Preoperative conversion therapy of camrelizumab combined with paclitaxel and nedaplatin for unresectable advanced esophageal squamous cell carcinoma: a phase 2 study Authors: Jinhong Zhu, Jinfeng Zhang, Xiaodong Ling, Changhong Wang, Yingnan Yang, Yangzhong Xin, Hao Jiang, Hao Liang, Luquan Zhang, Xionghai Qin, Jingle Lei, Chengyuan Fang, Xinxin Yang, Yuning Wang, Wei Meng, Xiaoyuan Wang, Hongxue Meng, Jianqun Ma Published: 2025 ABSTRACT AbstractBackgroundPatients with unresectable locally advanced esophageal squamous cell carcinoma (ESCC) face limited treatment options and poor outcomes. Preclinical evidence supports chemoimmunotherapy as a conversion therapy; therefore, this study aims to evaluate the efficacy and safety of preoperative camrelizumab with paclitaxel and nedaplatin.MethodsPatients received a combination of camrelizumab (200 mg), paclitaxel (155 mg/m2), and nedaplatin (80 mg/m2) every 3 weeks. An efficacy assessment was performed every two cycles using the RECIST v1.1 criteria, and patients in the resected group underwent surgery 4–6 weeks after the last dose. The primary endpoint was 1-year overall survival (OS) rate, while secondary endpoints included surgical conversion rates, pathological response rates, objective response rate (ORR), disease-free survival (DFS), and safety.ResultsThis trial enrolled 141 patients with unresectable ESCC (132 locally advanced [M0], 9 metastatic [M1]) with a median follow-up of 32.2 months. Post-induction ORR was 56.7%. The intention-to-treat (ITT) conversion rate reached 48.9% overall and 50.0% (66/132) in M0 patients (R0: 100%; pathological complete response (pCR): 20.9%; major pathological response: 55.2%). One-year OS was 78.7%. Surgical conversion significantly improved survival overall (median OS: not reached (NR) vs. 14.1 months; hazard ratio (HR) = 0.22; 95% CI, 0.13–0.37; P < 0.0001), with further benefit in objective responders (median OS: NR vs. 20.9 months; HR = 0.30; 95% CI, 0.14–0.61; P < 0.001). Approximately 75.9% of patients experienced treatment-related adverse events (TRAEs), predominantly mild to moderate. Pan-lactylation levels were significantly elevated in non-pCR ESCC compared to pCR tissues. A peritumoral radiomic features-derived model effectively stratified patients into distinct groups with significantly different OS and DFS.ConclusionsThe chemoimmunotherapy regimen shows promising efficacy as a conversion therapy in improving surgical resectability and survival outcomes with manageable TRAEs, and does not delay surgery.Trial registration numberChiCTR2100046355.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12916-025-04592-z.Keywords: Camrelizumab, Chemotherapy, Surgery conversion, Unresectable ESCC Background Globally, esophageal cancer incidence rank fell from 7th (2020) to 11th (2022) while maintaining 7th in cancer-related mortality [1]. It features high lethality, late-stage diagnosis, and geographic heterogeneity, histologically composed of two main types: adenocarcinoma and squamous cell carcinoma (ESCC), with ESCC exhibiting Asian predominance (> 90% in China) [1]. Surgery remains the curative cornerstone, yet ~ 50% ESCC is unresectable at diagnosis due to late presentation and no reliable early diagnostic markers [2]. Unresectable locally advanced (i.e., adjacent organ invasion) or metastatic diseases require definitive radiotherapy (dRT) or chemoradiotherapy (dCRT) [3], but face high toxicity, recurrence, and suboptimal 5-year survival (19.2–35.8%) [4–8]. Given therapeutic limitations, conversion surgery after induction therapy emerges for initially unresectable ESCC [9, 10]. This strategy may substantially improve survival, particularly in patients with anatomical constraints (tracheobronchial/vascular invasion) rather than distant metastases. Accumulating evidence has confirmed the benefits of conversion therapy in initially unresectable locally advanced ESCC [2, 9–15]. COSMOS phase 2 trial demonstrated that induction chemotherapy (ICT) with docetaxel plus 5-fluorouracil and cisplatin (DCF) achieved a 39.6% R0 rate (n = 48), conferring a 3-year survival advantage to R0 patients (71.4% vs. 30.1%) [11, 16]. Another multicenter randomized phase 2 trial (n = 99) indicated that R0 resection post-induction with chemoradiotherapy (CRT) and triplet chemotherapy (CT) correlates with survival advantage for T4b esophageal cancer (EC) [13, 17]. In the NEXUS-1 trial (n = 30) for unresectable locally advanced ESCC, an intensive sequential regimen of chemoradiation followed by immunochemotherapy yielded a 66.7% conversion rate. This approach resulted in a high R0 resection rate (95.2%) and improved survival outcomes [2]. Beyond CRT/CT, immunotherapy combined with chemotherapy in the neoadjuvant or first-line setting has revolutionized the treatment paradigm for locally advanced resectable/non-resectable or metastatic ESCC [18–25], suggesting its potential as an induction therapy for conversion surgery. First, neoadjuvant PD-1 blockade in combination with chemotherapy safely enhances pathological complete response (pCR), major pathological response (MPR), and survival without increasing preoperative adverse reactions, surgical complications, and operation complexity [18–22]. This efficacy and safety profile promotes tumor regression and downstaging for conversion resection in initially unresectable or borderline resectable ESCC. Second, the combination of PD-1 inhibitors with chemotherapy has become the standard treatment for advanced ESCC in recent years [23–26]. ESCORT-1st showed sustained efficacy for camrelizumab-chemotherapy over placebo-chemotherapy, with no cumulative toxicities, justifying its adoption as a standard therapy regimen for untreated advanced ESCC [23]. The CheckMate 648 Trial concluded that first-line treatment with nivolumab plus chemotherapy and nivolumab plus ipilimumab both significantly prolonged overall survival compared to chemotherapy alone in advanced ESCC [25]. However, preoperative chemoimmunotherapy for unresectable locally advanced or advanced ESCC lacks robust evidence [27–30], with only one prospective trial (n = 47) [30]. An urgent investigation is needed to determine if immune checkpoint inhibitors (ICIs) combined with chemotherapy facilitate sufficient downstaging for curative resection in initially unresectable ESCC. Camrelizumab, a checkpoint inhibitor targeting PD-1, has been approved as the second-line therapy [31], and its combination with chemotherapy was approved as the first-line therapy for patients with advanced or metastatic ESCC by the China National Medical Products Administration [26]. This trial aimed