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

Single Technology Appraisal

Regorafenib for previously treated metastatic colorectal cancer [ID4002] Committee Papers

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

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE SINGLE TECHNOLOGY APPRAISAL

Regorafenib for previously treated metastatic colorectal cancer [ID4002]

Contents:

The following documents are made available to consultees and commentators:

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

1. Company submission from Bayer Plc

2. Company response to NICE’s request for clarification

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

• a. Bowel Cancer UK

• b. NCRI-ACP-RCP-RCR

4. Evidence Review Group report prepared by Kleijnen Systematic Reviews 5. Evidence Review Group – factual accuracy check 6. Technical engagement response from Bayer Plc 7. Technical engagement response & expert statement from experts:  Jane Ashford – patient, nominated by Bowel Cancer UK

8. Technical engagement response from consultees and commentators:  Servier

9. Evidence Review Group critique of company response to technical engagement prepared by Kleijnen Systematic Reviews

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

fNATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

Single technology appraisal

Regorafenib for treating metastatic colorectal cancer (ID4002)

Document B

Company evidence submission

May 2022

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 1 of 178

1 of 178

Instructions for companies

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

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

Companies making evidence submissions to NICE should also refer to the NICE health technology evaluation guidance development manual.

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

Highlighting in the template (excluding the contents list)

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

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

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

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 2 of 178

2 of 178

Contents

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 3 of 178

3 of 178

List of Tables

Table 1: The decision problem ................................................................................. 14 Table 2: Technology being evaluated....................................................................... 16 Table 3: Tumour staging of CRC .............................................................................. 19 Table 4: NICE recommended treatments for previously treated mCRC (Subsequent or alternative therapy [NG151]) ................................................................................ 23 Table 5: Clinical effectiveness evidence ................................................................... 27 Table 6: Summary of trial methodology for CORRECT and CONCUR .................... 33 Table 7: CORRECT – baseline characteristics (ITT) ................................................ 41 Table 8: CONCUR – baseline characteristics (ITT) .................................................. 43 Table 9: Summary of statistical analyses for CORRECT and CONCUR .................. 48 Table 10: CORRECT – tumour response (ITT) ........................................................ 56 Table 11: CORRECT – patients completing the EQ-5D-3L questionnaire (ITT) ....... 58 Table 12: CONCUR – tumour response (FAS) ........................................................ 62 Table 13: Summary of the trials used to carry out the indirect treatment comparisons 66 Table 14: Results of the fixed effects NMA of OS .................................................... 68 Table 15: Results of the fixed effects NMA of PFS................................................... 68 Table 16. ITC sensitivity analyses ............................................................................ 69 Table 17: List of observational studies (real-world evidence) of regorafenib ............ 73 Table 18: Summary of efficacy outcomes in the real-world studies ......................... 79 Table 19: CORRECT – overview of TEAEs (SAS) ................................................... 80 Table 20: CORRECT – TEAEs occurring in ≥ 5% of patients in either group from start of treatment to 30 days after end of treatment (SAS) ....................................... 81 Table 21: CONCUR – overview of TEAEs (SAS) ..................................................... 87 Table 22: CONCUR – TEAEs occurring at any grade in ≥ 10% of patients, or at Grade ≥ 3 in any patients in either group, from the start of treatment to 30 days after the end of treatment (SAS) ....................................................................................... 88 Table 23: Summary list of published cost-effectiveness studies ............................ 100 Table 24: Features of the economic analysis ......................................................... 109 Table 25: Goodness-of-fit statistics of the pooled overall survival extrapolations ... 115 Table 26: Modelled trifluridine/tipiracil outcomes compared to reported outcomes in TA405 120 Table 27: Utilities used in the base case of the model ........................................... 124 Table 28: Utility systemic literature review results .................................................. 126 Table 29: Grade 3+ treatment related AEs reported in at least 2% of patients in any treatment arm ......................................................................................................... 128 Table 30: Disutilities per adverse event included in the model ............................... 130 Table 31: Summary of utility values for cost-effectiveness analysis ....................... 132 Table 32: Trifluridine/tipiracil dosing based on BSA ............................................... 134 Table 33: Dose intensity information used in the model ......................................... 135 Table 34: Monthly HRU rates in studies identified by the SLR ............................... 137 Table 35: HRU assumptions used in the model, based on TA405 ERG report and expert input ............................................................................................................ 139 Table 36: Costs per adverse event included in the model ...................................... 140 Table 37: Aggregate weekly adverse event costs used in the model ..................... 142 Table 38: Summary features of QALY shortfall analysis ........................................ 143 Table 39: Summary of health state benefits and utility values for QALY shortfall analysis .................................................................................................................. 144

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 4 of 178

Table 40: Summary of QALY shortfall analysis ...................................................... 144 Table 41: Summary of variables applied in the economic model ........................... 145 Table 42: Key model assumptions ......................................................................... 147 Table 43: Base-case results ................................................................................... 149 Table 44: Net health benefit ................................................................................... 149 Table 45: Probability sensitivity analysis results ..................................................... 152 Table 46: NMB and ICER results of regorafenib vs. trifluridine/tipiracil .................. 154 Table 47: Scenario analyses performed versus trifluridine/tipiracil ......................... 156 Table 48: Deterministic scenario results versus trifluridine/tipiracil ........................ 160 Table 49: Probabilistic scenario results versus trifluridine/tipiracil .......................... 161 Table 50: Deterministic scenario results versus BSC ............................................. 162 Table 51: Probabilistic scenario results versus BSC .............................................. 162 Table 52: Model outcomes compared to published trial results .............................. 166

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 5 of 178

List of Figures

Figure 1: Treatment pathway in mCRC .................................................................... 22 Figure 2: The proposed placement of regorafenib in the clinical pathway in England 25 Figure 3: CORRECT study schematic ...................................................................... 30 Figure 4: CONCUR study design ............................................................................. 32 Figure 5: CORRECT – Kaplan–Meier curve of overall survival (ITT) ....................... 54 Figure 6: CORRECT – Kaplan–Meier curve of progression-free survival (ITT) ........ 55 Figure 7: CONCUR – Kaplan–Meier curve of overall survival (FAS) ........................ 60 Figure 8: CONCUR – Kaplan–Meier curve of progression-free survival (FAS) ........ 61 Figure 9: Direct meta-analysis – overall survival – CORRECT and CONCUR ......... 65 Figure 10: Direct meta-analysis – progression-free survival – CORRECT and CONCUR ................................................................................................................. 65 Figure 11: Network diagram for OS and PFS ........................................................... 67 Figure 12: Overview of sensitivity analysis results for OS ........................................ 71 Figure 13: Overview of sensitivity analysis results for PFS ...................................... 71 Figure 14: Model structure ..................................................................................... 107 Figure 15: Short-term parametric fits of regorafenib overall survival extrapolations116 Figure 16: Long-term parametric fits of regorafenib overall survival extrapolations 117 Figure 17: Short-term parametric fits of BSC overall survival extrapolations .......... 118 Figure 18: Long-term parametric fits of BSC overall survival extrapolations .......... 119 Figure 19: Regorafenib and BSC survival curves used in the model base case .... 121 Figure 20: Trifluridine/tipiracil survival curves used in the model base case .......... 123 Figure 21: Cost effectiveness plane of regorafenib vs trifluridine/tipiracil (NMB calculated using a WTP of £30,000)....................................................................... 151 Figure 22: Cost effectiveness plane of regorafenib vs BSC. .................................. 151 Figure 23: Cost-effectiveness acceptability curve .................................................. 152 Figure 24: NMB results of regorafenib vs. trifluridine/tipiracil (WTP of £30,000) .... 154 Figure 25: Tornado diagram of the 10 most influential scenarios versus trifluridine/tipiracil (NMB calculated using a WTP of £30,000) ................................ 158 Figure 26: Regorafenib modelled OS compared to pooled trial results .................. 165 Figure 27: BSC modelled OS compared to pooled trial results .............................. 165

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 6 of 178

Abbreviations

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 7 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 8 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 9 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 10 of 178

10 of 178

Executive summary

Colorectal cancer (CRC), also known as bowel cancer, is the fourth most common cancer in the UK. It is a heterogeneous disease characterized by cancerous growths in the large intestine (colon) and rectum[1-3] . Stage IV metastatic CRC (mCRC) is an advanced form of CRC that has metastasized beyond the large intestine and nearby lymph nodes, typically spreading first to the liver.[4, 5] Patients with Stage IV mCRC have a poor prognosis, with 1-year survival rates of approximately 44%, and 5-year survival rates of less than 10%.[6, 7] Survival outcomes in the ≥ 3L setting are particularly poor, ranging between 6–12 months.[8, 9]

In England and Wales, first and second-line treatments for mCRC are chemotherapy and biological therapy. After failure of these therapies patients move to ‘subsequent or alternative therapy’ (NG151) i.e. NICE TA405 recommends trifluridine/tipiracil, if fluoropyrimidine-, oxaliplatin-, or irinotecan-based chemotherapies, anti-VEGF agents and anti-EFGR agents have failed or when these agents are not suitable.

Bayer is making this submission on the request of physicians who want an alternative to the chemotherapy trifluridine/tipiracil. Physicians consider trifluridine/tipiracil and regorafenib to be of comparable efficacy and comparable tolerability, but with the key difference of having different adverse effect profiles. A medicines adverse effect profile is a significant consideration when making individual treatment decisions.

This is a restricted submission where a recommendation for regorafenib is sought in a ≥3[rd] line setting i.e. a similar position to trifluridine/tipiracil. As such trifluridine/tipiracil is the key comparator in this submission and is the focus of the economic evidence.

An indirect treatment comparison shows regorafenib and trifluridine/tipiracil to be of comparable efficacy in the key outcomes of overall survival (HR 0.99 CI 0.84, 1.17) and progression free survival (0.93 CI 0.85, 1.03) - the base case point estimates favouring regorafenib to a small but non-significant degree for both outcomes. The findings of the ITC were robust to scenario analyses, with individual analyses numerically, but not statistically, favouring regorafenib or trifluridine/tipiracil.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 11 of 178

Regorafenib is available to the NHS with a confidential PAS. ''''''''' '''''''''''''''''' '''''''''''' '''' ''''''''''''''''''''''''''' ''''''''''''''''''' '''''''''' ''''' ''''''''''''''''''''''''''''' '''''''''''''''''''' '''' '''''''''''''''''''''''''''''''''''''' '''''''''' '''''''''' ''' ''''''''''''''''''''' ''''''''''''''''''' ''''''''''''''' '''''''''''''''''' ''''''''''''''' '''''''''''''''''''''''' ''''''' ''''''''''''''''''' ''''''''' '''' ''''''''''''' '''''''''''''''''''''''''' '''' '''' ''''''''' '''''''''''' '''''''''''' '''''''''' '''''''''''''''''' ''''' ''''''''''''''''''''''''''' ''''''''''''''''''''''''''''' '''''''''''''''''''''''' ''''''''''''''''''''''''''''''''''''' '''''''''''''''''''''''''''''' ''''''''''' '''''''''' ''''''''''''''' ''''''''''''''''''''''''''''''''''''''''''''' '''''' '''''''''''''''

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 12 of 178

Decision problem, description of the technology and clinical care pathway

B.1.1. Decision problem

Since regorafenib received approval in mCRC in 2013, treatment options have increased and clinical practice has changed. Patients are now treated with trifluridine/tipiracil where previously they would have received best supportive care.

Due to these changes in treatment practices, Bayer is not seeking a recommendation for the whole licensed population. This is a restricted submission where a recommendation for regorafenib is sought in a ≥3[rd] line setting i.e. a similar position to trifluridine/tipiracil. As such trifluridine/tipiracil is the key comparator in this submission and is the focus of the economic evidence.

The proposed position in the treatment pathway is narrower than the marketing authorisation. We are restricting the submission to these patients because physicians have requested an alternative to trifluridine/tipiracil at the third or later line setting. Physicians have indicated that regorafenib, with its comparable efficacy to trifluridine/tipiracil, and different adverse event profile, would provide a valuable additional option for patients.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 13 of 178

13 of 178

Table 1: The decision problem

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 14 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 15 of 178

B.1.2. Description of the technology being appraised

The summary of product characteristics (SmPC) is in Appendix C.

Regorafenib is an oral tumour deactivation agent that potently blocks multiple protein kinases, including kinases involved in tumour angiogenesis (VEGFR1, -2, -3, TIE2), oncogenesis (KIT, RET, RAF-1, BRAF, BRAF[V600E] ), metastasis (VEGFR3, PDGFR, FGFR) and tumour immunity (CSF1R). In particular, regorafenib inhibits mutated KIT, an oncogenic driver, and thereby blocks tumour cell proliferation. In preclinical studies regorafenib has demonstrated potent antitumour activity in a broad spectrum of tumour models including colorectal, gastrointestinal stromal and hepatocellular tumour models which is likely mediated by its anti-angiogenic and anti-proliferative effects. In addition, regorafenib reduced the levels of tumour associated macrophages and has shown anti-metastatic effects in vivo. Major human metabolites (M-2 and M-5) exhibited similar efficacies, compared to regorafenib in in vitro and in vivo models.

Regorafenib was approved by the European Medicines Agency (EMA) in August 2013 as monotherapy for the treatment of adult patients with metastatic colorectal cancer (mCRC) who have been previously treated with, or are not considered candidates for, available therapies. ‘Available therapies’ are defined as those available prior to 2013 and include fluoropyrimidine-based chemotherapy, antiVEGF therapy and anti-epidermal growth factor receptor (anti-EGFR) therapy (Table 2).[10]

Table 2: Technology being evaluated

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 16 of 178

16 of 178

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

B.1.3.1. Disease overview and epidemiology

Colorectal cancer (CRC; also known as bowel cancer) is a heterogeneous disease characterized by cancerous growths in the large intestine and rectum.[1] The majority of CRCs arise from precursor lesions that have transformed to adenocarcinoma; > 90% are adenocarcinomas originating from epithelial cells.[1] Molecular carcinogenesis pathways have been implicated in the development of CRC including

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 17 of 178

17 of 178

chromosomal instability (observed in 65–70% of sporadic colorectal tumours) and microsatellite instability (MSI; observed in nearly 15% of sporadic colorectal tumours).[11] In addition, CRC tumours can present with identifiable gene mutations that can act as prognostic factors of prognosis (e.g. BRAF and RAS V600E mutations).[12]

CRC is the fourth most common cancer in the UK; there were 34,825 new cases in England in 2017, which accounted for 11% of all new cancer cases.[2, 3 20] The current incidence of CRC in England is 68 incidence cases per 100,000.[3]

CRC patient symptomology includes bleeding from the rectum, changes in bowel habit, pain in the abdomen or rectum, involuntary weight loss and anaemia.[13, 14] The incidence of CRC is strongly related to age with approximately 44% of new cases in people aged 75 and over.[2] Other risk factors for developing CRC include male gender, poor diet and obesity.[15-17] However, the overall clinical presentation of CRC depends on the size and location of the tumour, as well as the absence or presence of any metastases. Tumour staging is most often done with the tumour node metastasis (TNM) classification system[18] (Table 3), although other classifications including the Dukes staging[19] system can also be used. The TNM classification system was used to stage patients in the pivotal CORRECT and CONCUR trials (see Section B.2).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 18 of 178

Table 3: Tumour staging of CRC

Stage IV mCRC is an advanced form of CRC that has metastasized beyond the large intestine and nearby lymph nodes, typically spreading first to the liver (Table 3).[4, 5] Over the course of the disease, according to the ESMO guidelines, approximately 55% of all patients with Stage II and III CRC will develop metastases and progress to Stage IV mCRC.[21] At initial diagnosis, 23–26% of patients present with Stage IV disease, in which synchronous CRC liver metastases are present in 15–25% of cases, and metastases are confined to the liver in 70–80% of these cases.[22, 23] In England, the estimated annual number of mCRC cases is

approximately 22,016 patients; of these, 20–25% will reach the third-line or beyond (≥ 3L) setting (the population of interest to this submission).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 19 of 178

19 of 178

B.1.3.2. Survival outcomes for mCRC

Patients with Stage IV mCRC have very poor prognosis, with 1-year survival rates of approximately 44%, and 5-year survival rates of less than 10%.[6, 7] Furthermore, mortality rates for Stage IV mCRC are markedly higher than for those patients diagnosed at earlier stages of the disease.[7] For patients with Stage IV mCRC who do not receive active treatment (i.e. who are given best supportive care [BSC] without chemotherapy), life expectancy is only approximately 6 months.[24-26]

Even when treatment is given, patient prognosis decreases as the treatment line increases. For patients diagnosed at an early stage of the disease, surgery and adjuvant chemotherapy are the main treatments and can be curative in 50% of patients.[27] Upon disease progression or recurrence, FOLFOX (oxaliplatin, leucovorin, and 5-fluorouracil [5-FU]) or FOLFIRI (infusional folic acid/5-FU, and irinotecan) are accepted first- and second-line chemotherapy regimens, and these have demonstrated improved overall survival (OS) outcomes to an average of 20 months in first line, and 9–10 months in second line.[27] Survival outcomes in the ≥ 3L setting (the setting of interest to this submission) are typically in the region of 6–12 months.[8, 9]

The lack of treatment options for ≥ 3L mCRC is well-recognized, with clinicians in agreement that there is a high unmet clinical need for additional treatment options in ≥ 3L mCRC where different tolerability profiles impacts on the choice of medicine.

