths related to CRS occurred.[1] The most common symptoms of CRS worst Grade 3 or higher were hypotension (18 patients; 11%), pyrexia (14 patients; 9%) and hypoxia (13 patients; 8%). The most common serious CRS symptoms by any grade were pyrexia ('''''' patients; ''''''%), hypotension ('''''' patients; ''''%) and hypoxia ('''''''''''' patients; ''''%).[44]

The median time to the onset of CRS was 3 days (range: 1–10) after the infusion, and the median duration was 7 days (range: 2–43). At the data cut-off date, all the CRS events were resolved.[1]

Table 13: Summary of treatment-emergent CRS and CRS symptoms occurring in ≥ 5% of patients in the axi-cel group, SAS

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 64 of 162

B.2.10.4.3. Cytopenia events

Table 14 presents cytopenia events following treatment with axi-cel and SOC.

The most common Grade 3 or higher cytopenia events in the axi-cel group were thrombocytopenia (''''''' patients; ''''''%), neutropenia ('''''''''' patients; ''''''%) and anaemia ('''''' patients; '''''''%).[44] The most common Grade 3 or higher cytopenia events in the SOC group were thrombocytopenia ('''''' patients; '''''''%), neutropenia ('''''' patients; ''''''%) and anaemia ('''''' patients; ''''''%). No patients had Grade 5 CRS.

Prolonged cytopenia events of Grade 3 or higher that were present at or after 30 days after the initiation of definitive therapy (from receipt of the axi-cel infusion or first dose of high-dose chemotherapy) occurred in 49 patients (29%) who received axicel, '''''''' patients ('''''%) in the overall SOC group, and in 12 of 62 patients (19%) in the SOC group who underwent auto-SCT (see Appendix F).[1, 44]

Table 14: Summary of treatment-emergent cytopenia events in either treatment group, SAS

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 65 of 162

B.2.10.4.4. Infections

Infections were experienced by 70 patients (41%) in the axi-cel group and 51 patients (30%) in the SOC group, of which 24 patients (14%) and 19 patients (11%) had worst Grade 3 or higher infections, respectively.[1] ''''''''''''''' patients (''''%) in the axi-cel group and ''''''' patients ('''%) in the SOC group had worst Grade 4 infections.[44] ''''''''''' patients ('''%) in the axi-cel group had a Grade 5 infection ('''''''' patients with COVID-19, ''''''''' '''''''''''''''' with progressive multifocal leukoencephalopathy, ''''''''' '''''''''''''''' with hepatitis B reactivation and ''''''''' '''''''''''''' with sepsis). ''''''' ''''''''''''''''' in the SOC group experienced a Grade 5 infection.

The most common infections in the axi-cel group were: unspecified ('''''' patients; ''''''%); viral infections ('''''' patients; ''''''%); bacterial infections ('''''' patients; '''%); upper respiratory tract infections ('''''' patients, ''''%); and opportunistic infections ('''''''''''' patients; '''%).[44] The most common infections in the SOC group were: unspecified ('''''' patients; ''''''%); bacterial infections ('''''' patients; ''''%); and viral infections ('''''''''''' patients; ''''%).

The most common worst Grade 3 or higher treatment-emergent infections by preferred term (excluding COVID-19) were pneumonia ('''''' patients; '''%) and upper respiratory tract infection (''''''''''''' patients; '''%) in the axi-cel group, and pneumonia and sepsis ('''''''''' '''''''''''''''''''' ''''''''''''; '''%) in the SOC group.[44] COVID-19 infections were reported as TEAEs for ''''''''''' patients (''''%) in the axi-cel group, all of whom had worst

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 66 of 162

Grade 3 or higher, and '''''''' ''''''''''''''' in the SOC group had a Grade 1 COVID-19 infection.

B.2.10.4.5. Hypogammaglobulinaemia

Hypogammaglobulinaemia includes preferred terms of hypogammaglobulinaemia and decreased blood immunoglobulin G. The severity of an event was graded by the investigator.[44]

Hypogammaglobulinaemia during treatment occurred in 19 patients (11%) who received axi-cel and in one patient (1%) who received SOC; all the events were Grade 1 or 2.[1]

A summary of hypogammaglobulinaemia TEAEs are presented in Appendix F.

B.2.10.5. Concomitant medications

Among patients who received axi-cel, '''''' patients (''''''%) received corticosteroids (with or without tocilizumab), ''''''''' patients (''''''%) were treated with tocilizumab (with or without corticosteroids) and ''''''' patients (''''''%) were treated with corticosteroids and tocilizumab.[44] ''''''''''''''''''''''' patients (''''''%) were treated with vasopressors and '''''' patients (''''''%) were treated with immunoglobulins.

B.2.10.6. Safety overview

The safety profile observed in ZUMA-7 was manageable and generally consistent with the safety profile of axi-cel treatment as a third-line therapy for patients with r/r DLBCL (ZUMA-1) and real-world use of axi-cel, as summarised in Table 15. Since the approved access of axi-cel and tisa-cel through the CDF in NHS England, clinicians are increasingly comfortable with toxicity management for this CD19directed CAR T-cell therapy class.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 67 of 162

Table 15: TEAEs observed in patients receiving CAR T-cell therapy

As recommended in the SmPC for axi-cel (Appendix C), patients should be monitored for the first 10 days following infusion for signs and symptoms of potential CRS, neurological events and other toxicities. After the first 10 days, the patient can be monitored at the physician’s discretion, but patients should remain within proximity of a qualified clinical facility for at least 4 weeks following infusion. At least one dose of tocilizumab in the event of CRS and emergency equipment must be available prior to axi-cel infusion, and the treatment centre must have access to an additional dose of tocilizumab within 8 hours of each previous dose. In the exceptional case where tocilizumab is not available due to a shortage that is listed in the MHRA (Medicines & Healthcare products Regulatory Agency) Central Alerting System, suitable alternative measures to treat CRS instead of tocilizumab must be available before infusion takes place.

Blood counts should be monitored after axi-cel infusion and patients should also be monitored for signs and symptoms of infection before, during and after axi-cel

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 68 of 162

infusion (and treated appropriately). Prophylactic antimicrobials should be administered according to standard institutional guidelines. Immunoglobulin levels should also be monitored after treatment with axi-cel and managed using infection precautions, antibiotic prophylaxis and immunoglobulin replacement.

A recent review summarised the latest data on potential late or prolonged effects of CAR T-cell therapy, including prolonged cytopenia events, hypogammaglobulinaemia, late neurological effects, late immune-related adverse events, second cancers, late infections and cardiac toxicities.[50] Very few late or prolonged effects presented after one year following treatment, which suggests that the available safety data from ZUMA-7 (providing 24.9 months of potential follow-up) will have captured any such event. Conversely to what has been reported for stem cell transplants, this review also reported that HRQL in long-term survivors is comparable to that of the general population, although this conclusion is based on limited data.

B.2.11. Ongoing studies

The ZUMA-7 study is ongoing with follow-up through to 60 months; the potential follow-up represented in the submission is 25 months. No other studies are investigating axi-cel in adults with primary refractory or early relapse DLBCL who are intended for transplant.

A Phase II study investigating axi-cel as a second-line therapy in patients with r/r B- cell NHL who are non-candidates for transplant is currently recruiting (ALYCANTE; NCT04531046), with primary data estimated to be available in May 2022.[51] This study is sponsored by the Lymphoma Academic Research Organisation and therefore Gilead has no early sight of data or access to patient-level data from this study.

B.2.12. Innovation

Axi-cel is a personalised, transformative, single-infusion medicine in which the patient’s own T-cells are engineered to target and kill cancer cells. Axi-cel was the first of the breakthrough class of CAR T-cell therapies to receive European Medicines Agency (EMA) and US Food and Drug Administration (FDA) approval,

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 69 of 162

and its innovative nature is well established and accepted across the healthcare community.

