Background: For ESCC patients who receive nCRT, high pathological complete response (pCR) rates could be achieved. An active surveillance strategy has been proposed for patients with clinically complete response (cCR) after nCRT. However, clinical assessment may misclassify non-pCR as cCR. This study aimed to investigate the value of ctDNA in predicting tumor response and residual disease after nCRT in ESCC patients. Methods: This was a side study of the prospective, diagnostic pre-SINO trial (NCT03937362). After completion of nCRT, all patients underwent one or two clinical response evaluations (CREs) before planned surgery. The first (CRE‐1) was 4–6 weeks after completion of nCRT. A second examination (CRE‐2) was 10–12 weeks after nCRT for patients without histological evidence of residual tumor at CRE1. Serial ctDNAs were analyzed by ultra-deep unique molecular identifiers -based next-generation sequencing using 168 cancer-related genes at three time points: at baseline (B0), CRE1, CRE2. pCR was defined as ypT0N0M0. cCR was defined as no evidence of residual tumor during CRE1 & CRE2 in endoscopic biopsies and endoscopic ultrasonography with fine-needle aspiration cytology. We correlated quantitative ctDNA levels and ctDNA presence with pCR and major locoregional residual disease (MLRD) ( > 10% residual carcinoma or any residual nodal disease). Results: At the time of this initial analysis, 62 consecutive ESCC patients received per-protocol treatments had sufficient sample to test ctDNA. The pCR rate was 32.3% (20/62). ctDNA was detectable in 73.77% at B0, 27.12% at CRE1 and 17.24% at CRE2. Maximum allelic fraction (maxAF) with a > 2-fold decrease post-nCRT (B0-CRE1) was seen in 67% patients with pCR, and 65.2% patients without MLRD. At B0, the high frequency mutated genes detected were TP53 (74%), PIK3CA (13%), CCND1 (11%), FGF19 (11%), FGF3 (11%), and FGF4 (11%). Baseline TP53 mutation status was significantly associated with pathological evaluations. A combining model, which incorporated both CRE1&2 positivity (cCR) and TP53 missense mutation at B0, was developed and demonstrated lower false-negative rate (FNR) in predicting non-pCR (4.9%, 2/41) and MLRD (2.8%, 1/36) compared with clinical assessment only (14.3%, 6/42) and (13.5%, 5/37), respectively. Of note, the FNR of combining clinical assessment and dynamic ctDNA change (continuous drop in maxAF ≥80% [B0-CRE1-CRE2]) was 7.3% (3/41) for non-pCR and MLRD (5.6%, 2/36). Conclusions: ctDNA testing combining with clinical assessment further optimized the tumor response and residual disease evaluations after nCRT for ESCC. ctDNA provides complementary information of response to nCRT and might become useful in an active surveillance strategy for the management of ESCC. Tumor-informed ctDNA analysis designed to track patient-specific somatic variants is ongoing. Clinical trial information: NCT03937362.