B.1.3.3. Burden of disease

Along with a significantly poor prognosis compared with earlier stages of the disease (see Section B.1.3.2), patients with Stage IV mCRC experience several burdensome symptoms including change in bowel habits, nausea, vomiting, constipation and fatigue. For those patients with liver metastases, pain on the right side of the abdomen, nausea, ascites and jaundice are also common.[28, 29] The high morbidity and mortality rates in mCRC are associated with a substantial patient burden with patients having a substantially impaired health-related quality of life (HRQL).[30] Psychological distress is common in mCRC with psychological symptoms of depression (ranging between 1.6% and 57.0%) and anxiety (1.0–47.2%) being highly prevalent.[31]

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 20 of 178

Another burden faced by patients is due to adverse events. Patients differ in their susceptibility and tolerance of different AEs, and it is often tolerability issues with prior therapies that heavily influences the next treatment choice.

Metastatic CRC also places a substantial impact on caregivers. Patients with mCRC experience multiple symptoms that affect their daily living; therefore, family caregivers have to cope with demanding role changes that may affect their own mental and physical health.[32] Furthermore, increased caregiving responsibilities to the neglect of caregiver vital activities (such as self-care) have been associated with impaired QoL in caregivers.[32]

B.1.3.4. Clinical pathway of care

The management of mCRC is focused on improving QoL and survival through the management of local and metastatic disease.[33]

Figure 1 illustrates the treatment pathway for mCRC in England, based on guidance issued by the National Institute for Health and Care Excellence (NICE). Clinicians have confirmed that they follow the treatment pathway as outlined by NICE. In the NHS in England, treatment decisions for mCRC are based on genetic testing (biomarker driven) and treatment in later-lines is informed by prior therapy. First and second-line treatment of mCRC is dominated by chemotherapy combination regimens, which are typically FOLFOX or XELOX (oxaliplatin and capecitabine), and less commonly FOLFOXIRI (oxaliplatin, leucovorin, 5-FU, and irinotecan) which accounts for only 10% of all front-line treatments for mCRC). For patients with specific mutations, clinicians have the additional choice of including biologics in the first-line setting (Figure 1).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 21 of 178

Figure 1: Treatment pathway in mCRC

==> picture [452 x 193] intentionally omitted <==

Key: BSC, best supportive care; chemo, chemotherapy; dMMR, mismatch repair deficient; EGFR, epidermal growth factor receptor; mCRC, metastatic colorectal cancer; MSI-high, microsatellite instability-high; MSS, microsatellite stable.

Notes:[a ] typically FOLFOX or XELOX, less common FOLFOXIRI;[b ] irinotecan (after FOLFOX) or FOLFIRI (after FOLFOX or XELOX);[c ] cetuximab or panitumumab;[d ] encorafenib + cetuximab in second- or third-line depending on prior oxaliplatin and irinotecan exposure;[e ] nivolumab + ipilimumab after fluoropyrimidine-based combination chemotherapy;[f ] pembrolizumab; Figure adapted from the NICE pathway for colorectal cancer[18] and mCRC.[34]

Table 4 presents an overview of the NICE-recommended treatments for later-line mCRC. Trifluridine/tipiracil is recommended for adults who have had previous treatment with available therapies, or for whom these available therapies are not suitable. Trifluridine/tipiracil is the main comparator for regorafenib in previously treated mCRC.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 22 of 178

Table 4: NICE recommended treatments for previously treated mCRC

(Subsequent or alternative therapy [NG151])

B.1.3.5. Other clinical guidelines

The European Society for Medical Oncology (ESMO)[21, 27] , the American Society of Clinical Oncology (ASCO)[35] and the National Comprehensive Cancer Network (NCCN)[33, 36] have guidelines for the treatment of mCRC. These guidelines are consistent with NICE’s treatment pathway with the exception that regorafenib is recommended in later-line mCRC. The ESMO and ASCO guidelines position trifluridine/tipiracil and regorafenib alongside each other as options in the ≥3L setting.

In the ESMO 2016 consensus guidelines, regorafenib is recommended as a thirdline treatment for patients previously treated with fluoropyrimidines, oxaliplatin, irinotecan, bevacizumab and EGFR antibodies (in RAS wild-type patients).[27] The ASCO guidelines for late-stage CRC published in 2020 also recommend regorafenib

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 23 of 178

23 of 178

as a third and fourth-line treatment.[35] The NCCN 2021 guidelines for colon cancer recommend regorafenib as a third and fourth-line treatment for patients with mCRC refractory to chemotherapy given before or after trifluridine/tipiracil.[33] Similarly, the NCCN 2021 guidelines for rectal cancer recommend regorafenib as a third and fourth-line treatment under the continuum of care for metastatic disease.[36]

B.1.3.6. Limitations with current treatments and remaining unmet needs in ≥ 3L mCRC

In England and Wales, patients who fail treatment with standard first and second-line therapies (chemotherapy and biological therapy) have limited options i.e.

• 1. biomarker targeted treatment: MSI or MMR deficiency - nivolumab and ipilimumab; BRAF V600E mutation positive - encorafenib plus cetuximab
• 2. the chemotherapy trifluridine/tipiracil.

Biomarker targeted therapies

The combination therapies of encorafenib plus cetuximab, and nivolumab plus ipilimumab are biological treatments that are restricted to patients whose tumours contain the relevant molecular alteration ( BRAF V600E mutation and MSI/mismatch repair [MMR] deficiency, respectively). Thus, these targeted therapies are not available to all patients, as most patients with mCRC do not have these genetic alterations (it is estimated that only 5–21% of patients with mCRC have the BRAF mutations[12] and only approximately 4–5% have MSI/MMR deficiency[37] ).

Trifluridine/tipiracil

Trifluridine/tipiracil is clinically effective and provides a meaningful increase in overall and progression free survival. However, being a chemotherapy, it is associated with multiple toxicities which are typical of these treatments. A proportion of patients will be reluctant/unable to start another chemotherapy based on their prior experience with this class of therapies and this thereby limits their options and ability to benefit from additional active treatment.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 24 of 178

Unmet need

There is a lack of treatment options for patients in the ≥ 3[rd] line setting. For patients for whom biomarker targeted therapies are not options there is only the chemotherapy trifluridine/tipiracil. There is a need for a chemotherapy-free alternative therapy with a different adverse event profile which would increase the options available.

B.1.3.7. Proposed positioning of regorafenib

Bayer is making this submission on the request of physicians who want an alternative to the chemotherapy trifluridine/tipiracil in the ≥ 3[rd] line treatment setting. As such, this is a restricted submission and we are seeking a recommendation for regorafenib in a similar position to trifluridine/tipiracil.

Figure 2 presents the proposed positioning of regorafenib in the treatment pathway.

Figure 2: The proposed placement of regorafenib in the clinical pathway in

England

==> picture [452 x 195] intentionally omitted <==

Key: BSC, best supportive care; chemo, chemotherapy; dMMR, mismatch repair deficiency; EGFR, epidermal growth factor receptor; L, line; MSI, microsatellite instability; MSS, microsatellite stable. Notes :[a ] typically FOLFOX or XELOX, less common FOLFOXIRI;[b ] irinotecan (after FOLFOX) or FOLFIRI (after FOLFOX or XELOX);[c ] cetuximab or panitumumab;[d ] encorafenib + cetuximab;[e ] nivolumab + ipilimumab;[f ] pembrolizumab. Figure adapted from the NICE pathway for colorectal cancer[18] and metastatic colorectal cancer.[34]

B.1.4. Equality considerations

No equality issues relating to the use of regorafenib have been identified.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 25 of 178

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

Clinical effectiveness

B.2.1. Identification and selection of relevant studies

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

B.2.2. List of relevant clinical effectiveness evidence

Table 5 summarizes the clinical-effectiveness evidence supporting regorafenib for the treatment of adult patients with mCRC receiving ≥ 3L therapy. The key data comes from the two pivotal randomized controlled trials (RCTs): CORRECT and CONCUR. Both trials were used to inform the cost-effectiveness model (see Section B.3).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 26 of 178

26 of 178

Table 5: Clinical effectiveness evidence

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 27 of 178

The key efficacy data presented in this document is taken from the primary publications for both trials, supplemented with data from the unpublished CSRs. Ten observational/RWE (non-randomised) studies were identified in the clinical SLR (Appendix D). These studies were not used to inform the economic model, but the results are presented in Section B.2.10 as supporting evidence to the pivotal trial data for regorafenib. Further details of these studies are presented in Table 17.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 28 of 178

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

B.2.3.1. Trial methodology

Table 6 provides details of CORRECT and CONCUR trial methodology, and a brief narrative summary of each trial is presented below.

CORRECT

CORRECT was a Phase III, Global, randomized, double-blind, placebo-controlled trial, designed to evaluate the efficacy and safety of regorafenib and BSC (hereafter termed regorafenib ) versus placebo plus BSC (hereafter termed placebo ) in adult (≥ 18 years) patients with Stage IV mCRC (Figure 3). Patients had progressed after approved standard treatments available at the time (including fluoropyrimidine, oxaliplatin, irinotecan, bevacizumab and cetuximab/panitumumab). The study was conducted in 114 centres in 16 countries in North America, Europe, Asia and Australia.

Patient eligibility (Table 6) included histological or cytological documentation of adenocarcinoma of the colon or rectum, Eastern Cooperative Oncology Group performance status (ECOG PS) of 0 or 1, standard therapy discontinuation due to unacceptable toxicity, and disease progression during or within 3 months after the last administration of the last standard therapy. For patients who had received adjuvant oxaliplatin, trial eligibility required progression during or within 6 months of therapy completion, or retreatment with oxaliplatin-based therapy if progression was later than 6 months. Patients with previous regorafenib treatment or uncontrolled medical disorders were ineligible for the trial.

Randomized (2:1) patients received oral regorafenib 160 mg or matching placebo once daily (OD) on Days 1–21 of each 28-day cycle until the determination of Response Evaluation Criteria in Solid Tumours (RECIST)-defined disease progression (PD), clinical progression, the development of severe adverse events (AEs), withdrawal from the study, death, or a decision by the treating physician that discontinuation would be in the patient’s best interest (Figure 3). Upon

discontinuation of either regorafenib or placebo treatment, all patients were to enter

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 29 of 178

29 of 178

a 30-day safety follow-up period. Regardless of the reason for discontinuation, all patients were followed for survival until death was documented, except for those who specifically withdrew consent to follow-up. Patients who withdrew consent from study drug treatment were allowed to enter the survival follow-up period after signing a separate informed consent for survival follow-up.

All patients received BSC (excluding other investigational anti-tumour agents or antineoplastic chemotherapy, hormonal therapy, or immunotherapy). No crossover between treatment groups was allowed (Table 6). Predefined dose modifications were permitted to manage clinically significant treatment-related toxic effects (see Table 1 and Table 2 of the regorafenib SmPc). Patient randomization was stratified by:

• Previous treatment with VEGF-targeting drugs (yes or no)

• Time from diagnosis of metastatic disease (≥ 18 months versus < 18 months)

• Geographical region (North America, Western Europe, Israel, and Australia; Asia; and South America, Turkey and Eastern Europe)

The primary endpoint in CORRECT was OS, defined as the time from randomization to death from any cause. Secondary endpoints included progression-free survival (PFS), tumour response (objective tumour response rate [ORR]), disease control rate (DCR) and safety. Tertiary endpoints included duration of response (DOR), duration of stable disease and QoL. See Table 6 for details and definitions of all study endpoints.

Figure 3: CORRECT study schematic

==> picture [425 x 124] intentionally omitted <==

Key: BSC, best supportive care; od, once daily; po, per os (by mouth); VEGF, vascular endothelial growth factor.

Source: Figure 7-1 of CORRECT Clinical Study Report (Final).[45]

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 30 of 178

CONCUR

CONCUR was a Phase III, randomized, double-blind, placebo-controlled, parallelgroup trial designed to evaluate the efficacy and safety of regorafenib versus placebo in Asian patients with Stage IV mCRC. Patients had progressed after ≥ 2 lines of previous standard therapy. The trial was designed with a near identical protocol to the CORRECT trial but was exclusively focused on adult (≥ 18 years) Asian patients. The study was conducted at 25 hospitals in mainland China, Hong Kong, South Korea, Taiwan and Vietnam.

Patient eligibility (Table 6) was as per the CORRECT trial but specified Asian ethnicity and at least two previous treatment lines, including a fluoropyrimidine plus oxaliplatin or irinotecan. Previous treatment with bevacizumab, cetuximab or panitumumab was allowed but was not mandatory. Randomized (2:1) patients received oral regorafenib 160 mg or matching placebo OD on Days 1–21 of each 28day cycle until the determination of RECIST-defined PD, clinical progression, the development of severe AEs, withdrawal from the study, death, or a decision by the treating physician that discontinuation would be in the patient’s best interest (Figure 4). However, patients with PD could continue treatment at the investigator’s discretion. Upon discontinuation of either regorafenib or placebo treatment, all patients were to enter a 30-day safety follow-up period. Regardless of the reason for discontinuation, all patients were followed for survival until death was documented, except for those who specifically withdrew consent to follow-up. Patients who withdrew consent from study drug treatment were allowed to enter the survival follow-up period after signing a separate informed consent for survival follow-up.

