Ductal Lavage of Fluid-Yielding and Non–Fluid-Yielding Ducts in BRCA1 and BRCA2 Mutation Carriers and Other Women at High Inherited Breast Cancer Risk

Women with an inherited predisposition to develop breast cancer are a group at very high risk of the disease, one which has not been well-served by standard screening techniques. It is estimated that 9,000 to 18,000 cases of breast cancer in the U.S. per year are attributable to inherited risk. A large percentage of such cases are related to deleterious mutations in the breast cancer susceptibility genes BRCA1 and BRCA2, of which some 1 in 500 to 800 American women are estimated to be carriers (1, 2). Lifetime risks of breast cancer in women with BRCA mutations have been reported in the 45% to 82% range (3, 4). A minority of such women choose to undergo prophylactic mastectomy, which is the most effective available preventive method (5-10); bilateral salpingo-oophorectomy (generally done to decrease the high risk of ovarian cancer in BRCA mutation carriers) and tamoxifen are also used as breast cancer–risk reducing strategies in this population (11-15).

For the majority of women with an inherited predisposition to breast cancer who do not choose prophylactic mastectomy, intensive screening is an emerging alternative. Standard mammographic screening has been shown to be of inadequate sensitivity in this group of generally young women (16). A high incidence of interval cancers has been reported with mammographic screening in this population (17). Increasingly, breast magnetic resonance imaging (MRI) is being incorporated, both within and outside of research protocols, as a screening technique in these high-risk women, with encouraging reports of high tumor detection rates at early stages (18-23). We have recently reported on the use of mammography, high-quality breast MRI, clinical breast examination, and ductal lavage as a comprehensive screening protocol for women at high inherited risk of developing breast cancer (24). In this population, we have identified high-risk lesions by MRI screen detection and by cytologic assessment of ductal cells (24). Because of the high cancer risk in these patients, there is great need to develop and validate novel breast screening techniques, with the ultimate goal of improving their breast cancer outcomes through early detection.

Most breast cancers start in the breast ducts, and early, potentially premalignant alterations in the ductal epithelium are beginning to be defined (25-27). For these reasons, evaluation of breast duct cells is an emerging technique for breast cancer risk assessment, and for the discovery of potential biomarkers which may serve as intermediate end points in trials of cancer prevention agents (28-34). Various minimally invasive methods exist for collecting breast duct cells, including nipple suction aspiration and random periareolar fine-needle aspiration; the finding of atypia in cells collected by each of these techniques has been associated with increased subsequent breast cancer risk (35, 36). A more recently developed cell collection method, ductal lavage, uses a small catheter inserted into the nipple to collect cells lining the breast ducts (37). Advantages of ductal lavage include higher average cell yield than nipple suction aspiration, and relative anatomic specificity, with the ability to resample a specific abnormal duct over time (24, 37, 38). Potential disadvantages include reports of low cancer detection rate in patients with known malignancy, possibly because of duct occlusion by tumor (39, 40); this finding has led to speculation that ductal lavage may be a more appropriate technique for risk assessment than for cancer diagnosis. In a large, multicenter study evaluating ductal lavage for tolerability and cell yield, ductal lavage was attempted only in those ducts which yielded fluid on nipple suction aspiration (fluid-yielding ducts; ref. 37), a strategy based on previous reports that women with fluid-yielding ducts were at higher risk of breast cancer than those with non–fluid-yielding ducts (35, 41, 42).

Our breast screening protocol for women with high inherited risk of breast cancer, combining annual ductal lavage, breast MRI, mammography, and biannual clinical breast exam, was initiated with a goal of improving early detection of cancer and high-risk breast lesions (24, 38). Early in the course of this study, we observed that a higher proportion of our patients than the 16% reported in previous series (37) did not yield fluid on nipple suction aspiration, and we commenced lavage of non–fluid-yielding ducts, as well as fluid-yielding ducts. We now report results of ductal lavage of fluid-yielding ducts and non–fluid-yielding ducts, with associated reproductive and life-style characteristics, in women enrolled in this breast screening protocol.

