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

Single Technology Appraisal

Fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (Review of TA756) [ID5115]

Committee Papers

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

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

SINGLE TECHNOLOGY APPRAISAL

Fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (Review of TA756) [ID5115]

Contents:

The following documents are made available to stakeholders:

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

1. Company submission from Celgene UK – a BMS company: a. Full submission

• b. Summary of Information for Patients (SIP)

2. Clarification questions and company responses

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

• a. MPN Voice-Leukaemia Care

• b. National Disease Registration Service SACT report

4.

Expert personal perspectives from:

• a. Professor Tim Somervaille – clinical expert, nominated by Celgene UK – a BMS company

• b. Andy Tattersall – patient expert, nominated by MPN Voice

• c. Jonathan Mathias – patient expert, nominated by MPN Voice

5. External Assessment Report prepared by School of Health and Related Research, University of Sheffield

6.

External Assessment Report – factual accuracy check

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

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE

Single technology appraisal

Fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

Document B

Company evidence submission

March 2024

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

CONTENTS

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

TABLES

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

FIGURES

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

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

Fedratinib (INREBIC[®] ) in the treatment of disease-related splenomegaly or symptoms in myelofibrosis was evaluated by the National Institute for Health and Care Excellence (NICE) in December 2021 (TA756). During the evaluation, a number of uncertainties regarding the efficacy of fedratinib were outlined. The committee believed that fedratinib is clinically effective, but the lack of comparator data made assessing the comparative effectiveness challenging. Furthermore, the indirect treatment comparison suggested fedratinib improved response compared with best available therapy (BAT), but there were uncertainties due to the methodological approach. The committee also identified the ability of fedratinib to extend overall survival (OS) compared with BAT as a key uncertainty. After the evaluation, fedratinib was recommended for use within the Cancer Drugs Fund (CDF) to allow a period of managed access, supported by additional data collection to answer the clinical uncertainty.

During the CDF managed access stage, further evidence on fedratinib in the post-ruxolitinib setting was collected through the FREEDOM-2 trial to address these uncertainties. FREEDOM-2 is a multicentre, open-label, randomised phase 3 study to evaluate the efficacy and safety of fedratinib compared with BAT in participants with Dynamic International Prognostic Scoring System (DIPSS)–intermediate or high-risk primary myelofibrosis (PMF), postpolycythaemia vera myelofibrosis (post-PV MF), or post-essential thrombocythemia myelofibrosis (post-ET MF), and previously treated with ruxolitinib. In addition, supplementary data showing the real-world effectiveness of fedratinib in the CDF population has been routinely collected from Systemic Anti-Cancer Therapy (SACT) data set and has been analysed and reported here.

B.1.1 Decision problem

The marketing authorisation for fedratinib (INREBIC[®] ) is for “the treatment of disease-related splenomegaly or symptoms in adult patients with primary myelofibrosis, post-polycythaemia vera myelofibrosis or post-essential thrombocythaemia myelofibrosis who are Janus kinase (JAK) inhibitor naive or who have been treated with ruxolitinib.” This submission focuses only on those patients who have been exposed to ruxolitinib. The proposed position in the treatment pathway is narrower than the marketing authorisation because this position reflects the unmet need within the myelofibrosis treatment pathway and reflects where clinicians anticipate using fedratinib in UK practice due to the current lack of active treatments available and in line with subsequent CDF access.

The decision problem addressed is summarised in Table 1.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 1. The decision problem

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115] © BMS (2024). All rights reserved

BAT = best available therapy; CT = computed tomography; HRQOL = health-related quality of life; IWG-MRT = International Working Group-Myeloproliferative Neoplasms Research and Treatment; JAK = Janus kinase; MRI = magnetic resonance imaging; NHS = National Health Service; OS = overall survival; RBC = red blood cell; SACT = Systemic Anti-Cancer Therapy; SVR = spleen volume reduction; UK = United Kingdom.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115] © BMS (2024). All rights reserved

B.1.2 Description of the technology being evaluated

A summary description of fedratinib, including details of its mechanism of action and marketing authorisation, is provided in Table 2.

Appendix C provides a draft summary of the product characteristics (SmPC).

Table 2. Technology being evaluated

CHMP = Committee for Medicinal Products for Human Use; EMA = European Medicines Agency; JAK = Janus kinase; STAT = signal transducer and activator of transcription; TYK2 = tyrosine kinase 2; UK = United Kingdom. Source: Inrebic SmPC[1]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

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

B.1.3.1 Overview of the disease

Myelofibrosis is a rare haematological disorder characterised by abnormal cytopenias, bone marrow fibrosis, and extramedullary haematopoiesis; myelofibrosis often results in splenomegaly, constitutional symptoms, and shortened survival.[2-5] Most people with myelofibrosis have a gene variant that results in constitutive activation of the JAK/signal transducer and activator of transcription (STAT) signalling pathway.[6,7] Activation of this pathway results in cell proliferation, inhibition of cell death, and clonal expansion of myeloproliferative malignant cells. The abnormal proliferation of pluripotent

haematopoietic stem cells that release inflammatory cytokines and growth factors in the bone marrow leads to marrow fibrosis. Progressive bone marrow fibrosis results in release of the malignant stem cells into the circulation and may result in extramedullary haematopoiesis, manifesting as splenomegaly. Extramedullary haematopoiesis is not able to fully compensate for the loss of production of blood cells in the bone marrow; as a result, individuals experience a decrease in one or more blood cell types, i.e., cytopenias (most commonly anaemia and thrombocytopenia). Myelofibrosis may also undergo transformation to acute myeloid leukaemia (AML).[8]

The disease can present as PMF or secondary to polycythaemia vera or essential thrombocythaemia. Myelofibrosis is diagnosed and stratified by risk using one of the following scoring systems: the International Prognostic Scoring System (IPSS), the DIPSS or DIPSS Plus.[9] These are used to classify individuals into 1 of 4 risk groups (low, intermediate-1, intermediate-2, and high) based on factors such as age, presence of constitutional symptoms, and haematological parameters. Approximately half of people with myelofibrosis are found to have either intermediate-2 or high-risk myelofibrosis,[10] which is associated with a poor overall prognosis and very limited survival time.[11 ]

B.1.3.2 Epidemiology and prognosis

Myelofibrosis typically occurs more frequently with increasing age, with the median age at diagnosis being approximately 65 years.[12-14] It affects slightly more men (62%) than women (38%).[12] Epidemiological estimates for myelofibrosis in people in the United Kingdom (UK) suggest a 10-year prevalence of 3.2 per 100,000 and an annual incidence of 0.6 per 100,000.[15] This suggests that the total population of people with myelofibrosis is 2,130, half of whom are expected to have intermediate-2 and high-risk disease.

People within these risk groups represent a population with considerably worse outcomes compared with people with intermediate-1 or low risk disease.[8,16,17] Currently, only ruxolitinib is recommended by NICE for use in patients with intermediate-2 or high-risk disease.[10] When patients become relapsed, refractory or intolerant to treatment, survival outcomes are poor with several published reports demonstrating a median OS of 13-16 months post-ruxolitinib treatment.[10,18-20] The poor survival outcomes in these patients are attributable to the lack of effective treatment options in the relapsed, refractory and intolerant to ruxolitinib setting, with many patients on suboptimal treatment (see Section B.1.3.4).[18,19] Data from clinical trials indicate that 45% to 89% of patients who are relapsed and refractory to ruxolitinib continue suboptimal ruxolitinib treatment with limited benefits in the absence of other active treatment options.[21-24 ]

B.1.3.3 Physical and psychological burden of disease

Over 80% of people with myelofibrosis experience splenomegaly, while other clinical manifestations of myelofibrosis include symptoms associated with cytopenias (> 35% of people), fatigue (> 90%), and

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

constitutional symptoms (~ 30%).[16] Myelofibrosis is associated with a range of debilitating symptoms that may worsen as the disease progresses and can have a major impact on health-related quality of life (HRQOL).[2,3,25] These stem from the pathological changes in haematopoiesis and the bone marrow, as described above. Splenomegaly can lead to abdominal pain, early satiety, and portal hypertension, whereas progressive bone marrow fibrosis leads to worsening cytopenias, particularly thrombocytopenia and anaemia.[2] Anaemia is associated with fatigue, weakness, palpitations, bone pain, and dyspnoea[2] , whereas cytopenias, such as thrombocytopenia and neutropenia, result in complications such as

petechiae and infection, respectively. The risk of cytopenias increases with disease progression, resulting in more severe symptoms and an increased risk of leukaemic transformation.

Although extramedullary haematopoiesis predominantly occurs in the spleen and liver, it can also occur in other organs resulting in further complications such as chronic headache, spinal cord compression and pleural effusions.[2]

There are also a range of constitutional symptoms that result from abnormal cytokine production related to the proliferation of progenitor cells. These include fatigue, pruritis, night sweats, fever and cachexia (leading to weight loss) (Figure 1).[2,3,26]

Approximately 10% to 20% of people with PMF will progress to AML.[27] These people have dismal outcomes, with OS ranging from 3 to 8 months and a 1-year survival rate of 5% to 10%.[13]

Figure 1. Debilitating symptoms of myelofibrosis

==> picture [390 x 220] intentionally omitted <==

Adapted from Mesa et al.[26] ; Polverelli et al.[28] ; Devendra et al.[29] ; Kander et al.[30] ; Finazzi et al.[31]

Studies reporting on the impact of myelofibrosis symptoms on HRQOL suggest that myelofibrosis particularly impacts physical and social function, and this impact increases with disease progression.[25,26,32] The negative effect on HRQOL experienced by people with myelofibrosis is comparable with that reported for people with recurrent cancer and represents a clinically meaningful reduction in the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core) (EORTC QLQ-C30) global health score compared with the general population.[25] Many people reduce their working hours or take early retirement because of myelofibrosis.[26,33,34]

In patients who have been treated with ruxolitinib, the physical and psychological burden of myelofibrosis is particularly pronounced. A comparison of the HRQOL at baseline for JAK-naive patients from one of the ruxolitinib pivotal trials, COMFORT-II,[35] with baseline data for ruxolitinib-

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

exposed patients included in the fedratinib JAKARTA-2 trial suggests that HRQOL is worse in people who have been treated with ruxolitinib. Both studies assessed HRQOL using the EORTC QLQ-C30. In FREEDOM-2, HRQOL was assessed by EQ-5D-5L in ruxolitinib-experienced patients who were treated with fedratinib. The results showed a clinically meaningful change from cycle 2 day 1 (C2D1) through C5D1, with a mean (standard error of the mean [SEM]) change from baseline of 6.2 (18.97) for patients treated with fedratinib (see Section B.2.6.1.4).[36]

B.1.3.4 Clinical pathway of care

Allogeneic stem cell transplant is the only potentially curative treatment of myelofibrosis; however, it is only suitable for people who are fit enough to undergo treatment as it is associated with considerable morbidity and mortality.[37] Allogeneic stem cell transplant is generally only considered for people with intermediate-2 or high-risk myelofibrosis, of whom only 5% to 10% will meet eligibility criteria for such an intensive therapy.[38,39]

In first-line treatment, options aim to relieve debilitating symptoms, particularly splenomegaly and cytopenia, and improve HRQOL. This includes targeted therapy with JAK inhibitors such as ruxolitinib. Ruxolitinib is the only targeted treatment recommended for use in people with myelofibrosis (with intermediate-2 and high-risk disease) in clinical practice in the UK.[10]

There are considerable limitations associated with treatment with ruxolitinib. Of patients treated with ruxolitinib in clinical trials so far, only 28% to 42% have achieved the primary endpoint of 35% or more spleen volume reduction (SVR) from baseline.[32,40,41] Reports from the COMFORT long-term follow-up trials state more than 50% of patients discontinue ruxolitinib treatment after 3-5 years[24] ; however, this may not be reflective of UK clinical practice. Feedback from UK clinicians, including feedback from an advisory board, revealed that many patients continue to receive suboptimal treatment with ruxolitinib, despite being relapsed or refractory (Figure 2). Reasons for this include the lack of treatment options and concerns regarding the potential for a proinflammatory state and acute deterioration of the patient due to ruxolitinib withdrawal.[23,42] These withdrawal symptoms include acute relapse of disease symptoms, accelerated splenomegaly, worsening of cytopenias, and occasional haemodynamic decompensation (including a septic shock-like syndrome).[43 ]

Figure 2. Current treatment duration in patients who respond to ruxolitinib

==> picture [448 x 187] intentionally omitted <==

BAT = best available therapy.

This continuation of suboptimal ruxolitinib in UK clinical practice aligns with observations from PERSIST-2 and SIMPLIFY-2, where considerable proportions of participants in the BAT arms were receiving ruxolitinib (45% and 89%, respectively).[21-23] BAT includes treatment options that are largely

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

supportive and do not significantly alter the course of the disease. These may also include treatments such as hydroxycarbamide, other chemotherapies, androgens, splenectomy, radiation therapy, erythropoietin, and red blood cell (RBC) transfusion.

Patients relapsed and refractory to ruxolitinib have a reduced life expectancy, with an estimated median OS of 13-16 months after discontinuation.[10,18-20] Data on survival in patients who continue suboptimal ruxolitinib are uncertain; however, it is not expected to be significantly greater than observed in the literature, which is supported by clinical experts.[25]

Given that there are currently no disease-modifying treatment options available in the UK to patients who no longer respond to ruxolitinib, fedratinib was introduced under the CDF. The introduction of fedratinib to the pathway of care provides an opportunity for targeted therapy in a patient population otherwise associated with poor survival outcomes.

Figure 3 presents the clinical pathway of care for people with myelofibrosis in England and the position of fedratinib under the CDF within this pathway.

Figure 3. Clinical pathway of care for people with intermediate-2 and high-risk myelofibrosis in England

==> picture [452 x 224] intentionally omitted <==

ASCT = allogenic stem cell transplant; BAT = best available therapy; ET = essential thrombocythaemia; Int = intermediate; PV = polycythaemia vera; RBC = red blood cell.

