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

Single Technology Appraisal

Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B cell lymphoma [ID1576]

Committee Papers

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

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

SINGLE TECHNOLOGY APPRAISAL

Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B cell lymphoma [ID1576]

Contents:

The following documents are made available to consultees and commentators:

The final scope and final stakeholder list are available on the NICE website.

1. Company submission from Roche

2. Clarification questions and company responses

• a. Main response

• b. Economic appendix

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

• a. Lymphoma Action

• b. Royal College of Physicians (NCRI-ACP-RCP-RCR)

4. Expert personal perspectives from:

• a. Dr Sridhar Chaganti, Consultant Haematologist – clinical expert, nominated by NCRI-ACP-RCP

• b. Dr Kate Cwynarski, Consultant Haematologist – clinical expert, nominated by NCRI-ACP-RCP

• c. Mr Stephen Scowcroft, Director of Operations & External Affairs – patient expert, nominated by Lymphoma Action

5. Evidence Review Group report prepared by Kleijnen Systematic Reviews

6. Evidence Review Group report – factual accuracy check

7. Technical report for engagement

8. Technical engagement response from Roche

• a. Response form

• b. Appendix

9. Evidence Review Group critique of company response to technical engagement

• a. Reply form

• b. Addendum 1

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

ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma

Document B

Company evidence submission

July 2019

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 1 of 144

Instructions for companies

This is the template for submission of evidence to the National Institute for Health and Care Excellence (NICE) as part of the single technology appraisal (STA) process. Please note that the information requirements for submissions are summarised in this template; full details of the requirements for pharmaceuticals and devices are in the user guide.

This submission must not be longer than 150 pages, excluding appendices and the pages covered by this template. If it is too long it will not be accepted.

Companies making evidence submissions to NICE should also refer to the NICE guide to the methods of technology appraisal and the NICE guide to the processes of technology appraisal.

In this template any information that should be provided in an appendix is listed in a box.

Highlighting in the template (excluding the contents list)

Square brackets and grey highlighting are used in this template to indicate text that should be replaced with your own text or deleted. These are set up as form fields, so to replace the prompt text in [grey highlighting] with your own text, click anywhere within the highlighted text and type. Your text will overwrite the highlighted section.

To delete grey highlighted text, click anywhere within the text and press DELETE.

Grey highlighted text in the footer does not work as an automatic form field, but serves the same purpose – as prompt text to show where you need to fill in relevant details. Replace the text highlighted in [grey] in the header and footer with appropriate text. (To change the header and footer, double click over the header or footer text. Double click back in the main body text when you have finished.)

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 2 of 144

Contents

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 3 of 144

Tables

==> picture [455 x 693] intentionally omitted <==

----- Start of picture text -----

----- End of picture text -----

Table 43. Predicted long-term survival (cure fraction) from OS cure-mixture model extrapolations (OS not informed by PFS) ............................................................................ 91 Table 44. Predicted long-term survival (cure fractions) from OS informed by PFS curemixture model extrapolations .............................................................................................. 92 Table 45. Incidence of treatment-related AEs included in the model (CTCAE ≥Grade 3, serious) ............................................................................................................................... 97 Table 46. HRQoL and utility studies in R/R DLBCL identified in the SLR ............................ 98 Table 47. HRQoL and utility results from Wang et al. 2018 ................................................. 99 Table 48. Disutility values used in the cost-effectiveness model ....................................... 100 Table 49. Summary of utility values for cost-effectiveness analysis (base case) ............... 102 Table 50. Summary of utility values for cost-effectiveness analysis (scenario analyses) ... 102 Table 51. Drug acquisition costs for Pola+BR, BR and R-GemOx ..................................... 104 Table 52. NHS reference costs 2017–18 for chemotherapy administration ....................... 106 Table 53. Drug administration costs per cycle ................................................................... 107 Table 54. Supportive care resource use unit costs included in the model.......................... 107 Table 55. Annual frequency of resource use in PFS and PD ............................................. 109 Table 56. Per cycle supportive care costs for PFS and PD heath states ........................... 112 Table 57. Subsequent treatment costs based on GO29365 data ...................................... 114 Table 58. Unit costs of treatment-related AEs included in the economic model ................. 115 Table 59. Summary of variables applied in the economic model base case ...................... 117 Table 60: Key assumptions in the economic analysis........................................................ 119 Table 61. Base case deterministic results ......................................................................... 123 Table 62. PSA parameter inputs ....................................................................................... 124 Table 63. Mean probabilistic results .................................................................................. 127 Table 64. DSA results ....................................................................................................... 129 Table 65: Scenario analysis results ................................................................................... 131 Table 66. Comparison of model clinical outcomes vs GO29365 ........................................ 135

Figures

Figure 23. Base case PFS and OS extrapolations .............................................................. 93 Figure 24. KM for OS and PFS from pooled Pola+BR cohort (N=46) and model extrapolations ..................................................................................................................... 94 Figure 25. OS KM for Pola+BR from the ROMULUS study and model extrapolations ......... 95 Figure 26. Time to off-treatment KM plots (GO29365) ......................................................... 96 Figure 27. Overall survival for ITT patient population and population censored for those receiving a treatment with curative intent (SCT or CAR-T) ................................................ 113 Figure 28. Cost-effectiveness plane for Pola+BR versus BR ............................................. 127 Figure 29. Cost-effectiveness acceptability curve for Pola+BR versus BR ........................ 128 Figure 30. Distribution of PSA ICER values for Pola+BR versus BR ................................. 128 Figure 31. Deterministic sensitivity analysis – tornado diagram of the top 15 most influential parameters for Pola+BR versus BR .................................................................................. 130

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 6 of 144

Abbreviations

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 7 of 144

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 8 of 144

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 9 of 144

B.1 Decision problem, description of the technology and clinical care pathway

B.1.1 Decision problem

The submission covers the technology’s full marketing authorisation for this indication.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 10 of 144

Table 1: The decision problem

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B- cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 11 of 144

B.1.2 Description of the technology being appraised

The technology being appraised is described in Table 2. See Appendix C for details of the draft summary of product characteristics (SmPC) and European Public Assessment Report (EPAR).

Table 2: Description of the technology

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 12 of 144

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

treatment pathway

B.1.3.1 Disease overview

Non-Hodgkin lymphoma (NHL) is a heterogeneous group of lymphoproliferative malignancies, with 80–95% of cases arising from B-clls and the remaining from T-cells. NHL is divided between high and low grade NHL subtypes (8). Diffuse large B-cell lymphoma (DLBCL), a high grade B-cell NHL, represents approximately 40% of all lymphoma cases globally and 30–58% of NHL cases (9, 10).

DLBCL is itself heterogeneous and composed of large neoplastic B lymphoid cells that generally express pan B-cell antigens (CD19, CD20, CD22, CD79a) (11). The majority have genetic abnormalities, but there is no single cytogenetic change that is typical or diagnostic.

The clinical heterogeneity of DLBCL has also been recognised at a molecular level by assigning DLBCL into two cell-of-origin categories based on gene expression patterns indicative of different stages of B cell development. One subgroup expresses genes reminiscent of germinal centre B cells (GCB-like DLBCL), the second group expresses genes normally induced during the activation of peripheral blood cells (ABC-like DLBCL). Patients with GCB-like DLBCL have a significantly better prognosis than those with ABC-like DLBCL (12).

Incidence, prevalence and survival statistics

The Haematological Malignancy Research Network (HMRN) estimates that there will be 5,510 new cases of DLBCL each year in the UK, which accounts for approximately 40% of all UK NHL cases (13, 14). The median age at diagnosis of DLBCL in the UK is approximately 70 years (15) and there is a slightly higher incidence among males compared with females.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 13 of 144

The 10-year prevalence is estimated at 28,291 cases (43.4 patients per 100,000 people), again with more male patients affected (16).

Approximately 591 patients per annum are estimated to be treated for relapsed or refractory (R/R) DLBCL not suitable for hematopoietic stem cell transplant[1] .

Prognosis for first-line DLBCL patients

DLBCL is an aggressive, high grade lymphoma with a life expectancy of weeks to months if not treated (18). The prognosis is varied among DLBCL patients; overall response rates to standard chemoimmunotherapy are high, ranging from 88–91% (19), but 5-year overall survival varies significantly according to the Revised International Prognostic Index (R-IPI) score (51–96%) and NCCN-IPI score (33%-96%) (20, 21). Overall, the five-year survival rate following first-line treatment in the UK is approximately 61% (22).

Prognosis for relapsed/refractory (R/R) DLBCL patients

The prognosis is poor for patients with R/R DLBCL, with a median survival of 10 months. Fewer than half of relapsed patients (41%) survive for 12 months. Age is an important prognostic indicator in DLBCL patients who relapse; patients aged ≥65 years have a poorer prognosis compared to those aged <65 years (23).

Outcomes are even worse for patients who are refractory to first-line therapy. The SCHOLAR-1 study, the largest pooled retrospective analysis of patients with refractory DLBCL, showed that median overall survival was just 6.3 months for these patients, with 22% of patients alive at 2 years (24).

Impact on patients

Patients with DLBCL typically present with a rapidly enlarging symptomatic mass, most commonly a nodal enlargement in the neck or abdomen, or, in the case of primary mediastinal large B cell lymphoma, the mediastinum. Systemic "B" symptoms (i.e., fever, weight loss, drenching night sweats) are observed in approximately 30% of patients, with elevated serum lactate dehydrogenase, a well known poor prognostic factor for NHL, in over 50% of patients. Approximately 60% of patients present with advanced stage DLBCL (stage III or IV disease) while 40% have localised disease, usually defined as that which can be contained within one irradiation field (25).

1 This figure is based on the Office of National Statistics reported incidence of 6391 newly diagnosed patients with DLBCL 17. Office for National Statistics. Cancer Registration Statistics 2017. 2018.. Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 14 of 144

There are limited data on the impact of DLBCL on patients’ quality of life (QoL), however patients with high grade NHL demonstrate a lower QoL compared to patients with low grade NHL, including physical, social/family, and emotional factors, functional well-being, as well as higher anxiety (26). This is partly related to uncertainties towards the prognosis of their disease, side effects of treatment and fear of relapse (27), especially given the disappointing efficacy of standard salvage regimens prior to transplant (28). Patients who achieve a complete response (CR) after first-line treatment have demonstrated significant improvements in QoL compared to non-complete responders (29). Patients who are refractory to or relapse following first-line treatment experience even greater anxiety due to the poorer prognosis of their condition and the need for further, often more intensive treatment. This will also increase the demand on hospital services and the use of skilled nursing facilities and hospice services (30). Therefore, there remains an unmet need for additional treatments for R/R DLBCL patients that offer better outcomes over existing treatments, can reduce psychological distress and improve QoL (31).

B.1.3.2 Current treatment practice

Terminology

Salvage therapy: a treatment for cancer that has not responded to other treatments. Note, the use of this term is not consistent within the R/R DLBCL setting – UK clinical experts advised Roche that the term ‘salvage’ is reserved for more intensive therapy aimed at delivering a patient to a potentially curative transplant.

Conditioning regimen: transplant eligible patients who respond to salvage chemotherapy undergo conditioning treatment to consolidate their response. Conditioning regimens include chemotherapy +/- monoclonal antibody therapy or radiotherapy.

Autologous stem cell transplant (ASCT): a procedure in which blood-forming stem cells are removed, stored and later given back to the same patient.

Allogenic stem cell transplant: a procedure in which a patient receives blood-forming stem cells from a genetically similar, but not identical, donor.

A number of treatment guidelines are available for DLBCL including the NICE clinical pathway (NG52) (32), the British Society for Haematology (BSH) (33), ESMO (9) and National Comprehensive Cancer Network (NCCN) (34); however, advice obtained from UK clinical experts at an advisory board meeting[2] confirmed that there is no universal guideline

2 In October 2018, Roche Products Ltd. held an advisory board meeting with nine clinical experts from across the UK to discuss current treatment practice in the management of R/R DLBCL and to gain feedback on approaches to the cost-effectiveness analysis for this submission. Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 15 of 144

followed for the treatment of R/R DLBCL. The advisors confirmed that current clinical practice for this population is likely to vary across the country, depending on the expertise of the treatment centre and will also likely be informed by individual clinician and patient choice (35).

First-line DLBCL treatment

The R-CHOP regimen (rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone) has been the gold standard in the management of DLBCL for over 15 years (36). However, approximately 30–50% of patients are not cured by this treatment, depending on disease stage or prognostic index. Among patients for whom R-CHOP therapy fails, 20% suffer from primary refractory disease (disease does not enter complete remission and/or progresses during or soon after treatment) whereas 30% relapse after achieving complete remission (37).

Relapsed/refractory disease

Relapsed/refractory patients have a poor outcome and most will die from their disease (38). Relapses commonly occur within the first two years, however late relapses are possible for approximately 10% of patients (39) and may be associated with an initial favourable International Prognostic Index (IPI) score and extranodal involvement at diagnosis (40).

Refractory disease is defined as a <50% decrease in lesion size with initial therapy, or the occurrence of new lesions. Patients with progressive or relapsed disease present with new or enlarging lesions after the attainment of complete remission. Therefore, there are three groups of patients who fail first-line therapy:

1. Relapse after complete remission (relapse >3 months after CR)

2. Partial responders with persistent but not progressive disease

3. Refractory to first-line treatment (patients with stable disease or progressive disease, i.e. failure to achieve CR or relapse ≤3 months after CR) (41)

Prognosis varies among these groups, with refractory patients generally having a worse outlook, as demonstrated in SCHOLAR-1. This international multi-cohort retrospective study of pooled data from two Phase III clinical trials demonstrated a median OS of 6.3 months and response rate of 26% (CR 7%) to the next line of therapy among patients with refractory DLBCL (24).

The initial approach to R/R DLBCL is to assess whether the patient is fit for intensive salvage therapy and potentially autologous stem cell transplant (ASCT). The decision flow presented below was compiled following advice obtained from UK clinical experts and reflects how patients are identified as being eligible for transplant in UK clinical practice (35).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 16 of 144

Figure 1: Decision flow for transplant eligibility among R/R DLBCL patients

==> picture [433 x 219] intentionally omitted <==

*per institutional guidelines, considering salvage treatment tolerance, performance status, adequacy of organ function, satisfactory stem cell collection, patient choice, etc. 2L, second-line; 3L, third-line; ASCT, autologous stem cell transplant; R, rituximab; R/R, relapsed or refractory; SoC, standard of care

Treatment of clear or borderline candidates for transplant

In the PARMA trial, salvage chemotherapy followed by ASCT resulted in significantly superior event-free survival and OS in patients with relapsed DLBCL compared with salvage chemotherapy alone, which led to salvage chemotherapy plus ASCT being adopted as the SOC for R/R patients (42). However, ASCT is typically only available for younger, fit patients, although age alone should not be an absolute contraindication to ASCT (43). Eligibility for ASCT should also take into account other factors such as comorbidities e.g. severe pulmonary compromise or left ventricular dysfunction (44, 45). In UK clinical practice, there is no universal guidance on how to assess whether a patient is a suitable candidate for intensive therapy. This is typically an individualised decision, taking into account age (<70– 75 years), if the patient has chemo-sensitive disease, if stem cells can be harvested, and if the patient has sufficient organ fitness to receive such treatment (35).

For patients who are eligible for ASCT, the first approach is to administer a rituximab-based salvage chemotherapy regimen to minimise disease burden and demonstrate chemosensitivity, followed by consolidation with a high-dose regimen. Response to, and tolerance of, salvage chemotherapy may also confirm eligibility of borderline candidates to receive ASCT since demonstration of response to such treatment is a highly predictive factor of outcome following ASCT.

There are many salvage therapies available, mostly involving rituximab in combination with standard antineoplastic agents (43), as highlighted in the ESMO guidelines for treating R/R

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 17 of 144

DLBCL (although UK clinical experts confirmed that these are not routinely followed in UK clinical practice) (Table 3) (9). For patients fit enough to tolerate high-intensity salvage therapy, NICE guidance recommends offering salvage therapy with multi-agent immunochemotherapy, with R-GDP (rituximab with gemcitabine, dexamethasone and cisplatin) specifically mentioned due to its more tolerable toxicity profile compared to other salvage regimens (32).

Table 3: ESMO guidelines for patients with first and second relapse or progression

*Following advice from UK clinical experts, oxaplatin is the preferred platinum regimen for transplant-ineligible patients in UK clinical practice (35)

There remains no clear evidence regarding the superiority of one salvage regimen over another in randomised studies (Table 4). For instance, the Phase III Collaborative Trial in Relapsed Aggressive Lymphoma (CORAL) study, which compared the efficacy of R-ICE (rituximab with ifosfamide, carboplatin, etoposide) or R-DHAP (rituximab with cisplatin, cytarabine and dexamethasone) followed by ASCT with or without rituximab maintenance, demonstrated no difference in 2-year OS between salvage regimens, with only 50% of patients being able to proceed to ASCT (36).

Table 4: Salvage chemotherapy regimens in randomised studies for DLBCL

R-GDP, rituximab-gemcitabine, dexamethasone, cisplatin; DLBCL, diffuse large B cell lymphoma; O-DHAP, Ofatumumab- dexamethasone, cytarabine, cisplatin; PFS, progression-free survival; R-DHAP, rituximabdexamethasone, cytarabine, cisplatin; R-ICE, rituximab-ifosfamide, etoposide, carboplatin; RR, relative risk

Clinical experts confirmed to Roche that there is no standard of care in UK clinical practice

for patients who are considered fit for intensive salvage therapy (estimated to be 50–60% of all R/R DLBCL patients). Patients in the UK are typically treated with platinum-based

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 18 of 144

regimens, irrespective of whether they are clear or borderline candidates for ASCT, such as R-GDP, R-DHAP, R-ICE and R-ESHAP (rituximab with etoposide, methylprednisolone, cytarabine and cisplatin) (35). Clinical experts also reported that R-Gem-Ox (rituximab with gemcitabine and oxaliplatin), an option considered by NHS England for older patients, is not widely used in UK clinical practice although familiarity with the regimen may increase among treatment centres that are enrolling patients to the ARGO study (35, 48).

Intensive salvage chemotherapy is followed by a conditioning regimen, typically carmustine, etoposide, cytarabine and melphalan (BEAM) or lomustine, etoposide, cytarabine and melphalan (LEAM), although alternative, less toxic regimens e.g. carmustine, etoposide, cytarabine and cyclophosphamide (BEAC) and lomustine, cytarabine, cyclophosphamide and etoposide (LACE) may be used for older patients (>70 years of age) (9, 35, 43).

