Molecular markers for predicting prostate cancer stage and survival

To assess several molecular markers (detected by immunohistochemistry, IHC) to determine whether they can be used to improve the prognostic value of histological grade alone in predicting the behaviour of prostate cancer.

Tumour tissue was retrieved from 156 men in whom tumour grade, stage and survival were known. The outcome measures were: (i) local stage (T-stage, organ-confined vs extraprostatic); (ii) metastatic status (M-stage, bone metastasis vs no bone metastasis); and (iii) survival. The IHC markers used were chosen to provide a broad representation of various aspects of tumour biology, i.e. the androgen receptor (AR) and oestrogen receptor (ER), adhesion molecules (E-cadherin), proliferation markers (MIB-1), tumour-suppressor genes (TP53 and the retinoblastoma gene product, Rb) and other novel cancer-related proteins (cyclin D1 and the breast cancer susceptibility gene product, BRCA2). All factors were assessed using logistic regression and Cox proportional-hazards survival models for predictive value, after adjusting for effects.

MIB-1, ER, cyclin D1 and E-cadherin all showed close statistically significant univariate associations with histological grade. Univariate analysis also identified close statistically significant associations between T-stage and both MIB-1 and E-cadherin. Likewise, there were close univariate associations for both M-stage and survival, and MIB-1, cyclin D1 and ER. Logistic regression modelling identified MIB-1, cyclin D1 and ER as statistically significant predictors of M-stage and, once MIB-1 was entered into the model, the effects of grade no longer made a significant contribution. MIB-1 was a significant predictor for T-stage, but the effects of grade remained significant in this model. Cox proportional-hazards modelling identified MIB-1, cyclin D1 and ER as being statistically significant predictors of survival, after adjusting for grade. After adjusting for both grade and MIB-1, the effects of cyclin D1 and ER were no longer statistically significant. Excess MIB-1, cyclin D1 or ER expression tended to be present within the most poorly differentiated and advanced-stage lesions; this provides an inherent instability to the models described. TP53, Rb, AR and BRCA2 were of limited prognostic value.

MIB-1, ER and cyclin D1 provide prognostic information that is clearly independent of grade. However, their true clinical value is probably limited because they are expressed mainly in the most advanced lesions. Nevertheless, MIB-1 expression is of sufficient value to warrant inclusion in future prognostic models. Furthermore, the expression of cyclin D1 and ER may reflect aspects of tumour biology that individually are worthy of further investigation. However, none of the IHC markers used in this study can be recommended for use in routine histological preparations.