CADTH Reimbursement Review

Dostarlimab (Jemperli)

Sponsor: GlaxoSmithKline Inc.

Therapeutic area: Endometrial 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

Endometrial cancer (EC) is the most common gynecological cancer.¹ An estimated 7,000 to 8,000 new cases of EC are diagnosed in Canada every year, with approximately 1,400 deaths annually.^2,3 Approximately 80% of ECs are diagnosed at an early stage and are curable with surgery.¹ Recurrence occurs in 13% to 20% of patients.^4,5 The 5-year survival rate of patients diagnosed with metastatic or advanced disease is less than 20%.⁶ Patients with advanced, metastatic, or recurrent EC have limited effective therapeutic options after front-line standard treatment with a platinum-containing chemotherapeutic.

Mismatch repair-deficient (dMMR) and microsatellite instability-high (MSI-H) tumour status is a predictive biomarker of clinical benefit from checkpoint inhibitors, and represent approximately 25% of primary ECs and 13% to 30% of recurrent ECs.^7-9 At first recurrence or primary advanced disease, response rates with platinum-based combination regimens in the first-line setting ranges from 40% to 62%.^10-13 However, for patients with advanced or recurrent EC who have progressed on or after platinum-based chemotherapy, there is currently no standard second-line therapy. Single-drug chemotherapies or hormonal therapy may be administered, but these have low response rates and no clear survival benefit.¹⁴

Dostarlimab is an anti-programmed cell death protein-1 (PD-1) monoclonal antibody. It targets the cellular pathway between the PD-1 receptor and 2 ligands, PD-L1 and PD-L2, found on immune cells. Dostarlimab binds to the PD-1 receptor and blocks its interaction with PD-L1 and PD-L2, which allows the PD-1 pathway-mediated immune response and antitumour immune response to occur.¹⁵ On December 23, 2021, dostarlimab was approved by Health Canada for the treatment of adults with dMMR or MSI-H recurrent or advanced EC that has progressed on or after prior treatment with a platinum-containing regimen. The sponsor’s requested reimbursement criteria for dostarlimab align with the Health Canada–approved indication. Dostarlimab underwent review by Health Canada, which used an expedited review process (advance consideration under Notice of Compliance with conditions [NOC/c]). Dostarlimab has no other Health Canada–approved indication and has not previously been reviewed by CADTH. Dostarlimab was approved by the FDA and European Union, and is currently under review by the National Institute of Health Care Excellence (NICE) and the Scottish Medicines Consortium.^16,17

Dostarlimab is available as a 500 mg IV infusion and is administered as an IV infusion over 30 minutes. The recommended dosage is 500 mg every 3 weeks for dose 1 through 4, and 1,000 mg every 6 weeks for dose 5 onward. Treatment may continue until disease progression or unacceptable toxicity. Dose reductions are not recommended, but dosing delays and discontinuation may be required based on safety and tolerability. Patients should be selected for treatment based on dMMR or MSI-H tumour status, determined by an accredited laboratory using validated testing methods.

The objective was to perform a systematic review of the beneficial and harmful effects of dostarlimab for the treatment of adults with dMMR or MSI-H recurrent or advanced EC that has progressed on or after prior treatment with a platinum-containing regimen.

Stakeholder Perspectives

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

Patient Input

Patient and caregiver input used for this review was collected by the Canadian Cancer Society. The input was based on an online survey and patient and caregiver testimonials. A total of 6 testimonials and 22 survey responses were received (20 patients with current or previous EC and 2 caregivers).

Respondents indicated that EC symptoms affected their daily activities, causing detrimental effects on their health-related quality of life (HRQoL). Respondents reported that the most significant side effects related to their current cancer treatment were issues with libido, sexual function, and fatigue. Loss of income due to absence from work and travel costs for cancer treatment were important financial barriers.

Respondents reported that they expect the following key outcomes for any treatment: better HRQoL, longer periods of remission, better affordability, better access, and fewer side effects. Eight of 22 respondents indicated that they had received direct experience with dostarlimab, either receiving it or assisting a patient who received it. All respondents indicated that dostarlimab was easier to use than other therapies because it had few to no side effects, longer intervals between doses, and a shorter infusion time.

Clinician Input

Input From Clinical Experts Consulted by CADTH

All CADTH review teams include at least 1 clinical specialist with expertise in the diagnosis and management of the condition for which the drug is indicated. Clinical experts are a critical part of the review team and are involved in all phases of the review process (e.g., providing guidance on the development of the review protocol, assisting in the critical appraisal of clinical evidence, interpreting the clinical relevance of the results, and providing guidance on the potential place in therapy). The following input was provided by 2 clinical specialists with expertise in the diagnosis and management of EC.

The clinical experts consulted by CADTH indicated that currently there are no standard, funded second-line treatment options for advanced or recurrent EC. The clinical experts agreed that patients who would most benefit from dostarlimab include those with identified dMMR or MSI-H recurrent or advanced EC. One of the clinical experts noted that dostarlimab could be used as monotherapy in first-line settings or later in the absence of effective treatments. The clinical experts noted that treatment with dostarlimab would not be suitable in patients with the following characteristics: very poor performance status; history of severe autoimmune disease; prior immunotherapy use; known uncontrolled central nervous system metastases and/or carcinomatous meningitis; poor medical risk due to a serious uncontrolled medical disorder; nonmalignant systemic disease or active infection requiring systemic therapy; or microsatellite stable (MSS) EC.

In the opinion of the clinical experts consulted by CADTH, treatment with dostarlimab should be discontinued in the case of disease progression, severe toxicity, or intolerability. The clinical experts indicated that the following outcomes would best assess response to treatment: overall survival (OS); response rate based on clinical and radiological investigation; progression-free survival (PFS); reduction of cancer burden and symptom improvement in activities of daily living; HRQoL; durability of response; and response to subsequent therapies.

In terms of clinically meaningful responses, the clinical experts recommended that in addition to clinical assessment of disease symptoms and duration of disease control, the use of standard immune-related Response Evaluation Criteria in Solid Tumours (irRECIST) for the assessment of response to immunotherapeutic treatments may be useful.

Clinician Group Input

A total of 7 clinician group inputs were submitted from the following groups: British Columbia Cancer Provincial Gynecological Oncology Tumour Group; McGill University Health Centre (MUHC), Division of Gynecologic Oncology; Ontario Health-Cancer Care Ontario (OH-CCO) Gynecological Drug Advisory Committee; Princess Margaret Cancer Centre (PMCC), Gynecologic Cancers Disease Site Group, Medical Oncology Group; Saskatchewan Cancer Agency (SCA); the Society of Gynecologic Oncology of Canada (GOC); and Sunnybrook Health Sciences Centre (SBHSC).

The views of the clinician groups were overall consistent with those of the clinical experts consulted by CADTH. The clinician groups indicated that the most important treatment goals are achieving disease control, delaying worsening of symptoms, prolonging survival, maintaining HRQoL, delaying disease progression, and an acceptable safety profile. All the clinician groups indicated that all patients with recurrent EC would benefit from an effective immunotherapy, but patients with dMMR or MSI-H subtypes would be the most likely to benefit from immune checkpoint inhibitor therapy. All groups recommended that patients diagnosed with metastatic EC should be offered platinum-based chemotherapy as first-line therapy. However, the British Columbia Cancer Provincial Gynecological Oncology Tumour Group did acknowledge that treatment with an immune checkpoint inhibitor may be an appropriate first-line therapy when chemotherapy is contraindicated or not desirable.

Drug Program Input

The drug programs provide input on each drug being reviewed through CADTH’s reimbursement review processes by identifying issues that may impact their ability to implement a recommendation. The drug plans identified implementation issues related to considerations for the initiation of therapy, continuation or renewal of therapy, and generalizability. The clinical experts consulted by CADTH weighed evidence from the GARNET trial (Study 4010-01-001)¹⁸ and other clinical considerations to provide responses to implementation questions posed by drug programs. Refer to Table 4 for more details.

Clinical Evidence

Pivotal Study

Description of Study

The GARNET trial (Study 4010-01-001)¹⁸ is an ongoing nonrandomized, noncomparative, multi-centre, open-label, phase I dose-escalation and cohort-expansion study in patients with recurrent or advanced solid tumours. The objective of part 2B of the GARNET trial was to evaluate the safety and antitumour activity of dostarlimab in patients with advanced solid tumours. Cohort A1 included patients with advanced or recurrent dMMR or MSI-H EC that had progressed on or after prior treatment with a platinum-containing regimen. Patients were enrolled from 123 sites in 8 countries (including 8 Canadian sites). Enrolment started on April 10, 2017, and is ongoing.

To be eligible, patients had to be at least 18 years of age, diagnosed with recurrent or advanced dMMR or MSI-H EC, and had to have progressed on or after no more than 2 lines of prior systemic therapy, with at least 1 of these being platinum-based doublet therapy. In addition, patients had to have adequate organ function and an Eastern Cooperative Oncology Group (ECOG) performance status of less than 1. The study consisted of 3 phases: the screening phase (up to 35 days before treatment), the treatment phase, and the follow-up phase. A total of 129 patients were enrolled in cohort A1. All patients received dostarlimab by IV injection (500 mg every 3 weeks for cycles 1 to 4, and 1,000 mg every 6 weeks from cycle 5 onward) for up to 2 years or until disease progression, treatment discontinuation, or withdrawal.

The co-primary outcomes of the GARNET trial were objective response rate (ORR) and duration of response (DOR). The secondary outcomes were OS, disease control rate (DCR), immune-related DCR (irDCR), PFS, immune-related PFS (irPFS), immune-related ORR (irORR), and immune-related DOR (irDOR). HRQoL was an exploratory outcome assessed by the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 (EORTC QLQ-C30) and the EQ-5D 5-Levels (EQ-5D-5L). The safety outcomes assessed included treatment-emergent adverse events (TEAEs), serious adverse events (SAEs), immune-related adverse events (irAEs), ECOG Performance Status, clinical laboratory measures, vital signs, electrocardiogram, physical examination, serum pregnancy testing, and concomitant medications.

The GARNET trial has 3 planned interim analyses that are contingent on a combined enrolment of 100, 200, and 300 patients, respectively, in cohort A1 and cohort F and 24 weeks of follow-up. Only data from the second interim analysis (IA-2) was reviewed by CADTH; it included a subset of patients from the first interim analysis (IA-1) (N = 72). Data for this subset of patients is available in Appendix 4. The data cut-off dates were July 8, 2019, March 1, 2020, and November 1, 2021, for IA-1, IA-2 (N = 105), and the third interim analysis (IA-3) (N = 143), respectively. The GARNET trial was sponsored by GlaxoSmithKline Inc.

Because only 2 (< 2%) enrolled patients had mismatch repair (MMR)-unknown but MSI-H tumours (MMR-unk/MSI-H), these patients were included with patients with dMMR tumours. Median age was 64 years (range = 39 to 80 years), median weight was 71 kg (range = 34.0 to 141.4), and median body mass index (BMI) was 27.97 kg/m² (range = 13.6 to 53.9 kg/m²). Most patients were White (> 75%). The most common histology type of EC was type I endometrioid carcinoma (67.6%), with grade 2 being the most common histology grade at diagnosis (39%). More than two-thirds (67.6%) of patients had stage IV EC. ECOG Performance Status of 1 was the most common (60%), followed by ECOG 0 (40%). All patients received prior anti-cancer treatment. Most patients (88.6%) had 2 or fewer lines of prior anti-cancer regimens, and a smaller proportion had 2 or more lines (11.5%). More than half (56.2%) of patients had received prior regimens for metastatic disease. The patient subgroups of interest, as identified in the CADTH systematic review protocol, included the following: International Federation of Gynecology and Obstetrics (FIGO) stage, histology of tumour type (e.g., type I, type II) and subtype (e.g., clear cell carcinoma), number and type of prior systemic therapies (e.g., chemotherapy, hormonal therapy), prior radiation, and progression-free interval after the most recent platinum-containing anti-cancer therapy.

Outcomes

The key outcomes from cohort A1 of the GARNET trial are summarized in Table 2. At the time of IA-2, the median duration of follow-up was 16.3 months, and the median duration of treatment (DoT) was 26 weeks.

The proportion of patients who achieved an ORR (complete response [CR] or partial response [PR]) was 44.8% (95% confidence interval [CI], 35.0 to 54.8). The Best overall response (BOR) was CR in 11 patients (10.5%), PR in 36 patients (34.3%), and stable disease in 13 patients (37.1%). The DCR was 57.1% (95% CI, 35.0 to 54.8). Of those who responded, 89.5% had an ongoing response. The median DOR was not reached, but 79% of patients who achieved an objective response had a DOR of at least 6 months.

At the time of IA-2, a large proportion of patients (66.7% and 45.7%, respectively) had no OS or PFS events. The median OS was not reached, but Kaplan-Meier (KM) estimates for the probability of survival at 6, 9, and 12 months were 80.9% (95% CI, 71.7 to 87.4), 75.1% (95% CI, 65.2 to 82.6), and 68.9% (95% CI, 58.3 to 77.4), respectively. The median PFS was 5.5 months (95% CI, 3.2 to not reached), with KM estimates of PFS by RECIST 1.1 of 48.6% (95% CI, 38.6 to 57.9) at month 6 and 47.5% (95% CI, 37.4 to 56.8) at month 9 and at month 12. In terms of HRQoL, the EQ-5D-5L visual analogue scale (EQ-VAS) and EORTC QLQ-C30 scores appeared stable over time. Summary data for the EQ-5D-5L descriptive system were not provided.

As part of the sponsor’s feedback on this CADTH reimbursement review report, the sponsor provided CADTH with a summary of the updated analysis (data cut-off November 1, 2021) for certain baseline characteristics, efficacy, and safety outcomes in the GARNET trial. The results of the updated analysis were, overall, consistent with those reported in the previous analyses performed as of March 1, 2020, data cut-off date. The additional results from the November 1, 2021, data cut-off date are available in Appendix 5.

Harms Results

A summary of the key harms reported in cohort A1 of GARNET are summarized in Table 2. Almost all patients (95.3%) experienced at least 1 TEAE. The most common serious TEAEs were abdominal pain, acute kidney injury, sepsis, pulmonary embolism, pyrexia, and urinary tract infection. Grade 3 or higher TEAEs occurred in 48.1% of patients, with the most common being anemia (14.7%), abdominal pain (5.4%), and hyponatremia (3.9%). No patient withdrew due to an adverse event (AE) as a primary reason. Study treatment discontinuation due to AEs occurred in 11.6% of patients, whereas AEs that led to study treatment interruption occurred in 24% of patients. The most common AEs leading to study interruption were anemia (3.1%) and diarrhea (2.3%).

One patient died due to a TEAE (aspiration) during the treatment period, and 4 patients died due to TEAEs (i.e., pleural effusion, pneumonia, sepsis, and shock) during the 90-day safety follow-up. None of the TEAEs leading to death were considered treatment-related, and no TEAEs were the primary cause of death during the long-term follow-up period.

The notable harms associated with dostarlimab included immune-related toxicity. The incidence of irAEs was 34.9% in cohort A1. The most frequently reported irAEs (≥ 5%) were diarrhea and hypothyroidism. A total of 7.9% of patients had a serious irAE, 12.7% had an irAE of grade 3 or higher, and 4.8% had an irAE that led to study treatment discontinuation. Most of the irAEs were considered related to the study treatment.

Table 2: Summary of Key Results From the Pivotal Study (IA-2)

AE = adverse event; BICR = blind independent clinical review; BOR = best overall response; CI = confidence interval; CR = complete response; DCR = disease control rate; DOR = duration of response; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; EQ-5D-5L = EuroQol 5-Dimensions 5-Levels; EQ-VAS = EQ-5D-5L visual analogue scale; HRQol = health-related quality of life; IA-2 = second interim analysis; KM = Kaplan-Meier; max = maximum; min = minimum; NE = not evaluable; NR = not reported; ORR = objective response rate; OS = overall survival; PFS = progression-free survival; PR = partial response; SAEs = serious adverse events; SD = standard deviation; TEAEs = treatment-emergent adverse events.

Note: A “≥” sign indicates that patients’ response is ongoing.

^aData from patients with dMMR tumours and MMR-unk/MSI-H tumours are presented as a total due to the small proportion of patients with such tumours (3 in the safety analysis dataset and 2 in the primary efficacy analysis dataset); the sponsor noted that data remained similar when these patient groups were pooled together.