Salvage chemotherapy is an area of high unmet need given the poor rate and duration of response; only 30–40% will respond and proceed to ASCT (28, 46, 47). Furthermore, the outcome for patients who do not respond to salvage regimens is very poor, with a median OS for non-responding patients of only 4 months (49).

Treatment options for patients who fail salvage chemotherapy are limited to clinical trials of novel agents, if available, or an additional line of salvage chemotherapy for younger patients who are willing to receive another treatment (35). Novel therapies for R/R DLBCL are in development, including chimeric antigen receptor T-cell (CAR-T) therapies, which have demonstrated activity in DLBCL in single arm studies (50, 51). However, CAR-T therapies are only available to those patients who have had two or more prior lines of systemic therapy, have sufficient disease control to await the manufacturing times, and can tolerate the conditioning regimen (usually fludarabine/cyclophosphamide), treatment emergent cytokine-release syndrome and sometimes severe neurotoxicities (52)

For patients who do respond to salvage chemotherapy, ASCT offers a second chance of cure. However, the overall benefit of ASCT as an option is limited by the fact that a substantial proportion of patients will be deemed ineligible or will relapse after ASCT. The CORAL study, conducted in the pre-PET era, demonstrated a 3-year event free survival of just 21% for patients in the prior rituximab-treated group (28). A more recent study evaluating the prognostic value of PET prior to ASCT demonstrated improved long term outcomes for patients achieving a complete metabolic response to salvage therapy by contemporary Deauville scoring (DS) (3-year PFS 77% for DS 1-3 vs 49% for DS 4) (53). Nevertheless, the prognosis of those patients who relapse after ASCT is poor (median survival of approximately 8 months among patients who relapse within 12 months of transplant (54)) with very little consensus on the optimal subsequent therapy. Although

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 19 of 144

allogenic stem-cell transplant is an option for some patients (33), it is rarely used in UK clinical practice and is associated with treatment-related mortality and limited disease control (55-57).

Treatment of transplant-ineligible patients

A substantial proportion of patients are not eligible for intensive therapy followed by ASCT due to age, comorbidities or chemotherapy-insensitive disease – UK clinical experts estimate this to be 40–50% of all R/R DLBCL patients (35). The treatment approach is palliative for such patients in the second or subsequent line setting, although there is still a goal of improving survival, albeit not necessarily with curative intent.

There are no universally established therapies for patients with R/R DLBCL who are ineligible for transplant or who relapse after transplant. There is a considerable amount of variability on the selected regimen for these patients with bendamustine with rituximab and gemcitabine and/or platinum-based therapies (such as oxaliplatin) among the most commonly used regimens. However, outcomes of such transplant-ineligible patients (including patients who relapse after ASCT) remain poor, with median OS of approximately 6 months (24, 58) (Table 5). Furthermore, there is no evidence of one chemotherapy regimen demonstrating superiority over another. The combination of bendamustine with rituximab (BR) has been shown to be active in transplant-ineligible patients with R/R DLBCL with a manageable haematological toxicity profile; median PFS has been reported to be 3.5–6.7 months and median OS reported to be 6.7–9.5 months (59-62). It should be noted however that direct comparison with prior studies investigating chemotherapy-based regimens has several severe limitations; namely different inclusion/exclusion criteria (i.e., limitations on prior lines of therapy, refractoriness) as well as historical context (e.g., how many patients had prior exposure to rituximab, differences in assessing response or what the first-line therapy was), leading to potentially significant differences in prognostic factors between different trial cohorts.

Table 5: Selected regimens for transplant-ineligible R/R DLBCL patients

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 20 of 144

CD4, cluster of differentiation 4; CR, complete response; FFS, failure-free survival; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; PR, partial response; TTP, time-to-progression

Treatments beyond second-line are further limited and include gemcitabine, bendamustine and palliative oral combinations (66). Pixantrone monotherapy is recommended by NICE as a third- or fourth-line treatment option for adult patients with R/R DLBCL (TA306) (67). However, UK clinical experts confirmed that pixantrone is not widely used in the UK compared with the rest of Europe (an observation corroborated by the exclusion of pixantrone as a treatment option for patients with R/R DLBCL in the BSH guidelines (33)), with real-world data reporting disappointing efficacy (median OS 3.4 months) (66). Furthermore, a Phase III trial (PIX306) investigating the efficacy or R+pixantrone compared with R-gemcitabine failed to demonstrate superiority in terms of progression free survival (PFS) in patients with R/R DLBCL (68).

(50, 51)The novel CAR-T cell therapies represent an additional treatment option for R/R DLBCL patients (52)who have had two or more prior lines of systemic therapy. However, in

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 21 of 144

practice many patients with progressive DLBCL have a rapid clinical disease course rendering them unsuitable for CAR-T therapy. Importantly also, due to the complex manufacturing and distribution and the need for intense monitoring, this treatment modality is currently limited to specialised tertiary centres and not available to the broad population. Emerging real world evidence will continue to inform the safety and efficacy profile of these new treatment options.

NICE currently recommends two CAR-T therapies. Axicabtagene ciloleucel is recommended for use in the Cancer Drugs Fund (CDF) as an option for adult patients with R/R DLBCL or primary mediastinal large B-cell lymphoma who have previously received two or more systemic therapies (TA559) (69), although this will only be initially available for 200 patients per year in eight specialised centres that are able to administer it. Tisagenlecleucel is also recommended for use in the CDF as an option for treating R/R DLBCL in adults after 2 or more systemic therapies (TA567) (70).

Overall, the outcome for this large group of R/R DLBCL patients who are ineligible for ASCT is poor; patients tend to be older therefore conventional salvage regimens offer little benefit in disease control and have substantial morbidity (71). There are no established therapies for patients with R/R DLBCL who are ineligible for transplant or who relapse after transplant, therefore there is a significant need for new and more effective treatments that extend survival with at least acceptable, if not superior, safety and tolerability profiles for these patients.

B.1.3.3 Proposed position of polatuzumab vedotin in the treatment pathway

The proposed treatment pathway and position of pola in combination with bendamustine and rituximab (pola+BR) is summarised below in Figure 2. In summary, the following patients will be considered eligible for pola+BR:

• R/R patients who are clear non-candidates for transplant (unfit for intensive therapy based on physician assessment), either as second-line treatment or as a third-line treatment and beyond for patients who have relapsed following or are refractory to their last-line of therapy

• R/R patients who would be candidates for transplant but fail to respond to salvage therapy (and are therefore transplant ineligible)

• R/R patients who receive salvage therapy and ASCT but subsequently relapse

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 22 of 144

Figure 2: Proposed positioning of pola+BR in DLBCL treatment pathway

==> picture [401 x 229] intentionally omitted <==

Evidence for the efficacy of pola+BR in UK clinical practice is sourced from the GO29365 study, in which patients with R/R DLBCL were enrolled (NCT02257567) (72). Patients enrolled must have either relapsed or have been refractory to a prior regimen for DLBCL and were ineligible for stem cell transplant (as assessed by the physician). Sixteen patients with R/R DLBCL enrolled in GO29365 were refractory to or relapsed after prior transplant.

B.1.4 Equality considerations

No equality issues related to the use of pola+BR have been identified.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 23 of 144

B.2 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 6: Clinical effectiveness evidence

DLBCL, diffuse large B cell lymphoma; ECOG PS, Eastern Cooperative Oncology Group performance status; pola+BR, polatuzumab vedotin plus bendamustine and rituximab; R/R, relapsed/refractory

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 24 of 144

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

Unless otherwise stated, information on the GO29365 study was sourced from the interim clinical study report and protocol (79, 80).

B.2.3.1 Study design

GO29365 is a Phase Ib/II, multicentre, open-label study of pola in combination with BR in patients with R/R DLBCL, and pola in combination with bendamustine and obinutuzumab (BG) in patients with R/R follicular lymphoma. This submission will focus on the

combination of pola+BR in patients with R/R DLBCL only in accordance with the

proposed marketing authorisation indication. The study was conducted in accordance with the principles of the “Declaration of Helsinki” and Good Clinical Practice (GCP) according to the regulations and procedures described in the following sections of the protocol.

The R/R DLBCL component of the study consisted of two stages that ran sequentially (Figure 3):

• Phase Ib safety run-in stage: to determine the safety, tolerability, and PK of pola+BR and to identify the recommended Phase II dose (RP2D) of pola to be used in the Phase II stage

• Phase II randomised and expansion stage : to evaluate the efficacy, further assess the safety and tolerability, and to characterise the PK of pola+BR.

Figure 3: GO29365 study design schema (R/R DLBCL pola and BR populations only)

==> picture [290 x 206] intentionally omitted <==

*1:1 randomisation, stratified by DOR ≤12 months or >12 months

Lyo, lyophilised formulation

Treatment administered every 21 days x 6 cycles: pola 1.8 mg/kg, C1D2, then D1 for C2+; bendamustine: 90 mg/m2, C1D2/3 then D1/2 for C2+; rituximab: 375 mg/m2, D1 for C1+

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 25 of 144

During the Phase Ib safety run-in, six patients were treated with pola+BR and monitored for adverse events (AEs) during a safety observation period corresponding to one treatment cycle (from Cycle 1 Day 1 to Cycle 2 Day 1 for a minimum of 21 days). An Internal Monitoring Committee (IMC) performed a safety analysis after the six patients had completed the safety observation period and provided a recommendation on whether to continue into Phase II, and on the RP2D for pola+BR regimen to be used.

For the Phase II randomised portion, patients were randomised to either pola+BR (investigative arm) or to BR alone (control arm).

All patients had tumour assessments including 18F-fluorodeoxyglucose-positron emission tomography (FDG-PET) and a diagnostic-quality CT scan with both oral and IV contrast (referred to as PET-CT) at screening and at an interim response assessment (between Cycle 3 Day 15 and Cycle 4 Day 1), and at primary response assessment: 6–8 weeks after completion of study treatment (i.e., Day 1 of Cycle 6 or after last dose of study medication).

The primary objective of the Phase II portion of the study was to evaluate the efficacy of pola+BR compared with BR alone in patients with R/R DLBCL as measured by PET-defined CR rate using modified Lugano 2014 response criteria (PET-CT criteria) (81) at the primary response.

Data from the Phase Ib and the randomised Phase II portion of GO29365 was generated with a liquid formulation of pola; however, a lyophilised formulation of pola suitable for commercialisation and use in ongoing and future clinical studies was subsequently developed. In late 2017, the protocol was amended to add a new formulation (NF) cohort (Arm G [N=42]), which was designed primarily to assess pharmacokinetic and safety of the lyophilised formulation of pola in combination with BR in R/R DLBCL. Efficacy was evaluated as a secondary objective; xxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxx xxxxxxxxxx xxxxxx xxxx xxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxx xxxxxxxxx xxxxxxx xxxxxxxx xxxx xxxxx xxxxxx xxx xxxxx xxxxx xxxxxxx xxxxxxx xxxxxxx x. In October 2018, another arm was added to the NF cohort (Arm H) recruiting an additional 60 R/R DLBCL patients using the lyophilised formulation of pola in combination with BR. Xxxxxxxxxx xxxxxxxxxxxx xxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxx xxxxxxxxx xxxxxxxxx xxxxxxx xxxxxxxxxx xxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Results reported in this submission are from patients treated with the liquid formulation of polatuzumab vedotin; however, this is not anticipated to be different from that seen with the lyophilised formulation, as reflected by preliminary safety and PK data that has been submitted to EMA. Furthermore, the FDA and EMA have allowed Hoffmann-La Roche to file for marketing authorisation based on results from the liquid formulation.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 26 of 144

The BR regimen, chosen to be combined with pola as the investigative treatment, and as the comparator in the randomised Phase II portion of this study, has demonstrated clinical activity in transplant ineligible patients with R/R DLBCL and is associated with manageable haematologic toxicity (59-62). The bendamustine backbone was also selected to minimise the overlapping toxicity of peripheral neuropathy (PN) that may occur with platinum-based therapies.

The dose and schedule of bendamustine used in combination with rituximab in this study (90 mg/m[2] administered on two consecutive days for six 21-day cycles for patients with with R/R DLBCL) was consistent with the recommendations from an international consensus panel based on data in the relapsed setting at the time the study was initiated (82).

The dose of pola was limited to 1.8 mg/kg every 21 days for 6 cycles. Limiting the pola dose regimen for ≤8 cycles may enhance tolerability and mitigate the risk of PN compared with longer treatment durations and higher doses (83). Time-to-event modelling data suggest that 6–8 cycles of 1.8 mg/kg pola has a predicted incidence of grade ≥2 PN of 17.8–28.8%; this is comparable with other antimicrotubule agents for lymphoma treatment (84).

B.2.3.2 Summary of study methodology

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 27 of 144

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 28 of 144

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 29 of 144

• 10[9] /L (for thrombocytopenia) on or after Day 8 of the scheduled date for the next cycle. If prior bendamustine dose reduction had occurred, bendamustine dose could be further reduced to 50 mg/m[2] for recurrent Grade 3 or 4 neutropenia or thrombocytopenia. No more than two dose reductions of bendamustine were allowed.

• Pre-medications • All rituximab infusions were to be preceded by premedication with oral acetaminophen/paracetamol and an antihistamine 30–60 minutes before the start of each infusion (unless contraindicated) to minimise the risk of IRRs.

• Concomitant medications Permitted concomitant medications included: • Continued use of oral contraceptives, hormone-replacement therapy, or other maintenance therapies

• • Use of G-CSF for the treatment of neutropenia • Mandatory premedication with acetaminophen/paracetamol and antihistamine prior to administration of each rituximab infusion

• • Mandatory premedication with oral allopurinol or a suitable alternative treatment (with adequate hydration) prior to Cycle 1, Day 1 and subsequent cycles of treatment if deemed appropriate by the investigator for all patients with high tumour burden and considered to be at high risk for TLS

• • Anti-infective prophylaxis for viral, fungal, bacterial, or Pneumocystis infections

• • Necessary supportive measures for optimal medical care throughout study according to institutional standards, including growth factors (e.g., erythropoietin) and anti-emetic therapy, if clinically indicated

• Prohibited concomitant medications: • Cytotoxic chemotherapy, other than bendamustine and intrathecal chemotherapy for CNS prophylaxis

• • Immunotherapy or immunosuppressive therapy, other than study treatments

• • Radioimmunotherapy or radiotherapy • Hormone therapy, other than contraceptives, stable hormonereplacement therapy, or megestrol acetate

• • Biologic agents other than haematopoietic growth factors, which are allowed if clinically indicated and used in accordance with instructions provided in the package inserts

• • Any therapy (other than intrathecal CNS prophylaxis) intended for the treatment of lymphoma

• Primary outcome Primary endpoint: • PET-defined CR rate at the time of primary response assessment (6–8 weeks after Cycle 6 Day 1 or last dose of study medication) as defined by the IRC

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 30 of 144

ADA, anti-drug antibodies; ALT, alanine aminotransferase; ANC, absolute neutrophil count; AST, aspartate aminotransferase; BOR, best overall response; BR, bendamustine + rituximab; CNS, central nervous system; CR, complete response; DLBCL, diffuse large B-cell lymphoma; DOR, duration of response; ECOG PS, Eastern Cooperative Oncology Group performance status; EFS, event-free survival; G-CSF, granulocyte-colony stimulating factor; IRC, Independent Review Committee; IRR, infusion-related reaction; OR, overall response; OS, overall survival; PET-CT, positron emission tomography-computed tomography; PFS, progression-free survival; PN, peripheral neuropathy; Pola, polatuzumab vedotin; PR, partial response; (a)PTT, (activated) partial thromboplastin time; R/R relapsed/refractory; TINAS, Therapy-Induced Neuropathy Assessment Scale; TLS, tumour lysis syndrome; ULN, upper limit of normal

B.2.3.3 Patient demographics and baseline characteristics

Demographic characteristics were generally well balanced across cohorts of patients

between treatment arms. Any differences in incidence of demographic characteristics by category observed between BR and pola+BR treatment arms in the randomised Phase II was less than 10% (accounted for by four patients or fewer).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 31 of 144

In the randomised Phase II, a higher proportion of patients with R/R DLBCL in the BR arm, compared to the pola+BR arm had bulky disease (≥7.5 cm) (BR: 15/40 patients [37.5%] vs. pola+BR: 10/40 patients [25.0%]), ECOG PS 2 (8/40 [20.0%] vs 6/40 [15.0%], and IPI high risk (4 or 5 risk factors; 17/40 [42.5%] vs 9/40 [22.5%]).

The most common reasons for transplant ineligibility fell into two categories: patient characteristics such as age, comorbidity, or inadequate performance status (BR: 22/40 patients [55.0%]; pola+BR: 14/40 patients [35.0%]), and disease status, including inadequate response to salvage therapy or relapsing after prior autologous transplant (BR: 15/40 patients [37.5%]; pola+BR: 22/40 patients [55.0%]).

Table 7: GO29365 - key demographic and baseline disease characteristics

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 32 of 144

a Defined as no response or progression or relapse within 6 months of last anti-lymphoma therapy end date among patients whose last prior regimen contained anti−CD20

b Defined as no response or progression or relapse within 6 months of last anti−lymphoma therapy end date c Defined as time from end date of last anti−lymphoma therapy to first dose date

d Duration of response to prior therapy based on IxRS for randomised cohorts and CRF for non-randomised cohorts

ECOG PS, Eastern Cooperative Oncology Group performance status; IPI, international prognostic index; IxRS, Interactive Voice/Web Response System; SCT, stem cell transplant

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

Unless otherwise stated, information on the GO29365 study was sourced from the interim clinical study reports and protocol (79, 80). The participant flow and details on patient study and treatment withdrawal for GO29365 is presented in Appendix D.

Determination of sample size

In total, enrolment of approximately 224 patients was planned in order to evaluate the safety and efficacy of pola when combined with BR or BG in DLBCL and FL:

• Twenty-four patients in total were planned to be enrolled during the Phase Ib safety run-in portion of the study, with a minimum of six patients for the pola+BR DLBCL run-in. Less than two observed safety events in a given 6-patient cohort was considered to be deemed safe for the purpose of moving to the Phase II part of the study; this is consistent with requirements for identification of a candidate RP2D based upon a standard 3+3 design

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 33 of 144

• Forty patients were planned for each treatment arm in the Phase II randomisation phase in patients with R/R DLBCL in order to evaluate the safety and efficacy of pola+BR compared with BR with acceptable accuracy.

The primary analysis was an estimation of treatment-specific CR rates as well as the difference in PET CR rates between patients randomised to treatment with pola+BR and those randomised to treatment with BR alone.

