CADTH Reimbursement Review

Lorlatinib (Lorbrena)

Sponsor: Pfizer Canada

Therapeutic area: ALK-positive locally advanced or metastatic non–small cell lung cancer

This multi-part report includes:

Clinical Review

Pharmacoeconomic Review

Stakeholder Input

Clinical Review

Executive Summary

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

Introduction

Lung cancer is the most commonly diagnosed cancer in Canada and the leading cause of cancer-related deaths.² The majority of lung cancers are classified as non–small cell lung cancer (NSCLC). The 5-year survival for patients with NSCLC is 25%, and this estimate is even lower for patients with metastatic NSCLC.³ Approximately 5% of NSCLC tumours harbour the anaplastic lymphoma kinase (ALK) tumour mutation, which is a rare mutation that is responsive to small-molecule tyrosine kinase inhibitors (TKIs).^4,5 Patients with the ALK-positive NSCLC are highly susceptible to developing intracranial metastases and typically have a life expectancy of between 5 years and 10 years. Approximately 25% of patients develop intracranial metastases during the first year of their disease, and those who do not will develop intracranial metastases within 5 years.⁶ Treatment with TKIs may help increase the chance of patient survival.⁷

New-generation TKIs, such as alectinib and brigatinib, have become the new standards of care over crizotinib (a first-generation TKI) because of their improved efficacy and superior penetration of the blood-brain barrier for treating brain metastases. Alectinib is typically the first-line treatment choice for patients; however, the CADTH pan-Canadian Oncology Drug Review Expert Review Committee (pERC) recently issued a positive reimbursement recommendation for brigatinib for the same indication. Brigatinib is currently under consideration for price negotiations with the pan-Canadian Pharmaceutical Alliance. The clinical experts consulted by CADTH indicated there is an unmet need for treatments after patients progress on first-line therapies, and that lorlatinib should be made available for use in multiple lines of therapy. The clinical experts stated that their preference would be to reserve use of lorlatinib for a later line of therapy unless the patient presented with active central nervous system (CNS) metastases, in which case the preference would be to use it as first-line treatment.

Lorlatinib is a selective, brain-penetrating, small molecule that competes with adenosine triphosphate for binding of ALK and ROS1 tyrosine kinases. This mechanism of action addresses resistance mechanisms following previous treatment with ALK inhibitor therapy.¹ Lorlatinib was approved by Health Canada as monotherapy for the treatment of patients with ALK-positive locally advanced or metastatic NSCLC. The recommended dosage of lorlatinib is 100 mg orally once daily continuously. Treatment with lorlatinib should continue until disease progression or the advent of unacceptable toxicity.¹ Previous CADTH reviews for advanced or metastatic ALK-positive NSCLC include brigatinib,^8,9 lorlatinib,¹⁰ alectinib,^11-13 ceritinib,^14,15 and crizotinib.¹⁶ The previous CADTH review for lorlatinib was for the treatment of adult patients with ALK-positive metastatic NSCLC who had progressed on crizotinib and at least 1 other ALK inhibitor, or patients who had progressed on ceritinib or alectinib. The pERC review of lorlatinib for this indication in 2020–2021 resulted in a negative recommendation for reimbursement due to uncertain clinical benefits in this treatment setting stemming from limitations of the evidence, including a non-randomized clinical trial with no specific hypothesis testing.

The objective of this CADTH review is to perform a systematic review of the efficacy and safety of lorlatinib as monotherapy for the first-line treatment of patients with ALK-positive locally advanced (not amenable to curative therapy) or metastatic NSCLC.

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

Input was received from 2 patient groups: CanCertainty Coalition and Lung Cancer Canada (LCC). Patients did not contribute to the submission from the CanCertainty Coalition. The received input highlighted the financial burdens associated with oral lung cancer treatments, which are not funded in the same manner as IV therapies, and coverage varies by province. In Ontario and the Atlantic provinces, only individuals over the age of 65 are automatically covered for oral oncology medications. According to CanCertainty, for patients without private insurance, access to medication requires navigating a complicated process of funding applications that are associated with approval delays, and most often result in patients incurring out-of-pocket costs. The group also indicated that the high cost of oral therapies may result in medication nonadherence, particularly among younger and lower-income patients.

The submission from LCC was based on data retrieved through interviews, questionnaires, and environmental scanning of records of patients and caregivers of patients with ALK-positive NSCLC. Data were received from 17 patients, including 9 females and 8 males, most of whom were 35 years of age or older. Twelve of the respondents were patients and 5 were caregivers. The LCC respondents were from Spain, the US, Canada, the UK, Switzerland, Philippines, and online.

Respondents in the LCC submission emphasized the unmet need for treatments that provide a cure for their lung cancer. Currently, all treatment options are considered palliative. Unmet need was also highlighted for patients with brain metastases, as there are few effective treatment options to treat brain involvement. Respondents described their experiences receiving crizotinib, ceritinib, alectinib, and chemotherapy. Crizotinib, while an effective treatment option, was described as not as effective against brain metastases, resulting in the need for radiation therapy. Patients also reported difficult side effects with crizotinib and ceritinib. Alectinib was described by LCC as the current standard of care for patients with ALK-positive NSCLC due to its efficacy and reduced toxicity compared to crizotinib, and as an effective treatment for patients with brain metastases. Chemotherapy was described to be associated with toxic side effects and limited benefit. The LCC submission also described the burden of disease on caregivers who are often at the centre of their loved one’s care and who often require taking time off work, resulting in further financial burden.

The LCC highlighted the following goals for treatments: improvement in disease symptoms, preservation of patient quality of life, manageable toxicity profiles for treatments, delayed progression, and maintenance of patient’s functionality and independence. The LCC gathered the experiences of 17 patients who had experience with lorlatinib; however, only 1 of these patients (from Spain) had received lorlatinib as first-line treatment. The respondents reported positive experiences with lorlatinib treatment, noting that it showed efficacy against their disease, including metastases, and therefore provided a sense of hope. Patients also commented that lorlatinib had a tolerable toxicity profile, improved their disease symptoms, and preserved independence and quality of life, with patients reporting being able to return to work, engaging in social activities, and having more energy.

Clinician Input

Input From Clinical Experts Consulted by CADTH

The 2 clinical experts consulted by CADTH stated that, in Canada, alectinib is the first-line treatment used for most patients with ALK-positive NSCLC, although brigatinib is also an option. Lorlatinib would serve as another first-line option for patients; however, the clinicians highlighted that the use of lorlatinib would also be beneficial in later lines of therapy. According to the clinical experts, the goals of therapy are to prolong life, improve disease symptoms, maintain quality of life, delay progression, reduce the severity and frequency of symptoms, and reduce loss of cognition. Both clinical experts highlighted the need for curative therapies that are better tolerated and preserve patient quality of life. They also noted that improved biomarker-targeted therapies are needed to allow for multiple lines of therapy that provide patients with additional treatment options upon disease progression. Patients with brain metastases were highlighted as a group of patients with unmet needs, as currently there are few therapies that also have efficacy in the brain. Patients eligible for treatment are identified through imaging and ALK testing. Assessment varies by line of therapy, but typically patients are assessed every 3 months, with brain imaging conducted every 6 months. The clinical experts indicated that lorlatinib could be administered in inpatient and outpatient settings, and discontinuation of the drug would occur once patients experience clinical deterioration and cognitive dysfunction that affects quality of life.

Clinician Group Input

Two clinician group inputs were received, 1 from the Ontario Health (Cancer Care Ontario) Lung and Thoracic Cancer Drug Advisory Committee and 1 from LCC. In total, input was received from 26 clinicians. Identification of the patients who would be eligible for treatment was stated to occur up front, as ALK testing occurs at initial diagnosis. Both inputs identified alectinib and brigatinib, which is currently accessed through special access programs, as the available first-line treatments for patients with ALK-positive NSCLC. Both inputs noted that treatment goals include prolonging life; delaying disease and CNS progression; maintaining or improving quality of life; reducing severity of symptoms; minimizing adverse events (AEs); reducing the loss of cognition, memory and other sequalae of CNS metastases and its local treatments; and maintaining patient independence. The input from Cancer Care Ontario stated that many of these needs are addressed through alectinib; however, new treatments that provide longer control of symptomatic disease, improved progression-free survival (PFS) and overall survival (OS) are desired. Input from LCC indicated an unmet need for more effective therapies in the first line, alternative therapies to allow for individualization of therapy, convenient dosing of treatments, and more effective therapies that treat brain metastases. Both LCC and Cancer Care Ontario emphasized the need for more effective treatments in later lines of therapy, as patients eventually become refractory to currently available treatment options. While both groups acknowledged that lorlatinib would be a first-line option for patients, they also stated that lorlatinib could address treatment gaps in later lines of therapy. After patients progress on lorlatinib in the first line, TKIs are not typically available to patients in later lines of therapies and both clinician groups stated that use of ALK TKIs after first-line therapy would be preferential.

According to the clinician groups, assessment of a patient’s response to treatment is based on improvement of symptoms and assessment of radiographic response, as well as through PFS, OS, and intra- and extracranial PFS. The clinician groups agreed that testing for response should occur every 2 to 3 months, with imaging conducted every 2 to 6 months, or as needed. Patients would be discontinued from treatment due to disease progression or unmanageable toxicities. Lorlatinib would be administered in an outpatient setting, although community or inpatient settings were considered acceptable at times under the supervision of the prescribing oncologist.

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. For the CADTH review of lorlatinib, the drug programs provided input and/or had questions pertaining to the initiation of therapy, the prescribing of therapy, generalizability, funding algorithms, care provision, and system and economic issues. The clinical experts consulted by CADTH weighed evidence from the CROWN trial and other clinical considerations to provide responses to questions, which can be found in the Drug Program Input section.

Clinical Evidence

Pivotal Studies and Protocol-Selected Studies

Description of Studies

One multinational, multi-centre, randomized, active-controlled, open-label superiority trial met the criteria for the CADTH systematic review. The CROWN trial evaluated the efficacy and safety of lorlatinib compared to crizotinib as first-line treatment in adult patients with locally advanced or metastatic ALK-positive NSCLC who had not received previous systemic treatment for metastatic disease. Patients who were diagnosed with and treated for an earlier stage of disease were eligible for enrolment if their treatment was completed more than 12 months before randomization. Eligible patients were required to have their ALK status confirmed through an immunohistochemistry (IHC) test approved by the FDA or Pharmaceutical and Medical Devices Agency or marked by Conformité Européene performed on Ventana ULTRA or XT platforms and a good performance status defined as an Eastern Cooperative Oncology Group Performance Status (ECOG PS) of 0 to 2. Patients with brain metastases were eligible for enrolment.¹⁷

The trial recruited patients from 104 sites in 23 countries (in Asia, the European Union, and North America), including Canada. A total of 296 patients were randomized in a 1:1 ratio using an interactive web-based response technology system; 149 patients were randomized to the lorlatinib group and 147 patients were randomized to the crizotinib group. Randomization was stratified according to presence of brain metastases (yes versus no) and ethnic origin (Asian versus non-Asian). Patients randomized to the lorlatinib group received 100 mg daily, and patients randomized to the crizotinib group received 250 mg daily. Both lorlatinib and crizotinib were administered orally.¹⁷

The primary objective of the study was to determine whether lorlatinib was superior to crizotinib in prolonging PFS based on Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST 1.1) criteria according to a blinded independent central review (BICR). The trial was designed as a group sequential trial using a Lan-DeMets (O’Brien-Fleming) alpha spending function to determine efficacy boundaries. The overall significance level was preserved at 0.025 with a 1-sided stratified log-rank test. The trial results were based on an interim analysis (with a data cut-off date of March 20, 2020), after approximately 133 PFS events (75%) had occurred per BICR. A final analysis of PFS was specified only if the boundary for efficacy was not crossed at the interim analysis.

OS was planned as a secondary end point that was hierarchically tested if statistical significance was obtained for PFS. Other pre-planned exploratory end points of the trial included PFS according to investigator assessment, objective response rate (ORR), duration of response (DOR), time to response (TTR), and intracranial efficacy end points (intracranial objective response rate [IC-ORR], intracranial duration of response [IC-DOR], intracranial time to progression [IC-TTP], and intracranial time to response [TTR]); these end points were not part of the statistical testing hierarchy. Health-related quality of life (HRQoL) was measured using the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 (EORTC QLQ-C30), its corresponding survey for lung cancer (EORTC QLQ-LC13), and the EQ-5D 5-Levels questionnaire (EQ-5D-5L).

Patient characteristics at baseline were mostly balanced between the treatment groups. Patients’ mean ages were 59 years (standard deviation [SD] = 13) in the lorlatinib group and 56 years (SD = 14) in the crizotinib group. A higher proportion of patients in the lorlatinib group were 65 years or older compared to patients in the crizotinib group (39.6% versus 29.9%, respectively). There were more females in both the lorlatinib (56.4%) and crizotinib (61.9%) groups. Most patients were White (48.2% versus 49.0%, respectively) or Asian (43.6% versus 44.2%, respectively).¹⁷ Almost all patients had measurable disease at baseline (96.6% versus 97.3%, respectively), with approximately one-quarter of patients presenting with brain metastasis (25.5% versus 27.2%, respectively). Most patients had an ECOG PS of 0 (45.0% versus 38.8%, respectively) or 1 (53.0% versus 55.1%, respectively), with the adenocarcinoma type of NSCLC (94.0% versus 95.2%, respectively), which is stage IV metastatic disease (90.6% versus 94.6%, respectively). Most patients were classified as either never smokers (54.4% versus 63.9%, respectively) or former smokers (36.9% versus 29.3%, respectively).¹⁷

Efficacy Results

Key efficacy results from the CROWN trial are reported in Table 2. Two analyses for PFS were planned, an interim and a final analysis. At the data cut-off date (March 20, 2020), results for PFS crossed the pre-specified stopping boundary for statistical significance, which favoured the lorlatinib group (stratified hazard ratio [HR] = 0.28; 95% confidence interval [CI], 0.191 to 0.413; stratified log-rank 1-sided P < 0.0001). The results for PFS at the data cut-off date were considered final. At the data cut-off date, OS was also tested in accordance with the statistical testing hierarchy, with the results showing that the majority of patients remained alive; there were 23 deaths (15.4%) in the lorlatinib group and 28 deaths (19.0%) in the crizotinib group and the between-group difference was not statistically significant (HR = 0.72; 95% CI, 0.414 to 1.249). The intracranial efficacy end points, including IC-ORR, IC-DOR, IC-TTP, and IC-TTR, were also assessed, with the results demonstrating a consistently improved response among patients with brain metastases who were treated with lorlatinib compared to those treated with crizotinib. However, the CROWN trial was not powered to assess these end points, and the analyses of intracranial efficacy end points are therefore considered exploratory.¹⁷

Health-related quality of life was assessed as a pre-specified exploratory end point in the CROWN trial. No clinically meaningful differences between treatment groups, based on a difference of at least 10 points, were observed in any of the EORTC QLQ-C30 functioning domains. In general, the mean change in scores from baseline to the end of study period were similar for the EORTC QLQ-C30 and QLQ-LC13 in both treatment groups in the Global Health Scale and subscales. Also, the mean scores in the EQ-5D-5L index and Visual Analogue Scale (EQ VAS) were similar in both treatment groups. The time to deterioration (TTD) analysis conducted for symptom scales in the EORTC QLQ-C30 also showed no differences between the lorlatinib and crizotinib groups.¹⁷

Harms Results

In general, AEs were more commonly reported in patients treated in the lorlatinib group than in the crizotinib group. The most common AEs in the lorlatinib group were hypercholesterolemia (70.5% versus 3.5% in the lorlatinib group), hypertriglyceridemia (63.8% versus 5.6%, respectively), edema (55.0% versus 39.4%, respectively), increased weight (38.3% versus 12.7%, respectively), peripheral neuropathy (33.6% versus 14.8%, respectively), cognitive effects (21.5% versus 5.6%, respectively), diarrhea (21.5% versus 52.1%, respectively), and dyspnea (20.1% versus 16.2%, respectively). More AEs related to CNS effects were reported in the lorlatinib group than in the crizotinib group (cognitive effects: 21.5% versus 5.6%, respectively; mood effects: 16.1% versus 4.9%; speech effects: 4.7% versus 0; psychotic effects: 3.4% versus 0). Serious adverse events (SAEs) of any grade (34.2% versus 27.5%) and grade 3 or 4 AEs (22.8% versus 17.6%) were higher in the lorlatinib group than in the crizotinib group, respectively.¹⁷ More AEs related to CNS effects were reported in the lorlatinib group than in the crizotinib group (cognitive effects: 21.5% versus 5.6%, respectively; mood effects: 16.1% versus 4.9%; speech effects: 4.7% versus 0; psychotic effects: 3.4% versus 0), and SAEs of any grade (34.2% versus 27.5%) and grade 3 or 4 (22.8% versus 17.6%) were higher in the lorlatinib group than in the crizotinib group, respectively.¹⁷

