Reproducibility of cytologic atypia in repeat nipple duct lavage




It is believed that atypical cells identified by nipple duct lavage (NDL) indicate an increased risk for breast carcinoma similar to atypical ductal hyperplasia diagnosed by tissue biopsy, but many basic performance characteristics of NDL currently are undefined.


NDL was performed in 108 patients unselected for breast carcinoma risk and then was repeated after 2–14 months (median, 8 months) if the initial lavage was classified as atypical. Breast magnetic resonance images (MRIs) were obtained from a subset of patients who had atypical lavage results.


Marked atypia was diagnosed in 22% of 36 breasts with an incident carcinoma compared with 7% of 172 unaffected breasts (P = 0.01). After excluding breasts with an incident carcinoma, there were 32 patients (30%) with either mild or marked atypia. The lavage was repeated in 23 of these women, and the second lavage was classified as atypical in 48%. Neither marked atypia on the initial lavage nor a 5-year Gail risk ≥ 1.7% predicted atypia on repeat lavage, but there was a trend for improved reproducibility when the atypia initially was diagnosed in a fluid-producing duct. MRIs were abnormal in 13% of 24 breasts with an atypical lavage, and ductal carcinoma in situ was diagnosed subsequently in 1 breast.


Atypia frequently is diagnosed by NDL, but the reproducibility of repeat lavage is low. Lavage atypia may be physiologic or artifactual rather than pathologic in many instances. Marked atypia occasionally may represent mammographically occult ductal carcinoma in situ. Cancer 2005. © 2005 American Cancer Society.

Most breast carcinomas arise from the epithelial cells lining the ductal system, and atypical ductal epithelium is a marker of increased risk for the development of breast carcinoma. Wresch et al. followed 2300 women over 12.7 years and found that cytologic atypia in nipple aspirate fluid (NAF) was associated with a relative risk for breast carcinoma of 4.9.1 An updated analysis of those data adjusted the relative risk down to 2.8.2 Fabian et al. reported that women with a personal history of invasive or in situ breast carcinoma or with a 10-year Gail breast carcinoma risk > 4% who were found to have atypical cells on random fine-needle aspiration breast biopsy (FNAB) were 5 times more likely to develop breast carcinoma than women with a 10-year Gail risk < 4% and no atypia.3 Dupont et al. found that the risk of breast carcinoma was increased 4.3-fold in women who were diagnosed with atypical hyperplasia by surgical biopsy.4 Nipple duct lavage (NDL) has been proposed as a minimally invasive technique for obtaining breast epithelial cells, with the assumption that atypia identified by NDL confers the same relative risk for breast carcinoma as atypia identified by NAF, FNAB, or surgical biopsy. Although this has not been confirmed in a prospective trial to date, there is justifiable interest in developing and validating a minimally invasive procedure for the detection of atypical hyperplasia.

The National Surgical Adjuvant Breast Project (NSABP)-sponsored Breast Cancer Prevention Trial (BCPT-P1) demonstrated that 5 years of tamoxifen reduced the risk of breast carcinoma by ≈ 50% in increased risk women,5 but most eligible women refuse to take tamoxifen.6 Because women with atypical hyperplasia are at significantly increased risk for breast carcinoma and experience the greatest risk reduction with tamoxifen (86%), a test for atypia, such as NDL, may help eligible women decide to accept tamoxifen.

Although clinical guidelines for NDL were published previously.7, 8 many of the basic performance characteristics of the procedure remain unknown. A multiinstitutional study comparing NDL with NAF found that the insufficient sample rate was much lower for NDL than for NAF (22% vs. 73%) and that the atypia rate was much higher (24% vs. 10%).9 However, currently, it is unknown whether atypia diagnosed by NDL predicts an increased risk for breast carcinoma. Some atypical lavages may reflect underlying atypical hyperplasia, but others are likely to reflect reversible physiologic changes related to the hormonal milieu, benign intermediate-risk lesions (such as intraductal papilloma or papillomatosis), or fully developed ductal carcinoma in situ (DCIS). Lavage atypia that is not reproducible may be related to reversible physiologic changes in the breast epithelium, whereas atypia that is reproduced may be related to fixed, underlying pathologic alterations. We performed repeat NDL and magnetic resonance imaging (MRI) scans in women with lavage atypia to estimate the prevalence of persistent lavage atypia and the prevalence of mammographically occult DCIS or invasive carcinoma when atypical cells are identified.


