==> picture [193 x 21] intentionally omitted <==

Single Technology Appraisal

Nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small-cell lung cancer [ID3757]

Committee Papers

Committee papers

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

SINGLE TECHNOLOGY APPRAISAL

Nivolumab with chemotherapy for neoadjuvant treatment of resectable

non-small-cell lung cancer [ID3757]

Contents:

The following documents are made available to stakeholders:

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

1. Company submission from Bristol Myer Squibb

2. Clarification questions and company responses

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

• a. Roy Castle Lung Cancer Foundation

• b. Royal College of Pathologists

4. External Assessment Report prepared by Newcastle TAR

• a. EAG report

• b. EAG report addendum

5. External Assessment Report – factual accuracy check

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

[Insert footer here]

2 of 2

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

Single technology appraisal

Nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer [ID3757]

Document B

Company evidence submission Submitted by Bristol Myers Squibb Pharmaceuticals, Ltd.

20 October 2022

Company evidence submission template for Nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Contents

B.2.12 Ongoing Bristol Myers Squibb nivolumab randomised controlled trials ......... 73

Company evidence submission template for Nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Tables

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Abbreviations

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

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

B.1.1 Decision problem

This submission covers the full marketing authorisation for nivolumab in combination with platinum doublet chemotherapy (nivolumab + PDC) for the neoadjuvant treatment of resectable (tumours ≥ 4 cm or node positive) non-small cell lung cancer (NSCLC) in adults.[1] The company submission is consistent with the final National Institute for Health and Care Excellence (NICE) scope and the NICE reference case (Table 1).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 1. The decision problem

• EFS = event-free survival; DFS = disease-free survival; N/A = not applicable; NSCLC = non-small cell lung cancer; pCR = pathologic complete response; PD-1 = programmed cell death protein 1; PDC = platinum doublet chemotherapy; PD-L1 = programmed death-ligand 1.

• a Tumour resectability is assessed at diagnosis and again after the administration of a neoadjuvant treatment but before surgery. Therefore, some patients may be deemed potentially eligible for resection at diagnosis, but their resectability status may change before surgery. In the remainder of the document, potentially resectable is referred to as resectable when relating to neoadjuvant treatment.

• b Although in the NICE scope, NICE confirmed at the checkpoint meeting that atezolizumab is no longer a relevant comparator because it is only recommended for use within the Cancer Drugs Fund, not in routine commissioning and it is not therefore included in this submission.[2 ]

• c The original study design also included a study arm with nivolumab + ipilimumab, which was stopped in a protocol revision and is not relevant to this appraisal. This decision was based on evidence from the metastatic setting and external data from the NADIM trial, which demonstrated a more promising pCR benefit for nivolumab + PDC than seen for nivolumab + ipilimumab in the NEOSTAR trial. Therefore, the clinical development of nivolumab + PDC was prioritised in CheckMate 816. The population presented in this submission is the concurrently randomly assigned population, described as the “intention-to-treat” population in the remainder of this document.

B.1.2 Description of the technology being evaluated

As summarised in Section B.1.1, this appraisal is for nivolumab + PDC for adults with resectable (tumours ≥ 4 cm or node positive) NSCLC in the neoadjuvant setting.

Nivolumab + PDC (cisplatin with either gemcitabine or pemetrexed, or carboplatin with paclitaxel) for the neoadjuvant treatment of adults with resectable NSCLC received marketing authorisation in the United Kingdom (UK) on 16 August 2022.1 It has been compared with PDC (cisplatin with either gemcitabine, pemetrexed, vinorelbine, or docetaxel, or carboplatin with paclitaxel) in the CheckMate-816 clinical trial in adults with resectable IB-IIIA (American Joint Committee on Cancer [AJCC]/Union for International Cancer Control [UICC] seventh edition) NSCLC (Table 2).[3]

Table 2. Technology being evaluated

• MHRA = Medicines and Healthcare Products Regulatory Agency; NSCLC = non-small cell lung cancer; PAS = patient access scheme; PD-1 = programmed cell death protein 1; PDC = platinum doublet chemotherapy; PD-L1 = programmed death-ligand 1; PD-L2 = programmed death-ligand 2; UK = United Kingdom.

• a Cost of a course of nivolumab + PDC at list price based on 3 cycles of therapy as received in the CheckMate816 trial per protocol for all event-free patients.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Nivolumab is an immuno-oncology agent that acts to restore the body’s natural antitumour response by inhibiting the suppression that the tumour exerts on antitumour immune responses. Programmed cell death protein 1 (PD-1) is an immune checkpoint involved in T-cell differentiation and function. PD-1 is specifically involved in inhibiting T-cell destruction of healthy “self-cells” at the effector (later) stage of the immune response. Tumour cells can exploit this pathway by upregulating proteins (i.e., programmed death-ligand 1 [PD-L1], programmed death-ligand 2 [PD-L2]) that engage PD-1 to limit the activity of T cells at the tumour site.8 Used before surgery, nivolumab-based regimens are expected to help prime the body’s immune response not only to target primary tumour cell activity before surgery and promote responses against micro-metastases already present but to kill tumour cells released during surgery, limiting recurrence.[9]

Evidence from animal models, supported by clinical evidence in some tumours, suggests that surgery and the body’s response to the associated trauma can be a trigger for the development of metastases.[9] First, although cells from the primary tumour are present in the blood stream in most patients and are generally rapidly destroyed, the disruption of the primary tumour that occurs during surgery may result in increasing levels of circulating tumour cells by as much as 10-fold.[9]

Second, surgery has been found to reduce natural killer cell cytotoxic activity and impair macrophage functioning, both of which can promote the development of metastases.[9] Other perioperative factors such as the use of opioids for pain management, blood transfusions, hypothermia during surgery, and the effects of anaesthetics may also have immunosuppressive effects that reduce the body’s antitumour activity.[9]

Third, the acute inflammatory response to surgery has been found to provide conditions which increase the capture of tumour cells in locations which favour their growth. For example, in response to tissue injury, neutrophils form neutrophil extracellular traps which capture the tumour cells, thus promoting their growth.[9] Taken together, priming the antitumour activity of the immune system before surgery could decrease the presence of tumour cells in the blood stream and encourage an antitumour immune response to help reduce the risk of both local and distant recurrence after surgery.[9]

A further theoretical rationale for use of immuno-oncology agents in the neoadjuvant setting is that they may be more effective in the presence of a macroscopic tumour burden due to higher levels of endogenous tumour antigen present in the primary tumour that enhance T-cell priming.[10] Thus, the effects of immuno-oncology agents to promote the antitumour immune response are expected to be optimal when initiated before surgery because they would help overcome the adverse effects of surgery related to micro-metastatic spread and immunity impairment, thereby overcoming important limitations of current treatment approaches. On this note, preclinical models support better outcomes with neoadjuvant immuno-oncology therapy than adjuvant immuno-oncology therapy.[11]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

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

B.1.3.1 Disease background

Lung cancer is the second most common cancer worldwide, and is the most frequent cause of cancer deaths both worldwide and in the UK.[12,13]

There are 2 major groups of lung cancer that differ based on histology: NSCLC (80%-85% of lung cancers) and small cell lung cancer (15%-20% of lung cancers).[14] NSCLC is divided into 2 main histological subtypes: squamous cell carcinoma (25%-30%) and nonsquamous NSCLC (75%, primarily composed of adenocarcinoma [~40%] and large cell carcinoma [~5%-10%]).[14-16] A few other subtypes of NSCLC, such as adenosquamous carcinoma and sarcomatoid carcinoma, are much less common.[15]

NSCLC can also be categorised according to the presence of molecular markers; the patients with these mutations can be treated with targeted therapy and are generally not considered for immuno-oncology therapy in resectable disease.[17] These biomarkers include epidermal growth factor receptor (EGFR), BRAF, Kirsten rat sarcoma virus (KRAS), anaplastic lymphoma kinase (ALK), ROS proto-oncogene 1 (ROS1), and neurotrophic tyrosine receptor kinase (NTRK).[18-29]

The AJCC/UICC staging system for NSCLC classifies patients at diagnosis into stages of disease that predict survival outcomes. It is based on 3 features of the tumour:

• Size and extent of the main tumour (T)

• Spread to nearby lymph nodes (N)

• Spread (metastasis) to distant sites (M)

Numbers or letters after T, N, and M provide more details about each of these factors, describing the size of the tumour and how far it has spread. Within the T, N and M stages, higher numbers mean the cancer is more advanced. Letters are used as modifiers or to provide further parameters, such as the grade of the tumour, any invasion present, completeness of the operation, timing or method used to determine the tumour stage.

The seventh edition of the AJCC/UICC TNM staging system defines 9 stages of disease, from occult carcinoma to metastatic disease. In this classification, stages I-II are considered to correspond to early disease, whereas locally advanced disease corresponds to stage III. A broader category, non-metastatic disease, is considered to include stages I-III and includes all stages in which distant metastases are not present (i.e., M category is M0).

The eighth edition of the AJCC/UICC TNM staging system includes revised definitions for some of the stages and the addition of stage IIIC; broadly, tumour size in the eighth edition is generally smaller than that in the same stages of the seventh edition. Table 3 compares these editions of the AJCC/UICC.

In the CheckMate-816 trial, the patient inclusion criteria are based on the AJCC/UICC seventh edition criteria with included patients being required to have stage IB (with tumour size ≥ 4 cm) to IIIA disease. This largely corresponds to stages IB (with tumour size 4 cm) to

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

IIIB (non-N3 and non–N2-T4) in the eighth edition. The changes that are relevant to the nivolumab indication are as follows: some patients within IB are reclassified as IIA depending on tumour size, all IIA patients are reclassified as IIB, some patients within IIB are reclassified as IIIA depending on tumour size, and some patients within the IIIA category are reclassified as IIIB. Note that the seventh edition is used when referring to CheckMate-816 patients throughout this appraisal; patients have not been reclassified into the eighth edition.

Table 3. Summary of revisions from the seventh edition of the AJCC/UICC staging system to the eighth edition

AJCC = American Joint Committee on Cancer; UICC = Union for International Cancer Control.

Notes: Properties of the tumour are indicated as T = size and extent of the main tumour; N = spread to nearby lymph nodes; M = metastasis to distant sites.

0-4 denotes increasing severity of T, N, or M. Bold text indicates differences between the seventh and eighth editions of the AJCC/UICC TNM staging system.

Sources: Detterbeck et al. (2009)[30] ; Goldstraw et al. (2016)[31]

B.1.3.2 Diagnosis

Most lung cancers are diagnosed at an advanced stage when the cancer has spread to the lymph nodes and other organs in the chest (locoregional disease; stage III) or to other parts of the body (metastatic disease; stage IV).[32] Of all lung cancer cases, 26% were diagnosed at stages IB-IIIA in England in 2019.[i][,33,34]

i Note, although 2020 data are available from the NLCA, given the impact of COVID-19 on diagnosis and treatment in 2020, the 2019 data are presented and used in this submission. 2017 data are also used in certain cases where it is necessary to ensure alignment with the seventh edition of the AJCC/UICC TNM staging system used in the CheckMate-816 trial.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 1. Stage distribution of lung cancer for 2017 in England and Wales

==> picture [301 x 201] intentionally omitted <==

AJCC = American Joint Committee on Cancer; UICC = Union for International Cancer Control.

Note: Data are from the 2018 audit (2017 data) to ensure alignment with the seventh edition of the AJCC/UICC TNM staging system used in the CheckMate-816 trial.

Source: Royal College of Physicians (2018)[34]

B.1.3.3 Prevalence and incidence

Approximately 35,000 people were diagnosed with lung cancer in England and Wales in 2019; of these, 29,481 had NSCLC in England and 7,665 were diagnosed at stages IB-IIIA and, therefore, were potentially eligible for curative resection.[33,34][ii]

B.1.3.4 Mortality and survival

Lung cancer is the most common cause of cancer death worldwide and in the UK (Figure 2); the percentage of patients who die from lung cancer is similar to that of patients who die from prostate cancer, colon cancer, and breast cancers combined (20.6%).[12,13]

ii Note: number of patients eligible for curative resection were calculated using proportions of patients diagnosed by stage from the 2018 audit (2017 data), which were applied to 2019 data. This is to ensure alignment with the seventh TNM staging edition used in the CheckMate-816 trial.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 2. Causes of cancer deaths in the United Kingdom in 2020

==> picture [426 x 415] intentionally omitted <==

CNS = central nervous system. Source: GLOBOCAN (2020)[35]

According to National Lung Cancer Audit (NLCA) data from England and Wales, median survival (by stage, AJCC/UICC eighth edition) in 2019 was not reached (and therefore was greater than 1 year) for all patients with stage I and II disease, whereas median survival for all patients with stage III and IV disease was 362 and 100 days, respectively.[33] The overall 1-year relative survival rate for NSCLC in 2019 was 46% in England and 42% in Wales.[33,36] The 1-year survival rates for lung cancer in England decreased with increasing stage: 87.7%, 73.0%, 48.7%, and 19.3% for stages I, II, III, and IV, respectively (2013-2017 data, TNM staging edition not listed).[37]

In England only, 21.1% of patients with lung cancer were alive at 2 years, and 11.3% at 3 years (2018 data, AJCC/UICC eighth edition).[iii,38] The 5-year survival rates for lung cancer

iii This is a different cohort of patients, including those in the 2014-2017 annual reports.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

in England decreased with increasing stage: 56.6%, 34.1%, 12.6%, and 2.9% for stages I, II, III, and IV, respectively (2013-2017 data, TNM staging edition not listed).[37]

Variation in overall survival (OS) is determined by the extent of disease at diagnosis. Figure 3 presents the range in 5-year OS according to the AJCC/UICC seventh edition. For patients potentially eligible for nivolumab + PDC neoadjuvant therapy, 5-year (60-month) OS ranged from 36% to 66% according to the seventh edition (i.e., stages IB to IIIA).[31]

Figure 3. Overall survival according to AJCC/UICC seventh edition staging system

==> picture [433 x 273] intentionally omitted <==

AJCC = American Joint Committee on Cancer; MST = median survival time; N = patient number; NR = not reached; UICC = Union for International Cancer Control. Source: Goldstraw et al. (2016)[31]

B.1.3.4.1 Disease progression and survival

The current prognosis of resectable NSCLC remains poor despite the curative intent of surgery. Risk of disease progression increases by tumour stage, while OS decreases by stage (and severity of metastases in advanced disease).[33,39] This has been demonstrated in a retrospective study performed in France, Germany and the UK that included 831 patients diagnosed with stage IB-IIIA NSCLC (AJCC/UICC seventh edition).[39] Over a median followup of 26 months, 33% of patients developed recurrence, and 24% progressed to metastatic disease.[39] Median disease-free survival (DFS), a measure of the risk of relapse, decreased with increasing disease stage (DFS in stage IB disease: not reached; DFS in stage IIA: 42.3 months; DFS stage III disease: 28.5 months). Therefore, optimising systemic treatment in the resectable setting is important to prevent and/or delay disease progression to metastatic disease, which is associated with worse survival outcomes, reduced healthrelated quality of life (HRQOL), and increased healthcare costs.[31,33,39-41]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Cure

Patients with resectable disease may be considered cured if there is no disease relapse at a defined timepoint; at this stage, patients are no longer followed up regularly, and risk of disease recurrence and risk of death are similar to people in the age-matched general population (this was confirmed by clinical experts; Appendix N).[42,43]

Several cancers may be considered cured if diagnosed and treated in the early stage of disease; for example, 75% of breast cancers can be cured if found at an early stage. Other cancers, such as colon cancer, prostate cancer, and pancreatic cancer, may be cured if they are detected and treated in stages I-II.[44-46] This is also the case in resectable NSCLC; longterm evidence suggests that some patients who are treated for resectable non-metastatic NSCLC may remain recurrence-free at 5 years and thus be considered cured (see Section B.3.3.3 for further detail on cure).

B.1.3.5 Morbidity

Fatigue, dyspnoea, pain, and cough are the most troublesome symptoms associated with non-metastatic disease and impact HRQOL.[47-49]

Disease recurrence, and progression to locoregional or distant metastatic disease, are associated with worsening symptoms, including those associated with specific sites of metastases. Bone metastases (occurring in 30%-40% of patients with lung cancer[50] ) and brain metastases (occurring in 30%-40% of patients with NSCLC[50] ) are associated with specific debilitating symptoms such as bone pain, risk of fracture, headache, seizures, and other neurological complications, all of which substantially impact HRQOL.[51-55] Therefore, optimising systemic treatment in the non-metastatic setting is important to prevent the worsening of symptoms and deterioration of HRQOL that occur during disease progression.

Resectable NSCLC does not only affect the patient; caregivers for patients with NSCLC also experience a considerable burden associated with care.[40,41] Improved treatment options may therefore also help reduce the burden on caregivers.

B.1.3.6 Clinical pathway of care

Treatment options for patients with newly diagnosed, potentially resectable NSCLC are determined based on both the stage of disease and operability of the patients. NICE (2019)[17] recommends surgery, radiotherapy, chemotherapy, or a combination of these for resectable disease (Figure 4).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 4. Treatments used for resectable NSCLC in clinical practice in England

==> picture [432 x 302] intentionally omitted <==

NSCLC = non-small cell lung cancer; PDC = platinum doublet chemotherapy; PD-L1 = programmed death-ligand 1. a Atezolizumab was recommended for use within the Cancer Drugs Fund in August 2022. It is an option for adjuvant treatment after complete tumour resection in adults with stage II-IIIa NSCLC with PD-L1 ≥ 50% whose disease has not progressed after adjuvant PDC, but is not considered a comparator in this submission.[2]

Source: NICE (2019)[17]

Surgical resection is the standard of care (SOC) for most eligible patients with resectable NSCLC.[17] In addition to surgery, the patient may also receive neoadjuvant or adjuvant treatment with the aim of improving long-term patient outcomes, although options are limited.[17] Current neoadjuvant and adjuvant chemotherapy regimens offer only modest benefits to patients[56-60] and many patients choose not to have adjuvant chemotherapy because of these modest improvements (i.e., an absolute improvement in 5-year OS of approximately 5%[56,61] ), while wanting to avoid any toxicities, or are simply not fit enough to tolerate chemotherapy following surgery (Appendix N). Notably, chemotherapy is only recommended by NICE in the adjuvant setting.

Chemoradiotherapy (chemotherapy in combination with radiotherapy) followed by surgery may be considered for use in patients with stage IIIA-N2 disease, although such patients only comprise approximately 7% of all patients with NSCLC[62] and a smaller proportion of these are resectable; clinical experts have also confirmed that this subpopulation is small, and that only 7%-8% patients eligible for resection are treated with chemoradiotherapy (Appendix N).

Immuno-oncology therapies are being evaluated for the treatment of resectable NSCLC in the adjuvant or neoadjuvant settings, and have shown strong promise in other early-stage disease settings, and long-term effectiveness in metastatic NSCLC.[63] Atezolizumab after

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

adjuvant PDC has recently been reviewed by NICE for NSCLC and is now recommended for use in the Cancer Drugs Fund while additional data are collected.[2]

When immuno-oncology treatments (such as nivolumab) are initiated before surgery, they are designed to help prime the body’s immune response not only to target primary tumour cell activity before surgery and promote responses against micro-metastases already present but to kill tumour cells released during surgery, limiting recurrence.[9]

Based on the above, there is a clear unmet need for an active treatment that can effectively prevent disease recurrence and progression to metastatic disease. Importantly, metastatic disease is associated with worse survival outcomes, reduced HRQOL, and increased healthcare costs compared with non-metastatic disease.[31,33,39-41] An effective immunooncology therapy that can be used for all patients with stage IB-IIIA disease and that provides a clear treatment pathway with a short duration of treatment is expected to result in better adherence to the published NICE guidelines across English and Welsh National Health Service (NHS) trusts, and improved outcomes for patients with resectable NSCLC. The introduction of neoadjuvant nivolumab + PDC into the treatment pathway would provide a clear, evidence-based SOC therapy for patients with resectable, stage IB-IIIA NSCLC, with the potential to improve adherence to guidelines and clinical outcomes in England and Wales (Figure 5).

Figure 5. Potential position of nivolumab + PDC in the treatment pathway for resectable NSCLC in clinical practice in England and Wales

==> picture [432 x 302] intentionally omitted <==

NSCLC = non-small cell lung cancer; PDC = platinum doublet chemotherapy; PD-L1 = programmed death-ligand 1; SOC = standard of care.

a Atezolizumab was recommended for use within the Cancer Drugs Fund in August 2022. It is an option for adjuvant treatment after complete tumour resection in adults with stage II-IIIa NSCLC with PD-L1 ≥ 50% whose disease has not progressed after adjuvant PDC but is not considered a comparator in this submission.[2]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.1.3.6.1 Surgical treatment

Surgical resection is the SOC for most eligible patients with stage IB-IIIA NSCLC if the tumour is clinically assessed as being resectable and the patient is eligible for surgery.[17] Tumour resectability is assessed at diagnosis and again after the administration of a neoadjuvant treatment but before surgery.[17] Therefore, some patients may be deemed eligible for potential resection at diagnosis, but their resectability status may change before surgery.

Surgery may involve:

• Removal of a whole lung (pneumonectomy),

• Removal of a whole lobe (lobectomy), or

• Removal of part of a lobe—segmentectomy or wedge resection.

Surgery is associated with an immediate postoperative mortality of approximately 3% after lobectomy and 7% after pneumonectomy.[64,65,66] Lobectomy (a surgery to remove one of the lobes of the lungs, either via open surgery or thoracoscopy [a type of video-assisted surgery that is considered less invasive]) is offered to eligible patients.[67] Hilar and mediastinal lymphnode sampling (for staging purposes) or ‘en bloc’ resection (to remove any local invasion) are performed in parallel with lobectomy for all patients undergoing surgery with curative intent. More extensive surgery (pneumonectomy [removal of an entire lung], bronchoangioplasty, bilobectomy) is performed only when needed to obtain clear margins; these surgeries are associated with worse HRQOL and mortality.[17,64,68] For patients with T3 NSCLC with chest wall involvement who are undergoing surgery, the aim is complete resection of the tumour using either extrapleural or en bloc chest wall resection.[17]

Given the risk of recurrence after surgery, neoadjuvant or adjuvant treatment may be considered. Neoadjuvant immuno-oncology treatment may optimise surgical outcomes and facilitates minimally invasive surgery (and shorten the duration of surgery), which, in turn, is associated with improved patient HRQOL.[3,69] For example, minimally invasive surgery rates are higher in patients treated with neoadjuvant immuno-oncology therapies such as nivolumab + PDC, versus patients treated with PDC alone (29.5% vs. 21.5%, respectively). Further, a higher proportion of patients treated with PDC alone have been shown to undergo pneumonectomy rather than lobectomy versus those treated with nivolumab + PDC (lobectomy, 60.7% vs. 77.2% %; pneumonectomy, 25.2% vs. 16.8%, respectively). Therefore, nivolumab + PDC treatment in the neoadjuvant setting may confer substantial surgical outcome benefits to patients with resectable NSCLC.

