LMS

LMS predicts overall survival (OS) and cancer-specific survival (CSS) in patients with leiomyosarcoma (LMS) metastatic to the lungs using statistical models derived from the SEER database.

Key Features:

• Data source: Uses cases from the Surveillance, Epidemiology, and End Results (SEER) database.
• Cut-off determination: Applies X-tile analysis to determine optimal cut-offs for age and tumor size, converting continuous variables to categorical variables.
• Prognostic factor identification: Performs Cox regression analysis to identify independent prognostic factors.
• Nomograms: Constructs separate nomograms for overall survival (OS) and cancer-specific survival (CSS).
• Dataset composition: OS dataset comprised 228 cases (training n=160, validation n=68); CSS dataset comprised 183 cases (training n=129, validation n=54).
• Predictive performance assessment: Evaluates models using receiver operating characteristic (ROC) curves and calibration curves and reports area under the curve (AUC) values.
• AUC for OS: 1-, 2-, and 3-year AUCs of 0.783, 0.830, and 0.832, respectively.
• AUC for CSS: 1-, 2-, and 3-year AUCs of 0.889, 0.777, and 0.884, respectively.
• OS prognostic variables: Age, T stage, bone metastasis, surgery, chemotherapy, marital status, tumor size, and tumor site.
• CSS prognostic variables: Age, bone metastasis, surgery, chemotherapy, tumor size, and tumor site.

Scientific Applications:

• Survival prediction: Predicts 1-, 2-, and 3-year OS and CSS for leiomyosarcoma patients with lung metastasis.
• Prognostic factor analysis: Quantifies the impact of clinical variables such as age, T stage, bone metastasis, surgery, chemotherapy, tumor size, tumor site, and marital status on survival.
• Model validation: Provides internal validation using separate training and validation cohorts derived from SEER data.

Methodology:

X-tile analysis determined optimal cut-offs for age and tumor size and converted continuous variables to categorical variables; Cox regression identified independent prognostic factors; two nomograms were constructed and evaluated using receiver operating characteristic curves, calibration curves, and AUC statistics on training and validation cohorts.