^b95% CIs were generated using the method of Brookmeyer and Crowley. Biometrics, 1982;38:29-41.

^cExact 2-sided 95% CI for the binomial proportion.

^dNo summary results were provided for the EORTC QLQ-C30 and EQ-5D-5L scores.

^eAll responders who have not yet died or progressed (including clinical progression); the denominator for the percentage is the number of responders.

^fTEAEs experienced by ≥ 5% of patients with dMMR or MSI-H EC by preferred term (safety analysis dataset).

^gTEAEs were experienced by ≥ 15% of patients with dMMR or MSI-H EC by preferred term (safety analysis dataset).

^hWithin 90 days after the EOT visit or until the first follow-up anti-cancer therapy, whichever occurred first.

ⁱ90 days after the EOT visit or until the first follow-up anti-cancer therapy, whichever occurred first.

Source: Clinical Study Report.¹⁸

Critical Appraisal

Internal Validity

The main limitation of the GARNET trial is the single-arm design, which makes it challenging to interpret the data and determine whether the efficacy and safety events observed were attributable to dostarlimab. Formal hypothesis and statistical significance testing were not performed, limiting the ability to draw conclusions. Given that results were based on an interim analysis, some time-to-event outcomes, including median OS and DOR, were not reached due to data immaturity; therefore, the treatment effect observed with dostarlimab may be overestimated. The risk of overestimating HRQoL benefit and known subjective harms is also high, given the open-label trial design in which treatment was not blinded. To mitigate bias, the sponsor used a blind independent clinical review (BICR) to evaluate treatment response, with standardized criteria for certain efficacy outcomes (i.e., ORR, DOR, PFS, and DCR). Therefore, bias is less of a concern for these end points and OS, and more of a concern for subjective end points, including HRQoL and safety. It is also acknowledged that mature OS data will be confounded by the use of subsequent anti-cancer therapy received by some patients after progressive disease. No analyses were undertaken to account for the potential of confounding. Overall, the magnitude and direction of bias is unclear. The clinical experts agreed that in the absence of robust comparative data on PFS and OS, no firm conclusions could be drawn on how dostarlimab compares with other relevant treatment options, as causal inferences cannot be made from the results of a single-arm trial design.

HRQoL was identified as an important outcome by the patient and clinician groups providing input for this review. However, no conclusions could be drawn from the HRQoL data from the GARNET trial due to several limitations. Given the wide and overlapping CIs, the reduced number of patient responses over time, and the lack of statistical testing and a definition of what constituted a clinically meaningful response, it is not possible to draw conclusions with precision from the available data.

External Validity

Overall, the clinical experts consulted by CADTH agreed that the inclusion and exclusion criteria, baseline patient characteristics, concomitant medications, and prohibited medications present in cohort A1 of the GARNET trial were reflective of patients they see in clinical practice for the indication under review. There were no barriers to identifying patients who would most benefit from the treatment, given that testing for MMR and microsatellite instability (MSI) status is standard practice in Canada. The clinical experts indicated that no difference in treatment effect would be expected based on variation in disease-management practices across participating countries. In the opinion of the clinical experts, as long as patients have dMMR or MSI-H tumour status, dostarlimab would be appropriate to administer after any of the prior therapies received by patients in the trial. However, they noted that clinical benefit may be diminished in patients with more prior lines of systemic therapy. There were a limited number of patients included in the primary efficacy analysis dataset (n = 105) and very few patients from various ethnic backgrounds, which may reduce the generalizability of the results to a real-world practice setting. Furthermore, the subgroup analyses had no statistical comparisons and even smaller sample sizes, which limits generalizability to a broader population.

Indirect Comparisons

Description of Studies

The sponsor submitted 6 reports of indirect treatment comparisons (ITCs) — 3 reports of matching-adjusted indirect comparisons (MAICs), and 3 reports of inverse probability of treatment weighting (IPTW) analyses¹⁹ — which aimed to compare survival between dostarlimab from the phase I GARNET trial with the current treatment paradigm in advanced or recurrent EC.

Efficacy Results

The primary end point for all comparisons was OS. Other outcomes included PFS, ORR, DOR, time to treatment discontinuation (TTD), DoT, time to next treatment (TTNT), time to deterioration in HRQoL, and AEs; however, these were less frequently investigated, and outcomes specifically important to patients, including HRQoL, were not assessed. The results of the MAIC and IPTW analyses generally suggest that dostarlimab is favoured for OS over all the included comparators.

Harms Results

The sponsor-submitted MAIC and IPTW reports did not assess safety outcomes.

Critical Appraisal

Although the results of the MAIC and IPTW analyses generally suggest that dostarlimab is favoured for OS over all the included comparators, there was significant uncertainty in the results based on the clinical heterogeneity of the included populations, resulting in reduced sample sizes and wide CIs. There were important differences in the design of the comparator studies that limit the ability to draw strong conclusions about the effectiveness for dostarlimab compared with other treatments. An important limitation of all analyses was the fact that MMR and MSI-H status was unknown for all or most patients in the comparator trial, and it is therefore uncertain whether the comparator population in the ITC analyses would be eligible for treatment with dostarlimab, providing further uncertainty about the comparative effectiveness.

Conclusions

One phase I, singe-arm, open-label trial (GARNET)¹⁸ provided evidence of the efficacy and safety of dostarlimab in adults with dMMR or MSI-H recurrent or advanced EC (cohort A1) that had progressed on or after prior treatment with a platinum-containing regimen. The clinical experts consulted by CADTH felt that the response outcomes (ORR and DOR co-primary outcomes) observed in the trial were clinically meaningful and durable for this patient population and, in their opinion, were higher than what is observed with currently used second-line therapies in this setting. The trial results were based on an interim analysis; therefore, there is the possibility of overestimating clinical benefit and underestimating harms. There was uncertainty around the magnitude of the clinical benefit, given the limitations inherent in the single-arm trial design. The trial data on important long-term outcomes were immature, and interpretation of OS will be confounded by the use of subsequent anti-cancer therapies. The clinical experts noted that a randomized controlled trial (RCT) would be needed to directly compare dostarlimab with currently available therapies in the second-line setting to accurately evaluate its efficacy in this patient population. In the absence of a direct comparison of dostarlimab with relevant treatment options, the sponsor submitted multiple ITCs. However, the CADTH critical appraisal of these analyses identified significant limitations with the submitted MAICs and IPTWs, which restricted the ability to interpret the relative treatment-effect estimates obtained. Limitations of the ITCs included heterogeneity across study designs, high risk of confounding and effect modifiers, and uncertainty regarding the inclusion of dMMR or MSI-H status in the comparator groups. The results for HRQoL, an outcome important to patients and clinicians, remained inconclusive due to the lack of statistical analysis, the substantial decline in patients completing questionnaires over time, and the lack of a definition of what constituted a clinical meaningful change from baseline. The notable harms observed with dostarlimab, such as diarrhea and peripheral nephropathy, were considered manageable and consistent with other immunotherapies by the clinical experts and, in their opinion, appeared favourable when naively compared with currently available chemotherapy options. However, interpreting the safety events attributable to dostarlimab was challenging because all patients in cohort A1 received the same treatment. Overall, limitations of GARNET’s single-arm design prohibited the drawing of causal conclusions between the intervention and outcomes.

Introduction

Disease Background

EC is the most common gynecological malignancy among women in Canada.¹ EC malignant tumours arise from the cells of the uterine lining. More than 95% of all uterine cancers are endometrial.²⁰ Uterine cancer is the 17th leading cause of cancer death in Canada.^3,20 The Canadian Cancer Society estimated that 8,000 women would be diagnosed with uterine cancer in 2021 and that 1,400 women would die of the disease.^2,3 EC most often occurs in patients older than 50 years, with an average age of diagnosis at 60 years.²¹ Diagnosis of EC occurs at an early stage for approximately 80% of patients because of the early presenting symptom of uterine bleeding.^1,20 The most common route of diagnosis of EC is endometrial biopsy, followed by endometrial curettage and hysterectomy specimen.¹ EC uses the FIGO criteria to determine disease stage, which depends on the size of the tumour and the extent to which the tumour has spread to lymph nodes or distant sites (metastasis).²² Generally, the higher the stage number, the more the cancer has spread.²³ Tumour stage is fixed, regardless of tumour type.¹

The prognosis of EC is primarily based on stage of cancer, histology of the tumour, and grade. Five-year survival by FIGO stage is 80% to 90% for stage I, 70% to 80% for stage II, and 20% to 60% for stage III and IV.¹ In terms of histology, there are 2 subtypes: type I ECs represent 80% of patients and are low-grade (1 or 2) endometrioid tumours,^1,20,24 and type II accounts for 10% to 20% of ECs and includes grade 3 endometrioid tumours and tumours of nonendometrioid histology, such as serous clear cell, mucinous, squamous, transitional cell, mesonephric carcinosarcoma, and undifferentiated.^1,20 The 5-year survival of type I is around 80% to 90%, while 5-year survival of type II EC is as low as 20%. Other notable prognostic factors for EC include race, age, uterine tumour location, peritoneal cytology results, and lymphovascular space invasion.¹

Molecular testing of biomarkers during endometrial biopsy assists in the identification of treatment options and in risk stratification.¹ Standard testing includes immunohistochemistry (IHC) and polymerase chain reaction. IHC is used to test for dMMR, in which the cells’ ability to repair mistakes during the division process is impaired. The tumour is immunohistochemically assessed for the loss of at least 1 of the following MMR proteins: MLH1, MSH2, MSH6, and/or PMS2.²⁵ The dMMR proteins cause cellular hypermutations and high levels of microsatellite instability (MSI-H) in sections of DNA. If MMR status cannot be determined from tumour samples, the sample may undergo genetic testing that uses next-generation sequencing (NGS) to identify MSI status. dMMR and MSI-H tumour statuses are predictive of clinical benefit from PD-1 inhibitors and represent approximately 25% of primary ECs and 13% to 30% of recurrent ECs.^7-9 Recurrence occurs in approximately 13% to 20% of patients with EC, with rates varying greatly by FIGO stage at diagnosis; rates are highest among patients with stage IV EC (> 65%).^4,5,26 The prognosis of patients with recurrent EC is poor, with a median survival of approximately 12 months.⁴

Standards of Therapy

Treatment options for EC are dependent on stage and pathologic factors identified after initial surgery and are based on estimated risk of disease recurrence. Early-stage EC and/or type I tumour ECs can be cured with surgery alone.^5,27,28 Individuals diagnosed with advanced or recurrent EC may require adjuvant radiotherapy and/or chemotherapy, depending on the extent and location of spread and/or pathologic risk factors. CCO and Alberta Health Services recommendations favour combination chemotherapy over single-drug chemotherapy for individuals with advanced or recurrent EC, as combination therapy elicits a higher response rate while maintaining acceptable toxicity levels.^5,29-32 The current standard of care for patients with advanced or recurrent disease is platinum-based chemotherapy as a doublet or single drug.²⁹ A standard echoed by the European Society of Gynaecological Oncology, the European Society of Radiotherapy and Oncology, and the European Society of Pathology.³³ The most common platinum-based therapy is carboplatin plus paclitaxel.^5,29,33 For a subset of patients with low-grade recurrent or metastatic EC who are estrogen- or progesterone-receptor positive or for patients with poor tolerance to systemic therapy, hormonal therapy, such as megestrol, letrozole, and medroxyprogesterone, may be used.²⁹

For patients with newly diagnosed advanced or recurrent EC, response rates for standard first-line treatment range from 40% to 62%.^10-13 However, for patients with advanced or recurrent EC who have progressed on or after platinum-based chemotherapy, there is currently no standard effective or curative second-line therapy.^5,33 Patients with recurrent EC are typically re-treated with platinum-based chemotherapy with poor outcomes; response rates range from 10% to 15% for all available treatment options.³³ The clinical experts consulted by CADTH noted that median survival ranges from 12 to 15 months after re-treatment. Various single-drug chemotherapies may be administered to patients who are resistant or refractory to platinum-based chemotherapy, with response rates typically below 15% and no known clear survival benefit.¹⁴ Hormonal treatments may also be used for disease control but are not considered curative.

The clinical experts consulted by CADTH for this review indicated that there is great unmet need for effective therapies with acceptable toxicity profiles that achieve disease control, reduce disease-related symptoms, improve HRQoL, prevent disease progression, and prolong survival among patients with recurrent or advanced EC that has progressed on or after prior treatment with a platinum-containing regimen. There is currently no standard effective second-line therapy for recurrent or refractory disease, and commonly used therapies are noncurative. The clinical experts anticipated more promising benefit with biomarker-driven treatments for patients with dMMR or MSI-H cancers.

Drug

Dostarlimab is an anti-PD-1 monoclonal antibody that targets the cellular pathway between the PD-1 receptor and 2 programmed death ligands, PD-L1 and PD-L2, found on immune cells. Dostarlimab binds to the PD-1 receptor and blocks its interaction with PD-L1 and PD-L2. This suppresses the PD-1-pathway-mediated immune response in the tumour microenvironment, enhancing a patient’s antitumour immune response.¹⁵

On December 23, 2021, dostarlimab was issued a Notice of Compliance with conditions (NOC/c) by Health Canada for the treatment of adults with dMMR or MSI-H recurrent or advanced EC that has progressed on or after prior treatment with a platinum-containing regimen. The sponsor’s requested reimbursement criteria for dostarlimab align with the Health Canada indication. Dostarlimab underwent an expedited Health Canada review process (advance consideration under NOC/c); the market authorization comes with conditions, pending the results of trials that confirm its clinical benefit. Dostarlimab has no other Health Canada–approved indications and has not previously been reviewed by CADTH.

Dostarlimab received accelerated approval from the FDA in August 2021 for the treatment of adults with dMMR or MSI-H recurrent or advanced EC that has progressed on or after prior treatment with a platinum-containing regimen.³⁴ Dostarlimab received conditional authorization in the European Union in April 2021 for the same indication.³⁵ Currently, dostarlimab is being reviewed by NICE and the Scottish Medicines Consortium.^16,17

Dostarlimab is available as a 500 mg IV infusion and is administered with an IV infusion pump over 30 minutes.¹⁵ The recommended dosage of dostarlimab as monotherapy in adults is 500 mg once every 3 weeks for dose 1 through 4, and 1,000 mg once every 6 weeks for dose 5 onward (dose 5 occurs 3 weeks after dose 4). The product monograph states that treatment may continue until disease progression or unacceptable toxicity. Dose reductions of dostarlimab are not recommended, but dosing delays and discontinuation are permitted based on safety and the patient’s tolerability of the treatment. Patients should be selected for treatment based on MSI-H or dMMR tumour status, determined by an accredited laboratory using validated testing methods.

Key characteristics commonly used in the treatment of advanced or recurrent dMMR or MSI-H EC are presented in Table 3.

Table 3: Key Characteristics of Dostarlimab, Platinum-Based Therapy, and Hormonal Therapy Regimens

dMMR = mismatch repair-deficient; EC = endometrial cancer; PD-1 = programmed cell death protein-1; IgG4 = immunoglobulin G4; HRT = hormone replacement therapy; MSI-H = microsatellite instability-high.

^aHealth Canada–approved indication

Source: Application overview, product monograph for dostarlimab,¹⁵ carboplatin,³⁶ paclitaxel³⁷ and medroxyprogesterone acetate.³⁸

Stakeholder Perspectives

Patient Group Input

The information in this section is a summary of input provided by the patient groups that responded to CADTH’s call for patient input for the purpose of this CADTH review. The full patient input received is available in the Stakeholder Input section.

The patient and caregiver input received for this review was collected by the Canadian Cancer Society. The input was sourced from on an online survey and from patient and caregiver testimonials gathered from October 22 to November 3, 2021. Six testimonials and 22 survey responses were received (20 patients with current or previous EC, and 2 caregivers of someone with current or previous EC). Of the 22 survey respondents, 8 (6 patients and 2 caregivers) had experience with dostarlimab. Six of these respondents resided in Quebec and 2 resided in British Columbia. All patients and caregivers who had experience with dostarlimab reported receiving it through a clinical trial.