B.1.3.6.2 Neoadjuvant treatment

As discussed in Section B.1.3.6, neoadjuvant chemotherapy is not currently recommended by NICE outside a clinical trial,[17] although neoadjuvant and adjuvant chemotherapy regimens offer similar and only modest improvements in 5-year OS.[56-59]

Based on the NICE guidelines, chemoradiotherapy with surgery should be considered (with surgery scheduled 3-5 weeks after chemoradiotherapy) for patients with operable stage IIIA– N2 NSCLC who can have surgery and are healthy enough for multimodality therapy. For eligible patients with stage IIIA-N2 NSCLC who cannot tolerate or who decline

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

chemoradiotherapy (with or without surgery), radical radiotherapy (either conventional or hyperfractionated) should be considered.[17] However, patients with stage IIIA-N2 NSCLC account for only approximately 7% of patients with NSCLC,[62] which is confirmed by clinical experts (Appendix N). Further, stage IIIA-N2 NSCLC comprises a heterogeneous group of patients with both resectable and unresectable NSCLC; therefore, the number of patients with resectable stage IIIA-N2 NSCLC is likely to be limited further.

Patients are actively monitored for cancer recurrence. If the cancer comes back, treatment options and prognosis depend on the site of the recurrence (see Section B.1.3.4.1).

B.1.3.6.3 Adjuvant treatment

Based on the NICE guidelines, postoperative cisplatin-based combination chemotherapy should be:

• Offered to patients with good performance status (World Health Organization [WHO] 0 or 1) with T1a-4, N1-2, M0 NSCLC.

• Considered for patients with good performance status (WHO 0 or 1) and T2b-4, N0, M0 NSCLC with tumours greater than 4 cm in diameter.[17]

B.1.3.6.4 Real-world evidence on treatment of resectable lung cancer

Although the NICE guidelines provide guidance on the appropriate treatments for resectable lung cancer, data from the NLCA provide real-world evidence on the current treatment pathway in England. Based on data from 2019 (to ensure the impact of COVID19 on treatments is not considered), only 3,881 of 6,716 patients (58%) with stage IA-IIB and Performance Status (PS) 0-2 NSCLC (AJCC/UICC eighth edition) in England underwent surgical resection with curative intent in 2019.[33] Furthermore, according to clinical experts (Appendix N), approximately half of patients who undergo surgery go on to have adjuvant chemotherapy in the UK for a variety of reasons as described in Section B.1.3.6 above. Similarly, Felip et al. (2010)[70] assessed the treatment of patients with resectable NSCLC in Spain, Germany, Portugal, Sweden, and Switzerland and found that 34% of patients with NSCLC stages IB-IIIA who were allocated to receive adjuvant chemotherapy did not begin treatment.[70]

NLCA data from January 2017 to June 2018 showed that 15% of patients with stage IA to IIB NSCLC (TNM staging edition not listed) who did not have surgery declined surgery due to patient wishes, and 11% of patients did not have surgery due to their comorbidities. However, the reason for not having surgery was not documented in 75% of patients, and is thus unclear in most patients.[71]

According to the NLCA, in patients with stage IIIA disease and PS 0-2 (AJCC/UICC edition not listed), 11% underwent surgery alone and 12% underwent surgery followed by chemotherapy.[71] Of the remaining patients, 12% underwent concurrent chemoradiotherapy and 7% underwent sequential chemoradiotherapy (note, in addition, some patients only received best-supportive care or palliative therapies).[71] Although chemoradiotherapy and adjuvant chemotherapy are recommended for patients with stage IIIA NSCLC, there is currently no clear treatment pathway for these patients in NHS trusts across England.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Overall, only half of patients with stage IB-IIIA NSCLC are treated with any form of systemic anticancer therapy and it is unclear if these patients also have surgery (Table 4).

From the audit data, it is clear that there is no preferred treatment for patients with resectable NSCLC in England. For the basis of this appraisal, surgery alone, with some patients also receiving adjuvant platinum-based chemotherapy, most closely reflects clinical practice and is considered the most relevant comparator for this appraisal.

Table 4. Patients with stage IB-IIIA (AJCC/UICC eighth edition), PS 0-1 NSCLC who were treated with systemic anticancer therapy or targeted treatment in England in 2019

AJCC = American Joint Committee on Cancer; NSCLC = non-small cell lung cancer; PS = performance status; SACT = systemic anticancer therapy; UICC = Union for International Cancer Control. Source: Royal College of Physicians (2022)[33]

B.1.3.6.5 Goals of treatment

As with all cancers, improvement in OS is a key final endpoint of interest. However, because of the early nature of the disease and associated improved prognosis versus metastatic disease, OS data can be immature at the time of regulatory and health technology assessments for neoadjuvant therapies. At the time of this submission, CheckMate-816 does not have mature Kaplan-Meier OS data (see Section B.2.6.1.3). However, in resectable NSCLC, potential surrogate endpoints that are indicative of the survival benefit include pathologic complete response (pCR) and event-free survival (EFS).

An association between pCR and EFS, pCR and OS, and EFS and OS has been consistently demonstrated[72-75] as shown by the following studies conducted in patients with resectable NSCLC treated with neoadjuvant chemotherapy:

• CA2098Y9: a systematic literature review (SLR) and meta-analysis with the key objective of identifying studies that examined the association between pCR and EFS/OS and between EFS and OS and meta-analysing the magnitude of the association from the available literature.[72]

• CA2097 C4: a retrospective, observational real-world study that uses electronic health record data supplemented with chart review with the key objective of characterising the relationship between pCR and EFS/OS and between EFS and OS.[74]

• CA2097L8: a pooled meta-analysis of individual patient-level data from completed randomised controlled trials (RCTs) with the key objective of determining the magnitude of the association between pCR and EFS/OS and between EFS and OS.[73]

• An SLR and meta-analysis with the key objective of identifying studies that examined the association between EFS and OS following neoadjuvant therapy for NSCLC.[75]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

An association between pCR and EFS/OS and between EFS and OS for patients with resectable NSCLC treated with neoadjuvant chemotherapy was demonstrated (Table 5). Further, it has been found in several studies that patients who have a pCR have improved EFS and OS outcomes versus patients who do not achieve a pCR.[72,76-81]

Table 5. Analyses assessing association between pathologic complete response, event-free survival, and overall survival

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

CI = confidence interval; EFS = event-free survival; HR = hazard ratio; NR = not reached; OS = overall survival; pCR = pathologic complete response; RCT = randomised controlled trial; SLR = systematic literature review; US = United States.

Note: A trial-level analysis was conducted for studies CA2098Y9 and CA2097L8, and no association between treatment effect on pCR and improvement in OS or EFS was found. However, the number of trials included in the analysis was small.

• a Patient-level survival data were reconstructed from Kaplan-Meier curves to calculate an HR when authors did not report one.

• b OS HR by EFS status before landmark

B.1.4 Equality considerations

No equality issues are foreseen.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.2 Clinical effectiveness

• CheckMate-816 is the first phase 3 study to demonstrate significantly improved EFS and pCR for an immuno-oncology–based combination in the neoadjuvant setting of resectable stage IB-IIIA NSCLC.[3 ]

• With only 3 cycles of treatment, nivolumab + PDC reduced the risk of disease recurrence, progression, or death by 37% versus PDC alone.[3]

• Nivolumab + PDC demonstrated a significant improvement in pCR versus PDC alone (24% for nivolumab + PDC vs. 2% for PDC alone).[3]

• Nivolumab + PDC demonstrated a significantly longer EFS versus PDC alone (HR, 0.63; 97.38% confidence interval [CI], 0.43-0.91).[3 ]

• Longer EFS was observed in patients who achieved pCR versus those who did not across both arms of CheckMate-816 (HR, 0.13 [95% CI, 0.05-0.37] for nivolumab + PDC; HR was not computed for the PDC arm).[3]

• Median time to death or distant metastases (TTDM) was not reached for nivolumab + PDC and was 26.7 months for PDC alone.[3]

• 31% of patients treated with nivolumab + PDC had a radiographic downstaging of their disease versus 24% of those treated with neoadjuvant PDC alone.[3]

• A prespecified interim analysis for OS showed a promising early trend in OS with an HR of 0.57 (not statistically significant); OS will continue to be followed in upcoming analyses.[3]

• Feasibility of surgery was maintained with nivolumab + PDC versus neoadjuvant PDC alone: numerically more patients underwent surgery, had a less invasive surgery and had a shorter duration of surgery with similar length of hospital stay. Furthermore, nivolumab + PDC did not increase postsurgical complications.[3]

• In terms of safety, nivolumab + PDC showed no statistically significant detriment versus PDC alone (all-cause grade 3/4 adverse event [AE] rates were 41% vs. 44% for nivolumab + PDC and PDC alone, respectively). Finally, most immunemediated AEs reported were of low grade.[3]

B.2.1 Identification and selection of relevant studies

B.2.2 List of relevant clinical effectiveness evidence

An SLR was conducted to identify RCTs relevant to the decision problem. One RCT that evaluated nivolumab + PDC as neoadjuvant therapy for the treatment of patients with resectable NSCLC was identified: CheckMate-816 (Table 6). This is the key study relevant to the decision problem described in Section B.1.1. See Appendix D for full details of the process and methods used to identify and select the clinical evidence.

Table 6. CheckMate-816: clinical effectiveness evidence

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

• AE = adverse event; AJCC = American Joint Committee on Cancer; AUC = area under the curve; cRR = clinical response rate; DFS = disease-free survival; EFS = event-free survival; EFS2 = event-free survival on secondline therapy; HRQOL = health-related quality of life; MPR = major pathologic response; NSCLC = non-small cell lung cancer; OS = overall survival; pCR = pathologic complete response; PDC = platinum doublet chemotherapy; PD-L1 = programmed death-ligand 1; PK = pharmacokinetics; TMB = tumour mutational burden; TTDM = time to death or distant metastases; TTLR = time to locoregional recurrence; UICC = Union for International Cancer Control.

Note: Outcomes marked in bold are incorporated into the cost-effectiveness model.

• a In the intervention arm, PDC may include cisplatin (75 mg/m2 on day 1 of a 3-week cycle for up to 3 cycles) and either gemcitabine (1,000 mg/m[2] or 1,250 mg/m[2] [per local prescribing information] on days 1 and 8 of a 3-week cycle for up to 3 cycles, squamous histology) or pemetrexed (500 mg/m[2] on day 1 of a 3-week cycle for up to 3 cycles, nonsquamous histology); or carboplatin (AUC 5 or 6 on day 1 of a 3-week cycle for up to 3 cycles) and paclitaxel (175 or 200 mg/m[2] on day 1 of a 3-week cycle for up to 3 cycles, any histology).

• b In the comparator arm, PDC may include cisplatin (75 mg/m2 on day 1 of a 3-week cycle for up to 3 cycles) and either gemcitabine (1,000 mg/m[2] or 1,250 mg/m[2] [per local prescribing information] on days 1 and 8 of a 3-week cycle for up to 3 cycles, squamous histology), pemetrexed (500 mg/m[2] on day 1 of a 3-week cycle for up to 3 cycles, nonsquamous histology), vinorelbine (25 mg/m[2] or 30 mg/m[2] [per local prescribing information] on days 1 and 8 of a 3-week cycle for up to 3 cycles), or docetaxel (60 mg/m[2] or 75 mg/m[2] [per local prescribing information] on day 1 of a 3-week cycle for up to 3 cycles); or carboplatin (AUC 5 or 6 on day 1 of a 3-week cycle for up to 3 cycles) and paclitaxel (175 or 200 mg/m[2] on day 1 of a 3-week cycle for up to 3 cycles, any histology).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

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

B.2.3.1 CheckMate-816: methodology

CheckMate-816 was a randomised, open-label trial comparing nivolumab + PDC with PDC as neoadjuvant treatment of newly diagnosed resectable (stage IB [≥ 4 cm], stage II, or stage IIIA (N2), AJCC/UICC seventh edition) NSCLC.[3] See Section B.1.3.4 for a description of the AJCC/UICC staging system.

Nivolumab was evaluated in a 360 mg flat dose + PDC every 3 weeks up to 3 cycles versus PDC alone (Figure 6). Table 7 outlines the trial methodology.[3] Additional details of the statistical analyses and endpoints are provided in Section B.2.4.

The original study design also included a study arm with nivolumab + ipilimumab, which was stopped in a protocol revision and is not relevant to this appraisal. This decision was based on evidence from the metastatic setting and external data from the NADIM trial, which demonstrated a more promising pCR benefit for nivolumab + PDC than seen for nivolumab + ipilimumab in the NEOSTAR trial. Therefore, the clinical development of nivolumab + PDC was prioritised in CheckMate-816.[83]

Figure 6. CheckMate-816: study design

==> picture [408 x 162] intentionally omitted <==

• AE = adverse event; ALK = anaplastic lymphoma kinase; BICR = blinded independent central review; BIPR = blinded independent pathologic review; CT = computed tomography; ctDNA = circulating tumour DNA; ECOG PS = Eastern Cooperative Oncology Group performance status; EFS = event-free survival; EGFR = epidermal growth factor receptor; MPR = major pathologic response; MRI = magnetic resonance imaging; NSCLC = non-small cell lung cancer; ORR = objective response rate; OS = overall survival; pCR = pathologic complete response; PD-L1 = programmed death-ligand 1; Q3W = every 3 weeks;

RECIST = Response Evaluation Criteria in Solid Tumours; RT = radiotherapy; TMB = tumour mutational burden.

• a Determined by the PD-L1 immunohistochemistry 28-8 pharmDx assay (Dako).

• b Included patients with PD-L1 expression status not evaluable and indeterminate.

• c Nonsquamous: pemetrexed + cisplatin or paclitaxel + carboplatin; squamous: gemcitabine + cisplatin or paclitaxel + carboplatin.

• d Vinorelbine + cisplatin, docetaxel + cisplatin, gemcitabine + cisplatin (squamous only), pemetrexed + cisplatin (nonsquamous only), or paclitaxel + carboplatin.

• e Postoperative assessments with CT with contrast of the chest including the adrenal glands and CT or MRI of other additional suspected/known sites of disease. The first tumour assessment should occur 12 weeks (± 7 days) after definitive surgery per RECIST 1.1 and then should occur every 12 weeks (± 7 days) for 2 years (104 weeks), then every 6 months (24 weeks ± 7 days) for 3 years, and then every year (52 weeks ± 7 days) for 5 years or until disease recurrence or progression confirmed by BICR.

• f Performed using tumour-guided personalised ctDNA panel.

• Source: Forde et al. (2022)[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

CheckMate-816 included adults aged ≥ 18 years with stage IB (≥ 4 cm), stage II, or stage IIIA NSCLC and an Eastern Cooperative Oncology Group (ECOG) performance status of 0-1 who had not been previously treated with prior chemotherapy or any other cancer therapy for resectable NSCLC, and who had no active brain metastases, autoimmune disease, or known EGFR mutations or ALK translocations.[83]

Patients who had a resectable tumour at diagnosis were enrolled; at the time of study design, the study investigators considered the appropriate comparator to be neoadjuvant PDC.[iv] Therefore, 358 patients were concurrently randomly assigned 1:1 to treatment with nivolumab + PDC (n = 179) or PDC alone (n = 179). The stratification factors for randomisation were PD-L1 expression (≥ 1% vs. < 1%/not evaluable/indeterminate), disease stage (IB/II vs. IIIA), and gender/sex (male vs. female).[3,83]

Treatments administered in the study arms are presented in Table 7; 3 cycles of treatment were provided in both arms. In the nivolumab + PDC arm, PDC was investigator’s choice and consisted of either cisplatin (with either gemcitabine or pemetrexed depending on histology), or carboplatin and paclitaxel. In the PDC arm, PDC consisted of either cisplatin (and gemcitabine or pemetrexed depending on histology) combined with either gemcitabine, pemetrexed, vinorelbine, or docetaxel; or carboplatin and paclitaxel.[3]

Surgery was required within 6 weeks post-neoadjuvant treatment and adjuvant treatment with chemotherapy and/or radiotherapy was permitted at the discretion of the healthcare provider. No immuno-oncology therapy was allowed in the adjuvant setting.[83]

The independent primary endpoints were blinded independent central review (BICR)– assessed pCR, and EFS. Secondary endpoints were blinded independent pathologic review (BIPR)–assessed major pathologic response (MPR), OS and BICR-assessed TTDM. It is important to note that in CheckMate-816, the outcome TTDM is time to death or distant metastases, whereas in the economic model described in Section B.3 and the network metaanalysis that informs it described in Section B.2.9, TTDM refers to time to distant metastases only. Exploratory endpoints included EFS on the next line of therapy (EFS2) and safety.[3,83] Safety assessments were based on the frequency of deaths, serious adverse events, adverse events (AEs) leading to discontinuation or dose modification, overall AEs, clinical laboratory assessments (haematology, serum chemistry, liver, and thyroid function tests), and vital sign measurements.[83] The feasibility of surgery and rate of perioperative and postoperative complications (within 90 days of surgery) were additional exploratory objectives.[3,83]

iv Because neoadjuvant chemoradiotherapy is often used for borderline resectable or Pancoast tumours, this was not deemed an appropriate comparator for nivolumab + PDC in this indication.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 7. CheckMate-816: summary of trial methodology

• Location 111 sites in 14 countries (Argentina, Brazil, Canada, China, France, Greece, Italy, Japan, South Korea, Romania, Spain, Taiwan, Turkey, and US)

• Trial design International, multicentre, open-label, randomised, active-controlled phase 3 trial

• Eligibility criteria Inclusion criteria: Exclusion criteria: for participants ▪ Males and females aged ▪ Patients who have received prior ≥ 18 years chemotherapy or any other cancer

• ▪ Histologically confirmed, therapy for resectable NSCLC resectable, stage IB (≥ 4 cm), ▪ Patients with distant active brain stage II, or stage IIIA NSCLC metastases (according to AJCC/UICC seventh ▪ Patients with an active, known or edition) confirmed by PET/CT with suspected autoimmune disease contrast ▪ Known EGFR mutations or ALK

• ▪ If the CT component of the translocations PET/CT is of insufficient diagnostic quality for RECIST 1.1 measurements, an additional CT with contrast of the chest, abdomen, and other suspected areas of disease will be performed

• ▪ Lung function capacity capable of tolerating the proposed lung surgery

• ▪ ECOG performance status of 0-1

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

• Gemcitabine (1,000 mg/m[2] or 1,250 mg/m[2] on days 1 and 8 of a 3-week cycle for up to 3 cycles) (squamous histology)

• • Pemetrexed (500 mg/m[2] on day 1 of a 3-week cycle for up to 3 cycles) (nonsquamous histology)

• Vinorelbine (25 mg/m[2] or 30 mg/m[2] on days 1 and 8 of a 3-week cycle for up to 3 cycles)

AJCC = American Joint Committee on Cancer; ALK = anaplastic lymphoma kinase; AUC = area under the curve; BICR = blinded independent central review; BIPR = blinded independent pathological review; CT = computed tomography; ECOG = Eastern Cooperative Oncology Group; EFS = event-free survival; EGFR = epidermal growth factor receptor; IV = intravenous; MPR = major pathologic response; NSCLC = non-small cell lung cancer; pCR = pathologic complete response; PDC = platinum doublet chemotherapy; PD-L1 = programmed death-ligand 1; PET = positron emission tomography; RECIST = Response Evaluation Criteria in Solid Tumours; TMB = tumour mutational burden; TTDM = time to death or distant metastases; UICC = Union for International Cancer Control; US = United States.

Source: BMS data on file (2021)[83]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.2.3.2 CheckMate-816: baseline characteristics

Demographics and baseline disease characteristics in the 2 treatment arms in CheckMate-816 were balanced and generally representative of a resectable NSCLC population (Table 8); UK clinical experts confirmed that the study population is similar to that in the UK, other than expected differences between trial and real-world evidence. These included minor differences in median age and proportion of patients with squamous histology (see Appendix N).

Patients had a median age of 64 to 65 years, and 71% were male. Twenty-three percent of patients were from Europe in the nivolumab + PDC group and 14% in the PDC group. There were 65% to 69% of patients with an ECOG performance status of 0, and the remainder (31%-35%) had a performance status of 1. Overall, 49% to 53% of patients had squamous cell carcinoma, 43% to 44% of patients had PD-L1 < 1%, and 88% to 89% of patients were current or former smokers.[3]

Table 8. CheckMate-816: baseline characteristics of patients

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

ECOG PS = Eastern Cooperative Oncology Group performance status; mut/MB = mutations per megabase; PDL1 = programmed death-ligand 1; TMB = tumour mutational burden.

• a This category includes Argentina and Turkey only.

• b ECOG PS scores range from 0-5, with higher scores indicating greater disability.

• c Data for disease stage are from case report forms, with the TNM Classification of Malignant Tumours, seventh edition, used for classification.

• d 1 patient in the chemotherapy-alone group had stage IA disease, and 1 patient in each group had stage IV disease.

• e 1 patient in the chemotherapy-alone group had unknown smoking status.

• f Percentages are based on the primary analysis population. The status of PD-L1 expression was determined with the use of the PD-L1 immunohistochemistry 28-8 pharmDx assay (Dako); patients with tumour tissue that could not be assessed for PD-L1 expression (≤ 10% of all the patients who underwent randomisation) were stratified to the subgroup with a PD-L1 expression level of less than 1% at randomisation.

• g TMB was not analysed from patients in China, and these patients are included in the Not Reported category.

Source: Forde et al. (2022)[3] ; BMS data on file (2021)[83]

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

B.2.4.1 CheckMate-816

An analysis of pCR was planned when all patients were available for assessment of pCR, followed by 2 prespecified interim analyses for EFS after 148 EFS events (80% of events required for final analysis) and 167 EFS events (90% of events required for final analysis), respectively.[83] The 2 EFS interim analyses were planned in 358 randomly assigned patients to ensure an 82% power, assuming a hazard ratio (HR) of 0.65 between the 2 arms.[83] The first database lock for EFS, triggered after 148 EFS events occurred on 20 October 2021, and results from this analysis are presented in this submission, alongside the final analysis of pCR (database lock, 16 September 2020). Results from xxxxxxx xxxxxxxx x are expected to be available within the timelines of this appraisal and the xxxxxxxx xxxx will be provided once available. However, the timeframe of the appraisal will not allow ixxxxxxxxxxxx xxxx xxx xxxxxxxxxxxxxxxxx xxxxl. Figure 7 presents a summary of all planned analyses, the analysis that is the basis of this submission is highlighted in red.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 7. CheckMate-816: planned analyses

==> picture [437 x 237] intentionally omitted <==

EFS = event-free survival; FA = final analysis; IA = interim analysis; OS = overall survival; pCR = pathologic complete response.

