First-line tislelizumab plus chemotherapy versus placebo plus chemotherapy in unresectable advanced/metastatic esophageal squamous cell carcinoma: minimum 3-year survival follow-up of a randomized phase 3 trial (RATIONALE-306)

Harry H Yoon; Ken Kato; Eric Raymond; Richard Hubner; Yongqian Shu; Yueyin Pan; Sook Ryun Park; Takashi Kojima; Chen-Yuan Lin; Lucjan Wyrwicz; David Tougeron; Ryu Ishihara; Kaijun Wang; Yanyan Peng; Jianming Xu

doi:10.1038/s41467-026-77327-0

. 2026 Sep 3;17:10499. doi: 10.1038/s41467-026-77327-0

First-line tislelizumab plus chemotherapy versus placebo plus chemotherapy in unresectable advanced/metastatic esophageal squamous cell carcinoma: minimum 3-year survival follow-up of a randomized phase 3 trial (RATIONALE-306)

Harry H Yoon ¹, Ken Kato ², Eric Raymond ³, Richard Hubner ⁴, Yongqian Shu ⁵, Yueyin Pan ⁶, Sook Ryun Park ⁷, Takashi Kojima ⁸, Chen-Yuan Lin ⁹, Lucjan Wyrwicz ¹⁰, David Tougeron ¹¹, Ryu Ishihara ¹², Kaijun Wang ¹³, Yanyan Peng ¹⁴, Jianming Xu ^15,^✉

PMCID: PMC13639055 PMID: 42834033

Abstract

RATIONALE-306 previously demonstrated improved overall survival (OS) in participants treated with tislelizumab plus chemotherapy versus those receiving placebo plus chemotherapy as first-line treatment for advanced/metastatic esophageal squamous cell carcinoma, with a minimum follow-up of approximately 15 months. Exploratory long-term efficacy, safety, and quality of life (QoL) analyses after 36.0 months’ minimum follow-up are reported here. Patients (649) were randomized 1:1 to tislelizumab 200 mg or placebo with investigator-chosen chemotherapy. Tislelizumab plus chemotherapy improves OS versus placebo plus chemotherapy in all patients (17.2 versus 10.6 months; hazard ratio, 0.70; 95% confidence interval, 0.59-0.80). At 3 years, 15.0% of patients treated with tislelizumab plus chemotherapy had not progressed versus 2.9% of those receiving placebo plus chemotherapy. No new safety signals have been identified. Median time to deterioration of physical functioning—a key QoL domain—has not been reached with tislelizumab plus chemotherapy versus 18.8 months with placebo plus chemotherapy. With longer follow-up, tislelizumab plus chemotherapy demonstrates long-lasting efficacy, tolerable safety, and favorable patient-reported outcomes, making it a suitable option for this patient population. ClinicalTrials.gov: NCT03783442.

Subject terms: Chemotherapy, Cancer immunotherapy, Oesophageal cancer, Phase III trials

The RATIONALE-306 trial demonstrated improved overall survival in participants with advanced/metastatic esophageal squamous cell carcinoma treated with tislelizumab plus chemotherapy comparing with those who received placebo plus chemotherapy. Here they report its exploratory long-term efficacy/safety/quality of life analysis after 36.0 months follow-up.

Introduction

Esophageal cancer is the seventh leading cause of cancer-related mortality globally¹; esophageal squamous cell carcinoma (ESCC) is the most common subtype, accounting for up to 90% of cases worldwide, with incidences rising in Asia^2,3. Monoclonal anti–programmed cell death protein 1 (PD-1) antibodies in combination with platinum-based chemotherapy have shown superior survival outcomes compared with platinum-based chemotherapy alone as first-line treatment for ESCC^3–8.

A higher level of programmed death-ligand 1 (PD-L1) expression is associated with a greater magnitude of benefit to PD-1/PD-L1 inhibitor-based therapy. PD-1 inhibitors such as tislelizumab, nivolumab or pembrolizumab in combination with chemotherapy have been found to be beneficial in patients with tumors expressing PD-L1. However, health authorities have chosen varying PD-L1 expression cutoffs for approval according to different scoring methods based on tumor, desmoplastic stroma and immune cell expression of PD-L1. Further, there is a regional variability in the required scoring methods for assessment of PD-L1.

Some of the commonly used PD-L1 scoring methods include the tumor cell (TC) score, tumor proportion score (TPS), combined positive score (CPS) and Tumor Area Positivity (TAP) score. Tumor cell score and TPS are interchangeable and are routinely used in clinical practice for estimating PD-L1 levels. They are determined by the number of positive tumor cells divided by the total number of viable tumor cells multiplied by 100%^9–11. The TAP score is a more recently developed scoring system that involves a visual estimation of the tumor area occupied by PD-L1–expressing immune and tumor cells that was validated in the RATIONALE-306 (NCT03783442) trial^12–14.

Tislelizumab (BGB-A317) is a humanized immunoglobulin G4 (IgG4) monoclonal antibody engineered to target PD-1 with high affinity and specificity. It is designed to minimize binding to the Fc gamma receptor on macrophages, enhancing anti-tumor activity by preventing antibody-dependent phagocytosis, a potential mechanism of T-cell clearance^15,16. Tislelizumab plus chemotherapy is approved for the treatment of adult patients with unresectable locally advanced or metastatic ESCC in the first line in multiple countries and is under review by several health authorities globally^17–19. In the United States, tislelizumab in combination with platinum-containing chemotherapy has been approved by the US Food and Drug Administration (FDA) for the first-line treatment of adults with unresectable or metastatic ESCC whose tumors express PD-L1 ( ≥1)¹⁸, and in Europe, the European Medicines Agency (EMA) has approved tislelizumab in combination with platinum-based chemotherapy for the first-line treatment of adult patients with unresectable, locally advanced or metastatic ESCC whose tumors express PD-L1 with a TAP score ≥5%¹⁷.

Other PD-1 inhibitors have also received regulatory approval for the treatment of unresectable, locally advanced, or metastatic ESCC with tumors expressing PD-L1 evaluated using a range of cutoffs and scoring methods. Pembrolizumab and nivolumab, in combination with platinum- and fluoropyrimidine-based chemotherapy, are approved by the FDA for the treatment of unresectable locally advanced or metastatic esophageal cancer in the first line for patients whose tumors express PD-L1 (CPS ≥1 and PD-L1 ≥1, respectively)^20,21. However, the EMA has authorized the use of first-line pembrolizumab or nivolumab in combination with platinum- and fluoropyrimidine-based chemotherapy for patients with a PD-L1 CPS of ≥10 and TC PD-L1 expression of ≥1%, respectively^22,23.

RATIONALE-306 is a global, randomized, double-blind, multicenter phase 3 trial evaluating tislelizumab in combination with investigator-chosen chemotherapy as first-line treatment for patients with unresectable locally advanced or metastatic ESCC. Evaluation of PD-L1 status was not required at the time of enrollment.

The availability of long-term follow-up data is crucial for clinicians to understand the persistence of treatment effects, particularly for immune checkpoint inhibitors, where benefits may extend beyond initial response periods. Extended follow-up data can provide insights into long-term survival outcomes, maintenance of response, and emergence of late toxicities that might not be apparent in earlier analyses.

The primary analysis of the RATIONALE-306 trial demonstrated that patients treated with tislelizumab plus chemotherapy experienced significantly improved overall survival (OS) in all patients (stratified hazard ratio (HR), 0.66; 95% confidence interval (CI) 0.54–0.80); one-sided nominal P value < 0.0001) and those with a PD-L1 TAP score ≥10% (HR, 0.62 (95% CI 0.44–0.87); one-sided nominal P value 0.0029) compared with placebo plus chemotherapy, with a tolerable safety profile¹². In the CheckMate 648 trial all randomized patients treated with nivolumab plus chemotherapy experienced significantly longer OS, and those with CPS ≥1, ≥5 and ≥10 had a clinically meaningful improvement in OS compared with chemotherapy alone¹¹. After a 29-month minimum follow-up, the CheckMate 648 trial demonstrated continued improvement in OS (median 15.0 versus 9.1 months; HR, 0.59 (95% CI 0.46–0.76)) with nivolumab plus chemotherapy versus chemotherapy in patients with a PD-L1 tumor cell score ≥1%¹¹. Similarly, in the KEYNOTE-590 study, all randomized patients and those with CPS ≥10 who were treated with pembrolizumab plus chemotherapy experienced significantly longer OS than those treated with chemotherapy alone^5,24. These examples highlight the variety of cutoffs and methods used to describe the population of patients with ESCC that will most likely benefit from immuno-oncology therapy.

This update presents extended efficacy and safety findings from RATIONALE-306 after a minimum 3-year follow-up. We also report results from post hoc exploratory analyses of OS and progression-free survival (PFS) based on PD-L1 expression and the concordance between PD-L1 TAP score and CPS. The results show that at a minimum of 3 years of follow-up, first-line tislelizumab continued to show clinically meaningful improvements in OS and PFS compared with placebo in unresectable locally advanced or metastatic ESCC.