B.1.3.5 Unmet medical need

In the current clinical pathway of care, ruxolitinib is the only targeted treatment available and is associated with low response rates, with less than half of participants in clinical trials achieving the primary endpoint.[32,40] In patients who do respond, many will become relapsed or refractory to ruxolitinib over time. In lieu of alternative treatment options, relapsed and refractory patients remain on suboptimal therapy.[23,42] Outcomes in patients no longer responding to ruxolitinib are poor, with a loss of response associated with worse symptoms and an increased spleen size, causing detriments to HRQOL. There is a significant unmet need for a new therapy to address this and provide an alternative treatment option so that clinicians do not have to resort to using limited healthcare resources for suboptimal treatment.

Fedratinib, a targeted and novel therapy, offers an effective treatment option that has shown a clinically meaningful response in patients who have been treated with ruxolitinib. These benefits lead to

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

considerable HRQOL improvements and expected improvements in long-term survival in a patient population that would otherwise experience poor outcomes.

B.1.4 Equality considerations

No potential equality considerations have been raised for the use of fedratinib in people with myelofibrosis.

B.2 Clinical effectiveness

B.2.1 Identification and selection of relevant studies

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

In summary, a systematic literature review (SLR) was conducted in February 2020 for primary intervention trials (randomised controlled trials and prospective non–randomised controlled trials) assessing the efficacy and safety of fedratinib or comparator therapies in people with myelofibrosis. Since February 2020, no other pharmaceuticals have been approved by NICE for this indication. Therefore, no further evidence is anticipated to be found other than the pivotal trial FREEDOM-2 (providing head-to-head data); thus, an update of the February 2020 SLR would not affect this submission.

The SLR identified 2 key studies that evaluated fedratinib as an active intervention:

• The phase 3 trial, JAKARTA, investigated the safety and efficacy of fedratinib in the ruxolitinibnaive population.

• The phase 2 trial, JAKARTA-2, investigated the safety and efficacy of fedratinib in patients previously treated with ruxolitinib.

Since the SLR was conducted, the following clinical trials have been either completed or have had their preliminary results analysed:

• The phase 3 single-armed trial, FREEDOM, investigated the safety and efficacy of fedratinib in patients previously treated with ruxolitinib (completed).

• The phase 3 trial, FREEDOM-2, investigates the safety and efficacy of fedratinib compared with BAT in patients previously treated with ruxolitinib (the trial is still ongoing but not recruiting).

JAKARTA was a randomised, double-blind, placebo-controlled, phase 3 trial that compared 400 mg or 500 mg fedratinib with placebo in participants with intermediate-2 or high-risk PMF, post-PV MF, or postET MF with splenomegaly. Based on the lack of a head-to-head comparator and the study taking place in the ruxolitinib-naive population, the study is not relevant to this submission and is not included in the main dossier. Details and results of this study are presented in Appendix D as supportive evidence.

JAKARTA-2 was a phase 2, multicentre, open-label, single-arm study that evaluated the efficacy of a once daily, 400 mg dose of fedratinib in 97 participants previously treated with ruxolitinib and was included in the original NICE technology appraisal assessment of fedratinib (TA756). JAKARTA-2 has not been used to populate the model for this submission; therefore, details and results of this study are presented in Appendix D as supportive evidence.

FREEDOM was a phase 3, single-arm, US-based, open-label study of 38 participants previously treated with ruxolitinib who had with intermediate-2 or high-risk PMF, post-PV MF, or post-ET MF. This

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

study was not included in the economic model because it did not include a head-to-head comparison. A summary of this study is Appendix D as supportive evidence.

B.2.2 List of relevant clinical effectiveness evidence

The key clinical study used for this submission is FREEDOM-2, a phase 3, open-label, randomised study of 201 participants with intermediate-2 or high-risk PMF, post-PV MF, or post-ET MF, comparing fedratinib with BAT (Table 3).

Because FREEDOM-2 provides direct evidence for fedratinib compared with BAT in a patient population that has been treated with ruxolitinib, it forms the key source of clinical and economic evidence in this submission and is described in detail in the following sections.

Furthermore, National Health Service (NHS) England have evaluated the real-world treatment effectiveness of fedratinib in the CDF population during the managed access period. This report presents the results of the use of fedratinib in clinical practice in England, using the routinely collected SACT data set (Table 4).

The SACT data were not used to populate the economic model but has been included as supplementary evidence and summarised in Sections B.2.2-B.2.4 and B.2.6. This study was not included in the economic model because it does not report all the key outcomes, does not have sufficient time for OS follow-up, and does not compare the outcome of fedratinib with BAT.

Table 3. Primary source of clinical effectiveness evidence: FREEDOM-2 (data cut, December 2022)

BAT = best available therapy.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 4. Clinical effectiveness evidence: SACT data cohort study

SACT = Systemic Anti-Cancer Therapy.

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

B.2.3.1 FREEDOM-2

FREEDOM-2 was a phase 3, multicentre, open-label, randomised study that evaluated the efficacy and safety of a once daily 400-mg dose of fedratinib compared with BAT in 201 participants previously treated with ruxolitinib. The study included adult participants aged ≥ 18 years with a current diagnosis of intermediate-2 or high-risk PMF, post-PV MF, or post-ET MF.[36]

Participants included in FREEDOM-2 had been previously exposed to ruxolitinib with inadequate response as refractory or relapsed (< 10% SVR by magnetic resonance imaging [MRI] or < 30% decrease from baseline in spleen size by palpation or regrowth). Alternatively, participants treated with ruxolitinib for more than 28 days and experiencing a complication that requires either RBC transfusion or having grade ≥ 3 adverse events (AEs) of thrombocytopenia, anaemia, hematoma, and/or haemorrhage were also included in the study.[36]

The FREEDOM-2 trial design consisted of a screening period of up to 35 days with a 2:1 randomisation to fedratinib or BAT. For the fedratinib arm, the treatment phase consisted of 4 -week (28-day) cycles of fedratinib and a follow-up visit (approximately 30 days after the last dose of fedratinib). Participants could remain on fedratinib until disease progression or unacceptable toxicity. Participants were allowed to cross over from BAT to fedratinib after the sixth cycle response assessment, or earlier in the event of disease progression (confirmation of splenomegaly by MRI/computed tomography [CT] scan) (Figure 4).[36 ]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Figure 4. FREEDOM-2: study design

==> picture [441 x 235] intentionally omitted <==

AE = adverse event; AML = acute myeloid leukaemia; ANC = absolute neutrophil count; C6 = cycle 6; DIPSS = Dynamic International Prognostic Scoring System; GI = gastrointestinal; Int-2 = intermediate-2; LCM = left costal margin; MF = myelofibrosis; MPN = myeloproliferative neoplasm; MRI = magnetic resonance imaging; PB = peripheral blood; PO = orally; PRO = patient-reported outcome; PV = polycythaemia vera; QOL = quality of life; SVR = spleen volume reduction; WE = Wernicke’s encephalopathy.

Source: BMS data on file[36]

The primary outcome measure in FREEDOM-2 was spleen volume response rate, defined as the proportion of participants with a ≥ 35% SVR from baseline at the end of cycle 6 (EOC6) in the fedratinib and BAT arms. This was measured using an MRI or CT scan and assessed by blinded central review.[36] Splenomegaly is the main physical feature of myelofibrosis and the cause of many symptoms associated with the disease. As such, SVR is a key treatment goal in myelofibrosis (see Section B.2.6.1.2).

Secondary outcomes measured in FREEDOM-2 include[36] :

• Symptom response rate, proportion of participants with ≥ 50% reduction in total symptom score (TSS), measured by Myelofibrosis Symptom Assessment Form (MFSAF) at EOC6

• Spleen volume response rate by MRI or CT scan, proportion of participants who have ≥ 25% reduction in spleen volume at the EOC6 (RR25)

• Spleen response by palpation, ≥ 50% reduction in spleen size if spleen was > 10 cm below left costal margin (LCM) or non-palpable if spleen was palpable at 5 to 10 cm below the LCM

• Durability of spleen volume response, duration of ≥ 35% SVR by MRI/CT

• Durability of spleen response by palpation, ≥ 50% reduction in spleen size by palpation for participants with a palpable spleen at least 5 cm below LCM

• Durability of symptoms response, ≥ 50% reduction in TSS measured by MFSAF

• Spleen and disease progression-free survival (SDPFS)

• Overall survival (OS)

• HRQOL measured by EORTC QLQ-C30 domains

• Patient-reported outcomes (PROs) measured by EQ-5D-5L

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Exploratory outcomes measured in FREEDOM-2 include[36] :

• Time to spleen response by palpation: time from baseline to a ≥ 50% reduction in spleen size by palpation for participants with a palpable spleen at least 5 cm below the LCM at baseline

• Best spleen volume response rate during first 6 treatment cycles (BRR6): measured by MRI/CT scan during the first 6 treatment cycles and by MRI/CT scan from the start of study treatment to the end of study treatment

• Anaemia response: ≥ 2 g/dL increase in haemoglobin level in transfusion-independent participants or transfusion-dependent participants who become transfusion independent

A summary of the methodology of FREEDOM-2 is provided in Table 5.

Table 5. FREEDOM-2: summary of methodology

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

AE = adverse event; CT = computed tomography; DIPSS = Dynamic International Prognostic Scoring System; ECOG PS = Eastern Cooperative Oncology Group performance status; eCRF = electronic case report form; EOC6 = end of cycle 6; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); HRQOL = health-related quality of life; JAK = Janus kinase; LCM = left costal margin; MFSAF = Myelofibrosis Symptom Assessment Form; MRI = magnetic resonance imaging; PRO = patient-reported outcome; QOL = quality of life; RBC = red blood cell; SAE = serious adverse event; SVR = spleen volume reduction; TSS = total symptom score; UK = United Kingdom.

Sources: BMS data on file[36] ; BMS data on file[44]

B.2.3.1.1 Baseline demographics

The demographics and baseline disease characteristics in FREEDOM-2 are representative of a group of patients with advanced myelofibrosis and a high disease burden.

The demographic data matched what was anticipated based on the study design. The treatment groups did not differ significantly from one another. For the entire intention-to-treat (ITT) population, the median age was 70 years, ranging from 38 to 91 years. Most of the participants were White (n = 164 [81.6%]). Table 6 summarises the baseline characteristics in FREEDOM-2.

Table 6. FREEDOM-2: baseline characteristics (ITT population)

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

BAT = best available therapy; CT = computed tomography; ECOG PS = Eastern Cooperative Oncology Group performance status; IPSS = International Prognostic Scoring System; IWG-MRT = International Working GroupMyeloproliferative Neoplasms Research and Treatment; ITT = intention to treat; JAK2 = Janus kinase 2; MPN-SAF = Myeloproliferative Neoplasm Symptom Assessment Form; MRI = magnetic resonance imaging; RBC = red blood cell.

Notes: Spleen volume was measured by MRI/CT scan and reviewed in a blinded fashion by a central imaging laboratory. Spleen size was measured by palpation (i.e., length in cm).

a RBC transfusion dependence at baseline was defined per revised IWG-MRT criteria 2013.

• b A participant had constitutional symptoms if any of the symptoms were in the baseline MPN-SAF (> 10% weight loss in 6 months, night sweat, unexplained fever > 37.5 °C).

c Baseline spleen volume by MRI/CT scan based on central review.

d Below lower coastal region.

Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Prior myelofibrosis treatment

All 201 participants had received prior treatment with ruxolitinib over at least 3 months. The median time from the last ruxolitinib dose to the first dose of fedratinib was 21 days (minimum 3 days, maximum 2,671 days). Prior anticancer therapies other than ruxolitinib for MF were reported for 34 participants (16.9%). The compound reported most often was hydroxycarbamide (in 28 participants [13.9%]). Prior RBC transfusions were reported for 101 participants (50.2%) and prior platelet transfusions were reported for 8 (4.0%). There were no notable differences between the treatment groups. Red blood cell transfusion dependence at baseline was reported for 29 participants (21.6%) in the fedratinib group and 11 (16.4%) in the BAT group.[36]

Of the participants enrolled and treated in FREEDOM-2, 20.4% were intolerant to ruxolitinib, 27.9% had a loss of response to ruxolitinib, and 26.9% never responded to ruxolitinib.[45] A summary of the reasons for ruxolitinib discontinuation is provided in Table 7.

Table 7. FREEDOM-2: reasons for ruxolitinib discontinuation by investigator assessment (ITT population)

AE = adverse event; BAT = best available therapy; ITT = intention to treat.

Note: Intolerance: haematological toxicity (anaemia, thrombocytopenia, other), non-haematological toxicity. Source: BMS data on file[45]

Patient summary by crossover status

All randomly assigned participants qualified for the ITT and safety populations, and were treated according to their randomisation.[36] As previously described, the study allowed for crossover from BAT to fedratinib after the sixth cycle response assessment, or earlier in the event of disease progression (confirmation of splenomegaly by MRI/CT scan), which aligns with the timing of assessment in clinical practice.[44] The number of participants per treatment group and population is presented in Table 8.

Table 8. FREEDOM-2: analysis population (ITT population)

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

• BAT = best available therapy; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); HRQOL = health-related quality of life; ITT = intention to treat; PRO = patient-reported outcome; PRO-CTCAE = Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events.

• Notes: Percentages are based on the number of participants in the ITT population.

• a All participants who were randomly assigned.

• b All participants who received the treatment they were assigned to by randomisation, had no important violation of inclusion/exclusion criteria, and no other important protocol deviations that could impact on efficacy outcome.

• c All participants who had been treated and had evaluable symptom assessments (i.e., non-zero total symptom score) at baseline and at least 1 after baseline.

• d All participants who were administered at least 1 dose of study medication.

• e All participants from the BAT arm who crossed over to the fedratinib arm.

f All participants from the BAT arm who crossed over to the fedratinib arm and received at least 1 dose of fedratinib.

• g All participants who had an evaluable assessment of a given PRO/HRQOL measure at baseline and at least 1 evaluable assessment of a given PRO/HRQOL measure after baseline.

