Patients and methods

Two hundred and ninety five women presenting mainly via the Southampton and Salisbury NHSBSP service between January 1997 and December 1998 with impalpable mammographic abnormalities, and assessed in the Southampton University Hospitals NHS Trust Breast Unit, were included in our study. The techniques used for NCB and FNAC are described below.

FINE NEEDLE ASPIRATION CYTOLOGY (FNAC)

A breast FNAC was performed by one of two radiologists involved with the study (CR and MB) under ultrasonographic or conventional stereotactic guidance. The FNAC was obtained by an ultrasonographic guided 21 gauge 1.5 inch needle or a conventional stereotactic guided 22 gauge spinal needle without aspiration. When performing FNAC before NCB the needle was used to localise the lesion and a single FNAC was obtained opportunistically and not repeated to obtain an adequate sample. The cytology smears were classified by the cytologist into five categories according to the NHSBSP guidelines for cytology.⁵

NEEDLE CORE BIOPSY (NCB)

All core biopsies were performed by the two radiologists involved with our study using ultrasonographic and conventional stereotactic guidance. Fourteen gauge needles were used (18 gauge needles were used initially). The NCB was performed by first cleaning the skin overlying the lesion with alcohol; this was followed by skin and subcutaneous infiltration with approximately 1–2 ml of 1% lignocaine. Utilising a sterile technique, the needle was inserted into the lesion. Usually two or more biopsies were taken from each solid lesion with, towards the end of our study, six cores being taken from microcalcifications. The specimen was then x rayed and calcifications were identified within the tissue sample. The specimen was then transferred into formaldehyde and processed in the histopathology department.

All of the biopsies were reported by one of two consultant histopathologists with a special interest in breast pathology (ACB and JMT), who placed them into one of several categories, using a classification system based on that kindly provided by Dr C Wells at St Bartholomew's Hospital, London (table 1). We introduced two modifications to the original classification system, namely the B2r category (benign but uncertain whether the tissue was representative of the mammographic lesion) and the subdivision of the B3 category into B3a (benign lesions that might be associated with malignancy, such as radial scar/complex sclerosing lesion, intraduct papilloma) and B3b (atypical epithelial proliferations, such as atypical ductal hyperplasia).

The breast NCB results were discussed at one or both of two weekly meetings attended by the breast radiologists, the histopathologists, the cytopathologists, and the surgeons. Surgical excision biopsy was performed, where indicated, in a proportion of the cases.

DATA ANALYSIS

The data were entered into a Microsoft Excel^™ V5.0/95 database and analysed using the SPSS^™ V6 statistical program and Stata^™ V 6.0.

To calculate the sensitivity and specificity of breast NCB, the individual cases were categorised into positive and negative groups. Positive breast NCBs were defined as those within the B3b, B4, and B5 categories. The true positive cases were those confirmed as malignant on subsequent open biopsy. Negative breast NCBs were defined as those within the B2r, B2, and B3a categories; false negative breast NCBs were defined as those cases within the B2r, B2, and B3a categories in which malignancy was identified on subsequent excision biopsy.

The complete sensitivity of breast FNAC was defined as the number of cases scored as C3, C4, or C5 expressed as a percentage of the total number of cases subsequently confirmed as malignant on NCB and/or excision biopsy. The specificity of FNAC was defined as the number of correctly identified benign cases (that is, scored as C2) expressed as a percentage of the total number of benign lesions aspirated.

Sensitivity, specificity, and complete sensitivity are all presented together with their respective 95% confidence intervals (CI). Spearman rank correlation coefficients were obtained to assess the strength of the linear association between the breast NCB score and the radiology and cytology scores, respectively.

Results

The results from 298 consecutive breast NCBs from 295 women (mean age, 56 years; range, 46–80) were analysed along with the corresponding FNAC, radiological, and excision biopsy data. The mammographic indications for NCB, listed by NCB result category, are shown in table 2. The mean follow up period was 15.8 months (range, 5–28).

The breast NCB specimens were categorised as follows: unsatisfactory/non-representative (B1; n = 61; 20.5%), benign but uncertain whether representative (B2r; n = 52; 17.4%), benign (B2; n = 103; 34.6%), benign lesions that might be associated with malignancy (B3a; n = 9; 3.0%), atypical epithelial proliferations (B3b; n = 10; 3.4%), suspicious of malignancy (B4; n = 7; 2.3%), and malignant (B5; n = 56; 18.7%).