Results

Patient demographics and baseline characteristics

A total of 649 patients were randomly assigned to receive tislelizumab plus chemotherapy (n = 326) or placebo plus chemotherapy (n = 323) in the intent-to-treat (ITT) population. At data cutoff on 24 November 2023, the minimum study follow-up (defined as the difference between the date of cutoff and the date of last patient randomized) was 36.0 months (range, 0.1–59.2). The safety population included 99.4% of patients (645/649; ITT population) in the tislelizumab plus chemotherapy (324/326) and placebo plus chemotherapy (321/323) arms who received at least one dose of study treatment. Baseline demographics and characteristics were generally balanced between treatment arms (Supplementary Table S1).

Of the 649 patients randomized, 542 had evaluable TAP scores and 537 had evaluable post hoc CPS results (Supplementary Table S2). PD-L1 expression was balanced between the treatment arms and comparable across subgroups by TAP score or CPS at different thresholds. In total, 223 patients (41.1%) had a PD-L1 TAP score of ≥10%, 358 (66.1%) had a score of ≥5% and 481 (88.7%) had a score of ≥1%. Similarly, 228 patients (42.5%) had PD-L1 CPS of ≥10, 343 (63.9%) had a score of ≥5 and 480 (89.4%) had a score of ≥1. PD-L1 expression by TAP score (10%) and TPS (1%) was comparable between treatment arms (Supplementary Table S3). A total of 191 patients (29.4%) had a PD-L1 TAP score ≥10% and TPS ≥1%, while 108 patients (16.6%) had a PD-L1 TAP score <10% and TPS ≥1% (Supplementary Table S3).

Treatment exposure

Median durations of treatment exposure with tislelizumab plus chemotherapy versus placebo plus chemotherapy were 6.4 months (interquartile range (IQR), 3.3–11.1) and 4.9 months (IQR, 2.5–8.3). Exposure to the different chemotherapy doublets among patients who received at least one dose of study treatment was similar between treatment arms¹². Of the 645 patients (safety analysis set) who received at least one dose of study treatment, 45.0% (290/645) were treated with platinum plus fluoropyrimidine and 55.0% (355/645) with platinum plus paclitaxel. Overall, 4.3% (14/326) of patients in the tislelizumab plus chemotherapy arm and 0.3% (1/323) in the placebo plus chemotherapy arm remained on study treatment as of data cutoff (Supplementary Fig. S1).

In the tislelizumab plus chemotherapy and placebo plus chemotherapy arms, 51.5% (168/326) and 57.9% (187/323) of patients received subsequent anticancer therapy, including, respectively, 15.3% (50/326) and 24.8% (80/323) of patients who received subsequent immunotherapy (Supplementary Table S4).

The most common reason for treatment discontinuation was disease progression (54.3% in the tislelizumab plus chemotherapy arm and 69.0% in the placebo plus chemotherapy arm), followed by withdrawal by patient (12.6% versus 10.8%) and adverse events (12.0% versus 6.2%). Supplementary Table S4 provides an overview of subsequent therapies received.

Overall survival

The minimum study follow-up was 36.0 months and the median follow-up was 12.8 months (range, 0.1–59.2; data cutoff: 24 November 2023). After the longer follow-up, improvement in OS was maintained with tislelizumab plus chemotherapy versus placebo plus chemotherapy in the ITT population, with median OS 17.2 months (95% CI, 15.8–20.1) versus 10.6 months (95% CI, 9.3–12.0) respectively (stratified HR, 0.70; 95% CI, 0.59–0.83; one-sided nominal P value < 0.0001) (Fig. 1a). Higher OS rates were observed in the tislelizumab plus chemotherapy arm versus the placebo plus chemotherapy arm at 12 months (65.0% versus 44.7%), 24 months (37.9% versus 24.8%) and 36 months (22.1% versus 14.1%) (Fig. 1a). Median OS in the tislelizumab plus chemotherapy versus placebo plus chemotherapy arms in patients with a PD-L1 TAP score ≥10% was 16.6 months (95% CI, 15.3–23.4) and 10.0 months (95% CI, 8.6–13.3), respectively, with an unstratified HR of 0.71 (95% CI, 0.53–0.95) and one-sided nominal P value = 0.0113 (Fig. 1b). Post hoc analysis demonstrated a median OS of 17.2 months (95% CI, 15.3–24.1) and 9.4 months (95% CI, 8.5–12.3), with an unstratified HR of 0.64 (95% CI, 0.48–0.86) and one-sided nominal P value = 0.0013 in patients with a PD-L1 CPS ≥10 in the tislelizumab plus chemotherapy versus placebo plus chemotherapy arms, respectively (Fig. 1c). Similarly, post hoc analysis revealed an OS benefit in patients with a PD-L1 TAP score ≥5% (median OS: 19.1 versus 10.0 months for tislelizumab plus chemotherapy versus placebo plus chemotherapy; unstratified HR, 0.63; 95% CI, 0.49–0.79; one-sided nominal P value < 0.0001) (Fig. 1d) and in patients with a PD-L1 CPS ≥5 (median OS: 16.8 versus 10.0 months for tislelizumab plus chemotherapy versus placebo plus chemotherapy; unstratified HR, 0.67; 95% CI, 0.52–0.85; one-sided nominal P value = 0.0004) (Fig. 1e). An improvement in OS was also observed across multiple prespecified subgroups of patients in the tislelizumab plus chemotherapy arm compared with the placebo plus chemotherapy arm (Supplementary Fig. S2).

Fig. 1 — a ITT population and in patients with PD-L1, b TAP score ≥10%, c CPS ≥10, d TAP score ≥5% and e CPS ≥5. Data in (c–e) are from post hoc analyses. The one-sided nominal P value was estimated using a log-rank test stratified by pooled geographic region (Asia versus rest of world), prior definitive therapy (yes versus no) and investigator’s choice of chemotherapy (platinum with fluoropyrimidine versus platinum with paclitaxel) per interactive response technology (IRT). The HR was based on a stratified Cox regression model including treatment as a covariate, and pooled geographic region (Asia versus rest of world), prior definitive therapy (yes versus no) and investigator’s choice of chemotherapy (platinum with fluoropyrimidine versus platinum with paclitaxel) per IRT as strata. Source data are provided as a Source Data file. CI confidence interval, CPS combined positive score, HR hazard ratio, ITT intent-to-treat, OS overall survival, PD-L1 programmed death-ligand 1, TAP Tumor Area Positivity.

Exploratory post hoc analysis after a minimum 3-year follow-up (data cutoff: 24 November 2023) showed an improvement in OS favoring patients who were treated with tislelizumab plus chemotherapy compared with placebo plus chemotherapy in PD-L1 subgroups defined by TAP score and CPS at matched thresholds of 10%/10, 5%/5 and 1%/1 (Supplementary Fig. S3). A trend in OS favoring tislelizumab plus chemotherapy compared to placebo plus chemotherapy was observed across all PD-L1 subgroups analyzed, regardless of the definition used. In patients with a PD-L1 TAP score ≥1%, median OS was 16.8 months (95% CI, 15.3–20.8) with tislelizumab plus chemotherapy versus 9.6 months (95% CI, 8.9–11.8) with placebo plus chemotherapy (HR, 0.68; 95% CI, 0.55–0.83; one-sided nominal P value < 0.0001). Similarly, in patients with a PD-L1 CPS ≥1, median OS was 16.8 months (95% CI, 15.3–20.8) in the tislelizumab plus chemotherapy arm versus 9.6 months (95% CI, 8.9–11.8) in the placebo plus chemotherapy arm (HR, 0.67; 95% CI, 0.55–0.82; one-sided nominal P value < 0.0001). Kaplan–Meier curves of OS results defined by PD-L1 TAP scores <10% (Supplementary Fig. S4a) and CPS < 10 (Supplementary Fig. S4b), as well as TAP scores <5% (Supplementary Fig. S4c) and CPS <5 (Supplementary Fig. S4d) were similar. Consistent OS benefit was observed in patients receiving tislelizumab plus chemotherapy versus placebo plus chemotherapy across a spectrum of PD-L1 expression levels, as shown by Kaplan–Meier curves of OS by TAP scores ≥1% to <10% (Supplementary Fig. S4e) and CPS ≥1 to <10 (Supplementary Fig. S4f), as well as TAP scores ≥5% to <10% (Supplementary Fig. S4g) and CPS ≥5 to <10 (Supplementary Fig. S4h). Similar trends of OS benefit were observed with tislelizumab plus chemotherapy versus placebo plus chemotherapy in patients with TAP scores ≥1% to <5% (Supplementary Fig. S4i) and CPS ≥1 to <5 (Supplementary Fig. S4j).

Progression-free survival

At the 24 November 2023 data cutoff, investigator-assessed PFS was substantially improved in patients who were treated with tislelizumab plus chemotherapy compared with placebo plus chemotherapy. Median investigator-assessed PFS with tislelizumab plus chemotherapy versus placebo plus chemotherapy was 7.3 months (95% CI, 6.9–8.3) versus 5.6 months (95% CI, 4.9–6.0), respectively (HR, 0.60; 95% CI, 0.50–0.72; one-sided nominal P value < 0.0001). The PFS rate was higher among patients treated with tislelizumab plus chemotherapy than with placebo plus chemotherapy at 12 months (30.4% versus 15.7%), 24 months (18.1% versus 7.2%) and 36 months (15.0% versus 2.9%; Fig. 2a).

