Abstract 5P
Background
The identification of non-invasive prognostic stratification methods for breast cancer and the discovery of reliable biomarkers for precision therapy are of paramount importance. STAT3—a pivotal transcription factor integral to the regulation of numerous cellular processes, has been shown to be correlated with overall survival (OS) and can be predicted through radiomics potentially.
Methods
The research cohort of 101 patients with matched RNA-seq data from TCGA and DCE-MRI from TCIA. To evaluate STAT3 expression and prognosis, Kaplan-Meier survival analysis, Cox regression analysis, and subgroup analyses were implemented. Functional enrichment analysis and immune cell infiltration examination were conducted. Breast cancer IHC images from HPA database were analyzed by DIA via QuPath software. Radiomic features were extracted from DCE-MRI images using pyradiomics toolset. A predictive radiomics model for STAT3 expression was constructed by LASSO regression and binary logistic regression. The efficacy of the model was assessed by ROC curves, PR curves, goodness-of-fit test, and DCA. The correlation between Rad-scores and immune-related gene expression levels from ImmPORT database was examined by Spearman's rank correlation coefficient.
Results
Our findings indicated that reduced STAT3 expression in patients with breast cancer was associated with a poorer prognosis [Hazard ratio (HR) = 1.927, 95% CI: 1.369-2.712, p < 0.001]. STAT3 expression was significantly lower in tumor tissue compared to normal breast tissue (p < 0.001). The radiomic model exhibited an area under the curve (AUC) of 0.861 in the training set and 0.742 in the validation set (p = 0.348). PR, calibration and DCA curves all confirmed a robust predictive capability of the model. Furthermore, Rad-scores were found to be correlated with STAT3 expression and OS; higher Rad-scores were associated with increased STAT3 expression (p < 0.001) and shorter OS (p = 0.033).
Conclusions
The radiomics model based on DCE-MRI has the potential to non-invasively forecast STAT3 expression preoperatively, thereby providing novel insights into the survival prognosis and personalized treatment strategies for patients with breast cancer.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.