AEs that resulted in dose reductions were generally infrequent, occurring in 31 patients (20.8%) in the lorlatinib group and 22 patients (15.5%) in the crizotinib group. Grade 3 AEs resulting in dose reductions occurred in 9 patients (6.0%) in the lorlatinib group and 7 patients (4.9%) in the crizotinib group; no grade 4 AEs resulted in dose reductions in either treatment group. Dose interruptions due to AEs occurred in similar proportions of patients in both the lorlatinib (49.0%) and crizotinib (44.4%) treatment groups, with 32.9% and 36.6% of interruptions due to grade 3 or 4 AEs.¹⁷

Deaths occurred in 23 patients (15.4%) in the lorlatinib group and 18 patients (19.7%) in the crizotinib group, with most deaths considered to be due to disease progression (11.4% versus 16.2%, respectively).¹⁷

Notable harms identified by the sponsor included hypercholesterolemia, hypertriglyceridemia, edema, peripheral neuropathy, CNS effects, vision disorder, pneumonitis, weight gain, liver function test increase, QT prolongation, atrioventricular block, and pancreatitis, and these aligned with the CADTH protocol. Notable AEs were more common in the lorlatinib group, except for vision disorder and liver function test increase, which were more common in the crizotinib group than in the lorlatinib group (39.4% versus 18.1% and 37.3% versus 20.8%, respectively). The most common AEs of special interest were hypercholesterolemia (70.5% in the lorlatinib group versus 3.5% in the crizotinib group), hypertriglyceridemia (63.8% versus 5.6%, respectively), edema (55.0% versus 39.4%, respectively), weight gain (38.3% versus 12.7%, respectively), peripheral neuropathy (33.6% versus 14.8%, respectively), cognitive effects (21.5% versus 5.6%, respectively), liver function test increase (20.8% versus 37.3%, respectively), and mood effects (16.1% versus 4.9%, respectively).¹⁷

Critical Appraisal

The CROWN trial was a multinational, multi-centre, open-label, phase III trial that employed a group sequential design. A BICR was implemented to assess end points that involved judgments of patients’ clinical progression (i.e., PFS, ORR, and DOR). However, it is possible that the open-label design posed a greater risk of bias for end points involving subjective reporting, such as HRQoL and safety (e.g., CNS effects).

The primary and secondary end points of the CROWN trial were PFS and OS. Both end points were considered in power calculations, and OS was tested hierarchically at the time of the data cut-off if PFS was deemed to be statistically significant. Other secondary and exploratory end points were not included in the statistical hierarchy. The statistically significant findings on subgroup analyses were likely subject to multiplicity and an inflated type I error rate. At the time of the data cut-off, the interim analysis of PFS had crossed the pre-specified efficacy boundary and showed a statistically significant difference in favour of lorlatinib, and the analysis was therefore considered final by the sponsor. However, OS data were deemed immature, as only 26% of the 198 OS events required for the final analysis of OS had occurred.¹⁷ An improvement in PFS may not always correlate with a difference in OS in assessments of oncology treatment benefit. In fact, mainly due to disease progression, a higher proportion of patients in the crizotinib group compared to the lorlatinib group (17.4% versus 58.5%, respectively) withdrew from the study. This would largely bias the estimate of OS in the final analysis. Therefore, further evidence is required to confirm the superiority of lorlatinib over crizotinib in treatment efficacy in terms of OS.

The majority of patients included in the CROWN trial had an ECOG PS score 0 or 1. The generalizability of the results in terms of a PFS benefit to patients with a poor ECOG PS remains unknown. Moreover, the study excluded patients with potential vascular or cardiac diseases, or those with unfavourable laboratory testing of renal, liver, pancreatic, or bone marrow function. In reality, the safety profile of lorlatinib for patients with those comorbidities or abnormal testing may be even worse considering that lorlatinib increased the risk of hypercholesterolemia and hypertriglyceridemia. The CROWN trial allowed for enrolment of patients with brain metastases, who accounted for 25.5% of the lorlatinib group and 27.2% of the crizotinib group.¹⁷ Inclusion of patients with brain metastases is highly relevant as many patients with ALK-positive metastatic NSCLC develop brain metastases. The results for the assessed intracranial-specific efficacy end points consistently showed numerically improved outcomes in the lorlatinib group compared with the crizotinib group. Despite the limitations associated with exploratory end points, the clinical experts consulted by CADTH recognized the results of patients with brain metastases as noteworthy.

Indirect Comparisons

Description of Studies

Four indirect treatment comparisons (ITCs) were summarized and critically appraised, including 1 from the sponsor¹⁸ and 3 published ITCs by Chuang et al. (2021),¹⁹ Wang et al. (2021),²⁰ and Ando et al. (2021).²¹

The ITCs compared the safety and efficacy of lorlatinib to alectinib (600 mg and 300 mg), brigatinib, crizotinib, ceritinib, chemotherapy, and ensartinib among patients with ALK-positive metastatic NSCLC who had not received prior systemic treatment in the first line. While not all ITCs included comparisons to each of these treatments, all ITCs compared lorlatinib to alectinib, brigatinib, and crizotinib.

The sponsor’s ITC compared lorlatinib to alectinib (600 mg and 300mg), brigatinib, ceritinib (450 mg, 600 mg, 750 mg), crizotinib, chemotherapy, and ensartinib.¹⁸ Ando et al. (2021),²¹ compared lorlatinib to alectinib, brigatinib, ceritinib, crizotinib, and chemotherapy. Wang et al. (2021)²⁰ compared lorlatinib to alectinib and brigatinib. Chuang et al. (2021)¹⁹ compared lorlatinib to alectinib (600 mg and 300 mg), brigatinib, crizotinib, and ensartinib.

Efficacy Results

Efficacy results reported here focus on PFS as this was the primary end point of all trials included in the ITCs.

Results of the sponsor’s ITC¹⁸ favoured lorlatinib over all comparators, including alectinib at 600 mg (HR = 0.61; 95% credible interval [CrI], 0.38 to 0.99), brigatinib (HR = 0.57; 95% CrI, 0.34 to 0.95), ceritinib at 750 mg (HR = 0.22; 95% CrI, 0.13 to 0.37), ceritinib at 450 mg (HR = 0.31; 95% CrI, 0.15 to 0.66), ceritinib at 600 mg (HR = 0.25; 95% CrI, 0.12 to 0.54), crizotinib (HR = 0.0.28; 95% CrI, 0.19 to 0.41), ensartinib (HR = 0.55; 95% CrI, 0.32 to 0.93), and chemotherapy (HR = 0.12; CrI, 0.08 to 0.19) except for alectinib at 300 mg (HR = 0.83; 95% CrI, 0.36 to 1.85).

The results in the ITC by Ando et al. (2021)²¹ favoured lorlatinib over all comparators, including brigatinib (HR = 0.572; 95% CrI, 0.326 to 0.997), ceritinib (HR = 0.220; 95% CrI, 0.131 to 0.367), crizotinib (HR = 0.280; 95% CrI, 0.191 to 0.411), and chemotherapy (HR = 0.121; 95% CrI, 0.078 to 0.187), except for alectinib (HR = 0.742; 95% CrI, 0.4666 to 1.180).²¹

The ITC by Wang et al. (2021)²⁰ conducted comparisons among patients who were ALK inhibitor– and chemotherapy-naive, and patients who were ALK inhibitor–naive. Results favoured lorlatinib compared to alectinib (ALK inhibitor– or chemotherapy-naive patients: HR = 0.59; 95% CrI, 0.37 to 0.94; ALK inhibitor–naive: HR = 0.65; 95% CrI, 0.42 to 1.01) and brigatinib (ALK inhibitor– or chemotherapy-naive patients: HR = 0.54; 95% CrI, 0.31 to 0.94; ALK inhibitor–naive: HR = 0.57; 95% CrI, 0.34 to 0.95) in both groups of patients.

In the ITC by Chuang et al. (2021),¹⁹ lorlatinib was favoured over crizotinib (HR = 0.28; 95% CrI, 0.19 to 0.41), ensartinib (HR = 0.54; 95% CrI, 0.32 to 0.92), and brigatinib (HR = 0.57; 95% CrI, 0.32 to 0.95), but not over alectinib at 600 mg (HR = 0.68; 95% CrI, 0.42 to 1.08) or 300 mg (HR = 0.76; 95% CrI, 0.34 to 1.28).

Harms Results

The sponsor’s ITC¹⁸ conducted a safety analysis for grade 3 and 4 AEs. Grade 3 and 4 AEs were |||||||||||||||| reported in the lorlatinib group compared to alectinib at 300 mg (odds ratio [OR] = ||; 95%CrI, ||||||||||||||) or 600 mg (OR = ||; 95% CrI, ||||||||||||) and crizotinib (OR = ||; 95% CrI, ||||||||||||). |||||||||||||||| were observed between lorlatinib and brigatinib (OR = ||; 95% CrI, ||||||||), ceritinib (750 mg) (OR = ||; 95% CrI, ||||||||||||), ensartinib (OR = ||; 95% CrI, ||||||||||||), and chemotherapy (OR = ||; 95% CrI, ||||||||||||).

Ando et al. (2021)²¹ conducted safety analyses for any grade of AEs, SAEs, grade 3 or higher SAEs, and specific AEs including nausea, diarrhea, increased alanine transaminase (ALT), increased aspartate transaminase (AST), and pneumonitis. Because a different number of trials was included in the analysis of each safety end point, the comparators were different for each safety end point. In general, lorlatinib was not favoured over comparators. For any grade of AEs, no treatments were favoured between lorlatinib and alectinib (relative risk [RR] = 1.018; 95% CrI, 0.985 to 1.051), lorlatinib and brigatinib (RR = 1.041; 95% CrI, 1.001 to 1.083), or lorlatinib and crizotinib (RR = 1.010; 95% CrI, 0.985 to 1.035). Regarding SAEs, no treatments were favoured between lorlatinib and alectinib (RR = 1.614; 95% CrI, 1.041 to 2.503) or lorlatinib and crizotinib (RR = 1.249; 95% CrI, 0.881 to 1.768). Regarding grade 3 or higher AEs, no treatments were favoured between lorlatinib and alectinib (RR = 1.255; 95% CrI, 0.737 to 2.146) or lorlatinib and crizotinib (RR = 1.219; 95% CrI, 0.816 to 1.818). Regarding specific AEs (nausea, diarrhea, increased ALT and/or AST, and pneumonitis), lorlatinib was generally favoured over chemotherapy, crizotinib, or ceritinib, but not over alectinib or brigatinib.

Wang et al. (2021)²⁰ conducted safety analyses involving assessments of AEs, AEs leading to treatment discontinuation, and AEs leading to dose reduction. None of the treatments (lorlatinib, alectinib, and brigatinib) were favoured over another.

Chuang et al. (2021)¹⁹ conducted a safety analysis for AEs of grade 3 or higher. Lorlatinib had a higher risk of grade 3 or higher AEs compared to crizotinib (RR = 1.27; CrI 1.07 to 1.52), and alectinib at 600 mg (RR = 1.62; 95% CrI, 1.24 to 2.12) and 300 mg (RR = 2.09; 95% CrI, 1.48 to 2.95), but not brigatinib (RR = 1.07; 95% CrI, 0.84 to 1.37).

Critical Appraisal

All ITCs involve issues related to heterogeneity. Specifically, differences in baseline characteristics may limit the comparability of patients across trials. For example, there were differences in the proportions of patients with brain metastases, the enrolment of patients from Asian and non-Asian countries, and the inclusion of patients who may have received prior treatment with an ALK inhibitor and/or chemotherapy. These characteristics may serve as treatment-effect modifiers of the comparisons of efficacy and safety in the ITCs. In some cases, the ITCs conducted subgroup or sensitivity analyses that accounted for differences in some but not all of these characteristics. The sponsor’s ITC included the ASCEND-8 trial, which was a phase I, dose-ranging, active-controlled trial. The inclusion of this trial is likely to have introduced bias to the comparisons with ceritinib, although it is possible that the evidence base of the ITC was broadened by inclusion of this trial. In addition, some studies included in the ITCs assessed treatment at different doses; specifically, alectinib was assessed at 300 mg and 600 mg. While some ITCs considered the 2 doses to be different nodes in the overall networks of comparisons, 2 of the ITCs combined data from trials that assessed alectinib at different doses and included only 1 node for alectinib. The different doses of alectinib may not be equivalent in efficacy or safety, and comparisons against alectinib that include data from both doses (600 mg and 300 mg) may have introduced uncertainty. The efficacy end points of PFS, OS, and ORR were assessed in the ITCs. Only PFS was powered for all trials included in the ITCs, and interpretation of evidence should therefore be limited to this end point. Overall, due to limitations of the ITCs, it is not possible to determine the true magnitude and direction of comparative treatment effects between lorlatinib, alectinib, and brigatinib.

Conclusions

One ongoing, phase III, open-label, randomized superiority trial (CROWN) provided evidence regarding the efficacy and safety of lorlatinib as first-line treatment in adult patients with ALK-positive locally advanced (not amenable to curative therapy) or metastatic NSCLC. Compared to crizotinib, patients treated with lorlatinib showed a statistically significant improvement in PFS that was considered clinically meaningful by the clinical experts consulted by CADTH. A consistent improvement in PFS was observed in all patient subgroups assessed, most notably those with brain metastasis. The intracranial efficacy outcomes assessed in the trial (IC-ORR, IC-TTP, IC-DOR, IC-TTR), although exploratory, also showed a consistent treatment benefit for lorlatinib. These results suggest that, compared to crizotinib, lorlatinib may be capable of improved penetration of the blood-brain barrier and may be an option for patients with brain metastasis. Due to immaturity of the trial data, no evidence was available on OS. The trial is ongoing, and longer-term data are needed to determine whether the observed PFS benefit will translate to an improvement in OS. Lorlatinib may have a poorer safety profile compared to crizotinib, as the incidence of grade 3 and 4 AEs was higher (by approximately 17%) in patients treated with lorlatinib, although this did not appear to result in a higher rate of dose modification, interruption, or treatment discontinuation. In particular, patients and clinicians considering initiating treatment with lorlatinib should be aware that lorlatinib was associated with a higher incidence of neurologic AEs (i.e., cognitive and mood effects). The HRQoL analyses did not reveal any clinically meaningful differences between the treatment groups. In general, the results of the CROWN trial support the use of lorlatinib as another first-line treatment option for patients with ALK-positive locally advanced or metastatic NSCLC.

Comparison of efficacy and safety between lorlatinib and other TKIs, specifically alectinib and brigatinib, were considered more relevant in the Canadian first-line treatment setting for advanced or metastatic ALK-positive NSCLC. The ITCs included in this review showed that, similar to lorlatinib, both alectinib and brigatinib are associated with improved PFS compared to crizotinib, and therefore are more likely to be choices for first-line therapy than crizotinib. However, given the limitations related primarily to clinical and methodological heterogeneity across trials included in the ITCs, the magnitude and direction of comparative estimates of efficacy and safety between lorlatinib, alectinib, and brigatinib are uncertain.

Introduction

Disease Background

Lung cancers are the most commonly diagnosed cancers in both males and females in Canada. In 2020, they were expected to account for 29,800 cases. Lung cancer is also the leading cause of cancer-related death; approximately 25% of cancer deaths were expected to be due to lung cancer.² Lung cancers may be classified as either small cell lung cancer or NSCLC. The latter accounts for approximately 88% of lung cancer cases in Canada (excluding Quebec) and is classified into histological subtypes, including adenocarcinoma, squamous cell carcinoma, and large cell carcinoma; of these, adenocarcinomas are the most commonly diagnosed histological subtype. Patients with adenocarcinoma typically do not have any smoking history and may grow more slowly than other types of lung cancers.²² The 5-year survival of NSCLC varies depending on the stage, but on average, the estimated 5-year survival for NSCLC is 25%.³

Some lung cancers can harbour chromosomal rearrangements, including ALK rearrangement, which is responsive to small-molecule TKIs and occurs in approximately 5% of NSCLC tumours.^4,5 Testing for the ALK rearrangement occurs at initial diagnosis using an IHC test or ribonucleic acid panels that show the fusion mutation of ALK. Patients with the ALK-positive disease typically have a life expectancy of between 5 and 10 years. This patient group is highly susceptible to developing intracranial metastases. Approximately 25% of patients will typically develop intracranial metastases within the first year of their disease, and 70% to 80% will develop intracranial metastases approximately 5 years after their disease onset.⁶ Treatment with TKIs may help increase the chance of patient survival.⁷

Standards of Therapy

Current National Comprehensive Cancer Network guidelines suggest the following treatments as first-line options for patients with ALK-positive NSCLC: alectinib, brigatinib, ceritinib, crizotinib, or lorlatinib.²³

In addition to these treatments, the clinical experts consulted by CADTH indicated that ensartinib may also be used. These first-line treatments have been studied in clinical trials against crizotinib. However, the clinical experts noted that crizotinib is no longer commonly used as a first-line treatment for ALK-positive metastatic NSCLC due to its poor penetration of the blood-brain barrier.