to evaluate preoperative camrelizumab combined with paclitaxel and nedaplatin as a conversion therapy in patients with unresectable ESCC and determine the safety profile, conversion efficacy, and survival benefits of this regimen, while exploring biomarkers of therapeutic response. Methods Study design This was a single-arm, prospective phase II clinical trial (ChiCTR2100046355) assessing the efficacy and safety of camrelizumab-chemotherapy as conversion therapy to qualify patients for esophagectomy in ESCC. Recruitment started from May 2020 to April 2025. Participants Patients with histologically proven, locally advanced/advanced ESCC (stage: T2-4, N0-3, M0/M1) were enrolled if they were deemed unresectable by our institutional multidisciplinary team (MDT). Unresectability was defined as follows: (1) clinical T4b disease: radiographic evidence (via contrast-enhanced CT and/or endoscopic ultrasound) of tumor invasion into adjacent vital structures, such as the aorta, trachea or main bronchi, or vertebral fascia; (2) unresectable lymph node disease: metastatic lymph nodes that were bulky and fixed, precluding a safe R0 resection. Eligible participants should also be aged 18 to 70 years, have had at least one measurable lesion as defined by the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1, and have not received prior systemic therapy. Additional eligibility requirements included Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–1, expected good survival, adequate organ and bone marrow function, and voluntary participation with informed consent. Exclusion criteria included the presence of other malignancies, active tuberculosis or autoimmune disease, a history of interstitial lung disease or pneumonitis, abuse of psychotropic substances or drugs, HIV-positive status, recent vaccination with live vaccines, and any other conditions deemed unsuitable for clinical research participation. This study was performed at Harbin Medical University Cancer Hospital in China. All procedures were conducted in accordance with the Declaration of Helsinki. Reporting of this trial follows the CONSORT guidelines (see checklist, Additional file 1: CONSORT). Procedures On day 1 (d1) of each 3-week cycle, all patients received the following induction therapy regimen, consisting of camrelizumab (200 mg, iv, d1), paclitaxel (155 mg/m2, iv, d1), and nedaplatin (80 mg/m2, iv, d1). After every 2 cycles (6 weeks), an efficacy evaluation will be conducted according to the RECIST 1.1 criteria to determine eligibility for surgical resection. Patients who achieved operability were scheduled for esophagectomy within 4–8 weeks following the completion of therapy. Discontinuations, interruptions, and modifications to the chemotherapy regimen were permitted based on patient safety and tolerance, adhering to prespecified recovery criteria outlined in the protocol. Camrelizumab was suspended when patients experienced serious adverse reactions, but the dose reductions or increases were not permitted. Endpoints and assessment The primary endpoint was the 1-year overall survival (OS) rate. The secondary endpoints were defined as follows: the surgery conversion rate refers to the proportion of patients who underwent surgical resection among the total study population. Pathological complete response (pCR) is defined as the absence of residual viable tumor cells in the tumor bed of the postoperative tissue specimen. Major pathological response (MPR) indicates that the proportion of residual viable tumor cells in the tumor bed of the postoperative tissue specimen is ≤ 10%. R0 resection is confirmed by the absence of residual tumor at the surgical margin under microscopic examination. The objective response rate (ORR) is calculated as the sum of the complete response (CR) and the partial response (PR) rates based on RECIST V.1.1 criteria. The overall survival (OS) was calculated as the time from the start of treatment to the date of death or the last follow-up. Disease-free survival (DFS) was calculated from the date of surgical resection to the date of disease recurrence or death. Adverse events were graded and recorded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE) version 5.0, and the postoperative complications were classified using the Clavien-Dindo classification system. Immunohistochemistry (IHC) Formalin-fixed paraffin-embedded (FFPE) ESCC biopsy specimens underwent PD-L1 immunohistochemical staining using the FDA-approved 22C3 pharmDx assay (Dako North American). The combined positive score (CPS) was calculated as the number of PD-L1-positive cells (tumor cells, lymphocytes, macrophages) per 100 viable tumor cells, expressed as CPS = (PD-L1⁺ cells / viable tumor cells) × 100. Two blinded independent pathologists scored all slides, with third-pathologist adjudication for discordant cases (> 10% score variance). ESCC tumors were considered PD-L1 positive if CPS was ≥ 1; scores < 1 indicated PD-L1 negativity. Protein lactylation levels in ESCC tissues were evaluated using a rabbit monoclonal anti-pan-lactylation antibody (Catalog # PTM-1401RM, PTM Biolabs, Hangzhou, China). FFPE sections underwent standard IHC processing with positive and negative controls being systematically included. Two blinded pathologists independently assessed tumor regions using a semi-quantitative scoring system: staining intensity was graded as 0 (negative), 1 (weak positive; faint yellow), 2 (positive; brown-yellow), or 3 (strong positive; deep brown), while positive cell percentage was scored as 1 (0–25%), 2 (26–50%), 3 (51–75%), or 4 (76–100%). The final IHC score was calculated by multiplying the intensity score by the percentage score (range: 0–12). Intergroup differences in pan-Kla expression levels were analyzed using the nonparametric Mann–Whitney U test. Survival prediction with peritumoral radiomic features We previously developed a peritumoral radiomics (Rad_per) model to predict the pathological response in ESCC patients treated with camrelizumab plus paclitaxel/nedaplatin [32]. This model evaluates the possibility of MPR for each patient, assigning a score ranging from 0 to 1, with higher scores indicating a greater probability of achieving MPR. The comprehensive methods are described in our previous publication [32]. Here, we assessed the prognostic value of this model for survival in patients with ESCC. Statistical considerations For categorical variables, the number and percentage of patients in each category were calculated. The 95% confidence interval (CI) was computed using the Clopper-Pearson method. For quantitative (continuous) data, values are presented as