Introduction

Benign breast disease (BBD) is a broad category of pathologic diagnoses including, but not limited to fibroadenomas, cysts, fibrocystic disease, papillomas, and ductal epithelial proliferations with or without atypia. Studies have demonstrated a 4–5-fold increased risk of developing breast cancer in individuals demonstrating BBD with atypia and a 1.5–2-fold increased risk of developing breast cancer in individuals with BBD without atypia [1]. The underlying cause of BBD remains unknown, though the pathway from normal terminal ductal lobule to premalignant breast lesions, noninvasive, and invasive cancer has been well described morphologically [2]. The histologic evolution of ductal breast cancer was initially introduced by Wellings and Jensen. The modified Wellings Jensen model of breast cancer development describes how hyperplastic breast epithelial cells slowly enlarge the terminal duct lobular units; forming hyperplastic enlarged lobular units that may then lead to microcysts or development of usual ductal hyperplasia or atypical ductal hyperplasia (ADH). ADH may develop into ductal carcinoma in situ and then may develop eventually into invasive breast malignancy [3]. Genetic predisposition and environmental factors, such as diet, alcohol, and physical activity, may modify the development of benign lesions, though mechanisms are poorly understood [4–10].

Challenges in studying the epidemiology of BBD include the lack of widespread use of a standardized histologic classification system for BBD and the lack of true prevalence estimates in the general population. The cumulative incidence of biopsy-proven BBD is approximately 10–20 %, whereas autopsy studies have demonstrated much higher prevalence, at approximately 50 % [11]. Whether some of the conditions of BBD (such as proliferative disease with atypia) are direct precursors to premalignant and invasive malignancy, as postulated by the modified Wellings Jensen model, versus purely risk indicators, remains unclear [12]. Given the variable increased risk of breast cancer in women with BBD reported by multiple studies, a formal meta-analysis of the existing literature was performed to better quantify the risk estimate associated with BBD including both proliferative and non-proliferative conditions.

Methods

Protocol and registration

No standard protocol or registration was utilized; rather, we followed established methods and approaches to searching, defining eligibility, and analysis.

Eligibility criteria

Published observational studies including retrospective and prospective case–control and cohort studies evaluating benign breast disease and the risk of developing breast cancer were included in this study. There were no publication date restrictions. Studies published in English from multiple continents were included in this study. The primary outcome evaluated was the risk estimate of developing breast cancer in individuals with a history of confirmed benign breast disease. The risk estimates extracted from the studies included relative risk, odds ratio, standardized incidence ratios, rate ratio, hazards ratio, and incidence rate ratio.

Information sources and search

Studies were identified by searching MEDLINE (1966-present) via Pubmed, Google Scholar (1972-present), and EMBASE (1980-present). The date of the most recent search was January 2nd, 2011. Full articles available in English that evaluated the risk of breast cancer in individuals with benign breast disease were included in the meta-analysis.

The Pubmed database search was conducted using the search terms: benign breast disease and breast cancer. This yielded 320 search items. After review of these search items, 62 potential primary subject articles were identified. The EMBASE database search was conducted utilizing the search terms: benign breast disease and breast cancer, benign breast disease and breast cancer risk, benign proliferative breast disease and breast cancer, and benign proliferative breast disease, and breast cancer risk which yielded 1069, 185, 107, and 25 items, respectively. After review of these search items, 14 potential primary subject articles were identified. The original Google Scholar Search utilizing the terms: benign breast disease and breast cancer that yielded approximately 953,000 search items. The first 1,000 items of these were reviewed and this yielded 33 potential primary subject articles. Utilizing the same search engine, benign breast disease and breast cancer, the modified Google Scholar Search with the limitation of “Exact Phrase” yielded 541 search items and the additional modified Google Scholar Search with the limitation of “All in Title” yielded 155 search items. Utilizing the search engine: benign proliferative breast disease AND breast cancer, a modified Google Scholar Search with the limitation of “All in Title” yielded seven potential items. After these items from the modified Google Scholar Search were reviewed, an additional 19 potential primary subject articles were identified. Thus in total, the Google Search identified 52 potential primary subject articles. Throughout the three database searches, there were 128 potential primary subject articles identified. Twenty-three additional related articles were identified during data extraction from the bibliographies of the articles obtained from the primary database search, five of which were potential primary subject articles. During data extraction, 55 of these primary subject articles were found to be relevant to the focus of the meta-analysis. Fifteen of these 55 primary subject articles dealt specifically with one type of benign breast condition. After eliminating duplicate reports from a study patient population and eliminating studies that did not report a risk estimate with confidence intervals, there were 32 studies [13–44] that were used in quantitative synthesis of this meta-analysis and 23 studies [13–15, 17–23, 26–29, 31, 33–36, 38, 40, 42, 44] that were used specifically in quantitative synthesis of the categories: non-proliferative disease (NPD), proliferative without atypia, benign breast disease not otherwise specified (BBD), and atypical hyperplasia not otherwise specified (AHNOS). See Appendix A.

