Tislelizumab plus chemotherapy versus chemotherapy alone as first-line treatment for advanced squamous non-small-cell lung cancer: final analysis of the randomized, phase III RATIONALE-307 trial

Abstract

Purpose

First-line tislelizumab plus chemotherapy significantly improved progression-free survival (PFS) versus chemotherapy alone in advanced squamous non-small-cell lung cancer (sq-NSCLC) at the interim analysis of the phase III RATIONALE-307 trial. We present the final analysis of this trial.

Patients and methods

Patients with treatment-naive, stage IIIB/IV, sq-NSCLC were randomized (1 : 1: 1) to 21-day cycles of i.v.: tislelizumab plus paclitaxel and carboplatin (arm A); tislelizumab plus nab-paclitaxel and carboplatin (arm B); or paclitaxel and carboplatin (arm C). The primary endpoint was independent review committee-assessed PFS; overall survival was a secondary endpoint.

Results

In total, 360 patients were randomized; 355 received treatment. At the final analysis (median study follow-up: 16.7 months), tislelizumab plus chemotherapy had a manageable safety profile, consistent with that at the interim analysis. Improvement in PFS was maintained for arms A and B versus C {hazard ratio (HR) 0.45 [95% confidence interval (CI) 0.33-0.62] and 0.43 (95% CI 0.31-0.60), respectively}. Overall survival HRs for arms A and B versus C were 0.68 (95% CI 0.46-1.01) and 0.75 (95% CI 0.50-1.12), respectively.

Conclusions

The RATIONALE-307 final analysis demonstrated superior clinical benefit with addition of tislelizumab to chemotherapy, and a manageable safety profile, as first-line treatment of advanced sq-NSCLC.

Highlights

• •1L tislelizumab plus chemotherapy prolonged PFS versus chemotherapy in Chinese patients with advanced/metastatic sq-NSCLC.
• •PFS benefits with tislelizumab plus chemotherapy versus chemotherapy alone were consistent across patient subgroups.
• •Tislelizumab plus chemotherapy demonstrated a manageable safety profile, with no new safety signals identified.
• •Results support the potential for tislelizumab plus chemotherapy to become a 1L option for advanced sq-NSCLC.

Introduction

Most patients diagnosed with non-small-cell lung cancer (NSCLC) present with advanced-stage, unresectable disease.¹ Historically, treatment with standard platinum-based chemotherapy (cisplatin or carboplatin) in this setting has been associated with limited survival benefit and significant toxicities.2, 3, 4

Inhibitors of programmed cell death protein 1 (PD-1) or its ligand, programmed death-ligand 1 (PD-L1), have revolutionized the treatment of NSCLC.¹ As monotherapies for advanced/metastatic NSCLC in the first-line setting, PD-1/PD-L1 inhibitors have been reported to prolong progression-free survival (PFS) and overall survival (OS) versus chemotherapy, particularly in patients with tumors expressing high levels of PD-L1.5, 6, 7, 8 In addition, first-line PD-1/PD-L1 inhibitors plus chemotherapy have been shown to improve PFS versus chemotherapy alone in several phase III studies in patients with advanced/metastatic NSCLC, unselected for PD-L1 expression level.9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 In trials focusing specifically on patients with squamous (sq)-NSCLC, however, OS results have varied, with significant improvements in KEYNOTE-407 and CameL-Sq, but not in IMpower131.^9,10,17

Tislelizumab, a monoclonal antibody with high affinity and binding specificity for the PD-1 receptor, was specifically engineered to minimize Fcγ receptor binding on macrophages.20, 21, 22 In two early-phase studies, tislelizumab monotherapy was generally well tolerated and demonstrated promising antitumor activity in patients with advanced lung cancers.^23,24 Additionally, in a phase II trial, first-line tislelizumab plus platinum-based chemotherapy demonstrated encouraging antitumor activity in patients with advanced lung cancer, with an objective response rate (ORR) of 67%-80% for sq-NSCLC, depending on the chemotherapy regimen used.²⁵