Excision biopsy was performed in 43 cases within the B1–B3a categories (B1, n = 19; B2r, n = 8; B2, n = 8; B3a, n = 8 (data unavailable in one case)) because of radiological or cytological concerns, revealing malignancy in 18 cases (table 3). Of these 18 cases, the radiological assessment was equivocal (R3) in four cases, suspicious (R4) in six cases, and malignant (R5) in eight cases. The cytological assessment was equivocal (C3) in two cases (both assessed radiologically as R3) and malignant (C5) in two further cases (assessed radiologically as R4 and R5, respectively). Excision biopsy was performed in 65 of the 73 cases in the B3b, B4, and B5 categories (data unavailable in seven cases and one case of a lymphoma not requiring open biopsy), revealing malignancy in 64 cases (in situ carcinoma, n = 36, 56.3%; invasive carcinoma, n = 28, 43.7%) and a single benign lesion.

The sensitivity of breast NCB for malignancy in this series was 87.7% (CI, 77.9% to 94.2%), with a specificity of 99.4% (CI, 96.4% to 100.0%) and a positive predictive value of 98.5% (CI, 91.7% to 100.0%), based on the prevalence within the study population of 31.9%. The complete sensitivity of FNAC for malignancy was 34.5% (CI, 24.6% to 45.4%), with a specificity of 47.6% (CI, 37.8% to 57.6%) and an inadequacy rate of 58.7%.

Tables 4 and 5 show the association between the breast NCB score and the radiology and cytology, respectively; table 6 shows the association between the radiology and cytology scores. Ignoring all cases in which the NCB score was inadequate (B1 cases), the correlation between the NCB and radiology scores was r = 0.49 (n = 237; p < 0.0005). Similarly, ignoring cases with inadequate biopsy or cytology, the correlation between the breast NCB and cytology scores was r = 0.59 (n = 102; p < 0.0005). These correlation coefficients demonstrate that there is a moderate association between the breast NCB score and each of the radiology and cytology scores. Ignoring cases with inadequate cytology, the correlation between the cytology and radiology scores was r = 0.52 (n = 123; p < 0.0005). This indicates that when an adequate cytology specimen was available, there was a moderate correlation between the cytology and radiology scores.

Discussion

We have found the five category classification system for breast NCB to be very useful but have encountered two main drawbacks. First, a considerable proportion of breast NCB specimens show benign changes but the histopathologist cannot be certain whether the tissue is representative of the mammographic lesion. This occurred most frequently in our study when NCBs were performed for microcalcifications but only occasional microcalcifications were identified. However, we also encountered this problem when NCBs were performed for mammographic lesions other than microcalcifications. These NCB specimens could be classified as either B1 (normal/inadequate) or B2 (benign), but we feel that such cases should be highlighted to ensure that the breast team reviews them. Therefore, we have introduced the B2r category for this situation, indicating that benign breast tissue is present but that review is particularly important. Breast NCBs are often obtained in our group using a stereotactic method,¹² and therefore one can be relatively confident in such cases that the tissue is truly representative of the mammographic lesion.

The second problem that we encountered was that the B3 group includes a heterogeneous group of breast diseases, some of which are essentially benign, although possibly associated with an increased frequency of malignancy (for example, radial scar and intraduct papilloma). However, also included in the B3 category are atypical epithelial proliferations, which are strongly associated with the presence of malignancy on excision biopsy. This would include the presence on NCB of single ducts containing an atypical epithelial proliferation, but in which a firm diagnosis of either atypical ductal hyperplasia or ductal carcinoma in situ was not possible because the extent of the abnormal proliferation could not be assessed from the NCB alone. A review of the literature found 66 cases of “atypical ductal hyperplasia” identified on breast NCBs, which on subsequent excision were diagnosed as ductal carcinoma in situ in 27 cases and invasive ductal carcinoma in seven cases.¹³ Other studies have reported the identification of malignancy at open biopsy in 33–88% of cases in which atypia was identified on breast NCBs.^13–16 Of the 10 breast NCBs showing atypical epithelial proliferation in our study, open biopsy revealed carcinoma in situ in six cases, invasive ductal carcinoma in three cases, and one case of fibrocystic change with an incidental focus of atypical lobular hyperplasia. Because of the wide variation in the prognostic relevance of the breast diseases comprising the B3 category, these cases could not be included within either the positive or negative groups for the calculation of sensitivity and specificity, and previous studies have excluded them from statistical analysis.¹⁷ We believe that separation of these groups of conditions is also important for departmental and national audit programmes, which may otherwise produce apparently conflicting results when different centres are compared. We have subdivided the B3 group into essentially benign conditions such as intraduct papillomas and radial scars/complex sclerosing lesions (category B3a), which can then, as in our study, be considered as negative cases for statistical and audit purposes, and atypical epithelial proliferations (category B3b), which can be considered as positive cases for analysis.