The clinical experts consulted by CADTH indicated that, based on data from the ALEX trial, the first-line treatment for ALK-positive patients is typically alectinib. However, it was also noted that first-line treatment with brigatinib has shown similar efficacy to alectinib. The clinical experts agreed that lorlatinib could serve as another first-line option, given the improved efficacy observed in the CROWN trial. While no head-to-head studies comparing lorlatinib to available first-line treatments other than crizotinib have been conducted, lorlatinib is associated with the lowest HR against crizotinib.

The clinical experts consulted by CADTH indicated that no ALK inhibitors are currently publicly reimbursed for use as second-line treatment for patients with ALK-positive metastatic NSCLC. Consequently, second-line therapies typically include any treatment available through a special access program, or chemotherapy. Lorlatinib has received Health Canada approval for use as second- or third-line treatment for patients who have progressed on crizotinib and at least 1 other ALK inhibitor or patients who have progressed on ceritinib or alectinib. However, as previously mentioned, the CADTH review for this indication resulted in a negative recommendation for reimbursement from pERC. In the opinion of the clinical experts consulted by CADTH, lorlatinib should be made available for use in multiple lines of therapy, and their preference would be to reserve its use for later-line therapy unless the patient presents with active CNS metastases, in which case the preference would be to use it as first-line treatment. Lorlatinib is typically not used in combination with other therapies and should only be used as a monotherapy; although some patients may require concurrent radiotherapy for a variety of reasons, including symptomatic brain or bone metastases and oligoprogression.

The clinicians consulted by CADTH consider lorlatinib to be best suited for patients with brain metastases due to its ability to penetrate the blood-brain barrier. These patients would typically be identified using radiographic imaging or MRI. However, it was noted that MRI can be associated with challenges such as long wait times and high costs of travel to regional sites. Patients with asymptomatic brain metastases were considered to be candidates for treatment with lorlatinib, as evidence suggests that lorlatinib both delays the occurrence of brain metastases and treats it. One of the clinical experts stated that it may be possible for patients to be misdiagnosed with metastases; imaging for patients typically occurs around every 6 months, but many patients may not be diagnosed until symptoms appear.

The clinical experts consulted by CADTH indicated that a patient’s response to treatment is assessed through imaging, including CT, MRI, carcinoembryonic antigen scans, and chest X-rays, and assessment of clinical progression. Patients may be assessed for response to treatment every 3 months, with imaging occurring approximately every 6 months. Although assessments may vary depending on the line of therapy, the clinical experts indicated that a clinically meaningful response to treatment would be improved survival, reduced symptom severity, improvement in symptoms, and maintenance quality of life, and ability to complete tasks of daily living. Treatments would typically be discontinued if patients experience clinical deterioration, and when cognitive dysfunction affects their quality of life.

Overall, the goals of treatment for patients with ALK-positive metastatic NSCLC include improving survival; improving or maintaining quality of life, employment, and independence; decreasing disease progression, and reducing severity of symptoms.

No treatments for ALK-positive metastatic NSCLC patients were considered curative and all currently available treatments are palliative, although treatments do provide benefits of disease response, decreased disease progression, and improved survival. In terms of treatment gaps. In addition, treatments with improved toxicity profiles that allow patients to preserve their quality of life are needed. Both clinicians emphasized the need for treatments that target specific biomarkers to help direct treatment. Current targeted treatment options are limited to first-line therapies, and further lines of therapy are needed to provide options for patients whose disease will ultimately progress.

Drug

Lorlatinib is approved by Health Canada and indicated for use as monotherapy for first-line treatment of adult patients with ALK-positive locally advanced (not amendable to curative therapy) or metastatic NSCLC.¹ Lorlatinib is a selective adenosine triphosphate–competitive small molecule that can penetrate the blood-brain barrier and inhibit ALK and ROS1 tyrosine kinases.¹ The Health Canada–recommended dose of lorlatinib is 100 mg taken orally once daily. The sponsor has requested the reimbursement of lorlatinib according to the Health Canada indication.

Key characteristics commonly used in treatments of ALK-positive metastatic NSCLC are presented in Table 3.

Stakeholder Perspectives

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 4.

Clinical Evidence

The clinical evidence included in the review of lorlatinib is presented in 2 sections. The first section, 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 indirect evidence from the sponsor and indirect evidence selected from the literature that met the selection criteria specified in the review.

Systematic Review of Pivotal and Protocol-Selected Studies

Objectives

To perform a systematic review of the beneficial and harmful effects of lorlatinib at 100 mg orally once daily for the first-line treatment of adult patients with ALK-positive locally advanced (not amenable to curative therapy) or metastatic NSCLC.

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 5. Outcomes included in the CADTH review protocol reflect those 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 PRESS Peer Review of Electronic Search Strategies checklist.²⁷

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.

Published literature was identified by searching the following bibliographic databases: MEDLINE All (1946–) via Ovid and Embase (1974–) via Ovid. All Ovid searches were run simultaneously as a multi-file search. Duplicates were removed using Ovid deduplication for multi-file searches, followed by manual deduplication in EndNote. 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 lorlatinib. Clinical trials registries searched the US National Institutes of Health’s clinicaltrials.gov, the WHO International Clinical Trials Registry Platform (ICTRP) search portal, Health Canada’s Clinical Trials Database, and the European Union Clinical Trials Register.

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. Appendix 1 provides detailed search strategies.

The initial search was completed on July 21, 2021. Regular alerts updated the search until the meeting of pERC on November 10, 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.²⁸ 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.

Findings From the Literature

A total of 345 studies were identified from the literature for inclusion in the systematic review (Figure 1). The included studies are summarized in Table 6. A list of excluded studies is presented in Appendix 1.

Figure 1: Flow Diagram for Inclusion and Exclusion of Studies

Description of Studies

The CROWN trial was a phase III, multinational, multi-centre, open-label randomized controlled trial (RCT) comparing the efficacy and safety of lorlatinib to crizotinib among patients with locally advanced (stage IIIB, not amenable for multimodality treatment) or metastatic (stage IV) ALK-positive NSCLC who had not received previous systemic treatment for metastatic disease. Patients (N = 296) from 104 centres in 23 countries, including 2 patients from Canada (1 in each treatment group), were randomized in a 1:1 ratio using an interactive web-based response technology system to receive either lorlatinib or crizotinib. Patients had to receive study treatments no later than 7 days after they were randomized. Patents were stratified by the presence of brain metastases (yes versus no) and ethnic origin (Asian versus non-Asian).¹⁷ Patients were classified as having brain metastases if they had brain lesions that were present and observed at study entry or which were known from a patient’s medical history to have been present in the past even if the brain lesions were no longer visible (e.g., irradiated or surgically removed). Brain lesions were permitted to be either measurable or nonmeasurable and were classified as target or nontarget lesions. The ethnic origin of patients referred to the patient’s actual race and not where they were living or where they were being treated. Stratified randomization was centrally allocated across all participating study centres through the interactive response technology system. Crossover was not permitted between treatment groups.¹⁷

Patients were required to undergo molecular testing for the presence of ALK tumour aberrations. Patients who screened as ALK-negative were considered screen failures.¹⁷ All patients were required to undergo CNS imaging using MRI (unless contraindicated) at baseline. A BICR was used for assessment of imaging to determine patient’s overall tumour response according to RECIST 1.1 criteria. The BICR of patient intracranial tumour response used a modified version of RECIST 1.1 criteria.

The CROWN trial included 3 phases: screening, treatment, and follow-up. Assessments were conducted at screening and then repeated every 8 weeks (± 1 week) starting from randomization while on treatment until disease progression. Patients were eligible to be treated beyond progression. If patients continued treatment beyond disease progression, they continued receiving tumour assessments every 8 weeks (± 1 week). Assessments for bone progression were conducted every 16 weeks (± 1 week) only if evidence of bone metastases was observed at baseline.¹⁷ Patients who discontinued treatment for reasons other than disease progression were followed for disease progression regardless of receipt of new anticancer therapy. Tumour assessments were repeated after at least 4 weeks to confirm patient response (complete response [CR] or partial responses [PR]).¹⁸

The primary end point of the CROWN trial was PFS with 2 pre-specified analyses, including interim and final analyses. The interim analysis was pre-specified to occur when approximately 133 events occurred (75% of the 177 events planned at the end of the study). By the data cut-off date (March 20, 2020), 127 PFS events had occurred (72% of the 177 events planned for the final analysis of PFS). The pre-specified superiority boundary for PFS was crossed at the interim analysis and was considered final, although the trial is currently ongoing. The secondary end point of the trial was OS, which was hierarchically tested depending on statistical significance of PFS. By the data cut-off date, OS data were immature as the majority of patients remained alive. Two additional pre-specified analyses are planned for OS, with the next analysis planned to occur in early 2025. The CROWN trial was funded and conducted by Pfizer.

A representation of the CROWN trial study design is depicted in Figure 2.

Figure 2: CROWN Study Design

ALK = anaplastic lymphoma kinase; BICR = blinded independent central review; BID = twice a day; CNS = central nervous system; ECOG PS = Eastern Cooperative Oncology Group Performance Status; NSCLC = non–small cell lung cancer; PD = progressive disease; QD = every day; RECIST v1.1 = Response Evaluation Criteria in Solid Tumors Version 1.1. RP2D = recommended phase II dose.

Source: CROWN Clinical Study Report.¹⁷

Populations

Inclusion and Exclusion Criteria

Key inclusion and exclusion criteria used in the CROWN trial are summarized in Table 6. Briefly, the trial enrolled adult patients with histologically or cytologically confirmed diagnosis of locally advanced or metastatic ALK-positive NSCLC. ALK-positive status was required to have been determined by a Ventana ALK (D5F3) CDx IHC test performed on the Ventana ULTRA or XT platform. Patients were also required to have at least 1 extracranial measurable target lesion (per RECIST 1.1 criteria) that had not been previously irradiated. Patients could have treated or untreated CNS metastases, as long as they were asymptomatic and aligned with other criteria specified in Table 6. Patients were required to have an ECOG PS of 0, 1, or 2.¹⁷

Key exclusion criteria included the presence of chronic or uncontrolled conditions, such as vascular or nonvascular conditions, predisposing characteristics for acute pancreatitis, lung disease, or psychiatric conditions. Patients must not have received prior systemic treatment for advanced or metastatic NSCLC; although prior treatment for earlier stages of NSCLC was permitted if it was completed more than 12 months before randomization. Patients must also not have received major surgery within 4 weeks of randomization or have spinal cord compression unless the patient had good pain control attained through therapy and had stabilized or recovered neurological function within 4 weeks of randomization. Radiation therapy, including stereotactic or partial brain irradiation, was also not permitted within 2 weeks before randomization. Patients who completed brain irradiation within 4 weeks before randomization or who received palliative radiotherapy outside of the CNS within 48 hours before randomization were also not permitted in the trial.¹⁷

Baseline Characteristics

A summary of the baseline characteristics of patients in the CROWN trial is presented in Table 7. Demographic characteristics were generally balanced across the treatment groups. The mean age of patients was 59 years (SD = 13) in the lorlatinib group and 56 years (SD = 14) in the crizotinib group, with most patients being 45 years of age or older (83% versus 76.2%, respectively). A larger proportion of patients in the lorlatinib group were aged 65 years or older than patients in the crizotinib group (39.6% versus 29.9%, respectively), and a slightly smaller proportion of patients in the lorlatinib group were between 18 and 45 years of age compared with the crizotinib group (17.4% versus 23.8%, respectively). Most patients were female, (56.4% and 61.9%, respectively), White (48.2% and 49.0%, respectively) or Asian (43.6% versus 44.2%, respectively). Of the Asian patients, most were Japanese (38.5% versus 35.4%, respectively) or Chinese (40.0% versus 35.4%, respectively), followed by Korean (12.3% versus 20.0%, respectively) and other (9.2% in both groups).¹⁷

Baseline disease characteristics were balanced across the treatment groups. Almost all patients had measurable disease at baseline (96.6% in the lorlatinib group versus 97.3% in the crizotinib group), with approximately one-quarter of patients presenting with brain metastasis (25.5% versus 27.2%, respectively). Of patients with brain metastasis at baseline, 17 (11.4%) in the lorlatinib group and 13 (8.8%) in the crizotinib group had measurable brain disease. Most patients had an ECOG PS of 0 (45.0% in the lorlatinib group versus 38.8% in the crizotinib group) or 0 (53.0% versus 55.1%, respectively) and were classified as having adenocarcinoma (94.0% versus 95.2%, respectively) or stage IV metastatic NSCLC (90.6% versus 94.6%, respectively). Patients were classified as either never smokers (54.4% versus 63.9%, respectively) or former smokers (36.9% versus 29.3%, respectively); a greater proportion of patients in the crizotinib group were classified as never smokers, and more patients in the lorlatinib group were classified as former smokers.¹⁷

Table 7: Summary of Baseline Characteristics for the CROWN Study — Full Analysis Set

AJCC = American Joint Committee on Cancer; FAS = full analysis set; SD = standard deviation; TNM = tumour, nodes, metastases.

^aAge at screening (years) = (date of given informed consent − date of birth + 1)/365.25.

^bPer independent central neuroradiological review.

^cOne patient with locally advanced disease at study entry was staged according to AJCC version 8.0, instead of AJCC version 7.0 as required by protocol was therefore classified under Other.

Source: CROWN Clinical Study Report.¹⁷

Prior Treatments

In general, the prior therapies patients received were similar in each treatment group (Table 8). Prior radiotherapy (13.4% versus 13.6% in the lorlatinib and crizotinib groups, respectively) or surgery (18.1% versus 15.6%) were the most commonly reported prior anticancer therapies. Prior systemic anticancer treatments were reported infrequently among patients and were primarily received in the adjuvant treatment setting (6.7% versus 5.4%). Prior adjuvant and neoadjuvant therapies were infrequently reported among patients in the lorlatinib and crizotinib groups (6.7% versus 5.4% and 0.7% versus 0.7%, respectively). One patient in the lorlatinib group received carboplatin and paclitaxel for advanced disease violating eligibility criteria; this event was captured as an important protocol deviation.

Interventions

Patients were randomized to receive either lorlatinib or crizotinib as monotherapy. Lorlatinib was administered to patients orally at 100 mg using four 25 mg tablets once daily. Crizotinib was administered orally at a dosage of 250 mg twice daily; capsules of crizotinib were available at 200 mg or 250 mg.¹⁷

Patients were required to swallow the lorlatinib tablet or crizotinib capsule whole without manipulating or chewing before swallowing at the same time every day.¹⁷

Each cycle was specified to be 4 weeks (or 28 days) regardless of dose delays, dose interruptions, or missed doses. Treatments were administered to patients until disease progression confirmed by BICR, patient refusal, the patient was lost-to-follow-up, unacceptable toxicity, or if the study was terminated by the sponsor, whichever came first.¹⁷ Patients were eligible for treatment beyond progression (confirmed by BICR) if they were thought to still be deriving benefit from continued treatment, as long as the treating physician determined that the benefits and risks were favourable for the patient.¹⁸ Patients who received treatment beyond progression were required to undergo the same assessments during the active treatment period, including the following tumour assessments:

• For patients with only documented extracranial progression, intracranial lesions stable or in response, intracranial assessments were to be performed until intracranial disease progression.

• No further tumour assessments were required once intracranial disease progression was documented.

