Background: Perioperative chemotherapy (ChT) is a standard for operable gastroesophageal adenocarcinoma (GEA). The addition of immune checkpoint inhibitors (ICIs) has improved pathological complete response (pCR) with unclear survival advantages in an unselected population. In the advanced setting, PD-L1 is a validated biomarker of response to ICIs. We aimed to assess its predictive role in the early stages wherein remains unexplored. Methods: We conducted a systematic review of randomised clinical trials (RCTs) of neoadjuvant or peri-operative ChT +/- ICIs in non-metastatic GEA, published between 2006-2024. Biomarker-unselected phase 2/3 or phase 3 RCTs with an active comparator were selected. Odds ratios (ORs) for pathological response (defined as pCR, ypT0, or ypT≤1), the primary outcome of this meta-analysis, were extracted and analysed using a random-effect meta-analysis. Subgroup analysis (ICI-ChT versus ChT) was performed to estimate the pooled effect of peri-operative ICIs on pathological response. Sensitivity analyses were performed to account for between-study heterogeneity. We then extracted pCR data by PD-L1 (cutoffs 1, 5, 10) from phase 2 trials of neoadjuvant ICIs+ChT or chemoradiotherapy (CRT) published between 2006-2024. Weighted Pearson trial-level correlation and a meta-regression model were used to assess the relationship of PD-L1 with pCR. Results: Nine RCTs were eligible for the analysis, one of which had two experimental arms contributing to two comparisons. Of 5359 included patients, 1302 received ICI-ChT and 4057 systemic ChT. ICI-ChT showed a significant improvement in pathological response (pooled OR 2.90, 95% CI 1.81-4.63, p<0.001) with a larger effect compared to ChT-only trials (pooled OR 1.40, 95% CI 1.10-1.79, p=0.007). Results were consistent in sensitivity analyses of pCR-only, non-Asian, and since-FLOT studies. Fourteen studies were included in the phase 2 trials dataset (n=10 ICI+ChT and n=4 ICI+CRT) accounting for 708 patients receiving ICIs. PD-L1 showed a moderate positive correlation with pCR, stronger for higher cut-offs (weighted R for cut-offs 1, 5, and 10: 0.33, 0.51, and 0.65, respectively) and after exclusion of CRT trials (weighted R for cut-offs 5 and 10: 0.67 and 0.99, respectively). In a mixed-effect meta-regression model, PD-L1 ≥5 was a significant predictor of response (effect size [ES]: 0.71, 95% CI, 0.29-1.13, p=0.001), even after accounting for the backbone treatment (ChT versus CRT) (ES: 0.60, 95% CI, 0.20-1.00, p=0.003). Conclusions: Neoadjuvant ICI-ChT consistently improves pathological response in operable GEA, particularly in PD-L1-positive tumours. PD-L1 ≥5 appears a significant biomarker of pCR in non-metastatic GEA, supporting stratification by PD-L1 and biomarker-selected trials for evaluation at the individual patient level.