Source: BMS data on file[36]

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

B.2.4.1 FREEDOM-2

The primary objective of FREEDOM-2 was to compare the SVR at EOC6 in participants who received fedratinib versus those who received BAT. The primary outcome was spleen volume response rate, which was defined as the proportion of participants who had ≥ 35% SVR at EOC6 measured from baseline. The hypothesis was as follows:

==> picture [146 x 57] intentionally omitted <==

To consider the sample size needed for FREEDOM-2, data from JAKARTA-2 were used. In the JAKARTA-2 study, the reported response rate of fedratinib was approximately 35% for the per-protocol population. With the assumption of a 20% dropout rate, the response rate would be 28% based on the ITT population. Assuming a true ITT response rate (considering dropouts) in the BAT arm of 10%, 192 participants will yield approximately 90% power at a one-sided significance level of 0.025.[44]

The primary analysis for spleen volume response rate as measured by MRI/CT is based on the ITT population. The data cutoff for response rate occurred when the last randomly assigned participant had completed 6 cycles of fedratinib or BAT. Participants with a missing measurement of spleen volume at EOC6, including those who meet the criteria for progression of splenomegaly before EOC6, were considered as non-responders. For crossover participants, only data before crossover were included.[44]

A Cochran-Mantel-Haenszel (CMH) test to adjust for planned stratification factors (spleen size by palpation, platelet counts, and refractory/relapsed or intolerance to ruxolitinib treatment) was performed to compare fedratinib to BAT at a one-sided 2.5% alpha level. The response rates and 95% confidence

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

intervals (CIs) were provided for each arm, as well as for the difference in response rates and 95% CI of the difference for fedratinib to BAT.[44] Table 9 summarises the statistical analyses in FREEDOM-2.

Table 9. FREEDOM-2: summary of statistical analyses

AE = adverse event; BAT = best available therapy; CI = confidence interval; CMH = Cochran-Mantel-Haenszel; CRF = case report form; CT = computed tomography; max = maximum; min = minimum; MRI = magnetic resonance imaging; n = number of observations; SD = standard deviation; SVR = spleen volume reduction.

Source: BMS data on file[44]

Further information regarding the participant flow in FREEDOM-2 is presented in Appendix D.

B.2.4.2 SACT data set

The SACT data set was an evaluation of real-world treatment efficacy of fedratinib in the CDF population. There were 75 applications for CDF funding of fedratinib for the treatment of disease-related splenomegaly or symptoms in myelofibrosis between 17 November 2021 and 31 October 2022 in the NHS England Blueteq database. After de-duplication, 67 unique patients were identified. One patient was then excluded because they received fedratinib before the drug was available through the CDF, 1 patient did not receive treatment, 4 patients died before treatment, and 7 patients were missing from SACT, resulting in 54 patients being included in the analysis.[46]

The included patients were then followed up through the National Disease Registration Service (NDRS) routinely collected SACT data to provide SACT treatment history. The efficacy outcomes presented in the NDRS NHS England report were treatment duration and OS.[46] Table 10 presents the clinical treatment criteria for patients included in the SACT data set as well as a summary of the methodology.[46] Table 11 presents the completeness of the key SACT variables collected.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 10. SACT data set: summary of methodology

CDF = cumulative distribution function; ECOG PS = Eastern Cooperative Oncology Group performance status; NDRS = National Disease Registration Service; NHS = National Health Service; OS = overall survival; SACT = Systemic Anti-Cancer Therapy.

Source: NHS England data on file[46]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 11. SACT data set: completeness of key data variables for the fedratinib cohort (N = 54)

SACT = Systemic Anti-Cancer Therapy. Source: NHS England data on file[46]

B.2.4.2.1 Patient characteristics

The median age of the 54 patients receiving fedratinib for the treatment of disease-related splenomegaly or symptoms of myelofibrosis was 72 years. The median age in males and females was 73 and 72 years, respectively. Table 12 presents patient characteristics in the SACT data set.

Table 12. SACT data set: patient characteristics

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

SACT = Systemic Anti-Cancer Therapy. Note: Figures may not sum to 100% due to rounding. Source: NHS England data on file[46]

B.2.5 Critical appraisal of the relevant clinical effectiveness evidence

B.2.5.1 FREEDOM-2

This study is generally considered a high-quality study, being conducted in accordance with the ethical principles that have their origin in the Declaration of Helsinki. The rights, safety, and well-being of the study participants were the most important consideration and prevailed over the interests of science and society. The protocol, amendments, and participant informed consent received appropriate approval by the Institutional Review Board/Independent Ethics Committee before initiation of study at the site.[36]

Measures to minimise bias were considered in the study design and third parties were used for critical activities, such as[36] :

• Randomised treatment allocation

• Prespecified criteria for dose modification, dose adjustment, and management of nausea, vomiting, diarrhoea, and encephalopathy including Wernicke’s encephalopathy

• Use of centralised vendors: interactive response technology (IRT), central laboratories

• Selected study procedures performed in a blinded manner: MRIs/CTs relevant for the primary efficacy outcome, and further efficacy outcomes

• Supervision of the conduct of the trial by a Steering Committee, presided over by the coordinating Principal Investigator and representative Regional Investigators from participating countries

B.2.6 Clinical effectiveness results of the relevant studies

B.2.6.1 FREEDOM-2

B.2.6.1.1 Overview

The efficacy of fedratinib over BAT in patients who have been treated with ruxolitinib has been demonstrated in FREEDOM-2[36] and is supported by similar efficacy in the single-armed FREEDOM trial, the JAKARTA-2 trial, and the JAK inhibitor–naive patient population from JAKARTA.[41,47]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

The results for the primary endpoint (≥ 35% SVR at EOC6), as well as the results for both of the 2 key secondary endpoints (≥ 50% reduction in TSS at EOC6, ≥ 25% SVR at EOC6), proved superiority for fedratinib over BAT while controlling the family-wise Type I error rate at a one-sided significance level of 0.025.

In the ITT population of FREEDOM-2, 35.8% of participants receiving fedratinib and 6% receiving BAT achieved the primary outcome of spleen volume response rate defined as ≥ 35% SVR at EOC6. Similarly, 34.1% of participants receiving fedratinib and 16.9% receiving BAT achieved the key secondary outcome of symptom response rate defined as ≥ 50% reduction in TSS at EOC6. Furthermore, 47% of participants receiving fedratinib and 13.4% receiving BAT achieved the key secondary outcome of spleen volume response rate defined as ≥ 25% SVR at EOC6. An overview of the results for key outcomes from FREEDOM-2 is provided in Table 13. Full trial results for the ITT FREEDOM-2 population are presented in subsequent sections. Supporting results from JAKARTA, JAKARTA-2, and FREEDOM are provided in Appendix D.

Table 13. FREEDOM-2: overview of fedratinib efficacy (ITT population)

BAT = best available therapy; CI = confidence interval; EOC6 = end of cycle 6; ITT = intention to treat; SVR = spleen volume reduction; TSS = total symptom score.

a Participants with missing assessment at EOC6, including those who met the criteria for progression of splenomegaly before EOC6, were considered non-responders. They were included in the denominator.

b The 2-sided 95% CI was based on the exact Clopper-Pearson method.

Source: BMS data on file[36]

B.2.6.1.2 Primary outcome: spleen volume response rate (≥ 35% SVR) at EOC6

Treatment with fedratinib is associated with a significant spleen volume response rate, with 35.8% of participants achieving ≥ 35% SVR at EOC6 compared with 6.0% of participants on BAT. Superiority of fedratinib was proven in the stratified analysis based on electronic case report form (eCRF) data and based on IRT data, as well as in the unstratified analysis. After imputation of missing data, the percentage of participants with ≥ 35% SVR at EOC6 was 43.3% in the fedratinib group and 9.0% in the BAT group (Table 14).[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 14. FREEDOM-2: spleen volume response rates at EOC6 (≥ 35% SVR) by MRI or CT scan (ITT population)

• BAT = best available therapy; CI = confidence interval; CMH = Cochran-Mantel-Haenszel; CT = computed tomography; eCRF = electronic case report form; EOC6 = end of cycle 6; IRT = interactive response technology; ITT = intention to treat; MRI = magnetic resonance imaging; SVR = spleen volume reduction.

• a Participants with missing assessment at EOC6, including those who met the criteria for progression of splenomegaly before EOC6, were considered non-responders. They were included in the denominator.

• b The 2-sided 95% CI was based on the exact Clopper-Pearson method.

• c The stratified P value was one-sided based on CMH test using the Greenland and Robins method to adjust for stratification factors: spleen size by palpation and platelet counts. The third stratification factor ‘refractory/relapsed or intolerance to ruxolitinib treatment’ was dropped due to small cell count issue.

• d The 95% CI of the difference was based on Greenland and Robins method.

• e The unstratified P value was one-sided based on the z-test with an unpooled estimate of variance.

• f For participants with missing assessment at EOC6, data were imputed using multiple imputation methods. Thirty imputed data sets were created. The proportion with 95% CI for each arm and the difference in proportion with 95% CI were calculated for each imputed data set and combined using Rubin’s rules. The number of responders was backcalculated from the pooled adjusted response rate.

Source: BMS data on file[36]

In the ITT population, when considering individual changes in spleen volume for participants with measurements at baseline and EOC6, all participants except 5 in the fedratinib arm compared with 14 in the BAT arm showed a reduction in volume (Figure 5).[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Figure 5. FREEDOM-2: percentage change in spleen volume from baseline to EOC6 in participants with MRI/CT scan at EOC6 (ITT population)

==> picture [453 x 474] intentionally omitted <==

CT = computed tomography; EOC6 = end of cycle 6; ITT = intention to treat; MRI = magnetic resonance imaging.

Note: Each bar represents a participant with both baseline and post-baseline results. The dotted lines represent the 25% and 35% change from baseline.

Source: BMS data on file[36]

Additional analyses of spleen volume response rate (≥ 35% SVR)

The analysis of the primary efficacy endpoint was repeated based on data at EOC3. The advantage for fedratinib was supported by the results (Table 15).[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 15. FREEDOM-2: spleen volume response rates at EOC3 (≥ 35% SVR) by MRI or CT scan (ITT population)

• BAT = best available therapy; CI = confidence interval; CMH = Cochran-Mantel-Haenszel; CT = computed tomography; eCRF = electronic case report form; EOC3 = end of cycle 3; IRT = interactive response technology; ITT = intention to treat; MRI = magnetic resonance imaging; SVR = spleen volume reduction.

• a The 2-sided 95% CI was based on the exact Clopper-Pearson method.

• b The stratified P value was one-sided based on CMH test using the Greenland and Robins method to adjust for stratification factors: spleen size by palpation and platelet counts. The third stratification factor ‘refractory/relapsed or intolerance to ruxolitinib treatment’ was dropped due to the small cell count issue.

• c The 95% CI of the difference was based on Greenland and Robins method.

• d The unstratified P value was one-sided based on the z-test with an unpooled estimate of variance.

Source: BMS data on file[36]

The treatment effect was generally consistent across subgroups. A possibly larger treatment effect was observed in the subgroup with platelet counts of 50 to < 100 × 10[9] /L and the subgroup with ruxolitinib intolerance. In the BAT arm, most participants (n = 46) crossed over to fedratinib after EOC6. In participants remaining in the BAT arm, the SVR was at least 23%, as measured at EOC12 and EOC18. The mean percentage change in spleen volume from baseline is shown at every third cycle (see Figure 35).[36]

B.2.6.1.3 Secondary outcomes (key secondary outcomes)

Symptoms response rate

Symptoms response rate is a key secondary outcome, which requires at least 50% reduction in MF-associated symptoms, as measured by MFSAF TSS in the ITT population with non-zero baseline TSS at EOC6. This rate was reported by 43 participants (34.1%) in the fedratinib group and 11 (16.9%) in the BAT group. Superiority for fedratinib emerged from the stratified analyses, the unstratified analyses, and after imputation of missing data, as presented in Table 16.[36]

The completion rate at baseline of the MFSAF questionnaire was 100.0% in both treatment groups. At C6D1, the completion rate was 87.6% in the fedratinib group and 86.0% in the BAT group. Subsequent completion rates showed a steep decline because they included participants who had discontinued the study. At C6D1, the completion rate was 67.5% in the fedratinib group and 75.4% in the BAT group.[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 16. FREEDOM-2: symptom response rate at EOC6 (ITT population with non-zero baseline TSS)

• BAT = best available therapy; CI = confidence interval; CMH = Cochran-Mantel-Haenszel; CRF = case report form; eCRF = electronic case report form; EOC6 = end of cycle 6; IRT = interactive response technology; ITT = intention to treat; TSS = total symptom score.

• a Participants with missing assessment at EOC6, including those who met the criteria for progression of splenomegaly before EOC6, were considered non-responders. However, they were included in the denominator.

• b The 2-sided 95% CI was based on the exact Clopper-Pearson method.

• c The stratified P value was one-sided based on CMH test using the Greenland and Robins method to adjust for stratification factors: spleen size by palpation and platelet counts. The third stratification factor ‘refractory/relapsed or intolerance to ruxolitinib treatment’ was dropped due to the small cell count issue.

• d The 95% CI of the difference was based on the Greenland and Robins method.

• e The unstratified P value was one-sided based on z-test with unpooled estimate of variance.

• f Participants with missing TSS assessment at EOC6 were imputed using multiple imputation methods. Thirty imputed data sets were created. The proportion with 95% CI for each arm and the difference in proportion with 95% CI were calculated for each imputed data set and combined using Rubin’s rules. The number of responders was backcalculated from the pooled adjusted response rate.

Source: BMS data on file[36]

Spleen volume response rate, volume reduction ≥ 25%

The percentage of participants with ≥ 25% SVR at EOC6 was 47.0% in the fedratinib group and 13.4% in the BAT group. Superiority of fedratinib over BAT was shown in both stratified analyses, in the unstratified analysis, and after imputation of missing data (Table 17).[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 17. FREEDOM-2: spleen volume response rate at EOC6 (≥ 25% SVR) by MRI or CT scan (ITT population)

• BAT = best available therapy; CI = confidence interval; CMH = Cochran-Mantel-Haenszel; CRF = case report form; CT = computed tomography; eCRF = electronic case report form; EOC6 = end of cycle 6; IRT = interactive response technology; ITT = intention to treat; MRI = magnetic resonance imaging; SVR = spleen volume reduction.

• a Participants with missing assessment at EOC6, including those who met the criteria for progression of splenomegaly before EOC6, were considered non-responders. However, they were included in the denominator.

• b The 2-sided 95% CI was based on the exact Clopper-Pearson method.

