Kurniawan ED, Rose A, Mou A, Buchanan M, Collins JP, Wong MH, Miller JA, Mann GB. Risk Factors for Invasive Breast Cancer When Core Needle Biopsy Shows Ductal Carcinoma In Situ. Arch Surg. 2010;145(11):1098-1104. doi:10.1001/archsurg.2010.243
A core needle biopsy (CNB) diagnosis of ductal carcinoma in situ (DCIS) may be associated with a final diagnosis of invasive cancer. Preoperative radiologic, clinical, and pathological features may identify patients at high risk of diagnostic upstaging, who may be appropriate candidates for sentinel node biopsy at initial surgery.
Review of prospectively collected database.
Tertiary teaching referral hospital and a population-based breast screening center.
Consecutive patients from January 1, 1994, to December 31, 2006, whose CNB findings showed DCIS or DCIS with microinvasion.
Main Outcome Measures
Upstaging to invasive cancer.
Eleven of 15 cases of DCIS with microinvasion (73.3%) and 65 of 375 cases of DCIS (17.3%) were upstaged to invasive cancer. Ten of 21 palpable lesions (47.6%) were found to have microinvasion. For impalpable DCIS, multivariate analysis showed that noncalcific mammographic features (mass, architectural distortion, or nonspecific density) (odds ratio [95% confidence interval], 2.00 [1.02-3.94]), mammographic size of 20 mm or greater (2.80 [1.46-5.38]), and prolonged screening interval of 3 years or longer (4.41 [1.60-12.13]) were associated with upstaging. The DCIS grade on CNB was significant on univariate analysis (P = .04). The rate of upstaging increased with the number of significant factors present in a patient: 8.3% in patients with no risk factors, 20.8% in those with 1 risk factor, 39.6% in those with 2 risk factors, and 57.1% in those with 3 risk factors.
The risk of upstaging can be estimated by using preoperative features in patients with DCIS on CNB. We propose a management algorithm that includes sentinel node biopsy for patients with DCIS who have microinvasion on CNB, palpable DCIS, 2 or more predictive factors, and planned total mastectomy.
Percutaneous core needle biopsy (CNB) of suspicious breast lesions is now standard practice, replacing diagnostic open biopsy in most clinical situations.1 The advantages of this change have been to reduce the number of benign surgical biopsy results and to allow appropriate definitive surgical treatment to occur in a single visit to the operating room in most cases.2,3
A significant proportion of neoplastic breast lesions identified in a mammographic screening program are ductal carcinoma in situ (DCIS). Percutaneous CNB is used to differentiate between DCIS and invasive cancer. This distinction is important because it may determine the extent of surgery. Because DCIS is a preinvasive condition, axillary staging is not part of standard surgical treatment.3,4 There is inherent underestimation of invasive cancer in CNB because of the limited amount of tissue taken.5 Published series report that CNB underestimates the number of invasive cancers by 9% to 44%.6- 14 Patients with DCIS on CNB who are treated with breast surgery alone and found to have invasive cancer generally require additional surgery to evaluate their lymph node status. Some studies suggest that sentinel node biopsy (SNB) after breast surgery has a higher failure rate15,16; therefore, it is preferable to stage the axilla during the first surgical procedure if it is deemed necessary.
An SNB could be routinely performed at the time of the first surgical procedure, but the procedure carries some risk, SNB adds significantly to the cost of the procedure, and the rate of lymph node positivity is low in patients with pure DCIS (0%-7%).14,17- 22 It is therefore desirable to identify the subgroup of patients in whom the rate of invasive cancer is most likely to be underestimated at CNB, for whom SNB may be indicated at initial surgery.
A number of studies have attempted to define predictors of invasive cancer by looking at mammographic and histologic CNB findings. High cytologic grade,8- 10,13 a mass on the mammogram,10- 12 and larger lesion size8,9,12- 14 appear to correlate with accompanying invasive cancer, but there is no consensus.
We investigated preoperative demographic, mammographic, and pathological factors that may predict underestimation of invasive cancer when CNB shows DCIS. We restricted the study to breast cancers and DCIS detected within a population-based mammographic screening program. Screen-detected lesions tend to be smaller and are more likely to be impalpable than symptomatic lesions,23,24 and it is possible that results from a screening program would differ from those of both screen-detected and symptomatic lesions. We then assessed the impact of multiple risk factors to generate a possible management algorithm.
