Esophageal cancer remains a major global health concern, with an estimated 604,100 new cases and 544,076 deaths reported in 2020 (1). More than half of these cases occur in China, where esophageal squamous cell carcinoma (ESCC) is the predominant histologic subtype, in contrast to the predominance of esophageal adenocarcinoma in Western countries (2). The prognosis of ESCC remains poor, largely due to late-stage diagnosis stemming from nonspecific symptoms and limitations in early detection strategies (3).

Surgical resection is a mainstay of treatment for locally advanced ESCC, offering a 5-year overall survival rate of 52.9% (4). However, postoperative recurrence occurs in approximately 33.7% of patients, underscoring the need for improved preoperative strategies (4). The addition of immunotherapy to chemotherapy has shown synergistic antitumor effects across multiple cancer types (5–7), enhancing tumor immunogenicity and modulating immune responses through the activation of cytotoxic T cells and depletion of immunosuppressive cell populations (8–11). Recent important studies have explored cellular and molecular factors affecting immunotherapy. For example, glycolysis-driven histone lactylation in breast cancer is closely linked to tumor immune responses and immune-related pathway regulation (12). Solid tumors exhibit multiple subtypes due to cellular and molecular heterogeneity, with immune cell infiltration in the tumor microenvironment correlating with immunotherapy efficacy (13). Additionally, the stemness index is a potential biomarker for cancer progression and prognosis; studies on non-small cell lung cancer showed that patients with a high stemness index responded better to PD-1/PD-L1 blockade immunotherapy (14).

Recent phase II and III trials evaluating neoadjuvant immunochemotherapy (nICT) in locally advanced ESCC patients have demonstrated favorable surgical outcomes and manageable toxicity profiles (15–18). Despite these advances, a subset of patients remains ineligible for surgery following nICT due to inadequate tumor regression or disease progression (19, 20). Data from prospective trials suggest that up to 25% of patients fail to undergo surgical resection after completing nICT. For example, in the ESONICT-2 study, only 15 of 20 patients (75.0%) underwent surgery after receiving toripalimab in combination with chemotherapy (21).

Radiotherapy may increase resectability by reducing the tumor burden, decreasing viability, and facilitating clearer intraoperative margins through induced fibrosis at the tumor–normal tissue interface (22–25). On the basis of these observations, we hypothesized that secondary induction radiotherapy could improve the clinical response and surgical outcomes in patients with a poor response to nICT.

To our knowledge, this retrospective study is the first research to evaluate the efficacy and safety of secondary induction radiotherapy in patients with locally advanced ESCC who demonstrated poor response to nICT. The findings suggested that the secondary induction radiotherapy yielded high rates of R0 resection, encouraging pathological responses and acceptable toxicity in this populations of patients, offering a potential therapeutic regimen to curative surgery for this challenging subgroup.

Immunochemotherapy has emerged as a key component of neoadjuvant treatment for ESCC. In the phase III ESCORT-NEO/NCCES01 trial, patients receiving camrelizumab combined with chemotherapy achieved significantly higher pCR and MPR rates than those receiving chemotherapy alone did (28.0% and 59.1% in Group A vs. 4.7% and 20.9% in Group C, respectively) (26). As reported in the ESONICT-1 study, neoadjuvant sintilimab combined with chemotherapy achieved a pCR rate of 21.7% and a MPR rate of 52.2% (18). Similarly, the ESONICT-2 trial demonstrated that the regimen of toripalimab plus docetaxel and cisplatin yielded a pCR of 16.7% and a MPR of 41.7% (21). These findings are consistent with the broader landscape of phase II trials in this setting, which have reported pCR rates ranging from 20% to 50%, along with encouraging tumor downstaging and surgical feasibility (15–17). However, a proportion of patients showed limited tumor regression following nICT (27). These nICT-insensitive patients are considered unlikely to achieve a favorable pathological response or long-term survival after surgery. In a prospective phase II trial (ChiCTR2100045722) enrolling locally advanced ESCC patients who received nICT, the pCR rates in the responder (CR/PR) and non-responder (SD/PD) groups were 32.2% and 0.0%, respectively. Compared with the non-responder group, the responder group trended toward prolonged DFS (p=0.310) and OS (p=0.08) (28). A real-world retrospective study further showed that the 2-year DFS rate in locally advanced ESCC patients with CR/PR following nICT was significantly higher than that in patients with SD/PD (83.2% vs. 59.8%, P = 0.04), with a similar trend observed for the 2-year OS rate (89% vs. 63.8%, P = 0.02) (29). Therefore, the treatment strategy proposed in this study, which involves introducing a second induction radiotherapy to further improve clinical response rates in ESCC patients with SD/PD after nICT, is of great necessity. In our study, all 23 patients who showed poor response to nICT completed secondary induction radiotherapy prior surgery. The efficacy was assessed as CR (17.4%) in 4 patients, PR (34.8%) in 8 patients, and an ORR of 52.2% and a DCR of 100%.

