Immunotherapy, AI and an Aging Population Are Rewriting Esophageal Cancer Surgery

Esophageal cancer has long been one of the most lethal and unforgiving of the major malignancies, but the ground beneath the surgeon’s feet is shifting faster now than at any point in the past two decades. A comprehensive review published in Annals of Gastroenterological Surgery synthesizes pivotal evidence from 2024 and 2025 and argues that the discipline is moving from standardized, procedure-centered operations toward a new paradigm built on immunotherapy integration, biologically informed personalization, and intelligent surgical technology. Three converging forces are driving the transformation: the arrival of immune checkpoint inhibitors in the perioperative setting, a rapidly aging and increasingly frail patient population, and the emergence of artificial intelligence and extended reality tools that can see, predict, and teach. The authors contend that the esophageal surgeon’s role is evolving from technical operator to strategic decision-maker, one who must balance oncologic efficacy, physiologic reserve, and technological innovation in every individual case.

The most mature evidence concerns what happens after surgery. The phase III CheckMate 577 trial established adjuvant nivolumab as the global standard for patients with esophageal or gastroesophageal junction cancer who still had residual disease after neoadjuvant chemoradiotherapy and a microscopically complete R0 resection. In the original analysis, nivolumab extended disease-free survival to 22.4 months compared with 11.0 months for placebo, a hazard ratio of 0.69. The 2025 ASCO update confirmed that the benefit endures: disease-free survival remained superior at long-term follow-up (hazard ratio 0.76), distant metastasis-free survival improved as well (hazard ratio 0.75), and a favorable five-year overall survival trend emerged of 46 percent versus 41 percent. Toxicity stayed manageable, with grade 3 to 4 treatment-related adverse events in roughly 14 percent of patients and no new safety signals. Notably, the benefit was more pronounced in squamous cell carcinoma and in tumors with a PD-L1 combined positive score of at least 1, an early hint that biomarker selection will govern who truly benefits.

Alternative adjuvant strategies have fared less well, and the failures are as instructive as the successes. A randomized phase II trial of adjuvant durvalumab in esophageal squamous cell carcinoma showed no improvement in disease-free or overall survival overall, although exploratory analyses hinted at a striking benefit in PD-L1-positive tumors, where 36-month overall survival reached 94 percent versus 64 percent in the control arm. Meanwhile, the European VESTIGE trial tested intensified dual immunotherapy, nivolumab plus ipilimumab, against standard adjuvant chemotherapy in high-risk gastroesophageal junction adenocarcinoma, and the results were sobering: disease-free survival was actually inferior in the immunotherapy arm at 11.4 versus 20.8 months, forcing early termination of the study. Taken together, the review concludes that adjuvant nivolumab after neoadjuvant chemoradiotherapy remains the only strategy with consistent and durable benefit, and that in East Asia, where chemotherapy-based neoadjuvant regimens predominate for squamous disease, the postoperative immunotherapy question remains open pending confirmatory phase III evidence.

Before surgery, the picture is more heterogeneous and histology matters enormously. In adenocarcinoma, the evidence for adding checkpoint inhibitors to perioperative chemotherapy remains inconsistent. The phase III KEYNOTE-585 trial raised pathological complete response rates with perioperative pembrolizumab plus FLOT chemotherapy but failed to deliver a statistically significant event-free survival benefit, and the PERFECT trial of atezolizumab with the CROSS chemoradiation regimen achieved a 25 percent pathological complete response rate without improving survival over matched historical controls. Squamous cell carcinoma tells a different story. Early phase II studies of pembrolizumab-based neoadjuvant therapy reported pathological complete response rates around 45 to 55 percent, and next-generation PD-1 inhibitors have built on that foundation. The TD-NICE trial of tislelizumab plus chemotherapy achieved high pathological response rates without treatment-related deaths or surgical delays, while the RATIONALE-306 trial had already demonstrated an overall survival benefit for the drug in advanced disease.

The strongest phase III signal in the neoadjuvant setting comes from China. In the ESCORT-NEO/NCCES01 trial, 391 patients with resectable esophageal squamous cell carcinoma were randomized to chemotherapy with or without camrelizumab. The addition of the immunotherapy agent nearly doubled pathological complete response rates, reaching 28.0 percent with nanoparticle albumin-bound paclitaxel plus cisplatin and 15.4 percent with conventional paclitaxel plus cisplatin, compared with just 4.7 percent for chemotherapy alone. Crucially, this came without excess surgical morbidity, which ranged from 34 to 39 percent across arms. A network meta-analysis ranked camrelizumab plus chemotherapy highest for pathological responses, while pembrolizumab-based regimens showed the best objective response and disease control rates. Meta-analyses across histologies confirm that neoadjuvant immunochemotherapy and immunochemoradiotherapy achieve high response rates with surgical feasibility comparable to conventional regimens, and grade 3 or worse adverse event rates of roughly 20 to 35 percent, similar to established treatments.

For the operating room itself, timing and technique are being codified. Surgery across trials was typically scheduled four to six weeks after the last immunotherapy dose, an interval long enough to resolve immune-related inflammation but short enough to avoid progressive fibrosis. Common immune-related adverse events such as rash, thyroid dysfunction, and pneumonitis are usually low grade and manageable, but surgeons must watch for them, since operating through an unresolved pneumonitis or impaired wound healing would be hazardous. Real-world and multicenter analyses have shown that minimally invasive esophagectomy after neoadjuvant chemoimmunotherapy produces complication rates, operating times, pneumonia, and anastomotic leakage figures comparable to surgery alone. Perhaps most provocatively, a meta-analysis of 14 studies encompassing 3,212 patients found that complete lymph node regression and nodal downstaging after therapy were strongly associated with survival, with hazard ratios of roughly 0.47 for overall survival and 0.42 for disease-free survival, raising the prospect of tailoring the extent of lymphadenectomy to each patient’s pathological response.