• c The stratified P value was one-sided based on CMH test using the Greenland and Robins method to adjust for stratification factors: spleen size by palpation and platelet counts. The third stratification factor ‘refractory/relapsed or intolerance to ruxolitinib treatment’ was dropped due to the small cell count issue.

• d The 95% CI of the difference was based on the Greenland and Robins method.

• e The unstratified P value was one-sided based on z-test with unpooled estimate of variance.

• f For participants with missing assessment at EOC6, data were imputed using multiple imputation methods. Thirty imputed data sets were created. The proportion with 95% CI for each arm and the difference in proportion with 95% CI were calculated for each imputed data set and combined using Rubin’s rules. The number of responders was backcalculated from the pooled adjusted response rate.

• Source: BMS data on file[36]

B.2.6.1.4 Secondary outcomes (other secondary outcomes)

Spleen response rate by palpation

A spleen response according to the International Working Group-Myeloproliferative Neoplasms Research and Treatment (IWG-MRT) 2013 criteria was found by palpation at EOC6 in 38 participants (28.4%) in the fedratinib group and in 5 (7.7%) in the BAT group. The difference in the percentage was 19.9 (95% CI, 10.0-29.7) (Table 18).[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 18. FREEDOM-2: spleen volume response rate by palpation

BAT = best available therapy; CI = confidence interval; EOC6 = end of cycle 6; LCM = left costal margin.

a In 2 participants in the BAT group and no participant in the fedratinib group, spleen size at baseline was < 5 cm at LCM.

b Two-sided 95% CI was based on the exact Clopper-Pearson method.

c The 95% CI of the difference was based on the Greenland and Robins method.

Source: BMS data on file[36]

Durability of spleen response by MRI/CT scan

In the ITT analysis, based on a median follow-up time of 36.1 (range, 0.1-131.9) weeks for the fedratinib group and 18.65 (range, 0.1-65.1) weeks for the BAT group, a Kaplan-Meier (KM) analysis yielded a median durability of spleen volume response by MRI/CT scan of 86.3 weeks in the fedratinib group. Median durability in the BAT group was not estimable (NE) (ITT analysis). Among 72 participants in the fedratinib group with spleen volume response at any time, an event relevant for this analysis was reported for 19 participants (26.4%). Data for the other 53 participants (73.6%) in the fedratinib group were censored. In the BAT group, 8 participants had a spleen volume response at any time. None of these participants had an event relevant for this analysis so that data for all 8 participants was censored. The durability of the spleen volume response by MRI/CT scan is provided as KM plots in Figure 6.

Figure 6. FREEDOM-2: Kaplan-Meier plot of durability of spleen volume response by MRI/CT scan (ITT population)

==> picture [417 x 266] intentionally omitted <==

BAT = best available therapy; CI = confidence interval; CT = computed tomography; ITT = intention to treat; MRI = magnetic resonance imaging; NE = not estimable.

Source: BMS data on file[48]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Durability of spleen response by palpation

Based on a median follow-up time of 56.10 (range, 3.7-139.4) weeks for the fedratinib group and 21.70 (range, 3.9-60.3) weeks for the BAT group, a KM analysis yielded a median durability of spleen response by palpation of 118.3 (95% CI, 76.0-NE) weeks in the fedratinib group and 47.1 (95% CI, 21.7-NE) weeks in the BAT group (ITT analysis). Among 71 participants in the fedratinib group with spleen response by palpation at any time, an event relevant for this analysis was reported for 19 participants (26.8%). Data for the other 52 participants (73.2%) in the fedratinib group were censored. In the BAT group, 11 participants had a spleen response by palpation at any time. There were 2 participants (18.2%) who had an event relevant for this analysis so that data for the other 9 participants (81.8%) were censored. A KM plot of the durability of the spleen response by palpation is provided in Figure 7.[36,45,48]

Figure 7. FREEDOM-2: Kaplan-Meier plot of durability of spleen response by palpation (ITT population)

==> picture [453 x 289] intentionally omitted <==

BAT = best available therapy; CI = confidence interval; ITT = intention to treat; NE = not estimable. Source: BMS data on file[48]

Durability of symptoms response

Based on a median follow-up time of 11.50 (range, 0.1-102.1) weeks for the fedratinib group and 8.10 (range, 0.1-37.0) weeks for the BAT group, a KM analysis yielded a median durability of the symptom response of 12.1 (95% CI, 8.1-16.1) weeks in the fedratinib group and 10.1 (95% CI, 4.1-16.7) weeks in the BAT group (ITT analysis). Among 90 participants in the fedratinib group with symptoms response at any time, an event relevant for this analysis was reported for 70 participants (77.8%). In the BAT group, 32 participants had symptoms response at any time. There were 27 participants (84.4%) who had events relevant for this analysis. A KM plot of the durability of symptoms response is provided in Figure 8.[36,45,48]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Figure 8. FREEDOM-2: Kaplan-Meier plot of durability of symptom response (ITT population)

==> picture [453 x 284] intentionally omitted <==

BAT = best available therapy; CI = confidence interval; ITT = intention to treat; TSS = total symptom score. Source: BMS data on file[48]

Spleen and disease progression-free survival

Based on a median follow-up time of 46.20 (range, 0.1-148.9) weeks for the fedratinib group and 24.40 (range, 0.1-77.3) weeks for the BAT group, a KM analysis yielded a median time of SDPFS of 112.4 (95% CI, 75.0-NE) weeks in the fedratinib group and NE (95% CI, 30.4-NE) weeks in the BAT group (ITT analysis). Relevant events were reported for 42 participants (31.3%) in the fedratinib group and 12 (17.9%) in the BAT group. Figure 9 presents a KM plot of SDPFS.[36,45,48]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Figure 9. FREEDOM-2: Kaplan-Meier plot of spleen and disease progression-free survival (ITT population)

==> picture [453 x 291] intentionally omitted <==

BAT = best available therapy; CI = confidence interval; ITT = intention to treat; NE = not estimable. Source: BMS data on file[36]

Overall survival

There are clear methodological issues that may result in an underestimate of OS for the fedratinib group, which means that the OS data need to be considered with caution. In the patient population that received BAT, nearly 70% crossed over to fedratinib, leaving very few participants in the BAT-only population, something the study was not powered for. Due to these limitations, crossover adjustments for OS have been explored and are presented in Section B.2.6.1.6.

In FREEDOM-2, based on a median follow-up time of 64.50 (range, 2.3-150.1) weeks for the fedratinib group and 63.70 (range, 1.9-146.0) weeks for the BAT group, a KM analysis yielded a median OS time of NE (95% CI, 112.6-NE) weeks in the fedratinib group and 124.6 (95% CI, 98.9-NE) weeks in the BAT group (ITT analysis). Over the full course of the study (full treatment period + crossover treatment period + follow-up period), death was reported for 43 participants (32.1%) in the fedratinib group and 18 (26.9%) in the BAT group. Figure 10 presents a KM plot of OS.[36,45,48]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Figure 10. FREEDOM-2: Kaplan-Meier plot of overall survival (ITT population)

==> picture [453 x 292] intentionally omitted <==

BAT = best available therapy; CI = confidence interval; ITT = intention to treat; NE = not estimable. Source: BMS data on file[36]

EORTC QLQ-C30

At baseline, the completion of the EORTC QLQ-C30 questionnaires as reflected in the adherence rate was 100.0% in both treatment groups. At C6D1, the adherence rate was 90.5% in the fedratinib group and 89.1% in the BAT group. The completion rates declined more quickly over time due to participants discontinuing the study. At C6D1, the completion rate was 72.4% in the fedratinib group and 82.0% in the BAT group.[36]

At baseline, there was an imbalance in mean EORTC QLQ-C30 scores for most domains, demonstrating slightly better health status for participants in the fedratinib group than in the BAT group in all domains except for diarrhoea and appetite loss. The difference in baseline scores was clinically meaningful for the anaemia-related domains (physical functioning, fatigue, dyspnoea) as well as for the global health status, cognitive functioning, nausea and vomiting, pain, constipation, and financial difficulties, where the difference between the group means reached the threshold for small differences established in Cocks et al.[49] The results at baseline are presented in Table 19.[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 19. FREEDOM-2: EORTC QLQ-C30 results at baseline (HRQOL evaluable population)

BAT = best available therapy; HRQOL = health-related quality of life; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); SD = standard deviation.

Source: BMS data on file[36]

Results by treatment group, cycle, and subscores (function, symptom domain scores) are provided below. Results are summarised for timepoints up to EOC6 due to potential crossover of participants in the BAT group to fedratinib. A ≥ 10-point change, as reported by Osoba et al.,[50] has been commonly used as a threshold to define a meaningful change at a group level (i.e., within-group change and between-group difference). To aid in interpretation, this threshold was used to highlight clinically meaningful change in the mean change from baseline.[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Global health status

In both treatment groups, there were increases from baseline in the mean scores for global health status, indicating improvement, during the treatment period through C6D1. Clinically meaningful change was observed at C3D1 and C4D1 for the fedratinib group and at C2D1 and C3D1 for the BAT group. At C6D1, the mean (SEM) change from baseline was 5.4 (26.02) for the fedratinib group and 9.1 (27.56) for the BAT group (Figure 11). Clinically meaningful HRQOL response by cycle results were supportive of these findings with more than half of the participants in both groups at most timepoints demonstrating clinically meaningful improvement.[36]

Figure 11. FREEDOM-2: EORTC QLQ-C30 Global Health Status/QOL scores (HRQOL evaluable population)

==> picture [453 x 259] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; QOL = quality of life; SEM = standard error of mean.

Note: Horizontal reference lines indicate MID, considered a change of ± 10 points from baseline.

Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Physical functioning

In both treatment groups, there were increases from baseline in the mean scores for physical functioning, indicating improvement, during the treatment period through C6D1. Clinically meaningful change was observed at C3D1 for the fedratinib group and at C6D1 for the BAT group. At C6D1, the mean (SEM) change from baseline was 6.1 (21.22) for the fedratinib group and 10.7 (25.60) for the BAT group (Figure 12). Clinically meaningful HRQOL response results by cycle were supportive of these findings, with clinically meaningful improvement emerging for more than half of the participants in both groups at most timepoints.[36]

Figure 12. FREEDOM-2: EORTC QLQ-C30 Physical Functioning scores (HRQOL evaluable population)

==> picture [453 x 257] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Role functioning

In both treatment groups, there were increases from baseline in the mean scores for role functioning, indicating improvement, during the treatment period through C6D1. Clinically meaningful change was observed at C3D1 and C4D1 for the fedratinib group and at no timepoint for the BAT group. At C6D1, the mean (SEM) change from baseline was 5.3 (31.88) for the fedratinib group and 9.3 (34.97) for the BAT group (Figure 13). Clinically meaningful HRQOL response by cycle results were supportive of these findings with more participants in both groups demonstrating clinically meaningful improvement than either no change or worsening in both groups at most timepoints.[36]

Figure 13. FREEDOM-2: EORTC QLQ-C30, Role Functioning scores (HRQOL evaluable population)

==> picture [453 x 255] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Emotional functioning

In both treatment groups, there were increases from baseline in the mean scores for emotional functioning, indicating improvement, during the treatment period through C6D1. The fedratinib group did not show clinically meaningful change at any timepoint, but the BAT group did at C3D1 and C6D1. At C6D1, the mean (SEM) change from baseline was 8.6 (19.81) for the fedratinib group and 10.2 (23.04) for the BAT group (Figure 14). Clinically meaningful HRQOL response results by cycle were supportive of these findings, with clinically meaningful improvement emerging for more participants in both groups than either no change or worsening at most timepoints.[36]

Figure 14. FREEDOM-2: EORTC QLQ-C30 Emotional Functioning scores (HRQOL evaluable population)

==> picture [453 x 259] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Cognitive functioning

In both treatment groups, there were increases from baseline in the mean scores for cognitive functioning, indicating improvement, during the treatment period through C6D1. However, neither treatment group showed clinically meaningful change at any timepoint. At C6D1, the mean (SEM) change from baseline was 3.1 (18.80) for the fedratinib group and 4.5 (21.09) for the BAT group (Figure 15). Clinically meaningful HRQOL response by cycle results were supportive of these findings, with most participants demonstrating clinically meaningful improvement or no change in both groups at most timepoints.[36]

Figure 15. FREEDOM-2: EORTC QLQ-C30 Cognitive Functioning scores (HRQOL evaluable population)

==> picture [453 x 254] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Social functioning

In both treatment groups, there were initial increases from baseline in the mean scores for social functioning, indicating improvement, in both treatment groups during the treatment period which continued through C6D1. The fedratinib group showed clinically meaningful change at C4D1 and the BAT group had no clinically meaningful change. At C6D1, the mean (SEM) change from baseline was 7.2 (28.72) for the fedratinib group and 2.4 (27.02) for the BAT group (Figure 16). Clinically meaningful HRQOL response results by cycle were supportive of these findings, with most participants demonstrating clinically meaningful improvement or no change in both groups at all timepoints.[36]

Figure 16. FREEDOM-2: EORTC QLQ-C30 Social Functioning scores (HRQOL evaluable population)

==> picture [453 x 253] intentionally omitted <==

BAT = best available therapy; BL = baseline; QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = healthrelated quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Fatigue

In both treatment groups, there were decreases from baseline in the mean scores for fatigue, indicating improvement, during the treatment period through C6D1. Clinically meaningful change was observed at C2D1 through C5D1 for the fedratinib group and at C4D1 through C6D1 for the BAT group. At C6D1, the mean (SEM) change from baseline was −9.6 (28.07) for the fedratinib group and −11.1 (32.53) for the BAT group (Figure 17). Clinically meaningful HRQOL response by cycle results were supportive of these findings with clinically meaningful improvement being demonstrated for more participants than either no change or worsening in both groups at many timepoints.[36]

Figure 17. FREEDOM-2: EORTC QLQ-C30 Fatigue scores (HRQOL evaluable population)

==> picture [453 x 256] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Nausea and vomiting

In both treatment groups, there were mostly decreases from baseline in the mean scores for nausea and vomiting, indicating improvement, in both treatment groups during the treatment period through C6D1. However, neither group showed clinically meaningful change at any timepoint. At C6D1, the mean (SEM) change from baseline was −1.1 (16.18) for the fedratinib group and −7.3 (26.89) for the BAT group (Figure 18). Clinically meaningful HRQOL response by cycle results were supportive of these findings, with most participants in both groups demonstrating no change at all timepoints.[36]

