Computer-assisted complete three-dimensional reconstruction of the mammary ductal/lobular systems

Implications of ductal anastomoses for breast-conserving surgery

Authors


Abstract

BACKGROUND

The intraductal spread of breast carcinoma can occur along the mammary ductal/lobular systems (MDLS) with no invasion of tissues. Because ductal anastomoses in the MDLS are considered to be a possible risk factor for extensive intraductal spread of breast carcinoma, the architecture of the MDLS has important therapeutic implications for patients treated with breast-conserving surgery.

METHODS

An entire breast resected by subcutaneous mastectomy from a 69-year-old woman with ductal carcinoma in situ (DCIS) was examined in submacroscopic sections by stereomicroscopic and histologic techniques. Serial 2-mm sections underwent computer-assisted complete three-dimensional reconstruction of all MDLS.

RESULTS

The entire breast that was studied contained 16 MDLS that were arranged radially, with the nipple at the center. Of these 16 MDLS, 4 (25.0%) had ductal anastomoses whereas the remaining 12 MDLS had no ductal anastomoses and completely independent regional anatomy. Ductal anastomoses were observed at 11 sites in the 4 MDLS. The 2 of 11 ductal anastomoses that connected different MDLS (18.2%) were situated > 4 cm from the nipple. The remaining nine ductal anastomoses connected ducts within the same MDLS; their location varied from near the nipple to the peripheral region. In the specimen examined, DCIS extended only within a single MDLS and did not spread between different MDLS via ductal anastomoses.

CONCLUSIONS

To the authors' knowledge, the current study is the first time the complete architecture of all MDLS in an entire breast has been studied three-dimensionally. The risk of promoting the intraductal spread of disease during surgery may be greater when intraductal lesions extend more peripherally than centrally. The features of ductal anastomoses may provide a significant anatomic clue regarding negative surgical margins in breast-conserving surgery. Cancer 2001;91:2263–72. © 2001 American Cancer Society.

In recent years, the results of large randomized controlled trials of breast carcinoma conducted in Europe and the U.S.1–3 have favored the wide acceptance of breast-conserving therapy,4, 5 in which partial mastectomy with postoperative radiotherapy is substituted for total mastectomy. However, local recurrence developing from residual carcinoma cells that have spread intraductally is a common problem with breast-conserving therapy.6 Improving local tumor control after breast-conserving therapy requires more effective strategies for resecting noninvasively extending components of breast carcinoma without leaving residual tumor. However, even with the advanced diagnostic imaging techniques that have been developed in recent years, the accurate prediction of the extent of intraductal spread of breast carcinoma prior to surgery appears impossible.7 This difficulty also is evident from the fact that satisfactory local control of ductal carcinoma in situ (DCIS) is not obtained even with breast-conserving therapy including postoperative radiotherapy.8 In breast-conserving therapy for DCIS, even when histologic examination indicates a tumor-free surgical margin, reliable local control of the regional intraductal lesion is difficult to achieve because of intraductal spread.9 In breast-conserving therapy, pathologic study of intraductal spread of carcinoma is very important to determine the likelihood of residual carcinoma at the surgical margins, but even histologic techniques using serial two-dimensional (2-D) thin sections is not enough to identify intraductal spread reliably.

In the intraductal spread of carcinoma, noninvasive tumor components are considered to develop in terminal ducts or lobules such as terminal ductal lobular units (TDLU)10 or terminal end buds (TEB),11 and then extend in a central direction through the duct.12–15 In fact, the direction and extent of this intraductal extension are diverse,16 and intraductal spread must be interpreted precisely from a three-dimensional (3-D) viewpoint with regard to anatomic units, the mammary ductal/lobular systems (MDLS). To our knowledge, we are the first group of authors worldwide to develop and report a particularly accurate technique for the 3-D reconstruction of several MDLS using surgical specimens from breast-conserving surgery.17 By this method, the intraductal spread of breast carcinoma was found to occur in a continuous manner along MDLS, with ductal anastomoses between different MDLS representing a risk factor for the extensive intraductal spread of carcinoma through multiple MDLS. However, to our knowledge the anatomic details of all MDLS in an entire breast have not been investigated to date, and many details concerning the branching and courses of the mammary ducts in the adult female breast are unknown. Accurate anatomic description of MDLS including ductal anastomoses would be very useful for planning effective breast-conserving therapy, and should result in improved local tumor control.