Eligibility Criteria

The Institutional Review Board at the University of Texas Southwestern Medical Center at Dallas approved the protocol for this study, and written informed consent was documented for all participants. Patients from the Mary L. Brown Cancer Genetics and Risk Assessment Clinic at the University of Texas Southwestern Center for Breast Care were recruited for enrollment. A comprehensive risk assessment was performed for each patient. Patients with incident breast carcinoma and unaffected women age ≥ 18 years who presented for breast carcinoma risk assessment were offered duct lavage regardless of their calculated risk level. Exclusion criteria included the presence of an undefined palpable or mammographic breast lesion suspicious for malignancy; bilateral prophylactic mastectomy; any prior breast irradiation; any systemic chemotherapy in the past; a performance status that restricted normal activity for a significant portion of the day; current use of androgens, luteinizing hormone-releasing hormone analogs, prolactin inhibitors, antiandrogens, or corticosteroids (women were eligible if these drugs were discontinued 3 months prior to lavage); ever use of tamoxifen, raloxifene, or other selective estrogen receptor modulator (SERM) therapy; or pregnancy or lactation within 6 months.

NDL Procedure

Local anesthetic cream (EMLA; AstraZeneca, London, United Kingdom) was applied to the nipple and then covered with an occlusive patch 1–2 hours prior to the procedure. At the start of the procedure, the patient performed a self-breast massage, after which, the nipple was dekeratinized with a mild abrasive gel (Nuprep; D. O. Weaver and Company, Aurora, CO). The operator then continued the breast massage in an effort to express NAF. If no NAF was elicited manually, then a nipple aspirator (FirstCyte; Cytyc Health Corporation) was used. Fluid-producing ducts initially were cannulated with a tapered dilator coated with 2% lidocaine jelly, after which, a ductal lavage microcatheter (FirstCyte Microcatheter; Cytyc Health Corporation) was inserted. Saline (10 mL) was infused into the duct in 0.5-mL increments, and the effluent fluid was aspirated. An attempt was made to lavage all fluid-producing ducts and at least one nonfluid-producing duct from each breast. The location of each cannulated duct orifice was recorded on a circular grid with 45 cells, so that the orifice of any duct that yielded atypical cells could be identified in the future. Repeat lavage was offered to women whose initial lavage returned atypical cells. At the time of repeat lavage, every effort was made to recannulate the same ducts that were cannulated at the initial lavage. Breast MRI was recommended for all women whose initial lavage returned atypical cells but was performed only if third-party payor approval could be obtained.

Cytologic Evaluation

The lavage effluents from each duct were collected separately in 30 mL of CytoLyt solution (Cytyc Health Corporation). Cytology slides were prepared using the ThinPrep method and were stained using the Papanicolaou technique. All slides were evaluated by the same breast cytopathologist (R.A.), who classified each sample according to the most severe alterations identified: insufficient for diagnosis, normal epithelium or apocrine metaplasia only, typical epithelial hyperplasia, mild atypia, or marked atypia. Cytologic interpretation was performed according to the guidelines published by the Cytyc Health Corporation ( Briefly, mild atypia was defined as clusters of crowded, overlapping cells with slight nuclear enlargement, mild anisonucleosis, prominent nucleoli, occasional myoepithelial cells, and granular, evenly distributed chromatin. Marked atypia was diagnosed when these same features were more pronounced and included marked anisonucleosis, significantly increased nuclear:cytoplasmic ratios, and irregular, clumping chromatin.

Statistical Analysis

Proportions and atypia prevalence rates were compared using the Fisher exact test. The α value was set at 0.05.



Ductal lavage was performed in 377 ducts from 208 breasts in 108 female patients. On average, 1.8 ducts were lavaged per breast, and 3.5 ducts were lavaged per patient. There were 41 women with incident breast carcinoma and 67 women who were unaffected with breast carcinoma but who had completed a comprehensive breast carcinoma risk assessment. Among the women who were unaffected with breast carcinoma, 52% had a 5-year Gail risk < 1.7%, and 48% had a 5-year Gail risk ≥ 1.7%. The mean patient age was 46.3 years (range, 30–82 years), and 42% of patients were postmenopausal. Most of the patents were Caucasian (81%), and 35% of patients were using oral contraceptives or hormone replacement therapy at the time of initial sampling. NAF was expressible from 86% of the patients (Table 1).

