Adjuvant immunotherapy versus observation after neoadjuvant therapy and esophagectomy for esophageal cancer with residual pathologic disease: a systematic review and reconstructed individual participant data meta-analysis

Abstract

Background The role of adjuvant immunotherapy after neoadjuvant therapy and esophagectomy remains uncertain. We examined its association with survival in populations composed predominantly of patients with residual disease across neoadjuvant strategies. Methods This systematic review included comparative studies of postoperative immunotherapy versus observation or placebo. The majority of the intervention groups received adjuvant immune checkpoint inhibitor (ICI) monotherapy, while three cohorts included patients treated with concomitant postoperative chemotherapy. The reported hazard ratios (HRs) were synthesized using random-effects models and were estimated from reconstructed Kaplan–Meier data when unavailable. Reconstructed data were also used to generate the survival curves. Bayesian hierarchical models and sensitivity analyses assessed robustness. Results A total of 13 studies, including two randomized trials and 11 retrospective cohorts, were included. The overall survival (OS) data came from 12 studies with 4,225 patients, while the disease-free or recurrence-free survival (DFS/RFS) data were from 11 studies with 2,363 patients. Postoperative immunotherapy was associated with longer OS [HR = 0.64, 95% confidence interval (CI) = 0.53–0.77] and DFS/RFS (HR = 0.74, 95%CI = 0.66–0.84). The associations were favorable after neoadjuvant chemoradiotherapy (NCRT), chemoimmunotherapy (NCIT), and chemotherapy (NCT). Interaction testing was significant for OS, but not for DFS/RFS. The paradigm estimates were not direct comparisons as the histology, geography, and study design differed across strategies. Sensitivity analyses addressing residual disease eligibility and postoperative treatment composition yielded similar average estimates. Analyses restricted to studies with aligned postoperative time origins remained directionally favorable, but were based on only a few studies. The Bayesian pooled estimates were concordant, although the predictive intervals crossed 1 for both outcomes. Certainty was low. Conclusions Randomized evidence supports a DFS benefit from adjuvant immunotherapy after NCRT. However, a statistically significant randomized OS benefit has not been demonstrated. The associations after NCIT were favorable, but entirely observational. The favorable estimates after NCT were based on only two small retrospective studies, while mixed neoadjuvant treatment was represented by a single study with imprecise estimates. The esophageal squamous cell carcinoma (ESCC)-predominant and residual pathology findings were exploratory. Further randomized studies are needed to define the role of postoperative immunotherapy after NCIT or NCT and to determine whether pathologic features can inform patient selection. Systematic Review Registration https://www.crd.york.ac.uk/PROSPERO/, identifier CRD420261384875.