With 40 patients per arm, 95% exact Clopper-Pearson confidence intervals (CIs) for estimation of the true CR rate for would have a margin of error not exceeding ±17%. With 40 patients and an observed CR rate of at least 60%, a true CR rate below 43% can be ruled out with 95% confidence (Table 8). In addition, with 40 patients in each arm, assuming a 40% CR rate in the BR arm, and a 25% increase in CR rate when pola is added to BR, the 95% CI for the difference in CR rates is 3.8%, 46.2%.

Table 8: Clopper-Pearson exact 95% confidence intervals for assumed observed CR rates based on sample size of 40 patients

CI, confidence interval; CR, complete response

With respect to assessment of safety based on a sample size of 40 patients in each of the BR arms, there is at least an 87% chance of observing at least one AE with a true incidence of ≥5%.

Analysis populations

Efficacy analyses for the randomised component of the Phase II (BR and pola+BR) were based on the intent-to-treat (ITT) population and conducted in accordance with the ITT principle (i.e., including all randomised patients irrespective of whether they received study treatment, with patients grouped according to treatment assignment at randomisation).

Safety analyses were based on the safety-evaluable population which included all treated patients (i.e., patients who received any amount of study medication) according to actual treatment received.

Efficacy analysis

Analysis methodology for primary, secondary and exploratory efficacy endpoints in the GO29365 study is summarised below.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 34 of 144

Table 9: Efficacy outcome measures and analysis methodology

BOR, best overall response; CR, complete response; DOR, duration of response; EFS, event-free survival; IRC, Independent Review Committee; KM, Kaplan-Meier; OR, overall response; OS, overall survival; PET-CT, positron emission tomography-computed tomography; PFS, progression-free survival; PR, partial response

Handling of missing data and censoring methods

For response endpoints, patients with no response assessments (for any reason) were considered non-responders.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 35 of 144

For the PFS analyses, patients who did not have documented disease progression or death had observations censored on the date of the last tumour assessment or, if no tumour assessments were performed after the baseline visit, at the time of randomisation and enrolment +1 day.

For OS, patients for whom death had not been documented had observations censored on the last date at which they were known to be alive.

Patient-reported outcome analysis

The PRO analyses included patients in the intent-to-treat population and were analysed according to assigned treatment. For the total score and each of the Therapy-Induced Neuropathy Assessment Scale (TINAS) single symptom items, descriptive statistics for recorded values at each visit and changes from baseline were calculated.

In the event of patients not completing individual TINAS items, missing data were handled per developer scoring instructions, such that a prorated total score was calculated if ≥50% of items were answered using the following formula:

Prorated total score = [Sum of item scores] x [Total no. of items] / [No. of items answered]

B.2.5 Quality assessment of the relevant clinical effectiveness evidence

Critical appraisal of the included randomised clinical trials was performed using established risk of bias tools recommended for HTA submissions. The complete quality assessment is presented in Appendix D. A summary is presented below.

Table 10: Clinical effectiveness evidence quality assessment

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 36 of 144

B.2.6 Clinical effectiveness results of the relevant trials

• Pola+BR resulted in higher response rates and longer DOR, PFS, EFS, and OS compared to BR in the randomised Phase II portion of GO29365. Efficacy results in patients with R/R DLBCL (CCOD 30 April 2018) can be summarised as follows:

• The primary efficacy endpoint of CR rate at the primary response assessment (PRA) based on PET-CT, as determined by the IRC, was higher in the pola+BR arm (40.0% [16/40 patients]; 95% CI: 24.9%, 56.7%) compared with the BR arm (17.5% [7/40 patients]; 95% CI: 7.3%, 32.8%) (Δ22.5% in favour of pola+BR; 95% CI: 2.6%, 40.2%; p=0.0261).

• Secondary efficacy endpoints of response rates (CR and objective response [OR; CR or PR]) whether measured with or without PET and assessed by the investigator or by the IRC remained consistent with primary efficacy results, with a higher proportion of patients with R/R DLBCL achieving CR or OR in the pola+BR arm compared to the BR arm o INV-assessed CR (PET-CT): 42.5% vs. 15.0%

• IRC-assessed OR (PET-CT): 45.0% vs. 17.5%

• o IRC-assessed response rates (CT only): CR, 22.5% vs. 2.5%; OR, 42.5% vs. 15.0% o INV-assessed response rates (CT only): CR: 20% vs. 5.0%; OR, 45.0% vs. 15.0% o INV-assessed BOR (PET-CT or CT): CR: 57.5% vs. 20.0%; ORR: 70.0% vs. 32.5% o IRC-assessed median DOR 12.6 months (95% CI: 7.2, NE) vs. 7.7 months (95% CI: 4.0, 18.9) (stratified HR=0.47; 95% CI: 0.19, 1.14]; p=0.0889)

• o IRC-assessed median PFS: 9.5 months (95% CI: 6.2, 13.9) vs. 3.7 months (95% CI: 2.1, 4.5) (stratified HR=0.36; 95% CI: 0.21, 0.63; p=0.0004)

• • Exploratory time-to-event analyses demonstrated a consistent treatment effect favouring the pola+BR arm compared to the BR arm for DOR, PFS, EFS, and OS: o INV-assessed median DOR: 10.3 months (95% CI: 5.6, NE) vs. 4.1 months (95% CI: 2.6, 12.7) (stratified HR=0.44; 95% CI: 0.20, 0.95)

• o INV-assessed median PFS: 7.6 months (95% CI: 6.0, 17.0) vs. 2.0 months 95% CI: 1.5, 3.7) (stratified HR=0.34; 95% CI: 0.20, 0.57; p<0.0001). The updated INV-assessed PFS analysis (CCOD 11 October 2018) was consistent with that seen in the interim analysis: xxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxx xxxxxx xxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxx

• o Median EFS: 6.4 months (95% CI: 4.0, 11.1) vs. 2.0 months (95% CI: 1.5, 3.1) o Median OS was extended to 12.4 months (95% CI: 9.0, NE) in the pola+BR arm, from 4.7 months (95% CI: 3.7, 8.3) in the BR arm (stratified HR=0.42; 95% CI: 0.24, 0.75; p=0.0023. xxxxxxxxxxxx xxxxxxxxxxx xxxxxxxxxxx xxxxxxxxxx xxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxx xxxxxxxxxxx

• o The treatment effect for survival was consistently observed across all subgroups of patients with R/R DLBCL tested.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 37 of 144

The primary analysis presented in this submission is from the clinical cut-off date 30 April 2018, although data for PFS and OS from the most recent data cut (11 October 2018) is also presented.

B.2.6.1 Primary efficacy endpoint

In the randomised Phase II part of the GO29365 study, the proportion of patients with R/R DLBCL with a CR at the primary response assessment by PET-CT as assessed by the IRC was higher in the pola+BR arm (40.0% [16/40 patients]; 95% CI: 24.9%, 56.7%) compared to patients in the BR arm (17.5% [7/40 patients], 95% CI: 7.3%, 32.8%). The difference in CR rates between arms was statistically significant (Δ22.5%; p=0.0261, CMH chi-square).

Table 11: CR rate with PET at primary response assessment (IRC-assessed)

CCOD: 30 April 2018

B.2.6.2 Secondary efficacy endpoints

Response rates at primary response assessment based on PET

The objective response (CR or PR) rate at the primary response assessment based on PET by IRC was 45.0% (18/40 patients) in the pola+BR arm and 17.5% (7/40 patients) in the BR arm, a difference of 27.5% (p=0.0069, CMH chi-square test). Most patients who had an objective response at the primary response assessment in each treatment arm achieved a CR (BR arm: 7 CRs, 0 PRs vs pola+BR arm: 16 CRs, 2 PRs).

Table 12: Objective response (CR/PR) rates by PET at primary response assessment (IRC-assessed)

CR, complete response; OR, overall response; PR, partial response CCOD: 30 April 2018

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 38 of 144

INV-assessment of response by PET at the primary response assessment was highly consistent with IRC assessment. Complete response rates were 42.5% (17/40 patients) in the pola+BR arm and 15.0% (6/40 patients) in the BR arm, a difference of 27.5% (p=0.0061, CMH chi-square test). Objective response rates, driven by CRs, were 47.5% (19/40 patients) and 17.5% (7/40 patients), respectively (Δ30.0%; p=0.0036, CMH chi-square test).

Table 13: Complete response and objective response (CR/PR) rates by PET at primary response assessment (INV-assessed)

CR, complete response; OR, overall response; PR, partial response CCOD: 30 April 2018

Response rates at primary response assessment based on CT

Complete response rates for patients with R/R DLBCL, measured by CT and as assessed by IRC, were 22.5% (9/40 patients) in the pola+BR arm and 2.5% (1/40 patients) in the BR arm (Δ20.0%; p= 0.0078, CMH chi-square test).

Overall response rates were 42.5% and 15.0%, respectively (Δ27.5%; p=0.0051, CMH chisquare test) and compared with objective responses by PET, were made up of proportionately more partial responses (BR arm: 1 CR, 5 PRs vs. pola+BR arm: 9 CRs, 8 PRs).

Table 14: Complete response and objective response (CR/PR) rates by CT at primary response assessment (IRC-assessed)

CR, complete response; OR, overall response; PR, partial response CCOD: 30 April 2018

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 39 of 144

Investigator assessment of response by CT at primary response assessment was similar to the IRC assessment. The proportions of patients with CR and OR were 20.0% (8/40 patients) in the pola+BR arm versus 5.0% (2/40 patients) in the BR arm (Δ15.0%; p=0.0454, CMH chi-square test), and 45.0% (18/40 patients) in the pola+BR arm versus 15.0% (6/40 patients) in the BR arm (Δ30.0%; p= 0.0032, CMH chi-square test) respectively.

Table 15: Complete response and objective response (CR/PR) rates by CT at primary response assessment (INV-assessed)

CR, complete response; OR, overall response; PR, partial response CCOD: 30 April 2018

Best overall response while on study

For analyses of BOR, the tumour assessment result was based on PET-CT or CT results. A best response of CR or PR as assessed by IRC was achieved by 25/40 patients (62.5%) in the pola+BR arm and 10/40 patients (25.0%) in the BR arm while on study. The proportion achieving a best response of CR was 20/40 patients (50.0%) and 9/40 patients (22.5%), respectively.

Table 16: Best overall response rate (IRC-assessed)

BOR, best overall response CCOD: 30 April 2018

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 40 of 144

A best response of CR or PR as assessed by the investigator was achieved by 28/40 patients (70.0%) in the pola+BR arm and 13/40 patients (32.5%) in the BR arm while on study. The proportion achieving a best response of CR was 23/40 patients (57.5%) and 8/40 patients (20.0%), respectively.

Table 17: Best overall response rate (INV-assessed)

BOR, best overall response CCOD: 30 April 2018

Duration of response by IRC

Among patients who achieved an OR (CR/PR) at any time (23 patients in the pola+BR arm and 10 patients in the BR arm), 13 patients in the pola+BR arm (52.0% of responders) and eight patients in the BR arm (80.0% of responders) subsequently had progressive disease or died.

The median IRC-assessed DOR was 12.6 months in the pola+BR arm (95% CI: 7.2 months, NE]) compared to the BR arm (7.7 months [95% CI: 4.0 months, 18.9 months]). The risk of responders progressing or dying was reduced by 53% in patients treated with pola+BR compared to BR (stratified hazard ratio [HR]=0.47; 95% CI: 0.19, 1.14).

Table 18: Duration of response (CR/PR) (IRC-assessed)

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 41 of 144

Stratified HR % 0.47 (95% CI) (0.19, 1.14) p value (log-rank) p=0.0889

Tumour assessment is based on PET-CT wherever it is available and valid and uses CT only result if PET-CT is missing HR, hazard ratio CCOD: 30 April 2018

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxx

Progression-free survival by IRC

More patients with R/R DLBCL in the BR arm compared with the pola+BR arm had a PFS event (PD or death) at the time of the clinical cut-off (80.0% [32/40 patients] vs. 62.5% [25/40 patients]). Deaths accounted for most of the PFS events (19/32 patients) in the BR arm and 13 patients in the BR arm had disease progression. PFS events in the pola+BR arm included 11 deaths and 14 patients with disease progression.

The risk of PD or death was reduced by 64% in patients treated with pola+BR compared to BR (stratified HR=0.36; 95% CI: 0.21, 0.63; p<0.0002). Median PFS was over two-fold the duration in patients treated with pola+BR (9.5 months [95% CI: 6.2, 13.9]) compared to BR (3.7 months [95% CI: 2.1, 4.5]).

Table 19: Progression-free survival (IRC-assessed)

Tumour assessment is based on PET-CT wherever it is available and valid and uses CT only result if PET-CT is missing HR, hazard ratio CCOD: 30 April 2018

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 42 of 144

See section B.2.6.4 for sensitivity analyses of PFS by IRC based on two additional censoring rules that were applied.

B.2.6.3 Exploratory efficacy endpoints

Duration of response by investigator

Among patients who achieved an OR (CR/PR) at any time (28 patients in the pola+BR arm and 13 patients in the BR arm), 17 patients in the pola+BR arm (60.7% of responders) and 11 patients in the BR arm (84.6% of responders) subsequently had progressive disease or died.

The median DOR was over two-fold the duration in the pola+BR arm (10.3 months [95% CI: 5.6 months, NE]) compared to the BR arm (4.1 months [95% CI: 2.6 months, 12.7 months]). The risk of responders progressing or dying after response to treatment was reduced by 56% in patients treated with pola+BR compared to BR (stratified HR=0.44; 95% CI: 0.2, 0.95).

Table 20: Duration of response (INV-assessed)

Tumour assessment based on PET-CT wherever available and valid, CT only if PET-CT is missing HR, hazard ratio, NE, not estimated CCOD: 30 April 2018

Progression-free survival by investigator

More patients with R/R DLBCL in the BR arm, compared to the pola+BR arm of the randomised Phase II, had a PFS event (PD or death) at the time of the clinical cut-off (87.5% [35/40 patients] vs. 67.5% [27/40 patients]). The risk of PD or death was reduced by 66% in patients treated with pola+BR compared to BR (stratified HR=0.34 (95% CI: 0.20, 0.57); p<0.0001).

Median PFS was over three-fold the duration in patients treated with pola+BR (7.6 months [95% CI: 6.0, 17.0]) compared to BR (2 months [95% CI: 1.5, 3.7]).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 43 of 144

Table 21: Progression-free survival (INV-assessed)

Tumour assessment based on PET-CT wherever available and valid, CT only if PET-CT is missing HR, hazard ratio CCOD: 30 April 2018

The updated efficacy analysis (CCOD 11 October 2018) provide approximately 5 months of additional data, with an estimated median follow-up of xxxxxxxxxxx.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxx

Table 22: INV-assessed progression-free survival (updated analysis)

Tumour assessment based on PET-CT wherever available and valid, CT only if PET-CT is missing HR, hazard ratio CCOD: 11 October 2018

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 44 of 144

Figure 4: Kaplan-Meier plot of INV-assessed PFS (updated analysis)

==> picture [443 x 245] intentionally omitted <==

CCOD 11 October 2018

Event-free survival by investigator

As with PFS, more patients with R/R DLBCL in the BR arm, compared to the pola+BR arm of the randomised Phase II had an EFS event (new anti-lymphoma treatment [NALT], PD or death) at the time of the clinical cutoff (95.0% [38/40] vs. 72.5% [29/40]). The risk of PD, death or starting a new antilymphoma treatment was reduced by 70% in patients treated with pola+BR compared to BR (stratified HR=0.30; 95% CI: 0.18, 0.50; p<0.0001).

Table 23: Event-free survival (INV-assessed)

Tumour assessment based on PET-CT wherever available and valid, CT only if PET-CT is missing HR, hazard ratio CCOD: 30 April 2018

Overall survival

At the time of the clinical cutoff, a total of 51 patients with R/R DLBCL in the randomised Phase II had died; 28 patients (70.1%) in the BR arm and 23 patients (57.5%) in the

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 45 of 144

pola+BR arm. The main cause of death in both arms was disease progression (17 patients in the BR arm and 14 patients in the pola+BR arm). A further two patients with R/R DLBCL in the pola+BR safety run-in died of progressive disease.

The risk of death was reduced by 58% in patients treated with pola+BR compared to BR (stratified HR=0.42; 95% CI: 0.24, 0.75; p=0.0023). Median overall survival was 12.4 months (95% CI: 9.0, NE) in patients in the pola+BR arm compared to 4.7 months (95% CI: 3.7, 8.3) in the BR arm.

The duration of survival follow-up for patients with R/R DLBCL as assessed by reverse KM methodology was same in the two treatment arms (median follow-up 22.3 months in both treatment arms).

Table 24: Overall survival

HR, hazard ratio; NE, not estimated CCOD: 30 April 2018

At the time of the updated CCOD (11 October 2018),

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxx

Table 25: Overall survival (updated analysis)

==> picture [452 x 110] intentionally omitted <==

----- Start of picture text -----
pola+BR BR
n=40 n=40
Patients with event, n (%) xxxxxxxxx xxxxxxxxx
Median time to event, months xxxxx xxxx
95% CI xxxxxxxxxxxxx xxxxxxxxxxxx
Stratified HR % xxxx
(95% CI) xxxxxxxxxxxx
p value (log-rank) xxxxxxx
----- End of picture text -----

HR, hazard ratio; NE, not estimated CCOD: 11 October 2018

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 46 of 144

Figure 5: Kaplan-Meier plot of overall survival (updated analysis)

==> picture [443 x 245] intentionally omitted <==

CCOD 11 October 2018

B.2.6.4 PFS sensitivity analyses

To assess the robustness of the results of PFS by IRC in the randomised Phase II portion, two additional censoring rules were applied. For the patients who had missed one or more assessments before their recorded event of disease progression or death, the data were censored at the date of the last non-missing disease assessments prior to the events. For patients who started NALT prior to the disease progression, the data were censored at the date of the last non-missing diease assessments before the NALT. The estimates of the PFS by IRC using these additional censoring rules are consistent with the results of using the standard PFS censoring rule.

Table 26: Progression-free survival (IRC-assessed) – censoring for one or more missing responses

Tumour assessment is based on PET-CT wherever it is available and valid and uses CT only result if PET-CT is missing HR, hazard ratio, NE, not estimated CCOD: 30 April 2018

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 47 of 144

Table 27: Progression-free survival (IRC-assessed) – censoring for NALT

Tumour assessment is based on PET-CT if available and valid and uses CT only result if PET-CT is missing HR, hazard ratio, NE, not estimated CCOD: 30 April 2018

B.2.6.5 Multiple Cox-regression

To further assess the robustness of the treatment effects observed on pola+BR compared to BR in the randomised Phase II portion, multiple Cox-regression analyses were conducted for PFS and OS. The prognostic factors included in the final models were selected based on the statistical threshold (p<0.2) and clinical consideration while controlling for the multicollinearity among the factors.