Similar to CORRECT, all patients received BSC (excluding other investigational antitumour agents or antineoplastic chemotherapy, hormonal therapy, or immunotherapy; Table 6) and predefined dose modifications were permitted to manage clinically significant treatment-related toxic effects (see Table 1 and Table 2 of regorafenib SmPc). However, in CONCUR, patient randomization was stratified by:

• Metastatic site (single versus multiple organs); and

• Time from diagnosis of metastatic disease (< 18 months versus ≥ 18 months)

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 31 of 178

31 of 178

The primary endpoint in CONCUR was OS (defined as the time from randomization to death from any cause) and secondary endpoints included PFS, tumour response (ORR), DCR and safety. Tertiary endpoints included DOR, duration of stable disease and QoL. See Table 6 for details and definitions of all study endpoints.

Figure 4: CONCUR study design

==> picture [377 x 122] intentionally omitted <==

Key: BSC, best supportive care; Od, once daily; po, per os (by mouth). Source: Figure 7-1 of CONCUR Clinical Study Report (Final).[48]

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 32 of 178

Table 6: Summary of trial methodology for CORRECT and CONCUR

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 33 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 34 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 35 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 36 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 37 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 38 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 39 of 178

Trial number (acronym) CORRECT: NCT01103323 CONCUR: NCT01584830

cardiac failure, clinically significant bleeding and severe skin infections (Stevens-Johnson Syndrome, erythema multiforme and toxic epidermal necrolysis) and acute liver failure. Source: CORRECT Clinical Study Report[45] ; CONCUR Clinical Study Report[48] ; Grothey et al., 2013[25] ; Li et al., 2015.[26]

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 40 of 178

B.2.3.2.

Baseline characteristics

CORRECT

Table 7 provides a summary of the baseline characteristics, including demographic and clinical characteristics, for patients in CORRECT. Overall, baseline characteristics were balanced between the two arms; the majority of patients were men (regorafenib: 62%; placebo: 60%) and white (regorafenib: 78%; placebo: 79%). The median age was 61 years for both groups. The majority of patients had received four or more previous systemic anti-cancer therapies (regorafenib: 49%; placebo: 47%), and all patients had received prior bevacizumab (an anti-VEGF monoclonal antibody). There was a higher proportion of patients in the placebo group who had progression on bevacizumab, irinotecan and oxaliplatin (84%, 90% and 63%, respectively) than in the regorafenib group (80%, 80% and 55%, respectively). The rates of ECOG PS 1 (48% versus 43%) and a positive BRAF mutation (4% versus 2%) were similar between arms (Table 7).

Table 7: CORRECT – baseline characteristics (ITT)

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 41 of 178

41 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 42 of 178

42 of 178

CONCUR

Table 8 provides a summary of the baseline characteristics, including demographic and clinical characteristics, for patients in CONCUR (all patients were Asian). Overall, baseline characteristics were balanced between arms with two exceptions: sex and age category. There were more males than females (58% versus 42%) overall in the study; however, there was a higher percentage of males in the regorafenib group than in the placebo group (63% versus 49%). The mean age was 56.5 years; however, there were more patients aged < 65 years old in the placebo group than in the regorafenib group (85% versus 70%). The majority of patients had received four or more systemic anti-cancer therapies (regorafenib: 54%; placebo: 51%), and overall, 40% of patients had not previously received any biological treatment (anti-VEGF or otherwise) before randomization. However, more patients in the placebo arm (25%) than those in the regorafenib arm (18%) received an antiEGFR but not anti-VEGF as a prior targeted therapy. The rates of ECOG PS 1 (74% versus 78%) and a positive BRAF mutation (0% versus 1%) were similar between arms (Table 8).

Table 8: CONCUR – baseline characteristics (ITT)

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 43 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 44 of 178

44 of 178

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

B.2.4.1. Statistical analyses

Table 9 provides a summary of the statistical analysis and definitions of analysis sets in CORRECT and CONCUR. A brief narrative summary of the statistical analyses of each trial is presented below.

CORRECT

CORRECT was designed to have 90% power to detect a 33.3% increase in median OS, assuming a 4.5-month median OS for the placebo group (i.e. a hazard ratio [HR] of 0.75 for regorafenib over placebo). A total of 582 death events were required (one-sided Type 1 error rate of 0.025) and a total of 690 patients were planned for randomization (2:1). Two formal interim analyses were planned during the study when approximately 30% and 70% of the planned total number of required death events had occurred. Statistical evaluation was performed using the software package SAS[®] (Version 9.1 or higher). OS and PFS were compared between treatment groups with a stratified log-rank test; HRs (with 95% confidence intervals [CIs]) were calculated with the Cox model, adjusting for stratification factors; and Kaplan–Meier survival estimates were calculated for each treatment group. ORR and DCR were compared between treatment groups with the Cochran–Mantel–Haenszel test, adjusting for stratification factors. Safety data were listed and summarized descriptively for the safety population. AEs were coded according to the Medical Dictionary for Regulatory Activities (MedDRA) v14.1 and graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE) v3.0.

The formal efficacy analyses were based on the intention-to-treat (ITT) population (data cut-off 21 July 2011). At the time of the analysis no patients had crossed over from placebo to regorafenib therapy.

Patients were followed on a monthly basis for safety until all patients were off treatment. Safety analyses were based on the safety analysis set (SAS) – data cutoff 27 March 2014.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 45 of 178

CONCUR

CONCUR was designed to have 80% power to detect a 33.3% increase in median OS (i.e. an HR of 0.75). A total of 154 deaths were estimated to be required, with randomization planned for approximately 200 patients. Statistical evaluation was performed using the software package SAS. OS and PFS were compared between treatment groups with a stratified log-rank test; HRs (with 95% CIs) were calculated with the Cox model, adjusting for stratification factors; and Kaplan–Meier survival estimates were calculated for each treatment group. ORR and DCR were compared between treatment groups with the Cochran–Mantel-Haenszel test, adjusting for stratification factors. Safety data were listed and summarized descriptively for the safety population. AEs were coded according to MedDRA and graded according to NCI CTCAE v4.0.

The formal efficacy analyses were based on the full analysis set (FAS) – data cut-off 29 November 2013, which was defined as all randomized patients (Table 9).

Patients were followed on a monthly basis for safety until all patients were off treatment. All safety analyses were based on the safety analysis set (SAS) – data cut-off 14 January 2016.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 46 of 178

46 of 178

B.2.4.1.1. Patient withdrawals (CORRECT and CONCUR)

Patients were allowed to withdraw from treatment at any time at their own request or withdraw at the discretion of the investigator or sponsor due to safety or behavioural reasons. Patients withdrawn from the study treatment were not replaced.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 47 of 178

47 of 178

Table 9: Summary of statistical analyses for CORRECT and CONCUR

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 48 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 49 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 50 of 178

B.2.4.2. Patient disposition data

Appendix D provides the CONSORT diagrams for CORRECT and CONCUR. A brief overview of the patient disposition for each trial is summarized below.

CORRECT

A total of 1,052 patients completed screening, of whom 760 were randomized (regorafenib, n = 505; placebo, n = 255 [ITT population]) between 19 May 2010 and 22 March 2011. Most randomized patients were enrolled in Italy (n = 143), France (n = 116), Japan (n = 100), Belgium (n = 87), Spain (n = 83), US (n = 83) and Germany (n = 64). Overall, 753 patients initiated treatment (regorafenib: n = 500; placebo: n = 253 [safety population]). Of the seven patients who did not receive treatment, four patients (n = 2 for each group) had AEs that were associated with clinical disease progression and one patient (regorafenib group) had an AE that was not associated with clinical disease progression. Of the remaining two patients who did not receive treatment, one patient had an ECOG PS of > 1 after randomization, and the other patient withdrew consent.

In total, 88.7% and 95.7% of patients in each arm (regorafenib versus placebo), respectively, discontinued treatment. 540 patients (71.1%) entered post-treatment survival follow-up (regorafenib: n = 353; placebo: n = 187). In addition, a total of 21 patients were permitted to continue treatment after PD was proven as they derived clinical benefit from treatment. At the efficacy data cut-off date (21 July 2011), 61 patients (regorafenib: n = 52; placebo: n = 9) were still receiving treatment.

CONCUR

A total of 234 patients were screened, of whom 204 were randomized (regorafenib: n = 136; placebo: n = 68 [FAS population]) between 29 April 2012 and 19 January 2013. The primary reasons for screening failure were not meeting study eligibility criteria (15.2%) and patient withdrawal (0.8%). All 204 randomized patients were included in the FAS population and also in the safety population.

At the efficacy data cut-off of 29 November 2013, the majority of patients (regorafenib: 95.6%; placebo: 100%) had discontinued treatment, while six patients

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 51 of 178

51 of 178

in the regorafenib group continued treatment. By the safety data cut-off date (14 January 2016), all 204 patients in the study had terminated treatment. Safety followup data were available for 192 patients.

B.2.5. Critical appraisal of the relevant clinical effectiveness evidence

Quality assessment of CORRECT and CONCUR was conducted using the NICE checklist, the full details of which are provided in Appendix D. Both trials were approved by the institutional review board or independent ethics committee for each study centre and were both conducted according to Good Clinical Practice. Overall, both trials were considered to be methodologically robust, high-quality studies with a comprehensive approach to patient allocation, control of confounding factors, and an overall low risk of bias.[25, 26]

Overall, the results of both studies are expected to be generalizable to patients in England and Wales. There were a two main areas where the trials diverged from the baseline characteristic of patients in England and Wales i.e. in respect of ethnicity and prior treatment.

Ethnicity

Regarding applicability of the trials to English clinical practice, CONCUR enrolled exclusively Asian patients, while CORRECT enrolled European and Asian patients from across 16 countries including Western Europe. Thus, it was possible to observe whether Asian and non-Asian patients responded to regorafenib in a similar manner; indeed, subgroup analyses of CORRECT and CONCUR confirmed that race was not a treatment effect modifier for regorafenib[49] . Clinicians have confirmed that ethnicity is not expected to be a treatment effect modifier (see Section B.2.7). It is therefore reasonable to generalize the results of both studies to Western populations.

Prior treatments: Anti-VEGF

In CORRECT, all patients had previously received anti-VEGF targeted therapy (i.e. bevacizumab), whilst in CONCUR, patients were enrolled who had not received previous treatment with bevacizumab.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 52 of 178

Bevacizumab (an anti-VEGF) is not recommended by NICE (either as monotherapy or as part of a combination regimen) for patients with mCRC.[34] Thus, of the two trials, CONCUR more closely aligns with clinical practice in England and Wales as it includes a significant proportion of patients who have never received anti-VEGF therapy. Prior treatment with anti-VEGF is relevant as regorafenib has anti-VEGF activity (see Section B.1.2) - the implication of this prior therapy is that regorafenib could be expected to be less effective in patients who have already been treated with, and failed on, an anti-VEGF.

Prior treatments: number received for mCRC

CORRECT recruited a more heavily pre-treated population (Table 7) than CONCUR (Table 8). It is recognised that prognosis, and ability to benefit from treatment diminishes with each additional line of therapy. On this basis a difference in efficacy between CORRECT and CONCUR might be anticipated.

B.2.6. Clinical effectiveness results of the relevant trials

B.2.6.1. CORRECT

B.2.6.1.1. Primary efficacy outcome: overall survival

The median duration of treatment was 2.8 versus 1.8 months for regorafenib versus placebo. A total of 432 death events occurred in the ITT population (n = 760), with the majority occurring in the placebo group (regorafenib: 54.5%; placebo: 61.6%; Appendix M; Table 79).

In CORRECT, the primary endpoint was met. Regorafenib significantly prolonged OS compared with placebo (HR 0.77; 95% CI: 0.64, 0.94; p = 0.0052; Appendix M; Table 79 and Figure 5). ''''''''''''''''' ''''''''''''''''''''''''''''''' ''''''''' '''''''''''''''' ''''''''''''''''''''''' of patients had experienced an OS event (Appendix M; Table 79), and median OS was 6.4 versus 5.0 months, respectively. The OS rate was higher in the regorafenib group than in the placebo group at 3 (80.3% versus 72.7%), 6 (52.5% versus 43.5%), 9 (24.3% versus 24.0%) and 12 (24.3% versus 24.0%) months post-randomization, respectively. '''''''''''''''''''' ''''' ''''''''' ''''''''''''''''''''''' '''''''''''''''''''''' '''''''''''' ''''''''''''''''''''''''' ''''' ''''''''' '''''''''''''''''''''''''' ''''''''' '''''''' ''''''''''''''''' ''''''''''''''''''' '''' ''''''''' '''''''''''''''''''''' '''''' '''''''''''''' '''''''''.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 53 of 178

Figure 5: CORRECT – Kaplan–Meier curve of overall survival (ITT)

==> picture [435 x 202] intentionally omitted <==

Key: CI, confidence interval; HR, hazard ratio; ITT, intention-to-treat. Source: Figure 2. Grothey et al., 2013.[25]

B.2.6.1.2. Secondary efficacy outcomes

B.2.6.1.2.1. Progression-free survival

PFS was significantly longer in the regorafenib group compared with placebo (HR 0.49; 95% CI: 0.42, 0.58; p < 0.0001; Figure 6; and Appendix M; Table 81). ''''''''''''''''

'''''''''''''''''''''''''''''' '''''''''' ''''''''''''''' '''''''''''''''''''''' of patients had experienced a PFS event (Appendix M; Table 81), and the median PFS was 1.9 versus 1.7 months, respectively. PFS was consistently higher in the regorafenib group than in the placebo group at 3, 6 and 9 months post-randomization: '''''''''''''' '''''''''''''''' ''''''''''''''''''

''''''''''''''' ''''''''''''''' ''''''''''''' '''''''''' '''''''''''' ''''''''''''''' '''''''''''''' respectively. ''''''''''''''''''' '''' '''''''' '''''''''''''''''''''''' ''''''''''''''''''''''' ''''''''''''' ''''''''''''''''''''''''' ''''' ''''''''' '''''''''''''''''''''' ''''''''''' '''''''''''''' '''''''''''''''''''''' '''''''''''''''''''''' ''''''' '''''''''''' '''''''.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 54 of 178

54 of 178

Figure 6: CORRECT – Kaplan–Meier curve of progression-free survival (ITT)

==> picture [443 x 209] intentionally omitted <==

Key: CI, confidence interval; HR, hazard ratio; ITT, intention-to-treat. Note: Kaplan–Meier curves based on investigator assessment Source: Figure 3. Grothey et al., 2013.[25]

B.2.6.1.2.2. Objective response rate and disease control rates

In CORRECT, no patients had a complete response (CR); five patients assigned regorafenib and one patient assigned placebo had a partial response (PR), giving ORRs of 1.0% and 0.4%, respectively (p = 0.19; Table 10). Compared with placebo, regorafenib was associated with significant disease control (41% [n = 207] versus 15% [n = 38]; p < 0.0001). In addition, more patients in the regorafenib group compared with placebo had a shrinkage in tumour size (''''''% versus '''%; Appendix M; Table 83).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 55 of 178

Table 10: CORRECT – tumour response (ITT)

B.2.6.1.3. Tertiary outcomes

B.2.6.1.3.1. Duration of response and duration of stable disease

For the '''''''' '''''''''''''''''' in the regorafenib group and the '''''''''' ''''''''''''''''' in the placebo group who had a PR, the median duration of response could not be estimated in the regorafenib group due to the small number of patients in the group (range without censored values: ''''''''''''''' days) and was '''''' days for the one patient in the placebo group (Appendix M; Table 84). Median duration of stable disease was 2.0 months (interquartile range [IQR]: 1.7–4.0) in the regorafenib group and 1.7 months (IQR: 1.4–1.9) in the placebo group (Appendix M; Table 84).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 56 of 178

56 of 178

B.2.6.1.3.2. Patient-reported outcomes

Appendix M: Tables 85 and 86 provides a summary of the patient-reported outcomes (PROs); please refer to Table 6 for the PRO instruments ranking rules.