Respondents indicated a range of EC symptoms that affected their daily activities. The daily activities that were most commonly moderately or severely affected included the ability to preform household chores (46%), travel (41%), exercise (41%), work (36%), fulfill family obligations (32%) and spend time with family and friends (27%). Forty-four percent of patients were not currently being treated for EC and 39% were undergoing immunotherapy. The treatment side effects most commonly reported as having a moderate or severe impact on daily life were issues with libido and sexual function (45%) and fatigue (41%). Fifty-nine percent of patients reported a financial barrier related to their treatment; among these patients, a loss of income due to absence from work (31%) and travel costs for cancer treatment (31%) were the most common barriers.

According to the patient input received, respondents expect the following key outcomes to be improved with any new drug or treatment: quality of life, periods of remission, drug affordability, access across jurisdictions, and fewer side effects (such as skin issues, fatigue, bladder control, stamina, hair loss, pain, arthritis, vaginal dryness, vaginal bleeding after intercourse, and concentration problems). All survey respondents indicated that, compared with other therapies, dostarlimab was easier to use either because it had little to no side effects (75%), longer intervals between doses (13%), or a shorter infusion time (13%).

Clinician Input

Input From Clinical Experts Consulted by CADTH

All CADTH review teams include at least 1 clinical specialist with expertise in the diagnosis and management of the condition for which the drug is indicated. Clinical experts are a critical part of the review team and are involved in all phases of the review process (e.g., providing guidance on the development of the review protocol; assisting in the critical appraisal of clinical evidence; interpreting the clinical relevance of the results; and providing guidance on the potential place in therapy). The following input was provided by 2 clinical specialists with expertise in the diagnosis and management of EC.

Unmet Needs

Currently, there is no standard second-line therapy for individuals with recurrent EC who have progressed on or after prior treatment with a platinum-containing regimen. This represents a critical unmet need in this patient population. Both clinical experts agreed that biomarker-driven treatments are needed to guide selection of the most effective and durable treatment option.

Place in Therapy

Currently, patients with advanced or recurrent EC receive carboplatin plus paclitaxel (another platinum chemotherapy doublet) or a single drug as first-line therapy. For a subset of patients with estrogen- or progesterone-receptor-positive indolent low-grade metastatic or recurrent EC, hormonal treatments, such as megestrol, medroxyprogesterone, letrozole, and tamoxifen, are considered. For those who have progressed on platinum chemotherapy and are considered to have platinum-resistant or refractory disease, second-line therapy with single-drug chemotherapy is considered, but it has a low expected response rate and a short DOR with no known survival benefit. Hormonal treatments may also be used for disease control but are not considered curative.

Dostarlimab will likely cause a shift in the current treatment paradigm for patients with dMMR or MSI-H metastatic or recurrent EC. The mechanism of action of dostarlimab would address the underlying disease-specific process and biomarkers for patients with dMMR or MSI-H EC. The clinical experts felt it would be preferable to initiate treatment with dostarlimab before other therapies.

Patient Population

The clinical experts agreed that patients with dMMR or MSI-H EC tumours would most benefit from an immune checkpoint inhibitor. One clinical expert added that patients with dMMR or MSI-H metastatic or recurrent EC of any histology would benefit from dostarlimab, regardless of symptoms or prior treatment; however, the other clinical expert noted that a smaller magnitude of benefit may be observed with increased previous lines of systemic therapy.

The clinical experts mentioned that the diagnosis of EC typically relies on biopsy, which may be conducted in clinic. If technical issues make it difficult to obtain samples from the endometrial lining during an office biopsy, the procedure can be performed in the operating room using dilatation and curettage. IHC testing for MMR status is relatively inexpensive and considered standard practice. If the MMR status of a sample is unknown, it may be further analyzed with genomic testing to determine MSI status. A valid test would involve NGS.

The clinical experts noted that treatment with dostarlimab would not be suitable for patients any of the following characteristics:

• very poor performance status

• history of severe autoimmune disease

• prior therapy with anti-PD-1, anti-PD-L1, or anti-PD-L2 drugs

• known uncontrolled nervous system metastases and/or carcinomatous meningitis

• poor medical risk due to serious uncontrolled medical disorder

• nonmalignant systemic disease or active infection requiring systemic therapy

• history of immunosuppression

• MSS EC.

Assessing Response to Treatment

According to the clinical experts, the most important goals of treatment for recurrent or advanced EC would be to improve OS and PFS; reduce symptoms of cancer; improve functional status (i.e., ability to perform activities of daily living); improve HRQoL; and reduce the burden of disease on patients and caregivers. One clinical expert noted that many patients with advanced or recurrent EC suffer from pelvic symptoms (e.g., unresectable disease in the pelvis causing bleeding or pain), lung symptoms (e.g., dyspnea from metastases), and neurologic symptoms (e.g., brain metastases) or bone symptoms (e.g., painful bony metastases), and that alleviation of these symptoms would be a benefit of treatment.

In terms of a clinically meaningful response, the clinical experts recommended that in addition to clinical assessment of disease symptoms and duration of disease control, the use of standard irRECIST³⁹ for the assessment of response to immunotherapeutic treatments is useful. One clinical expert noted that the Common Terminology Criteria for Adverse Events⁴⁰ tool for CT imaging can be used to assess response to treatment. The same clinical expert also suggested that cancer antigen 125 (nonspecific) may be used to assess treatment response in addition to other methods, although it is not commonly used in EC. In terms of the timing of assessments, the clinical experts recommend that response to treatment should be assessed radiologically every 3 months, with blood work every month and clinical assessments every 2 to 3 months.

Discontinuing Treatment

According to the clinical experts, treatment with dostarlimab should be discontinued when there are radiological and clinical signs and symptoms of disease progression, treatment toxicities (e.g., grade 3 or higher adverse reactions), or intolerability to treatment.

Prescribing Conditions

According to the clinical experts consulted by CADTH, the diagnosis, treatment, and monitoring of patients with EC should be undertaken by a specialist, namely a gynecologist oncologist, medical oncologist, and/or surgeon. Biomarker testing to identify dMMR or MSI-H status is also recommended.

Clinical experts recommend that dostarlimab be administered in a hospital clinic that has multidisciplinary medical supports to manage potential immune-related side effects. The clinical experts noted that treatment may also be administered in outpatient clinics.

Additional Considerations

The clinical experts noted that re-treatment with dostarlimab is possible if patients experience recurrence, as long as they had no signs of toxicity or intolerability while using the drug. If a patient completed their treatment and achieved a durable response for up to 2 years, and a significant time period elapsed before they progressed, then re-treatment with dostarlimab could be considered. The clinical experts noted that 50% of patients will have disease recurrence within 1 year of completing treatment for recurrent EC and will need to restart treatment.

Clinician Group Input

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

The information in this section is a summary of 7 inputs provided by the registered-clinician groups that responded to CADTH’s call for clinician input for the purpose of this CADTH review. The full clinician group inputs received are available in the Stakeholder Input section. Input was received from the following clinical groups:

• British Columbia Cancer Provincial Gynecological Oncology Tumour Group

• MUHC, Division of Gynecologic Oncology

• OH-CCO Gynecological Drug Advisory Committee

• PMCC, Gynecologic Cancers Disease Site Group, Medical Oncology Group

• SCA

• GOC

• SBHSC, Division of Gynecologic Oncology.

Unmet Needs

The views of the clinician groups were overall consistent with the clinical experts consulted by CADTH, indicating that the most important treatment goals for advanced or recurrent dMMR or MSI-H EC are disease control, prolonged survival, delayed worsening of symptoms, maintenance of HRQoL, delayed disease progression, and an acceptable safety profile. Treatment for individuals diagnosed with recurrent EC is a critical unmet need. All clinician groups noted that although the prognosis for patients diagnosed with early-stage disease is good, for those with recurrent or metastatic EC, median OS is short. In fact, there is no effective second-line treatment for Canadians with EC. Response rates to second-line cytotoxic drugs are low (< 20%), with a median PFS of 3 to 4 months.

All clinician groups also noted that although Health Canada recently issued a NOC/c approval for pembrolizumab immunotherapy for dMMR or MSI-H EC, there is no funded access to this treatment. The lack of funding has created disparity in access to immunotherapy, as only those with insurance coverage or the capacity to self-pay can access pembrolizumab. Moreover, PMCC added that there is currently no patient-supported or compassionate-access program for immune checkpoint inhibitor therapies. The MUHC and SBHSC noted that the side-effect profile of combination therapy with pembrolizumab and lenvatinib was significant, with 66.9% of patients experiencing a grade 3 or 4 toxicity. The MUHC added that prolonging poor HRQoL with interventions associated with significant toxicity serves very little purpose.

Place in Therapy

All clinician groups indicated that all patients with advanced or recurrent EC would benefit from effective immunotherapy, but patients with MSI-H or dMMR subtypes would most benefit from treatment with an immune checkpoint inhibitor. All groups recommended that patients diagnosed with advanced or recurrent EC should be offered platinum-based chemotherapy as first-line therapy. However, the British Columbia Cancer Provincial Gynecological Oncology Tumour Group did acknowledge that treatment with immune checkpoint inhibitors may be an appropriate first-line therapy for patients who either refuse chemotherapy or for whom chemotherapy may be contraindicated or known to be poorly tolerated, provided dMMR or MSI-H status is confirmed. The MUHC further stated, “it is false economy to delay starting a highly effective treatment [dostarlimab] with minimal toxicity in this niche population.” The MUHC stated that trying second-line chemotherapy alone in this patient population is not advisable due to toxicity and lack of benefit.

MUHC indicated that when EC recurs in patients who initially responded well and dostarlimab is interrupted for any other reason, restarting treatment with dostarlimab would be appropriate. If patients progress on dostarlimab or other immunotherapies, palliative and supportive care options are considered.

Patient Population

Currently, the standard of care in Canada includes universal molecular characterization of all endometrial carcinomas for dMMR. Thus, identifying EC subtypes should not be a barrier to treatment. The British Columbia Cancer Provincial Gynecological Oncology Tumour Group and the GOC noted that patients whose EC is related to Lynch syndrome would benefit from treatment with dostarlimab, and PMCC added that those with polymerase-epsilon-mutation would benefit as well. The SCA noted that patients with EC suitable for treatment with dostarlimab should be PD-L1-naive and should not have undergone more than 2 prior treatments with a platinum-based therapy.

All clinician groups noted that patients who do not have a dMMR or MSI-H profile and/or have contraindications to immune checkpoint inhibitors are not suitable for treatment with dostarlimab. PMCC and the GOC also indicated that patients with MMR-proficient EC would not be suitable to receive dostarlimab as monotherapy treatment.

SBHSC noted that the EC patient groups less suitable for treatment with dostarlimab would be those that met all the following criteria:

• no previous treatment with platinum-based chemotherapy

• no dMMR seen on IHC

• poor medical risk due to a serious uncontrolled medical disorder

• poor performance status (ECOG Performance Status score of at least 3)⁴¹

• inadequate organ function

• a known immunodeficiency or current use of systemic steroids or other immunosuppressant medications.

Assessing Response to Treatment

The British Columbia Cancer Provincial Gynecological Oncology Tumour Group, COG, OH-CCO, and SBHSC noted that improvement in symptoms and physical findings related to advanced or recurrent EC (e.g., pain, bleeding, shortness of breath) would be considered a clinically meaningful treatment outcome. OH-CCO and PMCC stated that durable disease control with no adverse effects on HRQoL is an ideal treatment goal in this population. The SBHSC specified that PFS of at least 6 months would be a clinically meaningful outcome in this patient population. In addition, OH-CCO listed reduced caregiver burden as a meaningful treatment outcome.

The British Columbia Cancer Provincial Gynecological Oncology Tumour Group and the GOC also listed the following outcomes as clinically meaningful responses to treatment:

• maintenance or improvement of performance status and ability to perform activities of daily living

• evidence of disease regression from imaging studies

• disease stabilization (in patients with good baseline performance status and few disease-related symptoms).

OH-CCO and the GOC reported that standard clinical monitoring of therapy, including a physical examination, symptom review, and intermittent CT imaging, should be used to evaluate response to treatment. The MUHC recommended that after the first 3 cycles of dostarlimab, patients should be assessed for response to treatment with a CT scan. If there is no indication of progression and patients are well, the CT-monitoring interval may be spaced out to every 12 weeks. PMCC recommended that tumour assessment by CT scan or MRI be completed every 2 to 3 cycles (i.e., every 6 to 9 weeks). The British Columbia Cancer Provincial Gynecological Oncology Tumour Group and SBHSC stated that CT imaging should occur every 12 weeks, whereas the SCA suggested imaging every 3 to 6 months. The British Columbia Cancer Provincial Gynecological Oncology Tumour Group and SCA also added that laboratory values, as well as respiratory and pulmonary status, should be assessed to ensure that treatment response is occurring without toxicity.

Discontinuing Treatment

All clinician groups indicated that treatment with dostarlimab should be discontinued if patients experience disease progression or serious toxicities. SBHSC indicated that the following side effects warrant consideration of pausing or discontinuing dostarlimab:

• grade 2 anemia

• pneumonitis

• grade 3 colitis

• grade 2 asthenia

• grade 3 myalgia

• pemphigoid

• grade 3 increase in transaminases.

PMCC and SBHSC noted that due to the nature of immunotherapies and the possibility of pseudo-progression, patients with progression of disease on the first CT scan after initiation of treatment, but with no other symptoms, should continue treatment until further imaging demonstrates progression of disease. Similarly, the MUHC adds that in cases where disease progression occurs after 3 cycles with dostarlimab, patients should receive follow-up in 6 weeks to rule out pseudo-progression and discontinue treatment if progression is confirmed.

Prescribing Conditions

The British Columbia Cancer Provincial Gynecological Oncology Tumour Group, MUHC, OH-CCO, GOC, and SCA explained that dostarlimab is suitable to be delivered in the community and in outpatient and specialty clinics. Conversely, PMCC and SBHSC suggested that dostarlimab be delivered at cancer centres by gynecologic or medical oncologists. Both groups recommend that dostarlimab be administered in a chemotherapy suite with appropriate supervision by an oncologist familiar with gynecologic cancers and the management of immune-related adverse effects. The British Columbia Cancer Provincial Gynecological Oncology Tumour Group and the GOC noted that patients should be under the care of a treating physician or nursing staff with experience monitoring and evaluating patients for possible toxicities that may be caused by immune checkpoint inhibitor therapy. The GOC stated that this should not be a barrier to treatment because such therapies are now routinely used by most oncologists. The MUHC notes that for patients who do not live close to a treatment centre (e.g., patients in rural, remote, or Indigenous communities), it is appropriate to work closely with local physicians to share the responsibility of patient care.

Drug Program Input

The drug programs provide input on each drug being reviewed through CADTH’s reimbursement review processes by identifying issues that may impact their ability to implement a recommendation. Implementation questions from the drug programs and corresponding responses from the clinical experts consulted by CADTH are summarized in Table 4.

Table 4: Summary of Drug Plan Input and Clinical Expert Response

dMMR = mismatch repair-deficient; DOR = duration of response; EC = endometrial cancer; ECOG = Eastern Cooperative Oncology Group performance status; IHC = immunohistochemistry; MMR = mismatch repair; MSI = microsatellite instability; MSI-H = microsatellite instability-high; MSS = microsatellite stable; NGS = next-generation sequencing; OS = overall survival; PD-1 = programmed cell death protein-1; PD-L1 = programmed death ligand-1; PFS = progression-free survival.

Clinical Evidence

The clinical evidence included in the review of dostarlimab is presented in 2 sections. The first section, the systematic review, includes pivotal studies provided in the sponsor’s submission to CADTH and Health Canada, as well as studies that were selected according to the a priori CADTH 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 CADTH review.

Systematic Review (Pivotal and Protocol-Selected Studies)

Objectives

To perform a systematic review of the beneficial and harmful effects of dostarlimab in 500 mg to 1,000 mg doses administered intravenously for the treatment of adults with dMMR or MSI-H advanced or recurrent EC that has progressed on or after prior treatment with a platinum-containing regimen.

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. The systematic review protocol was established before the granting of the NOC/c from Health Canada. Outcomes included in the CADTH review protocol reflect outcomes considered to be important to patients, clinicians, and drug plans.

Table 5: Inclusion Criteria for the Systematic Review

AEs = adverse events; BOR = best overall response; DCR = disease control rate; dMMR = mismatch repair-deficient; DOR = duration of response; EC = endometrial cancer; FIGO = International Federation of Gynecology and Obstetrics; HRQoL = health-related quality of life; ICU = intensive care unit; MSI-H = microsatellite instability-high; ORR = objective response rate; OS = overall survival; PFS = progression-free survival; PLD = pegylated liposomal doxorubicin; RCT = randomized control trial; SAEs = serious adverse events; WDAEs = withdrawal due to adverse events.