Patient Compliance

As patients self-administered treatment at home, compliance was captured and completed by the patient using a diary. The diary was to be maintained by the patient and include unchanged, missed, or changed doses of study treatments received. Patients randomized to either group were required to return bottles of lorlatinib or crizotinib at the end of each 28-day cycle during their planned visit to the investigational site. The number of tablets of lorlatinib or capsules of crizotinib were documented and recorded at each clinic visit. Study sites were required to follow up with patients on day 5 of cycle 1 (± 3 days) to confirm the patient’s understanding and compliance with treatment dosing instructions. Patients were re-trained if necessary. If patients required a dose modification, the same follow-up procedures were applied.¹⁷

Dose Modifications

Dosing of treatments was permitted to be delayed and/or reduced in the event of significant toxicity experienced by the patient. Dose modifications were based on the worst toxicity observed if multiple toxicities occurred. Patients were instructed to inform their investigators at the first occurrence of any AE. In the event of AEs, trial investigators were instructed to employ the best supportive care practices of local institutions or follow pre-specified dose modifications for lorlatinib or crizotinib.¹⁷

Doses of lorlatinib were permitted to be modified at 2 levels; the first level involved a dose reduction to 75 mg, and the second level involved a dose reduction to 50 mg. Dose reductions beyond the second level (50 mg) were not permitted. Re-escalation of doses was also not permitted unless there was a discussion with the sponsor’s medical contact. Precautions surrounding AEs of hyperlipidemia and PR-interval prolongation were pre-specified in the CROWN trial protocol as these were expected AEs that have been observed in previous studies of lorlatinib.¹⁸

Doses of crizotinib were permitted to be modified at 2 levels; the first level involved a dosage reduction to 200 mg twice daily, and the second dose level involved a dosage reduction to 250 mg once daily. Dosage reductions beyond the second level (250 mg once daily) were not permitted. Re-escalation of crizotinib was not permitted except in specific circumstances. Precautions surrounding AEs of nausea and vomiting, diarrhea, bradycardia, pneumonitis, pneumonia, renal cysts, and severe visual loss were pre-specified.¹⁸

Patients were required to permanently discontinue study treatment if dose interruptions lasted longer than 6 weeks due to ongoing treatment-related toxicities. Patients who discontinued treatment permanently due to severe toxicity were to remain in the trial for ongoing tumour assessments until RECIST 1.1–defined disease progression as assessed by BICR. However, patients were able to continue treatment after a dose interruption of greater than 6 weeks if there was a discussion of the clinical circumstances with the sponsor’s medical monitor.¹⁸

If patients overdosed on study treatment (took > 100 mg of lorlatinib or > 250 mg twice daily of crizotinib), they were required to report this to the sponsor regardless of whether AEs were observed.¹⁸

Concomitant Treatments

Concomitant medications were permitted at any time during the trial if they were considered necessary for the patient’s well-being. Concomitant medications solely for supportive care (e.g., antiemetics, analgesics, megestrol acetate for anorexia, bisphosphonates, or RANK ligands for metastatic bone disease or osteoporosis) were permitted. Patients who were already on treatment with RANK ligands (i.e., denosumab) before study entry were required to be at a stable dose before randomization. No treatments were prohibited during the post-treatment follow-up phase. Other anticancer treatments, investigational therapies, radiation therapy (except for palliative radiotherapy to specific sites of disease), and herbal remedies with known anticancer properties were not permitted.¹⁸

The primary enzymes involved in the metabolism of lorlatinib include CYP3A and UGT1A4, along with involvement of other CYP enzymes in the metabolism, inhibition, and induction of lorlatinib (i.e., CYP2C19, CYP2C8, CYP3A, and CYP2B6). As inhibition or induction of these enzymes may alter the systemic exposure of lorlatinib to patients, safety precautions for certain concomitant medications were pre-specified.¹⁸

The CYP3A enzyme is predominantly involved in the metabolism of crizotinib. Concomitant medications that are CYP3A inhibitors or inducers may be expected to alter the plasma concentrations of crizotinib in patients; therefore, concurrent use of potent CYP3A inhibitors or inducers were not permitted in the trial. However, topical use of CYP3A inhibitors were permitted. Other CYP isoforms (i.e., CYP1A2, 2C8, 2C9, 2C19, and 2D6) may also be inhibited by crizotinib. In addition, treatments that operate through CYP3A4 were also limited during the trial.¹⁸

Outcomes

A list of the efficacy end points identified in the CADTH review protocol that were assessed in the clinical trials included in this review is provided in Table 9. These end points are further summarized in the following section. A detailed discussion and critical appraisal of the HRQoL outcome measures are provided in Appendix 3.

Safety Assessments

Adverse events were graded using the Common Terminology Criteria for Adverse Events Version 4.03. Safety end points were summarized while patients were on treatment (unless otherwise specified).¹⁸

Statistical Analysis

Sample Size

The statistical analysis plan specified that approximately 280 patients, 140 in each treatment group, were to be randomized in the CROWN trial. As of February 28, 2019, a total of 196 patients were randomized into the study. The primary objective of the study was to determine whether lorlatinib was superior to crizotinib in prolonging PFS based on RECIST 1.1 criteria as determined by a BICR. The trial was required to have 177 PFS events to have at least 90% power to detect an HR of 0.611 using a 1-sided stratified log-rank significance level of 0.025 and a 2-look group sequential design with a Lan-DeMets (O’Brien-Fleming) alpha spending function to determine the efficacy boundaries.¹⁸ Assuming an HR of 0.611 would indicate the superiority of lorlatinib, the median PFS was estimated to be 18 months in the lorlatinib group and 11 months in the crizotinib group. A 15% dropout rate was assumed within each treatment group at 30 months, along with a nonuniform patient accrual over a period of approximately 15 months with a follow-up period of approximately 18 months after randomization of the last patient.¹⁸

The expected sample size was planned to allow for a comparison of OS dependent on the observed statistical significance of lorlatinib over crizotinib for PFS.¹⁸ A total of 198 deaths were required to have 70% power using a 1-sided stratified log-rank test at a 1-sided significance level of 0.025, and a 3-look group sequential design with Lan-DeMets (O’Brien-Fleming) alpha spending function to determine the efficacy boundaries. The 198 OS events were expected to result in an HR of 0.70, which would support improved efficacy of lorlatinib over crizotinib and result in an expected median OS of 68.6 months and 48 months in the lorlatinib and crizotinib groups, respectively. A 15% dropout rate was assumed with each treatment group for OS at 120 months, with a follow-up of approximately 110 months after randomization of the last patient.¹⁸

Interim and Final Analyses

Interim and final analyses for PFS were pre-specified in the statistical analysis plan of the CROWN trial. The interim analysis was to be performed after approximately 133 PFS events (75% of 177 events) had occurred per BICR and was to be conducted by an independent statistician. A final analysis of PFS was specified only if the boundary for efficacy was not crossed at the interim analysis. The final analysis of PFS was specified to occur when 177 PFS events occurred according to a BICR, and to be performed by the sponsor’s clinical team.¹⁸

The nominal significance levels for the interim and final efficacy analyses for PFS were determined using the Lan-DeMets procedure with an O’Brien-Fleming stopping boundary; the overall significance level was preserved at 0.025 (1-sided). The superiority of lorlatinib over crizotinib was considered to have been met if the value of the test statistic exceeded the efficacy boundary (z < −2.337; P < 0.01) at the interim analysis (Table 10).¹⁸ However, as the number of observed events may not necessarily be the same as the number of expected events, the efficacy boundaries were specified to be updated based on the observed number of PFS events at the time of the interim analysis using pre-specified alpha spending functions. At the data cut-off date of March 20, 2020, the interim analysis was performed after 127 PFS events had occurred (72% of the 177 events planned for the final analysis of PFS). The efficacy boundaries were updated to reflect the number of PFS events that occurred, and are provided in Table 11; a P value of 0.0081 was used to determine statistical significance at the time of the interim analysis for PFS.¹⁷

As the secondary end point of the trial, OS was specified to be analyzed using a hierarchical testing procedure dependent on statistical significance of PFS favouring lorlatinib.¹⁸ A maximum of 3 analyses for OS were planned:

• at the time of the interim/final analysis for PFS (where PFS exceeded the efficacy boundary)

• at 70% of the 198 expected OS events

• at 100% of the 198 expected OS events (the final analysis).

An alpha spending function according to Lan-DeMets (O’Brien-Fleming) independent of the function used for the primary efficacy analysis was used to preserve the 0.025 overall level of significance across the repeated testing of the OS in the interim and final analyses. The nominal P values used for declaration of statistical significance at the time of the analysis of OS were dependent on the number of OS events occurring at the time of the analyses and the alpha spent for OS at the time of the earlier analyses. Early stopping of the trial was permitted for a superior OS result, such that PFS, the primary end point, was shown to be statistically significantly favouring lorlatinib. The interim analyses for OS were to be performed by an independent statistician.¹⁸ At the time of data cut-off (March 20, 2020), 51 OS events (26% of the 198 events planned at the end of the study) had occurred, and efficacy boundaries for OS were updated based on Lan-DeMets (O’Brien-Fleming), which included a P value of < 0.0001 (Table 11).¹⁷

Analysis Populations

The full analysis set (FAS) included all patients who were randomized to the CROWN trial. Patients were analyzed according to the treatment group to which they were randomized. This analysis set was used for evaluate efficacy end points and patient characteristics.¹⁸ The FAS was considered equivalent to the intention-to-treat population.

The safety analysis set included all patients who received at least 1 dose of study treatment. Patients were classified according to the treatment they actually received.¹⁸

The patient-reported analysis set included patients in the FAS who completed a patient-reported outcome (PRO) assessment at baseline and at least 1 post-baseline assessment. This analysis set was used for the analyses of change from baseline scores and TTD in patient-reported pain in the chest, dyspnea, and cough in the EORTC QLQ-LC13. The analysis sets were separate for the EORTC QLQ-C30, EORTC QLQ-LC13, and EQ-5D-5L descriptive system and EQ VAS questionnaires.¹⁸

Primary End Point of Progression-Free Survival According to a BICR

The primary objective of the study was to determine whether lorlatinib was superior to crizotinib in prolonging PFS based on RECIST 1.1 criteria as determined by a BICR.¹⁸

The following protocol was used to censor PFS:

• Patients were censored on the date of their last adequate tumour assessment if they:

  • did not experience an event (progression or death)

  • started a new anticancer treatment before experiencing a PFS event (progression or death)

  • experienced a PFS event after 2 or more missed tumour assessments.

• Patients were censored on the day of randomization if they:

  • did not have a baseline tumour assessment

  • did not have any post-baseline tumour assessments.

The analyses of PFS were conducted using the FAS and a BICR with a 1-sided log-rank test stratified by the factors used for randomization (i.e., presence of brain metastases and ethnic origin). The significance of the log-rank test was preserved at a 1-sided 0.025 alpha at the interim and/or final analyses.¹⁸ The Kaplan-Meier method was used to estimate the PFS times of each treatment group, with corresponding 95% CIs calculated according to Brookmeyer and Crowley³⁰ and results displayed graphically where appropriate. The CIs for the 25th, 50th and 75th percentiles were reported. A Cox proportional hazards model was used to determine HRs and corresponding 95% CIs. The log(-log) method according to Kalbfleisch and Prentice³¹ was used to estimate the CI for the survival function. The standard error (SE) was estimated using the Greenwood formula.¹⁸

Sensitivity Analyses for Progression-Free Survival

Sensitivity analyses were performed for the primary outcome of PFS per BICR and were considered exploratory.¹⁸ These analyses were conducted in the same manner as the primary analysis of this outcome and included:

• counting all events of disease progression and deaths as PFS events regardless of missing assessments or timing of the event (i.e., not censoring due to start of new anticancer therapy before event or due to missed assessments)¹⁸

• an unstratified analysis¹⁸

• a stratified analysis using the 2 randomization stratification factors and baseline ECOG PS value from the case report forms¹⁸

• multivariable Cox proportional hazards models to explore the potential influences of baseline patient characteristics (e.g., age, gender, ethnic origin, presence of brain metastasis at baseline per BICR intracranial assessment, smoking status, ECOG PS, extent of disease, and histology) on PFS.¹⁸

Secondary End Points

All secondary end points were analyzed using the FAS. Unless otherwise specified, tumour assessments for secondary end points were conducted by a BICR.¹⁸

Progression-Free Survival According to the Investigator

Analysis of PFS by investigator assessment was conducted using the same methods for PFS per BICR.¹⁸

Overall Survival

A 1-sided stratified log-rank test was used for the analysis of OS. A Cox proportional hazards model stratified for the presence of brain metastases and ethnic origin was used to estimate the HR for OS, and the corresponding 95% CI was calculated using the log(-log) method according to Kalbfleisch and Prentice.³¹ The Kaplan-Meier method was used to determine the OS of each treatment group, with CIs calculated according to Brookmeyer and Crowley³⁰ and displayed graphically. The CIs for the 25th, 50th, and 75th percentiles were reported. To account for alpha spending, a repeated CI was calculated. The OS rate was estimated at 12, 24, and 36 months, with corresponding 2-sided 95% CIs. The estimate of the SE was computed using the Greenwood formula.¹⁸

Censoring for OS was conducted if:

• patients were alive

• patients withdrew consent

• patients were lost-to-follow-up.

Sensitivity Analyses for Overall Survival

Sensitivity analyses for OS were considered exploratory and conducted in the same manner as the primary analysis for OS and included:

• an unstratified analysis¹⁸

• a stratified analysis using the 2 randomization stratification factors and baseline ECOG PS¹⁸

• multivariable Cox proportional hazards models to explore the potential influences of baseline patient characteristics (e.g., age, gender, race, ethnic origin, presence of brain metastasis at baseline per BICR of intracranial assessment, smoking status, ECOG PS, disease stage, extent of disease, and histology) on OS.¹⁸

Subsequent Therapies

The statistical analysis plan of the CROWN trial pre-specified analyses that would correct for the receipt of subsequent anticancer therapies by patients. It was acknowledged that subsequent anticancer therapies had the potential to affect estimates of OS; although the impact was considered uncertain. Based on the type, frequency, or distribution of subsequent therapies, analyses were pre-specified to correct for effects of subsequent treatments.¹⁸

Objective Response Rate

In the analysis of ORR, patients without a documented CR or PR were considered nonresponders. The ORR was determined by a BICR and investigator assessment. To determine the ORR, the number of patients with a CR or PR was divided by the number of patients randomized to each treatment group and the corresponding exact 2-sided 95% CIs were calculated. The Cochran-Mantel-Haenszel test was used to provide a comparison of ORRs between the treatment groups with the corresponding 2-sided 95% CI.¹⁸

Subgroup Analyses

Subgroup analyses were pre-specified and performed on the following groups for PFS, OS, and ORR¹⁷:

• age (< 65 years versus ≥ 65 years)

• gender (male versus female)

• ethnic origin (Asian versus non-Asian, from an interactive voice response system)

• presence of baseline brain metastasis (yes versus no, from BICR intracranial assessment)

• smoking status (never versus current/former)

• ECOG PS (0 or 1 versus 2)

• extent of disease (locally advanced versus metastatic)

• histology (adenocarcinoma versus non-adenocarcinoma).

