Chemoradiotherapy and Subsequent Immunochemotherapy as Conversion Therapy in Unresectable Locally Advanced Esophageal Squamous Cell Carcinoma: A Phase II NEXUS-1 Trial Free

Esophageal squamous cell carcinoma (ESCC) is the predominant histologic type of esophageal cancer in Eastern Asia, accounting for approximately 90% of cases (1, 2). Due to the absence of early symptoms and robust diagnostic markers, only approximately 50% of ESCC cases present with potentially resectable disease. Locally advanced esophageal cancer, especially cT4 disease, may preclude curative resection, resulting in an unfavorable prognosis. Guidelines have recommended for more than 15 years that unresectable locally advanced esophageal cancer should be treated with a combination of chemotherapy and radiation therapy (3, 4). However, the long-term prognosis for patients receiving definitive chemoradiation (dCRT) remains unfavorable, with the reported 5-year survival rate being a meager 35.8% (5).

Trimodality therapy has the potential to improve outcomes in patients with initially unresectable cT4b ESCC (6, 7). Multiple studies have also confirmed that patients with initially unresectable esophageal cancer who achieved R0 resection following conversion therapy have a significantly favorable prognosis (8–10). Therefore, enhancing the R0 resection rate and pathologic response rate through comprehensive treatment may be promising for improving the prognosis of unresectable esophageal cancer. Moreover, compared with chemotherapy alone, an improved outcome with chemoradiotherapy (CRT) as the initial therapy for conversion surgery in T4b esophageal cancer has been reported in a recent multicenter phase II trial, suggesting that the addition of radiotherapy plays a crucial role in conversion therapy for unresectable patients (8).

Immune checkpoint inhibitor (ICI)-based strategies have shown good efficacy in ESCC (11–13). In the PALACE-1 study, which is a phase I trial with a limited number of patients included, preoperative pembrolizumab plus CRT yielded an impressive pathologic response in resectable ESCC, with a pathologic complete response (pCR) rate of 55.6% (14). In a real-world study, immunochemotherapy (iCT) reached an R0 resection rate of up to 94% as conversion therapy for unresectable locally advanced ESCC (9). Nevertheless, a randomized controlled trial is needed to evaluate the application of CRT combined with iCT in conversion therapy for unresectable locally advanced ESCC.

Tislelizumab is a humanized IgG4 monoclonal antibody exhibiting high affinity and binding specificity for PD1. In the RATIONALE-306 phase III study, tislelizumab plus chemotherapy showed superior overall survival (OS) in advanced ESCC compared with the control arm using placebo plus chemotherapy (median, 17.2 months vs. 10.6 months), with a manageable safety profile (11). Additionally, tislelizumab plus chemotherapy/CRT as neoadjuvant therapy demonstrated promising antitumor activity for resectable ESCC, with high pCR and R0 resection rates and acceptable tolerability (15, 16). Taken together, these robust data provide a compelling rationale for evaluating tislelizumab-based regimens in the conversion therapy setting to improve the prognosis of unresectable locally advanced ESCC.

In this study, we designed a prospective phase II trial to evaluate the antitumor activity and safety of CRT followed by tislelizumab-based iCT and surgery in unresectable locally advanced ESCC and to explore potential biomarkers for predicting tumor response and survival outcomes.

The NEoadjuvant rX for borderline Unresectable esophageal Squamous cell carcinoma (NEXUS)-1 trial was a single-center, open-label, phase II study to investigate the safety and efficacy of dCRT followed by tislelizumab-based iCT as conversion therapy in patients with unresectable locally advanced ESCC. The Ethics Commission of Cancer Hospital of Chinese Academy of Medical Sciences approved the study protocol (21/133-2804). All participants gave written informed consent prior to the study. This trial was conducted in accordance with the Good Clinical Practice guidelines and the Declaration of Helsinki. It was registered with the Chinese Clinical Trial Registry (ChiCTR2100054327).

Patients, ages 18 to 75 years, were eligible for this trial if they had histologically proven unresectable locally advanced thoracic ESCC [T4, borderline resectable T3 (invasion of other organs, such as the trachea, bronchus, or aorta, cannot be ruled out on imaging; ref. 17) with/without unresectable lymph node metastasis invading adjacent organs, or clinically confirmed unresectable disease by surgeons] and no evidence of distant metastases (excluding supraclavicular lymph node metastasis). Eligible patients were also required to have an Eastern Cooperative Oncology Group performance status of 0 to 1 and adequate organ and bone marrow function. The exclusion criteria were as follows: distant or hematogenous metastasis beyond the supraclavicular lymph node region at diagnosis, existing or high risk of esophageal perforation, and having received any prior antitumor therapy, including but not limited to surgery, radiotherapy, chemotherapy, targeted therapy, and immunotherapy.