^aOnly indicated for patients with hormone-therapy-positive and low-grade endometrioid cancer.

^bApproved by Health Canada but not approved for reimbursement.

^cThese outcomes were identified as being of particular importance to patients in the input received by CADTH from patient groups.

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.⁴²

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 dostarlimab. Clinical trials registries were searched: the US National Institutes of Health’s clinicaltrials.gov, WHO’s International Clinical Trials Registry Platform (ICTRP) search portal, Health Canada’s Clinical Trials Database, 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. Refer to Appendix 1 for the detailed search strategies.

The initial search was completed on November 3, 2021. Regular alerts updated the search until the meeting of the CADTH pan-Canadian Oncology Drug Review Expert Committee (pERC) on March 9, 2022.

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.⁴³ Included in this search were the websites of regulatory agencies (FDA and European Medicines Agency). Google was used to search for additional internet-based materials. Refer to Appendix 1 for more information on the grey literature search strategy.

These searches were supplemented with a review of bibliographies of key papers and contacts with appropriate experts. In addition, the manufacturer of the drug was contacted for information regarding unpublished studies. Two CADTH clinical reviewers independently selected studies for inclusion in the review based on titles and abstracts, according to the predetermined protocol. Full-text articles of all citations considered potentially relevant by at least 1 reviewer were acquired. Reviewers independently made the final selection of studies to be included in the review, and differences were resolved through discussion.

A focused literature search for network meta-analyses dealing with EC was run in MEDLINE All (1946–) on November 3, 2021. No limits were applied to the search.

Findings From the Literature

Two reports of 1 study^18,44 were identified from the literature for inclusion in the systematic review (Figure 1); the study is summarized in Table 6. A list of excluded studies is presented in Appendix 2.

Figure 1: Flow Diagram for Inclusion and Exclusion of Studies

Description of Study

The GARNET trial (Study 4010-01-001)¹⁸ is an ongoing multi-centre, single-arm, open-label, phase I dose-escalation and cohort-expansion study of patients with recurrent or advanced solid tumours. The primary objective of part 2B of the GARNET trial was to evaluate the safety and antitumour activity of dostarlimab in patients with advanced solid tumours. The secondary objective was to characterize the pharmacokinetic profile and to evaluate immunogenicity and additional measures of clinical benefit of dostarlimab. Exploratory objectives were to characterize the pharmacodynamics of dostarlimab and patient-reported outcomes assessed with EQ-5D-5L and the EORTC QLQ-C30. Patients were assigned to different cohorts based on variety of factors, including cancer type, response to previous therapies, and/or clinical biomarkers. Cohort A1 included patients with recurrent or advanced dMMR or MSI-H EC who had progressed on or after prior treatment with a platinum-containing regimen. Cohort A2 included patients with recurrent or advanced MMR-proficient or MSS EC that had progressed on or after prior treatment with a platinum-containing regimen. Cohort F included patients with nonendometrial dMMR, MSI-H, or polymerase-epsilon-mutation cancers who had progressed after 2 prior lines of systematic therapy for recurrent or advanced disease and had no alternative treatment options. For the purpose of this review, only the results of cohort A1, which aligns with the Health Canada indication, are presented. Refer to Figure 2 for a schematic of the GARNET trial.

A total of 129 patients were enrolled into cohort A1. Enrolment started on April 10, 2017 and is ongoing. Patients were enrolled across 123 sites in 8 countries (refer to Table 6). Of the 123 sites, 8 were located in Canada. Refer to Figure 2 for a schematic of the GARNET trial. All patients received dostarlimab by IV injection (500 mg every 3 weeks for cycles 1 to 4, and 1,000 mg every 6 weeks from cycle 5 onward) for up to 2 years or until disease progression, treatment discontinuation, or study withdrawal. Patients were permitted to continue treatment with dostarlimab beyond 2 years if the treating physician and the sponsor agreed that the treatment continued to provide clinical benefit to the patient.

To be enrolled, patients had to have archival or newly obtained tumour tissue so the tumour microenvironment could be assessed for biomarkers; in cohort A1, this included IHC results from a certified local laboratory that could be used to determine dMMR status. The definition of dMMR was the loss of expression of 1 of the following proteins, determined with IHC testing: MLH1, MSH2, MSH6, and PMS2. In the absence of known MMR status, MSI-H was tested using NGS in a central laboratory. Assay results assessed with local polymerase chain reaction were also acceptable for the determination of MSI status.

The study consisted of 3 phases: the screening phase (up to 35 days before beginning treatment), the treatment phase, and the follow-up phase. During the screening phase, 479 patients were screened for enrolment in either cohort A1 or cohort A2, based on MMR/MSI status. There were 184 screening failures, and 5 patients were screened but not treated. During the follow-up phase, patients were followed for safety (final follow-up visit was 90 ± 7 days after the end of treatment [EOT]), as well as for OS and HRQoL. The date of the final analysis for OS had not been determined at the time of the data cut-off (March 1, 2020).

The GARNET trial has 3 planned interim analyses that are contingent on a combined enrolment of 100, 200, and 300 patients, respectively, in cohort A1 and cohort F. For IA-1, the data cut-off date was July 8, 2019, after a 100 patients with measurable disease at baseline and at least 24 weeks of follow-up had been enrolled in either cohort A1 or cohort F. For IA-2, the data cut-off date was March 1, 2020, after the combined patient enrolment of cohort A1 and cohort F reached approximately 200 patients with measurable disease at baseline and at least 24 weeks of follow-up. For IA-3, the data cut-off date was November 1, 2021, after the combined patient enrolment in cohort A1 and cohort F reached approximately 300 patients with measurable disease at baseline and at least 24 weeks of follow-up. Only data from the IA-2 data cut-off date were reviewed by CADTH, which included a subset of 72 patients from IA-1 (refer to Appendix 4 for further details). The GARNET trial was sponsored by GlaxoSmithKline Inc.

Amendments and Protocol Deviations

The study protocol of the GARNET trial was amended 6 times, although versions 4.0 and 5.0 were not implemented. Amendments to the protocol did not substantially affect the results of the study.

Amendment 2 (October 31, 2016) included updates to the inclusion criteria, tumour assessment criteria, definitions for safety data, and sample size justification of cohort A1 and cohort A2.

Amendment 3 (October 9, 2017) included the addition of HRQoL assessments and interim analyses for patients with dMMR or MSI-H cancer in cohort A1 and cohort F combined.

Amendment 4 (July 3, 2018) included revisions to allow clinically stable patients without major safety issues to continue with dostarlimab treatment after confirmation of progressive disease. Changes were also made to blood sample collection and to increase the number of patients in cohort A2.

Amendment 5 (May 10, 2019) implemented changes to the definition of cohort A1 and cohort A2, based on feedback from a regulatory agency. A recommendation was made to select patients based on the MMR IHC result (local or central test) instead of NGS; hence, the central testing vendor was changed. The sample size for cohort A1 was increased and the assessment of HRQoL was revised from a secondary to an exploratory objective.

Amendment 6 (January 7, 2020) updates to the primary objective included evaluation of the safety and tolerability of dostarlimab, as well as the addition of 2 interim analyses for the combined enrolment of cohort A1 and cohort F to reach 200 and 300 patients, respectively, and for all patients to have measurable disease at baseline and at least 24 weeks of follow-up.

Figure 2: Study Schema

DLT = dose-limiting toxicity; dMMR = mismatch repair-deficient; MMRp = mismatch repair proficient; MSI-H = microsatellite instability-high; MSS = microsatellite instability stable; mut = mutation; NSCLC = non–small cell lung cancer; PDy = pharmacodynamics; PK = pharmacokinetics; POLE = polymerase epsilon; PROC = platinum-resistant ovarian cancer; ; Q3W = every 3 weeks; Q6W = every 6 weeks; RP2D = recommended phase II dose.

Source: Clinical Study Report.¹⁸

Populations

Inclusion and Exclusion Criteria

The key inclusion and exclusion criteria used in cohort A1 of the GARNET trial are summarized in Table 4. Briefly, the trial enrolled adults aged 18 years and older who had been diagnosed with recurrent or advanced EC and who progressed on or after no more than 2 lines of prior systemic therapy (not including hormone therapy), with at least 1 of these being platinum-based doublet therapy. Patients had to have documentation of dMMR or MSI-H status and radiologically measurable disease that met Response Evaluation Criteria in Solid Tumors 1.1 (RECIST 1.1).⁴⁵ All EC histologies were permitted, except for endometrial sarcoma (including carcinosarcoma). At screening, patients had to have an ECOG Performance Status of 0 or 1, adequate organ function, a negative serum pregnancy test (if of childbearing potential), and 1 highly effective form of contraception throughout the study period. Prior therapy with an anti-PD-1, anti-PD-L1, or anti-PD-L2 drug was not permitted.

Baseline Characteristics

The baseline characteristics of patients who comprised the primary efficacy population of the GARNET trial are summarized in Table 7. At baseline, 105 patients in cohort A1 were identified as having dMMR or MSI-H EC (103 dMMR and 2 MMR-unk/MSI-H EC patients) for the primary efficacy analysis at IA-2. The demographic and baseline characteristics of patients with dMMR or MMR-unk/MSI-H tumours are presented as a total because of the small proportion of patients (n = 2) with MMR-unk/MSI-H tumours; the sponsor noted that the data remained similar when these patient groups were pooled. The median age was 64 years (range = 39 to 80 years), with 50.5% of patients between 39 and 65 years. Median weight was 71 kg (range = 34.0 to 141.4), and median BMI was 27.97 kg/m² (range = 13.6 to 53.9 kg/m²). Most patients were White (78.1%). The most common histology type was type I endometrioid carcinoma (67.6%); the most common histology grade at diagnosis was grade 2 (39.0%). At the time of study enrolment, more than two-thirds (67.6%) of patients had FIGO stage IV EC. An ECOG Performance Status of 1 was the most common (60%), followed by an ECOG Performance Status of 0 (40%). All patients received prior anti-cancer treatment that included surgery (90.5%), radiotherapy (70.5%), adjuvant or neoadjuvant treatment (53.3%), and/or bevacizumab use (4.8%). Most patients (88.6%) had previously received 2 or fewer lines of anti-cancer regimens; a small portion had previously received 2 or more lines (11.5%). More than half (56.2%) of patients had received prior regimens for metastatic disease.

Table 7: Summary of Baseline Characteristics of Patients in the GARNET Trial (IA-2, Primary Efficacy Analysis Dataset)

BMI = body mass index; ECOG = Eastern Cooperative Oncology Group; FIGO = International Federation of Gynecology and Obstetrics; IA-2 = second interim analysis.

^aPopulation total (n = 105) includes 103 dMMR and 2 MMR-unk/MSI-H patients. Data from patients with dMMR tumours and MMR-unk/MSI-H tumours are presented as a total due to the small proportion of patients (n = 2) with MMR-unk/MSI-H tumours; the sponsor noted that data remained similar when these patient groups were pooled together.

^bOther may include adenocarcinoma; adenocarcinoma with ambiguous differentiation; biopsies that showed a high-grade adenocarcinoma, which can be seen in tumours of Mullerian origin; carcinoma epidermoid; endometrial adenocarcinoma; endometrial adenocarcinoma not otherwise specified; endometrial neuroendocrine carcinoma; endometrioid adenocarcinoma; high-grade uterine carcinoma; moderately differentiated adenocarcinoma; papillary serous carcinoma; and undifferentiated clear cell carcinoma.

^cIncludes surgery, radiotherapy, or drug therapy, excluding hormonal drugs.

^dDoes not include neoadjuvant regimens, adjuvant regimens, or hormonal drugs.

Source: Clinical Study Report.¹⁸

Interventions

All patients in cohort A1 received dostarlimab as a 30-minute IV infusion (with a permitted window of –5 minutes and + 15 minutes). Doses were administered in hospitals, infusion centres, and outpatient clinics. Dosing schedules were as follows:

• cycles 1 to 4: 500 mg every 3 weeks (day 1 of each 21-day cycle)

• cycle 5 and onward: 1,000 mg every 6 weeks (day 1 of each 42-day cycle).

Patients continued with the study treatment for up to 2 years or until specific withdrawal criteria were met. Patients discontinued study treatment for the following reasons: AEs, including dose-limiting toxicities; progressive disease based on clinical criteria used by the investigator or outlined in the protocol; patient request or patient pregnancy; risk to patients, determined by the investigator and/or sponsor; and severe noncompliance with the protocol, determined by the investigator and/or sponsor. In some instances, study treatment continued beyond 2 years if the treating physician and the sponsor agreed that treatment with dostarlimab continued to provide clinical benefit to the patient.

Dose Modifications (Delay, Interruption)

Dosing delays were permitted in the case of medical or surgical events or for logistical reasons not related to study treatment (e.g., surgery, unrelated medical events, participant vacation). Continuation of the study treatment should have occurred within 28 days of the scheduled dostarlimab infusion. If a delay was more than 28 days, a patient may have been permitted to continue treatment after discussion with the sponsor.

Dosing interruptions were permitted in the event of toxicities. In general, treatment withheld for drug-related grade 3 toxicities could be resumed if toxicity resolved to grade 1 or lower. Similarly, for all irAEs, dostarlimab was held until the patient was considered clinically and metabolically stable and AEs resolved to grade 1 or lower. Dostarlimab was permanently discontinued after any drug-related grade 4 events and after some grade 3 immunologic-mediated AEs.

Dose reductions were not permitted.

Concomitant Medications

Concomitant medications were allowed to treat AEs and comorbidities. Concomitant medications included any medications other than study treatments taken on or after the initial study treatment dosing date. If systemic steroids were used as a part of irAE management, the total dose of daily steroids allowed was 10 mg or less of prednisone when dostarlimab treatment was resumed. Patients were also allowed to receive rescue medications and appropriate supportive care deemed necessary by the treating investigator.

To ensure accurate assessment of the safety and efficacy of dostarlimab, patients were prohibited from receiving the following therapies during the screening and treatment periods of the study:

• systemic anti-cancer or biological therapy

• immunotherapy not specified in the protocol

• chemotherapy not specified in the protocol

• investigational drugs other than dostarlimab

• radiation therapy within 3 weeks before study day 1 and during study treatment

  • note: palliative radiation therapy to a small field more than 1 week before day 1 of study treatment may have been allowed

• any surgery that involved tumour lesions

  • note: radiation therapy or surgery that involved tumour lesions was considered to be progressive disease at the time the procedure was performed.

Subsequent Therapies

Data were collected on whether patients received subsequent anti-cancer therapy after treatment discontinuation or progressive disease. However, these were not defined in the protocol or statistical analysis plan.

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 8. These end points are further summarized below. A more detailed discussion of the HRQoL and symptom severity outcome measures assessed in the trial is provided in Appendix 3.

Response assessment, according to RECIST 1.1,⁴⁵ was based on radiologic imaging and performed by a BICR at pre-specified time points for the outcomes of OS, DCR, PFS, DOR, and BOR. The first radiographic evaluation occurred 12 weeks (84 ± 10 days) after the first dostarlimab dose and every 6 weeks (42 ± 10 days) thereafter, independent of cycle delays and/or dose interruptions. After 48 weeks of radiographic assessments, imaging and assessment of serum-based tumour markers were performed every 12 weeks (84 ± 10 days) until progressive disease. If a patient discontinued treatment for any reason other than progressive disease, death, withdrawal of consent, or loss to follow-up, radiographic scans and appropriate testing of serum-based tumour markers continued at the specified intervals until progressive disease was confirmed or an alternate anti-cancer therapy was started, whichever occurred first.

Tumour response was also evaluated by the investigator based on irRECIST 1.1^46,47 for the outcomes of immune-related DCR (irDCR), immune-related PFS (irPFS), immune-related ORR (irORR), and immune-related (irDOR). Investigators used irRECIST-based assessment to make clinical decisions. Per irRECIST, immune-related complete response (irCR) or immune-related partial response (irPR) was confirmed with repeat radiographic evaluation, at the earliest, 4 weeks after the first indication of response or at the next scheduled scan, whichever was clinically indicated. Progressive disease was confirmed with radiographic evaluation a minimum of 4 weeks and up to 6 weeks after the first progressive disease assessment.

