Our analysis of serum PRL levels among population-based controls in the Polish study demonstrated significant relationships with three established breast cancer risk factors: nulliparity, among premenopausal women, and HRT and BMI among postmenopausal women. Consistent with previous reports, we found that PRL levels among parous premenopausal women were lower than those among nulliparous women (Musey et al, 1987
; Eliassen et al, 2007
). In addition, among premenopausal parous women, levels declined slightly with increasing parity. However, we did not find an association between parity and PRL concentrations among postmenopausal, which contrasts with some reports (Wang et al, 1988
; Eliassen et al, 2007
). Lowered PRL levels have been suggested as one of several possible mechanisms that mediate this risk.
In this study, PRL levels were inversely associated with BMI among postmenopausal women, whereas other analyses have shown null (Kwa et al, 1976
; Tworoger et al, 2007b
; Su et al, 2009
) or positive associations (Wang et al, 1988
; McTiernan et al, 2006
). Postmenopausal obesity is associated with higher circulating oestrogen levels and increased breast cancer risk in many studies (Key et al, 2001
). Given that the PRL gene contains an oestrogen response element and that in vitro
oestrogen upregulates expression of PRL (Duan et al, 2008
), our inverse association is unexpected. However, in previous analyses from this study, postmenopausal obesity was associated only with larger tumours, rather than breast cancer overall (Garcia-Closas et al, 2006
), so the subject requires further investigation, ideally by considering distribution of adiposity and concurrent measurements of serum oestrogens.
Although there have been reports demonstrating a positive association of oral contraceptive use with PRL levels (Mishell et al, 1977
; Scott et al, 1978
; Clevenger et al, 2003
), the data regarding HRT are largely null (Castelo-Branco et al, 1995
; Foth and Romer, 1997
; Schlegel et al, 1999
; Molitch, 2008
). In our population, use of both oral contraceptives and HRT were uncommon compared with the United States. Nonetheless, recent/current HRT in the Polish study was significantly associated with higher PRL levels in postmenopausal women. Mutual adjustment of BMI and HRT did not alter these interpretations substantively; both lower BMI and HRT remained related to PRL concentrations. However, these findings need careful interpretation as analyses were based on small numbers of users.
Previous data have suggested that a positive family history of breast cancer may be related to higher PRL levels, especially among premenopausal women (Hankinson et al, 1995
; Clevenger et al, 2003
; Eliassen et al, 2007
). Similarly, we observed increased risk related to increased levels of PRL among women with a family history of breast cancer, but women with a positive family history were relatively uncommon in this data set and results were not statistically significant. Associations of PRL levels with benign breast disease have been mixed and may depend on the particular underlying pathologic condition leading to the development of benign breast disease (Courtillot et al, 2005
). We did not find an association in Poland, but screening was less common than in some other populations.
In addition, we examined the association of serum PRL levels with tumour characteristics. We did not find significant difference in geometric mean PRL levels by either tumour size or the presence of lymph node metastases, suggesting that PRL levels may not be related to time of clinical diagnosis. In this population, we did identify a stronger relationship between high PRL levels and postmenopausal invasive lobular carcinoma. This finding is interesting and in contrast to previous reports in which no heterogeneity between invasive ductal and lobular cancers was detected (Tworoger et al, 2007a
). Some previous reports have suggested a relationship between HRT and risk of lobular cancer (Li et al, 2008
). Our finding of the association of higher PRL levels with invasive lobular carcinoma was independent of HRT use.
Prolactin levels were not related to ER, PR, or HER2 status. These data did not replicate the finding from NHS I and II where the association with PRL was stronger among ER+/PR+ tumours (Tworoger et al, 2007a
). Our study was truncated at age 74 years and in a largely unscreened population; therefore, the characteristics of our postmenopausal ER+/PR+ cancers may have differed from the NHS. Apart from this analysis, knowledge about relationships of PRL levels and HER2 status are limited and further studies are necessary.
The analyses presented herein have some limitations. Notably, our case–control results must be interpreted with caution as PRL is a stress hormone and we cannot exclude that the relationship with breast cancer was influenced by a stress responses (Freeman et al, 2000
). In addition, breast tumour cells have been shown to synthesise and secrete PRL in cell culture models (Ginsburg and Vonderhaar, 1995
). If PRL levels were affected by tumours or patient stresses, our case–control estimates might be inflated; however, our case–control associations are generally similar to those found in NHS (Tworoger et al, 2007a
In this study, PRL also was measured by an immunoassay that does not discriminate between PRL isoforms, some of which are reported to have varying biological activity (Freeman et al, 2000
). Regardless, this immunoassay is a widely accepted method for measuring PRL in clinical and epidemiology studies and is currently the only method that can be easily applied to large population-based studies. Our choice of immunoassay also provides the opportunity for our results to be compared with those obtained by others (Hankinson et al, 1999
; Tworoger et al, 2004
). Finally, high mammographic density, which is perhaps the strongest risk factor for non-familial breast cancer apart from age and gender, has been associated with higher PRL levels in some (Boyd et al, 2002
) but not all (Tamimi et al, 2005
; Johansson et al, 2008
) studies. We did not have the ability to examine the association of PRL levels with mammographic density in the current analyses.
The strengths of this study include its population-based design, and extensive collection of risk factor, pathologic and immunohistochemical data. Our analyses were based on incident cases from whom serum was collected at the time of diagnosis of breast cancer. In addition, PRL measurements have been shown to be reliable and most likely a reflection of cumulative exposure over time (Missmer et al, 2006
; Arslan et al, 2008
; Tworoger and Hankinson, 2008
; Kotsopoulos et al, 2010
), and hence can be considered a stable marker of exposure and potentially risk. We found that elevated PRL levels were associated with selected breast cancer risk factors and, with the caveats outlined above, also increased breast cancer risk among postmenopausal women. Consistent with previous prospective studies (Tworoger et al, 2004
) and case–control studies summarised in a recent review (Tworoger and Hankinson, 2008
), we found that PRL levels were unrelated to two factors reflecting progression, tumour size, and lymph node metastases. In conclusion, our data suggest that PRL levels may be related to several breast cancer risk factors and could potentially have value in understanding the mechanisms that mediate these factors. Accordingly, continued study of the importance of PRL in breast cancer is warranted.