Figure 18. FREEDOM-2: EORTC QLQ-C30 Nausea and Vomiting scores (HRQOL evaluable population)

==> picture [453 x 257] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Pain

In both treatment groups, there were decreases from baseline in the mean scores for pain, indicating improvement, during the treatment period through C6D1. Clinically meaningful change emerged at C2D1 through C5D1 for the fedratinib group and at C2D1 through C4D1 and C6D1 for the BAT group. At C6D1, the mean (SEM) change from baseline was −8.3 (27.15) for the fedratinib group and −11.8 (28.93) for the BAT group (Figure 19). Clinically meaningful HRQOL response by cycle results were supportive of these findings with clinically meaningful improvement being demonstrated in more participants than either no change or worsening in both groups at many timepoints.[36]

Figure 19. FREEDOM-2: EORTC QLQ-C30 Pain scores (HRQOL evaluable population)

==> picture [453 x 254] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Dyspnoea

In both treatment groups, there were decreases from baseline in the mean scores for dyspnoea, indicating improvement, during the treatment period through C6D1. Clinically meaningful change was observed at C3D1 and C4D1 for the fedratinib group and at C3D1 through C5D1 for the BAT group. At C6D1, the mean (SEM) change from baseline was −9.8 (29.90) for the fedratinib group and −5.7 (39.37) for the BAT group (Figure 20). Clinically meaningful HRQOL response by cycle results were supportive of these findings, with most participants demonstrating clinically meaningful improvement or no change in both groups at all timepoints.[36]

Figure 20. FREEDOM-2: EORTC QLQ-C30 Dyspnoea scores (HRQOL evaluable population)

==> picture [453 x 259] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Insomnia

In both treatment groups, there were decreases from baseline in the mean scores for insomnia, indicating improvement, during the treatment period through C6D1. Clinically meaningful change was observed at all timepoints for the fedratinib group and at C3D1 for the BAT group. At C6D1, the mean (SEM) change from baseline was −13.2 (28.84) for the fedratinib group and −5.7 (28.77) for the BAT group (Figure 21). Clinically meaningful HRQOL response by cycle results were supportive of these findings, with most participants demonstrating clinically meaningful improvement or no change in both groups at all timepoints.[36]

Figure 21. FREEDOM-2: EORTC QLQ-C30 Insomnia scores (HRQOL evaluable population)

==> picture [453 x 264] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Appetite loss

In both treatment groups, there were decreases from baseline in the mean scores for appetite loss, indicating improvement, during the treatment period through C6D1. Clinically meaningful change was observed at all timepoints for the fedratinib group and at C3D1 through C6D1 for the BAT group. At C6D1, the mean (SEM) change from baseline was −16.2 (30.55) for the fedratinib group and −17.9 (33.41) for the BAT group (Figure 22). Clinically meaningful HRQOL response by cycle results were supportive of these findings with most participants demonstrating clinically meaningful improvement or no change in both groups at all timepoints.[36]

Figure 22. FREEDOM-2: EORTC QLQ-C30 Appetite Loss scores (HRQOL evaluable population)

==> picture [453 x 259] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Constipation

In the BAT group, there were mostly decreases from baseline in the mean scores for constipation, indicating improvement, during the treatment period through C6D1 whereas the fedratinib group showed little change. However, neither group showed clinically meaningful change at any timepoint. At C6D1, the mean (SEM) change from baseline was −0.4 (27.48) for the fedratinib group and −5.7 (25.71) for the BAT group (Figure 23). Clinically meaningful HRQOL response by cycle results were supportive of these findings with most participants in both groups demonstrating no change at all timepoints.[36]

Figure 23. FREEDOM-2: EORTC QLQ-C30 Constipation scores (HRQOL evaluable population)

==> picture [453 x 254] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Diarrhoea

In both treatment groups, there were both small increases, indicating deterioration, and decreases, indicating improvement, from baseline in the mean scores for diarrhoea at various point during the treatment period through C6D1. However, neither group showed clinically meaningful change at any timepoint. At C6D1, the mean (SEM) change from baseline was 7.9 (24.87) for the fedratinib group and −1.6 (23.51) for the BAT group (Figure 24). Clinically meaningful HRQOL response results by cycle were supportive of these findings, with no change emerging for most of participants in both groups at all timepoints.[36]

Figure 24. FREEDOM-2: EORTC QLQ-C30 Diarrhoea scores (HRQOL evaluable population)

==> picture [453 x 255] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Financial difficulties

In both treatment groups, there were both small increases, indicating deterioration, and decreases, indicating improvement, from baseline in the mean scores for financial difficulties at various points during the treatment period through C6D1. Neither group showed clinically meaningful change at any timepoint. At C6D1, the mean (SEM) change from baseline was 2.2 (18.33) for the fedratinib group and 1.6 (16.59) for the BAT group (Figure 25). Clinically meaningful HRQOL response results by cycle were supportive of these findings, with no change being demonstrated for most participants in both groups at all timepoints.[36]

Figure 25. FREEDOM-2: EORTC QLQ-C30 Financial Difficulties scores (HRQOL evaluable population)

==> picture [453 x 256] intentionally omitted <==

BAT = best available therapy; BL = baseline; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire–Quality of Life of Cancer Patients (Core); EOT = end of treatment; HRQOL = health-related quality of life; MID = minimally important difference; SEM = standard error of mean. Note: Horizontal reference lines indicated MID, considered a change of ± 10 points from baseline. Source: BMS data on file[36]

EQ-5D-5L

At baseline, the completion of the EQ-5D-5L as reflected in the adherence rate was 100.0% in both treatment groups. At C6D1, the adherence rate was 90.4% in the fedratinib group and 87.5% in the BAT group. The completion rates declined more quickly due to inclusion of participants discontinuing from the study, with 72.8% in the fedratinib group and 80.8% in the BAT group. No remarkable differences in the baseline scores between the treatment groups were noted for the EQ-5D-5L visual analogue score and Utility Score. Results are summarised for timepoints up to EOC6 due to potential crossover of participants in the BAT group to fedratinib afterwards. For the visual analogue score, both treatment groups had increases from baseline in the mean scores, indicating improvement, during the treatment period through C6D1. The fedratinib group showed clinically meaningful change at C2D1 through C5D1. The BAT group had no timepoints with clinically meaningful change. At C6D1, the mean (SEM) change from baseline was 6.2 (18.97) for the fedratinib group and 3.5 (23.29) for the BAT group (Figure 26).[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Figure 26. FREEDOM-2: EQ-5D-5L visual analogue scale, mean change from baseline (HRQOL evaluable population)

==> picture [426 x 233] intentionally omitted <==

BAT = best available therapy; BL = baseline; EOT = end of treatment; HRQOL = health-related quality of life; SEM = standard error of mean; VAS = visual analogue score.

Source: BMS data on file[36]

In both treatment groups, there were increases from baseline in the mean scores for the Utility Score, indicating improvement, during the treatment period through C6D1. The fedratinib group showed clinically meaningful change at C2D1 through C5D1, and the BAT group showed clinically meaningful change at C3D1. At C6D1, the mean (SEM) change from baseline was 0.0594 (0.26864) for the fedratinib group and 0.0658 (0.24229) for the BAT group (Figure 27).[36]

Figure 27. FREEDOM-2: EQ-5D-5L utility index, mean change from baseline (HRQOL evaluable population)

==> picture [426 x 239] intentionally omitted <==

BAT = best available therapy; BL = baseline; EOT = end of treatment; HRQOL = health-related quality of life; SEM = standard error of mean.

Source: BMS data on file[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

B.2.6.1.5 Exploratory efficacy outcomes

Time to spleen response by palpation

The median time to spleen response by palpation was 20.3 weeks (95% CI, 12.6-45.9) in the fedratinib group and NE in the BAT group. These results were based on data of 71 participants (53.0%) in the fedratinib group and 11 (16.9%) in the BAT group. In 2 participants in the BAT group, the spleen size was less than 5 cm below at baseline, so that their data were not included in this calculation.[36]

Best spleen volume response

The proportion of participants with best spleen volume response rate during the first 6 treatment cycles (BRR6) of ≥ 35% SVR as well as the proportion of participants with best spleen volume response rate during the full treatment period (BRR) of ≥ 35% SVR was higher in the fedratinib group than in the BAT group (Table 20).[36]

Table 20. FREEDOM-2: BRR6 and BRR by MRI/CT scan (ITT population)

BAT = best available therapy; BRR = best spleen volume response rate during full treatment; BRR6 = best spleen volume response rate during the first 6 treatment cycles; CI = confidence interval; CMH = Cochran-Mantel-Haenszel; CT = computed tomography; eCRF = electronic case report form; ITT = intention to treat; MRI = magnetic resonance imaging.

a The 2-sided 95% CI was based on the exact Clopper-Pearson method.

b The stratified P value was one-sided based on CMH test using the Greenland and Robins method to adjust for stratification factors: spleen size by palpation and platelet counts. The third stratification factor ‘refractory/relapsed or intolerance to ruxolitinib treatment’ was dropped due to small cell count issue.

c The 95% CI of the difference was based on Greenland and Robins method.

Source: BMS data on file[36]

Anaemia response

Anaemia response was found in 20 participants (19.8%) in the fedratinib group and 12 (22.6%) in the BAT group.[36]

B.2.6.1.6 FREEDOM-2: overall survival crossover adjustment

A total of 68.7% of participants crossed over from BAT to fedratinib during FREEDOM-2. In an ITT analysis, the outcomes were analysed according to the intervention a participant was randomly assigned to, regardless of treatment switching. Patient crossover affects the OS analysis (for the ITT analysis), specifically the relative treatment effects, because outcomes in the control arm may be influenced by both the control and experimental treatment after crossover. In cases in which the experimental treatment is more effective than the control, this will result in an underestimation of the

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

true relative treatment effect of the experimental treatment. This can lead to some complex uncertainty. Therefore, crossover adjustment was performed.

A literature review was conducted to identify alternative crossover-adjustment methods. Five alternative OS crossover-adjustment methods were identified, including randomisation-based methods to estimate counterfactual survival (rank-preserving structure failure time [RPSFT] and iterative parameter estimation [IPE]), the 2-stage methods (simplified 2-stage estimation [TSEsimp] and complex 2-stage estimation with g-estimation [TSEgest]), and the propensity-based method (inverse probability of censoring weighting [IPCW]).

Three methodologies were applied for FREEDOM-2 to account for treatment switching from BAT to fedratinib: RPSFT, TSEsimp, and IPCW. Additional approaches (i.e., IPE and TSEgest) were not considered due to challenges with respect to the methods explored. The IPE method assumes that the only difference between randomly assigned groups is the treatment received and a common treatment effect and that the survival times follow a parametric distribution; as such, the IPE method was deemed unsuitable. TSEgest requires a strong assumption of no unmeasured confounding over time and, as such, was deemed unsuitable.[51]

Rank-preserving structure failure time (RPSFT) model

In the RPSFT model, the crossover-adjusted estimates improved the OS for BAT when predicting the counterfactuals had participants in BAT not crossed over to receive fedratinib. This finding was not consistent with the choice of clinicians to switch participants to fedratinib (69% crossed over) and therefore lacks face validity. Additionally, the simplified 2-stage-adjusted results were biased due to the acceleration factors lacking face validity. Finally, the short-term follow-up available in participants in BAT who did not cross over is a major limitation of the IPCW approach. The IPCW-adjusted BAT curve was closely aligned with the naive approach of censoring participants at crossover, suggesting that the weights had a minimal impact to account for informative censoring.[51] The contradictory results and the likely violation of the crossover-adjustment methods are the reasons why none of the estimates from the explored methods are recommended. As such, estimates obtained from the crossover-adjustment methods are not considered reliable in this context.[51] However, the results are presented for transparency.

A KM overlay of OS stratified by crossover status for the BAT arm is presented in Figure 28. Out of the 67 participants randomly assigned to receive BAT, 46 participants (69%) crossed over from BAT to fedratinib. Therefore, there were only 21 participants in BAT arm who did not cross over. Among the participants who crossed over, 43 (93%) did so after 6 cycles whereas the remaining 3 (7%) crossed over after progression, and the average time of treatment switch was 6.8 months (median, 6.3 months; range, 5.5-17.5).[51]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Figure 28. FREEDOM-2: Kaplan-Meier overlay of overall survival for ITT populations (fedratinib and BAT arms), stratified by crossover status

==> picture [447 x 194] intentionally omitted <==

BAT = best available therapy. Source: BMS data on file[51]

RPSFT-adjusted KM OS curves are presented in Figure 29. Typically, the observed OS estimates favour the new intervention versus the control arm, in which case the treatment effect estimated from the experimental arm and applied to the control arm results in OS predictions less favourable than observed estimates (i.e., are shrunken by acceleration factor) to reflect the hypothetical situation where participants in the control arm did not cross over. However, for FREEDOM-2, given that BAT has more favourable OS than fedratinib, the crossover-adjusted estimates improve the OS for BAT when predicting the counterfactuals had participants in BAT not crossed over to receive fedratinib. This finding was not consistent with the choice of clinicians to switch participants to fedratinib (69% crossed over) and therefore lacks face validity.[51]

Figure 29. FREEDOM-2: RPSFT-adjusted overall survival–adjusted Kaplan-Meier curves

==> picture [418 x 222] intentionally omitted <==

BAT = best available therapy; RPSFT = rank-preserving structure failure time. Source: BMS data on file[51]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Simplified 2-stage estimation (TSEsimp)

To estimate the effect of switching on survival, various models were tested (i.e., different accelerated failure time models), all of which resulted in an acceleration factor close to 0 (Table 21). This corresponds with a time ratio of 1:0, suggesting switchers have an infinitely longer survival time. Because we are unable to adequately estimate an acceleration factor that has any face validity, the TSEsimp method is not recommended, and any results may be biased.