Overall Survival

OS was a secondary outcome of the GARNET trial and was defined as the time from the first dose of study treatment to death by any cause. Patients last known to be alive were censored at the date of the last known contact, as follows: OS in days = date of death or censoring – date of first dose + 1.

OS was assessed at EOT, at the safety follow-up visit, and every 3 months after treatment.

Disease Control Rate and Immune-Related Disease Control Rate

DCR and irDCR were secondary outcomes of the GARNET trial. DCR, per RECIST 1.1., was defined as the proportion of patients achieving BOR of confirmed CR, PR, or stable disease. irDCR, per irRECIST 1.1., was defined as the proportion of patients achieving irBOR or irCR. irPR or immune-related stable disease as assessed by the investigator.

BOR was programmatically derived, based on reported time point response assessments determined by a central reader at different evaluation time points, from the first dose until the first documented progressive disease. The irBOR, according to irRECIST, was programmatically derived, based on reported overall time point responses determined by investigators after assessment of radiological scans, at different evaluation time points from the first dose until documented disease progression.

BOR was determined according to the following rules:

• CR — at least 2 consecutive determinations of CR more than 4 weeks apart, with no other assessment between the 2 determinations other than not evaluable, CR, or missing.

• PR — at least 2 consecutive determinations of PR or better more than 4 weeks apart, with no other assessment between the 2 determinations other than not evaluable, CR, PR, or missing before disease progression (and not qualifying for a CR).

• Stable disease — at least 1 stable disease or non-CR or nonprogressive disease assessment (or better) at least 12 weeks to 10 days (i.e., ≥ 74 days) after baseline and before progressive disease (and not qualifying for a CR or PR). For an unconfirmed PR or CR to qualify, it must still meet the requirement of at least 12 weeks to 10 days.

• Progressive disease — disease progression after baseline. Note that a determination of CR followed at least 4 weeks later by stable disease resulted in a BOR of progressive disease.

• No disease — for central imaging data, when the independent radiologist cannot identify any disease at baseline, all subsequent assessments were documented as no disease if not declared progressive disease or not evaluable.

irBOR followed similar rules, except outcomes were immune-related (e.g., irCR, irPR). Clinical deterioration was not considered to be progressive disease. Only tumour assessments performed before the start of any new anti-cancer treatment were considered in the assessment of BOR and irBOR.

Progression-Free Survival and Immune-Related Progression-Free Survival

PFS and irPFS were secondary outcomes of the GARNET trial. PFS was defined as the time from the first dose to the earlier assessment of progressive disease or to death by any cause in the absence of progressive disease. irPFS time was defined, per irRECIST 1.1, as the time from the first dose to the earlier date of assessment of immune-related disease progression (irPD) event or to death by any cause in the absence of disease progression.

irPD events were considered when:

• 2 consecutive irPDs were observed; the date of the first was considered to be the date of the irPD event.

• the only or most recent tumour assessment before treatment discontinuation was irPD; the date was considered to be the date of the irPD event.

PFS and irPFS times were defined as follows: PFS or irPFS in days = date of PD or irPD event or death ÷ censoring – date of first dose + 1 (refer to Table 9 for details on censoring rules for PFS and irPFS).

Objective Response Rate and Immune-Related Objective Response Rate

ORR was the primary outcome of the GARNET trial and was defined, per RECIST 1.1 based on BICR, as the proportion of patients that achieved a BOR of CR or PR. irORR was a secondary outcome of the GARNET trial and was defined, per irRECIST 1.1, as the proportion of patients that achieved irBOR of irCR or irPR. Nonresponders were patients who did not have a post-baseline radiographic tumour assessment, who received post-baseline antitumour treatments (including surgery or radiation to tumour lesions) other than the study treatments before reaching a CR or irCR or a PR or irPR, or who died, progressed, or dropped out for any reason before reaching a CR or irCR or a PR or irPR.

Duration of Response and Immune-Related Duration of Response

DOR was a co-primary outcome of the GARNET trial and was defined as the time from first documentation of overall response leading to a confirmed CR or PR until the time of first documentation of disease progression or death. Censoring rules for DOR were similar to those for PFS (refer to Table 9 for details on censoring rules). irDOR was a secondary outcome of the GARNET trial and was defined, per irRECIST 1.1, as the time from first documentation of response leading to a confirmed irCR or irPR until the time of the irPD event or death. As with BOR and irBOR, clinical deterioration was not considered to be documented progressive disease.

Health-Related Quality of Life

The HRQoL outcomes measured in the trial included EORTC QLQ-C30 and EQ-5D-5L scores. Data from both instruments were exploratory outcomes in the GARNET trial and assessed at IA-2. HRQoL assessments were collected at scheduled visits (at each cycle) every 3 weeks (± 7 days) for the first 12 weeks, beginning on cycle 1, day 1, and every 6 weeks (± 7 days) thereafter while the patient was receiving study treatment. After treatment discontinuation, HRQoL assessments were collected from the remaining patients at the EOT visit, at the safety follow-up visit (90 days after EOT), and every 90 days (± 14 days) during the post-treatment follow-up period. A detailed discussion and critical appraisal of the HRQoL measures are provided in Appendix 3.

The EORTC QLQ-C30 is a questionnaire developed specifically to assess HRQoL in cancer patients. The questionnaire consists of 30 questions, 5 function scales (physical, role, cognitive, emotional, and social), 1 global health status/global quality-of-life scale, 3 symptom scales (fatigue, pain, and nausea and vomiting), and 6 single items that assess additional symptoms (dyspnea, appetite loss, sleep disturbance, constipation, and diarrhea) and financial impact.⁴⁸ Scales and single items range in score from 0 to 100, with higher scores on the functional and global health status/quality-of-life scales indicating higher levels of functioning and health status/quality of life, respectively. Higher scores on symptom scales or items represent a greater presence of symptoms.⁴⁸

The EQ-5D-5L version 2.0 is a generic HRQoL questionnaire for assessing a patient’s health status in terms of a single index value or utility score. There are 2 components of the questionnaire: a descriptive system that allows patients to rate their level of problems (none, slight, moderate, severe, extreme/unable) in 5 areas (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression), and a visual analogue scale that allows patients to rate their overall health status from 0 (worst imaginable) to 100 (best imaginable).⁴⁹

Safety

Assessing the safety of dostarlimab was a primary objective of the GARNET trial. The primary safety variables included the following:

• TEAEs — all, SAEs, their relationship to the study treatment, and their intensity (based on National Cancer Institute Common Terminology Criteria for Adverse Events v4.03 severity grades)

• TEAEs leading to discontinuation

• TEAEs leading to death

• irAEs

• clinical laboratory measures (including hematology, chemistry, coagulation, thyroid function, and urinalysis) and vital signs

• ECG

• physical examination

• ECOG status

• serum pregnancy testing

• concomitant medications.

Statistical Analysis

Sample Size Determination

For the final analysis of the co-primary end point (ORR) in cohort A1, the sample size of approximately 100 patients with dMMR or MSI-H EC (with potential enrolment of up to 165 patients) was planned after protocol amendment 5. The null hypothesis that the true response rate was 20% or less (expected ORR for conventional therapy) was tested against a 1-sided alternative hypothesis of a response rate of at least 40%. With 65 patients in cohort A1, there was 92% power to rule out an ORR of 20% or less when the true ORR was 40% at the 2.5% type I error rate (1-sided). The increase in sample size allowed the lower-limit boundary of the exact 95% CI to exclude a response rate of 25% or less, assuming the observed ORR was 35%.

Interim Analyses

The 3 planned interim analyses were based on the combined enrolment in cohort A1 and cohort F (nonendometrial cancer). IA-1, IA-2, and IA-3 were conducted when the combined enrolment reached 100, 200, and 300 patients, respectively. All enrolled patients in cohort A1 and cohort F had dMMR tumours with measurable disease at baseline and at least 24 weeks of follow-up. IA-1 had a data cut-off date of July 8, 2019; IA-2 had a data cut-off date of March 1, 2020; and IA-3 had a data cut-off date of November 1, 2021. IA-2 and IA-3 were added after amendment 6.

All interim analyses evaluated the outcomes listed in Table 8, with the exception of IA-1, which included irORR as a primary end point and patient-reported outcomes as a secondary end point.

Primary Outcomes

The co-primary outcomes in the GARNET trial were ORR and DOR. No formal hypothesis-testing or inferential analyses were performed in cohort A1, and no statistical comparisons were planned between cohorts. The analysis of ORR included summary statistics, including the number of patients and percentage for categorical variables and, for continuous variables, the number of patients, mean, standard deviation (SD), median, minimum, and maximum. Two-sided exact 95% CIs based on the Clopper-Pearson method were reported for the ORR.

DOR analyses were performed using KM methods and summarized by minimum, maximum, 25th, 50th (median), and 75th percentiles with associated 95% CIs, the number and percentage of events, the number and percentage of censored observations, and the median duration of follow-up. Censoring rules for DOR are presented in Table 7.

Secondary Outcomes

The analyses for OS were based on the safety analysis and primary efficacy analysis datasets. OS analyses were performed using KM methods and summarized by the 25th, 50th (median), and 75th percentiles with associated 95% CIs, as well as the number and percentage of events and censored observations. Censoring for OS was set at the last known date of contact.

The analyses of DCR and irDCR involved summary statistics, including the number of patients and percentage for categorical variables and, for continuous variables, the number of patients, mean, SD, median, minimum, and maximum. Two-sided exact 95% CIs based on the Clopper-Pearson method were provided for DCR and irDCR.

The analyses of PFS and irPFS were performed using KM methods and summarized by the 25th, 50th (median), and 75th percentiles with associated 95% CIs, as well as the number and percentage of events and censored observations. Censoring rules for PFS and irPFS are presented in Table 9.

The analyses of irORR and irDOR were similar to the those of the co-primary outcomes of ORR and DOR.

Exploratory Outcomes

No analysis plan, objectives, or minimally important difference (MID) for the EORTC QLQ-C30 or the EQ-5D-5L instruments were specified a priori in the statistical analysis plan; however, according to the study protocol, changes from baseline to each pre-specified study time point were to be measured and summarized using descriptive statistics. HRQoL assessments (EQ-5D-5L and EORTC QLQ-C30) were collected during scheduled visits for all patients in cohorts A1 and F enrolled under amendment 3 and subsequent amendments (i.e., every 3 weeks ± 7 days for the first 12 weeks, in alignment with study drug administration, and every 6 weeks ± 7 days thereafter, in alignment with tumour imaging assessments) while the patient was receiving study treatment. Once a patient discontinued treatment, patient-reported outcome assessments were performed during the EOT visit, the safety follow-up visit, and the post-treatment follow-up period every 90 days (± 14 days).

For the EORTC QLQ-C30, scores and changes in scores from baseline for each domain were summarized by the number of patients with values for the mean, SD, median, range, first and third quartiles, and 95% CI. Statistical graphs were produced for each of the 15 domains to show mean scores over time, with 95% CIs. Similar graphic displays were produced for the mean change from baseline over time. In terms of missing data, if at least half of the items from a particular scale were answered, it was assumed that the missing items had values equal to the average of those items that were present. With this method of imputation, none of the single-item measures could be imputed.

Safety Outcomes

Safety and tolerability were evaluated by monitoring the frequency, duration, and severity of AEs and irAEs. In general, all by-visit summaries of safety parameters were only summarized up to and including month 6 and at the treatment discontinuation visit. AEs were organized based on the Medical Dictionary for Regularity Activities version 23.0. Only TEAEs were analyzed, but all AEs occurring during the study were listed. TEAEs were tabulated by preferred term and system organ class. Patients with the same TEAE more than once had that event counted only once for each system organ class and once for each preferred term. By-patient listings of irAEs were presented, if appropriate.

Sensitivity Analyses

No sensitivity analyses were conducted for any of the outcomes outlined in the protocol of the GARNET trial.

Subgroup Analyses

Subgroup analyses were performed for ORR, DOR, and TEAEs. The following subgroup analyses for DOR and ORR were planned a priori in the statistical analysis plan:

• MSI status (MSI-H versus MSS versus unknown/missing)

• histology (patients with dMMR EC only)

• number of prior anti-cancer therapy regimens (1 versus ≥ 2)

• prior radiation therapy (yes or no)

• prior bevacizumab use (yes or no)

• BOR from last platinum-containing prior anti-cancer therapy (CR or PR, SD, progressive disease, or missing)

• progression-free interval from last platinum-containing prior anti-cancer therapy (< 6 months versus ≥ 6 months versus missing).

The following subgroups aligned with those pre-specified in the protocol for this CADTH review: histology, number of prior anti-cancer therapy regimens, prior radiation therapy, and progression-free interval from last platinum-containing prior anti-cancer therapy. Only the subgroups identified in the CADTH review protocol are reported in the Outcomes section.

Analysis Populations

The primary efficacy analysis dataset, by RECIST 1.1 per BICR, was defined as all patients in the safety analysis dataset with measurable disease at baseline (defined as the existence of at least 1 target lesion at baseline tumour assessment identified by BICR) who had the opportunity for at least 24 weeks of tumour assessment at the time of analysis (i.e., patients whose first dose of dostarlimab administration was on or before September 15, 2019).

The secondary efficacy analysis dataset, by irRECIST per investigators’ assessment, was defined as all patients in the safety analysis dataset with measurable disease at baseline (defined as the existence of at least 1 target lesion at baseline tumour assessment identified by investigator assessment) who had the opportunity for at least 24 weeks of tumour assessment at the time of analysis (i.e., patients whose first dose of dostarlimab administration was on or before September 15, 2019).

The safety analysis dataset included all patients who received any amount of study drug. In addition, all patient-reported outcome analyses were conducted on patients using the safety analysis dataset enrolled under amendment 3 or subsequent amendments.

Results

Patient Disposition

Details of patient disposition in the cohort A1 safety analysis dataset of the GARNET trial are summarized in Table 10. A total of 479 patients were screened and, of those, 129 were enrolled into cohort A1. In cohort A1, there were 126 patients with dMMR EC and 3 with MMR-unk/MSI-H identified through NGS. All patients enrolled received dostarlimab. Of enrolled patients, 100.0%, 81.4%, and 87.6% were included in the safety, primary, and secondary efficacy analysis datasets, respectively. In IA-2, a subset of 72 patients from IA-1 were included. Refer to Appendix 4 for detailed outcomes data for this subset of patients.

At the time of IA-2, the median duration of follow-up was 16.3 months. A total of 47 (36.4%) patients had withdrawn from the study. The most common reason for study discontinuation was death, which occurred in 36 patients (27.9%); progressive disease was the main cause of death. Two patients were lost to follow-up (1.6%) and 71 (55.0%) discontinued treatment. The primary reasons for treatment discontinuation were confirmed progressive disease (38.0%) followed by AEs (10.9%). Twenty-eight (21.7%) patients continued treatment with dostarlimab beyond initial confirmation of progressive disease. Thirty-six patients (27.9%) died during the study, 31 (24.0%) due to progressive disease and 5 (3.9%) due to an AE.

Protocol Deviations

Protocol deviations in cohort A1 of the GARNET trial are summarized in Table 11. There were 168 important protocol deviations in 75 (58.1%) patients in the safety analysis dataset. Most of these deviations were related to study visits or procedures (n = 61; 42.1%), There were 17 significant protocol deviations in 11 (8.5%) patients, the majority of which were related to eligibility criteria (n = 8; 6.2%). The sponsor noted that “none of these significant protocol deviations were considered to affect the patients’ safety or well-being or the overall integrity of the study.”

Exposure to Study Treatments

Duration and Dose Intensity

Data on exposure to dostarlimab in cohort A1 at IA-2 are summarized in Table 12. Data were available for 129 patients with dMMR or MMR-unk/MSI-H. The median DoT with dostarlimab was 25.6 weeks. The maximum DoT with dostarlimab was 138.9 weeks. Approximately half of patients (53.5%) received dostarlimab treatment for up to 19 to 24 weeks. The median relative dose intensity was 100% throughout the study treatment period because dose reductions were not permitted. Twenty-eight patients continued study treatment despite initial progressive disease.