All patients with a CNB diagnosis of DCIS or DCIS with microinvasion (DCISM), from January 1, 1994, through December 31, 2006, were identified from the records of North Western BreastScreen, Melbourne, Australia. This is a population-based mammographic screening program serving the northern and western suburbs of Melbourne. Microinvasion was defined as invasive carcinoma measuring less than 1 mm.
Demographic, radiologic, clinical, and pathological data were collected prospectively on all cases. Most of the information was collected contemporaneously by the screening program, and this data collection was supplemented by information from pathology records. Mammograms were further characterized through a systematic review by 2 consultant breast radiologists (A.R. and A.M.).
Percutaneous biopsies were performed by 1 of 3 specialist breast radiologists (including A.R. and A.M.), and the tissues obtained were assessed at the Pathology Department of The Royal Melbourne Hospital, a tertiary teaching referral hospital. Lesions not seen on ultrasonography were subjected to biopsy by means of a 14-gauge automated needle (ProMag Automatic, MD Technologies, Galena, Illinois) on a prone stereotactic table (Lorad DSM, Hologic, Bedford, Massachusetts). After 2001, an 11-gauge vacuum-assisted biopsy device (Mammotome Biopsy System, Ethicon Endo-Surgery Inc, Cincinnati, Ohio) was used, on the same prone stereotactic table. Biopsy specimens of ultrasound-visible lesions were obtained with a 14-gauge automated needle (ProMag Automatic; Bard Magnum Biopsy, Bard Limited, Crawley, West Sussex, England; or Achieve Automatic Biopsy System, Cardinal Health Dublin, Ohio) with the aid of an ultrasound machine (Antares Sonoline, Siemens, New York, New York; or ATL HDI 3000, Universal Diagnostic Solutions, Oceanside, California). Impalpable lesions were then localized by specialist breast radiologists using a hookwire or a carbon track.25 Surgery was performed at The Royal Melbourne Hospital or 1 of 6 private hospitals in Melbourne.
Statistical analyses were performed with SPSS software (SPSS Inc, Chicago, Illinois). Pearson χ2 and Fisher exact tests were used to determine significant associations between patient characteristics and margin status. Multivariate analysis was performed by logistic regression.
A total of 396 women had a CNB diagnosis of DCIS or DCISM at the North Western BreastScreen from January 1, 1994, through December 31, 2006. Five had missing or incomplete data and 1 declined definitive surgery, leaving 390 cases for analysis. The majority of patients (77.4%) were aged 50 to 69 years, which is the target age group of the mammographic screening program at North Western BreastScreen.
The CNB results showed 375 patients (96.2%) who had pure DCIS and 15 (3.8%) who had DCISM (Table 1). After definitive surgery, the diagnosis of 65 (17.3%) of those with pure DCIS on CNB was upstaged to invasive cancer, 16 (4.3%) showed microinvasion, and the remaining 294 (78.4%) retained the diagnosis of pure DCIS (including 11 cases that showed small DCIS on CNB and yielded no pathological findings after definitive surgery). For those with DCISM on CNB, the diagnosis of 11 (73.3%) was upstaged to invasive cancer, and 4 (26.7%) showed microinvasion. The presence of microinvasion on CNB was strongly associated with upstaging to invasive cancer (P < .001) (Table 2). A palpable lump suggestive of a malignant neoplasm was an uncommon finding, accounting for only 21 of 390 cases. Ten of the 21 palpable lesions (47.6%) were upstaged to invasive cancer (P = .01).
Of 76 patients whose diagnosis was upstaged to invasive cancer, 71 (93.4%) underwent axillary assessment. Twenty-one (27.6%) had either SNB or axillary dissection together with their initial breast surgery owing to surgeon or patient preference. Fifty (65.8%) received axillary assessment as part of their reoperation after final pathological examination results showed invasive cancer. Five patients had no axillary assessment: 4 had cancers measuring less than 3 mm, and 1 elderly patient with a widely excised 10-mm cancer decided against further surgery.
