LASSO Regression

LASSO's higher specificity may reduce unnecessary investigations but could miss more cancers than LCA. LASSO regression was used as a supervised regularisation method that incorporated cancer diagnosis as the outcome. Variables with non-zero LASSO coefficients were included in the LASSO score, with points assigned by coefficient magnitude. LASSO was applied to the full 5,821-patient cohort using multiply imputed data. The LASSO score was likely most attuned to late-stage malignancy because it selected ALP as a high-weight marker in a cohort with many metastatic and late-stage cancers. The LASSO model treated biomarkers as continuous variables to preserve information and stabilize coefficient estimation.