To assess inclusion and exclusion criteria, all subjects underwent a comprehensive clinical assessment, including cervical, chest, and abdominal CT scans with contrast, ¹⁸F-fluorodeoxyglucose PET-CT, barium esophagography, esophagogastroduodenoscopy, and endoscopic ultrasonography. Enhanced CT scans were used for clinical response assessments. Nutritional assessment was planned before conversion therapy. The insertion of a feeding tube or parenteral feeding was offered to subjects at risk.

Eligible patients were treated with radiotherapy (50 Gy/25f, 5 days per week) for 5 weeks, nab-paclitaxel (100 mg) and cisplatin (25 mg/m²) on day 1 of a 7-day cycle for five cycles. Radiation was delivered using intensity-modulated radiotherapy or volumetric-modulated arc therapy. The gross tumor volume (GTV) was defined as visible primary tumor and metastatic regional nodes delineated by physicians using all possible resources (barium esophagography, CT, esophagogastroscopy, endoscopic ultrasonography, and PET-CT). The clinical target volume (CTV) consisted of the primary tumor plus a 0.5-cm circumferential margin, a 3-cm craniocaudal margin, and metastatic regional nodes, plus a 0.5-cm margin in all directions and covering the corresponding lymphatic drainage region. The planning target volume was defined as the CTV plus a uniform 0.3- to 0.5-cm margin. The elective lymph nodes, for tumors located in the upper esophagus according to the eighth edition of American Joint Committee on Cancer staging atlas (18), included supraclavicular, upper mediastinal, paratracheal paraesophageal, and subcarinal nodes at the same levels as the CTV. For tumors located in the middle of the esophagus, the elective lymph nodes were middle mediastinal, paratracheal, paraesophageal, and subcarinal nodes at the same levels as the CTV. For distal tumors, the elective lymph nodes included lower mediastinal, paracardial, left gastric, and celiac nodes at the same levels as the CTV.

Dose constraints to the organs at risk were as follows: the volume of lung tissue receiving 20 Gy or more should not exceed 25% of the total lung volume (V20 <25%) and the mean dose of lung tissue should be lower than 14 Gy (Dmean lung <14 Gy). Other dose constraints to organs at risk included the following: V40 heart <30%, V30 heart <40%, V40 stomach <30%, Dmax stomach <55Gy, V40 small intestine <30%, V30 liver <25%, V20 kidney <25%, and Dmax spinal cord PRV <45 Gy. Concurrent chemotherapy was discontinued until recovery to grade 1 if grade ≥2 toxicity developed, whereas radiotherapy continued. CRT was stopped if grade ≥3 toxicity occurred. Full details are available in Supplementary Methods.

Patients without clinical progression after dCRT continued to receive two cycles of tislelizumab (200 mg) on day 1 and nab-paclitaxel (150 mg/m²) and cisplatin (75 mg/m²) on day 2 per 21 days. Dose reduction was not allowed for tislelizumab. Full details of chemotherapy dose adjustment are available in Supplementary Methods.

Right transthoracic procedures (McKeown esophagectomy) with extended two/three-field lymph node dissection were performed within 4 weeks from the last iCT. Tumor invasion of adjacent organs was considered resectable, provided that en bloc resection of the entire tumor could be performed with a tumor-free margin according to the physicians’ discretion. The (circumferential) resection margins were evaluated using the College of American Pathologists protocol (19), and tumor regression grading was assessed using the Mandard scale (20).

Patients with any extent of residual disease received adjuvant therapy with tislelizumab for up to 1 year or until disease progression or intolerable toxicity. Postoperative follow-up CT scanning was performed every 3 months for 2 years and every 6 months thereafter to collect survival information (Fig. 1).