Previously, we published results from an interim analysis of a phase III trial of tislelizumab plus chemotherapy in patients with treatment-naive, advanced sq-NSCLC (RATIONALE-307).²⁶ At interim analysis (median study follow-up: 8.6 months), RATIONALE-307 met its primary endpoint of PFS assessed by an independent review committee (IRC). Tislelizumab plus chemotherapy (either paclitaxel and carboplatin or nab-paclitaxel and carboplatin) demonstrated superior PFS versus paclitaxel and carboplatin alone {hazard ratio (HR) 0.52 [95% confidence interval (CI) 0.37-0.74; P < 0.001] with tislelizumab plus paclitaxel and carboplatin and 0.48 (95% CI 0.34-0.68; P < 0.001) with tislelizumab plus nab-paclitaxel and carboplatin}.²⁶ Median PFS was prolonged with both tislelizumab-containing regimens (7.6 months for both) versus chemotherapy alone (5.5 months). Additionally, ORR was greater, and duration of response (DoR) was longer with tislelizumab plus chemotherapy versus chemotherapy alone. OS was a secondary study endpoint, but data were not mature at the time of interim analysis.²⁶

Here we present the efficacy and safety results from the final analysis of RATIONALE-307. We also report the results of subsequent ad hoc analyses of OS that were conducted to examine longer-term outcomes of tislelizumab plus chemotherapy using more mature data, and describe findings of a supportive analysis conducted to adjust OS outcomes for the effect of in-study crossover to tislelizumab in the chemotherapy alone arm.

Materials and methods

Study design

The study methodology for RATIONALE-307 has been previously published.²⁶ In brief, RATIONALE-307 (clinicaltrials.gov NCT03594747) was an open-label, randomized, multicenter, phase III trial conducted in China. RATIONALE-307 was conducted in conformance with Good Clinical Practice guidelines, the principles of the Declaration of Helsinki, and local regulatory requirements. Patients gave written informed consent before enrollment, and the participating institutions provided Institutional Review Board approval.

Patients

Eligible patients were aged 18-75 years, and were treatment-naive for histologically confirmed, locally advanced (stage IIIB) or metastatic (stage IV) sq-NSCLC, classified by the American Joint Committee on Cancer 7th Edition of the Cancer Staging Manual,²⁷ and had an Eastern Cooperative Oncology Group performance status ≤1. Patients were eligible if their disease was not amenable to curative surgery or radiotherapy, and was measurable [per Response Evaluation Criteria in Solid Tumors, version 1.1 (RECIST v1.1)]. Patients with known EGFR-sensitizing mutations or ALK gene translocations, a history of interstitial lung disease, or non-infectious pneumonitis were ineligible.

Interventions

Patients were randomized (1 : 1 : 1) to receive one of the following regimens i.v., in 21-day cycles (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop.2024.103727): arm A: tislelizumab (200 mg, day 1) plus paclitaxel (175 mg/m², day 1) and carboplatin [area under the plasma concentration-time curve (AUC) of 5, day 1]; arm B: tislelizumab (200 mg, day 1) plus nab-paclitaxel (100 mg/m², days 1, 8, and 15) and carboplatin (AUC of 5, day 1); arm C: paclitaxel (175 mg/m², day 1) and carboplatin (AUC of 5, day 1). Chemotherapies were administered for 4-6 cycles, at the investigator’s discretion. Randomization was stratified by disease stage (IIIB versus IV) and level of tumor cell (TC) PD-L1 expression (<1% TC expression versus 1%-49% versus ≥50%), and carried out using an interactive response technology system.

Endpoints and assessments

The primary endpoint was IRC-assessed PFS in arms A and B versus arm C. Secondary endpoints included OS, IRC- and investigator-assessed ORR and DoR, investigator-assessed PFS, health-related quality of life (not reported herein), PD-L1 expression as a response biomarker, and safety and tolerability. Endpoints were summarized for patients in arm A versus arm C and arm B versus arm C, separately. Safety and tolerability were assessed throughout the trial by monitoring adverse events (AEs) (graded per National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0 and coded using Medical Dictionary for Regulatory Activities version 23.0).

Subgroup analyses of PFS were conducted to determine whether treatment effect was consistent across various subgroups, including by disease stage and PD-L1 expression. During screening, TC PD-L1 expression was assessed using the VENTANA PD-L1 (SP263) assay at a central laboratory.