mean ± standard deviation (SD) if normally distributed, or as median (interquartile range, IQR) if non-normally distributed. The Kaplan–Meier product limit method is used to analyze DFS and OS, with corresponding 95% confidence intervals computed for each estimate. Fisher’s exact test or the chi-square test was utilized to assess the composition ratios of qualitative data. A Cox proportional hazards regression analysis was conducted to assess the impact of various baseline clinical characteristics on patient outcomes. The proportional hazards assumption for all Cox regression models was verified using the cox.zph() function in R software (version 4.4.0) with the survival package. A P value > 0.05 indicated no significant violations detected. All other statistical analyses were conducted using SPSS version 19.0 and GraphPad Prism version 8.0 without further specifications. A two-tailed test was applied for hypothesis testing, with a significance threshold set at P < 0.05. Safety data were reported in terms of occurrence and incidence rates. Results Patient characteristics and disposition Between May 13, 2020, and April 8, 2025, a total of 141 patients with locally advanced or advanced ESCC were enrolled in the study and received at least two cycles of combined immune-chemotherapy (Fig. 1). Of the enrolled patients, four protocol-deviating patients were excluded from subsequent analysis: one received radiotherapy, two opted for traditional Chinese medicine, and one underwent targeted chemotherapy. An additional 3 patients were removed from the analysis due to prolonged drug discontinuation. Among the remaining 134 participants, 69 patients achieved surgical conversion. Study flowchart of the trial As summarized in Table 1, the average age of patients was 58.5 years, with 139 patients (98.6%) being male. The ECOG PS was 0 in 98 (69.5%) patients and 1 in 43 (30.5%). Additionally, more than half of the patients reported a history of smoking and alcohol consumption. Middle esophageal tumors predominated (n = 63, 44.7%), followed by lower segments (n = 56, 39.7%) and upper segments (n = 22, 15.6%). Most patients presented with T3 stage disease and lymph node metastasis, while 9 (6.4%) patients had lymph node or organ metastasis. The median tumor length was 70 mm. Table 1Baseline characteristics of all eligible patientsCharacteristicsAll patients (n = 141)Age (years), mean ± SD58.5 ± 6.8Gender Male139 (98.6%) Female2 (1.4%)ECOG performance status 098 (69.5%) 143 (30.5%)Smoking Never59 (41.8%) Former or current78 (55.3%) Unknown4 (2.8%)Drinking Never47 (33.3%) Former or current90 (63.8%) Unknown4 (2.8%)Tumor location Upper22 (15.6%) Middle63 (44.7%) Lower56 (39.7%)Clinical T stage T212 (8.5%) T398 (69.5%) T4a27 (19.1%) T4b4 (2.8%)Clinical N stage N040 (28.4%) N171 (50.4%) N226 (18.4%) N34 (2.8%)Clinical M stage M0132 (93.6%) M + 9 (6.4%)Clinical stage (AJCC, 8th edition) II26 (18.4%) III73 (51.8%) IVA33 (23.4%) IVB9 (6.4%) Tumor length (mm), median (IQR)70 (40.0)PD-L1 CPS expression CPS < 135 (24.8%) CPS ≥ 17 (5.0%)Data are presented as n (%) or median (range). ECOG, Eastern Cooperative Oncology Group Baseline characteristics of all eligible patients Data are presented as n (%) or median (range). ECOG, Eastern Cooperative Oncology Group Treatment exposure A total of 141 patients received at least 2 doses of camrelizumab in combination with chemotherapy. Specifically, 58 patients completed 2 cycles of chemoimmunotherapy, 21 patients completed 3 cycles, 48 patients completed 4 cycles, and 14 patients received more than 5 cycles of treatment. The median number of treatment cycles was 3, with one patient receiving immunotherapy for up to 2 years. Among those who underwent surgical resection, the distribution of treatment cycles was as follows: 36 patients received 2 cycles, 9 received 3 cycles, 20 received 4 cycles, and 4 received more than 5 cycles of treatment. Radiographic response According to RECIST v1.1, two (1.5%) of the 134 patients achieved complete response (CR), 74 (55.2%) achieved partial response (PR), and 57 (42.5%) had stable disease (SD). The ORR was 56.7% (95% CI: 47.9–65.2%), and the DCR was 99.3% (95% CI: 95.9–100.0%). More treatment cycles correlated with better radiographic response (P < 0.05) and resulted in a higher ORR (Table 2). A waterfall plot illustrated the best radiographic responses of all ESCC patients (Fig. 2A). Of the 69 patients who underwent surgery, 42 exhibited PR, and 27 had SD. Table 2Radiographic responseThe second cycle (n = 54)The third cycle (n = 20)The fourth cycle (n = 46)The fifth and more cycle (n = 14)P valueTotal (n = 134)Best overall response Complete response001 (2.2%)1 (7.1%)/2 (1.5%) Partial response24 (44.4%)10 (50.0%)29 (63.0%)11 (78.6%)/74 (55.2%) Stable disease29 (53.7%)10 (50.0%)16 (34.8%)2 (14.3%)/57 (42.5%) Progressive disease1 (1.9%)00/1 (0.7%) Objective response24 (44.4%)10 (50.0%)30 (65.2%)12 (85.7%)0.0276 (56.7%) Disease control53 (98.1%)20 (100.0%)46 (100%)14 (100%)0.609133 (99.3%)Data presented as n (%). Final analysis cohort comprised 134 patients who completed radiographic assessment after exclusion of 7 participants (4 protocol violations, 3 medication discontinuations). Initial screening identified 141 potentially eligible candidatesP values < 0.05 are considered statistically significant Radiographic response Data presented as n (%). Final analysis cohort comprised 134 patients who completed radiographic assessment after exclusion of 7 participants (4 protocol violations, 3 medication discontinuations). Initial screening identified 141 potentially eligible candidates P values < 0.05 are considered statistically significant Waterfall plots of tumor response assessment. A Best radiographic response by RECIST 1.1 in all evaluable patients (n = 134). B Pathological tumor regression in surgical patients (n = 67). pCR, pathological complete response; MPR, major pathological response; PR, partial response; SD, stable disease; PD, progressive disease Surgical and pathological outcomes The surgical outcomes are summarized in Additional file 2: Table S1. Of the 69 patients who underwent surgery, complete surgical and pathological records were available for 67; the remaining 2 patients had no records because they underwent surgery in other hospitals. All 67 patients achieved R0 resection. The surgical conversion rate is 48.9% (69/141). Surgical methods were mainly robotic surgery (37.3%) and laparoscopic surgery (62.7%). No treatment-related surgical delays were recorded, and the median interval between the last administration of conversion therapy and surgery was 39.0 days (IQR, 9.0 