The primary endpoint was the 1-year progression-free survival (PFS) rate, defined as the proportion of patients without disease progression or death within 1 year from treatment initiation. Disease progression was defined as recurrence after R0 resection or disease progression resulting from unresectability. Secondary endpoints included the R0 resection rate, major pathologic response (MPR), pCR, postoperative complications, 2-year PFS rate, 1- or 2-year OS rate, locoregional recurrence-free survival (LRFS), distant metastasis-free survival (DMFS), and safety. R0 resection was defined as no tumor identified on microscopic examination of proximal, distal, or circumferential margins. The MPR was defined as tumor regression grading (TRG) 1 and 2. pCR was defined as no evidence of residual tumor cells in the resected specimen. The 1- or 2-year OS rate was defined as the proportion of patients who remained alive within 1 or 2 years from treatment initiation. LRFS was defined as the time from treatment initiation to the first local or regional relapse. DMFS was defined as the time from treatment initiation to distant metastasis. An exploratory endpoint was also preplanned to evaluate the correlation of biomarkers [such as PD-L1 expression and the dynamic monitoring of minimal residual disease (MRD)] with clinical efficacy.

Postoperative complications were described according to the Clavien–Dindo (CD) classification. Treatment-related adverse events (TRAE) were monitored and graded according to NCI Common Terminology Criteria version 5.0. Immune-related AEs (irAE) were defined as AEs of clinical interest with potentially drug-related immunologic causes.

Tumor PD-L1 status of pretreatment tissue samples were evaluated by IHC using the 22C3 assay (Dako North America), with PD-L1 expression determined using combined positive scores (CPS). Serial plasma ctDNA samples were collected at baseline (t1), post-CRT (t2), and preoperation (t3) for MRD dynamic monitoring. Plasma ctDNA samples at t1 and t2/t3 were subject to genomic profiling using the Radiotron and Shielding ULTRA panels (Nanjing Geneseeq Technology Inc.), respectively. Full details on plasma sample collection and targeted next-generation sequencing are available in Supplementary Methods.

The sample size calculation was based on the 1-year PFS rate (primary endpoint). In this trial, we hypothesized that CRT followed by iCT would increase the 1-year PFS rate from 42% to 65% in patients with unresectable locally advanced ESCC (21, 22). The enrollment period for patients was estimated to be 12 months. Accordingly, a total of 27 patients would be sufficient, with a one-sided α of 5% and a power of 80%. Considering a dropout rate of 10%, a sample size of 30 patients was determined.

Efficacy analysis was performed in the full analysis set and surgical analysis set. The full analysis set consisted of all participants who received study treatment and had at least one follow-up data record available. The surgical analysis set included all participants who received R0 resection. Safety was analyzed in all patients who received study medication and had at least one safety assessment. The Fisher exact test and linear regression models were used to compare the frequencies and means of the independent groups, respectively. For survival data, Kaplan–Meier curves were generated, with log-rank tests used to compare differences. Cox proportional hazards models were fitted to estimate HRs with 95% confidence intervals (CI), and the proportionality of hazards was assessed using log(−log) survival plots. Individuals without any plasma ctDNA sample were excluded from exploratory biomarker analyses. All reported P values were two-tailed, and P values <0.05 were considered statistically significant. Data were analyzed using R software (version 4.0.3), along with the survival, survminer, and epiR packages.

The data generated in this study are not publicly available due to patient privacy but are available upon reasonable request from the corresponding author Y. Li ([email protected]) for 10 years.

Between April 2021 and April 2023, 32 patients with unresectable locally advanced ESCC were screened, of which 2 were excluded due to declined participation (n = 1) and suspected lung metastasis (n = 1). Ultimately, 30 eligible patients were enrolled in this trial and received at least CRT (Fig. 2). Table 1 summarizes the baseline patient characteristics. The median age was 56 years (range, 42–73). Of the 30 included patients, 27 (90.0%) were male; 11 (36.6%) had borderline resectable T3 disease; 23 (76.7%) had stage IVa disease; and 28 (97.3%) had a baseline Eastern Cooperative Oncology Group performance status of 1, suggesting the representativeness of our study cohort (Supplementary Table S1). PD-L1 expression data were available for 26 (86.7%) patients, with 12 (40.0%) having a CPS ≥10.

