Diagnostic accuracy of stereotactic large-core needle biopsy for nonpalpable breast disease: Results of a multicenter prospective study with 95% surgical confirmation

Authors

  • Helena M. Verkooijen

    Corresponding author
    1. Department of Surgery, University Medical Center Utrecht, Utrecht, The Netherlands
    Current affiliation:
    1. Geneva Cancer Registry, Institute for Social and Preventive Medicine, Geneva University, Geneva, Switzerland
    • P.H.M. Peeters for the COBRA Study Group, Julius Center for Patient Oriented Research, University Medical Center Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands
    Search for more papers by this author
    • Fax: +31-30-250-5484


  • The members of the COBRA Study Group are listed in Appendix I.

Abstract

Stereotactic large-core needle biopsy is increasingly applied for the diagnosis of nonpalpable breast disease. Our study examines whether this minimally invasive technique is sufficiently accurate to replace surgical breast biopsy. In a prospective multicenter study, 973 consecutive women with 1,029 nonpalpable breast lesions were offered stereotactic 14-gauge needle biopsy. If the needle biopsy yielded breast cancer, the patient was offered therapeutic surgery. Surgical biopsy was proposed in cases of needle biopsies without malignancy. An expert panel reviewed all discrepancies in histologic diagnosis between the needle biopsy and open biopsy. Forty-five patients withdrew from participation and 113 (11%) planned needle biopsy procedures were cancelled. Of the 871 successful biopsy procedures, 95% were confirmed surgically. In 13 cases (1.5%), insufficient material was obtained for histologic assessment. Fifty-five percent of the needle biopsies were diagnosed as malignant (290 invasive cancers, 190 ductal carcinoma in situ). Thirteen of the 322 lesions (4%, 95% CI 2–7%) with a benign needle biopsy diagnosis contained malignancy after surgery. Six of the 26 (23%, 95% CI 9–44%) lesions with a high-risk diagnosis (atypical ductal or lobular hyperplasia or lobular carcinoma in situ) were diagnosed as malignant after surgery. Five of the 30 lesions containing normal breast tissue held malignancy (17%, 95% CI 6–35%). Guidelines for the management of different categories of needle biopsy diagnoses were made. Application of these guidelines to the present findings resulted in sensitivity and specificity rates of 97% (95% CI 95–98%) and 99% (95% CI 97–100%), respectively. Stereotactic large-core needle biopsy is an accurate diagnostic instrument for nonpalpable breast disease. It may safely replace needle localised open-breast biopsy provided that high-risk and normal breast tissue diagnoses are followed by needle or open-breast biopsy. © 2002 Wiley-Liss, Inc.

The widespread introduction of national screening programmes for breast cancer has led to a significant increase in the number of nonpalpable breast lesions detected.1 The gold standard diagnostic test for such lesions is the needle localised open-breast biopsy.2 In the past decade, percutaneous biopsy techniques have been introduced as less-invasive alternatives for this surgical procedure. One of these alternatives is the stereotactic large-core needle biopsy on a prone biopsy table.3

In the United States, the stereotactic large-core needle biopsy is now generally accepted as an alternative for open-breast biopsy.4, 5 Much has been written about the diagnostic accuracy of this technique, but the number of studies with adequate surgical or clinical follow-up is limited6 and in several studies patient selection took place. Sensitivity rates reported in studies with good methodology vary greatly, from 94–99%.4, 7, 8 Because of low prevalence of malignancy among women referred for breast biopsy in the United States (10–30%), the risk of having breast cancer despite a benign large-core needle biopsy diagnosis is minor whether the sensitivity rate is 94% or 99%.4, 6 In the Netherlands, as in other European countries, approximately 60% of the patients referred for breast biopsy of nonpalpable lesions have breast cancer.1 In these populations, a 5% difference in sensitivity might have a major impact on the risk of malignancy despite a benign biopsy result.6 It is therefore of particular interest for the European population whether the diagnostic accuracy of stereotactic large-core needle biopsy is equal to that of the needle localised open-breast biopsy.

We conducted the prospective multicenter COBRA study (COre Biopsy after RAdiological localisation) to assess the diagnostic accuracy of stereotactic large-core needle biopsy in a population of Dutch women referred for histologic confirmation of nonpalpable breast disease.