Exploratory End Points

Intracranial Objective Response Rate

A modified version of RECIST 1.1 was used to assess intracranial disease. Assessment of IC-ORR was conducted by a BICR. The IC-ORR was summarized in a manner similar to ORR with at least 1 baseline measurable intracranial lesion. Surgery or radiotherapy of extracranial lesions was stated not to affect the determination of IC-ORR.¹⁸

Intracranial Time to Progression

The IC-TTP was assessed by a BICR among the subgroups of patients with and without brain metastases at baseline. The same censoring rules used for PFS were applied to the analysis of IC-TTP, except that patients who died without experiencing disease progression were not censored, as death was not considered an event. Patients were also not censored in cases of surgery or radiotherapy if the surgery or radiotherapy involved an extracranial lesion. A 1-sided stratified log-rank test was conducted to test for treatment differences between treatment groups. The Kaplan-Meier method was used to summarize IC-TTP in each treatment group with corresponding 95% CIs, and the results were displayed graphically where appropriate. The CIs were reported for the 25th, 50th, and 75th percentiles. The Cox proportional hazards model was used to calculate treatment HRs and the corresponding 95% CIs.¹⁸

Probability of a First Event Being a CNS Progression, Non-CNS Progression, or Death

The competing risk approach to estimating cumulative incidence functions by treatment group was used to evaluate the probability of a first event being a CNS progression per BICR intracranial assessment, a non-CNS progression per BICR, or death. The time to first event being a competing event (which included a “CNS progression” or “non-CNS progression” or “death”) was defined as the time from randomization until the date of that specific event. Patients were censored on the date of their last tumour assessment if they did not have any competing events.¹⁸

Duration of Response

The censoring rules for analysis of DOR were the same as described for PFS. The Kaplan-Meier estimates were presented for each treatment group, including summary statistics (i.e., median DOR time, 2-sided 95% CI) depending on the number of patients who achieved a CR or PR and who subsequently had an event in each treatment group. The median DOR and associated CIs were calculated according to Brookmeyer and Crowley.³⁰

Intracranial DOR

The IC-DOR was summarized in the subgroup of patients with at least 1 measurable intracranial lesion. Censoring rules for IC-DOR were the same as those described for PFS, except that surgery or radiotherapy for an extracranial lesion was not considered.¹⁸

Time to Tumour Response

Simple descriptive statistics (i.e., mean, SD, median, range, and interquartile range) were used to summarize TTR.¹⁸

Intracranial Time to Tumour Response

Intracranial TTR was summarized using simple descriptive statistics (i.e., mean, SD, median, range, and interquartile range).¹⁸

Testing of the Proportional Hazard Assumption

Testing of the proportional hazard assumption was conducted to evaluate the validity of the models for time-to-event outcomes (i.e., PFS and OS). Analyses were conducted by plotting Schoenfeld residuals for the stratified Cox proportional regression models to graphically investigate violations of the proportional hazard assumption. Based on this form of analysis, a non-zero slope was considered evidence of departure of the proportional hazard assumption. The proportional hazard assumption was also formally tested using Schoenfel residual test, using a P value of less than 0.05 as evidence of departure of the proportional hazard assumption. Visual examination of the proportional hazard assumption within each stratum was conducted by plotting log(–log[S(t)]) versus log(t), where S(t) was the estimated survival function for PFS at a specified time, t. In cases where there was evidence of large departures from the proportional hazard assumption, then PFS per BICR assessment was also analyzed based on restricted mean survival time differences.^18,32

Patient-Reported Outcomes

Patient-reported outcomes were assessed using the EORTC QLQ-C30 (version 3.0), EORTC QLQ-LC13 (version 3.0), and EQ-5D-5L. The EORTC QLQ-C30, EORTC QLQ-LC13, and EQ-5D-5L were scored according to their respective validation papers and user guides. Higher scores on the EORTC QLQ-C30 represented “better” levels of functioning and/or “worse” level of symptoms. A minimal important difference (MID) of 10 points was used as the threshold for determining improvement or worsening from baseline in the EORTC QLQ-C30. Higher scores on the EORTC QLQ-LC13 represented “worse” level of symptoms. Higher scores in the EQ-5D-5L represented “better” self-rated health. Completion rates were summarized for each questionnaire, as well as summaries of descriptive statistics including change from baseline, mean, SE, and 95% CIs.¹⁸

A TTD analysis was conducted for pain in the chest, dyspnea, and cough individually and as a composite end point. This was defined as the time from randomization to the first time a patient’s score showed an increase of 10 points or more after baseline in any of the 3 symptoms. A stratified 1-sided log-rank test was used to compare the TTD in the 3 symptoms and composite end point between the treatment groups. Estimates of TTD were produced using Kaplan-Meier methods in each treatment group, including medians with associated 95% CIs calculated according to Brookmeyer and Crowley. The Cox proportional hazards model was used to determine treatment effects. Censoring occurred at the last day a patient completed a subscale assessment if they had not deteriorated, and death was not considered an event for the TTD.¹⁸

Results

Patient Disposition

A summary of the disposition of patients in the CROWN trial is provided in Table 12. The number of patients screened for eligibility was not reported, although the sponsor did report that 129 patients were considered screened failures. The reasons for screen failures were not collected.³³ A total of 149 patients were randomized to the lorlatinib group and 142 patients were randomized to the crizotinib group. In the treatment phase, fewer patients in the lorlatinib group discontinued from the study than in the crizotinib group (30.9% versus 78.2%, respectively). The main reasons for discontinuation from the study in the respective treatment groups were disease progression (17.4% versus 58.5%), followed by AEs (6.7% versus 8.5%), withdrawal by the patient (2.7% versus 5.6%), and death (4.0% versus 2.8%). During the long-term follow-up phase, fewer patients in the lorlatinib group discontinued from the trial (18.1%) than in the crizotinib group (32.7%). The main reasons for study discontinuation in both treatment groups were death (15.4% versus 19.0%, respectively) and withdrawal by the patient (2.7% versus 12.2%, respectively). A total of 122 patients (81.9%) in the lorlatinib group remain ongoing in the trial, compared to 99 patients (67.3%) in the crizotinib group.¹⁷

Protocol Deviations

Important protocol deviations were reported among similar proportions of patients in both the lorlatinib and crizotinib groups (43.0 versus 41.5%, respectively). Most important protocol deviations were due to informed consent (17.2%), procedures and tests (12.8%), and investigational product (10.1%). Similar proportions of important protocol deviations were reported across both treatment groups, and no important protocol deviation was reported for greater than 20% of patients.

Protocol Amendments

The original protocol of the CROWN trial, dated October 5, 2016, was amended a total of 4 times. The amendments were mainly to do with clarifications and minor updates. During the fourth amendment of the protocol, dated October 4, 2019, a change was made to the design of the study based on a primary end point of PFS assessed by BICR. The original design of the trial specified a planned futility analysis for PFS to occur when approximately 60% of events had occurred. However, the amendment replaced the futility analysis with an interim analysis to be conducted when approximately 75% of events had occurred. Accordingly, changes were made to the sample size calculations of the trial and the efficacy boundaries at the interim and final analysis of PFS were determined. The stated rationale for the change was based on the observation that PFS events in the CROWN trial were occurring at a rate slower than originally expected, and on the availability of alternative treatment options for patients.¹⁸

Exposure to Study Treatments

Duration and Dose Intensity

The mean duration of treatment was 16.6 months (SD = 8.32) in the lorlatinib group compared to 10.4 months (SD = 6.90) in the crizotinib group (Table 13). The median duration of treatment was not estimable (NE) in the lorlatinib group and 9.6 months (95% CI, 7.6 to 11.1) in the crizotinib group. At the data cut-off date, more patients in the lorlatinib group remained on treatment than in the crizotinib group (69.1% versus 21.8%, respectively).

As the duration of treatment was longer in the lorlatinib group, the dose intensity and cumulative dose intensity were greater in this group compared to the crizotinib group. The relative dose intensity was similar in the treatment groups.

Dose Modifications

A summary of the dose reductions and interruptions that occurred in the trial by treatment group is provided in Table 14. A dose reduction was defined as a non-zero dose that was less than the prior dose. Dose reductions were reported in 41 patients (27.5%) in the lorlatinib group and 32 patients (22.5%) in the crizotinib group. The proportion of patients requiring 1 dose reduction was similar between the lorlatinib (16.1%) and crizotinib (16.2%) groups; and the proportion of patients requiring 2 dose reductions was greater in the lorlatinib group (11.4% versus 6.3%, respectively). All dose reductions were due to AEs in the lorlatinib group, and almost all were due to AEs in the crizotinib group.

Dose interruptions were defined as a temporary discontinuation of dosing for more than 1 day. At least 1 dose interruption occurred in 79 patients (53.0%) in the lorlatinib group and 68 patients (47.9%) in the crizotinib group. The proportions of patients requiring any number of dose interruptions was similar across both treatment groups. Most patients in either treatment group required only 1 or 2 dose interruptions, and the majority of dose interruptions were due to AEs (49.0% in the lorlatinib group versus 44.4% in the crizotinib group). The dose interruptions due to “other” reasons were related to mistakes or oversight from patients and medical procedure protocols that required dose interruptions (6.7% in the lorlatinib group and 11.3% in the crizotinib group).

Concomitant Treatments

Most patients in each treatment group received concomitant medications (96.6% versus 95.1%, respectively); however, the frequency of some concomitant medications differed by treatment group. Rosuvastatin calcium (38.9%), paracetamol (36.9%), furosemide (28.9%), and rosuvastatin (24.8%) were the most commonly reported (≥ 20%) concomitant medications in the lorlatinib group, compared to paracetamol (39.4%) and metoclopramide (21.8%), which were most commonly reported in the crizotinib group.¹⁷

Subsequent Therapies

At the cut-off date, |||||||||||||| in the crizotinib group had received subsequent therapy compared with patients in the lorlatinib group (Table 15). |||||||||||||| in the crizotinib group |||||||| had received at least 1 type of follow-up systemic anticancer treatment compared with patients in the lorlatinib group ||||||||. |||||||||||||| in the crizotinib group received 1 |||||||||||||||||||||||||| or 2 subsequent therapies |||||||||||||||||||||||| compared to the patients in the lorlatinib group and |||||||||||||||||||||| of patients in each treatment group received 3 or more subsequent therapies. Radiotherapy was received by |||| of patients in the crizotinib group versus |||| of patients in the lorlatinib group, and no patients in either treatment group received subsequent anticancer surgery.¹⁷

The first and second subsequent therapy was |||||||||||||||| an ALK TKI compared with chemotherapy with or without an anti-angiogenic therapy, an immunotherapy or other (Table 16). A subsequent TKI was |||||||||||||| in the crizotinib group (|||||| compared with the lorlatinib group ||||||||; alectinib (||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||) and brigatinib (|||||||||||||||||||||| ||||||||||||) were |||||||||||||||| reported as a first subsequent TKI in the crizotinib group compared with the lorlatinib group, whereas crizotinib (||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||) and ceritinib (||||||||||||||||||||||||||||||||||||||||) were ||||||||||| reported as a first subsequent ALK TKI in the lorlatinib group. Chemotherapy with or without an anti-angiogenic therapy was |||||||||||||| in the lorlatinib group (|||||| compared with the crizotinib group (|||||||||||||||||||| in either group received either immunotherapy (||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||) or other (||||||||||||||||||||||||).¹⁷

An ALK TKI as a second subsequent therapy was |||||||||||||| in the lorlatinib group (||||||||||||||||||||||||||||||||||||||||||||||||||). Alectinib (||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||) and brigatinib (||||||||||||||||||) were both |||||||||||||| as a second subsequent TKI in the lorlatinib group compared with the crizotinib group, whereas ceritinib (|||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||) and lorlatinib (||||||||||||||||||) were |||||||||||||| as a second subsequent ALK TKI in the crizotinib group compared with the lorlatinib group. Chemotherapy with or without an anti-angiogenic was |||||||||||||| in the crizotinib group (|||||| compared to the lorlatinib group ||||||)¹⁷

|||||||||||||| received a third subsequent systemic anticancer therapy. A third subsequent ALK TKI was |||||||||||||| in the crizotinib group (||||) compared with the lorlatinib group (|||||| Chemotherapy with or without an anti-angiogenic therapy was reported in |||| |||||||||||||||||| in the lorlatinib group and |||||||||||||||||||||||| in the crizotinib group.¹⁷

Efficacy

At the data cut-off date (March 20, 2020), the CROWN trial met its primary end point and demonstrated a statistically significant difference in PFS that favoured treatment with lorlatinib over crizotinib. The median PFS follow-up duration was 18.3 months (95% CI, 16.4 to 20.1) in the lorlatinib group and 14.8 months (95% CI, 12.8 to 18.4) in the crizotinib group. Based on pre-specified statistical criteria, this was considered the final analysis for PFS, and no formal testing for PFS was planned for future time points.¹⁷ At the time of this analysis, OS, which was hierarchically tested depending on the statistical significance of PFS, did not show any statistically significant difference between the lorlatinib and crizotinib treatment groups.¹⁷ A detailed summary of the results for the primary and secondary end points of the CROWN trial are reported in Table 17. A summary of the results of intracranial efficacy end points, which were assessed as exploratory outcomes, are reported in Table 18.

The sponsor assessed the proportional hazards for PFS by performing tests for statistical significance using log(-log(Survival)) plots and a plot of Schoenfeld residuals from stratified Cox proportional regression models. The sponsor provided the log(-log(Survival)) plots and a plot of Schoenfeld residuals; however, the results of proportional hazards tests were not reported.

Primary End Point: Progression-Free Survival per BICR

A lower proportion of patients experienced a PFS event in the lorlatinib group (27.5%) than in the crizotinib group (58.5%). The majority of the PFS events were progression events (21.5% versus 55.8%, respectively), with deaths representing 6.0% and 2.7%, respectively). A statistically significant improvement in PFS was demonstrated in the lorlatinib group compared to the crizotinib group, corresponding to a 72% reduction in the risk of progression or death (stratified HR = 0.28; 95% CI, 0.191 to 0.413; stratified 1 sided log-rank P < 0.001). The median PFS in the lorlatinib group was NE (95% CI, NE to NE) compared to 9.3 months (95% CI, 7.6 to 11.1) in the crizotinib group (Figure 3).¹⁷

More patients in the lorlatinib group were censored in the analysis of PFS than in the crizotinib group (72.5% versus 41.5%, respectively). The primary reasons for censoring were due to patients continuing without an event (62.4% in the lorlatinib group versus 15.6% in the crizotinib group), beginning a new anticancer therapy (6.7% versus 16.3%, respectively) or withdrawing from the trial (2.0% versus 9.5%, respectively) (Table 19).¹⁷

Figure 3: Kaplan-Meier Plot of PFS (Primary Analysis) per BICR Assessment (RECIST 1.1) — Full Analysis Set

BICR = blinded independent central review; CI = confidence interval; HR = hazard ratio; NE = not estimable; PFS = progression-free survival; RECIST 1.1 = Response Evaluation Criteria in Solid Tumors Version 1.1; vs. = versus.

Source: CROWN Clinical Study Report.¹⁷

Subgroup Analysis

Subgroup analyses are depicted in Figure 4. All subgroup analyses supported the primary analysis of PFS, which favoured treatment with lorlatinib over crizotinib.¹⁷

Figure 4: Forest Plot of PFS per BICR Assessment (RECIST 1.1) by Subgroups — Full Analysis Set

CI = confidence interval; ECOG = Eastern Cooperative Oncology Group; FAS = full analysis set; RECIST 1.1 = Response Evaluation Criteria in Solid Tumors Version 1.1.

Note: Presence of brain metastases subgroup was based on modified RECIST 1.1 BICR baseline data. Hazard ratios were not calculated due to insufficient numbers of events (< 10 events on either treatment group within the defined subset), as dictated by the statistical analysis plan, for patients who had an ECOG Performance Status of 2 (2 versus 8), extent of locally advanced (5 versus 3) or histology of non-adenocarcinoma (5 versus 5). Stratified by presence of brain metastases (yes or no) and ethnic origin (Asian or non-Asian) at randomization. Percentages were calculated based on the number of patients in the FAS in each treatment group. Plot presented on a log scale (base = 2). P values were not adjusted for multiplicity.

Source: CROWN Clinical Study Report.¹⁷

Secondary End Point: Overall Survival

At the time of the interim analysis, the majority of patients in both treatment groups were alive. Of the 198 OS events required for the final analysis of OS, 51 (26%) occurred at the time of the primary analysis (final) of PFS, including 23 (15.4%) in the lorlatinib group and 28 (19.0%) in the crizotinib group (Figure 5). The statistical boundary of efficacy for OS was not crossed. The median OS was NE in either treatment group. Preliminary results for OS demonstrated an HR of 0.72 (95% CI, 0.414 to 1.249), which were not statistically significant.¹⁷

Similar proportions of patients were censored in the lorlatinib (84.6%) and crizotinib (81.0%) groups in the analysis of OS (Table 20). The reasons for censoring were primarily patients remaining alive (81.9% in the lorlatinib group versus 67.3% in the crizotinib group) and withdrawal of consent (2.7% versus 12.2%, respectively).¹⁷

Figure 5: Kaplan-Meier Plot of Overall Survival — Full Analysis Set

CI = confidence interval; HR = hazard ratio; NE = not estimable; OS = overall survival.

Source: CROWN Clinical Study Report.¹⁷

Exploratory End Point: Progression-Free Survival per Investigator Assessment

The results for PFS assessed by investigator were consistent with the results for PFS assessed by a BICR. Results favoured the lorlatinib group over the crizotinib group (HR = 0.21; 95% CI, 0.144 to 0.307). The median PFS was NE in the lorlatinib group (95% CI, NE to NE) and 9.1 months (95% CI, 7.4 to 10.9) in the crizotinib group (Figure 6).¹⁷

More patients in the lorlatinib group were censored (73.2%) than in the crizotinib group (29.3%) (Table 21). Censoring was mainly due to patients continuing without an event (64.4% in the lorlatinib group versus 15.6% in the crizotinib group), withdrawal of consent (2.0% versus 7.5%, respectively), and starting a new anticancer therapy (3.4% versus 5.4%, respectively).¹⁷

Figure 6: Kaplan-Meier Plot of PFS Based on Derived Investigator Assessment (RECIST 1.1) — Full Analysis Set

CI = confidence interval; HR = hazard ratio; NE = not estimable; PFS = progression-free survival; RECIST = Response Evaluation Criteria in Solid Tumors Version 1.1.