Of 30 patients receiving CRT, 26 (86.7%) completed radiotherapy and 21 (70.0%) completed at least four cycles of concurrent chemotherapy during CRT. Four patients did not complete radiotherapy due to economic reasons (n = 1) or AEs (n = 3). One patient refused subsequent iCT and underwent surgery ahead of schedule. The remaining five patients did not receive iCT due to economic reasons (n = 2), esophageal fistula (n = 1), myelosuppression (n = 1), or the COVID-19 pandemic (n = 1; Fig. 2). Of the 24 patients receiving subsequent iCT, 17 (70.8%) completed two cycles of tislelizumab plus chemotherapy, whereas seven (29.2%) discontinued iCT prematurely due to pneumonitis (n = 3), esophageal fistula (n = 1), grade 4 white blood cell or platelet toxicity (n = 2), or the COVID-19 pandemic (n = 1). Among these 24 patients with iCT, four (16.7%) did not undergo surgery due to unresectable disease assessed by surgeons (n = 2) and AEs (esophagomediastinal fistula, n = 1; pneumonitis, n = 1). Ultimately, 21 (70.0%) of the 30 patients were assessed for esophagectomy eligibility, with 20 achieving R0 resection (R0 resection rate: 95.2%, 20/21), whereas one patient was discovered to have persistent tumor invasion intraoperatively and received exploratory thoracotomy. The median numbers of lymph nodes resected and positive lymph nodes were 27 (range, 11–55) and 0 (range, 0–2), respectively.

Of 30 enrolled patients receiving CRT, 20 (66.6%) underwent definitive esophagectomy, with 13 of 30 (43.3%) achieving pCR and 16 of 30 (53.0%) achieving MPR. The response was evaluated according to the Mandard tumor regression scale, which classified patients into responders (TRG 1) and nonresponders (TRG 2, TRG 3, TRG 4, and TRG 5). Of the 20 patients who underwent esophagectomy, 14 (70.0%) were responders and 6 (30.0%) were nonresponders.

As of the data cut-off on September 30, 2023, the median follow-up time was 21 months (95% CI, 3–27). In the full analysis set, the median PFS was 26.3 months [95% CI, 17.8–not reached (NR)], with 1-year and 2-year PFS rates of 79.4% (95% CI, 59.8%–90.2%) and 57.6% (95% CI, 36.0%–74.2%), respectively (Fig. 3A). The median PFS was significantly longer in the R0 resection group compared with the nonsurgery group (NR vs. 8.4 months; HR = 0.28; 95% CI, 0.08–0.84; P = 0.02; Fig. 3B), with 1-year PFS rates of 94.7% (95% CI, 68.1%–99.2%) vs. 48.0% (95% CI, 16.1%–74.5%). There was no statistically significant difference in PFS between the pCR and non-pCR groups (P > 0.05; Supplementary Table S2).

The median OS in full analysis set was NR (95% CI, 25.4 months–NR), with OS rates of 89.6% (95% CI, 71.0%–96.5%) at 1 year and 72.6% (95% CI, 50.5%–86.0%) at 2 years (Fig. 3C). Patients achieving R0 resection had significantly longer OS compared with those not receiving surgery (median, NR vs. 19.2 months; HR = 0.18; 95% CI, 0.04–0.73; P < 0.01; Fig. 3D), with 1-year OS rates of 100% (95% CI, 100.0%–100.0%) vs. 68.6% (95% CI, 30.5%–88.7%). No statistically significant difference in OS was observed between the pCR group and non-pCR groups (P > 0.05; Supplementary Table S2). Patients with R0 resection showed significant improvements in LRFS and DMFS compared with those did not undergo surgery (all P < 0.05; Supplementary Fig. S1; Supplementary Table S3).

TRAEs during CRT and iCT are summarized separately in Table 2. TRAEs of any grade occurred in all patients. The most common TRAEs during CRT were radiation esophagitis (93.3%), anemia (76.7%), and leukopenia (76.7%). Grade 3 or higher TRAEs occurred in 14 (46.7%) patients, with radiation esophagitis, anemia, leukopenia, neutropenia, and vomiting observed most frequently, each at 6.7%. Of the 24 patients receiving iCT, 20.8% (n = 5) developed immune-related pneumonitis, with 8.3% (n = 2) experiencing grade 3 pneumonitis. One patient discontinued conversion surgery due to grade 3 immune-related pneumonitis, whereas the condition resolved in the other four patients.

Feasibility was assessed by examining the proportion of eligible patients who underwent conversion surgery without significant delay after completing stages I and II. Substantial treatment-related delay was defined as any delay exceeding 12 weeks. After completing stages I and II, all operative candidates except one proceeded to surgery without delay due to grade 3 irAEs.

All postoperative complications are summarized in Supplementary Table S4. Most postoperative complications were CD grade 3 or less, mainly including CD grade IIIa anastomotic leak (n = 2, 10.0%), pleural effusion (n = 2, 10.0%), and pneumonia (n = 1, 5.0%) and CD grade IIIb tracheal collapse (n = 1, 5.0%). There were no CD grade IV or higher complications or mortality within 90 days after surgery.