In the current study we successfully produced the first computer-assisted, complete 3-D reconstruction of all MDLS in an entire breast by applying stereomicroscopic techniques involving serial submacroscopic sections, and using this reconstruction to analyze the anatomic features of the mammary ducts. Furthermore, using computer-imaging techniques, we displayed the mode of spread of DCIS along MDLS in this breast specimen.

MATERIALS AND METHODS

Breast Specimen

The study was performed on a left subcutaneous mastectomy specimen from a 69-year-old postmenopausal woman. Subcutaneous total mastectomy had been performed for DCIS that presented with bloody nipple discharge. The cytologic examination of a specimen from the nipple discharge suggested that the lesion was malignant. In preoperative physical examinations, no tumor was palpable in the left breast. Findings on mammography included extensive microcalcification in the region corresponding to the left lateral inferior quadrant, in which pressure elicited the bloody nipple discharge. With a clinical diagnosis of noninvasive breast carcinoma, subcutaneous mastectomy was performed under general anesthesia. The skin was incised along the mammary fold of the left breast, and the entire breast was detached carefully along the subcutaneous tissue plane so the parenchyma of the mammary gland would not be incised. The converged collecting ducts were severed immediately under the nipple and the entire left mammary gland was resected together with the fascia from the pectoralis major muscle.

Preparation of Serial Submacroscopic Sections

To completely fix the entire breast specimen without incising it, a 20% formalin solution that also contained 10% methanol was used. This solution ensures good formalin permeation in specimens with abundant fat. To maintain the preoperative shape of the breast, immersion fixation was performed for approximately 2 weeks with the specimen in the horizontal position. To render the fixed specimen sufficiently firm for the mechanical preparation of serial submacroscopic sections, it was washed in cold running water for 4 hours and in warm running water for 15 minutes, and then immersed in a solution of 20% gelatin dissolved in hot water and maintained at 70 °C. The specimen was solidified overnight at 4 °C in a refrigerator where it was embedded in a cold gelatin block. The gelatin-embedded specimen was set on an electrically powered ham slicer (OMS-812C; Ohmichi, Maebashi, Japan) and serial parallel sections measuring 2 mm in thickness were prepared mechanically with equal spacing, resulting in a total of 86 sections. Washing in hot running water at 70 °C for 2 hours then removed the gelatin from around the sections. After 1 hour of washing in cold running water, all submacroscopic sections were stained with Gill hematoxylin No. 2 for approximately 30 minutes and then washed in cold running water for 1 hour with occasional agitation. After sections were decolorized for 5 minutes with 1% acid alcohol, all mammary glandular elements in each section remained clearly stained with hematoxylin, and mammary ducts and lobules could be recognized readily (Fig. 1). The sections then were washed overnight in cold running water, dehydrated in 95% ethanol for 8 hours, and dehydrated further in 99.5% ethanol for 48 hours. Dehydration and fat removal were performed for 2 days. In these serial submacroscopic sections in which the mammary ducts and lobules were clearly defined, the same side of each section was photographed in its entirety; these images were developed and enlarged at a magnification of ×5. Informative areas with respect to mammary ducts and lobules were marked on corresponding photographs for 3-D reconstruction. The dehydrated and defatted submacroscopic sections were rinsed with acetone for 30 minutes and were drained gradually and transferred to a methyl salicylate solution over a period of 2 hours. In submacroscopic sections permeated with methyl salicylate, the mammary stroma and fat tissues were rendered transparent, resulting in unobstructed 3-D visualization of the running ducts and lobules (Fig. 2a). The submacroscopic sections were packaged in heavy polyethylene bags containing sufficient methyl salicylate solution. Air bubbles were excluded carefully and each section was sealed tightly using a vacuum packing machine and sequentially numbered and labeled in the order of consecutive slicing.

Figure 1.