MATERIAL AND METHODS

Patients with nonpalpable breast lesions, requiring histologic examination, were eligible for inclusion in the COBRA study. The study was carried out in accordance with the Helsinki Declaration. The Dutch National Health Insurance Council and all local Institutional Review Boards approved the study protocol. Patients were enrolled in 19 Dutch hospitals. Exclusion criteria were: (i) coagulopathies or use of anticoagulants that could not be discontinued; (ii) inability to maintain prone position for 1 hr; (iii) inability to comprehend study protocol.

Stereotactic large-core needle biopsy procedures were carried out in 5 centres (University Medical Center Utrecht, Bosch Medical Center Den Bosch, Martini Hospital Groningen, Dr Daniel den Hoed Clinic Rotterdam and Antoni van Leeuwenhoek Hospital Amsterdam). Biopsies were performed according to a standardised protocol using a prone biopsy table (Fisher Imaging, Denver, CO, or Lorad Stereoguide, Danbury, CT) and a 14-gauge core needle, long throw (2.2 cm excursion) automated biopsy device with multiple passes (C.R. Bard, Covington, GA). We used digital mammography to localise the lesions. Under local anaesthesia, at least 5 biopsy specimens were obtained from each lesion.9 If stereotactic large-core needle biopsy was performed to assess the nature of mammographic microcalcifications, specimen mammography was carried out.10 Before performing stereotactic large-core needle biopsy procedures, radiologists underwent special training: first they attended 10 biopsy procedures and subsequently they performed another 10 biopsy procedures under the supervision of a radiologist with considerable experience in stereotactic large-core needle biopsy.

After informed consent, patients were referred to 1 of the 5 centres (mentioned above) to undergo stereotactic large-core needle biopsy. Subsequent to this biopsy procedure, patients returned to the original hospitals where all further diagnostic and therapeutic interventions took place in a routine setting. Whenever large-core needle biopsy yielded invasive breast cancer, definitive surgical therapy was planned (breast-conserving therapy or mastectomy including axillary dissection or sentinel node biopsy). In the case of ductal carcinoma in situ (DCIS), either local excision (with the aim to obtain tumour-free surgical margins) or mastectomy was performed. In all other cases, we recommended needle localised open-breast biopsy. We linked patients with a benign large-core needle biopsy, who refused to undergo open-breast biopsy, to the Dutch National Morbid-Anatomical Record Department (PALGA) in July 2000 to verify if they had developed breast cancer. PALGA is a national database in which more than 98% of the nationwide reports of cytologic and histologic examinations are registered.

The histopathologic assessment of the large-core needle biopsies and the surgical specimens were performed in a routine setting by a group of 42 different pathologists divided over 10 departments of pathology. Whenever there was a discrepancy between the large-core needle biopsy diagnosis and the diagnosis made on open biopsy, an expert panel (2 radiologists and 3 pathologists) reviewed the mammographies, digital images and the histologic slides to elucidate the reasons for disagreement.

The COBRA data centre at the University Medical Center Utrecht coordinated central patient registration and prospective data collection. At intake, we recorded demographic information, relevant medical history, breast cancer risk factors, mammographic findings and reasons for referral. Standardised registration of the large-core needle biopsy procedures, needle localisation and surgical procedures were performed. Pathologists reported their findings on a standard form.

Data handling and analysis

We evaluated the diagnostic accuracy of large-core needle biopsy using a methodology adapted from Burbank and Parker.11 Histologic diagnoses of the core biopsy specimens were divided into 5 categories: normal breast tissue, benign breast disease, high-risk lesions, DCIS and invasive breast cancer. Histologic diagnoses of the surgical specimens were divided into the same categories.

High-risk lesions were those known to have a high prevalence of carcinoma at excision (atypical hyperplasia [ductal or lobular] and lobular carcinoma in situ).4, 12, 13 Previously, it has been recommended to perform open-breast biopsy in all cases with a high-risk diagnosis on large-core needle biopsy.14 Some investigators classify large-core needle biopsies containing complex sclerosing lesions (radial scars) as high risk as well, because of their reported 40% risk of being malignant at excision biopsy.4 We classified them as benign but performed a subgroup analysis of this category.

Subsequently, the predictive value of a normal breast tissue diagnosis, the predictive value of a benign diagnosis, the high-risk underestimate rate and the DCIS underestimate rate were calculated.

For this purpose, we defined the predictive value of a normal breast tissue diagnosis as the proportion of lesions diagnosed as normal breast tissue (i.e., not explanatory for the mammographic abnormality) on large-core needle biopsy that did not reveal carcinoma (DCIS or invasive) at excision biopsy or follow-up.

Similarly, the predictive value of a benign diagnosis was defined as the proportion of lesions with a benign diagnosis on large-core needle biopsy that remained benign at excision biopsy.