Of the 20 patients receiving R0 resection, 18 had PD-L1 expression data. There was no statistically significant difference observed between patients with CPS ≤10 (n = 9) and CPS ≥10 (n = 9) in terms of pCR, MPR, and the postoperative pathologic stage (all P > 0.05; Supplementary Table S5). Statistically significant differences in PFS, OS, LRFS, and DMFS were not observed between these two subgroups (all P > 0.05; Supplementary Fig. S2).

A total of 75 serial plasma ctDNA samples were collected from 28 patients, including 28 t1, 28 t2, and 19 t3 samples. Two patients without plasma ctDNA samples were excluded from exploratory biomarker analyses. At t1, 67.9% (19/28) of patients were detected with at least one somatic genetic aberration and defined as MRD⁺. Their mutational landscapes are shown in Fig. 4A. TP53 mutations were observed in each MRD⁺ patient (67.9%), and other commonly mutated genes included NOTCH1 (21.4%), NFE2L2 (14.3%), and TET2 (10.7%). Four signaling pathways were frequently mutated with prevalence of more than 20%, including p53 (67.9%), Notch (32.1%), NRF2 (21.4%), and PI3K-AKT (21.4%).

No significant differences in clinical characteristics, such as clinical tumor–node–metastasis (cTNM) and PD-L1 expressions, were observed between MRD⁺ and MRD⁻ patients at t1 (Supplementary Table S6), and the t1 MRD status was unable to indicate pathologic responses (Supplementary Table S7). Compared with t1 MRD⁻ patients, MRD⁺ patients exhibited similar PFS (median PFS, 26.3 months vs. NR; HR = 1.23; 95% CI, 0.32–4.66; P = 0.76; Supplementary Fig. S3A) and OS (median OS, NR vs. NR; HR = 2.76; 95% CI, 0.33–23.32; P = 0.33; Supplementary Fig. S3B). Notably, compared with patients with a wild-type NRF2 pathway at t1, six patients harboring mutated counterparts exhibited a trend toward inferior PFS (median PFS, 8.5 months vs. NR; HR = 3.11; 95% CI, 0.90–10.73; P = 0.06; Fig. 4B) and OS (median OS, 13.4 months vs. NR; HR = 4.04; 95% CI, 0.89–18.38; P = 0.05; Fig. 4C). Other prevalent mutated genes and pathways with prevalence of more than 20% were screened, and the PI3K–AKT pathway had a trend toward association with PFS and OS (Supplementary Table S8). In multivariable Cox proportional hazards models including NRF2 and PI3K-AKT pathways, and surgical resection, which was statistically significant in univariate analyses (Supplementary Figs. S4 and S5), mutated NRF2 pathway remained associated with poorer PFS (HR = 3.81; 95% CI, 1.00–14.49; P = 0.05; Supplementary Fig. S6A) and had a trend toward association with OS (HR = 3.86; 95% CI, 0.74–20.25; P = 0.11; Supplementary Fig. S6B).

The treatment regimens, cTNM staging, t1 CPS of PD-L1 expression, MRD status, and clinical outcomes of 28 patients included in the biomarker analyses are summarized in Fig. 4D. After completing CRT, the MRD detection rate at t2 reduced sharply compared with baseline at t1 (14.3% vs. 67.9%; P < 0.01; Fig. 4E). The t2 MRD status was not associated with baseline clinical characteristics (Supplementary Table S6); however, MRD⁻ patients at t2 seemed to have a relatively high proportion of ypTNM stage I diseases compared with MRD⁺ patients (82.4% vs. 33.3%; P = 0.07; Supplementary Table S7). Notably, the only patient (P14) being MRD⁺ at t3, who was cTNM stage IVa at initial diagnosis and ypTNM stage III after completing neoadjuvant therapy and surgical treatment, failed to achieve pCR and MPR with TRG4 measurement (Fig. 4D).

The MRD dynamics from t1 to t3 showed decreases in the maximum and mean variant allele frequencies, as well as ctDNA concentration (Fig. 4F). Patients were categorized into an unfavorable group comprising those being MRD⁺ at t2 or t3 (n = 5) and a favorable group comprising those being MRD⁻ at both t2 and t3 (n = 23). When adjusting for t1 mutational status of the NRF2 pathway and surgical resection, the unfavorable MRD dynamic pattern was associated with a trend toward inferior PFS (HR = 4.89; 95% CI, 0.96–24.96; P = 0.06; Fig. 4G) and was associated with inferior OS (HR = 42.85; 95% CI, 2.38–770.9; P = 0.01; Fig. 4H).

