CADTH Reimbursement Review

Larotrectinib (Vitrakvi)

Sponsor: Bayer Inc.

Therapeutic area: Solid tumours with NTRK gene fusion

This multi-part report includes:

Clinical Review

Pharmacoeconomic Review

Clinical Review

Executive Summary

An overview of the submission details for the drug under review is provided in Table 1.

Introduction

The neurotrophic tyrosine receptor kinase genes, NTRK, encode the neurotrophin family of receptors.¹ Fusion of NTRK genes results from chromosomal rearrangements, which pre-clinical data indicates lead to constant activation of downstream signalling pathways without the need for ligands.² Although reported to be prevalent in 0.28% of all solid cancers, NTRK oncogenic fusions are observed at variable frequencies across a spectrum of pediatric and adult cancers, depending in part on the number of patients screened and NTRK fusion–detection techniques.¹ In Canadian adults, the 3 most common cancer diagnoses are lung, colorectal, and breast cancer.³ Childhood cancer accounts for less than 1% of all new cancer cases in Canada. Brain and central nervous system (CNS) cancers account for 19% of cancers and 34% of deaths, while neuroblastoma and other peripheral nervous cell tumours account for 11% of deaths.⁴ NTRK fusions are generally less prevalent in common cancers (approximately 0.1% to 1% in non–small cell lung cancer [NSCLC]^5-7 and 2% to 3% in sporadic colorectal cancers [CRCs⁵]), although NTRK fusion is more common in certain colorectal tumours with high levels of microsatellite instability (MSI)^8,9 and in primary thyroid cancers (6%).¹⁰ In contrast, NTRK fusions are nearly ubiquitous among rare cancer types, such as mammary analogue secretory carcinoma (MASC) and infantile fibrosarcoma (IFS).^5,11 In pediatric oncology, NTRK fusions are pathognomonic of specific, rare cancers including IFS (91% to 100%)¹² and cellular congenital mesoblastic nephroma (CMN) (83%).⁵ These fusions are also commonly observed in several other very rare pediatric cancers, including secretory breast cancer (SBC) (92%)¹³ and MASC of the salivary gland (100%).¹⁴ In addition, significant numbers of NTRK fusion cancers occur in children with papillary thyroid carcinoma (9.4% to 25.9%),^15,16 undifferentiated sarcomas (1%; frequency in adult versus pediatric patients not specified),¹⁷ high-grade gliomas (HGGs) (7.1%),⁵ and inflammatory myofibroblastic tumours, but rarely in acute leukemia.¹⁸

According to the clinical experts consulted for this review, no currently reimbursed drug targets the NTRK pathway. Among adult cancers, defining accepted clinical practice is difficult as NTRK fusions can be observed in a wide variety of solid cancers. Patients with locally advanced or metastatic solid tumours are currently largely treated with standard of care (chemotherapy, immunotherapy, and/or targeted therapy) as determined by the primary disease site.¹⁹ Ultimately, many of these cancers have a poor prognosis, and patients who progress on upfront therapies will have few subsequent therapeutic options. For patients with pediatric NTRK fusion cancers that are refractory to upfront therapy, there is no standard of care at relapse.²⁰ In addition, for infants with locally advanced, unresectable IFS, standard upfront conventional cytotoxic chemotherapy is poorly tolerated, has limited efficacy, and is associated with significant morbidity.⁵

Larotrectinib is an oral selective inhibitor of tropomyosin kinase receptors TrkA, TrkB, and TrkC, which are encoded by NTRK1, NTRK2, and NTRK3, respectively.²¹ Health Canada has issued marketing authorization for the use of larotrectinib for the treatment of adult and pediatric patients with solid tumours that have NTRK gene fusion without a known acquired resistance mutation or that are metastatic, or for whom surgical resection is likely to result in severe morbidity with no satisfactory treatment options. The marketing authorization was issued with conditions, pending the results of trials to verify its clinical benefit.²¹ The recommended dosage of larotrectinib in adults is 100 mg taken orally twice daily until the patient is no longer clinically benefiting from therapy or until unacceptable toxicity occurs. In pediatric patients, dosing is based on body surface area (BSA). The recommended dosage of larotrectinib in pediatric patients (1 month to 18 years of age) is 100 mg/m² taken orally twice daily, with a maximum of 100 mg per dose, until the patient is no longer clinically benefiting from therapy or until unacceptable toxicity occurs.²¹

The objective of this systematic review is to assess the beneficial and harmful effects of larotrectinib (adults: 100 mg taken orally twice daily; pediatric patients: 100 mg/m² taken orally twice daily up to a maximum of 100 mg per dose) for the treatment of adult and pediatric patients with solid tumours that have NTRK gene fusion without a known acquired resistance mutation or that are metastatic, or for whom surgical resection is likely to result in severe morbidity and who have no satisfactory treatment options.

Stakeholder Perspectives

The information in this section is a summary of input provided by the patient groups who responded to CADTH’s call for patient input and from clinical expert(s) consulted by CADTH for the purpose of this review.

Patient Input

Five patient group submissions were received for this review from prominent national patient advocacy groups, including Lung Cancer Canada (LCC), the Canadian Breast Cancer Network (CBCN), Colorectal Cancer Canada (CCC), the Sarcoma Cancer Foundation of Canada (SCFC), and the Canadian Cancer Survivor Network (CCSN). These submissions were based on information collected through surveys, interviews, literature reviews, and various patient engagement events. A total of 29 patients with NTRK fusion provided input, including 28 who received larotrectinib for their conditions. Responding patients were diagnosed at an advanced stage or had an aggressive form of rare cancer, were treated with multiple therapies with suboptimal success, and therefore needed a targeted therapy with a better efficacy and safety profile than conventional therapies. Patients as well as their caregivers reported significant negative impacts on virtually all aspects of their day-to-day life. Patients are often bound to limited physical work, resulting in limited employment opportunities; household activities and chores become difficult or impossible; and financial and personal stress becomes overwhelming, causing a significant drop in quality of life. The most important symptoms that need to be managed include pain in various parts of the body while minimizing or managing the side effects of chemotherapy treatment, fatigue and shortness (or loss) of breath, insomnia, loss of appetite, vomiting, diarrhea, wheezing, and coughing up blood. For pediatric patients, trouble swallowing and breathing, difficulty breathing, loss of appetite, and cachexia were noted as important symptoms to be managed. Overall, disease control, extension of life, improvement in quality of life (QoL), improvement of cancer and treatment-induced symptoms, longer-lasting and durable treatment, and minimal or tolerable toxicity were all noted as outcomes expected from a new therapy. Another important consideration for patients is accessibility and relative ease of administration, such that any additional burdens from travel, hospital visits, and resulting costs are minimized. Finally, a treatment that specifically targets the underlying NTRK mutation and minimizes the non-specific targeting of other tissues and organs was regarded as highly desirable.

Clinician Input

Input From Clinical Experts Consulted by CADTH

Five clinical experts from across Canada provided input for this review. Patients have unmet needs that vary by tumour type, but, in general, patients with metastatic solid malignancies eventually progress on currently available therapies. Certain tumours could be suitable candidates for larotrectinib treatment. Examples of such tumours in adults include MASC, which is a rare malignancy with a high prevalence of NTRK fusions and no established treatment options; differentiated thyroid cancer; anaplastic thyroid cancer, which has a dismal prognosis with standard cytotoxic treatments; and sarcoma with unresectable disease for which currently available chemotherapy often does not offer adequate response to facilitate surgery. In children and adolescents, this includes unresectable or metastatic IFS and soft tissue sarcoma (STS), which are chemorefractory and have no curative alternative therapy options; radioactive-iodine refractory NTRK fusion–positive thyroid cancers that require repeated invasive surgeries with no access to effective systemic therapies; unresectable and progressive/symptomatic low-grade gliomas (LGGs) that do not respond to conventional chemotherapy; several cancers with disease-related morbidity (e.g., loss of vision due to progressive optic pathway glioma) or that require multiple surgeries and ultimately radiation therapy (when old enough); and NTRK fusion HGGs that are unresectable or refractory to chemotherapy and for which no alternative curative options are available.

In both adult and pediatric patients, larotrectinib would be considered early in the course of NTRK fusion cancer treatment, even as first-line treatment for rare cancers without an effective option, or as a second-line treatment if the patient progressed on standard therapy. Examples include cancers for which no, limited, or inefficacious treatments are available, such as MASC, anaplastic thyroid cancer, or adult sarcomas with a NTRK gene fusion, IFS, pediatric unresectable or metastatic STS, NTRK fusion–positive thyroid carcinoma, unresectable and progressive/symptomatic LGG, or unresectable HGG in patients for whom radiation therapy is not possible.

Clinicians indicated that larotrectinib should be considered in adults with a good performance status and pediatric patients with advanced solid tumours that harbour an NTRK fusion. Patients with relapsed/refractory, unresectable, or metastatic NTRK fusion–positive STS (including IFS) and HGG are the patients with the greatest need. Conversely, patients with a poor performance status or those unable to tolerate oral therapies, a lack of NTRK gene fusions, or who have resectable disease would be least suitable. Response to treatment can be assessed by cross-sectional imaging modalities (MRI, CT, PET/CT) following Response Evaluation Criteria in Solid Tumours Version 1.1 (RECIST 1.1) or Response Assessment in Neuro-Oncology (RANO) criteria for CNS tumours, symptom improvement, tumour markers (thyroglobulin in differentiated thyroid cancers), treatment tolerability, and time to progression. Treatment response is typically assessed every 3 months in adult and pediatric patients; this interval may be prolonged once a response is established or remission is achieved. Treatment failure would be determined by disease progression, treatment intolerability, or patient request to discontinue treatment in adults, and if a response that facilitated curative surgery was achieved or there was objective evidence of disease progression on imaging evaluations in pediatric patients.

Clinician Group Input

A total of 9 clinician group inputs were received; all were joint clinician inputs, comprised primarily of 52 oncologists. Overall, the clinical groups shared mostly similar views as the clinical experts consulted by CADTH. Patients with advanced stages of cancer almost always fail on multiple therapies and have no or limited effective therapeutic options left. These patients would benefit most from a new treatment, particularly if it targets the underlying tumour cause, NTRK fusion in this case. While the current practice and the indication of larotrectinib is for use after all other treatment options have been exhausted, clinicians noted that larotrectinib could be prioritized in certain circumstances, depending on cancer type and underlying cause, stage, the availability of the treatment, and the efficacy and toxicity of treatment alternatives. Patients harbouring NTRK fusion were expected to be best suited for larotrectinib treatment, irrespective of the cancer type and location. Aside from NTRK status, patients with a good performance status were considered the most suitable candidates. In cases of pediatric cancers, patients with a poor prognosis were noted as suitable candidates for larotrectinib, unlike those with favourable prognosis. Conversely, the absence of NTRK fusion in adults, and children with an alternative, low-morbidity curative option would be least suitable for larotrectinib treatment. With respect to clinically meaningful response to treatment, clinician groups identified stabilization, no deterioration of symptoms, reduction in disease-related symptoms, overall reduction in tumour burden, improvement in survival, and QoL as important indicators of treatment efficacy. In pediatric cancers, clinically meaningful responses can vary, given the variety of histologies and presentations in the potential patient population. Patients initiating treatment are expected to be monitored frequently, at intervals ranging from 2 weeks up to 6 months, whereas those showing sustained responses may require assessment less frequently. Disease progression and certain unacceptable toxicities were noted as key factors when considering treatment discontinuation, which should be based on clinical status and objective progression on imaging.

Drug Program Input

Two main implementation issues were raised by the drug programs. First, the Provincial Advisory Group (PAG) inquired if larotrectinib can be used in patients with an Eastern Cooperative Oncology Group (ECOG) performance status of greater than 2, a Lansky Performance Scale (LPS) score of less than 40%, or a Karnofsky Performance Scale (KPS) score of less than 50%. Patients included in the trials (and therefore in the pooled analysis) had an ECOG status of 0 to 2, an LPS score greater than 40%, or a KPS score greater than 50%. To address the PAG’s question, the clinical experts consulted for this review indicated that patients with worse performance scores would not have been eligible for the phase I/II clinical trials, as per standard practice; however, children with worse performance scores may still benefit from larotrectinib. Second, in response to questions regarding the testing landscape, clinicians noted that testing in adults should be done at diagnosis or in a first-line treatment setting, with a provincial approach, on selected and/or enriched populations (with a known high prevalence of NTRK fusion, e.g., metastatic CRC with MLH1-promoter hypermethylation), or by incorporating NTRK testing in next-generation sequencing (NGS) for tumour sites that are routinely tested with NGS. For pediatric patients, testing depends on the diagnostic category for the underlying tumour; certain tumours (such as IFS and CMN) are routinely tested with fluorescence in situ hybridization (FISH) or NGS (for non-rhabdomyosarcoma STS) during diagnosis and can readily incorporate NTRK testing. For other tumours, molecular testing is conducted for a small subset of patients, e.g., unresectable or symptomatic or progressive radioactive-iodine refractory papillary thyroid carcinoma, which are gliomas that lack other clear oncogenic drivers. No information was provided on whether and how to incorporate NTRK testing for the latter tumour types.

Clinical Evidence

Pivotal Studies and Protocol-Selected Studies

Description of Studies

The CADTH resubmission of larotrectinib was based on a pooled analysis²² of 3 multi-centre, open-label, single-arm trials of larotrectinib in adult and pediatric patients with advanced or metastatic solid tumours: LOXO-TRK-14001 (phase I), LOXO-TRK-15003/SCOUT (phase I/II), and LOXO-TRK-15002/NAVIGATE (phase II basket trial). The trials are ongoing; the pooled analysis has been updated periodically for regulatory submissions, with updated data, larger sample sizes, and longer follow-ups at each update. The original larotrectinib submission to CADTH was based on the integrated dataset (data cut-off date: July 30, 2018).²³ The current resubmission was based on the following pooled datasets, with a data cut-off date of July 15, 2019: extended primary analysis set 4 (ePAS4) (n = 164 patients with NTRK gene fusion and a non-CNS primary tumour who had an independent review committee [IRC]-assessed response), safety analysis set 3 (SAS3) (n = 24 patients with NTRK-positive primary CNS tumours who had an investigator-assessed response), patients in the health-related quality of life (HRQoL) analyses (n = 126 patients with non-CNS primary solid tumours with an NTRK gene fusion (n = 74 adults, 24 children ≥ 2 years old, and 28 infants < 2 years old) for efficacy), a tropomyosin receptor kinase (TRK) fusion cancer labelled-dose safety analysis set (n = 196 patients with TRK fusion cancer), and an overall labelled-dose safety analysis set (n = 238 patients with or without TRK fusion cancer) for safety.²² A new data cut-off at July 2020 was used to create 4 updated datasets, which were submitted before this report being finalized, and included the following: extended primary analysis set 5 (ePAS5) (n = 192 patients, the most updated dataset with the highest number of patients and follow-up data, which was an update of the ePAS4 dataset), SAS3 (n = 33, an update of the previous data cut-off point), NTRK fusion cancer safety set (n = 260, an update of the TRK fusion cancer labelled-dose safety analysis set), and an overall safety set (n = 331, an update of the overall labelled-dose safety analysis set).

Most patients in the pooled analysis were treated with larotrectinib 100 mg orally twice daily in individuals with a BSA of at least 1 m², or 100 mg/m² orally twice daily for children with a BSA of less than 1 m², although some patients in the LOXO-TRK-14001 and LOXO-TRK-15003 trials received 50 mg/day to 400 mg/day and 150 mg twice daily of larotrectinib, respectively. The primary and secondary outcomes varied by trial, which was dependent on the objective, phase, and design of the trials. The primary efficacy end point in the pooled analysis was overall response rate (ORR), defined as the proportion of patients with a best overall response (BOR) of either complete response (CR) or partial response (PR), according to the RECIST 1.1 as determined by an IRC. Secondary efficacy end points included time to tumour response (TTR), duration of response (DOR), disease control rate (DCR), progression-free survival (PFS), and overall survival (OS). Additionally, HRQoL was assessed using the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 (EORTC QLQ-C30) and Pediatric Quality of Life (PedsQL) Version 4.0 Generic Core Scale.²²

The ePAS4 dataset included 13, 53, and 98 patients from the LOXO-TRK-14001, LOXO-TRK-15003, and LOXO-TRK-15002 trials, respectively. The included patients had a median age of 42.0 years; most patients had an ECOG performance status of 0 (49%) or 1 (38%); 49% were male; and 94% received 1 prior treatment of surgery (76%), systemic therapy (77%), or radiotherapy (46%). The most common tumour types were STS (22%), IFS (20%), thyroid tumours (16%), or salivary gland tumours (13%), and the following tumours were less frequent (< 10%): lung, colon, melanoma, breast, gastrointestinal stromal tumour (GIST), pancreatic, bone sarcoma, cholangiocarcinoma, appendix, hepatic, CMN, prostate, and unknown primary cancer. Most patients (74%) had metastatic disease at enrolment. In the SAS3 dataset, the primary CNS tumour types included glioblastoma (n = 7), glioma (n = 6), astrocytoma (n = 5), not otherwise specified (3), glioneuronal (1), neuronal and mixed neuronal-glial tumours (n = 1), and primitive neuroectodermal tumours (n = 1). The median age for patients in the SAS3 dataset was 8 years; 83% were pediatric patients, 46% were male, most had an ECOG status of 0 (63%) or 1 (29%), and most had received 1 to 2 lines of previous systemic treatment (67%). In terms of patient disposition, 84 of the 164 patients (51.2%) in ePAS4 dataset discontinued treatment primarily due to disease progression (29.9%). Disposition for the SAS3 dataset was not available.²²

Efficacy Results

All efficacy outcomes, with the exception of HRQoL, were reported in the ePAS4 dataset, which consisted of 164 patients with non-CNS solid tumours, and in the SAS3 dataset for 24 patients with primary CNS tumours. Outcomes for HRQoL were reported in 126 patients in the HRQoL analyses. Additionally, results of ePAS5 (n = 192) and SAS3 (n = 33), the newest datasets, were presented when data were available.

ORR: In the ePAS4 dataset, the ORR by IRC assessment was 73% (95% confidence interval [CI], 65% to 79%); 19% of patients achieved a CR, 5% achieved a pathological CR, and 49% achieved a PR. The DCR (defined as CR, pathological CR, PR, and stable disease lasting ≥ 16 weeks) was 84% (95% CI, 77% to 89%). Consistent with ePAS5, the ORR was 72% (95% CI, 65% to 79%); 23% and 7% patients achieved CR and pathological CR, respectively. In the original larotrectinib submission, as of the data cut-off date of July 30, 2018, the ORR was 81% (95% CI, 72% to 88%); 17% of patients achieved a CR and 63% achieved a PR.

In the SAS3 dataset, the ORR by investigator assessment was 21% (95% CI, 7% to 42%); 8% of patients achieved a CR, 13% achieved a confirmed PR, and the DCR was 63%. With the new data cut-off date of July 2020, the ORR was 24% (95% CI, 11% to 42%), with 9% and 15% of patients achieving a CR and PR, respectively. The estimated ORR in the original submission was 36% (95% CI, 13% to 65%), with a CR reported in 14% and PR in 21% of the patients.²²

The range of ORR across the different subgroups varied widely, although the subgroup analyses were noninferential and presented descriptively. The ORRs ranged from 91% (95% CI, 80% to 97%) in pediatric patients to 63% (95% CI, 54% to 72%) in adults; from 33% (95% CI, 1% to 92%) in patients with an ECOG status of 3 to 83% (95% CI, 72% to 90%) in patients with an ECOG status of 0; from 59% (95% CI, 41% to 75%) in patients who received 2 prior systemic treatments to 86% (95% CI, 71% to 95%) in patients who received 0 prior systemic treatments; from 67% (95% CI, 58% to 75%) in patients with metastatic cancer to 88% (95% CI, 74% to 96%) in patients with locally advanced metastatic cancer at baseline; and from 50% (95% CI, 7% to 93%) in patients with an NTRK2 gene fusion to 80% (95% CI, 71% to 88%) in patients with NTRK3 gene fusion. Across various tumour types, the ORRs ranged from 0% (95% CI, not estimated) for appendix and pancreas tumours and cholangiosarcoma (with ≤ 2 patients each) to 100% (95% CI, 40% to 100%) for GIST (with 4 patients). Although noninferential, the wide range of ORRs as well as the associated 95% CIs across the different tumour types limit the interpretability of the ORR for the individual tumour types. Given the small number of cases of any single tumour type, which was particularly true for cancers with fewer than 10 patients, it is unclear if the observed effects can be generalized to a larger sample of patients with the same cancer type. For example, based on 27 patients in ePAS4, the ORR in patients with thyroid cancer was 56% (95% CI, 35% to 75%), which was much lower than the 100% ORR (95% CI, 48% to 100%) seen in 5 patients with thyroid cancer based on a previous dataset, whereas the ORR improved from 25% to 38% in the case of colon cancer when 4 additional patients were included.²²

TTR: In the ePAS4 dataset, the median TTR was 1.84 months (range = 0.92 to 14.55 months). The percentage of patients experiencing a TTR of 2 months or less was 81% (96 of 119). In the SAS3 dataset, the median TTR was 1.82 months (range = 0.99 to 3.75); 67% of patients experienced a response in less than 2 months. The TTR was not reported in the original submission.²²

DOR: In the ePAS4 dataset, the median DOR was not estimable (NE) (95% CI, 27.6 months to NE) after a median follow-up of 15.7 months (interquartile range [IQR] = 6.6 months to 24.8 months). In the original submission, the median DOR had not been reached either. However, in the ePAS5 dataset, the DOR was 34.5 months (95% CI, 27.6 to 54.7) with a median follow-up of 20.3 months (IQR = data not applicable [NA]). In the SAS3 dataset, the median DOR had not been reached (95% CI, 3.8 months to NE) after a median follow-up duration of 5.3 months (IQR = 3.6 to 10.1 months).²²

PFS: In the ePAS4 dataset, the median PFS was 33.4 months (95% CI, 19.3 months to NE) after a median follow-up of 14.0 months (IQR = 7.9 months to 26.6 months). In the ePAS5 dataset, after a median follow-up of 22.1 months (IQR = data NA), the median PFS was 33.4 months (95% CI, 22.5 to 43.5). In the original submission, the median PFS was 28.3 months (95% CI, 9.9 to NE). In the SAS3 dataset, the median PFS was 11.0 months (95% CI, 5.4 to NE), with a median follow-up duration of 5.6 months (IQR = 3.6 to 13.1 months). With the new data cut-off, the median PFS was 18.3 months (95% CI, 6.7 to NE) after a median follow-up of 16.5 months.²²

OS: In the ePAS4 dataset, the median OS was NE (95% CI, 44.4 months to NE), and 85% of patients were alive after a median follow-up of 15.8 months (IQR = 9.3 months to 28.8 months). In the original submission, the median OS had not been reached after a median follow-up of 14.8 months. In the SAS3 dataset, the median OS was NE (95% CI, 9.4 months to NE), with a median duration of follow-up of 6 months.²² With the new data cut-off, the OS was still NE for ePAS5 and SAS3.