In patients with a final diagnosis of DCISM, 14 of 20 (70%) had axillary assessment: 6 (30%) during the initial surgical procedure and 8 (40%) during a subsequent procedure, leaving 6 without axillary assessment. Only 19 of 294 patients with pure DCIS (6.5%) received axillary assessment. Seventeen (5.8%) were during the initial surgery: 16 were in association with mastectomy, and 1 underwent axillary dissection after suspicion of invasive cancer owing to the discovery of extensive calcification during the preoperative workup. An additional 2 patients (0.7%) had an SNB in association with completion mastectomy during their subsequent surgical procedure.
Overall, of 390 patients, 109 (27.9%) had axillary assessment, and in 44 of them (40.3%) this was at the initial operation. Thirteen of the 109 patients (11.9%) had at least 1 positive lymph node; among them was 1 patient with 10 positive lymph nodes.
The rate of reoperation was highest in the group whose diagnosis was upstaged to invasive cancer (55 of 76 [72.4%]), followed by the DCISM group (9 of 20 [45.0%]), and lowest in the pure DCIS group (85 of 294 [28.9%]). Axillary assessment was involved in the majority of additional procedures in the first 2 groups, whereas reexcisions for unsatisfactory margins predominated in the last group.
In the 355 patients with impalpable lesions and no microinvasion on CNB, we examined preoperative factors that may be associated with a change in diagnosis after definitive surgery, to either DCISM or invasive cancer (Table 3).
The presence or absence of mammographic microcalcifications was not associated with a change in the final diagnosis (P = .47). Of the 91 cases in which the patient had a mass, density, or architectural distortion on mammography, the diagnosis in 28 cases (31%) was upstaged. By contrast, the diagnosis in only 44 of 264 patients (16.7%) without any of these features (ie, microcalcification only) was upstaged (P = .004). Thus, lesions with microcalcification as the sole mammographic abnormality were significantly less likely to have invasive cancer on final pathological examination. As the maximum diameter of the mammographic abnormality increased, the presence of invasion became more likely. Only 20 of 260 lesions (12.5%) smaller than 20 mm were upstaged, compared with 52 of 195 lesions (26.7%) measuring 20 mm or greater (P = .001).
There was no difference in upstaging rates between different CNB techniques: 37 of 184 (20.1%) for the stereotactic method (14-gauge needle), 27 of 134 (20.1%) for the vacuum-assisted method (11-gauge needle), and 8 of 37 (21.6%) for the ultrasound-assisted method (14-gauge needle) (P = .98).
A greater proportion of cancers were upstaged with increasing grade of DCIS (10.6% of low-, 16.2% of intermediate-, and 25.0% of high-grade DCIS), and this trend was significant (P = .04). The presence of comedonecrosis on CNB was more likely to be associated with invasive cancer on final pathological examination, but this was not statistically significant (P = .07).
The BreastScreen Victoria mammographic screening program sends invitations to women aged 50 to 69 years to attend screening every 2 years. In our series, if 3 years had passed since a woman's last screening mammogram, that woman was considered a “lapsed attendee.” Of 23 lapsed attendees, 10 (43.5%) had microinvasion or invasive cancer on final pathological examination. Among those whose positive screening mammogram was within 3 years of the previous one, 62 of 332 (18.7%) had an invasive component (P = .004).
On univariate analysis, 4 factors were found to be significantly associated with upstaging to microinvasive or invasive cancer. After adjustment for other factors, the presence of a mass, density, or architectural distortion on mammography; mammographic size 20 mm or greater; and screening interval 3 years or longer remained significant (Table 4). The DCIS grade on CNB was not significant on multivariate analysis.
We then examined the outcome of patients with variable numbers of the significant preoperative factors. The proportion of patients for whom the diagnosis was upstaged rose with each additional risk factor: 9 of 108 (8.3%) of those with no factors, 40 of 192 (20.8%) of those with 1 factor, 19 of 48 (39.6%) of those with 2 factors, and 4 of 7 (57.1%) of those with 3 factors (Table 5).