Treatment Compliance

Treatment compliance was based on number of infusions. At any time during the study, less than 6% of patients with dMMR EC had a missed infusion, and less than 20% had an infusion delay. From cycle 1 to cycle 4, there were 3 missed infusions (2.4%) and 11 infusion delays (8.7%) reported in patients with dMMR EC. From cycle 5 through the EOT, 7 missed infusions (5.6%) and 25 infusion delays (19.8%) were reported in patients with dMMR EC. One infusion interruption (0.8%) was reported between cycle 5 and EOT.

Table 12: Patients With dMMR or MSI-H EC on Treatment by Week Intervals in Cohort A1 of the GARNET Trial (IA-2, Safety Analysis Dataset)

dMMR = mismatch repair-deficient; DoT = duration of treatment; EC = endometrial cancer; IA-2 = second interim analysis; MSI-H = microsatellite instability-high; SD = standard deviation.

^aThe population total (N = 129) included 126 dMMR EC patients and 3 MMR-unk/MSI-H EC patients. Data from patients with dMMR tumours and MMR-unk/MSI-H tumours are presented as a total due to the small proportion of patients (n = 3) with MMR-unk/MSI-H tumours; the sponsor noted that data remained similar when these patient groups were pooled together.

^bIntervals are inclusive of the upper week number (i.e., week 1 to ≤ week 3 is equivalent to day 1 to day 21, inclusive).

^cIf the last cycle of treatment was ≤ 4 cycles, the DoT was calculated as follows: the last dose date − start date + 21 or death date – start date + 1 if death occurred < 21 days after last dose. If the last cycle of treatment was ≥ 5 cycles, the DoT was calculated as follows: last dose − start date + 42 or death date – start date + 1 if death occurred < 42 days after last dose.

Source: Clinical Study Report.¹⁸

Concomitant Medication

Nearly all patients in cohort A1 reported taking concomitant medications (99.2% of dMMR EC patients). The most common types of concomitant medications (used by ≥ 40% of patients) were in the Anatomical Therapeutic Chemical (ATC) classes of the following:

• other analgesics and antipyretics (76.2%)

• opioids (58.7%)

• antithrombotic drugs (49.2%)

• drugs for peptic ulcer and gastro-esophageal reflux disease (47.6%)

• anti-inflammatory and antirheumatic products, nonsteroids (42.1%)

• drugs for constipation (41.3%).

Subsequent Treatments

Thirty-three patients received subsequent anti-cancer therapy after treatment with dostarlimab. Of these patients, 27 had progressive disease. These therapies, from most to least frequent, included the following:

• radiotherapy (7.8%)

• single-drug chemotherapy, such as doxorubicin (4.7%), carboplatin (4%), paclitaxel (4%), and pegylated liposomal doxorubicin (4%)

• immunotherapy, such as pembrolizumab (4.0%) and bevacizumab (2.4%)

• hormone therapy, such as letrozole (3.1%), megestrol acetate (0.8%) medroxyprogesterone (0.8%), tamoxifen (0.8%), temsirolimus (0.8%)

• combination chemotherapy, such as paclitaxel plus carboplatin (2.4%) and carboplatin plus gemcitabine (1.6%)

• surgery (0.8%).

Outcomes

Only outcomes and analyses of the subgroups identified in the review protocol are reported below. Data tables for the subset of patients from IA-1 included in IA-2 (n = 72) can be found in Appendix 4. In addition, data tables for all immune-related outcomes (except irPFS) are available in Appendix 4. As part of the sponsor’s feedback on this CADTH reimbursement review report, the sponsor provided CADTH with an updated analysis (data cut-off of November 1, 2021) for certain baseline characteristics, efficacy, and safety outcomes in the GARNET trial. The results of the updated analysis were, overall, consistent with those reported in the previous analyses of data from the March 1, 2020, cut-off date. The additional results from the November 1, 2021, data cut-off date are available in Appendix 5.

Overall Survival

The results for OS from cohort A1 of the GARNET trial are summarized in Table 13. The median OS was not reached at the time of IA-2. The median follow-up time was 16.3 months. In total, there were 35 deaths (33.3%), and 70 patients (66.7%) were censored. The probability of patients in cohort A1 surviving to 6, 9, and 12 months was 80.9% (95% CI, 71.7% to 87.4%), 75.1% (95% CI, 65.2% to 86.6%), and 68.9% (95% CI, 58.3% to 77.4%), respectively.

Figure 3: KM Plot of OS for Patients With dMMR or MSI-H EC per RECIST 1.1 (IA-2, Secondary Efficacy Analysis Dataset)

CI = confidence interval; MMR = mismatch repair-deficient; EC = endometrial cancer; IA-2 = second interim analysis; KM = Kaplan-Meier; MMR-unk = unknown mismatch repair tumour status; MSI-H = microsatellite instability-high; OS = overall survival; RECIST 1.1 = Response Evaluation Criteria in Solid Tumours Version 1.1; NR = not reported.

Source: Clinical Study Report¹⁸

Disease Control Rate and Immune-Related Disease Control Rate

The DCR results (based on investigator assessment) of the GARNET trial for cohort A1 are summarized in Table 14. The DCR was 57.1%, with 11 patients (10.5%) having a BOR of CR, 36 patients (34.3%) having a BOR of PR, and 13 patients (12.4%) having a BOR of stable disease. Forty-two (89.4%) patients at the time of IA-2 had an ongoing response.

irDCR by investigator assessment per irRECIST 1.1 (n = 113) was 63.7% for patients with dMMR or MSI-H EC, which was slightly higher than the DCR. Eight patients (7.1%) had an irBOR of irCR, 44 patients (38.9%) had an irBOR of irPR, and 20 patients (17.7%) had an irBOR of irSD. Forty-three (82.7%) patients at the time of IA-2 had an ongoing response. Detailed irDCR results are available in Appendix 4.

Health-Related Quality of Life

EORTC QLQ-C30

EORTC QLQ-C30 data were available for 94 of the 129 patients in cohort A1 of the GARNET trial at the time of IA-2. Completion rates for the EORTC QLQ-C30 instrument declined over time. The completion rate at baseline was 100% and 58.5% by cycle 7 (n = 55). The mean scores and mean changes from baseline at each assessment point for EORTC QLQ-C30 (global health status or quality-of-life scale) are summarized in Table 15 and Figure 4. A definition for what constituted a clinically meaningful change from baseline in the study population was not provided. Overall, summary scores increased or remained stable from baseline to cycle 7 among patients who completed questionnaires.

A summary of the disease-related symptom subscales can be found in Figure 5. Patients who received dostarlimab and completed the questionnaire reported decreased symptoms or remained stable over time in key disease-related symptom subscales, including pain and fatigue.

Mean change in symptomatic AE response from baseline over time is illustrated in Figure 6. Most patients who experienced symptomatic AEs, including nausea, vomiting, constipation, diarrhea, or tiredness, remained stable or had improvement from baseline in these symptoms over the treatment course. Less than 25% of patients reported single-category worsening in these AE symptoms, and less than 6% reported 2- or 3-category worsening.

Figure 4: Mean Change in Global Health Status and QoL From Baseline (IA-2, Safety Analysis Dataset)

IA-2 = second interim analysis; QoL = quality of life.

Figure 5: Pain, Fatigue, and Physical Functioning Mean Change From Baseline (IA-2, Safety Analysis Dataset)

CxDx: cycle X, day X; IA-2 = second interim analysis.

Figure 6: Symptomatic AE Change in Response From Baseline (IA-2, Safety Analysis Dataset)

AE = adverse event; CxDx = cycle X, day X; IA-2 = second interim analysis; Q = question.

EQ-5D-5L

EQ-5D-5L data were available for 89 of 129 patients in cohort A1 at the time of IA-2. Summary data for the descriptive system of the EQ-5D-5L were not reported for any of the 5 dimensions. Overall, completion rates for the EQ-5D-5L and corresponding EQ-VAS instrument declined over time; there were 89 patients who completed questionnaires at baseline and 5 or fewer patients at each post-treatment visit.

EQ-VAS

At baseline, the mean EQ-VAS score was 69.3 (SD = 19.2) (refer to Figure 7). The following changes were observed in the EQ-VAS scores over time:

• At week 12, the mean score was 77.1 (SD = 18.0) and mean change from baseline was 5.0 (SD = 12.6).

• At week 18, mean score was 77.4 (SD = 17.4) and mean change from baseline was 4.0 (SD = 15.2).

• At week 42, the change from baseline was 4.0 (SD = 16.2).

Greater change in scores was observed after EOT; however, the number of patients at each post-treatment visit was markedly low (< 5 patients).

Figure 7: Adjusted Mean Change From Baseline in EQ-VAS (IA-2, Safety Analysis Dataset)

EOT = end of treatment; EQ-VAS = EQ-5D-5L visual analogue scale; IA-2 = second interim analysis; SUVF: survival follow-up; WX = week X.

Note: Data for the adjusted mean and 95% CI are derived from mixed model repeated measures with week of visit and ECOG Performance Status as factors and baseline score as continuous covariate as well as an unstructured covariance structure.

Progression-Free Survival and Immune-Related Progression-Free Survival

The results for PFS in cohort A1 of the GARNET trial, based on BICR, are summarized in Table 16. At IA-2, 57 (54.3%) patients had a PFS event and median PFS was 5.5 (95% CI, 3.2 to NR) months. KM estimates of PFS by RECIST 1.1 were 48.6% (95% CI, 38.6% to 56.8%) at month 6 and 47.5% (95% CI, 37.4% to 56.8%) at month 9 and at month 12. Refer to Figure 8 for the PFS KM curve.

In terms of irPFS, 57.3% of patients had an irPFS event at the time of IA-2. The median irPFS time was 10.3 months (95% CI 5.2 to 18.0). Based on investigator assessment, the probability of having irPFS based on KM estimates were 54.7% at 6 months, 50.7% at 9 months, and 47% at 12 months for patients with dMMR or MSI-H EC (95% CI not reported). KM estimates for irPFS were higher than that of PFS by BICR. Refer to Figure 13 in Appendix 4 or the irPFS KM curve.

Figure 8: KM Plot for PFS in Patients With dMMR or MSI-H EC per RECIST 1.1 Assessed by BICR (IA-2, Primary Efficacy Analysis Dataset)

BICR = blinded independent central review; CI = confidence interval; dMMR = mismatch repair-deficient; EC = endometrial cancer; IA-2 = second interim analysis; KM = Kaplan-Meier; MMR-unk = unknown mismatch repair tumour status; NR = not reported; PFS = progression-free survival; RECIST 1.1 = Response Evaluation Criteria in Solid Tumours Version 1.1.

Source: Clinical Study Report.¹⁸

Objective Response Rate and Immune-Related Objective Response Rate

The results for ORR (based on BICR) for cohort A1 of the GARNET trial are summarized in Table 12. The ORR results were generally consistent with the results for irORR.

At the time of IA-2, the ORR, per RECIST 1.1, was 44.8% (95% CI, 35.0% to 54.8%); 47 of 105 patients achieved either a CR (n = 11) or PR (n = 36) (Table 14). The median follow-up time was 16.3 months. The ORR results were generally consistent with the results of irORR; the irORR, per irRECIST 1.1, was 46% (95% CI, 36.6% to 55.6%), with 82.7% of responders having an ongoing immune-related response at the time of IA-2.

The ORR results for pre-specified patient subgroups of interest to the CADTH review are summarized in Table 17. The ORR was 49.2% in patients with 1 line of prior anti-cancer therapy, compared with 36.8% in patients who received 2 or more lines of prior anti-cancer therapy. The ORR was 46.6% in patients with prior radiation, compared with 40.0% in patients who had not received prior radiation. The ORR was 46.0% in patients with a progression-free interval of at least 6 months, compared with 39.5% in patients with a progression-free interval of less than 6 months. The ORR was 56.3% in patients with type II endometrioid carcinoma and 40.0% in patients with type I endometrioid carcinoma. Note that 95% CIs were not provided for any of the ORR subgroup analyses.

The sponsor noted that there were too few patients of each endometrioid carcinoma type II subtype to make a meaningful comparison based on histologic subtype.

Duration of Response and Immune-Related Duration of Response

The DOR results (based on BICR) for cohort A1 of the GARNET trial are summarized in Table 18.

Of the 47 patients who achieved an objective response, 42 (89.4%) had an ongoing response at the time of the IA-2 data cut-off. The median duration of follow-up was 16.3 months; however, the median DOR was not reached among responders. The DOR in these patients ranged from 2.63 to 28.09 months as at the time of the IA-2data cut-off, with 37 patients (78.7% of responders) achieving a DOR of at least 6 months. Based on these results, the probability of responders maintaining a confirmed objective response, per RECIST 1.1, was estimated to be 97.9% (95% CI, 85.8% to 99.7%), 90.9% (95% CI, 73.7% to 97.1%), and 80.1% (95% CI, 56.8% to 91.7%) at months 6, 12, and 18, respectively.

The sponsor noted that the small sample size of responders hampered the ability to estimate the DOR of patient subgroups, so these data were not reported.

The results of investigator-assessed irDOR were generally consistent with DOR assessed by BICR (refer to Appendix 4 for further details). At the time of the IA-2, 43 patients had an ongoing immune-related response out of the 52 who achieved an irCR or irPR. The median irDOR was not reached among responders after a median follow-up of 16.5 months. Approximately 76.9% of responding patients achieved an irDOR of at least 6 months. Based on KM estimates, the probability of maintaining a response, per irRECIST, at 6 months was 96.1% (95% CI, 85.2 to 99.0) and at 12 months and at 18 months was 79.2% (95% CI, 62.1% to 89.2%).

Harms

Only those harms identified in the CADTH review protocol are reported in this section. A summary of harms data is provided in Table 19.

Adverse Events

The most frequently reported TEAEs (≥ 15%) in patients with dMMR or MSI-H EC were nausea, diarrhea, anemia, fatigue, asthenia, constipation, vomiting, abdominal pain, cough, arthralgia, urinary tract infection, and back pain. With the exception of anemia, these TEAEs were reported to be mild or moderate in severity in most patients.

Grade 3 or higher TEAEs occurred in 48.1% of patients in cohort A1. The most commonly reported grade 3 or higher TEAEs were anemia (14.7%), abdominal pain (5.4%), and hyponatremia (3.9%). Grade 4 TEAEs were reported for hyponatremia, pulmonary embolism, pneumonia, and sepsis. Sepsis was reported as a grade 4 TEAE in 3 patients with dMMR EC and 1 patient with MMR-unk/MSI-H. Other grade 3 or higher TEAEs included acute kidney injury (3.1%) and back pain (3.1%).

Serious Adverse Events

The percentage of patients that experienced a serious TEAE was 34.1% in cohort A1. The most common serious TEAEs were abdominal pain, acute kidney injury, and sepsis, each occurring in 3.1% of patients, and pulmonary embolism, pyrexia, and urinary tract infection, each occurring in 2.3% of patients.

Withdrawals Due to AEs

None of the patients in cohort A1 who withdrew from the GARNET trial had an AE as a primary reason. AEs that led to study treatment discontinuation occurred in 11.6% of patients, whereas AEs that led to study treatment interruption occurred in 24.0% of patients. The most common AEs leading to study treatment interruption were anemia (3.1%) and diarrhea (2.3%).

Mortality

TEAEs leading to death occurred relatively rarely in cohort A1 during the treatment period (n = 1; 0.8%) and included aspiration. TEAEs leading to death during the 90-day safety follow-up occurred in 4 patients (3.1%), and included pleural effusion, pneumonia, sepsis, and shock. None of the TEAEs leading to death were considered by the investigators to be treatment-related. No TEAEs were the primary cause of death during the long-term follow-up period.

Notable Harms

Notable harms as specified in the CADTH review protocol included immune-related toxicity.

The incidence of irAEs was 34.9% in cohort A1. The most frequently reported irAEs (≥ 5%) were diarrhea and hypothyroidism. A total of 7.9% of patients had a serious irAE, 12.7% had an irAE that was grade 3 or higher, and 4.8% of patients with dMMR EC had an irAE that led to study treatment discontinuation. Most of the irAEs were considered by the investigators to be related to study treatment.