Fig. 2 — a ITT population (stratified analysis) and in patients with PD-L1, b TAP score ≥10%, and c TAP score ≥5% (unstratified analyses). Data in (b, c) are from post hoc analyses. For the ITT population (a), the one-sided nominal P value was estimated using a log-rank test stratified by pooled geographic region (Asia versus rest of world), prior definitive therapy (yes versus no) and investigator’s choice of chemotherapy (platinum with fluoropyrimidine versus platinum with paclitaxel) per IRT and the HR was based on Cox regression model including treatment arm as a covariate and stratified by the same factors. For PD-L1 subgroups (b, c), the HR was based on an unstratified Cox regression model including treatment as a covariate, and pooled geographic region (Asia vs rest of world), prior definitive therapy (yes versus no) and investigator’s choice of chemotherapy (platinum with fluoropyrimidine versus platinum with paclitaxel) per IRT as strata. For PFS in ITT, stratified HR was based on Cox regression model including treatment arm as a covariate and stratified by pooled geographic region (Asia versus rest of world) per IRT, prior definitive therapy (yes versus no) per IRT and investigator-choice chemotherapy (platinum with fluoropyrimidine versus platinum with paclitaxel) per IRT. For PFS by PD-L1 expression, TAP score ≥10% and TAP score ≥5%, HR was based on unstratified Cox regression model only including treatment as a covariate. Source data are provided as a Source Data file. CI confidence interval, CPS combined positive score, HR hazard ratio, ITT intent-to-treat, PD-L1 programmed death-ligand 1, PFS progression-free survival, TAP Tumor Area Positivity.

Similar to the OS results, post hoc analysis showed an improvement in PFS among patients who were treated with tislelizumab plus chemotherapy compared with placebo plus chemotherapy across all PD-L1 subgroups defined by TAP score cutoffs: 10%, 5% and 1% (Supplementary Fig. S5). Kaplan–Meier curves revealed improvement in PFS in patients with PD-L1 TAP scores ≥10% (Fig. 2b), ≥5% (Fig. 2c) and ≥1% (Supplementary Fig. S6a). Consistent improvement in PFS was also observed across a spectrum of PD-L1 expression levels using both TAP score and CPS, as shown by Kaplan–Meier curves for TAP scores ≥1 to <10% (Supplementary Fig. S6b) and CPS ≥1 to <10 (Supplementary Fig. S6c), as well as TAP scores ≥5 to <10% (Supplementary Fig. S6d) and CPS ≥5 to <10 (Supplementary Fig. S6e). Similar trends of PFS improvement were observed with tislelizumab plus chemotherapy versus placebo plus chemotherapy in patients with TAP scores ≥1 to <5% (Supplementary Fig. S6f) and CPS ≥1 to <5 (Supplementary Fig. S6g).

Concordance between TAP score and CPS in the PD-L1 subgroups

PD-L1 expression testing was not a requirement for RATIONALE-306 trial enrollment. Hence, sample collection and testing for PD-L1 status were done retrospectively, and 16.5% of the study population had unknown PD-L1 expression levels. In an exploratory post hoc analysis, Sankey diagrams were utilized to visualize the relationship between PD-L1 expression levels as defined by TAP score and CPS. These diagrams revealed a good visual alignment between the two scoring methods. The strongest agreement was observed at TAP scores ≥10% and <1%. For TAP scores ≥10%, 88.3% (196/222) of samples also had a CPS ≥10, with 11.7% (26/222) down-migrating to lower CPS categories. Similarly, for TAP scores <1%, 83.6% (51/61) of samples had a CPS <1, with 16.4% (10/61) up-migrating to higher CPS levels. However, agreement was less consistent in intermediate ranges. In the TAP score 1–4% category, 67.5% (81/120) of samples had a matching CPS of 1–4, with 27.5% (33/120) up-migrating to higher CPS levels and 5.0% (6/120) down-migrating to CPS < 1. The least alignment was observed in the TAP score 5–9% category, where 49.3% (66/134) of samples had a matching CPS of 5–9, with 19.4% (26/134) up-migrating to CPS ≥10 and 31.3% (42/134) down-migrating to CPS 1–4 (Fig. 3).

Fig. 3 — Diagram created using SankeyMATIC (post hoc analysis). Source data are provided as a Source Data file. CPS combined positive score, TAP Tumor Area Positivity.

Antitumor activity

Investigator-assessed objective responses were achieved by 63.5% (207/326) of patients treated with tislelizumab plus chemotherapy versus 42.4% (137/323) treated with placebo plus chemotherapy (a complete response was observed in 4.6% (15/326) versus 2.5% (8/323) of patients receiving tislelizumab plus chemotherapy versus placebo plus chemotherapy, respectively), with an odds ratio of 2.38 (95% CI, 1.73–3.27) (Supplementary Table S5). Investigator-assessed median duration of response (DoR) in the tislelizumab plus chemotherapy arm was 7.1 months (95% CI, 6.1–8.1) versus 5.7 months (95% CI, 4.4–7.1) in the placebo plus chemotherapy arm (Supplementary Fig. S7). At the time of the analysis, 13.0% (27/207) of patients in the tislelizumab plus chemotherapy arm and 2.9% (4/137) of patients in the placebo plus chemotherapy arm had an ongoing response (Supplementary Table S5). Disease control rate per investigator assessment was 89.0% (290/326) in the tislelizumab plus chemotherapy arm and 80.2% (259/323) in the placebo plus chemotherapy arm (Supplementary Table S5).

Patient-reported outcomes

Patient-reported outcomes, assessed using the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire - Core 30 (EORTC QLQ-C30) questionnaire, showed that for the physical functioning domain, median time to deterioration (TTD) was not reached in the tislelizumab plus chemotherapy arm and was 18.8 months in the placebo plus chemotherapy arm (HR, 0.78; 95% CI, 0.59–1.02). Excluding the physical functioning domain of the EORTC QLQ-C30, TTD analyses showed that the risk of clinically meaningful worsening across all patient-reported outcomes based on the EORTC QLQ-C30 and EORTC QLQ-Oesophageal Cancer Module (EORTC QLQ-OES18) questionnaires was similar between treatment arms (Tables 1 and 2).

Table 1.

Analyses of time to deterioration of EORTC QLQ-C30 (ITT Analysis Set)

		Tislelizumab plus chemotherapy (n = 326)	Placebo plus chemotherapy (n = 323)
EORTC QLQ-C30
GSH/QoL	Patients
	Worsened	109 (33.4)	98 (30.3)
	Censored	217 (66.6)	225 (69.7)
	Median TTD, months (95% CI)^a	NR (14.6–NE)	NR (9.5–NE)
	P value^b	0.4336
	Stratified HR (95% CI)^c	0.98 (0.74–1.29)
	12-Month TTD rate, % (95% CI)^d	57.5 (50.3–64.1)	54.4 (45.7–62.3)
Physical functioning	Patients
	Worsened	106 (32.5)	104 (32.2)
	Censored	220 (67.5)	219 (67.8)
	Median TTD, months (95% CI)^a	NR (11.9–NE)	18.8 (8.1–NE)
	P value^b	0.0360
	Stratified HR (95% CI)^c	0.78 (0.59–1.02)
	12-Month TTD rate, % (95% CI)^d	57.2 (49.6–64.1)	57.3 (49.7–64.2)

Open in a new tab

Data cutoff: 24 November 2023.

CI confidence interval, EORTC QLQ-C30 European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire - Core 30, GSH global health status, HR hazard ratio, NE not estimable, NR not reached, IRT interactive response technology, TTD time to deterioration.

Time to deterioration (clinically meaningful worsening) was defined as the time from randomization to the first time the difference between the current derived score and baseline was ≥10 in the worsening direction. A deterioration was not counted as an event if a subsequent improvement returned the overall worsening from baseline to <10 points. Patients without clinically meaningful worsening were censored at the last time the C30 was assessed. Source data are provided as a Source Data file.

^aEstimated based on Kaplan–Meier method.

^bOne-sided P value was estimated from log rank test stratified by pooled geographic region (Asia versus rest of world) per IRT, prior definitive therapy (yes versus no) per IRT and chemotherapy option (investigator choice of chemotherapy (platinum with fluoropyrimidine versus platinum with paclitaxel)) per IRT, for descriptive purposes only.

^cHR is based on Cox regression model including treatment as covariate and stratified by pooled geographic region (Asia versus rest of world) per IRT, prior definitive therapy (yes versus no) per IRT and chemotherapy option (investigator choice of chemotherapy (platinum with fluoropyrimidine versus platinum with paclitaxel)) per IRT.

^dTTD rate is the proportion of patients whose current derived score had not worsened by ≥10 points from baseline.

Table 2.