Two sets of factors were used for PFS and OS:

• PFS: Ann Arbor stage, baseline ECOG and IPI

• OS: Ann Arbor stage, baseline ECOG, bulky disease and IPI

After adjusting for the potential prognostic factors and baseline characteristics the treatment effect of pola+BR remains robust:

• For investigator-assessed PFS, the adjusted HR is between 0.34 (95% CI: 0.20, 0.58; p<0.0001) and 0.38 (95% CI: 0.22, 0.64; p=0.0003)

• For IRC-assessed PFS, the adjusted HR is between 0.37 (95% CI: 0.21, 0.66; p=0.0009) and 0.40 (95% CI: 0.23, 0.70; p=0.0012)

• For OS, the adjusted HR is between 0.43 (95% CI: 0.24, 0.78; p=0.005) and 0.46 (95% CI: 0.26, 0.82; p= 0.008)

Table 28: Mulitple Cox regression models for PFS

*stratification factor: duration of response ≤12 months;[†] two patients missed baseline ECOG CCOD: 30 April 2018

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 48 of 144

Table 29: Mulitple Cox regression models for OS

*stratification factor: duration of response ≤12 months †two patients missed baseline ECOG CCOD: 30 April 2018

B.2.6.6 Patient-reported outcomes

The severity of symptoms of neuropathy related to pola treatment and impact on daily functioning were self-reported by the patient on the TINAS v1.0 instrument (85). The TINAS is an 11-item questionnaire that assesses the severity of neuropathy-related symptoms in the last 24 hours. Each item is scored on a 0–10 scale, with 0 being the symptom is not present, and 10 being the symptom is as bad as the patient can imagine. Responses were entered electronically by the patient either while on-site or at home, using the patient’s own electronic device or a device provided to them for the purpose of this study.

Patient-reported outcome completion over the course of the trial was generally limited for both Phase Ib and Phase II parts of the study; the proportion of R/R DLBCL patients completing at least one item of the TINAS at each weekly assessment was typically not greater than 50%. During the Phase II randomisation phase, compliance in the BR arm declined more rapidly than the pola+BR arm.

As PN in the study is assumed to be specific to pola, data from pola+BG and pola+BR arms were pooled across the Phase Ib and Phase II stages to maximise the sample size available for analyses given the extent of missing data.

Mean scores for individual TINAS items were low (≤1.5) at the beginning of treatment in both the pola+BR/BG and BR arms, indicating that patients were experiencing relatively little PN burden (range is 0–10 for each item), and generally stayed low during treatment. By the end of treatment, the severity of PN symptoms was rated as low, with the highest means observed for numbness/tingling in hands/feet, although still ≤2.

When exploring mean TINAS scores over time, several sensory (i.e., hot/burning sensations in hands/feet, pins/needles in arms/legs, numbness/tingling in hands/feet, cramps in hands/feet) and motor items (i.e., trouble grasping small objects, trouble walking, and difficulty with balance) showed slight elevations during treatment with pola+BR/BG, whereas

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 49 of 144

trajectories for the remaining items were largely flat. However, the mean scores rarely exceeded 3.0 during treatment, indicating that overall, patients perceived PN symptoms to be mild. Trends for the BR arm across all items were relatively flat.

B.2.7 Subgroup analysis

Exploratory subgroup analyses of PFS and OS in patients with R/R DLBCL evaluated the potential impact of demographic and baseline disease characteristics, e.g. cell of origin category, duration of response to prior therapy, disease stage, and other prognostic factors on the treatment effect (see Appendix E).

Overall, the treatment effect for PFS was consistent and in the same direction as for the overall R/R DLBCL population with point estimates of HR <0.50 across all except two subgroups, patients with Ann Arbor Stage I/II disease at study entry (HR=0.7 [95% CI: 0.1, 4.98]) and patients with prior transplant (HR=0.86 [95% CI: 0.26, 2.88]). However, the number of patients in each of these subgroups was very small (10 and 16 patients, respectively).

Exploratory subgroup analysis of OS in patients with R/R DLBCL by baseline risk factor shows that for all patient subgroups tested, the treatment effect for survival was consistent with and in the same direction (point estimates of HR were generally ≤0.50) as for the overall R/R DLBCL population.

B.2.8 Meta-analysis

A meta-analysis was not feasible as only one study was identified.

B.2.9 Indirect and mixed treatment comparisons

There is no universally accepted standard of care regimen for treating patients with R/R DLBCL who are not candidates for ASCT. The feasibility of an indirect treatment comparison of Pola+BR with comparators other than BR identified in the NICE scope was investigated based on the results of the systematic review (Appendix D).

However, no connected network of RCTs to other potential treatment options identified in the final scope could be established. As shown below only a limited number of studies were identified for the relevant population of R/R DLBCL patients ineligible for transplant, that did not form a connected network.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 50 of 144

Figure 6: Network of evidence, black circles – ASCT ineligible and white circles ASCT eligible

==> picture [302 x 250] intentionally omitted <==

In the review of single arm studies, only one study of R-GemOx was identified that included a group of patients that had received rituximab in a prior treatment line (rituximab pre-treated patients) (58). However, the study did not report KM data for rituximab pre-treated and naïve patients separately and it was therefore not feasible to conduct a robust match-adjusted treatment comparison.

B.2.9.1 Uncertainties in the indirect and mixed treatment comparisons

Not applicable.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 51 of 144

B.2.10 Adverse reactions

• Overall, the safety and tolerability profile of pola+BR was acceptable within the context of this pre-treated population of patients with R/R DLBCL; the nature and frequency of the observed AEs were consistent with the known safety profiles of pola and of BR and in line with what might be expected in this patient population

• No new safety signals were identified relative to the known safety profile of pola, bendamustine, or rituximab

• The duration of exposure to study treatment was longer (median 5 vs. 3 cycles received), cumulative exposure higher, and more patients with R/R DLBCL completed their planned number of treatment cycles (46.2% vs. 23.1%) in the pola+BR arm compared to the BR arm because of fewer early discontinuations due to disease progression

• The safety profile of pola+BR in patients with R/R DLBCL, in the context of the clinically significant benefit observed and longer treatment duration, was acceptable and overall was consistent with that observed in the BR arm:

  • Grade ≥3 AEs were reported at a higher overall incidence in patients treated with pola+BR than with BR (84.6% vs. 71.8%), the difference driven mainly by a higher incidence of Grade 3 or 4 cytopenias (neutropenia, thrombocytopenia and anaemia) in the pola+BR arm.

  • Fewer patients died due to disease progression in the pola+BR arm compared to the BR arm (14 patients vs. 17 patients)

  • The overall incidence of serious AEs was almost similar in the pola+BR and BR arms (64.1% vs. 61.5%)

  • In the setting of longer treatment exposure in the pola+BR arm, treatment discontinuations due to AEs were more frequent in the pola+BR arm compared to the BR arm (33.3% vs. 15.4%)

• Of selected AEs/AEs to monitor defined as identified and potential risks of pola: o Neutropenia, thrombocytopenia, anaemia (all grade and Grade ≥3) were more frequently reported in the pola+BR arm compared to the BR arm but these were manageable

• o The incidence of infections (all grade, Grade ≥3 and serious) was similar between pola+BR and BR arms; four cases of infection in the pola+BR arm and three cases of infection in the BR arm were fatal.

  • Peripheral neuropathy events were, as expected, more frequently reported in the pola+BR arm compared to the BR arm (43.6% vs. 7.7%, all Grade 1 or 2). A single patient discontinued pola due to muscle atrophy (Grade 1) and 2 patients had their pola dose reduced due to Grade 2 PN

The safety evaluable population in the GO29365 study excluded two patients (one in each

treatment arm) who did not receive any study treatment (84 patients in the DLBCL cohort; all six patients in the Phase Ib safety run-in and 78 patients in the Phase II randomisation

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 52 of 144

cohort). All patients who were administered at least one dose of study drug, received their intended treatment.

Overall, no new safety signals were noted with the addition of pola to BR relative to the known safety profile of Pola, and the safety and tolerability profile of the pola+BR regimen was acceptable within the context of this pre-treated population of patients with R/R DLBCL. An overview of the safety profile of pola+BR in GO29365 is summarised below.

Table 30: Overview of safety profile in GO29365

AE, adverse event; PD, progressive disease CCOD: 30 April 2018

Extent of exposure to study treatment

Planned treatment of patients with R/R DLBCL given pola+BR or BR consisted of 6x21-day cycles during which a total of 6 doses of pola (1.8 mg/kg), 12 doses of bendamustine (90 mg/m[2] ), and 6 doses of rituximab (375 mg/m[2] ) were to be administered.

Patients with R/R DLBCL receiving pola+BR completed a median of 5 cycles (46.2% completing all 6 treatment cycles). Median treatment duration for each individual component of study treatment was approximately 2.42 months. Patients receiving bendamustine and rituximab completed a median of 3 cycles for both (23.1% completing all cycles of treatment).

The extent of exposure of patients to bendamustine and rituximab in the comparator BR arm was less than for patients in the pola+BR arm, largely due to a higher rate of early treatment

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 53 of 144

discontinuations in patients in the BR arm most frequently due to progressive disease. Patients in the BR arm received a median of 3 cycles of treatment; median treatment duration was 1.39 months.

Table 31: Exposure to polatuzumab vedotin, bendamustine and rituximab

*Adjusted for dose modification and delay NE, not estimated; SD, standard deviation CCOD: 30 April 2018

Common adverse events

All patients with R/R DLBCL treated with pola had at least one AE during the study. In the randomised Phase II, AEs of any grade were reported in 100% of patients (39/39) in the pola+BR arm and 97.4% of patients (38/39) in the BR arm.

The AEs by preferred term with a >10% higher incidence in the pola+BR arm compared to the BR arm were neutropenia, thrombocytopenia, anaemia, lymphopenia, diarrhoea, vomiting, upper abdominal pain, pyrexia, chills, pneumonia, hypokalaemia, hypoalbuminaemia, peripheral neuropathy, dizziness, peripheral sensory neuropathy, and pruritus.

Table 32: Most frequently reported adverse events (>10%) with pola+BR

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 54 of 144

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 55 of 144

Table only shows preferred terms reported in >10% of all patients with R/R DLBCL treated with pola+BR (at least 5/45 patients in Phase Ib/II combined) CCOD: 30 April 2018

Adverse events by intensity

The majority of patients had at least one Grade ≥3 AEs. The proportion of patients with Grade 4 events was slightly higher in patients treated with pola+BR compared to the BR (37.8% vs. 25.6%).

A total of 202 Grade 3–5 AEs were reported in 38/45 patients (84.4%) with R/R DLBCL treated with pola+BR (Phase Ib/II combined). The most frequently reported Grade 3–5 AEs at a frequency of >5% of patients (in at least 3/45 patients with R/R DLBCL treated with pola+BR in Phase Ib/II combined) were neutropenia, thrombocytopenia, anaemia and lymphopenia.

Most frequently reported Grade 3–5 adverse events (>5%)

AE preferred terms reported in >5% of all patients with R/R DLBCL treated with pola+BR (at least 3 patients out of total of 45 in Phase Ib/II combined)

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 56 of 144

*All AEs shown in this table of most frequently reported Grade 3-5 AEs (with onset from first dose of study drug through 90 days after last dose of study drug) were Grade 3 or 4 except for two fatal (Grade 5) events of pneumonia (one event each in pola+BR arm and BR arm of randomised Phase II) CCOD: 30 April 2018

Serious adverse events

A total of 57 serious AEs (SAEs) were reported in 29/45 patients (64.4%) with R/R DLBCL treated with pola+BR (Phase Ib /II combined). The most common SAEs by preferred term were febrile neutropenia (five patients, 11.1%), pneumonia, and pyrexia (four patients each, 8.9%).

In the randomised Phase II, the overall incidence of SAEs was similar between the arms (64.1% [25/39 patients] pola+BR arm vs 61.5% [24/39 patients] BR). SAEs by preferred term or by SOC were generally balanced with no more than three patients in either arm accounting for any difference observed.

Deaths

A total of 25/46 patients (54.3%) with R/R DLBCL in the pola+BR arms had died at the time of the clinical cut-off. The main cause of death was disease progression (in 16/25 deaths [64.0%]).

In the randomised Phase II, fewer deaths overall had occurred in the pola+BR arm (23 deaths), compared to the BR arm (28 deaths); the difference can be accounted for by fewer deaths due to disease progression in the pola+BR arm (14 vs 17 patients) and fewer deaths due to AEs (9 vs 11 patients).

Three deaths due to AEs in the pola+BR arm (pneumonia, haemoptysis, and pulmonary oedema) and 4 deaths due to AEs in the BR arm (septic shock and pneumonia) occurred during the study treatment period, after first dose of study treatment and before treatment discontinuation or completion.

Table 33: Summary of deaths in GO29365

CCOD: 30 April 2018

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 57 of 144

Selected adverse events to monitor in R/R DLBCL

This section describes the results of the analyses of selected AEs, defined as identified risks and potential risks of pola based on data available at the time of the assessment. Most patients with R/R DLBCL treated with pola+BR had at least one selected AE (95.6% [43/45 patients]). A higher incidence of selected AEs were reported in the pola+BR arm compared to the BR arm of the randomised Phase II (97.4% vs. 87.2%).

No protocol-defined adverse events of special interest (potential drug-induced liver injury, suspected transmission of an infectious agent by the study drug, TLS of any grade or second malignancies) requiring expedited reporting, even if non-serious, was reported for any patient in the GO29365 study.

Table 34: Selected adverse events in patients with R/R DLBCL

IRRs, infusion-related reactions CCOD: 30 April 2018

Adverse events leading to treatment withdrawal or dose modification/interruption

Adverse events led to discontinuation of any study drug in 14/45 patients (31.3%) with R/R DLBCL treated with pola+BR (Phase Ib/II combined), compared with 6/39 patients (15.4%) treated with BR. In most cases (12/14) all drugs (Pola, bendamustine, and rituximab) were discontinued permanently at the same time (in two patients, bendamustine only was

discontinued).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 58 of 144

The most common events leading to discontinuation of any treatment were cytopenias, (8/14 discontinuations; predominantly Grade ≥3), mainly thrombocytopenia, neutropenia, and pancytopenia.

Adverse events that led to the discontinuation of pola in 12/45 patients (26.7%) with R/R DLBCL treated with pola+BR (Phase Ib/II combined); the most frequent AE by preferred term was thrombocytopenia (four patients), followed by neutropenia (two patients). All other AEs leading to pola withdrawal were unique events affecting single patients.

Drug interruption (dose delays/withholding of dose) was the most common action taken in response to AEs. Approximately one half of patients with R/R DLBCL had any drug interrupted (23/45 patients [51.1%]) due to an AE. The most frequent events leading to drug interruption were neutropenia and thrombocytopenia.

Table 35: Incidence of adverse events leading to discontinuation, drug interruption or dose modification of study drugs in patient with R/R DLBCL

CCOD: 30 April 2018

Anti-drug antibodies directed against Pola

For all R/R DLBCL patients treated with Pola, the baseline prevalence of ADAs was 4.8% (2/42 evaluable patients). Post-baseline, ADAs were detected in 3/41 evaluable R/R DLBCL patients (7.3%) treated with Pola. All three patients had treatment-induced ADAs (i.e., ADAnegative at baseline or missing a baseline sample for ADA analysis and at least one positive

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 59 of 144

post-baseline ADA result); however, the emergence of ADAs to pola did not appear to impact efficacy; two of these patients had ongoing long-term responses.

Table 36: Incidence of anti-drug antibodies to Pola

ADA, anti-drug antibodies CCOD: 30 April 2018

Adverse events of special interest

In the randomised Phase II, the overall incidence of Grade ≥3 AEs was higher in patients with R/R DLBCL treated with pola+BR than with BR; the difference between arms was driven mainly by cytopenias (neutropenia, thrombocytopenia, and anaemia) in the pola+BR arm. However, descriptive comparisons of incidences of AEs for pola+BR vs BR should take into account the longer median duration of exposure and higher cumulative exposure to study drug in patients with R/R DLBCL treated with pola+BR relative to the BR arm (median 3.2 months vs 1.4 months, with patients completing a median of 5 cycles vs 3 cycles of treatment), due in part to the higher frequency of early treatment discontinuation in the BR arm as a result of disease progression.

Fewer patients treated with pola+BR died due to progressive DLBCL compared to those patients treated with BR. There were 9 deaths due to AEs (Grade 5) in the pola+BR arm and 11 deaths dues to AEs in the BR arm, with no clear pattern of fatal events emerging in any arm. In the pola+BR arm, 4/9 fatal AEs were due to infections, with pneumonia the cause of death of two patients with R/R DLBCL. Deaths due to pneumonia and other infections were also reported in the BR arm.

Neutropenia, an identified risk of Pola, was among the most commonly observed toxicities in patients treated with pola+BR in the study. While neutropenia events, including febrile neutropenia were mainly Grade 3 or 4, they were adequately managed by delaying study drug administration and use of G-CSF support, as mandated by the protocol (80).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 60 of 144

Neutropenia events were reversible and did not lead to a high rate of study treatment discontinuation or drug dose reduction. None of the events of neutropenia were fatal. Thrombocytopenia and anaemia are potential risks of Pola. Grade 3 or 4 thrombocytopenia and anaemia events were frequently reported in patients with R/R DLBCL (40.0% and 24.4% respectively). As with neutropenia, thrombocytopenia was manageable with treatment delays of all study drugs and dose reduction (bendamustine only). Grade ≥3 anaemia usually resolved without specific action taken to the study drug administration.

The incidence of infections (all grade, Grade ≥3, and serious) was comparable between the pola+BR and BR arms. Four cases of infection in the pola+BR arm and three cases of infection in the BR arm were fatal.

Peripheral neuropathy, including peripheral sensory and/or motor neuropathy, is an identified risk of Pola, consistent with the mechanism of action of MMAE and was reported in 39% patients receiving pola+BR (Grade 1: 21%; Grade 2: 18%] vs 3% (all Grade 2) in those receiving BR. However, PN infrequently led to study treatment discontinuation (1 patient) or drug dose reduction (two patients). Moreover, patient-reported severity of PN-related symptoms as captured by mean score of responses to items on the TINAS questionnaire were generally reported to be mild across the majority of the treatment period in both the pola+BR and BR arms (see Section 2.6.6).

Other selected events specifically analysed as potential risks of pola included hepatic toxicity events and diarrhoea, which were mainly Grade 1 or 2 and were tolerated, requiring no specific action to be taken with study treatment. No cases of potential drug-induced liver injury (according to Hy’s Law) were reported.