A submacroscopic section measuring 2 mm in thickness and stained with hematoxylin. Mammary glandular elements are demonstrated clearly, indicating the sites of all ducts and lobules in the section (arrows) (original magnification ×2).

Figure 2.

Submacroscopic and histologic photographs of the ductal carcinoma in situ (DCIS) lesion. (a) Stereomicroscopic view of the mammary glandular trees in a thick section, showing intraductal extension of DCIS (arrows) in a central direction, toward the nipple (hematoxylin, original magnification ×4). (b) Magnified histologic view of the lesion shown in the thick section. A noninvasive ductal carcinoma of comedo type shows central necrosis (H & E, original magnification ×150).

Stereomicroscopic Study of Serial Submacroscopic Sections

In the submacroscopic sections placed in heavy polyethylene bags, courses and branching of all mammary ducts and lobules in the entire breast were examined using a dissecting stereomicroscope (SZH-131; Olympus, Tokyo, Japan). Good continuity of observations was attained by focusing up or down. Specifically, stereomicroscopy was used to observe in detail how the cut profiles of all mammary ducts and lobules on the submacroscopic section diverged, converged, and were connected to the corresponding profiles of mammary ducts and lobules in the next submacroscopic section. After the submacroscopic examination, methyl salicylate was removed completely from each section using 99.5% ethanol. Then, each submacroscopic section was embedded in paraffin with the same side up, and 3-μm thin sections were prepared and stained with hematoxylin and eosin. The entire breast then was examined histologically at equally spaced intervals of 2 mm with parallel cutting to study how each mammary duct was related to lobules and converged and branched while passing to the subsequent submacroscopic section with the aid of stereomicroscopic 3-D observation. By combining the two observational techniques of histologic and submacroscopic examination, the details of all MDLS in the entire breast could be studied definitively with excellent continuity.

Three-Dimensional Reconstruction

For computer-assisted 3-D reconstruction, all contours of the mammary ducts and lobules of each submacroscopic section were input to a computer graphics system from the corresponding magnified photographs of the cut surface. The system included a desktop computer (PC-9801FS; NEC, Tokyo, Japan) with a 3-D reconstruction program (TRI; Ratoc, Tokyo, Japan), and a computer-imaging processor (Spicca II; Avionics, Tokyo, Japan) with a full-color memory-mapped graphics display of 512 × 480 pixel resolution (BD-410CA; Avionics, Tokyo, Japan). A sheet of serial photographs magnified from submacroscopic sections was placed on a digitizer in which all the contours of the mammary ducts and lobules were entered on the computer-imaging processor using a monochrome CCD video camera (XC-77; Sony, Tokyo, Japan) and the resulting polygon files were analyzed and stored for each section. This input manipulation was repeated for all photographs of submacroscopic sections, and the contours of the mammary ducts and lobules of the entire breast were converted to large polygon files. Next, all polygon files were interconnected and reconstructed three-dimensionally by the TRI program. The computer could display the 3-D images rotated at any given angle around the X-, Y-, and Z-axes. The plane of each submacroscopic section was regarded as the X-Y plane. In the current study, two computer graphics techniques described as the “solid model” and the “network model,” respectively, were used consecutively to analyze the 3-D reconstructed images. In the solid model, the external structural appearance of a reconstructed object was transformed to an infinitely smooth surface by computer processing to reproduce a 3-D figure more closely resembling the real object, permitting the observation of the courses and branching of the mammary ducts. Conversely, the network model visualized the complex intertwined duct network skeletally or schematically using fine lines, directing attention to the courses and branching conditions of the mammary ducts and lobules. Depending on the purpose of the observation, different colors were used to delineate normal mammary ductal trees, DCIS, and sites of ductal anastomoses. The MDLS is a continuous collective entity of mammary lobules and ducts progressively converging toward the nipple from the peripheral region of the mammary gland. When mammary ducts joined with other mammary ducts apart from the overall convergence pattern, the connection was defined as a ductal anastomosis.

