Prognostic Significance of Circulating Tumor Cells in Patients with Esophageal Cancer Who Attained Pathological Complete Response after Neoadjuvant Therapy

Abstract

Background

Pathological complete response (pCR) refers to the absence of residual tumor cells after neoadjuvant chemoradiotherapy (nCRT) and is associated with favorable outcomes. However, 25%−33% of such patients still experience disease recurrence. This study aimed to evaluate the prognostic value of circulating tumor cells (CTCs) for risk stratification in patients with esophageal carcinoma (EC) who achieved pCR following nCRT.

Methods

We measured post-nCRT CTC counts in the peripheral blood of patients with EC who attained pCR after undergoing nCRT. The impact of CTC counts on survival was then analyzed, accounting for the potential confounding effects of established clinical and pathological risk factors.

Results

The study included 24 patients, of whom 18 (75%) had detectable CTCs. The mean CTC concentration was 5.1 per mL of blood. Univariable Cox regression analysis showed that CTC count was the only independent predictor of disease-free survival (hazard ratio, 1.113; 95% confidence interval, 1.008−1.229; p=0.034). Receiver operating characteristic curve analysis determined an optimal cutoff value of 4.5 CTCs per mL. The 3-year disease-free survival rate was significantly higher in the low-CTC group (92.9%) than in the high-CTC group (50.0%, p=0.032).

Conclusion

In patients with EC who achieve pCR following nCRT, CTC quantification may help identify those at increased risk of poor survival outcomes.

Introduction

Esophageal cancer (EC) is an aggressive malignancy that generally carries a poor prognosis [1]. Evidence indicates that neoadjuvant chemoradiotherapy (nCRT) followed by surgery significantly improves survival in patients with locally advanced EC compared with surgery alone [2,3]. Notably, patients who respond favorably to nCRT derive the greatest benefit from this approach. Pathological complete response (pCR), which is defined as the absence of detectable cancer cells in the final pathology specimen, is widely accepted as a key indicator of a favorable response to neoadjuvant therapy [4].

Although patients who achieve pCR typically experience improved outcomes, a cure is not assured. Approximately 25% to 33% of these patients eventually develop disease recurrence [5-7]. While the recent CheckMate 577 trial demonstrated the benefit of adjuvant immunotherapy in reducing recurrence rates among EC patients treated with nCRT, the study population consisted specifically of non-pCR patients [8]. Thus, identifying reliable prognostic markers in both pCR and non-pCR patients is essential to guide treatment decisions and tailor therapy to individual risk profiles [9].

Over the past decade, circulating tumor cell (CTC) assessment has emerged as a promising biomarker for detecting residual disease after curative-intent treatment [10]. While available data support its prognostic value in patients undergoing upfront surgery, its utility in stratifying prognosis among those achieving pCR after nCRT remains uncertain [11]. We hypothesized that post-nCRT CTC assessment could identify a subgroup of high-risk patients within the pCR population. Therefore, we measured CTCs following nCRT and evaluated their prognostic value, while accounting for the possible confounding effects of established clinical and pathological risk factors.

Methods

Study design

This study was conducted at Chang Gung Memorial Hospital (CGMH), Linkou (Taiwan), where all visits and procedures took place. Eligible participants were patients diagnosed with locally advanced EC who had received nCRT as their first-line treatment, followed by scheduled esophagectomy. CTCs were measured once, immediately prior to post-nCRT surgery, and within 12 weeks of completing nCRT. From this group, we selected patients who achieved pCR after surgery and excluded those who experienced surgical mortality or for whom follow-up data were unavailable. All participants provided written informed consent, and the study was approved by the CGMH Institutional Review Board (IRB reference number: 2017022222B0).

All patients were staged according to the 8th edition of the American Joint Committee on Cancer staging system. Pretreatment tumor length was defined as the greatest longitudinal extent of the lesion measured on barium-swallow radiography when available, or otherwise on computed tomography imaging.

Neoadjuvant chemoradiotherapy

In accordance with our institutional guidelines, patients with locally advanced EC located in the non-cervical region (classified as cT2–4aNany or T1N+) who were deemed medically fit for surgery were offered nCRT followed by surgical resection. Two nCRT regimens were used: PF (fluorouracil/cisplatin regimen) and CROSS (paclitaxel/carboplatin regimen).