Table 1. Characteristics of the Study Sample
CharacteristicNo. of patients (%)
  1. DCIS: ductal carcinoma in situ; ADH: atypical ductal hyperplasia.

Total patients108 (100.0)
Age (yrs) 
 Caucasian87 (80.6)
 African American16 (14.8)
 Hispanic4 (3.7)
 Asian1 (0.9)
Expressible nipple aspirate fluid93 (86.1)
Menopausal status 
 Premenopausal56 (51.9)
 Perimenopausal7 (6.5)
 Postmenopausal45 (41.7)
Oral contraceptive use (premenopausal)16/56 (28.6)
Hormone replacement (perimenopausal and postmenopausal)22/52 (42.3)
Risk groups 
 Breasts ipsilateral to a breast carcinoma36 (100.0)
  DCIS only3/36 (8.3)
  Infiltrating ductal carcinoma29/36 (80.6)
  Infiltrating lobular carcinoma3/36 (8.3)
  Medullary carcinoma1/36 (2.8)
  Any associated DCIS29/36 (80.6)
 Breasts contralateral to a breast carcinoma38
Unaffected risk assessed patients67/108 (62.0)
 History of ADH1/67 (1.5)
 BRCA gene mutation3/67 (4.5)
 5-Yr Gail risk 
  0.01–0.8520/67 (29.9)
  0.86–1.6915/67 (22.4)
  1.70–2.5417/67 (25.4)
  > 2.5415/67 (22.4)

Frequency of Atypia

Table 2 summarizes the atypia rates for 36 breasts with an incident breast carcinoma, 38 breasts contralateral to an incident breast carcinoma, and 134 breasts from women who were unaffected by breast carcinoma. Results are reported separately for the right and left breasts of the unaffected women to permit comparisons between patients with breast carcinoma and unaffected patients on a per-breast basis. The insufficient sample rate was higher for ducts from breasts with an incident carcinoma (40%) than for ducts from breasts that were unaffected with breast carcinoma (27%; P = 0.06). Atypia of any degree was diagnosed in 36% of breasts with an incident breast carcinoma and in 24% of breasts that were unaffected with breast carcinoma (P = 0.19), but marked atypia was diagnosed more frequently in breasts with an incident breast carcinoma (22%) than in unaffected breasts (7%; P = 0.01). Among breasts that were unaffected with breast carcinoma, we diagnosed cytologic atypia in 18% of ducts, 24% of breasts, and 30% of patients. There were no trends in the insufficient sample rate or in the frequency of diagnosis of mild or marked atypia for initial lavages over time (Fig. 1).

Table 2. Frequency of Atypia by Sampling Group
VariableNo. of patients (%)
ICMDMild atypiaMarked atypiaAny atypiaICMDMild atypiaMarked atypiaAny atypia
  1. ICM: insufficient cellular material for diagnosis.

Cancerous breast24/60 (40.0)7/60 (11.7)8/60 (13.3)15/60 (25.0)11/36 (30.6)5/36 (13.9)8/36 (22.2)13/36 (36.1)
Contralateral breast19/57 (33.3)6/57 (10.5)3/57 (5.3)9/57 (15.8)10/38 (26.3)5/38 (13.2)3/38 (7.9)8/38 (21.1)
Unaffected right breast35/133 (26.3)15/133 (11.3)7/133 (5.3)22/133 (16.5)10/67 (14.9)12/67 (17.9)5/67 (7.5)17/67 (25.4)
Unaffected left breast32/127 (25.2)19/127 (15.0)6/127 (4.7)25/127 (21.3)13/67 (19.4)12/67 (17.9)4/67 (6.0)16/67 (23.9)
Figure 1.

Insufficient cellular material for diagnosis (ICMD) rate and frequency of mild and marked atypia for initial lavages over time. The entire study sample was divided into quintiles to detect trends in the insufficient sample rate or in the frequency of diagnosis of mild or marked atypia that might suggest systematic changes in the performance or interpretation of nipple duct lavage over time. No such trends were identified. Error bars bracket the 95% confidence intervals. ♦: ICMD; •: Mild atypia; ▪: Marked atypia.