Analyses of time to deterioration of EORTC QLQ-OES18 (ITT Analysis Set)

		Tislelizumab plus chemotherapy (n = 326)	Placebo plus chemotherapy (n = 323)
EORTC QLQ-OES18
Dysphagia	Patients
	Worsened	112 (34.4)	106 (32.8)
	Censored	214 (65.6)	217 (67.2)
	Median TTD, months (95% CI)^a	NR (13.6–NE)	NR (8.9–NE)
	P value^b	0.2647
	Stratified HR (95% CI)^c	0.92 (0.70–1.20)
	12-Month TTD rate, % (95% CI)^d	58.3 (51.3–64.5)	55.0 (46.8–62.4)
Eating	Patients
	Worsened	79 (24.2)	68 (21.1)
	Censored	247 (75.8)	255 (78.9)
	Median TTD, months (95% CI)^a	NR (NE–NE)	46.0 (19.6–NE)
	P value^b	0.5179
	Stratified HR (95% CI)^c	1.01 (0.73–1.40)
	12-Month TTD rate, % (95% CI)^d	70.5 (63.7–76.4)	73.4 (66.8–78.8)
Reflux	Patients
	Worsened	84 (25.8)	64 (19.8)
	Censored	242 (74.2)	259 (80.2)
	Median TTD, months (95% CI)^a	NR (NE–NE)	NR (17.3–NE)
	P value^b	0.8039
	Stratified HR (95% CI)^c	1.15 (0.83–1.60)
	12-Month TTD rate, % (95% CI)^d	68.7 (62.0–74.5)	72.7 (64.1–79.6)
Pain	Patients
	Worsened	64 (19.6)	64 (19.8)
	Censored	262 (80.4)	259 (80.2)
	Median TTD, months (95% CI)^a	NR (NE–NE)	25.7 (24.4–NE)
	P value^b	0.1003
	Stratified HR (95% CI)^c	0.79 (0.56–1.13)
	12-Month TTD rate, % (95% CI)^d	74.3 (67.3–80.1)	74.6 (67.4–80.5)

Open in a new tab

Data cutoff: 24 November 2023.

CI confidence interval, EORTC QLQ-OES18 European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire - Oesophageal Cancer Module, GSH global health status, HR hazard ratio, NE not estimable, NR not reached, IRT interactive response technology, TTD time to deterioration.

Time to deterioration (clinically meaningful worsening) was defined as the time from randomization to the first time the difference between the current derived score and baseline was ≥10 in the worsening direction. A deterioration was not counted as an event if a subsequent improvement returned the overall worsening from baseline to <10 points. Patients without clinically meaningful worsening were censored at the last time the OES18 was assessed. TTD rates (event-free rates) were estimated by Kaplan–Meier method with 95% CIs estimated using the Greenwood’s formula. Data shown in TTD analyses post hoc analyses. Source data are provided as a Source Data file.

^aEstimated based on Kaplan–Meier method.

^dTTD rate is the proportion of patients whose current derived score had not worsened by ≥10 points from baseline.

Safety and tolerability

At least one treatment-emergent adverse event (TEAE) was reported in 99.7% (323/324) of patients in the tislelizumab plus chemotherapy arm and in 99.4% (319/321) of patients in the placebo plus chemotherapy arm. Grade ≥3 TEAEs related to tislelizumab/placebo treatment were observed in 32.1% (104/324) versus 20.2% (65/321) of patients in the tislelizumab plus chemotherapy arm versus the placebo plus chemotherapy arm. A higher proportion of patients treated with tislelizumab plus chemotherapy than placebo plus chemotherapy had serious TEAEs related to any study treatment component: 29.9% (97/324) versus 19.6% (63/321) of patients. Discontinuation of study treatment (tislelizumab/placebo) due to TEAEs was reported in 13.3% (43/324) versus 6.5% (21/321) of patients in the tislelizumab plus chemotherapy arm versus the placebo plus chemotherapy arm. Dose modification of tislelizumab/placebo due to TEAEs occurred in 53.7% (174/324) versus 39.9% (128/321) of patients in the tislelizumab plus chemotherapy versus placebo plus chemotherapy arms (Supplementary Table S6).

Select TEAEs with an incidence difference greater than 5% between the two arms at the interim analysis (median follow-up, 16.3 months)¹² were decreased appetite, vomiting, pruritus and hypothyroidism. Exploratory post hoc analysis of these selected TEAEs over time revealed that the majority of events occurred within the first 12 months of treatment, with the tislelizumab plus chemotherapy group generally showing higher incidences of decreased appetite (42.9% versus 38.6%), pruritus (12.0% versus 5.9%) and hypothyroidism (9.3% versus 4.4%), while the placebo plus chemotherapy group had a higher incidence of vomiting (21.9% versus 27.4%), respectively. The frequency of these TEAEs typically decreased over time in both groups, though interpretation of long-term trends is limited by small patient numbers in later time periods (Fig. 4).

Fig. 4 — Data cutoff: 24 November 2023. Any TEAEs with an incidence difference greater than 5% between the two arms at the interim analysis in RATIONALE-306 safety analysis set are included. Data shown are from a post hoc analysis. n represents the number of patients who were receiving treatment during the specified time interval and therefore at risk for treatment-emergent adverse events. This number differs from the number at risk in survival analyses because patients may have discontinued treatment but remained in survival follow-up. Percentages were calculated based on n. Source data are provided as a Source Data file. TEAE, treatment-emergent adverse event.

Incidents of any-grade treatment-related adverse events (TRAEs) were similar among patients receiving tislelizumab plus chemotherapy and those receiving placebo plus chemotherapy (Supplementary Table S7). Grade ≥3 TRAEs were reported in 67.0% (217/324) of patients in the tislelizumab plus chemotherapy arm and 64.5% (207/321) in the placebo plus chemotherapy arm (Supplementary Table S7). The most common TRAEs occurring in ≥5% of patients who received at least one dose of study treatment in either treatment arm is shown in Supplementary Table S7. The most common grade ≥3 TRAEs were decreased neutrophil count (30.9% (100/324) in the tislelizumab plus chemotherapy arm versus 32.7% (105/321) in the placebo plus chemotherapy arm), anemia (14.8% (48/324) versus 12.8% (41/321)) and decreased white blood cell count (10.8% (35/324) versus 15.6% (50/321)) (Supplementary Table S7). TRAEs leading to death occurred in six (1.9%) patients in the tislelizumab plus chemotherapy arm and four (1.2%) patients in the placebo plus chemotherapy arm.

No new types of grade ≥3 TRAEs were identified in this extended follow-up compared to those reported in the interim analysis¹². Analysis of grade ≥3 TRAEs over time showed that the vast majority occurred within the first 12 months of treatment, with a substantial decrease in incidence after this period. For example, decreased neutrophil count, the most common grade ≥3 TRAE, occurred in 30.9% of patients in the tislelizumab plus chemotherapy arm during the first 12 months, but only in 5.0% during months 12–24 and in 0% thereafter. Similar patterns were observed for other common grade ≥3 TRAEs, including anemia (≤12 months: 14.8%; >12–24 months: 1.3%; >24 months: 0%) and decreased white blood cell count (≤ 12 months: 10.5%; >12–24 months: 1.3%; >24 months: 0%).

Immune-mediated adverse events (imAEs) were reported in 21.6% (70/324) of patients in the tislelizumab plus chemotherapy arm compared with 5.9% (19/321) in the placebo plus chemotherapy arm. The most common imAEs were hypothyroidism (8.0%), skin adverse reactions (4.0%), and pneumonitis (2.8%). Grade ≥3 imAEs occurred in 8.6% (28/324) versus 1.6% (5/321) of patients, respectively. imAEs leading to treatment discontinuation occurred in 3.7% (12/324) versus 0.6% (2/321) of patients. Systemic corticosteroids were used to manage imAEs in 13.3% (43/324) of patients in the tislelizumab plus chemotherapy arm, with high-dose corticosteroids required in 9.6% (31/324).

Discussion

Our long-term follow-up analysis confirms that the improvements in median OS and PFS were sustained and consistent with previous reports from the interim analysis¹² and minimum 2-year follow-up data²⁵. In addition, the 36-month OS and PFS rates were higher in the tislelizumab plus chemotherapy arm than in the placebo plus chemotherapy arm in the ITT population.

The OS benefit observed with tislelizumab plus chemotherapy in RATIONALE-306 (median OS, 17.2 months; HR, 0.70) represents one of the most favorable outcomes reported to date in global phase 3 trials of first-line treatment with anti-PD-1 antibody plus chemotherapy for advanced or metastatic esophageal cancer (or ESCC), with a minimum follow-up of 36.0 months. The extended 3-year data provide valuable new insights into the sustained clinical benefit of tislelizumab plus chemotherapy, with higher 36-month OS and PFS rates in the tislelizumab plus chemotherapy arm than in the placebo plus chemotherapy arm in the ITT population.