Table 21. FREEDOM-2: acceleration factor estimated from different accelerated failure time models

The TSEsimp-adjusted KM OS curves are presented in Figure 30. Note that these should be considered biased results due to the acceleration factors lacking face validity. Applying a near-zero acceleration factor to the post-crossover survival time resulted in truncating survival time in participants who crossed over to the point of treatment switch, driving the observed drop in the TSEsimp-adjusted KM curve (blue grey line) at 6 months. The recensored BAT arm (dark grey line) has a cluster of censored participants within the first month, which is a function of the acceleration factor being near zero, and as such the later timepoint results were driven by relatively few participants.[51]

Figure 30. FREEDOM-2: TSEsimp-adjusted overall survival–adjusted Kaplan-Meier curves

==> picture [469 x 224] intentionally omitted <==

BAT = best available therapy; TSEsimp = simplified 2-stage estimation. Source: BMS data on file[51]

Inverse probability of censoring weighting (IPCW)

The distribution of weights estimated for the IPCW method was considered reasonable (median weight, 0.99; range, 0.46-2.19). However, the short-term follow-up available in participants in the BAT arm who

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

did not cross over (n = 21) is a major limitation of the IPCW approach. To implement the method, switchers were censored at time of crossover, and the remaining observations were weighted to account for bias introduced by censoring. However, most of the crossover happened around 6 months, at which point there were only 14 BAT participants who did not cross over who were remaining at risk of event. Therefore, any longer-term outcomes were driven by relatively few patients, which may introduce bias. Additionally, the lack of events (i.e., deaths) after 12 months suggests that any weighting performed would likely have minimal effect on long-term results because reweighting all censored participants cannot account for the curve flattening that occurs after 12 months. As such, the IPCW method is not recommended. Figure 31 presents the IPCW-adjusted KM OS curves. Of note, the IPCW-adjusted BAT curve was closely aligned with the naive approach of censoring participants at crossover, suggesting that the weights had minimal effect to account for informative censoring.[51]

Figure 31. FREEDOM-2: IPCW-adjusted overall survival–adjusted Kaplan-Meier curves

==> picture [469 x 202] intentionally omitted <==

BAT = best available therapy; IPCW = inverse probability of censoring weighting. Source: BMS data on file[51]

Figure 32 presents the IPCW-adjusted KM OS curves from a sensitivity analysis assuming that there was a zero probability of switching before cycle 6. When excluding observations before cycle 6 in the weight calculation, the IPCW-adjusted KM curve (dark grey line) shifted more than in the base-case analysis, but the results were still largely in line with the naive-censored approach (pink line). The weights remained reasonably distributed and concentrated around 1 (mean, 0.99; median, 1.00 [range, 0.43-3.25]).

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Figure 32. FREEDOM-2: IPCW-adjusted overall survival–adjusted Kaplan-Meier curves (sensitivity analysis assuming probability of switching before cycle 6 was zero)

==> picture [469 x 204] intentionally omitted <==

BAT = best available therapy; IPCW = inverse probability of censoring weighting.

Conclusion of overall survival adjustment

When applied to FREEDOM-2, the RPSFT, TSEsimp, and IPCW methods all provided contradictory results. Of more importance, the underlying assumptions of the crossover-adjustment methods were likely violated. Overall, none of the estimates from the explored methods are recommended. Some of the trial characteristics of FREEDOM-2 contributed to the violations of the underlying assumptions of these methods. As such, estimates obtained from the crossover-adjustment methods are not considered reliable in this context, for instance, not having a universal cutoff applicable to all participants in order to accurately predict switching (in FREEDOM-2, crossover could occur either at postprogression [3 of 46 patients] or after cycle 6 [43 of 46 patients]). Also, the availability of data plays a crucial part, as a small sample size will limit the inclusion of feasible covariates. Because FREEDOM-2 was not powered for either OS or subgroup analysis, this results in an even lower sample size for the crossover analysis. Furthermore, a higher proportion of participants in the fedratinib arm received at least 1 prior systemic anticancer therapy (20.1%) compared with participants in the BAT arm (10.4%). Participants who received more than 4 prior anticancer therapies (5.2% in fedratinib, 0 in BAT) are presumably sicker compared with participants with less than 4 prior anticancer therapies. As such, OS estimates would favour the healthier population in the BAT arm.[51]

B.2.6.2 SACT data set

B.2.6.2.1 Primary outcome: treatment duration

Of the 54 patients included in the SACT data set analysis, 27 (50%) were classified as having completed treatment by 31 October 2022 (latest follow-up in SACT data set). Patients are assumed to have completed treatment if they have died, have an outcome summary recorded in the SACT data set, or they have not received treatment with fedratinib in at least 3 months. The median follow-up time in the SACT data set was 4.6 months (140 days). The median follow-up time in the SACT data set was the patients’ median observed time from the start of their treatment to their last treatment date in SACT plus the prescription length. Due to differences in data submission procedures, the maximum follow-up period was either 11 months (applied to 94% of patients, n = 132) or 12 months (applied to 6%, n = 9).[46] Table 22 presents treatment status, and Table 23 presents treatment duration.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 22. SACT data set: treatment status

SACT = Systemic Anti-Cancer Therapy. Source: NHS England data on file[46]

Table 23. SACT data set: treatment duration at 6- and 12-month intervals

CI = confidence interval; SACT = Systemic Anti-Cancer Therapy.

Source: NHS England data on file[46]

Figure 33 presents the treatment duration estimate as a KM curve. The median treatment duration for all patients was 5.7 months (95% CI, 3.9-9.7 months) (173 days).[46]

Figure 33. SACT data set: Kaplan-Meier treatment duration (n = 54)

==> picture [361 x 242] intentionally omitted <==

SACT = Systemic Anti-Cancer Therapy. Source: NHS England data on file[46]

Table 24 presents the outcomes recorded for patients who ended treatment.[46]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 24. SACT data set: treatment outcomes for patients who have ended treatment (n = 27)

SACT = Systemic Anti-Cancer Therapy. Source: NHS England data on file[46]

B.2.6.2.2 Primary outcome: overall survival

Of the 54 patients included in the initial SACT data set analysis, the minimum follow-up was 3.4 months (103 days) from the last CDF application. Patients were traced for their vital status on 13 February 2023. This date was used as the follow-up date (censored date) if a patient was still alive. The median follow-up time in the SACT data set was 7.5 months (228 days). Median follow-up was the patient’s median observed time from the start of their treatment to death or censored date. A reassessment of vital status was performed on 5 February 2024; the median follow-up time was then 15.5 months (471 days).[46] Table 25 presents OS at 6, 12, 18, and 24 months.

Table 25. SACT data set: overall survival at 6-, 12-, 18-, and 24-month intervals

CI = confidence interval; SACT = Systemic Anti-Cancer Therapy. Source: NHS England data on file[46]

Figure 34 presents the OS estimate as a KM curve, censored on 5 February 2024. The median OS was 15.4 months (468 days).[46]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

==> picture [397 x 276] intentionally omitted <==

----- Start of picture text -----
Figure 34. SACT data set: Kaplan-Meier survival plot (n = 54)
Kaplan Meier survival estimate
0 3 6 9 12 15 18 21 24 27
Survival in months
1.00
0.75
0.50
0.25
0.00
----- End of picture text -----

SACT = Systemic Anti-Cancer Therapy. Source: NHS England data on file[46]

B.2.7 Subgroup analysis

B.2.7.1 FREEDOM-2

Subgroup analyses were carried out on the primary outcome (spleen response rate, ≥ 35% SVR at EOC6) to determine the treatment effect of fedratinib on clinically important subpopulations. Subpopulations were by baseline demographic and disease characteristics, platelet and haemoglobin count, and Eastern Cooperative Oncology Group (ECOG) performance status (PS).[36]

The treatment effect was consistent across subgroups (Figure 35), where fedratinib showed clinical benefit on spleen response rate, supporting the robustness of the results of the primary analysis. A possibly larger treatment effect was observed in the subgroup with platelet counts of 50 to < 100 × 10[9] /L and the subgroup with ruxolitinib intolerance.[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Figure 35. FREEDOM-2: ≥ 35% SVR at EOC6, forest plot (ITT population)

==> picture [452 x 241] intentionally omitted <==

==> picture [452 x 238] intentionally omitted <==

BAT = best available therapy; CI = confidence interval; ECOG = Eastern Cooperative Oncology Group; EOC6 = end of cycle 6; ITT = intention to treat; JAK2 = Janus kinase 2; LCM = left costal margin; PMF = primary myelofibrosis; postET MF = post-essential thrombocythemia myelofibrosis; post-PV MF = post-polycythemia vera myelofibrosis; RBC = red blood cell; SVR = spleen volume reduction; ULN = upper limit of normal.

Source: BMS data on file[36]

B.2.8 Meta-analysis

B.2.8.1 FREEDOM-2

As a single study with a head-to-head comparison of BAT, data for fedratinib are provided for patients treated with ruxolitinib. Therefore, meta-analysis of intervention studies is not required.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

B.2.9 Indirect and mixed treatment comparisons

B.2.9.1 FREEDOM-2

Because FREEDOM-2 is a head-to-head study including a comparison with BAT, no indirect comparisons or mixed treatment comparisons are included in this submission for FREEDOM-2.

B.2.10 Adverse reactions

B.2.10.1 Treatment exposure

B.2.10.1.1 FREEDOM-2

The median number of treatment cycles was 11 (range, 1.0-38) in the fedratinib all-treated arm, 10 (range, 1-26) in the crossover arm, and 7 (range, 1-23) in the BAT arm. The mean duration of treatment exposure was 52.5 weeks (standard deviation [SD], 39.68) in the all-treated fedratinib arm, 41.7 weeks (SD, 29.95) in the crossover arm, and 27.7 weeks (SD, 15.29) in the BAT arm. Treatment discontinuation in the fedratinib arm before the completion of 6 cycles was most frequently due to AEs in 10 participants (7.5%), followed by the decision of the physician in 7 participants (5.2%). In the BAT arm, the most frequent reasons for treatment discontinuation before the end of 6 cycles were AEs and participant decision, with each having 3 participants (4.5%).[36] In the fedratinib all-treated and crossover population, approximately 50% of participants completed 12 cycles, and only 10% in the BAT arm completed 12 cycles. Relative dose intensity (RDI) was 86.9% and 90.5% in all-treated and crossover fedratinib arms, respectively, suggesting overall tolerability. Table 26 summarises treatment exposure in FREEDOM-2.

Table 26. FREEDOM-2: fedratinib exposure (all-treated population and crossover population)

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

BAT = best available therapy; C1D1 = cycle 1, day 1; max = maximum; min = minimum; NA = not applicable; SD = standard deviation.

a Treatment duration = (last dose date – first dose date + 1) ÷ 7, regardless of unplanned intermittent discontinuations. For participants in BAT group, C1D1 is treated as first dose date. For last dose date (treatment end date), treatment discontinuation date is used if participant discontinues the study treatment without crossover; (crossover date-1) is used if participant crosses over to fedratinib group; for ongoing participants, data cutoff date is used.

b Actual dose intensity = cumulative dose ÷ treatment duration.

c Relative dose intensity = actual dose intensity ÷ planned dose Intensity (of 2,800 mg/week), presented as a percentage. Source: BMS data on file[45]

B.2.10.2 Summary of safety data

In FREEDOM-2, treatment-emergent AEs (TEAEs) included any AEs with onset or worsening of grade between the date of first dose and 30 days after the date of last dose. At least 1 TEAE was experienced in more than 97% of the fedratinib all-treated and crossover populations and the BAT arm. Grade 3 or 4 TEAEs were reported in 76.9%, 67.4%, and 55.2% of participants in the fedratinib all-treated and crossover populations and BAT arm, respectively.[45] Table 27 presents an overview of the TEAEs associated with fedratinib in FREEDOM-2.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 27. FREEDOM-2: safety overview (all-treated fedratinib, BAT, and crossover fedratinib populations)

BAT = best available therapy; NR = not reported; SAE = serious adverse event; TEAE = treatment-emergent adverse event.

Note: For the fedratinib group, only data for participants who were initially treated with fedratinib are summarised. For crossover participants in the BAT arm, only data before crossover are included.

a No. of events refers to number of participants with at least 1 TEAE or 1 SAE.

Source: FREEDOM-2 CSR[45]

B.2.10.3 Common adverse event data

B.2.10.3.1 FREEDOM-2

The most common TEAEs by study arm were (number of participants [%])[45] :

• Fedratinib all-treated: diarrhoea (62 [46.3%]), anaemia (59 [44%]), nausea (54 [40.3%])

• Fedratinib crossover: anaemia (20 [43.5%]), diarrhoea (17 [37%]), thrombocytopenia (13 [28.3%])

• BAT all-treated: anaemia (24 [35.8%]), asthenia (16 [23.9%]), thrombocytopenia (12 [17.9%])

The most common treatment-related TEAEs by study arm were (number of participants [%])[45] :

• Fedratinib all-treated: diarrhoea (54 [40.3%]), nausea (45 [33.6%]), thrombocytopenia (32 [23.9%])

• Fedratinib crossover: diarrhoea (15 [32.6%]), anaemia (14 [30.4%]), nausea (11 [23.9%])

• BAT all-treated: anaemia (9 [13.4%]), thrombocytopenia (4 [6%]), constipation (3 [4.5%])

A summary of the common AEs reported in FREEDOM-2 is presented in Table 28 for fedratinib alltreated fedratinib and BAT populations and Table 29 for the fedratinib crossover population.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 28. FREEDOM-2: common treatment-emergent and treatment-related adverse events (all-treated population)

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

AE = adverse event; ALT = alanine aminotransferase; AST = aspartate aminotransferase; BAT = best available therapy; NR = not reported; TEAE = treatment-emergent adverse event.

Notes: Shown are any AEs with onset or worsening in grade between the date of the first dose and 30 days after the date of last dose occurring in more than 5% of participants in either arm.

For the fedratinib arm, only participants who initially were randomly assigned to fedratinib are included. For crossover participants in BAT arm, only data before crossover are included.

Source: BMS data on file[45]

Table 29. FREEDOM-2: common treatment-emergent and treatment-related adverse events (crossover fedratinib population)

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

AE = adverse event; NR = not reported; TEAE = treatment-emergent adverse event.

Note: Shown are any AEs with onset or worsening in grade between the date of the first dose and 30 days after the date of last dose occurring in more than 5% of participants.