The specificity of breast NCB for malignancy in our series (99.3%; CI, 96.4% to 100.0%) compares well with published reports of specificity ranging from 85% to 100%.^{14, 17–21} However, our sensitivity for NCB in diagnosing malignancy (87.7%; CI, 77.9% to 94.2) was less than the reported range of 93–100% in similar studies.^14–21 This might be because of the inclusion of different types of case within the “negative” and “positive” groups. For example, we included within the negative group those breast NCBs showing benign changes, but in which we could not be entirely certain from histological examination alone whether they were derived from the mammographic lesion (category B2r). We also included within the negative group breast NCB specimens showing lesions that were benign in themselves, but which might be associated with malignancy, such as radial scars (category B3a; excision biopsy was always performed in such cases). When we calculate sensitivity using only NCB specimens showing benign changes that were thought to be definitely representative of the mammographic lesions (category B2) the sensitivity rises to 97.0% (CI, 89.5% to 99.7%), well within the previously reported ranges.

Our series contained a high inadequacy rate for breast FNAC (58.7%), with a specificity of 47.6% (CI, 37.8% to 57.6%) and a complete sensitivity for malignancy of 34.4% (CI, 24.6% to 45.4%). In a similar study, Lifrange et al reported a 22% inadequacy rate, 96% specificity, and 57% sensitivity.²² However, we believe that multiple attempts to obtain an adequate FNAC are not usually justified if an adequate breast NCB has already been obtained, and therefore FNAC was done as a sighting shot with samples taken opportunistically, and was not often repeated even if the first aspirate produced little material. Furthermore, unlike our study, Lifrange et al calculated the sensitivity and the specificity of their series after excluding the inadequate cases.²² If we were to combine the data gained from the radiological assessment, breast FNAC, and NCB in our calculation of the sensitivity and the specificity, there would be no “false negative” cases and both the sensitivity and the specificity of preoperative assessment would rise to 100%. This is because when the breast NCB revealed benign tissue in the 18 cases where malignancy was subsequently identified on excision biopsy, the radiological assessment was always at least equivocal (R3 or above) and the FNAC score was at least equivocal in four cases (C3 or above) (table 3). This illustrates the requirement for good clinicopathological liaison and for regular multidisciplinary meetings before further treatment is planned. We cannot of course be certain that all cases in which a negative breast NCB specimen was obtained were definitely benign, because only 18 such cases proceeded to excision biopsy. However, we have followed the patients for a median of 15.8 months and during this period no further cases of malignancy have been identified within this patient group.

Although the relatively high inadequacy rate of breast FNAC in our series precludes a definitive assessment of its role alongside breast NCB in the preoperative assessment of impalpable breast lesions, we suggest that there are advantages to combining the two methods. When performed in a “one stop” setting, breast FNAC allows immediate definitive diagnosis in a proportion of patients, within the outpatients' department. Furthermore, FNAC may sample a larger or slightly different area of breast tissue than NCB, resulting in a smaller number of false negative cases when the two techniques are combined, as was evident in our study and other studies (table 3).²² On the other hand, the advantages of breast NCB over FNAC include a definitive histological diagnosis adding important prognostic factors, essential for planning future treatment.^{23, 24} Other studies have suggested that combining the two techniques is particularly important in certain situations—for example, cases of microcalcifications where multiple sampling is paramount¹¹ and cases in which the differential diagnosis lies between phyllodes tumour and fibroadenoma, because FNAC is unable to distinguish phyllodes tumours from fibroadenoma.^{25, 26} However, we believe that phyllodes tumours at the benign end of the spectrum cannot be reliably distinguished from cellular fibroadenomas using breast FNAC or NCB, and therefore recommend excision biopsy in this situation.

In conclusion, our study highlights the importance of a multidisciplinary approach in the preoperative assessment of impalpable breast lesions. We have also suggested two modifications to an otherwise very useful classification system for breast NCBs. We fully accept that the clinical decision facilitated by breast NCB in these patients relates solely to their further management and, in particular, to whether a mammographic abnormality requires surgical excision. However, we believe that it is essential that the intellectual pathway leading to patient management decisions is clearly identifiable, particularly for audit purposes, and hope that our refinements to the classification system will enable this to be achieved with greater accuracy.