The second pillar of the new paradigm is the aging patient. The benefit of neoadjuvant chemotherapy in the elderly depends far more on physiology than on birth certificates: a multicenter study showed that preoperative chemotherapy significantly improved five-year overall survival only among patients with a performance status of zero, from 38.1 to 56.5 percent, while those with any impairment of daily function saw no benefit. Real-world data indicate that dose reductions in later chemotherapy cycles are safe in patients over 70 without compromising long-term survival, and a 2025 nationwide Japanese database study found comparable perioperative morbidity between doublet and docetaxel-based triplet regimens in older patients, supporting selection by physiological reserve rather than chronological age. Body composition emerges as a decisive variable. Meta-analyses from 2024 and 2025 show that sarcopenia, the loss of skeletal muscle mass, nearly doubles the risk of poor overall and disease-free survival and increases postoperative complications by about 30 percent, with sarcopenic obesity carrying the worst prognosis of all.

Nutrition and conditioning complete the geriatric assessment. The Geriatric Nutritional Risk Index, categorized above 98, between 92 and 98, and below 92, has been repeatedly validated as a predictor of survival and postoperative morbidity, including in a 2024 pan-cancer analysis. Prehabilitation, structured exercise and nutrition programs delivered before surgery, reduced postoperative pneumonia by nearly 30 percent in a meta-analysis of upper gastrointestinal cancer studies and shortened hospital stays, and systematic reviews in esophagectomy specifically found pulmonary complication odds ratios around 0.35 to 0.48 in prehabilitation cohorts. For patients whose disease recurs after definitive chemoradiotherapy, salvage surgery, once feared for its mortality, has become a realistic curative option in selected cases. The multicenter SURGES study reported three-year overall survival of approximately 49 percent with 90-day mortality of just 3.3 percent, and minimally invasive salvage techniques cut overall complications and pneumonia roughly in half compared with open surgery. Even oligometastatic disease may benefit, with local therapy extending median survival from 7.8 to 35.2 months in one observational cohort.

The third pillar is technological. Artificial intelligence models trained on robotic surgical videos are learning to identify the recurrent laryngeal nerve, a structure whose injury causes vocal cord paralysis and whose preservation is among the most delicate steps of esophagectomy. One model trained on 120 robot-assisted procedures achieved respectable segmentation scores and, more strikingly, helped trainee surgeons recognize the nerve earlier in 81.3 percent of cases versus 46.9 percent for controls. A follow-up proof-of-concept study showed that AI could detect excessive traction on the nerve with 84.4 percent accuracy before changes appeared on conventional electromyographic monitoring, opening a window for corrective action before injury occurs. Beyond the operating theater, machine learning is reshaping planning: radiomics models predict lymph node metastasis with areas under the curve approaching 0.74 to 0.89, deep learning predicts pathological complete response after chemoimmunotherapy with AUCs of 0.83 to 0.92, and gradient boosting models outperform established risk scores for predicting 90-day mortality. Integration with circulating tumor DNA, whose postoperative positivity strongly signals recurrence risk, promises molecularly grounded risk stratification.

The review’s conclusion is a call for synthesis. Virtual reality simulators and AI-driven feedback systems are standardizing training in robotic esophagectomy, validated procedure-specific rehearsal tools and international consensus curricula are emerging, and automated frameworks can now evaluate suturing performance in real time. When the stomach cannot serve as the replacement conduit, free jejunal grafts with microvascular anastomosis offer a reliable alternative, and thoracoscopic esophagectomy has proven non-inferior to open surgery in the landmark JCOG1409 trial, with robotic and transmediastinal approaches broadening options for patients with poor pulmonary reserve. What unites these threads is a philosophy: the surgeon of the coming decade will not simply execute an operation but will select, time, and integrate surgery within a multidisciplinary pathway informed by tumor biology, ctDNA, radiomics, frailty indices, and machine intelligence. The future of esophageal cancer surgery, the authors argue, will be defined not by technical precision alone, but by the capacity to fuse biological insight, data intelligence, and clinical judgment into genuinely personalized care.

Subject of Research: Integrating immunotherapy, frailty assessment, and artificial intelligence into personalized surgical strategies for esophageal cancer

Article Title: Essential Updates 2024–2025: Surgical Strategy for Esophageal Cancer Toward a New Paradigm in the Era of Immunotherapy and Personalization

Article References: Oya, S., Yagi, K., & Baba, Y. (2026). Essential Updates 2024–2025: Surgical Strategy for Esophageal Cancer Toward a New Paradigm in the Era of Immunotherapy and Personalization. Annals of Gastroenterological Surgery, 10(5), 1368-1381. https://doi.org/10.1002/ags3.70197

Image Credits: AI Generated

DOI: 10.1002/ags3.70197

Keywords: esophageal cancer, immunotherapy, immune checkpoint inhibitors, esophagectomy, neoadjuvant therapy, nivolumab, sarcopenia, prehabilitation, artificial intelligence, robotic surgery, salvage surgery, personalized medicine