PF protocol

Patients received 5-fluorouracil (5-FU) at 1,000 mg/m²/day as a continuous 96-hour infusion on days 1–4 and 29–33, combined with cisplatin 75 mg/m² administered intravenously over 3 hours on days 1 and 29. Concurrent radiotherapy was delivered at 45 Gy from day 1 to day 35 in 1.8 Gy daily fractions, 5 days per week. Treatment fields encompassed the entire esophagus, regional lymphatics, and lymphatic basins in the celiac, pericardial, and supraclavicular regions, provided organ-at-risk constraints were observed.

CROSS protocol

Patients received weekly carboplatin (target area-under-the-curve, 2 mg/min/mL) and paclitaxel 50 mg/m² for 5 consecutive weeks, along with concurrent radiotherapy to 45 Gy in 1.8 Gy fractions, 5 fractions per week.

Surgical procedures and postoperative adjuvant therapy

Patients who completed clinical restaging were eligible for surgery if they met the following criteria: American Society of Anesthesiologists score of 3 or lower, absence of tracheoesophageal fistula, and no evidence of metastasis to solid organs or distant lymph nodes. Surgery involved a right-sided transthoracic esophagectomy, using either the Ivor Lewis procedure with intrathoracic anastomosis or the McKeown procedure with cervical anastomosis. All patients underwent 2-field lymph node dissection.

At CGMH, adjuvant therapy (AT) after nCRT and surgery is recommended for patients with a non-R0 resection or for those with an R0 resection but classified as pathological non-responders (defined as having a postoperative T stage equal to or greater than the pretreatment T stage, or persistent nodal disease). AT regimens were determined jointly by medical oncologists, radiation oncologists, and the patient, based on (1) prior nCRT-related toxicity, (2) current general condition, and (3) presence of comorbidities.

Histopathological analysis

Surgical specimens were opened lengthwise and fixed in 10% formalin overnight. In cases where residual tumors were identified, representative sections were examined in detail to determine the deepest point of invasion and the anatomical relationship between the esophagus and stomach. If no visible tumors were present, ulcerated or fibrotic areas were examined, and representative sections were processed for pathological assessment. All slides were stained with hematoxylin and eosin. Diagnosis of pCR was made by a single board-certified pathologist (Dr. Chi-Ju Yeh) according to the CAP criteria. pCR was defined as the complete absence of viable tumor cells in all operative pathological specimens, including both the primary tumor site and all resected lymph nodes.

Quantification of circulating tumor cells

Blood samples were collected from all patients immediately before surgery. Enumeration and gating for CTC detection followed previously described methods [12,13]. In brief, blood was obtained via venipuncture, discarding the initial 2–4 mL to minimize contamination by skin epithelial cells. Of the 8 mL collected, 4 mL were reserved as backup, and the remaining 4 mL were used for CTC quantification. Samples were first treated with red blood cell lysis buffer (155 mM NH₄Cl, 14 mM NaHCO₃, and 0.1 mM EDTA at a 10:1 ratio) and then depleted of CD45-positive leukocytes using the EasySep Human CD45 Depletion Kit (catalog number: 18259; STEMCELL Technologies Inc.). Immunomagnetically enriched specimens spiked with OECM-1 cells—a human squamous cell carcinoma cell line (Food Industry Research and Development Institute)—were labeled with an Alexa Fluor 488-conjugated anti-EpCAM antibody (catalog number: 5198S; Cell Signaling Technology). Hoechst 33342 dye (catalog number: 62249; Thermo Fisher Scientific) was used for nuclear staining. The gating strategy consisted of several critical steps to ensure accuracy. First, 2 mL samples containing Hoechst-positive cells were gated from all events to exclude debris and fragments from the negative selection process. Singlet cells were then selected to prevent false positives from cell aggregates. CD45-positive cells were excluded again to minimize residual white blood cell contamination. Before enumeration, EpCAM-positive cells (and corresponding isotype-positive cells) were independently gated. CTCs were defined as CD45-negative cells that were positive for both EpCAM and Hoechst 33342, and were quantified on a CytoFLEX flow cytometer (Beckman Coulter). The CTC count was calculated as the number of EpCAM-positive cells minus the number of cells gated using the isotype control.