Study selection

Eligibility assessment was performed independently in a standardized manner utilizing title, abstract, and full-text publication by a single reviewer and then reviewed by a second author as detailed in Fig. S1.

Data collection process

We developed a data extraction excel sheet, pilot tested it on approximately ten included studies, and refined it accordingly. One review author performed data extraction and a second author reviewed the extracted data. Disagreements were resolved by discussion between the authors.

Data items

Information was extracted from each included trial on (1) trial participants’ characteristics (publication date, acquisition date range, reported patient age, study location, reported follow-up, review method), (2) benign breast disease assessment, and (3) risk estimate for each sub-classification of benign breast lesions and number of adjusted factors.

Summary measures

Risk estimates including relative risk, odds ratio, standardized incidence ratios, rate ratio, hazards ratio, and incidence rate ratio were the primary measures of breast cancer risk in patients with benign breast disease.

Synthesis of results

Random effects meta-analysis was used to allow for the heterogeneity of results across studies [45]. Data were processed in SAS and the analyses were performed using R-package “meta”. Since most studies reported relative risks (RR) or odds ratios (OR) and their associated 95 percent confidence intervals (CI), these were chosen as summary statistics for each study. The standard error of log (RR or OR) using the 95 percent CI was derived with the expression: [log (upper limit) − log (lower limit)]/2 × 1.96, assuming that the logarithms of the risk estimates have a normal distribution. These standard errors were used as weights for summary effect estimates in the meta-analysis. To examine the homogeneity of the effect size across studies, we used Cochrane Q statistic, which is distributed as Chi square with degrees of freedom (number of study-1) under the null homogeneity hypothesis [45]. Publication bias was visually examined by Funnel plot and formally tested by rank correlation methods [46]. These analyses were performed separately for non-proliferative disease, proliferative disease without atypia, benign breast disease not otherwise specified, atypical hyperplasia not otherwise specified, atypical ductal hyperplasia, atypical lobular hyperplasia, adenosis, cysts not otherwise specified, fibroadenoma, and papilloma not otherwise specified.

Risk of bias

The possibility of publication bias was assessed by evaluating the rank correlation test of funnel plot asymmetry in regards to the analysis of non-proliferative disease, proliferative disease without atypia, benign breast disease not otherwise specified, and atypical hyperplasia not otherwise specified.

Additional analyses

The reported time from diagnosis of benign breast disease to development of breast cancer, the age at first biopsy for BBD, and the age at breast cancer diagnosis were also analyzed when reported to explore their potential as modifiers of risk estimates [13, 15, 16, 27, 30, 31, 33, 36–38, 40, 42, 47, 48].

Results

Study selection

Initial literature search yielded 3,409 potential articles with a final number of 34 studies included for quantitative synthesis of this meta-analysis after exclusion criteria were applied (Fig. S1).

Study characteristics

Published retrospective and prospective case–control studies from 1972 to 2010 evaluating benign breast disease and the risk of breast cancer were included in this study with patient data acquisition reported from 1930 to 2007 (Table S1). The mean reported age at benign breast biopsy was 46.6 years (32.7–57.3) and the mean age of developing breast cancer was 55.9 years (49–63.2) (Table S2). The mean follow-up length was 12.8 years (range 3.3–20.6) (Table S1).

