Flat Epithelial Atypia and Risk of Breast Cancer: A Mayo Cohort Study

Abstract

Background

Based on its cytologic features, and its co-occurrence with atypical hyperplasia and breast cancer, flat epithelial atypia (FEA) has been proposed as a precursor lesion on the pathway to breast cancer development. It is often referred to as an “atypical” or high-risk lesion. However, the long term risk of breast cancer in women with FEA is undefined.

Methods

Biopsies with FEA were identified in excisional breast biopsies in the Mayo Clinic Benign Breast Disease (BBD) Cohort, which includes 11,591 women who had benign biopsies at Mayo-Rochester 1967–2001. Breast cancer risk of FEA, non-proliferative, proliferative and atypical hyperplasia (AH) subsets was assessed using standardized incidence ratios (SIRs), relative to the Iowa Surveillance, Epidemiology, and End Results registry.

Results

FEA was identified in 282 women (2.4%); 130 had associated AH (46%) and 152 (54%) were classified as proliferative disease without atypia (PDWA). With median follow-up of 16.8 years, the SIR for breast cancer in AH + FEA was 4.74 (95% CI: 3.17–6.81) versus 4.23 (3.44–5.13) for AH without FEA (p=0.59). The SIR for PDWA + FEA was 2.04 (95% CI: 1.23–3.19) versus 1.90 (1.72–2.09) for PDWA without FEA (p=0.76).

Conclusions

FEA is an uncommon finding in women with BBD. FEA does not convey independent risk of breast cancer beyond that of the associated PDWA or AH.

Introduction

Flat epithelial atypia (FEA) is a benign proliferative breast lesion characterized by columnar cell changes with cytologic atypia. It is often referred to as an “atypical” lesion or a high-risk lesion. However, FEA is distinct from classical atypical hyperplasia (that includes atypical ductal hyperplasia [ADH] and atypical lobular hyperplasia [ALH]). Atypical hyperplasia (AH) is a long-established premalignant lesion with a four-fold increased risk of later breast cancer, which demonstrates not only cytologic atypia but also architectural abnormalities.^1–4

FEA is the most advanced subset of a group of breast lesions called “columnar cell alteration of lobules”, which exhibit abnormal polarization of luminal epithelium.⁵ Some molecular data support a possible role for FEA as a precursor lesion in a pathway of low-grade malignancy.^6–8 However, the association between FEA and subsequent breast cancer (BC) risk remains undefined. Thus, there is significant uncertainty regarding the optimal clinical management of women with FEA.^9–11

Using the Mayo Clinic Benign Breast Disease Cohort, our primary goal was to investigate the independent BC risk associated with FEA by evaluating its impact on BC risk within well-established risk groups of benign breast disease- 1) AH plus coexistent FEA versus AH without FEA, and 2) proliferative disease plus coexistent FEA versus proliferative disease without FEA. We explored the impact of FEA on BC risk in relation to other BC risk factors such as family history of BC, patient’s age at biopsy, and associated lobular involution. Also, we characterized the breast cancers that occurred among women with FEA in order to address the low-grade malignancy hypothesis.

Materials and Methods

Study Population

The Mayo BBD Cohort includes 13,434 women ages 18–85 who had a breast biopsy with benign findings between January 1967 and December 2001.⁴ In this report, we focused on the 11,591 women who had an excisional breast biopsy or needle biopsy followed by excision for several reasons. Excision of lesions with atypia has been and remains the current standard of practice.¹² Furthermore, inclusion of women with core-only biopsies would introduce selection bias and could result in missed cancers that would alter results on cumulative cancer risk.¹³

Demographic information, clinical findings, family history, follow-up and BC events (both invasive cancer and ductal carcinoma in situ) were obtained from the patients’ medical records, Mayo Tumor Registry, and a study-specific questionnaire. We categorized the family history of BC into three groups: 1) strong = at least 1 first degree relative with BC before the age of 50 years; or 2 or more relatives with BC, with at least 1 being a first degree relative; 2) weak = any family history less than the definition of strong; and 3) none = no family history of BC. The study protocol, including patient contact and follow-up methods on ascertainment of later breast cancers, was approved by the Mayo Clinic Institutional Review Board with methods previously described.¹

Breast Histology

Original archived slides of breast biopsy tissues (hematoxylin and eosin stain) for the entire cohort were characterized with an independent re-review by a single breast pathologist (DWV), who was blinded to both the original diagnosis and later patient outcomes. Histologic findings of the breast biopsy tissues were recorded and classified using Dupont and Page criteria² into the three primary histologic groups of benign breast disease: nonproliferative disease (NP), proliferative disease without atypical hyperplasia (PDWA), or atypical hyperplasia (AH) (includes both ADH and ALH).

At the time of histologic review for the study, the presence or absence of FEA (defined according to WHO criteria)¹⁴ was routinely noted. Histologically, FEA is defined by enlargement, unfolding and architectural simplification of the terminal duct lobular unit along with proliferation of monotonous cells with hyperchromatic nuclei.¹⁴. FEA is characterized by enlarged acini and terminal ducts lined by layer(s) of monotonous epithelial cells that show features of low grade cytologic atypia, including rounding of nuclei, loss of nuclear polarization, and prominent nucleoli (Figure 1A). Importantly, FEA does not have the requisite architectural atypia (e.g. rigid bars, arcades, punched out spaces) required for a diagnosis of AH (Figure 1B). In this study, the combined presence of monotypism, loss of polarity/nuclear rounding and nuclear hyperchromasia were most important in identifying FEA. We did not require prominent nucleoli or nuclear enlargement; however, cases with these features were included as long as the degree of atypia did not indicate clinging ductal carcinoma in situ.

Figure 1.

A. Flat epithelial atypia, characterized by dilated acini with round smooth contour lined by monotypic cells showing low grade cytologic atypia. The nuclei are round and uniform in appearance. Apical cytoplasmic blebs and small amount of luminal flocculent secretions are present. There is no associated architectural complexity (hematoxylin and eosin, × 400).

B. Atypical ductal hyperplasia (ADH, center) adjacent to FEA (upper left corner). The epithelial proliferation of ADH shows cytologic atypia as well as architectural atypia characterized by punched out spaces (hematoxylin and eosin, × 400).

Since FEA is a proliferative epithelial lesion, samples with FEA were classified as PDWA unless there was coexisting AH, in which case they were classified as AH. The degree of age-related lobular involution in the background breast tissue of biopsy tissues was assessed as previously described:¹⁵ 1) complete involution = 75% or more of lobules atrophic; 2) partial involution = 1%–74% of lobules atrophic; and 3) no involution (none) = no atrophic lobules. For breast cancers that occurred during follow-up, slides were obtained for confirmatory review, and details of tumor histology and grade were recorded.

Statistical Methods

Data were descriptively summarized using frequencies and percents for categorical variables and medians and interquartile ranges (IQRs) for continuous variables. We compared presence of FEA across levels of categorical variables (including age at biopsy, year of biopsy, histological impression, extent of lobular involution, and family history of breast cancer) using chi-square tests of significance.

The number of years from surgical biopsy of the BBD to BC diagnosis, death, or last contact defined the duration of follow up. Kaplan-Meier curves were used to visually examine BC risk by levels of FEA and histologic impression. We compared risk of BC in women with BBD to external, population-based expected rates using standardized incidence ratios (SIR) and corresponding 95% confidence intervals (95% CI). The Iowa SEER registry was used as the external reference population because of its demographic similarities to the Mayo Clinic population (80% of cohort members reside in the upper Midwest; >95% were Caucasian, equivalent to that reported in Iowa census data during the study period). We calculated these expected counts by apportioning each woman’s follow-up into five-year age and calendar period categories, thereby accounting for differences associated with these variables. SIRs were calculated both overall and by subgroups defined by FEA and other demographic and clinical characteristics. We assessed potential heterogeneity in SIRs across subgroups using Poisson regression analysis, with the log transformed expected event rate for each individual modeled as the offset term.

Internal (that is, within-cohort) comparisons of BC risk were carried out using Cox proportional hazards regression analysis. The following characteristics were examined: FEA, age at biopsy, year of biopsy, histological impression, extent of lobular involution, and family history of breast cancer. Each variable was initially examined univariately in separate Cox regression models. Following this, we fit one overall multivariate Cox model that simultaneously included all variables in order to assess the independent effects of each on BC risk. All statistical tests were postulated a priori and were two sided, and all analyses were conducted using the SAS software system (SAS Institute, Cary, NC).

Results

Patient characteristics

Among the 11,591 women with excisional breast biopsy in the Mayo Cohort, FEA was uncommon, present in only 282 (2.4%) patients (Table 1). In the 282 women with FEA, 130 (46%) had associated atypical hyperplasia. The remaining 152 (54%) women with FEA were classified as having proliferative disease without AH (PDWA). Examining the frequency of FEA in women with AH, among all 597 women with atypical hyperplasia, 22% had associated FEA. FEA was less frequent among all BBD samples with PDWA (n=3,736), present in only 152 (4.1%) of the PDWA specimens (chi-square p<0.001).

Table 1.

Clinical and Histologic Characteristics of Women with BBD, with and without FEA

Characteristic	BBD without FEA (N=11309)	BBD with FEA (N=282)	Percent with FEA per row	P-value
Histologic Impression				<0.0011
NP	7258	0	0.0
PDWA	3584	152	4.1
AH	467	130	21.8
Year of Biopsy				<0.001
1967–1981	4293	58	1.3
1982–2001	7016	224	3.1
Extent of Lobular Involution				<0.001
Not assessable*	802	6
None	1650	23	1.4
Partial	6313	208	3.2
Complete	2544	45	1.7
Age at Biopsy				≤0.001
<45	3742	58	1.5
45–55	3111	93	2.9
55+	4456	131	2.9
Family History of Breast Cancer				0.03
Unknown	400	9
None	7020	154	2.1
Weak	2747	83	2.9
Strong	1142	36	3.1

Percent indicates percent of subjects in entire row with FEA. P-value comparing the distribution of FEA across levels of the other characteristics based on chi-square test.

¹Assesses differences in frequency of FEA across the subgroups for each variable; comparisons are subset to women with PDWA or AH

NP = non-proliferative disease; PDWA = proliferative disease without atypia; AH = atypical hyperplasia

^*No normal terminal duct lobular units present on the slide, so involution could not be assessed

For women with coexisting FEA and AH, the histologic subtype of AH was ADH in the majority (60.8%), ALH in 30.8%, and both ADH and ALH in 8.5%. For the 152 women with FEA and PDWA, only three had FEA as the sole abnormality. Almost universally, admixtures of FEA with other PDWA lesions were seen: 92% had other columnar cell alterations, 73% had sclerosing adenosis, 59% had ductal hyperplasia, 14% had papillomas, and 14% had radial scars.

FEA was observed with increasing frequency after the adoption of screening mammography in the early 1980s (FEA present in 1.3% of biopsies before 1981 and 3.1% thereafter, p<0.001). FEA was less common in women under age 45 (1.5%) compared to those 45 and older (2.9%, p<0.001). FEA occurred slightly more often in women with a family history of BC (3.0%) versus no family history (2.1%, p<0.03). FEA was more frequent (although still rare overall) among women with partial age-related involution (3.2%) compared to either no involution (1.4%) or complete involution (1.7%) (p<0.001).

FEA and breast cancer risk

Our 35-year cohort has been followed for a median of 16.8 years. FEA appeared to have no independent impact on BC risk: subsequent breast cancer events in women with FEA were associated with the major categories of AH and PDWA, and the coexistence of FEA did not change risk within AH and PDWA subgroups. Specifically, the SIRs were similar for women with AH, stratified by the presence or absence of FEA: 4.74 (95% CI 3.17–6.81), and 4.23 (95% CI 3.44–5.13) respectively (test for heterogeneity p=0.59), Table 2. Among women with PDWA, SIRs were also similar when FEA was present or absent: 2.04 (95% CI = 1.23–3.19), and 1.90 (95% CI=1.72–2.09) respectively (p=0.76). Kaplan-Meier curves also show similar risk in the presence/ absence of FEA within categories of AH and PDWA (Figure 2).

Table 2.

Impact of FEA on relative risk of breast cancer in women with PDWA or AH

	FEA Absent				FEA Present
Overall Histologic Impression	No. Women	Observed Events	Expected Events	SIRs (95% CI)	No. Women	Observed Events	Expected Events	SIRs (95% CI)	p- value
PDWA	3584	412	217	1.90 (1.72, 2.09)	152	19	9	2.04 (1.23, 3.19)	0.76
AH	467	101	24	4.23 (3.44, 5.13)	130	29	6	4.74 (3.17, 6.81)	0.59

Standardized incidence ratios (SIRs) compare the observed number of breast cancer events with the number expected on the basis of Iowa SEER Data. All analyses account for the effects of age and calendar period. P-values are testing heterogeneity of SIRs across FEA status within categories of BBD.

FEA = flat epithelial atypia; PDWA = proliferative disease without atypia; AH = atypical hyperplasia

Figure 2.

Cumulative incidence of breast cancer by FEA status and overall category of BBD.

We also performed a multivariate Cox regression analysis and found that the following variables were significantly and independently associated with BC risk: overall histologic impression (non-proliferative, PDWA and AH), extent of lobular involution, family history of BC, and age at biopsy, Table 3. The presence of FEA was univariately associated with BC risk but attenuated to non-significance after adjustment for other risk factors (p=0.48).

Table 3.

Variations in Hazard Ratios from a Cox Proportional Hazards Regression Modeling Breast Cancer by Characteristics of BBD¹

			Univariate		Multivariate
Characteristic	Number of Censored Observations	Number of Events	Hazard Ratio (95% CI)	p-value	Hazard Ratio (95% CI)	p-value
FEA				<.0001		0.4781
Absent	10265 (97.8%)	1044 (95.6%)	1.00 (ref)		1.00 (ref)
Present	234 (2.2%)	48 (4.4%)	2.29 (1.71, 3.06)		1.12 (0.82, 1.54)
Overall Histologic Impression				<.0001		<.0001
NP	6727 (64.1%)	531 (48.6%)	1.00 (ref)		1.00 (ref)
PDWA	3305 (31.5%)	431 (39.5%)	1.77 (1.56, 2.01)		1.61 (1.40, 1.85)
AH	467 (4.4%)	130 (11.9%)	4.54 (3.74, 5.51)		3.80 (3.04, 4.74)
Age at Biopsy				<.0001		<.0001
<45	3520 (33.5%)	280 (25.6%)	1.00 (ref)		1.00 (ref)
45–55	2820 (26.9%)	384 (35.2%)	1.73 (1.48, 2.02)		1.68 (1.42, 1.98)
55+	4159 (39.6%)	428 (39.2%)	2.04 (1.74, 2.38)		2.25 (1.87, 2.71)
Year of Biopsy				0.8805		0.0016
1967–1981	3832 (36.5%)	519 (47.5%)	1.00 (ref)		1.00 (ref)
1982–2001	6667 (63.5%)	573 (52.5%)	1.01 (0.89, 1.15)		0.80 (0.70, 0.92)
Extent of Lobular Involution				0.0032		0.0006
None	1488 (15.2%)	185 (18.4%)	1.00 (ref)		1.00 (ref)
Partial	5857 (59.9%)	664 (66.1%)	1.17 (0.99, 1.38)		0.85 (0.71, 1.01)
Complete	2433 (24.9%)	156 (15.5%)	0.89 (0.71, 1.10)		0.62 (0.49, 0.80)
Family History of Breast Cancer				<.0001		<.0001
None	6596 (65.3%)	578 (53.5%)	1.00 (ref)		1.00 (ref)
Weak	2496 (24.7%)	334 (30.9%)	1.33 (1.17, 1.53)		1.41 (1.23, 1.63)
Strong	1010 (10.0%)	168 (15.6%)	1.58 (1.33, 1.88)		1.51 (1.26, 1.81)

¹Cox proportional hazards regression analysis. Multivariate hazard ratios compare associations of variable of interest with risk of BC after adjustment for all other variables listed in the table.

NP = non-proliferative disease, PDWA = proliferative disease without atypia, AH = atypical hyperplasia.

Features of breast cancers after FEA

Among the 282 women with FEA, 48 cancers occurred at a median of 12.0 years (range 1.4 – 31.0). Twenty nine of the cancers occurred in women with FEA and AH; 19 occurred in those with FEA but no AH. The majority of cancers (N=37, 77%) were invasive (28 invasive ductal and 9 invasive lobular), 8 (17%) cancers were DCIS, and the remaining 3 (6%) were unknown. The proportion of invasive versus in situ cancers was the same whether or not AH was present with the FEA (p=0.99). Tumor grade varied as follows among 37 invasive cancers: 7 (19%) low grade, 14 (38%) intermediate grade, 4 (11%) high grade, and 12 (32%) unknown. There was a trend toward more intermediate and high grade cancers in the FEA without AH group (p=0.09). Among the 8 DCIS cases, 3 (38%) were low grade, 2 (25%) were intermediate grade, and 3 (38%) were high grade. Among all 48 cancers, 31 were ipsilateral to the FEA, 16 were contralateral, and 1 bilateral. Estrogen and progesterone receptor staining were performed on 33 of the invasive cancers and 7 of the DCIS cancers. Estrogen receptor staining was positive in 30 (91%) of invasive cancers and 5 (71%) of DCIS cases, and progesterone receptor staining was positive in 26 (79%) of invasive cancers and 4 (57%) of DCIS cases.

Discussion

In this study of over 11,000 women with benign excisional breast biopsies spanning both pre- and post-mammography screening practices, we found that: 1) FEA is an uncommon lesion (2.4% incidence) that is associated with AH in about half of cases, and 2) FEA is not independently associated with an increased longterm risk of breast cancer. Furthermore, in a multivariate analysis there was no association of FEA with risk of breast cancer after accounting for the influence of major risk factors for women with benign breast disease, including major histologic category of BBD, age at benign biopsy, extent of lobular involution, and family history

This proliferative columnar lesion with cytologic atypia – now called “flat epithelial atypia” -- has been recognized by pathologists for many years but described by multiple terms [clinging carcinoma,¹⁶ columnar alteration with prominent apical snouts and secretions,¹⁷ atypical cystic lobules,^{18, 19} atypical cystic ducts,^{20, 21} flat ductal intraepithelial neoplasia,²² small ectatic ducts lined by atypical cells with apocrine snouts,²³ and columnar cell lesion with atypia²⁴]. “Flat epithelial atypia” was proposed as a consensus diagnostic term by the World Health Organization in 2003,¹⁴ leading to more uniform acceptance of a single nomenclature and therefore the ability to consistently evaluate and study this lesion.

