Pattern of recurrence and prognostic factors in patients with pT1-3 N0 esophageal squamous cell carcinoma after surgery: analysis of a single center experience

This study was approved by the Ethics Committee of the Peking University First Hospital. We retrospectively reviewed data from 301 patients with esophageal cancer (EC) who underwent esophagectomy for curative intent between January 2011 and June 2017 in the Thoracic Department. The following were excluded: (1) patients with a pathological type of nonsquamous cell carcinoma; (2) patients with pathologically confirmed LNM; (3) patients with tumors that had a positive margin (R1 or R2); (4) patients who received neoadjuvant or adjuvant chemotherapy or chemoradiotherapy; (5) patients who died of non-neoplastic causes; and (6) patients who were lost during follow-up.

Preoperative examinations, which included a thorough physical examination, a thoracic computed tomography (CT) scan, a bone scan, an endoscopic examination and a biopsy, a cervical ultrasound, an abdominal CT scan or an ultrasound, a cardiopulmonary function test, and a hematological profile, were performed to rule out distant metastasis and to evaluate the feasibility of surgery. Positron emission tomography–computed tomography (PET-CT) scans were not routinely performed in our department. They were only used for the preoperative examination of patients with highly suspected metastases, such as lymphadenopathy (> 1 cm in diameter on a CT scan).

All operations were performed under general anesthesia and double lumen tracheal intubation by qualified surgeons. The surgery consisted of transthoracic esophagectomy and at least two field en-bloc lymphadenectomies (mediastinal and upper abdominal lymph nodes). Cervical lymph node dissections were only reserved for patients with suspected supraclavicular LNM before surgery. Gastric conduit was used for esophageal reconstruction in all patients, except one. The tumor was resected with a negative margin distance > 5 cm. For the resection of the upper ESCC, however, the distance of the proximal margin from the tumor might be < 5 cm, such that the proximal clearance was > 3 cm at least. There were two surgical approaches: (1) the right-sided transthoracic approach, where esophagectomy and mediastinal lymphadenectomy were performed through the right thoracic cavity. Stomach mobilization and abdominal lymph node dissection were performed at the abdominal stage. An anastomosis was made at the top right thoracic cavity or the left neck. The right-sided approach was used for cases of thoracic ESCC, regardless of the tumor location, except for those tumors located at the esophagogastric junction (EGJ). In the left-sided transthoracic approach, esophagectomy and thoracic and abdominal lymph node dissection were performed through the left thoracic cavity. An esophagogastric anastomosis was made underneath the aortic arch or at the left neck. The left-sided approach was used for cases of middle and lower ESCC and for tumors located at the EGJ. The extent of lymph node (LNs) dissection was as follows. During the left-sided procedure, the LNs at the paraoesophageal region (middle and lower paraoesophageal sites), the mediastinum (subcarinal, tracheobronchial, supradiaphragmatic, and posterior mediastinal sites), and the upper abdominal region (paracardial sites; LNs along the celiac, left gastric, and splenic arteries) were removed. During the right-sided procedure, in addition to the LNs that were dissected in the left-sided approach, the LNs along the bilateral recurrent nerves were also removed. After surgery, the specimens were examined by an experienced pathologist to determine whether LNM was present or absent and whether the surgical margin was free of the tumor (R0 resection). When the distance of the proximal margin from the tumor was less than 1 cm, the pathologist reported the precise distance between the surgical margin and the tumor.

The following clinicopathologic parameters were collected:

Outpatient reviews or telephone follow-ups were conducted every 3–4 months for the first 2 years and every 6 months thereafter. The follow-ups were conducted until October 2018 or recurrence. Examinations included chest CT scans and abdominal and cervical ultrasounds. The collected information included the presence or absence of recurrence, date of recurrence, and site of recurrence. The diagnosis of recurrence was based on reports from radiographic tests. The overall survival (OS) was calculated from the operation to the last follow-up or death. The recurrence-free survival (RFS) was calculated from the date of surgery to the first recurrence. Based on the location of the relapse, recurrence was divided into locoregional recurrence or distant metastasis. Locoregional recurrence was defined as a recurrence restricted to the anastomotic site, the area of the original tumor bed or regional lymph nodes (including the cervical, mediastinum, and upper abdominal LNs). Distant recurrence was defined as any recurrence with features beyond those of a locoregional recurrence.