Immune-mediated AEs (imAEs) were classified retrospectively by a programmatic algorithmic approach and were based on a defined list of preferred terms, without manual medical adjudication. The original outlined methodology was updated at the request of regulatory organizations to mitigate potential bias resulting from the prespecified manual medical adjudication used at the interim analysis timepoint.²⁶

Statistical analyses

The study sample size was determined by the number of PFS events required to demonstrate superiority of the tislelizumab-containing arms versus the chemotherapy alone arm. Approximately 173 PFS events were required at the final analysis for the primary endpoint. Since the primary endpoint was met at the interim analysis,²⁶ however, no formal significance testing was carried out for this final analysis of PFS, therefore the P values presented are descriptive. RATIONALE-307 was not powered to demonstrate superiority of the tislelizumab-containing arms versus the chemotherapy alone arm for OS.

Efficacy analyses were assessed in the intent-to-treat (ITT) analysis set, which included all randomized patients. Safety and tolerability data were summarized from the safety analysis set, which included patients who received any dose of tislelizumab and/or chemotherapy. Categorical variables were summarized by the number and percentage of patients; continuous variables were reported using descriptive statistics. Time-to-event endpoints were estimated using Kaplan–Meier analysis, with the Brookmeyer and Crowley method used to construct 95% CIs for median PFS, OS, and DoR. HRs for comparisons between arms A or B with arm C were estimated using the stratified Cox proportional hazards model, and a stratified log-rank test was carried out to test the difference between treatment arms. Differences in ORR were assessed by the stratified Cochran–Mantel–Haenszel chi-square test.

A post hoc supportive analysis using a two-stage method²⁸ was carried out to estimate the in-study crossover effect on post-progression survival (PPS) using data from patients in the control arm (arm C) who progressed per IRC assessment before any subsequent anticancer therapy. Further information is included in the Supplementary Appendix, available at https://doi.org/10.1016/j.esmoop.2024.103727.

Results

Patients and treatment

Between 30 July 2018 and 30 September 2020, 360 eligible patients were randomized to arm A [120 (33.3%)], arm B [119 (33.1%)], or arm C [121 (33.6%)] (Supplementary Figure S2, available at https://doi.org/10.1016/j.esmoop.2024.103727). Five randomized patients (one in arm B and four in arm C) did not receive study treatment; these patients were included in the ITT analysis set but not in the safety analysis set. At the final analysis data cut-off (30 September 2020), 31 patients in arm A (25.8%) and 34 in arm B (28.6%) remained on treatment; all patients in arm C had discontinued treatment, of whom 81 patients (66.9%) had completed the chemotherapy regimen. Median study follow-up (ITT analysis set) was 16.7 months. Demographic and disease baseline characteristics were well balanced between arms (Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2024.103727), as previously published.²⁶

Median duration of exposure to tislelizumab increased from 10 cycles in both arm A (range: 1-20) and arm B (range: 1-19) at interim analysis²⁶ to 13 cycles (range: 1-32) in both arms at final analysis. At the time of final analysis, 50.8% of patients in arm A and 51.7% of patients in arm B had received >12 cycles of tislelizumab. All patients had completed or discontinued chemotherapy treatment by the interim analysis timepoint, and the number of chemotherapy cycles received was therefore consistent between the interim and final analyses,²⁶ with a median of 4.5 cycles (range: 1-6) completed in arm A and 4 cycles (range: 1-6) completed in both arm B and arm C.

Safety and tolerability

An overall summary of treatment-emergent AEs (TEAEs) for the safety analysis set at final analysis is presented in Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2024.103727. The incidence of TEAEs of any grade or ≥grade 3 was similar in all treatment arms (any TEAE: 100.0%, 99.2%, and 100.0% of patients in arms A, B, and C, respectively; ≥grade 3 TEAE: 89.2%, 87.3%, and 84.6% of patients, respectively). Although the proportion of patients with serious TEAEs was higher in arms A and B [43.3% (n = 52) and 42.4% (n = 50), respectively] compared with arm C [24.8% (n = 29)], exposure-adjusted event rates were lower in arms A and B (6.3 and 5.9 per 100 patient-months exposure, respectively) compared with arm C (10.0 per 100 patient-months exposure) (Supplementary Table S3, available at https://doi.org/10.1016/j.esmoop.2024.103727). In the two tislelizumab-based arms, the incidence of ≥grade 3 TEAEs and serious TEAEs was greatest during the first 4-6 cycles of treatment, when chemotherapy was administered in combination with tislelizumab (≥grade 3 TEAEs: 87.5% and 84.7% of patients in arms A and B, respectively; serious TEAEs: 30.0% and 31.4%, respectively). Relative to the combination period, the incidence of such events in the tislelizumab-based arms was lower during the subsequent period without chemotherapy (≥grade 3 TEAEs: 20.8% and 22.0% of patients in arms A and B, respectively; serious TEAEs: 18.3% and 16.9%, respectively).