Note: EFS and OS tested using each their own O’Brien-Fleming alpha-spending function.

a Analysis occurred based on 16 September 2020 database lock.

b Analysis occurred based on 20 October 2021 database lock.

Source: BMS data on file (2018)[84] ; BMS data on file (2021)[83]

Table 9 summarises the statistical analyses in CheckMate-816. The sample size of the study was calculated based on the primary endpoint of EFS and accounted for the 2 primary endpoints comparisons: pCR rate (per BIPR) and EFS (per BICR).

A total of xxx EFS events ensured that an overall 2-sided 5% significance level sequential test procedure with 2 interim analyses after 148 events (xx% of events required for final analysis) and 167 events (xx% of events required for final analysis) in 358 randomised subjects would have 82% power, assuming a HR of 0.65 between the 2 arms.[83]

For the pCR outcome, it was estimated that a sample of approximately 350 randomly assigned patients would provide more than 90% power to detect an odds ratio of 3.857 with a 2-sided type I error of 1%.[83]

A total of xxx OS events ensures that an overall 2-sided x% significance level sequential test procedure with 3 interim analyses after approximately xxx, xxx, and xxx events (xx%, xx%, and xx% of events required for final analysis) in 358 randomly assigned patients would have xx% power, assuming an exponential distribution with the median OS time in the chemo arm being 54 months and in the nivolumab + PDC arm being xx months (corresponding to a target HR of 0.65).[83]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 9. CheckMate-816: summary of the statistical analyses

AE = adverse event; AJCC = American Joint Committee on Cancer; BICR = blinded independent central review; CI = confidence interval; CTCAE = Common Terminology Criteria for Adverse Events; ctDNA = circulating tumour DNA; EFS = event-free survival; EFS2 = event-free survival on second-line therapy; HR = hazard ratio; IRT = Interactive Response Technology; KM = Kaplan-Meier; MPR = major pathologic response; NCI = National Cancer Institute; NIVO = nivolumab; NSCLC = non-small cell lung cancer; OS = overall survival; pCR = pathologic complete response; PDC = platinum doublet chemotherapy; PD-L1 = programmed death-ligand 1; SAE = serious adverse event; TMB = tumour mutational burden; TTDM = time to death or distant metastases; UICC = Union for International Cancer Control.

Sources: BMS data on file (2018)[84] ; BMS data on file (2021)[83] ; Forde et al. (2022)[3]

B.2.5 Critical appraisal of the relevant clinical effectiveness evidence

Table 10 presents the quality assessment for CheckMate-816. CheckMate-816 was an open-label study because of the differences in chemotherapy-related and immuno-oncology therapy–related toxicities; histology-dependent chemotherapy options; dose modification rules for safety management, including different dose-delay rules per treatment arm; and premedication requirements according to chemotherapy made blinding impractical. An openlabel design also helped ensure that immune-related toxicities in patients receiving immunooncology therapy were promptly identified and managed. Bristol Myers Squibb (BMS) was blinded to the aggregated safety and efficacy data by treatment assignments.

Patients meeting eligibility criteria were randomly assigned to one of the treatment arms through Interactive Response Technology. Demographics and baseline disease characteristics were balanced between treatment arms (see Section B.2.3.2) and generally representative of the population with resectable NSCLC. A BIPR was used to review pathological data and tumour assessment for all randomly assigned patients. During the review, the BMS personnel remained blinded to treatment group assignment. Personnel who conducted the PD-L1 testing, scoring for tumour cell PD-L1, and assessed tumour mutational burden data, were blinded to patient treatment group assignment. Unblinded data reviewed by the data monitoring committee (DMC) at regular safety review meetings were not shared with BMS until formal pCR and EFS analysis achieved the prespecified significance level and unblinding thresholds as specified in the statistical analysis plan/DMC charter. No unexpected imbalances in dropouts between groups were reported.

Although neoadjuvant PDC was the comparator used in CheckMate-816, the current SOC in England is surgery alone or adjuvant platinum-based chemotherapy (see Section B.1.3.6). Nonetheless, evidence suggests that neoadjuvant and adjuvant chemotherapy regimens offer similar improvements in 5-year OS.[56-59]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 10. CheckMate-816: quality assessment

Was randomisation carried out appropriately? Yes Was the concealment of treatment allocation No; open label adequate? Were the groups similar at the outset of the Yes; baseline characteristics of all randomly study in terms of prognostic factors? assigned patients were similar and balanced between treatment groups Were the care providers, participants, and No; open label outcome assessors blind to treatment allocation? Were there any unexpected imbalances in No dropouts between groups? Is there any evidence to suggest that the No authors measured more outcomes than they reported? Did the analysis include an intention-to-treat Yes analysis? If so, was this appropriate and were appropriate methods used to account for missing data? How closely does the RCT(s) reflect routine Unclear – current SOC in England is surgery clinical practice? alone or adjuvant platinum-based chemotherapy rather than neoadjuvant chemotherapy.

RCT = randomised controlled trial; SLR = systematic literature review; SOC = standard of care.

Note: this has been updated from the SLR report, and is now based on the full publication by Forde et al. (2022). Sources: Forde et al. (2022)[3] ; BMS data on file (2021)[83]

B.2.6 Clinical effectiveness results of the relevant studies

B.2.6.1 CheckMate-816

Results presented here are based on the first prespecified interim analysis of EFS and OS (database lock, 20 October 2021; minimum follow-up, 21 months; median follow-up, 29.5 months) and the final analysis of pCR (database lock, 16 September 2020).[3]

Because the comparison between nivolumab + PDC versus PDC in concurrently randomly assigned patients was statistically significant for EFS (as per the 20 October 2021 database lock), formal testing for the secondary objective, OS, was subsequently performed by the DMC. Although readouts were favourable for nivolumab + PDC, with a clear visual separation in the Kaplan-Meier survival curves, the P value for OS did not cross the significance boundary at the first interim analysis. Overall survival will be formally tested again at the second interim OS analysis.[3]

Results presented in this section represent all patients relevant to NICE’s decision problem.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.2.6.1.1 Summary of treatment

Table 11 presents a summary of treatments in CheckMate-816. All patients were no longer receiving treatment at the time of the database locks; 93.8% in the nivolumab + PDC group and 84.7% in the PDC group had fully completed the prespecified neoadjuvant treatment, while 6.2% and 15.3%, respectively, had stopped treatment due to toxicity, disease progression or other reasons. Adjuvant chemotherapy was received by 11.9% of patients treated with nivolumab + PDC and 22.2% of those treated with PDC alone.[3]

Table 11. CheckMate-816: treatment summary

NIVO = nivolumab; PDC = platinum doublet chemotherapy.

a Denominator based on patients receiving neoadjuvant treatment.

b Reasons were adverse event unrelated to study drug in 3 patients, patient request to discontinue study treatment in 5 patients, patient withdrew consent in 4 patients, and patient no longer met study criteria in 1 patient.

Source: Forde et al. (2022)[3]

Any subsequent cancer therapy (excluding adjuvant treatment) was received by 21.2% of the patients in the nivolumab + PDC group and 43.6% of those in the PDC group; subsequent systemic therapy was received by 17.3% and 36.3% of patients, respectively (Table 12).[3] Chemotherapy is the most commonly prescribed subsequent treatment for both nivolumab + PDC and PDC alone, representing 15.1% and 22.3% received, respectively; clinical experts confirmed that this would reflect clinical practice after treatment with nivolumab + PDC (Appendix N). These data reinforce the long-term benefit of nivolumab + PDC in delaying recurrence; a lower proportion of patients require subsequent treatment following nivolumab + PDC and a greater proportion of patients who have responded remain event-free (see Event-free survival in Section B.2.6.1.2).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 12. CheckMate-816: subsequent therapies

NIVO = nivolumab; PDC = platinum doublet chemotherapy.

Note: Subsequent therapy was defined as therapy started on or after first dosing date (randomisation date if patient never treated) outside the protocol-specified adjuvant therapy. Patients may have received more than 1 type of subsequent therapy.

a Any subsequent anticancer (non-small cell lung cancer) surgery. Most were for palliative reasons or in patients with oligo-metastatic disease; some patients underwent subsequent surgery for the primary tumour. Source: Forde et al. (2022)[3]

B.2.6.1.2 Primary outcomes

Pathologic complete response

Among all patients in the primary analysis population (database lock, 15 September 2020), 24.0% (95% confidence interval [CI], 18.0%-31.0%) treated with nivolumab + PDC had a pCR versus 2.2% (95% CI, 0.6%-5.6%) treated with PDC alone (odds ratio, 13.94; 99% CI, 3.49-55.75; P < 0.001) (Figure 8).[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 8. CheckMate-816: pathologic complete response according to BIPR

==> picture [411 x 231] intentionally omitted <==

BIPR = blinded independent pathologic review; CI = confidence interval; NIVO = nivolumab; PDC = platinum doublet chemotherapy. Source: Forde et al. (2022)[3]

Achieving a pCR (i.e., an absence of residual tumour in lung resected tissue and lymph nodes) is clinically meaningful because this has been shown to be associated with improved EFS and OS; early evidence of a benefit in these outcomes indicates a stronger EFS and OS benefit in the longer term than patients who did not achieve a pCR (see Section B.1.3.6.5 for further detail).[72-75] Thus, the statistically significant increases in pCR observed with nivolumab + PDC versus PDC alone suggest that nivolumab + PDC may improve survival outcomes versus PDC alone.[3]

Event-free survival

With a minimum follow-up of 21 months, the median EFS was 31.6 months with nivolumab + PDC and 20.8 months with PDC alone, a 37% reduction in the risk of recurrence or death (HR, 0.63; 97.38% CI, 0.43-0.91; P = 0.005) (Table 13 and Figure 9).[3] Early separation of the Kaplan-Meier curves favours patients treated with nivolumab + PDC (Figure 9). After this timepoint, nivolumab + PDC is associated with consistently longer EFS versus patients treated with PDC alone, with an increasing incremental gain versus PDC alone in landmark observed EFS rates (Table 13). The EFS benefit with nivolumab + PDC was maintained after adjusting for optional adjuvant therapy (HR for disease progression, disease recurrence, or death, 0.65; 95% CI, 0.47-0.90).[3] A total of 64.2% patients in the nivolumab + PDC arm and 51.4% patients in the PDC arm were censored at their last tumour assessment.[83]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 13. CheckMate-816: summary of event-free survival

CI = confidence interval; EFS = event-free survival; HR = hazard ratio; NIVO = nivolumab; NR = not reached; PDC = platinum doublet chemotherapy.

Note: 115 of 179 patients (64.2%) in the nivolumab + PDC arm and 92 of 179 (51.4%) in the PDC arm were censored for EFS per blinded independent central review at database lock.

Source: Forde et al. (2022)[3]

Figure 9. CheckMate-816: event-free survival according to BICR

==> picture [412 x 194] intentionally omitted <==

BICR = blinded independent central review. Note: Chemotherapy refers to platinum doublet chemotherapy. Source: Forde et al. (2022)[3]

These data further support a durable benefit with nivolumab + PDC. It is clear from the Kaplan-Meier curves and favourable HR that nivolumab + PDC offers a substantial, clinically relevant benefit for patients in terms of EFS versus PDC alone, a benefit that is expected to be sustained with longer term follow-up and ultimately is expected to translate into OS benefit, as suggested in other studies.[75] Although longer follow-up is required, the existing evidence on attaining a cure at 5 years in resectable NSCLC suggests this is likely to occur in patients treated with nivolumab + PDC with additional follow-up.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.2.6.1.3 Secondary outcomes

Overall survival

In the interim analysis, median OS was not reached in either the nivolumab + PDC group or the PDC alone group at the time of the interim database lock due to immaturity of the data (minimum follow-up of 21 months; HR for death, 0.57; 99.67% CI, 0.30-1.07; P = 0.008) (Figure 10). The first prespecified interim analysis was conducted in line with the statistical plan because of the number of EFS events that had occurred; however, few OS events had occurred and the P value for OS did not cross the boundary for statistical significance (0.0033).[3] However, the trend in data indicates the hazard of death is lower for patients treated with nivolumab + PDC versus patients treated with PDC alone.

The Kaplan-Meier curves show similar OS at start of treatment in both the nivolumab + PDC and PDC alone arms; they overlap until a clear and consistent separation is seen after approximately 15 months, which suggests improved survival with nivolumab + PDC.[3] Although these data are not yet mature, they suggest, alongside the statistically significant comparative difference in pCR and EFS for nivolumab + PDC versus PDC alone, that a continued relative benefit for nivolumab + PDC and a statistically significant difference in OS may be observed once more events have accrued.

Figure 10. CheckMate-816: overall survival

==> picture [437 x 227] intentionally omitted <==

CI = confidence interval; NR = not reached. Note: Chemotherapy refers to platinum doublet chemotherapy. Source: Forde et al. (2022)[3]

Major pathologic response rate by BIPR

The percentage of patients with an MPR was substantially higher with nivolumab + PDC than with PDC alone in the primary analysis population (36.9% vs. 8.9%; odds ratio, 5.70; 95% CI, 3.16-10.26).[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Time to death or distant metastases according to BICR

Time to death or distant metastases was defined as the time between the date of randomisation and the first date of distant metastasis or the date of death in the absence of distant metastasis.[3] Results favoured nivolumab + PDC over PDC alone (HR, 0.53; 95% CI, 0.36-0.77) (Figure 11).[3] The Kaplan-Meier curves are similar between treatment arms as the curves overlap; after approximately 6 months, the curves separate where nivolumab + PDC begins to improve TTDM.

Figure 11. CheckMate-816: time to death or distant metastases

==> picture [418 x 314] intentionally omitted <==

CI = confidence interval; HR = hazard ratio; NR = not reached; TTDM = time to death or distant metastases. Note: Chemotherapy refers to platinum doublet chemotherapy. Source: Forde et al. (2022)[3]

B.2.6.1.4 Exploratory outcomes

Clinical response rate by BICR

Rates of response according to BICR were higher with nivolumab + PDC than with PDC alone (Table 14).[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 14. CheckMate-816: objective response rate and best overall response

CI = confidence interval; NIVO = nivolumab; PDC = platinum doublet chemotherapy.

a Objective response rate per blinded independent central review was defined as a complete or partial response from baseline to the presurgery scan per Response Evaluation Criteria in Solid Tumours (RECIST) 1.1. Source: Forde et al. (2022)[3]

Incidence of radiographic downstaging

The incidence of radiographic downstaging (reduction of disease stage from baseline) was 30.7% and 23.5% with nivolumab + PDC versus PDC, respectively (Table 15).[3]

Table 15. CheckMate-816: radiographic downstaging before and after treatment by stage of disease

NIVO = nivolumab; PDC = platinum doublet chemotherapy. Source: Forde et al. (2022)[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Event-free survival 2

Event-free survival 2 (EFS2) was defined as time from randomisation to objectively documented progression, per investigator assessment, after the next line of therapy or to death from any cause, whichever occurs first; patients without documented progression on the next line who started a second next line of subsequent therapy were considered to have had an event at the start of second next line of therapy. This definition aligns with the standard progression-free survival 2 (PFS2) definition used in the metastatic setting.[83] Nivolumab + PDC demonstrated a 46% statistically significant reduction in risk of EFS2 (HR, 0.54; 95% CI, 0.37-0.80) (Figure 12).[3] The Kaplan-Meier curve displays overlapping curves representative of similar EFS2 rates between treatment arms until a separation is seen after approximately 9 months, where nivolumab + PDC begins to show improved EFS2. This is sustained after the separation, which supports the durability of the benefit seen with nivolumab + PDC.

Figure 12. CheckMate-816: event-free survival 2

==> picture [398 x 303] intentionally omitted <==

CI = confidence interval; EFS2 = event-free survival 2; HR = hazard ratio; NR = not reached. Note: Chemotherapy refers to platinum doublet chemotherapy. Source: Forde et al. (2022)[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Event-free survival according to pathologic complete response status

As shown in Figure 13, EFS was longer in patients with a pCR than in those without a pCR. Among patients with a pCR, median EFS was not reached in either treatment group.3 In patients without a pCR, the median EFS was 26.6 months with nivolumab + PDC and 18.4 months with PDC alone (HR for disease progression, disease recurrence, or death, 0.84; 95% CI, 0.61-1.17).[3] As a higher proportion of patients treated with nivolumab + PDC achieve pCR, we anticipate that EFS will continue to show an improved benefit versus chemotherapy with longer follow-up.

Figure 13. CheckMate-816: event-free survival by pathologic complete response

==> picture [445 x 361] intentionally omitted <==

CI = confidence interval; EFS = event-free survival; HR = hazard ratio; NR = not reached; pCR = pathologic complete response.

Note: Chemotherapy refers to platinum doublet chemotherapy.

• HR (95% CI) for nivolumab + PDC versus PDC in patients without a pCR was 0.84 (0.61-1.17).

• HR was not computed for the chemotherapy arm because only 4 patients had a pCR.

Source: Forde et al. (2022)[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Feasibility of surgery and rates of perioperative and postoperative complications

Feasibility of surgery was maintained with nivolumab + PDC versus neoadjuvant PDC alone: numerically more patients underwent surgery with nivolumab + PDC, had a less invasive and less extensive surgery, had a shorter duration of surgery and had a similar length of hospital stay.

Among all the patients who underwent concurrent randomisation, 149 (83.2%) in the nivolumab + PDC group and 135 (75.4%) in the PDC group underwent definitive surgery (Table 16). Surgery was cancelled for 28 (15.6%) and 37 (20.7%) of the patients, respectively; reasons for cancellation included disease progression (12 [6.7%] and 17 [9.5%], respectively), AEs (2 [1.1%] and 1 [0.6%]), and other (14 [7.8%] and 19 [10.6%] [including patient refusal, unresectability, and poor lung function]). The percentage of patients with delayed surgery was similar in the 2 treatment groups.[3]

The median duration of surgery was numerically shorter and the use of minimally invasive approaches was more common in the nivolumab + PDC group than in the PDC alone group. Notably, minimally invasive surgery rates were 29.5% and 21.5% in the nivolumab + PDC and PDC alone groups, respectively, and rates of conversion from minimally invasive to open surgery were higher in the PDC group (11.4% vs. 15.6%). Further, a higher proportion of patients underwent lobectomy rather than pneumonectomy in the nivolumab + PDC group than in the PDC group (lobectomy, 77.2% vs. 60.7%; pneumonectomy, 16.8% vs. 25.2%, respectively). Thus, in CheckMate-816, patients treated with nivolumab + PDC were more likely to undergo surgery and surgery was more likely to be minimally invasive, which is associated with reduced mortality and improved HRQOL as discussed in Section B.1.3.6.1, than patients treated with neoadjuvant PDC alone.[3]

For those patients who underwent definitive surgery, a higher proportion of those treated with nivolumab + PDC (83.2%) than PDC alone (77.8%) had a complete resection with no residual tumour (R0).

The median length of hospital stay for surgery was 10 days in both treatment arms.3 This is longer than the expected length of hospital stay in the UK (according to clinical expert opinion; Appendix N) and is driven by the SOC in some of the Asian countries included in the study, where longer hospital stays are the norm. Nonetheless, use of nivolumab + PDC did not increase length of stay versus neoadjuvant PDC alone.[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 16. CheckMate-816: surgical outcomes summary

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

IQR = interquartile range; NIVO = nivolumab; PDC = platinum doublet chemotherapy.

• a Definitive surgery was not reported in 2 patients in the NIVO+PDC group and 7 in the PDC group.

• b Other reasons were patient refusal in 9 patients in the NIVO+PDC arm and 8 patients in the PDC arm; consent withdrawal in 3 patients in the PDC arm; COVID-19 in 1 patient in the PDC arm; unfit for surgery due to poor lung function in 2 patients in the NIVO+PDC arm and 4 patients in the PDC arm; and unresectability in 2 patients in each arm.

• c Time from last dose to neoadjuvant surgery > 6 weeks.

• d Denominator based on patients with definitive surgery (n = 149 in the NIVO+PDC group; n = 135 in the PDC group).

• e Denominator based on patients with delayed surgery.

• f Patients with reported duration of surgery: NIVO+PDC, 122; PDC, 121.

• g Thoracoscopic/robotic.

• h Patients may have had more than 1 surgery type.

• I Includes bilobectomy, sleeve lobectomy, and other.

• Source: Forde et al. (2022)[3]

Patient-reported outcomes

The addition of nivolumab to PDC had no detrimental impact on HRQOL during the neoadjuvant period.[85] Health-related quality of life was assessed using the EQ-5D-3L. A mixed-effects model repeated measures analysis evaluated longitudinal changes from baseline in EQ-5D visual analogue scale (range 0 to 100) and utility index (UI; range −0.594 to 1) scores during the neoadjuvant period (week 4, week 7, and post-neoadjuvant visit 1); higher scores reflect better HRQOL.

EQ-5D-3L completion rates were > 80% in both treatment arms at baseline and during the neoadjuvant period. Baseline EQ-5D-3L visual analogue scale and UI scores were consistent with UK population norms. Scores during the neoadjuvant period were generally similar to baseline for both treatment arms; there were no clinically meaningful differences between nivolumab + PDC versus PDC (Table 17). In both treatment arms, most patients reported “no problems” for individual EQ-5D-3L dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) at baseline and during treatment.[85]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 17. EQ-5D-3L in the neoadjuvant period

CI = confidence interval; LSM = least squares mean; MID = minimally important difference; PDC = platinum doublet chemotherapy; UI = utility index; VAS = visual analogue scale. Source: Felip (2022)[85]

B.2.7 Subgroup analysis

Prespecified subgroup analyses were conducted in line with the study protocol. However, patient numbers in some subgroups are low and very few events had occurred by the time of the database lock. Therefore, these analyses are not appropriate for decision-making, and the decision problem population included in CheckMate-816 is more appropriate; this was also highlighted by clinical experts (Appendix N).

A benefit with nivolumab + PDC with respect to pCR was observed across all subgroups excluding patients who never smoked (Figure 14).[3] Of note, a pCR benefit was seen regardless of stage and PD-L1 status. Analyses have consistently demonstrated an association between pCR and EFS/OS as well as EFS and OS[72-75] ; therefore, early evidence of a benefit in these outcomes indicates an OS benefit in the longer term (see Section B.1.3.6.5 for further detail).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 14. CheckMate-816: pathologic complete response according to BIPR by subgroup

==> picture [452 x 395] intentionally omitted <==

BIPR = blinded independent pathologic review; CI = confidence interval; ECOG = Eastern Cooperative Oncology Group; PD-L1 = programmed death-ligand 1; TMB = tumour mutational burden.

