Fax: (011) 618 82223204
Core imprint cytology of screen-detected breast lesions is predictive of the histologic results
Version of Record online: 21 APR 2006
Copyright © 2006 American Cancer Society
Volume 108, Issue 3, pages 150–156, 25 June 2006
How to Cite
Farshid, G. and Pieterse, S. (2006), Core imprint cytology of screen-detected breast lesions is predictive of the histologic results. Cancer, 108: 150–156. doi: 10.1002/cncr.21939
- Issue online: 12 JUN 2006
- Version of Record online: 21 APR 2006
- Manuscript Accepted: 27 OCT 2005
- Manuscript Revised: 23 OCT 2005
- Manuscript Received: 28 JUN 2005
- imprint cytology;
- breast cancer screening;
In multidisciplinary assessment clinics for screen-detected breast lesions, onsite cytopathologists provide immediate results of fine-needle aspiration biopsies (FNABs) and this information is used for patient counseling and treatment planning. Such consultation is not possible for the increasing proportion of lesions that are being assessed by core biopsy. If core imprint cytology (CIC) of breast cores can be shown to be reliable in a significant proportion of screen-detected lesions, this technique may be of clinical value in such clinics.
In the setting of a large, accredited, population-based breast cancer screening program, prospective results of CIC were gathered on 567 lesions and correlated with the results of core biopsy to determine the performance indicators for CIC.
The positive predictive value of a diagnosis of malignancy on CIC was 98.2% and the negative predictive value was 77.8%. The absolute sensitivity was 42.2%, complete sensitivity (inclusive of suspicious and atypical results) was 86.4%, absolute specificity was 56.3%, and total specificity (inclusive of acellular imprints) was 83.7%. The 2 false-positive imprints had atypical ductal hyperplasia on core histology but were found to be ductal carcinoma in situ (DCIS) on excision. False-negative imprints are a greater challenge, with 13.6% of malignant lesions producing benign-appearing or acellular imprints. Low-grade DCIS, lobular, and special type cancers account for most such lesions. The results of the current study also demonstrated significant variations in the accuracy of CIC in microcalcifications versus parenchymal lesions. In particular, the results of acellular imprints are analogous to benign CIC findings for microcalcifications but not in parenchymal lesions.
The current study may be the largest prospective series on CIC to date, and the only direct comparison of its results for microcalcifications versus parenchymal lesions. Breast CIC is a reliable predictor of core histology. The information this study provides can be used to clinical advantage. Cancer (Cancer Cytopathol) 2006. © 2006 American Cancer Society.
Core biopsy is being used increasingly as the first-line diagnostic modality for the assessment of screen-detected breast lesions. Although histologic results of core biopsies are highly reliable,1 immediate diagnosis is not possible with this technique.
BreastScreen South Australia is part of a national breast cancer-screening program, accredited since 1991. The design of this program has been described previously.2 We use fine-needle aspiration biopsy (FNAB) as the first-line diagnostic modality for most parenchymal lesions. Microcalcifications are increasingly being assessed by core biopsy, mainly because the presence of calcium, and therefore representativeness of the sample, can be evaluated on core biopsy but not on FNAB. We also use core biopsy for investigating parenchymal lesions in which the FNAB results have been indefinite or are discordant with the radiology. Our program has a multidisciplinary design, with simultaneous participation of a radiologist, a cytopathologist, and a breast surgeon at each assessment clinic. This system provides for accurate and efficient assessment of most lesions during each clinic session. Because the smears are read on-site, women who undergo FNAB have a consultation with the attending breast surgeon regarding the specifics of their breast lesion and their further management. By contrast, women with lesions that require core biopsy have to wait at least until the following day to receive any results at all. They receive these results from a general practitioner and should they require surgery, further delays are involved while referral to a breast surgeon is arranged.
Imprints of core biopsies of breast lesions are relatively simple and inexpensive to prepare. They can be readily incorporated into the workflow of existing same-day breast assessment clinics. If they can provide a reliable preliminary indication of the likely core biopsy results, this information may be valuable for same-day patient counseling and management planning.