Axi-cel transformed the DLBCL management pathway[52] and now offers a second step change through earlier use in primary refractory or early relapse DLBCL patients intended for transplant. Providing access to axi-cel at first relapse would increase the number of patients receiving definitive therapy in the second-line setting, and thus improve cure rates specific to this patient population with high unmet need and a poor prognosis. Data from ZUMA-7 clearly demonstrate this with a three-fold increase in the number of patients receiving definitive therapy and a 2.5fold increase in the number of patients living event-free for at least 2 years.[1] While data that would robustly allow quantification of the improved cure rate with secondline CAR T-cell therapy use versus third- or later-line CAR T-cell therapy use are not available to date, clinical experts felt that an approximate 10% improvement as observed in unadjusted OS analyses of ZUMA-7 (where several patients in the SOC arm went on to receive CAR T-cell therapy in the third- or later-line setting) was not ‘unreasonable’.[29] Support for advancing axi-cel positioning such that patients have access to CAR T-cell therapy when they are likely to have lower tumour burden and comorbidities, and higher T-cell and general fitness, is strong across the clinical community.[29, 30, 42, 43]

While the main health-related benefits of axi-cel will be captured in the qualityadjusted life years (QALY) calculation, the true extent of the benefit associated with cure is likely to have been underestimated, including the emotional benefit of hope that receiving a potentially curative treatment option can provide.[53] Additional benefits associated with a single-infusion medicine compared with multiple cycles of immunochemotherapy followed by HDT and auto-SCT include reduced impact on the daily lives of patients and their carers, and capacity benefits to health services. These benefits may not be captured in the QALY calculation.

Anecdotal reports of the emotional consequences of auto-SCT are not captured in clinical trial safety outcomes, which include symptoms aligned to post-traumatic stress disorder.[54] Emotional consequences of treatment are not formally captured in the QALY calculation, but with no similar reports of these consequences of CAR T-

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 70 of 162

cell therapy, this could represent a significant health-related benefit of axi-cel to patients, carers and healthcare services alike.

B.2.13. Interpretation of clinical effectiveness and safety evidence

B.2.13.1. Principal findings from the clinical evidence

The ZUMA-7 trial supports previous trial observations (Table 3) that only a minority of primary refractory or early relapse DLBCL patients intended for transplant go on to receive definitive therapy with current second-line SOC, and that the associated overall cure rate in this population remains low. It also shows the potential benefit of axi-cel in improving definitive therapy and cure rates, with three times as many patients randomised to axi-cel receiving definitive therapy, and 2-3 times as many patients living event-free for at least 2 years versus SOC.[1]

ZUMA-7 was designed to reflect the future decision-making process at first relapse in clinical practice, with patients intended for transplant enrolled and randomised to follow the current second line SOC pathway or the potential CAR-T service pathway. Despite intent, only 36% of patients randomised to SOC in ZUMA-7 actually went onto receive definitive treatment with HDT plus auto-SCT following re-induction therapy .[1] In contrast, axi-cel was successfully manufactured for 100% of patients who underwent apheresis and 94% of patients randomised to axi-cel received definitive treatment. The overall 2-year EFS rates were 16% in the SOC group versus 41% in the axi-cel group.

The percentage of patients with a response to axi-cel was significantly greater than in the SOC group (83% versus 50%; p < 0.001) and a CR was observed in twice as many patients (65% versus 32%).[1] At the time of analysis (median follow-up of 24.9 months), '''''''''' of responding patients had an ongoing response to axi-cel treatment compared with '''''''''' of patients who had an ongoing response to SOC.[44] In the interim survival analysis, the estimated 2-year OS rate was 61% in the axi-cel group versus 52% in the SOC group, notably, with subsequent cellular immunotherapy received off-protocol in 56% of patients); 29% of patients in the axi-cel group died from progressive disease compared with 36% of patients in the SOC group.[1] In a world without CAR T-cell therapy available in the third- or later-line setting, we would

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 71 of 162

expect survival to be significantly lower than observed in the SOC arm of ZUMA-7; this is further discussed in Section B.2.13.3.

B.2.13.2. Strengths and limitations of the evidence base

ZUMA-7 is the first and largest RCT investigating the efficacy and safety of CAR T- cell therapy versus current second line SOC for primary refractory or early relapse DLBCL patients intended for transplant. ZUMA-7 provides robust data that is directly relevant to the decision problem being addressed, that is, should patients eligible for intensive therapy at first relapse be treated with re-induction therapy for potential HDTauto-SCT or axi-cel? Data clearly demonstrate that the current pathway of care is lacking and that axi-cel offers a much higher chance of receiving definitive therapy at this crucial stage than current SOC and thus an associated higher chance of cure (see above).We should be aiming to treat patients with the most effective therapies at the earliest stage in their treatment pathway, and there is strong support for advancing axi-cel positioning across the clinical community.[29, 30, 42, 43]

The primary analyses of ZUMA-7 provide more than two years of potential follow-up. Although these data are still relatively immature given the curable disease setting, two years is considered a significant and clinically meaningful milestone in r/r DLBCL.[43] Clinical experts estimate that 95% of patients living event-free at 2 years will achieve long-term remission, and that most patients who would relapse after CAR T-cell therapy or auto-SCT would have done so by this 2-year timepoint.[29] This has been formally explored in the de novo DLBCL setting. A prospective study demonstrates that patients with DLBCL who were treated with immunochemotherapy and who were living event-free at 2 years had an equivalent OS to that of the ageand sex-matched general population.[55] Applying the estimated 95% long-term remission rates to the 2 year EFS rates observed in the ZUMA-1 trial suggests that 38% of patients treated with axi-cel in the second-line setting will be cured, compared with 15% of patients treated with current second-line care.

As noted in Section B.2.12, robust data quantifying the improved cure rate with second-line versus third- or later-line use are not available to date. However, unadjusted OS analyses of ZUMA-1 provide some insight into this potential benefit, taking into consideration that over half of patients who were randomised to the SOC arm went on to receive subsequent cellular immunotherapy. In the interim analyses,

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 72 of 162

we observe an approximate 10% improvement in 2-year OS (61% vs 52%), but the immaturity of the data prevents us from drawing longer-term conclusions.[1] With the requisite caveat around heterogeneity of patient populations and naïve comparisons, a similar difference is observed between 2-year OS rates of the axi-cel arms of ZUMA-7 and ZUMA-1 (61% vs 50%).[56]

In recognition of the current uncertainty around the longer-term benefit with axi-cel treatment in the second-line setting, we acknowledge that axi-cel is likely to be a CDF candidate. With interim funding through the CDF, earlier access to axi-cel would be available for a patient population with a high unmet need and a poor prognosis. This would happen alongside ongoing data collection, which will robustly assess the cost-effectiveness of axi-cel in this second-line treatment setting.

B.2.13.3. Applicability of clinical evidence to practice

The primary endpoint of ZUMA-7, EFS, is an established endpoint that classes a best ‘response’ of SD and new therapy commencement prior to radiographic disease progression as an event, alongside radiographic disease progression and death. This is the most clinically relevant endpoint in a curable disease setting where stable disease is not an acceptable outcome, and where patients with a suboptimal response to treatment will be moved onto a new therapy for potential cure at the earliest opportunity.[29, 43, 57] In comparison, PFS data collected are subject to informative censoring, as patients who receive a new therapy before disease progression (and who are then censored in PFS analyses) are not random and directly relate to patient prognosis, which may lead to bias.[58] Using EFS in appraisals of potentially curative treatment has previously been deemed appropriate for decision-making.[59]

EFS is also the most representative endpoint for cure in this setting. It is shown to be a valid surrogate endpoint for OS in haematological malignancy across several correlation analyses identified through systematic literature review (SLR).[60] In DLBCL specifically, the correlation between EFS and OS was found to be stronger than the correlation between PFS and OS in a large-scale surrogacy analysis that was based on 30 clinical trials and 47 retrospective studies.[61] Exploratory analyses of OS by EFS status in the ZUMA-1 trial (third- or later-line axi-cel treatment for r/r DLBCL) further support the usefulness of EFS rates as surrogate endpoints for long-

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 73 of 162

term OS specific to r/r DLBCL.[56] Significant associations are observed between EFS and HRQL, and between EFS and healthcare resource use.[62, 63]

The target population for reimbursement is aligned to the evidence base supporting the use of axi-cel in the second-line setting. Generally, the ZUMA-7 trial population is also directly applicable to the proposed use of axi-cel in clinical practice, which is for the treatment of primary refractory or early relapse DLBCL patients intended for transplant: a patient population with high unmet need and a poor prognosis for whom clinical experts believe CAR T-cell therapy could play an important role.[29, 30, 42, 43] However, as is often the case in a clinical trial versus real-world setting, the ZUMA-7 trial population is a select group of primary refractory or early relapse DLBCL patients who would be expected to tolerate and respond well to intensive therapy (i.e. a generally younger, fitter population than the real-world population) despite their generally poor prognosis (74% of patients had primary refractory disease, 45% had a high sAAIPI and 79% had Stage III/IV disease).[29] Trial outcomes may therefore be slightly optimistic compared to outcomes that might be expected in practice. However, this would apply to both arms in equal measure, i.e. to patients randomised to axi-cel or SOC, and therefore comparative effect estimates remain applicable to real-world expectations. There is also close alignment in 2-year EFS estimates from the ZUMA-7 SOC arm (16%) to historical trial 2-year EFS estimates (16-17%) in similar patient groups.[1, 25, 27]