In CheckMate 648, patients treated with nivolumab plus chemotherapy experienced improved OS, with a stratified HR of 0.78 in the overall population and an unstratified HR of 0.72 in the CPS ≥10 population, after a minimum follow-up of 29 months¹¹. Likewise in KEYNOTE-590, after a median follow-up of 34.8 months, a favorable OS benefit was observed in all patients with ESCC (HR, 0.73), and those with ESCC and CPS ≥10 (HR, 0.59) treated with pembrolizumab plus chemotherapy versus chemotherapy alone^5,24. However, cross-trial comparisons should be interpreted with caution. Differences in study design, patient populations, PD-L1 assay methods, chemotherapy backbones and post-trial therapies limit the validity of direct comparisons between studies. The differences in median OS observed between trials may be attributable to these variations rather than reflecting true differences in treatment efficacy.

These extended follow-up data, with a minimum 3-year follow-up from the RATIONALE-306 trial, provide further support for the therapeutic advantages of tislelizumab plus chemotherapy over placebo plus chemotherapy as first-line treatment for patients with unresectable locally advanced or metastatic ESCC. Among the global trials, RATIONALE-306 is the only study that allowed for different chemotherapy backbones. In both CheckMate 648 and KEYNOTE-590, the chemotherapy backbone was platinum/fluoropyrimidine. In RATIONALE-306, two main chemotherapy backbones were allowed and served as stratification factors (platinum/taxane and platinum/fluoropyrimidine). Comparable improved efficacy of tislelizumab was observed regardless of chemotherapy backbone. Platinum/taxane could be preferred options for patients in certain clinical settings (eg, dihydropyrimidine dehydrogenase deficiency, preference to avoid continuous-infusion 5-FU).

Our exploratory post hoc analysis revealed that both TAP score and CPS are viable options for the measurement of PD-L1 in patients with unresectable locally advanced or metastatic ESCC as visualized by the Sankey analysis. The KEYNOTE-590 study also used the CPS method to assess PD-L1 levels. The CheckMate 648 trial was prespecified to use TC scoring to assess PD-L1 expression status, whereas the CPS methodology was evaluated retrospectively.

In RATIONALE-306, comparable OS and PFS benefits with tislelizumab plus chemotherapy were observed in PD-L1 subgroups by CPS at matched thresholds, with good correlation between TAP score and CPS, as indicated by an interclass correlation coefficient value of 0.85. Additionally, TAP score and CPS at matched thresholds (10% versus 10, 5% versus 5 and 1% versus 1) exhibited substantial concordance^13,14. In our post hoc analysis using Sankey diagrams, we observed high concordance between TAP score and CPS at ≥10% and <1%, contrasting with a higher magnitude of discordance in intermediate categories (5–9% and 1–4%). It is important to note that the scoring of TAP and CPS was done by two separate pathologists, suggesting the discordance may be due to inter-pathologist variability. This discordance could lead to patient misclassification, potentially affecting treatment access based on country-specific testing requirements and eligibility cutoffs.

This discordance represents a major limitation and significant challenge for biomarker-driven therapy in ESCC. Given that regulatory approvals in different regions rely on specific PD-L1 expression cutoffs (eg, TAP score ≥5% in Europe as mentioned above), this variability could directly impact which patients receive treatment. Patients with tumors in the intermediate expression range might be eligible for treatment in one region but ineligible in another, solely due to differences in scoring methods rather than true biological differences. This highlights the need for standardized assessment methods and, potentially, the development of more robust biomarkers to guide treatment decisions in ESCC. Clinicians should be aware of these limitations when interpreting PD-L1 test results, particularly for patients with expression levels in these intermediate ranges.

Treatment with tislelizumab plus chemotherapy did not show any new safety signals with longer follow-up. The proportion of patients experiencing any-grade TRAEs was similar between the treatment arms, and the most common TRAEs (any grade and grade ≥3) were consistent with known adverse events of the individual study treatment components.

While grade ≥3 TRAEs were more common in the tislelizumab plus chemotherapy arm compared with the placebo plus chemotherapy arm (67.0% versus 64.5%), no new types of grade ≥3 TRAEs were identified compared with the interim analysis¹², suggesting a stable safety profile with extended follow-up. Importantly, temporal analysis of grade ≥3 TRAEs revealed that most events occurred within the first 12 months of treatment, with minimal occurrences beyond this timeframe. This pattern suggests that patients who tolerate the initial treatment period are unlikely to experience new severe adverse events with continued therapy, which is valuable information for clinical decision-making regarding long-term treatment with tislelizumab plus chemotherapy.

The observed higher rates of hypothyroidism in the placebo plus chemotherapy arm are likely attributable to the small number of patients remaining on treatment in later time periods, which can magnify the percentage impact of each event. Additionally, patients with ESCC often have a history of prior radiation therapy, which can lead to delayed onset of hypothyroidism that may manifest during long-term follow-up.

TTD analyses based on EORTC QLQ-C30 and QLQ-OES18 questionnaires showed that tislelizumab plus chemotherapy maintained overall quality of life compared with placebo, with a potential benefit in preserving physical functioning (median TTD not reached versus 18.8 months; HR, 0.78). These findings suggest that the addition of tislelizumab can improve survival outcomes without compromising patient-reported quality of life and may even help preserve physical function during treatment.

Limitations to be considered while interpreting the results include the scoring of CPS and the study design stratification. CPS was scored retrospectively on slides stained using an investigational use–only version of the VENTANA PD-L1 (SP263) CDx Assay (Roche Diagnostics, Indianapolis, IN) that were previously assessed by TAP score. This retrospective nature of CPS scoring introduces potential bias that may impact the interpretation of concordance between TAP and CPS. Additionally, while we observed good correlation between these scoring methods as reported by Moehler et al.¹⁴, we did not perform formal interaction tests between PD-L1 subgroups in the current analysis, which limits the statistical rigor of these comparisons.

This report also includes multiple post hoc exploratory analyses examining various PD-L1 cutoffs. These analyses were not adjusted for multiplicity, which increases the risk of false-positive findings. Cautious evaluation is necessary to avoid overinterpretation of these exploratory results that were not prespecified in the original analysis plan. These limitations should be considered when interpreting the clinical implications for regulatory approvals that rely on specific PD-L1 scoring methods.

It is important to note that the trial was not powered to evaluate efficacy in the small subgroup of patients with low PD-L1 expression (CPS <1 or TAP score <1%). The limited sample size in these subgroups results in wide confidence intervals and reduced statistical precision, which should be considered when interpreting the results for these specific patient populations.

In conclusion, after a minimum of 3 years of follow-up, patients with unresectable locally advanced or metastatic ESCC who received first-line treatment with tislelizumab plus chemotherapy continued to experience clinically meaningful improvements in OS, as well as improved PFS and long-lasting antitumor response, compared with those who received placebo plus chemotherapy. These findings from the extended follow-up of the RATIONALE-306 study align with the 2-year follow-up and interim analysis, reinforcing the sustained efficacy and tolerable safety profile of tislelizumab plus chemotherapy and providing further support for the improved outcomes with this combination as first-line treatment of unresectable locally advanced or metastatic ESCC.

Methods

Ethics statement

Relevant institutional review board or independent ethics committee approvals were obtained for each study site (see Supplementary Data 1 for full details). The study adhered to International Council for Harmonisation Good Clinical Practice Guidelines, the Declaration of Helsinki and local regulations. All patients provided written informed consent, and safety was monitored by an independent data monitoring committee. Participants were not compensated for this study. This study was compliant with the Guidance of China’s Ministry of Science and Technology for the Review and Approval of Human Genetic Resources.

Study design

RATIONALE-306 (NCT03783442) is a global, randomized, double-blind, parallel-arm, placebo-controlled phase 3 trial. Detailed study protocol (see Supplementary Note 2), eligibility criteria and endpoint assessments have been reported previously^12,26. In brief, adults (aged ≥18 years) with histologically confirmed unresectable locally advanced or metastatic ESCC with no prior systemic therapy for advanced disease, and who were not amenable to definitive therapies, including surgery or radiation, per local investigator, were recruited. Patients must have had a treatment-free interval of at least 6 months if prior neoadjuvant or adjuvant therapy with platinum-based chemotherapy was administered. Patients were required to have an Eastern Cooperative Oncology Group performance status of 0 or 1 and measurable or evaluable disease per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1.

Patients

Patients were randomized 1:1 to receive either tislelizumab plus chemotherapy or placebo plus chemotherapy. Randomization was stratified by investigator-chosen chemotherapy (platinum plus fluoropyrimidine versus platinum plus paclitaxel), region (Asia (excluding Japan) versus Japan versus other regions) and receipt of previous definitive therapy (yes versus no). Sex was not considered in the study design and was self-reported by participants.