RESULTS

Stereomicroscopic observation permitted the examination of the courses and branching of all MDLS in the breast in its entirety, from the terminal ducts and lobules to the collecting ducts. In the region of the DCIS, the mammary ducts were expanded in association with the intraductal spread of carcinoma (Fig. 2a). Not only could the courses and branching of normal mammary ducts be observed stereomicroscopically with good continuity by the submacroscopic technique, but all intraductal spread of carcinoma could be traced.

In the case of the current study, bloody nipple discharge was found prior to surgery. However, in our submacroscopic and stereomicroscopic examination of the resected specimen, all the tumor lesions were of typical comedo-type DCIS. There was no sign of intraductal papillomas (which often are found in the lesions with bloody nipple discharge) in either the MDLS with DCIS or in the other normal MDLS.

Figure 3 is a computer graphic image depicting the current study case as a solid model including all MDLS obliquely from beneath. The nipple is at the center of the image and the axilla is at the furthest right. Yellow was used to depict the mammary duct with normal epithelia and red was used for the DCIS lesion including extensive intraductal spread of the carcinoma. The majority of MDLS were arranged radially with respect to the nipple. DCIS of comedo type was confirmed histologically in MDLS from which a bloody discharge could be expressed from the nipple (Fig. 2b). Continuous and extensive intraductal spread of DCIS filled one MDLS from the terminal ducts to the collecting ducts. No multicentric or invasive malignancy was observed in the specimen.

Figure 3.

Complete three-dimensional solid model of the full extent of all mammary ductal/lobular systems (MDLS) in the entire breast. All MDLS are oriented radially, with the nipple at the center (arrow). Note a single MDLS involved by ductal carcinoma in situ (DCIS) (DCIS is shown in red and the normal ducts are shown in yellow).

Figure 4 is a network model of the same specimen viewed from immediately above, with complete 3-D reconstruction of MDLS. In this model individual systems are identified by different colors. The individual MDLS showed a repeated and irregular branching pattern and a sector-shaped overall distribution. The breast examined in the current study included 16 MDLS. The majority of MDLS showed a configuration that branched repeatedly, converging from the margins of the breast to the nipple, whereas some small MDLS showed branching only near the nipple. Several MDLS overlapped one another in the same region of the breast.

Figure 4.

Complete three-dimensional network model showing all mammary ductal/lobular systems (MDLS) individually identified by different colors. This breast specimen included 16 sector-shaped MDLS exhibiting repeated and irregular branching of ducts, here shown skeletally or schematically.

The current anatomic study of complete MDLS disclosed ductal anastomoses at various sites in these systems. Figure 5 indicates the positions of ductal anastomoses in different colors in the network model of the specimen. The network representations of the complete MDLS were shown in yellow, whereas ductal anastomoses connecting different MDLS were shown in red and ductal anastomoses within the same MDLS were shown in blue. In the current study, 11 ductal anastomoses were found in the entire breast. Only 2 of these anastomoses (18.2%) connected different MDLS and thus posed a risk of extensive intraductal spread of breast carcinoma. The remaining nine ductal anastomoses were within the same MDLS. The MDLS in which DCIS had developed showed no ductal anastomosis communicating with other MDLS; its ductal anastomoses were only observed within it. Although the intraductal spread of DCIS was extensive in both central and peripheral directions, the lesion was confined within the MDLS in which it originated.

Figure 5.

Complete three-dimensional network model of all mammary ductal/lobular systems (MDLS) showing sites of ductal anastomoses. A total of 11 ductal anastomoses were found at various sites considering all MDLS (branching pattern of MDLS is shown in yellow, ductal anastomoses connecting different MDLS are shown in red, and ductal anastomoses within a single MDLS are shown in blue).

Table 1 shows the distribution of ductal anastomoses in the individual MDLS. Among 16 MDLS, ductal anastomoses were present in 4 (25.0%), occasionally at a few sites in the same system; no ductal anastomosis was found in the remaining 12 MDLS. Ductal anastomoses connecting different MDLS were shown only in 2 of 16 MDLS (12.5%), and the remaining 14 MDLS were anatomically independent from each other.