At final analysis, TEAEs leading to death occurred in 3.3% (n = 4), 5.9% (n = 7), and 4.3% (n = 5) of patients in arms A, B, and C, respectively. One TEAE leading to death in arm A (hydrocephalus) and two TEAEs leading to death in arm B (one case of hepatic failure and one case reported only as death) were considered by investigators to be related to all three study regimen components (tislelizumab plus both chemotherapy components). These cases may have been confounded by the patients’ disease characteristics at baseline (brain metastases for the hydrocephalus case and liver metastases for the hepatic failure case), baseline comorbidities (pneumonia for the case of ‘death’), or the underlying disease (for both the hepatic failure and ‘death’ cases). No other TEAEs leading to death were considered to be related to tislelizumab.

TEAEs reported in ≥20% of patients are presented in Supplementary Table S4, available at https://doi.org/10.1016/j.esmoop.2024.103727. TEAEs of all grades with an incidence that was ≥10% higher in arm A relative to arm C included: increased aspartate aminotransferase (AST), increased alanine aminotransferase (ALT), rash, decreased appetite, blood bilirubin increased, platelet count decreased, pneumonia, constipation, and pain in extremity. TEAEs of all grades with an incidence that was ≥10% higher in arm B relative to arm C included: increased AST, rash, platelet count decreased, nausea, decreased appetite, anemia, thrombocytopenia, and increased ALT. Treatment-related TEAEs (all grades) were reported in 99.2% (n = 119) of patients in arm A, 99.2% (n = 117) of patients in arm B, and 100.0% (n = 117) of patients in arm C, and ≥grade 3 treatment-related AEs were reported in 86.7%, 83.9%, and 80.3% of patients, respectively (Supplementary Table S5, available at https://doi.org/10.1016/j.esmoop.2024.103727).

ImAEs of all grades were reported in 53 patients (44.2%) in arm A and 60 patients (50.8%) in arm B (Supplementary Table S6, available at https://doi.org/10.1016/j.esmoop.2024.103727). Skin adverse reactions (4.2% and 2.5% of patients) and pneumonitis (3.3% and 5.1% of patients) were the most common categories of ≥grade 3 imAEs in arms A and B, respectively (Supplementary Table S6, available at https://doi.org/10.1016/j.esmoop.2024.103727).

Efficacy

At the final analysis data cut-off, 245 PFS events per IRC across all arms were observed in the ITT analysis set. Tislelizumab plus chemotherapy (arms A and B) versus chemotherapy (arm C) continued to show significant, meaningful improvements in IRC-assessed PFS (Figure 1A and B), consistent with the interim analysis.²⁶ At the final analysis, the stratified PFS HR was 0.45 (95% CI 0.33-0.62) for arm A versus arm C, and 0.43 (95% CI 0.31-0.60) for arm B versus arm C, both P < 0.0001. Median PFS was longer in both arm A [7.7 months (95% CI 6.7-10.4 months)] and arm B [9.6 months (95% CI 7.4-10.8 months)] compared with arm C [5.5 months (95% CI 4.2-5.6 months)]. PFS benefit observed with the addition of tislelizumab to chemotherapy was largely consistent across subgroups, including disease stage and PD-L1 expression subgroups (Figure 1C and D). Investigator-assessed PFS results were consistent with IRC-assessed PFS results, with a stratified HR of 0.34 (95% CI 0.25-0.47) for arm A versus arm C and 0.40 (95% CI 0.29-0.56) for arm B versus arm C (Supplementary Table S7, available at https://doi.org/10.1016/j.esmoop.2024.103727).

Figure 1.

Progression-free survival per RECIST version 1.1 by independent review committee: (A) for arm A versus arm C; (B) for arm B versus arm C; (C) by subgroup for arm A versus arm C; (D) by subgroup for arm B versus arm C (all in the ITT analysis set; final analysis data cut-off). CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; HR, hazard ratio; ITT, intent-to-treat; nPC, nab-paclitaxel plus carboplatin; PC, paclitaxel plus carboplatin; PD-L1, programmed death-ligand 1; PFS, progression-free survival; RECIST, Response Evaluation Criteria in Solid Tumors; TC, tumor cell. ^aStratification factors were disease stage (IIIB versus IV) and PD-L1 expression in TCs (≥50% TCs versus 1%-49% TCs versus <1% TCs). ^bPatients with unevaluable PD-L1 status were removed from the <1% subgroup. Data cut-off: 30 September 2020.