Procedures

Patients received 200 mg of tislelizumab or matching placebo intravenously every 3 weeks on day 1 of each 21-day cycle together with an investigator-chosen chemotherapy regimen. Chemotherapy options were a platinum agent (cisplatin 60–80 mg/m² or oxaliplatin 130 mg/m² intravenously on day 1) combined with fluoropyrimidine (fluorouracil 750–800 mg/m² intravenously on days 1–5), capecitabine (1000 mg/m² orally twice daily on days 1–14) or paclitaxel (175 mg/m² intravenously on day 1). Cisplatin was used in China, Taiwan and Japan (oxaliplatin substitution was not permitted per local treatment guidelines). Crossover between treatment arms or between fluoropyrimidine and paclitaxel during the study treatment period was prohibited, even after unmasking. Treatment was continued until investigator-assessed disease progression per RECIST version 1.1, unacceptable toxicity, death or withdrawal of consent. Concomitant medications prohibited during the study included any concurrent antineoplastic therapy such as chemotherapy, hormonal therapy, immunotherapy (including interleukin-2, interferon, thymosin), or standard or investigational agents (including Chinese herbal and patent medicines) for cancer treatment. Restricted concomitant medications and procedures included immunosuppressive agents (except to treat drug-related adverse events) and systemic corticosteroids >10 mg daily (prednisone or equivalent), except for treating drug-related adverse events or short-term prophylactic use. Comprehensive details of prohibited and restricted concomitant medications or procedures are described in the study protocol (see Supplementary Note 2)¹².

When patients discontinued study treatment, they entered survival follow-up, during which detailed timelines of subsequent therapies were not systematically recorded. Reasons for discontinuation included disease progression, adverse events, withdrawal of consent, or other reasons per protocol.

Endpoints and assessments

Complete lists of study endpoints and assessments have been published¹². In summary, the primary endpoint was OS in the ITT population. Secondary endpoints included investigator-assessed PFS, ORR, DoR, OS in the PD-L1 TAP score ≥10% subgroup, safety and health-related quality of life. Health-related quality of life data will be reported in a separate publication. Minimum study follow-up time was defined as a difference between the date of cutoff and the date of last patient randomized.

Exploratory post hoc assessments included a subgroup analysis of OS and PFS using PD-L1 expression defined by cutoffs with TAP scores (≥10%, <10%, ≥5%, <5%, ≥1% and <1%) and CPS (≥10, <10, ≥5, <5, ≥1 and <1)¹³. Concordance between TAP score and CPS at multiple cutoffs and the interclass correlation coefficient were also investigated¹³. All available tumor samples were centrally assessed retrospectively for PD-L1 TAP score by pathologists using an investigational use–only version of the VENTANA PD-L1 (SP263) Assay. Central laboratory pathologists were trained and certified to evaluate SP263-stained samples using the TAP score method, in line with the manufacturer’s training programs, prior to evaluation of clinical patient samples. TAP score is a visual-based estimation method, defined as the total percentage of tumor area (tumor and any desmoplastic stroma) covered by tumor cells with PD-L1 membrane staining at any intensity and tumor-associated immune cells with PD-L1 staining at any intensity²⁷. For exploratory purposes, pathologists in the central laboratory, who were trained and certified to use the CPS method according to the manufacturer’s training programs, scored the same previously SP263-stained samples according to CPS. CPS is a cell counting method, calculated by dividing the number of PD-L1–expressing tumor cells, lymphocytes and macrophages by the total number of viable tumor cells, multiplied by 100, and the tumor cell score is the percentage of viable tumor cells with any membrane staining above background²⁸. Of note, PD-L1 expression quantified using the CPS method is not an approved use of the VENTANA PD-L1 (SP263) CDx Assay.

The PD-L1 analyses based on TAP score or CPS below certain thresholds (eg, TAP score <1% or CPS <1) had small sample sizes, which limited the interpretation of the efficacy data.

Key patient-reported outcome (PRO) endpoints were protocol prespecified from the EORTC QLQ-C30 (global health status/QoL, physical functioning, and fatigue) and the QLQ-OES18 (symptoms of dysphagia, difficulty eating, reflux, and pain, and the symptom index score), based on their relevance to disease- and treatment-related effects and the most prevalent ESCC-related symptoms. TTD was defined as the time from randomization to the first clinically meaningful worsening event (defined as a ≥ 10-point change from baseline). Patients who did not experience a clinically meaningful worsening in that domain were censored at the date of their last available assessment for that domain. This approach handled missing data through right-censoring in the time-to-event framework rather than by imputation.

Statistics and reproducibility

Statistical analyses for the RATIONALE-306 study have been published in previous reports¹². In brief, efficacy analyses were conducted in the ITT analysis set, which included all patients randomized to treatment. The safety population included all patients who received at least one dose of study treatment. OS and PFS were tested using a one-sided nominal P value, estimated with a log-rank test stratified by factors used in randomization. HRs and two-sided 95% CIs were estimated with a Cox regression model including treatment as a covariate, with the same stratification factors used for OS and PFS as prespecified in the protocol and statistical analysis plan. The Kaplan–Meier method was used to estimate median OS and PFS (in the ITT population and each of the PD-L1 TAP score and CPS subgroups). For median OS and PFS, two-sided 95% CIs were constructed using a generalized Brookmeyer and Crowley method. Investigator-assessed ORR was tested using the Cochran–Mantel–Haenszel test, adjusting for prespecified stratification factors; the two-sided 95% CI for odds ratio was calculated along with Clopper–Pearson 95% CIs for ORR in each treatment arm. Efficacy and safety data were analyzed using descriptive statistics and all P values reported are nominal. All calculations and analyses were done using SAS (SAS Institute, Cary, NC), version 9.4 or higher. RATIONALE-306 is registered with ClinicalTrials.gov (NCT03783442). Sankey diagrams were built using the SankeyMATIC online tool (https://sankeymatic.com/build/; accessed 23 May 2025).

Specific timing relative to the primary analysis is not available. No additional analyses were performed to compare differences in survival between arms based on subsequent therapies, as this was not prespecified in the study protocol and the lack of detailed timing data would limit the interpretation of such analyses.

For more information on the study design and protocol, please refer to the previous publication and Supplementary Note 2¹².

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information^{(66.8MB, pdf)}

41467_2026_77327_MOESM2_ESM.pdf^{(14.9KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1^{(50.6KB, docx)}

Reporting Summary^{(2.3MB, pdf)}

Transparent Peer Review file^{(469.1KB, pdf)}

Source data

Source Data^{(227.4KB, xlsx)}

Acknowledgements

The authors thank the patients and their families, investigators, co-investigators, and the study teams at each of the participating centers. BeOne Medicines was involved in the study design, data collection, analysis and interpretation. Medical writing support was provided by Nitya Venkataraman, PhD, of Parexel, and supported by BeOne Medicines. Patients or the public were not involved in the study design, conduct or reporting of the trial. Meeting presentation: Data presented in part at the 2024 American Society of Clinical Oncology Annual Meeting, Chicago, IL, USA, 31 May–4 June, 2024²⁶, and European Society for Medical Oncology Gastrointestinal Cancers Annual Congress, Munich, Germany, 26–29 June 2024¹³.

Author contributions

H.H.Y., K.K., E.R., R.H., Y.S., K.W. and J.X. were steering committee members; H.H.Y., E.R., R.H., K.W. and J.X. contributed to the conception and design of the study; H.H.Y., K.K., E.R., R.H., Y.Pan, S.R.P., T.K., C-Y.L., L.W., D.T., R.I. and J.X. acquired the data; K.W. analyzed the data; H.H.Y., K.K., E.R., R.H., Y.S., Y.Pan, S.R.P., T.K., C-Y.L., L.W., D.T., R.I., K.W., Y.Peng and J.X. interpreted the data.

Peer review

Peer review information

Nature Communications thanks Manabu Muto and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This study was funded by BeOne Medicines, Ltd.

Data availability

On request, and subject to certain criteria, conditions, and exceptions, BeOne Medicines, Ltd. will provide access to individual de-identified participant data from applicable BeOne Medicines-sponsored studies. Qualified scientific and medical researchers with appropriate competencies who are engaged in novel scientific research may submit a request for participant-level data with a research proposal for BeOne review. Research teams must include a biostatistician and sign a Data Sharing Agreement prior to receiving access to clinical trial data. BeOne Medicines shares data only when permitted by applicable data privacy and security laws and regulations, shares when it is feasible to do so without compromising the privacy of the study participants, and other considerations. Data requests may be submitted to ClinicalTrials@beonemed.com. Estimated time to response is 2–4 weeks. The study protocol is available in Supplementary Note 2. Source data are provided with this paper.