After study approval by Institutional Review Boards at both centers, in accordance with assurances filed and approved by the Department of Health and Human Services, participants were recruited from cancer genetics clinics at Stanford University School of Medicine and the Dana-Farber Cancer Institute. Inclusion criteria and patient enrollment procedures at both centers were similar. Eligibility criteria at both centers included a documented BRCA1 or BRCA2 mutation; at Stanford University School of Medicine, patients were also eligible if they had no BRCA mutation, but had a >10% risk of developing breast cancer at 10 years based on the Claus model, which incorporates only family history of breast cancer (43-45). If patients had a personal history of breast cancer and no mutation in BRCA1 or BRCA2, the Claus model was used to calculate predicted risk for an unaffected sister; if this risk was >10%, the patient was eligible for participation. Only the unaffected breast was eligible for lavage in patients with prior breast cancer history. Participants had to be at least 25 years of age, or 5 years younger than the earliest age at which a relative was diagnosed with breast cancer. Patients with a history of breast cancer or ovarian cancer had to have completed adjuvant therapy at least 1 year previously. Patients who had had prior breast surgery which seemed to distort the duct system, including incisions near or involving the nipple, were not eligible for ductal lavage of that breast, given our concern for potential increase in infection risk under those circumstances. Informed consent was obtained from all patients, and all study procedures were compliant with regulations of the Health Insurance Portability and Accountability Act of 1996. Alternatives to study participation were offered to all patients.

Participants were enrolled in a pilot breast screening study incorporating mammography, MRI, and ductal lavage, with the goal of evaluating these combined techniques for their ability to detect high-risk and malignant breast lesions. The breast screening protocol and its preliminary results have been described in detail previously (24, 38, 46). The protocol included twice yearly clinical breast exam, yearly mammogram, MRI, and ductal lavage. Abnormality detected on clinical breast exam required 3 to 4 months follow-up clinical breast exam or biopsy, as determined by clinical features; further imaging, including ultrasound and additional mammographic views, was done as prompted by clinical findings. Abnormal MRI or mammogram required 6 months of follow-up or biopsy, as determined by radiographic features. Atypical cells on ductal lavage required 6-month interval follow-up ductal lavage and 6-month follow-up MRI of the affected breast. Enrollment began in September of 2001, and accrual continues.

Participants were anesthetized topically with 4% lidocaine cream applied to the nipple 20 to 30 minutes prior to the procedure. Nipple suction aspiration was done to identify any fluid-yielding ducts. Attempts were made to cannulate any duct, regardless of fluid status, which could be identified using a dilator coated in 1% xylocaine gel, and subsequently a catheter (Cytyc Health Corporation, Boxborough MA; Acueity, Palo Alto, CA). If resistance was met on attempt to catheterize a duct, gentle pressure was applied; if further resistance was encountered, or if the patient experienced discomfort, no further attempt was made to catheterize that duct. Once the catheter was inserted into the duct, 3 to 5 mL of 1% lidocaine was injected, followed by approximately 15 mL of normal saline, in aliquots of 5 mL per injection. Following each injected aliquot of normal saline, breast massage was done and fluid collected via the lavage catheter. The location of each lavaged duct was marked in all cases by assigning a location on a two-dimensional grid, and in most cases by inserting a metal clip provided for this purpose (Acueity) and recording its location via photograph. A cytologic diagnosis of normal cells, insufficient cellular material for diagnosis, mild atypia, marked atypia, or malignant cells was made for each specimen. A representation of a benign and an atypical cytologic reading is presented in Figs. 1 and 2. Time constraints limited attempted cannulation to approximately two to three ducts per breast. Both medical oncologists performing the ductal lavage procedure (A.W. Kurian and A.R. Hartman) and both pathologists interpreting the cytologic specimens (L.C. Collins and K.W. Nowels) were trained by the same methods, as published by Dooley et al. (37).

Univariate analysis of patient characteristics associated with the results of ductal lavage was done using Fisher's exact test for categorical data, and the Mann-Whitney U test for continuous data. All P values are two-sided. Logistic regression was used to identify those variables which are most significant independent predictors of fluid-yielding versus non–fluid-yielding duct status.