Critical Appraisal

Internal Validity

The main limitation of the GARNET trial stems from the single-arm trial design and the lack of a comparator group. Such a design makes it challenging to interpret the efficacy and safety events attributable to dostarlimab, because all patients in cohort A1 received the same treatment. The lack of comparison with an active comparator or with standard of care or placebo precludes the ability to assess the relative therapeutic benefit or safety of dostarlimab. Formal statistical significance and hypothesis-testing were not performed, except for ORR. The risk of selection bias cannot be ruled out, given the lack of transparency regarding the selection and timing of the enrolment of patients in the trial. In the absence of such information, the risk of bias is unclear. GARNET is an open-label trial, and the study investigators and patients were aware of their treatment status, which increases the risk of detection and performance bias; this has the potential to influence results and outcomes in favour of dostarlimab if the assessor (investigator or patient) believes the study drug is likely to provide a benefit. Furthermore, if study personnel and patients knew that the treatment was dostarlimab (which is known to cause irAEs and other neurotoxicity), this also could have influenced the reporting of harms outcomes. To mitigate the impact of such bias, the investigators used a BICR to evaluate responses using standardized criteria for certain efficacy outcomes, such as ORR, DOR, PFS, and DCR. Bias is therefore less a concern for these efficacy end points and OS, and more of a concern for subjective end points such as HRQoL and safety. Overall, the magnitude and direction of this bias remain unclear.

Median OS was not reached at the time of IA-2, so the survival data from the trial were immature. Because results were based on an interim analysis, the treatment benefit may be overestimated and harms may be underestimated. In addition, the potential for confounding cannot be ruled out, given the lack of adjustment in the analysis for known confounders and treatment-effect modifiers that would typically be accounted for in an RCT. Interpretation of time-to-event end points such as OS or PFS is limited in single-arm studies. The clinical experts agreed that in the absence of robust comparative data on PFS and OS, no firm conclusions could be drawn on how dostarlimab compares with other relevant treatment options.

The open-label, single-arm design also limits the ability to interpret HRQoL data from the GARNET trial. Loss of patients completing HRQoL questionnaires over the course of the study led to small sample sizes at many assessment time points (in some cases < 5 patients) and resulted in imprecise measurement, with wide and overlapping CIs. Statistical analyses of changes in scale scores over time were not conducted, and a definition of what was considered a clinically meaningful response or minimally important difference was not reported. As well, summary data were not reported for the EQ-5D-5L descriptive system. For these reasons, no conclusions should be drawn from the HRQoL data. This is considered a significant limitation of the GARNET trial, because HRQoL was identified as an important outcome by the patient and clinician groups providing input for this review.

External Validity

Overall, the clinical experts consulted by CADTH agreed that the inclusion and exclusion criteria, baseline patient characteristics, concomitant medications, and prohibited medications in cohort A1 of the GARNET trial were reflective of patients they see in clinical practice for the indication under review. There were no barriers to the identification of patients who would most benefit from the treatment, given that testing for MMR and MSI status is standard practice in Canada. The population enrolled in the trial was consistent with the population expected to be treated in Canadian clinical practice, but the clinical experts noted that patients in the GARNET trial were slightly younger than those seen in clinical practice. The clinical experts also noted that patients with certain comorbidities (e.g., obesity) and of certain racial groups (e.g., Black) may have poorer prognosis; however, the findings of this study may not necessarily be generalizable to these subgroups, given the small sample size and lower level of representation of ethnic groups in the study (n = 2). Furthermore, the clinical experts indicated that no different treatment effect would be expected based on different disease-management practices in participating countries. In the opinion of the clinical experts, as long as patients have dMMR or MSI-H tumour status, dostarlimab would be appropriate to administer after any of the prior therapies received by patients in the trial. The majority of patients (88%) in cohort A1 of the GARNET trial had received 2 or fewer prior lines of systemic therapy before trial enrolment, and 11% of patients had received 3 or more lines of prior systemic therapy. However, the clinical experts noted that experience from clinical practice suggests that dostarlimab may have a lower magnitude of treatment benefit in patients with more prior lines of systemic therapy. This aligns with the results of the ORR subgroup analysis, in which the ORR was higher in patients with 1 line of prior anti-cancer therapy than in patients with 2 or more lines of prior anti-cancer therapy (49.2% versus 36.8% [95% CI, NR to NR]). However, these results are exploratory, and conclusions should be interpreted with caution. The clinical experts and clinician groups agreed that patients should not have previously been treated with immunotherapies. In cohort A1, 3 patients could not have their MMR tumour status determined, but their tumours were classified as MSI-H. The clinical experts noted that these 2 groups are generally thought to be synonymous, and data could be pooled together, especially given the small number of patients with MSI-H status. Concomitant medications received by patients in the trial appeared to reflect the medications that patients would receive in Canadian clinical practice, according to the clinical experts. It is also acknowledged that mature OS data will be confounded by the use of subsequent anti-cancer therapy received by some patients after progressive disease.

According to the clinical experts consulted by CADTH, OS, PFS, and HRQoL are clinically meaningful end points for patients with dMMR or MSI-H recurrent or advanced EC that has progressed on or after treatment with a platinum-containing regimen. Tumour response outcomes are also important in this patient population because of the accompanying delay in the worsening of symptoms and the slower decline in ECOG performance status.

There were a limited number of patients in the primary efficacy dataset (N = 105). The magnitude of the treatment-effect estimates observed in a small study sample may not be replicable in a larger study sample or generalizable to the target population in real-world clinical practice. Furthermore, the subgroup analyses had no statistical comparisons and even smaller sample sizes, which limits the generalizability to a broader population. These data should be considered exploratory in nature, and no conclusions should be drawn based on the results.

Summary Indirect Evidence

Objectives and Methods for the Summary of Indirect Evidence

The objective of this section is to provide an appraisal and summary of the indirect evidence submitted by the sponsor, which included 6 reports of ITCs (3 reports of MAICs and 3 of IPTW analyses),¹⁹ which compare dostarlimab with other treatments used for patients with advanced or recurrent EC that has progressed on or after a platinum-based regimen. Results for the MAIC and IPTW analyses are presented separately.

A focused literature search for network meta-analyses dealing with EC was run in MEDLINE All (1946–) on November 3, 2021. No limits were applied to the search. The literature search identified 31 citations, none of which met the eligibility criteria. Given the lack of comparative evidence, and the phase I, single-arm nature of the GARNET trial, the sponsor-submitted ITCs were used to inform the pharmacoeconomic model. The submitted analyses were appraised and summarized.

Matching-Adjusted Indirect Comparisons

Description of MAIC Analyses

The sponsor submitted 3 MAIC reports to demonstrate the efficacy and safety of dostarlimab compared with relevant treatments for advanced or recurrent EC¹⁹:

• Two MAIC reports that compared individual patient data (IPD) from the phase I GARNET trial with a GARNET-like real-world evidence (RWE) cohort in the UK that received current treatment paradigms (hereafter called MAIC report 1), and a selection of 5 specific treatments for advanced or recurrent EC (hereafter called MAIC report 2) from the National Cancer Registration and Analysis Service (NCRAS) database.¹⁹

• One MAIC report that compared IPD from the phase I GARNET trial with relevant comparator data from the published literature (hereafter called MAIC report 3).¹⁹

Methods of MAIC Analyses

Objectives

In the absence of direct comparative evidence from trials, the aim of each analysis was to compare the efficacy and safety (e.g., OS, PFS, time on treatment, ORR, and/or SAEs) of dostarlimab with current UK treatment paradigms based on RWE cohorts and published literature, using the MAIC method.¹⁹

Study Selection Methods

The index trial in all cases was based on IPD from cohort A1 of part 2B of the phase I, single-arm GARNET trial (dMMR or MSI-H EC cohorts, safety population set; n = 129).¹⁹ Study selection methods to identify relevant comparators varied in the 3 sponsor-submitted MAIC reports.

For the 2 reports that used RWE as a comparator (MAIC reports 1 and 2), no systematic literature review (SLR) was conducted. Instead, the investigators created GARNET-like cohorts based on a subgroup of patients from the NCRAS in the UK, including a base-case analysis of current treatment paradigms, an analysis of 5 treatment-specific cohorts, and a sensitivity analysis of patients with an ECOG Performance Status of 0 or 1 at index. Notably, information on dMMR or MSI-H status (an inclusion criterion in GARNET and the indication under review) was not available in the RWE cohorts. In MAIC report 3, a SLR of clinical and observational studies was conducted in accordance with the highest standard for evidence synthesis (date last searched, April 5, 2020).⁵¹ Although there were no limitations to study eligibility by risk of bias, the authors noted that the included studies were generally at low risk of bias per the NICE tool,⁵² and bias was either high or unclear per the CASP tool.^19,53

The selection characteristics used to develop the GARNET-like cohorts and the population, intervention, comparison, outcomes, and study (PICOS) criteria used to identify published literature are summarized in Table 20.

Table 20: Selection Criteria for GARNET-Like Subgroups and for Relevant Comparator Studies for Sponsor-Submitted MAICs

AEs = adverse events; CR = complete response; DCR = disease control rate; DFS = disease-free survival; DOR = duration of response; EC = endometrial cancer; EORTC QLQ-C30 = European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30; EQ-5D = EuroQol 5-Dimensions; HRQoL = health-related quality of life; MAIC = matching-adjusted indirect comparison; MMR = mismatch repair; MSI = microsatellite instability; NA = not applicable; NCRAS = National Cancer Registry Analysis System; ORR = objective response rate; OS = overall survival; PD-L1 = programmed death (ligand) 1; PFS = progression-free survival; PLD = pegylated liposomal doxorubicin; PR = partial response; RCT = randomized controlled trial; RWE = real-world evidence.

Source: Sponsor-Submitted ITC Reports¹⁹

The outcomes evaluated in the MAICs are summarized in Table 21. OS was the pre-specified primary outcome in all MAICs. In MAIC report 3, PFS was also included as a pre-specified primary outcome. PFS and TTD were the pre-specified secondary outcomes of MAIC reports 1 and 2. These were summarized descriptively, as the measurement definitions for PFS and the time period evaluations associated with the RWE cohort were too dissimilar to those in GARNET. As PFS was not recorded in the NCRAS database, TTNT and TTD were used as proxies. Pre-specified secondary end points in MAIC report 3 included ORR, DOR, HRQoL, and SAEs; however, no definitions were provided for these outcomes in the report.¹⁹

MAIC Analysis Methods

Given that the GARNET trial was a single-arm trial, unanchored MAICs were necessary to conduct the ITCs because the single-arm nature precluded the use of anchored ITC methods. A summary of the analysis methods in each MAIC is provided in Table 23.

GARNET Versus NCRAS RWE Cohort MAIC Methods (MAIC Reports 1 and 2)

Analysis methods for the MAICs conducted using RWE data (MAIC reports 1 and 2) were identical, only differing in the RWE cohorts used. The GARNET trial was a phase I, open-label, dose-escalation study, whereas the RWE study was a descriptive, noninterventional study of patient-level data available through the NCRAS to describe characteristics, treatments, and outcomes of patients diagnosed with advanced or recurrent EC. The dataset for analysis consisted of IPD for all patients in the GARNET safety cohort (n = 129), and aggregated summary information from the initial RWE cohort of incident patients at second-line treatment (n = 999; n = 501 for patients with an ECOG Performance Status of 0 or 1 at index). Treatment-specific RWE cohorts included patients who received paclitaxel monotherapy (n = 116), carboplatin plus paclitaxel (n = 279), carboplatin plus liposomal doxorubicin (n = 141), liposomal doxorubicin monotherapy (n = 130), and carboplatin monotherapy (n = 93).¹⁹

Key differences were noted by the sponsor between the GARNET trial and the NCRAS RWE cohort’s inclusion criteria. One of the main differences between GARNET and the RWE cohort was the MMR or MSI status of patients. All patients in GARNET had dMMR or MSI-H EC, whereas this information was not available in the RWE cohort, so patients were included regardless of MMR or MSI status. Another key difference between GARNET and the RWE cohort’s inclusion criteria was that the RWE cohort used platinum doublet therapy as the key milestone regarding lines of therapy. The RWE cohort did not analyze the lines of therapy before the diagnosis of advanced or recurrent EC. The first-line therapy received for advanced or recurrent diagnosis was platinum doublet therapy for nearly all (99.3%) patients. Thus, the incident population at second-line therapy gave rise to the base case sample size (n = 999). Last, although all patients in GARNET had an ECOG Performance Status of 0 or 1, the inclusion criteria on ECOG Performance Status were partially met in the RWE cohort. In GARNET, ECOG Performance Status was ascertained at the index date of the trial (i.e., the start of second-line therapy), whereas in the RWE cohort, ECOG Performance Status was recorded at diagnosis. Also, there were many patients with ECOG Performance Status documented as “not recorded” in the RWE cohort. A sensitivity analysis conducted with the RWE cohort was restricted to patients with a known ECOG Performance Status of 0 or 1. This was referred to as the RWE cohort (ECOG ≤ 1).¹⁹

A total of 8 analyses (4 base-case and 4 scenario analyses) were conducted in the initial RWE cohort of MAIC report 1. Hazard ratios (HRs), 95% CIs, and corresponding P values were calculated only for the OS outcome (comparisons were descriptive for other outcomes), as follows¹⁹:

• OS in GARNET (N = 129) versus OS in RWE cohort (base case, n = 999; ECOG ≤ 1, n = 501)

• PFS in GARNET (N = 129) versus TTNT in RWE cohort (base case, n = 999; ECOG ≤ 1, n = 501)

• PFS in GARNET (N = 129) versus TTD in RWE cohort (base case, n = 999; ECOG ≤ 1, n = 501)

• TTD in GARNET (N = 129) versus TTD in RWE cohort (base case, n = 999; ECOG ≤ 1, n = 501).

A total of 4 analyses were performed for each treatment-specific RWE cohort in MAIC report 2. HRs, 95% CIs, and P values were calculated only for the OS outcome (comparisons were descriptive for other outcomes), as follows¹⁹:

• OS in GARNET (N = 129) versus OS in treatment-specific RWE cohorts

• PFS in GARNET (N = 129) versus TTNT in treatment-specific RWE cohorts

• PFS in GARNET (N = 129) versus TTD in treatment-specific RWE cohorts

• TTD in GARNET (N = 129) versus TTD in treatment-specific RWE cohorts.

A range of prognostic variables typically associated with survival in EC were identified in a targeted literature review (conducted May 2020), and validated through a panel of oncologists in Canada, Germany, and the UK, as follows¹⁹:

• race and/or ethnicity (Black, other, unknown versus White)

• age category (≥ 65 years versus < 65 years)

• ECOG Performance Status at treatment initiation (1 versus 0)

• histology at initial diagnosis (nonendometrioid, unknown versus endometrioid)

• FIGO stage (stages III and IV versus stages I and II)

• grade of disease at diagnosis (grades 3 and 4, unknown versus grades 1 and 2)

• number of prior platinum-based therapies (0 or 1 versus ≥ 2)

• prior surgery for study indication (yes versus no).

In MAIC report 1, assessment of the effect modification and prognostic value of each potential matching variable was done using a Cox proportional hazard model for each outcome (OS, PFS, and TTD). Cox regression models were fit separately for the GARNET data and the RWE data. Patient characteristics that exhibited an association at a level of significance of at least 0.1 in at least 1 of the 2 datasets were considered prognostic. P values evaluated in a multivariable model were obtained from a full model (including all variables listed above), and then backward stepwise selection, with a P ≤ 0.1 threshold, was applied. Effect-modifying status was assessed semi-empirically by comparing regression coefficients (or HRs) for the variable of interest from the 2 regression models fitted to the GARNET and the RWE dataset, where large differences between coefficients were considered to be potentially effect-modifying. The following variables were considered prognostic for each outcome¹⁹:

• OS: ECOG, race and/or ethnicity, grade, histology, stage at diagnosis, and prior surgery

• PFS: ECOG, and histology

• TTNT: Stage at diagnosis, grade, and prior surgery

• TTD: ECOG, race and/or ethnicity, grade, histology, and prior surgery.

