Analysis of terminal duct lobular unit involution in luminal A and basal breast cancers

Data and biologic specimens for the current analysis came from a population-based breast cancer case-control study conducted in Warsaw and Łodz, Poland from 2000 to 2003, as previously described in detail [17]. Eligible cases were women between the ages of 20 and 74 years diagnosed pathologically with incident in situ or invasive breast carcinoma. Control subjects were randomly selected using a population-based database, frequency-matched to cases on city and age in five-year categories. A total of 2,386 cases (79% of eligible) and 2,502 controls (69% of eligible) agreed to participate in the study and provided informed consent. The study was approved by the National Cancer Institute and local Institutional Review Boards in Poland. Since the goal of the current study was to compare TDLU involution between tumor subtypes, controls were not included in the analysis.

At the time of routine pathological evaluation, we collected one tumor block (T) and two "grossly" benign tissue blocks, from peritumoral (PT; adjacent to but not touching the tumor) and distant regions from the tumors (DT; at the periphery of the specimen) when possible. We restricted the current analysis to luminal A and CBP cancers because these tumors are viewed as most divergent with regard to etiology and pathogenesis. Benign tissue blocks (PT or DT) were available for 804 cases having these two tumor subtypes (83% of total luminal A and CBP cases). Tumor characteristics and risk factors did not differ significantly among cases with and without benign tissue blocks. We further restricted our analysis to 385 cases younger than age 55 years (317 luminal A and 68 CBP) because at older ages, TDLUs are often not identifiable in tissue sections, particularly in the absence of exogenous hormone use or obesity (data not shown).

We assessed breast cancer risk factors through a detailed personal interview as described elsewhere [17, 18]. Factors evaluated included education, age at menarche, age at menopause, parity, breastfeeding, measured body mass index (BMI), age at first full-term birth, smoking, alcohol drinking, hormone replacement therapy (HRT) use among postmenopausal women, and family history of breast cancer.

We assessed tumor size, histologic type, grade, and axillary lymph node status via clinical reports and independent review (MES) [17].

Procedures used for construction of tissue microarrays (TMAs) from invasive cancer tissues, immunohistochemical staining and scoring are described in detail elsewhere [18, 19]. Previously, we assessed the expression of ER, PR, HER2, EGFR and cytokeratin 5 (CK5) using three different methods: (i) clinical reports of ER and PR (ii) immunohistochemical staining for cases included in TMAs, and (iii) AQUA™(HistoRx, Branford, CT), which is a quantitative immunofluorescent method that provides continuous measurement of expression levels [20]. Different measurement techniques showed strong agreement as previously reported [19]. For the current analysis, we classified tumors as luminal A (ER+ and/or PR+, HER2-) or CBP (ER-, PR-, HER2-, CK5+, and/or EGFR+) as previously described [18].

One pathologist (MES) microscopically reviewed hematoxylin and eosin stained sections of benign blocks to assess characteristics of TDLUs, masked to all clinical and epidemiologic data, including tumor marker expression and subtype classification. To determine which TDLUs were suitable for assessment, we applied a modification of criteria proposed by Milanese et al. [16]. Specifically, we assessed TDLUs which: 1) consisted of acini lined by single luminal and myoepithelial cell layers; 2) displayed a limited extent of metaplasia or dilatation of acini (we excluded TDLUs in which 50% or more of the acini showed lumina three times the normal diameter or metaplastic epithelium), and 3) did not display benign breast disease, such as adenosis or other entities as implied from criteria above. However, we did not exclude TDLUs based on epithelial cell size per se or number of acini. Accordingly, some acini included in this analysis resemble the hyperplastic enlarged lobular units illustrated by Lee et al. [21].

We focused on two agnostic metrics of involution: average number of acini per TDLU and average TDLU diameter (both inversely related to involution). For each slide, we recorded the presence or absence of TDLUs; TDLU "diameter", defined as the maximal linear span of epithelium (relative to 400 × microscopic fields, approximated in 0.25 increments) and number of acini per TDLU (0 to 10, 11 to 20, 21 to 30, 31 to 50, 51 to 100, > 100) (Figure 1). We first conducted a pilot study of 200 slides and measured all identifiable TDLUs present (> 1,000). The average number of acini per TDLU and TDLU diameters across all measured TDLUs did not vary significantly when six or more TDLUs were evaluated (Additional file 1, Figure S1). For the full review, we microscopically reviewed slides beginning at one end and sequentially categorizing up to 10 unselected TDLUs masked to other data.