Statistical analysis

Continuous variables are reported as mean (standard deviation) or median (interquartile range, IQR), and categorical variables as proportions. Data collected up to June 2025 were included in the analysis. Overall survival (OS) was defined as the time from initiation of nCRT to death. Recurrence was identified through surveillance imaging, without mandatory biopsy confirmation. Disease-free survival (DFS) was defined as the interval from surgery to documented recurrence or death. Censoring was applied at the date of last follow-up. Given the limited number of events, only univariable Cox proportional hazards models were used to minimize the risk of overfitting. Results are presented as hazard ratios (HRs) with 95% confidence intervals (CIs). Time-dependent receiver operating characteristic (ROC) curves for 3-year DFS were used to identify the optimal cutoff value, determined as the point with the highest Youden index (true positive rate minus false positive rate across all possible cutoffs). Survival probabilities were estimated using Kaplan-Meier curves and compared with the log-rank test. All statistical analyses were performed in JASP ver. 0.19.3.0 (https://jasp-stats.org/) and cross-validated in IBM SPSS ver. 22.0 (IBM Corp.), except for time-dependent ROC curve analysis, which was conducted in Python ver. 3.11 (pandas 2.2.0, scikit-learn 1.4.2, lifelines 0.28.0; https://www.python.org/). All tests were 2-sided, with a significance threshold of p<0.05.

Results

Patient characteristics

The study included 24 patients with EC who achieved pCR. Most participants were male (22/24; 91.7%), and the majority presented with stage II or III disease (71%) as their predominant clinical stage. The CROSS regimen was the most frequently used neoadjuvant approach, administered to 20 patients, while the PF protocol was given to 4 patients. The general characteristics of the study population are summarized in Table 1.

Table 1.

General characteristics of the study patients (n=24)

Characteristic	Value
Male sex	22 (91.7)
Age (yr)	57.50±8.753
Histologic type
Squamous cell carcinoma	23 (95.8)
Adenocarcinoma	1 (4.2)
Tumor location
Upper third	4 (16.7)
Middle third	10 (41.7)
Lower third	10 (41.7)
Tumor length (cm)	4.938±3.359
Clinical staging before nCRTa)
II	4 (16.7)
III	12 (50.0)
IVa	8 (33.3)
Clinical Ta)
T1	1 (4.2)
T2	5 (20.8)
T3	16 (66.7)
T4	2 (8.3)
Clinical Na)
N0	1 (4.2)
N1	8 (33.3)
N2	10 (41.7)
N3	5 (20.8)
Clinical Ma)
M0	24 (100.0)
Chemotherapy regimen
Fluorouracil/cisplatin	4 (16.7)
Paclitaxel/carboplatin	20 (83.3)
Response to nCRT
Partial response	16 (66.7)
Stable disease	7 (29.2)
Disease progression	1 (4.2)
No. of dissected nodes	24.25±10.27
Circulating tumor cell countb)	5.104±6.222

Values are presented as number (%) or mean±standard deviation.

nCRT, neoadjuvant chemoradiotherapy.

^a)According to the American Joint Committee on Cancer staging criteria, eighth edition, before nCRT. ^b)Number of cells per mL of blood.

Survival outcomes

After a median follow-up of 73.7 months (IQR, 66.7–79.8 months), 15 patients were alive, while 9 had died during the study period. Among the deceased, 4 deaths were due to cancer recurrence, 2 were attributed to treatment-related complications (pneumonia), and 2 were unrelated to EC: 1 from COVID-19 (coronavirus disease 2019) infection and 1 from progression of a second primary head and neck cancer. The 3-year OS rate was 87.5%, and the 5-year OS rate was 61.0%. Disease recurrence occurred in 7 patients. Of these, 1 experienced isolated local recurrence (LR), 2 had both LR and distant recurrence (DR), and the remaining 4 presented with DR alone.