The high-risk underestimate rate was defined as the proportion of lesions diagnosed as high risk by large-core needle biopsy that was upgraded to DCIS or invasive cancer in the surgical specimen.

The DCIS underestimate rate was defined as the proportion of lesions diagnosed as DCIS by large-core needle biopsy that was upgraded to invasive cancer in the surgical specimen.

By means of the results of the expert panel review, we were able to clarify the reasons for the occurrence of clinically relevant discrepancies between large-core needle biopsy and open-breast biopsy. We considered lesions with a benign diagnosis on large-core needle biopsy and a malignant diagnosis (DCIS or invasive cancer) on excision biopsy as clinically relevant because failing to diagnose breast cancer might seriously affect the prognosis of the individual patient. In addition, we considered lesions with a diagnosis of DCIS or invasive breast cancer on large-core needle biopsy, whereas only benign disease or normal breast tissue was seen at open-breast biopsy as clinically relevant. In these cases morbidity might be induced unnecessarily by performing wide surgical excision or mastectomy (sometimes with axillary dissection or sentinel node biopsy) for only benign disease. We did not consider large-core needle biopsies containing DCIS and open biopsies containing high-risk lesions as clinically relevant, because for either diagnosis on large-core needle biopsy surgical excision is indicated. Also, we did not classify the finding of a high-risk lesion on open biopsy, while the needle biopsy contained only benign disease, as a clinically relevant discrepancy. Although the presence of atypical hyperplasia or lobular carcinoma in situ in a surgical specimen is associated with an increased risk of breast cancer,15–17 it does not have clinical consequences.

Finally, we made guidelines for management of the different categories of large-core needle biopsy diagnoses based on the results of our study. This way, we were able to calculate the sensitivity and specificity rates of the stereotactic large-core needle biopsy. The sensitivity rate was defined as the proportion of malignancies that was identified as abnormal by large-core needle biopsy, and the specificity rate was defined as the proportion of benign lesions that was not categorised as carcinoma (DCIS or invasive) by large-core needle biopsy.

RESULTS

Between April 1997 and February 2000, we planned 973 consecutive patients with 1,029 nonpalpable lesions for stereotactic large-core needle biopsy (Fig. 1). Twenty-four patients withdrew from participation and in 21 cases the nonpalpable breast lesion was reassessed as benign prior to the stereotactic large-core needle biopsy procedure. A total of 984 stereotactic biopsy procedures (in 928 patients) were actually initiated, but 113 (11%) procedures were terminated prematurely because of various reasons (i.e., breast too small for adequate compression, negative stroke margin, patient did not endure prone position, localisation of the lesion too close to the chest wall). Consequently, 871 biopsy procedures (in 826 patients) were successfully completed. The mean age of the patients was 58 years (range 29–85). Other characteristics of these patients and their lesions are presented in Table I.

Figure 1.

Study profile. DCIS, ductal carcinoma in situ.

Table I. Characteristics Of The 826 Patients With Successful Large-Core Needle Biopsy Procedures Of 871 Nonpalpable Lesions
Characteristicn (%)
  • 1

    Based on valid observations, <4% missing.–2Mammographic features of the 871 nonpalpable lesions.

Family history of breast cancer249 (31)1
History of breast cancer112 (14)1
Nulliparity141 (18)1
Referred by national screenings programme493 (60)1
Radiology2
 Mass lesion (spiculated or well defined)310 (36)
 Microcalcifications409 (47)
 Microcalcifications with tissue distortion124 (14)
 Architectural distortion26 (3)
 Focal asymmetry2 (0.2)

In 13 cases (1.5%), the pathologist was not able to make a histologic diagnosis due to insufficient quality of the needle biopsy specimens.

In Table II the histologic diagnoses of the large-core needle biopsies and the open-breast biopsies are presented. Based on these figures, we calculated the predictive value of a normal breast tissue diagnosis, the predictive value of a benign diagnosis, the high-risk underestimate rate and the DCIS underestimate rate.