Table 1. Distribution of Ductal Anastomoses Classified by Serially Numbered Individual MDLS
MDLS No.Histologic findingsNo. of ductal anastomoses
Connection within the same MDLSConnection with an adjacent MDLS
  1. MDLS: mammary ductal/lobular systems; DCIS: ductal carcinoma in situ.

1Normal ducts0]—0
2Normal ducts0]—0
3Normal ducts0]—0
4Normal ducts3]—0
5Normal ducts0]—0
6DCIS3]—0
7Normal ducts0]—0
8Normal ducts0]—0
9Normal ducts0]—0
10Normal ducts0]—0
11Normal ducts0]—0
12Normal ducts3]—0
13Normal ducts0]—0
14Normal ducts0]—0
15Normal ducts0]—0
16Normal ducts0]—0
Total anastomoses92

Table 2 shows the distribution of ductal anastomoses according to distance from the nipple. Ductal anastomoses connecting mammary ducts in the same MDLS appeared to be distributed randomly with respect to distance from the nipple, being scattered at various sites from near the nipple to the peripheral region. In contrast, ductal anastomoses connecting different MDLS did not occur within 4 cm of the nipple; any such connections were more peripheral.

Table 2. Ductal Anastomoses According to Distance From the Nipple
Distance from nipple (cm)No. of ductal anastomoses
Connection within the same MDLSConnection with an adjacent MDLS
  1. MDLS: mammary ductal/lobular systems.

< 230
2–420
4–632
≧ 610
Total anastomoses92

DISCUSSION

In the adult female breast, 15–20 collecting ducts open at the nipple. Subsegmental mammary ducts join to form segmental ducts, which in turn unite to form these collecting ducts. Ultimately, the subsegmental ducts arise from TDLU10 or TEB,11 which are comprised of terminal mammary ducts and lobules. The dimension of the collecting ducts averages approximately 2 mm. The mammary ducts are correspondingly more narrow. In the peripheral portions of the mammary gland, 1 terminal duct subtends some 20–40 lobules, and 1 lobule includes 10–100 acini.18 A single MDLS representing the anatomic segment from the lobules and terminal ducts to the collecting duct is arranged radially with respect to other MDLS with the nipple at the center; presumably the number of MDLS is equal to that of the collecting ducts opening at the nipple.19, 20 Prior to our previous study of partial mastectomy specimens,17 to our knowledge no report had described anastomoses between branches of mammary ducts within an MDLS, and anastomoses between mammary ducts of different MDLS had not been demonstrated. An MDLS is widely known to drain independently from the nipple via its collecting duct, and to show a regular anatomic pattern.

DCIS is a proliferative process that develops in the epithelia of terminal ducts and lobules such as TDLU or TEB; the lesion then extends centrally toward the nipple along mammary ducts without infiltrating stromal tissues.10, 11, 15 Even when stromal infiltration occurs and invasive tumor is present, intraductal spread of noninvasive tumor components may be extensively and continuously present within the invasive tumor, extending to mammary ducts that in some cases are relatively far from the invasive tumor.17 The invasive tumor usually is palpable, and diagnostic localization is performed readily using mammography or ultrasonography. However, accurate localization of the intraductal spread of breast carcinoma is very difficult to achieve, and this problem might be a major cause of residual carcinoma at resection margins after breast-conserving surgery.21–26 In many previous reports including detailed pathologic study of serial histologic sections of the resected specimen, intraductal spread of breast carcinoma was found to extend too widely to be explained in terms of a single MDLS.27–31 We have demonstrated that ductal anastomoses connecting different MDLS are related to the extensive intraductal spread of carcinoma through multiple MDLS.17 Therefore an MDLS is not always entirely anatomically independent of all others. This imperfection in regional anatomy may result in extensive noninvasive components of invasive breast carcinoma remaining at the surgical margin after regional resection such as partial mastectomy and may be the basis for similar residual tumor after breast-conserving surgery for DCIS. Therefore, when planning a dependably effective resection of breast carcinoma that includes the removal of all intraductally spreading carcinoma, the likely anatomic features of all MDLS in the breast should be anticipated.

