In this review, anatomical unresectability is defined as a thoracic surgical and multidisciplinary assessment indicating that R0 resection, namely complete tumor removal with microscopically negative margins, is not expected to be achievable with acceptable safety. Assessment should integrate contrast-enhanced imaging, endoscopy, staging investigations, and bronchoscopy when airway invasion is suspected, while considering the structures involved and the feasibility of en bloc resection. cT4b tumor invasion of the aorta, trachea or bronchi, or vertebral body is a typical anatomical reason; TNM staging alone cannot replace an assessment of surgical resectability [7][8]. The Japan Esophageal Society guidelines explicitly state that consideration of conversion surgery after induction therapy or dCRT should be based on an expectation of R0 resection and acceptable operative risk [8]. Initial unresectability should therefore be recorded separately from resectability reassessed after treatment.

Based on eligibility criteria and baseline data, studies are classified as involving anatomically unresectable disease, mixed nonoperative populations, or insufficient information for classification. Mixed nonoperative populations may include patients with technically resectable tumors but contraindications to surgery, patients who decline surgery, and patients with cervical esophageal cancer treated for organ preservation. Nonsquamous histology, postoperative recurrence, and specific extra-regional nodal metastases are identified separately. Even if a study uses the term “unresectable”, its entire population cannot be classified as anatomically unresectable if the eligibility criteria also permit the other circumstances described above. Overall results from mixed cohorts cannot substitute for separately reported results from the anatomically unresectable subgroup.

Information sources for this review include published or publicly available PubMed-indexed literature, original publisher reports and supplementary materials, ClinicalTrials.gov records, and professional guidelines. Targeted searches used trial names and combinations of keywords including esophageal squamous cell carcinoma, definitive chemoradiotherapy, immune checkpoint inhibitor, induction, concurrent, consolidation, and maintenance; relevant references and follow-up reports were also checked. Prospective studies of immune checkpoint inhibitors combined with curative-intent dCRT in adults with ESCC were prioritized. Greater interpretive weight was given to randomized studies, mature follow-up, and studies with clearly reported eligibility criteria and safety data. Retrospective studies, mechanistic studies, and guidelines were used for discussion of the background, biomarkers, and toxicity management.

dCRT refers to radiotherapy and chemotherapy delivered concurrently with curative intent, without planned routine surgery. RTOG 85-01 established the superiority of concurrent chemoradiotherapy over radiotherapy alone; INT 0123 found no improvement in survival or local control with 64.8 Gy compared with 50.4 Gy; ARTDECO likewise found no improvement in major outcomes when the primary tumor dose was increased to 61.6 Gy [9]-[11]. Therefore, 50.0 - 50.4 Gy in 25 - 28 fractions of 1.8 - 2.0 Gy can serve as an evidence-based reference for cross-study comparisons. Some regimens in Japan and China use 50 - 60 Gy for unresectable disease, particularly T4b lesions, but routine dose escalation should not be equated with better efficacy. Modern radiotherapy should use intensity-modulated radiotherapy or volumetric modulated arc therapy with image guidance and fully report the total dose, fractionation, treatment interruptions, and cardiac and pulmonary dose exposure [12].

Radiation fields are generally classified as involved-field irradiation (IFI) or elective nodal irradiation (ENI). IFI covers the primary tumor, radiographically positive lymph nodes, and a margin for subclinical extension; ENI additionally covers high-risk regional lymphatic drainage areas. Current evidence is insufficient to support the same strategy for all patients. Selection should consider tumor location, nodal distribution, normal-organ doses, and the trial protocol, and should be explicitly reported [12]. Conventional concurrent chemotherapy uses cisplatin plus fluorouracil; FOLFOX is an alternative supported by randomized studies [13], while taxane-platinum combinations are commonly used in Chinese and other Asian studies. Before evaluating the additional benefit of an immunotherapy combination, the radiotherapy dose, radiation fields, intensity of concurrent drug treatment, and treatment completion rate should first be confirmed as comparable to those of the reference regimen.

Confirmatory evidence for ICI monotherapy comes mainly from advanced or recurrent disease. ATTRACTION-3 showed that nivolumab improved overall survival (OS) in patients with previously treated advanced ESCC; KEYNOTE-181 suggested that the benefit of pembrolizumab was concentrated mainly in patients with high programmed death-ligand 1 (PD-L1) combined positive scores [14][15]. These studies established the antitumor activity of programmed cell death protein 1 (PD-1)/PD-L1 pathway blockade, but their disease stages, treatment intent, and treatment backbones differ. They therefore do not directly demonstrate the additional benefit of ICIs in unresectable locally advanced disease treated with dCRT.