Of 355 women with impalpable DCIS on CNB, 55 had 2 or more significant risk factors for upstaging. Twenty-three of the 55 (41.8%) had invasion on final pathological examination, and if they had not been offered SNB as part of the initial surgical procedure, an additional procedure would have been necessary to assess the axilla. However, 15 of them had positive margins and would have needed another operation regardless. Therefore, if those with 2 or more risk factors were offered SNB as part of the initial surgical procedure, we would theoretically have saved 8 reoperations (14.5%). However, the 32 of 55 patients (58.2%) whose findings were not upstaged would have undergone an unnecessary SNB.
Preoperative identification of invasive cancer or DCIS has improved the management of impalpable breast neoplasia. It has increased the proportion of cases in which definitive surgical therapy can be administered during a single operation. Definitive management of the primary tumor occurs with both invasive cancer and DCIS; however, axillary assessment is required only when invasion is present. The dilemma the surgeon faces is that a diagnosis of DCIS on CNB is prone to underestimation, which is demonstrated after definitive surgery. Patients whose cancer is upstaged to invasive cancer will then need an additional procedure to stage the axilla. In our study, 17.3% of patients with a CNB finding of DCIS were found to have invasive cancer. This rate is comparable to those in previous studies, most of which cite rates around 20%.13 Many authors have used this information to recommend SNB in a substantial proportion of cases of DCIS to reduce the need for subsequent axillary staging in cases of underestimation.26 Our study sought to identify subgroups in which underestimation was more common.
The management of DCISM remains controversial. It is uncommon, it is typically seen in approximately 5% of cases of DCIS,27 and reports suggest that the rate of axillary nodal involvement in DCISM is in the range of 4% to 17%.6,10,11,28- 32 However, most reports have small sample sizes. In our study, none of the 15 patients with a CNB finding of DCISM had a positive axillary node. We found that the cancers of 73.3% of patients with DCISM were upstaged and that microinvasion was the strongest risk factor for upstaging. Recent reports have shown a similar association between DCISM and upstaging.14,33 This rate of upstaging supports the routine use of SNB in DCISM, even if the rate of nodal involvement in DCISM is low.
In our series, palpability was a strong predictor of invasion, with nearly half of the palpable tumors being upstaged. Previous studies have also reported the link between palpability and invasion11,34; therefore, it is reasonable to suggest routine SNB for palpable cancers.
The majority of screen-detected lesions are impalpable, and predicting upstaging in impalpable lesions showing DCIS on CNB is important. Preoperative factors associated with upstaging in multivariate analysis were the presence of a mass, density, or architectural distortion on mammogram; larger mammographic size (≥20 mm); and a screening interval of 3 years or longer. High cytologic grade on CNB was predictive of upstaging in univariate but not multivariate analysis. Several authors have reported grade as a significant risk factor,8,10,13 and it may remain significant in a larger study. The presence of a mass on a mammogram has been consistently reported as a predictor of invasion.10- 12,35 Previous studies have shown that mammographic size is significantly associated with upstaging,9,13,35 and our study agrees with this finding. The larger volume to sample by biopsy may increase the chance of a sampling error, or the larger volume of disease may be more likely to have focal progression from DCIS to invasive cancer. Results of previous studies were summarized in a recently published meta-analysis,26 and our findings are consistent with the most commonly reported significant predictors of invasive disease.
A potential novel predictor for invasive cancer in screen-detected DCIS relates to the interval between mammograms. Breast screening has been shown to reduce morbidity and mortality due to breast cancer,36 but the optimal screening interval is debated. In Australia, women between 50 and 69 years of age are invited every 2 years for screening. The National Health Service Breast Screening Program in the United Kingdom has a 3-yearly screening program, whereas the average interval between screenings in the United States is 18 months.37 One UK study suggests that shortening the interval to 2 years is beneficial,38 while another disagrees.39 An interesting aspect of our study was that those lapsed attendees with a screening interval of 3 years or longer showed a significant increase in the rate of upstaging. A recent report of a screened population from the United Kingdom by Goyal et al11 also showed a higher rate of upstaging to invasive cancer (38%) than in other studies. Because invasive carcinoma often results from progression of DCIS, an increased screening interval may allow more time for progression. An argument can be made for more vigilance when a lapsed attendee with a preoperative diagnosis of DCIS is examined.
