Risk factors for ductal and lobular breast cancer: results from the nurses' health study

Epidemiologic studies have shown that reproductive, lifestyle and anthropometric exposures are predictive of subsequent breast cancer risk [1, 2]; however, whether these factors are differentially associated with risk on the basis of histologic subtype is not clear. Among invasive breast cancer cases, ~75% are ductal carcinomas, 15% are of the lobular type and the remainder are a mix of other, less common histologies [3]. Differences in clinical, molecular and pathologic features of ductal and lobular tumors [4] suggest a distinct etiology and may have implications with respect to their prevention, diagnosis and treatment. Further, lobular cancers are more likely than ductal cancers to be estrogen receptor (ER)-positive and progesterone receptor (PR)-positive [5, 6]. This might explain why several studies have demonstrated differences in the association between postmenopausal hormone use (PMH) and risk of breast cancer by histology, with a stronger effect seen for lobular carcinomas [7].

Besides the studies of PMH and breast cancer risk by histology, few large studies have evaluated whether the association of other well-established breast cancer risk factors varies by histology [8‐10]. Moreover, limitations of prior studies include the lack of formal statistical testing across disease subtypes, small sample size, no control for ER or PR status or the use of retrospective data. Therefore, we prospectively examined the relationship between known breast cancer risk factors and their association with the two most common types of invasive breast cancers, ductal and lobular, diagnosed among postmenopausal women from the Nurses' Health Study (NHS) cohort.

At baseline, our analysis included 37,127 postmenopausal women and then expanded to 107,759 women by the end of follow-up. During 1,774,075 person-years of follow-up, we observed 6,226 incident cases of invasive breast cancer. Of these, 4,655 were ductal and 659 were lobular carcinomas (see Table 1 for characteristics by case type). The remaining 1,002 were excluded because of no pathology report (8%), no primary structure evident (1%), mixed ductal and lobular histologies (4%) or unknown or other histology (1%).

The results from this large prospective study further substantiate that well-established breast cancer risk factors differentially affect risk on the basis of tumor histology. Specifically, age at menarche, AFB and PMH use were more strongly associated with lobular versus ductal tumors, while there was no evidence for heterogeneity for age at menopause, nulliparity, parity, menopausal type, alcohol consumption, adult BMI, BMI at age 18, family history of breast cancer and a personal history of BBD. Collectively, these data suggest different etiologies among the tumor subtypes and that lobular carcinomas may represent a more hormonally responsive subtype. When our analysis was limited to ER-positive and PR-positive tumors, the findings were maintained, suggesting true differences in the underlying tumor biology and not a result of differential rates of ER and PR positivity.

A recent prospective analysis examined reproductive and hormonal factors and risk of breast cancer by histology among 17,923 ductal and 3,332 lobular cancers from the Million Women Study [18]. Similar to our results, the effect of age at menarche was greater for lobular compared with ductal tumors (RR per 5-year increase = 0.65 for lobular and 0.93 for ductal). Further, in the Reeves et al. [18] meta-analysis of six cohort and three case control studies, the summary estimates were similar. No formal tests for heterogeneity were performed. Although suggestive of differences in the effect of age at menarche by histologic subtype, most studies had low power and only one study was restricted to postmenopausal women [19]. Overall, prior studies and our observations support a stronger protective effect of a later age at menarche for lobular cancer.

The overall RRs for AFB ≥30 vs. <20 from a meta-analysis of nine studies (six cohort and three case control) were 1.24 for ductal cancer and 1.66 for lobular cancer; however, individually, most studies had limited power, and only the Million Women Study reported a statistically significant difference in the RRs by subtype [18]. Phipps et al. [20] recently reported no difference in risk with AFB across subtypes. In our analysis, nulliparity was associated with an increased risk of both cancer subtypes, while we observed significant heterogeneity in the association only for AFB, but not for parity per se. When we modeled the continuous values, the P trend for increasing parity was significant only for ductal tumors; however, the effect of AFB was strongest for lobular tumors. We cannot preclude that the discrepancy between the categorical and continuous results for parity may be attributed to residual confounding by AFB, which does not occur when modeled as categorical or nonoverlapping variables. In other words, it may be that women with higher parity are also likely to have a later AFB. Our RR estimates for a 5-year delay in AFB were 1.48 for lobular cancer vs. 1.12 for ductal cancer. Other studies have reported no difference in risk for parity by subtype [18, 21, 22]. Prior reports are in agreement with our findings regarding parity. A recent prospective analysis including 1,211,238 women reported a lower risk of ductal but not lobular cancers with parity [20]. Two additional reports have suggested that increasing parity was associated with decreased risk of ductal but not lobular cancers, although one study did not report P values [23] and the association was not statistically significant in the second study [24]. On the basis of our data and prior data, a later AFB and not parity per se may play an important etiological role for lobular cancers, while both AFB and number of births appear to be important for ductal carcinomas. Similar to our findings, most prior reports of the relationship between age at menopause or menopausal status and risk have reported no significant heterogeneity by histology [18, 21, 23, 25, 26].