Overall, the combination of 1.8 mg/kg pola with BR for 6 cycles in a pre-treated R/R DLBCL population was acceptable, and consistent with the known safety profiles for each treatment, with no new safety signals identified.

B.2.11

Ongoing studies

The study is ongoing and is scheduled to end at the time point at which all patients enrolled in the study have either had at least 2 years of follow-up from the time of the treatment completion visit or have discontinued the study.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

xxxxxxxxxxxxxxxxxxxx

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 61 of 144

Two cohorts were added to confirm the efficacy, safety, and pharmacokinetics of the new lyophilised formulation of pola in combination with BR (Arm G [N=42] and Arm H [N=60]).

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxx. Xxxxxxxx xxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxx xxxx xxxxxx xx xxxxx xxx xxxxxxxxxxxx xxxxxxx xxxxxxxx xxxxxxxxxxxx.

B.2.12 Innovation

Antibody-drug conjugates (ADCs) are an innovative class of anticancer treatment agents that comprise a monoclonal antibody targeted to a tumour antigen, a chemical linker, and a potent cytotoxic agent, which is often too toxic to be given as conventional chemotherapy (86). The characteristics of the linker component of an ADC are key to ensuring that the ADC molecule remains relatively stable in the circulation to prevent non-specific release of the cytotoxic agent, yet allowing the linker to be cleaved to release the cytotoxin within a specific microenvironment within the tumour cell (87). Improvements to linker technology associated with highly potent cytotoxic payloads have permitted the development of targeted ADCs that offer meaningful efficacy while minimising side effects (88).

Polatuzumab vedotin is the only ADC targeting CD79b. In doing so it preferentially delivers a potent anti-mitotic agent (monomethyl auristatin E [MMAE]) to B cells, resulting in anticancer activity against B cell malignancies. The pola molecule consists of MMAE covalently attached to a CD79b directed humanised IgG1 monoclonal antibody through a protease cleavable linker, maleimidocaproyl valine citrulline p aminobenzyloxycarbonyl (1-3).

Figure 7: Polatuzumab-vedotin

==> picture [310 x 150] intentionally omitted <==

CD79b is a signalling component of the B cell receptor expressed on the surface of B cells and is found in abundance in people with DLBCL. As such, CD79b expression is restricted to

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 62 of 144

normal cells within the B cell lineage (with the exception of plasma cells) and malignant B- cells; therefore targeted delivery of MMAE is expected to be restricted to these cells (3).

Antibodies bound to CD79b are rapidly internalised, which makes CD79b ideally suited for the targeted delivery of cytotoxic agents (2). After internalisation, the conjugate is cleaved by lysosomal enzymes to release MMAE, which binds to tubulin, disrupts the microtubule network, and results in the inhibition of cell division and cell growth, and induction of apoptosis (3-5). MMAE has a mechanism of action that is similar to vincristine, a cytotoxic agent used in DLBCL therapy.

Figure 8: Mechanism of action of antibody-drug conjugates

==> picture [240 x 196] intentionally omitted <==

Although R-CHOP is the standard of care for patients with previously untreated DLBCL, approximately 30–50% of patients are not cured by this treatment, depending on the stage of disease and prognostic index (37) (see Section B.1.3.2). While high-dose chemotherapy followed by ASCT offers a second chance for cure in some of those patients, approximately half of patients will not respond to subsequent therapy because of refractory disease (28), and a significant number are ineligible for this intensive therapy because of age, comorbidities or chemotherapy-insensitive disease (24, 28). Therefore, there is a significant need for new and more effective treatments with at least acceptable, if not superior, safety and tolerability profiles for patients with DLBCL that relapses or is refractory to treatment.

In the randomised phase of GO29365, pola+BR has clearly demonstrated a significant survival benefit in comparison to BR across all lines of therapy in the R/R DLBCL setting (see Section B.2.6.3; xxxxxxxxxxx xxxxxxx xxxxxxxxxx xxxxxxxx xxxxxxxxx xxxxxx xxxx xxxxxx xxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx]), while a clinically meaningful benefit was also observed in terms of IRC- and investigator-assessed PFS. A high CR rate (40%) was also observed with Pola+BR treatment, which has been associated with

improved outcomes in DLBCL and may in part explain the durable responses of at least 20 Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 63 of 144

months observed in a proportion of patients receiving pola+BR in the GO29365 study. In view of the survival advantage, high CR rate and prolonged disease control, pola+BR represents a major therapeutic innovation for a patient population with high unmet medical need.

The pola treatment regimen consists of an intravenous infusion of 1.8 mg/kg pola every 21 days for 6 cycles; the first infusion takes 90 minutes, which if tolerated can then be followed by 30 minute infusions for subsequent doses. This infusion duration, coupled with the fact that the individual components of the pola+BR regimen can be administered in any order (unlike some other chemotherapy combinations) ensures optimal administration of pola+BR with same day delivery, negating the need for any additional service provision or healthcare resource. Furthermore, administration of pola for 6 cycles may improve tolerability by reducing PN risks versus longer treatment durations and higher doses (84).

The potential of pola+BR to address the high unmet need in DLBCL was recognised by the EMA and FDA when PRIME and Breakthrough Therapy was granted in June 2017 and September 2017 respectively, which was followed by FDA approval in June 2019. Moreover, pola+BR for the treatment of adult patients with R/R DLBCL was granted a Promising Innovative Medicine (PIM) designation by the Medicines and Healthcare Products Regulatory Agency (MHRA) in December 2018, indicating that this treatment combination has the potential to address an unmet clinical need for patients with a life-threatening condition. In June 2019, the MHRA issued a Positive Final Scientific Opinion for the pola+BR Early Access to Medicines Scheme (EAMS) for adult R/R DLBCL patients who are ineligible for transplant. The lyophilised formulation of pola is being used for this EAMS.

B.2.13 Interpretation of clinical effectiveness and safety evidence

The Phase Ib/II study GO29365 evaluated the efficacy and safety of pola 1.8 mg/kg in combination with rituximab or obinutuzumab plus bendamustine in relapsed or refractory follicular lymphoma or DLBCL. This submission focuses on the combination of pola+BR in patients with R/R DLBCL only in accordance with the proposed marketing authorisation indication.

Polatuzumab vedotin has previously demonstrated an acceptable safety and tolerability profile in patients with R/R DLBCL as a monotherapy in a Phase I study (1) as well as clinical efficacy in combination with rituximab in the Phase II Romulus study, with an overall response rate of 54% (median duration of follow-up 17.4 months). The Phase II Romulus study also evaluated the efficacy of a second ADC in combination with rituximab – pinatuzumab vedotin (pina). Following this study, pola was selected for future development

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 64 of 144

in NHL, based on a longer duration of response and an overall benefit/risk ratio favouring pola in combination with rituximab compared with pina with rituximab (89).

There is no universal standard of care regimen for patients with R/R DLBCL who are ineligible for transplant or who relapse after transplant, as no prior randomised trials have established the superiority of one regimen over another for this population. For instance, one of the largest studies to date, a randomised Phase III study of inotuzumab ozogamicin + rituximab compared with investigator’s choice (BR or gemcitabine + rituximab) that included 306 patients with R/R DLBCL, failed to show a benefit in terms of ORR, PFS or OS for patients in the experimental compared to control arms (59). It is noteworthy that 80% of control patients in this study received BR and 20% received gemcitabine + rituximab, perhaps suggesting that investigators perceive bendamustine to have greater efficacy than gemcitabine

Cross trial comparisons of survival outcomes between different regimens have significant limitations in terms of differences in enrolled or observed populations such as numbers of refractory vs relapsed patients, numbers of prior lines of therapy, and when trials were conducted (some earlier trials occurred at a time when a significant proportion of patients had not been exposed to rituximab in first-line therapy). Thus, it remains extremely difficult to definitively conclude that regimens based on platinum or gemcitabine are truly superior to BR in the contemporary setting for transplant-ineligible R/R DLBCL patient population without a randomised trial.

The combination of BR has been shown to be active in transplant-ineligible patients with R/R DLBCL with a manageable haematologic toxicity profile; median PFS has been reported to be 3.5–6.7 months and median OS reported to be 6.7–9.5 months (59-62). Furthermore, guidelines such as those put forth by the NCCN include BR as a treatment option for patients with DLBCL who are not candidates for high-dose therapy with ASCT (34), therefore BR was chosen to be combined with pola and is considered to be a relevant comparator. Furthermore, the bendamustine backbone was selected to minimise the overlapping toxicity of PN that may occur with platinum-based therapies. The bendamustine dose of 90 mg/m[2] on Days 1 and 2 in combination with rituximab was selected due to concerns of additive myelotoxicity of bendamustine at 120 mg/m[2] when used in combination with rituximab. This dose recommendation was also supported by recommendations established by an international consensus panel (82).

The randomised phase of GO29365 demonstrated a clear and consistent clinically meaningful benefit of pola+BR in a pre-treated R/R DLBCL population. The primary endpoint of IRC-assessed CR rates by PET-CT at the primary response assessment was significantly

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 65 of 144

higher in the pola+BR arm compared to the BR arm (40.0% vs. 17.5%; Δ22.5%, p=0.0261). Most patients who responded to treatment achieved a CR. Objective response rates as measured by the IRC were also higher in the pola+BR arm compared to the BR arm (45.0% vs.17.5%; Δ27.5%, p=0.0069). Of particular interest, a proportion of patients receiving pola+BR had durable responses; xxxxxxxxxx xxxxxxxxxxxxxxxxxxxx xxxxxx xxx xxxxxx xxxxxx xxxxx xxxxxxxxx xxxxxx xxxx xxxx xxxxxx xxxxxx xxxxxxx xxxxxxxx xxxxxxxx xxxxxx

xxxxxxxxxxxxxxxxxxxxxxxx There were no obvious clinical predictors of response, as all subgroups examined appeared to benefit, including patients with refractory disease and patients who have received multiple prior lines of therapy. It is also notable that at an advisory board meeting, UK clinical experts commented that the CR rate seen in the control arm of GO29365 is in-line with what they would expect to see with other current treatment options for this population (35).

GO29365 is the first randomised study to show survival benefit in transplant-ineligible R/R DLBCL patients as a clinically meaningful benefit was seen in PFS and OS for pola+BR compared with BR. At the latest data cut (11 October 2018), the risk of death was reduced by xxx in patients treated with pola+BR compared to BR (xxxxxxxxxxxxxxxxxx xxxxxxxxxxx), with a median OS of xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx with pola+BR compared to xxxxxxxxxxxxxxxxxxxxx with BR alone.

Forty patients were enrolled into each treatment arm of the randomised Phase II cohorts of GO29365, with the sample size being sufficient to demonstrate a benefit in OS. The number of patients enrolled was calculated to provide meaningful estimates; 40 patients randomised to each treatment arm provided a margin of error not exceeding 17% for the 95% exact CIs for estimation of the true CR rate (primary efficacy endpoint), with at least an 87% chance of observing at least one adverse event with a true incidence of ≥5% (primary safety endpoint). The robustness of the benefit of pola+BR in R/R DLBCL is demonstrated by consistent efficacy findings between IRC and investigator and assessment, while OS benefits with pola+BR were observed across different subgroups defined based on demographic and baseline disease characteristics, duration of response to last therapy and cell of origin, with no difference between ABC- and GCB-like DLBCL.

The CHMP acknowledged that based on the April 2018 CCOD, the benefit-risk profile of pola+BR observed in GO29365 is positive and it was therefore acceptable to file for marketing authorisation based on these data.

Randomisation and stratification are methods used to construct comparable treatment arms by reducing the baseline imbalance over known and unknown factors. However, some baseline imbalance may arise by chance, as seen in the BR treatment arm where more

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 66 of 144

patients with bulky disease, categorised as IPI high risk and, ECOG PS 2 and refractory to last prior treatment were enrolled compared with the pola+BR arm. However, stratified multiple Cox regression analysis adjusting for this potential imbalance in baseline characteristics also supported the robustness of the treatment effect of pola+BR with the OS HR remaining in the 0.24–0.78 range post-adjustment.

The combination of pola+BR was associated with additional toxicity as expected when an additional therapeutic agent is added to the BR combination, but for the most part was clinically manageable; no new safety signals were identified relative to the known safety profile of pola, bendamustine, or rituximab. The toxicity profile observed with pola+BR in the GO29365 study is also in keeping with adverse events observed with other ADCs for NHL (90). Moreover, the tolerability of the pola+BR combination is demonstrated by the fact that more patients completed all six treatment cycles than those patients in the BR arm (46.2% vs 23.1%). A higher rate of Grade 3–4 cytopenias was observed with pola+BR compared with BR, but this did not result in a higher risk of infection or need for transfusion. This higher rate of AEs was likely contributed to by increased susceptibility in these pre-treated patients, as well as disease progression and subsequent anti-lymphoma therapy in this high-risk population of R/R DLBCL.

Peripheral neuropathy is a recognised toxicity associated with MMAE based antibody-drug conjugates, and was closely monitored during GO29365. The majority of PN observed was low grade and reversible, and led to few patients experiencing dose reduction or delay. Furthermore, patient-reported severity of PN-related symptoms as captured by mean score of responses to items on the TINAS questionnaire were generally reported to be mild across the majority of the treatment period in both the pola+BR and BR arms for patients with R/R DLBCL.

Conclusions

On the basis of the data from the randomised Phase II GO29365 study, the overall benefitrisk assessment of pola+BR in patients with R/R DLBCL ineligible for transplant is considered to be positive. The benefits of durable and higher PET-based CR response rates after completion of treatment and longer median PFS and OS with pola+BR relative to BR are significant and clinically meaningful. Moreover, the analysis of the efficacy findings between investigator and IRC were consistent, thus further supporting the robustness of the benefit of pola+BR in R/R DLBCL.

Overall, pola+BR offers a new treatment option for transplant ineligible patients with a high unmet medical need, who may be rapidly progressing and need urgent disease control. Furthermore, the safety profile of pola+BR is considered to be acceptable as the additional

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 67 of 144

toxicities observed with the addition of pola to BR are manageable and readily monitored, therefore this regimen may also provide an alternative treatment option for those patients who are unable to tolerate intensive treatments after progressing on platinum-based salvage regimens or who are refractory to or relapse following ASCT.

Table 37: End-of-life criteria

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 68 of 144

B.3 Cost effectiveness

Summary of Cost-Effectiveness Analysis for Pola+BR vs BR

• No published economic analyses were identified for polatuzumab vedotin or the comparators identified in the NICE final scope in R/R DLBCL, therefore a de novo cost-effectiveness model was developed.

• • A partitioned survival model was built with three mutually exclusive health states: PFS, PD and Death. The proportion of alive patients falling into PFS or PD was defined by extrapolated PFS and OS survival curves from GO29365.

• The model possesses a cycle length of 1 week, a lifetime (45 year) time horizon and costs and benefits discounted at 3.5% as per the NICE reference case (91).

• In the base case, BR was selected as the comparator to Pola+BR, as it was deemed representative of current standard of care for transplant-ineligible R/R DLBCL patients in the UK, and a robust comparison to Pola+BR using data from the randomised GO29365 trial was possible.

• Survival analysis was performed to identify appropriate parametric survival functions to extrapolate PFS and OS (92). Standard survival functions and cure-mixture models were explored, on the basis of clinical expert opinion and evidence from the literature that patients who achieve 2-years PFS following treatment are likely to experience long-term survival aligned with that of the age- and sex-matched general population.

• Based on visual fit to the GO29365 KM data and plausibility of the long-term extrapolations, curemixture models using the generalised gamma distribution were selected for the base case for PFS and OS. The OS extrapolation was informed by the ‘cure fraction’ for PFS, on the basis that achievement of long-term PFS is an indicator of long-term survival.

• Base case utilities were modelled by health state and were sourced from the recent manufacturer submission for axicabtagene ciloleucel in R/R DLBCL (69). AE disutilities were applied based on CTCAE Grade ≥3 AEs from GO29365 and were sourced from recent NICE appraisals. Patients who remained in PFS for >2 years reverted to age- and sex-matched general population utilities (93).

• • Categories of costs included in the model were acquisition, administration, supportive care and subsequent treatment costs. Costs were sourced from NHS Reference Costs 2017–18, PSSRU 2018, and the BNF and eMIT (both accessed June 2019). Patients who remained in PFS for >2 years no longer accrued supportive care costs.

• • The base case acquisition costs for polatuzumab vedotin were based on the availability of both 140 mg and 30 mg vials, the latter of which is due to be available in xxxxxxxxx. Alternative scenarios, including interim arrangements with compounding services, are explored in scenario analyses.

• • The base case results of the analysis demonstrated that Pola+BR is cost-effective vs BR at an ICER of £26,877 per QALY. This was driven by the substantially greater QALY gain vs BR.

• • The DSA and scenario analyses demonstrated the robustness of the base case results. DSAs identified no input parameters that resulted in a range of ICER values greater than 12% of the base case. Scenario analyses identified that in general, the ICER value remained relatively unchanged, except where survival modelling extrapolations of reduced clinical plausibility were used.

• • Variation in the PSA from the base case was observed, which may be attributed to the parameter uncertainty around the use of the generalised gamma distribution for modelling survival and the independent variation of input parameters for long-term survival and long-term remission.

• • The results of the cost-effectiveness analysis support that Pola+BR is a cost-effective treatment option vs BR, which may be considered representative of standard of care for transplant-ineligible R/R DLBCL patients in the UK.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 69 of 144

B.3.1 Published cost-effectiveness studies

No published cost-effectiveness analyses were available for the technology or comparator regimens identified in the scope. Further details on the methodology and results of the SLR are presented in Appendix G.

B.3.2 Economic analysis

As noted in Section B.3.1, no prior cost-effectiveness analyses that assessed Pola+BR in patients with R/R DLBCL who are ineligible for transplant were identified in the SLR. A de novo economic model was therefore built to inform decision making.x

B.3.2.1 Patient population

Patients with R/R DLBCL ineligible for SCT were included in the economic evaluation. This patient population is in line with the expected licensed indication for Pola+BR, the decision problem addressed in this submission (see Section B.1), and the patient population included in the GO29365 trial.

B.3.2.2 Model structure

An Area-Under-the-Curve (AUC) or partitioned survival analysis model was developed in Microsoft Excel. The AUC model structure is in line with NICE Decision Support Unit (DSU) guidance (94) and is consistent with previous appraisals conducted in this disease setting (67, 69, 70). An important benefit of the partitioned survival approach is that modelling of OS and PFS is based on study-observed events, which is expected to accurately reflect disease progression and the long-term expected survival profile of patients treated with Pola+BR.

