With the increasing application of endoscopic resection in T1-stage esophageal cancer, the ability to reliably predict nodal disease has become especially important to select the best treatment modality. In this study, a nomogram model was developed to predict LNM in patients with pT1-stage ESCC. Our nomogram model incorporated three clinicopathological factors, including tumor invasion depth, tumor differentiation, and tumor size, which were significantly associated with LNM. Based on the nomogram, a surgical treatment strategy was established according to the LNM risk stratification.
The depth of tumor invasion is probably the most important factor influencing LNM. The LNM rate of ESCC in T1a stage is 0%-15.4% [3–11]. In our series, the incidence of LNM in T1a patients was 7.3%, consistent with previous results. It is generally believed that the LNM rate of T1b ESCC is much higher than that of T1a ESCC, ranging from 15–51% [3–11]. Most previous studies have reported an incidence of about 20%, similar to the LNM rate of 22% for T1b tumors in our study. Some studies have divided the submucosa into three layers: sm1, sm2, and sm3[18]. Nentiwch et al. [3] showed that tumors in all three layers had a higher and similar incidence of LNM [19]. However, Akutsu et al. [4] revealed that the metastasis and relapse rate increased with infiltration depth, at 16%, 35%, and 62% for tumors in sm1, sm2, and sm3, respectively. The current study also confirmed tissue differentiation as an independent risk factor, which had been revealed by many previous reports [2, 5, 6, 8]. The risk of LNM increased with that of increasing tumor size [2. 5]. The best cutoff value for tumor size was revealed as 2.5 cm in this study.
The current study also confirmed cervical, thoracic, and abdominal LNM in pT1-stage ESCC with bidirectional and skipping metastasis along the esophagus [15, 20–21]. The most common LNMs were located at the recurrent laryngeal nerve and the left gastric artery [4, 22–23]. The extent of lymph dissection in surgery should respect the LNM topography, even in early-stage ESCC.
Thus far, a few researchers have developed nomogram models to predict the risk of LNM in esophageal cancer [6, 11, 24–26]. For example, Zheng et al. [11] recently developed a nomogram for the prediction of LNM in early-stage ESCC patients. They included four items, namely depth of tumor invasion, tumor differentiation, tumor size, and lymphovascular invasion (LVI), and also validated the model with another case cohort. The rate of LVI in our study was 7.2% lower than that of 12.8% in a previous study [11] and another study of 20.5% [5]. The low reporting rate may be the reason why LVI was not confirmed as an independent prediction factor in our study. An external validation was also performed using an independent case cohort from our cancer center.
Based on the nomogram model we constructed, LNM risk was stratified into different intensities starting from zero, for cases with well-differentiated T1a-stage tumors less than 2.5 cm in size, to the highest risk value, at 59%, for those with poorly differentiated T1b-stage tumors bigger than 2.5 cm in size. The nomogram model can be used to help clinical decision-making to determine when endoscopic resection is adequate, or surgery is needed. In early-stage ESCC patients with metastatic nodes in preoperative evaluation, the standard treatment involves neoadjuvant therapy followed by esophagectomy. In patients without metastatic nodes, esophagectomy or endoscopic resection should be chosen as the primary treatment, according to the risk stratification. If sufficient stratification information cannot be obtained, endoscopic resection should be considered as the first treatment option, allowing a subsequent decision on whether additional esophagectomy is needed. Clinical decisions also need to consider other clinical factors, including patient age, physical condition, willingness, socioeconomic factors, and the level of comprehensive knowledge of the surgeon or endoscopy physician.
The limitations of this study must be considered. First, the retrospective single institution study obviously may be accompanied by the potential risks of patient selection bias. Second, most patients underwent two-field lymphadenectomy without cervical lymph dissection. Third, the reported LVI rate is lower than that reported in the literature, which may have led to its exclusion as a risk factor in the nomogram model. Moreover, some studies have classified T1b tumors into three layers (sm1, sm2, and sm3). However, detailed data on classification were not obtained in the present study. In this group, insufficient lymph node dissection was performed in a few patients, which may underestimate the frequency of LNM.