Perioperative immune checkpoint inhibitor therapy across tumors: Insights and shared lessons from a rapidly evolving field

The integration of immune checkpoint inhibitors into perioperative management marks a major evolution in curative-intent oncology. This review examines the current evidence for perioperative immunotherapy encompassing neoadjuvant, adjuvant, and combined strategies in melanoma and non-melanoma skin cancers, non-small cell lung cancer (NSCLC), triple-negative breast cancer (TNBC), esophageal and gastroesophageal junction cancer, renal cell carcinoma, urothelial carcinoma, head and neck squamous cell carcinoma, colorectal cancer, gynecological malignancies, and hepatocellular carcinoma. Neoadjuvant immunotherapy demonstrates biological advantages by exposing the immune system to intact tumor antigens, consistently improving event-free survival and pathological response rates across tumor types. Notable successes include CheckMate 816 in NSCLC, KEYNOTE-522 in TNBC, and emerging trials in melanoma that show superior outcomes compared to adjuvant-only approaches. Pathological complete response and major pathological response have emerged as robust surrogate endpoints correlating with long-term survival. In contrast, adjuvant immunotherapy shows more variable results, with demonstrated recurrence-free survival benefits but inconsistent overall survival (OS) advantages-particularly concerning given the risk of overtreatment in patients potentially cured by surgery alone. Critical challenges include the absence of predictive biomarkers in most cancer types, immune-related adverse events occurring in up to 30% of patients, substantial healthcare costs, and insufficient OS follow-up duration in many approved indications. Future priorities include biomarker development, adaptive trial designs incorporating response-guided therapy, and long-term toxicity assessment. Although perioperative immunotherapy is reshaping curative-intent cancer treatment, optimal patient selection, treatment sequencing, and safety optimization remain essential for widespread implementation.