Table II. Histological Diagnoses Based On Large-Core Needle Biopsy And Surgical Specimen
Core biopsySurgical specimen
NormalBenignHigh riskDCISInvasiveTotal (%)
  • 1

    Thirty-nine benign and 4 normal needle biopsies were not followed by surgical biopsy. Final diagnosis was imputed from follow-up.– 2Two patients with DCIS and 4 patients with invasive cancer on needle biopsy did not undergo surgery. Reasons: distant metastases (n=2), radiotherapy for non-Hodgkin lymphoma (n=1), advanced age (n=2) and patient refused (n=1). Final diagnoses were classified into the same categories as large-core needle biopsy diagnoses.– 3Two open-biopsy specimens were incorrectly diagnosed as DCIS (i.e., expert panel reclassified final histologic diagnosis as benign).– 4Ten lesions diagnosed as DCIS and 3 diagnosed as invasive cancer on large-core needle biopsy contained only benign or normal breast tissue at excision biopsy. Panel review showed that the needle biopsies of 5 of these lesions (1 invasive and 4 DCIS) were incorrectly diagnosed as malignant by the pathologist.

Normal112313230 (3)
Benign16289121332322 (38)
High risk10105126 (3)
DCIS24464514332190 (22)
Invasive2344283290 (34)
Total (%)11 (1)331 (38)28 (3)168 (20)320 (37)858 (100)

Thirty large-core needle biopsy specimens revealed only normal breast tissue and provided no explanation for the mammographic abnormality. On excision biopsy, 5 of these lesions proved to be malignant (3 DCIS presenting as microcalcifications without tissue distortions and 2 invasive breast carcinomas, presenting as architectural distortions). Accordingly, the predictive value of a normal breast tissue diagnosis is 83% (95% CI 65–94%). Reasons for missing the malignancies were technical failures in all 5 cases: the needle biopsy specimens of the calcified lesions did not contain calcifications on specimen X-ray and review of the digital images of the architectural distortions revealed that the biopsy needle had not penetrated the lesion. Four patients with a diagnosis of normal breast tissue refused to undergo open-breast biopsy. Because none of these patients was diagnosed with breast cancer after a mean follow-up of 25 months, we categorised the final histologic diagnosis as benign.

A benign diagnosis based on large-core needle biopsy was observed in 322 lesions. Fifteen of these turned out to contain malignancy on excision biopsy (13 DCIS presenting as microcalcifications and 2 invasive carcinomas presenting as mass lesions). After panel review, we were able to explain 13 of these 15 discrepancies (Table III). In 3 cases, the cancers were missed due to technical failures. Five times the pathologist misinterpreted the large-core needle biopsies (i.e., DCIS or high-risk elements were present in the needle biopsies but not recognised as such). Two times, the excision biopsies were misinterpreted (i.e., the excision biopsy did not contain malignancy according to the expert panel, thus no cancer was actually missed). In 3 cases, the finding of malignancy in the excision biopsy specimen was incidental. In 2 cases, we were not able to elucidate the reasons for missing the malignancies: the procedure was performed state of the art and the histologic findings correlated perfectly with the mammographic features. Consequently, stereotactic large-core needle biopsy diagnosed 13 cancers incorrectly as benign. This gives rise to a predictive value of a benign diagnosis of 96% (95% CI 93–98%). Thirty-nine patients refused to undergo open-breast biopsy following benign large-core needle biopsy. After a mean follow-up of 17 months (range 5–36), none of the patients had developed breast cancer. The final histology was therefore categorised as benign. None of the 11 benign lesions with the histologic diagnosis “complex sclerosing lesion” (radial scar) turned out to be malignant at excision biopsy.

Table III. Reasons for clinically relevant discrepancy between large-core needle biopsy and open-breast biopsy
Reason for clinically relevant discrepancyDetailsn
  • 1

    Invasive breast cancer missed.

  • 2

    Histological misinterpretation of large-core needle and open-biopsy specimens by pathologist. DCIS, ductal carcinoma in situ; ILC, invasive lobular carcinoma.

Large-core needle biopsy benign–excision biopsy malignant (13 DCIS, 2 invasive breast cancers1)
 Technical failure (n=3)Microcalcifications not present on specimen X-ray2
Digital images revealed that mass lesion was not penetrated by biopsy needle11
 Misinterpretation of the large-core needle biopsies2 (n = 5)Expert panel diagnosed large-core needle biopsies as DCIS2
Expert panel diagnosed large-core needle biopsies as high risk3
 Misinterpretation of the excision biopsy specimens2 (n = 2)Expert panel diagnosed excision biopsies as benign2
 Accidental finding (n=3)DCIS present in vicinity of microcalcifications caused by benign disease, DCIS not visible mammographically2
ILC in vicinity of microcalcifications caused by benign disease, ILC not visible mammographically11
 Missed cancer of unknown origin (n=2)Large-core needle biopsy procedure technically adequate, histology correlated with mammography2
Large-core needle biopsy malignant (10 DCIS, 3 invasive cancers)–open-breast biopsy benign
 Improperly executed needle localised lumpectomyIncorrect needle localisation, surgical excision or histologic assessment of lumpectomy7 (2 IDC)
 Lesion completely removed by large-core needleFocally present DCIS was entirely removed during large-core needle biopsy procedure1
 Misinterpretation of the large-core needle biopsiesExpert panel diagnosed large-core needle biopsies as benign (false-positive large-core needle biopsy results)5 (1 IDC)