The IPD from the GARNET trial were matched to the RWE cohort using the selected matching variables to estimate the above comparisons. A method of moments, as introduced by Signorovitch et al. (2010),⁵⁴ was used to allow a propensity score logistic regression model to be estimated without IPD for the RWE cohort.¹⁹ After matching, re-weighted outcome data from the GARNET trial and RWE cohort were compared. This included calculating KM survival estimates for OS, PFS, TTD, and TTNT at given time points after index treatment initiation, as well as creating KM curves. The primary end point analysis used a Cox proportional hazard model, with weights obtained using the MAIC method, to re-estimate HRs for dostarlimab in the GARNET trial versus the RWE cohort. In the treatment-specific RWE cohort, weighted Cox proportional hazard models were fitted, and OS HRs for dostarlimab versus the 5 comparator treatments of interest were re-estimated, together with 95% CIs and P values. For use in these regression models, OS data from the RWE cohort were digitized, and pseudo-IPD were constructed from KM curves using the method described by Guyot et al. (2012).⁵⁵ Log-cumulative hazards, as well as Schoenfeld residuals, were plotted for OS to examine the proportional hazards assumption (PHA) for the treatment group variable before and after matching.¹⁹

To assess the impact of each potential covariate on effective sample size (ESS) and estimated treatment effect, a “leave 1 out” approach was applied. This approach started with the calculation of ESS and a measure of treatment effect (HR for OS in the case of this study) for a full model, and then the calculation of the same 2 measures after removal of the covariate from the model; all remaining matching variables in the model were kept.¹⁹

The following matching scenarios were considered in both MAICs of the RWE cohorts¹⁹:

• Scenario 1 included all matching variables considered most relevant by expert opinion (grade, histology [1 patient from GARNET with unknown histology was removed], number of prior platinum-based therapies in the advanced or recurrent setting [patients from GARNET with 0 or > 2 therapies were removed]).

• Scenario 2 included all matching variables from scenario 1 except grade, because of poor reporting of grade in the RWE cohort (histology, number of prior platinum-based therapies in the advanced or recurrent setting). Scenario 2 was considered to be a sensitivity analysis of scenario 1 that was supposed to maintain a larger ESS.

• Scenario 3 included the matching variables that were considered prognostic, based on the empirical regression analyses, with the exception of grade. Moreover, ECOG Performance Status was only used as a matching variable in the sensitivity analysis for patients with an ECOG Performance Status of 0 or 1 from the RWE cohort.

GARNET Versus Published Literature MAIC Methods (MAIC Report 3)

Three studies were eligible for the MAIC (MAIC report 3) based on the SLR. GARNET was a phase I single-arm trial of dostarlimab. McMeekin et al. (2015)⁵⁶ was a global, open-label, phase III RCT comparing ixabepilone with paclitaxel or doxorubicin, Julius et al. (2013)⁵⁷ was a retrospective review of medical records in the US focusing on pegylated liposomal doxorubicin, and Mazgani et al. (2008)⁵⁸ was a retrospective analysis of carboplatin plus paclitaxel in patients with relapsed EC at 1 centre in Vancouver, British Columbia. The individual comparator studies were assessed for eligibility based on reporting of outcomes and appropriateness of the MAIC based on sample size and population (had to be adequately comparable to GARNET).¹⁹ Differences in eligibility criteria across studies were not discussed; however, MMR and MSI-H status was not accounted for in the comparator studies.

The specific (regression-type) modelling approach used in the analyses depended on the outcome data type. All P values presented were exploratory in nature and not conclusive. The tests performed were conducted post hoc, were not formally powered, and were not adjusted for multiplicity, so the reported P values are not confirmatory. When available, KM curves were digitized (using GetData Graph Digitizer 2.26) and the algorithm of Guyot et al. (2012)⁵⁵ was used to produce reconstructed IPD for the aggregate trial data. As the GARNET trial had less than 36 months of follow-up for OS, KM data from each digitized comparator study were restricted (censored) to 36 months or less to improve the comparability of results.¹⁹

For all studies, each KM curve was digitized separately and the reconstructed IPD were synthesized. Curves created from these reconstructed IPD were plotted against curves displaying the raw digitized survival probability time points, and visual comparisons were made against the published curves. These pseudo-IPD were considered alongside the IPD from the GARNET trial. A weighted Cox regression was then applied to the dataset, combining the reconstructed IPD with the GARNET data (no covariates other than treatment). To assess the assumption of proportional hazards in the Cox survival modelling between the treatment groups compared, a (MAIC-weighted) log-cumulative hazard plotted against log time was visually assessed for parallelism, and Schoenfeld residuals plots were produced. If the PHA was violated, an accelerated-failure-time model with a Weibull distribution was used.¹⁹

A method of moments analytical technique was used to produce individual patient-specific weights for patients from the GARNET trial to produce weighted prognostic means. Comparison of means between the aggregate trial data before and after adjustment of the GARNET trial, as well as the ESS were presented.¹⁹

In MAIC report 3, pre-specified prognostic factors were identified in the same manner as for the RWE cohort MAICs. Additionally, BMI, number of prior anti-cancer regimens, and MMR or MSI molecular profile were considered to be prognostic factors but were not included in weighting. Because of limitations of data availability, the GARNET data could only be matched on a small number of prognostic factors for McMeekin et al. (2015)⁵⁶ (age, ECOG, histology, race), Julius et al. (2013)⁵⁷ (age, race, number of prior chemotherapies), and Mazgani et al. (2008)⁵⁸ (histology).¹⁹ The distribution of baseline characteristics by treatment group were summarized before and after matching.

To minimize heterogeneity between separate study populations before statistical adjustments, the comparator study inclusion and exclusion criteria were applied to the GARNET IPD. The definition of outcomes in the comparator studies were also compared. Disease expertise was sought to identify the most likely criteria to have been used in each study.¹⁹ The inclusion criteria of GARNET that were modified to match the comparator studies and the resulting modified sample sizes are summarized in Table 22.

No sensitivity or subgroup analyses were conducted for MAIC report 3.

Results of MAIC Analyses

Summary of Included Studies

Key design characteristics and sources of heterogeneity of the studies included in the RWE cohort MAICs and the published literature MAICs are summarized in Table 24.

GARNET Versus NCRAS RWE Cohort (MAIC Reports 1 and 2)

In total, the GARNET cohort included 129 patients and the base-case RWE cohort included 999 patients. The RWE cohort used for the sensitivity analysis (ECOG ≤ 1) included 501 patients.¹⁹

There was some variation in the baseline characteristics of patients in GARNET and MAIC reports 1 and 2 (initial and treatment-specific RWE cohorts, respectively). Age range in the study cohorts varied, with 31.2%% to 51.2% of patients younger than 65 years and with 48.8% to 68.8% of patients older than 65 years. The majority of patients were White (GARNET = 76.0%; RWE cohorts = 75.0% to 92.5%); however, the GARNET trial had a greater proportion of unknown race (15.5% versus 0.7% to 5.4%). The proportion of patients with endometrioid histology was greater in GARNET (69.8%) than in the initial RWE cohort and treatment-specific cohorts (37.6% to 44.6%). In general, the GARNET trial had a notably higher proportion of patients with an ECOG Performance Status of 0 or 1 (ECOG status for many patients in the RWE cohorts was unknown), FIGO stage I/II (44.2% versus 15.8% to 35.4%), and grade 1/2 disease (67.4% versus 22.4% to 29.0%). The number of prior platinum-based therapies in the advanced or recurrent setting was 1 for all patients in the RWE cohort, whereas in the GARNET ITT cohort, it was 0 or 2 or more for some patients, based on the method of recording prior treatment for the RWE cohort. All patients in the GARNET trial were dMMR or MSI-H. There was no information on MMR or MSI-H status in the initial RWE cohort population.¹⁹

GARNET Versus Published Literature (MAIC Report 3)

In MAIC report 3, comparing GARNET with published literature, 23 records representing 14 trial publications and 13 unique studies were included in the SLR. Three unique studies — McMeekin et al. (2015),⁵⁶ Julius et al. (2013),⁵⁷ and Mazgani et al. (2008)⁵⁸ — were eligible for the MAIC. It was noted that several identified studies were not deemed appropriate for a MAIC, primarily due to either a low sample size or the study population being systemically different than the GARNET trial.¹⁹

The GARNET, Julius et al. (2013),⁵⁷ and Mazgani et al. (2008)⁵⁸ studies were nonrandomized studies of dostarlimab, pegylated liposomal doxorubicin, and carboplatin plus paclitaxel, respectively, whereas the McMeekin et al. (2015)⁵⁶ study was an open-label, RCT comparing ixabepilone with paclitaxel or doxorubicin. Only the treatment arm of paclitaxel or doxorubicin was included in the MAIC with McMeekin et al. (2015).⁵⁶ The sample size of included studies ranged from 31 to 248 patients.¹⁹

An assessment of the baseline characteristics before weighting was not provided in MAIC report 3. The majority of baseline characteristics that were considered important, including age range, FIGO stage, disease grade, prior platinum-based therapy, and surgery, were not reported in the published literature. When reported, race was relatively similar between GARNET and the published literature trials. When reported, the proportion of patients with endometrioid histology and the proportion of patients with an ECOG Performance Status of 0 and 1 was greater in GARNET than in the studies included in the published literature MAIC.¹⁹

Results

GARNET Versus NCRAS Initial RWE Cohort (MAIC Report 1)

OS, PFS, and TTD in the GARNET ITT and initial RWE cohort are summarized in Table 25. For the initial RWE cohort, matching scenario 1, where studies were matched on histology, grade, and number of prior platinum-based therapies, resulted in the smallest ESS, with just 34 patients. The ESS was higher for scenario 2 and scenario 3, which did not include grade for 74 and 75 patients, respectively.¹⁹

Under all matching scenarios, dostarlimab was favoured over the current treatment paradigm for OS (scenario 1 HR = 0.52 [95% CI, 0.29 to 0.92]; scenario 2 HR = 0.35 [95% CI, 0.22 to 0.55]; scenario 3 HR = 0.31 [95% CI, 0.20 to 0.49]). Results for the sensitivity analysis of patients with an ECOG Performance Status of 0 or 1 were consistent with the base case. For PFS, 2 analyses were conducted: PFS from GARNET compared with TTNT from the RWE cohort; and PFS from GARNET compared with TTD from the RWE cohort. In general, the median PFS and PFS rate at 6, 12, and 18 months in all scenarios was higher than the comparator of current treatment paradigm, particularly for the analysis comparing GARNET PFS with RWE TTD. The median TTD and TTD rates at 6, 12, and 18 months were higher in all scenarios for dostarlimab than in the RWE cohort.¹⁹

The analysis of the leave-1-out approach to assess the impact of each potential covariate on ESS and estimated treatment effect is summarized for OS in Table 26. Omission of disease grade had the largest impact on ESS and estimated treatment effect, as a substantial proportion of patients in the RWE cohort had unknown grade.¹⁹

Table 25: Survival Outcomes Before and After Matching — GARNET Versus Initial RWE Cohort (MAIC Report 1)

CI = confidence interval; ECOG = Eastern Cooperative Oncology Group performance status; ESS = effective sample size; HR = hazard ratio; ITT = intention to treat; MAIC = matching-adjusted indirect comparison; NE = not estimable; OS = overall survival; PFS = progression-free survival; RWE = real-world evidence; vs. = versus; TTD = time to treatment discontinuation.

^aScenario 1: Matching variables are histology, grade, and number of prior platinum-based therapies.

^bScenario 2: Matching variables are histology and number of prior platinum-based therapies.

^cScenario 3: Matching variables are race/ethnicity, histology, stage at initial diagnosis, and surgery.

^dP values are exploratory and not adjusted for multiplicity, thus at an increased risk of type I error.

Source: Sponsor-submitted ITC reports¹⁹

Table 26: Summary of Leave-1-Out Approach — Initial RWE Cohort (Base-Case and Scenario Analyses, MAIC Report 1)

CI = confidence interval; EC = endometrial cancer; ECOG = Eastern Cooperative Oncology Group performance status; ESS = effective sample size; FIGO = Federation of Gynecology and Obstetrics; HR = hazard ratio; MAIC = matching-adjusted indirect comparison; OS = overall survival; RWE = real-world evidence.

^aFor the base-case RWE cohort, ECOG Performance Status is not used as a matching variable because 49.8% of patients have an unknown ECOG status and adjusting for this imbalance would correspond to performing the analysis in the RWE ECOG ≤ 1 cohort (where patients with unknown ECOG have been removed).Source: Sponsor-submitted ITC reports.¹⁹

GARNET Versus NCRAS Treatment-Specific Cohorts (MAIC Report 2)

For the treatment-specific MAICs, Table 27 summarizes the OS for dostarlimab compared with the 5 specific treatments in the RWE cohort. All results for scenario analyses using unweighted GARNET data were consistent with the primary analyses for OS. For paclitaxel monotherapy, the ESS in scenarios 1, 2, and 3 was 30, 72, and 63, respectively. Under all matching scenarios, the median OS (and HRs) favoured dostarlimab over paclitaxel monotherapy (scenario 1: HR = 0.36; 95% CI, 0.19 to 0.65; scenario 2: HR = 0.24; 95% CI, 0.15 to 0.40; scenario 3: HR = 0.18; 95% CI, 0.11 to 0.30), indicating that patients taking dostarlimab had a lower risk of death than patients receiving paclitaxel. The OS rates at 6, 12, and 18 months were greater for dostarlimab than for paclitaxel in the RWE cohort.¹⁹

For carboplatin plus paclitaxel, the ESS in scenarios 1, 2, and 3 was 36, 74, and 69, respectively. Dostarlimab was favoured over carboplatin plus paclitaxel only in scenarios 2 and 3 (scenario 2: HR = 0.48; 95% CI, 0.29 to 0.78; scenario 3: HR = 0.42; 95% CI, 0.25 to 0.68), but not scenario 1 (HR = 0.70; 95% CI, 0.39 to 1.28). The OS rates at 6, 12, and 18 months were greater for dostarlimab than for carboplatin plus paclitaxel in the RWE cohort.¹⁹

For carboplatin plus liposomal doxorubicin, the ESS in scenarios 1, 2, and 3 was 26, 69, and 74, respectively. Dostarlimab was favoured over carboplatin plus liposomal doxorubicin only in scenarios 2 and 3 (scenario 2: HR = 0.45; 95% CI, 0.27 to 0.76; scenario 3: HR = 0.40; 95% CI, 0.24 to 0.67), but not scenario 1 (HR = 0.74; 95% CI, 0.39 to 1.41). After matching, the 6-, 12-, and 18-month OS rates were greater for dostarlimab.¹⁹

For liposomal doxorubicin monotherapy, the ESS in scenarios 1, 2, and 3 was 37, 78, and 76, respectively. Under all matching scenarios, the median OS (and HRs) favoured dostarlimab over liposomal doxorubicin monotherapy (scenario 1: HR = 0.20; 95% CI, 0.09 to 0.44; scenario 2: HR = 0.17; 95% CI, 0.10 to 0.29; scenario 3: HR = 0.16; 95% CI, 0.09 to 0.27), indicating that patients taking dostarlimab had a lower risk of death than patients taking liposomal doxorubicin. Under all matching scenarios, the OS rates at 6 and 12 months were greater in the GARNET ITT cohort than in the RWE cohort.¹⁹

For carboplatin monotherapy, the ESS in scenarios 1, 2, and 3 was 23, 67, and 69, respectively. Dostarlimab was favoured over carboplatin in all matching scenarios (scenario 1: HR = 0.53; 95% CI, 0.29 to 0.98; scenario 2: HR = 0.32; 95% CI, 0.19 to 0.55; scenario 3: HR = 0.28; 95% CI, 0.16 to 0.48), indicating that patients taking dostarlimab had a lower risk of death than patients taking carboplatin monotherapy. Under all matching scenarios, the OS rates at 6, 12, and 18 months were greater in the GARNET ITT cohort compared with in the RWE cohort.¹⁹

Results for PFS for all comparators are summarized in Table 28. When compared with the RWE cohort TTNT and TTD, the median PFS was longer for dostarlimab than for all relevant comparators in all scenarios except scenario 1 of the TTNT comparison for paclitaxel monotherapy, carboplatin plus paclitaxel, carboplatin plus liposomal doxorubicin, and carboplatin monotherapy. For the TTD comparison, dostarlimab was favoured over all comparators in all scenarios except carboplatin monotherapy in scenario 1. The PFS rate for dostarlimab was greater in all scenarios and at all time points for the TTD comparison for all relevant comparators. In contrast with TTNT from the RWE cohort, the PFS rate for dostarlimab was greater than both paclitaxel and liposomal doxorubicin monotherapy at all time points and in all scenarios. The PFS rate was not greater at the 6-month assessment in any scenario for carboplatin plus paclitaxel, carboplatin plus liposomal doxorubicin, or carboplatin monotherapy. Otherwise, dostarlimab had a greater PFS rate at the 12- and 18-month time points in all scenarios.¹⁹