days). Complications were graded by the Clavien-Dindo classification system (Additional file 2: Table S2). Affected patients received appropriate management. No complications required reoperation. There was no 30-day or 90-day mortality among the patients with complete follow-up. One patient was lost to follow-up after 30 days, but was confirmed to be alive at that time. In the non-converted cohort (n = 72), 7 were excluded from efficacy analysis due to protocol violations and discontinued medication. Among 65 efficacy-assessed patients, non-conversion was due to either completed therapy with insufficient downstaging (n = 20), patient refusal (n = 16), progression (n = 7), or therapy discontinuation because of adverse events, financial constraints, or other factors (n = 22). Table 3 summarizes the baseline characteristics of patients undergoing surgical resection (n = 69) versus non-surgical management (n = 72). No statistically significant differences were observed in age or gender distribution. Surgical candidates demonstrated significantly superior ECOG performance status. Non-smokers or non-drinkers were more likely to undergo surgery. Intriguingly, surgical patients had more advanced disease. This paradox may stem from the preference of earlier-stage patients for surveillance over surgery after achieving an OR. The pathologic profile of the 67 patients after tumor resection is summarized in Table 4. Fourteen (20.9%, 95% CI, 11.9 to 32.6%) patients achieved pCR (ypT0N0); 37 (55.2%, 95% CI, 42.6 to 67.4%) patients achieved MPR. In addition, in terms of tumor regression grade (TRG), 17 patients reached grade 0 (ypT0), and 10, 20, and 20 patients reached TRG grade 1, 2, and 3, respectively. Detailed tumor regression results are shown in a waterfall plot (Fig. 2B), including the cycle of neoadjuvant therapy, radiographic response, clinical stage (cT and cN), and postoperative pathological stage (ypT and ypN). Fifty patients (74.6%) achieved T downstaging, and 20 (29.9%) achieved N downstaging, with 17 (25.4%) showing pathological T0 and 34 (50.7%) reaching pathological N0. Table 3Baseline characteristics of non-surgery and surgery patientsCharacteristicsAll patients (n = 141)Non-surgery (n = 72)Surgery (n = 69)P valueAge (years), mean ± SD58.5 ± 6.858.9 ± 7.358.0 ± 6.30.406Gender0.238 Male139 (98.6%)72 (100%)67 (97.1%) Female2 (1.4%)02 (2.9%)ECOG performance status0.003 098 (69.5%)42 (58.3%)56 (81.2%) 143 (30.5%)30 (41.7%)13 (18.8%)Smoking0.012 Never59 (41.8%)22 (30.6%)37 (53.6%) Former or current78 (55.3%)46 (63.9%)32 (46.4%) Unknown4 (2.8%)4 (5.6%)Drinking0.008 Never47 (33.3%)16 (22.2%)31 (44.9%) Former or current90 (63.8%)52 (72.2%)38 (55.1%) Unknown4 (2.8%)4 (5.6%)Tumor location0.126 Upper22 (15.6%)14 (19.4%)8 (11.6%) Middle63 (44.7%)35 (48.6%)28 (40.6%) Lower56 (39.7%)23 (31.9%)33 (47.8%)Clinical T stage0.000 T212 (8.5%)5 (6.9%)7 (10.1%) T398 (69.5%)62 (86.1)36 (52.2%) T4a27 (19.1%)5 (6.9%)22 (31.9%) T4b4 (2.8%)0 (0)4 (5.8%)Clinical N stage0.224 N040 (28.4%)16 (22.2%)24 (34.8%) N171 (50.4%)36 (50%)35 (50.7%) N226 (18.4%)16 (22.2%)10 (14.5%) N34 (2.8%)4 (5.6%)0Clinical M stage0.533 M0132 (93.6%)66 (91.7%)66 (95.7%) M + 9 (6.4%)6 (8.3%)3 (4.3%)Clinical stage (AJCC, 8th edition)0.000 II26 (18.4%)12 (16.7%)14 (20.3%) III74 (52.5%)48 (66.7%)26 (37.7%) IVA33 (23.4%)7 (9.7%)26 (37.7%) IVB9 (6.4%%)6 (8.3%)3 (4.3%)Tumor length (mm), median (IQR)70 (40.0)70.0 (44.3)70.0 (30.0)0.862‡Continuous variables: Normally distributed data expressed as mean ± standard deviation (independent samples t-test); ‡Non-normally distributed data expressed as median [interquartile range] (Wilcoxon rank-sum test). Categorical variables expressed as numbers (percentages) (χ2 test or Fisher’s exact test as appropriate). ECOG, Eastern Cooperative Oncology Group; IQR, interquartile range (25th–75th percentile)P values < 0.05 are considered statistically significant Baseline characteristics of non-surgery and surgery patients Continuous variables: Normally distributed data expressed as mean ± standard deviation (independent samples t-test); ‡Non-normally distributed data expressed as median [interquartile range] (Wilcoxon rank-sum test). Categorical variables expressed as numbers (percentages) (χ2 test or Fisher’s exact test as appropriate). ECOG, Eastern Cooperative Oncology Group; IQR, interquartile range (25th–75th percentile) P values < 0.05 are considered statistically significant Table 4Postoperative pathological response (n = 67)CharacteristicsPatientsPathological response pCR14 (20.9%) MPR37 (55.2%)Tumor regression grade TRG 017 (25.4%) TRG 110 (14.9%) TRG 220 (29.9%) TRG 320 (29.9%)Pathological T stage pT017 (25.4%) pTis4 (6.0%) pT113 (19.4%) pT214 (20.9%) pT319 (28.4%)Pathological N stage pN034 (50.7%) pN121 (31.3%) pN211 (16.4%) pN31 (1.5%)pTNM I23 (34.3%) II14 (20.9%) IIIA14 (20.9%) IIIB15 (22.4%) IVA1 (1.5%)Downstaging of the T stage Yes50 (74.6%) No17 (25.4%)Downstaging of the N stage Yes20 (29.9%) No47 (70.1%)Downstaging of the TNM stage Yes44 (65.7%) No23 (34.3%)Data presented as n (%). pCR pathological complete response, MPR major pathological response, TRG tumor regression grade, assessed by the four-tier College of American Pathologists grading system. Pathological response assessment included 67 patients who underwent institutional procedures. Two patients receiving surgery at external facilities were excluded from analysis due to incomplete data documentation Postoperative pathological response (n = 67) Data presented as n (%). pCR pathological complete response, MPR major pathological response, TRG tumor regression grade, assessed by the four-tier College of American Pathologists grading system. Pathological response assessment included 67 patients who underwent institutional procedures. Two patients receiving surgery at external facilities were excluded from analysis due to incomplete data documentation Association of PD-L1 CPS and protein pan-lactylation with pathological response Of 67 surgically treated patients with pathological evaluation, 42 had paired pre-treatment tissues available. For all patients, available preoperative (baseline) and postoperative ESCC tissues were subjected to IHC staining for PD-L1 expression and pan-lactylation (pan-Kla) levels. In pre-treatment samples, patients with PD-L1 CPS < 1 (n = 33) showed a lower MPR rate of 48.48% compared to 71.43% in patients with PD-L1 CPS ≥ 1 (n = 7). The pCR rate was lower in PD-L1 negative patients (CPS < 1; 17.14%, n = 35) versus PD-L1 positive patients (CPS ≥ 1; 42.86%, n = 7). Post-treatment, PD-L1 CPS distribution was similar (CPS < 1: n = 50; CPS ≥ 1: n = 14), with MPR rate