Note: Chemotherapy refers to platinum doublet chemotherapy. Source: Forde et al. (2022)[3]

Event-free survival across most subgroups in the subgroup analyses favoured nivolumab + PDC (Figure 15).[3] However, subgroups were underpowered and an insufficient number of events have occurred to demonstrate statistical significance across all subgroups. The pCR benefit seen across all subgroups, as described above, is anticipated to result in longer term EFS benefit once more events have occurred with longer follow-up.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 15. CheckMate-816: event-free survival according to BICR by subgroup

==> picture [452 x 390] intentionally omitted <==

BICR = blinded independent central review; CI = confidence interval; ECOG = Eastern Cooperative Oncology Group; NR = not reached; PD-L1 = programmed death-ligand 1; TMB = tumour mutational burden. Note: Chemotherapy refers to platinum doublet chemotherapy. Source: Forde et al. (2022)[3]

Although CheckMate 816 was not powered for subgroup analyses, nivolumab + PDC was favoured versus PDC in a descriptive, exploratory subgroup analysis of EFS and pCR across disease stages. While a greater EFS and pCR benefit was observed in patients with stage IIIA disease, a numerical benefit was observed in patients with stage IB-II (AJCC/UICC seventh TNM edition). Clinical experts highlighted the considerable benefit seen in all stages (including IB-II), which they noted is as substantial as that of other treatments (Appendix N). Further, they considered that from a patient perspective, nivolumab + PDC would be beneficial, particularly with only 3 cycles of treatment.

Nivolumab + PDC was also favoured versus PDC regardless of PD-L1 expression in the descriptive, exploratory subgroup analysis. While a greater EFS and pCR benefit was observed in PD-L1 ≥ 1%, a numerical benefit was also observed in PD-L1< 1%; clinical experts confirmed this and reiterated the fact that the CheckMate-816 trial is not powered to make conclusions based on these subgroup data (Appendix N).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.2.8 Meta-analysis

Only 1 phase 3 RCT (CheckMate-816) was identified via the SLR that has investigated the efficacy and safety of nivolumab + PDC for the neoadjuvant treatment of resectable NSCLC. As such, a meta-analysis could not be conducted, as this would require 2 or more studies that contained the intervention of interest.

B.2.9 Indirect and mixed treatment comparisons

In the absence of head-to-head trial evidence of nivolumab + PDC versus all UK relevant comparators of interest, an indirect treatment comparison was necessary to enable a comparison for this submission. A summary of the performed analysis is presented in this section with more details presented in Appendix M.

B.2.9.1 Evidence base

Treatment comparisons were informed by a network meta-analysis (NMA) based on data extracted from a previously conducted SLR of RCTs. The evidence base identified included both RCTs where potentially resectable patients were randomised before surgery (mainly neoadjuvant RCTs) and RCTs where completely resected patients were randomised after surgery (mainly adjuvant RCTs). It is also worth noting that completely resected patients are a subset of all potentially resectable patients, as not all surgeries are successful in achieving negative margins. Performing robust indirect treatment comparisons when time of randomisation and patient populations between different RCTs differ substantially is challenging and may lead to biased results. For this reason, the base-case NMA focused on evidence from RCTs conducted among potentially resectable patients, to be aligned with patients eligible for nivolumab + PDC. In addition, in order to reduce heterogeneity in the chemotherapy regimens considered in RCTs, only 3rd generation chemotherapy regimens were included in the base-case analysis, as they were deemed the most relevant to the decision problem. These inclusion criteria were relaxed in sensitivity analysis, where RCTs enrolling completely resectable patients and evaluating 2[ nd] generation chemotherapies were admitted into the NMA network of evidence. Extending the analysis to RCTs enrolling completely resected patients after surgery enabled the comparison between nivolumab + PDC and adjuvant PDC for all endpoints of interest, which is part of the scope of this appraisal.

The comparators of interest included in this analysis reflect the comparators considered in the decision problem addressed in this submission and presented in Table 18. Please note the terminology used for the different treatment options in the NMA. It should be noted that the analysis was conducted from a broader perspective than required for this appraisal and thus also included neoadjuvant chemotherapy (3rd generation PDC).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 18. Comparators included in the network meta-analysis

Note: comparator abbreviations used in this section correspond to full terms in the remainder of the submission; abbreviations are used here to align with data figures.

The primary outcomes of interest were EFS and OS. Secondary outcomes of interest were time to locoregional recurrence (TTLR), TTDM, pCR and safety endpoints. Within the NMA section, TTDM refers to time to distant metastases only, as opposed to the CM816 trial definition of time to death or distant metastases. This is because deaths have been censored to allow for accurate transition probabilities to be calculated for the economic model. For EFS, the following author-reported endpoints were included: DFS, relapse-free survival, and progression-free survival (PFS).

B.2.9.2 Network meta-analysis method assessment

Event-free survival, OS, TTLR and TTDM were analysed as time-to-event data, using HRs while pCR was reported as odds ratios. Where relevant, Kaplan-Meier curves were digitised and individual patient-level data were generated. Proportional hazards (PHs) assumptions were evaluated by reconstructing Kaplan-Meier curves and examining log-cumulative hazard plots, Schoenfeld residual plots, and Grambsch-Therneau tests. The NMA conduct and reporting were in alignment with good practice guidelines published by NICE Decision Support Unit (DSU) Technical Support Documentation and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidance. Random and fixed effect models were implemented for both time-to-event and binary endpoints. All analyses were performed using R version 4.0.2 and WinBUGS version 1.4.3.

Both random- and fixed- effects models were developed. However, the sparse evidence base available to inform the network did not provide sufficient evidence to calculate the between-study standard deviation; therefore, the fixed effect results were preferred.

B.2.9.3 Summary of results

Of the 58 RCTs identified for inclusion in the SLR (see Appendix D), xx were eligible for inclusion in the NMA. Among these, x were included in the base case as they were conducted among patients deemed candidates for surgery and evaluated 3rd generation chemotherapies.[3,70,86-91] In addition, x xxxx were included in the sensitivity analyses expanding to xxx xxxxxxxxxx xxxxxxxxxxxxxx,[92-96] x xxxx were included in the sensitivity analyses expanding to xxxxxxxxxx xxxxxxxx xxxxxxxx.[97-100] A total of x xxxx met the NMA

PICOS (patients, intervention, comparator, outcomes, and study design) criteria but were not included in any analyses; these RCTs included completely resected patients and

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

investigated a 2nd generation chemotherapy or tegafur and uracil (UFT, a therapy relevant only in Japan).[101-104]

B.2.9.3.1 Disposition and baseline/demographic characteristics

Among the x xxxx included in the base-case NMA, median age ranged from xx xx xx xxxxx There was some heterogeneity in terms of: xxxxxxxxxx xx xxxxx xxxxx xx xx xx xxxxxxx xxxxxxxxxxxxxxx xxxxxxxxxxxxxx xx xxxxxxxxxxxxxxxx xxxxxxxx xxxxxxxxx xxxxxx x xx x xx xxxxxxxxx xxxxxx xxxxxxxxxx xx xxxxxxxxxx xxxxxxxxxxxx xxxxxxxx xxxxxxxxx xxxxx xx xx xx xxx xxxxx xxxxxxxxx xxxxxxx xxxxxxxx xxxx).

B.2.9.3.2 Primary endpoint results

The network of evidence describing the totality of the evidence included in the base-case analysis for EFS and OS is presented in Figure 16; results for EFS and OS are presented in Figure 17 and Figure 18, respectively. Given that the treatment effect in the model was applied via the common node in the network (neoCT), the tables present results versus neoCT to be consistent with values used in the economic model. The NMA results suggest that neoNIVO-CT is associated with improved EFS and OS relative to neoCT, neoCRT, S, and adjCT. Results of the sensitivity analyses were generally in line with the base-case results.

Figure 16. Network of evidence for event-free survival and overall survival

==> picture [313 x 210] intentionally omitted <==

adjCT = adjuvant chemotherapy; neoCRT = neoadjuvant chemoradiotherapy; neoCT = neoadjuvant chemotherapy; neoNIVO-CT = neoadjuvant nivolumab-chemotherapy; S = surgery.

Sources: Forde et al. (2022)[3,] Felip et al. (2010)[70,] Girard et al. (2010)[86,] Katakami et al. (2012)[87,] Pless et al. (2015)[88,] Li et al. (2009)[89,] Pisters et al. (2010)[90,] Scagliotti et al. (2012)[91]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 17. Event-free survival hazard ratio estimates of all relevant comparators vs. neoadjuvant chemotherapy (fixed effect estimates)

==> picture [460 x 105] intentionally omitted <==

CrI = credible interval; HR = hazard ratio; neoNIVO-CT = neoadjuvant nivolumab-chemotherapy.

Figure 18. Overall survival hazard ratio estimates for all relevant comparators vs. neoadjuvant chemotherapy (fixed effect estimates)

==> picture [457 x 107] intentionally omitted <==

CrI = credible interval; HR = hazard ratio; neoNIVO-CT = neoadjuvant nivolumab-chemotherapy.

B.2.9.3.3 Secondary endpoint results

xxxxxxxxx xxx xxxxxxxxxx xxxx xxxxxxxxxxxxx xxxxxxxxxxx xxxxxx xxxx xx xxxxxxxxx xxx

xxxxxxxx xx xxxxx xxx xxxxxx (Figure 19). The network of evidence for pCR does not include S and adjCT, given that pCR is an endpoint specific to neoadjuvant treatments. Evidence with respect to TTLR (Figure 20) xxx xxxxxx xxxxxxxxx xxxxxxx xxxxxx xxx xxxxxxxxx xxxx xxxxxx xxxxxxxx Evidence with respect to TTDM (Figure 21) shows a xxxxxx xxx xxxxxxxxxxxxx xxxxxxxxxxx xxxxxxx xxx xxxxxxxxx xxxxxxxx xx xxxxx xxxxxxxxx xx xxx xxxx xxxxxxxxxxx xxxxxxxxx xx xxxxxxxxxx xx xxxxxxxx xxxxxxx xxxxxxxxxx

Figure 19. Pathologic complete response odds ratio estimates for all relevant comparators vs. neoadjuvant chemotherapy (fixed effect estimates)

==> picture [457 x 85] intentionally omitted <==

CrI = credible interval; neoNIVO-CT = neoadjuvant nivolumab-chemotherapy; OR = odds ratio.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 20. Time to locoregional recurrence hazard ratio estimates for all relevant comparators vs. neoadjuvant chemotherapy (fixed effect estimates)

==> picture [457 x 96] intentionally omitted <==

CrI = credible interval; neoNIVO-CT = neoadjuvant nivolumab-chemotherapy; OR = odds ratio.

Figure 21. Time to distant metastases hazard ratio estimates for all relevant comparators vs. neoadjuvant chemotherapy (fixed effect estimates)

==> picture [458 x 98] intentionally omitted <==

CrI = credible interval; HR = hazard ratio; neoNIVO-CT = neoadjuvant nivolumab-chemotherapy.

Scenario analysis results

As shown in the results above, the xxxx xxxxxxxx xxx xxx xxxxx xxx xxxxxxxxxx xxxx xxxxxxxx xxxxxxxxxxxx xxx xxxx xxx xxxx xxxxx xxx xxxx xx xxxxxxx xxxxxxxxxx xxxxxxxxx xx xxxx xxxxxxxxxxx xxxxxxxxxx xxxxxxxx xxxx xxxx xxx xxxx xxx xxx xxxxxxxx xxxxxxxx xxxx xx xxx xxxxxxxx xxxxx xx xx xxxxx xxx xxxxxxxxxx xxxx xxxxxxxx xxxxxxxxxxxx xx xxxxxxxxx xxxxxx xxx xxxx xxxxx xxxxxxxx xxxx xxxxxxxxx xx xx xxxxxxxx xxxxxxxx xxxx xxxxxxxxx xxxx xxxxxxxxx xxxxxxxxxxx xx x xxxxxxxx xxxxxxx xxxxxxxxxx Figure 22 shows

the evidence network for TTLR and Figure 23 shows the results of the analysis, while Figure 24 and Figure 25 show the evidence network and results for TTDM.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 22. Network diagram for TTLR for Scenario Analysis: Potentially resectable & completely resected, 3rd generation chemotherapies

==> picture [334 x 243] intentionally omitted <==

adjCT = adjuvant chemotherapy; neoCRT = neoadjuvant chemoradiotherapy; neoCT = neoadjuvant chemotherapy; neoNIVO-CT = neoadjuvant nivolumab-chemotherapy; S = surgery; TTLR = time to locoregional recurrence.

Note: The reference treatment is neoadjuvant chemotherapy.

Sources: Forde et al. (2022)[3,] Girard et al. (2010)[86,] Katakami et al. (2012)[87,] Pless et al. (2015)[88] Pisters et al. (2010)[90,] Douillard et al. (2006)[98,] Ou et al. (2010)[99]

Figure 23. Network meta-analysis output: vs. reference treatment (CT) for TTLR for Scenario Analysis: Potentially resectable & completely resected, 3rd generation chemotherapies (fixed effect model)

==> picture [435 x 108] intentionally omitted <==

adjCT = adjuvant chemotherapy; CrI = credible interval; HR = hazard ratio; neoCRT = neoadjuvant chemoradiotherapy; neoCT = neoadjuvant chemotherapy; neoNIVO-CT = neoadjuvant nivolumabchemotherapy; S = surgery; TTLR = time to locoregional recurrence.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 24. Network diagram for TTDM for Scenario Analysis: Potentially resectable & completely resected, 3rd generation chemotherapies

==> picture [335 x 241] intentionally omitted <==

adjCT = adjuvant chemotherapy; neoCRT = neoadjuvant chemoradiotherapy; neoCT = neoadjuvant chemotherapy; neoNIVO-CT = neoadjuvant nivolumab-chemotherapy; S = surgery; TTDM = time to death or distant metastases.

Note: The reference treatment is neoCT.

Sources: Forde et al. (2022)[3,] Girard et al. (2010)[86,] Katakami et al. (2012)[87,] Pisters et al. (2010)[90,] Douillard et al. (2006)[98,] Ou et al. (2010)[99]

Figure 25. Network meta-analysis output: vs. reference treatment (neoCT) for TTDM for Scenario Analysis: Potentially resectable & completely resected, 3rd generation chemotherapies (fixed effect model)

==> picture [435 x 111] intentionally omitted <==

adjCT = adjuvant chemotherapy; CrI = credible interval; HR = hazard ratio; neoCRT = neoadjuvant chemoradiotherapy; neoCT = neoadjuvant chemotherapy; neoNIVO-CT = neoadjuvant nivolumabchemotherapy; S = surgery; TTDM = time to death or distant metastases.

Adverse events

Quantitative synthesis of safety data was not conducted, as it was considered

inappropriate given the sparseness of the data and the differences in treatment regimens across the base-case studies. Only 3 studies reported grade 3 and 4 AEs; the proportion

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

of patients experiencing AEs ranged from 11% (S, Scagliotti et al. (2012)[91] ) to 60% (neoCT, Pless et al. (2015)[88] ).

B.2.9.4 Uncertainties in the indirect and mixed treatment comparisons

This quantitative analysis produced estimates of comparative efficacy of neoNIVO-CT relative to a broad range of therapies used for individuals with potentially resectable NSCLC.

Key strengths of the NMA included the use of an evidence base informed by a comprehensive SLR, a homogeneous base-case analysis (restricted to 3rd generation chemotherapies and potentially resectable patients) that was validated with extensive sensitivity analyses, and alignment with best practice guidance for the conduct of systematic reviews and network meta-analyses. Several limitations were apparent, however, mainly due to the limited body of evidence in resectable NSCLC. In the base-case network, connections were informed by at most 3 studies. Of the studies included, some were dated, were stopped early and were underpowered to detect a statistically significant benefit, had relatively short follow-up, and were somewhat heterogeneous in terms of stage (including the staging version) and outcome definitions. Across all analyses, the sparseness of the evidence also led to insufficient data to estimate the between-study standard deviation with enough precision, precluding the consideration of the random effects model.

Despite these limitations, the available evidence suggests that neoNIVO-CT confers a strong added benefit in terms of survival-based endpoints and pathological complete response versus neoadjuvant therapies and surgery alone. Analyses conducted in this NMA suggest that neoNIVO-CT lowers patients’ risk of experiencing disease recurrence/progression (as per CM816 EFS definition) and death relative to neoCT +/- radiotherapy, surgery alone, or adjCT. However, due to the sparseness and limited sample size within the evidence base, some comparisons were associated with considerable uncertainty, leading to nonstatistically significant estimates of relative effect in some cases. NeoNIVO-CT was also associated with statistically significant higher odds of achieving pCR relative to neoCT +/radiotherapy.

There is also evidence of benefit in terms of progression-based endpoints, such as time to locoregional recurrence and distant metastases. For TTDM, the findings consistently demonstrated a trend suggesting that neoNIVO-CT may be associated with approximately half the risk of distant metastasis relative to all comparators in the base case (neoCT, neoCRT, surgery alone; with HRs ranging from XXXX to XXXX), although estimates of relative effect were not statistically significant for the comparison with neoCRT. This finding is consistent with the hypothesis that utilising immunotherapy in the neoadjuvant setting is the earliest opportunity to treat micro-metastases. Similar trends were observed relating to TTLR, except for a change in the direction of association for TTLR between neoNIVO-CT and neoCRT (in favour of neoCRT), although it did not reach statistical significance.

The results from the TTDM and TTLR analyses for neoCRT align with the expectations that adding radiotherapy to neoCT may reduce the risk of locoregional recurrence relative to neoCT alone, via its localized delivery, but does not reduce the risk of distant metastases, as has been seen in other indications. Conversely, neoNIVO-CT conferred a similar magnitude of effect over neoCT for both TTDM and TTLRR. By extension, the magnitude of effect between neoNIVO-CT relative to neoCRT is stronger for TTDM than for TTLRR, although in the current analysis, none of the estimates were statistically significant due to small sample sizes and studies that were not powered to detect differences in these endpoints.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

In terms of AEs, no quantitative synthesis was conducted due to the paucity of data, however, the addition of NIVO to neoCT backbone in Forde 2022 did not lead to an increase in grade 3-4 treatment-related AEs.

B.2.10 Adverse reactions

Overall, nivolumab + PDC as neoadjuvant therapy was well tolerated, having a similar incidence of treatment-related AEs, surgery-related AEs, AEs leading to discontinuation, and serious AEs versus PDC. No new safety signals were observed.[3]

Adverse events of any cause occurred in 92.6% of the patients in the nivolumab + PDC group and in 97.2% of those in the PDC group.[3] The incidence of grade 3 or 4 treatmentrelated AEs was 33.5% and 36.9% in the respective groups (Table 19).[3]

Treatment-related AEs of any grade leading to discontinuation of treatment occurred in 10.2% of the patients in the nivolumab + PDC group and in 9.7% of those in the PDC alone group.[3]

Table 19. CheckMate-816: adverse events

AE = adverse event; NIVO = nivolumab; PDC = platinum doublet chemotherapy.

a Included are events reported between the first neoadjuvant dose and 30 days after the last neoadjuvant dose.

• b Treatment-related deaths in the PDC-alone group were due to pancytopenia, diarrhoea, and acute kidney injury (all in 1 patient); enterocolitis; and pneumonia.

c The denominators are based on patients who underwent definitive surgery. Included are events reported up to 90 days after definitive surgery. Grade 5 surgery-related AEs (defined as events that led to death ≤ 24 hours after the onset of an AE) were reported in 2 patients in the nivolumab + PDC group and were deemed by the investigator to be unrelated to the trial drugs (1 each due to pulmonary embolism and aortic rupture). Source: Forde et al. (2022)[3]

The most common grade 3 or 4 treatment-related AEs across both treatment arms were neutropenia (8.5% with nivolumab + PDC and 11.9% with PDC alone) and decreased neutrophil count (7.4% and 10.8%, respectively) (Table 20).[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 20. CheckMate-816: most frequent treatment-related adverse events (≥ 15% of patients in any treatment group) and surgery-related adverse events (≥ 5% of patients in any treatment group)

AE = adverse event; CTCAE = Common Terminology Criteria for Adverse Events; NIVO = nivolumab; PDC = platinum doublet chemotherapy.

a Included events reported between the first neoadjuvant dose and 30 days after the last neoadjuvant dose as per CTCAE Version 4.0; Medical Dictionary for Regulatory Activities (MedDRA) Version 23.0.

b Includes events reported up to 90 days after definitive surgery. CTCAE Version 4.0; MedDRA Version 23.0.

c Denominator based on patients with definitive surgery (n = 149 in the nivolumab + PDC group; n = 135 in the PDC group).

Source: Forde et al. (2022)[3]

Overall, the incidence of immune-mediated AEs was low, and events were mainly of grade 1 or 2. The most common immune-mediated AE of any grade with nivolumab + PDC was rash (in 8.5% of the patients); 2 patients (1.1%) had grade 1 or 2 pneumonitis. Three treatmentrelated deaths were noted, all in the PDC alone group.[3]

Adverse events of any grade led to delayed surgery in 3.4% of the patients receiving nivolumab + PDC and in 5.1% of those receiving PDC alone and led to cancellations in 1.1% and 0.6%, respectively. Adverse events of any grade that were identified as surgical complications occurred in 41.6% of the patients in the nivolumab + PDC group and in 46.7% of those in the PDC alone group; grade 3 or 4 surgery-related AEs occurred in 11.4% and 14.8% of the patients in the respective groups. Grade 5 surgery-related AEs were reported in 2 patients treated with nivolumab + PDC and were deemed to be unrelated to the trial drugs by the investigator (1 each due to pulmonary embolism and aortic rupture).[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 21. CheckMate-816: adverse events leading to surgery delay and/or cancellation

AE = adverse event; NIVO = nivolumab; PDC = platinum doublet chemotherapy. Source: Forde et al. (2022)[3]

B.2.11 Supporting studies

The following studies provide evidence for the efficacy of nivolumab in the neoadjuvant setting and support the CheckMate-816 trial.