The use of steroid-only bridging therapy as mandated in the ZUMA-7 study may also have been a factor in the selection of patients who entered the study. Bridging therapy was restricted to glucocorticoids to isolate the effects of CAR T-cell therapy as second-line therapy in ZUMA-7 but this may have restricted enrolment of patients with rapidly progressing disease that would otherwise have warranted more aggressive bridging therapy. In clinical practice, bridging therapy with outpatient chemotherapy is expected to be used in approximately two-thirds of patients.[29] While the use of bridging therapy should not have a direct impact on effect estimates and is intended to keep patients stable while axi-cel is being manufactured, it does impact cost estimates and is thus further explored in the economic analysis. As noted above, any impact on effect estimates relating to patient selection would be applied to both arms in equal measure.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 74 of 162

The SOC immunochemotherapy regimens used in ZUMA-7 are reflective of those that are most used at first relapse in clinical practice. Although some differences are observed in the proportion of patients receiving each regimen in the trial versus realworld treatment patterns (see Appendix M), clinical experts confirmed that they would expect equivalence in effect across regimens, as demonstrated in historical clinical trials.[29] Gisselbrecht, 2010 #41;Crump, 2014 #81} NICE makes a specific recommendation to consider R-GDP based on equivalent effectiveness and reduced toxicity[64] . The proportion of patients receiving an R-GDP regimen was very similar across the ZUMA-7 SOC group (25%) and real-world data (23%), however clinical experts advised that there is an increasing use of R-GDP in practice as it is possible to administer this regimen in an outpatient setting.[1, 17, 29] Importantly, '''''''' '''''''' ''''''''' ''''''' ''''''''''' ''''''''''''''''' ''''''''''''''''''''' '''''''''''''''' ''''''' '''''''''''''''''''' ''''' ''''''' '''''''''''''''''''''''' ''''''''''''''' '''''''''' '''''' ''''''''''''''''''' ''''''''''''''''''' ''''''''''''''''' ''''''''' ''''''' ''''''''' '''''''''''''''' ''''' ''''''''''''''''''''''''''''' '''''''''''''''''''''''''' ''''''''' ''''''''''''''''''' '''''''''''' '''''' ''''''''''' ''''''''''''''''''''' '''''''''''''''''' '''''''''''''''''''''''''''' '''''''''''''''''''''''''''''' ''''' '''''''' ''''''''''''''''' ''''''''''''''''''' '''' '''''''' '''''''''''''''''''''''''''

The high proportion of subsequent cellular immunotherapy use in the SOC group of the ZUMA-7 trial (56% off-protocol treatment switching) should be considered when interpreting the (unadjusted) interim OS analyses. Although CAR T-cell therapy is available in the third- or later-line setting to patients in England, it is currently funded through the CDF, and such treatments should not be considered in a treatment sequence for new technology appraisals according to the NICE position statement on this topic.[65] There is also arguably a narrower window of opportunity to select patients for CAR T-cell therapy in the real-world setting vs a clinical trial environment such that the proportion of patients receiving subsequent cellular immunotherapy in the ZUMA-7 SOC arm is higher than we would expect in practice.[30] Aligning to NICE recommendations, post-hoc sensitivity analyses that adjust OS estimates for subsequent CAR T-cell therapy use in the SOC arm of ZUMA-7 are used in the costeffectiveness base case (see Section B.3.3.4.1).

In a world without CAR T-cell therapy available in the third- or later-line setting, we would expect OS to be significantly lower than observed in the (unadjusted) interim OS analyses in the SOC arm of ZUMA-7. Indeed, OS estimates in the SOC arm were higher than observed in historical trials conducted before CAR T-cell therapies

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 75 of 162

were introduced at third- or later-lines. For example, the 2-year OS rate in patients with primary refractory or early relapse DLBCL in the ORCHARRD study was 31%, and the 2-year OS rate in patients with primary refractory DLBCL in the SCHOLAR-1 study was 24% (Table 3).[24, 27] Clinical experts agreed that patient survival in a world without CAR T-cell therapy would be significantly lower, and that patients who

relapse after current second-line care would follow a steep downward trajectory, with EFS and OS curves estimated to align by 5 years at the latest, perhaps even as early as by 1 year in the primary refractory or early relapse poor risk patient cohort.[29, ] 30

B.2.13.4. Service implications

Axi-cel is already reimbursed for the treatment of adult patients with r/r DLBCL and primary mediastinal large B-cell lymphoma (PMBCL) after two or more lines of systemic therapy. NHS England service provision for CAR T-cell therapies are well established, and there is no or very minimal impact on further site qualification, patient referral or management expected with the advancement of axi-cel in the clinical care pathway. Approval of axi-cel for earlier use would have an impact on patient numbers, but plans are already in place to increase the number of CAR T-cell therapy centres throughout 2022.

B.2.13.5. Axi-cel as an end-of-life therapy

Primary refractory or early relapse DLBCL patients intended for transplant have a poor prognosis with current second-line care, and in the absence of CAR T-cell therapy, these patients are not expected to survive beyond 2 years. Data supporting axi-cel as an end-of-life therapy in this population are summarised in Table 16.

Clinical experts agree with these data and state that in a world without CAR T-cell therapy, patients who relapse after current second-line care would follow a steep downward trajectory, with EFS and OS curves estimated to align by 5 years at the latest, perhaps even as early as by 1 year in the primary refractory or early relapse poor risk patient cohort.[29, 30]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 76 of 162

Table 16: End-of-life criteria

==> picture [29 x 13] intentionally omitted <==

Cost effectiveness

B.3.1. Published cost-effectiveness studies

A systematic search was conducted to identify existing published cost-effectiveness studies in adults ≥ 18 years of age with r/r DLBCL after first-line therapy only. Full details of the search methods and results are presented in Appendix G. The search

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 77 of 162

identified five modelling studies conducted across various geographies.[66-70] Only one study was conducted in the UK and thus can be considered relevant to decisionmaking in England.[70] This study is summarised in Table 17.

Table 17: Summary list of published cost-effectiveness studies

Furthermore, a search of the NICE website identified four previous single technology appraisals for adults with r/r DLBCL. These are:

• TA649: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma[2]

• TA306: Pixantrone monotherapy for treating multiply relapsed or refractory ‐

• aggressive non-Hodgkin's B cell lymphoma[71]

• TA559: Axicabtagene ciloleucel for treating diffuse large B-cell lymphoma and primary mediastinal large B-cell lymphoma after 2 or more systemic therapies[52]

• TA567: Tisagenlecleucel for treating relapsed or refractory diffuse large B-cell lymphoma after 2 or more systemic therapies[72]

Modelling approaches for the two CAR T-cell therapy appraisals (TA559 and TA567)

are summarised in Table 18. Although these consider a later-line population than the current appraisal, TA559 and TA567 were used as a basis for the modelling approach, inputs and assumptions.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 78 of 162

Table 18: Summary of previous TAs of CAR T-cell therapies in DLBCL

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 79 of 162

B.3.2. Economic analysis

B.3.2.1. Patient population

In line with the ZUMA-7 trial, the patient population considered in this analysis is adults with primary refractory or early relapse (≤ 12 months) DLBCL after 1 systemic therapy who are intended for transplant.

Population characteristics in the model are aligned with those of the ZUMA-7 trial population; the mean age at baseline is 57.2 and the proportion female is 34%.

Section B.2.3.1 provides further details on the baseline characteristics of ZUMA-7 participants. and Section B.2.13.3 discusses the applicability of ZUMA-7 evidence to clinical practice.

B.3.2.2. Model structure

A de novo cost-effectiveness model was developed in Microsoft Excel[®] . As health economic experts highlighted a preference for a simpler model structure, a partitioned survival approach with three health states was specified (Appendix R). The model structure is presented in Figure 18.

Figure 18: Model structure

==> picture [452 x 172] intentionally omitted <==

Key: EFS, event-free survival; OS, overall survival, TTNT, time to next treatment.

As shown in Figure 18, the partitioned survival model has three mutually exclusive health states defined by three stages of the disease:

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 80 of 162

• Event-free (split into ‘on treatment’ and ‘off treatment’ states)

• Post-event (split into ‘on next treatment’ and ‘off treatment’ states)

• Death

All outcomes are modelled independently of each other with transitions between health states derived directly from OS, EFS and time to next treatment (TTNT) projections. As shown in Figure 18, the proportion of patients who are dead in each model cycle is estimated by one minus estimated survival (1-OS), the proportion of those in the post-event state is estimated by the area between OS and EFS projections (OS-EFS), and the proportion in the event-free state is the area under the EFS curve. The TTNT curve is used to further partition the post-event health state into those receiving and not receiving subsequent treatment, thus is important in determining post-event treatment costs.