The model includes three mutually exclusive health states: “Progression-Free Survival (PFS)”, “Progressed Disease (PD)” and “Death” as shown in Figure 9.

Figure 9: Economic model structure

==> picture [326 x 169] intentionally omitted <==

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 70 of 144

All patients enter the model in the PFS health state and remain in this health state until they progress. Upon progression, patients either transition into the PD health state or enter the absorbing health state of Death. Patients in the PD health state stay in that health state until Death. Patients cannot transition to an improved health state (i.e. from PD to PFS); a restriction that is consistent with previous economic modelling in oncology.

The proportion of patients in each health state at any time is defined by the partitioning of alive patients alive into “PFS” and “PD” at discrete time points, based on the PFS and OS survival curves from GO29365. The proportion of patients falling into the “PD” health state is the difference between OS and PFS, as illustrated in Figure 10. The “PD” health state also includes any further lines of treatment, as described in Section B.3.5.5.

Figure 10. Example of a partitioned survival model

==> picture [395 x 237] intentionally omitted <==

Features of the de novo analysis

The model has been designed to use a weekly cycle, with the proportion of patients in each health state calculated each week. Transition between health states can occur at any time within the cycle, therefore a half-cycle correction was applied to account for the over- or under-estimation of transitions occurring at the beginning or end of the cycle.

Costs and health-related utilities are allocated to each health state and multiplied by state occupancy to calculate the weighted costs and quality adjusted life years (QALYs) per cycle. Costs and health outcomes are discounted at 3.5% and the perspective of the NHS and Personal Social Services (PSS) is assumed, as per the NICE reference case (91). The model inputs for the Pola+BR versus BR comparison (efficacy, safety and tolerability) are based on the results of the randomised phase II study GO29365 (see Section B.2).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 71 of 144

The economic model uses a time horizon of 45 years, which is considered to be appropriate as a lifetime horizon for patients with R/R DLBCL, taking into account the average age of patients at the start of treatment (69 years). This time horizon ensures all benefits and costs accrued by patients are captured from start of treatment to death, and is consistent with the anticipated survival based on the economic model, with less than approximately 1% of patients still alive at 45 years for Pola+BR.

Model results are reported in terms of costs per QALY gained, reflecting the decision problem.

An overview of how the economic analysis for Pola+BR compares to other NICE appraisals in R/R DLBCL is provided in Table 38 below.

Table 38. Features of the economic analysis

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 72 of 144

BNF, British National Formulary; CMM, cure-mixture model; eMIT, drugs and pharmaceutical electronic market tool; EQ-5D, EuroQol Five Dimensions; OS, overall survival; PD, progressed disease; PFS, progression-free survival; PSSRU, Personal Social Services Research Unit; SOC, standard of care

B.3.2.3 Intervention technology and comparators

Intervention – Pola+BR

The model intervention is Pola+BR, as described in Section B.1.2. Pola+BR was modelled to follow the dosing schedule implemented in GO29365, see Section B.2.3, and in accordance with the anticipated marketing authorisation (7).

Comparators – BR (base case) and R-GemOx (scenario)

In the base case, Pola+BR was compared to BR, the comparator in the randomised GO29365 study (see Section B.2). A scenario analysis was also performed in which R- GemOx was included as a comparator, under the assumption of equivalent efficacy to BR.

As discussed in Sections B1.1 and B.1.3, there is no universally accepted standard of care regimen for R/R DLBCL patients not eligible for SCT, with patients typically prescribed one regimen from a range of available gemcitabine and/or platinum-based therapies, or BR. Choice of chemotherapy in SCT-ineligible patients is dependent upon clinician preference (35), and there is no strong evidence that one regimen is superior to another (see Section B.1.3.2).

In the NICE final scope, a number of potential regimens used in NHS clinical practice were identified (R-GemOx, R-Gem, R-P-MitCEBO and [R])DECC), in addition to BR (95). However, in the clinical SLR, studies were only identified for R-GemOx, please see Appendix D. The possibility of performing a robust indirect comparison between Pola+BR and R-GemOx was found to be unfeasible, as no connected network of randomised studies was identified (see Section B.2.9). A robust unanchored comparison was similarly found to be not feasible due to significant or unknown differences in prognostic factors in the study populations for GO29365 and captured R-GemOx studies, including proportion of refractory patients, prior rituximab exposure and number of prior lines of treatment. A lack of robust and comparable studies assessing therapies for DLBCL is an inherent limitation of the disease area, as identified by the NICE technology appraisals for tisagenlecleucel and

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 73 of 144

axicabtagene ciloleucel, in which the indirect comparisons performed to chemotherapy (necessitated by the single-arm trials for the interventions) were deemed to be associated with substantial bias, significantly limiting the use of the comparative evidence to inform decision making (69, 70).

As such, given the lack of evidence demonstrating superiority of one chemotherapy regimen over another, clinical opinion that the range of available chemotherapy regimens for DLBCL are considered to be equally effective, and the ability to make a robust comparison with Pola+BR using the data from the GO29365 randomised controlled trial, BR was selected as the comparator to Pola+BR in the base case analysis.

In order to supplement the evidence presented in the base case, a scenario analysis was also performed in which R-GemOx was included as an additional comparator. This scenario assumed equivalent efficacy with BR, which is supported by recent real-world evidence demonstrating no OS difference between people with R/R DLBCL treated with BR and R- GemOx. Limited UK-based data are available, however, in a cohort of DLBCL patients from the US Veterans Health Association database that had been treated with either second-line BR or R-GemOx, and followed-up for 11.3 and 11.7 months, respectively, median OS was estimated at 11 months for BR and 13 months for R-GemOx. Univariate and multivariate analyses both found no significant difference in OS between either regimen (96).

In addition to this recent real-world data, reported outcomes in prospective studies fall into a similar range. In a Phase II study based in France, Mounier et al. reported median PFS and OS values for R-GemOx for transplant-ineligible R/R DLBCL patients who had previously received rituximab of 4 months and 8 months, respectively (58). More recently, Hong et al. reported median PFS and OS values for transplant-ineligible R/R DLBCL patients treated with BR of 3.9 months (95% CI: 2.4–4.5) and 6.7 months (95% CI 4.7–8.7), respectively (60). The GO29365 study BR arm showed a median PFS (investigator-assessed) of 2.0 months (95% CI: 1.5–3.7) and OS of 4.7 months (95% CI: 3.7–8.3) (78).

Finally, BR and R-GemOx are expected to acquire similar acquisition costs, as presented in Table 39. Therefore, the base case analysis for Pola+BR versus BR, supplemented by the scenario including R-GemOx as an additional comparator, is considered to provide a representative analysis of the cost-effectiveness of Pola+BR vs standard of care chemotherapy regimens used in the UK for R/R DLBCL patients who are ineligible for transplant.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 74 of 144

Table 39. Drug acquisition costs for BR and R-GemOx

aCosts sourced from BNF 2019 (rituximab, bendamustine) (97) and eMIT 2019 (gemcitabine, oxaliplatin) (98); bDosing regimens sourced from GO29365 (BR) and Mournier 2013 (R-GemOx) (58); cFull details for how the acquisition costs for each regimen have been calculated can be found in Section B.3.5.2.

BR, bendamustine + rituximab; R-GemOx, rituximab + gemcitabine + oxaliplatin

B.3.3 Clinical parameters and variables

The primary source of clinical data for the Pola+BR and BR arms of the economic model is the GO29365 study. Data from the latest available data cut (October 2018) have been used to inform the clinical parameters for PFS and OS (results data for which are reported in Section B.2.6.3). For treatment duration and treatment-related AE rates, the latest available data were from the clinical cut-off date of April 2018. All patients had completed treatment with Pola+BR or BR by this date.

B.3.3.1 Survival inputs and assumptions

PFS and OS results from GO29365 were extrapolated to the model lifetime time horizon, as lifetime results are not available for patients who participated in this study. NICE DSU Technical Support Document (TSD) 14 (92), which provides guidance on survival analysis, was followed to identify appropriate parametric survival models to model both outcomes. Specifically, the following points were performed:

1. Visual inspection of the OS and PFS log-cumulative hazard plots, based on patient level data for the two arms of GO29365, to test for the plausibility of the proportional hazards assumption and to examine the hazard of progression or death in each arm over time

2. The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) goodness-of-fit statistics were calculated to assess statistical fit of the models to both arms of the PFS and OS KM data from GO29365

3. The clinical plausibility of the long-term extrapolations for the base case parametric models was validated by comparing the long-term behaviour of the models with suitable data sources and the expectations of clinical experts

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 75 of 144

For both PFS and OS, application of standard parametric survival models (exponential, Weibull, Gompertz, log-normal, generalised gamma and log-logistic) was explored, in addition to the fitting of cure-mixture models. Cure-mixture models represent an approach to modelling cancer therapies for which there is evidence to support that a proportion of treated patients enter long-term remission, and subsequently experience mortality aligned with that of the general population. This is reflected in the parameterisation of cure-mixture models, which assume the patient population comprises two subpopulations; the first subpopulation is considered to be at the same risk of mortality as the age- and sex-matched general population (sourced from the Office for National Statistics for this model (99), whilst the mortality rate of the second subpopulation is defined by a selected standard parametric survival curve. The proportion of patients falling into the first population (known as the ‘cure fraction’) is estimated through logistic regression of trial data. The extrapolations for each subpopulation are then combined via the cure fraction to obtain extrapolations for the population as a whole.

Accordingly, evidence to support the exploration of cure-mixture survival modelling in the context of this appraisal is as follows:

A study of the natural history of newly diagnosed DLBCL patients treated with immunochemotherapy identified that patients who did not experience a progression or death event after two years went on to experience subsequent survival equivalent to that of the age- and sex-matched general population (93). Whilst an equivalent study has not been performed in the R/R DLBCL setting, clinical experts confirmed that patients who achieve two years PFS are at very low risk of subsequent progression, and their risk of death can be assumed to have returned to a level close to that of the matched general population (35).

Nevertheless, with current standard of care options, the proportion of patients achieving sustained remission in the transplant-ineligible R/R DLBLC setting is small. Of these, clinical experts estimated that the proportion of patients achieving long-term remission and subsequent long-term survival is approximately 5–10% (35). Similarly, the SCHOLAR-1 study, a multi-national study which combines outcomes from two randomised controlled trials and two academic databases, reported a two-year OS of 11% for refractory DLBCL patients who had not undergone SCT (24).

PFS and OS data from the GO29365 study demonstrate that compared to current standard of care, Pola+BR is likely to offer patients an improved probability of achieving long-term remission (and therefore long-term survival), as evidenced by the statistically significantly improved rate of PFS vs BR. Of note, a very low risk of relapse or death can be observed in the KM plots for PFS and OS for Pola+BR towards the end of follow-up, indicative of a very

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 76 of 144

low risk of relapse or death for patients who were still alive towards the end of follow-up (Sections B.3.3.2 and B.3.3.3).

Finally, precedent of cure-mixture modelling in NICE appraisals for R/R DLBCL was established in TA567 and TA559, where the respective Committees accepted that patients who are able to demonstrate sustained remission are likely to benefit from long-term survival (69, 70).

B.3.3.2 Extrapolation of PFS

For the extrapolation of PFS in the model, INV PFS from GO29365 was selected over IRC PFS. The rationale for this selection is that treatment decisions for patients included in the trial, for example, to move a patient to the next line of treatment, were based on progression as measured by the investigator. As such, these data are more consistent with the treatment pathway experienced by patients in the trial and therefore deemed more suitable for inclusion in the model.

Assessment of the proportional hazards assumption

Visual inspection of the log-cumulative hazard plots for PFS from GO29365 (Figure 11) indicates that making the proportional hazards assumption is plausible. This is evidenced by the approximately parallel lines that can be observed for Pola+BR and BR for INV PFS, indicating that the ratio of hazard rates between arms remains approximately constant over the follow-up period.

Figure 11. Log-cumulative hazard for PFS (INV; GO29365)

==> picture [348 x 250] intentionally omitted <==

BR, bendamustine + rituximab; INV, investigator assessed; Pola+BR, polatuzumab + bendamustine + rituximab; PFS, progression-free survival

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 77 of 144

As such, investigating the fitting of proportional hazards survival functions (exponential, Gompertz or Weibull) to model PFS for Pola+BR and BR was considered appropriate. The fitting of the log-normal, log-logistic and generalised gamma parametric survival functions to the observed data was also explored.

The standard extrapolations were fitted to the GO29356 data using two approaches. The first approach was to model curves independently, and the second, to use a dependent approach, whereby the comparator survival curves were estimated via a treatment effect applied to the intervention curve. For the dependent approach, extrapolations were fitted for both treatment arms in SAS, with treatment as a covariate.

Assessment for cure-mixture modelling

The suitability of the GO29365 PFS data to the application of cure-mixture modelling was assessed. As discussed in Section B.3.3.1, to support the use of cure-mixture modelling, the trial data must indicate that a proportion of patients enter long-term remission (PFS) and are therefore likely to experience long-term survival. In line with this, the KM data for INV PFS presented in Figure 12 demonstrate a very low rate of relapse for both Pola+BR and BR around the 24-month timepoint, suggesting there is a fraction of patients in the GO29365 trial who have achieved long-term remission. As discussed previously (Section 3.3.1), evidence from the literature and clinical opinion suggest that patients remaining progressionfree for two years are expected to demonstrate survival aligned with that of the general population. A drop-off in the ‘plateau’ shape of the KM data for the two arms can be observed towards the end of follow-up, however, this can be attributed to the low patient numbers at risk in the trial towards the end of follow-up. Later data cuts are expected to provide more data around this time point.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 78 of 144

Figure 12. KM plot for INV PFS (GO29365; data cut: October 2018)

==> picture [674 x 346] intentionally omitted <==

BR, bendamustine + rituximab; CI, confidence interval; INV, investigator-assessed; KM, Kaplan Meier; PFS, progression-free survival; Ph, phase; Pola+BR, polatuzumab + bendamustine + rituximab

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B- cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 79 of 144

Both progression and death are considered as events when assessing PFS. Accordingly, progression events may be analysed alone (i.e. excluding any death events), as presented in Figure 13, in order to further assess the suitability of using cure-mixture modelling. It can be observed from Figure 13 that most progression events occur within the first 12 months in both arms of GO29365, and that patients are at a very low risk of progression after 24 months, further supporting the exploration of using cure-mixture modelling to extrapolate PFS.

Figure 13. Cumulative incidence of progression (INV) from GO29365 a) Pola+BR and b) BR

==> picture [261 x 9] intentionally omitted <==

----- Start of picture text -----
a) b)
----- End of picture text -----

==> picture [228 x 157] intentionally omitted <==

==> picture [217 x 163] intentionally omitted <==

BR, bendamustine + rituximab; INV, investigator assessed; Pola+BR, polatuzumab + bendamustine + rituximab

Finally, additional rationale for exploring cure-mixture modelling is provided by the logcumulative hazard plot presented previously to assess the proportional hazards assumption (Figure 11). The plot indicates a decline in the hazard of progression for both interventions at the end of the follow-up period, again suggesting that a proportion of patients enter longterm remission.

The parameterisation of cure-mixture models means this model type is better suited to reflect the potential for a proportion of patients to achieve long-term survival. Accordingly, in addition to the standard parametric models, cure-mixture models were investigated and fitted independently to the two arms of the model.

The proportion of patients achieving long-term remission is a parameter that can be fitted from the observed GO29365 data via logistic regression. The ‘cure fraction’ of patients are assumed not to progress or be susceptible to cancer-related death.

Statistical fit of models to the observed data

Table 40The AIC and BIC goodness of fit results for the functions used to model PFS for Pola+BR and BR in GO29365, as well as a qualitative impression of visual fit to the

observed KM curve are provided below.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 80 of 144

In the selection of suitable survival functions for PFS, for clinical plausibility, consideration was given to consistency with the extrapolations being explored for OS (discussed in Section B.3.3.3). For all survival functions explored for OS for both arms, parameterisation for the Gompertz model did not converge. Accordingly, given that a Gompertz extrapolation would ultimately not be selected for OS, this function was excluded from consideration for PFS. As such, AIC/BIC values are therefore not presented for this extrapolation.

Table 40. Ranking of PFS distributions for Pola+BR and BR based on AIC, BIC and assessment of their visual fit

aA ✓ symbol indicates a model has a good fit to the KM data; A ~ symbol indicates a model has an average fit to the KM data; a × indicates a model has an unsuitable fit to the KM data.[b] The presented statistics represent the overall fit of the dependent model to both arms of the trial.[c] The Gompertz extrapolation was not considered for PFS for either arm due failure of parameterisation for this function for OS; AIC/BIC statistics are therefore not presented. AIC, Akaike Information Criterion; BIC, Bayesian Information Criterion; BR, bendamustine + rituximab; KM, Kaplan-Meier; NA, not available; Pola+BR, polatuzumab + bendamustine + rituximab

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 81 of 144

Among the standard models (dependently and independently fit), the AIC and BIC statistics indicated that all models had a similar statistical fit to the KM data in both arms. The topranking models (both arms) for both dependent and independently-fit extrapolations were the log-normal, generalised gamma and log-logistic. Similarly, for the cure-mixture models, minimal variation was observed among the statistics; the best ranking models were the generalised gamma and log-logistic for Pola+BR and the log-normal, log-logistic and exponential for BR.

The fit of the dependently modelled extrapolations was inspected visually (Figure 14). The exponential and Weibull models appeared to overestimate PFS in both the Pola+BR and BR arms early in the extrapolation, and did not capture the decline in progression at the end of the follow-up period. In the Pola+BR arm, the log-logistic, log-normal and generalised gamma appeared to provide a better fit early in the extrapolation, but similarly did not capture the decline in progression in both arms towards the end of follow-up well. The generalised gamma, log-logistic and log-normal curves offered reasonable fits to the BR arm.

Figure 14. PFS standard extrapolation functions (dependent fit)

==> picture [441 x 299] intentionally omitted <==

The Gompertz extrapolation was not considered for either arm for PFS due failure of parameterisation for this function for OS; this extrapolation is therefore not presented.

BR, bendamustine + rituximab; KM, Kaplan-Meier; Pola+BR, polatuzumab + bendamustine + rituximab

Similar conclusions were made from the visual inspection of the independent fit

extrapolations as the dependent fit extrapolations; in the Pola+BR arm, functions typically

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 82 of 144

either overestimated PFS stages in the earlier months and/or underestimated the decline in patient progression towards the end of follow-up. Of all the functions, the generalised gamma provided the most reasonable fit in the Pola+BR arm. In the BR arm, the generalised gamma, log-logistic and log-normal appeared to fit the observed data reasonably well.