B.2.11.1 Study CA209-159

The phase 1 study, CA209-159 (NCT02259621), investigated nivolumab monotherapy and nivolumab + ipilimumab as neoadjuvant therapy in patients with resectable NSCLC.[105] Early results indicated that nivolumab as neoadjuvant therapy is generally well tolerated in patients with newly diagnosed resectable stage I–IIIA disease and does not impede feasibility of surgery. Nivolumab neoadjuvant therapy induced MPR in 45% of patients.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Updated 5-year outcomes have recently been published.[106] At a median follow-up of 63 months, 3-, 4- and 5-year survival rates were 85%, 80%, and 80% respectively and recurrence-free survival rates at 3-, 4- and 5-years were 65%, 60%, and 60% respectively. Furthermore, at 5-year follow-up, 8 of 9 (89%) patients with a MPR were alive and no cancer deaths had occurred. Amongst patients with a MPR (n = 9), 1 had a cancer recurrence in the mediastinum treated successfully with definitive chemoradiotherapy. Both patients with a pCR were alive and without recurrence.

B.2.11.2 NEOSTAR

NEOSTAR (NCT03158129) was a phase 2 randomised trial comparing neoadjuvant therapy consisting of nivolumab, nivolumab + ipilimumab, or nivolumab + PDC, followed by surgery in patients with operable NSCLC (stage IA–IIIA, AJCC/UICC seventh edition).[78] The arms investigating nivolumab alone and nivolumab + ipilimumab are not relevant to this submission and are not described here. In the nivolumab + PDC arm, nivolumab was given at a dosage of 3 mg/kg on days 1, 15, and 29 or 360 mg intravenously + PDC (cisplatin on days 1, 22, and 43 with docetaxel or pemetrexed on days 1, 22, and 43, all intravenously).[107]

The results of the NEOSTAR study support the results from CheckMate-816. In the nivolumab + PDC group (n = 22), 86% of patients completed the planned neoadjuvant therapy; all of these patients underwent curative surgery, with 91% achieving complete resection. In the overall population, 32% of patients achieved an MPR, 4 patients (18%) had a pCR, and the objective response rate was 41%, including 9 partial responses. No new safety concerns were noted.[78]

B.2.11.3 NADIM

The phase 2 NADIM trial (CA209-547, NCT03081689) provides supporting evidence relating to the feasibility and safety profile of neoadjuvant nivolumab + PDC, and the pCR rates achieved with the treatment in patients with newly diagnosed resectable stage IIIA disease.[76]

All 46 patients completed the planned course of neoadjuvant therapy and 41 (89%) proceeded to undergo surgery, all of whom achieved complete resection, while pathological downstaging was observed in 37 (90%) of patients.[108] There was no operative mortality at either 30 or 90 days. Of patients who underwent surgery, 37 (90%) received adjuvant nivolumab for a median of 10.8 months,[76] and 12 (29%) experienced postoperative complications.

A total of 34 (83%) patients of those who underwent surgery achieved at least an MPR, including 26 (63%) who had a pCR. 88.4% of patients with an MPR were progression-free at 2 years, versus 77.1% of the total modified intention-to-treat (ITT) population. Similarly, of the 26 patients (63%) who achieved a pCR, 96.2% were progression free at 2 years, a difference that was statistically significant versus patients with a MPR ( P = 0.041) or an incomplete pathological response ( P = 0.0023).[76] These data thus support the expected survival benefit of achieving pCR.

An updated analysis with a median follow-up time of 37.9 months demonstrated that PFS at 36 and 42 months in the ITT population were 69.6% (95% CI, 54.1-80.7) in both cases. Similarly, PFS at 36 and 42 months in the per-protocol population were 81.1% (95% CI, 64.4-90.5) in both cases. The percentage of patients who were alive at 36 and 42 months in the modified ITT population were 81.86% (95% CI, 66.8-90.6) and 78.94% (95% CI, 63.188.6), respectively.[109]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Neoadjuvant therapy was generally well tolerated. Approximately a third (30%) of patients experienced ≥ grade 3 AEs, with the most frequently reported being increased lipase (7%) and febrile neutropenia (7%). None of the AEs experienced during neoadjuvant treatment led to treatment discontinuation, dose reduction, surgery delay or death. The most frequently reported AEs of any grade (> 30% of patients) were asthenia or fatigue (24 patients), alopecia (17 patients) and neurotoxicity (15 patients).[76]

B.2.11.4 NADIM II

NADIM II (NCT03838159) is an open-label, randomised, two-arm, phase II, multicentre clinical trial conducted in 87 patients with resectable locally advanced disease, and supports the superiority of neoadjuvant nivolumab + PDC versus PDC in terms of pCR, as well as the feasibility of surgery, with a moderate increase in grade 3-4 toxicity.[110] Nivolumab + PDC in the neoadjuvant setting significantly increased the pCR rate versus PDC (36.2% vs. 6.8%; relative risk 5.25 [99% CI, 1.32-20.87]; P = 0.0071). Nivolumab + PDC also improved MPR rates versus PDC (52 % vs. 14 %), as well as objective response rate (74 % vs. 48%). Definitive surgery occurred for 91% of patients treated with nivolumab + PDC and 69% with PDC; surgery was cancelled rarely due to AEs (1 patient/experimental arm) and due to disease progression in 1 and 4 patients in the experimental and control arm respectively. Grade 3-4-related AEs were reported in 24 % vs. 10% in the nivolumab + PDC versus PDC arms, respectively. In the ITT experimental arm, patients with pCR had higher PD-L1 Tumour Proportion Score (TPS) (median 70%, interquartile range 5%-90%) versus nonresponders (median 0%, interquartile range 0%-37.5%, P = 0.0035). Area under the curve to predict pCR was 0.734 (95% CI, 0.59-0.88; P = 0.005). The pCR rate rose across increasing categories of PD-L1 TPS (< 1% 14.3%; 1%-49% 41.7%; ≥ 50% 61.1%; P = 0.008).[110]

A further analysis with a median follow-up time was 21.9 months (95% CI, 18.7-23.3). showed that PFS at 24 months was 67.3% (95% CI, 55.5-81.6) for patients treated with nivolumab + PDC versus 52.6% (95% CI, 36.8- 75.2) for patients treated with PDC (HR, 0.56; 95% CI, 0.28-1.15; P = 0.117). Overall survival at 24 months was 85.3% (95% CI, 75.7-96.1) with nivolumab + PDC versus 64.8% (95% CI, 47.4-86.4) with PDC (HR, 0.37; 95% CI, 0.14-0.93; P = 0.003). In the experimental arm, PD-L1 expression (≥ 1%) significantly identified patients with improve PFS (HR, 0.26; 95% CI, 0.08-0.77; P = 0.015). Pathological complete response rate was 36.2% in the experimental arm versus 6.8% in the control arm ( P = 0.007). None of the patients showing a pCR has progressed or deceased ( P log-rank < 0.001 and P log-rank = 0.013 for PFS and OS, respectively).[111]

B.2.11.5 NCT03366766

NCT03366766 was an investigator-initiated trial including 13 patients with newly diagnosed locally advanced NSCLC with a potentially resectable tumour.[112] Pre-surgical grade 3 toxicity occurred in 2 of 13 patients treated with nivolumab + PDC, 1 of whom was changed to carboplatin for courses 2 and 3. Grade 3 toxicities were neutropenia (2/13), anaemia (1/13), and renal (1/13). One patient developed hypothyroidism 4 months after surgery. One patient died 6 weeks after surgery from complications unrelated to study drugs. The primary endpoint was met: 11/13 (85%) patients had at least an MPR with 6/13 (46%) patients and 5/13 (38%) patients having an MPR and pCR respectively. Radiologic response rate was 46% (partial response 5, complete response 1). Patients with either PD-L1-positive or PD-L1-negative status had MPRs. No recurrences were seen within the median follow-up of 10 months.[112]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.2.12 Ongoing Bristol Myers Squibb nivolumab randomised controlled trials

CheckMate-816 is ongoing. Results from xxxxxxx xxxxxxxx x are expected to be available within the timelines of this appraisal and the xxxxxxxx xxxx will be provided once available. However, the timeframe of the appraisal will not allow xxxxxxxxxxxxx xxxx xxx

xxxxxxxxxxxxxxxxx xxxxx.

Further support for nivolumab + PDC as neoadjuvant therapy for resectable NSCLC is likely to come from the ongoing phase 3 trial CheckMate-77T, which is being performed in sites in North America, South America, Europe, Asia, and Australia.[113] This trial is a randomised and double-blinded, whereby patients are randomly assigned to receive either:

• Neoadjuvant nivolumab + PDC followed by surgery followed by adjuvant nivolumab

• Placebo + PDC followed by surgery followed by placebo

The trial involves newly diagnosed patients with resectable stage IIA–IIIB (T3N2 only, eighth edition of AJCC/UICC) NSCLC and having an ECOG PS of 0-1. Patients with EGFR/ALK mutations, brain metastasis, previous systemic anticancer treatment or radiotherapy, and autoimmune disease are excluded. The trial aims to enrol approximately 452 patients. The primary endpoint is EFS, assessed by BICR. Secondary endpoints include OS, pCR, and MPR assessed by BIPR, as well as safety and tolerability.

Nivolumab is also being investigated in the ANVIL trial, a randomised, open-label, phase 3 study of nivolumab after surgical resection in patients with stage IB-IIIA NSCLC being conducted by the National Cancer Institute.[114,115] ANVIL is being performed in over 600 sites in the United States (US), and patients are randomly assigned to either nivolumab or SOC observation and stratified by stage, histology, prior adjuvant treatment, and PD-L1 status (≥ 1% or < 1%). Adjuvant chemotherapy and/or radiotherapy is allowed but not required. The trial involves patients who have undergone complete surgical resection of their tumour (stage IB [≥ 4 cm], II, or IIIA NSCLC [AJCC/UICC seventh edition]) and have had negative surgical margins. The trial aims to enrol approximately 900 patients; DFS and OS are primary endpoints, and the incidence of AEs is a secondary endpoint.[114,115]

B.2.13 Interpretation of clinical effectiveness and safety evidence for nivolumab + PDC in CheckMate-816

Nivolumab is an immuno-oncology agent that acts to restore the body’s natural antitumour response by inhibiting the suppression that the tumour exerts on antitumour immune responses. Used before surgery, nivolumab-based regimens are expected to help prime the body’s immune response not only to target primary tumour cell activity before surgery and promote responses against micro-metastases already present but to kill tumour cells released during surgery, limiting recurrence.[9]

CheckMate-816 is the first phase 3 study to demonstrate significantly improved EFS and pCR for an immuno-oncology–based combination in the neoadjuvant setting of resectable stage IB-IIIA NSCLC.[3] Further, the efficacy of the regimen is achieved with only 3 cycles of treatment, which is expected to reduce burden on patients, limit the occurrence of treatment-

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

related AEs, improve adherence, and reduce costs versus other immuno-oncology regimens which often require up to 2 years of therapy in later lines.

Feasibility of surgery was maintained with nivolumab + PDC versus neoadjuvant PDC alone: numerically more patients underwent surgery, had a less invasive surgery and had a shorter duration of surgery with similar length of hospital stay.[3]

Nivolumab + PDC demonstrated a significantly longer EFS versus PDC alone (HR, 0.63; 97.38% CI, 0.43-0.91).[3] Significantly more patients treated with nivolumab + PDC had a pCR versus those treated with PDC alone; longer EFS was observed in patients who achieved pCR versus those who did not across both arms of CheckMate-816. Therefore, the higher rates of pCR for patients treated with nivolumab + PDC may result in increased EFS benefit with longer follow-up.[3] Of note, as described in Section B.1.3.6.5, analyses using data from clinical trials of neoadjuvant treatment in patients with stage IB-IIIA NSCLC have demonstrated an association between pCR and both improved EFS and improved OS, which is consistent across subgroups.[72-75]

Further, a prespecified interim analysis for OS in CheckMate-816 showed a promising early trend in OS. Although these OS data are not yet mature, they suggest, alongside the significant comparative difference in pCR and EFS for nivolumab + PDC versus PDC alone and the demonstrated association between pCR and OS and between EFS and OS, that a significant difference in OS may be observed once more events have occurred.

The safety profile of 3 cycles of nivolumab + PDC was consistent with safety profiles in other disease settings and tumours. Furthermore, nivolumab + PDC did not increase postsurgical complications.[3]

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.3 Cost-effectiveness

• A de novo four-state semi-Markov model was developed to assess the costeffectiveness of nivolumab with chemotherapy for the neoadjuvant treatment of resectable NSCLC compared with current SOC in the UK. The modelled health states were Event-Free (EF), Locoregional Recurrence (LR), Distant Metastasis (DM), and Dead.

• The modelled population and key clinical inputs were based on the CheckMate816 trial. Comparators considered were, neoadjuvant chemoradiation (CRT), surgery alone, and adjuvant PDC. As CheckMate-816 did not include individual treatment arms for neoadjuvant CRT, adjuvant PDC, or surgery alone, efficacy estimates for these treatments are based on an indirect treatment comparison conducted using data available in the published literature.

• The model considered costs related to treatment acquisition and administration, surgery, AEs, terminal care, ongoing medical resource use (MPR) (e.g., tests, scheduled medical specialist visits).

• ▪ Health-state utility values were based on CheckMate-816. The model also considers the disutility of grade 3 and 4 AEs.

• The results of the cost-effectiveness analysis showed improved survival for patients treated with nivolumab + PDC, versus all comparators. This resulted in an increase of xxx, xxx and xxx quality-adjusted life-years (QALYs) versus surgery alone, neoadjuvant CRT, and adjuvant PDC respectively. Based on the current simple patient access schemes for nivolumab, approved by the Department of Health, this resulted in an incremental cost-effectiveness ratio (ICER) of £2,685 per QALY versus surgery alone and being dominant versus both neoadjuvant CRT and adjuvant PDC. The results were robust for all scenario analyses conducted with only one extreme scenario resulting in an ICER above £20,000 per QALY.

• In conclusion, nivolumab + PDC offers an innovative, clinically effective treatment and cost-effective option in resectable non-metastatic NSCLC.

B.3.1 Published cost-effectiveness studies in non-metastatic NSCLC

In the targeted review for health economic models in resectable non-metastatic NSCLC, only 2 studies were identified to potentially inform the model design (see Appendix G). One study sought to investigate the cost-effectiveness of adjuvant vinorelbine plus cisplatin versus observation only,[116] and the other sought to investigate the relative cost-effectiveness of PDC options in the neoadjuvant setting.[117] Both studies were based directly on their respective trials, and neither study presented an explicit description of their model structure. As such, they were not found to be useful for aiding in the development of a costeffectiveness analysis for nivolumab + PDC in non-metastatic NSCLC. In addition, 2 other sources were identified through a review of the health authority bodies of Canada and the UK; both studies focused on adjuvant osimertinib versus active monitoring in EGFR mutation-positive NSCLC after complete tumour resection (Table 22).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 22. Summary list of published cost-effectiveness studies

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

• CAD = Canadian dollar; CEA = cost-effectiveness analysis; CI = confidence interval; EGFR = epidermal growth factor receptor; ERG = evidence review group; ICER = incremental cost-effectiveness ratio;

  • ICUR = incremental cost-utility ratio; LYG = life-year gained; NR = not reported; NSCLC = non-small cell lung cancer; QALY = quality-adjusted life-year; UK = United Kingdom.

• a Patients with stage IB disease and a tumour size of at least 4 cm were treated with adjuvant chemotherapy, while those with stage IB disease and a smaller tumour size were observed.

• b All patients have their genes profiled, and those that were deemed high risk were given adjuvant treatment, while those with low-risk disease were observed.

• c Patients with stage IB disease had gene profiling performed and high-risk patients received adjuvant treatment along with patients who had stage II disease.

• Source: BMS data on file (2022)[122]

B.3.2 Economic analysis

A de novo model was developed for this analysis because no appropriate model was identified through the targeted literature review (see Section B.3.1). This section describes the de novo economic model constructed for the submission, and the rationale for the model development.

B.3.2.1 Patient population

The population evaluated in the analysis is aligned with the indication for nivolumab in the neoadjuvant treatment of resectable (tumours ≥ 4 cm or node positive) NSCLC in adults.[1] Patient characteristics for the model were based on those in CheckMate-816 and have been validated by clinical expert feedback to be relevant for England (Appendix N). Table 23 presents specific baseline characteristics, such as age, sex, and disease stage.

Table 23. CheckMate-816: population characteristics

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

NIVO = nivolumab; PDC = platinum doublet chemotherapy; ECOG = Eastern Cooperative Oncology Group; NE = not estimable; PD-L1 = programmed death-ligand 1; TPS = Tumour Proportion Score. Source: BMS data on file (2021)[83]

B.3.2.2 Model structure

Given that no published models identified through the literature review were deemed suitable for the current submission (see Section B.3.1) a de novo model was developed. To inform the model development a supplementary review of recent health technology assessment (HTA) submissions for adjuvant/neoadjuvant treatments in non-metastatic solid tumours was conducted in 2021 to help inform the model structure. In addition to the review, input from experts in clinical and health economics and outcomes research was sought through several advisory boards and expert meetings to inform the model development. A summary of each meeting is provided in Appendix N.

1. UK HE expert meeting, February 2022

2. UK HTA clinical expert meeting, March 2022

3. Global HTA advisory board meeting, May 2022 (included 1 UK clinical expert and 1 UK HE expert)

4. UK clinical advisory board, May 2022

5. UK HTA clinical validation meeting, August 2022

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

For the supplementary review of HTA submissions, no submissions in the neoadjuvant or adjuvant treatment setting were available for NSCLC;[v] therefore, the search examined other solid-tumour cancers, where surgery combined with neoadjuvant or adjuvant treatment were part of the treatment pathway. The NICE website was searched first, as a baseline, and the same agents and indications were subsequently sought out in other HTA websites. The specific HTAs included were NICE (UK), Canadian Agency for Drugs and Technologies in Health (CADTH), Pharmaceutical Benefits Advisory Committee (PBAC; Australia), Scottish Medicines Consortium (SMC), and French National Authority for Health (HAS).Table 24 summarises the submissions identified.

Table 24. HTA submissions for adjuvant/neoadjuvant treatments in nonmetastatic solid tumours

CADTH = Canadian Agency for Drugs and Technologies in Health; EGFR = epidermal growth factor receptor; HAS = French National Authority for Health; HER2 = human epidermal growth factor receptor 2; HTA = health technology assessment; NSCLC = non-small cell lung cancer; PBAC = Pharmaceutical Benefits Advisory Committee; SMC = Scottish Medicines Consortium.

a The only submission for pancreatic cancer was in advanced/metastatic disease. There were no HTA submissions for kidney cancer, and the only relevant HTA submission for colorectal cancer was older than 5 years and thus excluded.

v Note, since the original review was conducted, the NICE appraisal of atezolizumab after adjuvant PDC has been published,[2] but it was not considered during the early stages of model development.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Among the 21 submissions reviewed, Markov models were the most frequent model structure, accounting for 17 of the reviewed models. The remaining 4 models took a partitioned survival analysis (PartSA) approach, and all were related to submissions in support of darolutamide as a treatment of patients who are at high risk of non-metastatic, castration-resistant prostate cancer. NICE feedback regarding the darolutamide submission also suggested that the treatment pathway would have been better accommodated by a Markov model.[123]

Given the early stage of disease and associated immature time-to-event data from CheckMate-816, a key limitation of PartSA framework is that the extrapolations of frequently used outcomes such as EFS and OS could generate unrealistic estimated hazards (i.e., the risk of an event in the OS curve being higher than the risk of an event in the EFS curve). Moreover, incomplete OS data from CheckMate-816 presents challenges in generating a robust surrogate relationship with EFS which would be required for a traditional PartSA. In early-stage NSCLC, subsequent treatments are expected to have a considerable impact on long-term outcomes given the likelihood of different recurrence events (conditional on locoregional or distant metastasis) and multiple lines of therapy. For this reason, a model framework with more than 3 health states is required to accurately map the prognostic nature of intermediate health states in the extrapolation period, and for differential treatment effects, resource use and costs to be applied to different states of the disease process.[124]

Since the targeted literature review was performed, there have been 2 NICE appraisals submissions within the non-metastatic NSCLC treatment pathway. The first NICE submission was for osimertinib as an adjuvant treatment in non-metastatic NSCLC. Although it was released after the search, it was considered to be highly relevant because it focused on resectable NSCLC and was published during the development stages of the model. For this submission, a 5-state Markov model was developed. Although the evidence assessment group (EAG) objected to specific aspects of the approach (e.g., how different types of treatment costs were applied), it found that the model structure and approach were generally appropriate to the disease area and decision problem. The second appraisal was in support of the use of atezolizumab as adjuvant treatment of patients with resected NSCLC. This appraisal was published very close to the submission of the current submission, but NICE feedback on the atezolizumab appraisal was carefully considered to ensure that the costeffectiveness model (CEM) for nivolumab + PDC in neoadjuvant non-metastatic NSCLC would be appropriate for NICE decision-making. Table 25 presents key feedback and how it is addressed in the neoadjuvant nivolumab + PDC model.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 25. Appraisal committee meeting feedback on atezolizumab in adjuvant resected NSCLC

BMS = Bristol Myers Squibb; CEM = cost-effectiveness model; DM = distant metastasis; DSU = Decision Support Unit; EFS = event-free survival; EAG = evidence assessment group; I-O = immuno-oncology; ITC = indirect treatment comparison; LY = life-year; NHS = National Health Service; NSCLC = non-small cell lung cancer; PDC = platinum doublet chemotherapy; QALY = quality-adjusted life-year.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

In summary, the results of the HTA review suggested that a Markov approach with 4 or more health states has a strong precedent in early disease modelling and is particularly useful when later lines of treatment could be important to the decision question.

Based on the findings from the HTA review, both clinical and health economics and outcomes research experts were consulted in the Global HTA advisory board meeting, May 2022, to inform the development of the economic model (see Appendix N). The health economics and outcomes research experts surveyed noted that current data limitations could impact the feasibility of PartSA and thus a Markov model were seen to be preferable. Further, the experts expressed that a 4-state model is well aligned with the natural history of the disease versus a 3-health-state model. Patients who progress may experience significantly different outcomes on the basis of the type of progression experienced. For example, patients with local or regional recurrence will, on average, live longer than patients with distant metastasis, and experience a higher quality of life. Conversely, patients with distant metastasis would be anticipated to require more intensive care (both in terms of healthcare resource utilisation and treatment received), and also experience a greater reduction in quality of life than patients with local recurrence. Additionally, the duration of treatments in the 2 states varies significantly: advisory board feedback indicated that patients with LR would be anticipated to receive limited care, whereas patients with distant metastasis would need lifelong treatment.

In concordance with the natural history of the disease, previous neoadjuvant and adjuvant HTA submissions, and feedback from clinical and economic experts, a 4-state semi-Markov cohort model was developed and implemented in Excel (Figure 26).

Figure 26. Four-state semi-Markov model diagram

==> picture [341 x 268] intentionally omitted <==

DM = Distant Metastasis; LY = life-year; QALY = quality-adjusted life-year.