The choice to capture EFS in the model structure rather than PFS was driven by several factors. EFS is the primary endpoint of the ZUMA-7 trial (for which the trial is powered) and defined as the time until disease progression, initiation of a next line of therapy or death. As outlined in Table 1 and Section B.2.13.2, EFS is the most clinically relevant endpoint for DLBCL given the curative intent of treatment. In DLBCL it is common practice to move patients to the next line of therapy in this setting if their best response is stable disease, given the severe nature of the condition. Furthermore, the use of the alternative outcome, PFS, would be biased by informative censoring[58] as, for the assessment of PFS in ZUMA-7, patients who receive a new treatment are censored if this occurs before progression. As initiation of a new treatment is not random and is related to a patient’s prognosis, this results in an overestimation of PFS as the outcome is reflective of patients with a better prognosis. There is also precedent for the use of EFS as an outcome on which to base a partitioned survival model. The modelling approach for TA554 (tisagenlecleucel for treating relapsed or refractory B-cell acute lymphoblastic leukaemia in people aged up to 25 years),[73] used EFS as again this was the primary endpoint in the key trials. The structure of the company’s model was deemed appropriate for decision-making by the committee.[59]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 81 of 162

The partitioned survival model structure is both simple and flexible enough to extrapolate survival using various methods and can incorporate relative efficacy in numerous ways. The approach is considered to reflect the patients’ disease pathway in r/r DLBCL and allows for key trial endpoints to be modelled directly (EFS is the primary outcome in ZUMA-7).

Partitioned survival modelling is a widely used and accepted approach in oncology appraisals, particularly for end-stage cancer treatments. It is also consistent with the model structures used for previous CAR T-cell therapies in r/r DLBCL (Table 18). In both prior CAR T-cell therapy appraisals in DLBCL the committee accepted the model structure as appropriate for decision-making.[74, 75] Specifically, the models for TA559 and TA567 incorporated a decision tree to account for costs and outcomes for patients who undergo leukapheresis but do not go on to receive CAR T-cell therapy infusion. This was not required for the current submission as ZUMA-7 is a randomised (rather than single-arm) trial, patient outcomes were measured from randomisation and therefore capture the range of events that can occur before axicel infusion. On consultation, clinical experts agreed that the model structure appropriately reflected the disease pathway for r/r DLBCL patients.[29]

B.3.2.2.1. General model settings

The analysis perspective is that of the NHS and Personal Social Services (PSS) in England for costs and direct health effects on individual patients for outcomes, in line with the NICE reference case.[76]

The model uses a 1-month cycle length (30.44 days). A half-cycle correction is applied throughout the model to both costs and health outcomes; to better account for the fact that some (costs) can occur at any point during the cycle, while others (health outcomes) are spread across time.

The analysis assumes a lifetime time horizon (50 years), which is sufficient to capture the plausible maximum life expectancy for the ZUMA-7 ITT population (mean age 57.2 years). Shorter time horizons are explored in the scenario analysis in Section B.3.8.3.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 82 of 162

A discount rate of 3.5% per annum is applied to costs and QALYs, as also specified in the NICE reference case. All costs are presented in British pound sterling (GBP) and the cost year is 2021.[76]

General model settings are summarised in Table 19, with features of previous CAR T-cell therapies presented in Table 20 alongside features of the current appraisal. It is important to note that CAR T-cell therapies presented are in the third-line population and thus slightly less fit than patients considered in the current secondline appraisal.

Table 19: General model settings

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 83 of 162

Table 20: Features of the economic analysis

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 84 of 162

B.3.2.3. Intervention technology and comparators

B.3.2.3.1. Intervention

The intervention, axi-cel, is implemented in the model as per the expected marketing authorisation, anticipated '''''''''''''''''''''''''''' ''''''''''', and is reflective of the decision problem described in Section B.1.1.

Axi-cel is an autologous anti-CD19 CAR T-cell product, that recognises and eliminates all CD19 expressing target cells, including B-cell malignancies and normal B-cells. The mechanism of action and process for manufacturing and administering axi-cel is described in Section B.1.2.

Axi-cel is a single-infusion product, for autologous and intravenous use only. Each single-infusion bag contains a target dose of 2 x 10[6] CAR-positive viable T-cells per kg of body weight. Before infusion, all patients will receive lymphodepleting chemotherapy consisting of intravenous cyclophosphamide 500 mg/m[2] and intravenous fludarabine 30 mg/m[2] intravenous on the 5th, 4th and 3rd day before axi-cel infusion, and some patients are treated with bridging chemotherapy.

B.3.2.3.2. Comparator

As described in Section B.1.3.4, patients who are fit enough to tolerate intensive therapy should be offered further multi-agent immunochemotherapy (re-induction therapy) at relapse to try to obtain sufficient response for HDT-auto-SCT consolidation.

Aligned with the control arm of ZUMA-7, the comparator considered within the analysis comprises a basket of the most commonly given treatments for transplantintended patients at second-line, including platinum-containing salvage chemotherapy (R-ICE, R-ESHAP, R-GDP or R-DHAP) followed by HDT (e.g. BEAM) and auto-SCT in responders.

Clinical expert opinion was sought to determine the regimens given in NHS England, in addition to estimates of the distribution across these regimens.[29][29] It was stated that the type of chemotherapy regimen use was centre dependent, however both R- ESHAP and R-DHAP was rarely used in England. Clinicians also stated that a lot of centres in England were moving towards using R-GDP, as it is possible to administer

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 85 of 162

in an outpatient setting, and therefore had the possibility of using less inpatient beds. There was general consensus that R-ICE and R-GDP are the most commonly used regimens in England, therefore an equal split was assumed across these two regimens. In addition, clinicians stated that it is reasonable to assume equal efficacy across all four of the platinum-based chemotherapy regimens, therefore the distribution of use was only expected to affect costs.[29][29]

Importantly, CAR T-cell therapies, axi-cel and tisagenlecleucel are approved to treat patients with DLBCL after at least two prior therapies (i.e. third-line+). In line with NICE’s position statement on the inclusion of CDF-funded treatments as comparators or subsequent therapies[77] , third-line CAR T-cell therapies are excluded from this analysis. Further details are provided in Section B.3.3 and Section B.3.5. Other subsequent treatment options, based on the final scope of the previous thirdline CAR T-cell therapy appraisals, include nivolumab, pembrolizumab, polatuzumab, lenalidomide, auto-SCT, allo-SCT, and best supportive care (including radiotherapy). However, some of these treatments are given in an experimental setting only, therefore have been excluded from the analysis. Further details are provided in Section B.3.5.[78, 79]

B.3.3. Clinical parameters and variables

B.3.3.1. Clinical effectiveness data overview

EFS, OS and TTNT expectations for axi-cel and SOC were based on the latest available data for the FAS population of the ZUMA-7 trial (data cut-off date 18 March 2021). The median potential follow-up time was 24.9 months and the median actual follow-up time was '''''''''' months.[1, 44]

Previous CAR T-cell therapy appraisals have focussed on the mITT population for the axi-cel arm (i.e. patients receiving CAR T-cell therapy infusion). However, this approach has been critiqued due to the need to determine outcomes for patients who failed to receive treatment due to events occurring prior to infusion, such as death during the manufacturing period, adverse events associated with pretreatment, or manufacturing failures.[52, 72] Therefore, the FAS population is used to determine outcomes in the current appraisal, providing data for 180 patients receiving infusion with axi-cel and 179 patients receiving SOC.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 86 of 162

Kaplan–Meier data for EFS, OS and TTNT for both axi-cel and SOC are presented in Figure 19 to Figure 22. Two different Kaplan–Meier data is presented for OS as one shows the mITT population (Figure 20) and the second shows the crossover adjusted Kaplan–Meier data (Figure 21) as this is used in the base case. Extrapolation of trial survival data was required to capture lifetime outcomes following guidance in NICE Decision Support Unit (DSU) Technical Support Documents (TSDs).[80, 81]

Figure 19: Kaplan–Meier plot for EFS as per central assessment, FAS

==> picture [452 x 244] intentionally omitted <==

Key: EFS, event-free survival; FAS, full analysis set. Source: Locke et al. 2021.[1 ]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 87 of 162

Figure 20: Kaplan–Meier plot for OS, FAS

==> picture [452 x 223] intentionally omitted <==

Key: FAS, full analysis set; OS, overall survival. Source: Locke et al. 2021[1]

Figure 21: Kaplan–Meier plot of OS – analysis using RPSFT model, FAS

==> picture [452 x 226] intentionally omitted <==

Key: CI, confidence interval; FAS, full analysis set; HR, hazard ratio; mo, months; NE, not estimable; NR, not reached; OS, overall survival; RPSFT, rank-preserving structural failure time; SOC, standard of care.