To date, 10 case control and 7 cohort studies have examined the relationship between PMH use and risk of developing ductal and lobular carcinoma [7, 20]. Of these, all but two cohort studies reported a stronger increased risk for lobular compared with ductal carcinoma [27, 28]. In general, most studies have shown that combined E+P was more strongly related to risk of lobular vs. ductal carcinoma than use of E alone [7, 29]. Lobular cancers are more likely to be ER-positive and/or PR-positive than ductal tumors [30, 31]. Thus, it is not surprising that the effect of PMH may be strongest for a hormone receptor-positive tumor. In fact, a recent study reporting a decrease in the incidence of invasive cancers, particularly a decline in lobular carcinomas [32], attributed this change to the concurrent decrease in PMH use [33] following publication of the results from the Women's Health Initiative [34]. Unlike what we observed among all tumors, the association with PMH did not differ significantly by histologic subtype when restricted to ER-positive and PR-positive lobular and ductal tumors; however, this could be attributed to the smaller case numbers, since the association was still suggestively stronger for lobular tumors.

Fewer studies have examined the relationship between duration of hormone use and risk [7, 23, 29, 35‐38]. In the Million Women Study, Reeves et al. [29] reported no significant difference in risk of ductal, lobular or tubular cancer with increasing duration of use. Results from the French E3N cohort also suggested no significant heterogeneity [38]. Calle et al. [7] reported an increased risk with ≥2 years of use among current users for both ductal and lobular cancers. In contrast, E alone was associated with an increased risk of lobular cancer after 5 years of use, but not with risk of ductal cancer; however, use of E alone for 5-10 years was associated with risk of both cancer types, although the association was stronger for lobular tumors. Similarly, Li et al. [35] documented an increased risk of both histologic subtypes with ≥3 years of use of E+P. Other case control studies have also found an increased risk of lobular cancer with longer duration of combined E+P use [10, 36, 37]. With our large study population and repeated assessment of PMH use, our study is adequately powered to assess a role of both timing and duration of PHH use. We similarly observed a stronger association with use of combined E+P and the lobular type. Although there was no significant heterogeneity in the results when we modeled PMH duration as a continuous variable, this was expected, since we previously reported that breast cancer risk is increased only with long-term use (for example, at least 5 years) [39]. Nonetheless, others have shown a significant increase in risk with shorter durations of use [7, 28, 40].

Epidemiological studies have demonstrated a positive association between BMI and postmenopausal breast cancer risk [41], although for the most part anthropometric factors appear to be similarly related to risk by histology in some [8, 25] but not all studies [42]. Despite this, BMI was more strongly associated, although not statistically different, with ductal cancer compared with lobular cancer in one study [21], while another group reported an inverse association between BMI and mixed ductal-lobular carcinomas (P-het = 0.008), but no significant associations for the other subtypes [26]. We did not observe significant heterogeneity for the association between BMI and risk of cancer by subtype.

To our knowledge, ours is the first study that has evaluated an association between BMI at age 18 and risk of postmenopausal breast cancer by histologic subtype. In general, adolescent obesity is inversely associated with breast cancer risk [41]. In a prospective analysis of the NHS, we previously reported that higher BMI at age 18 was associated with lower breast cancer incidence prior to and after menopause [15]. In the current study, the inverse association between BMI at 18 was similar for ductal and lobular cancers.