• A one-off cost, life-year and quality-adjusted life-year consequence is applied to patients entering distant metastasis. The subsequent transition from DM to Dead is implicitly captured by the one-off cost, LY and QALY amount applied to patients entering that state but is not explicitly tracked in the model.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

This model includes 4 health states: Event-Free (EF), Locoregional Recurrence (LR), Distant Metastasis (DM), and Dead. All patients enter the model in the EF health state, where patients may experience 1 of 2 types of progression: LR or DM. Additionally, patients in the EF health state may also die, moving to the Dead health state. Patients in the LR health state may experience further progression, moving to the DM or Dead health states.

When patients experience distant metastasis, a one-off cost, QALY, and life-year (LY) total representing subsequent treatment mix is applied; further outcomes are no longer explicitly tracked, and the patient does not make any further state transitions. Patients in all health states except DM are subject to a probability of death each cycle (for patients in DM, this probability is implicitly considered in the LY and QALY total applied). The rationale for providing “oneoff” outcomes for the distant metastasis health state it outlined in Section B.3.3.2.

Health states were selected based on the CheckMate-816 trial endpoints and the current understanding of the disease area (see Section B.2.3.1). The EF health state was designed to align with the definition of EFS used in CheckMate-816, where EFS was a primary endpoint (see Section B.2.2). Event-free survival begins from the time of randomisation, rather than from the time of surgery. This allows the model to indicate the possibility that some patients who are unresponsive to neoadjuvant therapy could see their disease progress prior to surgical resection and the prognoses of these patients will be captured in the time-to-event data.

The definitions of LR and DM also follow their given definitions from the CheckMate-816 trial. Specifically, patients with LR had disease progression per Response Evaluation Criteria in Solid Tumours (RECIST) 1.1 criteria but did not have any identified lesions outside the thorax; conversely, patients with DM were those who had developed 1 or more new lesions outside the thorax. Separation of the composite trial endpoint of EFS from CheckMate-816 into time to LR and DM also lends credibility to any modelled survival benefit. This is because separate health states for LR and DM allow the model to accurately capture real-life patterns of progression, by considering distinct survival and progression trajectories, quality of life, and costs associated with each type of progression.

In general, the transition probabilities used by the model are computed based on results from CheckMate-816. One exception is for transitions out of the LR state, which are informed by publicly available data in the literature; this transitional probability was not possible to inform from the trial because further study imaging was not required for patients who experienced documented locoregional recurrence without a distant lesion. Greater detail on the analyses used to derive transition probabilities is provided in Section B.3.3.1.

Once patients experience an event, appropriate treatment is initiated based on the type of event. Costs were assigned to each health state, and utilities were applied according to patients’ disease progression status, treatment received, and any AEs experienced.

Following NICE (2019)[17] , patients in the LR state can receive treatment in up to 3 modalities: PDC, radiotherapy, and surgery. Because patients may receive more than one type of treatment (e.g., patients could concurrently receive both PDC and radiotherapy), the

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

distribution is not mutually exclusive and can total a value greater than 100%. Additional detail on the costs of care considered in the LR is provided in Section B.3.5.2.1. The care received in LR is not explicitly considered in terms of how it may affect the mortality and likelihood of experiencing subsequent DM, which is assumed to be the same for all patients in the model, regardless of the treatment received or time spent in the EF state. Additional detail on the inputs used to describe the likelihood of DM for patients with LR is provided in Section B.3.3.1.4, and the mortality for patients with LR in Section B.3.3.1.6.

As discussed during the decision problem meeting with NICE, the model applies one-off cost, QALY, and LY values to patients entering the DM state to capture the outcomes for this population; no further explicit modelling of outcomes was implemented to inform the costs and consequences of transitions into this state (e.g., treatment received, duration of therapy, PFS, and OS). The specific cost, QALY, and LY totals accrued are based on discounted values from existing NICE appraisals of first-line metastatic NSCLC treatments provided by NICE. The treatments are weighted such that the shares of patients who are receiving immuno-oncology and chemotherapy treatment are reflective of UK clinical practice.

B.3.2.3 Perspective

In accordance with NICE guidelines,[125] the model takes an NHS perspective, considering direct costs incurred by the NHS, i.e., drug acquisition and administration costs, surgery costs, routine medical resource use (MRU) costs, terminal costs, and treatment-related AE costs.

B.3.2.4 Time horizon

The model takes a lifetime horizon in the base case, although it has the flexibility to consider shorter time horizons (e.g., 1 year, 5 years, 15 years, or 20 years). There are 2 key reasons why a lifetime horizon is employed in the base case. First, nivolumab + PDC is anticipated to extend patient lifespans based on the strong EFS performance we have observed in the IA2 data. Second, resected patients will require continued check-ins and other care over time; many will require subsequent treatment if their disease progresses. Thus, a lifetime horizon can fully capture the costs and benefits of nivolumab + PDC as neoadjuvant treatment of patients with resectable non-metastatic NSCLC.

Lifetime is implemented as the point in time when fewer than 1% of patients are alive in the model engine. Based the modelled survival, we found that this threshold was crossed at approximately 34.5 years. Rounding up to the nearest integer, the lifetime time horizon used in the base-case analysis was 35 years.

B.3.2.5 Cycle length and half-cycle correction

The model adopts a 21-day (i.e., 3-week) cycle length. This aligns with the treatment schedule for nivolumab and PDC in the CheckMate-816 trial; treatments are administered once every 3 weeks (see Section B.2.3.1). The 3-week cycle length also aligns with the dosage schedule for many current treatment options in the adjuvant and post-progression settings (e.g., docetaxel, pemetrexed, cisplatin, carboplatin, paclitaxel, atezolizumab, and pembrolizumab; see Table 7).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Most results are adjusted using a half-cycle correction, distributing the costs, LYs, and QALYs accrued across the cycle duration. Half-cycle correction is not applied to drug acquisition and administration costs because all patients received pharmacological treatment at the start of each cycle.

B.3.2.6 Discounting

Both cost and health outcomes are discounted based on the time at which they are accrued in the model to capture the opportunity cost of money and to properly evaluate the present value of future outcomes. This is set to 3.5% per annum for both cost and health outcomes in the base case, per NICE recommendations.[12] 5 The model includes the flexibility to adjust both discounting rates to allow for adaptation to settings that may recommend different rates other than the 3.5% per annum recommended by NICE, and for the purposes of sensitivity analysis.[125]

B.3.2.7 Intervention technology and comparators

As described in Section B.1.3.6 and the decision problem (see Table 1), surgical resection is the SOC for early-stage disease if the tumour is resectable and the patient is suitable for surgery. In addition to surgery alone, the patient may also receive neoadjuvant chemoradiotherapy or adjuvant PDC treatment. As a result, the current analysis investigates the cost-effectiveness of neoadjuvant nivolumab + PDC versus surgery, neoadjuvant CRT, and adjuvant PDC, as outlined in Table 26 and as set out by the final NICE scope. Note that analyses comparing nivolumab + PDC to adjuvant atezolizumab (TA10751]) have not been generated as adjuvant atezolizumab has entered the Cancer Drugs Fund as is not a relevant comparator for this appraisal.

Table 26. Comparators in the economic analysis

CRT = chemoradiation; PDC = platinum doublet chemotherapy.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.3.3 Clinical parameters and variables

Table 27 presents a high-level summary of transition probabilities considered by the model, and sources to inform the base-case analysis.

Table 27. Summary of clinical inputs

DM = Distant Metastasis; EF = Event-Free; HTA = health technology assessment; KOL = key opinion leader; LR = Locoregional Recurrence; LY = life-year; NSCLC = non-small cell lung cancer; QALY = quality-adjusted life-year.

B.3.3.1 Parametric survival modelling for Event-Free and Locoregional Recurrence transition probabilities

B.3.3.1.1 Approach to parametric fitting and assessment

To allow the model to accurately reflect clinical data from CheckMate-816, parametric survival modelling was conducted. This allows for estimation of time-dependent transition probabilities beyond the end of the existing Kaplan-Meier trial data. Parametric survival modelling is the common approach for extrapolation of time-to-event outcomes and is recommended by the UK’s NICE DSU for the analysis of survival outcomes for economic evaluations, alongside clinical trials for projection.[126] Figure 27 presents the approach to parametric survival modelling.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 27. Survival model selection process algorithm presented by NICE Decision Support Unit and referenced by other HTA agencies

==> picture [353 x 400] intentionally omitted <==

AFT = accelerated failure time; AIC = Akaike information criteria; BIC = Bayesian information criteria; HTA = health technology assessment; PH = proportional hazard.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Parametric survival modelling assumes that the time-to-event outcome follows a parametric distribution. The following distributions were investigated: exponential, Weibull, Gompertz, log-logistic, log-normal, gamma, and the generalised gamma distribution. If none of these distributions produced a reasonable fit, more sophisticated fittings, specifically, spline models, were explored. The properties of these distributions have been described by Ishak et al. (2013)[127] and can be found in standard textbooks on survival analysis, such as in the book by Collett (2003)[128] . They cover a broad range of possible shapes of hazards. The analytical process involves testing various potential statistical distributions and assessing fit over the observed data period and beyond to ensure reliable projection. The approach to fitting the curves is outlined in Ishak et al. (2013)[127] , which provides the technical details behind the steps of the analyses described below. The process of selecting a best-fitting distribution involves both statistical and clinical considerations, as well as considerations based on the observed data in assessing goodness-of-fit and plausibility of results. Thus, although the process involves distinct steps, it is not necessarily algorithmic. The comparator arm in CheckMate-816, neoadjuvant PDC, is not considered SOC in the UK and thus not a comparator within scope in the current appraisal. Nevertheless, extrapolations were conducted for both arms of the trial. The rationale for extrapolation of both arms was that the neoadjuvant PDC arm is the common comparator in the evidence network and thus can be used for extrapolation of other comparators within scope through the indirect treatment comparison–derived HRs.

Fitting of models

Equations based on distributions were fitted to the data. This was done using the flexsurv package in R. Spline fittings were conducted using the STPM2 package in STATA. This procedure produces estimates of the scale and shape parameters of the distributions and allows the scale parameter to be regressed on explanatory variables. The Akaike information criteria (AIC) and Bayesian information criteria (BIC) were compared among fitted models to determine best statistical fit (with lower values indicating better fit). Regression models were fitted both by including treatment (nivolumab + PDC vs. PDC alone) as a categorical explanatory variable and by conducting fully stratified analyses where separate models are fitted for each treatment.

Graphical assessment of fits

For all parametric distributions, graphical assessment of model fit will be made by comparing the Kaplan-Meier survival curves with the fitted survival curves. Furthermore, for the exponential, Weibull, log-logistic, and log-normal distributions, assessment of fit will also be made based on diagnostic plots associated with each of these distributions. For example, a plot of log-survival probabilities against time can be used to assess fit for an exponential model; log-negative log-survival versus log of time can be used for Weibull, etc. A linear pattern in these graphs indicates that the distribution may be adequate and, conversely, deviation from linearity indicates poor fit. Because of the parametric formulation of the generalised gamma distribution, assessment of fit based on diagnostic plots is not possible. Similarly, graphical tests for the Gompertz distribution are problematic because they do not rely purely on the observed data but also on unverifiable assumptions.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

The statistical and graphical fit to the Kaplan-Meier data is meant to provide a preliminary assessment of suitability of each distribution for consideration in further assessments; some distributions may be cut from further consideration, but it is possible that all are carried forward despite poor fit for comparison purposes.

Assessment of clinical plausibility

Fitted parametric models must provide good fit over the observed period and plausible extrapolations beyond. The former is assessed in the previous step with the AIC and BIC statistics and diagnostic plots. The latter must rely on clinical judgement of plausibility of extrapolations and may be assessed by examining the shape of the long-term projection of the curve, and based on measures derived from the predicted curve. For example, the estimate of the median and/or mean event time, or 3-year survival can be a way of judging validity; estimates that contradict common perceptions would be indicative of inadequate fit. To assess long-term plausibility, extrapolations were compared against relevant KaplanMeier curves from the literature for individual endpoints. Clinician input was also sought to check whether extrapolated curves were concordant with expert opinion. Finally, the OS predicted by the model (including all extrapolations) was compared with a constructed conditional survival built using available evidence (e.g., meta-analyses of neoadjuvant and adjuvant trials, and registry data).

This information was used to determine the best-fitting parametric model. However, best fitting does not necessarily imply good fit; the best-fitting distribution may still deviate from the observed data or produce clinically implausible long-term projections. If this is the case, alternate methods may need to be considered.

Output of standard parametric survival models

The output from the standard parametric survival models includes:

• Diagnostic graphs for each of the commonly used distributions, including probability plots

• A table summarising estimated values of the parameters for the fitted distributions, including variance covariance matrix, fit statistics, and projected median and lifeexpectancy estimates

• Observed versus predicted plots for each fitted model

• Long-term projection plots for each fitted model

B.3.3.1.2 Time to Locoregional Recurrence health state

Observed data

Estimates for transition probabilities characterising TTLR were based on data collected from the CheckMate-816 trial. In total, xx LR events were observed in the nivolumab + PDC arm (xxx), and xx LR events were observed in the PDC arm (xx). The median time to LR could not be estimated for either arm, although a lower 95% CI could be estimated for the PDC arm. The data translated to an overall HR of 0.65 for nivolumab + PDC versus PDC alone. A

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

summary of the observations, and the Kaplan-Meier curves, are depicted in Table 28 and Figure 28, respectively.

Table 28. CheckMate-816: time to locoregional recurrence—summary of observed data

CI = confidence interval; NE = not estimable; NIVO = nivolumab; PDC = platinum doublet chemotherapy.

Figure 28. CheckMate-816: time to locoregional recurrence: observed data

==> picture [455 x 312] intentionally omitted <==

chemo = chemotherapy; CI = confidence interval; HR = hazard ratio; KM = Kaplan-Meier; NE = not estimable; NIVO = nivolumab.

Diagnostic plots assessing whether PH or accelerated failure time (AFT) assumptions hold between the 2 treatment arms are presented in Figure 29. The log-log plot (top left panel in Figure 29), where points representing observations in the nivolumab + PDC and PDC treatment arms were relatively parallel during the entire follow-up suggested that the PH assumption holds. This finding was further supported by Schoenfeld residuals test ( P value 0.46) (bottom left panel in Figure 29). Additionally, the points forming a relatively straight line in the QQ-plot (top right panel in Figure 29) suggested that both sets of observed quantiles in the nivolumab + PDC and PDC treatment arms came from the same AFT distribution and

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

that the AFT assumption holds. Therefore, the use of jointly fitted distributions with treatment arm as predictor was recommended for the ITT population.

Figure 29. Time to locoregional recurrence in the intention-to-treat population: loglog plot, Schoenfeld residuals plot and QQ-plot

==> picture [473 x 306] intentionally omitted <==

The statistical tests suggested that both AFT and PH assumptions hold. Therefore, the use of jointly fitted distributions with the treatment arm as a predictor was selected to extrapolate TTLR.

Extrapolations

Because the PH assumption was found to hold, it was determined that jointly fitted curves would provide the best fit to the data. Figure 30 and Figure 31 present the short-term projections of the standard parametric functions for the nivolumab + PDC and PDC alone arms, respectively. Table 29 summarises goodness-of-fit statistics.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 30. Time to locoregional recurrence: within-trial fittings (nivolumab + PDC)

==> picture [375 x 300] intentionally omitted <==

PDC = platinum doublet chemotherapy.

Figure 31. Time to locoregional recurrence: within-trial fittings (PDC)

==> picture [388 x 302] intentionally omitted <==

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

PDC = platinum doublet chemotherapy.

Table 29. Time to locoregional recurrence: goodness-of-fit statistics

AIC = Akaike information criteria; BIC = Bayesian information criteria.

All of the curves were considered to fit the Kaplan-Meier data reasonably well during the within-trial period. The log-normal and exponential distributions were considered to provide the best statistical fit of the data based on AIC and BIC.

Long-term extrapolated outcomes and context

Figure 32 and Figure 33 present long-term extrapolations versus Kaplan-Meier data for the nivolumab + PDC and PDC alone arms, respectively. Table 30 and Table 31 detail the median and mean TTLR months for all the distributions, along with the percentage of patients who were LR free at 1, 3, 5, 10, 20, and 30 years for the nivolumab + PDC and PDC alone arms.

Figure 32. Time to locoregional recurrence: long-term extrapolations (nivolumab + PDC)

==> picture [398 x 292] intentionally omitted <==

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

PDC = platinum doublet chemotherapy.

Figure 33. Time to locoregional recurrence: long-term extrapolations (PDC arm)

==> picture [422 x 310] intentionally omitted <==

PDC = platinum doublet chemotherapy.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 30. Time to locoregional recurrence: predicted median, mean, and landmarks (nivolumab + PDC)

PDC = platinum doublet chemotherapy.

Table 31. Time to locoregional recurrence: predicted median, mean, and landmarks (PDC)

NE = not estimable; PDC = platinum doublet chemotherapy.

As noted previously, all extrapolations tested were found to fit the data well during the withintrial period. However, in the long-term, significant divergence was observed in the curve tails for various extrapolations. For example, at 20 years in the PDC arm, the predicted survival based on the exponential extrapolation was found to be xxx% versus xxxx% in the lognormal extrapolation.

Therefore, to ensure that the long-term projections for PDC were realistic, the estimates were firstly compared against external sources (Figure 34). Published evidence on TTLR in this population is limited, and only 2 external sources were identified (see Table 27):

• Estimates of TTLR collected as part of the 2014 NSCLC Collaborative Group metaanalysis[129]

• The base-case extrapolation used in NICE TA761[121] (osimertinib for adjuvant treatment of EGFR mutation-positive NSCLC after complete tumour resection)

The comparison with TA761 was conducted due to the lack of available data from sources more closely approximating the CheckMate-816 population, despite the difference in patient population (TA761 considered all patients with EGFR mutation-positive NSCLC, whereas patients with known EGFR mutation-positive NSCLC were excluded from CheckMate-816). For the comparison (Figure 34) exponential and log-normal distributions were selected for comparison given that they provide the best statistical fit to the observed data based on BIC and AIC, respectively.

Figure 34. Time to locoregional recurrence: comparison versus external data - PDC

==> picture [436 x 321] intentionally omitted <==

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

• CM-816 = CheckMate-816; KM = Kaplan-Meier; NSCLC = non-small cell lung cancer; PDC = platinum doublet chemotherapy; SOC = standard of care; TA = technology appraisal.

The extrapolations of TTLR based on CheckMate-816 data were generally more pessimistic than those observed in the NSCLC Collaborative Group meta-analysis. This discrepancy is likely primarily due to the difference in disease stage in CheckMate-816 (which included more patients with stage IIIA) versus the NSCLC Collaborative Group meta-analysis (which included more patients with stage I/II). In the long-term, the log-normal distribution was similar to the log-normal distribution for TTLR from TA761.

Summary: base-case input selection

The standard parametric distributions (exponential, Weibull, log-normal, log-logistic, Gompertz, generalised gamma, and gamma) were fit to the TTLR data from CheckMate-816. Visual inspection of the within-trial data suggested all tested extrapolations generally fit well. Assessment of AIC and BIC suggested the exponential and log-normal distributions were the best fitting based on BIC and AIC, respectively. Comparison of the curve tails against external data suggested (Figure 34) the log-normal distribution would provide the best fit in the long-term. This was also confirmed from the feedback received where the clinical experts consulted suggested the log-normal distribution was plausible for TTLR in the long-term. Therefore, the log-normal distribution was used in the base case to describe TTLR for EF patients for both the neoadjuvant nivolumab + PDC and neoadjuvant PDC arms of the model.

B.3.3.1.3 Time to any progression and derivation of time to distant metastasis health state

Observed data for TTDM from the CheckMate-816 trial were immature with a relatively low event count. In total, only xx DM events were observed in the nivolumab + PDC arm (xxx and xx DM events were observed in the PDC arm (xxx Median was not reached in either arm. Figure 35 presents the Kaplan-Meier curves, and Table 32 summarises the observed data.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 35. CheckMate-816: time to distant metastasis—observed data

==> picture [457 x 299] intentionally omitted <==

chemo = chemotherapy; CI = confidence interval; HR = hazard ratio; KM = Kaplan-Meier; NE = not estimable; NIVO = nivolumab.

Table 32. CheckMate-816: time to distant metastasis—summary of observed data

CI = confidence interval; HR = hazard ratio; NE = not estimable; NIVO = nivolumab; PDC = platinum doublet chemotherapy.

The immature TTDM data from the CheckMate-816 trial prevented the development of reliable parametric fittings and long-term extrapolations. Instead, a constructed TTDM curve was calculated as the difference between the long-term estimates from time to any progression and TTLR. The consensus opinion of the experts consulted as part of the during the Global HTA advisory board meeting, May 2022 (Appendix N) was that this would be an appropriate approach. The estimates of time to any progression and the derivation of TTDM extrapolations are presented in the following sections.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Time to any progression

Observed data

Estimates for time to any progression were based on data collected from the CheckMate-816 trial. The observed time to any progression curves were derived from the EFS data but with death events censored. In total, xx progression events were observed in the nivolumab + PDC arm (xxx and xx progression events were observed in the PDC arm (xxx The median time to progression for the PDC arm was xxx months and the median xxxxx xxx xx xxxxxxxxx for the nivolumab + PDC arm. The data translated to an overall HR of 0.60 for nivolumab + PDC versus PDC alone. The Kaplan-Meier curves and a summary of the observations are depicted in Figure 36 and Table 33, respectively.

Figure 36. CheckMate-816: time to any progression—observed data

==> picture [455 x 312] intentionally omitted <==

chemo = chemotherapy; CI = confidence interval; HR = hazard ratio; KM = Kaplan-Meier; NE = not estimable; NIVO = nivolumab.

Table 33. CheckMate-816: time to any progression—summary of observed data

CI = confidence interval; HR = hazard ratio; NE = not estimable; NIVO = nivolumab; PDC = platinum doublet chemotherapy.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Diagnostic plots assessing whether PH or AFT assumptions hold between the 2 treatment arms are presented in Figure 37. The log-log plot (top left panel in Figure 37), where points representing observations in the NIVO+PDC and PDC treatment arms were relatively parallel during the entire follow-up suggested that the PH assumption holds. This finding was further supported by Schoenfeld residuals test ( P value of 0.44; bottom left panel in Figure 37). Additionally, the points forming a relatively straight line in the QQ-plot (top right panel in Figure 37), except some deviation between months 10 and 13 (for PDC), suggested that both sets of observed quantiles in the NIVO+PDC and PDC treatment arms came from the same AFT distribution and that the AFT assumption holds. Therefore, the use of jointly fitted distributions with treatment arm as predictor was recommended for the ITT population.