Twenty-six large-core needle biopsies were diagnosed as high risk. Six of these lesions revealed malignancy at excision biopsy (1 invasive carcinoma, 5 DCIS). Accordingly, the high-risk underestimate rate was 23% (95 CI 9–44%).

Large-core needle biopsy revealed DCIS in 190 cases. After surgical therapy, an invasive component was present in 32 cases, resulting in a DCIS underestimate rate of 17% (95% CI 12–22%). Ten lesions were diagnosed as normal or benign after surgical therapy. These discrepancies were explained by 5 improperly executed needle localised open-biopsy procedures and 1 complete removal of a DCIS lesion during the large-core needle biopsy procedure (Table III). In addition, 4 large-core needle biopsy specimens were misinterpreted by the pathologists and incorrectly diagnosed as DCIS. These 4 cases were considered as false-positive large-core needle biopsy results.

Two hundred ninety large-core needle biopsy specimens contained invasive breast cancer. Three times, no malignancy was found in the excision biopsy. Reasons were 2 improperly executed lumpectomies and, once, the pathologist misinterpreted the large-core needle biopsies and incorrectly diagnosed them as invasive breast cancer (Table III). We classified the latter as false-positive as well.

Based on these results, we recommend the following guidelines for the management of the different categories of large-core needle biopsy diagnoses (Table IV). Whenever large-core needle biopsy yields only normal breast tissue (i.e., when the needle biopsy diagnosis does not explain the mammographic lesion), a repeat biopsy or excision biopsy is recommended. In the case of a benign large-core needle biopsy result, the risk of the presence of breast cancer is 4%, and in these cases mammographic follow-up is advised (for example, biennial bilateral mammography in the context of a national screening program for breast cancer). A high-risk diagnosis is always indicative for open-breast biopsy. Whenever DCIS is diagnosed, wide excision (aiming for tumour-free margins) should be executed, although there is a small risk (2%) that this is a false-positive diagnosis. In addition, a risk of 17% exists that invasive cancer is present in the surgical specimen. In the case of invasive breast cancer diagnosed on large-core needle biopsy, definitive surgical therapy (with axillary dissection or sentinel node biopsy) can be performed. The risk of a false-positive diagnosis is less than 1%.

Table IV. Recommendations For Management Of Different Large-Core Needle Biopsy Diagnoses
Large-core needle biopsy diagnosisStrategy
  • DCIS, ductal carcinoma in situ.

  • 1

    Including complex sclerosing lesion (radial scar).

  • 2

    Atypical (ductal or lobular) hyperplasia and lobular carcinoma in situ.

Normal breast tissueRepeat large-core needle biopsy or open-breast biopsy
Benign1Mammographic follow-up
High risk2Open-breast biopsy
DCISSurgical excision
Invasive breast cancerBreast-conserving therapy/mastectomy (including axillary dissection/sentinel node biopsy)

We applied this strategy to our study results and calculated the sensitivity and specificity rates of stereotactic large-core needle biopsy. On a total of 499 malignancies (DCIS and invasive carcinomas), 13 lesions would have been diagnosed as benign and would not have received adequate therapy. This results in a sensitivity rate of 97% (95% CI 95–98%). Of the 359 nonmalignant lesions (normal breast tissue, benign and high risk), 5 were incorrectly diagnosed as malignant and these patients would have been overtreated. Accordingly, the specificity rate was 99% (95% CI 97–100%).