Prognostic impact of circulating tumor cells

Post-nCRT blood samples revealed detectable CTCs in 18 patients (75%), with a mean concentration of 5.1 cells per mL. Univariable Cox regression analysis showed that CTC count was the only independent predictor of DFS (HR, 1.113; 95% CI, 1.008–1.229; p=0.034) (Table 2). To increase the translational relevance of these findings, we converted the CTC count into an ordinal variable using ROC curve analysis for time-to-event data. Table 3 lists the corresponding sensitivity and 1–specificity values calculated for various CTC cutoffs (range, 0–23.5). Fig. 1 shows the ROC curve for 3-year DFS based on these values, summarizing multiple distinct CTC thresholds. The area under the ROC curve (AUC) was 0.72, indicating moderate accuracy for the CTC count. The optimal cutoff value was identified as 4.5 cells per mL, where both the AUC and diagnostic performance were highest. In the low CTC group, the median DFS was not reached, whereas in the high CTC group it was 32.9 months. The 3-year DFS rate was significantly higher in the low CTC group compared with the high CTC group (92.9% vs. 50.0%, p=0.032) (Fig. 2A). The median OS was not reached in the low CTC group, while it was 43.4 months in the high CTC group. The 3-year OS rate was numerically higher for the low CTC group compared with the high CTC group (100% vs. 70.0%), although this difference was not statistically significant (p=0.235) (Fig. 2B). Supplementary Table 1 provides the general characteristics of patients in the high and low CTC groups.

Table 2.

Univariable Cox regression analysis for the predictors of overall survival and disease-free survival

Risk factor	Overall survival			Disease-free survival
	HR (95% CI)	p-value		HR (95% CI)	p-value
Age	0.976 (0.907−1.050)	0.514		0.995 (0.931−1.063)	0.871
Tumor length	1.133 (0.934−1.374)	0.204		1.058 (0.870−1.287)	0.572
Clinical stage		0.357			0.248
II+III	Reference			Reference
IVa	1.866 (0.495−7.033)			2.027 (0.612−6.713)
Chemotherapy regimen		0.719			0.971
PF	Reference			Reference
CROSS	0.749 (0.155−3.619)			1.029 (0.222−4.770)
No. of dissected nodes	0.936 (0.861−1.017)	0.117		0.972 (0.908−1.041)	0.423
Circulating tumor cell count	1.095 (0.987−1.214)	0.087		1.113 (1.008−1.229)	0.034

HR, hazard ratio; CI, confidence interval; PF, fluorouracil/cisplatin regimen; CROSS, paclitaxel/carboplatin regimen.

Table 3.

Receiver operating characteristic curve analysis (true- and false-positive rates) of the circulating tumor cell count at various cutoff points

Circulating tumor cell count cutoff point	True-positive ratea)	False-positive rateb)
0.0	1.0	1.0
1.0	0.8333	0.7222
1.5	0.8333	0.6111
2.0	0.8333	0.4444
3.0	0.8333	0.3333
4.5	0.8333	0.2778
6.0	0.5	0.2778
10.0	0.3333	0.2222
17.0	0.3333	0.0
22.5	0.1667	0.0
23.5	0.0	0.0

^a)Sensitivity. ^b)1–specificity.

Fig. 1.

Receiver operating characteristic (ROC) curve analysis, designed for time-to-event data, of circulating tumor cell count. DFS, disease-free survival; AUC, area under the ROC curve.

Fig. 2.

Disease-free survival (A) and overall survival (B) of patients with high (≥4.5/mL) and low (<4.5/mL) circulating tumor cell (CTC) counts.

Discussion

This study examined the prognostic significance of CTCs in patients with EC who achieved pCR following nCRT. Two key findings emerged. First, despite achieving pCR, 75% of patients still had detectable CTCs, suggesting that complete tumor eradication may not have occurred. Second, patients with elevated CTC counts (≥4.5/mL) experienced significantly reduced DFS. These findings indicate that CTC quantification could serve as a useful stratification tool for personalized prognostic assessment in pCR patients after nCRT. Although current guidelines and available evidence do not recommend routine AT for this patient population [14], individuals with elevated CTC counts may warrant more intensive surveillance and could be considered for AT in carefully selected or investigational settings. Further case-control studies are required to determine whether AT provides a survival benefit in this subset.

