BackgroundPrimary small cell carcinoma of the oesophagus (PSCCE) is a gastrointestinal tumour of rare onset. The current study was to investigate the role of a novel risk stratification system (RSS) for PSCCE.MethodsThe study included patients with PSCCE attending any of five medical institutions in China in 2008-2021, four of which served as a training set (n = 422) for construction of the RSS while the other served as a separate cohort (n = 256) for validation of the model. The RSS was established based on covariates associated with overall survival (OS) with a two-sided P-value of < 0.05 in multivariable regression. Survival discrimination of RSS was assessed.ResultsIn the training cohort, multivariate regression analysis revealed age, Eastern Cooperative Oncology Group score, and initial lymph node metastasis to be independent prognostic factors for OS in non-distant metastatic PESCC; concurrent hepatic metastasis was the only significant predictor of distant metastatic PESCC. Accordingly, the RSS was developed and could classify patients into four subgroups: low-risk localized disease (LLD, defined as non-distant metastasis PESCC without risk factors, n = 58); high-risk localized disease (HLD, defined as non-distant metastasis PESCC with ≥ 1 risk factor, n = 199); low-risk metastatic disease (LMD, defined as metastatic PESCC without concomitant liver metastases, n = 103); and high-risk metastatic disease (HMD, definded as metastatic disease with synchronous liver metastases, n = 63). Three-year OS rates were 52.5%, 29.5%, 14.4%, and 5.7% for LLD, HLD, LMD, and HMD, respectively. When compared with the tumor-node-metastasis (TNM) system, RSS showed a consistently superior ability to predict OS in both the training and validation cohorts.ConclusionThe RSS is a reliable stratification model that could be used to optimize treatment for PESCC.