Source: BMS data on file[45]

B.2.10.4 Treatment-emergent serious adverse events

B.2.10.4.1 FREEDOM-2

Treatment-emergent serious AEs (SAEs) were reported in 72 participants (53.7%) in the fedratinib alltreated population, 21 (31.3%) in the BAT all-treated population, and 16 (34.8%) in the fedratinib crossover arm. The most common SAE was infections and infestations, reported in 17 participants (12.7%) in the fedratinib all-treated arm and 10 (14.9%) in the BAT all-treated arm. For the fedratinib crossover population, the most common SAE was also infections and infestations, reported in 9 participants (19.6%).

Treatment-related SAEs were reported in 25 participants (18.7%) in the fedratinib all-treated population, 2 (3%) in the BAT all-treated population, and 5 (10.9%) in the fedratinib crossover arm. The most common SAE was renal and urinary disorders, reported in 8 participants (6%) in the fedratinib alltreated population. The 2 SAEs in the BAT arm reported were a respiratory, thoracic, and mediastinal disorder (pleural effusion) and an infection and infestation (lymph node tuberculosis). For the fedratinib crossover population, the most common SAE was infections and infestations, reported in only 2 participants (4.3%).

A summary of treatment-emergent and treatment-related SAEs is presented in Table 30 for all-treated fedratinib and BAT populations and in Table 31 for the fedratinib crossover population.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 30. FREEDOM-2: treatment-emergent and treatment-related SAEs (all-treated population)

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

AE = adverse event; ALT = alanine aminotransferase; AST = aspartate aminotransferase; BAT = best available therapy; NR = not reported; SAE = serious adverse event.

Note: Shown are any AEs with onset or worsening in grade between the date of the first dose and 30 days after the date of last dose.

Source: BMS data on file[45]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 31. FREEDOM-2: treatment-emergent and treatment-related SAEs (crossover fedratinib population)

AE = adverse event; NR = not reported; SAE = serious adverse event.

Note: Shown are any AEs with onset or worsening in grade between the date of the first dose and 30 days after the date of last dose.

Source: BMS data on file[45]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

B.2.10.5 Adverse events leading to treatment discontinuation

B.2.10.5.1 FREEDOM-2

Treatment-emergent AEs leading to permanent treatment discontinuation occurred in 24 participants (17.9%) in the fedratinib population, 4 (6%) in the BAT all-treated population, and 7 (15.2%) in the fedratinib crossover population. The most common reason for treatment discontinuation for the fedratinib all-treated population was blood and lymphatic system disorders, which occurred in 5 participants (3.7%), followed by renal and urinary disorders in 4 participants (3%). In the BAT alltreated arm, the most common reason for treatment discontinuation was infections and infestations, which occurred in 2 participants (3%). The most common reasons for treatment discontinuation in the fedratinib crossover population were blood and lymphatic system disorders and infections and infestations, which occurred in 2 participants (4.3%) in each of those System Organ Classes. Table 32 summarises TEAEs leading to permanent treatment discontinuation in FREEDOM-2.

Table 32. FREEDOM-2: TEAEs leading to treatment discontinuation (all-treated and crossover population)

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

AE = adverse event; AML = acute myeloid leukaemia; BAT = best available therapy; MedDRA = Medical Dictionary for Regulatory Activities; NR = not reported; TEAE = treatment-emergent adverse event.

Note: Shown are any AEs with onset or worsening in grade between the date of the first dose and 30 days after the date of last dose.

• a System Organ Class and Preferred Terms were coded using the MedDRA version 25.1. If multiple TEAEs were reported within a given Preferred Term, only 1 event was counted per participant. The table is sorted by decreasing frequency of System Organ Class and Preferred Term in the all-grades column of TEAEs (without consideration of relatedness).

Source: BMS data on file[45]

Table 33. FREEDOM-2: TEAEs leading to dose reduction and interruption (all-treated population)

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

AE = adverse event; ALT = alanine aminotransferase; AST = aspartate aminotransferase; BAT = best available therapy; NR = not reported; TEAE = treatment-emergent adverse event.

Note: Shown are any AEs with onset or worsening in grade between the date of the first dose and 30 days after the date of last dose.

Source: BMS data on file[45]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 34. FREEDOM-2: TEAEs leading to dose reduction and interruption (crossover fedratinib population)

AE = adverse event; NR = not reported; TEAE = treatment-emergent adverse event.

Note: Shown are any AEs with onset or worsening in grade between the date of the first dose and 30 days after the date of last dose.

Source: BMS data on file[45]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

B.2.10.6 Adverse events leading to death

In FREEDOM-2, a total of 21 participants (15.7%) on fedratinib, 4 (6%) in the BAT arm, and 4 (8.7%) who crossed over to fedratinib from BAT died during the treatment course.[36] Table 35 summarises TEAEs leading to death.

Table 35. FREEDOM-2: TEAEs leading to death (all-treated population and crossover population)

AE = adverse event; BAT = best available therapy; MedDRA = Medical Dictionary for Regulatory Activities; NR = not reported; TEAE = treatment-emergent adverse event.

Notes: Shown are any AEs with onset or worsening in grade between the date of the first dose and 30 days after the date of last dose.

The table is sorted by decreasing frequency of System Organ Class and Preferred Term.

a System Organ Classes and Preferred Terms were coded using the MedDRA version 25.1. If multiple TEAEs were reported within a given Preferred Term, only 1 event was counted per participant.

Source: FREEDOM-2 CSR[45]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

B.2.10.7 Safety overview

B.2.10.7.1 FREEDOM-2

The safety results of FREEDOM-2 were also aligned with the known profile of fedratinib. The mitigation strategies for gastrointestinal and thiamine-related toxicities were generally effective.[36] Within the first 6 treatment cycles, there were similar rates of at least 1 TEAE each reported in the fedratinib and BAT treatment arms. Grade 3 or 4 TEAEs related to the study drug were reported for 38.8% of participants in the fedratinib arm and 11.9% in the BAT arm. Permanent discontinuation of the study drug resulted from TEAEs in 9.7% of participants in the fedratinib arm and 6.0% in the BAT arm.[36]

During the full treatment period, 98.5% of participants in the fedratinib arm and 97.0% in the BAT arm had at least 1 TEAE. Grade 3 or 4 TEAEs related to the study drug were reported for 46.3% of participants in the fedratinib arm and 14.9% in the BAT arm. Permanent discontinuation of the study drug resulted from TEAEs in 17.9% of participants in the fedratinib arm and 6.0% in the BAT arm.[36]

B.2.11 Ongoing studies

The phase 3 FREEDOM-2 study of fedratinib compared with BAT in participants with DIPSS intermediate or high-risk PMF, post-PV MF, or post-ET MF who were previously treated with ruxolitinib is currently ongoing but no longer recruiting. The primary clinical study report was finalised in August 2023. The estimated completion date is June 2025.

B.2.12 Interpretation of clinical effectiveness and safety evidence

FREEDOM-2 demonstrated that treatment with fedratinib, compared with BAT, is associated with considerable reductions in spleen volume and size, as well as marked improvements to symptoms in individuals previously treated with ruxolitinib.

Splenomegaly is the key physical feature and cause of symptoms of myelofibrosis; as such, SVR is an important treatment goal. Internationally recognised research groups have identified ≥ 35% SVR as the appropriate threshold for defining response in patients with myelofibrosis.[52 ] In FREEDOM-2, just over one-third of participants (35.8%) on fedratinib in FREEDOM-2 achieved this response versus 6% of participants on BAT.[36,53]

In lieu of availability of curative treatments, the relief of debilitating symptoms is another important treatment goal in myelofibrosis. The clinically meaningful threshold for symptom response is ≥ 50% reduction in TSS,[52 ] with 34.1% of participants receiving fedratinib in FREEDOM-2 having achieved this versus 16.9% of participants on BAT.[36,53] Alleviating these symptoms provides patients with an improved ability to carry out normal daily functions and relieves some of the physical and psychological burden associated with myelofibrosis.

Furthermore, the FREEDOM-2 subgroup analysis shows that treatment effects are consistent across all subgroups, such as baseline demographics, disease characteristics, platelet count, haemoglobin count, and ECOG PS.

Over the full course of the FREEDOM-2 study (full treatment period + crossover treatment period + follow-up period), death was reported for 43 participants (32.1%) in the fedratinib group and 18 (26.9%) in the BAT group.[36,45,48] There are clear methodological issues that may result in an underestimate of OS for the fedratinib group, which means that the OS data need to be considered with caution. In the participant population that received BAT, nearly 70% crossed over to fedratinib, leaving few participants in the BAT-only population, something the study was not powered for. Crossover analyses were attempted; however, none of the methods were suitable because they provided contradictory results

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

and violated underlying assumptions. Therefore, the results of the crossover analysis are not considered reliable.

The proposed position of fedratinib in the treatment pathway is narrower than the marketing authorisation because the population of patients who have been treated with ruxolitinib represents the greatest unmet need in myelofibrosis for which the clinical and cost-effectiveness of fedratinib is most demonstrable. The survival outcomes in patients who have been treated with ruxolitinib are poor, with studies indicating a median OS of 13 to 16 months after ruxolitinib discontinuation.[10,18-20] Furthermore, should fedratinib be included in the current treatment landscape in NHS England, additional treatment options would be available for the heterogenous patient population with myelofibrosis.

B.3 Cost-effectiveness

B.3.1 Published cost-effectiveness studies

An SLR was performed for TA756, to identify published cost-effectiveness studies in myelofibrosis to support the development of a de novo economic model for fedratinib. Since the submission of TA756, no other pharmaceuticals have been approved by NICE for this indication. Therefore, no further evidence is anticipated to be found other than the pivotal trial FREEDOM-2 (providing head-to-head data); thus, an update of the SLR would not impact this submission. The search strategy and study selection criteria are described in detail in Appendix G.

In total, 1,126 potentially relevant articles were identified in database searches. After exclusion of irrelevant articles (n = 1,120) and the addition of relevant articles from bibliographic (n = 1) and health technology assessment (HTA) (n = 8) searches, a total of 15 publications were included. Because some studies were associated with multiple publications, secondary publications were combined; this resulted in inclusion of 9 studies identified from 15 publications.[10,54-61] A summary of the 9 cost-effectiveness studies identified by the SLR are presented in Appendix G.

B.3.2 Economic analysis

The SLR included 9 studies from 15 publications that investigated the cost-effectiveness of therapies in patients with myelofibrosis. All 9 studies assessed the cost-effectiveness of ruxolitinib relative to either BAT or placebo. The studies included 5 ruxolitinib HTA submission documents:

• Canada: Canadian Agency for Drugs and Technologies in Health (CADTH), 2013[59]

• Ireland: National Centre for Pharmacoeconomics (NCPE), 2013[60]

• England and Wales: National Institute for Health and Care Excellence (NICE), 2016[10]

• Scotland: Scottish Medicines Consortium (SMC), 2015[58]

• Australia: Pharmaceutical Benefits Advisory Committee (PBAC), 2015[57]

The remaining 4 studies were published in peer-reviewed journals.[54-56,61]

Where reported, alive health states were often defined by the treatment received (n = 3).[10,56,58] This tended to consist of ruxolitinib, BAT, or supportive care. Alternatively, response or non-response were used to define health states (n = 3).[57,60,61] One study included leukaemic transformation as a health state (n = 1).[61] The omission of AML as a distinct health state was queried by the evidence review group in NICE TA386 (ruxolitinib for treating disease-related splenomegaly or symptoms in adults with myelofibrosis).[10] Three studies did not explicitly report health states (n = 3).[54,55,59]

Where reported, effectiveness outcomes were informed by 1 or more of 3 studies: COMFORT-I,[62] COMFORT-II,[63] and NCT00509899.[64] The primary outcomes of COMFORT-I and COMFORT-II were

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

≥ 35% reduction in spleen volume from baseline at 24 weeks and 48 weeks, respectively. COMFORT-I also investigated symptom response, as assessed by the TSS of the modified MFSAF v2.0. NCT00509899 measured the proportion of participants with ≥ 35% reduction in spleen volume from baseline at time intervals up to 48 weeks, and the change in TSS from baseline at 24 weeks. In the economic models, 6 studies used one or both of the COMFORT-I and COMFORT-II studies to inform treatment response.[10,55-58,60] One study used both COMFORT-II and NCT00509899, but did not leverage the symptom score data.[61] Another study used NCT00509899 trial data alone and considered both spleen volume and symptom response to produce their economic recommendations.[59] One study did not report the data source used.[54]

Cohort models were commonly applied (n = 5).[54-56,60,61] Patient-level discrete-event simulation (DES) was also leveraged (n = 2). Two studies did not report the model type (n = 2).

A single submission to NICE was made in myelofibrosis (TA386[10] ) and a DES modelling approach was used to estimate cost-effectiveness of ruxolitinib. In line with the modelling approach used in NICE TA386[10] and in the previous submission of fedratinib (NICE TA756), this submission takes a similar DES modelling approach to evaluate the cost-effectiveness of fedratinib in patients with myelofibrosis who have been treated with ruxolitinib.

B.3.2.1 Patient population

The main population in the economic analysis comprises adults with disease-related splenomegaly caused by PMF, post-PV MF, or post-ET MF who have been treated with ruxolitinib and are classified as intermediate-2 or high risk by DIPSS.

The data used for this analysis were derived from the FREEDOM-2 clinical trial. FREEDOM-2 is an ongoing, phase 3, multicentre, randomised, 2-arm, open-label trial that included participants with intermediate-2 or high-risk PMF, post-PV MF, or post-essential thrombocythaemia myelofibrosis previously treated with ruxolitinib.[36] A top-level summary of the trial is given in Table 36. All analyses were performed using the 27 December 2022 data cut for FREEDOM-2.

Table 36. FREEDOM-2: summary of characteristics

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

BAT = best available therapy; CT = computed tomography; DIPSS = Dynamic International Prognostic Scoring System; ITT = intention to treat; LCM = left costal margin; MRI = magnetic resonance imaging; SVR = spleen volume reduction.