DISCUSSION

Our study prospectively evaluated the diagnostic accuracy of stereotactic large-core needle biopsy in a large group of consecutive patients with nonpalpable breast lesions. More than 95% of the needle biopsy specimens were surgically confirmed and all discrepancies between large-core needle biopsy and the reference standard were reviewed by an expert panel. This way, errors of the reference standard were recognised as well. Based on the results of our study, we believe that stereotactic large-core needle biopsy can safely replace the needle localised open-breast biopsy, since its diagnostic accuracy (97% sensitivity rate and 99% specificity rate) is comparable to that of the reference standard (sensitivity rate 96–99%).2, 18

The results of our study are comparable to those of other large-sample studies.5, 19–21 However, in these studies there was no surgical confirmation or long-term follow-up of benign large-core needle biopsies. In a recently conducted meta-analysis, a pooled sensitivity rate of 97% was calculated.6

By evaluation of all discrepancies by an expert panel, we were able to elucidate reasons for missing malignancies by large-core needle biopsy. Some of these missed cancers could have been prevented. Three biopsy procedures had not been properly executed technically. By careful examination of the digital images and X-rays of the biopsy specimens, technical failures can be recognised. Radiologists with limited large-core needle biopsy experience need to pay particular attention to this. In addition, correlation of every large-core needle biopsy diagnosis to the mammographic findings, preferably in a multidisciplinary setting, is of utmost importance to identify inadequate biopsy procedures.

Five malignancies were missed because the pathologist was not able to recognise a DCIS lesion as such (or as high risk), although these features were in fact present in the large-core needle biopsy specimen. This problem, however, is not specific for large-core needle biopsy, since it also occurs in the assessment of open-biopsy specimens of nonpalpable breast lesions. Whenever a pathologist has reservations whether an atypical or in situ component is present in a needle biopsy specimen, (s)he should not hesitate to consult a colleague with extensive expertise in breast pathology.

The 2 missed breast cancers of unknown origin (technically perfect biopsy procedure, histologic findings correlating with mammography) and the 3 accidental findings (malignancies that were not visible mammographically in the vicinity of a benign nonpalpable lesion) are the unavoidable shortcomings of the technique. Fortunately, these events were not observed frequently.

Our study indicates that false-positive results may occur due to misinterpretation of the large-core needle biopsy specimens by the pathologist. In our series, 5 needle biopsy specimens were incorrectly diagnosed as malignant, leading to 1 patient undergoing unnecessary axillary lymph node dissection.

Only 1.5% of the large-core needle biopsy specimens was of insufficient quality to establish a diagnosis. This low insufficiency rate is an advantage of large-core needle biopsy over fine-needle aspiration, which is frequently applied in some European countries for the diagnosis of nonpalpable breast disease. With fine-needle aspiration, insufficiency rates of up to 33% have been reported.22 Nevertheless, a rather large proportion of stereotactic large-core needle biopsy procedures in our study had to be cancelled (11%). This is in accordance with the findings of others, who reported a proportion of cancelled procedures of 16%.23 Very dense breast tissue, axillary localisation of the lesion, localisation of the lesion close to the chest wall, low body mass index and more than 1 nonpalpable lesion are reported to be risk factors for cancellation of the biopsy procedure.24

A small group of patients would not benefit from replacement of the open biopsy by the large-core needle biopsy. For patients with a high-risk diagnosis on large-core needle biopsy, the needle biopsy procedure would only be an additional intervention, since needle localised open-breast biopsy has to be executed anyway. Similarly, patients with a diagnosis of normal breast tissue will have to undergo repeat or open biopsy. Fortunately, the proportion of patients with such a diagnosis on large-core needle biopsy is small (in our study we observed 3% high-risk lesions and 3.5% normal breast tissue on large-core needle biopsy). In addition, in case of a DCIS underestimate, patients do not benefit from the availability of a preoperative histologic diagnosis, because a second surgical procedure (axillary dissection or sentinel node biopsy) always has to be performed at a later point in time.

Complex sclerosing lesions (radial scars) are benign breast lesions, known to be associated with a 2-fold increased risk of breast cancer.25 The finding of a complex sclerosing lesion on large-core needle biopsy was reported to be associated with a 60% risk of finding breast cancer at excision biopsy,4 and it was therefore recommended to perform excision biopsy in all such cases. In contrast, none of the 11 lesions diagnosed as complex sclerosing lesion in our series revealed malignancy after surgical excision. A recent study of Philpotts et al.26 supports our findings (none of the 9 needle biopsies containing complex sclerosing lesions turned out to be malignant). We believe, however, that additional research is needed to get more insight in the exact management of large-core needle biopsies containing complex sclerosing lesion, since the size of our series is quite small and the literature rather ambiguous.