Several prior studies have evaluated the breast cancer risk attributable to FEA with variable and inconclusive results. Using the Nashville Breast Cohort, investigators performed a nested case-control study of the various categories of columnar alterations, which included 52 women with FEA (14 cases and 38 controls). Women with associated AH were excluded.²⁵ The women with FEA had no increased risk compared to women with other types of columnar cell lesions. In a retrospective series of 1,751 core needle biopsies, Martel et al identified 63 FEA (3.6%) without associated carcinoma or atypical hyperplasia.. Nine women in the FEA group developed breast carcinoma, with seven occurring in the ipsilateral breast at a mean of only 3.7 years after the core biopsy, raising a question of missed breast cancers in this series.¹¹ No comparison group was included in this study. Of note, in another retrospective study of 84 patients who had surgical biopsy with “pure” FEA (i.e. no atypical hyperplasia or lobular carcinoma in situ present), none developed subsequent invasive cancer after median follow up of 13.3 years.¹⁰ Thus, previous work was unable to answer the question of FEA and breast cancer risk conclusively.²⁶

While we see no evidence for FEA as an independent precursor to breast cancer, its relationship to AH and breast cancer remains unclear. Among women in our series with FEA, 46% had co-existing AH; reciprocally, among women with AH, 22% had FEA. When FEA and AH are both present, they usually are seen in the same lobule or in the same tissue section. Others ^{13, 25} have also observed the common co-occurrence of FEA and AH. These data imply that FEA and AH may be related, possibly sharing a common histogenesis. However, the lack of independent breast cancer risk among women with FEA argues against it as a precursor of breast cancer. Moreover, given that FEA is less common than AH, it is unlikely to be a precursor of AH.

There has been speculation that FEA may be linked to the development of low grade cancers based primarily on studies of FEA and synchronous low grade cancer.^{6, 27–29} Our study design enables the testing of a true precursor hypothesis – namely by identifying FEA on benign biopsy and then following women long-term for later cancer events. As described here, we see no evidence that FEA is associated with a higher breast cancer risk than other proliferative lesions. Moreover in studying the features of the breast cancers that developed in women with FEA without AH, we see no predilection for low-grade cancers.

An issue in the current clinical management of FEA is the question of surgical excision when FEA is diagnosed with core needle biopsy. Although outside the scope of this investigation, other retrospective studies have reported that cancer is found in 3–13% of women with FEA in needle core biopsy who subsequently undergo surgical excision.^{13, 30–36} This finding is not surprising, given the association between FEA and AH that we observed, and the known risk of upgrade to cancer with surgical excision of AH. In addition to a small risk of finding cancer at surgical excision, multiple studies have reported finding other high risk lesions (AH or LCIS) in 20–52% of women undergoing excision of FEA found at core biopsy.^{13, 30–33, 36} While most studies recommend surgical excision for FEA on core needle biopsy, others have suggested that this might not be warranted. In the study by Uzoaru et al.,³⁰ 3 of 95 (3.2%) patients with pure FEA on core needle biopsy had carcinoma on subsequent excision and none of 33 patients without excision developed malignancy after a mean follow up of 5 years (range 3 to 8 years). Our study does not specifically address the issue of management for patients with FEA in needle core biopsy, as all of our subjects had undergone surgical excision. Based on existing literature, we believe there is insufficient data for patients with FEA on core biopsy to make categorical recommendations for all patients or to accurately stratify patients based on multiple criteria. As with any finding of atypia on core biopsy, excision should be strongly favored if there are residual worrisome findings on imaging or if the patient is constitutionally at high risk for developing breast cancer.

In summary, FEA is an uncommon finding in this large cohort of women with benign breast disease that spans both pre-mammographic and post-mammographic screening eras. Approximately half of the samples with FEA had concomitant atypical hyperplasia. The presence of FEA does not convey independent elevation in breast cancer risk beyond that of the risk conferred by the co-existing histologic findings, either proliferative disease without atypia or atypical hyperplasia. Due to the frequent association of FEA and atypical hyperplasia, a finding of FEA should prompt pathologists to search for concomitant atypical hyperplasia. Longterm breast cancer surveillance and risk management of women with FEA should follow guidelines tailored to the final diagnostic biopsy categorization of either proliferative disease without atypia or atypical hyperplasia.