In brief, the mean EORTC QLQ-C30 score at baseline (Day 1, Cycle 1) was 62.6 and 64.7 in the regorafenib and placebo groups, respectively. The mean scores at the end of treatment (EOT) visit were 48.9 and 51.9, respectively. The mean EQ-5D™ index scores were similar between the two groups at baseline (regorafenib: 0.73; placebo: 0.74) and EOT (0.59 for both groups). The mean EQ-5D visual analogue scale (VAS) scores were also similar between arms at baseline (regorafenib, 65.4; placebo, 65.8) and at EOT (regorafenib, 55.5; placebo, 57.3).

The five dimensions of the EQ-5D-3L were further evaluated.[38] In the analysis conducted by Su et al., the authors calculated the proportion of ITT patients reporting ‘no problems’ in each of the five dimensions (mobility, self-care, usual activities, pain/discomfort and anxiety/depression; dimensions can be graded as ‘no problem’, ‘some problems’ or ‘severe problems’). At baseline, 95% of the ITT population completed the EQ-5D-3L questionnaire; patients reported the highest percentage of ‘no problems’ in self-care (87%) and lowest in pain/discomfort (35%) (Table 11). EQ-5D completion rates declined with each cycle of treatment, as did the percentages of patients reporting ‘no problems’ in each dimension; notably, the declines were more rapid for placebo than for regorafenib. Both completion rate and percentages of ‘no problems’ were somewhat higher for placebo than for regorafenib at baseline. This continued only until Cycle 2, and reversed completely from Cycle 3 onward – showing an increasingly diminished proportion of those on placebo able to complete the questionnaire and reporting ‘no problem’ in each dimension. The differences between arms became larger with each cycle beginning Cycle 3 (Table 11). Overall, the results indicate that regorafenib appears to enable patients to maintain their mobility, self-care and usual activity, and to remain pain-free and anxiety-free, as well as their ability to complete the EQ-5D questionnaire, at much higher rates than placebo.[38]

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 57 of 178

Table 11: CORRECT – patients completing the EQ-5D-3L questionnaire (ITT)

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 58 of 178

B.2.6.1.4. Overview of efficacy

The Phase III CORRECT trial demonstrated a significant prolongation of survival outcomes with the addition of regorafenib to BSC in patients with relapsed/refractory mCRC who had had the option of receiving all approved standard therapies that were available at the time of study initiation. The primary endpoint of OS was in favour of regorafenib compared with BSC alone with an HR of 0.77 (95% CI: 0.64, 0.94), which translates into a 23% reduction in risk of death with regorafenib during the course of the study. Similarly, analyses of PFS (p < 0.0001) and disease control (p < 0.0001) were consistently in favour of regorafenib over BSC alone, with the data indicating that the main effect of regorafenib on mCRC appears to be disease stabilization. Sensitivity analyses (both stratified and unstratified) of OS and PFS were consistent with the main findings and predefined subgroup analyses of OS and PFS were aligned with the main findings, demonstrating consistent benefit of regorafenib over BSC.

Overall, the observed benefit in OS and PFS is clinically significant and meaningful in a patient population with very poor prognosis and limited treatment options for whom prior therapies have failed.

B.2.6.2. CONCUR

B.2.6.2.1. Primary efficacy outcome: overall survival

The median duration of treatment was 2.4 versus 1.6 months for regorafenib versus placebo. A total of 155 death events occurred in the FAS population (n = 204), with the majority occurring in the placebo group (regorafenib: 70%; placebo: 88%). Median follow-up for the OS analysis was 7.4 months (IQR: 4.3–12.2).

In CONCUR, the primary endpoint was met. Regorafenib significantly prolonged OS compared with placebo (HR 0.55; 95% CI: 0.40, 0.77; p = 0.00016; Figure 7; and Appendix M; Table 89). ''''''''''''''''' '''''''''''''''''''''''''''' ''''''''' '''''''''''''''' '''''''''''''''''''''''' of patients had experienced an OS event (Appendix M; Table 89), and median OS was 8.8 versus 6.3 months, respectively. The OS rate was higher in the regorafenib group than in the placebo group at 3 (''''''''''% versus '''''''''''%), 6 (''''''''''% versus '''''''''''%), 9 (''''''''''% versus ''''''''''''%) and 12 (''''''''''% versus '''''''''''%) months post-randomization,

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 59 of 178

59 of 178

respectively. '''''''''''''''''''' '''' '''''''' ''''''''''''''''''''''' '''''''''''''''''''' ''''''''''' ''''''''''''''''''''''''' ''''' ''''''''' ''''''''''''''''''''''

''''''''''' ''''''' '''''''''''''''' '''''''''''''''''''''' '''' '''''''' '''''''''''''''''''''''''' '''''' '''''''''''''' '''''''''

Figure 7: CONCUR – Kaplan–Meier curve of overall survival (FAS)

==> picture [423 x 195] intentionally omitted <==

Key: CI confidence interval; FAS, full analysis set; HR, hazard ratio. Source: Figure 2. Li et al., 2015.[26]

B.2.6.2.2. Secondary efficacy outcomes

B.2.6.2.2.1. Progression-free survival

PFS was significantly longer in the regorafenib group compared with placebo (HR 0.31; 95% CI: 0.22, 0.44; p < 0.0001; Figure 8; and Appendix M; Table 91). '''''''''''''''' '''''''''''''''''''''''''''''' '''''''' ''''''''''''''''' ''''''''''''''''''''' of patients had experienced a PFS event (Appendix M; Table 91), and the median PFS was 3.2 versus 1.7 months,

respectively. '''''''''' '''''''''' '''''''''''''''''''''''''' '''''''''''''''' ''''' ''''''''' ''''''''''''''''''''''''''' '''''''''''' ''''''''''' '''' '''''''' ''''''''''''''''''' ''''''''''''' ''''' ''''' ''''' ''' '''''''''' '''''' ''''''''''''''''''' '''''''''''''''''''''''''''''''''''''''''' '''''''''''''''''''''' ''''' '''''''' '''''''''''''''''''''''' ''''''''''''''''''''' ''''''''''' ''''''''''''''''''''''''''' ''''' ''''''''' ''''''''''''''''''''''' ''''''''' ''''''''''''' ''''''''''''''''''' '''''''''''''''''''''''''' ''''''' '''''''''''' '''''''''

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 60 of 178

60 of 178

Figure 8: CONCUR – Kaplan–Meier curve of progression-free survival (FAS)

==> picture [430 x 196] intentionally omitted <==

Key: CI confidence interval; FAS, full analysis set; HR, hazard ratio. Note: Kaplan–Meier curves based on investigator assessment Source: Figure 3. Li et al., 2015.[26]

B.2.6.2.2.2. Objective response rate and disease control rates

In CONCUR, no patients had a CR. Six patients assigned regorafenib had a PR, while no patients receiving placebo had a PR. More patients receiving regorafenib than placebo had stable disease as best overall response (45.6% versus 7.4%), and the ORR was 0% versus 4.4%, respectively (Table 12). Compared with placebo, regorafenib was associated with significant disease control (51.5% [n = 70] versus 7.4% [n = 5]; p < 0.0001). In addition, more patients in the regorafenib group compared with the placebo group had a shrinkage in tumour size ('''''% versus ''''%; Appendix M; Table 93).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 61 of 178

Table 12: CONCUR – tumour response (FAS)

B.2.6.2.3. Tertiary outcomes

B.2.6.2.3.1. Duration of response and duration of stable disease

The median DOR in the six patients in the regorafenib group who had a PR was 4.8 months. The median duration of stable disease was 3.0 months in the regorafenib group and 1.7 months in the placebo group (Appendix M; Table 94).

B.2.6.2.3.2. Patient-reported outcomes

Appendix M; Tables 95 and 96 provide a summary of the PROs; please refer to Table 6 for the PRO instruments ranking rules.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 62 of 178

62 of 178

In brief, the mean EORTC QLQ-C30 score at baseline (Day 1, Cycle 1) was 66.7 and 58.0 in the regorafenib and placebo groups, respectively. The mean scores at the EOT visit were 51.1 and 52.2, respectively. The mean EQ-5D index scores were similar in the two groups at baseline (regorafenib: 0.84; placebo: 0.75) ) and EOT (0.57 for both groups). The mean EQ-5D VAS scores were also similar between arms at baseline (regorafenib: 73.4; placebo: 71.4) and EOT (regorafenib: 61.5; placebo: 62.6).

B.2.6.2.4. Overview of efficacy conclusions

The Phase III CONCUR trial demonstrated that, for patients with relapsed/refractory mCRC, the addition of regorafenib to BSC resulted in a clinically relevant and statistically significant prolongation of survival outcomes. Compared with placebo, regorafenib significantly prolonged OS (HR 0.55; 95% CI: 0.40,0.77; p = 0.000159) and PFS (HR 0.31; 95% CI: 0.22,0.44; p < 0.0001) and was associated with statistically significantly improved disease control rates (p < 0.0001). Overall, the subgroup and sensitivity analyses were consistent and supportive of the main survival (OS and PFS) outcomes.

CONCUR and CORRECT were similarly designed studies, however a larger benefit was observed in CONCUR. Clinical experts have advised that efficacy is not expected to differ according to ethnicity so the difference is not explained by CONCUR being conducted in an Asian population. Two possible explanations, aside from random fluctuation (i.e. the confidence intervals are overlapping), are related to prior treatment i.e. the total number of prior treatments received; and the proportion of patients pre-treated with anti-VEGF.

• CORRECT recruited a more heavily pre-treated population (Table 7) than CONCUR (Table 8). It is understood that prognosis, and ability to benefit from treatment diminishes with each additional line of treatment. Ceteris paribus you might anticipate a better efficacy from CONCUR on this basis.

• All patients in CORRECT had previously received anti-VEGF therapy (i.e. bevacizumab), whilst in CONCUR, patients were enrolled who had not received previous treatment with bevacizumab. Prior treatment with antiVEGF is relevant as regorafenib has anti-VEGF activity (see Section B.1.2) - the implication of this prior therapy is that regorafenib could be expected to be

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 63 of 178

63 of 178

less effective in patients who have already been treated with, and failed on, an anti-VEGF (as was the case in the CORRECT study). As a consequence, the results seen in CONCUR could be more representative of expectations for patients in the NHS.

B.2.7. Subgroup analysis (CORRECT and CONCUR)

In both CORRECT and CONCUR, subgroup analyses of OS and PFS were preplanned and were prespecified, as listed in the statistical analysis plan (SAP). For both trials, descriptive statistics and HR estimates with 95% CIs for OS and PFS were calculated for each baseline stratification factor and other relevant baseline variables, provided there was a sufficient number of events in total within the subgroup across the treatment arms. Forest plot representations were also provided. Subgroup analyses were performed based on demographic information (e.g. race, sex, age group [< 65 years, ≥ 65 years]), region (Region 1: North America, Western Europe, Israel, and Australia; Region 2; Asia; Region 3, South America, Turkey, and Eastern Europe), time from diagnosis of metastatic disease (≥ 18 months and < 18 months), prior systemic anti-cancer therapies, historical KRAS mutation status, and baseline cancer characteristics of primary site of tumour (e.g. ECOG PS: 0 or 1). Subgroup analyses of AEs were also conducted.

Overall, in both trials, the efficacy subgroup analyses demonstrated consistent survival benefits with regorafenib over placebo, with OS and PFS outcomes that were generally comparable with those observed in the overall populations.

A summary of results for subgroups analysed in both trials are provided in Appendix E.

B.2.8. Meta-analysis

A direct meta-analysis using fixed and random effects models for OS and PFS data from CORRECT and CONCUR was performed. The generic inverse variance method for meta-analysis was performed using the ‘ meta ’ package in R.[50] Results are shown in Figure 9 and Figure 10, respectively.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 64 of 178

The fixed and random effects models estimated similar OS and PFS HRs for regorafenib versus placebo. For OS, the fixed effects HR was 0.68 (95% CI: 0.58, 0.79) and for PFS it was 0.42 (95% CI: 0.39, 0.45).

The I2 statistic was estimated at 82% in the analysis of OS and 97% in the analysis of PFS, which suggests a high level of heterogeneity between the studies. Study heterogeneity was explored in detail in the context of indirect treatment comparisons (B.2.9) and studies were considered sufficiently similar for meta-analysis (see Appendix D for further details).

Figure 9: Direct meta-analysis – overall survival – CORRECT and CONCUR

==> picture [452 x 117] intentionally omitted <==

Key: CI, confidence interval; HR, hazard ratio; PBO, placebo; REG, regorafenib; seTE, standard error of treatment effect; TE, treatment effect.

Figure 10: Direct meta-analysis – progression-free survival – CORRECT and CONCUR

==> picture [446 x 111] intentionally omitted <==

Key: CI, confidence interval; HR, hazard ratio; PBO, placebo; REG, regorafenib; seTE, standard error of treatment effect; TE, treatment effect.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 65 of 178

B.2.9. Indirect and mixed treatment comparisons

In the absence of direct head-to-head evidence comparing the efficacy of regorafenib directly with that of trifluridine/tipiracil, indirect treatment comparisons (ITCs) were performed in line with NICE Decision Support Unit (DSU) technical support document (TSD) 2 to estimate relative treatment effects for OS and PFS.

Evidence for the efficacy and safety of regorafenib and trifluridine/tipiracil were identified through a clinical systematic literature review (SLR). Studies included in the ITCs were placebo controlled RCTs. Table 13 summarises the studies included in the ITCs and a summary of the excluded studies can be found in Appendix D (Section B.3.1.8). Data from the primary publications were used for both regorafenib and trifluridine/tipiracil to avoid bias from open-label follow-up. All studies included a best supportive care arm.

Table 13: Summary of the trials used to carry out the indirect treatment

comparisons

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 66 of 178

Assessment of study heterogeneity (see section B.2.9.1) identified some differences in study design, inclusion/exclusion criteria, and baseline characteristics between the five studies outlined in Table 13, however, the studies were considered to sufficiently satisfy the basic assumptions of homogeneity, similarity, and consistency as reported in Song et al. 2009.[54] Fixed effects NMAs were therefore performed using the network presented in Figure 11. A fixed effects NMA, rather than a random effects NMA, was deemed more suitable given the small number of studies included in the network. Full details of the methods are included in Appendix D (B.3.1.8.2)

Figure 11: Network diagram for OS and PFS

==> picture [401 x 209] intentionally omitted <==

Key: OS, overall survival; PFS, progression-free survival.

