Esophageal carcinoma is a common malignant obstructive esophageal disease with an incidence and mortality of 604 and 544 thousand worldwide and ranks seventh and sixth in terms of incidence and mortality, respectively (1). It is histologically classified into esophageal adenocarcinoma, esophageal squamous cell carcinoma (ESCC), and other subtypes, whereas greater than 90% of esophageal cancer cases are ESCC in Asia countries (2). Although current treatments for ESCC can acquire favorable outcomes in given conditions, the 5-year survival rate is only approximately 29.7% due to lymph node metastasis (LNM) being affected at a rate of approximately 50% (3).

The recurrent laryngeal nerve (RLN) lymph nodes (LNs) are the crucial indicator during esophagectomy, which may decide the necessity of cervical lymphadenectomy in some previous studies (4, 5). The RLN LNM in patients with ESCC is frequently observed at a rate of 18–63% in the previous studies (6–13). Routine esophagectomy with three-field lymph node resection for all ESCC patients is still controversial due to the balance of complication and survival (9, 14). Additionally, the status of RLN LNs was demonstrated to be closely associated with prognosis (15, 16). Hence, the evaluation of the status of RLN LNs using an accurate prediction model before an operation could be valuable to conduct the optimal treatment and improve the prognosis for ESCC patients.

Previous studies reported the risk factors for RLN LNM in esophageal cancer, but the predictive value is limited for a single risk factor (4, 17, 18). Although several reports of a prediction model is based on clinicopathological characteristics for RLN LNM, the performance of a preoperative prediction model needs further evaluation (6, 19, 20). This retrospective study was conducted to assess the preoperative clinical characteristics of ESCC patients and to develop and validate a preoperative nomogram for predicting RLN LNM. Furthermore, we first evaluated the prognostic value of the nomogram for overall survival.

Considering the significant value of RLN LNs status of patients with ESCC for treatment guidance, we sought to investigate the independent risk factors and construct a preoperative model for RLN LNM and survival prediction. In this study, the following findings were revealed. (a) Clinical T stage, bioptic tumor differentiation, endoscopic tumor length, and preoperative CEA were independent risk factors for RLN LNM. (b) The nomogram developed by the preoperative characteristics could predict the RLN LNs status accurately, which was validated internally and externally. (c) The risk of RLN LNM was significantly associated with the prognosis of patients with ESCC. This nomogram may help to make a more rational clinical treatment decision.

Since the RLN is a common metastasis site, RLN LNs are routinely resected in esophagectomy. However, surgical complications increase RLN injury and permanent RLN palsy with rates of 36 and 12%, respectively (9). Current evidence confirms the effects of RLN LNM on the 5-year survival of ESCC patients at a rate of only 21.7% (23). Other studies also indicated that RLN LNM was a predictor of worse survival and mortality (7, 23, 24). These findings emphasized the necessity of RLN LNM prediction and the balance between extended lymph node dissection and RLN-related complications.

Currently, radiological examinations are usually used to diagnose RLN LNM status before surgery, like CT and positron emission tomography (PET)/CT. In CT images, a diameter of RLN LNM could reflect the status of RLN LNM and short diameter RLN LNs greater than 10 mm are usually judged as metastasis. However, previous studies demonstrated that the cut-off of the short diameter of RLN LNs to diagnose metastasis was shorter than 10 mm, like 6.5 mm (19) and 7.5 mm (6). Additionally, it is well-known that using a CT image to distinguish benign inflammation of LNs from malignant metastasis and to diagnose metastasis in LNs less than 5 mm is difficult. However, a previous study reported that the LNM diameter less than 5 mm accounted for a relatively large proportion (25). These limitations make CT insufficient for preoperative diagnosis of RLN LNM. Although PET/CT is regarded as a more accurate method to diagnose LNM than CT, it faces low sensitivity and is limited in small LNM (26, 27). Therefore, accurate RLN LNM prediction needs to combine with preoperative clinical characteristics of patients with ESCC.

The depth of invasion is important in the staging of esophageal carcinoma. In the different infiltrating stratifications of esophageal carcinoma, there are completely different lymphatic drainage systems, namely, lymphatic vessels in the submucosa containing transverse and longitudinal vessels (28). When the tumor infiltrates the submucosa or deeper, it may further metastasize to the RLN LNs through longitudinal lymphatic vessels (29). Longfei et al. (30) reported that tumor invasion depth significantly influenced RLN LNM. Additionally, almost all relevant previous studies showed that the T stage was associated with RLN LNM (19, 20). In our study, the clinical T stage was regarded as an independent risk factor for RLN LNM, consistent with the current understanding that a higher probability of RLN LNM generally occurs in advanced T-stage ESCC (7).

Existing definite evidence has shown that a poor degree of postoperative tumor differentiation contributes to a high rate of RLN LNM (30, 31). Tumor cells with poor differentiation are similar to immature tissue, where the tumor is highly malignant. In this case, the probability of the transition increases two-fold. Yu et al. (20) reported that well, moderately, and poorly differentiated ESCC had RLN LNM rates of 10.35, 14.15, and 45.46%, respectively. However, our study first introduced bioptic tumor differentiation as one of the potential risk factors which may be more rational for predicting RLN LNM before surgery. In this current study, the well, moderately, and poorly bioptic tumor differentiation of ESCC presented RLN LNM rates of 19.2% (34 of 177), 29.1% (60 of 206), and 63.8% (30 of 47), respectively, which were roughly consistent with previous literature (19, 20).

