Predicting preoperative lymph node metastasis in esophageal cancer: Advancement and challenges

Abstract

Accurate preoperative prediction of lymph node metastasis is crucial for developing clinical management strategies for patients with esophageal cancer. In this letter, we present our insights and opinions on a new nomogram proposed by Xu et al. Although this research has great potential, there are still concerns regarding the small sample size, limited consideration of biological complexity, subjective image segmentation, incomplete image feature extraction and statistical analyses. Furthermore, we discuss how to achieve more robust and accurate predictive performance in future research.

Key Words: Esophageal cancer; Radiomics; Lymph node metastasis; Nomogram; Machine learning; Computed tomography

Core Tip: The nomogram is very valuable for predicting preoperative lymph node metastasis in patients with esophageal cancer (EC), but current research still has limitations. Improving the study design and statistical analyses of the current research are crucial for assessing EC prognosis and developing personalized treatment plans.

TO THE EDITOR

We read with great interest a recent paper, entitled "Nomogram based on multimodal magnetic resonance combined with B7-H3mRNA for preoperative lymph node (LN) prediction in esophagus cancer"[1]. In this study, Xu et al[1] developed a radiomic nomogram that included radiomic features and B7-H3 mRNA expression, enabling convenient preoperative individualized prediction of LN metastasis (LNM) in esophageal cancer (EC) patients. They emphasized that the combination of magnetic resonance imaging (MRI)-based radiomic features with tumour factors had significant potential for accurately predicting LNM in patients with EC.

LNM AND EC

EC is one of the most common malignant tumours worldwide, causing approximately 598300 new cases annually and accounting for 3.1% of all cancer cases[2]. EC has a poor prognosis, and once metastasis occurs, the five-year survival rate is only 5%[3]. LN status is one of the most important prognostic factors in EC, and accurately predicting LN status before surgery is crucial for formulating treatment strategies[4]. Previous studies have focused primarily on the diagnosis and prediction of LN status on the basis of clinical and radiological features, whereas research based on tumour factors has been relatively limited[5,6]. Therefore, Xu et al[1] focused on the ability of multimodal MRI combined with B7-H3 to predict preoperative LNM in patients with EC. We acknowledge that Xu et al[1] provided valuable insights for predicting preoperative LNM in patients with EC. However, we also noted some existing problems that could impact quality.

First, the expression level of the B7-H3 mRNA did not directly correspond to the functional manifestation of the protein. The actual role of B7-H3 may be influenced by posttranslational modifications or protein degradation, which are not reflected in the mRNA expression levels. Therefore, future studies should adopt other methods, such as western blotting or immunohistochemistry, to verify the expression of B7-H3 to ensure the rigor of the study. Additionally, owning to the area under the curve of only 0.76 and the low accuracy, it is difficult to apply this model widely in clinical practice. Among the molecular markers for predicting LNM in EC, P53 mutations are currently more commonly reported, whereas B7-H3 could be a new potential marker[7]. Therefore, combining B7-H3 expression with other biomarkers or clinicopathological factors may improve the prediction accuracy.

Second, the authors developed a nomogram using MRI radiomics data, B7-H3 mRNA levels, and computed tomography (CT)-reported LN size. Although the authors mentioned that an LN ≥ 10.0 mm in the CT report was considered a positive result, the article lacked prescription of the radiological diagnostic criterion. Previous studies have indicated that a majority of the involved LNs in EC have a diameter of less than 10 millimetres[8,9]. However, quantity, location, and fusion are also key features of the LNs. A detailed description of LNs may further enhance the significance of this study. The image segmentation in this study was manually performed by an operator, which may be subjective. This study lacked quality control and validation of the segmentation results. Therefore, adding quality control measures for segmentation to ensure accuracy and reliability is recommended. In addition to imaging techniques, such as CT and MRI, endoscopic ultrasound plays a significant role in assessing LNM in patients with EC[10].

LMN of EC is a multifactorial process that is influenced by various factors, including tumor biology, the immune microenvironment, and complex tumor-host interactions[11,12]. Tumor cell characteristics, such as genetic mutations and epithelial-to-mesenchymal transition, as well as the local immune landscape, play pivotal roles in determining the ability of cancer cells to invade and colonize LNs[13,14]. Furthermore, the host immune response, including inflammatory pathways and immune cell infiltration, significantly influences the metastatic potential. Given these complexities, the current tools may not fully account for the multifactorial nature of LNM[15].

Finally, this was a single-centre study, and the small sample size may have introduced bias. The authors used only the area under the receiver operating characteristic curve to evaluate the predictive performance of the model, which was insufficient. It is recommended that the sample size be further expanded and that calibration metrics be added to improve the accuracy and credibility of the results[16].

CONCLUSION

In summary, we hope that our findings are useful for further research to promote technological advancements in this field, to help assess EC prognosis and to develop personalized treatment plans.