Competing interests

H.H.Y. was an invited speaker and reports writing engagements with BeOne Medicines; has provided expert testimony to MJH Life Sciences; has served on advisory boards to Amgen, Astellas, AstraZeneca, BeOne Medicines, Bristol Myers Squibb, Elevation Oncology, Macrogenics, Merck, Novartis, OncXerna and Zymeworks; and has received research grants from Bristol Myers Squibb, CARsgen Therapeutics, Merck and BeOne Medicines and other travel support from BeOne Medicines. K.K. has received consulting fees from AstraZeneca, Bayer, BeOne Medicines, Bristol Myers Squibb, Janssen, Merck Bio, Merck & Co., Novartis, Ono Pharmaceutical and Roche; payment for expert testimony from Bristol Myers Squibb and Ono Pharmaceutical; and has participated on data safety monitoring boards or advisory boards for Bristol Myers Squibb, Chugai, Merck & Co. and Ono Pharmaceutical. E.R. reports no conflicts of interest. R.H. has received honoraria from Eisai and has participated on data safety monitoring boards or advisory boards for BeOne Medicines, Ipsen and Novartis. Y.S. reports no conflicts of interest. Y.Pan reports no conflicts of interest. S.R.P. has received research grants from ImmuneOncia Therapeutics, Incyte and Ono Pharmaceutical Korea; consulting fees from BeOne Medicines and Pfizer Pharmaceuticals; and honoraria from MSD Korea. T.K. has received grants from Amgen, BeOne Medicines, Bristol Myers Squibb, Chugai Pharmaceutical, EPS Corporation, Merck Biopharma, MSD K.K., Ono Pharmaceutical, Parexel International, Shionogi & Co and Taiho Pharmaceutical; honoraria from Bristol Myers Squibb, Covidien Japan, MSD K.K., Oncolys Biopharma, Ono Pharmaceutical and Taiho Pharmaceutical; and has participated on data safety monitoring boards or advisory boards for Astellas Pharma, Bristol Myers Squibb, Merck Biopharma, MSD K.K. and Oncolys BioPharma. C-Y.L. reports no conflicts of interest. L.W. has received honoraria from AstraZeneca, BeOne Medicines, Bristol Myers Squibb, MSD, Roche and Servier Laboratories; has acted as a committee member for Bristol Myers Squibb and Servier Laboratories; and owns stocks with BeOne Medicines. D.T. has received grants from Gilead, MSD, Pierre Fabre, Roche, Servier Laboratories and Takeda; consulting fees/honoraria from Amgen, Bristol Myers Squibb, Merck Serono, MSD, Pierre Fabre, Roche, Takeda and Servier Laboratories; support for attending meetings from MSD, Pierre Fabre, Roche and Servier; and has held a leadership or fiduciary role at Federation francophone de cancerologie digestive. R.I. has received honoraria from AI Medical Service, AstraZeneca, Daiichi Sankyo, FUJIFILM Medical, Miyarisan Pharmaceutical, MSD, Olympus and Ono Pharmaceutical; and has participated on a data safety monitoring board or advisory board for Astellas Pharma, Bristol Myers Squibb, Merck Biopharma, MSD K.K. and Oncolys Biopharma. K.W. and Y.Peng are employees of BeOne Medicines, Ltd. J.X. reports no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77327-0.

References

1.Bray, F. et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin.74, 229–263 (2024). [DOI] [PubMed] [Google Scholar]
2.Cancer Genome Atlas Research Network et al. Integrated genomic characterization of oesophageal carcinoma. Nature541, 169–175 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Obermannová, R. et al. Oesophageal cancer: ESMO Clinical Practice Guideline for diagnosis, treatment and follow-up. Ann. Oncol.33, 992–1004 (2022). [DOI] [PubMed] [Google Scholar]
4.Doki, Y. et al. Nivolumab combination therapy in advanced esophageal squamous-cell carcinoma. N. Engl. J. Med.386, 449–462 (2022). [DOI] [PubMed] [Google Scholar]
5.Sun, J. M. et al. Pembrolizumab plus chemotherapy versus chemotherapy alone for first-line treatment of advanced oesophageal cancer (KEYNOTE-590): a randomised, placebo-controlled, phase 3 study. Lancet398, 759–771 (2021). [DOI] [PubMed] [Google Scholar]
6.Luo, H. et al. Effect of camrelizumab vs placebo added to chemotherapy on survival and progression-free survival in patients with advanced or metastatic esophageal squamous cell carcinoma: the ESCORT-1st randomized clinical trial. JAMA326, 916–925 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Lu, Z. et al. Sintilimab versus placebo in combination with chemotherapy as first line treatment for locally advanced or metastatic oesophageal squamous cell carcinoma (ORIENT-15): multicentre, randomised, double blind, phase 3 trial. BMJ377, e068714 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Wang, Z. X. et al. Toripalimab plus chemotherapy in treatment-naïve, advanced esophageal squamous cell carcinoma (JUPITER-06): a multi-center phase 3 trial. Cancer Cell40, 277–288.e273 (2022). [DOI] [PubMed] [Google Scholar]
9.de Ruiter, E. J. et al. Comparison of three PD-L1 immunohistochemical assays in head and neck squamous cell carcinoma (HNSCC). Mod. Pathol.34, 1125–1132 (2021). [DOI] [PubMed] [Google Scholar]
10.Ulas, E. B. et al. Predictive value of combined positive score and tumor proportion score for immunotherapy response in advanced NSCLC. JTO Clin. Res Rep.4, 100532 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Kato, K. et al. Nivolumab plus chemotherapy or ipilimumab versus chemotherapy in patients with advanced esophageal squamous cell carcinoma (CheckMate 648): 29-month follow-up from a randomized, open-label, phase III trial. Cancer Med.13, e7235 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Xu, J. et al. Tislelizumab plus chemotherapy versus placebo plus chemotherapy as first-line treatment for advanced or metastatic oesophageal squamous cell carcinoma (RATIONALE-306): a global, randomised, placebo-controlled, phase 3 study. Lancet Oncol.24, 483–495 (2023). [DOI] [PubMed] [Google Scholar]
13.Raymond, E. et al. Tislelizumab (TIS) + chemotherapy (CT) vs placebo (PBO) + CT in locally advanced unresectable or metastatic esophageal squamous cell carcinoma. PD-L1 biomarker analysis from RATIONALE-306. Ann. Oncol.35, S159 (2024). [Google Scholar]
14.Moehler, M. et al. Concordance between the PD-L1 Tumor Area Positivity score and combined positive score for gastric or esophageal cancers treated with tislelizumab. Mod. Pathol. 38, 100793 (2025). [DOI] [PubMed]
15.Zhang, T. et al. The binding of an anti-PD-1 antibody to FcγRΙ has a profound impact on its biological functions. Cancer Immunol. Immunother.67, 1079–1090 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Dahan, R. et al. FcγRs modulate the anti-tumor activity of antibodies targeting the PD-1/PD-L1 axis. Cancer Cell28, 285–295 (2015). [DOI] [PubMed] [Google Scholar]
17.European Medicines Agency. TEVIMBRA 100 mg concentrate for solution for infusion. Summary of product characteristics. https://www.ema.europa.eu/en/documents/product-information/tevimbra-epar-product-information_en.pdf (2025).
18.US Food and Drug Administration. TEVIMBRA™ (tislelizumab-jsgr) injection, for intravenous use. Prescribing information. https://www.accessdata.fda.gov/drugsatfda_docs/label/2025/761232s004lbl.pdf (2025).
19.Department of Health and Aged Care Therapeutic Goods Administration. TEVIMBRA Australian Prescription Medicine Decision Summary. https://www.tga.gov.au/resources/auspmd/tevimbra-tislelizumab (2024).
20.US Food and Drug Administration. KEYTRUDA® (pembrolizumab) injection, for intravenous use. Prescribing information. https://www.accessdata.fda.gov/drugsatfda_docs/label/2025/125514s178lbl.pdf (2025).
21.US Food and Drug Administration. OPDIVO® (nivolumab) injection, for intravenous use. Prescribing information. https://www.accessdata.fda.gov/drugsatfda_docs/label/2025/125554s131lbl.pdf (2025).
22.European Medicines Agency. KEYTRUDA 25 mg/mL concentrate for solution for infusion. Summary of product characteristics. https://www.ema.europa.eu/en/documents/product-information/keytruda-epar-product-information_en.pdf (2025).
23.European Medicines Agency. OPDIVO 10 mg/mL concentrate for solution for infusion. Summary of product characteristics. https://www.ema.europa.eu/en/documents/product-information/opdivo-epar-product-information_en.pdf (2025).
24.Metges, J.-P. et al. First-line pembrolizumab plus chemotherapy versus chemotherapy in advanced esophageal cancer: longer-term efficacy, safety, and quality-of-life results from the phase 3 KEYNOTE-590 study. J. Clin. Oncol.40, 241–241 (2022). [Google Scholar]
25.Hubner, R. et al. Randomized, global, phase 3 study of tislelizumab + chemotherapy versus placebo + chemotherapy as first-line treatment for advanced/metastatic esophageal squamous cell carcinoma: RATIONALE-306 update. Ann. Oncol.34, S853 (2023). [DOI] [PubMed]
26.Yoon, H. H. et al. Global, randomized, phase III study of tislelizumab plus chemotherapy versus placebo plus chemotherapy as first-line treatment for advanced/metastatic esophageal squamous cell carcinoma (RATIONALE-306 update): minimum 3-year survival follow-up. J. Clin. Oncol.42, 4032 (2024). [DOI] [PubMed] [Google Scholar]
27.Liu, C., Fang, F., Kong, Y. & ElGabry, E. A. Tumor Area Positivity (TAP) score of programmed death-ligand 1 (PD-L1): a novel visual estimation method for combined tumor cell and immune cell scoring. Diagn. Pathol.18, 48 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Paintal, A. S. & Brockstein, B. E. PD-L1 CPS scoring accuracy in small biopsies and aspirate cell blocks from patients with head and neck squamous cell carcinoma. Head Neck Pathol.14, 657–665 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information^{(66.8MB, pdf)}