The present study also found that an endoscopic tumor length more than 3 cm was considered to be related to an increased probability of RLN LNM compared to a smaller tumor. Eloubeidi et al. (32) reported similar findings and suggested that tumor size should be suffixed in the T stage as a modification of the TNM staging system. A previous study reported an optimal cut-off point of 3.5 cm, demonstrating that larger tumors were related to a higher positive LNM, including RLN LNM (33).

In the current study, we found that preoperative CEA level was associated with RLN LNM as shown in the multivariable analysis. However, CEA was reported as having no statistical relationship with RLN LNM in a previous study (20). The reason why it has differed from our result may be that they set CEA as categorical data with a cut-off of 5 ng/ml. Although no specialized literature has confirmed the relationship between CEA and RLN LNM, CEA was shown to be correlated with LNM (34). Tetsuro et al. (35) found that CEA mRNA in the blood is useful for early detection of recurrence in esophageal cancer. Additionally, CEA could be a noninvasive useful diagnostic tool if it can effectively predict RLN LNM since endoscopic ultrasonography, computed tomography, and positron emission tomography cannot exclude RLN LNM completely (36).

Several studies reported the prediction model to estimate the risk of RLN LNM for patients with ESCC. Liu et al. (19) and Yu et al. (20) developed the prediction model for RLN LNM. However, both models are based on the postoperative characteristics of patients with ESCC, which limited the application of this model. Recently, Zhang et al. (6) reported a preoperative nomogram to predict left and right RLN LNM, but their nomogram did not include several critical clinical characteristics like clinical T stage and tumor differentiation. Moreover, all previous studies lacked internal or external validation, which may result in inaccurate performance assessment. In the current study, we included four preoperative risk factors to construct this model. The discrimination, calibration, and clinical usefulness of the model highlighted its performance. Finally, the bootstrap method and hold-out method (external validation cohort) were used to validate this model. The AUCs of our model in the internal and external validation was greater than 0.750, indicating that the nomogram had acceptable discrimination for predicting RLN LNM. Additionally, the predicted probability and actual probability of RLN LNM were roughly equal, and the curve was close to the diagonal in the calibration curve, indicating a good prediction effect. DCA further demonstrated that the nomogram outperformed any single risk factor in predicting RLN LNM in terms of clinical usefulness. This nomogram can be easily applied in clinics with visualization. For example, consider a patient with clinical T1–2 stage (0 points), G3 (22 points), the endoscopic tumor length ≥3 cm (15 points), and the preoperative CEA was 6 ng/ml (30 points). The variables listed above yielded a score of 67 points for this patient, and the corresponding risk of RLN LNM was close to 70% ( Supplementary Figure 1 ).

In our study, we further confirmed that RLN LNM was significantly associated with the overall survival of ESCC patients. Longfei et al. (30) reported that patients with RLN LNM had worse overall survival and disease-free survival at any stratification. To the best of our knowledge, our study is the first to assess the prognostic value of the preoperative prediction model of RLN LNM in patients with ESCC. We found that the predicted RLN LNM risk of the nomogram had a comparable prognostic value with RLN LNM, which further demonstrated the great effectiveness of this nomogram to aid in clinical treatment decision making. After adjusting other preoperative factors, the predicted RLN LNM risk is still significantly associated with overall survival. Due to the aim of preoperative prediction, we did not analyze the influence of the adjuvant treatment or other postoperative factors. Theoretically, the patients with high-risk are more likely to diagnose RLN LNM, which may impact the postoperative prognostic factors like adjuvant treatment and further cause a worse overall survival.

The limitations of this current study are presented as follows. First, this study design was retrospective, and there was some inevitable selection bias. A further randomized controlled trial may be performed in our further study. Second, although we performed this study at two centers, the relatively small sample size limited the application of this nomogram. Therefore, investigations with a large sample size are warranted in the future. Third, important endpoints, namely, recurrence and disease-free survival, were not evaluated in this study. However, these endpoints are vital in the outcomes of malignant tumors. We analyzed the overall survival of ESCC patients, which is one of the most important endpoints, and our study can be considered as a starting point for subsequent studies. Fourth, when evaluating the clinical T stage, only limited patients executed PET/CT examination in our current cohort due to high cost. We performed enhanced CT, endoscopic ultrasonography and ultrasonography of the cervical and abdominal region to assess the clinical T stage. Finally, we only included CEA in this study, and other tumor biomarkers may be associated with RLN LNM. Due to limited reports, the tumor biomarkers to predict RLN LNM need further study.

We constructed a preoperative nomogram that can effectively predict RLN LNM and overall survival with a good performance. Both internal and external validation indicated the great performance of this model. Data from more centers are needed to validate this model further.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by The Ethics Committees and Review Board of the Affiliated Hospital of North Sichuan Medical College. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Conception, study design and protocol: H-JY, W-JM, R-XY, H-YW, and DT. Data collection and extraction: K-YJ, HH, Z-DZ, and Q-CQ. Methodology assessments: H-JY, W-JM, R-XY, and X-GG. Statistical analyses: H-JY, W-JM, and R-XY. Writing: H-JY, W-JM, and R-XY. Project oversight and supervision: H-YW and DT. Critical revisions for important intellectual content: H-YW and DT. All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.

This research was supported by the Clinical Research Foundation of the Affiliated Hospital of North Sichuan Medical College (2021LC004) (to DT) and Nanchong City-School Cooperation Research Fund (No. 20SXZRKX0001) (to H-YW).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.