After completing the slide review, we performed a masked repeat review of 30 randomly selected slides to assess reproducibility; ratings were significantly correlated (number of acini per TDLU, concordance = 77%, P = 0.002; TDLU diameter, concordance = 71%, P = 0.008). TDLU involution variables were not significantly different in PT and DT blocks (P > 0.1 for pair-wise differences between PT and DT for both number of acini per TDLU and TDLU diameter); therefore, we took the average values from PT and DT blocks for each TDLU variable if both were available. As expected, the average number of acini per TDLU and the average TDLU diameter showed strong correlation (r = 0.86, P < 0.0001).

To assess whether TDLU involution varied by molecular characteristics of the cancers, we evaluated the relationships between immunohistochemical status of ER, PR, CK5 and EGFR in tumors (explanatory variables) and the average number of acini per TDLU or the average TDLU diameter (outcome variables) using likelihood ratio tests and unconditional polytomous logistic regression. We defined tertiles of the distribution of the average number of acini per TDLU and the average TDLU diameter using cut-points based on distributions among luminal A tumors. Results, based on using maximal, median and average values across TDLUs, were similar and, therefore, we present only the average values. We also compared continuous AQUA scores for ER, PR, EGFR and CK5 expression in invasive tumors [19], to the number of acini per TDLU or TDLU diameter categories using the Kruskal-Wallis test. To assess our main hypothesis, we evaluated whether the level of TDLU involution varied between luminal A and CBP cancers using adjusted logistic regression models with tumor subtype (CBP vs. luminal A) as the outcome variable and number of acini per TDLU or TDLU diameter as explanatory variables. Adjustment variables included age (five-year interval), study site, breast cancer risk factors (education, age at menarche, age at menopause, parity, age at first full-term birth, breastfeeding, HRT use among post-menopausal women, BMI, and family history of breast cancer). To evaluate which of several correlated tumor features (tumor size, histology, grade, and nodal status, tumor subtype) were most important in driving the associations with TDLU involution variables, we also fitted regression models with average acini per TDLU or average TDLU diameter as the outcome variable and tumor characteristics as the explanatory variables.

We used SAS (version 9.1.3, SAS Institute, Inc., Cary, NC, USA) software for all analyses.

Among the 385 cases younger than age 55 years, TDLUs were identified in 281 cases (232 luminal A and 49 CBP). Distributions of risk factors and pathologic characteristics did not differ significantly among cases with and without evaluable TDLUs, overall or within subtype (Additional file 1, Table S1).

First, we compared TDLU involution characteristics in tumor subtypes defined by individual markers (ER, PR, CK5, and EGFR), dichotomously categorized as negative or positive, and adjusted for age and study site. Compared with ER negative cases, ER positive cases were associated with a significantly reduced number of acini per TDLU (odds ratio (OR) = 0.34, 95% confidence interval (CI) = 0.17 to 0.70, P = 0.003, comparing the highest to the lowest group), whereas EGFR positive tumors were associated with a greater number of acini per TDLU than EGFR negative tumors (OR = 2.78, 95% CI = 1.21 to 6.39, P = 0.02). CK5 positive cancers were also associated with greater number of acini per TDLU than CK5 negative cancers, but the comparison was not statistically significant. The number of acini per TDLU was not associated with PR expression, tumor size, histological type, grade or axillary node status (Table 1). Results for TDLU diameter were similar to those for the number of acini per TDLU (for ER: OR = 0.38, 95% CI = 0.19 to 0.76, P = 0.006; for EGFR: OR = 2.95, 95% CI = 1.24 to 7.03, P = 0.01; comparing the highest to the lowest group).

Next, we evaluated relationships between TDLU involution characteristics and these tumor markers assessed as continuous variables. Analyses comparing number of acini per TDLU to continuous AQUA scores for ER, PR, and EGFR expression demonstrated associations consistent with those found in analyses in which markers were classified as negative or positive (Kruskal-Wallis test, P = 0.02 for ER and EGFR, P = 0.74 for PR, Figure 2).

The number of acini per TDLU and TDLU diameter were significantly greater for CBP as compared with luminal A tumors (Table 2). The associations remained significant after adjustment for age and study site (number of acini per TDLU: OR = 3.36, 95% CI = 1.36 to 8.32, P = 0.009; TDLU diameter: OR = 2.49, 95% CI = 1.08 to 5.74, P = 0.03; comparing the highest to the lowest group). Associations were equivalent or stronger in analyses adjusted for multiple breast cancer risk factors (education, age at menarche, age at menopause, parity, age at first full-term birth, breastfeeding, HRT use among post-menopausal women, BMI and family history of breast cancer; for number of acini per TDLU: OR = 4.44, 95% CI = 1.58 to 12.51, P = 0.005) (Table 2). When analyzing the correlated tumor characteristics simultaneously, the associations with both number of acini per TDLU and TDLU diameter were only observed for tumor subtype (CBP vs. luminal A) but not for tumor grade, histology, size, or nodal status (data not shown).