Source: CROWN Clinical Study Report.¹⁷

Exploratory End Point: ORR per BICR

The ORR was numerically higher in the lorlatinib group at 75.8% (95% CI, 68.2 to 82.5) compared to 57.8% (95% CI, 49.4 to 65.9) in the crizotinib group. Similarly, more patients in the lorlatinib group than in the crizotinib group achieved a CR (2.7% versus 0, respectively) or PR (73.2% versus 57.8%, respectively).¹⁷

Exploratory End Point: DOR per BICR

The median DOR was NE (95% CI, NE to NE) in the lorlatinib group, and 11.0 months (95% CI, 9.0 to 12.9) in the crizotinib group.¹⁷ There was a large amount of censoring in the analysis of DOR; 84.1% of patients in the lorlatinib group and 47.1% in the crizotinib group were censored, mainly due to patients continuing without an event (75.2% versus 22.4%, respectively), or starting a new anticancer therapy (5.2% versus 18.8%, respectively). Fewer patients in the lorlatinib group (15.9%) had an event compared with the crizotinib group (52.9%), and most were events of progressive disease (12.4% versus 52.9%, respectively).

Exploratory End Point: Time to Tumour Response per BICR (RECIST 1.1)

Among patients with a confirmed objective response (CR or PR) per BICR assessment, the mean TTR was similar in the treatment groups, at 2.13 months in the lorlatinib group (SD = 1.370) and 2.12 months (SD = 1.252) in the crizotinib group.¹⁷

Exploratory End Point: Intracranial Time to Progression per BICR (Modified RECSIT 1.1)

The time to CNS progression was numerically longer in patients treated in the lorlatinib group than the crizotinib group (Figure 7). The median IC-TTP was NE (95% CI, NE to NE) in the lorlatinib group and 16.6 months (95% CI, 11.1 to NE) in the crizotinib group (HR = 0.07; 95% CI, 0.026 to 0.170; nominal 1-sided P < 0.0001).¹⁷

In the subset of patients with brain metastases at baseline, fewer patients in the lorlatinib group than in the crizotinib group experienced an event (10.5% versus 65.0%, respectively). The median IC-TTP per BICR was NE in the lorlatinib group (95% CI, NE to NE) and 7.3 months (95% CI, 3.7 to 9.3) in the crizotinib group (Table 22). The stratified HR was 0.08 (95% CI, 0.026 to 0.227) (Figure 8).¹⁷

In the subset of patients without brain metastases at baseline, 1 patient (0.9%) in the lorlatinib group experienced an event compared to 19 patients (17.8%) in the crizotinib group (Table 23). The median IC-TTP among patients without brain metastases at baseline was NE in either treatment group. The stratified HR favoured the lorlatinib group over the crizotinib group (HR = 0.03; 95% CI, 0.004 to 0.230) (Figure 9).¹⁷

Figure 7: Kaplan-Meier Plot of IC-TTP Based on BICR Assessment (Modified RECIST 1.1) — Full Analysis Set

BICR = blinded independent central review; CI = confidence interval; HR = hazard ratio; IC-TPP = intracranial time to progression; NE = not estimable; RECIST 1.1 = Response Evaluation Criteria in Solid Tumors Version 1.1.

Source: CROWN Clinical Study Report.¹⁷

Figure 8: [Redacted]

Note: This figure has been redacted at the sponsor’s request.

Source: CROWN Clinical Study Report.¹⁷

Figure 9: [Redacted]

Note: This figure has been redacted at the sponsor’s request.

Source: CROWN Clinical Study Report.¹⁷

Alumni Exploratory End Point: Probability of First Event Being a CNS Progression, Non-CNS Progression, or Death per BICR (RECIST 1.1 and Modified RECIST 1.1)

Time to CNS progression and time to development of CNS metastases were end points specified in the CADTH protocol as being of interest to patients and clinicians, but these end points were not reported in the CROWN trial. However, the CROWN trial did report results for the probability that a patient’s first event was a CNS progression, non-CNS progression, or death, and data on these end points were considered clinically important by the clinical experts consulted by CADTH. The probability of a first event being a CNS progression after adjusting for non-CNS progression and death was lower (i.e., favoured) in the lorlatinib group compared with the crizotinib group (HR = 0.06; 95% CI, 0.022 to 0.182). The probability of a first event being a non-CNS progression after adjusting for CNS progression or death was also lower in the lorlatinib group than in the crizotinib group (HR = 0.30; 95% CI, 0.185 to 0.474). There was no difference in the probability of a first event being death after adjusting for prior CNS or non-CNS progression between the treatment groups (Table 24).¹⁷

Exploratory End Point: Intracranial ORR per BICR (Modified RECIST 1.1)

Patients With Any (Measurable or Nonmeasurable) Brain Metastases at Baseline

A total of 38 patients in the lorlatinib group and 40 patients in the crizotinib group had any measurable or nonmeasurable brain metastasis. Among these patients, the IC-ORR was numerically higher in the lorlatinib group (65.8%; 95% CI, 48.6 to 80.4) than in the crizotinib group (20.0%; 95% CI, 9.1 to 35.6) (Table 25). More patients in the lorlatinib group had an intracranial CR than patients in the crizotinib group (60.5% versus 15.0%, respectively).¹⁷

Patients With at Least 1 Measurable Brain Metastasis at Baseline

A total of 17 patients in the lorlatinib group and 13 patients in the crizotinib groups had at least 1 measurable brain metastasis (Table 26). Among these patients, the IC-ORR was numerically higher in the lorlatinib group (82.4%; 95% CI, 56.6 to 96.2) than in the crizotinib group (23.1%; 95% CI, 5.0 to 53.8). The proportion of patients achieving an intracranial complete response (IC-CR) was higher in the lorlatinib group than the crizotinib group (70.6% versus 7.7%, respectively).¹⁷

Exploratory End Point: Intracranial DOR per BICR (Modified RECIST 1.1)

Among patients with any measurable or nonmeasurable brain metastases at baseline who had a confirmed CR or PR, the median IC-DOR was NE (95% CI, NE to NE) in the lorlatinib group and 9.4 months (95% CI, 6.0 to 11.1) in the crizotinib group (Table 27).¹⁷

Among patients with at least 1 measurable brain metastasis who had confirmed CR or PR, the median IC-DOR was NE (95% CI, NE to NE) in the lorlatinib group and 10.2 months (95% CI, 9.4 to 11.1) in the crizotinib group (Table 28).¹⁷

Exploratory End Point: Intracranial TTR per BICR (Modified RECIST 1.1)

Among patients with any baseline brain metastasis who had a confirmed CR or PR, the mean IC-TTR was numerically longer in the lorlatinib group at 3.04 months (SD = 2.800) than in the crizotinib group at 2.21 months (SD = 0.844). Among patients with at least 1 measurable brain metastasis at baseline, the mean IC-TTR was numerically longer in the lorlatinib group at 2.39 months (SD = 0.886) compared to the crizotinib group at 1.86 months (SD = 0.083) (Table 29).¹⁷

Patient-Reported Outcomes

Data for PROs are presented in this report for the first 18 cycles of treatment because the number of patients completing PRO assessments in each treatment group was smaller during later treatment cycles (≤ 20%). Patient attrition was due primarily to patients experiencing disease progression. Due to the small sample sizes in each treatment group at later cycles, the sponsor expressed concern regarding the interpretability of the data.¹⁷

Completion Rates

More than 90% of patients completed at least 1 question of the EORTC QLQ-C30 and QLQ-LC13 questionnaires at each assessment time point. More than 88% of patients completed the entire questionnaire (all questions answered) until the data cut-off date. After cycle 19, less than 50% of patients were eligible for completion of the questionnaires in the lorlatinib group, resulting in small sample sizes; in the crizotinib group, less than 50% of the sample were eligible for completion of the questionnaires after cycle 11. Compliance in the EQ-5D-5L questionnaire was also high, with more than 90% of patients completing at least 1 question of the or the entire questionnaire until cycle 38. In the lorlatinib group, less than 50% of patients were eligible for completion of the survey after cycle 19. In the crizotinib group, less than 50% of patients were eligible for completion of the survey after cycle 11.¹⁷

EORTC QLQ-C30 Functional and Symptoms Scales

At baseline, the mean scores for the global quality-of-life scale were 64.6 (SE ± 1.82) in the lorlatinib group and 59.8 (SE ± 1.90) in the crizotinib group. Numerically, there was a greater overall improvement in global quality of life from baseline up until the data cut-off date; the estimated mean global quality-of-life score was 70.79 (95% CI, 63.57 to 78.01) in the lorlatinib group and 66.79 (95% CI, 60.36 to 73.22) in the crizotinib group; and the estimated mean difference between the lorlatinib and crizotinib groups was 4.65 (95% CI, 1.14 to 8.16; P = 0.0096) (Figure 10). This between-group difference did not meet the pre-specified threshold for the MID of at least 10 points. A similar proportion of patients in the lorlatinib and crizotinib groups achieved an improvement of 10 points or more (41.8% versus 42.6%, respectively) or a stable score (39.7% versus 38.1%, respectively) in the global quality-of-life subscale.

The mean changes from baseline scores were numerically higher in the lorlatinib group compared to the crizotinib group for the QLQ-C30 functioning subscales (physical functioning, role functioning, emotional functioning, social functioning), except for the cognitive functioning scale, for which the change from baseline scores were numerically higher in the crizotinib group over the lorlatinib group.¹⁷ For the symptoms sales (fatigue, pain, nausea and vomiting, dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties), the mean changes from baseline scores were numerically lower in the lorlatinib group for the fatigue, nausea and vomiting, insomnia, appetite loss, constipation, diarrhea, and financial difficulties subscales, except for the pain and dyspnea subscales, which were numerically higher in the lorlatinib group compared to the crizotinib group. For all functioning and symptom domain subscales of the EORTC QLQ-C30, none of the between-group differences met the MID (≥ 10 points) threshold.¹⁷

Figure 10: Plot of Mean Change From Baseline (± SE) Over Time for EORTC QLQ-C30

EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; SE = standard error.

Note: Baseline was defined as the last assessment performed on or before date of the first dose of study treatment. Based on EORTC QLQ-C30 PRO analysis set within each treatment group. Mean change from baseline is shown through cycle 18 to correspond with the median follow-up time

Source: CROWN Clinical Study Report.¹⁷

EORTC QLQ-LC13 Symptom Scales and Time to Deterioration

Time to deterioration in EORTC QLQ-LC13 symptom scales was evaluated for individual symptoms, including pain in the chest, dyspnea, and cough, and as a composite end point of these 3 symptoms. Pain in the chest, dyspnea, and cough were chosen as they are commonly reported disease-related symptoms experienced by patients with lung cancer. The mean scores for pain in the chest, dyspnea, and cough at baseline were similar between the lorlatinib and crizotinib groups. The TTD in the composite end point of lung cancer symptoms (cough, dyspnea, or pain in the chest) was not different between the 2 treatment groups; the median TTD was 3.3 months (95 CI, 2.1 to 4.7) in the lorlatinib group and 3.7 months (95% CI, 0.211 to 0.382) in the crizotinib group (HR = 1.09, 95% CI, 0.822 to 1.444; 1-sided P value = 0.7293) (Figure 11).¹⁷ Similarly, the TTD in individual symptoms was not different between the 2 treatment groups for the pain in the chest, dyspnea, and cough scales.¹⁷

Figure 11: Kaplan-Meier Plot of Time to Deterioration in Composite of Pain in the Chest, Dyspnea, and Cough

CI = confidence interval; HR = hazard ratio.

Source: CROWN Clinical Study Report.¹⁷

EQ-5D VAS and Index Values

The mean scores of patients in the lorlatinib and crizotinib groups |||||||||||| for the EQ-5D-5L and EQ VAS scores (|||||||||||||||||||||||||||||||||) and index values (mean = |||||||||||||||| versus ||||||||||||||||) at baseline and throughout the trial until cycle 38. The proportion of patients reporting slight, moderate, severe, or extreme problems in the functional dimensions of the questionnaire (mobility, self-care, usual activities, pain or discomfort, anxiety or depression) was ||| in both treatment groups at baseline; the proportion of patients reporting slight, moderate, severe, or extreme problems |||||||||||||||||||||||||| in later cycles from baseline.¹⁷

Harms

Only those harms identified in the CADTH review protocol are reported below. A summary of harms data is provided in Table 30. Further details regarding AEs are provided in the text that follows.

Adverse Events

All patients in the lorlatinib group (100%) and almost all patients in the crizotinib group (98.6%) experienced an AE, most of which were grade 1 or 2. The most common AEs in the lorlatinib group were hypercholesterolemia (70.5%), hypertriglyceridemia (63.8%), edema (55.0%), increased weight (38.3%), peripheral neuropathy (33.6%), cognitive effects and diarrhea (21.5% each), and dyspnea (20.1%); all these AEs occurred more frequently in the lorlatinib group, except for diarrhea, which was more common in the crizotinib group. The most common AEs in the crizotinib group were nausea (52.1%), diarrhea (52.1%), edema (39.4%), vision disorder (39.4%), vomiting (38.7%), increased ALT (33.8%), constipation (29.6%), increased AST (27.5%), and decreased appetite (24.6%); all these AEs occurred more frequently in the crizotinib group, except for edema, which occurred more frequently in the lorlatinib group.¹⁷

Approximately three-quarters of the AEs in the lorlatinib group were grade 3 or 4 (72.5%) compared to just over half of AEs in the crizotinib group (55.7%). The grade 3 or 4 AEs that occurred more frequently in the lorlatinib group were increased weight (16.8%), hypercholesterolemia (15.4%), and hypertriglyceridemia (12.8%). The most common grade 3 AEs in the crizotinib group included neutropenia (8.5%), increased gamma-glutamyl transferase, decreased neutrophil count, hypoalbuminemia (4.2% each), increased ALT, increased AST, increased blood creatine phosphokinase, and increased lipase (3.5% each). Neutropenia was the only grade 3 event to occur more frequently in the crizotinib group than in the lorlatinib group (8.5% versus 0.7%, respectively). Grade 4 AEs were generally infrequent in both treatment groups; hypertriglyceridemia (7.4%) was the most common grade 4 AE in the lorlatinib group, and decreased neutrophil count (2.8%) was the most common grade 4 event in the crizotinib group (Table 31).