The prevalence of breast cancer in our population was 58% (499 invasive and in situ carcinomas). In addition to sensitivity and specificity, prevalence of malignancy is a determinant of the predictive value of a benign diagnosis. In our study, the predictive value of a benign diagnosis was 96%. Consequently, women with a benign diagnosis on large-core needle biopsy have a risk of 4% that a carcinoma is actually present. If the prevalence of malignancy would have been much lower, i.e., 20% (as in the United States), the risk of missing breast cancer despite a benign diagnosis would have been less than 1%, applying the same sensitivity and specificity rates (97% and 99%).18 Until recently, the general policy in Western Europe was to apply stringent selection criteria before referring a patient with nonpalpable breast disease for breast biopsy. This way, the number of surgical interventions for benign breast lesions was kept to a minimum (40%).1 One could question whether such a stringent policy should be continued now that a minimally invasive, less-expensive diagnostic test is available. Cost effectiveness analyses may help in making such decisions.

A significant number of lesions in our study were visible by means of ultrasound. Because we wanted to evaluate the merits of stereotactic large-core needle biopsy, we did not use ultrasound for guidance of the large-core needle biopsy procedures. Nevertheless, ultrasound guidance is the imaging technique of first choice for percutaneous breast biopsies: it gives minimal inconvenience to the patient, there is real-time visualisation of the biopsy needle entering the targeted lesion, the procedure time is shorter and, consequently, the costs are relatively low.27 A considerable proportion of lesions, however, cannot be identified by means of ultrasound, including many clusters of microcalcifications and some small masses, especially those surrounded by fatty tissue or located deep within large breasts. Stereotactic guidance is needed for large-core needle biopsy of this category of lesions.

In recent years, more advanced sampling techniques, such as the vacuum-assisted 11- or 14- gauge stereotactic breast biopsy probes, have been developed. These were also beyond the scope of our study. Advantages of these new techniques might be that fewer biopsy procedures need to be cancelled and that histologic underestimate rates can be reduced.28 Until now, however, the added value (in terms of cost effectiveness) of these much more expensive techniques in comparison to the 14-gauge large-core needle biopsy has not been sufficiently elucidated.

In conclusion, stereotactic large-core needle biopsy on a prone biopsy table is an accurate diagnostic test for nonpalpable breast disease and can safely replace the needle localised open-breast biopsy. A high-risk diagnosis is always a reason for open-breast biopsy. Whenever only normal breast tissue is obtained by large-core needle biopsy, a repeat needle biopsy or open-breast biopsy is recommended.

APPENDIX I – COBRA STUDY GROUP

COBRA executive and writing committee

HM Verkooijen, PHM Peeters, IHM Borel Rinkes, WPThM Mali, ThJMV van Vroonhoven

COBRA study group

APE Besnard, IHM Borel Rinkes, E Buskens, WAH Gelderman, AD Groote, JHCL Hendriks, R Holland, PKH Hut, HJ de Koning, WPThM Mali, M Oudkerk, PHM Peeters, JL Peterse, RM Pijnappel, EJT Rutgers, MEI Schipper, P Schipper, JG van den Tweel, HM Verkooijen, ThJMV van Vroonhoven, PFGM van Waes, T Wiggers

Data coordination centre

R van Biljouw, LE Hoorntje, EJ Polder, MMA Tillmann, HM Verkooijen (coordinator)

Expert panel

JHCL Hendriks, R Holland, MEI Schipper, JL Peterse, RM Pijnappel

Radiologists performing stereotactic large-core needle biopsy procedures

APE Besnard, A van Dalen, C Holt, W Koops, TH Lo, J Noordzij, AIM Obdeijn, DB Piers, RM Pijnappel, P Schipper, LJ Schultze-Kool, RK Storm, E Tetteroo, D Urich, KVJM Vanghillewe, PFGM van Waes, PJW Wensing, A Zwemmer

Safety monitoring committee

DJ Dronkers, RWM Giard, PW de Graaf, ALM Verbeek

Participating centres and investigators

University Medical Center Utrecht: M Boere, IHM Borel Rinkes, T Buijs-van der Woude, E Buskens, A van Dalen, M van Dam, R Goldschmeding, J de Groot, C Haaring, C Harderwijk, J Heijboer, A Hennipman, GHM Hogenbrik, GN Jonges, S van Kanten, VCM Koot, SHJ Loyson, WPThM Mali, PGJ Nikkels, J Noordzij, PHM Peeters, SCJ van den Putten, M Raatgever, I Rollema, MS Schlooz-Vries, MEI Schipper, DMDS Sie-Go, RKJ Simmermacher, PJ Slootweg, WGM Spliet, JG van den Tweel, HM Verkooijen, C de Vries, ThJMV van Vroonhoven, O van Vulpen, PFGM van Waes, PJW Wensing, E van Zuijdam, D Zuidema