Table 14 presents the results from the fixed effects NMA of OS comparing regorafenib with trifluridine/tipiracil and placebo in the study ITT populations. Regorafenib was associated with a reduction in the hazard of death compared with placebo (HR: 0.68 [95% CrI: 0.59, 0.78]). This is consistent with the results observed in the clinical trials investigating regorafenib; CORRECT HR: 0.77 (95% CI: 0.66, 0.94) and CONCUR HR: 0.55 (95% CI: 0.39, 0.81). Trifluridine/tipiracil was also associated with a reduction in the hazard of death compared with placebo (HR: 0.68 [95% CrI: 0.62, 0.76]). Finally, results from the NMA suggested that the effectiveness of regorafenib and trifluridine/tipiracil for OS was similar, with a small but nonsignificant numerical advantage for regorafenib (HR was close to 1 and the CrI included 1).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 67 of 178

Table 14: Results of the fixed effects NMA of OS

Table 15 presents the results from the fixed effects NMA of PFS comparing regorafenib with trifluridine/tipiracil and placebo for the ITT population. Regorafenib was associated with a reduction in the hazard of progression compared with placebo (HR: 0.42 [95% CrI: 0.39, 0.45]). This is consistent with the results observed in the clinical trials; CORRECT HR: 0.49 (95% CI: 0.42, 0.58) and CONCUR HR: 0.31 (95% CI: 0.22, 0.44). Trifluridine/tipiracil was also associated with reduction in the hazard of progression compared with placebo (HR: 0.45 [95% CrI: 0.42, 0.48]). Finally, results from the NMA suggested that the effectiveness of regorafenib and trifluridine/tipiracil for PFS was similar (HR was close to 1 and the CI included 1).

Table 15: Results of the fixed effects NMA of PFS

To test the sensitivity of the fixed-effects NMAs including the five studies (CORRECT, CONCUR, RECOURSE, TERRA and Yoshino 2012), the analyses as outlined in Table 16 were performed to allow for differences between studies that were identified in the assessment of study heterogeneity.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 68 of 178

Table 16. ITC sensitivity analyses

 - Sensitivity analysis 1: Fixed effects NMA including CORRECT, CONCUR, RECOURSE and TERRA

Yoshino 2012 was removed from the network to explore whether the difference in phase of study impacted the results (CORRECT, CONCUR, RECOURSE and TERRA were all Phase III studies compared with Yoshino 2012 which was Phase II). Yoshino 2021 was also the only study to be conducted in one country only (i.e. in Japan).

 Sensitivity analysis 2: Bucher ITC including CORRECT and RECOURSE

All patients in CORRECT and RECOURSE had received a prior targeted biologic treatment (prior bevacizumab) compared with CONCUR, TERRA and Yoshino 2012 where around 40%, 50% and 20% of patients had received prior targeted biologic treatment, respectively. In the UK bevacizumab is not recommended. Prior targeted

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 69 of 178

biological treatment was identified as a potential treatment effect modifier by clinical experts. However, further sensitivity analyses adjusting for this variable (using matching-adjusted indirect treatment comparison methods) had very little impact on results (more information on identification of treatment effect modifiers and results from the matching-adjusted indirect treatment comparison is detailed in Appendix D [Section B.3.1.8.1]).

 - Sensitivity analysis 3: Fixed effects NMA including CONCUR, TERRA and Yoshino 2012

CORRECT and RECOURSE included patients from across the world (14% and 34% of patients were Asian, respectively), whereas CONCUR, TERRA and Yoshino 2012 only included Asian patients. Despite there being no evidence that ethnicity is prognostic or a treatment effect modifier this sensitivity analysis was conducted for completeness. In addition, some patients in each of these trials had not received prior treatment with anti-VEGF (anti-VEGF not recommended in the UK) and therefor these 3 trials are more representative of treatment in the UK setting.

 Sensitivity analysis 4: Bucher ITC including CONCUR and TERRA

The patient characteristics from TERRA and CONCUR are considered to be most similar to those seen in UK clinical practice in terms of patients being naïve to targeted biologic treatment (52.7% had no prior targeted therapy).

 Sensitivity analysis 5: Bucher ITC including CONCUR and Yoshino 2012 Conducted for completeness in respect of Asian-only trials.

Results for OS and PFS from these sensitivity analyses are summarized in Figure 12 and Figure 13, respectively. Overall, sensitivity analysis results for both OS and PFS were consistent with the primary analysis and indicated comparable efficacy.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 70 of 178

Figure 12: Overview of sensitivity analysis results for OS

==> picture [404 x 133] intentionally omitted <==

==> picture [404 x 133] intentionally omitted <==

Key: CI, confidence interval; Crl, credible interval; FE, fixed effects; ITC, indirect treatment comparison; NMA, network meta-analysis; OS, overall survival.

Figure 13: Overview of sensitivity analysis results for PFS

==> picture [399 x 133] intentionally omitted <==

==> picture [399 x 133] intentionally omitted <==

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 71 of 178

71 of 178

Key: CI, confidence interval; Crl, credible interval; FE, fixed effects; ITC, indirect treatment comparison; NMA, network meta-analysis; PFS, progression-free survival.

B.2.9.1. Uncertainties in the indirect and mixed treatment comparisons

The credible intervals for the comparisons of regorafenib with placebo and trifluridine/tipiracil with placebo from the NMA are slightly smaller than the CIs observed in the clinical trials. This is likely due to the fixed effects NMA

underestimating the uncertainty of the point estimates and is a limitation of this NMA. While a random effects NMA could address this slight underestimation, it was not deemed feasible to conduct this NMA given the small number of studies included in the analysis, which would likely lead to implausibly wide credible intervals when using a non-informative prior to estimate the between-trial heterogeneity.

Assessment of study heterogeneity identified some differences across the CORRECT, CONCUR, RECOURSE, TERRA and Yoshino 2012 studies. The main differences related to phase of study (phase II/III), ethnicity of patients (some studies were conducted in Asians only) and prior targeted biologic treatments (see Appendix D) . Only prior targeted biologic treatment was identified as a potential treatment effect modifier and differences in the distributions of patients with prior targeted biologic treatment across studies may have introduced some bias into the analyses. However, across sensitivity analyses (that [1] used studies with similar patients with/without prior targeted biologic treatment and that [2] matched on this variable using matching-adjusted indirect treatment comparison methods – Appendix D) results remained consistent indicating that the primary analysis results were robust despite the differences observed.

B.2.10. Supportive evidence

Ten observational/real-world studies, presenting evidence on the effectiveness of regorafenib in clinical practice, were identified in the clinical SLR (Appendix D). An overview of real-world evidence from these observational studies is presented in Table 17, with a brief overview of the findings in Section B.2.10.1.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 72 of 178

Table 17: List of observational studies (real-world evidence) of regorafenib

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 73 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 74 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 75 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 76 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 77 of 178

B.2.10.1. Real-world evidence

The real-world studies demonstrate efficacy outcomes consistent with those of CORRECT and CONCUR – but, notably, with numerically better median OS and PFS rates across the studies (Table 18). The safety data from these studies are also consistent with that of CORRECT and CONCUR (Section B.2.11).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 78 of 178

78 of 178

Table 18: Summary of efficacy outcomes in the real-world studies

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 79 of 178

B.2.11. Adverse reactions

B.2.11.1. CORRECT

B.2.11.1.1. Extent of exposure

Treatment compliance was high in both treatment groups: patients assigned regorafenib received ''''''''''''''' of the planned dose during the course of the study (mean daily dose 147.1 mg) compared with 90.1% for the placebo group (mean 159.2 mg).

B.2.11.1.2. Overview of adverse events

Table 19 presents an overview of the safety data up to the data cut-off date of 21 July 2011. Most patients in each group (regorafenib, 99.6%; placebo, 96.8%)

experienced at least one treatment-emergent AE (TEAE), the majority of which were mild or moderate events. SAEs were reported at a similar rate in both groups (regorafenib, 43.8%; placebo, 39.5%), and, while low overall, the incidence of treatment-emergent SAEs that were considered drug-related was slightly higher with regorafenib (11.8% versus 3.6%).

Although significantly more patients in the regorafenib arm had dose modifications because of AEs (66.6% versus 22.5%), the difference in the incidence of AEs leading to permanent treatment discontinuation was relatively small (17.6% versus 12.6%), indicating that most AEs could be managed by dose modification and did not result in a permanent discontinuation of study drug.

Table 19: CORRECT – overview of TEAEs (SAS)

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 80 of 178

80 of 178

==> picture [452 x 257] intentionally omitted <==

----- Start of picture text -----
Any AE, n (%) Drug-related AE, n (%)
Regorafenib Placebo Regorafenib Placebo
Event (n = 500) (n = 253) (n = 500) (n = 253)
CTC Grade 1 '''''' ''''''''''' ''''''' ''''''''''''' ''''''' '''''''''''' '''''' '''''''''''''''
CTC Grade 2 ''''''' '''''''''''''' ''''''' ''''''''''''''' '''''''' ''''''''''''' '''''' '''''''''''''
CTC Grade 3 ''''''''' '''''''''''''' ''''''' '''''''''''''''' ''''''''' ''''''''''''''' '''''' '''''''''''''''
CTC Grade 4 ''''''' '''''''''''' '''''' '''''''''' '''''' ''''''''''' ''' ''''''''''''
CTC Grade 5 ''''''' ''''''''''''''' ''''''' '''''''''''''' '''' ''''''''''' ''''
AE leading to '''''' ''''''''''''''' ''''''' ''''''''''''''' ''''''' ''''''''''' ''' '''''''''''
permanent
discontinuation
AE leading to ''''''''' '''''''''''''' ''''''' '''''''''''''' ''''''''' ''''''''''''' '''''' ''''''''''''
dose
modification
Key: AE, adverse event; CTC, Common Toxicity Criteria; SAE, serious adverse event; SAS, safety
analysis set; TEAE, treatment-emergent adverse event.
Source: Table 10-2 of CORRECT Clinical Study Report (Final). [45]
----- End of picture text -----

B.2.11.1.3. Treatment-emergent adverse events

Table 20 presents TEAEs that occurred in ≥ 5% of patients in either treatment group. The most frequent AEs of any grade were fatigue and hand–foot skin reaction (17% each) in the regorafenib group, and fatigue (28%) and anorexia (15%) in the placebo group. Grade 3 or 4 TEAEs occurred at a higher rate in the regorafenib group than in the placebo group (54% versus 14%). The most frequent Grade ≥ 3 regorafenibrelated AEs (affecting ≥ 5% of patients) were hand–foot skin reaction (17%), fatigue (< 10%), diarrhoea (< 8%), hypertension (7%), and rash or desquamation (6%) (Table 20).

Table 20: CORRECT – TEAEs occurring in ≥ 5% of patients in either group from

start of treatment to 30 days after end of treatment (SAS)

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 81 of 178

B.2.11.1.4. Serious adverse events and deaths

There was a similar incidence of SAEs between groups (regorafenib, 43.8%; placebo, 39.5%; Appendix F; Table 35. The differences regarding Grade 4 and 5 treatment-emergent SAEs between the two treatment groups was small and clinically not relevant. Grade 3 treatment-emergent SAEs were reported at a higher incidence in the regorafenib group than in placebo (18.2% versus 13.8%), although the incidences of Grade 3 and 4 SAEs were similar between groups.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 82 of 178

Overall, there were 110 deaths (regorafenib, 13.8% [n = 69]; placebo, 16.2% [n = 41]) reported during the study (i.e. up to within 30 days of last dose). The most common reason for death was PD (regorafenib, 11.6%; placebo, 13.8%); other reasons were reported as due to an AE not associated with clinical disease progression (1.6% versus 1.2%), unknown cause (0.4% versus 0.4%), and other cause (0.8% versus 0.2%). In the regorafenib group, the AEs not associated with disease progression that contributed to death were: pneumonia (n = 2), gastrointestinal bleeding (n = 2), intestinal obstruction (n = 1), pulmonary haemorrhage (n = 1), seizure (n = 1) and sudden death (n = 1). In the placebo group, these AEs were pneumonia (n = 2) and sudden death (n = 1). Occurrence of thromboembolism did not differ between groups (2% for both groups).

B.2.11.1.5. AEs of special interest

Liver dysfunction

The incidence of liver dysfunction was low in both groups (regorafenib, '''''''%; placebo, ''''''''%). More patients in the regorafenib group than placebo had AEs of liver dysfunction that resulted in fatal outcomes: '''''' patients versus ''''''''''' patients. Three events of liver dysfunction that had a fatal outcome in the regorafenib group were assessed by the treating Investigator as related to study drug; for two of these patients, the cause of death as stated by the Investigator was disease progression.

Cardiac ischaemia/infarction and bleeding events

The incidence of cardiac ischaemia/infarction was low in both groups (regorafenib, ''''''''%; placebo, '''''''''%). In most of the cases, patients had existing cardiovascular risk factors. In both groups, there was '''''''''' death reported, which was not assessed as drug-related. Regarding bleeding events, the incidence was higher in the regorafenib group than in the placebo group (regorafenib, ''''''''''%; placebo, ''''''''%); however, the majority of events were Grade 1 nose bleeds in the regorafenib arm. The incidence of serious bleeding was low in both groups (regorafenib, ''''''''%; placebo, ''''''''%). In total, ''''''''' ('''''''%) patients in the regorafenib group had a bleeding event resulting in death, while no deaths due to bleeding were reported in the placebo group.

Hand–foot skin reaction (palmar–plantar erythrodysesthesia)

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 83 of 178

The incidence of hand–foot skin reaction was higher in the regorafenib group than in the placebo group (regorafenib, '''''''''''%; placebo, '''''''%). This is expected, as hand– foot skin reaction is a known toxicity for patient treatment in this drug class. The majority of these events were Grade 1 or 2; however, ''''''''''% of patients in the regorafenib group had Grade 3 events, compared with ''''''''% in the placebo group. The incidence of drug-related hand–foot skin reaction events was notably higher in the regorafenib group compared with the placebo group ('''''''''''% versus '''''''%). The incidence of SAEs was low ('''''''% versus '''%, respectively). Permanent

discontinuation of treatment due to hand–foot skin reaction was low (regorafenib, ''''''''%; placebo, ''''%), with the majority of reactions being managed by dose reductions or interruptions.

Rash

The incidence of rash was higher in the regorafenib group than in the placebo group (regorafenib, ''''''''''%; placebo, ''''''''%). This is expected, as rash is a known toxicity for patient treatment in this drug class. The majority of these events were Grade 1 in both treatment groups. The incidence of Grade 3 events was low (regorafenib, ''''''''%; placebo, ''''''''%). Rash treatment-emergent AEs could usually be managed by dose reductions or interruptions, and these TEAEs led to permanent discontinuation of treatment in only '''''''' ('''''''%) regorafenib-treated patients and '''''' placebo-treated patients.

Renal failure

The incidence of renal failure was low in both groups (regorafenib, '''''''''%; placebo, '''''''''%). Most events were Grade 3, with ''''''''' Grade 4 AE (in the placebo group) and '''''''''' Grade 5 AE (in the regorafenib group) reported. TEAEs of renal failure that were assessed as related to treatment were reported at the same incidence in regorafenib-treated patients and placebo-treated patients (''''''''% in each group).