To the best of our knowledge, this is the first trial to evaluate the multidisciplinary treatment of CRT followed by iCT plus surgery for unresectable locally advanced ESCC. Clinically significant benefit was observed with a 1-year PFS rate of 79.4%, meeting the primary endpoint. Particularly noteworthy was the encouraging 1-year PFS rate of 94.7% observed in patients achieving R0 resection.

dCRT has traditionally yielded suboptimal outcomes for locally advanced ESCC, particularly in stages III to IV of disease. Surgery is emerging as a critical strategy for enhancing local tumor control. Studies of concurrent dCRT in ESCC with T4 tumors and/or M1 lymph nodes have reported 3-year OS rates ranging from 25.9% and 40.4% (22–24). Our previous national multicenter study showed a 5-year survival rate of 23.5% for stage IVa (T4NxM0 or anyTN3M0) ESCCs treated with dCRT (median OS, 21.6 months; ref. 5). Locoregional recurrence was the predominant cause of treatment failure. The majority of local failures after dCRT occurred within the GTV (107/119, 90%), significantly impacting OS (median, GTV failure vs. non-GTV failure, 23.3 months vs. 31.6 months; P < 0.01; ref. 25). In cases of large localized tumors, incorporating surgery could enhance local tumor control. The Fédération Francophone de Cancérologie Digestive (FFCD) 9102 trial demonstrated that neoadjuvant CRT followed by surgery significantly reduced the local tumor control rate compared with dCRT alone in ESCC responsive to neoadjuvant CRT (33.6% vs. 43.0%; P < 0.01; ref. 26). Similarly, lower locoregional recurrence rates were observed in locally advanced ESCCs receiving induction chemotherapy followed by neoadjuvant CRT plus surgery compared with those with dCRT alone (2-year local PFS; 64.3% vs. 40.7%; HR = 2.1; 95% CI, 1.3–3.5; P < 0.01); however, survival improvement was not observed (27).

Nowadays, immunotherapy has emerged as the preferred first-line treatment of advanced or metastatic ESCC (12, 13, 28). Combining CRT with immunotherapy has shown promising antitumor activity in ESCC. The PALACE-1 study demonstrated that preoperative pembrolizumab plus CRT achieved notable pathologic response in resectable ESCC, with pCR and MPR rates of 55.6% and 88.9%, respectively (14). Similarly, the NEOCRTEC1901 trial reported a pCR rate of 50% (21/42) with toripalimab combined with neoadjuvant CRT in locally advanced ESCC, indicating the potential for higher pCR rates with this combination (29). Recent studies have also highlighted the efficacy of combining dCRT with immunotherapy in unresectable ESCC. A multicenter phase II trial by Ai and colleagues (30) reported a 2-year local control rate of 81.7% (95% CI, 72.7%–90.7%) in locally advanced ESCC under the additional induction of sintilimab to standard concurrent CRT as a nonsurgical treatment option, which was higher than that in patients treated with concurrent CRT alone (71.3%). The EPOC 1802 trial, which has not been published yet, showed that patients with unresectable ESCC (stage IV disease, 97.5%) who received sequential immunotherapy after dCRT therapy achieved a 1-year PFS rate of 29.6% and a 1-year OS rate of 65.8% (31). Induction chemotherapy plus camrelizumab followed by dCRT in the ImpactCRT trial reported a 1-year PFS rate of 74.2% and a 1-year OS rate of 87.6% in a similar population (32). Additionally, there are multiple clinical trials investigating the efficacy of combining immunotherapy with definitive CRT in patients with unresectable esophageal cancer, such as KUNLUN (33), RATIONALE 311 (34), KEYNOTE-975 (35), and TENERGY (36). The 1-year PFS and OS rates in our trial were 79.4% and 89.6%, respectively, which were higher than those of the EPOC1802 and ImpactCRT trials. Notably, all enrolled patients in our trial had unresectable locally advanced ESCC or borderline resectable T3 ESCC clinically evaluated as unresectable.