of 62.0% vs. 35.71% and pCR rate of 24% vs. 14.29% (Fig. 3A). However, there was no significant association between CPS and clinical outcomes. Moreover, the pre-treatment pan-lactylation level (pan-Kla score) showed no significant correlation with either MPR or pCR. In contrast, analysis of post-treatment resection specimens revealed that the pan-Kla score was significantly elevated in tumor tissues from non-pCR patients compared to tissues from treated tumor beds of patients achieving pCR (Fig. 3B). Assessment of the association of PD-L1 combined positive score (CPS) and pan-lactylation (pan-Kla) score with MPR and pCR in pre-treatment biopsies and post-treatment surgical samples. Immunohistochemistry staining was performed to detect PD-L1 and pan-Kla expression. A No significant associations were observed between PD-L1 CPS and pathological response (MPR or pCR) in either pre-treatment biopsies or post-treatment samples (all P > 0.05). Statistics were performed using the chi-square test. B Evaluation of the relationship between pan-Kla levels and MPR/pCR. Pan-Kla levels showed no significant association with MPR in pre-treatment or post-treatment samples. Non-pCR patients exhibited significantly higher pan-Kla levels than pCR patients (P = 0.034). Data are shown as median ± IQR (Mann–Whitney U test). C Representative contrast-enhanced CT images of two pCR (C1, C2) and two non-pCR patients (C3, C4) before and after conversion chemoimmunotherapy. Red ovals highlight target tumor regions, showing remarkable regression in pCR cases versus persistent lesions in non-pCR cases. Postoperative pan-lactylation immunohistochemical staining of resected tumor or tumor bed specimens from corresponding patients. Enhanced pan-Kla (brown signals) is observed in tissues of non-pCR patients. Scale bars: 100 pixels Patient follow-up and survival analysis With a median follow-up time of 32.2 months (95% CI: 21.7–42.6 months), the intention-to-treat (ITT) population demonstrated a median overall survival (OS) of 38.5 months (95% CI: 20.8–56.2 months). The estimated OS rates were 78.7% at 1 year, 57.7% at 2 years, 50.3% at 3 years, and 42.1% at 4 years (Fig. 4A). Compared to non-surgical patients, those undergoing surgical resection achieved superior outcomes with median OS not reached (NR), demonstrating 1-year, 2-year, 3-year, and 4-year OS rates of 95.3%, 81.3%, 71.7%, and 62.3%, respectively. In contrast, non-surgical patients exhibited significantly reduced median OS of 14.1 months (95% CI: 11.3–16.8 months), with corresponding survival rates of 60.7%, 30.8%, 26.4%, and 20.6% (P < 0.0001; Fig. 4B). Surgical patients maintained durable disease control with a median DFS not reached, and DFS rates were 83.1% at 1 year, 63.6% at 2 years, and 54.0% at 3 years and 4 years (Fig. 4C). Mortality analysis revealed disease progression as the predominant cause, with four non-cancer-related deaths (myocardial infarction, esophageal fistula, COVID-19, and postoperative bleeding from brain surgery). Kaplan–Meier survival analysis comparing clinical outcomes. A Overall survival (OS) in the entire cohort (n = 141) with a median follow-up of 32.2 months. B Stratified OS analysis: surgical vs. non-surgical patients (n = 69/72; log-rank P < 0.0001). C Disease-free survival (DFS) in the surgical subgroup (n = 69) Moreover, 76 imaging responders (CR + PR) trended toward longer median OS (mOS: 46.6 vs. 22.0 months in 58 non-responders; HR = 0.61; 95% CI, 0.35–1.06, P = 0.0578; Fig. 5A). Importantly, among the OR subgroup, patients undergoing surgery showed superior OS versus non-operated responders (mOS: NR vs. 20.9 months; HR = 0.30, 95% CI, 0.14–0.61, P < 0.001; Fig. 5B). Pathological responders (pCR/MPR) trended toward improved OS/DFS versus non-responders, but did not reach statistical thresholds (Additional file 2: Fig. S1). Pathologically confirmed lymph node-positive (pN +) patients showed a trend toward lower OS (P = 0.075) and had significantly worse DFS (P < 0.05) compared to pN − patients (Fig. 5C and D). Next, based on the median Rad_per score of 0.427, the cohort undergoing surgery was dichotomized into high- and low-score groups (> 0.427 vs. ≤ 0.427). Survival analysis revealed that those with high scores demonstrated significantly superior OS and DFS compared to patients with low scores (log-rank test, P < 0.05) (Fig. 5E and F). Combined analysis of survival outcomes by treatment response, surgical intervention, and pathological lymph node status. A Overall survival (OS) analysis by objective response (OR) status. Responders (n = 76) showed a trend toward improved OS versus non-responders (n = 58) (log-rank P = 0.0578). B OS in responders stratified by surgical approach. Among responders, patients undergoing R0 resection (n = 42) had significantly prolonged OS compared to non-surgical patients (n = 34) (log-rank P = 0.0005). C OS by pathological lymph node (pN) status. Patients with node-negative disease (pN − , n = 34) had superior OS versus pN + patients (n = 33). D Disease-free survival (DFS) in pN + vs. pN − patients. Survival curves demonstrate the prognostic value of the Rad_per model for E OS and F DFS. Patients with a score above the median (0.427) had significantly better outcomes than those with a low score Multivariate COX regression analyses incorporating key baseline characteristics (age, ECOG performance status, smoking history, alcohol consumption, tumor location, clinical T stage, N stage, M stage, TNM stage, and tumor length) revealed that an ECOG score of 0 was associated with better prognosis compared to 1. No significant association was found between other characteristics and OS (Fig. 6). Forest plot of multivariate Cox proportional hazards analysis for OS. Covariates included clinically relevant baseline factors as indicated (n = 141). In multivariate Cox regression analysis, an ECOG performance status of 1 was independently associated with significantly worse overall survival compared to a status of 0. No other baseline characteristics showed a significant association with OS Safety One hundred and seven patients (75.9%, 107/141) experienced treatment-related adverse events (TRAEs) (Table 5). TRAEs classified as grade ≥ 3 were reported in only 2 patients (1.4%), with the predominant events being elevated levels of aspartate aminotransferase and alanine aminotransferase, each occurring in 1.4% of patients. Reactive cutaneous capillary endothelial proliferation was observed in 44 patients (31.2%), with the majority presenting at grade 1–2. Other common grade 1–2 adverse events included alopecia (22.7%), pruritus (16.3%), rash (12.1%), anemia (11.3%), and myalgia (9.9%). None of the adverse events required dose adjustments, treatment discontinuations, or resulted in fatalities. Table 5Treatment-related adverse events (TRAEs)TRAEsAll patients (n = 141)Any gradeGrade 1–2Grade 3 or higherAny adverse events107 (75.9%)105 (74.5%)2 (1.4%)Reactive cutaneous capillary endothelial proliferation44 (31.2%)44 (31.2%)0Alopecia32 (22.7%)15 (22.7%)0Pruritus23 (16.3%)23 (16.3%)0Rash17 (12.1%)17 (12.1%)0Anemia16 (11.3%)16 (11.3%)0Myalgia14 (9.9%)14 (9.9%)0Arthritis11 (7.8%)11 (7.8%)0Fatigue10 (7.1%)10 (7.1%)0Leukopenia9 (6.4%)9 (6.4%)0Hyperbilirubinemia9 (6.4%)9 (6.4%)0Hypothyroidism9 (6.4%)9 (6.4%)0Aspartate aminotransferase increased8 (5.7%)6 (4.3%)2 (1.4%)Nausea8 (5.7%)8 (5.7%)0Neutropenia7 (5.0%)7 (5.0%)0Lymphopenia7 (5.0%)7 (5.0%)0Thrombocytopenia7 (5.0%)7 (5.0%)0Alanine aminotransferase increased7 (5.0%)5 (3.5%)2 (1.4%)Sensory neuropathy7 (5.0%)7 (5.0%)0Anorexia4 (2.8%)4 (2.8%)0Pneumonia4 (2.8%)4 (2.8%)0Increased γ-GT3 (2.1%)3 (2.1%)0Vomiting3 (2.1%)3 (2.1%)0Hyperthyroidism3 (2.1%)3 (2.1%)0Fever1 (0.7%)1 (0.7%)0Gum bleeding1 (0.7%)1 (0.7%)0Oral mucositis1 (0.7%)1 (0.7%)0Diabetes1 (0.7%)1 (0.7%)0Headache1 (0.7%)1 (0.7%)0Blurred vision1 (0.7%)1 (0.7%)0Edema1 (0.7%)1 (0.7%)Proteinuria1 (0.7%)1 (0.7%)0Data are expressed as n (%). TRAEs were assessed in all patients completing ≥ 2 cycles of preoperative chemoimmunotherapy (n = 141) Treatment-related adverse events (TRAEs) Data are expressed as n (%). TRAEs were assessed in all patients completing ≥ 2 cycles of preoperative chemoimmunotherapy (n = 141) Discussion This single-arm, open-label, phase II trial evaluated the efficacy and safety of camrelizumab in combination with paclitaxel and nedaplatin as a conversion therapy in unresectable ESCC (n = 141). This combined therapy demonstrated substantial efficacy, with an ORR of 56.7% and a DCR of 99.3%. Overall, 48.9% (69/141) of patients achieved resectability by ITT analysis. After excluding M1 patients (n = 9), the cohort achieved 50.0% surgical conversion (66/132). Despite immature mOS, surgery led to significantly superior survival (1-/2-/3-/4-year OS: 95.3%/81.3%/71.7%/62.3%). Surgery survival benefit persisted in patients achieving an objective response. With manageable toxicity, the current regimen demonstrates preoperative conversion potential through enhanced resection feasibility and survival. PD-1 inhibitors (e.g., nivolumab, pembrolizumab, camrelizumab) plus dual-agent chemotherapy have demonstrated promising efficacy in unresectable advanced/metastatic ESCC, primarily in metastatic/recurrent settings (ESCORT-1st and CheckMate 648 trials) [23, 25]. Our study extends this evidence to preoperative conversion therapy. Surgical OR patients had significantly longer OS than non-surgical OR patients. Furthermore, the survival benefit was evaluated among 76 objective responders to isolate the effect of surgery from the initial therapy response. The superior survival in the surgery group confirms an additive benefit beyond response to chemoimmunotherapy, robustly supporting R0 resection’s survival benefit in the initially inoperable advanced ESCC. Consistently, in the ESCORT-1st study, first-line chemoimmunotherapy achieved a 72.1% ORR in unresectable advanced ESCC but yielded only 25.6% 3-year OS [23]. This closely matches the 26.4% 3-year OS in our non-surgical cohort, indicating radical surgery’s critical contribution to survival benefit. Importantly, a recent retrospective study (n = 90) confirmed the feasibility of combined organ resection (trachea/bronchus: 75.6%; large vessels: 24.4%; both: 5.6%) in EC invading adjacent organs after induction chemotherapy or chemoradiotherapy [33]. Such extended resections may convert even more initially inoperable patients to operable status, potentially improving survival. Our findings are timely and relevant among the rapid advances in ESCC immunochemotherapy. Regarding neoadjuvant immunochemotherapy in resectable ESCC, the randomized phase III ESCORT-NEO study reported that the pCR of 28% and the MPR of 59.1%, significantly surpassing the 4.7% and 20.9% rates observed with neoadjuvant chemotherapy alone [20]. In our prior neoadjuvant study, the same regimen yielded a pCR of 27.4% and an MPR of 45.2% in resectable ESCC [18]. We now show that it can achieve a comparable efficacy (pCR: 20.9%; MPR: 55.2%) in the unresectable setting. This comparable pCR rate is highly significant, as it was achieved in patients with a much higher tumor burden and more advanced disease at baseline, supporting the potency of the conversion regimen. In non-converters, besides 37.5% of primary resistance, 22.2% (n = 16) of patients refused surgery despite eligibility. The remaining patients were lost to real-world challenges. These findings indicate that robust supportive care and patient education are essential for improving outcomes. Unlike neoadjuvant studies, clinical trials focusing on induction therapy and conversion surgery are minimal. The NEXUS-1 trial (n = 30) reported an exceptional surgical conversion rate of 66.7% and a pCR of 65.0% (R0: 95.2%; median follow-up: 21 months) [2]. This remarkable outcome is likely attributable to its highly intensive, prolonged (15-week) sequential regimen of chemoradiation and chemoimmunotherapy. While impressive, this regimen’s intensity may pose challenges for broader clinical application, particularly in terms of patient compliance and tolerability. In contrast, our moderate chemoimmunotherapy regimen offers a more readily applicable and potentially safer alternative. The observed correlation between treatment cycles and response rates [34] suggests that our reported efficacy may be further improved by optimizing treatment duration, as over half of our cohort received a relatively brief course of therapy. Overall, 41.1% (58/141), 14.9% (21/141), and 34.04% (48/141) of patients completed 2, 3, and 4 cycles of treatment, respectively. Another prospective study with tislelizumab plus cisplatin/5-FU (n = 47) reported a conversion rate of 48.94% (pCR: 25.5%) (median follow-up: 18 months), which is similar to our results [30]. This consistency across studies strengthens the validity of our findings. Moreover, our study, with a larger cohort and a longer median follow-up of 32.2 months, provides more robust evidence regarding the feasibility and efficacy of this approach. Some retrospective studies support preoperative chemoimmunotherapy for unresectable ESCC [27–29]. The RICE-Retro real-world study (n = 155) reports high conversion rates of 74.8% with preoperative chemoimmunotherapy; however, the use of diverse PD-1 