Radiotherapy may enhance antitumor immunity through tumor antigen release, antigen presentation, and cytosolic DNA sensing, and may upregulate PD-L1, providing a rationale for immune checkpoint blockade. Conversely, larger radiation fields and low-dose exposure may also cause lymphopenia or immunosuppression [16]-[23]. Mechanistic studies therefore help explain treatment sequencing and target-volume optimization, but cannot replace randomized clinical evidence. Radiation dose, fractionation, and nodal irradiation may all affect immune responses, and a single immunologically optimal radiotherapy regimen cannot be assumed across different treatment settings [24].

Concurrent combination treatment aims to cover the window of radiotherapy-induced antigen release, and some studies continue maintenance immunotherapy after dCRT. A phase Ib study of camrelizumab combined with docetaxel, cisplatin, and radiotherapy enrolled 20 patients with ESCC and reported 1-year OS and PFS rates of 85% and 80%, respectively; however, the original study provided limited stratification of anatomical unresectability and mainly supports treatment feasibility [25]. The population described as “inoperable” in the study of durvalumab plus tremelimumab also included patients who were medically or surgically unsuitable for surgery, had cervical tumors, or had postoperative local recurrence, while excluding T4b disease. It therefore cannot serve as direct evidence for anatomically unresectable ESCC [26].

All 42 patients in EC-CRT-001 had ESCC, but approximately 62% were considered inoperable; the remaining patients could include those who declined surgery or had cervical tumors. The overall cCR rate of 62%, 3-year OS rate of 44.8%, and PFS rate of 35.7% should therefore be interpreted as results from a mixed nonoperative cohort and cannot be directly extrapolated to a purely anatomically unresectable subgroup [27][28]. NOBEL likewise involved a mixed population, including 25 patients with resectable and 16 with unresectable ESCC. The overall cCR rate was 73%; in the unresectable subgroup, the cCR rate was 56% and the 1-year OS rate was 81.3%, so interpretation should prioritize the subgroup results [29]. These single-arm studies suggest the feasibility of concurrent combination treatment but have not demonstrated its superiority over standard dCRT.

Another single-arm phase II study evaluated QL1706, which includes both PD-1 and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) blockade components, combined with dCRT [30]. This study extends investigation to dual-checkpoint strategies but has not demonstrated superiority over PD-1 blockade alone or standard dCRT.

Induction chemotherapy plus immunotherapy may shrink the tumor before radiotherapy, improve swallowing, and provide early response information, but may also increase bone marrow suppression or delay dCRT. The sintilimab induction study enrolled 75 patients with locally advanced esophageal cancer receiving nonoperative treatment, predominantly ESCC; however, anatomical resectability was not stratified in a way that permits independent extrapolation. The 2-year local control rate of 81.7% should therefore be viewed only as an exploratory result of this treatment pathway [31].

ImpactCRT enrolled 49 patients with ESCC. Although the study described them as unresectable, eligibility included inoperability, contraindications to surgery, or refusal of surgery; 16 patients (32.7%) had stage IVB disease with metastasis confined to the supraclavicular lymph nodes. The study should therefore be classified as a mixed nonoperative cohort. In the intention-to-treat (ITT) population, the 1-year OS/PFS rates were 85.7%/71.4%, and the 2-year rates were 72.5%/60.5%. Its primary endpoint, the 1-year OS rate, was assessed in the per-protocol population of 46 patients, using a different denominator from the ITT results. The single-arm design and occurrence of 1 treatment-related death warrant cautious interpretation [32].

EC-CRT-002 randomized 114 patients equally to tislelizumab maintenance or no maintenance, with 57 patients per group. The registry eligibility criteria allow nonoperative treatment for medical reasons or patient choice, so the entire population cannot be labeled anatomically unresectable [33][34]. Both groups first received 2 induction cycles every 3 weeks: paclitaxel 135 - 175 mg/m², cisplatin 75 mg/m², and tislelizumab 200 mg. This was followed by intensity-modulated radiotherapy at 50.4 Gy in 28 fractions, with concurrent weekly paclitaxel 45 mg/m² and cisplatin 25 mg/m² for 5 weeks, and 2 additional doses of tislelizumab every 3 weeks. Group A received 12 additional maintenance cycles, for a total of 16 immunotherapy cycles; group B stopped immunotherapy after 4 cycles [33][34].