Two prior studies reported a significant positive association between alcohol consumption and risk of breast cancer overall, which was stronger for lobular compared with ductal cancer among postmenopausal women (odds ratio (OR) = 1.8 vs. 1.2 [43] and 1.5 vs. 1.2 [26] for lobular and ductal cancers, respectively). In a third study, Rosenberg et al. [23] found a strong association between alcohol consumption and risk of tubular but not ductal or lobular cancers. Although we did not observe significant heterogeneity between daily alcohol consumption and risk overall, we did observe a positive association between increasing intake and risk which was suggestively stronger for lobular cancers (P-het = 0.11). Despite these findings, there was significant heterogeneity for this association when we limited the analysis to ER-positive and PR-positive tumors (P-het = 0.02). Given the inconsistent results, further research is needed for this exposure.

In our analysis, family history of breast cancer and a personal history of BBD were similarly associated with increased risk of breast cancer, regardless of subtype. Three prior studies have evaluated the relationship between family history and risk by histology [9, 20, 26]. Recently, Phipps et al. [20] reported similar positive associations across subtypes in their cohort study. In a population-based case control analysis, family history was associated with a twofold increased risk of breast cancer across all histologic subtypes except for mucinous tumors [26], while family history was similarly associated with a two- to threefold increase in risk of all subtypes except for tubular carcinomas [9]. No confidence intervals or formal tests for heterogeneity were reported in the latter study. To our knowledge, two prior studies have evaluated the relationship between BBD and breast cancer risk stratified by histologic subtype [20, 23]. A previous operation for BBD was associated with a slightly higher risk of lobular cancer (OR = 1.9) compared with ductal cancer (OR = 1.5), although this difference did not achieve statistical significance. In contrast, a recent cohort study found similar significant positive associations with BBD across subtypes with increasing density (hazard ratio for highest density vs. reference for both ductal and lobular cases was 1.62) [20]. On the whole, there do not appear to be large differences in the relationship between family history of breast cancer or personal history of BBD by histology.

Overall, the relationship between the exposures evaluated in the current study and the risk of breast cancer was within the expected range and both subtypes showed similar directions; however, the associations for age at menarche, AFB and PMH use were stronger for risk of lobular vs. ductal tumors. Since lobular tumors are more often hormone receptor-positive, it is not entirely unexpected that various hormonal or reproductive risk factors may be more strongly associated with this subtype. The proposed mechanism underlying the association between an earlier age at menarche and risk likely is via its effect on the number of lifetime ovulatory cycles, which in turn influences lifetime exposure to endogenous ovarian hormones [44]. Parity not only leads to substantial changes in hormone levels but also induces differentiation of terminal duct lobular units, making the breast epithelial cells less susceptible to carcinogenic insult [45]. Given that both exogenous and endogenous hormones have been shown to strongly influence breast cancer risk, an altered sex hormone profile associated with PMH use as an exogenous hormone source would also be expected to increase risk [2]. Collectively, the data suggest that lobular breast tumors are more hormonally responsive than ductal breast tumors, thus making this histologic subtype more responsive to hormonal stimulation [9]. Furthermore, exposure to reproductive hormones (that is, estrogen or progesterone) has been shown to promote lobular rather than ductal differentiation, subsequently increasing the number of lobular cells at risk [9]. In addition, various groups have shown differences in gene expression between these two histologic subtypes, providing insight into the biology of these tumors, which may in turn be utilized as diagnostic or therapeutic targets [46, 47]. Future studies should continue to investigate the biological basis for the reported differential relationships.

Major strengths of our study include the large number of participants and high participation rate, resulting in adequate statistical power to conduct stratified analyses. Further, the prospective nature of the NHS cohort allows for the analysis of repeated measures of most exposures and confounders, limiting the effect of measurement error. In addition, the methods used in this analysis allowed for the estimation of separate associations with each subtype simultaneously, as well as formal tests for differences in risk across subtypes. Although our analysis included a large number of invasive breast cancer cases, we did not have enough cases to assess risk with less common histologic subtypes (for example, tubular). Further, our analysis of ER-positive and PR-positive tumors was less well-powered, given that hormone receptor status was not available for all cases. An additional limitation was that we classified histologies on the basis of a review of pathology reports and did not conduct a centralized pathology review. This could lead to nondifferential misclassification of tumor histology and attenuate the associations.

In summary, the results from this study provide evidence that some reproductive or hormonal and nonreproductive factors are differentially associated with risk of postmenopausal breast cancer on the basis of histologic subtype. For the most part, our results are consistent with prior reports, complementing the growing body of evidence suggesting etiologic heterogeneity between ductal and lobular tumors. Future risk factor analyses should continue to investigate associations stratified by histology.