inhibitors and short follow-up limit the interpretability of these findings [28]. Another study demonstrated a higher conversion rate with preoperative immunochemotherapy (81.3%) than chemotherapy alone (66.7%) in cT4b ESCC (64 vs. 135 patients) [27]. However, such real-world evidence is inherently limited by selection bias, heterogeneous treatments, and short follow-up, precluding definitive conclusions. We also compared our trial with prospective conversion surgery using induction chemoradiotherapy (CRT) for initially unresectable ESCC. Previous chemotherapy-based conversion studies with 47 to 48 participants (e.g., using the DCF regimen) reported conversion rates of 37.5–57.4% and ORR of 31.3–65.9% [11, 14]. A multicenter phase II trial comparing induction chemoradiotherapy (CRT) versus DCF in cT4b esophageal cancer reported higher conversion rates with CRT [69% (34/49) vs. 50% (25/50)] [17]. Ultimately, 84% of patients in both groups underwent conversion surgery; however, the 4-year OS rate among those with R0 resection was only 40% [17]. In this context, our chemoimmunotherapy regimen represents a promising approach, achieving a 48.9% conversion rate and a 56.7% ORR, with a 4-year OS of 62.3% in surgically converted patients. Additionally, the regimen provided a more favorable safety profile by utilizing paclitaxel/nedaplatin, which is supported by established tolerability data in Asian populations [35–37]. Unfortunately, propensity score-matched analysis with a definitive CRT cohort was not feasible due to evolving standards of care and potential temporal biases. However, when viewed against these historical outcomes, our results affirm the potential of our strategy to markedly improve survival in a substantial patient population with otherwise poor prognoses. Lactate, a metabolic byproduct of glycolysis, is also a key signaling molecule through lactylation (Kla), a novel post-translational modification [38]. This metabolite-driven modification plays a critical role in cancer biology, profoundly influencing tumor initiation, progression, therapeutic resistance, and the immunosuppressive tumor microenvironment (TME) [39], with significant implications for ESCC [40, 41]. For instance, lactate confers cuproptosis resistance in ESCC via AARS1-mediated NUDT21 lactylation. Combined LDHA inhibition and copper ionophore treatment can overcome this resistance, suggesting lactylation’s role in treatment evasion [41]. Moreover, lactylation drives immunotherapy resistance by remodeling the immunosuppressive TME [42]. It promotes this resistance by inducing CD8 + T cell exhaustion, enhancing Treg suppressive function, and upregulating immune checkpoints like PD-L1, which collectively undermine anti-tumor immunity [43–47]. We observed significantly lower pan-lactylation levels in the treated tumor beds of patients with pCR compared to viable tumor regions from non-pCR patients. As pCR samples contain only stromal and immune cells without viable tumor, detected lactylation signals likely originate from non-tumor sources, including immune infiltrates, fibroblasts, dead tumor debris, or endothelial/stromal cells. We propose that low lactylation enables a therapeutic metabolic “reset” favoring anti-tumor immunity in pCR patients, whereas high lactylation in non-pCR tissues drives resistance by maintaining an immunosuppressive niche. Lactylation may be a promising modulator of chemoimmunotherapy efficacy. The observed toxicity profile was consistent with the well-established safety spectrum of camrelizumab-chemotherapy combinations [26, 48–51]. The current trial demonstrated a favorable safety profile, with only two grade 3 TRAEs (1.4%) observed, both of which were manageable. Immune-related AEs were primarily low-grade (e.g., rash, pneumonitis, hyperthyroidism). This represents a marked improvement over other key trials: (1) NEXUS-1 trial, where 40% of patients experienced grade ≥ 3 AEs following intensive sequential CRT and immunochemotherapy [2]; (2) a previous conversion study using tislelizumab/cisplatin/5-fluorouracil-based therapy (14.9% grade ≥ 3 AEs, n = 12) [30]; (3) the ESCORT-1st trial for advanced disease (63.4% grade ≥ 3 AEs) [26]. Similarly, our trial demonstrated a lower TRAE rate compared to previous neoadjuvant therapy regimens [50, 51]. The low toxicity profile may be attributable to the lower dose or frequency of chemotherapy administered in our cohort. In our study, patients received paclitaxel via intravenous infusion at a dose of 155 mg/m2 on day 1 of each treatment cycle. In contrast, other neoadjuvant ESCC studies involved nab-paclitaxel at doses of 100 mg/m2 on days 1, 8, and 15 of each cycle, or 260 mg/m2 on day 1 of each cycle [50, 51]. Furthermore, the overall incidence of surgical complications was 92.5%, which is consistent numerically with previous reports on neoadjuvant regimens for ESCC [18, 20, 50]. Pleural effusion was the most frequent complication, potentially associated with malnutrition, hypoproteinemia, pulmonary infection, or postoperative bleeding. All complications were mild, requiring no intervention, with no perioperative mortality or reoperations. While neoadjuvant/first-line chemoimmunotherapy has improved outcomes in ESCC, the efficacy is frequently limited by therapeutic resistance. The critical role of intratumoral heterogeneity in stromal cells and infiltrated immune cells in driving these divergent clinical responses remains largely unexplored. Jiang et al. recently identified two distinct CAF subpopulations that expand post-therapy and confer resistance in ESCC: (1) CD248+ mechanoresponsive CAFs, which form a physical barrier limiting CD8 + T cell infiltration and drug delivery; (2) IL6+CCL2+ immunomodulatory CAFs, which promote therapeutic resistance via IL-6 signaling [52]. Moreover, Alcindor et al.’s phase 2 trial of perioperative chemoimmunotherapy for gastro-esophageal adenocarcinoma (GEA) provides compelling parallel evidence for the role of innate immune populations in treatment response. Their study demonstrated that while no correlation was found between therapy response and PD-L1 or mismatch repair status, non-responding tumors showed significant proliferation of M2-tumor-associated macrophages [53]. These findings suggest that chemoimmunotherapy resistance likely involves interactions between multiple stromal components to establish an immunological barrier to effective treatment, including immunosuppressive CAFs, CD8+ T cells, and M2 macrophages. Future studies are essential to elucidate their relative contributions and potential synergistic effects in driving treatment resistance. This