HRQoL: HRQoL outcomes were only measured in Study 15002 and Study 15003; analyses were not pre-specified in the sponsor-submitted statistical analysis plan for the pooled analyses, and it was not clear which analyses or subgroups were pre-specified, if any. Nonetheless, HRQoL data were presented descriptively, as part of conference proceedings; the most notable results are presented here. For adults and pediatric patients, the proportions of patients with HRQoL scores that were within the normal reference range or above the norm (norm/above norm) and below the norm (below norm) were determined at baseline and at best response using the EORTC QLQ-C30 global health score (GHS) and PedsQL total score. The mean GHS of the EORTC QLQ-C30 for the US general population (63.9)²⁴ minus 10 points (the estimated minimal important difference [MID]) was used to construct the norm/above norm (≥ 53.9) and below norm (< 53.9) score categories for adults.²⁵ The average score for the combined self- and proxy-reported PedsQL questionnaire for healthy US children (85.0)²⁶ minus 4.5 points (the estimated MID) was used to construct the norm/above norm (≥ 80.5) and below norm (< 80.5) score categories for children 2 years of age and older. The sponsor-identified MID for a PedsQL total score for children less than 2 years of age was 7.2 points.²²

Of the 52 adult patients with a norm/above norm EORTC QLQ-C30 GHS at baseline, 98% remained in this category at the best response, and of the 22 patients below norm at baseline, 9% remained in this category and 91% improved to norm/above norm. Of the 9 pediatric patients 2 years of age and older with a PedsQL total score norm/above norm at baseline, 100% remained in this category at the best response, and of the 15 patients below norm at baseline, 33% remained in this category and 67% improved to norm/above norm. Data were not available for pediatric patients younger than 2 years of age.²²

When compared with baseline, the mean of best changes in total score (standard deviation [SD]) was 17.5 (20.0), 20.7 (17.2), and 12.0 (13.8) for adults, children 2 years of age or older and children younger than 2, respectively. The changes exceeded the sponsor-identified (and literature-supported) MIDs for the respective instruments: 10 points for an EORTC QLQ C-30 GHS in adults and 4.5 points and 7.2 points for a PedsQL total score in pediatric patients older and younger than 2 years of age, respectively. Among patients (or parents and/or caregivers) who completed the respective questionnaires (74 adults, 24 children ≥ 2 years, and 28 children < 2 years), 59% of adult patients, 79% of children 2 years of age or older, and 57% of children less than 2 years of age had improvements in the best post-baseline score at or above the estimated MIDs. Of the patients evaluable for a sustained improvement (i.e., with a baseline and 2 or more post-baseline assessments, magnitude not defined), the improvement was sustained for at least 2 consecutive cycles for 47% of adult patients, 75% of children 2 years of age or older, and 43% of children younger than 2 years of age, and the improvement was sustained until the end of assessments in 30%, 50%, and 29% of patients, respectively.²²

The HRQoL results were associated with various methodological limitations, which are discussed in the critical appraisal section, limiting the interpretability of the data. Because HRQoL was not assessed in the previous pooled analyses, a comparison of change in HRQoL over time (different cut-off data for pooled analysis) could not be conducted.

Harms Results

The majority of the reported adverse events (AEs) in the pooled analysis were grade 1 or 2, most commonly reported as increased aspartate aminotransferase (ALT) and alanine aminotransferase (AST), cough, constipation, dizziness, fatigue, nausea, vomiting, pyrexia, anemia, and constipation. Treatment-related grade 3 or 4 AEs occurred in 15% or less of patients, which represents an increase in the incidence in comparison to the original CADTH submission, in which treatment-related grade 3 or 4 AEs occurred in less than 5% of patients. The most common grade 3 or 4 AEs included anemia, increase in liver enzyme (ALT and AST) levels, and decreased neutrophil count. Larotrectinib treatment interruptions or dosage modifications attributable to treatment occurred in 19% of patients in the TRK fusion cancer labelled-dose safety set (n = 196) and in 15% of patients in the overall labelled-dose safety set (n = 238), while permanent discontinuation of larotrectinib for treatment-emergent adverse events (TEAEs), regardless of attribution, occurred in 5% of patients and 8% of patients, respectively, in the 2 safety sets.²²

Overall, despite the increase in grade 3 or 4 AEs in the larger sample size and longer follow-up time compared to the original CADTH submission, the clinical experts consulted by CADTH agreed that the safety and tolerability of larotrectinib was acceptable. Larotrectinib resulted in minimal significant toxicity. Additionally, the availability of a liquid formulation, which facilitates dosing in young children as well as in adults, which was noted as a positive factor considering most treatment alternatives involve invasive surgery or IV therapy. Results were consistent in the newly submitted overall NTRK fusion cancers safety set (N = 260) and the overall safety set (N = 331).²²

Critical Appraisal

Internal Validity

Although the rationale to pool data from the 3 trials was based on the rarity of NTRK fusion cancer, the pooled analysis had several methodological limitations. First, interpretation of pooled analysis results remains difficult due to the between-study heterogeneity resulting from the differences in the design feature of the included trials, including different objectives, phases, outcome measures, and eligibility criteria across trials. The sponsor provided a sensitivity analysis showing the inclusion or exclusion of phase I data with phase II data did not meaningfully affect the primary outcome (ORR). The sponsor also submitted 3 additional analyses to address concerns about the inherent heterogeneity of the trials (details are provided in a following section), 1 of which mitigated concerns about patient heterogeneity due to the lack of controls (if the underlying assumptions are considered valid). However, none of the supplementary analyses addressed concerns about heterogeneity in tumour types, as well as heterogeneities across studies due to design features. Second, the following uncertainties around the pooled analysis results should be considered: the sample size for each individual cancer type was too small (in some cases fewer than 10 patients per tumour type) and the consequent 95% CI was too wide to make any conclusion about larotrectinib’s effect on different cancer types separately; the ongoing nature of the trial data means results could change as data on more patients and longer follow-up become available; the efficacy analyses for the SAS3 dataset used investigator-assessed outcomes, which is particularly problematic given the small sample sizes in this dataset and the open-label design of the trial. In addition, a number of survival outcomes (PFS, OS, and DOR) were analyzed using the Kaplan-Meier method to pool data across the 3 trials, which could be problematic as traditional survival analysis methods such as Kaplan-Meier curves rely on the assumption that a single survival distribution can be used to estimate the survival outcome for all patients included in the analysis.

The HRQoL analyses had a different set of limitations, aside from the lack of pre-specification in the sponsor’s statistical analysis plan. Patients’ best responses were considered in evaluating effects on HRQoL, not their last reported responses. As HRQoL could vary over time, it is unclear whether the best change was, in fact, a transient improvement in HRQoL. The categorization of norm/above norm or below norm means patients could move from 1 category to another without actually exceeding the estimated MIDs for the respective instruments, particularly if their baseline values were closer to the chosen cut-off (53.9 for EORTC QLQ C-30 GHS in adults and 80.5 for PedsQL in children ≥ 2 years of age). Additionally, patients could experience a clinically relevant decline in HRQoL and still be in the norm/above norm category as long as they did not fall below norm (e.g., if they started with a high baseline HRQoL level, dropped to a level that exceeded the respective MIDs after treatment, and were still above the norm). Although a sensitivity analysis conducted using a higher threshold (58.9 for adults and 82.75 for children) showed similar results, the aforementioned limitations remained. Given these issues, and the lack of a statistical comparison between baseline and post-baseline levels of HRQoL, the maintenance and/or improvement in HRQoL is unclear.

External Validity

Overall, the 3 trials included in the pooled analysis consisted of patients with tumours at various sites, although not all solid tumour types were represented in the studies. While this suggests the results are generalizable to the various tumour types, some of these tumour types were under-represented in the study population, resulting in wide confidence intervals, reducing confidence in their generalizability. Patients included in the pooled analysis mostly had an ECOG status of 0, 1, or 2 (and their respective performance status equivalence for the pediatric population). There are insufficient data to support generalizing to patients with poorer performance status as the number of patients with an ECOG status of 3 in the pooled analysis was too small.

Other Relevant Evidence

Three analyses were submitted by the sponsor to address sources of heterogeneity using the available data. The submitted analyses include a Bayesian hierarchical model (BHM), a permutation analysis, and an intra-person Growth Modulation Index (GMI) analysis. The first and second analyses attempt to quantify and account for the heterogeneity in ORR across tumour locations and studies, and possibly for other differences in histology, while the third attempts to mitigate for the lack of control and therefore heterogeneity between participants due to histology and location, as well as many other factors, by using the individual’s time to progression or treatment failure (TTPF) under previous treatment to compare to their PFS under their later treatment with larotrectinib.

The BHM analysis, although appropriately performed, is presented as a means of accounting for, rather than investigating heterogeneity in, ORR across tumour types. The BHM clearly identifies heterogeneity across tumour types, and no further analysis is presented to show that this heterogeneity can be explained by participant-level characteristics outside of tumour type, i.e., the subject-level characteristics that are controlled for in the GMI analysis. The BHM analysis does not provide evidence to support the combined analysis of the data for approval across all tumour types. The permutation analysis does not seem to add to the evidence to support or reject the pooling of all data in a single analysis, as the presented distribution across all subgroups may merely be an artifact of smaller groups having lower ORRs and larger groups having higher ORRs. Additionally, in both of these analyses, ORR and not PFS or OS is used as the outcome. This may not answer the more relevant question of the effect on survival, as ORR may not be a reasonable surrogate for PFS or OS across all tumour types, participants, or treatments.

The intra-patient GMI analysis provides the best evidence that larotrectinib is effective over a number of patient types. If the assumptions underlying this analysis are valid, it helps mitigate concerns about patient heterogeneity due to the lack of controls. However, it is not clear if the needed assumptions hold, and the lack of information about how the end point was calculated and how well the analysis performed raises further questions about the results. Additionally, this analysis does not directly address concerns about heterogeneity across tumour types. This is particularly evident as most tumour types with a low ORR are not included in the GMI analysis. None of the analyses provides strong evidence that larotrectinib is as active in these smaller groups defined by particular tumour types.

In June 2021, the sponsor submitted a summary of key information on NTRK gene fusion testing in response to the PAG’s request for reconsideration of the draft recommendation issued by the CADTH pan-Canadian Oncology Drug Review (pCODR) Expert Committee (pERC) in May 2021, and the implementation questions raised by the PAG. In this summary document, the sponsor provided a summary of key considerations for NTRK gene fusion testing along with information on the current situation and future directions of NTRK testing in Canada (see Appendix 6 for more details).

Conclusions

The clinical data supporting the efficacy of larotrectinib in a histology-agnostic patient population with NTRK fusion–positive cancer is derived from a pooled analysis of 3 open-label, single-arm trials; including a phase I trial, a phase I/II trial, and a phase II basket trial. The current CADTH resubmission is based on a larger sample size than that of the previous submission, with longer patient follow-up. Additionally, alternative methods for evaluating ORR and PFS, and observational data submitted by the sponsor, were considered in this submission. In total, 164 adults and pediatric patients with NTRK-positive cancer of different histologies were included in the most recent pooled analysis.

Results showed that, among patients with a non-CNS tumour, larotrectinib treatment was associated with a 73% improvement in ORR; half of the responders had a PR, 19% patients achieved a CR, and the median time to response was less than 2 months. Across different tumour types, the ORRs varied widely, with a similarly wide range of uncertainty. Combined with differences in sample sizes across the different tumour types, some of which had fewer than 10 patients, these factors limit the generalizability of the findings of the mixed cancer population. The ORRs also varied between adults and pediatric patients, as well as by patient baseline state and previous treatment history. Effects on other outcomes important for clinicians and patients, including PFS, OS, DOR, and HRQoL, remain uncertain or inconclusive among patients with primary CNS or non-CNS tumours due to the lack of accrual of sufficient events and multiple methodological limitations. Among patients with primary CNS tumours, of which the majority were children, the ORR was 21%. Other measures of effect varied widely, with small sample sizes and consequently low event numbers adding to the uncertainty. Overall, these results and methodological limitations remained largely unchanged compared to the previous review. While the rarity of NTRK fusion creates practical and ethical challenges to conducting a randomized controlled trial, the lack of comparative evidence as well as multiple methodological limitations means the results should be interpreted based on clinical judgment.

Three additional analyses were submitted in response to concerns raised in the previous submission regarding the inherent heterogeneity across tumour types as well as patients included in the aforementioned trials and the lack of a comparator group. Results from the intra-patient GMI analysis mitigated concerns about between-patient heterogeneity due to the lack of controls. However, none of the alternative analyses supported uniform effectiveness of larotrectinib across tumour types. Four real-world evidence (RWE) studies with thousands of cancer samples supported the oncogenicity and mutual exclusivity of NTRK fusion in certain cancers, but showed no increase or reduction in PFS or OS among NTRK-positive cancer patients.

Introduction

Disease Background

The NTRK genes encode the neurotrophin family of receptors.¹ A recent study estimated that NTRK gene fusions occur in 0.28% of all solid cancers. Oncogenic fusion of NTRK genes arise from exact intrachromosomal or interchromosomal rearrangements that juxtapose the kinase domain–containing 3' region of NTRK with the 5' region of NTRK gene partners. Pre-clinical data demonstrated that chimeric oncogenic fusions may lead to partial or complete deletion of the immunoglobulin-like domain of TRK, which inhibits downstream signalling pathways in the absence of activating ligands.² Available literature demonstrates that NTRK gene fusions are oncogenic drivers in various cancers.¹⁷ Appendix 4 provides further details.^17,28-30

Although reported to be prevalent in 0.28% of all solid cancers,³¹ NTRK oncogenic fusions are observed at variable frequencies across a spectrum of pediatric and adult cancers, with some uncertainty regarding exact frequencies.¹ Different studies have reported varying frequencies, possibly due to the number of patients screened and fusion detection techniques.

Lung, colorectal, and breast cancer represent the 3 most common cancer diagnoses in Canada (Table 4).

• In NSCLC, NTRK fusions are less common (with an occurrence rate of approximately 0.1% to 1%)^5-7 compared with other oncogenic gene rearrangements that involve the anaplastic lymphoma kinase (ALK) gene, ROS proto-oncogene 1 (ROS1), and RET proto-oncogene, which occur at frequencies of approximately 4% to 6%, 1% to 2%, and 1% to 2%, respectively.^33-35

• The NTRK mutation is also quite rare in breast cancer, with the exception of the rare subtype of SBC, in which the prevalence of NTRK fusion has been reported to be 92%.

• NTRK gene fusions are also rare in sporadic colorectal cancers (2% to 3%),⁵ and appear to be more common in colorectal tumours with high levels of MSI, and mutually exclusive of RAS and BRAF mutations, which represent about 55% of metastatic CRCs.⁹

• The NTRK mutation is uncommon in adult sarcomas (1%); it is found at a higher frequency in GIST,³⁶ particularly wild-type GIST (lacking mutations in KIT and PDGFRA).

• NTRK gene fusions are observed in 6% of adults with primary thyroid cancers.¹⁰

While the frequency of NTRK fusions is low in common cancer types, NTRK3 fusions are nearly ubiquitous among rare cancer types such as MASC and IFS.^5,11 In pediatric oncology, NTRK fusions are pathognomonic of specific, rare cancers including IFS (91% to 100%)¹² and cellular CMN (83%).⁵ Fusions of NTRK genes are also commonly observed in several other rare pediatric cancers, including SBC (92%)¹³ and MASC of the salivary gland (100%).¹⁴ In addition, significant numbers of NTRK fusion cancers have been reported among children with papillary thyroid carcinoma (9.4% to 25.9%),^15,16 undifferentiated sarcomas (1%; frequency in adult versus pediatric patients not specified),¹⁷ HGGs (7.1%),⁵ inflammatory myofibroblastic tumours, and (rarely) acute leukemia.¹⁸

Childhood cancer accounts for less than 1% of all new cancer cases in Canada. Between 2009 and 2013, there were 4,715 new cases of cancer in children 0 to 14 years of age in Canada (average of 943 cases per year). Between 2008 and 2012, there were 595 cancer deaths in children 0 to 14 years of age in Canada (average of 119 deaths per year). Brain and CNS cancers account for 19% of cancers and 34% of deaths, respectively, whereas neuroblastomas and other peripheral nervous cell tumours account for 11% of deaths.⁴

The most common types of solid tumours found in adolescents and young adults (15 to 29 years) include thyroid (16%), testicular (13%) and melanoma (8%), while the majority of cancer deaths (from solid tumours) in this same age group are attributed to brain and CNS cancers (15%) and bone cancers (11%).⁴

Standards of Therapy

There is currently no reimbursed drug that targets the NTRK pathway. For adult cancers, defining accepted clinical practice is difficult, as NTRK gene fusions can be observed in a wide variety of solid cancers. Patients with locally advanced or metastatic solid tumours are currently treated largely with standard of care (chemotherapy, immunotherapy, and/or targeted therapy) according to the primary disease site.¹⁹ Ultimately, many of these cancers have a poor prognosis, and patients who progress on upfront therapies will have limited subsequent therapeutic options.

For patients with pediatric NTRK fusion cancers that are refractory to upfront therapy, there is no standard of care at relapse.³⁷ In addition, for infants with locally advanced, unresectable IFS, standard upfront conventional cytotoxic chemotherapy is poorly tolerated, has limited efficacy, and is associated with significant morbidity.

Selected Disease Site-Specific Burden and Need Considerations

Pediatric Solid Cancers

Despite their relative rarity in pediatric oncology, NTRK fusions are pathognomonic of specific, rare cancers (IFS and cellular CMN). Several other very rare pediatric cancers, including SBC and MASC of the salivary gland, are also expected to carry NTRK fusions. In addition, there are significant numbers of NTRK fusion cancers among children with papillary thyroid carcinoma, undifferentiated sarcomas, HGGs, and inflammatory myofibroblastic tumours, but they rarely occur in those with acute leukemia.²⁰

Infantile fibrosarcoma is the most common STS in infants younger than 1 year of age. It is rare, and many patients present with localized disease, in which case upfront resection with negative margins is the treatment of choice if feasible. In the largest retrospective review of IFS, surgical resection was the initial treatment for 68% of patients.³⁸ For children with locally advanced (unresectable) or metastatic IFS, the historical approach to therapy has been conventional cytotoxic chemotherapy. In Canada, the most common regimen used would include vincristine or actinomycin with or without cyclophosphamide.²⁰ The response rate to chemotherapy is approximately 68% to 71% but requires a central line and carries acute and long-term toxicity risks. An estimated 9% to 10% of children ultimately required disfiguring amputations to achieve a cure in the era before TRK inhibitors.³⁸

Among other childhood non-rhabdomyosarcoma, examples of STS, NTRK fusions are rare. Similarly, surgical resection is the mainstay for IFS that is localized and resectable, with adjuvant or neoadjuvant cytotoxic chemotherapy (typically doxorubicin and ifosfamide) used if it is unresectable or metastatic. The response rate to chemotherapy in a pediatric STS is approximately 55%.³⁹ According to the clinical experts consulted for this review, anthracycline and alkylator therapies come with significant late toxicities, including cardiomyopathy and infertility, when used in young children. There are no standard conventional chemotherapeutic approaches to second-line therapy for chemorefractory disease.

Fusions of NTRK genes are found in approximately 26% of pediatric patients with differentiated thyroid carcinomas, which often present at a more advanced stage than in adults. Surgery (thyroidectomy) is the primary therapy, and a subgroup also receives radioactive-iodine therapy based on the risk groups defined by the American Thyroid Association. The clinical experts consulted by the review team indicated that, based on their experience, outcomes of this first-line treatment are typically good, and few children need systemic therapy. Selective NTRK-inhibitor therapy would be considered as second-line therapy for NTRK fusion–positive thyroid cancer that is radioactive-iodine refractory and either symptomatic or progressive.²⁰

Finally, NTRK fusions are rare, accounting for 0.4% to 3.1% of pediatric brain tumours, both in LGGs and HGGs, and often in younger children. For LGGs, resection is recommended if it is in a resectable location. However, if the tumour is unresectable and progressive, first-line systemic therapy would be vinblastine, which is associated with an ORR of approximately 26%, or vincristine and carboplatin.⁴⁰ The clinical experts consulted by the review team noted that radiation therapy is often not possible due to the young age of children (< 5 years) and profound neurocognitive toxicity. Selective NTRK-inhibitor therapy would be considered second-line for NTRK fusion–positive patients if no clinical trial is available. High-grade gliomas are also treated with surgery, when possible, and accompanied by intensive cytotoxic chemotherapy in young children. Again, radiation therapy is standard in older children, but it is often not possible in young children due to morbidity. Recent work has shown that NTRK fusion accounts for a large number of alterations of hemispheric gliomas in infants. While this population is small in size, the management of these tumours is extremely challenging and most survivors have profound neurologic and cognitive deficits. There is a clear need to change the management of these tumours in this age group.⁴¹

Secretory Breast Cancer

Fusion of NTRK genes is quite rare in breast cancer.^13,18 A number of standard therapy options offer considerably improved survival rates are available for patients with advanced breast cancer,⁵ but many patients will ultimately go on to exhaust available therapies and be left with no suitable therapeutic options.¹⁹

Secretory breast carcinoma is a rare histologic subtype of breast cancer that is seen in less than 1% of invasive breast cancer in children and adults and is associated with a generally favourable prognosis and a low likelihood of metastases.^13,18 However, for patients with advanced, inoperable disease, treatment options are limited. Secretory breast carcinomas are also associated with a greater than 90% prevalence of NTRK gene fusions.⁵

Sarcoma and Gastrointestinal Stromal Tumour

Sarcomas are a relatively rare tumour subtype comprising more than 100 hundred subtypes. They are often categorized into STS varieties and bony sarcomas. The former are associated with a less-favourable prognosis, and in the adult population, are often not curable. Limited effective cytotoxic therapies exist for STS, particularly in the metastatic setting or upon relapse.⁴²

In adults with sarcoma and NTRK fusions, standard therapies include radiation and surgery, and cytotoxic therapy (i.e., doxorubicin) for those with advanced disease. However, in the advanced setting, traditional chemotherapy is essentially ineffective.¹⁹ The clinical exerts consulted by the review team stated that there are examples of fusion-positive pediatric patients with locally advanced disease who show enough response to the drug to facilitate curative surgical resection.