41467_2026_77327_MOESM2_ESM.pdf^{(14.9KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1^{(50.6KB, docx)}

Reporting Summary^{(2.3MB, pdf)}

Transparent Peer Review file^{(469.1KB, pdf)}

Source Data^{(227.4KB, xlsx)}

Data Availability Statement

[CR1] 1.Bray, F. et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin.74, 229–263 (2024). [DOI] [PubMed] [Google Scholar]

[CR2] 2.Cancer Genome Atlas Research Network et al. Integrated genomic characterization of oesophageal carcinoma. Nature541, 169–175 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR3] 3.Obermannová, R. et al. Oesophageal cancer: ESMO Clinical Practice Guideline for diagnosis, treatment and follow-up. Ann. Oncol.33, 992–1004 (2022). [DOI] [PubMed] [Google Scholar]

[CR4] 4.Doki, Y. et al. Nivolumab combination therapy in advanced esophageal squamous-cell carcinoma. N. Engl. J. Med.386, 449–462 (2022). [DOI] [PubMed] [Google Scholar]

[CR5] 5.Sun, J. M. et al. Pembrolizumab plus chemotherapy versus chemotherapy alone for first-line treatment of advanced oesophageal cancer (KEYNOTE-590): a randomised, placebo-controlled, phase 3 study. Lancet398, 759–771 (2021). [DOI] [PubMed] [Google Scholar]

[CR6] 6.Luo, H. et al. Effect of camrelizumab vs placebo added to chemotherapy on survival and progression-free survival in patients with advanced or metastatic esophageal squamous cell carcinoma: the ESCORT-1st randomized clinical trial. JAMA326, 916–925 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Lu, Z. et al. Sintilimab versus placebo in combination with chemotherapy as first line treatment for locally advanced or metastatic oesophageal squamous cell carcinoma (ORIENT-15): multicentre, randomised, double blind, phase 3 trial. BMJ377, e068714 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.Wang, Z. X. et al. Toripalimab plus chemotherapy in treatment-naïve, advanced esophageal squamous cell carcinoma (JUPITER-06): a multi-center phase 3 trial. Cancer Cell40, 277–288.e273 (2022). [DOI] [PubMed] [Google Scholar]

[CR9] 9.de Ruiter, E. J. et al. Comparison of three PD-L1 immunohistochemical assays in head and neck squamous cell carcinoma (HNSCC). Mod. Pathol.34, 1125–1132 (2021). [DOI] [PubMed] [Google Scholar]

[CR10] 10.Ulas, E. B. et al. Predictive value of combined positive score and tumor proportion score for immunotherapy response in advanced NSCLC. JTO Clin. Res Rep.4, 100532 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Kato, K. et al. Nivolumab plus chemotherapy or ipilimumab versus chemotherapy in patients with advanced esophageal squamous cell carcinoma (CheckMate 648): 29-month follow-up from a randomized, open-label, phase III trial. Cancer Med.13, e7235 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Xu, J. et al. Tislelizumab plus chemotherapy versus placebo plus chemotherapy as first-line treatment for advanced or metastatic oesophageal squamous cell carcinoma (RATIONALE-306): a global, randomised, placebo-controlled, phase 3 study. Lancet Oncol.24, 483–495 (2023). [DOI] [PubMed] [Google Scholar]

[CR13] 13.Raymond, E. et al. Tislelizumab (TIS) + chemotherapy (CT) vs placebo (PBO) + CT in locally advanced unresectable or metastatic esophageal squamous cell carcinoma. PD-L1 biomarker analysis from RATIONALE-306. Ann. Oncol.35, S159 (2024). [Google Scholar]

[CR14] 14.Moehler, M. et al. Concordance between the PD-L1 Tumor Area Positivity score and combined positive score for gastric or esophageal cancers treated with tislelizumab. Mod. Pathol. 38, 100793 (2025). [DOI] [PubMed]

[CR15] 15.Zhang, T. et al. The binding of an anti-PD-1 antibody to FcγRΙ has a profound impact on its biological functions. Cancer Immunol. Immunother.67, 1079–1090 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Dahan, R. et al. FcγRs modulate the anti-tumor activity of antibodies targeting the PD-1/PD-L1 axis. Cancer Cell28, 285–295 (2015). [DOI] [PubMed] [Google Scholar]

[CR17] 17.European Medicines Agency. TEVIMBRA 100 mg concentrate for solution for infusion. Summary of product characteristics. https://www.ema.europa.eu/en/documents/product-information/tevimbra-epar-product-information_en.pdf (2025).

[CR18] 18.US Food and Drug Administration. TEVIMBRA™ (tislelizumab-jsgr) injection, for intravenous use. Prescribing information. https://www.accessdata.fda.gov/drugsatfda_docs/label/2025/761232s004lbl.pdf (2025).

[CR19] 19.Department of Health and Aged Care Therapeutic Goods Administration. TEVIMBRA Australian Prescription Medicine Decision Summary. https://www.tga.gov.au/resources/auspmd/tevimbra-tislelizumab (2024).

[CR20] 20.US Food and Drug Administration. KEYTRUDA® (pembrolizumab) injection, for intravenous use. Prescribing information. https://www.accessdata.fda.gov/drugsatfda_docs/label/2025/125514s178lbl.pdf (2025).

[CR21] 21.US Food and Drug Administration. OPDIVO® (nivolumab) injection, for intravenous use. Prescribing information. https://www.accessdata.fda.gov/drugsatfda_docs/label/2025/125554s131lbl.pdf (2025).

[CR22] 22.European Medicines Agency. KEYTRUDA 25 mg/mL concentrate for solution for infusion. Summary of product characteristics. https://www.ema.europa.eu/en/documents/product-information/keytruda-epar-product-information_en.pdf (2025).

[CR23] 23.European Medicines Agency. OPDIVO 10 mg/mL concentrate for solution for infusion. Summary of product characteristics. https://www.ema.europa.eu/en/documents/product-information/opdivo-epar-product-information_en.pdf (2025).

[CR24] 24.Metges, J.-P. et al. First-line pembrolizumab plus chemotherapy versus chemotherapy in advanced esophageal cancer: longer-term efficacy, safety, and quality-of-life results from the phase 3 KEYNOTE-590 study. J. Clin. Oncol.40, 241–241 (2022). [Google Scholar]

[CR25] 25.Hubner, R. et al. Randomized, global, phase 3 study of tislelizumab + chemotherapy versus placebo + chemotherapy as first-line treatment for advanced/metastatic esophageal squamous cell carcinoma: RATIONALE-306 update. Ann. Oncol.34, S853 (2023). [DOI] [PubMed]

[CR26] 26.Yoon, H. H. et al. Global, randomized, phase III study of tislelizumab plus chemotherapy versus placebo plus chemotherapy as first-line treatment for advanced/metastatic esophageal squamous cell carcinoma (RATIONALE-306 update): minimum 3-year survival follow-up. J. Clin. Oncol.42, 4032 (2024). [DOI] [PubMed] [Google Scholar]

[CR27] 27.Liu, C., Fang, F., Kong, Y. & ElGabry, E. A. Tumor Area Positivity (TAP) score of programmed death-ligand 1 (PD-L1): a novel visual estimation method for combined tumor cell and immune cell scoring. Diagn. Pathol.18, 48 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Paintal, A. S. & Brockstein, B. E. PD-L1 CPS scoring accuracy in small biopsies and aspirate cell blocks from patients with head and neck squamous cell carcinoma. Head Neck Pathol.14, 657–665 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

First-line tislelizumab plus chemotherapy versus placebo plus chemotherapy in unresectable advanced/metastatic esophageal squamous cell carcinoma: minimum 3-year survival follow-up of a randomized phase 3 trial (RATIONALE-306)

Harry H Yoon

Ken Kato

Eric Raymond

Richard Hubner

Yongqian Shu

Yueyin Pan

Sook Ryun Park

Takashi Kojima

Chen-Yuan Lin

Lucjan Wyrwicz

David Tougeron

Ryu Ishihara

Kaijun Wang

Yanyan Peng

Jianming Xu

Abstract

Introduction

Results

Patient demographics and baseline characteristics

Treatment exposure

Overall survival

Fig. 1. Kaplan–Meier plots of overall survival (stratified analyses).

Progression-free survival

Fig. 2. Kaplan–Meier plot of progression-free survival by investigator assessment.

Concordance between TAP score and CPS in the PD-L1 subgroups

Fig. 3. Sankey diagram of PD-L1 TAP score versus CPS.

Antitumor activity

Patient-reported outcomes

Table 1.

Table 2.

Safety and tolerability

Fig. 4. Incidence of selected TEAEs over time (safety analysis set).

Discussion

Methods

Ethics statement

Study design

Patients

Procedures

Endpoints and assessments

Statistics and reproducibility

Reporting summary

Supplementary information

Source data

Acknowledgements

Author contributions

Peer review

Peer review information

Funding

Data availability

Competing interests

Footnotes

Supplementary information

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases