BackgroundNeoadjuvant immune checkpoint inhibitor in combination with chemotherapy (nICT) or chemoradiotherapy (nICRT) has been tested in resectable esophageal cancer. Nevertheless, efficacy and safety for this new strategy have not been clearly demonstrated.Patients and methodsPubMed, Embase, Cochrane Library, Web of Science, and scientific meetings were searched for eligible trials until June 30, 2022. The primary outcome of interest was pathological complete response (pCR). The random-effect model was used for statistical analysis.ResultsTwenty-seven trials with 809 patients were identified. The estimated rates of pCR for nICRT and nICT were comparable (32.7%, 95% CI: 20.3%-45.1% vs 26.3%, 95% CI: 19.8%-32.8%; P = 0.37). As for safety, surgical resection rate, R0 resection rate, surgical delay rate, and surgical mortality rate were similar between nICRT and nICT, while more grade ≥3 treatment-related adverse events were observed for nICRT (52.6%, 95% CI: 30.7%-74.5% vs 19.9%, 95% CI: 8.8%-31.0%; P = 0.01). In subgroup analysis, nICRT achieved higher pCR rate compared to nICT (56.2%, 95% CI: 41.0%-71.3% vs 27.2%, 95% CI: 20.2%-34.1%; P < 0.001) for squamous cell carcinoma (SCC) but adenocarcinoma. In patients receiving nICT, PD-L1 expression CPS ≥1 showed higher pCR rate compared to CPS <1 (51.3%, 95% CI: 41.4%-61.2% vs 26.6%, 95% CI: 8.6%-44.5%; P = 0.02); regimen of paclitaxel plus carboplatin/cisplatin (PC/TP) and 3-4 cycles of nICT did not lead to an significantly improved pCR rate compared to other chemotherapy regimens and 2 cycles of nICT, respectively, despite without increased toxicity.ConclusionBoth nICT and nICRT achieved promising pCR rates with acceptable tolerability, and nICRT was likely to have more antitumor efficacy compared to nICT for patients with SCC. PD-L1 status seemed to be predictive of pCR in patients receiving nICT; pCR rate did not appear to be greatly affected by CT regimen and increasing cycles of nICT.