Implications of the pathologic finding of an “extensive intraductal component” (EIC)32 have been investigated widely. EIC may be among the most useful predictive factors for local recurrence after breast-conserving therapy.33 Residual carcinoma due to noninvasive components of breast carcinoma occurred more frequently at the surgical margin in patients with compared with patients without EIC.34 However, EIC is an impression formed on the basis of 2-D histologic findings of invasive tumor, and EIC does not incorporate all 3-D aspects of the intraductal spread of breast carcinoma. In a study by Anscher et al.,35 EIC was a significant predictive factor for local recurrence after breast-conserving therapy on univariate analysis but not on multivariate analysis. This limited prediction capacity may affect the 2-D nature of EIC.

Wellings et al. established a 3-D method with which to study the architecture of mammary glandular trees using submacroscopic and stereomicroscopic techniques.10, 15 As a result, spatial continuity in examining these structures was achieved for what to our knowledge is the first time. The intraductal spread of breast carcinoma must be understood within the context of 3-D branching of MDLS, and spatial continuity is essential for accurate observations. The submacroscopic technique permits the observation of 3-D architectural features of mammary glands with far less effort than conventional histologic methods.36

In the past, serial conventional histologic sections of the resected specimen have been used to define complicated branching structures.37 However, the time and labor required essentially preclude 3-D reconstruction of a large specimen such as an entire breast. Computer-assisted 3-D reconstruction has reduced the time and effort involved in a reconstruction study greatly.38 All past 3-D reconstruction studies using computer graphics systems have been based on a “wire-framing model” in which the contours of the object are built up serially in the manner of playing with building blocks.39, 40 A wire-framing image simply provides 3-D expression of a crude external view of the object, and is ill-suited for delineating complex branching patterns. To reconstruct the course and branching of structures such as MDLS accurately, stereomicroscopic observation of continuous sections is needed. Such an approach is the network model produced in the current study with a submacroscopic technique.

In the current study one breast contained 16 MDLS, a finding that is consistent with conventional estimates of 15–20.18–20 Collecting ducts at the nipple numbered the same as MDLS, helping to confirm the finding that MDLS from the terminal ducts and lobules to the collecting ducts constituted a continuous, repeatedly converging anatomic unit. In nearly 25% of MDLS, the mammary ducts not only converged on the nipple from the periphery, but also joined other mammary ducts to form ductal anastomoses. The choice of an optimal resection procedure for breast carcinoma depends on anticipating whether MDLS are anatomically independent. In the current study, 2 MDLS (12.5%) in the breast communicated with other MDLS.

No ductal anastomoses connecting different MDLS were observed near the nipple, so MDLS may be anatomically independent in the area surrounding the nipple. In this respect, the risk at surgery of increasing the intraductal spread of breast carcinoma to multiple MDLS may be higher in cases in which the intraductal spread of carcinoma occurs in a peripheral as opposed to a central direction. In the majority of cases of invasive breast carcinoma, the intraductal spread of carcinoma predominantly extends centrally from the invasive tumor.17 Therefore, future study should concentrate on identifying those cases in which the intraductal spread of carcinoma extends laterally or peripherally from an invasive tumor.

Most ductal anastomoses in the current study connected with other mammary ducts in the same MDLS; very few ductal anastomoses connected different MDLS, even overlapping ones. In the entire breast, the majority of MDLS showed anatomically independent regional features as believed in the past. Therefore, in cases with no sign of multiple tumor development, the intraductal spread of breast carcinoma is unlikely to extend to multiple MDLS. However, in the current study, the spread of DCIS was very extensive both in the central direction toward the nipple and in the peripheral direction, essentially filling a single MDLS with tumor extension. Had ductal anastomoses connected this MDLS with adjacent ones, the DCIS lesion may have spread much more widely. Actually, the intraductal spread of the carcinoma extending widely from the primary invasive tumor through a ductal anastomosis connecting adjacent MDLS has been demonstrated in our previously described report.17 With this in mind, in cases of breast carcinoma with significant intraductal spread of the disease from an invasive or noninvasive tumor extending laterally or peripherally, the extent of resection must be determined with particular consideration of the risk of missing or causing the lesion to spread to multiple MDLS.