Proteinuria

The incidence of proteinuria was low in both groups (regorafenib, '''''''%; placebo, ''''''''%). Most events were Grade 1 or 2, with Grade 3 events reported for '''''''''% and ''''''''%, respectively, in each group. No Grade 4 or Grade 5 events were reported.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 84 of 178

84 of 178

B.2.11.1.6. AEs leading to discontinuations and dose modifications

Rates of permanent discontinuations due to TEAEs were low and similar in both groups (regorafenib, ''''''''''%; placebo, '''''''''''%). The incidence of Grade 3, Grade 4 and Grade 5 AEs leading to permanent discontinuation was similar between the two groups. Most toxicities could be managed by dose modifications.

Overall, more patients receiving regorafenib had AEs that led to dose medications than those on placebo (66.6% versus 22.5%). Of these, dose reductions occurred in 38% versus 3%, respectively, and dose interruption in 61% versus 22%, respectively. The most frequent AEs necessitating dose modification were dermatological, gastrointestinal, constitutional, and metabolic or laboratory events.

B.2.11.1.7. Additional safety data

An addendum to the CORRECT Final CSR was released on 29 August 2014. This addendum presented the cumulative safety results as of the final lock date of the clinical database (27 March 2014), including all data collected through the database cut-off date of 21 July 2011 (as reported in the CORRECT Final CSR), together with the additional safety data collected until all patients were off active treatment (last patient visit on 22 January 2014).

Overall, the results of the addendum were similar (with only slight or no changes) to those reported at the primary efficacy cut-off. In the safety update, there was a similar overall incidence in TEAEs as well as treatment-emergent SAEs for regorafenib versus placebo (Appendix F; Section B.5.1.2). Rates of permanent discontinuations and dose modifications remained low, and no new death events were reported.

B.2.11.1.8. Summary of AEs

Overall, no new safety signals were observed with regorafenib. There was a comparable incidence of TEAEs between groups, and the majority of patients in both groups had ≥ TEAE (regorafenib, 99.6%; placebo, 96.8%). The incidence of SAEs was also comparable between the two treatment groups (regorafenib, 43.8%; placebo, 39.5%). The incidence of TEAEs considered related to study treatment were higher with regorafenib than placebo, but this was as expected for its drug class. The rate of permanent discontinuations due to AEs was low, indicating that

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 85 of 178

most toxicities could be managed by dose modifications. Overall, there were 110 deaths (regorafenib, n = 69 [13.8%]; placebo, n = 41 [16.2%]) reported during the study. The majority of these cases were due to progression of underlying disease. The additional safety data did not reveal clinically significant changes from the initial data analysis, and no new safety concerns were identified.

In summary, the data suggest that regorafenib was well tolerated in CORRECT. The AEs associated with regorafenib were recognizable and manageable.

B.2.11.2. CONCUR

B.2.11.2.1. Extent of exposure

Treatment compliance was high in both treatment groups: patients assigned regorafenib received '''''''''''''' of the planned dose during the course of the study (mean daily dose: 145.4 mg) compared with '''''''''''% for the placebo group (mean daily dose: 160.0 mg).

B.2.11.2.2. Overview of adverse events

Table 21 presents an overview of the safety data up to the data cut-off date of 29 November 2013. Most patients in each group (regorafenib, 100%; placebo, 88.2%) experienced at least one TEAE, the majority of which were mild or moderate events. Of these events, 97.1% and 45.6% in each respective treatment group were

considered to be drug-related. SAEs were reported at a similar rate in both groups (regorafenib, 31.6%; placebo, 26.5%), and, while low overall, the incidence of treatment-emergent SAEs that were considered drug-related was higher with regorafenib (8.8% versus 3.4%).

Although more patients in the regorafenib arm had dose modifications because of AEs (71.3% versus 16.2%), the incidence of AEs leading to permanent treatment discontinuation was relatively small (14.0% versus 5.9%), indicating that most AEs could be managed by dose modification and did not result in a permanent discontinuation of study drug.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 86 of 178

Table 21: CONCUR – overview of TEAEs (SAS)

B.2.11.2.3. Treatment-emergent adverse events

Table 22 presents TEAEs that occurred in ≥ 10% of patients in either treatment group. The most frequent AEs of Grade ≥ 3 associated with regorafenib were hand– foot skin reaction (22 [16%]), hypertension (15 [11%]), hyperbilirubinaemia,

hypophosphataemia, and alanine aminotransferase concentration increases (nine [7%] each).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 87 of 178

Table 22: CONCUR – TEAEs occurring at any grade in ≥ 10% of patients, or at Grade ≥ 3 in any patients in either group, from the start of treatment to 30 days after the end of treatment (SAS)

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 88 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 89 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 90 of 178

B.2.11.2.4. Serious adverse events and deaths

There was a similar incidence of SAEs between groups (regorafenib, 31.6%; placebo, 26.5%; Appendix F; Table 44). The differences regarding Grade 4 and 5 treatment-emergent SAEs between the two treatment groups was small and clinically not relevant. Grade 3 treatment-emergent SAEs were reported at the same incidence in both groups (11.8%). Most treatment-emergent SAEs were not related to study drug treatment.

Overall, there were 19 deaths reported during the study (regorafenib, n = 12 [8.8%]; placebo, n = 7 [8.8%]). One additional patient in the placebo group died during the follow-up period. Fourteen of these cases were due to progression of underlying disease (regorafenib, n = 8; placebo, n = 6). In the regorafenib group, the deaths of two (1%) patients were deemed to be drug-related within 30 days after the last dose.[26] Brief narratives of the two patients are as follows:

The first patient was a 65-year-old woman who stopped regorafenib treatment during her first cycle as a result of a non-serious Grade 2 increase in bilirubin. One week after stopping treatment, she collapsed at home and had a cardiac arrest

The second patient was a 67-year-old man who received regorafenib for 2 days. On the next day, he had a Grade 4 cardiac arrest, resulting in admission to hospital and death[26]

B.2.11.2.5. AEs of special interest Acute liver failure

There were no reports of hepatic failure, hepatic necrosis or Grade 2–4 drug-related hepatobiliary disorder AEs in either treatment group. Overall, treatment-emergent hepatobiliary/pancreas SAEs (any grade) were reported with similar incidence in both groups (regorafenib, '''''''%; placebo, ''''''''%). In the regorafenib group, there was one Grade 4 AE of bile duct stenosis that led to permanent study discontinuation. There were no deaths in either treatment group that resulted from liver dysfunction. No cases of significant transaminase increase or severe drug-induced liver injury have been identified from the ongoing hepatotoxicity monitoring from this study.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 91 of 178

Acute cardiac failure and Grade ≥ 3 bleeding events

There was one report of heart failure in the placebo group and one report of cardiac arrest in the regorafenib group. There were no reports of cardiac ischaemia/infarction-related AEs. There was a low incidence of cardiac disorders reported in each group (regorafenib, ''''''''%; placebo, '''''''%).

Regarding Grade ≥ 3 bleeding events, the overall incidence was higher in the regorafenib group compared with the placebo group (''''''''''% versus ''''''''%); however, the majority of the bleeding events in the regorafenib group were Grade 1 or 2 anaemia. Serious bleeding AEs were only reported in the regorafenib group ('''''''' events). No Grade 5 bleeding events were reported, and there were ''''' haemorrhage/bleeding events that were the cause of permanent discontinuation of study medication in either treatment group.

Hand–foot skin reaction (palmar–plantar erythrodysesthesia)

The incidence of hand–foot skin reaction was higher in the regorafenib group than placebo (regorafenib, ''''''''''%; placebo, ''''''''%). This is expected, as hand–foot skin reaction is a known toxicity for patient treatment in this drug class. The majority of these events were Grade 1 or 2, although Grade 3 events were reported in the regorafenib group ('''''''''''% versus '''%, respectively). There were '''''' SAEs of hand– foot skin reaction in either treatment group. These events led to the permanent discontinuation of treatment in only one (0.7%) regorafenib-treated patient and no placebo-treated patients.

Rash

The incidence of rash was ''''''''''''''' ''''' '''''''' '''''''''''''''''''''''''' ''''''''''''''' ''''''''''' '''''''''''''''''

(''''''''''''''''''''''''''''''' ''''''''''''''''''''' '''''''''''''' ''''''''''''''''''''''''''''''''''' ''''''''''''''''; placebo, ''''''''''''''''''''''' '''''''''''''

'''''''''''''''''''''''''''''''''''''' ''''''''''''). This is expected, as rash is a known toxicity for patient

treatment in this drug class. The majority of events were Grade 1. No TEAEs related to rash led to permanent discontinuation in either treatment group. SAEs of rash were low (regorafenib, ''''''''%; placebo, '''%), and all were reported as Grade 3 events and were considered related to the study drug.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 92 of 178

Acute renal failure (any grade) or severe proteinuria (Grade ≥ 3)

The incidence of proteinuria was ''''''''''''''' ''''' ''''''' ''''''''''''''''''''''''' ''''''''''''''' '''''''''' ''''' '''''''' '''''''''''''''''

''''''''''''' ''''''''''''''''''''''''''''' ''''''''''''''''' ''''''''''''''''''''' ''''''''''''''' ''''''''''''' ''''' ''''''''''''' ''''''''''''''''' ''''''''''''' '''''''''''''' ''' '''' '''' ''''''''''''''' ''''''''''' '''''''''''''''''''' '''''''''''''''''' ''''''''''''''' ''''''' '''''''''' ''''''''''''''''''' '''''''' ''''''''''''''' '''' ''''' ''''''''''''' ''''

'''''''''''''''' '''''''''''' '''''''''''''''''''''.

B.2.11.2.6. AEs leading to dose modifications

Rates of permanent discontinuations due to TEAEs were low (regorafenib, 14.0%; placebo, 5.9%). The incidence of Grade 3 and 4 AEs leading to permanent discontinuation were comparable between groups. There were no Grade 5 AEs leading to permanent discontinuation in either treatment group, and most toxicities could be managed by dose modifications.

Overall, more patients receiving regorafenib had AEs that led to dose modifications than those on placebo (75.0% versus 22.1%). Of these, dose reductions occurred in 39.7% and 0% of patients, and dose interruptions occurred in 62.5% and 16.2% of patients, respectively.

B.2.11.2.7. Additional safety data

An addendum to the CONCUR final CSR was released on 21 March 2017. This addendum presents the cumulative updated safety results, including all data collected as of 29 November 2013 together with the additional safety data collected until all patients were off active treatment (database cut-off date of 14 January 2016). This CSR addendum includes the safety results for all patients, including the six patients (all in the regorafenib group) who were still receiving treatment as of the 29 November 2013 database cut-off date.

At the safety data cut-off date all 204 patients in the study had terminated treatment; the six regorafenib patients who continued treatment past 29 November 2013 discontinued treatment by this point due to disease progression based on radiological evaluation. Since the primary analysis the additional available safety data did not reveal clinically significant changes from the initial data analysis; no new safety concerns were identified from the data updated since the initial analysis (Appendix F; Section B.5.2.2).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 93 of 178

B.2.11.2.8. Summary of AEs

Overall, no new safety signals were observed with regorafenib. There was a comparable incidence of TEAEs between groups, and the majority of patients in both groups had one of more TEAEs (regorafenib, 100%; placebo, 88.2%). The incidence of SAEs was also comparable between the two treatment groups (regorafenib, 31.6%; placebo, 26.5%). The incidence of TEAEs considered related to study treatment was higher with regorafenib than with placebo, but this was as expected for its drug class. The rate of permanent discontinuations due to AEs was low, indicating that most toxicities could be managed by dose modifications. Overall, there were 19 deaths (regorafenib, n = 12 [8.8%]; placebo, n = 7 [8.8%]) reported during the study. One additional patient in the placebo group died during the followup period. The majority of these deaths were due to progression of underlying disease. The additional safety data from the updated data cut (14 January 2016) did not reveal clinically significant changes from the initial data analysis, and no new safety concerns were identified.

In summary, the data suggest that regorafenib was well tolerated. The AEs associated with regorafenib were recognizable and manageable. The safety profile overall was similar to that observed in earlier Phase I and II studies as well as the Phase III CORRECT study after which the study design for this study was modelled.

B.2.12. Ongoing studies

Both CORRECT and CONCUR are completed trials. No other RCTs of regorafenib in ≥ 3L mCRC are reading out in the next 12 months.

B.2.13. Interpretation of clinical effectiveness and safety evidence

B.2.13.1. Principal findings from the clinical evidence

Data from both the CORRECT and CONCUR trials demonstrated consistent clinically meaningful survival outcomes with regorafenib over placebo in ≥ 3L mCRC. Significantly prolonged OS (CORRECT, HR 0.77 [95% CI: 0.64, 0.94]; CONCUR, HR 0.55 [95% CI: 0.40, 0.77]; primary endpoint met for both trials) and PFS (CORRECT, HR 0.49 [95% CI: 0.42, 0.58]; CONCUR, HR 0.31 [95% CI: 0.22, 0.44]) were demonstrated for both trials. Subgroup analyses of OS and PFS were

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 94 of 178

94 of 178

consistent with these findings. While the ORR was similar between arms in both trials (CORRECT: p = 0.19; CONCUR, p = 0.045), regorafenib was associated with significant disease control compared with placebo (p < 0.0001 for both trials). Furthermore, regorafenib appears to improve QoL, with patients more likely to stay on treatment when compared with placebo (when assessed by EQ-5D-3L).[38] Indeed, this finding in QoL enhancement is in line with that reported in the literature for regorafenib, where it has been demonstrated that, compared with placebo, regorafenib delays the time until definitive deterioration in HRQL for multiple tumour types (including mCRC using data from CORRECT and CONCUR).[65]

The safety profile of regorafenib in CORRECT and CONCUR was both recognizable and manageable, and no new safety signals were reported. The overall incidence of TEAEs was comparable between the treatment arms in both trials (CORRECT, 99.6% versus 96.8%; CONCUR, 100% versus 88.2%). Dose modifications owing to AEs were higher with regorafenib compared with placebo (CORRECT, 67% versus 23%; CONCUR, 71% versus 16%); however, permanent discontinuations due to TEAEs were relatively small (CORRECT, 17.6% versus 12.6%; CONCUR, 14.0% versus 5.9%). Grade 3 TEAEs events were higher in the regorafenib arm than in the placebo arm (CORRECT, 56.0% versus 26.5%; CONCUR, 52.9% versus 32.4%). In both CORRECT and CONCUR, hand–foot skin reaction and hypertension were common regorafenib-related Grade ≥ 3 TEAEs. A total of 110 (regorafenib, n = 69 [13.8%]; placebo, n = 41 [16.2%]) and 18 (regorafenib, n = 12 [8.8%]; placebo, n = 6 [8.8%]) deaths were reported in CORRECT and CONCUR, respectively. The majority of these cases were due to progression of underlying disease.

Results from network meta-analyses of OS and PFS show similar efficacy between regorafenib and trifluridine/tipiracil. For OS, the hazard ratio (HR) was 0.99 (95% CrI: 0.84, 1.17) and for PFS the HR was 0.93 (95% CrI: 0.85, 1.03). Sensitivity analyses on the NMA confirmed efficacy was similar between regorafenib and trifluridine/tipiracil for OS and PFS. In addition, a MAIC comparison confirmed the results of the standard NMA (Appendix B.3.1.8.2).