Source: Locke et al. 2021[1]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 88 of 162

Figure 22: Kaplan–Meier plot for TTNT, FAS

==> picture [452 x 192] intentionally omitted <==

Key: CI, confidence interval; FAS, full analysis set; HR, hazard ratio; NE, not estimable; SOCT, standard of care therapy; TTNT, time to next treatment. Source: Figure 21. ZUMA-7 CSR.[44]

Despite the relatively short follow-up period and small number of patients at risk, the flattening tails observed in the EFS and OS data suggest a proportion of r/r DLBCL patients at this line experience long-term remission and survival. The following sections illustrate how, in comparison to standard parametric survival approaches described in TSD 14, more flexible ‘mixture cure’ methodologies as described in TSD 21 better fit both these data and expectation of long-term prospects for patients responding to CAR T-cell therapy. There is also empirical support for the use of mixture cure modelling to extrapolate trial OS estimates with ZUMA-1 data (axi-cel in 3L).[82] Most importantly, validation of predicted survival estimates was performed using data from ZUMA-1 (axi-cel in 3L) and the insights of clinical experts.[29, 56]

Importantly, 56% patients on the SOC arm received subsequent CAR T-cell therapies, which are currently only available to patients in NHS England via the CDF. Aligned with NICE’s position statement on this issue,[65] crossover analysis was conducted, in line with guidance from NICE DSU TSD 16.[83] This analysis attempts to remove the confounding effect of subsequent CAR T-cell therapies on SOC survival estimates (see Section B.3.3.4.1). Note that the impact of crossover only affects OS estimates, as treatment switch to subsequent lymphoma therapy was considered an event as per the definition of EFS and TTNT in the ZUMA-7 trial. Here, we present

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 89 of 162

both crossover adjusted analyses for SOC (base case, in line with NICE guidance) as well as a scenario analysis which uses the unadjusted ITT data. The latter, reflecting the state of the world where subsequent CAR T-cell therapies receive a positive recommendation following reassessment later this year. Crossover adjusted analyses were interpreted within the context of observational datasets for r/r DLBCL patients, prior to the availability of CAR T-cell therapies. Aligned with previous appraisals of CAR T-cell therapies in DLBCL, ORCHARRD and SCHOLAR-1 datasets summarised in Table 3 were used as validation of SOC OS extrapolation.

B.3.3.2. Mixture cure models

NICE TSDs 14 and 21[80, 81] discuss the potential benefits of using more flexible models when standard parametric curves do not provide a good fit to the observed data. Mixture cure models represent an alternative, more flexible approach to modelling EFS, OS and TTNT for axi-cel that can potentially account for more complex hazard functions. The use of these models can be beneficial over standard parametric models where there is evidence to support that a proportion of patients have more favourable outcomes (i.e. experience long-term survivorship) following treatment, and a proportion do not.

Following NICE TSDs 14 and 21[80, 81] , mixture cure models were estimated using the ZUMA-7 patient-level data, for which a logistic regression with maximum likelihood estimation using R and the package flexsurvcure was used to model the probability that patients experienced a ‘statistical cure’.[84] Associated cure fractions are presented in later sections (B.3.3.3 and B.3.3.4). Applying this survivor fraction splits the ZUMA-7 population into two groups: patients who experience a ‘statistical cure’ and those who do not. Mortality for ‘statistically cured’ (hereafter known as ‘cured’) patients is captured by standardised mortality ratio (SMR)-adjusted age- and gender-matched general population mortality data (derived from UK Life Table data)[85] ; for patients in the latter group, risk of progression was defined by the standard parametric survival model fits to ZUMA-7 data as reported in Appendix O.

In line with previous appraisals of CAR T-cell therapies, including the 3L DLBCL appraisals, an SMR of 1.09, derived from the publication by Maurer (2014) was used in the base case to adjust for excess mortality in long-term survivors.[52, 55, 72] Assuming the same excess mortality as per the 3L indication could arguably be

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 90 of 162

considered a conservative approach, given the better prognosis of less heavily pretreated patients.

The survival estimates for the overall population treated with a potentially curative intervention is the weighted average of the survival among the ‘cured’ and ‘noncured’ patients. The survival function is described as:

==> picture [159 x 13] intentionally omitted <==

Where S(t) denotes survival probability at time t, S* is the survival in the general population associated with background mortality, Su is the survival probability associated with the excess disease-related risk, and p denotes the cure fraction.[86]

The use of mixture cure models is statistically feasible regardless of the intervention used, as the model will determine a cure fraction based on the observed trial data and exogenous mortality data. However, good practice dictates that it should only be used when a “cure” is clinically feasible.

In a recent study looking at the accuracy of different extrapolation techniques in the ZUMA-1 trial (a phase II single-arm study of patients [N=101] given axi-cel in 3L LBCL) found that mixture-cure models were the most accurate models for predicting OS over the long-term.[56] This study fitted spline, mixture cure, non-mixture and single distribution models to the 12-month ZUMA-1 data cut. Extrapolations were then evaluated against the 24-, 36- and 48-month follow-up data using a range of metrics, including AIC and BIC. Single parametric models poorly predicted long-term survival in axi-cel treated patients. Therefore, the use of mixture cure models can be justified in this case.

Mixture cure models have been used for decision making in multiple previous CAR T-cell therapy appraisals, where, similar to this appraisal, the observed data were immature and where there was clinical expectation of a plateau in PFS/OS.[52, 72, 87]

B.3.3.3. Event-free survival analysis

This section details the approaches to modelling EFS for the axi-cel and SOC treatment arms. Patients randomised to the axi-cel arm experience a clear benefit in

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 91 of 162

terms of EFS in comparison to SOC patients as demonstrated by the HR of 0.398 (95% CI: 0.308, 0.514).

Kaplan-Meier plots and Cox regression results for ZUMA-7 EFS are reported in Appendix O. Although a treatment effect for axi-cel was observed and the proportional hazards assumption seems to be valid, the parallelism between curves was lost towards the end of the log-log plot for EFS. Therefore, the proportional hazards assumption was assumed not to hold for EFS across the entire time horizon and independent survival models have been fitted for axi-cel and SOC as per the NICE DSU guidance.[80]

As specified in NICE TSDs 14 and 21[80, 81] , standard parametric distributions and the mixture cure models were fit to each arm of the trial data, as well as spline models. Results for mixture cure models are described in this section, while standard parametric models and spline models are reported in Appendix O.

As described in sections B.3.3.1 and B.3.3.2, a more flexible approach to modelling EFS, mixture cure models, is considered appropriate and would better fit both these data and expectation of long-term prospects for patients responding to CAR T-cell therapy. This approach was validated by UK clinical and health economic experts.[29, ] 43

Table 21 reports the cure fraction as estimated by the mixture cure models, derived based on methods described in section B.3.3.2. The cure fractions represent the proportion of patients that experience adjusted general population mortality as determined by data on the pattern of death observed in ZUMA-7. For axi-cel the predicted cure fractions were between 35% and 39% with predicted cure fractions for SOC of between 14% and 16%.

EFS projections for each mixture cure model are presented for axi-cel and SOC in Figure 23 and Figure 24, respectively, with smoothed hazard plots presented in Appendix O. AIC and BIC statistics and landmark survival estimates are presented in Table 22.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 92 of 162

B.3.3.3.1. Base case EFS models

For both axi-cel and SOC, results show that the predicted EFS outcomes are very similar across different mixture cure models. Due to similar predictions across all models, the best-fitting models were selected in the base case analysis. These were the log-logistic and exponential models for axi-cel and SOC, respectively.

Alternative EFS curve selection is tested in the scenario analysis.

Table 21: EFS, mixture cure model, implied cure fractions

Figure 23: Axi-cel EFS, mixture cure models

==> picture [454 x 93] intentionally omitted <==

==> picture [454 x 94] intentionally omitted <==

==> picture [454 x 93] intentionally omitted <==

Key: EFS, event-free survival; KM, Kaplan–Meier.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 93 of 162

Figure 24: SOC EFS, mixture cure models

==> picture [451 x 104] intentionally omitted <==

==> picture [451 x 104] intentionally omitted <==

==> picture [451 x 104] intentionally omitted <==

Key: EFS, event-free survival; KM, Kaplan–Meier; SOC, standard of care.

