Classification tree models for predicting cancer status.

摘要

Early detection of cancers might improve the clinical outcomes. Multiple biomarkers with a novel LabMAP technology were used as the laboratory method to develop the diagnostic assay for ovarian, breast, endometrial, and lung cancer. To evaluate the accuracy of early stage detection, logistic regression (with forward selection) and classification tree models were applied as statistical methods. Furthermore, complexity parameters and the number of bootstrap samples were varied to assess the effect on sensitivity and specificity. The receiver operating characteristic curves reflected high sensitivities and specificities.