At a median follow-up of 22.7 months, the 1-year PFS rate was 52.6% (95% CI 41.4 - 67.3) in group A and 71.9% (61.1 - 84.6) in group B, compared with 56.4% (44.7 - 71.1) in the historical control. The PFS hazard ratios for groups A and B relative to that historical control were 1.06 (95% CI 0.67 - 1.68) and 0.54 (0.32 - 0.94), respectively; the OS hazard ratio for group B relative to the historical control was 0.42 (0.22 - 0.82) [33]. These hazard ratios derive from historical-control comparisons and are not estimates of the effect between the randomized groups A and B. The study did not demonstrate the superiority of the immunotherapy combination through a concurrently randomized dCRT control, and its results do not support routine extension to 16 cycles. It also cannot demonstrate equivalence between the two treatment durations or establish 4 cycles as a universally optimal regimen.

The consolidation strategy includes only patients who have completed dCRT without progression, reducing the complete overlap of acute toxicities but also introducing selection of treatment completers. All 40 patients in EPOC1802 had unresectable ESCC; 95% had cT4 disease, and 25% had extra-regional M1 lymph nodes. Patients were enrolled after dCRT and received atezolizumab for 1 year. From study enrollment, median PFS and OS were 3.2 months and 31.0 months, respectively; when retrospectively measured from the start of dCRT, they were 5.5 months and 33.0 months [35]. These two time origins must not be conflated with those used in concurrent or induction studies.

A single-arm study of consolidative camrelizumab after involved-field dCRT provides additional feasibility evidence [36]. However, enrollment and follow-up were conditional on completion of the preceding treatment, and the study has not resolved the comparative benefit of the post-dCRT consolidation strategy.

SKYSCRAPER-07 randomized 760 patients after completion of dCRT without progression. Atezolizumab plus tiragolumab did not meet the primary PFS endpoint versus placebo (HR 0.82, 95% CI 0.65 - 1.03; P = 0.0947), and the prespecified hierarchical testing therefore stopped. Atezolizumab monotherapy showed a favorable OS signal (HR 0.69, 95% CI 0.52 - 0.91; descriptive P = 0.0085) and PFS signal (HR 0.74, 95% CI 0.58 - 0.93), but both are descriptive results under the prespecified multiplicity framework [37]. This qualification preserves the positive signal for monotherapy while avoiding its interpretation as confirmatory evidence that all patients should routinely receive consolidation therapy.

KEYNOTE-975 includes squamous cell carcinoma and adenocarcinoma of the esophagus or gastroesophageal junction and therefore involves a histologically mixed population. EFS is measured from randomization, and its endpoint includes protocol-defined radiographic local, regional, or distant progression or recurrence and death [38][39]. Before the sample size, interaction test, and results for the prespecified unresectable ESCC subgroup are formally reported, conclusions for this specific subgroup cannot be drawn solely from results in the overall or mixed-histology population. RATIONALE-311, KUNLUN, and ESCORT-CRT focus on ESCC, but “inoperability” in some protocols may still include refusal of surgery or other nonanatomical reasons [40]-[42]. Final reports should therefore present the anatomically unresectable subgroup separately and report its PFS/EFS, OS, and local control outcomes using consistent time origins.

Table 1.Representative Prospective Studies and Interpretation of Results.

Note: Abbreviations: cCR, clinical complete response; CI, confidence interval; HR, hazard ratio; ITT, intention-to-treat; NR, not explicitly reported in the reviewed materials. Population classification follows Section 1.1. The cCR assessment window differs from the time origin of survival endpoints; NR does not mean that the study did not collect the endpoint. This review did not calculate pooled effects or indirect comparative effects.

Table 2 compares reported severe toxicity, treatment delivery, fistula, and treatment-related death. These data are descriptive observations and cannot be used to rank the safety of the three strategies. Concurrent studies include the acute CRT phase, induction studies add an earlier period of treatment exposure, and consolidation studies begin counting events only after further selection among patients who have completed dCRT. Differences in the extent of T4b invasion, radiation dose, laboratory monitoring, adverse-event attribution, and safety-analysis denominators can substantially affect the reported incidence. In particular, the incidence of severe lymphopenia must not be equated with the overall incidence of clinically symptomatic severe toxicity.