B.2.13.2. Strengths and limitations of the evidence base

The CORRECT and CONCUR trials demonstrate the clinical benefit of regorafenib in a group of patients with limited options - patients with mCRC who have had two or

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 95 of 178

more prior therapies. While the reported survival outcomes were modest over placebo (BSC), they were nevertheless clinically meaningful findings in the ≥ 3L setting in mCRC, a population of patients where survival prognosis is very poor and there is strong clinical consensus of there being a strong clinical need for additional treatment options.

Both CORRECT and CONCUR were relatively large studies (n = 760 and n = 204, respectively) that were methodologically robust, enabling them to read-out highquality clinical trial data. Furthermore, data from the two trials were consistent with each other, highlighting the efficacy of regorafenib across patients of different ethnicities. In addition, both trials reported HRQL data using validated instruments. This is a notable strength of the evidence base, as QoL assessments for active treatment in ≥ 3L mCRC are not commonly reported in the literature and were not reported for trifluridine/tipiracil in its NICE technology assessment. For regorafenib, evaluation of HRQL demonstrated that it was not associated with impaired HRQL over that of placebo (BSC) but rather that it likely has a positive effect on QoL, as well as on the length of survival.[38]

The evidence base is strengthened by the availability of several observational, realworld studies of regorafenib (Section B.2.10). These trials demonstrate efficacy and safety findings that are consistent with CORRECT and CONCUR.

A limitation of the evidence base is that there are no head-to-head data for regorafenib versus trifluridine/tipiracil, the key comparator in the scope of this appraisal. The CORRECT and CONCUR trials were completed before trifluridine/tipiracil regulatory approval and, therefore, did not include trifluridine/tipiracil as a comparator. To address this limitation, a meta-analysis and ITC were conducted, the findings of which demonstrate that both regorafenib and trifluridine/tipiracil have comparable efficacy (Section B.2.8 and B.2.9).

B.2.13.3. Applicability of clinical evidence to practice

As discussed in Section B.2.5, in general, the trial populations of CORRECT and CONCUR are reflective of patients with ≥ 3L mCRC in England and Wales. Furthermore, as ethnicity is not a recognized prognostic factor in mCRC, there is clinical consensus that of the two trials, the CONCUR cohort (specifically those with

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 96 of 178

96 of 178

no prior anti-VEGF therapy), despite being conducted solely in Asian patients, is more generalizable to the UK mCRC population. As UK patients will not be as highly pre-treated as the population of CORRECT (given that several targeted therapies, including bevacizumab, are not licensed in the UK but were used in CORRECT), it is expected that their survival outcomes will be slightly better and more closely aligned with those reported in CONCUR.

Regorafenib is an oral therapy that is not associated with any NHS clinical service changes or with any companion diagnostic requirements (Section B.2.3). This means that its approval would introduce a new chemotherapy-free therapy option in ≥ 3L mCRC without additional burden on the patient or NHS (oral medicines are associated with reduced exposure to healthcare facilities).

The primary treatment goal in mCRC is to prolong survival while maintaining QoL - incorporating a chemo-free alternative treatment option into the clinical management pathway will help achieve this goal.[66] In particular, treatments with different adverse event profiles provide valuable options where particular AEs have been troublesome in the past. This means that introducing regorafenib in ≥ 3L mCRC will help ensure patients gain the maximum benefit available to them from active treatment while their performance status is still good.

In conclusion, regorafenib offers a treatment with a different mode of action compared to the chemotherapy tipiracil/trifluridine which is the only alternative in the ≥ 3L mCRC setting. Regorafenib provides a chemotherapy-free treatment option that has the potential to make a substantial impact on patient health-related benefits, as it has a different AE profile compared to chemotherapy and provides meaningful survival outcomes[25, 26]

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 97 of 178

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

Cost effectiveness

B.3.1. Published cost-effectiveness studies

• In appendix G, describe and compare the methods and results of any published cost-effectiveness analyses available for the technology and/or the comparator technologies (relevant to the technology evaluation).

• See section 3.1 of the user guide for full details of the information required in appendix G.

An SLR was conducted to identify and evaluate the available economic evidence in ≥ 3L mCRC. Full details of the SLR methodology are presented in Appendix G. In total 23 cost-effectiveness and 3 budget impact studies, were identified (see Appendix G). All 23 identified cost-effectiveness studies for treatments used in ≥ 3L mCRC, along with a short overview of the approach, data input, and outcomes are shown in Table 23.

Of the identified studies, several studies investigated the cost-effectiveness of trifluridine/tipiracil in previously treated mCRC. These are also the most relevant studies for this submission as they are in line with the decision problem of this appraisal, in terms of comparator and indication. However, none of these studies provide reliable cost-effectiveness results to be used in this appraisal, for reasons detailed below.

Firstly, none of the studies used efficacy data from the 5 trials available for regorafenib and trifluridine/tipiracil in the cost-effectiveness analysis. More specifically, none of the identified studies used data from both CORRECT and CONCUR to inform regorafenib’s efficacy, except for Giuliani (2021). [67] However, this study only performed a simple cost calculation, based on median OS and treatment duration, which is too simplistic to accurately assess cost-effectiveness. Similarly, none of the identified studies used the full set of available trifluridine/tipiracil data (i.e. RECOURSE, TERRA, and Yoshino (2012)). Consequently, none of the identified cost-effectiveness studies provide a reliable efficacy comparison using all available regorafenib and trifluridine/tipiracil effectiveness data.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 98 of 178

Secondly, none of the identified studies use the appropriate cost data for this decision problem. This stems from the fact that all studies use the regorafenib list price without patient access scheme (PAS) discount for the analysis. Consequently, all identified studies will over-estimate the regorafenib costs, from an NHS perspective. In addition, most of the identified studies do not have a UK perspective, resulting in differences in treatment costs, healthcare resource use (HRU) costs, and AE costs used in these models, further limiting the accuracy of the cost results.

As none of the cost effectiveness studies provide reliable outcomes that are appropriate for the current decision problem, a de-novo model will be required to address the decision problem for this submission.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 99 of 178

Table 23: Summary list of published cost-effectiveness studies

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 100 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 101 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 102 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 103 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 104 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 105 of 178

B.3.2. Economic analysis

As discussed in Section B.3.1 none of the identified cost-effectiveness studies appropriately address the current decision problem as 1) none use data from each of the 5 RCTs identified in the SLR, and 2) none uses the regorafenib PAS price. As a result, a de novo cost-effectiveness analysis tailored to the current decision problem was performed.

B.3.2.1. Patient population

In line with the restricted submission and the CORRECT and CONCUR trials[45, 48] , the model features adult patients with mCRC who have been previously treated with, or are not considered candidates for, available therapies (i.e. fluoropyrimidine-based chemotherapy, anti-VEGF therapy and anti-EGFR therapy).

B.3.2.2. Model structure

A de novo partitioned survival (area under the curve) model was constructed in Microsoft Excel[®] . A partitioned survival model was chosen because it allows for a straightforward unbiased estimation of costs, life years (LYs) and quality-adjusted life years (QALYs). Unlike a Markov model, which also uses health states, partitioned survival models do not require the estimation and use of transition probabilities. Instead, the number of patients in each health state is calculated directly from each (extrapolated) OS and PFS curves.

The model consists of the following mutually exclusive health states (see Figure 14):

Progression-free : a patient’s disease is stable or responding, and not actively progressing. Costs in this health state are associated with treatment, administration and management of disease and AEs. QoL is higher compared with patients with progressed disease

Progressed : a patient’s disease is assumed to have progressed. Costs in this health state are only associated with management of disease as the model assumes no active treatment is given after progression. Patients have a lower QoL than in the progression-free state.

Death : an absorbing health state

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 106 of 178

Model health states were selected in accordance with the clinical pathway of care and are typical of modelling in advanced oncology. This structure has also been used in previous later-line mCRC NICE technology appraisals (TA242[88] , TA307[89] , TA405[9] , TA668[8] ). This structure is identical for all patients, regardless of their treatment, as it is based on disease progression rather than treatment. The possible routes patients may flow through the model are presented in Figure 14, below.

Figure 14: Model structure

==> picture [349 x 190] intentionally omitted <==

The likelihood of patients transitioning between the health states is determined via PFS and OS curves that were fitted to the clinical trial data. All patients enter the model in the progression-free health state and remain in this state until disease progression or death. The progression-free health state is designed to capture the benefits from an active treatment whilst the disease is controlled prior to progression, leading to relatively higher QoL. Following progression, patients are unable to transition back from the progressed health state to the progression-free health state. The progressed disease state is designed to capture the relatively poor QoL following disease progression and prior to death. Hence, the model captures the changes in QoL between patients who are progression-free and progressed. Patients can transition to the death health state from any other health state.

The model evaluates the cost-effectiveness of regorafenib in mCRC from a UK NHS and Personal Social Services (PSS) perspective. The model uses a weekly cycle length. This is in contrast to some previous mCRC models that used cycle lengths ranging from 1 day (TA405[9] ) to 1 month (TA242[88] , TA668[8] ). A weekly cycle length

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 107 of 178

was chosen to provide a good balance between more granularity but without overcomplicating the model. The time horizon of the model is 10 years, which corresponds to lifetime as < 1% of patients are alive after 10 years. Costs and QALYs are discounted with an annual rate of 3.5%, in line with the NICE reference case.[90] Half-cycle correction was also applied, to correct for mid-cycle progressions.

Table 24 below summarizes the different elements of the model structure and compares these to past appraisals in later-line mCRC. Not all previous appraisals are equally representative, as they include appraisals in ≥ 2L, ≥ 3L, and mutationspecific later-line mCRC. The only past appraisal in the same ≥ 3L mCRC population as regorafenib is TA405, which assesses trifluridine/tipiracil in ≥ 3L mCRC. TA405 is therefore the most relevant source of comparison and validation for this appraisal.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 108 of 178

Table 24: Features of the economic analysis

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 109 of 178

B.3.2.3. Intervention technology and comparators

The economic model allows the costs and efficacy of regorafenib to be compared with those of trifluridine/tipiracil and BSC.

A full set of economic analyses is presented against trifluridine/tipiracil as this is the main comparator and in keeping with the requested position i.e. we are seeking a similar position to trifluridine/tipiracil in the ≥3[rd] line setting. As stated in section B.1.1, Bayer is making this submission on the request of physicians who are seeking another treatment option alongside trifluridine/tipiracil.

A limited set of analyses is presented against BSC, mainly because BSC was the comparator in the pivotal trials for regorafenib, rather than it being considered directly relevant to the requested position. Treatment practices have changed since the completion of the pivotal trials and trifluridine/tipiracil is now used at the third-line or later setting. The model includes the same functionality for comparing against trifluridine/tipiracil or BSC.

Regorafenib is implemented in the model for patients with mCRC at a recommended dose of 160 mg (4 x 40mg tablets) once daily for 3 weeks followed by one week off therapy. This is reflective of the decision problem described in Section B.1.1. Regorafenib efficacy is informed by the two registration Phase III trials: the global CORRECT trial[25] and the CONCUR study.[26] Further details of these trials are discussed in Section B.3.3.

All consulted clinical experts agreed that trifluridine/tipiracil would be the key comparator for regorafenib in mCRC. Trifluridine/tipiracil is an orally administered combination of trifluridine, a thymidine-based nucleic acid analogue, and tipiracil hydrochloride, a thymidine phosphorylase inhibitor.[91] Trifluridine/tipiracil is administered at a dose of 35 mg/m[2] twice daily, 5 days a week, with 2 days of rest, for 2 weeks, followed by a 14-day rest period. This treatment cycle is repeated every 4 weeks.[91] Trifluridine/tipiracil efficacy is informed by the registration Phase III trials RECOURSE[51] and TERRA[52] , and the Japanese Phase II trial Yoshino (2012).[53] In the model, trifluridine/tipiracil is implemented as a comparator via an indirect comparison (see Section B.3.3).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 110 of 178

As per the trial protocols and licence in mCRC, treatment with regorafenib and trifluridine/tipiracil is continued until the determination of RECIST-defined disease progression, clinical progression, the development of severe AEs, withdrawal from the study, death, or a decision by the treating physician that discontinuation would be in the patient’s best interest.[91, 92] To accurately model treatment duration as observed in the CORRECT and CONCUR trials, pooled time on treatment (ToT) data are used to calculate the total active drug costs. Further details on regorafenib and trifluridine/tipiracil drug cost calculation are discussed in Section B.3.5.1.

The model also compares regorafenib against BSC alone (reduced analysis set): the comparator in both the CORRECT and CONCUR studies. BSC in CORRECT and CONCUR included any concomitant medications or treatments: antibiotics, analgesics, radiation therapy for pain control (limited to bone metastases), corticosteroids, transfusions, psychotherapy, growth factors, palliative surgery, and any other symptomatic therapy necessary to provide BSC. Other investigational antitumour agents or antineoplastic chemotherapies / hormonal therapies / immunotherapies were not included in BSC.

B.3.3. Clinical parameters and variables

Efficacy data for regorafenib in mCRC are available from two Phase III trials:

• 1. The CORRECT study (n = 760)
• 2. The CONCUR study (n = 204)

B.3.3.1. Data pooling

To maximize the data informing the model and to accurately represent the available evidence, data from CORRECT and CONCUR were pooled. However, neither trial is 100% generalisable to the UK setting – the two main differences are described below.

1) Prior anti-VEGF use

All patients in CORRECT received prior therapy with an anti-VEGF agent (i.e. bevacizumab), compared with 41.2% of patients receiving regorafenib and 36.8% of

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 111 of 178

patients receiving placebo in CONCUR. As anti-VEGF is not recommended by NICE, the CONCUR trial is more reflective of UK clinical practice.

Prior exposure to anti-VEGF therapy may reduce the treatment effect associated with regorafenib over BSC, as regorafenib is a multi-kinase inhibitor with targets that include VEGF.[92] This possibility is supported by the higher point estimate HRs for patients who received prior targeted therapy in CONCUR compared with patients who were naïve to targeted treatment. Similarly, it was supported by nine UK clinical experts who attended a recent advisory board meeting. As a result of this effect, and the absence of anti-VEGF therapy in UK clinical practice, it is possible that the CORRECT study would underestimate the efficacy of regorafenib in the UK population compared with CONCUR.

2) Ethnicity

All patients in CONCUR were Asian which makes CONCUR less representative of the ethnic mix who would receive treatment in the UK. However, none of the consulted clinical experts considered ethnicity to be a treatment effect modifier. This conclusion is supported by the CORRECT study, which reported comparable OS HRs in Asian (0.79) and non-Asian (0.76) groups. CONCUR therefore likely provides efficacy estimates generalizable to the UK population.

Appropriateness of using pooled data

We consider that pooling both trials represents the most robust evidence base for decision-making. Using pooled CORRECT and CONCUR data is also in line with approach in the appraisal of trifluridine/tipiracil (TA405).[9] In TA405 the Committee preferred an economic model that pooled the global RECOURSE study and Yoshino study, as opposed to alternatives based on the single studies themselves despite similar differences in ethnicity and prior treatment as present between CONCUR and CORRECT as described above.

Pooled CORRECT and CONCUR data were therefore used to inform the efficacy in the base case. Where pooled data were not readily available (e.g. relative dose intensity [RDI] data or AE probabilities), a weighted average of data from CORRECT

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 112 of 178

and CONCUR was used to ensure the model consistently uses pooled data for all inputs.

