Evaluation of Prognostic Factors in Prostate Cancer with Partial Least Squares Analysis

OBJECTIVE This study was carried out to explore the capability of partial least squares (PLS) analysis, a multivariate projection method, in the evaluation of investigative prognostic biomarkers. MATERIAL AND METHODS Using PLS analysis, survival and life expectancy were prognosticated in 53 patients with prostate cancer treated with transurethral resection. The age of the patient, stage and grade of the tumours, and immunoreactivity in the tumours for p53, Bcl-2, and TGF-beta1 and its receptors type I and type II (TGFbeta-RI and TGFalpha-RII), and markers for proliferation, vascular density, and macrophage density were included as prognostic factors in the analysis. RESULTS The prognostic impact of the factors, in decreasing order, was tumour grade, proliferation, stage, vascular density, TGFbeta-RI, metastasis, TGF-beta1, and TGFbeta-RII. Macrophage staining, p53, Bcl-2, and patient age did not contribute to the model. The model explained 62% of the total variance in survival, and the standard deviation of error of prediction was 29 months. CONCLUSIONS PLS analysis may become a useful tool for evaluation of putative prognostic factors.