Radiotherapy has long been recognized for its capacity to induce tumor shrinkage and improve resectability (22–25). In our study, 23 patients with a poor response to nICT underwent secondary induction radiotherapy. More than half (52.2%) of patients achieved an objective response, and all of them proceeded to surgery, with an R0 resection rate of 100%. Postoperative histopathological analysis revealed a pCR rate of 21.7% and an MPR rate of 43.5%. These results were comparable to historical data on nICT alone in ESCC (4, 15–18, 21, 26, 30) and suggested that radiotherapy represented a trend of restoring treatment sensitivity and surgical eligibility in this refractory cohort. Disease recurrence occurred in 5 patients, with 1 death during follow-up. Although long-term survival outcomes still needed support by further follow-up data, the universal achievement of R0 resection which was an established prognostic factor in esophageal cancer was encouraging (31). In one retrospective study, patients who underwent R0 resection had significantly better survival outcomes than those who underwent noncurative (R1 or R2) resection (P < 0.0001) (32).

The optimal radiation dose for neoadjuvant treatment of esophageal cancer remains debated. A number of retrospective studies have evaluated the impact of dose intensity, but the results have been inconsistent (33–37). A trial of 141 patients with stage III ESCC reported that a higher dose (50–50.4 Gy) was associated with improved pCR and 3-year survival compared with a lower dose (36 Gy) (31). However, a meta-analysis using biologically effective dose (BED) stratification reported superior survival and lower toxicity in the low-dose group (BED ≤48.85 Gy) (33). Subgroup analyses further favored the 40-41.4 Gy dose range, which was associated with improved progression-free survival (PFS) and overall survival (OS) and lower rates of distant metastasis. Our findings supported the use of 41.4 Gy as the promising dose for secondary induction radiotherapy in ESCC patients who demonstrated poor response to nICT. This regimen was associated with high rates of resection and pathological response while maintaining an acceptable safety profile. The events of postoperative complication were rare, with only 2 patients (8.7%) experiencing grade 3–4 surgical complications. Radiation-induced toxicity was also manageable, with grade 3–4 TRAEs, including myelosuppression (13.0%) and gastrointestinal symptoms (8.7%). A safety comparison with the nCRT group from the NEOCRTEC5010 trial revealed a markedly lower incidence of severe myelosuppression in our cohort (13.0% vs. 48.9%, P < 0.001) (4). In contrast, the rate of severe gastrointestinal toxicity did not differ significantly from the historical benchmark (8.7% vs. 4.0%, P = 0.274). Collectively, these data suggest an accepted safety tolerance for our treatment strategy of adding secondary induction radiotherapy for ESCC patients with a poor response to nICT.

Regarding long-term survival and recurrence patterns, our study found that incorporating secondary induction radiotherapy did not change the recurrence profile after neoadjuvant immunochemotherapy. The distribution of first recurrence sites remained similar, with distant metastases continuing to predominate over locoregional failure. The spectrum of involved metastatic sites was also comparable to that reported in other study (38).

This study has several limitations. Owing to its retrospective design and sample size, exemplified by the wide confidence interval for the pCR rate (21.7%, 95% CI 7.5%–43.7%), potential for selection bias should be acknowledged, and broader generalizability of the findings may require validation in larger cohorts. The lack of a comparative group limits direct inference on treatment efficacy, though the study underscores the feasibility of secondary induction radiotherapy in this specific patient population. Without a direct comparator cohort (e.g., patients undergoing surgery alone or continuing with nICT alone), the observed clinical benefits cannot be definitively isolated from potential selection biases. Consequently, the outcomes may partly reflect a “pseudo-benefit”, attributable to the enrollment of patients with favorable baseline biology or those already in a state of disease control after prior therapy, rather than solely to the therapeutic effect of radiotherapy. These observations highlight the need for prospective investigations to further characterize the role of such multi-modal approaches in clinical practice. Furthermore, the follow-up duration was insufficient to assess long-term survival outcomes. Given the study’s limitations, including susceptibility to selection bias, small sample size uncertainty, and a short follow-up that precludes assessment of long-term survival benefit, the observed promising pathological responses should be interpreted with caution. Although the observed R0, pCR and MPR rates in our cohort appeared to fall within the spectrum of rates reported in prior prospective trials of nICT alone, these findings must be regarded as preliminary and require validation in larger, prospective studies whether these favorable responses translate into long-term survival benefit.

Our work conducted an initial, hypothesis-generating assessment in this challenging patient population (poor responders to nICT). Therefore, building on these preliminary findings, future research should aim to optimize the integration of immunotherapy, chemotherapy, and radiotherapy in a sequential or concurrent fashion by using a controlled study design. Further efforts are needed to refine treatment sequencing, identify biomarkers predictive of response, and personalize treatment on the basis of tumor biology. These strategies are critical for improving long-term outcomes and quality of life in patients with locally advanced ESCC.

This study provides preliminary evidence that secondary induction radiotherapy is a feasible and effective approach for patients with locally advanced ESCC who fail to respond to nICT. The high R0 resection rate, encouraging pathological response, and manageable toxicity observed in this cohort support further investigation of this strategy in prospective clinical trials.