In summation, the following data were used in the base case:

• PFS was modelled using pooled data from CORRECT and CONCUR

• OS was modelled using pooled data from CORRECT and CONCUR

• ToT (time on treatment) was modelled using pooled data from CORRECT and CONCUR (for drug costing)

• RDI was modelled using a weighted average of data from CORRECT and CONCUR (for drug costing)

• Utility data was modelled using pooled EQ-5D data from CORRECT and CONCUR

• AE probabilities were modelled using a weighted average of data from CORRECT and CONCUR

B.3.3.2. Data extrapolation

Clinical data for time-to-event outcomes were extrapolated to facilitate modelling of survival over a lifetime horizon (i.e. 10 years). Treatment effects were modelled by extrapolating patient-level data from the primary cut-off for each arm separately (in line with NICE Decision Support Unit Technical Support Document 14[93] ). The proportional hazards assumption holds for all endpoints, as shown in Appendix O.

For each of the above-mentioned outcomes (PFS, OS, ToT), seven standard parametric models (i.e. exponential, Weibull, log-logistic, log-normal, generalized gamma, gamma and Gompertz) were fitted for each treatment group. To determine the best model fit, the following steps were undertaken:

• Akaike information criterion (AIC) and Bayesian information criterion (BIC) – model fits were evaluated using AIC and BIC statistics. Lower AIC and BIC figures are indicative of a better statistical fit of the survival curve

• Visual inspection – visual inspection was carried out by plotting the projected survival curves overlaid with the Kaplan–Meier survival functions.

• Clinical validity – the clinical plausibility of the extrapolated outcomes was assessed using expert opinion

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 113 of 178

Inclusion and extrapolation of each clinical outcome is detailed in the following sections.

B.3.3.2.1. Progression-free survival

As discussed in Section B.2.6, PFS in both CONCUR and CORRECT was very mature: 85.1% and 94.5% of regorafenib and BSC patients in CORRECT had experienced a PFS event at the time of primary efficacy cut-off, as had 88.2% and 95.6% of regorafenib and BSC patients, respectively, in CONCUR.[45, 48] Similarly, PFS Kaplan–Meier data was mature, with pooled PFS data available until 2.33% of regorafenib patients and 0.69% of BSC patients are progression-free. The KM data are used directly in the model after which an exponential extrapolation informs the remainder of the model (as the exponential survival estimates are closest to the Kaplan–Meier data at 9.2 and 16.1 months).

The fully parametric survival curves for PFS are shown in appendix N. A scenario using a fully parametric PFS approach is explored in sensitivity analysis.

B.3.3.2.2. Overall survival

Pooled OS data were extrapolated by applying seven standard parametric models (i.e. exponential, Weibull, log-logistic, log-normal, generalized gamma, gamma and Gompertz) to the pooled Kaplan–Meier data. Log-logistic was the model with the best statistical fit for regorafenib, whereas the log-normal curve showed the best statistical fit for BSC (see Table 25, below). Figure 15 to Figure 18 compare the pooled OS Kaplan–Meier data with different extrapolated curves for regorafenib and BSC, respectively. For BSC, the log-normal, log-logistic, and generalized gamma curves each performed well in terms of both visual inspection and statistical fit (see Table 25 and Figure 17). For regorafenib, visual inspection suggests the log-normal distribution fits the Kaplan–Meier data best, particularly in the tail end of the curve, followed by the log-logistic and generalized gamma (see Figure 15).

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 114 of 178

Table 25: Goodness-of-fit statistics of the pooled overall survival extrapolations

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved

115 of 178

Figure 15: Short-term parametric fits of regorafenib overall survival extrapolations

==> picture [699 x 300] intentionally omitted <==

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

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 116 of 178

Figure 16: Long-term parametric fits of regorafenib overall survival extrapolations

==> picture [715 x 307] intentionally omitted <==

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

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved

117 of 178

Figure 17: Short-term parametric fits of BSC overall survival extrapolations

==> picture [717 x 307] intentionally omitted <==

Key: BSC, best supportive care; KM, Kaplan–Meier; OS, overall survival.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved

118 of 178

Figure 18: Long-term parametric fits of BSC overall survival extrapolations

==> picture [721 x 309] intentionally omitted <==

Key: BSC, best supportive care; KM, Kaplan–Meier; OS, overall survival.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved

119 of 178

The validity of the long-term OS estimates was discussed during a clinical advisory board. Considering the small variation in OS estimates, especially for the best fitting OS curves (log-logistic, log-normal, and generalized gamma), the experts felt a case could be made for all curves. The experts considered that consistency with TA405 was important, as trifluridine/tipiracil was the key comparator. They therefore recommended log-logistic to be used in the base-case, in line with the committee and ERG preference in TA405. This approach was confirmed by the consulted health economic expert, who also agreed that log-logistic shows the best statistical fit for regorafenib.

The trifluridine/tipiracil OS results from our model were compared with the trifluridine/tipiracil results reported in TA405.[9] Median OS results were similar across OS distributions and in line with TA405 (see Table 26). For mean OS, the different extrapolations resulted in bigger differences in predicted values, with the log-logistic and log-normal extrapolations predicting mean OS most in line with TA405 ('''''''''' versus ''''''''''' months; Table 26). These data further confirm that the log-logistic and log-normal extrapolations provide the most appropriate OS input for the model. The base-case therefore applies a log-logistic OS extrapolation, with other curves explored in scenarios.

Table 26: Modelled trifluridine/tipiracil outcomes compared to reported

outcomes in TA405

B.3.3.2.3. Time on treatment

As per the licensed indication, patients treated with regorafenib are expected to be treated until disease progression or unacceptable toxicity. Treatment duration can therefore differ slightly from PFS due to early discontinuations caused by AEs and

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved

120 of 178

other reasons for discontinuations before progression (e.g. patients declining therapy). Therefore, a post hoc analysis of ToT (time on treatment) was performed, to be able to model treatment use more accurately. Here, ToT was defined as ‘Date of treatment end’ – ‘Date of treatment start’, as recorded in CORRECT and CONCUR.

Similar to PFS and OS, the model uses pooled ToT data from CORRECT and CONCUR to determine the duration of treatment in the base case. Due to the maturity of ToT, Kaplan–Meier data were used directly to inform ToT in the base case, with a parametric approach to ToT explored in the scenario analyses. The model uses a log-logistic extrapolation to inform ToT for the remainder of the model, when Kaplan–Meier data were no longer available, as the log-logistic estimates are closest to the end of the Kaplan–Meier data. By using this approach, the model assumes that ToT on average is similar but slightly shorter than PFS, as would be expected in clinical practice (see Figure 19).

Figure 19: Regorafenib and BSC survival curves used in the model base case

==> picture [334 x 335] intentionally omitted <==

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved

121 of 178

Key: BSC, best supportive care; KM, Kaplan-Meier; OS, overall survival; PFS, progression-free survival; ToT, time on treatment.

B.3.3.3. Comparators not included in CORRECT/CONCUR:

As discussed in Section B.2.9, there is no direct evidence comparing the efficacy of regorafenib directly to that of trifluridine/tipiracil. To be able to compare against trifluridine/tipiracil, an ITC using the primary publications for both treatments was conducted for OS and PFS (see Section B.2.9). The resulting ITC HRs versus regorafenib were applied to the OS and PFS curves used for regorafenib, to model trifluridine/tipiracil’s efficacy.

For OS and PFS, the full NMA using all data for regorafenib (CORRECT and CONCUR) and trifluridine/tipiracil (RECOURSE, Yoshino (2012), TERRA) showed both treatments to be comparable, with a small non-significant numerical benefit in favour of regorafenib, with OS and PFS HRs (CrIs) of 0.99 (0.84, 1.17) and 0.93 (0.85, 1.03), respectively. These point estimates for OS and PFS are used to model trifluridine/tipiracil OS, PFS, and ToT in the base case. The resulting

trifluridine/tipiracil survival curves used in the model base case are shown in Figure 20.

For ToT, no ITC could be performed, as insufficient ToT data for trifluridine/tipiracil were publicly available. However, considering the similarity between ToT and PFS in clinical practice, since patients are treated until progression, it is reasonable to assume the ToT and PFS HR will be similar as well. This is also illustrated by the PFS and ToT data used in the model, which shows a high degree of similarity (Figure 19). The model therefore assumes the estimated indirect PFS HR of 0.93 also applies to ToT. Similar to OS and PFS, this PFS HR versus regorafenib was applied to the ToT curves used for regorafenib, to model trifluridine/tipiracil’s ToT. The resulting trifluridine/tipiracil survival curves are shown in the Figure 20 overleaf.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 122 of 178

Figure 20: Trifluridine/tipiracil survival curves used in the model base case

==> picture [340 x 187] intentionally omitted <==

Key: OS, overall survival; PFS, progression-free survival; ToT, time on treatment.

However, as discussed in Section B.2.9, there were some limitations to the ITC, most notably around population differences between the included trials. To explore these uncertainties, the NMA includes several sensitivity analyses, which focussed on a different subsets of the studies. Specifically, this refers to the NMAs of CORRECT vs RECOURSE, CONCUR vs TERRA, CONCUR vs TERRA and Yoshino (2012), and the NMA that excluded Yoshino (2012). The resulting HRs are explored in the model as scenarios. In addition, considering that neither the NMA nor the sensitivity analyses showed any significant difference in OS or PFS, and that the consulted clinical and health economic experts agreed it was plausible to assume regorafenib and trifluridine/tipiracil are equivalent, we also explore a scenario that applies a HR of 1 to OS and PFS.

B.3.4. Measurement and valuation of health effects

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

In the CORRECT and CONCUR trials, patients’ QoL and health utility values were measured using the EORTC QLQ-C30 and EQ-5D-3L questionnaires, respectively. As discussed in Section B.2.6, only minor differences in EQ-5D-3L were observed between regorafenib and BSC in both trials, both for patients receiving treatment and at the end of treatment. Although there were some data indicating that patients on regorafenib had better quality of life, most notably from the higher response rates in

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved

123 of 178

CORRECT and data suggesting a superior QoL for patients on regorafenib,[38] no significant difference in QoL was observed in either study.[45, 48] The model therefore assumes that there is no utility benefit attributable to the treatment itself. Utilities are health-state specific with disutilities applied in the presence of adverse events. A average of all EQ-5D-3L index scores from CORRECT and CONCUR was

calculated, for all measurements while patients were on treatment, and for patients at the end of treatment. This pooled on-treatment utility value is assumed to represent the mean utility for the pre-progression health state, whereas the end of treatment utility is used for post-progression patients (Table 27).

None of the trifluridine/tipiracil studies included in the model captured EQ-5D. The only reported QoL data for trifluridine/tipiracil was captured in PRECONNECT, an open-label single arm post-registration study, designed to evaluate safety and QoL. This study was used by Sabater to calculate a pre- and post-progression utility for trifluridine/tipiracil of 0.72 and 0.59, respectively.[94] These values match the pre and post-progression values calculated from CORRECT and CONCUR i.e. also 0.72 and 0.59. In the economic model the values from CORRECT and CONCUR are used in the base case for both regorafenib and trifluridine/tipiracil (Table 27).

The use of utility values in the economic model are aligned to the reference case as 1) the EQ5D is preferred by NICE 2) the values are directly elicited from patients 3) the UK tariff was used.

Table 27: Utilities used in the base case of the model

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved

124 of 178

B.3.4.2. Mapping

No mapping was required, as EQ-5D-3L HRQL data were collected in both CORRECT and CONCUR.

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

In appendix H describe how systematic searches for relevant health-related quality-of-life data were done.

An SLR was performed to identify utility data for patients receiving ≥ 3L treatment for mCRC, and to validate the utility data observed in CORRECT and CONCUR trials. Full details of the search strategy and outcomes are described in Appendix H, and the identified unique studies are described below. Notably, many of the identified economic evaluations applied utility values derived from prior utility sources. Where this is the case, solely the originator study is reported.

In general, the utility values from the SLR are in line with the pooled CORRECT and CONCUR values used in the model (see Table 28). None of the alternative values was considered to be ‘better’ than using values derived from the CORRECT and CONCUR studies.

Several studies reported slightly higher progression-free utility values (0.70–0.81 in the SLR versus 0.72 in the model). In addition, the identified studies reported similar progression-free utility values between arms, with the active treatment sometimes showing a higher utility value than BSC.

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 125 of 178

Table 28: Utility systemic literature review results

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 126 of 178

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 127 of 178

B.3.4.4. Adverse reactions

To account for the effects of AEs, the model uses data from the CORRECT and CONCUR CSRs for regorafenib and BSC, and the main publications of RECOURSE, TERRA, and Yoshino for trifluridine/tipiracil to calculate AE rates per cycle.[45, 48, 51-53] The model considers AEs of Grade 3 and higher that occurred in at least 2% of patients in any treatment arm. This cut-off was chosen to ensure that infrequent but costly or severe AEs (e.g. febrile neutropenia, which occurred in 3.8% of trifluridine/tipiracil patients, or anaemia and thrombocytopenia, which occurred in 2.8% of regorafenib patients)[45, 51] are also considered in the model. The average AE rate per cycle per treatment across the different trials was calculated by combining the observed AEs and number of patients per arm, for the different trials. This was divided by the weighted average treatment duration to get an AE rate per treatment cycle and then converted to a probability per week, to be used in the model. These probabilities were combined with the costs per AE (discussed in Section B.3.5.3) to get the average AE cost per patient per week (i.e. per model cycle), which were applied to each model cycle until progression.

The pooled AE rates and resulting probabilities per week used in the model base case are shown in Table 29, below. As discussed in Section B.2.11, AEs in CORRECT and CONCUR were infrequent and mostly well tolerated.[45, 48] Trifluridine/tipiracil showed a different AE profile, with high rates for haematological AEs, as typically observed for chemotherapy. The overall rate of grade 3+ AEs of trifluridine/tipiracil was comparable to regorafenib.

Table 29: Grade 3+ treatment related AEs reported in at least 2% of patients in any treatment arm

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved 128 of 178

B.3.4.5. AE utility decrement

To capture the impact of AE on the patient’s QoL, utility decrements were also included in the model. Although it is likely that AE disutility is already captured within the utilities observed in the trials, most of the AEs were transient in nature. It is therefore uncertain whether patients would have had any disutility from the experienced AE on the day the EQ5D was administered. AE disutilities were

Company evidence submission template. Regorafenib for treating metastatic colorectal cancer (ID4002) © Bayer Plc (2022). All rights reserved

129 of 178

therefore included in the base case for all treatments, to ensure the different AE profiles are reflected in the utilities used in the model.

The disutilities per AE used in the model are shown in Table 30 below. These utilities were aligned with disutilities used in past TAs. The disutilities were combined with the pooled weekly AE probabilities, as reported in Section B.3.4.4, to calculate the average AE disutility per treatment. In line with TA405, the model assumes an AE duration of 1 week, so the average AE disutility was directly subtracted from the preprogression utility to generate treatment specific utilities. The resulting AE utilities in the base case were -0.00361, -0.00770, and -0.00124 for regorafenib, trifluridine/tipiracil, and BSC respectively. This is similar to TA405, where the ERG subtracted a disutility of 0.01 from the trifluridine/tipiracil pre-progression health statue utility.[9] The effect of including AE disutilities will also be explored in a scenario where disutilities were excluded.

Table 30: Disutilities per adverse event included in the model