Reproducibility of Atypia

Among the 32 patients who had an atypical lavage from a breast that was unaffected with breast carcinoma, repeat lavage was performed for 23 patients. Four patients with breast carcinoma received chemotherapy after the initial lavage, which rendered them ineligible for repeat lavage of the contralateral breast; two women without breast carcinoma moved out of the area; two women declined repeat lavage; and one woman underwent mastectomy without repeating the lavage after an MRI was interpreted as highly suspicious (Fig. 2). Repeat lavage was performed 2.3–14.3 months (median, 8.3 months) after the initial lavage. Because every effort was made to relavage all of the ducts that had been lavaged initially in the patients who had at least 1 duct diagnosed as atypical, a total of 78 ducts from 45 breasts were relavaged in these 23 patients. If any duct was classified as atypical on the repeat lavage, then the atypia was scored as “reproducible” for that patient. The repeat lavage was classified as atypical for 11 of 23 patients (48%), 11 of 32 breasts (34%), and 8 of 42 ducts (19%) that were diagnosed initially as atypical (Table 3). Failure to reproduce the atypia was due to an insufficient sample on the second lavage in 13% of patients and due to a diagnosis of normal epithelium or typical hyperplasia only in 39% of patients. Marked atypia on the initial lavage was no more predictive of an atypical second lavage than mild atypia (44% vs. 50%; P = 1.0).

Figure 2.

An interesting patient. (a) A screening mammogram revealed a suspicious cluster of microcalcifications in the right breast. The left breast was interpreted as normal. (b) Nipple duct lavage of the left breast at the time of right mastectomy returned markedly atypical cells. (c) This left breast magnetic resonance image demonstrates three areas of abnormal enhancement. (d) A subsequent left total mastectomy revealed 10 cm of ductal carcinoma in situ.

Table 3. Classification of Repeat Lavage Results According to Initial Lavage Results
Repeat lavageInitial lavage: No. of patients (%)
Mild atypiaMarked atypiaAny atypia
MarkedMildNormal or EHICMDAny atypiaMarkedMildNormal or EHICMDAny atypia
  1. EH: typical epithelial hyperplasia; ICMD: insufficient cellular material for diagnosis.

By patient3/14 (0.21)4/14 (0.29)5/14 (0.36)2/14 (0.14)7/14 (0.50)2/9 (0.22)2/9 (0.22)4/9 (0.44)1/9 (0.11)4/9 (0.44)11/23 (0.48)
By breast4/22 (0.18)5/22 (0.23)9/22 (0.41)4/22 (0.18)9/22 (0.41)1/10 (0.10)1/10 (0.10)6/10 (0.60)2/10 (0.20)2/10 (0.20)11/32 (0.34)
By duct3/30 (0.10)4/30 (0.13)13/30 (0.43)10/30 (0.33)7/30 (0.23)0/12 (0.00)1/12 (0.08)9/12 (0.75)2/12 (0.17)1/12 (0.08)8/42 (0.19)

Among the patients who produced NAF, atypia was diagnosed on the second lavage in 55%, compared with 0% for the 3 patients who did not produce NAF (P = 0.25). Among the patients who had a 5-year Gail risk ≥ 1.7%, atypia was diagnosed on the second lavage in 22% of patients, compared with 70% of the patients who had a 5-year Gail risk < 1.7% (P = 0.10). Reproducibility rates were similar for premenopausal women compared with perimenopausal or postmenopausal women, for women who were taking hormonal medications compared with women who were not taking these medications, and for women who underwent repeat lavage < 8.3 months after the initial lavage (the median interval for this series) compared with women who underwent repeat lavage ≥ 8.3 months after the initial lavage.