The IRC-assessed ORR was higher in arm A [74.2% (95% CI 65.4% to 81.7%)] and arm B [73.9% (95% CI 65.1% to 81.6%)] compared with arm C [47.9% (95% CI 38.8% to 57.2%)] (Table 1). The median time to response per IRC was 6.1 weeks (range: 5.1-51.7 weeks) in arm A, 6.3 weeks (range: 5.1-40.6 weeks) in arm B, and 6.2 weeks (range: 5.3-32.4 weeks) in arm C. Among responders, the median IRC-assessed DoR was also longer in arm A and arm B than in arm C [8.4 months (95% CI 5.0-15.8 months) in arm A, 8.6 months (95% CI 7.1-12.5 months) in arm B, and 4.3 months (95% CI 2.9-5.4 months) in arm C]. Among responders, 39/89 patients in arm A, 34/88 patients in arm B, and 12/58 patients in arm C still had complete or partial response at final analysis. When ORR was assessed in subgroups defined by PD-L1 expression, there was a trend towards increased ORR in arm A and arm B compared with arm C, regardless of PD-L1 expression subgroup (Supplementary Figure S3, available at https://doi.org/10.1016/j.esmoop.2024.103727). Investigator-assessed ORR and DoR analyses in the ITT analysis set showed higher ORRs and longer DoR in arm A and arm B compared with arm C (Supplementary Table S7, available at https://doi.org/10.1016/j.esmoop.2024.103727), consistent with IRC-assessed results.

Table 1.

Analysis of independent review committee-assessed disease response per RECIST v1.1 criteria (ITT analysis set; final analysis data cut-off)

	Arm A Tislelizumab + PC (n = 120)	Arm B Tislelizumab + nPC (n = 119)	Arm C PC (n = 121)
Best overall response – unconfirmed, n (%)
Complete response	7 (5.8)	8 (6.7)	1 (0.8)
Partial response	82 (68.3)	80 (67.2)	57 (47.1)
Stable disease	16 (13.3)	20 (16.8)	39 (32.2)
Progressive disease	12 (10.0)	5 (4.2)	11 (9.1)
Not evaluable	0 (0.0)	0 (0.0)	0 (0.0)
Missing	3 (2.5)	6 (5.0)	12 (9.9)
ORR, % (95% CI)	74.2 (65.4-81.7)	73.9 (65.1-81.6)	47.9 (38.8-57.2)
ORR difference versus arm C, % (95% CI)	27.0 (15.4-38.7)	26.1 (14.3-37.9)	—
Disease control rate, % (95% CI)	87.5 (80.2-92.8)	90.8 (84.1-95.3)	81.0 (72.9-87.6)

Data cut-off: 30 September 2020. Some best overall responses were categorized as missing because patients had no post-baseline tumor assessment. 95% CIs were calculated using the Clopper–Pearson method. ORR differences between arms were calculated using the Cochran–Mantel–Haenszel chi-square test, with actual stratification factors as strata. The paclitaxel + carboplatin arm was the reference group.

CI, confidence interval; ITT, intent-to-treat; nPC, nab-paclitaxel plus carboplatin; ORR, objective response rate; PC, paclitaxel plus carboplatin; RECIST, Response Evaluation Criteria in Solid Tumors.

At the final analysis data cut-off, there were 48 deaths (40.0% of patients) in arm A, 47 (39.5%) in arm B, and 52 (43.0%) in arm C. Median OS for arms A, B, and C was 22.8 months [95% CI 19.1 months-not estimable (NE)], NE (95% CI 18.6 months-NE), and 20.2 months (95% CI 16.0 months-NE), respectively. Stratified OS HRs were 0.68 (95% CI 0.46-1.01) between arms A and C and 0.75 (95% CI 0.50-1.12) between arms B and C (Figure 2A and B). At the final analysis data cut-off, 11 (9.2%) patients in arm A, 8 (6.7%) patients in arm B, and 74 (61.2%) patients in arm C had received subsequent immunotherapy. In arm C, all instances of subsequent immunotherapy included treatment with a PD-1/PD-L1 inhibitor, including 68 (56.2%) patients in arm C who crossed over to tislelizumab monotherapy (i.e. in-study crossover). In the two-stage analysis (carried out to evaluate the impact of in-study crossover to tislelizumab on PPS), the estimated shrink parameter was 0.581 (indicating the PPS time if patients did not receive crossover tislelizumab following progression was 58.1% of the original value), and the stratified OS HRs were 0.57 (95% CI 0.35-0.93) between arms A and C and 0.57 (95% CI 0.34-0.95) between arms B and C (Figure 3A and B).