SPSS 22.0 software (IBM Corporation, USA) was used for statistical analysis. Continuous variables were expressed as the mean ± standard deviation (SD) or median (range). Categorical variables were expressed as a percentage. All variables, including the demographic data, operative information and tumor characteristics, were analyzed. Kaplan–Meier methods were used to construct the RFS curves. The survival difference of each variable on RFS was analyzed by the log-rank test. The number of patients at risk was calculated for the beginning of each time period. Univariate and multivariate analyses were performed using the Cox proportional hazards regression model. Variables that had a significance level of P < 0.1 in univariate analysis were included in multivariate Cox regression analysis. All statistical tests were 2-sided, and a P-value < 0.05 was considered statistically significant.

There were 301 patients with EC who were treated with surgery during this period. The following were excluded: (1) patients with a pathological type of nonsquamous cell carcinoma (n = 8); (2) patients with pathologically confirmed LNM (n = 147); (3) patients with tumors that had a positive margin (R1 or R2) (n = 1); (4) patients who received neoadjuvant or adjuvant chemotherapy or chemoradiotherapy (n = 34); (5) patients who died of non-neoplastic diseases (n = 8); and (6) patients who were lost during follow-up (n = 2). As a result, 101 patients were enrolled in this study.

Of these patients, 78 were male (77.2%), with a median age of 64 years (range: 47–85 years). The mean S-CEA and S-SCC levels were 2.6 ± 1.5 μg/ml (range: 0.6–8.7 μg/L) and 2.5 ± 7.0 μg/ml (range: 0.3–70.0 μg/ml), respectively. The tumor was located in the upper third of the esophagus in 11 patients (10.4%), the middle segment of the esophagus in 58 patients (56.5%), and the lower third of the esophagus and the gastroesophageal junction in 32 patients (33.0%). Surgery was performed using left and right approaches in 44.3 and 55.7% of patients, respectively. Complications developed in 27.7% of patients after surgery. The TLN was 23 (8–56) for the entire study population. The median number of TLNs stratified by surgical approach was 23 (9–55) for the left-sided approach and 22.5 (8–56) for the right-sided approach, respectively. Among the 101 patients, 9 (8.9%) underwent cervical LN dissection or sampling, and the median number of removed LNs was 3 (1–9). All patients underwent thoracic and abdominal LN dissection, and the median numbers of removed LNs were 13 (1–41) and 9 (1–25), respectively. VI and NI were present in 5.9 and 14.9% of patients, respectively. The tumor grade was well differentiated (G1) in 16 patients (15.8%), moderately differentiated (G2) in 70 patients (69.3%), and poorly differentiated (G3) in 15 patients (14.9%). There were 51 patients (50.1%) with stage I disease and 50 patients (49.9%) with stage II disease. Among the 101 patients, 19 (18.8%) patients experienced recurrence. Compared with patients without recurrence, the prognosis of patients with recurrence was significantly worse (P < 0.001) (Fig. 1). The clinical characteristics of the patients are summarized in Table 1.

Follow-up was performed until October 2018 or recurrence. The median follow-up time for the 101 patients was 41 months (range: 4–92 months). For the 19 patients who developed recurrence, recurrence (17 patients, 89.5%) usually occurred within the first three years after surgery. Recurrence peaked within the first year (8 patients, 42.1%) and then dropped rapidly after the third year after surgery (1 patient per year). The most common pattern of relapse was local recurrence (n = 11, 57.9%), followed by distant recurrence (n = 7, 36.8%); one patient developed local and distant recurrence simultaneously. The details of the recurrences are presented in Table 2.

The results of univariate analysis indicated that tumor location (P = 0.011), presence of VI (P < 0.001), CEA level (P = 0.003), and tumor grade (P = 0.011) were significantly associated with the RFS of patients. The results of multivariate analysis using the Cox proportional hazards regression model with stepwise selection also demonstrated that the independent prognostic factors for decreased RFS in patients with node-negative disease were a tumor located in the upper chest (P = 0.007, odds ratio [OR] = 4.767), CEA ≥5 μg/ml (P = 0.007, OR = 5.466) and the presence of VI (P = 0.037, OR = 3.534). The details are summarized in Tables 3 and 4.

We estimated the 3- and 5- year RFS rates using the Kaplan–Meier method stratified by the aforementioned prognostic factors (Table 5). Poorly differentiated tumors had much worse prognosis than those with well- and moderately differentiated tumors in the log-rank test (P = 0.007) (Fig. 2a), although the degree of tumor differentiation was not an independent prognostic factor in multivariate analysis. In regard to the preoperative CEA level, a CEA level ≥ 5 μg/ml was strongly correlated with decreased RFS (P = 0.026) (Fig. 2b). Tumors located in the upper segment of the esophagus had a much worse prognosis than those located in the lower and middle chest (P = 0.006) (Fig. 3a). The presence of VI was also strongly correlated with the recurrence after surgery (P = 0.001) (Fig. 3b).