Fusions of NTRK genes are also seen in 3% to 4% of GIST tumours.³⁶ For GIST tumours with cKIT and PDGFRA mutations, targeted therapies represent the current standard of care. For the 10% to 15% of GISTs that are wild-type, there is a significant unmet need for effective therapies.⁴³

Thyroid Cancer

For patients with advanced, inoperable thyroid cancer that has progressed on radioactive-iodine therapy, current treatments include small-molecule tyrosine kinase inhibitors.¹⁹ An estimated 6% of thyroid cancers may involve NTRK gene fusions.¹⁰

Gastrointestinal Cancers

For patients with advanced CRC, there is an unmet need for better therapies in patients with chemorefractory disease (i.e., those who have progressed on 2 or more prior lines of therapy). While NTRK gene fusions are uncommon in CRC,^8,9 the clinical experts consulted by the review team noted that there is a significant unmet need for better therapies for patients with non-colorectal gastrointestinal cancers, particularly pancreatic cancer and cholangiocarcinoma.

Lung Cancer

Lung cancer remains the most common cancer in Canada.⁴⁴ NTRK fusions are estimated in up to 1% of NSCLCs⁵ (as compared to ALK fusions in 3% to 5%, ROS1 fusions in 1% to 2%, and epidermal growth factor receptor (EGFR) gene mutations in 20%).⁷ Systemic treatment options for advanced NSCLC include chemotherapy, immunotherapy, combinations, and biomarker-directed targeted therapies, with response rates ranging from 45% to 60% in those without ALK/EGFR/ROS1/BRAF-deranged lung cancer. While these current therapies have improved outcomes for patients with NSCLC, patients will ultimately become refractory and/or intolerant to available therapies, and effective and tolerable therapies are need for pre-treated patients.⁴⁴

Drug

Larotrectinib is an oral selective inhibitor of tropomyosin kinase receptors TrkA, TrkB, and TrkC, which are encoded by NTRK1, NTRK2, and NTRK3, respectively. Larotrectinib is a highly selective, potent, adenosine triphosphate–competitive, and small-molecule pan-TRK inhibitor with a half-maximal inhibitory concentration (IC50) in the low nanomolar range.²¹ Health Canada has issued marketing authorization for the use of larotrectinib for the treatment of adult and pediatric patients with solid tumours who have NTRK gene fusion without a known acquired resistance mutation, are metastatic or where surgical resection is likely to result in severe morbidity, and have no satisfactory treatment options.²¹ The marketing authorization was issued with conditions, pending the results of trials to verify its clinical benefit.²¹The recommended dosage of larotrectinib in adults is 100 mg taken orally twice daily until the patient is no longer clinically benefiting from therapy or until unacceptable toxicity occurs.²¹ In pediatric patients, dosing is based on BSA. The recommended dosage of larotrectinib in pediatric patients (1 month to 18 years) is 100 mg/m² taken orally twice daily, with a maximum of 100 mg per dose until the patient is no longer clinically benefiting from therapy or until unacceptable toxicity occurs.²¹

Stakeholder Perspectives

Patient Group Input

This section was prepared by CADTH staff based on the input provided by patient groups.

About the Patient Groups and Information Gathered

Five patient group submissions were received for this review: LCC, the CBCN, CCC, the SCFC, and the CCSN. All participating organizations are prominent national patient advocacy groups in their respective cancer area. LCC is a charitable organization providing resources for lung cancer education, patient and caregiver support, research, and advocacy. The CBCN is a patient-directed organization that aims to voice the issues and concerns of breast cancer patients through education and advocacy activities, and promote a national network through information sharing. CCC is a charitable not-for-profit organization dedicated to colorectal cancer awareness and education, supporting patients and caregivers, and advocacy on their behalf. The SCFC is a charitable organization dedicated to providing patient support, including education and advocacy, disease awareness, and support for Canadian sarcoma cancer research. Finally, the CCSN provided a collective patient input submission, with input from the following patient advocacy groups: the Advocacy for Canadian Childhood Oncology Research Network, Colorectal Cancer Resource and Action Network, and GIST Sarcoma Life Raft Group Canada. A disclosure of any conflicts of interest for the patient groups is available on the CADTH website. The CCSN collective patient input submission was coordinated and authored by the Blue Ribbon Project (Filomena Servidio-Italiano). The remaining submissions were prepared by the respective organizations without any external help.

Lung Cancer Canada collected information through interviews and environmental scans between November and December of 2020 in addition to information from a previous larotrectinib submission. The thoughts and experiences of 6 patients (4 males and 2 females; age range = 33 years to 66 years) were included. The CBCN submission collected information through a 2017 online survey, with data from 180 Canadians living with metastatic breast cancer. A similar online survey was conducted in 2012, with input from 71 patients with metastatic breast cancer and 16 caregivers. In addition, a review of current studies and grey literature was conducted to identify issues and experiences that are commonly shared among many women living with metastatic breast cancer. The information for the CCC submission was collected by a combination of an online survey and patient outreach by CCC representatives. An online patient/caregiver survey conducted between October and November 30, 2020, was used to gather data from 6 patients and 5 caregivers across Canada, the US, and Brazil. The patients and caregivers of patients included in the survey had different cancer types and stages, with most patients having cancer that had metastasized from the organ of origin. In addition, CCC reached out to patients via 2 conferences (1 in 2018 and 1 in 2020) and personal interviews (individually and as groups) to gather input on patient perspectives and experience with larotrectinib. Eight patients were interviewed; 5 female patients (age range: 30 years to 69 years) and 3 male caregivers (age range: 50 years to 69 years). The 5 caregivers surveyed provided care for pediatric and adult patients; 3 cared for a female patient (age range: 0 years to 39 years), and 2 cared for a male patient (age range: 0 years to 59 years). In 2019, CCC also submitted input based on a survey conducted between January 7, 2019, and January 29, 2019. Input from the SCFC was collected through multiple interviews with 3 patients, 2 caregivers, and several physicians across Canada. Additionally, general information on sarcoma cancers collected from personal experiences and the collective experience of SCFC members and the sarcoma community were shared. The CCSN submission was based on interviews conducted between October and November 2020 with 8 adult patients experienced with larotrectinib treatment solicited through the LOXO-101 clinical trial principal investigator, an online CRC support group in US and Canada, an online NTRKers support group, and Canadian clinicians who have been prescribing larotrectinib. In addition, CCSN received information from multiple interviews with 4 pediatric patients conducted by the Advocacy for Canadian Childhood Oncology Research Network, between February 2019 and October 2020. Patients included in the CCSN submission had cancers of 6 different sites. The patient group submissions from the CBCN had no reported cases of patients with NTRK mutations, whereas the SCFC, LCC, CCC, and CCSN incorporated input from 3, 3, 9, and 12 patients with the gene fusion, respectively.

Disease Experience

Patients’ experiences with the various cancers shared a common theme. In most cases, patients were diagnosed at an advanced stage, had an aggressive or rare form of cancer, or both. Most patients had suboptimal success with conventional therapies and longed for a targeted approach with a better safety and efficacy profile than that of conventional therapy. While the disease presentation, symptoms, progression, and how it was experienced varied by patients, they overwhelmingly indicated that the negative effects cancer had on their day-to-day life as well as their caregivers were debilitating. Most patients experienced a number of symptoms induced by the cancer as well as by the treatment they received, with the following symptoms noted as the most urgently in need of management: pain in various parts of the body, the side effects of chemotherapy treatment, fatigue and shortness (or loss) of breath, insomnia, loss of appetite, vomiting, diarrhea, wheezing, and coughing up blood. The severity of these symptoms varied, although most patients reported that their symptoms were severe or extreme. In some instances, patients reported being confined to bed or a wheelchair due to severe pain, having to stop driving, and difficulty walking or climbing stairs. One group that expressed the views of pediatric patients noted the need to manage the following cancer-induced symptoms: trouble swallowing and breathing, difficulty breathing, loss of appetite, and cachexia. Patients also experienced various disease- and treatment-specific complications. For example, sarcoma patients must sometimes contend with surgical amputation of a limb as a treatment, which is associated with a significant and lengthy rehabilitation process and the need to become comfortable with prosthetics. Some patients develop growths in various parts of the body that lead to appearance anxiety, annoyance, and pain.

The patient input stated that the debilitating symptoms cancer patients experience invariably affect their everyday life and activities. Patients are often restricted to limited forms of physical work, which can greatly reduce or completely diminish their employment opportunities. In addition, household activities and chores become difficult and sometimes impossible to carry out. Patients reported an inability to participate in social activities and hobbies, play sports, participate in family gatherings, and maintain a personal life. Their professional careers are often ruined, or severely limited, financial burdens become overwhelming, and some find themselves in significant debt to pay for treatments, resulting in the loss of their home, breakup of their marriage, and end of their career. All of these inevitably affect their psychological well-being, and patients reported a constant feeling of anxiety, depression, fear, and poor mental health. Overall, all aspects of patients’ lives are severely affected by the cancer and treatment-induced symptoms, resulting in a poor quality of life. One patient expressed her current state with breast cancer in the following way:

I’m 43 now and I will be in treatments for the rest of my life. I have a very difficult time still trying to figure out how to move forward while taking advantage of all the wonderful moments I still have. I have no choice but to continue to battle this war that my body has bombarded my family and me with… the most difficult aspect is planning for my mortality and trying to keep my chin up and not burden my family.

It was noted that the negative impacts that patients experience in their daily lives, mental health, and overall quality of life invariably seep into the lives of their caregivers. Caregivers often must devote a significant portion of their time to caring for family members or friends suffering from cancer. As a result, all aspects of their daily lives are negatively affected. Caregivers consulted by the patient groups described their experience as physically and emotionally draining; they felt stress in their own lives, feared for the survival of patients they cared for, and suffered from a poor quality of life.

Experiences With Currently Available Treatments

As with patient experiences with disease, treatment regimens depend on the underlying cancer location, stage, type and subtype, availability and access of effective alternative treatment, and toxicity profile. The following section provides a general description of the various treatments patients receive to manage their disease and symptoms, as well as the benefits and challenges of the existing therapy options.

Overall, patients participating in the surveys received 1 or more of the following therapies — alone, in combination, or sequentially (based on guideline or consensus-driven algorithm): systemic chemotherapy, radiation therapy, immunotherapy, (targeted) surgery, and/or oral adjuvant and oral adjunct therapies. Chemotherapy was by far the most common form of treatment irrespective of cancer type, histology, and metastatic status, in patients with NSCLC, SBC, CRC, sarcoma, thyroid cancer, salivary gland tumour, glioblastoma, IFS, and other less-commonly reported cancers. While chemotherapy is reportedly a viable treatment option in some instances, this was described as ineffective in halting tumour growth or progression, and was associated with various side effects, some of which were unmanageable, such as extreme fatigue, pain, nausea, diarrhea, weight loss, skin issues, food allergies, and edema. For a few patients, treatment was associated with rare but excruciating eczema, failure to thrive, or memory loss. In addition, patients and their caregivers reported that the need for hospitalization, commuting, and managing side effects were added burdens of chemotherapies, particularly from a financial standpoint. The low effectiveness and high toxicity of chemotherapies invariably resulted in loss of productivity and poor quality of life for the patients and their caregivers.

Immunotherapy, particularly for patients with NSCLC, temporarily demonstrated an improvement in managing disease progression and minimizing side effects; however, this treatment was not able to prevent disease progression.

Radiation therapy was seen as an important treatment option, particularly for patients with colorectal cancer, thyroid cancer, salivary gland cancer, and glioblastoma. Similar to chemotherapies, the effectiveness of radiation therapy varied by patient depending primarily on cancer location, type and subtype, stage, and treatment history pre- and post-radiation. Radiation therapy was also associated with significant side effects; patients reported treatment-induced toxicities similar to those for chemotherapies, most notably fatigue, pain, nausea, diarrhea, weight loss, and skin issues.

Partial or complete surgical resection of tumours was described as a relatively effective treatment option in the early stages of several cancer types. However, as noted by some patients, surgery is not an infallible method to halt or prevent disease progression, even when cancer is diagnosed early. Further, surgery is not feasible in metastatic cases and in some cancer types, such as STS. Surgery is often supplemented with chemotherapy or radiation therapy before or after, resulting in the addition of the side effects associated with these treatments. Furthermore, surgery is often associated with the undesired outcome of the amputation of a critical limb or body part, which adds the burdens of long-term rehabilitation, adjustment with prosthetics, and physical and mental challenges associated with the predicament. The following quote captures the plea of a patient with little to no effective treatment available:

Had you asked me some of these questions 4 years ago, the answers would have been different. My oncologist tells me that I am running out of treatment options. […] It is very scary to face the day (soon) when I will have no treatment and the cancer will be allowed to run its course.

In addition to the treatments described previously, a number of adjuvant oral medications were described by the patients as supplemental therapies to prevent or halt tumour growth and the progression of disease, and/or increase the chances of survival. These medications have varying degrees of effectiveness and side effects, although the side effects were described as manageable overall. These adjuvant therapies were noted as particularly efficacious when guided by the underlying mutational status of a patient’s tumour. However, mutational diagnosis was reportedly not always part of the standard treatment procedure.

Improved Outcomes

The aspects of disease and outcomes highlighted by patients as important were largely similar irrespective of tumour type, and included disease control, extension of life, improvement in quality of life, and minimal or tolerable toxicity. Patients expected a new therapy to improve PFS and OS, delay disease progression, reduce cancer- and treatment-induced symptoms, and provide longer-lasting, durable, and well-tolerated treatment. The presentation of cancer- and treatment-associated side effects varied by individual and cancer type; however, all patients overwhelmingly shared a common desire for no or minimal cancer-induced symptoms and treatment-induced side effects. Trade-offs commonly considered when choosing a therapy included extended OS and PFS and quality of life. Some patients reported that an improved quality of life is no less important than extension of survival, whereby one can resume a daily life as normal as possible. Another important consideration for patients is accessibility and relative ease of administration, so that any additional burden from travel, hospital visits, and resulting costs are minimized. Patients strongly value personal and professional life experiences from family, friends, personal interests, and careers, making a treatment option that allows them to prolong survival, manage symptoms and the underlying cancer, and improve quality of life with no or minimal and manageable side effects highly desirable. Patients also expressed a desire for a treatment that specifically targets the underlying NTRK mutation and minimizes the non-specific targeting of other tissues and organs is highly desired. Below are some quotes from the patients that captures their views from a new treatment:

• “I never really had a good response to any of the previous therapies. My tumours just kept growing and spreading which is why I am so happy to be on Vitrakvi. I became so ill where I was given a timeline for death.”

• “Nothing worked for me. My disease progressed on every therapy I have been on before larotrectinib.”

Experience With Drug Under Review

Input from 28 patients experienced with larotrectinib (3 from LCC, 13 from CCC, and 12 from the CCSN) was scarce given the relative rarity of cancer types with an NTRK mutation, with even fewer patients eligible or having had access to this treatment. Patients had access to the drug via clinical trials, special access programs from the sponsor or their government, scholarships, and insurance plans. In all instances, patients received larotrectinib after failing on multiple therapies, sometimes as a last treatment option, and had no possible alternative.

Overall, patients reported a reduction in tumour size, spread, and a general stabilization in cancer progression. After being on the treatment for a period ranging from a few months to a few years, most patients reported that they had achieved a complete or partial remission in all or nearly all cancer sites. In 1 particular case, a patient noted improved symptoms within days, and nearly full tumour shrinkage within a month. Few cases of recurrence were reported, although most noted a histological disappearance of cancer tissues with supplementary radiation or other therapies. In all cases, patients perceived larotrectinib as vastly superior to current treatment alternatives.

Patients also reported an improvement in their cancer- and treatment-induced symptoms after treatment with larotrectinib. Regardless of cancer type, patients reported that they felt both physically and mentally better, that their pain was resolved or greatly reduced, and that they were able to resume or return to daily activities, function, and participate in social activities and work. Some patients indicated larotrectinib treatment allowed them to progress from being in the last stage of life to being disease-free. Larotrectinib showed a similar effectiveness in pediatric patients, where the response was described as life-saving, with markedly improved disease response rates seldom observed with traditional therapies. In addition to the histological, physical, and mental improvement, both pediatric and adult patients greatly appreciated the convenience of the oral route of administration of larotrectinib, which they considered a feature that could make the uptake of this drug very high. The benefit of circumventing the need for IV chemotherapy or invasive surgery was also regarded as a key advantage of larotrectinib over standard of care, as the cost and inconvenience of commuting, hospitalization, and various complications associated with dosing and administration were causes of concern. Patients also welcomed the drug due to its targeted nature, particularly as larotrectinib is the only known therapy targeted at NTRK gene fusion.

Larotrectinib was also described as a relatively well-tolerated treatment, with few and manageable side effects. The most common side effects reported with larotrectinib treatment were myalgia, dizziness, diarrhea, mild constipation, withdrawal symptoms as the next dose approached, nerve sensations, skin tenderness, ear pain, minor or moderate fatigue, minor cough, general unwellness that resolved with a dose adjustment, and elevated liver enzymes. Following are some quotes from patients describing their experience with larotrectinib:

• “The worry that it won't be covered by our private insurance or our government insurance. If the tumours are inoperable and the other therapies won't work, the worry that Vitrakvi will not be available in Canada is very difficult.”

• “To be able to have a normal life and go to work, to the theatre, dinner, social events, charitable events, go to the gym, and go out with my partner is priceless.”

• “I had the most extraordinary response. I had so many tumours – they couldn’t count how many tumours I had in my lymph nodes, liver, kidneys etc. and they certainly couldn’t remove them surgically. I remember, at one point, they had doubled in size and number in just 5 days. They were growing and spreading so fast. I recall being so ill in hospital bed just before starting this Vitrakvi. I was told I had 3-4 weeks to live. I couldn’t even sit in a chair. They started me on the drug and in just one week, I felt remarkably better and I kept feeling better and better after that. After the first CT scan, all my tumours were gone except for one which had shrunk by 65%. Now that tumour has disappeared and is merely scar tissue. I no longer have any sign of cancer detectable through CT and there have been no new tumours. I am an NED patient because of Vitrakvi. I am fully restored.”

• “I wish I could have been spared dizziness, short-term memory loss and so much more from the other therapies I had to endure. This is a therapy that spares patients from toxic effects. I have had to give up driving because of brain radiation affecting cognitive skills and memory loss. I just didn’t want to kill someone or myself. This could have been avoided had I been able to access Vitrakvi. Others can be spared this.”

Companion Diagnostic Test

All patient groups unequivocally indicated the importance of having a testing platform to detect NTRK fusion mutations. The respondents indicated that testing early in the treatment scheme may allow them to receive this targeted therapy, thereby saving precious time that may be wasted on less-effective treatment. Given the indication of larotrectinib, an established testing platform with costs covered by public or private insurance was seen as imperative. Patients spoke of the ease of receiving genomic testing, with several patients indicating that faster results would be a welcome change.

Additional Information

Patients perceived the effectiveness and safety profile of larotrectinib as substantially better than current alternatives and valued its ease of administration. Among patients who often have no other effective treatment option, larotrectinib was considered a great step forward that offered much-needed hope.

Clinician Input

Input From the Clinical Expert Panel Consulted by CADTH

All CADTH review teams include at least 1 clinical specialist with expertise in the diagnosis and management of the condition for which the drug is indicated. Clinical experts are a critical part of the review team and are involved in all phases of the review process, providing guidance on the development of the review protocol, assisting in the critical appraisal of clinical evidence, interpreting the clinical relevance of the results, and providing guidance on the potential place in therapy. In addition, as part of the larotrectinib review, a panel of 5 clinical experts from across Canada was convened to characterize unmet therapeutic needs, assist in identifying and communicating situations where there are gaps in the evidence that could be addressed through the collection of additional data, promote the early identification of potential implementation challenges, acquire further insight into the clinical management of patients living with a condition, and explore the potential place in therapy of the drug (e.g., potential reimbursement conditions). A summary of this panel discussion follows, along with specific input received from the clinical experts consulted by CADTH.

Unmet Needs

Overall, the clinicians felt that it was difficult to specify the unmet need given the breadth of potential advanced solid tumours. However, they agreed that, in the case of metastatic solid malignancies, virtually all patients eventually progress on currently available therapies, with the possible exception of select patients receiving immunotherapies in select cancer types.

One clinician noted that the data to date for larotrectinib suggest a response rate that generally exceeds many available standard treatments. In some settings, there are no established and efficacious treatments, and larotrectinib would fill an unmet need. The following examples were provided in the clinician input: MASC, a rare malignancy with a very high prevalence of NTRK fusions and no established treatment options; differentiated thyroid cancer; anaplastic thyroid cancer, which has a dismal prognosis with standard cytotoxic treatments; and unresectable sarcomas in adults for whom currently available chemotherapy often does not offer enough response to facilitate surgery. The clinician also noted that chemotherapy results in toxicity not associated with NTRK inhibitors, further affecting quality of life.

The clinicians commented that, in the absence of access to selective NTRK inhibitors suitable for use in pediatric patients (which are currently available only through clinical trials and patient support programs in Canada), there is specifically unmet needs for the following groups:

• Children with unresectable or metastatic IFS and STS undergo cytotoxic chemotherapy with acute and long-term toxicities, and some may be unable to avoid disfiguring surgeries. Those with chemorefractory disease have no curative alternative therapy options.

• Children and adolescents with radioactive-iodine refractory NTRK fusion–positive thyroid cancers who may need repeated invasive surgeries but have no access to effective systemic therapies.

• Children with unresectable and progressive and/or symptomatic LGG who do not respond to conventional chemotherapy and may have additional disease-related morbidity (e.g., loss of vision due to progressive optic pathway glioma) or require multiples surgeries, and ultimately radiation therapy (when old enough).

• Children with NTRK fusion HGGs that are unresectable or refractory to chemotherapy and who have no alternative curative options.

Overall, the clinicians felt that an ideal treatment would prolong survival, which was noted as the most important goal for adult patients with advanced, incurable disease. Additional important treatment goals included minimizing toxicity, decreasing cancer-related symptoms (i.e., pain and shortness of breath), maintaining or improving quality of life, delaying disease progression, improving performance status, prolonging life, maintaining independence, and reducing burdens on caregivers. Additionally, the clinicians noted that an ideal treatment would be accessible to all age groups and would result in disease response facilitating curative surgery. This could prevent the need for aggressive and disfiguring surgeries (e.g., in patients with IFS or STS), neurocognitively devastating radiation therapy (e.g., in patients CNS tumours in young children) while avoiding the known significant long-term side effects of conventional alkylator and anthracycline chemotherapy in young children. The clinicians felt that achieving durable remissions and a cure in children with NTRK fusion cancers would reduce short- and long-term burdens on health care systems, and improve quality of life and survival.

Place in Therapy

In adult patients, given the risk of patient attrition across lines of therapy, the clinicians recommended that larotrectinib be considered early in the course of NTRK fusion cancer treatment. In most NTRK fusion cancers, this would be after a first line of standard therapy. One clinician added that this would be particularly relevant in settings with no, limited, or inefficacious treatments (e.g., when treating MASC, anaplastic thyroid cancer, or sarcoma with NTRK gene fusion).