Results of individual studies

Twenty-four of these studies were used specifically for analysis of the categories: non-proliferative disease, proliferative disease without atypia, benign breast disease not otherwise specified, and atypical hyperplasia not otherwise specified. Consistent with the variability of published results, the reported risk estimates for developing breast cancer in the 8 studies evaluating non-proliferative disease ranged from 0.75 to 1.60 (Fig. 1). Risk estimate reported in the 15 studies evaluating proliferative disease without atypia ranged from 1.20 to 7.26 (Fig. 2). Risk estimates reported in the 10 studies evaluating benign breast disease not otherwise specified ranged from 1.70 to 3.50 (Fig. 3). Reported risk estimates in the 13 studies evaluating atypical hyperplasia not otherwise specified ranged from 2.10 to 25.20 (Fig. 4).

Fig. 1
figure 1

Non-proliferative

Full size image
Fig. 2
figure 2

Proliferative disease without atypia

Full size image
Fig. 3
figure 3

Benign breast disease not otherwise specified (BBD)

Full size image
Fig. 4
figure 4

Atypical Hyperplasia not otherwise specified (AHNOS)

Full size image

Syntheses of results

The summary risk estimate of developing breast cancer following a biopsy showing non-proliferative disease was 1.17 (95 % CI 0.94–1.47), based on 8 studies (Fig. 1). The summary risk estimate for breast cancer following proliferative disease without atypia was 1.76 (95 % CI 1.58–1.95), based on 15 studies (Fig. 2). The summary risk estimate for benign breast disease not otherwise specified was 2.07 (95 % CI 1.64–2.61), based on 10 studies with significant heterogeneity (I 2 = 97.8 %; df = 9, p < 0.0001) (Fig. 3). The summary risk estimate for atypical hyperplasia not otherwise specified was 3.93 (95 % CI 3.24–4.76), based on 13 studies (Fig. 4 ). There was significant heterogeneity in the studies evaluating non-proliferative disease (I 2 = 79.7 %; df = 7, p = <0.0001) and benign breast disease not otherwise specified (I 2 = 97.8 %; df = 9, p < 0.0001). No significant heterogeneity was identified in the studies evaluating proliferative disease without atypia (I 2 = 40.1 %, df = 14, p = 0.0542) and atypical hyperplasia not otherwise specified (I 2 = 33.2 %; df = 12, p = 0.1166).

Risk of bias

Publication bias was observed in non-proliferative disease (Fig. S2), benign breast disease not otherwise specified (Fig. S3), and minimal in proliferative disease without atypia (Fig. S4). Publication bias was not observed in atypical hyperplasia not otherwise specified (Fig. S5).

Additional analyses

Histologic characteristics

Analysis of the 7 studies reporting adenosis demonstrated a summary risk estimate of 2.00 (95 % CI 1.46–2.74) (Fig. S6). The 6 studies reporting atypical ductal hyperplasia demonstrated a summary risk estimate of 3.28 (95 % CI 2.54–4.23) (Fig. S7). The 6 studies reporting atypical lobular hyperplasia demonstrated a summary risk estimate of 3.92 (95 % CI 2.81–5.47) (Fig. S8). The 9 studies that reported cysts not otherwise specified demonstrated a summary risk estimate of 1.55 (95 % CI 1.26–1.90) (Fig. S9). The 11 studies reporting fibroadenoma demonstrated a summary risk estimate of 1.41 (95 % CI 1.11–1.80) (Fig. S10). The 8 studies reporting papilloma not otherwise specified demonstrated a summary risk estimate of 2.06 (95 % CI 1.38–3.07) (Fig. S11). Significant heterogeneity existed for all of these studies except for atypical ductal hyperplasia and atypical lobular hyperplasia (Table S3).

Time from biopsy to diagnosis of breast cancer

Several of the studies also reported the average time to developing breast cancer after the initial biopsy of BBD, the age at the first biopsy, and the age at breast cancer diagnosis, which are clinically relevant (Table S2). The mean duration from initial diagnosis of benign breast disease to diagnosis of breast cancer was 9.4 years, based on the 10 studies that reported the average duration.

Discussion

This is the first meta-analysis, to the best of our knowledge, performed on multiple published studies evaluating the association of biopsy-proven benign breast disease with risk of developing breast cancer. We found that proliferative benign breast disease with or without atypia is associated with an increased risk of developing breast cancer with the highest measured relative risk of nearly 4-fold for atypical hyperplasia not otherwise specified. There was no heterogeneity among study results for proliferative benign breast disease or atypical hyperplasia not otherwise specified.