The design of our study, which administers CRT before iCT and then surgery, is based on two primary rationales: (i) administering CRT first can cause significant tumor regression, and neoantigens induced by radiation therapy can broaden the immunotherapeutic window (37, 38). Chemotherapy and radiotherapy have the potential to enhance the immunogenicity of the tumor microenvironment, thereby increasing the probability of triggering an antitumor immune response (39–41). (ii) We drew inspiration from the CheckMate 577 study’s protocol for esophageal or gastroesophageal junction cancer, which applied nivolumab as adjuvant therapy in patients receiving neoadjuvant CRT with residual pathologic disease (42), as well as the PACIFIC study (43). The positive outcomes of this sequence suggest that a similar approach might be beneficial in the conversion therapy of unresectable locally advanced ESCC. However, there is no consensus on the sequence of immunotherapy and radiotherapy or whether a combination of radiotherapy, chemotherapy, and immunotherapy should be administered concurrently, which requires further investigation.

The National Comprehensive Cancer Network guidelines recommended a preoperative dose range of 41.4 to 50.4 Gy, with 50 to 50.4 Gy considered a definitive radiotherapy dose for dCRT in esophageal cancer. Two randomized controlled trials have shown that there was no significant difference in disease control and survival outcomes between the 50 and 60.6 Gy in patients with esophageal cancer treated with dCRT (44, 45). The radiotherapy dose in our trial ensured that patients who were not subsequently converted to resectable could be treated to a definitive dose, as 50 Gy could be considered a definitive dose. Patients in our trial who received CRT but were unable to convert to surgery achieved a median OS of 19.2 months, comparable to our previous data (T4NxM0 or anyTN3M0; median OS, 21.6 months; ref. 5), suggesting that CRT followed by iCT therapy could achieve promising survival benefits.

In terms of safety, our trial observed that 12 patients (40%) developed at least one grade 3 or worse AEs, which was higher compared with the NEOCRTEC1901 trial of immunotherapy plus CRT in ESCC. However, these events in our trial were generally manageable and tolerable. It is important to note that the addition of ICIs to CRT, as highlighted by the EC-CERT-001 trial, might increase the risk of pulmonary toxicity (46). In the present trial, five (20.8%) patients experienced immune-related pneumonitis, including two (8.3%) with grade 3 immune-related pneumonitis. This incidence rate was higher than that reported in other studies (all grade, 0%–6.7%; grade ≥3, 0%–5%; refs. 47–50), with four of the cases resolving after appropriate treatment, allowing patients to undergo subsequent surgery. Notably, the single patient who developed grade 3 pneumonitis and was unable to receive surgery had been diagnosed with pulmonary interstitial fibrosis before participating in this trial. We recommended to perform pneumonitis surveillance in the application of this combination regimen, particularly in patients with baseline pulmonary comorbidities. As reported, the most common AEs with PD1 inhibitors were pneumonitis, arthralgia, and hypothyroidism (51, 52); however, there were no other irAEs observed in our trial with the exception of immune-related pneumonitis, probably due to using tislelizumab for only two cycles (6 weeks). It has been reported that immune-related pneumonitis is an AE occurring relatively early (median onset ∼ 2.8 months; range, 9 days–19.2 months; ref. 53), whereas other irAEs manifest later (54).

Our exploratory biomarker analyses on MRD demonstrated the prognostic potential of serial plasma ctDNA under conversion CRT followed by iCT in unresectable locally advanced ESCC. Consistently, multiple studies have revealed the clinical applicability of ctDNA surveillance among patients with ESCC (55–57). For instance, NG and colleagues suggested that postneoadjuvant ctDNA alterations and ctDNA alterations at 6 months postoperation could identify patients with poorer PFS after undergoing neoadjuvant CRT (HR = 3.16; 95% CI, 1.17–8.52; P = 0.02) and patients receiving curative surgical treatment with inferior OS (HR = 5.71; 95% CI, 1.81–17.97; P < 0.01), respectively (46). A previous study by our team demonstrated the prognostic value of plasma ctDNA at the fourth week of neoadjuvant/definitive CRT and 1 to 3 months postneoadjuvant/dCRT (56). An open-label, single-arm, phase II trial investigating neoadjuvant sintilimab and chemotherapy treatment also suggested that ctDNA low-releasers before neoadjuvant therapy were more likely to achieve pCR (P < 0.01; ref. 58). Our results revealed the potential of negative t2 MRD status in predicting better response after surgical treatment and the applicability of serial plasma ctDNA collection for MRD monitoring; however, further studies with larger study cohorts are required.