MRI Findings

MRI was performed in 24 breasts from 17 women whose initial lavage was interpreted as atypical. The atypia was marked in 9 breasts, and the MRI was abnormal in 1 of those breasts (11%). Total mastectomy revealed 10 cm of DCIS in this patient (Fig. 2). Repeat lavage was not performed prior to the MRI and subsequent surgery. The initial lavage was interpreted as mildly atypical in 15 breasts; and of those, the MRI was abnormal in 2 breasts (13%). The MRI was interpreted as borderline suspicious in both of those breasts. Repeat MRI in one patient demonstrated resolution of the region of abnormal enhancement, and repeat lavage in this patient was classified as typical epithelial hyperplasia only. In the second patient with mild atypia, a repeat lavage was classified as marked atypia, and a targeted ultrasound examination was unremarkable. Repeat MRI in this patient was interpreted as entirely normal.


NDL has been proposed as a secondary risk-stratification procedure for women who are determined to be at increased risk for breast carcinoma based on epidemiologic models. It is believed that the atypical cells identified by NDL confer the same breast carcinoma risk as atypia found in nipple aspirates, needle biopsies, and surgical biopsies; however, currently, there are no data to confirm this, and many of the essential performance characteristics of NDL are yet to be elucidated.

Although NDL currently is recommended for risk stratification, and not for early detection of breast carcinoma, it is reasonable to suppose that, if lavage atypia is a strong predictor of breast carcinoma risk, then it would occur at a high frequency in breasts with an incident breast carcinoma. Marked atypia, in fact, was more common in breasts with an incident breast carcinoma (22%) than in unaffected breasts (6–8%). This is a higher marked atypia rate than that reported for a series of 28 mastectomy patients.10 In that study, atypia rates were reported per lavage sample (ducts) rather than by breast, and marked atypia was identified in only 4 of 29 (14%) adequate samples from patients with an incident breast carcinoma. The exclusion of patients who had insufficient samples from the current series resulted in a marked atypia rate of 8 of 25 (32%) for breasts with an incident carcinoma compared with 12 of 139 (9%) for unaffected breasts (P = 0.003). In the prior study, the extent of carcinoma in situ appeared to correlate with the degree of atypia identified in the lavage samples, but DCIS was identified in 23 of 28 patients (82%), and it was not possible to determine whether or not lavage atypia was associated primarily with DCIS. Similarly, 29 of 36 patients (81%) with incident breast carcinoma in the current series had a DCIS component, but marked atypia was identified in 3 of 7 patients (42%) who had no DCIS component. It is clear that the marked atypia rate for breasts with an incident breast carcinoma is considerably higher than that previously reported and is considerably higher than that observed for breasts that are unaffected by breast carcinoma.

The prevalence of lavage atypia measured in our series of patients, who were unselected for breast carcinoma risk, was nearly identical to that reported in the first large validation series, a series that was limited to women who were at increased risk for breast carcinoma9 (Table 4). To make this comparison, we have considered only breasts that were unaffected by breast carcinoma (38 breasts contralateral to a breast carcinoma and 134 breasts from women who were unaffected with breast carcinoma). Results for 3 women with breast carcinoma who did not undergo lavage of the contralateral breast were excluded, leaving a total of 105 patients for comparison.

Table 4. Comparison of Insufficient Cellular Material for Diagnosis and Atypia Rates in the Current Series with Previously Reported Rates
ResultsNo. of patients (%)
Dooley et al., 20019Current series
  1. ICMD: insufficient cellular material for diagnosis.

By patient  
 ICMD84/383 (22)17/105 (16)
 Mild atypia66/383 (17)19/105 (18)
 Marked atypia26/383 (7)13/105 (12)
By duct  
 ICMD173/591 (29)86/317 (27)
 Mild atypia77/591 (13)40/317 (13)
 Marked atypia28/591 (5)16/317 (5)

Atypia was reproduced on repeat lavage in only 8 of 42 ducts (19%), 11 of 32 breasts (34%), and 11 of 23 patients (48%). A recent series of duct lavages in 38 high-risk women reported an atypia prevalence of 23% for those with an adequate sample and reproducibility of the atypia in only 1 of the 4 patients who underwent repeat lavage.11 The reproducibility of any cytology-based screening test will be related to the physiologic factors that affect the cytologic features of the cells collected, variability in the sampling procedure, and variability in the cytologic interpretation. The breast is an exquisitely hormone-responsive organ, and fluctuations in the hormonal milieu may affect cytology-based screening tests. Exogenous estrogens have been associated with atypical hyperplasia in humans,12 monkeys,13 and rodents,14 but it is not known whether these lesions are reversible in humans when the hormones are withdrawn. It is noteworthy that nine of our patients who underwent repeat lavage were using hormonal medications at the time of the initial lavage. Atypia was reproduced in 2 of the 4 patients (50%) who discontinued these medications between the first and second lavages. Endogenous estrogens also may influence the cytologic appearance of breast epithelial cells, although Mitchell et al. found no significant changes in breast epithelial cells recovered from weekly NAF samples that were collected over two menstrual cycles.15