Patient characteristics are presented in Table 1. A total of 75 patients underwent attempted lavage; 24 patients were enrolled from the Dana-Farber Cancer Institute, and 51 from Stanford University Medical Center. Comparison of baseline clinical characteristics between patients from the two participating centers revealed no statistically significant differences in median age, BRCA1 or BRCA2 mutation status, prior breast or ovarian cancer, prior chemotherapy or radiation therapy, prior breast biopsy, prior use of tamoxifen or other selective estrogen response modulator, hormone replacement therapy, or oral contraceptive pill use, premenopausal status, or fluid yield on nipple suction aspiration. Compared with patients at Stanford University Medical Center, patients at the Dana-Farber Cancer Institute were significantly more likely to be parous (87.5% versus 62.7%, P = 0.03), to have breastfed (79.2% versus 43.1%, P = 0.006), and to have had a prior bilateral salpingo-oophorectomy (70.8% versus 39.2%, P = 0.01, data not shown). A catheter could be inserted into one or more ducts in 60 patients [80%, 95% confidence interval (69.5-87.7%)]. Four patients were African-American, one was Asian-American, and 70 were Caucasian. Given the very small number of patients who were not Caucasian, analyses by race were not done. The median age of all patients in whom ductal lavage was attempted was 43 years. Ductal lavage was considered successful if a catheter could be inserted into a duct, and saline instilled. In all patients who underwent successful catheterization, median age was 41 years, whereas in all patients who could not be catheterized, median age was 53 years (P = 0.0003). Patients who could be successfully catheterized were more likely to be premenopausal than patients who could not (51.7% versus 20%, P = 0.041). Only 1 of the 15 patients who could not be catheterized yielded fluid on nipple aspiration, compared with 29 of the 60 successfully catheterized patients (6.7% versus 48.3%, P = 0.0029). No significant differences in BRCA mutation status, prior breast or ovarian cancer, prior chemotherapy or radiation therapy, prior breast biopsy, prior selective estrogen response modulator use, hormone replacement therapy or oral contraceptive pill use, parity, breastfeeding, or prior bilateral salpingo-oophorectomy were noted between patients who could and could not be successfully catheterized.

Patient characteristics according to ductal lavage cytology are summarized in Table 2. Eight patients had insufficient cellular material for diagnosis [13.3% (6.7-24.5%)]; compared with the 35 patients with benign cytology [58.3% (45.8-70.0%)], patients with insufficient cellular material for diagnosis were more likely to have had prior breast cancer (62.5% versus 14.3%, P = 0.01), to have had prior chemotherapy (62.5% versus 20%, P = 0.028) or to have taken tamoxifen or another selective estrogen response modulator (50% versus 11.4%, P = 0.028). The median age of patients with insufficient cellular material for diagnosis was 44.5 years, and the median age of patients with benign cytology was 42 years (P = 0.072). Of 8 patients with insufficient cellular material for diagnosis, 1 yielded fluid on nipple suction aspiration, compared with 19 of 35 patients with benign cytology (12.5% versus 54.3%, P = 0.05). Seventeen patients were found to have mildly atypical cytology [28.3% (18.5-40.9%)]. Of 17 patients with atypia, 9 yielded fluid from any duct on nipple suction aspiration [52.9% (31.1-74.0%)]; in 12 of these 17 patients, the ducts which produced atypia were non–fluid-yielding. No significant differences in mean age, BRCA mutation status, prior breast or ovarian cancer, prior chemotherapy or radiation therapy, prior breast biopsy, prior selective estrogen response modulator, hormone replacement therapy or oral contraceptive pill use, parity, breastfeeding, menopausal status, or prior bilateral salpingo-oophorectomy were noted between patients with atypical and benign cytology.