Figure 2.

Overall survival: (A) for arm A versus arm C, and (B) for arm B versus arm C (all in the ITT analysis set; final analysis data cut-off). CI, confidence interval; HR, hazard ratio; ITT, intent-to-treat; NE, not estimable; nPC, nab-paclitaxel plus carboplatin; OS, overall survival; PC, paclitaxel plus carboplatin; PD-L1, programmed death-ligand 1; TC, tumor cell. ^aStratification factors were disease stage (IIIB versus IV) and PD-L1 expression in TCs (≥50% TCs versus 1%-49% TCs versus <1% TCs). Data cut-off: 30 September 2020.

Figure 3.

Post hoc analysis of overall survival using the two-stage method for: (A) arm A versus arm C, and (B) arm B versus arm C (in the ITT analysis set; final analysis data cut-off). CI, confidence interval; HR, hazard ratio; ITT, intent-to-treat; NE, not estimable; nPC, nab-paclitaxel plus carboplatin; OS, overall survival; PC, paclitaxel plus carboplatin; PD-L1, programmed death-ligand 1; TC, tumor cell. ^aStratification factors were: disease stage (IIIB versus IV) and PD-L1 expression in TCs (≥50% TCs versus 1%-49% TCs versus <1% TCs). ^bHR 95% CI estimated by bootstrap method. Data cut-off: 30 September 2020.

At the updated analysis data cut-off (15 July 2022) with longer study follow-up time (median study follow-up 20.5 months), stratified OS HRs for arm A and arm B versus arm C were 0.69 (95% CI 0.50-0.95) and 0.84 (95% CI 0.61-1.14), respectively (Figure 4A and B). Subsequent immunotherapy had been received by 18 (15.0%) patients in arm A, 13 (10.9%) patients in arm B, and 77 (63.6%) patients in arm C. In arm C, 71 patients (58.7%) had crossed over to tislelizumab. Among the 77 patients in arm C who crossed over to immunotherapy, the median time from last dose of chemotherapy to subsequent immunotherapy was 10.3 weeks (minimum time to crossover was 0.1 weeks). At this later data cut-off, updated stratified OS HRs from the two-stage analysis for arms A and B versus C were 0.57 (95% CI 0.37-0.87) and 0.65 (95% CI 0.40-1.05), respectively (Figure 4C and D).

Figure 4.

Ad hoc updated overall survival analyses for: (A) arm A versus arm C; (B) arm B versus arm C; (C) using the two-stage method for arm A versus arm C; (D) using the two-stage method for arm B versus arm C (all in the ITT analysis set; ad hoc updated analysis data cut-off). CI, confidence interval; HR, hazard ratio; ITT, intent-to-treat; nPC, nab-paclitaxel plus carboplatin; OS, overall survival; PC, paclitaxel plus carboplatin; PD-L1, programmed death-ligand 1; TC, tumor cell. ^aStratification factors were: disease stage (IIIB versus IV) and PD-L1 expression in TCs (≥50% TCs versus 1%-49% TCs versus <1% TCs). ^bHR 95% CI estimated by bootstrap method. Data cut-off: 15 July 2022. In arms A, B, and C, the 12-month OS rates (95% CI) were 72.7% (63.7% to 79.9%), 75.9% (67.1% to 82.7%), and 69.8% (60.3% to 77.4%), respectively; the 24-month OS rates were 53.0% (43.6% to 61.6%), 48.3% (39.0% to 57.0%), and 40.4% (31.2% to 49.4%), respectively; and the 36-month OS rates were 38.2% (29.4% to 46.9%), 30.9% (22.7% to 39.4%), and 29.1% (20.9% to 37.8%), respectively. In the two-stage analysis, in arms A, B, and C, the 12-month OS rates (95% CI) were 72.7% (63.7% to 79.9%), 75.9% (67.1% to 82.7%), and 61.6% (51.9% to 69.9%), respectively; the 24-month OS rates were 53.0% (43.6% to 61.6%), 48.3% (39.0% to 57.0%), and 33.3% (24.7% to 42.1%), respectively; and the 36-month OS rates were 38.2% (29.4% to 46.9%), 30.9% (22.7% to 39.4%), and 25.1% (15.3% to 36.2%), respectively.