MATERIALS AND METHODS
Since 2001 we have prepared imprints of core biopsies performed at our breast clinics and have assessed these prospectively, using the same reporting scheme as for FNAB samples.3 Our radiographers prepare the imprints by holding 1 pole of the core gently with forceps and touching the surfaces of the cores onto glass slides. Generally, 1 imprint is made per lesion. In cases of microcalcifications, the imprints are prepared from the cores that bear microcalcifications, as demonstrated by the specimen radiographs. Fewer cores are prepared from parenchymal lesions, typically no more than 3 cores in each case. The surfaces of these cores are touched onto 1 glass slide, so that generally 1 imprint is prepared from each set of cores. The imprints are air-dried and stained with Diff-Quik (Baxter Diagnostics, McGaw Park, IL), a rapid staining method. The pathologist reports the imprints onsite. The same pathologist reports the histology of the cores the following day. All results have been recorded prospectively and are audited routinely. In the current study we wished to assess the degree to which core imprint cytology (CIC) is predictive of core histology. Because we use cores for 2 main indications, namely, microcalcifications and parenchymal lesions with indefinite or discordant FNAB results, the results of CIC were assessed separately for each lesion subset.
We retrieved data pertaining to all lesions assessed by core biopsy at BreastScreen SA from June 2001 to June 2003. Through review of client charts, we tabulated patient demographics, radiologic grade, and radiologic category of lesion using the Tabar grading scheme.4 Results of any FNAB performed, the original CIC results, core biopsy histology, and the final histology of the excised specimens were recorded. In addition to the prospectively recorded data, the 2 study pathologists each reviewed the imprints independently and 1 of us also reviewed the core histology. Each set of results was recorded separately.
The performance indicators of CIC were determined based on the methods of the National Health Service Breast Screening Programme (NHSBSP).3 Using standard definitions,5 the positive predictive value (PPV) of CIC was calculated as the fraction of cases with positive CIC results that had malignant findings on core biopsy. The negative predictive value (NPV) of CIC was calculated as the fraction of lesions with a benign CIC result that had a benign core biopsy diagnosis. Absolute sensitivity was the proportion of malignant cases with a positive CIC diagnosis. Complete sensitivity was calculated as the proportion of malignant cases with positive, suspicious, or atypical CIC diagnoses. Absolute specificity was the proportion of benign lesions with benign-appearing imprints and complete specificity was calculated as the proportion of benign lesions with benign or acellular imprints.
Between June 2001 to July 2003, core biopsy was performed on 665 lesions. Imprints were available in 630 cases. The original pathologist reported the imprints in 567 cases and these are the focus of the study. The mean age of these women was 57.9 years (standard deviation [SD] 8.6 years; range, 40-84). Of these, 78.5% were in the screening target age bracket of 50 years to 69 years. In 37.8% of perimenopausal or postmenopausal women, hormone replacement therapy was used.
Imaging Characteristics, Cytologic Diagnoses, and Histologic Outcomes
The dominant radiologic abnormality constituted microcalcifications in 344 cases (60.7%) and a parenchymal lesion in the remaining cases. The parenchymal lesions included 105 stellate lesions, 78 discrete masses, 39 asymmetrical densities, and 1 clinically detected but mammographically occult mass. The dominant lesion was assigned radiologically to Tabar Grade 3 in 49.6% of lesions, Grade 4 in 19.4% of lesions, and Grade 5 in 30.9% lesions.
The core biopsy constituted the first-line diagnostic modality in 53.8% of cases. In 42.5% cases core biopsy was performed after inconclusive or discordant FNAB. In 14 cases (2.5%) a prior positive FNAB diagnosis of malignancy was available, core biopsy being performed to assess invasion. In 7 cases (1.2%) core biopsy was mandated by our protocols (e.g., papillary lesion suggested on FNAB, or positive cytology from Grade 3 masses).
Overall, the core biopsy findings indicated malignancy in 258 lesions (45.5%) and a benign process in 245 lesions (43.2%). The remaining 64 cases (11.3%) cases were classified as atypical (including atypical ductal hyperplasia [ADH] and atypical lobular hyperplasia [ALH]), suspicious for malignancy, or nonrepresentative, mainly due to the absence of microcalcifications.
The question of whether the preparation of core imprints has a detrimental effect on the core histology has been raised previously.6 We have not found any appreciable effects on the cores resulting from the preparation of the imprints.
PPV and NPV
Table 1 shows the overall concordance of CIC with the core biopsy results. In 109 of the 111 cases with positive CIC findings, the core biopsy confirmed malignancy. The remaining 2 cases were classified as ADH on core biopsy but the excised specimens showed ductal carcinoma in situ (DCIS) in both instances. The PPV of a positive CIC diagnosis was therefore 98.2%, whereas suspicious CIC was associated with an 83.3% chance of finding malignancy on core biopsy. Cases with atypical CIC were indeterminate, being found to have malignant cores in 38.6% of cases and benign cores in 40.9% of cases.
|Core Result||Imprint Cytology Diagnosis|
|Atypical (ADH, ALH)||2||5||10||17||5||39|
Of the 180 cases with benign CIC findings, 138 had benign core biopsy diagnoses, representing an NPV of a benign CIC diagnosis of 76.7%. The likelihood of finding benign histology when the imprints were acellular was close to this figure, at 72.8%. However, 11.1% of cases with benign CIC findings and 16.3% of those with acellular smears had malignant core biopsy results. From another perspective, 5.8% of malignant lesions had acellular imprints and 7.8% had benign-appearing imprints, together accounting for the 13.6% false-negative rate of CIC in malignant lesions.