For pediatric patients, 1 panellist indicated that larotrectinib would be considered as the first-line treatment of rare cancers with NTRK fusion, such as in patients with IFS or another STS who have unresectable or metastatic disease; as a second-line treatment in NTRK fusion–positive thyroid carcinoma patients after surgery and radioactive-iodine therapy; as a first- or second-line treatment (in the absence of a clinical trial) in patients with NTRK fusion–positive LGGs that are unresectable and progressive or symptomatic after failure of conventional chemotherapies; as a first-line treatment in patients with NTRK fusion–positive HGGs that are unresectable; and in a young child for whom radiation therapy is not feasible.

The clinicians noted that the appropriateness of recommending that patients try other treatments before initiating treatment with larotrectinib would depend on the cancer subtype and efficacy of front-line therapy. For example, in melanoma, immunotherapy can result in an excellent and durable response, and this would take precedence over larotrectinib.

Patient Population

Clinicians indicated that larotrectinib should be considered in adult and pediatric patients with a good performance status and advanced solid tumours that harbour an NTRK fusion. Patients with a CNS tumour are in high need of an intervention, and larotrectinib offers a targeted therapy with minimal morbidity when there are few or no other options. Patients with a relapsed and/or refractory, unresectable, or metastatic NTRK fusion–positive STS (including IFS) and HGG have the greatest need.

When asked how to identify the patients best suited for treatment with larotrectinib, the clinicians noted that, aside from the rare tumours in which NTRK fusions are common (such as IFS or MASC), they are relatively rare in common cancers, and a diagnostic testing algorithm would therefore be required. Given the rarity, targeting tumour NTRK assessment in enriched populations may also be required. Specifically, routine clinical testing (i.e., FISH) establishes NTRK fusion status in classic histologies (e.g., IFS and CMN). Fusion panels in NGS are routinely used for other STS cases at diagnosis and include NTRK coverage. In patients with glioma who lack BRAF (V600E) or BRAF fusion, H3K27M mutations will require NGS testing that includes NTRK1, NTRK2, and NTRK3 gene fusions. Patients with differentiated thyroid cancer that is either unresectable or progressive and/or symptomatic after surgery and refractory to radioactive-iodine therapy will receive NGS screening. Reflex testing should also be considered. The clinicians noted that patients with poor performance status, those who are unable to tolerate oral therapies, patients who lack NTRK gene fusions, and patients who have resectable disease would be least suitable for treatment with larotrectinib. In adult patients, the clinicians were not aware of any additional predictive biomarkers of efficacy for larotrectinib beyond the presence of NTRK gene fusion.

Assessing Response to Treatment

In terms of outcomes that are used to determine whether a patient is responding to treatment in clinical practice, the clinicians indicated that typical metrics of treatment efficacy include disease evaluation by cross-sectional imaging modalities (MRI, CT, PET/CT) to assess response by RECIST (for solid tumours) or RANO (for CNS tumours), symptom improvement, treatment tolerability, and time to progression. In addition, for patients with differentiated thyroid cancer, thyroglobulin levels are monitored.

The clinicians noted that, in adults, objective response, non-progression, patient-reported improvements in their ability to perform activities, improved survival, stabilization, improvement, or reduced severity of symptoms would all be considered clinically meaningful outcomes. In pediatric patients, a clinically meaningful response to treatment would include improved survival; avoidance of aggressive and disfiguring surgeries or amputation; avoidance of neurocognitively devastating radiation therapy; improvement in symptoms, quality of life, and other patient- or parent-reported outcomes; and avoidance of long-term side effects of therapy for survivors of childhood cancer, such as cardiomyopathy (anthracycline), infertility (alkylators), second malignancy (alkylators, radiation, and radioactive iodine), pulmonary fibrosis (alkylators and radiation), or neurocognitive impairment (radiation therapy).

Clinicians noted that treatment response is typically assessed every 3 months in adults and pediatric patients, and once response is established or remission is achieved, this interval may be prolonged.

Discontinuing Treatment

In adult patients, the clinicians noted that treatment failure would be determined by disease progression, treatment intolerability, or patient request to discontinue treatment. In pediatric patients, the clinicians noted that therapy could be stopped if a response was achieved that facilitated curative surgery, or imaging evaluations produced objective evidence of disease progression. However, the optimal duration of treatment in children with a CNS tumour is still not known.

Prescribing Conditions

The clinicians reported that all treatment settings are appropriate for larotrectinib administration in adults, given that it is an oral therapy. They also noted that larotrectinib therapy for pediatric patients should be supervised by a subspecialty of pediatric oncology, or an endocrinology team based at 1 of 16 Canadian pediatric oncology programs.

The clinicians also noted the requirement for NTRK fusion status of the tumour to be documented. This can be done with FISH (which is routine in IFS or CMN) or NGS panels (for non-IFS, STS, glioma, and differentiated thyroid tumours), and that immunohistochemistry (IHC) screening has utility in non-CNS histologies.

Clinician Group Input

This section was prepared by CADTH staff based on the input provided by clinicians.

A total of 9 clinician group inputs were received for the review of larotrectinib for solid tumours harbouring a NTRK gene fusion. All were joint clinician inputs, comprised primarily of oncologists, from the following groups: Advanced Thyroid Cancer (ATC; 4 clinicians); Canadian Gastrointestinal Oncology Evidence Network (CGOEN) with the Medical Advisory Board of Colorectal Cancer Canada (CCC) and other GI cancer-treating clinicians (19 clinicians and pathologists); Ontario Health Cancer Care Ontario (OH-CCO) Gastrointestinal Cancer Drug Advisory Committee (DAC) (4 clinicians); OH-CCO Head, Neck, and Thyroid Cancer (HNTC) DAC (3 clinicians); LCC (1 clinician on behalf of 18 clinicians), OH-CCO Lung and Thoracic Cancer (LTC) DAC (4 clinicians), Pediatric Oncology Group of Ontario (POGO) (8 clinicians), Pediatric Oncology Group (POG) (3 clinicians); and OH-CCO Skin Cancer DAC (6 clinicians). In total, input was received from 52 oncologists.

Unmet Needs

A common theme was echoed across clinician groups with respect to treatment goals and unmet needs. Irrespective of tumour histology and subtype, the goal of treatment was to minimize cancer and treatment-induced symptoms, as well as to achieve meaningful improvement in the following areas: life expectancy and survival, PFS, response rate, HRQoL, control or delay of disease progression, and duration of response. The tolerability of treatment was another key treatment goal shared unanimously by the clinician groups. Several clinician groups, including the OH-CCO LTC DAC and CGOEN, indicated the need for a therapy that specifically targets the underlying tumour biology or disease-causing factor(s), in this case, NTRK fusion. In addition, several clinician groups referred to a number of treatment goals specific to some cancer types. These included prevention of brain metastases and skeletal-related events in the case of advanced thyroid cancer and minimizing the life-long impact of surgery, particularly in pediatric cases (i.e., IFS), even when such interventions were feasible. Clinicians noted that improvements in patients were expected to reduce caregiver burden, another key treatment goal.

All clinician groups agreed that patients with advanced stages of cancer almost always failed on multiple therapies, leaving them with limited effective therapeutic options (i.e., therapies that extend survival without compromising quality of life), or any treatment at all. The following cancer types were noted as particularly nonresponsive to current therapies or as types that inevitably become refractory over time: stage IV or advanced NSCLC, advanced colorectal cancer, gastrointestinal cancer, and some forms of skin cancer. Example of tumours with little or no available treatment options include thyroid cancer, salivary gland tumour, and MASC of the salivary glands. For some cancer types, treatment options were based on low-quality evidence and associated with significant toxicity in addition to low efficacy. Pediatric patients have similar unmet needs. Infants, children, and adolescents with metastatic tumours such as sarcomas and thyroid malignancies have minimal treatment options to provide even transient disease response; the need for an effective and low-toxicity therapy is not met. Treatments for other HGGs, such as palliative chemotherapy and radiation therapy for intracranial HGGs, were considered less effective in sustaining response, stabilizing, or reducing disease burden. Some patients have an unmet need for an effective therapy at second or subsequent lines. Finally, the lack of a targeted therapy for NTRK gene fusion was a notable unmet need. Clinicians agreed that a treatment based on tumour genomics, rather than the tissue of origin, would constitute an important advance.

Place in Therapy

All clinicians agreed that larotrectinib should be used as a monotherapy. With respect to which line of therapy larotrectinib should be used in, their responses varied, depending on cancer type and underlying cause, stage, and the availability of the treatment, as well as the efficacy and toxicity of treatment alternatives. While the current practice and the indication of larotrectinib is for use after all other treatment options have been exhausted, clinicians noted that larotrectinib could be prioritized in certain circumstances.

First, some clinicians suggested that larotrectinib could be used as a first-line therapy for NTRK fusion cases because its mechanism of action targets the driver mutation (e.g., in NTRK-positive NSCLC, gastrointestinal cancer, differentiated thyroid cancer, or advanced anaplastic thyroid cancer).

Second, clinicians suggested that larotrectinib may be used early in a treatment scheme when 1 or more of the following circumstances are present: absence of any treatment alternative, certain cases of refractory tumours, or cancers with poor prognosis. Examples include salivary gland tumours, advanced radioactive-iodine refractory thyroid cancers, and anaplastic thyroid cancers, respectively. The LCC group noted that first-line use of larotrectinib would be more likely if diagnosed with NGS analysis, and that it provides an opportunity to avoid the potential toxicities of CNS radiotherapy as CNS metastasis is a common occurrence in these patients.

With the exception of these circumstances, clinicians indicated that larotrectinib would generally be used as second- or subsequent-line therapy. For example, larotrectinib was expected to be used after chemotherapy and immunotherapy in lung cancer, or after conventional therapies in gastrointestinal cancer. In pediatric cases, larotrectinib was expected to be used as a second-line treatment in diseases with good prognoses and as a front-line option in diseases with poorer prognoses.

Patient Population

All clinician groups indicated that the patient populations with the greatest unmet need for larotrectinib treatment would be those harbouring NTRK fusion, and that this is also the group expected to be best suited for this treatment. The targeting of the driver mutation by larotrectinib was regarded as the primary reason for the clinicians’ expected benefit of the drug. This was noted irrespective of the cancer type and location, including melanoma and salivary gland, gastrointestinal, lung, and thyroid cancers. The rarity of NTRK fusion in otherwise common cancers such as those of the lung, thyroid, and gastrointestinal system were noted as an important reason for the lack of treatment options in this subpopulation. In some cases, (e.g., in metastatic CRC and advanced radioactive-iodine refractory anaplastic thyroid cancer with NTRK fusion), patients reportedly have an extremely poor prognosis, highlighting the unmet need for treatment for these populations. Aside from NTRK status, patients with a good performance status were considered the most suitable candidates. In the case of pediatric cancers, patients with a poor prognosis were candidates for larotrectinib, unlike those with favourable prognosis. This includes infants with IFS or high-grade intracranial gliomas, if NTRK fusion mutation was present.

As with treatment suitability, the overwhelming reason for not considering larotrectinib treatment was the absence of an NTRK mutation. With respect to treatment unsuitability due to contraindication, clinicians in the LCC group noted that patient age, performance status, or comorbidities were more frequently a contraindication to current standard of care options. In children, patients with an alternative low-morbidity curative option may be least suitable for front-line therapy with larotrectinib given the lack of long-term efficacy data.

Patients would be selected for treatment by testing for NTRK through companion diagnostics. Clinicians noted that a number of testing platforms and programs are available in Canada. Testing NTRK using NGS panels was described as the optimal test, although several clinician groups reported different methods for optimizing patient selection. In addition to tumour testing for NTRK fusion, clinical judgment was considered necessary to identify patients in an efficient manner. Some clinician groups indicated that IHC is carried out first to identify NTRK mutations, followed by confirmation with NGS. The CGOEN group noted a diagnostic framework to improve the selection of patients with NTRK fusion involving testing in high-risk groups such as patients with BRAF/RAS wild-type, high-level MSI colorectal carcinoma, and MLH1-promoter hypermethylation. Clinical judgment was also noted as key to identifying patients with a higher likelihood of NTRK fusion (i.e., those with characteristics in line with larotrectinib treatment eligibility). For pediatric patients, NTRK testing is routinely carried out for histologies in which fusion is common, specifically IFS, CMN, SBC, and MASC. However, in cancers with a low frequency of NTRK alterations (i.e., HGGs, metastatic sarcomas, and metastatic papillary thyroid cancers), NTRK positivity may be detected as part of broader tumour-sequencing efforts or by targeted assessment by IHC or a genetic assessment.

Among the currently available platforms, Bayer’s FAST-TRK program, and several provincial and local genomic testing programs, such as the Ontario-wide Cancer TArgeted Nucleic Acid Evaluation, were noted by the clinicians. Some groups indicated that many other centres currently perform NGS panels, and they expected that testing will increase with time as multiple molecular subtypes of various cancers (e.g., EGFR, ALK, ROS1, BRAF, KRAS G12C, Her2, c-Met exon14, RET fusion and NTRK fusions) will be incorporated in the same testing platforms.

Assessing Response to Treatment

Clinicians noted that the outcomes used in clinical practice align with the those typically used in clinical trials. These include a combination of any of the following: clinical assessment of OS, PFS, treatment tolerance, and quality of life; blood work; imaging (CT, MRI or PET); tumour markers (thyroglobulin in differentiated thyroid cancers); radiologic response; and performance status.

The following were noted by clinicians to be clinically meaningful responses to treatment: stabilization, no deterioration of symptoms, reduction in disease-related symptoms (e.g., pain, dyspnea, fatigue, and weight loss), overall reduction in tumour burden, and improvement in survival. In addition, an improved sense of well-being, being more active clinically, and being able to carry out more activities of daily living were important indicators of treatment efficacy. In pediatric cancer, clinically meaningful responses can vary given the variety of histologies and presentations in the potential patient population. For example, in patients with IFS, clinically meaningful responses would include the facilitation of a low-morbidity resection. In patients with an HGG, a clinically meaningful response may be stable disease and/or improvement of existing neurologic deficits.

With respect to the frequency of monitoring treatment responses, clinicians noted a period of 2 to 6 months, depending on the clinical presentation and type of assessment performed. Clinical assessment should be carried out every 2 to 4 weeks, although 4 to 8 weeks was noted by some clinicians. Assessments should be conducted at the following time points: every 2 to 3 months for imaging (although CT scans could be done every 3 to 6 months), every 4 weeks for tumour markers, every 8 to 12 weeks for radiology; and every 4 weeks for bloodwork and physical exams. Early assessment (i.e., 2 to 4 months) was favoured in patients initiating therapy. Patients showing sustained responses may require assessment less frequently, particularly if there are challenges such as sedation requirements around radiologic assessment.

Discontinuing Treatment

Overall, clinicians noted disease progression and certain unacceptable toxicities as the key factors when considering treatment discontinuation. Progression of disease should be based on deterioration in clinical status and objective progression on imaging. Significant AEs requiring dose interruption or discontinuation could vary by cancer and treatment type; the advanced thyroid cancer group noted grade 3 or higher elevated liver enzymes, and grade 3 or higher neurologic AEs (i.e., delirium, dysarthria, dizziness, gait disturbance, and/or paresthesia) as examples of unacceptable side effects. The Pediatric Oncology Group of Ontario noted that decisions to stop therapy in patients with a clinical response should be made in conjunction with patients and their families, considering both the clinical impact of therapy and the acceptability of associated toxicities.

Prescribing Conditions

Overall, clinicians noted disease progression and certain unacceptable toxicities as the key factors when considering treatment. Patients will be assessed and prescribed larotrectinib at cancer centres, with the medication administered at home in an outpatient setting as it is an oral medication and does not require a medical setting for administration. For pediatric patients, a specialized pediatric cancer program may be necessary.

Drug Program Input

The drug programs provide input on each drug being reviewed through CADTH’s reimbursement review processes by identifying issues that may affect their ability to implement a recommendation. The implementation questions and corresponding responses from the clinical experts consulted by CADTH are summarized in Table 6.

The PAG includes representatives from provincial cancer agencies and provincial and territorial Ministries of Health participating in pCODR. The complete list of PAG members is available on the CADTH website (www.cadth.ca/pcodr). The PAG identifies factors that could affect the feasibility of implementing a funding recommendation.

Overall Summary

Input was obtained from all 9 provinces (ministries of health and/or cancer agencies) participating in pCODR. The PAG identified the following as factors that could affect implementation:

• clinical factors:

  • place in therapy for larotrectinib

• economic factors:

  • additional health care resources may be required to monitor and treat toxicities

  • number of patients requiring and access to NTRK gene fusion testing.

More details are supplied in the following section.

Currently Funded Treatments

The PAG identified that there is no standard of care for the treatment of adult and pediatric patients with locally advanced or metastatic solid tumours harbouring NTRK gene fusion. Treatment is dependent on the specific type of solid tumour, and clinical trials may be offered to patients harbouring NTRK gene fusion. For patients who have experienced disease progression on all available treatment options, best supportive care would be available. The PAG concluded that the relevant comparator for this drug submission would be best supportive care. Because best supportive care is meant to alleviate patient discomfort, the PAG appreciates a demonstration that larotrectinib improves quality of life or other outcomes meaningful to patients, relative to usual care.

Eligible Patient Population

The 3 pivotal trials of larotrectinib are a phase I study in adult patients with solid tumours, the SCOUT trial in advanced pediatric solid or primary CNS tumours, and the NAVIGATE trial in patients with NTRK fusion–positive solid tumours. The PAG observed that this resubmission to CADTH includes an expanded pool of patients, including those with an ECOG status of 0 to 2, an LPS score of greater than 40%, or a KPS score of greater than 50%. The PAG is seeking guidance on the use of larotrectinib in patients with a poor performance status (i.e., an ECOG status > 2, an LPS score < 40%, and a KPS score < 50%).

If larotrectinib is recommended for reimbursement, the PAG noted that patients harbouring NTRK gene fusion and currently on other treatments would need to be addressed on a time-limited basis. The PAG identified potential indication creep in the use of larotrectinib in earlier lines of treatment or when other lines continue to be available, and as an alternative to surgical resection (i.e., not only for patients in whom surgical resection would result in severe morbidity).

Implementation Factors

Depending on other treatment options, larotrectinib may be associated with less chair time, which would be an enabler to implementation. Larotrectinib is available as a capsule or oral solution formulation. The PAG noted that the oral solution formulation, particularly for pediatric patients or those unable to take the capsule form, would be an enabler of implementation. However, dispensing larotrectinib would require additional pharmacy resources.

Additional health care resources (e.g., frequent clinic visits while patients are on therapy) are required to monitor adverse effects and tolerability with larotrectinib.

Larotrectinib is an oral drug that can be delivered to patients more easily than can IV therapy in both rural and urban settings, as patients can take oral drugs at home. PAG identified that the oral route of administration is an enabler of implementation. However, in some jurisdictions, oral medications are not funded through the same mechanism as IV cancer medications. This may limit accessibility of treatment for patients in these jurisdictions as they would first require an application to their pharmacare program and these programs can be associated with copayments and deductibles, which may cause a financial burden on patients and their families. Other coverage options in those jurisdictions that fund oral and IV cancer medications differently include private insurance coverage or full out-of-pocket expenses.

Sequencing and Priority of Treatments

PAG recognizes that larotrectinib would be positioned as last-line therapy. As such, patients would experience a customary series of tumour-specific treatments, culminating with this drug upon failure of all options.

Companion Diagnostic Testing

PAG noted that NTRK gene fusion testing is not routinely available in all provinces. Some jurisdictions do not currently have testing for NTRK gene fusion available, and other options, such as sending tissue samples out of province, would need to be explored. As there is no formalized testing process or funding in place for NTRK gene fusion testing in all jurisdictions, this would be a barrier to implementation. Health care resources and coordination to conduct NTRK gene fusion testing will also be required. The increase in costs for NTRK gene fusion testing is a barrier to implementation.

PAG had concerns related to:

• the turnaround time for NTRK gene fusion testing

• guidelines on criteria for testing and whether all patients should be tested

• the expected number of patients eligible for larotrectinib (e.g., anticipated number of patients requiring testing per year, with tumours harbouring a NTRK gene fusion, and who would receive larotrectinib treatment)

• timing of testing and whether patients should be tested at diagnosis or at relapse.

The number of patients requiring access to NTRK gene fusion testing may therefore be a barrier to implementation.

Additional Information

None.

Clinical Evidence

The clinical evidence included in the review of larotrectinib is presented in 3 sections. The first section, the systematic review, includes pivotal studies provided in the sponsor’s submission to CADTH and Health Canada, as well as those studies that were selected according to an a priori protocol. The second section includes alternative methodologies to evaluate the ORR and PFS from the sponsor. No indirect evidence met the inclusion criteria for this review.

Systematic Review (Pivotal and Protocol-Selected Studies)

Objectives

To perform a systematic review of the beneficial and harmful effects of larotrectinib (adults: 100 mg taken orally twice daily; pediatric patients: 100 mg/m² taken orally twice daily up to a maximum of 100 mg per dose) for the treatment of adult and pediatric patients with solid tumours that have NTRK gene fusion without a known acquired resistance mutation, are metastatic or for whom surgical resection is likely to result in severe morbidity, and have no satisfactory treatment options.

Methods

Studies selected for inclusion in the systematic review included pivotal studies provided in the sponsor’s submission to CADTH and Health Canada, as well as those meeting the selection criteria presented in Table 7. Outcomes included in the CADTH review protocol reflect outcomes considered to be important to patients, clinicians, and drug plans.

The literature search for clinical studies was performed by an information specialist using a peer-reviewed search strategy according to the Peer Review of Electronic Search Strategies checklist (https://www.cadth.ca/resources/finding-evidence/press).⁴⁵

Published literature was identified by searching the following bibliographic databases: MEDLINE All (1946‒) via Ovid and Embase (1974‒) via Ovid. The search strategy comprised both controlled vocabulary, such as the National Library of Medicine’s MeSH (Medical Subject Headings), and keywords. The main search concept was Vitrakvi (larotrectinib). Clinical trials registries searched included the US National Institutes of Health’s clinicaltrials.gov, WHO’s International Clinical Trials Registry Platform (ICTRP) search portal, Health Canada’s Clinical Trials Database, European Union Clinical Trials Register, and Canadian Partnership Against Cancer Corporation’s Canadian Cancer Trials.

No filters were applied to limit the retrieval by study type. Retrieval was not limited by publication date or by language. Conference abstracts were excluded from the search results. See Appendix 1 for the detailed search strategies.

The initial search was completed on December 10, 2020. Regular alerts updated the search until the meeting of the pERC on April 15, 2021.

Grey literature (literature that is not commercially published) was identified by searching relevant websites from the Grey Matters: A Practical Tool For Searching Health-Related Grey Literature checklist (https://www.cadth.ca/grey-matters).⁴⁶ Included in this search were the websites of regulatory agencies (US FDA and European Medicines Agency). Google was used to search for additional internet-based materials. Appendix 1 provides more information on the grey literature search strategy.