Discussion

This final analysis of the phase III RATIONALE-307 trial of first-line tislelizumab plus paclitaxel or nab-paclitaxel and carboplatin in patients with advanced/metastatic sq-NSCLC confirmed the improvements in PFS, ORR, and DoR versus paclitaxel and carboplatin alone, as previously reported at interim analysis.²⁶ Notably, with longer follow-up, median PFS in the tislelizumab plus nab-paclitaxel and carboplatin arm increased by 2 months from interim (7.6 months) to final analysis (9.6 months). Furthermore, subgroup analyses of PFS indicated that the clinical improvements were greater with tislelizumab plus chemotherapy versus chemotherapy alone regardless of PD-L1 expression status, disease stage, age, sex, smoking, and presence of liver metastases at baseline. These findings are broadly consistent with those of other studies of first-line PD-1/PD-L1 inhibitors plus chemotherapy in patients with advanced/metastatic sq-NSCLC.^9,10,17,18

RATIONALE-307 was designed and powered to assess the superiority of tislelizumab plus chemotherapy versus chemotherapy alone for PFS. OS was a secondary endpoint, but the study was not designed with sufficient power and sample size to formally test for superiority. OS analysis results favored both tislelizumab plus chemotherapy arms versus chemotherapy alone at final analysis. The OS benefits in the overall study population in the tislelizumab plus chemotherapy arms were maintained at the subsequent updated ad hoc analysis.

Although OS improvement is often considered the most relevant measure of clinical benefit in oncology drug trials, OS assessments require large sample sizes for sufficient power and can be confounded by patient withdrawal, loss to follow-up, and use of effective subsequent lines of therapy, including in-study crossover.^29,30 Consequently, OS benefits may be difficult to detect.³⁰ At the latest data cut-off, a high proportion of patients in the chemotherapy alone arm had received subsequent immunotherapy [63.6% of patients, of whom most (71/77) had in-study crossover to tislelizumab]. To assess the impact of in-study crossover on OS a two-stage method was used,²⁸ which showed a favorable OS in the tislelizumab plus chemotherapy groups versus chemotherapy alone. At the latest data cut-off, OS HRs using the two-stage method (0.57 and 0.65 for arms A and B versus C, respectively) were improved relative to those from the main analysis that did not account for in-study crossover (0.69 and 0.84, respectively). These findings suggest a more pronounced survival benefit for patients who initially received tislelizumab plus chemotherapy versus chemotherapy alone. This benefit may be more substantial than the OS values reported in the main analysis, primarily due to a high proportion of in-study crossover.

Previous phase III trials of PD-1/PD-L1 inhibitors plus chemotherapy in patients with stage IIIB/IV sq-NSCLC include KEYNOTE-407, IMpower131, and CameL-Sq.^9,10,17,18 KEYNOTE-407 enrolled patients with stage IV sq-NSCLC and demonstrated improved PFS and OS with pembrolizumab plus chemotherapy (carboplatin with paclitaxel or nab-paclitaxel) versus placebo plus chemotherapy.^10,18 Subsequent immunotherapy treatment was permitted for the placebo plus chemotherapy arm in KEYNOTE-407 (49.1% of patients received a PD-1/PD-L1 inhibitor, including 40.1% with in-study crossover to pembrolizumab),¹⁸ but occurred in fewer patients than in RATIONALE-307 (in which 63.6% of patients received subsequent immunotherapy at the latest data cut-off, including 58.7% with in-study crossover to tislelizumab). More recently, in the CameL-Sq trial, patients with stage IIIB/IV sq-NSCLC demonstrated improved PFS and OS with camrelizumab plus chemotherapy (carboplatin plus paclitaxel) versus chemotherapy alone.¹⁷ Again, subsequent immunotherapy treatment was permitted for the chemotherapy alone arm (53% of patients received a PD-1/PD-L1 inhibitor, including 47% with in-study crossover to camrelizumab),¹⁷ but occurred in slightly fewer patients than in RATIONALE-307. In contrast to KEYNOTE-407 and CameL-Sq, the IMpower131 trial, which enrolled patients with stage IV sq-NCSLC, demonstrated a significant PFS improvement with atezolizumab plus chemotherapy (carboplatin with nab-paclitaxel), but not in OS—IMpower131 prohibited in-study crossover to atezolizumab in the chemotherapy alone arm, but permitted the use of other subsequent immunotherapies, which were received by 43% of patients.⁹

