Thesis title: STUDIO RADIOMICO E DEI FATTORI PROGNOSTICI/PREDETTIVI DI RISPOSTA NELLE METASTASI POLMONARI DA PRIMITIVO COLON RETTALE TRATTATE MEDIANTE SBRT
Introduction: Stereotactic body radiation therapy (SBRT) is an effective and minimally invasive therapeutic strategy for the treatment of colorectal cancer (CRC) lung metastases, especially in patients with oligometastatic disease. The aim of this retrospective study is to evaluate the predictive factors for response to SBRT and the potential role of radiomics as a tool for predicting local control.
Methods: A single-center retrospective observational study was conducted on patients with lung metastases from CRC treated with SBRT from 2008 to 2023. Inclusion criteria included histological confirmation of the primary tumor, Karnofsky performance status >60, and age >18 years. A total of 149 metastases in 88 patients were treated. Metastases were classified as synchronous (6%) or metachronous (94%), subdivided into oligorecurrent (38%), oligoprogressive (51%), and oligopersistent (5%) lesions.
A radiomic analysis was also developed on pre-treatment CT images, with extraction of quantitative features and training of five machine learning models (Random Forest, SVM, KNN, MLP, logistic regression) for the classification of lesions as progressive or non-progressive.
Results: Local control (LC) rates at 1 and 2 years were 76.8% and 64.0%, respectively; time to polymetastatic conversion (tPMC) rates were 76.0% and 73.0%; progression-free survival (PFS) was 76.0% and 73.0%; and overall survival (OS) was 80.7% at 1 year and 53.0% at 2 years.
In multivariate analysis, lesion diameter <16 mm was a positive prognostic factor for local control (p<0.05), while no significant predictors emerged for tPMC and PFS. For OS, ECOG status 0 and female gender were associated with a better prognosis (p<0.05). No >G3 toxicities were observed.
The best radiomic model achieved an AUC of 61% (95% CI: 51–70), accuracy 69%, sensitivity 36%, specificity 80%, positive predictive value 44%, and negative predictive value 78% (p<0.05 for AUC; p<0.005 for accuracy, specificity, and NPV).
Conclusions: SBRT is a safe and effective technique for treating lung metastases from CRC, with good local control rates and negligible toxicity. Tumor diameter, functional status, and gender are significant clinical prognostic factors. The radiomic approach shows potential as a non-invasive imaging biomarker for predicting response to SBRT, although performance is still moderate. Prospective multicenter studies and standardized radiomic pipelines are needed to improve its generalizability and clinical integration.