Figure 15. PFS standard extrapolation functions (independent fit)

==> picture [443 x 283] intentionally omitted <==

The Gompertz extrapolation was not considered for either arm for PFS due failure of parameterisation for this function for OS; this extrapolation is therefore not presented. BR, bendamustine + rituximab; Pola+BR, polatuzumab + bendamustine + rituximab

Introducing cure-mixture models improved the visual fit of all models to both arms of the KM data (Figure 16), with log-logistic, log-normal and generalised gamma cure-mixture models providing good fits to the observed data in the Pola+BR arm. The Weibull and exponential overestimated PFS early in the extrapolation. In the BR arm, the Weibull appeared to overestimate PFS at the earliest stages of follow-up, with the exponential overestimating progression at later stages. All other models appeared to provide a relatively good fit to the KM data.

Table 41 presents the cure fractions (i.e. the proportion of patients achieving long-term remission) predicted by each of the cure-mixture extrapolations for each arm. The proportion of patients achieving long-term remission falls into a narrow range from 20.8% to 25.9% in the Pola+BR arm, and 0.0% to 4.4% in the BR arm. A narrow range of values demonstrates consistency in the cure fraction estimation across parametric models, further supporting the suitability of this modelling approach.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 83 of 144

Figure 16. PFS cure-mixture extrapolation functions

==> picture [443 x 303] intentionally omitted <==

The Gompertz extrapolation was not considered for either arm for PFS due failure of parameterisation for this function for OS; this extrapolation is therefore not presented. BR, bendamustine + rituximab; KM, Kaplan-Meier; Pola+BR, polatuzumab + bendamustine + rituximab

Table 41. Predicted long-term remission (cure fraction) from PFS cure-mixture model extrapolations

The Gompertz extrapolation was not considered for either arm for PFS due failure of parameterisation for this function for OS; cure fractions for this extrapolation are therefore not presented. BR, bendamustine + rituximab; NA, not available; Pola+BR, polatuzumab + bendamustine + rituximab

Based on visual fit, plausibility of the long-term extrapolation, and alignment with the

selected OS distribution (see Section B.3.3.3), the cure-mixture generalised gamma survival curve was selected for the base case for both arms, whilst the log-normal and log-logistic extrapolations were (applied in both arms) were explored in scenarios.

The final base case extrapolations are shown alongside the selected OS extrapolation in Figure 23 in Section B.3.3.3, where the long-term plausibility of the selected extrapolations for both outcomes is also discussed.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 84 of 144

In the scenario where R-GemOx is explored as a comparator, the base case PFS extrapolations for BR are adopted.

B.3.3.3 Extrapolation of OS

Assessment of the proportional hazards assumption

Visual inspection of the log-cumulative hazard plot for OS from GO29365 (Figure 17) indicates that making the proportional hazards assumption is plausible. This is evidenced by the approximately parallel lines that can be observed between Pola+BR and BR, indicating that the ratio of hazard rates between arms remains approximately constant over the followup period.

Figure 17. Log-cumulative hazard plot for OS in study GO29365

==> picture [400 x 195] intentionally omitted <==

Pola + BR, polatuzumab + bendamustine + rituximab; R-Benda, rituximab + bendamustine

As such, investigating the fitting of proportional hazards survival functions (exponential, Gompertz or Weibull) to model OS for Pola+BR and BR was considered appropriate. The fit of the log-normal, log-logistic and generalised gamma curves was also explored.

As for PFS, the standard extrapolations were fitted to the GO29356 data using both independent and dependent approaches.

Assessment for cure-mixture modelling

Cure-mixture models were also investigated for the modelling of OS, as long-term remission in R/R DLBCL would be expected to be associated with long-term survival. As was observed for PFS, a decline in the hazard of death can be seen towards the end of follow-up in the OS KM data from GO29365 (Figure 18) for both arms of the trial. The majority of death events can be seen to take place prior to 12 months.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 85 of 144

Figure 18. KM plot for OS (GO29365; data cut: October 2018)

==> picture [706 x 346] intentionally omitted <==

BR, bendamustine + rituximab; CI, confidence interval; INV, investigator-assessed; KM, Kaplan Meier; OS, overall survival; Ph, phase; Pola+BR, polatuzumab + bendamustine + rituximab

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B- cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 86 of 144

Further justification for the exploration of cure-mixture modelling is provided by the logcumulative plots presented in Figure 17. A decline in the hazard of death can be observed towards the end of the follow-up period, similar to the decline in the hazard of progression observed for PFS. Again, this is consistent with a proportion of patients experiencing longterm remission and subsequent long-term survival in the trial.

In prior economic evaluations (69, 70), it was stated that only patients who have not yet progressed can be considered to be long-term survivors, based on the view from clinical experts. As the OS rates in GO29365 were higher than PFS rates at the end of the observed follow-up period, a proportion of patients (those in progression) were expected to be at increased mortality risk, and not long-term survivors. Therefore, two cure-mixture model approaches were implemented, both of which ensured a conservative estimate of long-term survivor rates. Firstly, an approach was implemented whereby the proportion of long-term survivors was constrained to the proportion of patients in long-term remission, as fitted from the PFS data, i.e. the cure-mixture mode was informed by PFS. Treatment arms were fitted independently using this approach. Secondly, a dependent model was explored, whereby OS was not informed by PFS but the survival for patients who were not long-term survivors was assumed to be similar in the Pola+BR and BR arm, to achieve a more conservative estimate of long-term survival in the Pola+BR arm based on OS data alone. Further background on cure-mixture statistical models is provided in Appendix M.

Statistical fit of models to the observed data

Table 42 provides the AIC and BIC goodness of fit results for the functions used to model OS for Pola+BR and BR in GO29365, as well as a qualitative impression of visual fit to the observed KM curve for each arm. For all extrapolations, parameterisation of the Gompertz extrapolation for both arms did not converge, and therefore AIC and BIC statistics are not presented for this extrapolation.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 87 of 144

Table 42. Ranking of OS models for Pola+BR and BR based on AIC, BIC and assessment of their visual fit

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 88 of 144

aA ✓ symbol indicates a model has a good fit to the KM data; A ~ symbol indicates a model has an average fit to the KM data; a × indicates a model has an unsuitable fit to the KM data.[b] The presented statistics represent the overall fit of the dependent model to both arms of the trial.[c] Parameterisation did not converge for the Gompertz extrapolation in all four modelling approaches; AIC/BIC statistics are therefore not presented. AIC, Akaike Information Criterion; BIC, Bayesian Information Criterion; BR, bendamustine + rituximab; KM, Kaplan-Meier; NA, not available; Pola+BR, polatuzumab + bendamustine + rituximab

As was the case for PFS, AIC and BIC values indicated a similar statistical fit to the KM data for the standard models (dependently and independently fit) for both arms. The best ranking models in both arms were the log-normal, log-logistic and generalised gamma. For the two cure-mixture models, the AIC/BIC statistics also indicated a similar statistical fit among the extrapolations, with the log-logistic and log-normal curves suggesting the best fit in both arms.

Based on visual inspection, none of the standard models (applied dependently or independently) fitted the observed OS data in Pola+BR arm particularly well, as they tended to overestimate OS early on and did not capture the decline in the observed mortality rate at the end of the follow-up period. In the BR arm, only the log-logistic, log-normal and generalised gamma extrapolations provided a reasonable visual fit (Figure 19 [dependent fit], Figure 20 [independent fit]).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 89 of 144

Figure 19. OS standard extrapolation functions (dependent fit)

==> picture [462 x 270] intentionally omitted <==

Parameterisation did not converge for the Gompertz extrapolations. BR, bendamustine + rituximab; OS, overall survival; Pola+BR, polatuzumab + bendamustine + rituximab

Figure 20. OS standard extrapolation functions (independent fit)

==> picture [467 x 275] intentionally omitted <==

Parameterisation did not converge for the Gompertz extrapolations. BR, bendamustine + rituximab; KM, KaplanMeier; OS, overall survival; Pola+BR, polatuzumab + bendamustine + rituximab

Cure-mixture models were subsequently fitted to the OS GO29365 data (Figure 21).

However, all functions were found to have a poor fit to the KM data; in the Pola+BR arm, all Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 90 of 144

curves underestimated OS early in the extrapolation. In the BR arm, all extrapolations overestimated OS early on, and the majority did not capture the decline in mortality late in follow-up well.

Figure 21. OS cure-mixture extrapolation functions (OS not informed by PFS)

==> picture [468 x 279] intentionally omitted <==

Parameterisation did not converge for the Gompertz extrapolations. BR, bendamustine + rituximab; KM, Kaplan Meier; OS, overall survival; Pola+BR, polatuzumab + bendamustine + rituximab

The cure rates predicted by each model are presented in Table 43. Cure-mixture models fitted directly to OS data produced estimated proportions of patients with long-term survival in the Pola+BR arm ranging from 27.8% to 36.0%. In the BR arm, rates ranged from 0.0% to 11.5%.

Table 43. Predicted long-term survival (cure fraction) from OS cure-mixture model extrapolations (OS not informed by PFS)

Parameterisation did not converge for the Gompertz extrapolation, therefore the cure fraction is not presented. BR, bendamustine + rituximab; OS, overall survival; Pola+BR, polatuzumab + bendamustine + rituximab

Given the poor fit of all models explored to this point and guidance provided by clinical experts, cure-mixture models for which the long-term survivor fraction was informed by the

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 91 of 144

long-term remission fraction (i.e. the OS cure fraction was informed by the PFS cure fraction) were investigated. These models provided an improved fit to the KM data in both arms, with functions providing a closer fit to the OS KM data early on in the extrapolation, and an improved fit to the decline in mortality later in follow-up compared to the standard functions. The best fitting functions for both arms based on visual inspection were the log-normal and generalised gamma (Figure 22).

Figure 22. OS cure-mixture extrapolation functions (OS informed by PFS)

==> picture [466 x 270] intentionally omitted <==

Parameterisation did not converge for the Gompertz extrapolations. BR, bendamustine + rituximab; OS, overall survival; PFS, progression-free survival; Pola+BR, polatuzumab + bendamustine + rituximab

The predicted cure fractions for OS informed by PFS are presented in Table 44.

Table 44. Predicted long-term survival (cure fractions) from OS informed by PFS curemixture model extrapolations

Parameterisation did not converge for the Gompertz extrapolation, therefore the cure fraction is not presented. BR, bendamustine + rituximab; NA: not available; Pola+BR, polatuzumab + bendamustine + rituximab

Based on the fit to the observed data of all extrapolations to both the Pola+BR and BR arms,

the cure-mixture model informed by PFS approach was selected. The generalised gamma

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 92 of 144

and log-normal offered the best fits under this approach. Given that of the two extrapolations, the generalised gamma offered the most conservative cure fraction, this function was chosen as the base case for both arms. The log-normal and log-logistic curves (applied in both arms) were investigated in scenario analyses. The selected base case parametric extrapolation functions for PFS and OS are shown in Figure 23.

In the scenario where R-GemOx is explored as a comparator, the base case OS extrapolations for BR are adopted.

Figure 23. Base case PFS and OS extrapolations

==> picture [443 x 258] intentionally omitted <==

BR, bendamustine + rituximab; KM, Kaplan-Meier; PFS, progression-free survival; Pola+BR, polatuzumab + bendamustine + rituximab; OS, overall survival; R-Benda, rituximab + bendamustine

Clinical plausibility of long-term extrapolations for PFS and OS

Whilst statistical tests and visual inspection are useful in determining which models best fit the observed data, they cannot provide information on how suitable a parametric model is for the time period beyond the final trial follow-up. Therefore, the clinical validity of the selected extrapolations for PFS and OS was assessed by comparing the long-term predictions of the model with expected long-term outcomes.

To evaluate the clinical validity of the selected extrapolations for OS and PFS, two additional data sets in R/R DLBCL patients treated with a polatuzumab vedotin regimen over a longterm follow-up were considered.

Firstly, a cohort of six patients received Pola+BR in the safety run-in period prior to start of the randomised phase of GO29365. Data from these patients were pooled with the Pola+BR

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 93 of 144

arm of the randomised phase of the trial, resulting in a cohort of 46 patients that had been followed for up to 46 months at the October 2018 cut-off date. The KM data for the pooled cohort demonstrates a marked plateau for PFS and OS at the end of the follow-up period (Figure 24), which is more pronounced than that observed in the KM data from the randomised phase of GO29365, due to the longer follow-up time. These data thus further support the presence of a group of patients among the trial population who were treated with a regimen of polatuzumab vedotin and went on to experience sustained remission.

With regards to plausibility of the long-term extrapolations, close alignment can be seen between the model extrapolation and the long-term KM data for OS between 0 and approximately 15 months; following this timepoint, the model extrapolation may be considered a conservative estimation of long-term OS. A close fit between the PFS extrapolation and the KM data can also be seen between 0 and 28 months.

Figure 24. KM for OS and PFS from pooled Pola+BR cohort (N=46) and model extrapolations

==> picture [444 x 260] intentionally omitted <==

INV, investigator-assessed; KM, Kaplan-Meier; OS, overall survival; PFS, progression-free survival; Pola+BR, polatuzumab + bendamustine + rituximab

Long-term data are also available from the ROMULUS (Phase I/II) study, which included a cohort of R/R DLBCL patients with similar characteristics to those in GO29365 (89) and

xxxxxxxx xxxxxxxxxxx xxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxx xxxxxx xxxxx xxx xx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx. Patients were treated with polatuzumab vedotin with rituximab (Pola+R), with polatuzumab vedotin given at a higher dose of 2.4 mg/kg. Overall survival data from ROMULUS are presented in Figure 25alongside the model extrapolations for OS. It can be seen that the OS model extrapolation for Pola+BR appears to be Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 94 of 144

conservative relative to the longer-term ROMULUS data, and that an extended long-term survival on treatment with a polatuzumab vedotin-based regimen is plausible.

Figure 25. OS KM for Pola+BR from the ROMULUS study and model extrapolations

==> picture [443 x 245] intentionally omitted <==

BR, bendamustine + rituximab; KM, Kaplan-Meier; OS, overall survival; Pola-BR, polatuzumab + bendamustine + rituximab; Pola+R, polatuzumab + rituximab

With regards to the plausibility of the long-term BR extrapolations, as discussed in Section B.3.3.1, for transplant-ineligible R/R DLBCL patients treated with current standard of care (represented by the comparator BR arm in this analysis), it is expected that only a small proportion of patients go on to achieve long-term PFS, with clinical experts consulted by Roche indicating estimates of between 5–10% (35). In the BR arm of the model, estimates of the long-term remission rate for the cure-mixture models investigated ranged from 0.0% to 4.4%, indicating that model estimates are in line with clinical expectations.

B.3.3.4 Time on treatment

Time-to-off-treatment (TTOT) data from the GO29365 study were mature, as the Pola+BR and BR arm comprised of treatment for up to 6 cycles only. TTOT KM estimates were therefore used directly in the model base case, using separate curves for each medicine in the respective regimens. The TTOT KM plots for Pola+BR and BR are presented in Figure

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 95 of 144

26. For the scenario comparing Pola+BR to R-GemOx, 3 cycles of R-GemOx were assumed, based on the assumption used in TA567 (70).

Figure 26. Time to off-treatment KM plots (GO29365)

==> picture [443 x 245] intentionally omitted <==

BR, bendamustine + rituximab; KM, Kaplan-Meier; Pola+BR, polatuzumab + bendamustine + rituximab; TTOT, time-to-off-treatment

B.3.3.5 Adverse events

For Pola+BR and BR, treatment-related AEs of CTCAE Grade 3 or greater from GO29365 that were deemed to be serious were included in the model (data cut-off, April 2018). Serious AEs were defined as those that would require NHS resources to treat them.

The type and frequency of AEs experienced with R-GemOx treatment were derived from Grade 3–5 AEs affecting >5% of patients in a Phase II study on the treatment of R/R DLBCL patients with R-GemOx (58).

Duration of AE data were sourced primarily from GO29365 and also TA306 (67). If duration data were not available from either of these two sources, then the longest duration of an AE from GO29365 was selected (72 days).

Disutilities and costs were applied for each AE to the relevant arm (see Sections B.3.4.4 and B.3.5.6, respectively).

Treatment-related AEs included in the model, their incidence for each arm and duration (and associated source) are reported in Table 45.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 96 of 144

Table 45. Incidence of treatment-related AEs included in the model (CTCAE ≥Grade 3, serious)

a‘Maximum’ duration indicates equivalence to the longest AE duration from GO29365. AE, adverse event; BR, bendamustine + rituximab; CTCAE, Common Terminology Criteria for Adverse Events; Pola+BR, polatuzumab + bendamustine + rituximab; R-GemOx, Rituximab + gemcitabine + oxaliplatin

B.3.4 Measurement and valuation of health effects

Health-related quality-of-life (HRQoL) data for the model health states were based on values identified in the literature (see Section B.3.4.3), as HRQoL data were not collected in GO29365.

B.3.4.1 HRQoL data from clinical trials

EuroQol Five Dimension (EQ-5D) data, or HRQoL data that could be mapped onto EQ-5D utility values, were not collected in the GO29365 study.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 97 of 144

B.3.4.2 Mapping

Mapping from an HRQoL scale to the EQ-5D was not feasible as no HRQoL data were collected in the GO2935 trial.

B.3.4.3 Health-related quality-of-life studies

An SLR was performed to identify studies reporting HRQoL and health state utility data in patients with DLBCL (for detailed methodology and results of the SLR, please see Appendix H). Seven studies reporting HRQoL or utility data in patients with relapsed or refractory disease were identified in the literature review; the results of these studies are presented in Table 46. All studies identified had some limitations with regards to applicability to the costeffectiveness model, as discussed in Table 46. An additional relevant study was identified after the searches had been performed in September 2018, which was Wang et al . 2018 (100). The results of this study are presented in Table 47.