Influence of Lesion Type
As shown in Table 2, the PPV of positive cytology was 100% for parenchymal lesions but was slightly lower (96.6%) for microcalcifications due to the 2 ADH cases described earlier. The NPV of a benign CIC diagnosis was nearly identical for microcalcifications and parenchymal lesions, at 76.4% and 77.4%, respectively. The absolute sensitivity and absolute specificity were also very close between the 2 groups.
|PPV Positive CIC, %||NPV Benign CIC, %||Absolute Sensitivity, %||Complete Sensitivity, %||Absolute Specificity, %||Complete Specificity, %|
There are differences in the complete sensitivity and complete specificity between the 2 categories of lesions being assessed. The complete sensitivity, which includes all positive, suspicious, and atypical imprints from malignant cores, was 81.5% for microcalcifications and was significantly higher, at 91.9%, for parenchymal lesions. This implies that 18.5% of malignant microcalcifications and 8.1% of malignant parenchymal lesions, together amounting to 13.6% of all malignant cores, had benign or acellular findings on CIC. Review of the diagnoses shows that low-grade DCIS, lobular, and special type cancers are overrepresented among the cases with false-negative CIC results.
Although the absolute specificity of CIC was very close in the 2 groups, the complete specificity, which includes acellular as well as benign imprints, was significantly different, being 92.3% for microcalcifications and 64.5% for masses. These differences reflect the varying outcomes for acellular imprints between the 2 groups of lesions. Although acellular imprints were from benign lesions in 76.6% of microcalcifications, they originated from benign processes in only 53.3% of parenchymal lesions.
There were no benign cases in which the CIC was regarded as positive. In 4 benign cases the imprints were reported as suspicious. These included 2 cases of adenosis, 1 of which had a papillary component as well, 1 fibroadenoma, and a case of duct ectasia.
11G Needle Vacuum-Assisted versus 14G Core Biopsies
The differences in the performance indicators of CIC are mirrored when the results are tabulated by type of core used (Table 3). The lesions being assessed by 11G needle cores were microcalcifications in 95.2% of the cases. By contrast, parenchymal lesions comprised 89.2% of the cases assessed by 14G needle cores. The NPV, PPV, absolute sensitivity, and absolute specificity are very similar, but the complete sensitivity was 14.2% higher among 14G needle cores, whereas the pattern was reversed for complete specificity, with a 27.4% higher value noted among 11G needle cores.
|No.||Absolute Sensitivity, %||Complete Sensitivity, %||Absolute Specificity, %||Complete Specificity, %||PPV Positive CIC Diagnosis, %||NPV Benign CIC Diagnosis, %|
|11G VA core biopsy||331||38.7||79.0||57.7||92.3||98.0||76.4|
|14G core biopsy||232||45.9||93.2||52.7||64.9||98.4||76.5|
CIC in Cases with Indefinite Core Biopsy Results
In 11.3% of cases the core biopsy results were indefinite. These included 36 cases of ADH, 2 cases of ALH, 1 fibroadenoma with atypical features, 13 cases that were suspicious but not diagnostic of malignancy, and 12 nonrepresentative biopsies. Two cases had positive CIC results and both were proven to be malignant on excision. Eight of the 12 cases with suspicious CIC were also malignant, but the findings of the remaining imprints were not particularly helpful in these difficult lesions. Among the 12 cases with nonrepresentative cores, 3 had suspicious CIC results. These were malignant on excision.
Clinical Value of CIC
When core biopsy was performed as the first-line diagnostic modality, 41.7% of malignant cases had positive core imprints. Among cases requiring cores biopsy after indefinite or discordant FNAB, the generally greater cellularity of the imprints enabled us to provide a positive cytologic diagnosis in 36.8% of malignant cases (Figs. 1 and 2). The imprints were particularly useful in these cases. Overall, 40% of malignant lesions assessed by core biopsy had immediate positive CIC results.
