Predicting the patient at low risk for lymph node metastasis with localized prostate cancer: an analysis of four statistical models.

Abstract

Statistical models using preoperative Prostate-Specific Antigen, Gleason primary grade or score of the biopsy specimen, and clinical stage have been developed to predict those patients with clinically localized prostate cancer at low risk for lymph node metastasis. It has been recommended that these patients do not require pelvic lymph node dissections. Four such models were evaluated to assess their accuracy in identifying this subgroup of patients. We reviewed the records of 214 patients with clinically localized prostate cancer who underwent pelvic lymph node dissections. Data from these patients were entered into the four models. Lymph node metastasis was detected in 14% of patients. The results showed the following for each of the proposed models respectively: 78, 50, 76, and 42% of the patients were identified as low risk and, hence, would be spared pelvic lymph node dissections. The false negative rates are 13 (7.8%), 5 (4.6%), 14 (8.6%), and 1 (1.1%). Sensitivities are 56.7, 83.3, 53.3, and 96.7%. While the pelvic lymph node dissection is the most accurate method of detecting occult nodal metastasis, statistical models can identify a cohort of low risk patients that may be spared lymphadenectomy.