In spite of the complexity of the biological basis of the hormonal regulation of breast cancer, clinical studies tend to simplify the information by mainly categorizing continuous variables related to hormonal status and not considering the interactions between variables. The present study was planned to examine the presence of an interaction between cathepsin D (Cath-D) and pS2 in patients treated with adjuvant tamoxifen in a homogeneous subset of node-positive postmenopausal patients and to evaluate the contribution of the interaction to the predictive ability of the model. Steroid receptors (ER and PgR) were measured in cytosol using the dextran-coated charcoal method, while Cath-D and pS2 were determined using commercially available immunoradiometric assays. The prognostic role of each variable and their joint effect were investigated using a Cox regression model. Biological variables were analyzed as continuous and when their prognostic relationship did not seem linear, a restricted cubic spline regression smoothing approach was adopted. The logarithm of hazard showed a linear relationship with the log(ER), while it i) remained almost constant up to about 20 fmol/mg and subsequently decreased for PgR; ii) was almost constant up to about 50 pmol/mg and subsequently decreased for Cath-D; iii) decreased for increasing log(value) up to about 33 ng/mg and subsequently increased for pS2. In the multivariate analysis both PgR and the interaction between pS2 and Cath-D retained a significant prognostic role. For low values of pS2, the prognosis worsened with the increase in Cath-D levels and this relationship reversed for high values of pS2. From the results of the present study we can conclude that i) a significant interaction between Cath-D and pS2 was found in this case series; ii) the prognostic relationship should not be underestimated in clinical decision making; iii) a predictive score obtained considering the contribution of PgR, pS2 and Cath-D could be useful for clinical use.