Serious Adverse Events

In general, SAEs were more commonly reported in the lorlatinib group than in the crizotinib group (34.2% versus 27.5%, respectively) although the majority of SAEs occurred in 1 patient. The most common SAEs of any grade in the lorlatinib group were pneumonia (4.7%), dyspnea, respiratory failure (2.7% each), cognitive effects and pyrexia (2.0% each). In the crizotinib group, the most common SAEs of any grade were pneumonia (3.5%) and pyrexia (2.1%). Grade 3 or 4 SAEs occurred more frequently in the lorlatinib group than in the crizotinib group (22.8% versus 17.6%, respectively). The most common grade 3 SAEs in the lorlatinib group were dyspnea (2.7%) and respiratory failure (2.0%); and the most common grade 3 SAE in the crizotinib group was pneumonia (2.1%). All grade 4 SAEs occurred at a low frequency of less than 2% of patients.¹⁷

Dose Modifications

A summary of dose interruptions and modifications is provided in the Exposure to Study Treatments section of this report. Further details regarding AEs resulting in dose reductions and interruptions are provided in Table 33. Adverse events that resulted in dose reductions occurred in 31 patients (20.8%) in the lorlatinib group and 22 patients (15.5%) in the crizotinib group. The most commonly occurring AEs that resulted in a dose reduction in the lorlatinib group were edema (5.4%), hypertriglyceridemia (4.0%), and peripheral neuropathy (3.4%). The most commonly occurring AEs resulting in a dose reduction in the crizotinib group were edema and decreased neutrophil count (3.5% each). Peripheral neuropathy was the most common grade 3 AEs resulting in dose reductions in the lorlatinib group, occurring in 1.3% of patients; in the crizotinib group the most common grade 3 AE was decreased neutrophil count (2.1%). No grade 4 AEs resulted in dose reductions in either treatment group.¹⁷

Dose interruptions resulting from AEs occurred in a similar proportion of patients in the lorlatinib (49.0%) and crizotinib (47.2%) treatment groups, of which 32.9% and 36.6% were attributable to grade 3 or 4 AEs. The most common AEs of any grade that resulted in a dose interruption in the lorlatinib group included hypertriglyceridemia (7.4%) edema (5.4%), pneumonia (4.7%), and cognitive effects (4.0%). In the crizotinib group, the most common AEs of any grade resulting in dose interruption were due to decreased neutrophil count (7.0%), increased ALT (4.3%), and pneumonia (3.5). Grade 3 AEs resulting in dose interruption occurred in less than 3% of patients in the lorlatinib group, and the most common grade 4 AE was hypertriglyceridemia, which occurred in 5.4% of patients. In the crizotinib group, the grade 3 AEs that resulted in dose interruption were neutropenia (7.7%) and decreased neutrophil count decrease (4.2%). Grade 4 AEs resulted in a dose interruption in less than 3% of patients due to decreased neutrophil count (2.8%) and pneumonia (0.7%).¹⁷

Withdrawals Due to Adverse Events

Adverse events leading to permanent treatment discontinuation occurred in 10 patients (6.7%) in the lorlatinib group and 13 patients (9.2%) in the crizotinib group. Grade 3 AEs resulting in permanent treatment discontinuation occurred in 6 patients (4.0%) and 7 patients (4.9%) in the lorlatinib and crizotinib groups, respectively. Grade 4 AEs resulting in permanence discontinuation from study treatment occurred in 2 patients (1.3%) in the lorlatinib group and 1 patient (0.7%) in the crizotinib group. The most common AE that resulted in permanent discontinuation of lorlatinib was cognitive effects (N = 2; 1.3%). The most common AEs resulting in permanent discontinuation of crizotinib were pleural effusion (N = 3; 2.1%) and pneumonitis (N = 2; 1.4%).¹⁷

Mortality

Deaths occurred in similar proportions of patients in the lorlatinib and crizotinib groups: 23 patients (15.4%) and 18 patients (19.7%), respectively. Most deaths were attributed to disease progression, which occurred in 17 patients (11.4%) in the lorlatinib group and 23 patients (16.2%) in the crizotinib group. Two deaths in the lorlatinib group were considered by the investigator to be treatment-related; both of these patients had a history of prior heart disease and diabetes. One patient died of cardiac failure approximately 2 months after receiving their last dose of lorlatinib, and the second patient died of respiratory failure in the context of infectious pneumonia.¹⁷ The cause of death of 3 patients, including 1 in the lorlatinib group and 2 in the crizotinib group, was categorized as unknown. The cause of 2 deaths, including 1 patient in each treatment group, were categorized as other; 1 of these patients experienced worsening of decompensatio cordis, and the other patient experienced sepsis.

Three deaths in the lorlatinib group were related to SAEs, but they were not considered by the sponsor in the overall count of grade 5 AEs as they occurred after the patients began new anticancer therapies.¹⁷

Notable Harms

A list of AEs of special interest in the CROWN trial are reported in Table 35. The AEs of special interest in the CROWN trial aligned with notable harms listed in the CADTH systematic review protocol. In general, these AEs were more common in the lorlatinib group, except for vision disorder and increased liver function test, which were more common in the crizotinib group than in the lorlatinib group (39.4% versus 18.1% and 37.3% versus 20.8%, respectively). The most common AEs of special interest were hypercholesterolemia (70.5% in the lorlatinib group versus 3.5% in the crizotinib group), hypertriglyceridemia (63.8% versus 5.6%, respectively), edema (55.0% versus 39.4%, respectively), weight gain (38.3% versus 12.7%, respectively), peripheral neuropathy (33.6% versus 14.8%, respectively), cognitive effects (21.5% versus 5.6%, respectively), increased liver function test (20.8% versus 37.3%), and mood effects 16.1% versus 4.9%).¹⁷

Critical Appraisal

Internal Validity

The CROWN trial was a multinational, multi-centre, open-label, group sequential superiority trial. Open-label trials by design are associated with biases that may affect patient outcomes and reporting, and favour investigational products over standard of care, as patients and investigators are aware of the assigned treatment of patients. However, the choice of an open-label design was considered appropriate given that crizotinib has been a standard treatment for patients with metastatic ALK-positive NSCLC, and clinicians are likely knowledgeable of toxicity profiles. The primary end point of the CROWN trial, PFS, may have been affected by bias from an open-label trial design. However, BICR was implemented for the assessment of PFS, as well as other end points that involved judgment of patient clinical progression (i.e., ORR and DOR), which mitigated the potential for bias introduced through the trial design. The risk of bias from an open-label design is likely to be of greater concern for subjective end points, such as HRQoL and safety, as the knowledge of treatment assignment from patients and investigators could have influenced the reporting and assessment of these outcomes.

Patients were randomized to the CROWN trial in a 1:1 ratio using an interactive web-based response technology system. Randomization was stratified by the presence of brain metastases and ethnic origin.¹⁷ This method of randomization was considered to result in a low risk of bias, as relatively equal proportions of patients were randomly assigned to receive either lorlatinib or crizotinib. The clinical experts consulted by CADTH indicated that presence of brain metastases is a particularly important factor for consideration in this patient population, and was accounted for in the randomization scheme to avoid an imbalance of this disease characteristic between the treatment groups. Overall, baseline demographic and clinical characteristics of patients were well balanced across the groups, with some exceptions (age and smoking status). The clinical experts suggested that it is unlikely that these slight imbalances affected patient outcomes.

The primary and secondary end points of the CROWN trial were PFS and OS. Both end points were considered in power calculations, and OS was tested hierarchically at the time of the data cut-off depending on the statistical significance of PFS. Other secondary and exploratory end points were not included in the statistical hierarchy. The statistically significant findings on subgroup analyses, such as in patients with brain metastasis were likely subject to multiplicity and inflated type I error rate. At the data cut-off date, the interim analysis for PFS had crossed the pre-specified efficacy boundary and demonstrated a statistically significant difference in favour of lorlatinib in prolonged PFS, and the analysis was therefore considered final by the sponsor. However, OS data were deemed immature, as only 26% of the 198 OS events required for the final analysis of OS had occurred.¹⁷ However, improvements in PFS may not always correlate to a difference in OS in the assessment of oncology treatment benefit. In fact, mainly due to disease progression, a higher proportion of patients in the crizotinib group than in the lorlatinib group (17.4% versus 58.5%, respectively) withdrew from the study. This would largely bias the estimate of OS in the final analysis. Further evidence is therefore required to confirm the superiority of lorlatinib over crizotinib in treatment efficacy with respect to OS. The next analysis for OS, occurring once 70% of events have occurred, was projected to occur in early 2025, assuming that the number of OS events expected to occur for this analysis happen by the end of 2024.³³

At the data cut-off date, |||||||||||||| in the crizotinib group compared with the lorlatinib group had received a subsequent systemic anticancer therapy. There were disproportional differences in the frequency and type of subsequent systemic anticancer therapies administered to patients. For example, while an ALK TKI was the |||||||||||||||| administered subsequent therapy received by patients in both the lorlatinib and crizotinib groups (||||||||||||||||||||||||||||||||||||||||||), |||||||||||||| in the lorlatinib group received subsequent treatment with chemotherapy with or without an anti-angiogenic therapy compared with patients in the crizotinib group (||||||||||||||||||||||||||||||||||||||||).³³ These differences in subsequent therapies are expected to introduce bias into the efficacy analyses of OS and other patient outcomes. However, the direction and extent of this bias is difficult to predict.

Treatment crossover was not permitted within the CROWN trial. However, at the data cut-off date, |||||||||||||| in the lorlatinib group discontinued treatment and received crizotinib as a first subsequent therapy and |||||||||||||||| received crizotinib as a second subsequent therapy; |||||||||||||||| discontinued in the crizotinib group and received lorlatinib as a first subsequent therapy, |||||||||||| received lorlatinib as a second subsequent therapy, and |||||||||||||| received it as a third subsequent therapy.^17,33 Treatment crossover may introduce bias into efficacy analyses as effects of previous treatment may carry over into the next line of therapy. However, as there were |||||||||||||| in both treatment groups who received subsequent treatment with the assigned therapy of the opposite treatment group, it is unlikely that biases related to crossover affected efficacy results of the trial.

Most patients in each treatment group received concomitant medications (96.6% in the lorlatinib group versus 95.1% in the crizotinib group). although the frequencies of some types of concomitant medications received by patients varied across treatment groups. The clinical experts consulted by CADTH for this review indicated that the differences in concomitant medications may be indicative of real-world practices. For example, the greater use of statins in the lorlatinib group than in the crizotinib group may be indicative of AEs related to cholesterol in the lorlatinib group. While differences in the types of concomitant medications were observed between the lorlatinib and crizotinib groups, these differences were considered unlikely to affect efficacy outcomes.

Statistical analyses of OS and PFS were conducted using Cox proportional hazards models, which rely on the assumption of proportional hazards in both treatment groups. The sponsor assessed proportional hazards for PFS by performing tests for statistical significance using log(-log(Survival)) plots and a plot of Schoenfeld residuals from stratified Cox proportional regression models. However, the sponsor did not conduct a formal assessment of the proportional hazards assumption and did not provide a conclusion regarding its violation. Based on the plots provided, it is possible that there was a potential for violation of the assumption for the analysis of PFS. An analysis of proportional hazards for OS was not reported, and it is not known whether the analysis of OS violated the proportional hazards assumption; the data for OS were immature.

In general, due to patients continuing in the trial, the analyses of efficacy end points involved censoring of many patients. While data for the primary end point, PFS, was considered mature, large amounts of censoring may indicate that longer-term data are required to capture the true benefit of lorlatinib compared to crizotinib. Particularly, longer-term data would benefit analyses of OS as the median had not been reached in either treatment group. Results of exploratory end points (i.e., PFS according to investigator assessment, ORR, DOR, and IC-TTP) at the data cut-off date favoured treatment with lorlatinib over crizotinib; however, as the trial was not powered to formally test these outcomes, the obtained results should be considered as supportive evidence only.

Data on HRQoL were captured using the EORTC QLQ-C30, EORTC QLQ-LC13, and EQ-5D-5L. The EORTC QLQ-C30 had demonstrated adequate reliability and validity among lung cancer patients. The EORTC QLQ-LC13 is a lung cancer–specific module that may better reflect the changes in HRQoL among patients with lung cancer compared to general HRQoL questionnaires. The EQ-5D-5L is a generic tool used to measure HRQoL of patients across many diseases; the validity and reliability of this tool has been demonstrated across many patient populations. It uncertain whether the EQ-5D-5L is best equipped to capture the impacts of treatment and disease on lung cancer patients, especially among a subgroup of ALK-positive advanced or metastatic patients who are likely to develop brain metastases. The development of brain metastases may affect their HRQoL differently than it would patients for whom the likelihood of developing such metastases and associated side effects is lower. In particular, the EORTC QLQ-C30, EORTC QLQ-LC13, and EQ-5D-5L showed a lack of reliability regarding assessment of cognitive function (Appendix 3). It is possible that the HRQoL instruments used in the CROWN trial may not have adequately captured the impact of patient disease and treatment specifically related to cognitive function.

A summary of dose exposure showed that patients in the lorlatinib group received treatment for a longer mean period of time than patients in the crizotinib group. Consequently, the dose intensity and cumulative dose intensity were greater in patients treated with lorlatinib compared to those treated with crizotinib. These differences in exposure of treatment between the treatment groups are likely to have resulted in improved treatment effects in the lorlatinib group. However, the longer exposure to treatment should also be considered when interpreting safety and HRQoL data, as longer exposure to treatment may influence toxicities and quality of life for patients.

External Validity

The CROWN trial compared lorlatinib to crizotinib for patients with ALK-positive metastatic NSCLC. Lorlatinib was administered at a dose of 100 mg orally once daily, while crizotinib was administered at a dose of 250 mg twice daily orally; these doses are in alignment with Health Canada–approved indications. At the time of the trial’s inception, crizotinib was considered the standard of care. However, newer-generation TKIs such as alectinib and brigatinib have demonstrated improved efficacy and ability to penetrate the blood-brain barrier. They have become the standards of care for patients in the first-line treatment setting of locally advanced and metastatic ALK-positive NSCLC and are therefore more relevant comparators against lorlatinib. The comparative efficacy and safety of these treatments is not known because they have not been directly compared in RCTs. As such, the sponsor provided an ITC that compared lorlatinib to these treatments. Three ITCs published in the literature that compared lorlatinib to other ALK TKIs were also identified by the CADTH team. The ITCs are summarized and critically appraised in the Indirect Evidence section.

The majority of patients in the CROWN trial had an ECOG PS score of 0 or 1. The generalizability of the efficacy results (i.e., improvement in PFS) to those patients with a poorer ECOG PS remains unknown. Moreover, the study excluded patients with potential vascular or cardiac diseases, or patients with unfavourable lab testing on renal, liver, pancreatic, or bone marrow function. In reality, the safety profile of lorlatinib for patients with those comorbidities or abnormal testing may be even worse, particularly in the event that lorlatinib increased the risk of hypercholesterolemia and hypertriglyceridemia compared to crizotinib. The CROWN trial allowed for the enrolment of patients with brain metastases, who accounted for 25.5% of the lorlatinib group and 27.2% of the crizotinib group.¹⁷ Inclusion of patients with brain metastases is highly relevant as many patients with ALK-positive metastatic NSCLC develop brain metastases. Given the limitations of subgroup analysis results, intracranial-specific efficacy end points showed numerically improved patient outcomes in the lorlatinib group over the crizotinib group. The clinical experts consulted by CADTH recognized the results of the CROWN trial for patients with brain metastases.

In general, the baseline characteristics of patients in the CROWN trial were considered representative of patients typically treated in clinical practice. However, the clinical experts consulted by CADTH noted that ALK-positive NSCLC is more common in younger patients with a history of light smoking or no smoking history. In the CROWN trial, 36.9% of patients in the lorlatinib group and 29.3% of patients in the crizotinib group were reported to have been former smokers at baseline. While this is unlikely to have affected patient outcomes in the CROWN trial, the clinical experts suggested that the proportions of patients captured as former smokers may be lower in clinical practice. The clinical experts noted that the eligibility criteria for RCTs are typically more restrictive and may not capture the breadth of patients that clinicians see in clinical practice. It is unclear whether the treatment effects of lorlatinib are generalizable to patients who were excluded from the CROWN trial.

The CROWN trial allowed for enrolment of patients with brain metastases, who accounted for 25.5% of the lorlatinib group and 27.2% the crizotinib group.¹⁷ The CROWN trial demonstrated improved PFS among all patients, including those with brain metastases. In addition, intracranial-specific efficacy end points (IC-ORR, IC-TTP, IC-DOR, and IC-TTR) showed numerically improved results suggestive of improved patient outcomes in the lorlatinib group over the crizotinib group. The clinical experts consulted by CADTH supported the results of the CROWN trial, which suggested improved efficacy with lorlatinib for patients with brain metastases.

In the CROWN trial, assessments of efficacy were conducted every 8 weeks while on treatment until disease progression. The clinical experts consulted by CADTH indicated that patients are typically assessed every 3 months for response in their first year of treatment, and imaging for brain metastases generally occur every 6 months. Assessments for efficacy were conducted more frequently in the CROWN trial, which may have allowed for greater detection of treatment response in patients. It is not expected that the frequency of assessments in the CROWN trial negatively affected patient outcomes; however, the assessments were likely conducted too frequently and are not reflective of clinical practice.

Indirect Evidence

Objectives and Methods for the Summary of Indirect Evidence

The CROWN trial compared lorlatinib to crizotinib for the first-line treatment of adult patients with ALK-positive locally advanced (not amenable to curative therapy) or metastatic NSCLC. Crizotinib was considered the standard of care for patients with ALK-positive locally advanced or metastatic NSCLC; however, other treatments such as alectinib or brigatinib have been introduced for treatment of these patients and are also considered relevant comparators in the first-line treatment setting. This section summarizes and critically appraises the available indirect evidence comparing lorlatinib to other relevant treatments for ALK-positive advanced or metastatic NSCLC.