Recent trends in radiology to use vacuum-assisted biopsy with a larger needle may mean that invasive cancer is more often identified and that upstaging will be less of an issue.40 However, these new techniques have shown no significant advantage in recent reports.41,42 Our study found that vacuum-assisted 11-gauge needles have underestimation rates similar to those of 14-gauge needles.
The results of our analysis according to the number of risk factors suggest a possible management algorithm. If these findings are confirmed when applied to independent data sets, then in those with no risk factors, the chance of invasive cancer is less than 10%, and therefore SNB could be avoided. Patients with 2 or 3 of the risk factors have an incidence of underestimation of around 40%, and therefore SNB would be appropriate. Management of cases in which the patient has a single risk factor would require individual judgment, and SNB may be appropriate in selected cases. Previous studies have also highlighted the role of SNB in patients undergoing total mastectomy21,43,44 or a wide excision close to the axilla19,45 because axillary assessment may not be possible after this procedure,14,46,47 although recent evidence suggests that SNB can be successfully and reliably performed after previous breast procedures.16,48- 50
As expected, the rate of additional procedures was higher in patients found to have invasive cancer than in those who were not. This finding underlines the need to increase the preoperative detection of invasion, so that there is less need for reoperation to assess nodal status. However, it is not appropriate to recommend SNB to all patients with DCIS, especially those at very low risk of having invasive cancer. While SNB is undoubtedly a great advance over routine axillary dissection in clinically node-negative patients, it is not a trivial or harmless procedure. Sentinel node biopsy adds significantly to the cost and inconvenience of surgery, and there is definite morbidity, with a 5% incidence of lymphedema recently reported at 5 years of follow-up.51 Recommending SNB to those with 2 or more risk factors means that, in our series, 15.5% of cases (55 of 355) would have received this recommendation and 23 of them (41.8%) would have had invasive cancer at final pathological examination. Although 32 patients (58.2%) would have undergone an unnecessary axillary procedure, it is reassuring to know that there is no disease in the sentinel node of such high-risk cases of DCIS because there is likely to be an incidence of pathological sampling error in the assessment of extensive DCIS.
In our analysis, although 23 of 55 high-risk patients would have avoided reoperation for axillary assessment if they had been offered SNB as part of the initial surgical procedure, 15 would still have had to return because of positive margins, reducing the number in whom a second procedure would have been avoided to 8. Positive margins after the first procedure are a common indication for reoperation,52 especially in patients with pure DCIS. Therefore, attempts to reduce the number of additional procedures performed must take into consideration both the predictors of upstaging and predictors of unsatisfactory margins.
In summary, the finding of invasive cancer in the surgical specimen after a preoperative CNB shows DCIS is common. Microinvasion on CNB and palpability strongly predict upstaging. In patients with impalpable screen-detected cancers with DCIS on CNB, factors significantly associated with upstaging are the presence of a mass, density, or architectural distortion on mammogram; larger mammographic size (≥20 mm); and screening interval of 3 years or longer. This information may allow a more informed discussion of the role of SNB at the initial surgical procedure. We propose an algorithm whereby patients with 2 or more risk factors would be considered high risk and whereby routine SNB would be offered to this group as well.
Correspondence: G. Bruce Mann, MBBS, PhD, FRACS, Breast Unit, Ste 12, Royal Women's Hospital, Parkville 3052, Victoria, Australia (firstname.lastname@example.org).
Accepted for Publication: September 14, 2009.
Author Contributions:Study concept and design: Kurniawan, Rose, Mou, Collins, and Mann. Acquisition of data: Kurniawan, Rose, Mou, Buchanan, and Wong. Analysis and interpretation of data: Kurniawan, Rose, Mou, Collins, Wong, Miller, and Mann. Drafting of the manuscript: Kurniawan and Mann. Critical revision of the manuscript for important intellectual content: Kurniawan, Rose, Mou, Buchanan, Collins, Wong, Miller, and Mann. Statistical analysis: Kurniawan. Administrative, technical, and material support: Collins, Wong, and Mann. Study supervision: Rose, Mou, Buchanan, Collins, Miller, and Mann.
Financial Disclosure: None reported.
Previous Presentation: This study was presented in part as a poster at the San Antonio Breast Cancer Symposium; December 11, 2009; San Antonio, Texas.