Patient characteristics according to fluid-yielding status are summarized in Table 3. Twenty-nine patients had one or more fluid-yielding ducts on at least one occasion [48.3% (36.2-60.8%)] and 31 patients [51.7% (39.3-63.9%)] had only non–fluid-yielding ducts on all occasions. Patients with non–fluid-yielding ducts were significantly more likely than patients with fluid-yielding ducts to have had prior breast or ovarian cancer (48.4% versus 17.2%, P = 0.014) or prior chemotherapy (45.2% versus 17.2%, P = 0.027). No significant differences in mean age, BRCA mutation status, prior radiation therapy, prior breast biopsy, prior selective estrogen response modulator, hormone replacement therapy or oral contraceptive pill use, parity, breastfeeding, menopausal status or prior bilateral salpingo-oophorectomy were noted between patients with fluid-yielding ducts and with non–fluid-yielding ducts on univariate analysis. Logistic regression was used to identify independently predictive variables. Forward stepwise selection procedure, starting with the model with no predictors, was used for model-building purposes. Analysis was done using the statistical software S-PLUS version 6.1 for Windows (Insightful Corporation, Seattle, WA). All variables in Table 3 except for premenopausal status were allowed to enter the model; the premenopausal variable was removed because it was perfectly correlated with the bilateral salpingo-oophorectomy variable. As a result of this stepwise procedure, a model with a single predictor, prior breast or ovarian cancer, was selected. This variable had the smallest P value using Fisher's exact test. The same model was selected using backward stepwise selection, starting with a model containing 11 variables (all the variables in Table 3 excluding premenopausal status, prior breast and prior ovarian cancer variables). The variables of prior breast or ovarian cancer and prior chemotherapy were highly correlated: only one woman with prior history of cancer was not treated with chemotherapy, but for all others, a prior history of cancer implied having received chemotherapy. Given this fact, the effects of these two variables were difficult to separate, although both model selection procedures preferred the prior cancer variable.

Characteristics of patients with atypia, by fluid-yielding status, are described in Table 4. Of 17 patients with atypical cells, 12 had atypia from non–fluid-yielding ducts only [70.6% (46.8-87.2%)]. Of the remaining five patients, three had atypia from fluid-yielding ducts only [17.7% (5.5-42.1%)] and two had atypia from both fluid-yielding ducts and non–fluid-yielding ducts [11.8% (2.1-35.9%)]. Given the small numbers, patients with atypia from any fluid-yielding ducts [n = 5; 29.4% (13.1-53.7%)] were analyzed as a group. Patients with atypia from non–fluid-yielding ducts only were significantly more likely than patients with atypia from any fluid-yielding ducts to have had prior breast or ovarian cancer (58.3% versus 0%, P = 0.044) or to have had prior chemotherapy (58.3% versus 0%, P = 0.044). No significant differences in mean age, BRCA mutation status, prior radiation therapy, prior breast biopsy, prior selective estrogen response modulator, hormone replacement therapy or oral contraceptive pill use, parity, breastfeeding, menopausal status or prior bilateral salpingo-oophorectomy were noted between patients with atypia from non–fluid-yielding ducts only and patients with atypia from any fluid-yielding ducts.

To our knowledge, this is the first characterization of ductal lavage of non–fluid-yielding ducts in high-risk women, with a report on associated atypical cells. The present finding of atypical cells associated with non–fluid-yielding ducts, at least as frequently as in fluid-yielding ducts, suggests that non–fluid-yielding ducts in women with an inherited predisposition to breast cancer might be associated with higher risk than previously supposed.

