Investigating esophageal sarcomatoid carcinoma and its comparison with esophageal squamous cell carcinoma on clinicopathological characteristics, prognosis, and radiomics features: a retrospective study

Binbin Hu, Kejia Zhao, Yu‐Shang Yang, Yi Zhang, Guihong Liu, Haiyan Zeng, Bingwen ZouFrontiers in Oncology 2024 · DOI

Abstract

Esophageal sarcomatoid carcinoma (ESC) is a rare malignancy with a "biphasic" nature. This study demonstrates that ESC carries a significantly better prognosis than the more common esophageal squamous cell carcinoma (ESCC). Furthermore, the researchers developed a CT-based radiomics prediction model that provides a reliable and non-invasive method for the differential diagnosis of ESC from ESCC.

Introduction

Esophageal sarcomatoid carcinoma (ESC) is an elusive malignancy, accounting for less than 3% of all esophageal cancers. It is characterized by a unique "biphasic" nature, containing both epithelial (carcinomatous) and mesenchymal (sarcomatous) components. This complexity leads to significant clinical challenges.

Firstly, diagnostic ambiguity is common; because biopsies only sample a small portion of the tumor, they often miss one of the two components, leading to a misdiagnosis of standard ESCC in over 60% of cases. Secondly, due to its rarity, the aggressive nature and prognosis of ESC compared to ESCC have remained unclear, complicating treatment planning. This study sought to clarify the clinicopathological features of ESC and develop a radiomics-based tool to assist in pre-operative differentiation.

Methods and Study Design

The researchers conducted a retrospective study of 67 ESC patients treated at West China Hospital between 2008 and 2018. To ensure a rigorous comparison with the more common ESCC, Propensity Score Matching (PSM) was used at a 1:2 ratio, balancing variables such as age, sex, and tumor stage.

Study design and patient selection flowchart

A key technical component of the study was the development of a radiomics pipeline. Radiologists manually outlined tumor volumes on contrast-enhanced CT scans, from which 851 features (including shape, intensity, and texture) were extracted. Using LASSO (Least Absolute Shrinkage and Selection Operator) regression, the team identified the most predictive features to create a "Rad-score," providing a quantitative probability of a tumor being ESC versus ESCC.

Key Results and Insights

  • Superior Prognosis: Contrary to some earlier reports, this study found that ESC patients have a significantly better survival rate than ESCC patients. The 5-year overall survival was 61.1% for ESC versus 43.6% for ESCC, with a death risk Hazard Ratio of 0.59.
  • Diagnostic Accuracy: The radiomics "Rad-score" performed impressively, achieving an Area Under the Curve (AUC) of 0.823 in the training group and 0.828 in the validation group. This suggests that radiomics can capture intratumor heterogeneity—such as pixel pattern irregularity—that the human eye might miss.
  • Morphological Clues: On CT scans, ESC tends to manifest as a bulky mass expanding into the lumen with an eccentric, crescent-like appearance. In contrast, ESCC typically shows more uniform, annular wall thickening.
  • Biopsy Limitations: The study confirmed the difficulty of traditional diagnosis, noting that only 37.3% of patients received a correct ESC diagnosis via pre-operative endoscopic biopsy.
  • Predictors of Survival: For ESC patients, a tumor length greater than 3 cm was found to be a more significant independent predictor of poor survival than several traditional staging metrics.

Discussion and Limitations

While this study provides one of the largest ESC cohorts to date, it has several limitations. As a retrospective, single-center study, there is inherent selection bias, and the sample size is still relatively small for training robust machine learning models. Furthermore, the radiomics model requires multi-center validation across different CT scanner manufacturers and populations to ensure clinical readiness.

Future research should explore whether the specific ratio of sarcomatous to carcinomatous components affects patient response to modern therapies, such as immunotherapy or targeted treatments.

Summary updated

Abstract

ESC has a better prognosis when compared with ESCC. By developing a radiomics prediction model, we provide reliability and convenience for the differential diagnosis of ESC from ESCC.
BibTeX
@inproceedings{hu2024investigating,
  author = {Binbin Hu and Kejia Zhao and Yu‐Shang Yang and Yi Zhang and Guihong Liu and Haiyan Zeng and Bingwen Zou},
  title = {{Investigating esophageal sarcomatoid carcinoma and its comparison with esophageal squamous cell carcinoma on clinicopathological characteristics, prognosis, and radiomics features: a retrospective study}},
  booktitle = {Frontiers in Oncology},
  year = {2024},
  doi = {10.3389/fonc.2024.1398982},
}
Figure