Table 2.Descriptive Safety Comparison of Representative Treatment Strategies.

AE, adverse event; TRAE, treatment-related adverse event; G, grade; NR, not separately reported in the main text or available abstract. NR does not equal 0. Percentages retain the analysis populations of the original reports, and overall adverse events are distinguished from individual toxicities. Failure to complete treatment cannot invariably be attributed to toxicity; events occurring before enrollment for consolidation cannot be counted as events during consolidation treatment.

In EC-CRT-002, G≥3 adverse events occurred in 86.0% of the maintenance group and 80.7% of the nonmaintenance group, with corresponding rates of severe lymphopenia of 77.2% and 73.7% [33]. These data alone cannot establish the reasons for the difference in efficacy. Safety after dCRT is also affected by eligibility criteria: before atezolizumab was initiated, EPOC1802 excluded 1 patient with a tracheal fistula after CRT and some patients with persistent, unresolved CRT-related toxicity [35]. The absence of fistula in a small, selected cohort cannot demonstrate that immunotherapy prevents fistula.

Safety data should retain the version of the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) used in each original study; historical v4.0 or v5.0 events should not be retrospectively regraded using a later version [48]. Management should distinguish CRT-related injury, immune-related adverse events (irAEs), infection, and tumor progression. Table 3 summarizes the main differential diagnoses and the basis for management.

Table 3.Major adverse events and management principles.

Suspected immune-related pneumonitis should first be graded according to CTCAE. Grade 1 refers to radiographic changes without symptoms and requires close monitoring; ICIs may be temporarily withheld, with reassessment of symptoms, oxygen saturation, and chest CT. For grade 2 pneumonitis with symptoms requiring medical intervention, ICIs should be withheld; prednisone 1 - 2 mg/kg/day or an equivalent dose is commonly used and tapered over at least 4 - 6 weeks after improvement. If infection has not been excluded, pathogen testing should be performed concurrently, with empirical anti-infective treatment when appropriate. Grade 3 pneumonitis involves severe symptoms, impairment of self-care, or a need for supplemental oxygen; grade 4 is life-threatening. Both require hospitalization, permanent discontinuation of ICIs, and intravenous methylprednisolone 1 - 2 mg/kg/day. If there is no improvement within 48 - 72 hours, a multidisciplinary team should consider second-line immunosuppressive treatment such as mycophenolate mofetil or intravenous immunoglobulin [55][56]. Rechallenge may be considered on an individual basis only after a grade 2 event has completely or substantially resolved to grade ≤ 1, corticosteroids have been reduced to a physiological or low dose, and infection has been excluded. Reuse of ICIs is generally avoided after grade 3 - 4 pneumonitis.

Endocrine irAEs are an exception to the general principle of permanent discontinuation. Emergencies such as adrenal crisis or severe hyperglycemia should first be stabilized; thyroid, pituitary, or adrenal insufficiency is managed primarily with appropriate hormone replacement. Resumption of ICIs may be considered once the condition is stable, and treatment should not be permanently discontinued solely because long-term hormone replacement is needed [55][56]. High-dose corticosteroids are generally not used for isolated hypothyroidism. When adrenal insufficiency is suspected, glucocorticoid deficiency should be corrected before thyroid hormone replacement is initiated.

Radiation esophagitis and nutritional support likewise require standardized, evidence-based management. Body weight, intake, and swallowing function should be assessed continuously before and during treatment. Initial measures include individualized dietary guidance, analgesia, acid suppression and mucosal protection, and oral nutritional supplements. Enteral nutrition is preferred when gastrointestinal function is intact but oral intake remains inadequate; parenteral nutrition is considered only when enteral nutrition is not feasible or remains insufficient [49][50]. Esophageal stents should not routinely be placed during dCRT to prevent obstruction. The European Society of Gastrointestinal Endoscopy (ESGE) recommends covered self-expanding metal stents to seal malignant tracheoesophageal or bronchoesophageal fistulas, but placement before or during definitive chemoradiotherapy may increase adverse events. Establishing a route for nutritional support is generally prioritized, with individualized decisions made by a multidisciplinary team [52].