Prospective evaluation of outcomes associated with breast duct cytology was first done using nipple suction aspiration (35, 41, 42). Collection of nipple aspirate fluid has been reported in a clinic and population-based sample of women at varying levels of breast cancer risk; women who did not yield nipple aspirate fluid (15% of the studied population) were chosen as the reference group, based on previous observations suggesting that such women would have the lowest risk of breast cancer (47). At a mean 12.7 years of follow-up, relative breast cancer risk of 1.8 was reported in women with normal nipple aspirate fluid cytology, and relative risk of 10.3 was reported in women with atypical nipple aspirate fluid cytology, versus those without nipple aspirate fluid (35). With increased patient numbers and years of follow-up, the authors reported relative breast cancer risk of 1.2 to 1.6 with normal nipple aspirate fluid cytology, and 2.0 to 2.8 with abnormal nipple aspirate fluid cytology, compared with a relative risk of 1.0 in women without nipple aspirate fluid (42). Most subsequent studies of ductal lavage have reported cannulating fluid-yielding ducts only (30, 31, 34), and thus little is known about the prevalence of abnormal cytology from non–fluid-yielding ducts. A recent study reported atypical cells in non–fluid-yielding ducts of lower-risk women (48). Our finding that 12 of 17 patients [70.6% (46.8-87.2%)] with atypical cells produced them from non–fluid-yielding ducts provides evidence that fluid yield is not a prerequisite for cytologic abnormality in high-risk women. It is consistent with a recent report that atypical cells have been collected by random periareolar fine-needle aspiration in patients with non–fluid-yielding ducts on suction (49). It may also provide some explanation for reports of ductal lavage's poor performance as a diagnostic tool in patients with known breast cancer: several of the breast cancer cases considered to have been missed by ductal lavage in a prior publication occurred in patients who had non–fluid-yielding ducts only, in whom ductal lavage was not attempted (40). Given the 2- to 5-fold increase in subsequent breast cancer risk observed in women with atypia in nipple aspirate fluid or on random periareolar fine-needle aspiration (36, 42), our results suggest that non–fluid-yielding ducts should be evaluated when ductal lavage is done in high-risk women. Furthermore, they suggest that reassessment of ductal lavage's performance as a diagnostic tool in women with known breast cancer is warranted, including lavage of non–fluid-yielding ducts as well as fluid-yielding ducts.

One potential explanation for our finding of atypia from non–fluid-yielding ducts might relate to the study population: patients in our breast screening protocol were selected because of their strong inherited predisposition to develop breast cancer. Notably, 7 of the 12 patients with atypia from non–fluid-yielding ducts had mutations in BRCA1, a finding which seems consistent with the high incidence of atypical hyperplasia reported in prophylactic mastectomy specimens of BRCA mutation carriers (50). Patients who carry BRCA1 mutations have a high incidence of estrogen and progesterone receptor–negative tumors; it could be that breast cancer risk in these patients, for which ductal atypia may be a biomarker, is less related to the hormonal factors thought to associate with fluid-yielding ducts than it is in other patient populations. However, one caveat to this hypothesis is the reduction in breast cancer risk seen in BRCA1 mutation carriers after oophorectomy (11-13, 51). It is important to note that the presence of mildly atypical cells collected via ductal lavage has not yet been prospectively associated with increased breast cancer risk, in BRCA mutation carriers or in other patient groups, and that studies with longer follow-up, in larger numbers of women, will be necessary to determine whether such an association exists. It will also be important to determine whether women with atypia from non–fluid-yielding ducts are at risk for different kinds of cancer (for example, a higher incidence of estrogen receptor–negative tumors) than women with atypia from fluid-yielding ducts. If so, then the combination of atypia and fluid-yielding status could have value as a prognostic biomarker, and as a surrogate end point for trials of targeted chemopreventive agents.

On analysis of successfully catheterized patients by fluid-yielding status, factors which differed significantly were: having a history of prior breast or ovarian cancer and having received chemotherapy. History of breast or ovarian cancer remained a significant predictor in multivariate analysis (given the very close correlation between cancer history and chemotherapy, the ability of chemotherapy to add to a model incorporating prior cancer was limited). Consideration of our results and of those previously reported suggests that nipple fluid production may be associated with reproductive and hormonal factors such as ovarian function; our finding of higher breast and ovarian cancer incidence among patients with non–fluid-yielding ducts may reflect the antihormonal maneuvers (selective estrogen response modulator use, bilateral salpingo-oophorectomy, and potential for chemotherapy-induced amenorrhea) used to treat these cancers. The potential relation between fluid-yielding ducts and ovarian function may partially explain the previously observed association of nipple aspirate fluid with increased breast cancer risk (35, 41, 42), given that longer exposure to higher levels of hormones produced by the ovary is likely a mechanism of this observed effect.