Table 22: EFS, mixture cure model AIC and BIC statistics and landmark survival estimates

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 94 of 162

Figure 25 shows the modelled base case EFS curves for both axi-cel and SOC.

Figure 25: Modelled base case EFS curves

==> picture [446 x 128] intentionally omitted <==

==> picture [446 x 129] intentionally omitted <==

Key: EFS, event free survival; KM, Kaplan-Meier; SOC, standard of care Note: Axi-cel is modelled using log-logistic mixture-cure model; SOC is modelled using exponential mixture-cure model

B.3.3.4. Overall survival analysis

As discussed in Section B.3.3.1, the interpretation of OS based on data from ZUMA7 is challenging as a significant proportion of patients on the SOC arm of the trial went on to receive CAR Ts as subsequent therapies. As per NICE guidance, the

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 95 of 162

model base case attempts to adjust for the impact of subsequent CAR T treatments, however for completeness, unadjusted estimates of OS for SOC patients (as per the ITT population of ZUMA-7) are also presented and reflect a state of the world where subsequent CAR T-cell therapies are routinely commissioned within NHS England.

B.3.3.4.1. Crossover analysis

A significant proportion of patients in the SOC arm of ZUMA-7 went on to receive subsequent CAR T-cell therapies: 56% are estimated to receive, which are currently only available to patients in NHS England via the CDF. Aligned with NICE’s position statement on this issue,[65] crossover analysis was conducted, in line with NICE DSU TSD 16,[83] to attempt to adjust survival estimates for SOC patients to remove the confounding effect of subsequent CAR T-cell therapies. Full details of the methods and results of the crossover analyses conducted are presented in Appendix S. Different models were explored. Rank preserving structural failure time models (RPSFTM) results were summarised below. Results from Inverse Probability of Censoring Weighting (IPCW) models were reported in Appendix S. A two-stage model is applicable if there is an identifiable secondary baseline time when patients switch. A suitable secondary baseline could not be identified for this study, since patients switched to cell therapy at various different points. Therefore, the two-stage model is not appropriate in ZUMA-7 given the switching mechanism, and it is not considered.

B.3.3.4.1.1. RPSFTM

As discussed in section B.2.6.3.3, pre-specified analyses using inverse probability of censoring weights and rank preserving structural failure time models (RPSFTM) were explored as part of the ZUMA-7 analyses conducted by KITE, with RPSFT model stratified (HR: 0.58; 95% CI: 0.42, 0.81; see Figure 8). The analysis shows the sensitivity of the treatment effect (hazard ratio) as a result of treatment crossover.

The pre-specified analysis is based on recesoring switchers only. The recommendation around recensoring[88] is to present results both without any recensoring and with full recensoring of all control arm patients, not just recensoring switchers. Further RPSFTM based on censoring is explored and summarised in Table 23.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 96 of 162

It is worth noting that there was some evidence of non-proportional hazards in the “no recensoring” and “recensoring switchers only” models, but in the full recensoring analyses the proportional assumptions held. This is likely to be because these first two models exhibit a plateau in the control arm survival after around 18 months, whereas the full recensoring analysis recensors these patients prior to 18m and no plateau is seen.

The model applies the HR based on RPSFTM recensoring (0.425) in the base case analysis.

Table 23 Summary of OS results from ITT and standard RPSFTM analyses

(stratified)

A summary of the crossover adjusted Kaplan–Meier data for SOC is presented in Figure 26. Due to the inherent uncertainty in all methods for adjusting for crossover, it was important to validate analysis outcomes through external datasets and clinical expert consultation. The ORCHARRD and SCHOLAR-1 provide outcomes data for r/r DLBCL patients in the absence of CAR T-cell therapy. As summarised in Table 3, there are differences in the patient and study characteristics of ORCHARRD and

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 97 of 162

SCHOLAR-1 compared with ZUMA-7. Therefore, the adjusted OS from ZUMA-7 crossover analysis would not completely align with observation from ORCHARRD and SCHOLAR-1. This is further confirmed by clinical experts that the expected OS trend among DLBCL patients eligible for 2L treatments is likely to be between the observation from ORCHARRD and SCHOLAR-1.[29, 43]

Figure 26: OS estimates adjusted for crossover, compared to external datasets

==> picture [452 x 416] intentionally omitted <==

Key: IPCW, inverse-probability-of-censoring weighting; RPSFTM, rank preserving structural failure time model; SOCT, standard of care therapy.

B.3.3.4.2. Mixture cure models

As described in section B.3.3.1 and B.3.3.4.1, the model base case attempts to adjust for the impact of subsequent CAR T treatments, to reflect the current practice

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 98 of 162

where subsequent CAR T-cell therapies are not routinely commissioned within NHS England. Only the crossover (base case) analysis results are reported in this section. Results from ITT population are described in Appendix Q.

This section details the approaches to modelling OS for the axi-cel and SOC treatment arms. Kaplan-Meier plots and Cox regression results for ZUMA-7 OS are reported in Appendix O, showing that the proportional hazards assumption was not held for OS. Independent survival models have been fitted for axi-cel and SOC as per the NICE TSD14.[80]

Following the guidance from NICE TSDs 14 and 21,[80, 81] standard parametric distributions, mixture cure models and spline models were fit to each arm of the trial data. Results for mixture cure models are described in this section, while standard parametric models and spline models are reported in Appendix O.

As described previously in sections B.3.3.1 and B.3.3.2, mixture cure models is considered appropriate and would better fit both these data and expectation of longterm prospects in OS for patients responding to CAR T-cell therapy. This approach was validated by clinical and health economic experts.[29, 43][29]

The cure fraction, reported in Table 24, is derived based on methods described in section B.3.3.2. For axi-cel the predicted cure fractions were between 24% and 54% and predicted cure fractions for SOC between 32% and 49%.

OS extrapolation for each mixture cure model are presented for axi-cel and SOC in Figure 27 and Figure 28, respectively, with smoothed hazard plots presented in Appendix O. AIC and BIC statistics and landmark survival estimates are presented in Table 25.

B.3.3.4.3. Base case OS models

For both axi-cel and SOC, the log-logistic, generalized gamma and log-normal models provide the best statistical fit based on AIC/BIC.

As described in Section B.3.3.4.1, clinical experts expected that in the absence of CAR T-cell therapies, SOC OS would likely be falling somewhere between ORCHARRD (as presented) and SCHOLAR-1, with a survival plateau similar to that

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 99 of 162

of EFS at ~20%. Clinical experts further indicated that SOC OS and EFS are likely to converge by 5 years.[29] Table 25 and Figure 28 show that at 5 years, the predicted SoC OS by two of the best fitting models (log-logistic and log-normal) would not reflect this clinical expectation. Therefore, log-logistic and log-normal models are not considered due to lack of clinical plausibility.

It is worth noting that the base case analysis described in this section and presented in Figure 28 has taken into account the crossover adjustment in the SOC arm, as described in section B.3.3.4.1. The model base case applied the HR based on the RPSFTM full recensoring analysis (HR=0.425). This is different to the pre-specified analysis from ZUMA-7 (see section B.2.6.3.3, B.3.3.4.1.1 and Figure 8) where the analysis is for RPSFTM, recensoring with switchers only (HR=0.58). The crossover adjustment is not relevant for the axi-cel arm.

For axi-cel arm, published evidence reports sustained plateau in long-term OS. Based on the most recent ZUMA-1 data, 44% of patients treated with axi-cel at 3L were still alive 4 years later, supporting the emerging plateau in ZUMA-7 OS.[56]

There is also empirical support for the use of mixture cure modelling to extrapolate trial OS estimates. A recent study using five years’ worth of follow-up data from ZUMA-1 demonstrated that cure models, when fitted to ‘immature’ OS data, most accurately and reliably predicted long-term survival of axi-cel treated DLBCL patient versus spline based and standard parametric models, the latter of which substantially underestimated lifetime survival.[82]

In the base case, generalized gamma mixture cure model is selected for axi-cel as it provided the best statistical fit and had the most clinical plausibility and the HR relative to axi-cel is used in the SOC arm.

Alternative OS curve selection is tested in the scenario analysis.