Although every effort was made to recannulate the same ducts that were diagnosed as atypical on the initial lavage, technical problems with resampling may have contributed to the low reproducibility. The shear numbers of duct orifices clustered near the center of the papilla (11–48 orifices; median, 27 orifices)16 presents a challenge for recannulation of a specific duct. This may have been compounded by our study design, which permitted cannulation of duct orifices that were not producing NAF, because it is likely that NAF production, in conjunction with location information recorded on a grid, provides valuable visual cues for reidentifying a specific duct orifice. The insufficient sample rate on relavage for ducts that initially returned atypical cells was 12 of 42 samples (29%), essentially identical to the insufficient sample rate for the initial series of lavages. Possible reasons for the failure to obtain an adequate sample from ducts initially yielding atypical cells include inadvertent cannulation of a different duct orifice, regression of an atypical proliferative lesion either as a consequence of the initial lavage procedure or for reasons unrelated to the initial lavage, and ductal injury at the time of the initial lavage that precluded satisfactory recannulation and lavage. The same operator performed all of the lavage procedures, and insufficient sample rates were stable over time (Fig. 1), excluding differences in operator experience or technique as factors in the assessment of reproducibility. It has been suggested that the use of normal saline for NDL can induce artifactual atypia and that plasmolyte is a superior lavage solution. Because all initial and repeat lavages were performed using normal saline, this is unlikely to have influenced reproducibility rates.

Finally, interobserver and intraobserver variation in cytology scoring can impact the reproducibility of cytologic screening tests. Interobserver variability was excluded by having the same cytopathologist evaluate all of the samples from this study. It is possible that atypia was over-called in the earliest period of the study, resulting in lower atypia rates for the repeat lavages, but this is unlikely, because the frequency of diagnosis of mild or marked atypia for the initial lavages was stable over time (Fig. 1).

It is likely that a combination of physiologic and technical factors accounted for the low reproducibility of repeat lavage measured in this series. Neither marked atypia on the initial lavage nor a 5-year Gail risk ≥ 1.7% predicted atypia on repeat lavage, but there was a trend toward improved reproducibility when the atypia initially was diagnosed in a NAF-producing duct. We previously reported that the frequency of lavage atypia is similar for patients with a 5-year Gail risk < 1.7% and ≥ 1.7% and for ducts that produce NAF compared with ducts that do not produce NAF.17 The high prevalence of lavage atypia noted in this and prior studies, combined with a low reproducibility, makes it unlikely that a single NDL demonstrating either mildly or markedly atypical cells will predict a high risk for breast carcinoma. Adjunctive tests, such as tumor suppressor gene methylation status,18 loss of heterozygosity analysis,19 or chromosome copy number determination,20, 21 are feasible for NDL samples and may improve the predictive value of NDL cytology.

Lavage atypia was associated with significant MRI abnormalities in only 1 of 24 breasts. The atypia in this breast was classified as marked and ultimately was diagnosed as ductal carcinoma in situ. In 23 breasts with atypical lavage results, there were no reproducible MRI findings. This is in marked contrast to a recent series that identified MRI abnormalities in six of seven breasts with atypical lavages.11 Only one of those breasts was biopsied, and the diagnosis was atypical ductal hyperplasia. We performed MRI only in women whose third-party payors agreed to reimburse for the test (17 of 32 women). This is likely to have biased our results; however, given the low frequency of MRI findings, we currently are performing MRI only if marked atypia is confirmed on repeat lavage.

Ductal lavage is an excellent tool for retrieving breast epithelial cells, but the reproducibility of serial sampling is poor. In addition, lavage atypia is associated only infrequently with MRI findings but may represent mammographically occult DCIS. Until prospective studies validate lavage atypia as a marker for breast carcinoma risk, it is best to use it in the context of clinical trials.