One limitation of this meta-analysis includes the lack of uniform reporting on specific BBD pathologies when comparing studies from multiple countries. We addressed this, in part, by using four main histologic categories. Given the lack of uniform reporting on specific BBD pathologies, there may be overlap between the four main and six additional histological types. Also, follow-up of patients included in these studies did not always account for patient relocation, allowing the possibility of additional breast cancers diagnosed in individuals not included in the follow-up studies. The reported review method and number of adjusted factors in the reported risk estimate were also extracted. Given the large number of studies included in this meta-analysis and the variable number of adjusted factors and reported review method, additional statistical analysis was not performed and is a potential limitation of this meta-analysis.

Uniform classification and reporting of benign breast disease is needed to better delineate the relationship of specific benign breast disease pathologies and increased risk of breast cancer [11]. Furthermore, uniform measures of reporting risk estimates are needed to improve the clarity of implications for clinicians. For this purpose, Elmore and Gigerenzer recommend reporting results in terms of absolute risk as opposed to relative risk [12].

These summary estimates of breast cancer risk following benign breast biopsy have the potential to help improve clinician knowledge and patient education and guide screening recommendations. Women with benign proliferative breast disease should more closely adhere to the currently recommended American Cancer Society (ACS) guidelines for breast cancer screening which support annual screening beginning at 40 years old. It remains unclear if the increased risk associated with proliferative benign breast disease is significant in magnitude to justify additional screening approaches, such as breast MRI, which is recommended by the ACS for women with a >20 % calculated lifetime risk of breast cancer. The Gail Model is a commonly used clinical tool to calculate an individual woman’s risk of developing breast cancer over the next 5 years and by the age of 90. Both the number of prior breast biopsies and the presence of atypia on biopsy are incorporated in the Gail model. However, Pankratz et al. showed that the Gail model significantly underestimates the risk of breast cancer in women with atypia and directed clinicians to use caution when using the Gail model to counsel these women [49]. Recognizing the increased risk of breast cancer in patients with benign proliferative breast disease may help to improve current risk assessment models used for determining screening recommendations [50].

Further research is also necessary to determine the potential beneficial effects of chemoprevention strategies such as anti-estrogen medications. Khan et al. demonstrated that individuals with BBD and breast cancer were more likely to have estrogen receptor-positive epithelium than those patients with BBD without cancer and thus have the potential to be sensitive to anti-estrogens such as tamoxifen that block estrogen receptor signaling [51]. Data supporting this hypothesis come from the small subset of women with atypical hyperplasia from the National Surgical Adjuvant Breast Project P-1 trial, who had an 86 % reduction in breast cancer risk with the use of tamoxifen [52]. Tan-Chiu et al. in 2003 demonstrated that tamoxifen treatment also reduced the risk of benign breast disease by 28 % (RR = 0.72, 95 % CI 0.65–0.79) and reduced the risk of biopsy by 29 % reduction (RR = 0.71, 95 % CI 0.66–0.77). Additionally, in this study that examined the medical records of the 13,203 women with follow-up participating in the NSABP Breast Cancer Prevention Trial, Tan-Chiu et al. also found that tamoxifen therapy resulted in statistically significant reductions in the risk of adenosis (RR = 0.59, 95 %CI 0.47–0.73), cysts (RR = 0.66, 95 % CI 0.58–0.75), duct ectasia (RR = 0.72, 95 % CI 0.53–0.97), fibrocystic disease (RR = 0.67, 95 % CI 0.58–0.77), hyperplasia (RR = 0.60, 95 % CI 0.41–0.62), metaplasia (RR = 0.51, 95 % CI 0.41–0.62), fibroadenoma (RR = 0.77, 95 % CI 0.56–1.04), and fibrosis (RR = 0.86, 95 % CI 0.72–1.03). Overall, the authors concluded that women taking tamoxifen, especially women younger than 50 years old, had reduced incidence of clinically detected benign breast disease and underwent fewer biopsies [53]. The current American Society of Clinical Oncology recommendation to discuss chemoprevention with women at increased 5-year risk for beast cancer supports consistent use of benign breast biopsy morphology to classify risk [54].

Benign breast disease is a common entity that places women at an elevated risk for breast cancer as evaluated and proven by the many studies from multiple countries included in this meta-analysis. Investigation of management strategies to determine which patients could potentially benefit from closer adherence to existing screening recommendations, from additional screening modalities, or from chemoprevention is needed.