Clinicians agreed that despite the crossover adjustment, the modelled survival for the SOC arm predicted optimistic outcomes for patients who will not receive subsequent cell therapy in third line setting. In order to address this, a separate scenario analysis was also included to explore SOC OS and EFS converging at 5 years.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 100 of 162

Table 24 : Axi-cel OS, implied cure fractions

Figure 27 : Axi-cel OS, mixture cure models

==> picture [448 x 107] intentionally omitted <==

==> picture [448 x 106] intentionally omitted <==

==> picture [448 x 106] intentionally omitted <==

Key: KM, Kaplan–Meier; OS, overall survival.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 101 of of 162

101 of of 162

Figure 28 : SOC OS, mixture cure models, crossover

==> picture [458 x 103] intentionally omitted <==

==> picture [458 x 103] intentionally omitted <==

==> picture [458 x 103] intentionally omitted <==

Key: CR, complete response; KM, Kaplan–Meier; OS, overall survival; SOC, standard of care. . Notes: ORCHARRD data presented for population with the CR≤ 12 months

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 102 of of 162

102 of of 162

Table 25: OS mixture cure model AIC and BIC statistics and landmark survival estimates, crossover

Figure 29 shows the modelled base case curves for OS in the axi-cel and SOC arms.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 103 of of 162

103 of of 162

Figure 29: Modelled base case overall survival curves

==> picture [450 x 129] intentionally omitted <==

==> picture [450 x 129] intentionally omitted <==

Key: KM, Kaplan–Meier; OS, overall survival; SOC, standard of care Note: Axi-cel is modelled using the generalized gamma mixture-cure model; SOC is modelled using the crossover adjusted curve

B.3.3.5. TTNT analysis

This section details the approaches to modelling Time to Next Treatment (TTNT) for the axi-cel and SoC treatment arms. TTNT curves were used in the model to determine the time at which patients receive subsequent therapy costs.

As per the approach for OS and EFS, Kaplan-Meier plots and Cox regression results for TTNT are reported in Appendix O. The parallelism between curves was lost at several timepoints of the log-log plot for TTNT. Therefore, the proportional hazards assumption was assumed not to hold for TTNT across the entire time horizon and independent survival models have been fitted for axi-cel and SoC as per the NICE DSU guidance.[80]

Standard parametric distributions and the mixture cure models were fit to each arm of the trial data. Spline models were not explored. Results for mixture cure models are described in this section, while standard parametric models are reported in Appendix O.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 104 of 162

Mixture cure models for axi-cel and SoC TTNT are presented alongside ZUMA-7 TTNT Kaplan–Meier data in Figure 30 and Figure 31, respectively, with cure fractions presented in Table 26. Smoothed hazard plots are presented in Appendix O. AIC/BIC statistics and landmark estimates are presented in Table 27.

B.3.3.5.1. Base case TTNT models

For both axi-cel and SOC, results show that the predicted TTNT overtime are similar across different mixture cure models. Due to similar predictions across all models, the best-fitting models were selected in the base case analysis. The mixture cure models using a Loglogistic function provided the best fit for both axi-cel and SoC. The long-term TTNT extrapolations aligned with feedback from clinical experts.

Alternative TTNT curve selection is tested in the scenario analysis.

Table 26: Axi-cel TTNT, implied cure fractions

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 105 of of 162

105 of of 162

Figure 30 : Axi-cel TTNT, mixture cure models

==> picture [410 x 137] intentionally omitted <==

==> picture [410 x 136] intentionally omitted <==

Key: KM, Kaplan–Meier; TTNT, time to next treatment.

Figure 31 : SOC TTNT, mixture cure models

==> picture [379 x 138] intentionally omitted <==

==> picture [379 x 137] intentionally omitted <==

Key: KM, Kaplan–Meier; SOC, standard of care; TTNT, time to next treatment.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 106 of of 162

106 of of 162

Table 27: TTNT mixture cure model AIC and BIC statistics and landmark survival estimates

Figure 32 shows the modelled TTNT base case curves for both axi-cel and SOC.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 107 of of 162

107 of of 162

Figure 32: Modelled base case TTNT curves

==> picture [433 x 125] intentionally omitted <==

==> picture [433 x 125] intentionally omitted <==

Key: KM, Kaplan–Meier; SOC, standard of care; TTNT, time to next treatment. Note: Axi-cel is modelled using the log-logistic mixture-cure model; SOC is modelled using the loglogistic mixture-cure model

B.3.3.6. General population mortality

To ensure the hazard of death in the r/r DLBCL population was never less than that of the general population, background mortality was incorporated into the model based on age and sex matched general population mortality estimates from UK National Life Tables published by the Office for National Statistics.[85] The choice of pre-2020 mortality rates was intentional, to ensure that excess mortality associated with COVID-19 was not captured, thus reflecting typical mortality rates for the population receiving axi-cel.

As discussed throughout section B.3.3, an SMR of 1.09 was applied, to general population mortality estimates, to reflect excess mortality experienced by long-term survivors. This approach is aligned with previous technology appraisals for CAR T- cell therapies.[52, 72]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 108 of 162

B.3.4. Measurement and valuation of health effects

Section B.1.3.3 describes the negative impact of r/r DLBCL on patients’ quality of life.

B.3.4.1. Health-related quality of life data from clinical trials

Health-related quality of life (HRQL) data were collected in the ZUMA-7 QoL analysis set using both the EORTC QLQ-C30 and the EQ-5D-5L. The NICE reference case stipulates that the EQ-5D is the preferred measure of HRQL in adults[76] , and as such the EQ-5D results were used to derive utility values in the event-free state of the cost-effectiveness model.

In the axi-cel arm of ZUMA-7, data were collected at screening, the first day of conditioning chemotherapy, the day of axi-cel administration, and Months 2, 3, 5, 9, 12, 15, 18, 21 and 24 after randomisation. In the SOC arm, the data were collected at screening, approximately 5 days after randomisation (during the first cycle of salvage chemotherapy), at the time of disease assessment (assumed to be approximately Day 50/Month 2), the day of transplant for those receiving auto-SCT, and then Day 100 and Day 150 post-randomisation (Months 3 and 5) as well as Months 9, 12, 15, 18, 21 and 24.

Of the 359 patients enrolled in the ZUMA-7 study, 165 in the axi-cel arm and 131 in the SOC arm had baseline HRQL responses and ≥1 follow-up measure and were included for analysis in the ZUMA-7 QoL analysis set.

The health state utility values used in the economic analysis (derived from ZUMA-7 data in the event-free state) are presented in Section B.3.4.5. Notably, patients in the QoL analysis set of ZUMA-7 were not mandated as per the protocol to complete patient reported outcome questionnaires after an EFS event, and as such, data is sparse and potentially biased, introducing both statistical and clinical uncertainty. Details on the sparsity of the data is highlighted in the PRO report (in Appendix T). Therefore, for the base case, the utility for the post-event state is assumed to be equal to the utility derived from the JULIET trial data for tisagenlecleucel in R/R LBCL, as applied in the NICE submission following a mapping exercise.[52, 72] Alternative utility values sourced from the ZUMA-1 trial are explored in scenario analysis.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 109 of 162

B.3.4.2. Mapping

As described in Section B.3.4.1, the EQ-5D-5L questionnaire was administered to patients in the ZUMA-7 trial. As recommended by NICE in their updated position statement in October 2019[77] , the crosswalk algorithm developed by van Hout et al. (2012) was used to convert EQ-5D-5L scores into EQ-5D-3L utility values.[89]

The resulting EQ-5D-3L utility value of 0.785 was used in the model base case for the event-free health state. Furthermore, utility values for the event-free state were segregated further based on treatment status. This resulted in an “on-treatment” utility of 0.772 for the SOC arm, and for axi-cel patients a lower utility of 0.780. These were derived according to the following criteria:

• Axi-cel on treatment, pre-event: All visits that were after the axi-cel treatment start date and prior to the axi-cel treatment end date or date of event (whichever is sooner)

• SOC on treatment, pre-event: All visits that were after the SOC start date and prior to the SOC treatment end date or date of event (whichever is sooner).

Applying on-treatment specific utility values was explored in a scenario analysis, where they were applied for one month in the axi-cel arm to account for the fact that patients have recently relapsed from first-line treatment (average time between leukapheresis and infusion with axi-cel), and three months in the SOC arm. This approach has been taken in previous models for other CAR T-cell therapies in a third-line setting.[52, 72] Health state utilities in the base case are summarised in Section B.3.4.5.

B.3.4.3. Health-related quality of life studies

In line with the search for economic evaluations (described in Section B.3.1), a systematic search was conducted to identify HRQL evidence in adults with r/r DLBCL after first-line therapy only. The study identification process, search strategies and a description of the included studies is provided in Appendix H.

