Classification and regression tree analysis for prediction of recurrence risk in patients with surgically resected lymph node-positive penile cancer

Abstract

Objective: To assess the risk of recurrence in patients with surgically resected lymph node-positive penile cancer by classification and regression tree (CART) analysis. Methods: The medical records and follow-up data of sixty patients with surgically resected lymph node-positive penile cancer between 1990 and 2008 were retrospectively analyzed. The node-related prognostic factors were collected, such as the number of positive lymph nodes, unilateral or bilateral lymph node metastasis, pelvic nodal involvement and extranodal involvement. The CART analysis was used to generate a regression tree model for identifying recurrence risk of surgically resected lymph node-positive penile cancer, and the tree was validated in this study population. Results: The regression tree model for predicting different recurrence risk of patients with surgically resected lymph node-positive penile cancer was generated by CART. The positive lymph-node number and the extranodal involvement were both important prognostic factors. The patients with node-positive penile cancer were divided into four groups according to recurrence risk classified by CART. The survival curves showed the distinct difference in recurrence-free survival among these four gourps (P＜0.05). Conclusion: A regression tree model generated by CART can accurately identify and classify the recurrence risk of patients with surgically resected lymph node-positive penile cancer, and this regression tree may aid individualized decision-making in clinical practice. DOI:10.3781/j.issn.1000-7431.2011.07.014