Antoni van Leeuwenhoek Hospital Amsterdam: APE Besnard, J Bonnema, A Boom, MAM de Bree, F van Coevorden, PHP Davids, R Degeling, E Deurloo, WAH Govaert, CJ Kimmings, W Koops, J Merkus, J Molewijk, H Oldenburg, JL Peterse, EJT Rutgers, LJ Schultze-Kool, M Sosef

Bosch Medical Center Den Bosch: S van Asseldonk, RJ Bolhuis, JM Broekman, E van Geffen, WAH Gelderman, S Janssens, SJMM Ketelaars, JV van der Linden, HA Meijer, L Obertop, PB Piers, JMC Rottier, P Schipper, TJMJ Schiphorst, AJM van Unnik, S Veltkamp, M van Vught, J Wever, FGJ Willekes, JC Wissing, A Zwemmer

Dr Daniel den Hoed Clinic Rotterdam: CCM Bartels, M Bertheux, FH Beverdam, E Brouwers, LM Budel, P van Dijk, AN van Geel, RJJ Heyboer, SC Henzen-Logmans, J Hoftijzer, JC den Hollander, R Koelemij, D Koopmans, RR de Krijger, ThH van der Kwast, MBE Menke-Pluijmers, WJ Mooi, AIM Obdeijn, M Oudkerk, M Piskorz, S Ramnarain, A Stam, MMA Tilanus, D Urich, H Wesseling, T Wiggers

Martini Hospital Groningen: PC Baas, AW Brascamp-van der Hoek, HR van Dop, AD Groote, PWJM Hellema-Hagen, C Holt, ThEAM de Jong, AJ Julius, W Knol, G van der Kolk, FCP Moll, AUCC Nootenboom, WAH Nugteren, A Olieman, MW van Oven, RM Pijnappel, EFS Sclerrenborg, JDJ Steinberg, KVJM Vanghillewe, W Visser-Palsma, J de Vries, B Vrugt, AH Wieringa, JM de Wolf

University Medical Center Groningen: EW Nijboer, ATMG Tiebosch, W Timens

Antonius Hospital Nieuwegein: TJ Bast, FJEM Blomjous, JWBM Casparie, JRJ Elbers, HW van Es, PMNYH Go, HDWM van de Pavoort, B van Ramshorst, CA Seldenrijk, R Sybrandy

Catharina Hospital Eindhoven: M van Beek, PJJM Klinkhamer, MMY Lammers, G Nieuwenhuizen, H Peters, HJT Rutten, IWN Tan, AH van der Veen

St Clara Hospital Rotterdam: JK Boldewijn, RWM Giard

Diakonessenhuis Utrecht: JMHH van Gorp, JWTH Muller, CI Perre, A Pronk, HM Ruitenberg, J Weitz

Hospital Eemland Amersfoort: CE Albus-Lutter, H Barroclough, RA Buiskool, ARA Dijkema, A Enklaar, A van Es, RAM Gruijters, BGF Heggelman, BW Ike, C Kooijman, I van Meerwijk, L Meiss, B van Ooijen, JMA van Oosterhout, AJ van Overbeeke, L de Vries

St. Elisabeth Hospital Tilburg: JFMM Misere, JA Roukema, JLJM Teepen

Erasmus University Rotterdam, dept. of Public Health: ME van de Akker-Marle, E Birnie, R Boer, JH Groenewoud, HJ de Koning

Gelderse Vallei Wageningen: W de Roos

Havenhospital Rotterdam: JH Driebeek-van Dam

Ikazia Hospital Rotterdam: HF Veen

St Joseph Hospital Veldhoven: PAA Pauwels, RMH Roumen

Lorentz Hospital Zeist: PW Jiskoot, LJ de Kanter, WNM Steur, WJ van der Ven

Hospital Mesos Utrecht: JP Gelissen, WJC Geurts, EBM Teunissen

Hospital Rijnstate Arnhem: TIFM Bloemen, HE Deenik, IB Dulmus, WF Eggink, R Gilkes, WHA Govaert, GB ten Haken, IMC Janssen, FOHW Kesselring, MJ van Kints, JGH Klinkenbijl, JWR Meyer, H Mul, CB Reuvers, EJ Spillenaar, SMM van Sterkenburg, EH Strubbe, JM Wiersma, WR de Vries

Hospital Rivierenland Tiel: HAJ Bartholomeus, RKKM Hoenderop, PF Liqui Lung, EA Nooitgedacht, KH Ong, JE Rütter, F Ugahari

Ancillary