These searches were supplemented by reviewing bibliographies of key papers and through contacts with appropriate experts. In addition, the manufacturer of the drug was contacted for information regarding unpublished studies.

Two CADTH clinical reviewers independently selected studies for inclusion in the review based on titles and abstracts, according to the predetermined protocol. Full-text articles of all citations considered potentially relevant by at least 1 reviewer were acquired. Reviewers independently made the final selection of studies to be included in the review, and differences were resolved through discussion.

Findings From the Literature

No studies were identified from the literature for inclusion in the systematic review (Figure 1). Three unique studies and 1 pooled analysis of the studies were identified as part of the CADTH submission. The included studies are summarized in Table 8.

Figure 1: Flow Diagram for Inclusion and Exclusion of Studies

Description of Studies

This CADTH review included 3 open-label, single-arm trials of larotrectinib in adult and pediatric patients with advanced or metastatic solid tumours: the LOXO-TRK-14001, LOXO-TRK-15003 (SCOUT), and LOXO-TRK-15002 (NAVIGATE) trials. All 3 trials are ongoing; however, the reimbursement submission for larotrectinib is supported with pooled analyses of efficacy and safety data from NTRK fusion cancer patients enrolled in these 3 trials. After consideration of the methodological challenges attributable to the rarity of NTRK-positive solid tumours and the multiplicity of tumour types in which NTRK gene fusions can occur, the sponsor made the decision to pool efficacy data across the 3 trials from patients with NTRK fusion cancer based on global regulatory advice. The sponsor’s rationale for pooling the data included the common eligibility criteria and study procedures as well as the consistency of treatment response, safety, and tolerability across tumours and age groups. The current report represents a resubmission to CADTH with a larger pooled dataset and longer patient follow-up than what was provided for the original submission to CADTH.²³

LOXO-TRK-14001

The LOXO-TRK-14001 study is a multi-centre, open-label, ongoing (initiated in May 2014) phase I dose escalation and expansion trial in adult patients with an advanced or metastatic solid tumour.⁶¹ The trial was conducted at 8 centres in the US. The first part of the trial (dose escalation) enrolled adults with metastatic solid tumours (regardless of NTRK gene fusion status) to 6 cohorts. The expansion part, which is ongoing, includes 2 expansion cohorts: 1 in patients with an alteration in the NTRK1, NTRK2, or NTRK3 genes (rearrangement, fusion, or mutation), and 1 in patients without known NTRK alterations. During the dose escalation phase, patients received larotrectinib in doses of 50 mg once daily to 200 mg twice daily. During the expansion phase, patients received larotrectinib at 100 mg twice daily.

LOXO-TRK-15003 (SCOUT Trial)

The SCOUT trial is a multi-centre, open-label, phase I/II trial in pediatric patients with locally advanced or metastatic solid tumours or primary CNS tumours.⁶² The study was conducted internationally and consisted of 2 parts. The phase I dose escalation part of the study was designed to identify the maximum tolerated dose (MTD) through characterization of safety, specifically dose-limiting toxicity (DLT). The phase I portion of the trial used a sequential rolling 6-dose escalation design. Escalation in phase I was to proceed through 5 planned dose levels or until the MTD was reached. When the optimal dose was identified, a phase I expansion cohort of up to 18 patients was planned to further define the safety profile. Phase II aimed to determine the ORR in pediatric patients with an advanced cancer harbouring NTRK fusion who received larotrectinib and was conducted in 3 selected cohorts of pediatric patients with tumours with NTRK fusions: IFS, other extracranial solid tumours, and primary CNS tumours. The recommended phase II dose was 100 mg/m² twice daily, not to exceed a dose of 100 mg twice daily.

LOXO-TRK-15002 (NAVIGATE Trial)

The NAVIGATE trial is an ongoing open-label, phase II, multi-centre trial in patients with advanced cancer harbouring fusion of NTRK1, NTRK2, or NTRK3.⁶³ The trial consisted of 9 cohorts of patients with solid tumours bearing NTRK fusions, including: NSCLC, thyroid cancer, sarcoma, colorectal cancer, salivary gland cancer, biliary cancer, primary CNS tumour, all other solid tumour types with evaluable but not measurable disease, and patients with NTRK gene fusion without laboratory Clinical Laboratory Improvement Amendments (CLIA)-equivalent certification. Patients were administered larotrectinib at 100 mg twice daily in 28-day cycle increments.

Pooled Analysis

The pooled analyses included data from patients with NTRK gene fusions enrolled in each of the 3 clinical trials.²² The datasets are detailed later in this report. Briefly, datasets included a primary analysis dataset (PAS), extended PAS (ePAS), ePAS4, and ePAS5; a primary CNS tumour analysis set (SAS3); and safety analysis sets (TRK fusion cancer safety analysis set, overall safety analysis set, TRK fusion cancer labelled-dose safety analysis set, and overall labelled-dose safety analysis set). Figure 2 provides an overview of these different datasets.

The current resubmission is based primarily on ePAS5, ePAS4, and SAS3 for efficacy outcomes and on the TRK fusion cancer labelled-dose safety set and the overall labelled-dose safety set for safety outcomes.

Populations

Inclusion and Exclusion Criteria

LOXO-TRK-14001

The LOXO-TRK-14001 trial included adult patients (≥ 18 years of age) with a locally advanced or metastatic solid tumour that had progressed, was nonresponsive to available therapies, was unfit for standard chemotherapy, or for which no standard or available curative therapy existed.⁶¹ Patients also must have had an ECOG performance status of 0 to 2 and a life expectancy of at least 3 months. Although NTRK gene fusion status was not an inclusion criterion for the trial, only patients with NTRK-positive gene fusion were prospectively selected for inclusion for the pooled analysis informing the main clinical evidence in the CADTH review. For the NTRK expansion cohort, evidence of NTRK gene fusion was assessed before enrolment in a CLIA-certified or equivalent local laboratory. Patients were excluded if they had symptomatic brain metastases without CNS disease or active spinal cord compression, clinically significant active cardiovascular disease or a history of prolonged QT interval, or if they were on treatment with a strong cytochrome P450 (CYP) 3A4 inhibitor or inducer.

LOXO-TRK-15003 (SCOUT Trial)

In phase I, the SCOUT trial included: infants, children, and adolescents aged 1 month to 21 years with locally advanced or metastatic solid tumours or CNS tumours that had relapsed, progressed, or were nonresponsive to available therapies and for which no standard or available systemic curative therapy existed; infants from birth with a diagnosis of malignancy and documented NTRK fusion that had progressed or was nonresponsive to available therapies, and for which no standard or available curative therapy existed; and patients with locally advanced IFS that would otherwise require disfiguring surgery or limb amputation to achieve a complete surgical resection.⁶² In the phase I dose expansion, in addition to the previously listed inclusion criteria, patients must have had a malignancy with documented NTRK gene fusion, with the exception of patients with IFS, CMN, or SBC. Patients with IFS, CMN, or SBC may have been enrolled with documentation of an ETV6 rearrangement by FISH or reverse transcriptase–polymerase chain reaction (RT-PCR), or documented NTRK fusion by NGS. In phase II, the SCOUT trial included: patients from birth with either locally advanced or metastatic IFS that required disfiguring surgery or limb amputation to achieve a complete surgical resection; patients from birth through 21 years with locally advanced or metastatic solid tumours or CNS tumours that had relapsed, progressed, or were nonresponsive to available therapies and for which there was no standard or available systemic curative treatment and with documented NTRK gene fusion; and patients older than 21 with a tumour diagnosis with histology typical of a pediatric patient and NTRK fusion who could have been considered for enrolment following discussion between the local site investigator and the sponsor’s medical monitor.

To be eligible patients were required to have a Karnofsky (if ≥ 16 years of age) or Lansky (if < 16 years of age) performance status score of 50 or more, evaluable or measurable disease (as defined by RECIST, RANO, or International Neuroblastoma Response Criteria), adequate organ function, and full recovery from the acute toxic effects of all previous anticancer therapy. NTRK gene fusion status was not part of the eligibility criteria for all patients of the trial.⁶²

LOXO-TRK-15002 (NAVIGATE Trial)

The NAVIGATE trial included: patients 12 years of age and older with locally advanced or metastatic cancer with an NTRK1, NTRK2 or NTRK3 gene fusion, identified through molecular assays routinely performed at laboratories certified by CLIA or similar laboratories (exception: patients with NTRK gene fusion identified in a laboratory where certification of the laboratory could not be confirmed by the sponsor at the time of consent were eligible for cohort 9); patients who had received prior standard therapy or, in the opinion of the investigator, would be unlikely to tolerate or derive clinically meaningful benefit from appropriate standard of care therapy; and those who had at least 1 measurable lesion as defined by RECIST 1.1 (exception: patients without RECIST 1.1–measurable disease [evaluable disease only] were eligible for enrolment in cohort 8, regardless of tumour type).⁶³ Patients in cohort 7 (primary CNS tumours) must have received prior treatment, including radiation and/or chemotherapy, have at least 1 site of bi-dimensionally measurable disease (per RANO criteria), and have had an imaging study performed within 28 days before enrolment. Patients must have had an ECOG status of 0 to 3 (for patients ≥ 16 years old) or an LPS score of 40% or greater (if < 16 years old). If the patient was enrolled with a primary CNS tumour to be assessed by RANO, a Karnofsky (if ≥ 16 years of age) or Lansky (if < 16 years of age) score of 50 or more was required. Patients were excluded if they had symptomatic or unstable brain metastases (patients with asymptomatic brain metastases and patients with primary CNS tumours were eligible), unstable cardiovascular disease, and an inability to discontinue treatment with a strong CYP3A4 inhibitor or inducer before treatment initiation.

Pooled Analysis

Adults and pediatrics enrolled across the three3 larotrectinib studies (LOXO-TRK-14001, LOXO-TRK-15003 [SCOUT], and LOXO-TRK-15002 [NAVIGATE] trials) were included in the pooled efficacy analysis if they met all the following criteria²²:

• documented NTRK gene fusion as determined by local testing

• non-CNS primary tumour with 1 or more measurable lesions at baseline that could be assessed according to RECIST, version 1.1

• received 1 or more doses of larotrectinib.

The PAS consisted of the first 55 consecutively enrolled patients across the 3 larotrectinib studies (LOXO-TRK-14001, LOXO-TRK-15003, and LOXO-TRK-15002) meeting the aforementioned criteria.⁵⁷ The ePAS4 (the pooled efficacy dataset relevant for this review) included patients in the PAS plus patients who subsequently started treatment by February 19, 2019, and fulfilled the definition of the PAS in the Description of Studies section. The initial evaluation of ePAS4 had been performed using data from the July 15, 2019, cut-off date. The ePAS5 dataset, the most updated efficacy dataset, had a July 2020 cut-off.

All patients were included if they met these criteria. Patients in ePAS4 as well as ePAS5 must also have had an IRC-assessed response. Of the 208 patients included in the TRK fusion cancer safety analysis set (the pooled safety dataset relevant for this review), patients were excluded from ePAS4 for the following reasons: 11 patients had not yet received an IRC assessment, 9 did not have RECIST-measurable disease at baseline, and 24 had a primary CNS tumour. The primary CNS tumour patients were included in a separate CNS tumour dataset (SAS3).

The SAS3 population consisted of patients with primary CNS tumours.²² These patients were excluded from the overall analysis sets and analyzed separately for the following reasons:

• Patients with a primary CNS tumour were evaluated using either RANO or RECIST 1.1, whereas all other tumour types were evaluated using RECIST 1.1 only.

• The effect of edema, inflammation, and scarring from surgery and radiation on the radiological assessment was a concern in these patients.

• Data for these patients were not IRC-assessed (only investigator assessment).

Baseline Characteristics

Pooled Analysis

ePAS4: A total of 164 patients (13, 53, and 98 patients from LOXO-TRK-14001, LOXO-TRK-15003, and LOXO-TRK-15002, respectively) were included in ePAS4 (Table 9).²² Patient ages ranged from 0.1 to 84.0 years, with a median of 42.0 years. Most patients had an ECOG status of 0 (49%) or 1 (38%). Approximately half of the patients were male (49%) and had received 1 to 2 previous systemic chemotherapies, (41%). The most common tumour types were STS (22%), IFS (20%), thyroid (16%), or salivary gland (13%). Most patients (74%) had metastatic disease at enrolment.

SAS3

A total of 24 patients with an investigator-assessed data cut-off of July 15, 2019, (Table 10) were included in SAS3.²² The primary CNS tumour types included glioblastomas (n = 7), gliomas (n = 6), astrocytomas (n = 5), not otherwise specified (n = 3), glioneuronal (n = 1), neuronal and mixed neuronal-glial tumours (n = 1), and primitive neuroectodermal tumours (n = 1). The median age was 8 years (range = 1.3 to 79 years), and pediatric patients accounted for 83% (n = 20) of patients. Approximately half (46%) of patients were male, and most had an ECOG status of 0 (63%) or 1 (29%). Most patients had received 1 to 2 lines of previous systemic treatment (67%) and had an NTRK2 gene (67%). Baseline characteristics were not available for SAS3 (n = 33) with a July 2020 data cut-off.

Safety Sets

As of the July 15, 2019, data cut-off date, the median time on treatment for the TRK fusion cancer labelled-dose safety set was 9.3 months (range = 0.1 to 51.6 months) and 7.4 months (range = 0.03 to 51.6 months) in the overall labelled-dose safety set.²² The median age was 37.5 years (range = 0.1 to 84.0 years) with 37% of patients younger than 18 years of age in the TRK fusion cancer labelled-dose safety set and was 46.0 years (0.1 to 84.0 years) with 33% of patients younger than 18 years of age in the overall labelled-dose safety set.

Interventions

LOXO-TRK-14001

In the dose escalation phase, increasing dose levels (i.e., 50 mg daily, 100 mg daily, 100 mg twice daily, 200 mg daily, 150 mg twice daily, 200 mg twice daily) were used according to the occurrence of DLT in cycle 1, or until the MTD was reached or the sponsor determined that a suitable dose had been achieved (based on available safety, clinical activity, and pharmacokinetic exposure).⁶¹ The starting dose of 50 mg once daily was determined based on data from animal toxicity studies. Patients in the expansion cohorts were treated at the MTD, or at a dose level lower than the identified MTD but deemed by the sponsor to provide significant TRK inhibition. Larotrectinib was administered orally once or twice daily, based on 28-day cycles. Treatment was continued until progression, unacceptable toxicity, or patient withdrawal. Dose interruptions were allowed of up to 4 weeks for clinically significant AEs. After recovery, patients could either continue at the assigned dose of larotrectinib or receive a reduced dose. Patients with drug-related toxicities requiring more than 4 weeks to recover were to be withdrawn from the trial unless there was compelling evidence of response and no alternative treatment.

LOXO-TRK-15003 (SCOUT Trial)

During phase I dose escalation, patients received larotrectinib orally (capsule or liquid formulation).⁶² For cohorts 1 and 2, the starting dose was determined by pharmacokinetic (SimCyp dose escalation) modelling, which took both age and BSA into consideration. Because interim analysis of pharmacokinetic data from cohorts 1 and 2 indicated that a BSA-based dose would provide a more consistent pharmacokinetic response, doses for patients enrolled in phase I dose escalation cohort 3, phase I dose expansion cohort, and phase II cohorts were dosed based on the calculated BSA at day 1 of each cycle visit. The recommended phase II dosage (100 mg/m² twice daily, not to exceed a dosage of 100 mg twice daily) was declared on April 13, 2017. In patients aged between 0 and 1 month, a dose of 1 mg/kg was used. If no DLTs were reported during cycle 1, patients were increased to the recommended phase II dose on day 1 of cycle 2.

Larotrectinib was administered in 28-day cycles.⁶² Treatment was continued until progression, unacceptable toxicity, or other reasons as outlined in the protocol. Patients with locally advanced disease and a response of at least a PR for a period of 1 year after PR was initially confirmed may have had their larotrectinib treatment held. Patients with a confirmed PR were permitted to undergo optional biopsy and/or a PET scan at that time to better assess the response and to gather additional information regarding their disease. If study treatment was held in these patients, clinical assessments and disease assessments were performed. If patients who held treatment in this “wait and see” period exhibited evidence of radiographic disease progression, the patient was permitted to restart larotrectinib. At the time of larotrectinib re-initiation, the patients began assessments per protocol as delineated for patients on active treatment. Patients who re-initiated larotrectinib treatment were permitted to have their larotrectinib treatment held again. Patients who underwent surgical resection for local control may have continued to receive larotrectinib after surgical recovery and discussion between the investigator and the sponsor. These patients should have stopped larotrectinib for 24 hours before surgery and resumed no sooner than 24 hours after surgery. If the resection surgery resulted in negative margins and study treatment was stopped, disease assessments were performed every 3 months following the last dose of study medication. These patients are permitted to restart study treatment if they experienced progressive disease after discussion with the medical monitor.

LOXO-TRK-15002 (NAVIGATE Trial)

Larotrectinib was administered at a dosage of 100 mg orally twice daily in 28-day cycles.⁶³ Treatment was continued until disease progression, unacceptable toxicity, patient withdrawal of consent, or death. Patients who developed disease progression were allowed to continue larotrectinib if, in the opinion of the investigator, the patient was deriving clinical benefit from continuing the study drug and continuation of treatment was approved by the sponsor. For patients who experienced a clinically significant hematologic or non-hematologic TEAE (greater than grade 2, or an increase of more than 1 grade from baseline if the baseline was grade 2 or higher) larotrectinib dosing was held for up to 4 weeks to evaluate the AE and to allow for recovery (to grade 1 or baseline). In patients who had previously experienced a clinical benefit from larotrectinib, larotrectinib dosing could be held for more than 4 weeks to allow for resolution of AEs, with the sponsor’s permission. Upon resolution of the AE and/or symptoms related to the AE, and if the AE was considered unrelated to larotrectinib, dosing may have been restarted at the same dose. If the AE was considered related to the dose of larotrectinib, then dose modifications were made.

Pooled Analysis

The pooled analysis included patients with NTRK gene fusions who were enrolled in 1 of the 3 larotrectinib trials.²² Most patients were treated with larotrectinib 100 mg orally twice daily (in individuals with a BSA equal to or greater than 1 m²) or 100 mg/m² orally twice daily (for children with a BSA larger than 1 m²); however, some patients from the dose escalation phases of the trials were also included.²⁵

Outcomes

A list of efficacy end points identified in the CADTH review protocol that were assessed in the clinical trials included in this review is provided in Table 11. These end points are further summarized in the following section.

LOXO-TRK-14001

The primary end point of study LOXO-TRK-14001 was the safety of larotrectinib (including DLT) and identification of the MTD.⁶¹ Secondary end points were to characterize its pharmacokinetic properties and antitumour activity, which included ORR and DOR, and other measures of antitumour efficacy as determined by investigator assessment.

Disease assessments were performed by the investigator using RECIST 1.1, RANO, or a comparable assessment method depending on location of tumour.⁶¹ The estimate of the ORR was calculated based on a crude proportion of patients with a confirmed BOR of confirmed CR or PR. The DOR was calculated for patients who achieved a CR or PR and was defined as the number of months from the start date of CR or PR (whichever was observed first) to the first date that recurrent or progressive disease was objectively documented. Repeat tumour assessments were conducted on or before day 1 of cycles 3, 5, 7, 9, 11, and 13, and every 3 cycles thereafter until the onset of progressive disease. The severity of each AE was graded by the investigators using version 4.03 of the National Cancer Institute’s Common Terminology Criteria for Adverse Events. Any AEs occurring after cycle 1 or other cumulative toxicities were considered in the final definition of MTD and the ultimate dose was selected for further investigation. Patients who received at least 75% of the planned total dose in cycle 1 were considered to have sufficient study drug exposure to be evaluated in support of dose escalation.

Long-term follow-up assessments occurred every 3 months (± 1 month) and included: date of progressive disease (if not already documented), subsequent anticancer therapies, and survival status.⁶¹

LOXO-TRK-15003 (SCOUT Trial)

The primary objective of the phase I portion of the study was to determine the safety and DLTs of oral larotrectinib in pediatric patients with advanced solid or primary CNS tumours.⁶² Secondary objectives included pharmacokinetic characterization and identification of MTD and/or the appropriate dose of larotrectinib, to describe antitumour activity, and to study pain and HRQoL outcomes. The primary objective of the phase II portion of the study was to determine the ORR by IRC assessment in pediatric patients with an advanced cancer harbouring NTRK fusion who received larotrectinib. Secondary objectives included the study of other efficacy parameters including the DOR, and further assessment of the safety and tolerability of larotrectinib. Assessment of potential biomarkers (including TRK fusion status) of response and resistance to larotrectinib was an exploratory objective.

Disease assessments consisted of CT and/or MRI scans performed pre-treatment, at the beginning of odd cycles through cycle 13, and every third cycle thereafter.⁶² The antitumour activity of larotrectinib included measurement of ORR, clinical benefit rate (CBR), DOR, PFS, and OS. Response criteria were based on RECIST 1.1 for non-CNS solid tumours and RANO criteria for CNS tumours. The CBR was defined as the sum of patients with either a CR, PR, or stable disease. The DOR, which was calculated for patients who achieved a CR or PR, was defined as the number of months from the start date of CR or PR (whichever was observed first) to the first date that recurrent or progressive disease was objectively documented. If a patient died, irrespective of cause, without documentation of recurrent or progressive disease beforehand, then the date of death was used to denote the response end date. Progression-free survival was defined as the number of months from the date of the first dose of larotrectinib to the earliest date of documented evidence of progressive disease or death (whatever the cause) without evidence of prior progression. Overall survival was defined as the number of months from the date of the first dose of larotrectinib to the date of death (due to any cause).

Pain and HRQoL assessments were collected every cycle. In patients 3 years of age or older, pain was assessed by the Wong-Baker FACES Scale. For all patients, HRQoL was assessed using the PedsQL for the patient or their parent/caregiver.⁶²

Long-term follow-up assessments occurred every 3 months (± 1 month) and included the date of progressive disease (if not already documented), subsequent anticancer therapies, and survival status.⁶²

LOXO-TRK-15002 (NAVIGATE Trial)

The primary end point of the trial was to determine the IRC-assessed ORR following treatment with larotrectinib in patients age 12 years and older with an advanced cancer harbouring fusion involving NTRK1, NTRK2, or NTRK3 (collectively referred to as NTRK fusions) for each tumour-specific disease cohort.⁶³ Secondary end points included investigator-assessed ORR, DOR, CBR, PFS, OS, and safety. Exploratory objectives included determining the relationship between pharmacokinetic and treatment effects, including efficacy and safety, and evaluating changes from baseline in quality of life and health-utilities measures.