Figure 37. Time to progression in the intention-to-treat population: log-log plot, Schoenfeld residuals plot, and QQ-plot

==> picture [518 x 331] intentionally omitted <==

chemo = platinum doublet chemotherapy; NIVO = nivolumab.

Extrapolations

Figure 38 and 0 present the short-term projections of the standard parametric functions for the nivolumab + PDC and PDC alone arms, respectively. Table 34 summarises goodnessof-fit statistics. Because the curves are jointly fitted, the same goodness-of-fit statistics apply to both arms.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 38. Time to any progression: within-trial fittings (nivolumab + PDC)

==> picture [431 x 335] intentionally omitted <==

PDC = platinum doublet chemotherapy.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 39. Time to any progression: within-trial fittings (PDC)

==> picture [431 x 320] intentionally omitted <==

PDC = platinum doublet chemotherapy.

Table 34. Time to any progression: goodness-of-fit statistics

AIC = Akaike information criteria; BIC = Bayesian information criteria.

All of the curves were considered to fit the Kaplan-Meier data reasonably well during the within-trial period. The log-normal distribution was considered to be the best fitting, based on AIC and BIC.

Long-term extrapolated outcomes and context

Long-term extrapolations are shown versus Kaplan-Meier data for the nivolumab + PDC and PDC arms in Figure 40 and Figure 41, respectively. Table 35 and Table 36 detail the predicted median and mean months of time to any progression for all the distributions, along

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

with the percentage of patients who were EF at 1, 3, 5, 10, 20, and 30 years for the nivolumab + PDC and PDC alone arms.

Figure 40. Time to any progression: long-term extrapolations (nivolumab + PDC)

==> picture [419 x 313] intentionally omitted <==

PDC = platinum doublet chemotherapy.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 41. Time to any progression: long-term extrapolations (PDC)

==> picture [424 x 324] intentionally omitted <==

PDC = platinum doublet chemotherapy.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 35. Time to any progression: predicted median, mean, and landmarks (nivolumab + PDC)

NE = not estimable; PDC = platinum doublet chemotherapy.

Table 36. Time to any progression: median, mean, and landmarks (PDC)

NE = not estimable; PDC = platinum doublet chemotherapy.

Based on statistical assessment, the jointly fitted log-normal distribution provided the best fit to the observed time to any progression data. However, in the long-term, substantial differences were observed in the trials for various extrapolations similarly to long-term projections of TTLR.

Given the lack of literature specifically reporting time to progression, and that time to any progression represents a subset of EFS outcomes (i.e., all progression events, excluding death), a comparison of EFS extrapolations against available external sources was conducted to aid in the selection of clinically plausible distributions for time to progression. Only a small proportion of CheckMate-816 EFS events were deaths (xx across nivolumab + PDC and PDC arms); therefore, it was assumed that the same shape selected for EFS would be equally applicable to the projection of time to any progression outcomes. The EFS long-term projections (details of EFS estimation are presented in Appendix O) were compared with several external sources, including:

• The NSCLC Collaborative Group meta-analysis conducted in 2014[130]

• A patient-level meta-analysis conducted by BMS (CA2097L8)[73]

• A US oncology real-world study[131]

An additional constructed curve was created from 2 selected studies in the meta-analysis with up to 5 years of follow-up and weighted to reflect patients’ stage distribution in CheckMate-816. One study—Pless et al. (2015)[88] —was a phase 3 randomised clinical trial that enrolled 232 patients with stage IIIA NSCLC, randomly assigned to chemoradiotherapy (3 cycles of neoadjuvant chemotherapy [100 mg/m[2 ] cisplatin and 85 mg/m[2] docetaxel]) or neoadjuvant chemotherapy alone. The other study—Felip et al. (2010)[70] —was a phase 3 trial including 624 patients with stage IA, IB, or II NSCLC, randomly assigned to surgery alone (212 patients), 3 cycles of preoperative paclitaxel-carboplatin followed by surgery (201 patients), or surgery followed by 3 cycles of adjuvant paclitaxel/carboplatin

(211 patients). The constructed EFS curve is generated by weighting EFS curves from Felip (stage I-II) and Pless (stage III based on the stage distribution from CheckMate-816 baseline (35.7% stage I-II and 64.3% stage III). Table 37 presents a comparison of landmarks, while Figure 42 presents a selection of curves Appendix P presents additional details on the sources used for the validation.

NSCLC Collaborative meta-analysis and BMS patient-level meta-analysis (CA2097L8) data suggest almost 40% of patients remain in EFS at 5 years, and approximately 35% at 7 years, while the EFS derived from US oncology data is significantly lower: 25% of patients were in EFS at 5 years and 22% at 7 years. Based on the constructed Felip/Pless curve, 28% of patients are in EFS at 5 years, an estimate similar to that from the US oncology study.

Based on this comparison, the exponential, Weibull and gamma distributions appear to be overly pessimistic in the long run (EFS of approximately 7% at 7 years). Therefore, they lack clinical plausibility and are not suitable for EFS extrapolation. Event-free survival at 5 years based on Gompertz, log-normal and generalised gamma projections (25%-28% at year 5) is within 5% of the estimated Felip/Pless EFS (27.4% at year 5), and close to the BMS US Oncology study, while log-logistic is more pessimistic (22.9%). Based on this assessment,

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Gompertz, log-normal and generalised gamma could be considered to be the best candidates, but there was no single obvious best candidate.

Input from 6 clinicians consulted during the Global HTA advisory board meeting, May 2022 (Appendix N) all agreed that the log-normal extrapolation provided is clinically plausible longterm estimates of EFS.

Table 37. Event-free survival: landmark comparison versus external data

EFS = event-free survival; NSCLC = non-small cell lung cancer; US = United States.

• a Constructed by weighting EFS from Felip (stage I-II) and Pless (stage III); weights are based on stage distribution from CM-816 baseline (35.7% stage I-II and 64.3 stage III).

Figure 42. Event-free survival: visual comparison versus external data for PDC

==> picture [455 x 258] intentionally omitted <==

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

BMS = Bristol Myers Squibb; CM-816 = CheckMate-816; EFS = event-free survival; MA = meta-analysis; NSCLC = non-small cell lung cancer; PDC = platinum doublet chemotherapy; US = United States.

Note: Weibull and gamma are not presented in the figure given their close similarity with exponential. Similarly, log-logistic is not presented, given it closely resembles log-normal.

Summary: base-case input selection

The log-normal distribution is used in the base case to describe time to any progression for EF patients for both the neoadjuvant nivolumab + PDC and neoadjuvant PDC arms of the model.

The standard parametric distributions (exponential, Weibull, log-normal, log-logistic, Gompertz, generalised gamma, and gamma) were fit to the time to any progression data both jointly and independently to each arm of CheckMate-816.

Jointly fit parametric extrapolations were selected based on diagnostic testing.

Visual inspection of the within-trial data suggested all extrapolations tested generally fit well. Assessment of AIC and BIC suggested the log-normal distribution was the best fitting.

Comparison of the EFS curve against external data suggested the log-normal, generalised gamma or Gompertz distribution would provide plausible fit in the long-term.

Expert feedback from the Global HTA advisory board meeting held May 2022 suggested that the log-normal distribution was plausible for EFS, which is understood to be representative of time to any progression, given that progressions represent a large proportion of all EFS events (approximately 85%).

Derivation of time to Death or Distant Metastases health states

Instead of directly projecting the TTDM curve based on the immature data from CheckMate-816, TTDM was derived as the difference of time to any progression and TTLR at any model cycle, such that:

𝐻𝑎𝑧𝑎𝑟𝑑𝐷𝑖𝑠𝑡𝑎𝑛𝑡 𝑀𝑒𝑡𝑎𝑠𝑡𝑎𝑠𝑖𝑠 = 𝐻𝑎𝑧𝑎𝑟𝑑𝐴𝑛𝑦 𝑃𝑟𝑜𝑔𝑟𝑒𝑠𝑠𝑖𝑜𝑛 – 𝐻𝑎𝑧𝑎𝑟𝑑𝐿𝑜𝑐𝑜𝑟𝑒𝑔𝑖𝑜𝑛𝑎𝑙 𝑅𝑒𝑐𝑢𝑟𝑟𝑒𝑛𝑐𝑒

The hazards of time to any progression and TTLR were based on the best-fitting extrapolation curves for the corresponding transitions. In the base case, the log-normal distribution was selected for both transitions. Figure 43 and Figure 44 illustrate the resulting long-term TTDM estimates for the nivolumab + PDC and PDC alone arms, respectively.

This modelling approach was considered appropriate by several health economics and outcomes research experts consulted as part of the Global HTA advisory board meeting, May 2022 (see Appendix N).

Note that in the ITC and CEM sections, TTDM refers to time to distant metastases only, as opposed to the CM816 trial definition of time to death or distant metastases. This is because deaths have been censored to allow for accurate transition probabilities to be calculated for the economic model.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 43. Time to Death or Distant Metastasis: long-term extrapolations (nivolumab + PDC)

==> picture [454 x 294] intentionally omitted <==

DM = Distant Metastasis; KM = Kaplan-Meier; LR = Locoregional Recurrence; PDC = platinum doublet chemotherapy.

Figure 44. Time to Death or Distant Metastasis: long-term extrapolations (PDC)

==> picture [455 x 295] intentionally omitted <==

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

DM = Distant Metastasis; KM = Kaplan-Meier; LR = Locoregional Recurrence; PDC = platinum doublet chemotherapy.

B.3.3.1.4 Locoregional Recurrence to Distant Metastasis health states

In CheckMate-816, patient monitoring for progression events after the first progression was not required in the trial protocol; hence, estimates of the probability of transition between the LR and DM health states were not available for all patients at risk. Therefore, the model had to rely on external sources to inform this estimate.

Published data captured in the SLR conducted for MRU and utilities (see Appendix H) were used to address this gap. Specifically, data from LuCaBIS (Lung Cancer Burden of Illness Study),[39] a retrospective study of NSCLC conducted in the UK, France, and Germany, were found to be suitable to estimate the rate at which patients with LR experience DM because it included patients with resectable NSCLC and provided sufficient follow-up to track them to the time of LR and through to DM. Table 38 presents key outputs used to derive the rate of transition from LR to DM.

Table 38. Key trial outputs from LuCaBIS used to derive transition probability from Locoregional Recurrence to Distant Metastasis

DM = Distant Metastasis; LR = Locoregional Recurrence.

Based on these parameters, 16.3% of patients with LR were estimated to experience progression in 26 months. This was converted to an annual transition probability of 7.7% (0.46% per model cycle). In the model, this probability was applied as a constant hazard, and does not change over time.

This input was subsequently reviewed by clinicians (in the Global HTA advisory board meeting, May 2022 and in the UK validation meeting, August 2022) to evaluate its appropriateness. The unanimous consensus among the clinicians consulted was that the calculated probability estimate of LR to DM from LuCaBIS was too low, and that these patients were likely to experience DM at a faster rate than suggested by the LuCaBIS data. Six clinicians provided estimated annual transition probabilities. The suggested probabilities are summarised in Table 39. The estimates were used to estimate an overall transition probability (the mid-point of any range provided was used to calculate the average). The average estimate of 20% (which is the same as the average from the 2 UK key opinion leaders [KOLs]) was used in the base-case analysis with the values from LuCaBIS tested in a scenario analysis.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 39. Estimates of annual transition probability from Locoregional Recurrence to Distant Metastasis per key opinion leader input

KOL = key opinion leader.

B.3.3.1.5 Mortality for event-free patients

Observed data

Data characterising mortality risks for patients who had not yet experienced a progression event were available from CheckMate-816. In general, due to relatively immature follow-up data available for OS due to the early stage of disease and positive treatment impact, the number of pre-progression deaths were low; there were xx pre-progression deaths out of 179 patients in the nivolumab + PDC arm and xx pre-progression deaths out of 179 patients in the PDC arm. Kaplan-Meier curves for both treatment arms were overlapping (Figure 45), which suggested no difference in mortality, at the current trial follow-up, among EF patients between treatment arms. Therefore, data from both the treatment arms were pooled for conducting the parametric survival analyses of EF mortality with a larger sample size and therefore less uncertainty. It should be acknowledged however that assuming pooled OS for EFS patients is a conservative assumption; nivolumab + PDC may be prognostic on survival compared with chemotherapy alone in areas additional to delayed disease progression alone. The pooled Kaplan-Meier curve is illustrated in Figure 46. These data were still immature despite a larger sample size; xxx% of patients were event-free at the end of the trial follow-up.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 45. CheckMate-816: mortality in event-free patients—observed data

==> picture [443 x 299] intentionally omitted <==

chemo = chemotherapy; CI = confidence interval; HR = hazard ratio; KM = Kaplan-Meier; NE = not estimable; NIVO = nivolumab.

Figure 46. CheckMate-816: mortality in event-free patients—observed data (pooled)

==> picture [440 x 296] intentionally omitted <==

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

CI = confidence interval; KM = Kaplan-Meier; NE = not estimable

Extrapolations

The goodness-of-fit statistics (AIC and BIC) of all distributions fitted are presented in Table 40. Based on the goodness-of-fit statistics, the generalised gamma (AIC = 301.2 and BIC = 312.9) and exponential (AIC = 304.1 and BIC = 308) distributions provided the best fit to the observed data. However, the difference in the AIC and BIC between all distributions was minimal (< 6 points). Therefore, external data were considered for selecting the best clinically plausible distribution.

Table 40. Mortality in event-free patients: goodness-of-fit statistics (pooled data)

AIC = Akaike information criteria; BIC = Bayesian information criteria

Visual inspection of observed vs. predicted pre-progression survival curves in Figure 47 showed good fit during the observed follow-up for all distributions. The long-term projections (Figure 48) for the distributions differed substantially, as expected. Long-term survival estimates from the log-normal and log-logistic distributions were similar. Long-term projections from Gompertz and generalised gamma suggested a plateau after 100 months (approximately 8 years). Long-term survival estimates from exponential, Weibull and gamma distributions were relatively shorter. Table 41 details the predicted median and mean alive months from EF for all the distributions, along with the percentage of those who were EF at 1, 3, 5, 10, 20 and 30 years for the pooled population.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 47. Mortality in event-free patients: within-trial fittings (pooled data)

==> picture [421 x 313] intentionally omitted <==

KM = Kaplan-Meier.

Figure 48. Mortality in event-free patients: long-term extrapolations (pooled data)

==> picture [406 x 303] intentionally omitted <==

KM = Kaplan-Meier.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 41. Mortality in event-free patients: predicted median, mean, and landmarks (pooled data)

NE = not estimable.

Context and selection of base-case input

Given that most EF mortality extrapolations fit reasonably well in the short-term while diverging over the long-term, the predicted outcomes based on the extrapolations from CheckMate-816 were compared against observed outcomes from a separate patient-level meta-analysis conducted by BMS.[73] The meta-analysis pooled results observed in published neoadjuvant chemotherapy trials, and was sufficiently granular to allow for an analysis of EF mortality among those patients. Expert feedback collected from the Global HTA advisory board meeting, May 2022, confirmed that the meta-analysis is a reasonable source that can be used to better understand long-term mortality outcomes for EF patients To ensure the long-term CheckMate-816 extrapolations were plausible, estimates were capped to UK general population mortality (i.e., adjusted such that the minimum hazard of death is never lower than what would be expected among the general population for that age), with a sex distribution based on CheckMate-816 (Figure 49).

Figure 49. Patient-level meta-analysis versus long-term event-free mortality extrapolations from CheckMate-816

==> picture [443 x 328] intentionally omitted <==

BMS = Bristol Myers Squibb; CM816 = CheckMate-816; KM = Kaplan-Meier.

When a general population mortality cap is applied, most of the projections estimated that patients would die more slowly versus what would have been expected based on the patientlevel meta-analysis. Considering both good statistical fit (in terms of AIC/BIC) and the external data, the exponential distribution is selected in the model base case. The

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

exponential is also the predicting the most severe long-term event-free mortality which would be a conservative assumption given higher proportion of EFS in the nivolumab arm.

B.3.3.1.6 Mortality for patients with locoregional recurrence

Observed data

As in the estimates for EF mortality, estimates of mortality for patients who have experienced LR are pooled and assumed to be the same across treatment arms due to a relatively small number of events in the treatment (n = xx) and control arms (n = xx) respectively. Additionally, the Kaplan-Meier curves, developed using data collected from CheckMate-816, overlapped for the first 12 months following progression, after which nivolumab + PDC showed a benefit. However, the 95% CI of the estimated HR for the 2 curves crossed 1 (range, xxxxxx), meaning that this difference was not statistically significant. Figure 50 presents the treatment-specific Kaplan-Meier curves, and Figure 51 presents the pooled Kaplan-Meier curve.

Figure 50. CheckMate-816: mortality after locoregional recurrence—observed data

==> picture [455 x 306] intentionally omitted <==

chemo = chemotherapy; CI = confidence interval; HR = hazard ratio; KM = Kaplan-Meier; NE = not estimable; NIVO = nivolumab.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 51. CheckMate-816: mortality after locoregional recurrence—observed data (pooled)

==> picture [456 x 306] intentionally omitted <==

CI = confidence interval; KM = Kaplan-Meier; NE = not estimable

Extrapolations

Table 42 presents the goodness-of-fit statistics (AIC and BIC) of all fitted distributions. Based on the goodness-of-fit statistics, the log-logistic (AIC = 278.3 and BIC = 283) and exponential (AIC = 279.7 and BIC = 282.1) distributions provided the best fit to the observed data. However, the difference in the AIC and BIC between all distributions was minimal (< 6 points). Therefore, similar to all other extrapolations discussed, long-term projections and clinical expert opinion should be considered when selecting the most plausible distribution.

Table 42. Mortality after locoregional recurrence: goodness-of-fit statistics (pooled data)

AIC = Akaike information criteria; BIC = Bayesian information criteria.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Visual inspection of observed versus predicted post-LR survival curves in Figure 52 shows that none of the distributions provided a good fit during the observed follow-up period. The long-term projections (Figure 53) for the distributions differed substantially. Long-term projections from generalised gamma, Gompertz, log-normal, and log-logistic were similar. Long-term survival estimates from exponential, Weibull, and gamma distributions were relatively shorter.

Figure 52. Mortality after locoregional recurrence: within-trial fittings (pooled data)

==> picture [456 x 339] intentionally omitted <==

KM = Kaplan-Meier; LR = Locoregional Recurrence.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 53. Mortality after locoregional recurrence: long-term extrapolations (pooled data)

==> picture [456 x 306] intentionally omitted <==

KM = Kaplan-Meier; LR = Locoregional Recurrence.

Because none of the distributions provided a good fit to the observed post-LR mortality data, splines analyses of post-LR mortality data were performed. Up to 3 knots (i.e., up to 4 degrees of freedom [DFs]) and 3 scales (cumulative hazards, cumulative odds, or normal equivalent deviate [probit] scale) were used to fit post-LR mortality data in the pooled data from the nivolumab + PDC and PDC alone treatment arms. Table 43 presents goodness-offit of each fitted spline model.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 43. Mortality after locoregional recurrence: goodness-of-fit statistics for spline fitting (pooled data)

AIC = Akaike information criteria; BIC = Bayesian information criteria; NIVO = nivolumab; PDC = platinum doublet chemotherapy.

For post-LR survival, the models with 3 DFs on the hazard, odds and normal scale provided similar outcomes and the best fit to the observed data based on AIC and BIC. The predicted vs. observed plots and long-term extrapolations from these models with 2-4 DFs are shown in Figure 54. The models with 3 DFs provided a much better fit to the observed post-LR survival versus standard parametric models and models with 2 DFs by visual inspection. The models with 4 DFs produced similar fit to that with 3 DFs. Based on AIC and BIC and visual inspection, and a preference for simplicity (i.e., favouring fewer knots when the fit is similar), models with 3 DFs (2 knots) were selected to be the best-fitting spline.

The estimated median and mean post-LR survival were xxx and xxxx months for the 3-DF model on the hazard scale, 27.1 and not estimable for the 3-DF model on the odds scale and xxx and xxxx months for the 3-DF model on the normal scale, respectively. The plausibility of the estimated mean survival time confirmed that the 3-DF spline with hazard scale was the most appropriate fitting to be included in the model.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 54. Mortality after locoregional recurrence: observed and predicted survival from 2-4 degrees of freedom on hazard, odds, and normal scales for short-term (top), long-term (bottom), and pooled data

DF = 2

==> picture [450 x 219] intentionally omitted <==

DF = 3

==> picture [450 x 219] intentionally omitted <==

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

DF = 4

==> picture [450 x 220] intentionally omitted <==

DF = degree of freedom.

Context and selection of base-case input

The extrapolated estimates were compared against estimates from external data to assess the plausibility of the generated extrapolations. A source that precisely replicated the postLR progression patients from CheckMate-816 was not available. Therefore, 2 sources were explored relating 3 populations that were different from the post-LR progression patients from CheckMate-816 but similar enough to serve as benchmarks. Specifically, this included the placebo arm from PACIFIC[132] (a trial of durvalumab in patients with stage III, unresectable NSCLC) and Goldstraw et al. (2016)[31] , which reported long-term survival estimates for patients by stage (specifically, patients with stage IIIA disease and patients with stage IIIB were most relevant for this purpose). Figure 55 presents these curves, along with the best-fitting parametric and spline models for mortality in LR.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 55. Comparison of mortality in Locoregional Recurrence health state versus PACIFIC[132] and Goldstraw et al. (2016)[31]

==> picture [448 x 266] intentionally omitted <==

CM816 = CheckMate-816; DF = degree of freedom; LR = Locoregional Recurrence

The post-LR survival projected by log-logistic distribution was the most pessimistic estimation. The estimation by log-normal distribution and spline fitting with 2 DFs were similar. The 4 curves align well with the data from PACIFIC placebo arm. Spline fittings with 3 DFs and 4 DFs provided more optimistic and almost identical projection and align better with the curves from Goldstraw et al. (2016)[31] in the longer term.

In the Global HTA advisory board meeting, May 2022, the consensus among the consulted clinicians (see Appendix N) was that the stage IIIA population as reported in Goldstraw et al. (2016)[31] was the most appropriate population; therefore, the spline (DF = 3, hazard) should be favoured because it appears to adhere more closely to the Goldstraw stage IIIA KaplanMeier curve than the log-logistic extrapolation in the long run. Thus, the spline (DF = 3, hazard) is used in the base-case setting, and the log-logistic distribution is explored in sensitivity analysis.