study has several limitations. First, its single-arm design precludes direct comparison with standard therapies. Second, the moderate cohort size (n = 141) may limit statistical power and generalizability, while the follow-up may be insufficient for long-term outcomes. Third, potential selection bias exists due to specific patient inclusion criteria. Fourth, 7 patients (4.9%) were excluded from resectability assessment; sensitivity analysis indicates the true conversion rate lies between 48.9 and 53.9%. Fifth, conversion required both technical resectability and patient consent. We excluded some responders who declined surgery or those with possible R1/R2 outcomes. This strategy may have led to an underestimate of the true conversion rate. Sixth, despite the promise of individual biomarkers, integrating radiomics, pan-lactylation, and PD-L1 to define consensus ESCC molecular subtypes remains a key challenge. Prospective, large-scale, multicenter trials are essential to validate this multi-modal approach. Seventh, due to the inherent challenge of obtaining substantial pre-treatment biopsies in ESCC, this study lacks the comprehensive histological data (e.g., tumor grade, baseline TILs) needed to correlate with treatment response. Finally, a comprehensive, deep mechanistic investigation utilizing spatial transcriptomics and multiplex immunofluorescence on paired biopsies, patient-derived organoids, and PDX models is warranted to delineate the immunologic impact of this regimen. Conclusions This phase 2 trial demonstrates that camrelizumab plus paclitaxel/nedaplatin is an active and feasible conversion therapy for unresectable locally advanced/advanced ESCC, potentially expanding curative-intent opportunities for historically inoperable patients. While compelling, definitive validation requires phase 3 randomized trials with rigorous correlative analyses. Future studies should also address key translational questions regarding optimal treatment duration, biomarkers, immune microenvironment modulation, and resistance mechanisms. If validated, this strategy may redefine the standard of care for borderline resectable ESCC in the immunotherapy era. Supplementary Information Additional file 1: CONSORT. Additional file 2: Figure S1, Tables S1 and S2. Fig. S1 Association between pathological response and survival outcomes. Kaplan–Meier curves comparing overall survival (OS) and disease-free survival (DFS) in pathological responders versus non-responders. (A, B) Responders with major pathological response (MPR) and pathological completer response (pCR) showed a trend toward improved OS compared to non-responders, though the difference did not reach statistical significance. (C, D) MPR and pCR responders exhibited suggestive but statistically nonsignificant DFS benefits. Table S1 Surgical outcomes of patients who underwent surgery (n = 67). Table S2 Postoperative complications graded by Clavien-Dindo (n = 67). Acknowledgements None. Abbreviations Combined positive score Confidence intervals Complete response Chemoradiotherapy Docetaxel plus 5-fluorouracil and cisplatin Disease-free survival Eastern Cooperative Oncology Group performance status Esophageal squamous cell carcinoma Immune checkpoint inhibitors Interquartile range Major pathological response Overall survival Pan-lactylation Partial response Pathological complete response Programmed death receptor-1 Response Evaluation Criteria in Solid Tumors National Cancer Institute Common Terminology Criteria for Adverse Events Treatment-related adverse events Tumor regression grade Authors' contributions J.M. designed the clinical trial; J.Z. designed the experiments; J.Z. and J.M. wrote the manuscript; J.M., J.Z., and H.M. provided the funding; H.M. and X.Y. collected tissue samples and performed IHC staining; JF.Z., X.L., C.W., Y.Y., Y.X., H.J., H.L., L.Z., X.Q., J.L., and C.F. collected clinical and pathological information; J.M. registered the trial and secured ethics approval; J.M., JF.Z., X.L., and X.W.were in charge of clinical treatment of enrolled patients; W. M. contributed to radiomic model development. J.M., JF.Z., X.L., C.W., X.W., Y.Y., Y.X., H.J., H.L., L.Z., X.Q., J.L., and C.F. re-examined the clinical efficacy. Y.W. analyzed IHC data. All authors have read and approved the article. Funding This work was supported by the Natural Science Foundation of China (82573679), National Key Research and Development Program of China (2024ZD0520904), Heilongjiang Provincial Natural Science Foundation of China (ZD2024H003), Beijing Xisike Clinical Oncology Foundation (Y-2024AZ(EGFR) MS-0187), and the Climbing Program of Harbin Medical University Cancer Hospital (PDYS2024-02), National Medical High Level Talent Program (to Hongxue Meng), Heilongjiang Province Innovation Base Award Project (JD2023SJ03), and Nn10 Project of Harbin Medical University Cancer Hospital (Nn102024-05). Data availability The data generated in this study are available upon request from the corresponding author. Declarations Ethics approval and consent to participate The study protocol has been approved by the China Ethics Committee of Registering Clinical Trials (ChiECRCT20210128). All patients or their legal surrogates provided written informed consent before participating in the study. Consent for publication Informed consent was obtained from all patients’ legal surrogates. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Jinhong Zhu, Jinfeng Zhang, Xiaodong Ling, and Changhong Wang equally contributed to the manuscript. Contributor Information Wei Meng, Email: articlemengwei@163.com. Xiaoyuan Wang, Email: 601749@hrbmu.edu.cn. Hongxue Meng, Email: menghongxue@hrbmu.edu.cn. Jianqun Ma, Email: jianqunma@hrbmu.edu.cn. References Associated Data Supplementary Materials Additional file 1: CONSORT. Additional file 2: Figure S1, Tables S1 and S2. Fig. S1 Association between pathological response and survival outcomes. Kaplan–Meier curves comparing overall survival (OS) and disease-free survival (DFS) in pathological responders versus non-responders. (A, B) Responders with major pathological response (MPR) and pathological completer response (pCR) showed a trend toward improved OS compared to non-responders, though the difference did not reach statistical significance. (C, D) MPR and pCR responders exhibited suggestive but statistically nonsignificant DFS benefits. Table S1 Surgical outcomes of patients who underwent surgery (n = 67). Table S2 Postoperative complications graded by Clavien-Dindo (n = 67). Data Availability Statement The data generated in this study are available upon request from the corresponding author.