While cross-trial comparisons must be interpreted cautiously, median OS in the chemotherapy alone comparator arm in RATIONALE-307 (20.2 months at the final analysis cut-off) was longer than in the chemotherapy alone comparator arms of CameL-sq (14.5 months), IMPower131 (13.5 months), KEYNOTE-407 (11.6 months), CheckMate 9LA (10.9 months), and CHOICE-1 (17.1 months),^9,17, 18, 19^,31 suggesting that trial-specific variables (e.g. nuances of patient populations and differences in subsequent treatment in the chemotherapy alone arm) may have influenced the respective OS findings. Notably, RATIONALE-307, CameL-Sq, and CHOICE-1 were conducted in China only, whereas KEYNOTE-407 and IMPower131 were global studies with comparatively few Asian patients (19% of patients were enrolled in East Asia in KEYNOTE-407 and 11% of patients were of Asian ethnic origin in IMPower131).^9,10,17,18 Among patients with NSCLC, Asian populations have a better prognosis versus non-Asian populations,32, 33, 34, 35 and as such, the longer OS observed in RATIONALE-307 and CameL-Sq compared with KEYNOTE-407 and IMPower131 could in part be related to the inclusion of only Asian patients in these trials. Nevertheless, despite the better prognosis in Asian versus non-Asian patients, meta-analyses report that the magnitude of treatment benefit (i.e. OS HR) of first-line therapy with immune checkpoint inhibitors plus chemotherapy versus chemotherapy alone is similar in Asian and non-Asian populations with advanced NSCLC.^36,37 Thus, both Asian and non-Asian patients benefit from such treatment.

Compared with the interim analysis,²⁶ no new safety signals were identified, and no major changes in the safety profile of tislelizumab-based therapy were associated with the longer duration of treatment and follow-up at the final analysis. Despite the increase in tislelizumab exposure between interim and final analyses (median three additional cycles), few additional patients experienced serious AEs [eight (6.7%) additional patients in arm A and five (4.3%) additional patients in arm B].²⁶ The results of the retrospective programmatic imAE reporting method were consistent with those reported during the interim analysis for which the prespecified manual adjudication method was employed, confirming that the programmatic methodology is preferable for the consistency and reproducibility of the reporting of imAEs; the rationale for using the programmatic methodology will be published at a later date.

Potential bias due to the open-label study design was mitigated by randomization using an interactive response technology system and blinded IRC review of all radiological images for the efficacy analysis. Although the supportive OS analysis using the two-stage method is a recognized methodology for adjustment of crossover effect on OS,²⁸ however, which is applicable when treatment switching can only occur after a specific disease-related time-point, such as disease progression, it was subject to some limitations. To effectively apply the two-stage method, covariates that could impact PPS and that were not balanced between switchers and non-switchers were identified and fitted in an accelerated failure time model to estimate the effect of crossover. The potential bias of the two-stage method under this assumption was reduced somewhat after justifying all the covariates collected from the study. The challenge of identifying potential risk factors and the availability of corresponding data at the time of disease progression could have affected the outcome of this approach. In addition, since data from relatively few patients with disease progression per IRC before any subsequent anticancer therapy in the chemotherapy alone arm were used in the estimation process for the two-stage method (83 patients at the final analysis cut-off and 86 at the subsequent updated ad hoc cut-off), the results should be interpreted with caution.

In conclusion, the final analysis of the RATIONALE-307 trial provides evidence for the sustained clinical benefit of tislelizumab plus platinum-based chemotherapy versus chemotherapy alone in the first-line treatment of patients with advanced sq-NSCLC. The safety of the tislelizumab plus chemotherapy regimen remained manageable with a longer follow-up in the final analysis and was consistent with that established at the interim analysis.²⁶