The selection of the spline with 3 DF’s and hazard scale is further justified by comparing it against other spline fittings in the long-term (Figure 56). The best-fitting spline with 2 DFs (odds scale) produces outcomes that align with parametric fittings in the long-term, and these did not fit well to the data (see Section B.3.3.1.6). The best-fitting spline with 4 DFs (hazard scale) produces outcomes similar to those obtained with the 3 DF (hazard scale) spline. Due to the similar outcomes between the 4 DF and 3 DF outcomes, the spline with fewer knots is preferred to prevent overfitting.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 56. Long-term splines and parametric extrapolations

==> picture [450 x 330] intentionally omitted <==

DF = degree of freedom; KM = Kaplan-Meier.

B.3.3.2 Distant Metastasis health state

Patients entered the DM health state upon experiencing a progression to distant metastasis from the EF or LR state and remained in the DM state until death. In the DM state, patients were expected to receive a mix of therapies for first-line metastatic disease, in line with UK clinical practice. Instead of explicitly modelling the outcomes for post-DM treatments, one-off LYs, QALYs, and costs were applied upon entry into the DM state (Figure 57). The one-off approach was selected to avoid developing a series of metastatic models that track progression and survival time in the DM state for various first-line metastatic therapies, which substantially increases the computational complexity and data burden of this neoadjuvant NSCLC CEM and is outside the scope of the current decision problem, which pertains to resectable non-metastatic NSCLC. The same approach was considered pragmatic and reasonable by the NICE Appraisal Committees in the 2018 submission of dabrafenib + trametinib in the adjuvant treatment of patients with melanoma.[133] In addition, the one-off modelling approach in the DM state was validated by clinical experts and health economists during the Global HTA advisory board meeting in May 2022, see Appendix N. Finally, in the recent adjuvant atezolizumab NSCLC NICE appraisal,[2] the company submission was criticised for having produced an adjuvant NSCLC CEM which was overly complex and produced metastatic NSCLC outcomes that were inconsistent with previous

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

NICE appraisals in first-line metastatic NSCLC; the one-off approach addresses this issue by allowing the outcomes from previous appraisals to be used directly as model inputs.

Figure 57. One-off approach for Distant Metastasis health state

==> picture [450 x 160] intentionally omitted <==

• DM = Distant Metastasis; EF = Event-Free; I-O = immuno-oncology; LR = Locoregional Recurrence; LY = lifeyear; QALY = quality-adjusted life-year.

• 1 Multiple I-O therapies approved in England captured, with their respective LYs, QALYs and costs.

• 2 This one-off cost captures all costs associated with the DM state, including costs for subsequent treatment (e.g., second line), associated resource use, and terminal care.

To populate the cost, life-year and QALY inputs associated with each of the first-line metastatic appraisals we scraped the published committee papers from the NICE website associated with each of the following STAs:

• TA770: Pembrolizumab with carboplatin and paclitaxel for untreated metastatic squamous non-small-cell lung cancer

• TA531: Pembrolizumab for untreated PD-L1-positive metastatic non-small-cell lung cancer

• TA683: Pembrolizumab with pemetrexed and platinum chemotherapy for untreated, metastatic, nonsquamous non-small-cell lung cancer

• TA584: Atezolizumab in combination for treating metastatic non-squamous nonsmall-cell lung cancer

Due to being consistently redacted it was not possible to extract the costs, life-years and QALY inputs from the respective committee papers. Therefore, to ensure the outcomes in our DM health state were reflective of the published costs, QALYs and life-years associated with the above approved first-line metastatic NSCLC NICE TAs, BMS proposed a collaboration with NICE so that NICE could directly incorporate the confidential outcomes into our economic model. By facilitating this approach, we could reduce uncertainty in the subsequent treatments section of our economic model.

NICE has agreed to this approach and is preparing the inputs for consideration during this submission’s upcoming committee meeting. However, given the confidential nature of these values, NICE will not provide the inputs to BMS and the results of the economic model that incorporates these inputs will only be available as a confidential appendix that will be shared solely with the EAG and appraisal committee.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Because we cannot be provided the confidential inputs by NICE, BMS has sourced alternative input values for the cost, life-year and QALY input values, which can be used as placeholder values within the BMS base case cost effectiveness analysis for the purpose of this submission. These values were sourced from a previous NICE STA for nivolumab + ipilimumab in untreated advanced non-small cell lung cancer (TA724), specifically using the ERG-preferred values from the ACD document (Table 44)[134] . However, 2 relevant subsequent treatments were not included in this submission (pembrolizumab + carboplatin + paclitaxel in squamous NSCLC; atezolizumab monotherapy). Inputs for pembrolizumab + carboplatin + paclitaxel in squamous NSCLC were instead sourced from the nivolumab + ipilimumab submission for untreated advanced NSCLC submission to SMC.[135] Atezolizumab monotherapy was not included in either the NICE or SMC submission for nivolumab + ipilimumab in untreated advanced NSCLC and no publicly available information about costs, life-years and QALYs for this treatment could be identified. In the absence of data, input values for atezolizumab monotherapy were set equal to the other monotherapy immunooncology agent available in the UK, pembrolizumab (Table 44). Given that pembrolizumab has an approval in PD-L1> 50% only and atezolizumab is approved for any PD-L1, this assumption is likely to be biased. However, given the very limited uptake of atezolizumab in the UK at the time of submission (estimated to be xx), the impact of this limitation on the model results is expected to be negligible. Finally, following clinical expert elicitation, it was advised that 25% of patients that are eligible for first-line metastatic NSCLC treatment may only receive best supportive care (BSC; palliative care only). Due to the paucity of data for the life-years and QALYs associated with BSC, we elicited expert opinion on the hazard ratio associated with BSC compared to patients receiving PDC. We then applied this hazard ratio (0.9) to the PDC life-years and QALYs to generate the input values for BSC.

Changes to the DM inputs are explored in sensitivity and scenario analyses, shown in Sections B.3.9.2 and 0.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 44. Inputs for Distant Metastasis health state

ACD = appraisal consultation document; DAD = detailed advice document; DM = Distant Metastasis; LY = life year; NSCLC = non-small cell lung cancer; PDC = platinum doublet chemotherapy; PD-L1 = programmed death-ligand 1; QALY = quality-adjusted life-year; SMC = Scottish Medicines Consortium. Source: NICE (2021)[134,] SMC (2021)[135]

In the model, immuno-oncology therapy retreatment restrictions are considered. In the basecase analysis, patients who progress on or within 6 months after treatment completion with nivolumab + PDC in the neoadjuvant setting are not eligible for further treatment with immuno-oncology therapies; for those patients, immuno-oncology therapy weights are set to zero and redistributed across remaining treatment options. A retreatment restriction of 6 months was selected following input from Peter Clark during the first appraisal committee meeting for adjuvant atezolizumab (TA10751), who notified the committee that the NHS were considering retreatment within 6-12 months. Six months is the most conservative of the two time points for our economic model results.

Based on data from CheckMate-816, xxx of patients treated with nivolumab + PDC experienced an event while on treatment or within 6 months of treatment completion and were not eligible for further treatment with immuno-oncology therapy. A scenario analysis was performed in which the retreatment restriction was extended to 12 months: xxx of patients treated with nivolumab + PDC experienced an event while on treatment or within 12 months.

Table 45 presents the distribution of treatments received in the DM health state based on market data collected by BMS. The distribution of treatments in the DM state are equally adjusted for nivolumab + PDC to reflect retreatment criteria, that is, xxx of patients treated with neoadjuvant nivolumab + PDC were not eligible for further immune-oncology treatment because they experienced an event while on treatment or within 6 months of treatment completion with nivolumab + PDC (Table 45).

Table 45. Distribution of treatment in Distant Metastasis health state

CRT = chemoradiation; PDC = platinum doublet chemotherapy; BSC = best supportive care. Source: BMS market share data on file

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.3.3.3 Cure assumption

As presented in Section B.1.3.4, long-term evidence exists that suggests patients who are treated for resectable non-metastatic NSCLC may be able to achieve cure, defined as (1) no risk of progression and (2) no excess cancer-related mortality versus an age- and sexmatched population. In general, inclusion of cure in the model rested on 3 key pillars:

• Engagement with clinical experts, among whom there was consensus that the cure assumption was reasonable in this indication

• Precedent from previous NICE appraisals, namely, inclusion of the cure assumption in NICE TA761[121] and NICE TA10751[2] , and NICE’s finding that this inclusion was indeed appropriate

• Empirical evidence, that is, trial data showing reduction in hazard of progression at 5 years, and EFS data among patients treated with neoadjuvant PDC from beyond 5 years showing “plateau”

B.3.3.3.1 Clinical expert feedback for cure assumption

The key rational for implementation of cure within the model relies on to the clinical plausibility and relevance. Therefore in terms of expert engagement, clinical experts with experience treating non-metastatic NSCLC were asked both as part of the UK HTA clinical expert meeting, March 2022 and the Global HTA advisory board meeting, May 2022 (Appendix N) about the plausibility of cure in this setting, as well as their assessment of the likely timepoint and proportion of patients achieving cure. There was broad consensus that cure is a plausible outcome that the model should consider, and that 5 years is an appropriate timepoint to consider cure as having occurred. However, there was no clear consensus on the percentage of patients achieving cure.

B.3.3.3.2 Precedent for cure assumption

In terms of precedent, 2 recent NICE appraisals in early-stage NSCLC included a cure assumption in the model based on clinical input and published evidence. The first of these was NICE TA761, which assessed adjuvant osimertinib with or without PDC in patients who had undergone surgical resection and had EGFR mutation-positive, early-stage NSCLC.[121] Specifically, in that submission, it was assumed that 95% of patients who were progressionfree at 5 years would be cured (i.e., experience no further risk of progression) and return to mortality expected for an age- and sex-matched population without cancer.[121] While the EAG did note that the 5-year timepoint used in the submission might be “too generous,” it did not criticise the use of the cure assumption itself, instead opting to use an 8-year timepoint in its recommended analysis.[121] The second recent appraisal is the appraisal of atezolizumab or adjuvant treatment of resected NSCLC.[2] In that appraisal a cure proportion of t 91.5% was assumed based on published literature and a cure timepoint of 5 year was applied in the company base case. The committee agreed that there were uncertainty around both cure timepoint and proportion but agreed that a cure timepoint of 6 or 7 years would be plausible for atezolizumab and 5 years for active monitoring in relation to the preferred survival extrapolations in that submission.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.3.3.3.3 Empirical evidence

In terms of empirical data, 2 bodies of evidence were assessed that suggest cure. The first of these was a paper published by Demicheli et al. (2012)[42] . In this study, the authors sought to investigate how the hazard of different types of progression changes over time among patients with resected, early-stage NSCLC. As depicted in Figure 58, results of the study appear to suggest that the risk of LR or DM fluctuates over the first 5 years after resection, approaching zero at approximately 5 years.

Figure 58. Hazard of progression in early-stage resected NSCLC

==> picture [430 x 241] intentionally omitted <==

NSCLC = non-small cell lung cancer. Source: Demicheli et al. (2012)[42] ; figure reprinted here without permission

Long-term EFS outcomes were also assessed across studies evaluating neoadjuvant PDC. Additional detail on the sources cited are provided in Appendix P. In general, the trend across all of the included studies showed that the EFS curves flatten out after approximately 5 years (Figure 59).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Figure 59. Long-term event-free survival from neoadjuvant PDC studies

==> picture [450 x 284] intentionally omitted <==

NSCLC = non-small cell lung cancer; PDC = platinum doublet chemotherapy; US = United States.

Sources: Pless et al. (2015)[88] ; Felip et al. (2010)[70] ; Scagliotti et al. (2012)[91] ; Pisters et al. (2010)[90] ; NSCLC Metaanalyses Collaborative Group (2010)[56] ; BMS data on file (2021)[73] ; BMS data on file (2021)[131]

B.3.3.3.4 Implementation of cure in model

There are 2 potential methods that may be used to account for cure in survival extrapolations.[136] These are an “uninformed” approach, wherein the cure fraction is a model output based on observations from the relevant trial data, and an “informed” approach, wherein the cure fraction is a model input applied on the basis of known long-term survival estimates. Because no plateau suggesting cure was yet observed in the Kaplan-Meier data from CheckMate-816, the uninformed approach was ruled out as a method to consider cure in this model. Instead, based on the long-term observations and clinical expert input discussed in Section B.3.3.1.1, the informed approach was deemed to be feasible and was used in the model.

The implementation of cure in the model uses 3 key inputs:

• The proportion of patients achieving cure

• The timepoint at which cure is applied

• The period over which cure occurs

The pool of patients eligible for cure consists of those who have not yet experienced progression at the cure timepoint. Once the cure starting timepoint is reached, cure is applied using a constant rate over a period (with the specific rate calculated as a function of the period and cure proportion) until the full cure proportion is reached. The same cure parameters are applied to every treatment considered in the model; but the cured proportion

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

will differ across the treatment and comparator arm because transition probabilities from EF differs between treatments, so state membership in EF is different between arms over time. Table 46 presents the base-case cure parameters. As presented above and as discussed in the appraisals of osimertinib and atezolizumab there seems to be a clinical consensus that patients would be considered cured in a 5-8-year timeframe. However, there are still uncertainties around the exact cure timepoint as well as the proportion of patients being cured. Therefore, these assumptions were extensively tested in scenario analyses to investigate the impact of these assumptions on the results.

Table 46. Base-case cure parameters

EFS = event-free survival.

B.3.3.4 Adverse reactions

Adverse events may impact the costs associated with taking a given drug, as well as the quality of life of patients receiving treatment. Therefore, the safety profile of a given drug may be an important differentiating factor in a cost-effectiveness or cost-utility model. Grade 3 and 4 AEs were collected from CheckMate-816 trial data for neoadjuvant nivolumab + PDC and neoadjuvant PDC. Neoadjuvant CRT was assumed to have the same AE profile as neoadjuvant PDC. Adverse events for adjuvant PDC were collected from an SLR previously conducted by BMS.[137] Adverse events were considered not to be applicable to the surgery only arm, as these patients do not receive systemic therapy.

Lower-grade AEs (i.e., grade 1 to 2) were not considered, as they are generally not understood to have significant cost or quality of life implications (typically, grade 1 to 2 AEs are manageable by the patient, e.g., via over-the-counter medication, whereas grade 3 or 4 AEs require inpatient management).

Specific events included were those experienced by at least 5% or more of patients in at least 1 of the comparators included in the model. Rates were generally similar between nivolumab + PDC and PDC alone. The consequences of AE captured by the model were expressed in terms of their management cost (see Section B.3.5.3.1 for more detail) and utility (see Section B.3.4.4.2 for more detail). In addition, only AEs associated with initial (i.e., current line) treatment were considered, and AEs associated with subsequent lines were not considered. Table 47 presents the percentage of patients experiencing a grade 3 or 4 AE by treatment arm.

Table 47. Percentage of patients experiencing grade 3 or 4 adverse events

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

CRT = chemoradiation; NIVO = nivolumab; PDC = platinum doublet chemotherapy; SLR = systematic literature review.

B.3.4 Measurement and valuation of health effects

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

B.3.4.1.1 Methods

EQ-5D-3L utility data were collected in the CheckMate-816 clinical study in line with the clinical study protocol. The utility analysis used the EQ-5D-3L index score (utility index) at all timepoints in the study.

As per protocol, patients completed the EQ-5D-3L on the first day of each 3-week treatment cycle at baseline (day 1 of the first 3-week cycle) and on day 1 of every cycle during the neoadjuvant period (baseline and 2 on-treatment assessments), then post-neoadjuvant visits 1 (30 days from last dose and before surgery) and 2 (70 days from post-neoadjuvant visit 1). EQ-5D-3L was also completed in the adjuvant period (3 months after post-neoadjuvant visit 2 or after surgery) every 3 weeks for up to 4 cycles. The timing and number of EQ-5D-3L assessments were the same during the neoadjuvant period (baseline, week 4, week 7, and post-neoadjuvant visit 1) but after that could vary within patients and between treatment arms depending on whether patients underwent surgery or received adjuvant treatment. The dates of the EQ-5D-3L assessments were used to assign EQ-5D-3L assessments to health states (days were calculated relative to the date of randomisation + 1 day).

B.3.4.1.2 Estimating utility values for health states

For patients with progression or recurrence, EQ-5D-3L assessments were grouped by the date of the EQ-5D-3L assessment relative to the date of progression or recurrence and by the type of recurrence (locoregional or distant metastases) and classified as preprogression, locoregional recurrence, or distant metastases. Patients with progression type recorded as “not reported” were classified as locoregional recurrence and those with “both locoregional and distant metastases” were classified as distant metastases.

To estimate mean values of EQ-5D-3L for each health state, a mixed-model approach was used to account for repeated EQ-5D-3L measurements per patient within a health state (mixed model for repeated measures). The initial model (intercept only) did not include any health states in order to estimate the mean utility overall and by treatment group. For each

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

health-state model, 2 statistical models were fit: one with and one without treatment. The variables defining health states, treatment, and their interaction, if any, were included in the model as fixed effects. The model without treatment included the health states only. The model with treatment included treatment, health state, and the interaction of treatment and health state in the model. A random intercept was used to account for repeated measurements within each patient. An unstructured covariance structure was used. The baseline EQ-5D-3L was considered a pre-progression value (not included as a covariate), and there was no imputation of missing data.

B.3.4.1.3 Utility results

Among the 353 patients eligible for analysis, there were xxxx utility index observations available, with xxxx observations before progression or recurrence in xxx patients and xxx observations after progression or recurrence in xxx patients. Of the xxx post-progression observations, xxx (xx patients) were after locoregional recurrence, and xxx (xx patients) were after distant metastases.

Table 48 presents the mean utility estimates for type of recurrence by health-state utility (pre-progression or recurrence, locoregional recurrence, and distant metastases). The comparison of models with and without treatment did not show any statistically significant difference in model fit ( P = 0.355), indicating no significant difference between treatments.

Table 48. EQ-5D-3L utility index (UK weights): number of patients, observations, and least squares mean estimates

CI = confidence interval; DM = Distant Metastasis; LR = Locoregional Recurrence; NIVO = nivolumab; PDC = platinum doublet chemotherapy; UK = United Kingdom.

B.3.4.2 Mapping

EQ-5D data were collected in CheckMate-816 in line with the NICE reference case. Utility values for AEs for which CheckMate-816 data could not be used were obtained from the literature. Therefore, there was no need to use mapping techniques.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

B.3.4.3 Health-related quality of life studies

An SLR was undertaken to identify HRQOL studies relevant to the decision problem from the published literature. In particular, costs and utility data related to neoadjuvant, peri-adjuvant, and adjuvant treatment of early-stage NSCLC in Australia, Canada, France, Germany, Italy, Spain, the UK, and the US were sought. The SLR followed established best practices used in systematic review research[138,139] and was conducted according to the PRISMA guidelines.

Searches were performed in the MEDLINE and MEDLINE in-process (via PubMed) database, the National Health Service’s Economic Evaluation Database (NHS EED), and the Embase database to identify articles on human subjects published from the start of database indexing to May 2022. The SLR also captured a review of the grey literature, which included data from sources that were not indexed in the literature databases but were available from various scientific conferences or website.

The SLR was performed using the inclusion and exclusion criteria and the search strategy. In summary, it included observational studies, RCTs, and primary data from economic analyses reporting cost and utility data for adults with early-stage NSCLC. Appendix H presents full details of the search and a summary of the studies identified.

A total of 23 studies, reporting data on 20 unique study populations, were identified that met the eligibility criteria for the review.[23,63-81] However, none of the studies evaluated reported on EQ-5D in an appropriate population. Because no studies using the EQ-5D or mapping utilities were identified, detailed data extraction of the identified studies is not presented in this submission. Utilities from the CheckMate-816 study are used in the CEM in line with the reference case.

B.3.4.4 Health-related quality of life data used in the costeffectiveness analysis

Table 49 summarises the utilities used in the model base case. Overall, non–treatmentspecific utilities by health state were used and were applicable for all comparators equally.

Table 49. Summary of utility values for cost-effectiveness analysis

CI = confidence interval.

As noted above, the model applies a one-off QALY consequence in the DM health state; therefore, no utility value was associated with this health state.

The utility values from the CheckMate-816 trial are higher than might be expected for patients with NSCLC in the UK given that the age-adjusted utility value for the general population has been reported at 0.833.[140]

Clinical input received as part of the UK HTA clinical validation meeting, August 2022 indicated that they thought that the absolute values for both EF and LR seemed a higher than expected and that specifically the decrement from EF to LR was smaller than expected.

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

To at account for the higher absolute values for utilities in both health states the EF value was therefore capped at general population mortality in the base case and LR based on the EF to LR decrement observed in CheckMate-816. Given the clinical input received on the decrement likely being smaller than clinically expected this could be seen as a conservative assumption with regards to treatment effect. The trial values without adjustment were tested in a scenario analysis. Table 50 presents the alternative health-state utilities included in the base case and scenario analysis.

Table 50. Alternative utility estimates used in the base case and scenario analysis

CM-816 = CheckMate-816; EF = Event-Free; LR = Locoregional Recurrence.

B.3.4.4.1 Impact of adverse events on utility

Utility decrements were included in the model to capture the effect of grade 3/4 AEs on HRQOL, reflecting the safety profile of each treatment.

The model estimated the average QALY loss due to AEs for each treatment by considering the treatment-specific AE rates, the mean utility decrements associated with these AEs, and the mean duration of each AE episode.

Only grade 3/4 AEs occurring in ≥ 3% of patients in the study were included. The total mean QALY loss associated with AEs for each treatment was determined by calculating the sum of individual QALY loss associated with each AE. The total QALY loss due to AEs was applied once at the start of the model, assuming that AEs occurred within the early period of treatment. Utility decrements associated with AEs were not explicitly collected in the CheckMate-816 utility study; these values were sourced from the published study by Nafees et al. (2008)[141] . The study by Nafees et al. (2008)[141] considered HRQOL, as measured by the EQ-5D, in patients with metastatic NSCLC. However, there is a more recent study from Nafees et al. (2017)[142] . Of note, the 2008 study used the standard gamble (SG) valuation method to determine utility scores, whereas the 2017 study used the time–trade-off (TTO) method. Evidence suggests that TTO and SG methods do not produce the same estimates, and differences between these 2 approaches may be greater in more severe health states[143] . In addition, the TTO method tends to produce lower utilities than the SG method[143] . The application of higher disutilities from the Nafees et al. (2008)[141] publication should be seen as a conservative assumption for this analysis. If there were no data for certain AEs, utility decrements were based on assumptions (Table 51).

Company evidence submission template for nivolumab with chemotherapy for neoadjuvant treatment of resectable non-small cell lung cancer

Table 51. Adverse event–related disutilities