The current study was performed on a case of typical comedo-type DCIS, which showed extensive microcalcifications on mammography. Ductal anastomoses were found ubiquitously in the normal MDLS. DCIS also developed as if it was attempting to fill the MDLS in which ductal anastomoses were present. Ductal anastomoses were scattered not only in MDLS in which neoplasms were present, but also in many of the MDLS comprised of normal mammary ducts and lobules. It would be reasonable to consider that the ductal anastomoses are anatomic features that may have been induced by some aspect of the process of mammary gland maturation. The anatomic features of MDLS and ductal anastomoses found in the current study are simply the results of analysis of a single case. This does not necessarily provide definite and effective clinical evidence to aid in making the decision concerning the extent of resection for DCIS or the intraductally spreading lesion associated with invasive breast carcinoma. However, the fact that the ductal anastomoses were present apparently is an important anatomic risk factor suggesting the residual carcinoma is present at the surgical margins after breast-conserving surgery.

In 1991, the National Cancer Institute in the U.S. concluded that for clinical Stage I/II breast carcinoma, the most appropriate therapeutic method combined local surgical excision (with a tumor-free margin of 1 cm), axillary lymph node dissection, and postoperative radiotherapy to the remaining breast tissues.5 Even when postoperative radiotherapy and systemic adjuvant chemotherapy were added after lumpectomy, local recurrence was controlled in up to 3% or 6% of cases, a poorer result than that obtained with total mastectomy.41 Randomized comparisons between mastectomy and breast-conserving therapy as practiced in the past have demonstrated that local recurrence is responsible for only a small overall decrease in the survival rate.42 Accordingly, breast carcinoma treatment has shifted toward breast-conserving therapy worldwide. However, Leborgne et al. reported that in 796 patients treated with breast-conserving therapy, the 8-year disease-free survival rate among those patients who required salvage mastectomy due to local recurrence was 47%, which was significantly lower than the therapeutic result obtainable at the onset of disease.43 In recent years, breast carcinoma patients who undergo salvage mastectomy after local recurrence have been acknowledged to have a poorer prognosis.44–46 Whether local recurrence has a statistically significant influence on distant recurrence or overall survival in patients with breast carcinoma needs to be reexamined closely. At the current time, the degree of local tumor control achieved by breast-conserving therapy strongly influences the therapeutic result in a given case. Based on a multivariate analysis of 586 patients undergoing lumpectomy and postoperative radiotherapy for clinical Stage I or II breast carcinoma, Kurtz et al. suggested that the only prognostic factor significantly related to local recurrence was residual carcinoma at the surgical margins.47 Other studies also have stressed that surgical margins must be tumor-free to minimize the risk of local recurrence.35, 48–51

The results of previous studies examining local control after breast-conserving therapy for DCIS reported a local recurrence rate with wide excision of only 18% over an average observation period of 71 months. In the patients treated with radiotherapy in addition to wide excision, the local recurrence rate was 9%, although the observation period was relatively shorter and the results of local control for DCIS were not satisfactory.8 However, in recent years an attempt to maintain negative surgical margins has been made, and the results of local control for DCIS after breast-conserving therapy have shown results similar to those reported for local control in patients with invasive breast carcinoma.9, 52–54

The findings of the current study indicate that achieving local tumor control is essential in breast-conserving therapy. Although postoperative radiotherapy or systemic adjuvant chemotherapy and/or endocrine therapy may be a part of this treatment, sufficient surgical resection of the tumor, including reducing the risk of intraductal spread, is vital. The anatomic features of MDLS (including ductal anastomoses and likely patterns of intraductal spread) may provide an important basic clue with regard to achieving tumor-free surgical margins in breast-conserving surgery.

Acknowledgements

The authors thank Mr. M. Katoh, Ms. T. Kanno, Ms. S. Suzuki, Ms. Y. Mori, Miss M. Mashiko, Miss M. Kikuchi, Miss R. Yamamoto, and Miss J. Watanabe of the Second Department of Surgery, Fukushima Medical University School of Medicine, for their excellent technical assistance.

Ancillary