A supplemental search of the medical literature for publicly available ITCs was conducted by CADTH. A focused literature search for network meta-analyses (NMAs) dealing with non–small cell lung cancer was run in MEDLINE All (1946–) on July 21, 2021. No limits were applied to the search. Three published ITCs were identified in the CADTH literature search: Chuang et al. (2021),¹⁹ Wang et al. (2021),²⁰ and Ando et al. (2021).²¹

The sponsor conducted a systematic literature review to identify relevant studies for its submitted ITC. A systematic literature review by BresMed was conducted on October 31, 2019. Details of the systematic literature review are included in Table 36. Inclusion criteria for the systematic literature review were stated to be subject to the “80% rule,” under which 80% of the patients in a study must have met the inclusion criteria of the systematic literature review. Studies were limited to patients treated in the first-line setting; and studies that included patients who were not treatment-naive were not considered unless there were subgroup analysis involving patients who were treatment-naive.¹⁸ A grey literature search was also conducted to identify conference proceedings for studies that were ongoing or completed but unpublished. The search for grey literature was restricted to the proceedings published between 2018 and 2019. In addition, the bibliographies of key systematic reviews and meta-analysis articles were screened by the sponsor to ensure all relevant clinical studies were captured.¹⁸

Ando et al. (2021)²¹ conducted a systematic literature review to identify relevant literature for their NMA. The systematic literature review conducted on May 6, 2021, used 4 databases to capture published reports from 1946 to date. In addition to the 4 databases, the search strategy used for PubMed was also used in the Embase, CENTRAL, and SCOPUS databases. The reference lists of retrieved studies were also checked for relevant studies and the corresponding authors of studies were contacted when sufficient information or data could not be retrieved from the literature. To avoid publication bias, a manual search for relevant articles was conducted to supplement studies found in the literature search. No details regarding restrictions on searches were provided.

Wang et al. (2021) conducted a literature search to identify publications for inclusion into their NMA. Publications were captured if they were published until January 2021. In addition to publications retrieved from the literature search, reference lists of selected studies were hand-searched. Although Wang et al. (2021) did not specify inclusion criteria related to study design, based on studies included in the NMA, the CADTH team assumed that results were limited to RCTs.

Chuang et al. (2021) conducted a comprehensive literature search to identify publications for inclusion into their NMA. All databases were searched from inception until December 12, 2020. Abstracts in the main oncology congresses databases, including the American Society of Clinical Oncology, the American Association for Cancer Research, the European Society for Medical Oncology, and the World Conference on Lung Cancer, were reviewed. Reference lists of studies were also searched to identify any additional relevant literature.

Description of Indirect Treatment Comparisons

A description of the study selection methods used to identify relevant studies in each of the ITCs are reported in Table 36. All ITCs included studies of adult patients with advanced or metastatic ALK-positive NSCLC who were treated in the first-line setting; however, some of these also included patients who had received previous treatment for metastatic disease with chemotherapy or a previous ALK inhibitor. In these studies, the proportion of patients who had received previous treatment was generally less than 20%. All the ITCs compared lorlatinib with alectinib, brigatinib, and crizotinib; other comparators included ceritinib, ensartinib, and chemotherapy. In the sponsor’s ITC,¹⁸ quality assessments were performed based on the recommendations of the National Institute for Health and Care Excellence (NICE). while the published ITCs^19-21 conducted quality assessments using the Cochrane Collaboration risk of bias tool.

Methods of Sponsor’s Indirect Treatment Comparison

Objectives

The aim of the sponsor’s submitted ITC was to compare the relative efficacy of lorlatinib with alternative treatments for patients with ALK-positive advanced or metastatic NSCLC in the first-line treatment setting.

Study Selection Methods

A literature search was conducted based on the criteria reported in Table 36. Studies were screened by title and abstract, and the full-text screen resulted in the inclusion of 50 articles that captured 8 RCTs, and 75 articles that captured 54 non-RCTs. The following 8 RCTs were retrieved from the systematic literature review and considered for inclusion into the sponsor’s ITC: ALEX, J-ALEX, ALESIA, ALTA-1L, ASCEND-4, ASCEND-8, PROFILE 1014, and PROFILE 1029. The CROWN trial was also considered for inclusion, as well as another study identified by the sponsor, the eXalt3 trial. The 54 non-RCTs were not considered for inclusion into the ITC due to their study design.

ITC Analysis Methods

Details of the methodology used for the sponsor’s ITC are provided in Table 37.

Treatments

Treatments included in the studies of the ITC were lorlatinib, alectinib, brigatinib, ceritinib, crizotinib, ensartinib, and chemotherapy. Included single trials evaluated lorlatinib (CROWN), brigatinib (ALTA-1L), and ensartinib (eXalt3); these oral treatments were administered at 100 mg daily, 170 mg daily, and 225 mg daily, respectively. Three trials assessed alectinib (ALEX, J-ALEX, and ALESIA) at doses of 600 mg twice daily orally or 300 mg twice daily orally. Two trials provided information about ceritinib (ASCEND-4 and ASCEND-8) and assessed ceritinib at 3 doses: 750 mg twice daily, 600 mg twice daily, and 450 mg twice daily, all administered orally. Crizotinib was included in 8 trials (CROWN, ALEX, ALESIA, J-ALEX, ALTA-1L, PROFILE 1014, PROFILE 1029, and eXalt3), all of which assessed crizotinib at 250 mg twice daily orally. Chemotherapy regimens were included in 3 trials (PROFILE 1014, PROFILE 1029 and ASCEND-4), which assessed the same regimen of pemetrexed (500 mg/m²) in combination with carboplatin (area under the curve of 5 mg/min/mL or 6 mg/min/mL) or cisplatin (75 mg/m²) administered intravenously every 3 weeks. The J-ALEX trial was included in the NMA as a separate node; comparisons to alectinib at 300 mg were therefore made separately. Multiple doses of ceritinib were also assessed in the ASCEND-4 and ASCEND-8 trials, including 750 mg, 600 mg, and 450 mg. The different doses of ceritinib were also assessed as separate nodes.

Crossover was not permitted in the CROWN, ALEX, ALESIA, J-ALEX ASCEND-8, or eXalt3 trials.¹⁸ The presence of treatment crossover was considered as it may introduce bias into the assessments of efficacy. Crossover was permitted in the ALTA-1L, ASCEND-4, PROFILE 1014, and PROFILE 1029 trials. In the ALTA-1L trial, 35 patients (25.5%) randomized originally to the crizotinib group crossed over to receive brigatinib. The sponsor noted that, while no methods to adjust for crossover were reported in the primary publication of the ALTA-1L trial, investigations reported in the sponsor’s submission to NICE deemed the rank-preserving structural failure time (RPSFT) method to be the most appropriate method of adjusting for crossover in the assessment of OS and such an analysis was conducted on a later data cut-off date of the ALTA-1L trial. Therefore, HRs from the NICE submission were used in the sponsor’s NMA for scenarios that did and did not include treatment crossover. The ASCEND-4 trial reported 80 patients (43.3%) randomized to the chemotherapy group who crossed over to receive ceritinib. Similar to crossover scenarios for brigatinib in the ALTA-1L trial, the RPSFT method was used to adjust for crossover for OS as a sensitivity analysis in the sponsor’s submission to NICE. Information from these crossover analyses were used in the sponsor’s NMA. In the PROFILE 1014 trial, 72.4% of patients randomized to chemotherapy crossed over to receive crizotinib. To account for crossover in the PROFILE 1014 trial, the sponsor used the RPSFT methodology. In the PROFILE 1029 trial, 70% of patients randomized to receive chemotherapy crossed over to receive crizotinib. No adjustments for crossover were conducted for this trial. The sponsor noted that removal of trials that included crossover (without RPSFT adjustment) was not possible, as it resulted in a disconnected network. Sensitivity analyses were performed when crossover-adjusted data were substituted into the network, where available, for OS.

End Points

OS and PFS were the primary end points analyzed in the ITC.

Overall Survival: OS data were available from all trials except the J-ALEX and ASCEND-8 trials; exclusion of these trials did not break the overall network. In addition, no OS data were available for ceritinib at 600 mg and 450 mg, or for alectinib at 300 mg. The overall network for OS is depicted in Figure 12.

Figure 12: OS Network for Sponsor’s ITC

BID = twice daily; OS = overall survival; QD = once a day.

Source: Sponsor’s indirect treatment comparison.¹⁸

Overall Survival Sensitivity Analyses: Sensitivity analyses for OS were conducted that excluded Asian patients. The ALESIA, J-ALEX, ASCEND-8, and PROFILE 1029 trials enrolled only Asian patients; therefore, these trials were excluded from the sensitivity analysis.

Another sensitivity analysis only included patients who were treatment-naive. The ALTA-1L and eXalt3 trials were therefore excluded as they did not report OS data for subgroups of patients who did not receive prior treatment.

Progression-Free Survival: PFS data assessed by independent radiological review were available for all 8 trials; and PFS data assessed by investigators were available in 6 of the trials (CROWN, ALEX, ALESIA, J-ALEX, ALTA-1L, and ASCEND-4). Of the studies included in the NMA, PFS data reported by race subgroup were available from 7 trials (CROWN, ALEX, ALTA-1L, ASCEND-4, ASCEND-8, PROFILE 1014, and eXalt3).

However, as intracranial PFS was an outcome interest; this outcome was only reported for the CROWN trial, based on patient-level data for IC-TTP and OS.

Data for IC-TTP were available from all trials except the ASCEND-4, ASCEND-8, and eXalt3 trials.

The overall network for PFS is depicted in Figure 13.

Figure 13: PFS Network for Sponsor’s ITC

BID = twice daily; PFS = progression-free survival; QD = once a day.

Source: Sponsor’s indirect treatment comparison.¹⁸

Progression-Free Survival Sensitivity Analysis: Three trials — ALESIA, J-ALEX, and PROFILE 1029 — enrolled only Asian patients. These trials were excluded from the sensitivity analyses.

Additional sensitivity analyses were also conducted on patients who did not receive any prior treatment. All trials had available PFS data for subgroups of patients who were treatment-naive except for the eXalt3 trial, which was excluded from this analysis.

Subgroup PFS data for patients with brain metastases were available in all trials except the ASCEND-8, PROFILE 1029, and eXalt3 trials. These trials were excluded in sensitivity analyses to determine the effect of brain metastases at baseline on PFS.

Other End Points

Other end points for consideration in the sponsor’s ITC included ORR, IC-ORR, IC-CR, intracranial progression response (IC-PR), grade 3 and 4 AEs, and discontinuations due to AEs. These end points were binary end points considered in the sponsor’s ITC. The availability of data for each of these binary end points is reported in Table 38.

Data for ORR was reported in all trials and assessed by independent radiological review, except in the ALEX and ALESIA trials, for which data reported for ORR were assessed by an investigator. All studies were included in the network for ORR (Figure 14).

Data for IC-ORR was reported in all studies except for J-ALEX, ASCEND-8, and PROFILE 1029. Removal of these trials did not disconnect the overall network (Figure 15).

Data for IC-CR was reported in most trials, except the J-ALEX, ASCEND-8, PROFILE 1014, and PROFILE 1029 trials. Excluding the studies that did not report IC-CR resulted in a disconnected link between chemotherapy and ceritinib; the remaining links in the network remained connected (Figure 16).

Data for IC-PR was only reported in the CROWN, ASCEND-4, ASCEND-8, and eXalt3 trials. However, because the counts for IC-PR were zero for both treatment groups in the eXalt3 trial, the eXalt3 trial was excluded from the network. Overall, due to lack of available data, comparisons for IC-PR were not feasible (Figure 17).

A network diagram for grade 3 and 4 AEs is depicted in Figure 18. Data regarding grade 3 and 4 AEs were available in all trials except the ASCEND-8, PROFILE 1029, and eXalt3 trials. Removal of trials not reporting grade 3 and 4 AEs did not result in a disconnected network. The network diagram for AEs resulting in treatment discontinuation is depicted in Figure 19. Because all trials reported data regarding AEs resulting in treatment discontinuation, all trials were included in the network.

Figure 14: ORR Network Diagram in Sponsor’s ITC

BID = twice daily; ORR = objective response rate; QD = once a day.

Source: Sponsor’s indirect treatment comparsion.¹⁸

Figure 15: IC-ORR Network Diagram in Sponsor’s ITC

BID = twice daily; IC = intracranial; ORR = objective response rate; QD = once a day.

Source: Sponsor’s indirect treatment comparison.¹⁸

Figure 16: IC-Complete Response Network Diagram in Sponsor’s ITC

BID = twice daily; IC = intracranial; QD = once a day.

Source: Sponsor’s indirect treatment comparison.¹⁸

Figure 17: IC-Progression Response Network Diagram in Sponsor’s ITC

BID = twice daily; IC = intracranial; QD = once a day.

Source: Sponsor’s indirect treatment comparison.¹⁸

Figure 18: Grade 3 and Grade 4 AEs Network Diagram in Sponsor’s ITC

AE = adverse event; BID = twice daily; QD = once a day.

Source: Sponsor’s indirect treatment comparison.¹⁸

Figure 19: AEs Resulting in Discontinuation Network Diagram in Sponsor’s ITC

BID = twice daily; QD = once a day.

Source: Sponsor’s indirect treatment comparison.¹⁸

Both OS and PFS were estimated using published HRs and associated 95% CIs for data inputs. If only Kaplan-Meier curves were available, data from the Kaplan-Meier curves were digitized using methods by Guyot et al. (2012) to generate pseudo-patient-level data, and then estimate HRs. The proportional hazards assumption was assessed through visual inspection of Kaplan-Meier curves for both OS and PFS in each trial. For PFS, the sponsor treated the assumption as reasonable based on Kaplan-Meier plots as there was no clear change in HRs over time. The sponsor stated the proportional hazards assumption was reasonable for OS as well.¹⁸

Analyses of other end points (i.e., ORR, IC-ORR, IC-CR, IC-PR, grade 3 and 4 AEs, and AEs resulting in treatment discontinuation) used the number of patients with an event and the total number of patients in each treatment group of a study. Treatment effects were assessed using ORs and associated 95% CrIs.¹⁸

Results of Sponsor’s Indirect Treatment Comparison

Summary of Included Studies

Ten trials were included for analyses in the sponsor’s ITC, including CROWN, ALEX, J-ALEX, ALESIA, ALTA-1, ASCEND-4, ASCEND-8, PROFILE 1014, PROFILE 1029, and eXalt3. A summary of trial characteristics is provided in Table 39. Six of the included trials included treatment of patients in the first line, while the remaining included patients in multiple lines of therapy. Three of the trials enrolled Asian populations. All trials provided data on OS and PFS data except for the J-ALEX and ASCENT-8 trials, which did not have available OS data.

A summary of inclusion and exclusion criteria of the RCTs included in the ITC is provided in Table 40. Trials included in the ITC were published between 2017 and 2020. Enrolled patients included adults with stage IIIB or IV ALK-positive NSCLC (CROWN, ALEX, J-ALEX, ALESIA, ALTA-1L, ASCEND-8, and eXalt3); the remaining trials included patients with advanced, metastatic, or recurrent disease. Patients in all trials were naive to previous treatment with ALK inhibitors. Three trials (J-ALEX, ALTA-1L, and eXalt3) included patients who could have received prior treatment with chemotherapy and who were chemotherapy-naive, and 1 trial (ASCEND-8) included patients who could have received prior treatment with any treatment (including crizotinib) or who were naive to any treatment. All trials enrolled patients with an ECOG PS of 0 to 2. All trials permitted the enrolment of patients with brain metastases; however, some trials included criteria about patients with brain metastases, often specifying that their CNS disease must be asymptomatic and stable. All trials specified that disease must be measurable using RECIST 1.1 criteria; the CROWN trial was the only trial to specify patients must have had at least 1 extracranial measurable target lesion. Tumour requirements were not reported for the eXalt3 trial.

The J-ALEX, ALTA-1L, ASCEND-8, and eXalt3 trials enrolled patients who had received prior treatment; a summary of prior treatments received by these patients is provided in Table 41. In all trials, previous treatment with ALK inhibitors was not permitted, except for the ASCENT-8 trial, which allowed previous treatment with crizotinib. Previous treatment with chemotherapy was allowed within all 4 trials; previous treatment with chemotherapy was the only previous treatment permitted within the J-ALEX, ALTA-1L, and eXalt3 trials.¹⁸

There were no imbalances in the proportions of patients who received prior treatments across treatment groups, except for the ASCEND-8 trial, in which more patients randomized to receive ceritinib at 600 mg received a prior treatment than patients randomized to the 450 mg or 750 mg ceritinib groups. Across the trials, the proportions of patients with previous therapy were similar, at between 41% and 24%. Following the 80% rule used by the sponsor for its systematic literature review, patients in studies were required to fit 80% of eligibility criteria of the systematic literature review; the majority of patients enrolled in these trials were therefore treatment-naive.¹⁸