Patients underwent radiographic evaluation of their disease at the end of cycle 2, and every other cycle for the first 12 months, followed by 3 cycles thereafter. Patients with primary CNS disease underwent radiographic evaluation of their disease at the end of each cycle between cycle 1 and cycle 4, and every 2 cycles between cycle 5 and cycle 12, and every 3 cycles thereafter.⁶³ The ORR was defined as the rate of patients with a BOR of confirmed CR or PR using RECIST 1.1 or RANO criteria, as appropriate to tumour type. A confirmed CR or PR is defined as a CR or PR that was confirmed by repeat assessments performed no less than 28 days after the criteria for response was first met. The BOR was defined as the best response designation (in the order of CR, PR, stable disease, and progressive disease) for each patient recorded between the date of the first dose of treatment and the date of documented progressive disease per RECIST 1.1 or RANO criteria or the date of subsequent anticancer therapy or cancer-related surgery (i.e., surgical resection of tumour), whichever occurs first. The DOR was calculated for patients who achieved a CR or PR and was defined as the number of months from the start date of CR or PR (whichever was observed first) and subsequently confirmed as the first date that radiological recurrent or progressive disease was objectively documented or when death occurred due to any cause. The CBR was defined as the BOR of confirmed CR, PR, or stable disease lasting 16 or more weeks following initiation of larotrectinib. Stable disease was measured from the date of the first dose of larotrectinib until the criteria for progressive disease are first met. Progression-free survival was defined as the number of months from initiation of larotrectinib to the earliest of documented radiological evidence of progressive disease or death due to any cause without prior progression. Overall survival was defined as the number of months from the initiation of larotrectinib to the date of death due to any cause.

Patient-reported outcomes were measured using the EORTC QLQ-C30 and EuroQol 5 Dimensions 5-Levels questionnaire (EQ-5D-5L) for patients age 18 years and older, and PedsQL for patients age 12 to 17.⁶³ The HRQoL measures were collected at the same cycle visits as the disease assessment and at the end of treatment period.

Long-term follow-up assessments occurred every 3 months (± 1 month) and included date of progressive disease (if not already documented) and subsequent anticancer therapies.⁶³

Pooled Analysis

The primary end point of the pooled analysis was ORR (IRC-assessed in ePAS4, investigator-assessed in SAS3).²² The ORR was defined as the proportion of patients with a BOR of either a CR or a PR according to RECIST 1.1. A CR was defined as no radiological evidence of disease, negative surgical margins, and no viable tumour cells. A PR was defined according to RECIST criteria as a minimum decrease from baseline of at least 30% in the sum of target lesion diameters. The BOR was defined as the best response designation for each patient that was recorded between the date of the first dose of larotrectinib and the date of documented progressive disease by RECIST 1.1, International Neuroblastoma Response Criteria, or the date of subsequent therapy or cancer-related surgery, whichever occurs first.⁵⁷ Patients who underwent surgical resection and had no viable tumour cells and negative margins on a post-surgical pathology report were considered complete responders by surgery and/or pathology. Response (PR or CR) must have been confirmed by a repeat assessment performed no less than 28 days after the criteria for response were first met. When stable disease was believed to be the best response, it must also have met the minimum interval of 6 weeks (42 days) from the start of study treatment. If the minimum time was not met when stable disease was otherwise the best time point response, the patient’s best response depended on the subsequent assessments.

Secondary end points include ORR based on local investigator assessment, as well as the following (both investigator- and IRC-assessed): TTR, time to best response (TTBR), DOR, DCR, PFS, and OS.⁵⁷ The TTR was defined as the number of months between the date of the first dose of larotrectinib and the first documentation of objective response (CR or PR, whichever occurred earlier) that was subsequently confirmed. The TTR was only defined for patients with a BOR of a confirmed CR or confirmed PR. The DOR was defined as the number of months from the start date of a PR or CR (whichever response was recorded first), and subsequently confirmed, to the date of progressive disease or death, whichever occurred first. The TTBR was defined as the number of months between the date of the first dose of larotrectinib and the first documentation of a CR (if the patient’s BOR is a confirmed CR) or PR (if the patient’s BOR is a confirmed PR) that was subsequently confirmed. The TTBR was only defined for patients with a BOR of a confirmed CR or confirmed PR. The DCR was calculated based on the proportion of patients with a BOR of a confirmed CR, PR, or stable disease lasting 16 weeks or more. Stable disease was measured from the first date of larotrectinib administration until the criteria for progressive disease were first met. Progression-free survival was defined as the number of months from date of the first dose of larotrectinib to the earliest date of documented progressive disease or death. Overall survival was defined as the number of months that elapsed between the first dose of larotrectinib and death.

Analyses of HRQoL outcomes were not pre-specified in the sponsor-submitted statistical analysis plan for the pooled analyses, and it was not clear which analyses or subgroups were pre-specified (if any); however, analyses were performed and presented as part of conference preceedings.^59,60 Such outcomes were secondary end points in the NAVIGATE and SCOUT trials, for which the pooled analysis only included the EORTC QLQ C-30 and PedsQL. The NAVIGATE trial used the EORTC QLQ C-30 for patients 18 years of age and older, and the PedsQL 4.0 Generic Core Scale for patients aged 12 to 17 years.^64,65 The SCOUT trial used the PedsQL 4.0 Generic Core Scale. The HRQoL questionnaires were completed at the baseline and planned cycle visits (NAVIGATE: day 1 of cycles 3, 5, 7, 9, 11, 13, and every 3 cycles or months thereafter until disease progression; SCOUT: every cycle).

The EORTC QLQ C-30 is a cancer-specific measure of HRQoL, consisting of 30 items, including 5 multi-item functional scales, 3 multi-item symptom scales, 6 single-item symptom scales, and a 2-item QoL scale (also known as the GHS score).⁶⁶ The first 3 scales are rated on a 4-point scale, whereas the GHS is rated on a 7-point Likert scale. Responses for each scale are then converted to a standardized score ranging from 0 to 100, with higher scores indicating better function and QoL, whereas a decline in the symptom scales score reflects an improvement.⁶⁷ The PedsQL is a modular instrument with a 23-item Generic Core Scale, which can be combined with various disease-specific modules, to measure HRQoL in children and adolescents aged 2 to 18 years. The generic scale consists of 4 dimensions, including physical functioning (8 items), emotional functioning (5 items), social functioning (5 items), and school functioning (5 items).⁶⁸ Each item is rated on a 5-point Likert scale, then reverse-scored and linearly transformed to a 0-to-100 scale; higher scores indicate better HRQoL. The total score is the sum of all the items over the number of items answered on all the scales.⁶⁹ Both instruments have been validated extensively and are available in many languages for various cancer types. The MID for these measures was estimated in separate studies using a mixed cancer population and was defined as a change in score of 10 points for all EORTC QLQ-C30 scales, and 4.4 to 4.5 points for PedsQL total scores for patients ≥ 2 years of age.^65,70,71 No literature could be found to obtain an MID for infants younger than 2 years of age; the MID was therefore defined by the sponsor as half the SD of the baseline total PedsQL score for this group (7.2).²² A detailed discussion of these HRQoL measures is provided in Appendix 5.

Statistical Analysis

LOXO-TRK-14001

The dose escalation portion of the study employed a classical “3 + 3” dose escalation design, with 3 to 6 patients enrolled in each cohort.⁶¹ The total number of patients to be enrolled in the dose escalation phase was anticipated to be approximately 60, depending on the observed safety profile, which would determine the number of patients per dose cohort, as well as the number of dose escalations required to achieve the MTD. For the dose expansion portion of the study, approximately 20 patients were targeted for enrolment in each cohort, largely based on clinical consideration for phase I studies. The rationale provided was that if the observed ORR was high (exceeding 50%) within a cohort, then the corresponding lower limit of a 1-sided exact 90% CI would exclude true response rates that were considered marginal or uninteresting (e.g., < 30%).

For patients with NTRK1, NTRK2, or NTRK3 gene fusion, imaging was evaluated by site Investigators and by an IRC.⁶¹ For such patients, ORR and DOR was based on the time point and BORs as determined by the IRC, and a secondary analysis was based on site investigator assessments. Differences between the IRC and investigator assessments of response were reported. The statistical analysis plan stated that efficacy tables may be presented for subgroups, including patients with or without NTRK fusion cancer. The estimate of the ORR was calculated based on the maximum likelihood estimator (i.e., crude proportion of patients with a BOR or CR or PR, based on RECIST 1.1). The estimate of the ORR included 1- and 2-sided CIs with various coverage probabilities (e.g., 80% to 95%). The DOR was summarized descriptively using the Kaplan-Meier method with a 95% CI and calculated using the Greenwood formula. Median follow-up was estimated according to the Kaplan-Meier estimate of potential follow-up. The Greenwood formula was used to calculate the standard errors of the Kaplan-Meier estimate and the upper and lower limits of the 95% CI. The DOR was right-censored (the progression or censoring date was determined based on described conventions from the FDA [2007]) for patients who met 1 or more of the following conditions:

• initiated subsequent anticancer therapy in the absence of documented progressive disease

• died or had progressive disease after missing 2 or more consecutively scheduled disease assessment visits

• were last known to be alive and without documented progressive disease on or before the data cut-off date.

Tabular summaries were provided for all TEAEs, including summaries by relationship to study drug, maximum severity grade, study drug action taken, serious adverse events (SAEs), and fatal AEs. Laboratory variables and vital signs were assessed by means of standard shift tables and summary statistics.⁶¹

Unless noted otherwise, missing data were not imputed, and all analyses were based only on observed data.⁶¹ The effective sample sizes at each assessment visit was based on the total number of patients with non-missing data for the parameter of interest at that visit.

One interim analysis was performed on July 15, 2019.⁶¹ Additional unplanned interim analyses may be performed, but these will be considered as exploratory in nature. Unplanned interim analyses will be detailed in a separate interim statistical analysis plan and described in a subsequent Clinical Study Report.

LOXO-TRK-15003 (SCOUT Trial)

In the phase I dose escalation stage of the study, up to 36 patients were anticipated to be enrolled to define the MTD or larotrectinib; up to 6 patients evaluable for safety were to be enrolled in each dose cohort based on a phase I, rolling 6 design, with the exception of cohorts 3, 4, and 5, which may have enrolled up to 9 patients each to fulfill the requirement for at least 3 patients who meet minimum BSA criteria.⁶² In the phase II expansion cohort stage of the study, approximately 12 to 18 patients with specific abnormalities in NTRK genes or proteins were planned for enrolment to provide additional safety, pharmacokinetic, and antitumour activity information when larotrectinib is administered to such patients at the recommended phase II dose. The rationale provided was that the sample size was chosen based on clinical considerations for phase I studies. In phase II of the study, up to 40 patients were to be enrolled in each cohort. The number of patients planned for each phase II cohort was determined largely by feasibility considerations owing to the rarity of pediatric NTRK fusion cancers. It was anticipated that the ORR may be high (≥ 50%) for each cohort evaluated and that if such response rates were observed for phase II, then the lower limit of the CI about the observed response rate could be used to identify true response rates considered promising enough to warrant further development of larotrectinib within a disease indication. For example, if among 10 patients enrolled within a cohort, there were 5 (50%), 6 (60%), 7 (70%), or 8 (80%) patients with a CR or PR, then the 80% CI about this response rate would be (27% to 73%), (35% to 81%), (45% to 88%), or (55% to 95%), respectively. Similarly, there is a 90% confidence that the response rate observed in similar future trials will not be lower than 27% to 35%, 45%, or 55%, respectively.

The estimate of the ORR was calculated based on the maximum likelihood estimator (i.e., the crude proportion of patients with a BOR or a CR or PR, based on RECIST 1.1).⁶² Patients with a non-confirmed CR or PR were included in the ORR calculation, but were not considered as a responder. Patients without measurable disease at baseline or who continued but had no evaluable post-baseline assessment were excluded from the calculation of ORR. Two-sided exact binomial 95% CIs (Clopper-Pearson) were presented. The CBR was analyzed similarly.

The DOR was summarized descriptively using the Kaplan-Meier method with a 95% CI, calculated using the Greenwood formula. Median follow-up was estimated according to the Kaplan-Meier estimate of potential follow-up.⁶² The DOR was right-censored as follows (the progression or censoring date was determined based on described conventions from the FDA [2007]) for patients who met 1 or more of the following conditions:

• underwent an amputation or surgical resection of tumour in the absence of documented progressive disease; patients who underwent surgical resection and had no viable tumour cells and negative margins on post-surgical pathology report were excluded from this category; such patients were considered to have achieved CR by surgery/pathology and continued to undergo assessments for disease recurrence following surgery, and patients who underwent resection of lesions that were determined to be benign were excluded from this censoring condition

• initiated subsequent anticancer therapy in the absence of documented progressive disease

• died or had progressive disease after missing 2 or more consecutively scheduled disease assessment visits

• were last known to be alive and without documented progressive disease on or before the data cut-off date.

Progression-free survival was analyzed according to the same methods as DOR analyses with the addition that patients with no baseline or post-baseline disease assessments were right-censored unless death occurred before the first planned assessment (in which case the death was considered a PFS event).⁶² Kaplan-Meier estimates of PFS rates at 6 and 12 months, with corresponding 95% CIs calculated using the Greenwood formula, were presented.

For OS analyses, patients who were alive or lost to follow-up as of the data cut-off date were right-censored.⁶² The censoring date was determined from the date the patient was last known to be alive. The duration of OS was summarized descriptively using the Kaplan-Meier method, with a 95% CI about the median calculated using the Greenwood formula. Median follow-up for OS was estimated according to the Kaplan-Meier estimate of potential follow-up. Kaplan-Meier estimates of OS at 6 and 12 months, with corresponding 95% CIs calculated using the Greenwood formula, were presented. Analyses at additional time points (e.g., 24 months) may be performed as the follow-up data mature.

Analyses of HRQoL, which was assessed using the PedsQL, were analyzed descriptively by time point and mean changes over time in scores.⁶² Pain scores were analyzed descriptively by time point.

Safety data were tabulated and presented by nominal-dose cohort across both phase I and phase II and by all patients.⁶² For certain presentations (such as laboratory analyses and vital signs), data were presented across all dosed patients only. A tabulation of all AEs by NTRK fusion status was also presented. Adverse events were summarized based on the number and percentage of patients experiencing the event mapped to the Medical Dictionary for Regulatory Activities preferred term and grouped by system organ class. The causal relationship between the occurrence of an AE and the study drug was judged by the investigator. If a patient experienced repeat episodes of the same AE, then the event with the highest severity grade and strongest causal relationship to the study drug was used for the incidence tabulations. All deaths were reported in a patient listing, and included the treatment cohort, primary cause of death, and the number of days between the date of the last dose of study drug and death.

For data summarized over time by visit, no imputations were performed on missing data and all analyses were based on observed data only.⁶² The effective sample sizes at each assessment visit were based on the total number of patients with non-missing data for the parameter of interest at that visit.

Two interim analyses have been performed.⁶² Additional unplanned interim may be performed, but these will be considered exploratory in nature. Unplanned interim analyses will be detailed in a separate interim statistical analysis plan and described in a subsequent Clinical Study Report. Although, according to the protocol, the primary analysis of ORR was to be based on IRC assessments, investigator assessment was used for the interim Clinical Study Report. In the final Clinical Study Report analysis and in the Integrated Summary of Efficacy, the central IRC assessment was/will be used and the concordance between the methods was/will be examined.

LOXO-TRK-15002 (NAVIGATE Trial)

For the cancer-specific cohorts (1 through 7), Simon’s 2-stage design was used to determine whether larotrectinib had sufficient anticancer activity to warrant further development for that tumour type, and enrolment within a cohort could have been terminated early if larotrectinib was deemed to be insufficiently effective.⁶³ The decision to terminate or continue enrolment within a cohort is made independently of the other cohorts. For each cohort, a true ORR of 10% or less is considered insufficient to warrant further study (null hypothesis), whereas a true ORR of 30% or more is considered sufficiently effective (alternative hypothesis). The number of patients evaluated in each stage and the minimum number of responders needed to continue to the next stage were determined based on the optimum version of the aforementioned design with an 80% power and a 1-sided significance level of 10%. Based on these design considerations, up to 7 patients could be enrolled in each cohort (stage 1). If no patients achieved a CR or PR (confirmed or unconfirmed) within a cohort, then enrolment within that cohort would terminate. Otherwise, 11 additional patients would be enrolled within the cohort (second stage). Up to 18 patients per tumour-specific cohort (cohorts 1 through 7), and up to 25 patients in the other histologic tumour types cohort (cohort 8) or patients without measurable disease cohort (cohort 9), were estimated, for a maximum total sample of approximately 176 patients. In case 7, patients with the same histology were enrolled without a response, and no further patients would therefore be enrolled with that histology. Up to 18 additional subjects may have been enrolled in cohorts 1 through 8 once the second stage of the Simon 2-stage design had been fully completed for cohorts 1 through 7 and once cohort 8 had been fully enrolled. These patients constituted a “post–stage 2” enrolment group. Unless otherwise specified, data were summarized by cohort and in total. Patients enrolled in the post–stage 2 group were included in each respective cohort for the analyses.

The point estimate of the ORR was calculated based on the maximum likelihood estimator (i.e., the crude proportion of patients with the BOR of CR or PR), based on all evaluable subjects, where evaluable subjects are defined as subjects with measurable disease at baseline and with evaluable post-baseline disease assessment(s).⁶³ Patients with measurable disease at baseline who discontinued from study treatment were considered evaluable even if they had no post-baseline disease assessment. The point estimate of ORR was accompanied by a 1-sided 90% exact binomial CI using the Clopper-Pearson method. Additionally, 1-sided 97.5% exact binomial CIs for the ORR may also be presented. A forest plot of the ORRs and the associated 1-sided 90% CIs was presented by cohort.

The BOR was to be summarized descriptively and derived from the time point responses as determined by the investigator.⁶³ Response (PR or CR) must have been confirmed by a repeated assessment performed no less than 28 days after the criteria for response is first met. When stable disease is believed to be the best response, it must also have met the minimum interval of 6 weeks from the start of study treatment. If the minimum time was not met when stable disease was otherwise the best time point response, the subject’s best response was dependent on the subsequent assessments. For patients with measurable solid tumours, the sum of the longest diameters (shortest for lymph nodes) was calculated for target lesions. The baseline and the minimum post-baseline values of the sum of diameters as well as the best percentage change in the sum of diameters from baseline were summarized using descriptive statistics for subjects with measurable solid tumours. For subjects with primary CNS tumours in cohort 7, the sum of the product of diameters was calculated for target lesions. The sum of the product of diameters was summarized in the same manner as for the sum of the diameters of measurable solid tumours.

Time-to-event efficacy variables including DOR, PFS, and OS were summarized descriptively using the Kaplan-Meier method for medians and corresponding 2-sided 95% CIs, calculated using the Greenwood formula and presented by cohort.⁶³ Rates and the corresponding 2-sided 95% CIs were provided by 6-month intervals. Plots of the Kaplan-Meier estimate of the survival distribution function over time were presented. The median follow-up for each end point was estimated according to the Kaplan-Meier estimate of potential follow-up, calculated in the same way as the Kaplan-Meier estimate of the corresponding survival function, but with the meaning of the status indicator reversed. Exploratory subgroup analyses of selected efficacy end points were planned, subject to the availability of data. The subgroups would be defined based on the patient and disease characteristics as well as treatment history (e.g., extent of prior therapy), and were not predefined in the protocol

The DOR was right-censored as follows for⁶³:

• patients who had an amputation or surgical resection of tumour in the absence of documented progressive disease; those who underwent surgical resection and had no viable tumour cells and negative margins on post-surgical pathology report were excluded from this category, and such patients were considered a CR by surgery/pathology and continued to undergo assessments for disease recurrence following surgery; patients who underwent resection of lesions that were determined to be benign were also excluded from this censoring condition

• patients who initiated subsequent anticancer therapy in the absence of documented progressive disease

• patients who died or had progressive disease after missing 2 or more consecutively scheduled disease assessment visits

• patients who are were last known to be alive and without documented progressive disease on or before the data cut-off date.

Progression-free survival was analyzed according to the same methods as DOR analyses, with the addition that patients with no baseline or post-baseline disease assessments were right-censored unless death occurred before the first planned assessment (in which case the death was considered a PFS event).⁶³ For OS analyses, patients who were lost to follow-up or who were still alive at the time of analysis were censored at the last day the subject was known to be alive.

Scores from the EORTC QLQ-C30 and changes from baseline were summarized descriptively over time at scheduled time points by cohort for subjects 18 years and older.⁶³ The PedsQL generic module total scores and changes from baseline were summarized descriptively over time at scheduled time points by cohort for subjects aged 12 to 17 years. Frequency tables were used to display the number and percentage of subjects of reported problems for each level and each dimension of the EQ-5D-5L by scheduled time point.

Safety was assessed by clinical review of all relevant parameters, including AEs, SAEs, laboratory values, vital signs, and electrocardiogram results.⁶³ Safety analyses in general were descriptive and were presented in tabular format with the appropriate summary statistics. Tabulations were provided by cohort and in total. No formal statistical comparison was applied. TEAEs were defined as those that started on or after the first administration of the study drug. For the number and percentage of subjects with TEAEs, drug-related TEAEs, treatment-emergent SAEs, drug-related treatment-emergent SAEs, TEAEs leading to discontinuation of the study drug, drug-related TEAEs leading to discontinuation of study drug, TEAEs leading to dose reduction and/or interruption of the study drug, and drug-related TEAEs leading to dose reduction and/or interruption of the study drug summaries, subjects with multiple AEs were counted only once per system organ class and preferred term.

Unless noted otherwise, missing data for individual data points were not imputed, and all analyses were based on observed data only.⁶³ A single interim analysis was conducted based on an amended protocol. The interim analysis was performed for the purposes of supporting an application for marketing approval of larotrectinib. The interim Clinical Study Report included all subjects enrolled as of the data cut-off of July 15, 2019.

Pooled Analysis

The statistical analysis plan for the integrated summary of efficacy was based on interactions with the FDA and other global regulatory agencies.^25,57 The plan was finalized before the database was closed and before IRC assessment of tumours was conducted; however, time points for additional data cut-offs were chosen to fulfill requests for updated data from regulatory agencies. The plan included: sample size; the PAS; the rationale for lack of randomization, comparator arm, and balanced cohort recruitment; pooling (reasons, appropriateness, and design); and procedures for modifying the study design. The datasets represent different data cut-offs of the patients enrolled in the larotrectinib clinical trials who were eligible for analysis at a given point in time (i.e., these datasets are not derived from different patient cohorts, and the later dataset included the patients from earlier datasets). Due to different regional requirements for required sample sizes by regulatory agencies as well as the interest in longer-term follow-up of patients, regular data cut-offs, including incremental datasets of patients, were planned.⁵⁷

The point estimate of the ORR was calculated based on the crude proportion of patients’ BOR of a confirmed CR or confirmed PR. The point estimate will be accompanied by a 2-sided 95% exact binomial CI using the Clopper-Pearson method.⁵⁷ The BORs (CR, surgical CR, PR, stable disease, progressive disease, and unevaluable) were summarized descriptively to show the number and percentage of patients in each response category. The agreement rate between IRC assessment and investigator assessment was evaluated by frequency table. Heterogeneity of the estimated ORR by study was evaluated based on the quantity I², which describes the percentage of total variation across studies due to heterogeneity rather than chance. The analysis of DCR was based on the methods described for ORR.