No studies that reported health state utility values for the population of interest were identified in the search for HRQL evidence, therefore insights were drawn from the

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 110 of 162

two previous completed NICE single technology appraisals in r/r DLBCL (TA559 and TA567).[52, 72]

In TA559, health state utility values were based on EQ-5D data collected in a safety management cohort of the ZUMA-1 trial (n = 34, with 87 observations) were used in the base case analysis. In TA567, health state utility values were derived from SF-36 data collected in the JULIET study, mapped to EQ-5D. In both TA559 and TA567, health state utilities taken from NICE TA306 were tested in the scenario analysis.

Table 28 presents health state utility values sourced from prior NICE appraisals in patients with r/r DLBCL. The utility values reported in TA559 from the ZUMA-1 study are explored in a scenario analysis.

Table 28: Health state utility values from prior NICE appraisals

B.3.4.4. Adverse reactions

B.3.4.4.1. Adverse event data

As reported in NICE TA677, clinicians have become increasingly comfortable with toxicity management for CAR T-cell therapies.[87] However, following treatment with axi-cel, it is acknowledged that there may still be short-term impactful Aes. Table 29 presents Grade 3+ Aes that occurred in ≥10% of patients for axi-cel and SOC captured in the cost-effectiveness model, sourced from the ZUMA-7 study.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 111 of of 162

111 of of 162

Table 29: Adverse event data (Grade 3+, ZUMA-7)

B.3.4.4.2. Adverse event disutility

Adverse events associated with axi-cel and SOC are expected to occur in the shortterm after initial treatment, therefore a one-off QALY decrement is applied in the first model cycle.

Utility decrements for anaemia, febrile neutropenia, neutropenia, platelet count decrease and thrombocytopenia were obtained from the pixantrone submission to NICE.[71] For patients experiencing CRS, it is assumed that patients have a quality of life of zero (i.e. the utility decrement is set to be the negative value of the event-free health state). This is in line with the York study[90] , which was the method adopted for the third-line DLBCL axi-cel NICE submission (TA559).[52] Disutilities associated with the remaining AEs were not identified, therefore for each of these AEs, a disutility equal to the maximum of the identified non-CRS AE disutilities was assumed. This is in line with the pixatrone submission to NICE (TA306)[71] as well as the third-line DLBCL axi-cel NICE submission (TA559).[52] The duration of CRS and neurologic events was obtained from ZUMA-7, whilst the remaining duration of adverse events were sourced from ZUMA-1 patient level data, in line with the axi-cel third-line

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 112 of of 162

112 of of 162

DLBCL NICE submission (TA559).[52] Table 30 presents the AE disutilities and durations, and their respective data sources.

Table 30: Adverse event disutilities

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 113 of of 162

113 of of 162

B.3.4.5. Health-related quality of life data used in the cost-effectiveness analysis

As described in Section B.3.4.1 and B.3.4.2, utility values for the event free states are derived using trial-based EQ-5D-5L scores mapped to EQ-5D-3L values. As PRO questionnaires were not administered in ZUMA-7 post-event, it is assumed the progressed disease utility value from the JULIET study (reported in TA567) is applicable to the post-event state (0.710). The progressed-disease utility was used as this health state represents all post-event patients including those that have progressed after 3L treatment. ZUMA-1 utility data were also considered, however given the small number of post-progression observations (<5% of the sample with data), JULIET data were preferred. This is aligned with ERG feedback on TA559.[52]

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 114 of of 162

114 of of 162

In line with previous appraisals of CAR T-cell therapies, it is assumed that the healthrelated quality of life for long-term survivors, remaining event free (for both SOC and axi-cel) would eventually return to that of the age- and gender-matched general population values, reflective of the fact that patients would be effectively cured. Historically, there has been debate around when this may occur. The company submission for TA559 assumed this would happen after 2 years, however this was challenged by the ERG.[52] In this appraisal we assume a more conservative estimate of 5 years, in line with latest committee preferences for CAR T-cell therapies.[87]

General population utility estimates, applied to long-term survivors remaining event free after 5 years, were obtained from national publications for the UK, as shown in Table 31.[94]

Table 31: UK general population utility[94]

A summary of the utility values applied in the model base case is presented in Table 32.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 115 of of 162

115 of of 162

Table 32: Summary of utility values for cost-effectiveness analysis

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 116 of 162

Key: DLBCL, diffuse large B-cell lymphoma; EFS event-free survival; HRQL, health-related quality of life; N/A, not applicable; PRO, patient reported outcomes; SOC standard of care; TA, technology appraisal.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 117 of 162

B.3.5. Cost and healthcare resource use identification, measurement and valuation

B.3.5.1. Cost and resource use estimates identified in the literature

A systematic search for published healthcare costs and resource use studies for patients with early relapsed or primary refractory DLBCL was conducted alongside the search for published cost-effectiveness studies as detailed in B.3.1 and Appendix G. Full details of the search methods and results are presented in Appendix I.

As with the cost-effectiveness model structure, costs and resource use inputs are largely aligned with prior single technology appraisals of CAR T-cell therapies in DLBCL (TA559 and TA567).

B.3.5.2. Intervention and comparators’ costs and resource use

B.3.5.2.1. Axi-cel costs and resource use

For axi-cel, treatment-related costs included in the model are

• Leukapheresis

• Bridging therapy

• Conditioning chemotherapy

• Axi-cel acquisition costs

• Axi-cel infusion and monitoring (hospitalisation)

As described in Section B.2.4.1, in the ZUMA-7 FAS population, '''''''''' patients were randomised to the axi-cel arm. Subsequently, ''''''''' ''''''''''''''''' underwent leukapheresis, '''''' '''''''''''''''''' received bridging therapy, '''''''''' ''''''''''''''''' received conditioning therapy and ''''''''' ''''''''''''''''''' received axi-cel treatment. A further ''' '''''''''''' received axi-cel re-

treatment.

All axi-cel related costs included in the analysis have been scaled according to these proportions.

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 118 of 162

Leukapheresis costs

The cost of leukapheresis was obtained from NHS reference costs 2019/2020 and is based on the cost for peripheral blood stem cell harvest and bone marrow harvest for all HRGs. Table 33 details the costs of leukapheresis applied in the model. As discussed previously, ''''''''''''''' of patients on the axi-cel arm of ZUMA-7 received leukapheresis, therefore leukapheresis costs were weighted according to this proportion.

Table 33: Unit costs of leukapheresis

Bridging therapy and conditioning chemotherapy costs

In ZUMA-7, patients were permitted to receive bridging therapy after leukapheresis and up to 5 days before the administration of axi-cel. Bridging therapy was considered for any patient but particularly for those with high disease burden at screening, to maintain stable disease during the manufacturing process.

Bridging therapy in ZUMA-7 consisted of corticosteroid treatment (for example, dexamethasone at a dose of 20 to 40 mg or equivalent, either orally or IV daily for 1– 4 days). The choice of corticosteroid and dosing was based on clinical judgement. As discussed previously ***** of patients on the axi-cel arm of ZUMA-7 received bridging therapy. On consultation, clinical experts explained that the proportion of patients expected to receive bridging therapy in clinical practice in NHS England was closer to two thirds. Furthermore, rather than oral dexamethasone given in ZUMA-7, it is likely that one or two cycles of R-GDP chemotherapy would be administered in an outpatient setting. Therefore, the model base case was amended to reflect UK

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 119 of of 162

119 of of 162

expert opinion, and two thirds of patients received two cycles of R-GDP in an outpatient setting.[29]

Table 34: Bridging therapy cost calculations

As described in Section B.1.2, patients received lymphodepleting chemotherapy consisting of cyclophosphamide 500 mg/m[2] intravenous and fludarabine 30 mg/m[2] intravenous on the 5[th] , 4[th] and 3[rd] day before axi-cel infusion. This is aligned with the anticipated licence for axi-cel. The costs for conditioning chemotherapy were taken from the electronic market information tool (eMIT). In line with previous CAR T-cell therapy appraisals, conditioning chemotherapy was assumed to be administered in the inpatient setting, costs for which are documented in Section B.3.5.2.2. Unit costs for conditioning and bridging therapies are presented in Table 35 with dosing assumptions and final costs presented in Table 36. Final costs were weighted by the proportions receiving bridging and conditioning in ZUMA-7 as documented above.

Table 35: Unit costs conditioning chemotherapy

Company evidence submission for axicabtagene ciloleucel for treating r/r DLBCL after 1 systemic therapy [ID1684] © Gilead (2022). All rights reserved 120 of of 162

120 of of 162

Table 36: Conditioning chemotherapy cost calculations