Mutated NRF2 signaling pathway suggested worse prognosis in our trial. NRF2 is a key transcription factor in regulating antioxidant stress response, playing a crucial role in inducing the body’s antioxidant response (59). A comprehensive integrative analysis, for the first time, stratified ESCC into four distinct molecular subtypes: cell-cycle pathway activation, NFR2 oncogenic activation, immune suppression, and immune modulation. Among patients treated with immunotherapy, those with progressive disease clustered with the NFR2 oncogenic activation subtype, exhibiting activated NRF2 signaling pathway related with NFE2L2 copy-number amplifications and NFE2L2, CUL3, and KEAP1 mutations (60). This aligns with our finding that patients with t1-mutated NRF2 signaling pathway had potentially poor prognosis. Additionally, Xia and colleagues (61) have found that NFR2 promotes resistance to radiation therapy in ESCC through the activation of autophagy related to CaMKIIα. Therefore, to achieve a more thorough evaluation of the mutated NRF2 signaling pathway, it is imperative to conduct further validation in studies with a larger sample size.

Several limitations of our trial should be considered when interpreting the results. First, it was a single-arm phase II trial with a relatively small sample size, which limited our ability to detect the prognostic value of pCR with sufficient statistical power. Thus, the similar prognosis between patients achieving pCR and non-pCR should be interpreted with caution. Second, this trial was conducted in a single center, potentially limiting the generalizability of the results beyond the trial population, particularly to patients treated at low-volume centers. Research has reported that patients at high-volume centers have a higher survival rate after esophageal cancer surgery, (62) just like our high-volume center, which has more than 1,200/year esophageal cancer surgeries and rich experience in surgery, postoperative complications, and postoperative care. Third, following conversion therapy, our surgeons rely solely on CT scans for assessment of resectability, without mandatory endoscopic ultrasonography or bronchoscopy. Despite this, our surgeons’ extensive 30-year experience resulted in a notably high R0 resection rate of 95.2%. Although additional modalities such as ultrasound, bronchoscopy, MRI, PET-CT, and others could enhance the evaluation of R0 resection feasibility, their use may increase patient waiting times and economic burden. Fourth, biomarker and MRD analyses were exploratory, and long-term ctDNA monitoring after surgery was not performed. Lastly, in this trial, we had used multiple treatment modalities, making it challenging to pinpoint which method played a crucial role (CRT or iCT) and determine the optimal treatment sequence. Currently, a three-arm trial comparing the efficacy of the sequence of three combinations of treatment modalities (CRT followed by iCT and surgery vs. iCT followed by CRT and surgery vs. CRT followed by surgery) for unresectable ESCC is ongoing.

In summary, the primary endpoint of this trial has been met, indicating the potential efficacy and acceptable safety profile of conversion CRT and subsequent iCT followed by surgery for unresectable locally advanced ESCC. Based on the findings from our trial, further studies with larger sample sizes are warranted to evaluate the efficacy and safety of this combination regimen in the future.

X. Zhao reports personal fees from Nanjing Geneseeq Technology Inc., during the conduct of the study and outside the submitted work. Q. Ou reports other support from Geneseeq Technology Inc. during the conduct of the study. No disclosures were reported by the other authors.

X. Wang: Resources, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. X. Kang: Resources, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. R. Zhang: Data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. L. Xue: Data curation, formal analysis, investigation, writing–review and editing. J. Xu: Data curation, formal analysis, investigation, writing–review and editing. X. Zhao: Software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. Q. Ou: Formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. N. Yu: Data curation, formal analysis, investigation, writing–review and editing. G. Feng: Data curation, formal analysis, investigation, writing–review and editing. J. Li: Data curation, formal analysis, investigation, writing–review and editing. Z. Zheng: Data curation, formal analysis, investigation, writing–review and editing. X. Chen: Data curation, formal analysis, investigation, writing–review and editing. Z. Wang: Data curation, formal analysis, investigation, writing–review and editing. Q. Zheng: Data curation, formal analysis, investigation, writing–review and editing. Y. Li: Data curation, formal analysis, investigation, writing–review and editing. J. Qin: Data curation, formal analysis, investigation, writing–review and editing. N. Bi: Conceptualization, resources, data curation, supervision, funding acquisition, methodology, project administration, writing–review and editing. Y. Li: Conceptualization, resources, data curation, supervision, funding acquisition, methodology, project administration, writing–review and editing.

The authors thank all the patients who participated in this study. This study was supported by Beijing Xisike Clinical Oncology Research Foundation (Y-Young2023-0114, Y-MSDPU2022-0309, Y-Young2021-0145, Y-MSDPU2022-0822), and Special Program for Basic Resource Survey of the Ministry of Science and Technology (2019FY101101).