The TTR was summarized descriptively by calculating the median, quartiles, and minimum and maximum values.⁵⁷ The number and percentage of patients with TTR, measured relative to the date of the first dose of larotrectinib, were tabulated by the following time points: 2 months or less, more than 2 months to 4 months, more than 4 months to 6 months, more than 6 months to 9 months, and more than 9 months. Kaplan-Meier curves were used to graphically present the TTR distribution over time. The TTBR was summarized descriptively in the same manner as the TTR.

The DOR was summarized descriptively using the Kaplan-Meier method with the 95% CI about the median calculated using the Greenwood formula.⁵⁷ Median follow-up for DOR was estimated according to the Kaplan-Meier estimate of potential follow-up. Using the Kaplan-Meier method, the rates of continuing responders (irrespective of censoring status of patient) for at least 6, at least 12, at least 18, and at least 24 months were estimated, including 95% CIs. A separate table assessing DOR by tumour type was provided, including 12- and 24-month DOR rates. The DOR was right-censored as follows for:

• patients who had an amputation or surgical resection of tumour in the absence of documented progressive disease; those who underwent surgical resection and had no viable tumour cells and negative margins on post-surgical pathology report were excluded from this category; such patients were considered a CR by surgery/pathology and continued to undergo assessments for disease recurrence following surgery; patients who underwent resection of lesions that were determined to be benign were also excluded from this censoring condition

• patients who initiated subsequent anticancer therapy in the absence of documented progressive disease

• patients who died or had progressive disease after missing 2 or more consecutively scheduled disease assessment visits

• patients who were last known to be alive and without documented progressive disease on or before the data cut-off date.

Progression-free survival was analyzed according to the same methods as the DOR analyses, with the addition that patients with no baseline or post-baseline disease assessments were right-censored unless death occurred before the first planned assessment (in which case the death was considered a PFS event).⁵⁷

The duration of OS was summarized descriptively using the Kaplan-Meier method with the 95% CI about the median calculated using the Greenwood formula.⁵⁷ Using the Kaplan-Meier method, the rates of patients alive for at least 6, at least 12, at least 18, and at least 24 months were estimated, including 95% CIs. Median follow-up for OS was estimated according to the Kaplan-Meier estimate of potential follow-up. Patients who are alive or lost to follow-up as of the data cut-off date were right-censored. The censoring date was determined from the date the patient was last known to be alive.

For HRQoL, the proportion of adults and children with norm/above norm and below norm QoL scores was compared with values in published literature for the US general population.⁶⁰ The mean EORTC QLQ-C30 GHS for the US general population (63.9)²⁴ minus 10 points (the MID) was used to construct the norm/above norm (≥ 53.9) and below norm (< 53.9) score categories for adults.²⁵ The average score for the combined self and proxy-reported PedsQL questionnaire for healthy US children (85.0)²⁶ minus 4.5 points (the MID) was used to construct the norm/above norm (≥ 80.5) and below norm (< 80.5) score categories for children 2 years of age or older. As no established normative PedsQL total score for infants younger than 2 years of age was identified, no comparisons were performed in this group. For patients who exhibited sustained improvement or deterioration for at least 2 consecutive cycles, the following method was used to calculate sustained improvement/deterioration. The duration of improvement was calculated as the duration (in months) for those patients with improvement from the date of the first increase from baseline greater than or equal to the MID for the respective HRQoL instrument until the date of the last adjacent increase from baseline greater than or equal to the MID (i.e., an increase from baseline greater than or equal to the MID after a decrease was ignored). A patient was censored at the date of last available questionnaire if there was no visit with an increase less than the MID before the last available measurement.

Safety was assessed based on the incidence and severity of all AEs and AEs that were deemed SAEs.⁵⁷ Changes in clinical laboratory tests, vital signs, body weight, performance status, and 12-lead electrocardiograms were also assessed. These data presentations were performed for each of the analysis sets, with the exception of the analysis set of only patients treated at the recommended dose, which was only used to analyze AEs, deaths, and laboratory abnormalities (according to toxicity grading). All deaths that occurred while on treatment or after the last dose of the study drug were displayed in a patient-listing format.

Supportive analyses were performed to assess ORR for consistency across selected subgroups and special populations.⁵⁷ The point estimate of the ORR (and 95% CI), based on IRC assessment, was calculated for the subgroups and special populations defined by the following:

• age at enrolment (1 month to < 2 years, 2 to < 6 years, 6 to < 12 years, 12 to < 18 years, 18 to < 65 years, ≥ 65 years)

• pediatric (age at enrolment < 18 years) versus adults (age at enrolment ≥ 18 years)

• sex (male, female)

• race (White, Black, Asian, other)

• ECOG performance status at baseline (0 to 1, 2, or 3)

• NTRK fusion (NTRK1, NTRK2, NTRK3)

• NTRK fusion partner (e.g., ETV6)

• primary cancer diagnosis (according to standardized term)

• cancers considered pathognomonic for NTRK fusions (IFS and MASC)

• disease status (metastatic or locally advanced)

• number of prior systemic regimens or treatment courses (0, 1, 2, ≥ 3)

• BOR to most recent prior systemic regimen or treatment course (CR, PR, stable disease, progressive disease, unknown or unevaluable, or not applicable)

• starting dose of larotrectinib and frequency of administration (separately for adult versus pediatric patients).

Depending on the requests or requirements of individual regulatory agencies, age groupings other than those specified above may have been used for supplemental analyses.⁵⁷ The subgroups and special populations were determined based on the data recorded for the screening or baseline assessments. Subgroups based on larotrectinib dosing parameters were determined from the actual dose, frequency, and formulation recorded for study day 1.

As a sensitivity analysis to assess the potential impact of requiring 6 months of follow-up for inclusion into the respective latest analysis set on response rate, the subset of patients of that were immature for the dataset that discontinued or died within less than 6 months after recruitment were assessed by IRC and combined with the analysis set for a sensitivity analysis.⁵⁷ Response rates were determined and evaluated using a method similar to that previously described.

Unless noted otherwise, missing data were not imputed^.57 For efficacy outcome data summarized over time by visit, no imputations were performed on missing data. All analyses were based on observed data only. The effective sample sizes at each assessment visit were based on the total number of patients with non-missing data for the parameter of interest at that visit. For safety outcomes, all analyses were based on observed data only. The effective sample sizes at each assessment visit were based on the total number of patients with non-missing data for the parameter of interest at that visit.

Analysis Populations

LOXO-TRK-14001

The following analysis sets were used in the trial⁶¹:

• A full analysis set (FAS) was used primarily for the analysis of tumour response and other efficacy-related data. It included all enrolled patients who received 1 or more doses of larotrectinib. For the final analysis, the protocol-specified FAS was used and included the following: patients in either the escalation or expansion phase who received 1 or more doses of larotrectinib, patients with a documented NTRK abnormality, and patients with 1 or more measurable lesions at baseline as determined by RECIST 1.1 or RANO criteria as appropriate.

• A safety analysis set was used primarily for the analysis of safety data. It consisted of all enrolled patients who received at least 1 dose of larotrectinib.

LOXO-TRK-15003 (SCOUT Trial)

The following analysis sets were used in the trial⁶²:

• An FAS was used primarily for the analysis of tumour response and other efficacy-related data. It included all enrolled patients who receive 1 or more doses of larotrectinib.

• A safety analysis set was used primarily for the analysis of safety data. It consisted of all enrolled patients who received at least 1 dose of larotrectinib.

LOXO-TRK-15002 (NAVIGATE Trial)

The following analysis sets were used in the trial⁶³:

• An FAS was used for all efficacy analyses in the interim analyses and included all enrolled patients who received 1 dose of larotrectinib, including patients enrolled as part of the post–stage 2 enrolment group. The FAS designation for the interim analysis was more inclusive than the FAS specified in the protocol as it also included post–stage 2 patients who were dosed. For the final analysis, the protocol-specified FAS will be used.

• A safety analysis set was used for safety data and consisted of all enrolled patients who received at least 1 dose of larotrectinib. A baseline measurement and at least 1 laboratory or other safety-related measurement obtained after at least 1 dose of the study drug may have been required for inclusion in the analysis of a specific safety parameter.

Pooled Analysis

The following datasets were used for the pooled analyses²²:

• The PAS (July 17, 2017, data cut-off) included the first 55 patients across the 3 studies (8 patients form LOXO-TRK-14001, 12 patients from LOXO-TRK-15003, and 35 patients from LOXO-TRK-15002) who met the inclusion criteria for the pooled analysis. This dataset constituted the FDA submission.³¹

• The ePAS (February 19, 2018, data cut-off) included an additional 18 patients (9 patients from LOXO-TRK-15003 and 9 patients from LOXO-TRK-15002) who met inclusion criteria for the pooled analysis. A total of 73 patients with NTRK gene fusions were analyzed at this data cut-off date. The results of the ePAS were submitted to Health Canada for regulatory review and to CADTH for the original consideration.

• Extended primary analysis set 4 (July 15, 2019, data cut-off) included a total of 164 patients (13 patients from LOXO-TRK-14001, 53 patients from LOXO-TRK-15003, and 98 patients from LOXO-TRK-15002) who met inclusion criteria for the pooled analysis (including having an IRC-assessed response). The results of ePAS4 were submitted to CADTH and L'Institut national d’excellence en santé et en services sociaux (INESSS) for reconsideration.

• Extended primary analysis set 5 (July 2020 data cut-off) included a total of 192 patients. This was the most updated dataset with the largest number of patients and follow-up data.

• Safety analysis set 3 (July 15, 2019, data cut-off) included 24 patients with primary CNS tumours. Additionally, a new SAS3 dataset with a longer follow-up (July 2020 data cut-off) and 33 patients was submitted.

• Patients included in the HRQoL analyses (July 15, 2019, data cut-off) had non-CNS primary solid tumours with NTRK gene fusion and measurable disease from LOXO-TRK-15003 and LOXO-TRK-15002. This analysis included 126 patients who had received larotrectinib and completed a baseline and at least 1 post-baseline questionnaire (74 adults, 24 children ≥ 2 years old, and 28 infants < 2 years old).

• The TRK fusion cancer labelled-dose safety analysis set (July 15, 2019, data cut-off) included 196 patients with TRK fusion cancer who received at least 1 dose of larotrectinib at the labelled or recommended dose.

• The overall labelled-dose safety analysis set (July 15, 2019, data cut-off) included 238 patients (with or without TRK fusion cancer) who received at least 1 dose of larotrectinib at the labelled or recommended dose.

• The integrated dataset (July 30, 2018, data cut-off) consisted of 122 patients from the combined extended primary and supplementary datasets, i.e., larotrectinib-treated patients with NTRK gene fusions who had their outcomes assessed by the investigator. This dataset was the basis for the 2019 CADTH submission.²³

• The safety dataset (July 30, 2018, data cut-off) encompasses the entire larotrectinib safety database (n = 207), which includes 122 patients with NTRK gene fusion cancer and 70 patients without confirmed NTRK gene fusions.

• The NTRK fusion cancer safety set (July 2020 data cut-off) consisted of updated safety data from 260 cancer patients with NTRK fusion.

• The overall safety set (July 2020 data cut-off) consisted of the entire larotrectinib safety database with a larger sample (n = 331).

Figure 2 provides an overview of the different pooled datasets submitted to different regulatory agencies at different time points, with different data cut-offs and sample sizes.

Figure 2: Overview of Pooled Datasets for Larotrectinib

EMA = European Medicines Agency; ePAS = extended primary analysis set; ePAS4 = extended primary analysis set 4; ESMO = European Society for Medical Oncology; INESSS = Institut national d’excellence en santé et en services sociaux; IRC = independent review committee; NICE = National Institute for Health and Care Excellence; PAS = primary analysis set; pCODR = CADTH pan-Canadian Oncology Drug Review; SAS3 = safety analysis set 3.

Source: Note to clinical reviewer (2020).²²

Results

Patient Disposition

A total of 164 eligible patients with NTRK gene fusions from the 3 aforementioned trials were included in ePAS4 (Table 13); of those, 84 patients (51.2%) had discontinued treatment as of the July 15, 2019, data cut-off. Reasons for discontinuation included disease progression in 49 patients (29.9%), AEs in 4 patients (2.4%), physician decision in 7 patients (4.3%), subject decision in 6 patients (3.7%), protocol deviation in 2 patients (1.2%), death in 3 patients (1.8%), and other reasons in 7 patients (7.9%).²²

Exposure to Study Treatments

The exposure to study treatment was not reported for the patients included in the pooled analysis.

Efficacy

Only those efficacy outcomes and analyses of subgroups identified in the review protocol are reported here. As noted earlier, the results of the ePAS4 dataset (July 15, 2019, data cut-off) were submitted to CADTH for reconsideration, and therefore they are discussed here. Results from the sponsor’s primary pooled analysis (PAS [n = 55]; July 17, 2017, data cut-off) used for the FDA submission, the original CADTH submission (integrated dataset [n = 122]; July 30, 2018, data cut-off), and the newest ePAS5 dataset (July 2020 data cut-off) are also presented, where available, for comparison.

Overall Response Rates

ePAS4 and ePAS5 Datasets

The primary efficacy end point of the pooled analysis was the ORR. The ORR from the PAS (n = 55; July 17, 2017, data cut-off) was 75% (95% CI, 61% to 85%); 20% of patients achieved a CR, 5% achieved a pathological CR, and 49% achieved a PR (Table 14).²² At the time of the data cut-off date of July 15, 2019, for ePAS4 (n = 164), the ORR by IRC assessment was 73% (95% CI, 65% to 79%); 19% of patients achieved a CR, 5% achieved a pathological CR, and 49% achieved a PR (Table 15). The DCR (defined as CR, pathological CR, PR, and stable disease lasting ≥ 16 weeks) was 84% (95% CI, 77% to 89%). At the time of the data cut-off of July 2020 for the ePAS5 (n = 192), the ORR was 72% (95% CI, 65% to 79%); 23% and 7% patients achieved CR and pathological CR, respectively. In the original submission, the ORR as of the July 30, 2018, data cut-off date was 81% (95% CI, 72% to 88%); 17% of patients achieved a CR and 63% achieved a PR (Appendix 3).

Subgroup Analyses

Subgroup analyses of ePAS4 (July 15, 2019, data cut-off date) are summarized in this section.²² The ranges of ORRs across the different subgroups were tumour type: 0% (95% CI to NE) to 100% (95% CI, 40% to 100%) (Table 15); age: 91% (95% CI, 80% to 97%) in pediatrics and 63% (95% CI, 54% to 72%) in adults (Table 16); baseline ECOG status: 33% (95% CI, 1% to 92%) in patients with an ECOG status of 3 to 83% (95% CI, 72% to 90%) in patients with an ECOG status of 0 (Table 17); number of prior systemic treatment regimens: 59% (95% CI, 41% to 75%) in patients who have received 2 prior systemic treatments to 86% (95% CI, 71% to 95%) in patients who have received 0 prior systemic treatments (Table 18); baseline disease status: 67% (95% CI, 58% to 75%) in patients with metastatic cancer at baseline to 88% (95% CI, 74% to 96%) in patients with locally advanced metastatic cancer at baseline (Table 19); an NTRK gene involved in gene fusion: 50% (95% CI, 7% to 93%) for patients with NTRK2 gene fusion to 80% (95% CI, 71% to 88%) for patients with NTRK3 gene fusion (Table 20); and NTRK gene fusion for the 3 most common fusion partners: 62% (95% CI, 32% to 86%) in patients with LMNA-NTRK1 gene fusion to 85% (95% CI, 75% to 92%) in patients with ETV6-NTRK3 gene fusion (Table 21).

Across the different tumour types, ORR varied between the 3 datasets at 3 different cut-off points (July 17, 2017, July 15, 2019, and July 2020 for PAS, ePAS4, and ePAS5, respectively). For tumours found in more than 10 patients, the differences in ORR estimates were most noticeable for STS and thyroid cancer.

SAS3

At the July 15, 2019 data cut-off date for SAS3 (n = 24), which included patients with primary CNS tumours, the ORR by investigator assessment was 21% (95% CI, 7% to 42%); 8% of patients achieved a CR and 13% achieved a confirmed PR (Table 22).²² With the new data cut-off date of July 2020 (n = 33), the ORR was 24% (95% CI, 11% to 42%); 9% and 15% of patients achieved a CR and PR, respectively. In the original 2019 submission, among patients with primary CNS tumours (n = 18), the ORR was estimated to be 36% (95% CI, 13% to 65%); with CR in 14%, PR in 21%, and stable disease in 64% of patients (Appendix 3).

Time to Tumour Response

ePAS4

As of the data cut-off date for the ePAS4 dataset (July 15, 2019), the median TTR was 1.84 months (minimum = 0.92 months; maximum = 14.55 months) (Table 23, Figure 3).²² The percentage of patients experiencing a time to response of 2 months or less was 81% (96 of 119). The TTR was not assessed in the original 2019 submission.

Figure 3: Swimmer Plot of Time to Response and Overall Treatment Duration for ePAS4 (Database Cut-Off: July 15, 2019)

ePAS4 = extended primary analysis set 4.

Source: Note to clinical reviewer (2020).²²

Figure 4: Swimmer Plot of Time to Response and Overall Treatment Duration for ePAS5 (Database Cut-Off: July 2020)

ePAS5 = extended primary analysis set 5.

Source: Updated ePAS5 data.²⁷

SAS3

The median TTR in patients with primary CNS tumours was 1.82 months (range = 0.99 to 3.75) (Figure 5).²² The majority of patients experienced a response in less than 2 months (67%). At the time of data cut-off, 5 patients had a confirmed response; of these, 4 (80%) did not have disease progression at the time of data analysis.

Figure 5: Swimmer Plot of Time to Response and Overall Treatment Duration for SAS3 (Database Cut-Off: July 15, 2019)

SAS3 = safety analysis set 3.

Source: Note to clinical reviewer (2020).²²

DOR

ePAS4 and ePAS5

As of the data cut-off date for ePAS4 (July 15, 2019), after a median follow-up of 15.7 months (IQR = 6.6 months to 24.8 months), the median DOR per IRC assessment was NE (95% CI, 27.6 months to NE) (Table 24, Figure 6). Ninety patients (76%) had an ongoing response during this period. The percentage of participants with responses lasting longer than 24 months was 18% (21 of 119). At the time of the July 2020 data cut-off for ePAS5 (n = 192), the DOR was 34.5 months (95% CI, 27.6 to 54.7) with a median follow-up of 20.3 months (IQR = data NA), 79% and 66% patients had a DOR of 12 and 24 months, respectively.

The PAS (n = 55) was also evaluated for the DOR as of the July 15, 2019, data cut-off (Table 25).²² After a median follow-up of 28.2 months, the median DOR was 32.9 months (95% CI, 14.8 months to NE). A DOR of 24 months or more was seen in 62% (95% CI, 46% to 78%) of patients. In the original CADTH submission, the median DOR had not been reached as of the July 30, 2018, data cut-off date (integrated dataset). At that time point, 88% and 75% patients had an ongoing response at 6 months and 12 months from the start of response, respectively (Appendix 3).

Figure 6: Kaplan-Meier Plot of DOR Based on IRC Assessment in ePAS4 (Subset of Patients With Confirmed Response; n = 119; Database Cut-Off: July 15, 2019)

DOR = duration of response; ePAS4 = extended primary analysis set 4; IRC = independent review committee; No. = number.

Source: Note to clinical reviewer (2020).²²

SAS3

At the time of the data cut-off, the median DOR had not been reached (95% CI, 3.8 months to NE), with a median duration of follow-up of 5.3 months (IQR = 3.6 to 10.1 months) (Figure 5).²²

Progression-Free Survival

ePAS4 and ePAS5

As of the data cut-off date for ePAS4 (July 15, 2019), after a median follow-up of 14.0 months (IQR = 7.9 months to 26.6 months), the median PFS per IRC assessment was 33.4 months (95% CI, 19.3 months to NE) (Table 25, Figure 7).²² The percentage of patients with progressive disease was 34%; however, 60% were alive without documented disease progression, 3% had a surgical resection of their tumour without CR, and no evaluable post-baseline disease assessment was available for 3%. At 2 years or more, 58% (95% CI, 48% to 67%) of all patients remained progression-free. As of the data cut-off for ePAS5 (July 2020), after a median follow-up of 22.1 months (IQR = data not available), the median PFS per IRC assessment was 33.4 months (95% CI, 22.5 to 43.5), with 12- and 24-month PFS rates of 67% and 57%, respectively (Figure 7).

The PAS patient dataset (n = 55) was also evaluated for PFS as of the July 15, 2019, data cut-off (Table 25).²² After a median follow-up of 30.3 months, the median PFS was 29.4 months (95% CI, 10.9 months to NE). At 2 years or more, 52% (95% CI, 38% to 66%) of all patients remained progression-free. In the original submission, the median PFS as of the July 30, 2018, data cut-off date (integrated dataset) was 28.3 months (95% CI, 9.9 to NE) (Appendix 3).

Figure 7: Kaplan-Meier Plot of PFS Based on IRC Assessment in ePAS4 (n = 164; Database Cut-Off: July 15, 2019)

ePAS4 = extended primary analysis set 4; IRC = independent review committee; No. = number; PFS = progression-free survival.

Source: Note to clinical reviewer (2020).²²

Figure 8: Kaplan-Meier Plot of PFS Based on IRC Assessment in ePAS5 (n = 192; Database Cut-Off July 2020)

ePAS5 = extended primary analysis set 5; IRC = independent review committee; No. = number; PFS = progression-free survival;

Source: Updated ePAS5 data.²²

SAS3

At the time of the data cut-off, 58% of patients had not progressed, and the median PFS was 11.0 months (95% CI, 5.4 to NE), with a median duration of follow-up of 5.6 months (IQR = 3.6 to 13.1) (Figure 9).²² With the new data cut-off of July 2020 (n = 33,), the median PFS was 18.3 months (95% CI, 6.7 to NE) after a median follow-up of 16.5 months; the PFS rates at 12 and 24 months were 56% and 42%. respectively (data not presented).

Figure 9: Kaplan-Meier Plot of PFS Based on Investigator Assessment in SAS3 (n = 164; Database Cut-Off: July 15, 2019)

No. = number; PFS = progression-free survival; SAS3 = supplementary analysis set 3.

Note: Tick marks represent censored patients.

