As previously mentioned, Puglisi et al. [
18] did not analyse clinical outcome according to PN expression, while Zhang et al. [
20] observed that PN expression was correlated with tumour stage at diagnosis. Only Xu’s study has previously evaluated the correlation of PN epithelial expression and BCa-specific mortality, showing that high epithelial PN expression was significantly associated with a higher mortality risk. In our study, a strong correlation between positive epithelial PN expression and both all causes and BCa-specific mortality was also found but only in patients also characterized by a low stromal expression of the protein (Figs.
4,
5 and
6). The trends shown in Fig.
6a &
b were confirmed by competitive risks analysis. This analysis demonstrated a strong correlation of each PN epithelial/stromal PN phenotype with BCa-related deaths, suggesting a specific role for PN expression and interaction according to its compartmentalization in affecting BCa lethality. The lack of any association of PN expression with clinical pathological variables and the maintenance of statistically significant trends observed after adjusting for the co-variates predictive in univariate models confirm the independent prognostic value of PN staining and suggest that different biological mechanisms than those merely correlated with the clinical pathological features at diagnosis might play a more crucial role in the later phases of the natural history of the disease. In a BCa murine model, Malanchi et al. [
27] demonstrated that stromal PN is crucial for metastatic colonisation through the interaction with epithelial stem cells (which were also found by these investigators to overexpress PN) at the metastatic niche level. Tumour cells may stimulate or send a signal to the surrounding tissues to produce PN at the ECM level. A return signal may also exist. These biological premises might help in interpreting the results of our study where no strong association between either epithelial or stromal PN expression and mortality was observed when these variables were analysed on an individual basis while a significant association between mortality and different epithelial/stromal PN expression patterns was found. The possibility of defining groups with hugely different outcomes on the basis of the pattern of PN expression in both the epithelial and stromal compartments was already demonstrated by our group in prostate cancer [
22]. Here again we were able to demonstrate that stromal expression could display different prognostic implications in the patients depending on protein epithelial expression [
22]. In our previous study on prostate cancer, the type of interaction between epithelial and stromal PN expression appears to differ from that observed in the present study. However, again we should take into adequate account that different criteria have been used to categorize PN expression in the two studies (IRS, i.e., a score obtained by multiplying the percentage of coloured cells by the degree of staining intensity in the study on prostate cancer and the percentage of coloured cells in the present study) and, above all, that the intrinsic biologic differences correlated with tumour type might not allow for the comparison of the results achieved in the two studies. In fact, it is well known that different PN isoforms are expressed by different tumours and are most likely involved in carcinogenetic processes [
30,
31], though, to our knowledge, information about PN isoforms specifically involved in prostate or breast cancer is currently unknown. Clearly, our findings should be regarded as merely exploratory and, as such, they should be taken with caution because they might present several limitations related to the retrospective nature of our study and methodological issues. This was, in fact, a retrospective cohort study performed on a relatively small patient cohort. Both conditions might represent a bias and affect the statistical power of the study as well as the possibility to perform appropriate sub-group comparisons. Moreover, our women may not represent the current average population of women newly diagnosed with BCa [
32], and therapeutic standards and attitudes have also been changing over time. In fact, our patient cohort was referred to us between 1985 and 1990 and, therefore, show a relatively high (approximately 50 %) prevalence of women with tumours >2 cm in size a/o with nodal involvement. For this reason, transferring our findings to current clinical situations might also be difficult. The differences in patient populations and follow up duration might also render it difficult to compare our findings with the few ones previously reported in the literature [
18,
19]. Other aspects that might bias the comparison of our findings with those previously reported in the literature are likely to be related to methodological issues. While the same polyclonal rabbit antibody was used in our study and in those by Puglisi et al. [
18] and by Xu et al. [
19], a different scoring method was used by Puglisi’s group based exclusively on the intensity of coloration. Though we have demonstrated a direct relationship between staining intensity and the percentage of coloured cells (Fig.
1), there is no doubt that determining staining intensity is much more subject to individual interpretation, especially in the absence of referee standards. However, our choice to score PN expression only on the basis of the percentage of coloured cells might be questioned, and indeed it might not represent the most appropriate scoring method, especially relative to stromal cells. In fact, we realize that it might not always be easy to distinguish between specific staining of the cellular component of the stroma from the non-specific staining of PN that accumulates in the extracellular matrix. The PN expression at the cellular level in the stroma might not be appropriate even from the biological point of view because the PN secreted in the extracellular matrix is not “biologically inert” but rather contributes to the functional role of the protein. While it is intuitive that the expression of PN in the extracellular matrix can be “quantified” only on the basis of staining intensity, it might be reassuring that the coefficient of correlation between the percentage of coloured cells in the stroma and the intensity of immune staining did not change when the intensity of the immune staining was recalculated on the basis of the intensity observed both in the cellular and in the extracellular component of the stroma (c:0.97;
p < 0.000).