Previous authors have reported that age is related to fluid yield (47); a recent Australian study has confirmed the finding of higher fluid-yield and cell count on ductal lavage in premenopausal women (52). Our results show similar trends. The median age of patients who could not be successfully catheterized was 53 years; the median age of patients who could be successfully catheterized, but had only insufficient cellular material for diagnosis, was 44.5 years. Both numbers were larger (in the former case, significantly so) than the median ages of patients who could be successfully catheterized (41 years) or had benign cytology (42 years), respectively. Patients who could not be catheterized were significantly less likely to yield fluid on nipple suction aspiration or to be premenopausal than patients who could. Patients who yielded only insufficient cellular material for diagnosis on catheterization had a higher likelihood of prior breast or ovarian cancer, chemotherapy, selective estrogen response modulator use, and had a nonsignificant trend toward a lower rate of fluid on nipple suction aspiration than patients with benign cytology. Our findings and those of others suggest a decline in patency and fluid production of the ductal system, initially manifested by decreased cellularity of lavage specimens, and associated with falling levels of estrogen and progesterone (which would likely decline after treatment with chemotherapy or selective estrogen response modulators, and with rising age). They are consistent with known proliferative effects of estrogen on the mammary epithelium at various stages in development, as observed in murine models (53). Future studies of high-risk women who are postmenopausal or aged 50 or older should evaluate other methods of assessing ductal cytology, such as random periareolar fine-needle aspiration, which do not rely upon fluid production on aspiration or duct patency. Exploration of methods to increase duct patency, including use of topical nitroglycerin as has been previously reported, might also be effective in cytologic evaluation of this population (54).

Strengths of our study include its prospective, multi-institutional nature. Patients enrolled at the two institutions had generally similar clinical characteristics; the finding of a higher rate of bilateral salpingo-oophorectomy in the Dana-Farber Cancer Institute group may represent a higher percentage of BRCA1 (70.8% versus 50.9%, P = 0.14) and BRCA2 (29.1% versus 11.8%, P = 0.10) mutation carriers, although this did not reach statistical significance. The higher rate of parity and breastfeeding among the Dana-Farber Cancer Institute patients might reflect their slightly older median age (45 versus 43 years, P = 0.22), a difference which also did not reach statistical significance. Given the overall similarity of populations at both institutions, and the equivalent procedural and diagnostic methods used, these differences seem unlikely to have significantly affected the combined results. One limitation is the absence of a normal control group; future larger studies should include a group of average-risk women who will similarly be evaluated for atypia on ductal lavage, and for subsequent breast cancer incidence. As previously noted, the diagnosis of mild atypia by ductal lavage has not yet been prospectively associated with increased breast cancer risk (unlike atypical findings on nipple aspirate fluid or random periareolar fine-needle aspiration), and artifacts related to the lavage technique cannot be excluded as a contributor to the present findings (35, 36, 42). However, it is notable that a similar atypia rate was found in our population (28.3%) as has been found in other studies of ductal lavage (21%), and random periareolar fine-needle aspiration (24%) among cohorts of high-risk women, which suggests some consistency between these techniques (36, 37). Another limitation is the small number of women of races other than Caucasian, particularly given previous reports of racial differences in nipple aspirate fluid yield (47). Finally, time and technical limitations permitted cannulation of only 2 to 3 ducts per breast, from an estimated total of 6 to 12 (24); uncertainty remains as to whether ductal atypia represents a field effect, involving multiple breast ducts in an affected woman, or is specific to one or a few affected ducts only. With eventual improvements in ductal lavage technology, it is anticipated that the majority of breast ducts in one woman might be lavaged and their cytology compared, which would help to address this question.

The present results show that non–fluid-yielding ducts produce atypical cells in women with an inherited predisposition to develop breast cancer. They suggest that fluid-yielding status is inversely associated with prior cancer and its treatment by chemotherapy (perhaps consistent with an antihormonal mechanism), and that other strategies than ductal lavage may be preferable for cytologic evaluation of postmenopausal women, in whom successful catheterization was less often possible. Future studies of ductal lavage should include evaluation of non–fluid-yielding ducts, and alternate methods, such as random periareolar fine-needle aspiration, for the evaluation of women whose ducts cannot be cannulated, or yield only insufficient cellular material for diagnosis. Longer follow-up of a larger number of patients will be necessary to establish the clinical significance of ductal atypia in women at high inherited risk; we are currently embarked on a prospective, multi-institutional breast cancer screening trial which will address this question.