Source: Note to clinical reviewer (2020).²²

Overall Survival

ePAS4 and ePAS5

As of the data cut-off date for ePAS4 (July 15, 2019), after a median follow-up of 15.8 months (IQR = 9.3 months to 28.8 months), the median OS was NE (95% CI, 44.4 months to NE) and 85% of patients were alive (Table 26, Figure 10).²² At 2 and 2 years of follow-up, the probability of survival was 90% and 82%, respectively. As of the data cut-off for ePAS5 (July 2020), after a median follow-up of 24.0 months (IQR = data not available), the median OS per IRC assessment was NE (95% CI, NE to NE), with 12- and 24-month OS rates of 89% and 82%, respectively (Figure 11).

The PAS (n = 55) was also evaluated for OS as of the July 15, 2019, data cut-off (Table 26).²² After a median follow-up of 32.5 months, the median OS was NE (95% CI, 44.4 months to NE). At 12 months or more, 90% of patients were alive, and at 24 months or more, 82% of patients were alive. The OS results were not available from the integrated analysis (original submission); However, in the analysis of the ePAS (February 19, 2018, data cut-off), which was used for the Health Canada submission, the median OS had not been reached after a median follow-up of 14.8 months (Appendix 3).

Figure 10: Kaplan-Meier Plot of Overall Survival Based on the ePAS4 (n = 164; Database Cut-Off: July 15, 2019)

ePAS4 = extended primary analysis set 4; No. = number.

Source: Note to clinical reviewer (2020).²²

Figure 11: Kaplan-Meier Plot of Overall Survival Based in ePAS5 (n = 192; Database Cut-Off: July 2020)

ePAS5 = extended primary analysis set 5; No. = number.

Source: Updated ePAS5 data.²⁷

SAS3

At the time of the data cut-off, 96% of patients were still alive and the median OS was NE (95% CI, 9.4 months to NE), with a median follow-up duration of 6 months (Figure 12).²² The estimated 12-month OS rate was 88% (95% CI, 65 to 100%). With the new data cut-off of July 2020 (n = 33), the median OS was NE (95% CI, 16.9 to NE) after a median follow-up of 16.5 months; the OS rates at 12 and 24 months were 85% and 58%, respectively (data not presented).

Figure 12: Kaplan-Meier Plot of Overall Survival Based on Investigator Assessment in SAS3 (Database Cut-Off: July 15, 2019)

No. = number; SAS3 = supplementary analysis set 3.

Source: Note to clinical reviewer (2020).²²

Health-Related Quality of Life

As previously described in the Analysis Populations section, the HRQoL analyses dataset (July 15, 2019, data cut-off) included patients with non-CNS primary solid tumours, NTRK gene fusion, and measurable disease from the LOXO-TRK-15003 and LOXO-TRK-15002 trials.⁶⁰ Analyses of HRQoL outcomes were not pre-specified in the sponsor-submitted statistical analysis plan for the pooled analyses, and it was not clear which analyses or subgroups were pre-specified (if any); however, analyses were performed and presented as part of conference preceedings.^59,60 The analysis included 126 patients who had received larotrectinib and completed a baseline and at least 1 post-baseline questionnaire (74 adults, 24 children ≥ 2 years old, and 28 infants < 2 years old). As of the data cut-off, 126 patients were included in the analyses as they had received larotrectinib and completed questionnaires at baseline and at least 1 post-baseline follow-up visit (74 adult patients for EORTC QLQ-C30, 24 patients ≥ 2 years of age completed by the patient or their parent/caregiver for PedsQL and 28 infants < 2 years of age completed by their parent/caregiver for PedsQL).

The numbers and proportions of patients with above norm/norm and Below Norm HRQoL scores for the EORTC QLQC-30 GHS and PedsQL total score at baseline and at best response are displayed in Table 27.⁶⁰ Of the 52 adult patients with an EORTC QLQ-C30 GHS at norm/above norm at baseline, 98% remained in this category at the best response, and of the 22 patients at below norm at baseline, 9% remained in this category and 91% improved to norm/above norm. The GHS is rated on a 7-point Likert-type scale, which is then converted to a standardized score ranging from 0 to 100, with higher scores representing better QoL.⁶⁷ The mean EORTC QLQ-C30 GHS for the US general population (63.9)²⁴ minus 10 points (the MID) was used to construct the norm/above norm (≥ 53.9) and below norm (< 53.9) score categories for adults.²⁵ Of the 9 pediatric patients 2 years of age or older with a PedsQL total score at norm/above norm at baseline, 100% remained in this category at the best response, and of the 15 patients at below norm at baseline, 33% remained in this category and 67% improved to norm/above norm. The PedsQL total score is the sum of all 23 items over the number of items answered on all the component scales; higher scores indicate better HRQoL.⁶⁹ The average score for the combined self and proxy-reported PedsQL questionnaire for healthy US children (85.0)²⁶ minus 4.5 points (the MID) was used to construct the norm/above norm (≥ 80.5) and below norm (< 80.5) score categories for children 2 years of age and older. The sponsor identified that the MID for the PedsQL total score for children younger than 2 years of age was 7.2 points.

The analysis results for the best change from baseline in HRQoL scores is summarized in Table 28. The majority of adult patients (59%), children 2 years of age and older (79%), and children younger than 2 years of age (57%) had improvements in the best post-baseline score at or above the sponsor-identified (and literature-supported) MID. Of the patients evaluable for a sustained improvement (i.e., with a baseline and at least 2 post-baseline assessments; magnitude not defined), the improvement was sustained for at least 2 consecutive cycles for 47% of adult patients, 75% of children 2 years of age and older, and 43% of children younger than 2 years of age, and the improvement was sustained until the end of assessments in 30% to 50%, and 29%, respectively.

EORTC QLQ-C30 Results

The best changes from baseline (magnitude not defined) in EORTC QLQ-C30 GHS are summarized in Table 28.⁶⁰ The mean of best changes in total score from baseline was 17.5 (SD = 20.0) points. Of the 74 adult patients who completed an EORTC QLQ-C30 questionnaire, 69% had an improvement from baseline scores, with 59% reporting a best post-baseline score that reached or exceeded the sponsor-identified MID of 10 points. Among patients evaluable for a sustained improvement (i.e., those with a baseline and at least 2 post-baseline assessments), 47% had a sponsor-identified improvement in an EORTC QLQ-C30 GHS score that lasted for at least 2 consecutive cycles and 30% had a sponsor-identified improvement that was sustained until the end of assessments. Sustained improvements (magnitude not defined) occurred by 2 months of treatment in 69% of adult patients, and the median duration of sustained improvement was 12.0 months (range = 1.7 to 20.3 months).

PedsQL Total Score Results

The best changes from baseline in PedsQL total score are summarized in Table 28.⁶⁰ The mean of best changes in total score from baseline was 20.7 points (SD = 17.2) in patients 2 years of age or older and 12.0 points (SD = 13.8) in patients younger than 2 years of age. Of the 24 patients 2 years of age or older and the 28 patients younger than 2 years of age who completed PedsQL questionnaires, 88% and 82%, respectively, had an improvement from baseline scores, with 79% and 57%, respectively, reporting a best post-baseline score that reached or exceeded the sponsor-identified (and literature-supported) MID of 4.5 points. Among patients evaluable for a sustained improvement (i.e., with a baseline and at least 2 post-baseline assessments; magnitude not defined), 75% and 43%, respectively, had an improvement in PedsQL total score that lasted for at least 2 consecutive cycles, and 50% and 29%, respectively, had an improvement that was sustained until the end of assessments. Sustained improvements occurred by 2 months of treatment in 75% of patients 2 years of age and older and in 61% of patients younger than 2 years of age, and the median durations of sustained improvement were NE (range = 1.1 to 20.3 months) and NE (range = 0.9 to 16.7 months), respectively.

Harms

A summary of reported AEs in the TRK fusion cancer labelled-dose safety set and the overall labelled-dose safety set (descriptions provided in the Analysis Populations section) as of the July 15, 2019, data cut-off is presented in Table 29.²² Most AEs occurring in either safety set were grade 1 or 2.

In the TRK fusion cancer labelled-dose safety set (n = 196), the most common AEs (regardless of grade) were increased ALT (32%), cough (31%), fatigue (30%), constipation (29%), diarrhea (28%), increased AST (27%), dizziness (26%), pyrexia (24%), anemia (23%), vomiting (23%), and nausea (22%).²² Grade 3 or 4 AEs occurred in 96 patients (49%); however, grade 3 or 4 AEs related to treatment occurred in only 27 patients (14%). The most common grade 3 or 4 AEs were anemia (8%), decreased neutrophil count (8%), and decreased lymphocyte count (5%). Larotrectinib treatment was interrupted or the dosage was modified in 79 patients (40%); however, only 37 patients (19%) had a treatment interruption or modification attributable to treatment. Permanent discontinuation of larotrectinib for TEAEs, regardless of attribution, occurred in 5% of patients receiving the recommended dose. Results were consistent in the newly submitted overall NTRK fusion cancers safety set (N = 260).

In the overall labelled-dose safety set (n = 238), the most common AEs (regardless of grade) were fatigue (31%), cough (29%), increased ALT (29%), constipation (26%), diarrhea (26%), dizziness (26%), increased AST (25%), anemia (24%), nausea (24%), vomiting (24%), and pyrexia (24%).²² Grade 3 or 4 AEs occurred in 120 patients (50%); however, grade 3 or 4 AEs related to treatment occurred in only 35 patients (15%). The most common grade 3 or 4 AEs were anemia (9%), decreased neutrophil count (7%), and decreased lymphocyte count (5%). Larotrectinib treatment was interrupted or the dosage was modified in 97 patients (41%); however, only 49 patients (21%) had a treatment interruption or modification attributable to treatment. Permanent discontinuation of larotrectinib for (TEAEs), regardless of attribution, occurred in 8% of patients. Results were consistent in the newly submitted overall safety set (N = 331).

Critical Appraisal

Internal Validity

Heterogeneity in Design Elements of Studies Included in the Pooled Analysis

Interpretation of pooled analysis results remains difficult in the presence of between-study heterogeneity due to several factors:

• Different phased studies: Given the rare nature of NTRK fusion–positive solid tumours and methodological challenges, the sponsor rationalized that the conduct of a randomized trial was not feasible.⁷² The submitted data were therefore pooled from 3 single-arm trials: a phase I adult trial (LOXO-TRK 14001), a phase I/II pediatric trial (LOXO-TRK-15003 [SCOUT]), and a phase II basket trial (LOXO-TRK-15002 [NAVIGATE]) in adults and adolescents. The phase II part of the SCOUT trial, investigating long-term safety and efficacy of larotrectinib in pediatric patients, is ongoing and results are yet to be published. The objectives of the 3 included trials differed: LOXO-TRK-14001 – safety; SCOUT phase I – safety; SCOUT phase II – ORR; and NAVIGATE – ORR.

• Different primary outcomes: The primary objective of the dose escalation phases of the LOXO-TRK 14001 and SCOUT studies was to determine the safety and tolerability of larotrectinib, while the primary objective of the NAVIGATE trial was to determine the efficacy of larotrectinib by measuring the best ORR.

• Different requirements for outcome measurement: In the LOXO-TRK 14001 and SCOUT trials, the ORR was assessed by the investigator using RECIST 1.1 or RANO criteria, as appropriate to tumour type; whereas in the NAVIGATE trial, the ORR was determined by an independent radiology review committee using RECIST 1.1 or RANO criteria.

• Different eligibility criteria: As mentioned earlier in this section, LOXO-TRK 14001 included adult patients, SCOUT included pediatric patients, and NAVIGATE enrolled adults and adolescent patients. Additionally, the LOXO-TRK-15002 trial included patients with an ECOG status of 3, while LOXO-TRK 14001 included patients with an ECOG status of 0 to 3, and SCOUT did not specify ECOG status as part of its inclusion criteria. In addition, the presence of confirmed NTRK fusion was mandated before enrolment in the NAVIGATE trial, while an NTRK-positive status was not a requirement for eligibility in the LOXO-TRK 14001 and SCOUT trials. Fusions of TRK genes were identified prospectively in the 2 latter trials. These sources of heterogeneity in the patient selection criteria may introduce bias to the results of the pooled analysis.

Given the heterogeneities across studies, findings from the pooled analysis are associated with methodological limitations, even though there may be a rationale (rarity of NTRK fusion–positive cancer) for pooling data. The sponsor submitted 3 additional analyses to address concerns about the inherent heterogeneity of the baskets, described in the Other Relevant Evidence section, which should be consulted in this regard.

Uncertainty Around the Pooled Analysis Results

The following limitations should be considered when interpreting the pooled analysis results:

• The pooled analyses were performed as post hoc decisions for the included trials. While a statistical analysis plan was provided for the pooled analyses, the trials contributing to the pooled sets were not designed to provide statistical power for the post hoc pooled analyses. As previously stated, the objectives of the 3 included trials differed: LOXO-TRK-14001– safety; SCOUT phase I – safety; SCOUT phase II – ORR; and NAVIGATE – ORR. As a result, not all the trials that contributed to the pooled datasets were designed to investigate efficacy (i.e., ORR) as the primary objective of the trial. It is also unclear if the pooled analysis has sufficient power to detect a clinically meaningful effect for the primary and secondary outcomes. Nonetheless, a small proportion of the total patients in ePAS4 were obtained from phase I portions of Study 14001 and NAVIGATE (29 of 164 patients or 17.7%); the majority (135 patients) came from the phase II studies. The sponsor provided a separate sensitivity analysis showing that the inclusion or exclusion of phase I study data did not meaningfully affect the ORR. The ORR was 73% (95% CI, 65 to 79%) including phase I data, and 70% (95% CI, 61 to 77%) excluding phase I data.

• The estimates of response outcomes were pooled. Due to the small sample size, there is uncertainty regarding the magnitude of the treatment effect of larotrectinib in any 1 histologic subtype of solid tumours with an activating NTRK rearrangement. The clinical experts consulted for this review agreed that the pooled ORR estimate for treatment effect may be generalizable to all of the subgroups. However, the pooled treatment effect varied widely across tumour types. The reported point estimates for the ORR across tumour types ranged from 100% in GIST, CMN, and cancers of unknown primary origin down to 0% in appendix, pancreas, hepatic, and prostate cancers, and cholangiocarcinoma. Additionally, unbalanced and small sample sizes for each tumour type resulted in significant uncertainty in the subgroup results by tumour type, which was reflected in the wide confidence intervals around the point estimates. Further, because these subgroups were descriptive in nature, any interpretation would be inconclusive for comparative purposes. In the aforementioned subgroup analysis, 1 patient was enrolled in each of the cancers of unknown primary origin, CMN, appendix, hepatic, and prostate tumour subgroups.

• Pooling data on survival outcomes (i.e., PFS and OS) can be even more problematic if there is variability in the PFS or OS across tumour types. This is because traditional survival analysis methods such as Kaplan-Meier curves rely on the assumption that a single survival distribution can be used to estimate the survival of all study participants.

• The efficacy analyses for SAS3 used investigator-assessed outcomes, which is particularly problematic given the small sample sizes in this dataset and the open-label design of the trial.

• All 3 larotrectinib trials are ongoing. The LOXO-TRK-14001 trial stopped enrolment in 2017; however, SCOUT and NAVIGATE are still enrolling patients. The results of the pooled analysis are therefore subject to change as more data become available.

• The HRQoL analyses were not pre-specified as part of the sponsor-provided statistical analysis plan, limiting the ability to interpret the results including what decisions were made post hoc.

  • It was not clear whether the subgroup analyses provided in the results were pre-specified or if the subgroups were determined after the analyses were run.

  • The analyses considered the patient’s HRQoL results at their best response, not at their last reported response. The differences between the 2 could be significant; for example, patients may respond well at a given time point or study visit, but the benefits may not continue in subsequent visits or at their last reported response. Without any information on when the best change was determined, it is unclear if the best change was, in fact, a transient improvement in HRQoL.

  • The analyses using norm/above norm and below norm categories had several limitations. First, norm/above norm could be almost 10 points lower than the population normal to begin with. And patients could experience a clinically relevant decline in HRQoL as long as they did not fall below the norm. Second, patients who were close to the threshold (a score of 53.9) only required a small change to move above or below the norm, even though that would not be considered clinically relevant as the estimated MID was 10 points. Although a sensitivity analysis conducted using a higher threshold (58.9 for adults and 82.75 for children) showed similar results, the aforementioned limitations still remained.

  • Last, the HRQoL results were presented descriptively, not analyzed statistically, which limits the interpretability of the results.

External Validity

Potential limitations to the external validity of the pooled analysis include:

• The larotrectinib trials included patients with NTRK-positive solid tumours regardless of their tumour type. However, not all solid tumour types were represented in the studies.

• The pooled analysis excluded patients with primary CNS tumours; however, these patients were included in another dataset.

• The eligibility criteria for the 3 larotrectinib trials did not restrict the number of previous lines of systematic therapy.

Other Relevant Evidence

The sponsor included 3 additional analyses in this submission to address concerns about inherent heterogeneity across tumour type and patients. These include a BHM evaluating ORR across tumour types, a permutation analysis of ORR of all possible subsets across study and tumour type, and an intra-patient comparison of GMI.

• Three main sources of heterogeneity are of concern: the heterogeneity of the effects and background prognosis by tumour location or type; the heterogeneity of effects and background prognosis by tumour histology; and the inherent heterogeneity of the effects of treatment between patients, which is uncontrolled given the lack of a control arm. Controlled and randomized trials stratified by histology and tumour location for an outcome of OS, rather than a surrogate such as ORR, could potentially reduce issues of heterogeneity of study participants across studies or prognosis by tumour histology or location. However, the rarity of NTRK fusion cancers makes such an approach difficult and it was not deemed possible for the sponsor’s subject patient population.

• Three alternative analyses were submitted by the sponsor to addresses these sources of heterogeneity using the available data. The submitted analyses include the BHM⁷³ and the permutation analysis,⁷⁴ which attempt to quantify and account for the heterogeneity in ORR across tumour locations and studies, and possibly account for other differences in histology and the intra-person GMI analysis.^75,76 This analysis attempts to mitigate for the lack of control — and therefore heterogeneity — between participants due to histology and location, as well as many other factors, by using individual times to progression under previous treatment to compare to their PFS under their later treatment with larotrectinib.

• The populations for the 3 analyses were not the same across all subanalyses and are presented for each analysis separately.

• The presentation of these analyses can be found in the note to clinical reviewer, section 5.10, pages 67 through 79.²²

Bayesian Hierarchical Modelling

Populations and Methods

The primary analysis used ePAS4, which included 164 participants over the 3 clinical trials. The data used in the analysis are presented in Table 31. Two sensitivity analyses were performed on ePAS4, with the primary CNS added along with a subset of the data removing all tumour histology subgroups with 3 or fewer participants, and combining both breast tumour types into 1.

Interventions

• Larotrectinib

Outcomes

• ORR

Points of Statistical Analysis Critical for Bayesian Hierarchical Modelling

The outcome of the model was the number of responders out of the number of the total participants per tumour type, as presented in Table 31. Responders are anyone classified as having either a CR or PR according to RECIST 1.1. A CR is defined as no radiological evidence of disease, negative surgical margins, and no viable tumour cells. A PR is defined as a minimum decrease from baseline in the sum of target lesion diameters of at least 30%. This gives a single parameter, P_j, for each tumour type (j), which is then logit-transformed and assumed to come from a normal with mean (µ) and SD (σ). This is the assumed distribution of the transformed parameter logit (P_j), called the prior. Hyper-priors are placed on µ, again normal with a mean of −0.85, corresponding to 0.3 on the ORR scale, with a variance of 10 and a uniform value (0,5) for σ. These are based on priors suggested by Cunanan et al. (2019).⁷⁷ The same analysis was also performed on a subset of ePAS4; all model specifications remain the same.

All reported estimates for parameters are medians taken over the samples from the posterior. The median of the posterior of is referred to as the pooled estimate and is given on the ORR scale. This is the median of the posterior of µ, transformed, which is the centre of the logit distribution (P_j) and serves to shrink the tumour type P_j toward a common centre value. The measure of the heterogeneity of the logit (Pj) over the tumour types is given by σ, and this is reported in the results as the measure of heterogeneity across the tumour types. The reported estimates for this are medians. All credible intervals are 95% wide between the 2.5 percentile and the 97.5 percentile of the posterior samples. The reported posterior probabilities of exceeding a given ORR level are the proportion of the samples from the posterior that exceed the given ORR value. The results for each of the tumour types is the median for the specific P_j posterior sample. The posterior predictive distribution is simply the posterior distribution for µ because there are no fixed effects or other terms in the model. This gives predictions for a new tumour type that are considered exchangeable, coming from the same posterior distribution as the currently evaluable tumour types used to build the model.

Markov chain Monte Carlo sampling was used to generate draws from the posterior, the distribution generated by combining the prior and likelihood. A total of 55,000 iterations of 2-chains were run, using 5,000 burn-ins and a thinning of 2. Convergence was assessed via twice the binomial log-likelihood with zero responder correction. Exact zero values are not used in the log-likelihood, as this would cause it to fail, and the models are stated to have converged. The analysis also corrects for zero responders in the group by setting any exact zero number of responders or expected number of responders to 0.1, which is needed due to the logit transformation.

Results

Complete ePAS4 Data

The posterior median of σ, which is the measure of the heterogeneity of the logit (Pj), was 1.55 (95% credible interval, 0.70 to 3.35). The sponsor reported that this value demonstrates the heterogeneity of the data. The pooled P estimate, the median of the posterior of µ, transformed to be on the ORR scale (expit(μ)) is 0.624 (95% CI, 0.331 to 0.812). Table 2 from the submitted technical report⁷³ is reproduced in Table 32, which includes the results from this model for each of the tumour types.

The posterior predictive ORR for a new tumour type based on this model is presented in Figure 13. The results suggest that a new tumour type with the same posterior distribution as the currently evaluable types would be predicted to have a 0.629 probability of having a 50% or higher ORR, and a 0.793 probability of having a 30% or higher ORR.

Sensitivity Analysis

Two sensitivity analysis were performed. One was performed on a combined dataset including patients from ePAS4 and patients from the primary CNS cancer dataset using the same BHM. The primary CNS tumours were treated as a single tumour type. The response rate for primary CNS tumours was estimated to be 23.0% (95% CI, 9.7 to 41.8). Estimated response rates per tumour type were reported to be similar to the main model in principle, with negligible corrections for tumour types with high observed response rates and differences of 4% to 5% for tumour types with low numbers of patients studied and no responses observed. The pooled mean response from the combined model was estimated to be 58.2% (95% CI, 30.9 to 78.5). An estimate of the heterogeneity is not provided for this analysis.

The second sensitivity analysis removed tumour types with only 1 or 2 patients studied. Therefore, an identical BHM as the main model was simulated, but the tumour types “bone sarcoma,” “cholangiocarcinoma,” “pancreas,” “CMN,” and “unknown primary cancer” were excluded. All breast cancer tumour types were combined into a single type for this analysis, resulting in a pooled estimate of ORR of 74.8% (95% CI, 51.2 to 89.7), and a reduced estimate of heterogeneity of 1.23.