Table 46. HRQoL and utility studies in R/R DLBCL identified in the SLR

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 98 of 144

==> picture [454 x 380] intentionally omitted <==

----- Start of picture text -----
to DLBCL. Source publications
relatively dated.
Knight 2004 NR Non- - EuroQoL-based utilities sourced from
(103) responders/ an unpublished data source, therefore
relapsed validity and reliability could not be
patients, assessed. Utility weights were
treatment sourced from a large UK community
with sample. Source publication relatively
CHOP: 0.38 dated.
R-CHOP:
0.38
Kymes 2012 NR NR Day before Utility values sourced from published
(104) transplant European studies of R/R NHL
(patients (Doorduijn 2005) and aggressive NHL
undergoing (van Agthoven 2001). Values not
apheresis): specific to DLBCL, or PFS/PD health
0.75 states. Source publications relatively
14 days post- dated.
transplant
(during high-
dose
chemotherap
y and
engraftment):
0.53
3 months
post-
transplant
(post
engraftment):
0.78
----- End of picture text -----

3L, 3-level; 5L, 5-level; AE, adverse event; CAR-T, chimeric antigen receptor-T cell; CR, complete response; (R-) CHOP, (rituximab-) cyclophosphamide, doxorubicin, vincristine, prednisolone; DLBCL, diffuse large B cell lymphoma; EQ-5D, Euro-QoL-5 Dimensions; HL, Hodgkin’s lymphoma; NHL, non-Hodgkin’s lymphoma; NR, not reported; Pola+BR, polatuzumab + bendamustine + rituximab; PD, progressed disease, PFS, progression-free survival; R/R, relapsed/refractory; US, United States

Table 47. HRQoL and utility results from Wang et al. 2018

3L, 3-level; 5L, 5-level; EQ-5D, Euro-QoL-5 Dimensions; PD, progressed disease, PFS, progression-free survival; UK, United Kingdom

B.3.4.4 Adverse reactions

As HRQoL data were not collected in GO29365, the impact of treatment-related AEs was modelled by applying disutilities derived from previous NICE appraisals in R/R DLBCL (67, 69) and brentuximab vedotin in R/R systemic anaplastic large cell lymphoma (TA478) (105). Table 48 presents the disutilities included in the model. As discussed in Section B.3.3.5,

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 99 of 144

treatment-related AEs of CTCAE Grade 3 or higher that were deemed to be serious were included in the model. Disutilities from AEs for the respective treatments were applied in the model as a weighted average value, with the estimated disutilities (Table 48) being weighted by their corresponding incidence and duration as outlined in Table 45.

Table 48. Disutility values used in the cost-effectiveness model

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 100 of 144

AE, adverse event

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

In the base case, the recent health state utility values used by the manufacturer in TA559 for PFS and PD were adopted (Table 49) (69). These values were sourced from EQ-5D-5L data captured in the ZUMA-1 study and cross-walked to -3L values, and thus align with the NICE reference case and position statement on the use of the EQ-5D-5L valuation set for England (110). The patient population of ZUMA-1 may be considered similar to that of GO29365; ZUMA-1 contained a subgroup of R/R DLBCL patients (PMBCL and TFL histologies were also included in the trial), the majority of patients had an ECOG performance status of 1 and had received three or more lines of therapy (111).

These values are more conservative than the majority of values identified in the SLR, and the PFS value may be considered to possess face validity given it is below the average utility value for the general population (0.79) (112) at the average baseline age of patients in GO29365.

In agreement with the assumptions adopted in TA559 and TA567, in the base case, patients who have remained in the PFS state for two years revert to age- and sex-matched general population utilities for the UK, which were based on Ara and Brazier 2010 (112). This assumption aligns with clinical expert feedback on the natural history of R/R DLBCL and evidence from Maurer et al. 2014 (93) (as discussed in Section B.3.3.1), that patients who achieve sustained remission for up to two years are considered to experience long-term survival aligned to that of the general population. It is therefore assumed that a similar utility to the general population is accrued in these patients.

Scenario analyses performed with respect to utilities are presented in Table 50. The PFS and PD health state utilities used in TA306 and TA567 were both explored in scenarios. In addition, to explore uncertainty around the time point at which patients are considered to be in long-term remission, application of age- and sex-adjusted general population utility was applied to patients in the PFS state after five years, instead the two-year time point used in the base case. Finally, a scenario was performed in which a decrease in utility in the three months before death was implemented to capture the decline in utility before the end of life. The utility value for this final scenario were sourced from Färkkilä et al. 2014 (113).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 101 of 144

AE-related disutilities were applied by treatment arm, as described in Section B.3.4.4.

Table 49. Summary of utility values for cost-effectiveness analysis (base case)

AE, adverse events; DLBCL, diffuse large B-cell lymphoma EQ-5D: EuroQoL 5 Dimensions; PD, progressed disease; PFS, progression-free survival; R/R, relapsed refractory

Table 50. Summary of utility values for cost-effectiveness analysis (scenario analyses)

PD, progressed disease; PFS, progression-free survival

B.3.5 Cost and healthcare resource use identification,

measurement and valuation

An SLR was conducted to identify data to inform relevant costs and resource use associated with the treatment of patients with R/R DLBCL.

Full details of the SLR search strategy and process for study selection is reported in Appendix I. Of 235 unique records, the SLR identified a single study by Wang et al. 2017 that met the inclusion criteria (116).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 102 of 144

This study reports UK population-based treatment costs for DLBCL patients from a simulation model based upon clinical data from the UK Haematological Malignancy Research Network (HMRN) database. Parameter input costs for the model were derived from NHS reference costs and the unit costs of chemotherapy according to data obtained from the Leeds Teaching Hospital NHS Trust.

B.3.5.1 Costs included in the model

The economic analysis was conducted from the NHS and PSS perspective, with appropriate unit cost sources such as NHS reference costs (2017–18) , British National Formulary (BNF) online (accessed June 2019) and electronic Marketing Information Tool (eMIT) (accessed June 2019) used to inform model cost inputs (97, 98, 117).

The following resource use and cost elements were included for the PFS and PD health states present in the model:

• PFS: drug acquisition and administration, treatment-related AEs and routine supportive care (professional and social services, health care professionals and hospital resource use, treatment follow-up) and subsequent treatment costs

• PD: drug acquisition and administration (for further interventions received), and supportive care (professional and social services, health care professionals and hospital resource use, treatment follow-up) and subsequent treatment costs

The assumptions used for deriving the resource use and costs for supportive care in both PFS and PD health states were aligned with those specified in the previous relevant submissions TA306 (67) and TA559 (69). Based upon the ESMO guidelines recommending routine follow-up of up to 24 months (9), it is assumed that patients remaining in PFS for two years would be discharged and therefore would not incur the further supportive costs that are associated with DLBCL.

B.3.5.2 Intervention and comparators’ costs and resource use

Drug acquisition costs and cost per cycle for all interventions in the model are presented in Table 51.

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 103 of 144

Table 51. Drug acquisition costs for Pola+BR, BR and R-GemOx

aCalculated from vial combinations required to match the GO29365 patient dose distribution, informed by weight and BSA;[b] Vial size available in xxxxxxxxxx[c] Polatuzumab vedotin, planned list price;[d] Dosing source, GO29365; eCost source, BNF 2019; fAssumed discount of 50% applied, based on national tendering process for biosimilar rituximab;[g] Cost source, eMIT 2019;[h] Dosing source, Mounier et al. 2013 (58). BNF, British National Formulary; BR, bendamustine + rituximab; BSA, body surface area; Pola+BR, polatuzumab + bendamustine + rituximab; R- GemOx, rituximab + gemcitabine + oxaliplatin

Pola+BR drug acquisition costs and dose calculations

Drug acquisition costs and cost per cycle for Pola+BR are presented in Table 51. For the Pola+BR regimen, patients were assumed to receive up to six cycles (21 days per cycle) of Pola+BR, administered at mean doses of 1.8 mg/kg for polatuzumab vedotin and 375 mg/m[2] for rituximab (both on day 1 of each cycle), with 90 mg/m[2] of bendamustine administered on days 1 and 2 of each cycle. The mean treatment doses were derived from the weight and body surface area (BSA) distribution of patients enrolled in the GO29365 study.

It is planned for polatuzumab vedotin to be available in 140 mg and 30 mg vials (lyophilised product prepared for reconstitution prior to infusion). Due to earlier than anticipated

marketing authorisation (expected xxxxxxxxxxxxxx), polatuzumab vedotin will initially be available only with a 140 mg vial size at a list price of xxxxxxxxxx per vial. The 30 mg vial is in development and is planned to be available at an equivalent per mg price (xxxxxxxxx per 30 mg vial) in xxxxxxxxx.

The use of the 140 mg vial alone prior to the availability of the 30 mg vial could initially create waste for individual NHS Trusts due to a lack of flexibility in vial sizes to tailor the

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 104 of 144

dose to patients’ individual weights. In consultation with NHS compounding service providers, Roche is planning to put arrangements in place so hospitals can obtain bags ready for infusion with the correct patient-specific dosing from these service providers without incurring any wastage costs. Trusts would therefore only be charged on a per mg basis for the drug acquisition costs, resulting in a ‘no waste’ or ‘full vial sharing’ scenario. The use of compounders is already common practice for other chemotherapies in an increasing number of NHS Trusts. Upon availability of the 30 mg vial, it is envisaged that NHS Trusts will be able to prepare doses in-house, incurring minimal wastage. Details of the compounding arrangements for polatuzumab vedotin are being discussed with NHS England.

Based on the above, costs per cycle in the model base case were therefore calculated based on the availability of 140 mg and 30 mg vials under the conservative assumption of ‘no vial sharing’, representing the way in which polatuzumab vedotin will be supplied the long-term. Based on the weight distribution of patients enrolled in the GO29365 study, a mean weight of 74.86 kg resulted in a mean per cycle dose of 143.9 mg polatuzumab vedotin at an average cost of xxxxxxxxxx per cycle.

A further scenario was also included for completeness, representing the use of 140 mg vials only, with no vial sharing.

Rituximab is available as a biosimilar at a list price of £157.17 for the 100 mg vial and £785.84 for the 500 mg vial (Rixathron[®] /Truxima[®] , BNF 2019 (97)). For the economic analysis base case, an estimated discount of 50% was applied to the biosimilar rituximab list price, based on the national tendering process for rituximab biosimilar medicines (precise discount values are kept in confidence by the NHS). In the model base case, the rituximab dose is calculated based on the BSA distribution of the GO29365 patient cohort. Patients were assumed to receive a dose of 375 mg/m[2] of rituximab administered on day 1 of each cycle. Assuming no vial sharing, the average cost per cycle for rituximab was calculated to be £581.52.

Bendamustine is now available as a generic formulation in vials of 25 mg and 100 mg at a cost of £6.85 and £27.77 per vial respectively (BNF 2019 (97)). Patients were assumed to receive a dose of 180 mg/m[2] per cycle (90 mg/m[2] on days 1 and 2 of the cycle) based on the BSA distribution of the GO29365 patient cohort. Assuming no vial sharing, the cost per cycle for bendamustine was calculated to be £95.95.

Comparator drug acquisition costs and dose calculations

Drug acquisition costs for BR are presented in Table 51, with calculations for per cycle cost

for bendamustine and rituximab the same as those specified for Pola+BR. Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 105 of 144

Drug acquisition costs for R-GemOx are presented in Table 51. In the treatment regimen, gemcitabine and oxaliplatin were assumed to be administered on day 1 of each cycle (14 days per cycle) at doses of 1,000 mg/m[2] and 100 mg/m[2] , respectively, as reported by Mounier et al. (58). Based on an assumption of no vial sharing, an average cost per cycle was calculated at £17.84 for gemcitabine and £13.87 for oxaliplatin, based on the BSA distribution of the GO29365 patient cohort.

B.3.5.3 Drug administration costs

Administration costs for chemotherapy included in the model are presented in Table 52, with the unit cost per resource as reported in the NHS reference cost schedule 2017–18.

Pharmacy costs for the preparation of IV infusions were not considered separately in previous TAs in R/R DLBCL (67, 69, 70), likely on the basis that there is no unbundled NHS tariff to cover pharmacy service costs in relation to the preparation of IV infusions. In this analysis it was assumed that preparation of each cycle of a regimen containing polatuzumab or rituximab required 39 minutes of pharmacy time, as estimated in a UK-based time and motion study of rituximab in non-Hodgkin’s lymphoma (118). An hourly cost for a hospital pharmacist is £48 (119), resulting in a per cycle cost of £31.20.

Table 52. NHS reference costs 2017–18 for chemotherapy administration

aNHS Improvement. NHS Reference Cost Schedule, 2017–18. HRG, healthcare resource group

The total per cycle drug administration costs for the Pola+BR, BR and R-GemOx treatment regimens are summarised in Table 53. For Pola+BR, the same administration tariff costs are applied up to a maximum of the first six cycles (as determined by the TTOT KM estimate data). Administration tariff costs for BR are separated into administration costs for the first administration (first cycle) and subsequent administrations (subsequent cycles).

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 106 of 144

Table 53. Drug administration costs per cycle

aNHS Improvement. NHS Reference Cost Schedule, 2017–18. BR, bendamustine + rituximab; HRG, healthcare resource group; Pola+BR, polatuzumab + bendamustine + rituximab; R-GemOx, rituximab + gemcitabine + oxaliplatin

Expected costs per treatment cycle were calculated using the total administration cost per cycle and TTOT KM estimate data (Section B.3.3.3).

B.3.5.4 Health-state unit costs and resource use

The type and frequency of resource utilisation in the PFS and PD health states is based upon data from the manufacturer’s submission for TA306, which were derived from questionnaire responses from a set of UK physicians selected based upon publication record in the field of aggressive non-Hodgkin’s lymphoma (NHL), prior collaboration, and referrals from other physicians (67). The resources listed consist of three separate categories (professional and social services, healthcare professionals and hospital resource use, and treatment follow-up). Table 54 presents the cost per unit for each type of resource included in the model, whilst Table 55 presents the annual frequency of resource use in each health state. Where required, resource use frequency per model cycle was calculated from the 28day frequency values as below:

==> picture [290 x 75] intentionally omitted <==

Table 54. Supportive care resource use unit costs included in the model

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 107 of 144

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 108 of 144

aNHS Improvement. NHS Reference Cost Schedule, 2017–18. CT, computed tomography; GP, General Practitioner; LDH, lactate dehydrogenase test; ECG, electrocardiogram; MRI, magnetic resonance imaging; MUGA, multiple-gated acquisition scan; PET-CT, positron emission tomography–computed tomography; PD, progressed disease; PSSRU, Personal Social Services Research Unit; R, rituximab

Resource use was assumed to be the same for both arms, in accordance with clinical expert opinion (35). Clinical expert opinion also considered that patients remaining in PFS for longer than two years were in long-term remission, and it was therefore assumed that no additional supportive costs were incurred beyond this time point (9).

Based on the unit costs and the annual frequencies presented above, the average per cycle supportive care costs for each health state were calculated (Table 56) as shown below:

Per cycle supportive care cost = Per cycle frequency ∗Resource unit cost

For the PFS health state, resource use was specified for patients whilst they were on- or offtreatment.

Table 55. Annual frequency of resource use in PFS and PD

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 109 of 144

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 110 of 144

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 111 of 144

One-off costs, PD (Proportion of patients requiring resource)[b]

aTA306 (67). bOne-off costs weighted by the proportion of patients requiring the respective resource. BR, bendamustine + rituximab; CT, computed tomography; ECG, electrocardiogram; ERG, Evidence Review Group; GP, General Practitioner; LDH, lactate dehydrogenase test; MRI, magnetic resonance imaging; MUGA, multiplegated acquisition scan; PD, progressed disease; PET-CT, positron emission tomography–computed tomography; PFS, progression-free survival; Pola + BR, polatuzumab + bendamustine + rituximab; R-GemOx, rituximab + gemcitabine + oxaliplatin

Table 56. Per cycle supportive care costs for PFS and PD heath states

PD, progressed disease; PFS, progression-free survival

B.3.5.5 Subsequent treatment costs

As discussed in Section B.1.3.3, a small number of third-line and beyond options exist for R/R DLBCL patients, including post-treatment SCT, CAR-T therapy and palliative care.

With regards to post-treatment SCT, GO29365 was not designed to investigate Pola+BR or BR as a salvage regimen for potential transplant candidates, and the comparator regimens identified in the scope are similarly unlikely to be considered as salvage therapies prior to

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 112 of 144

SCT in clinical practice, as per NICE and ESMO guidance (see Section B.1.3.2). In GO29365, each treatment arm contained only a single patient who received a transplant (2.5% in each arm), both of whom were from the same treatment centre, indicating that subsequent treatment with SCT was not a widespread treatment choice for patients. It is therefore not expected that a significant proportion of patients who meet the decision problem would proceed to transplant after Pola+BR or BR treatment in UK clinical practice. Accordingly, costs for post-treatment SCT were not included in the base case model.

In terms of post-treatment with CAR-T therapy, similarly to SCT, GO29365 was not designed to investigate either Pola+BR or BR as a bridging regimen to CAR-T therapy, although patients could potentially receive CAR-T treatment after progression. An imbalance between the use of curative treatments between treatment arms could potentially bias OS survival estimates. In the Pola+BR treatment arm two patients received CAR-T therapy, one of whom subsequently died, whereas no patients received CAR-T therapy in the BR arm.

The influence of treatments with curative intent (such as CAR-T or SCT) was explored by comparing the GO29365 ITT patient population with a population censored for patients who had received SCT or CAR-T therapies at the time this was received. No difference between the ITT population and the censored population was observed (Figure 27) indicating that OS in the patient population was not affected by post-progression treatments in either arm.

Figure 27. Overall survival for ITT patient population and population censored for those receiving a treatment with curative intent (SCT or CAR-T)

==> picture [443 x 245] intentionally omitted <==

BR, bendamustine + rituximab; CAR-T, chimeric antigen receptor-T cell; ITT, intention-to-treat; OS, overall survival; Pola-BR, polatuzumab + bendamustine + rituximab; SCT, stem cell transplant

Company evidence submission template for ID1576: Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma. © Roche Products Ltd. (2019). All rights reserved Page 113 of 144

As CAR-T therapies are currently funded by the Cancer Drug Fund (CDF), they are not considered as part of standard NHS clinical practice. Accordingly, post-treatment CAR-T costs were not included in the base case model.

In study GO29365, the majority of patients in the randomised phase xxxx), did not receive any subsequent therapy after Pola+BR or BR. Of those receiving treatment, the majority received chemotherapy with or without rituximab (xxxxxxxxxxxxxxxxxxxxxxxxx). For the purpose of the economic analysis, the subsequent treatment costs for patients in PD who come off of Pola+BR or BR treatment, were estimated based on the proportion receiving chemotherapy, chemotherapy with rituximab, rituximab alone or radiotherapy. Other regimens, which included investigative treatments or SCT/CAR-T, were not costed. Based on clinical opinion (35) the base case assumes the same subsequent treatments are given in both arms, and pooled estimates across arms from GO29365 for the proportion of patients receiving treatments in the aforementioned categories were used. For the cost of chemotherapy with or without rituximab, the costs of three cycles of GemOx with and without rituximab were assumed, as chemotherapies are available as generic medicines, and costs of different regimens are broadly similar. A weighted average cost was calculated as shown in Table 57. The total cost of subsequent treatments was applied as a one-off cost at the time point of progression in the model.

Table 57. Subsequent treatment costs based on GO29365 data