In recent years, liquid biopsy approaches—such as CTCs and circulating tumor DNA (ctDNA)—have emerged as potential biomarkers for residual disease following upfront esophagectomy [15]. However, the utility of CTCs and ctDNA in patients treated with nCRT remains debated. Our research group has previously demonstrated that evaluating CTC counts alongside clinical response assessment may help identify EC patients who respond favorably to nCRT [12], even when findings contrast with major histopathological response rates. Regarding post-nCRT prognostic stratification, prior studies have shown an association between elevated post-neoadjuvant CTC counts and an increased risk of early recurrence [16]. However, those studies did not specify the proportion of patients who achieved pCR, and none of the participants received radiotherapy; instead, they underwent only neoadjuvant chemotherapy [16]. We believe that the presence of CTCs in pathological complete responders is clinically relevant, as it suggests residual disease, particularly in patients with high CTC counts. Conversely, while patients with minimal or absent CTCs generally had more favorable outcomes, recurrences still occurred in 2 cases within the low-CTC group. This may reflect the limitations of current CTC detection methods. The Food and Drug Administration-approved approach, which uses anti-EpCAM and CK antibodies, can identify only epithelial CTCs and thus fails to detect those lacking epithelial markers. This limitation is significant because tumor heterogeneity can lead to underestimation of CTC counts. Cells that have undergone the epithelial–mesenchymal transition (EMT) may acquire stem cell–like properties, enabling them to escape hematologic detection by downregulating epithelial markers and upregulating mesenchymal markers, potentially resulting in false-negative results. To address these limitations and develop more accurate predictive models, future research should focus on dynamic CTC monitoring and the characterization of CTCs with mesenchymal phenotypes. Incorporating EMT-related CTC analysis could provide valuable prognostic insights and improve risk stratification [17].

This study has several limitations. First, the small sample size and absence of external validation substantially limit the robustness of our findings. Accordingly, these results should be regarded as preliminary evidence that requires confirmation in larger cohorts before informing clinical practice. Second, because most participants had esophageal squamous cell carcinoma, the findings may not be generalizable to other histological subtypes. Third, our analysis quantified CTCs only in the post-nCRT phase, leaving gaps in understanding dynamic fluctuations of CTC counts over time. Future research should assess CTCs at multiple stages of treatment to develop more precise predictive models. Fourth, although we identified a cutoff value of 4.5 CTCs/mL based on ROC curve analysis, comparisons with other studies should be made with caution due to substantial methodological heterogeneity in CTC detection. Our approach—immunofluorescence-based flow cytometry with CD45 depletion and EpCAM labeling—represents a semi-automated, label-dependent method distinct from commercially available systems such as CellSearch (https://www.cellsearchctc.com/). While such methodological differences prevent direct comparison of absolute cutoff values across studies, a consistent observation remains: elevated CTC counts, regardless of detection platform, are associated with poor oncologic outcomes [12,13,18-21]. This reinforces the idea that CTC enumeration can serve as a surrogate marker for minimal residual disease in the perioperative setting. Finally, because of the small number of events in this cohort, only univariable analyses were performed to avoid overfitting, which restricted adjustment for potential confounders. Future investigations should include prospective, multi-center studies that track CTC counts across multiple time points (pre-, mid-, and post-nCRT, pre-surgery, and during follow-up) in both squamous cell and adenocarcinoma patients. Standardized, cross-validated assays would improve reproducibility and enable the establishment of uniform cutoff values. In addition, case-control studies should test whether intensified surveillance or AT in patients with persistently high CTC levels can improve survival outcomes.

In conclusion, in patients with EC who achieve pCR following nCRT, CTC quantification may help identify individuals at increased risk of poor survival outcomes. If validated in larger cohorts, this approach could be used to pinpoint patients at heightened risk of post-nCRT recurrence and inform decision-making regarding the potential use of AT.

Supplementary materials

Supplementary materials can be found via https://doi.org/10.5090/jcs.25.055. Supplementary Table 1. Comparison of patients with high and low CTC counts.