B.3.2.1.1 Patient characteristics

The demographics and baseline disease characteristics in FREEDOM-2 are representative of a group of people with advanced myelofibrosis and a high disease burden, with 16.9% of participants having received prior anticancer therapies other than ruxolitinib for myelofibrosis. All participants in FREEDOM-2 had received prior treatment with ruxolitinib over at least 3 months.[36]

There were no notable differences between the characteristics of participants receiving BAT and those of the participants receiving fedratinib, with the median age being 70.0 years (range, 40-86) in the fedratinib arm and 68.0 years (range, 38-91) for the BAT arm. Male participants comprised 56.0% of the fedratinib arm and 44.8% of the BAT arm. Finally, height and weight were similar across both arms; participants in fedratinib had a median height of 168.0 cm (range, 146.2-192.0) compared with 165.0 cm (range, 144.0-191.0) in the BAT arm. Their median weight was 71.70 kg (range, 43.0-112.2) for the fedratinib arm and 66.20 kg (range, 45.5-108.0) for the BAT arm.

Table 37 presents the baseline characteristics in FREEDOM-2 used in the cost-effectiveness model.

Table 37. FREEDOM-2: baseline characteristics

SD = standard deviation.

In the economic model, the median is used for age, height, and weight; it is considered to be closer to the true average value of a sample than the mean, which could be skewed because of outliers. Considering that the total median value was not available from FREEDOM-2, a weighted mean and median parameters based on both treatment arms has been calculated to be inputted into the model. The values currently used in the economic model are reported in Table 38.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 38. Patient characteristics included in the economic model

CSR = clinical study report.

B.3.2.2 Model structure

B.3.2.2.1 Model type

The cost-effectiveness analysis for fedratinib was conducted using a DES model built in Excel.

DES modelling, in line with the approach taken in TA756 and TA386, allow for individual patient pathways to be estimated by sampling directly from time-to-event curves. When more than 1 event may occur (such as treatment discontinuation or death), event times for each event are sampled, and the event with the lowest sampled time is simulated as occurring next. An advantage of the DES approach is that it does not impose assumptions that force events to only occur at defined intervals known as “time cycles,” which is the standard approach in cohort-based models and many patient-level simulations. This model type also enables “memory,” meaning a patient’s experience of a treatment pathway (both their previous health states or treatments received along with time in their current health state) is recorded for accurate calculations of costs, utilities, and transitions to future health states. Based on the feedback from the External Assessment Group (EAG) in NICE TA756, the costeffectiveness model was further updated with a supportive care health state and worsening health utility over time, as well as the replacement of the palliative health state with a one-off end-of-life cost.

DES modelling allows enhanced tracking of patients, enables memory to be implemented in the model, and provides flexibility in dealing with transitions. As such, the DES structure allows users to track the patients’ time in health states and implement the health utilities for responders after 4 weeks in the health state, which is one of the primary endpoints of the FREEDOM-2 clinical trial. In addition, use of a DES structure is more accurate than a simpler area-under-the-curve model because it allows for estimation of the precise time of events (as opposed to assuming that they occur at discrete time cycles) as well as capturing patient heterogeneity in these outcomes. A potential drawback of DES models is that they can be computationally intensive, but this is not the case for the adapted model. Finally, because of the possibility for the model to stratify between responders and non-responders, a decision was made to keep the initial DES model structure to accurately reflect the progression of the disease.

B.3.2.2.2 Model structure

Figure 36 shows the model structure. The health states in the model are divided into 4 categories: treatment states, response assessment states, progressed states, and end-of-life states. In the treatment states, patients receive either fedratinib or BAT. Patients accrue costs according to the treatment received and accrue quality-adjusted life-years (QALYs) in line with their response to treatment. Patients enter the response assessment state after 24 weeks of treatment with either fedratinib or BAT. Patients who discontinue treatment or die before reaching this state are labelled as “early discontinuation” or “early death.” In the assessment state, patients undergo an instantaneous response assessment. The potential definitions of response used in the model are:

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

• Spleen volume reduction (SVR): ≥ 35% SVR from baseline at 24 weeks

• Symptom response (TSS): ≥ 50% TSS reduction from baseline at 24 weeks

• Spleen or symptom response: ≥ 35% spleen volume or ≥ 50% TSS reduction from baseline at 24 weeks

Further details on the response assessments are given in Section B.3.3. For both fedratinib and BAT, time to discontinuation beyond 24 weeks is estimated separately for responders and non-responders.

Figure 36. Model structure for fedratinib in myelofibrosis

==> picture [483 x 263] intentionally omitted <==

BAT = best available therapy.

B.3.2.2.3 Model implementation

The model includes the explanation sheets (for model structure, model layout in Excel, background of the disease area), inputs and result sheets, as well as engine sheets. The user can set the settings of the model (i.e., time horizon, discounting, sources used in the model for survival, costs, and health utilities) in the ‘Control’ sheet. Most of the calculations are programmed in Visual Basic for Applications (VBA), which enables faster calculations. The VBA code contains several modules clearly labelled, with descriptions of the purpose and functioning of the code. The modelling methodology emulates from the approach taken in the previous technology appraisal of ruxolitinib (TA386)[10] , the first JAK inhibitor approved for myelofibrosis, and remains the same as the modelling approach from the previous fedratinib in myelofibrosis NICE appraisal (TA756).[65]

Table 39 summarises the different events in the model and their descriptions.

Table 39. Implementation of events in the model

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

BAT = best available therapy; JAK = Janus kinase; OS = overall survival; TTD = time to treatment discontinuation.

B.3.2.2.4 Model features

Key health economic outputs in line with NICE’s reference case are included in the model. These include discounted and undiscounted costs, life-years (LYs), and QALYs as totals and as values that are disaggregated by health state. Costs are also presented by component (e.g., costs for drug acquisition, administration, resource utilisation, and AEs).

To assess cost-effectiveness, pairwise incremental cost-effectiveness ratios (ICERs) are presented. Additionally, incremental net monetary benefit, with a user-amendable willingness-to-pay (WTP) threshold, is also calculated by the model.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

The model has been developed from the perspective of the NHS and, as such, only considers direct costs in the base case.

Because the model outcomes focus on survival, a lifetime horizon was adopted. In the base case, the model considers a 30-year time horizon based on the median age in FREEDOM-2 of 69.3 years old. Shorter time horizons can be explored on scenario analysis.

In the base case, the model considers a 3.5% discount rate for costs and health effects in line with the NICE methods guide.[66]

Table 40 presents features of the economic analysis.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 40. Features of the economic analysis

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115] © BMS (2024). All rights reserved

AE = adverse event; AML = acute myeloid leukaemia; BAT = best available therapy; BNF = British National Formulary; eMIT = electronic market information tool; HRQOL = healthrelated quality of life; JAK = Janus kinase; MF-8D = Myelofibrosis 8 Dimensions; MIMS = Monthly Index of Medical Specialities; NHS = National Health Service; PSSRU = Personal Social Services Research Unit; UK = United Kingdom.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115] © BMS (2024). All rights reserved

B.3.2.3 Intervention technology and comparators

Fedratinib is an oral kinase inhibitor with activity against wild-type and mutationally activated JAK-2 and FMS-like tyrosine kinase 3 (FLT3). Fedratinib has a higher potency for JAK-2 over family members JAK-1, JAK-3, and tyrosine kinase 2 (TYK2). Abnormal activation of JAK-2 is associated with myeloproliferative neoplasms, including myelofibrosis and polycythaemia vera. Fedratinib has been compared with BAT in the FREEDOM-2 clinical trial, as presented in Section B.2.

In the cost-effectiveness model, the comparator of interest is BAT. The use of BAT as a comparator aligns current clinical practice, with the design of comparative clinical trials in myelofibrosis and previous ruxolitinib economic modelling. The composition of BAT from FREEDOM-2 is shown in Table 41. This table also shows how BAT changes in the model after fedratinib is discontinued to reflect a reduction in prescribing JAK inhibitors after discontinuation of both ruxolitinib and fedratinib. Of note, the proportion of patients receiving ruxolitinib in BAT was highlighted as an area of uncertainty in TA756; use of the observed proportion from FREEDOM-2 now resolves this uncertainty.

In the model, suboptimal fedratinib corresponds to fedratinib given in subsequent BAT treatment for patients who initially responded to fedratinib treatment. For the base case, it is assumed that suboptimal fedratinib does not occur (the rate is 0%). A scenario analysis explored suboptimal fedratinib given to 32.1% of responding patients receiving BAT. This was based on the discontinuation rate of fedratinib of 67.9% at the cutoff date.[36] It was assumed that these patients would not receive fedratinib, if fedratinib was to be considered as part of BAT. Therefore, it was assumed that all patients who had not discontinued fedratinib (32.1%) would continue fedratinib as part of BAT. This is a strong conservative assumption.

Table 41. FREEDOM-2: composition of BAT

BAT = best available therapy. Source: BMS data on file[36]

B.3.3 Clinical parameters and variables

B.3.3.1 Response assessment

The key types of response in myelofibrosis are spleen response and symptom response. The myelofibrosis-related symptoms evaluation is performed using the MFSAF version 4.0 using a 7-day recall period. Response assessments are assumed to occur at the end of cycles 3, 6, 12, 18, and 24 (and at the end of every sixth cycle as applicable afterwards; assessment of spleen size by palpation

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

and symptoms score only). Cycles are defined for administrative purposes as 4-week (28-day) periods irrespective of the assigned treatment arm.[36]

Spleen volume reduction and TSS are assessed at the end of the sixth cycle, which is equivalent to 24 weeks of treatment.

B.3.3.2 Spleen volume reduction response rate

A ≥ 35% SVR was used, considered by the IWG-MRT and European LeukemiaNet (ELN) as appropriate for response in patients with myelofibrosis.[52,68] Treatment with fedratinib is associated with a statistically significant improvement in spleen response rate ( P < 0.0001), with 35.8% of patients achieving ≥ 35% SVR at the end of the sixth cycle in the fedratinib arm. This compares to an SVR response of 6.0% in the BAT arm.

The values presented in Table 42 were retrieved from the FREEDOM-2 clinical trial and included in the model for SVR response rates.

Table 42. FREEDOM-2: SVR response rates

BAT = best available therapy; CSR = clinical study report; SVR = spleen volume reduction. Source: BMS data on file[36]

B.3.3.3 Total symptom score response rates

Treatment with fedratinib was associated with considerable symptom relief, with an improvement in TSS in 34.1% of patients in the fedratinib arm, whereas 16.9% of patients in the BAT arm achieved the clinically meaningful threshold for response of ≥ 50% reduction. This improvement was statistically significantly ( P < 0.0001).

The values presented in Table 43 were retrieved from the FREEDOM-2 clinical trial and included in the model.

Table 43. FREEDOM-2: TSS response rates at EOC6

BAT = best available therapy; CSR = clinical study report; EOC = end of cycle; TSS = total symptom score. Source: BMS data on file[36]

B.3.3.4 Spleen volume reduction or total symptom score response rates

Spleen or symptom response rate is defined as the number of patients achieving either ≥ 35% SVR or ≥ 50% reduction in TSS. A combined endpoint of spleen or symptom response was strongly recommended as a modelling input by experts at an advisory board, with the rationale that this outcome would be reflective of UK clinical practice given that the SVR and TSS track together.[23] The FREEDOM-2 clinical trial did not report such data directly. Consequently, SVR or TSS response rate was calculated using available data from the clinical study report.[36]

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

The values presented in Table 44 are included in the model and demonstrate improved levels of response with fedratinib.

Table 44. SVR or TSS response rates calculated from FREEDOM-2

BAT = best available therapy; SVR = spleen volume reduction; TSS = total symptom score. Source: BMS data on file[36]

B.3.3.5 Adverse events

Only non-haematological AEs grade ≥ 3 are explicitly modelled. The impacts of thrombocytopenia, anaemia, and neutropenia (common AEs in myelofibrosis) on costs and utilities are assumed to be captured elsewhere by the model. Costs of haematological AEs are considered by the “resource utilisation” of patients on either JAK inhibitors or BAT; and the impact on utilities of such AEs is assumed to be captured within the health-state utility values. Transformation to AML is also an important aspect of the progression and natural history of myelofibrosis, therefore, the economic model quantifies its impact by including AML as an AE. The model was set up to account for other “placeholder” AEs to facilitate updates in local adaptations.

Adverse event data were taken from FREEDOM-2 for both the fedratinib and BAT arm.[36] Adverse events were retrieved using the exposure-adjusted incidence rate per 100 person-years. The exposure-adjusted incidence rate was calculated by dividing the number of patients with specified TEAEs by the total exposure time (in years) to the event, and then dividing the result by 100. Exposure time was the TEAE follow-up time for patients without the event and the time up to the first event start date for patients with the event. The overall exposure time was 146.47 patient-years for the fedratinib group and 39.62 patientyears for the BAT group. This approach was used to obtain the AEs for the full treatment period rather than using the data for the first 6 treatment cycles, as longer follow-up makes more use of the available evidence, reducing uncertainty (by including more events). The reported data for the first 6 treatment cycles were also deemed to be less suitable as these data do not adjust for duration of exposure.

The exposure-adjusted incidence rate (per 100 people years) is then used in the model to calculate the annual incidence rate.

For AEs in BAT after fedratinib, the user has the option to choose between COMFORT-II or FREEDOM-2. FREEDOM-2 did not report AEs for BAT after fedratinib; however, no strong discrepancy is expected to be experienced by patients receiving BAT after fedratinib when compared with patient receiving BAT as the comparator treatment. Hence, the AE rates for BAT as the comparator were used for AE rates when receiving BAT after fedratinib.

The AEs from FREEDOM-2 and COMFORT-II are reported in Table 45 and Table 46, respectively.

There was no observed transformation to AML in FREEDOM-2. Nevertheless, this was included in the model AEs list to reflect the potentiality that myelofibrosis may also transform to AML. The incidence for transformation to AML was taken from the NICE 2016 TA386 Committee papers, used in TA756 as well[65] , and included in the model.[10] Rates of transformation to AML were assumed to be the same for both fedratinib and BAT, consistent with the committee’s preferred assumptions in TA756.

Company evidence submission for fedratinib for treating disease-related splenomegaly or symptoms in myelofibrosis (CDF review of TA756) [ID5115]

© BMS (2024). All rights reserved

Table 45. FREEDOM-2: grade 3+ adverse events experienced by participants