Regarding benign lesions, 64.6% of first-line core biopsies and 92.7% of second-line cores, together amounting to 83.5% of benign lesions, were correctly identified on the basis of benign or acellular imprints.
The above analysis was based on the prospectively accrued, original CIC results, read by any 1 of 6 program pathologists. As part of this study, 2 of these pathologists each reviewed the core imprints retrospectively. The results for PPV and NPVs varied within very narrow ranges.
To our knowledge, this is the largest series to date assessing the value of breast core imprint cytology exclusively for screen-detected lesions and with prospective data. It is also to our knowledge the first study that provides comparison data on the use of this technique for microcalcifications and parenchymal lesions. Our results show that CIC is quite predictive of the results of breast core biopsies, with good interobserver agreement rates. Overall, 40% of malignant lesions assessed by core biopsy had immediate positive CIC results, with a PPV of 98.2%. This is an acceptable value for PPV, particularly because the 2 “false positive” cases did have DCIS on the final excision specimen. When positive, suspicious, and atypical CIC results are combined, 86.4% of all patients with malignant lesions on core biopsy are afforded the opportunity to discuss their specific findings with an on-site breast surgeon.
A similarly high proportion of benign lesions (83.5%) had benign or acellular imprints. These women may be reassured by discussing these preliminary results with a breast surgeon. There were no false-positive diagnoses, but 1.6% of patients with benign core biopsy findings had suspicious imprints. As with FNAB, false-positive diagnoses are to be expected with this cytologic technique, and the same types of lesions cause diagnostic difficulties.7, 8 These included duct adenoma, fibroadenoma, columnar cell change, and sclerosing adenosis. Unlike FNAB, with CIC the delay until the receipt of the definitive core biopsy histologic results is usually only 1 day, with no surgical management implications.
False-negative diagnoses are a larger challenge than false-positives, with 13.6% of malignant lesions having acellular or benign-appearing imprints. Because core imprints are a form of exfoliative cytology, it is not surprising that the same group of lesions that are problematic for FNAB—namely, low-grade DCIS and desmoplastic and special type cancers—produce most false-negative CIC results.
Approximately 10% of cases had indefinite core biopsy results. CIC was generally not useful in these challenging cases, but it is of interest that a small subset of these lesions had positive or suspicious CIC and nonmalignant cores, but the final histology was indeed malignant. One explanation for such results is that the area involved by the malignant process may be sheared by the core biopsy, yielding insufficient material for a histologic diagnosis, but because the atypical cells are shed on the imprint, a cytologic diagnosis of malignancy may be possible.
In FNAB, acellular smears are regarded as inadequate. By extrapolation, several series have excluded acellular smears from further assessment.9, 10 Acellular smears have comprised 2.7% to 29%9, 10 of all cases in different series. In this study, 16.2% of the smears were acellular. The performance characteristics of acellular imprints were similar to those of benign imprints, particularly for microcalcifications, in which the NPV was nearly identical, at 77.4% and 77.6%, respectively. Newman et al.,6 addressing the role of imprint cytology in microcalcifications, also found that 11 of 14 of their acellular imprints (78.6%) originated from benign lesions. This suggests that the failure of a lesion to shed its cells onto the imprints, particularly in cases of microcalcifications, is itself informative, favoring a benign process. Acellular imprints in parenchymal lesions are less reassuring because desmoplastic cancers, including lobular lesions, do not shed their cells readily in cytologic samples.
In comparing the results of CIC for microcalcifications versus parenchymal lesions, we found that, whereas the positive and negative predictive values are very similar, there are significant differences in absolute sensitivity and absolute specificity between the 2 groups. One explanation may be the method used to obtain the cores. Certainly these differences are echoed in the results of the technique for 14G versus 11G needle cores. However, it is more likely that the differences are attributable to the underlying nature and specifically the varying rates of malignancy among the lesions being assessed by the 2 types of core biopsies. We use 11G needle vacuum-assisted core biopsies chiefly for microcalcifications, which as a group were malignant in 39.2% of the cases. By contrast, 14G ultrasound-guided cores are used mostly for parenchymal lesions, among which the rate of malignancy was higher, at 55.2%. It is well established that the prior probability of malignancy influences the performance indicators of a test.5
Table 4 summarizes the reported literature on CIC. The criteria for patient selection have varied among various series, as some have focused on screen-detected lesions,8, 10, 11 or subsets thereof.6 Others assessed symptomatic patients,9, 12, 13 and a combination of the 2 groups has been presented in other series.14 Because the prevalence of malignancy is different among screen-detected versus symptomatic lesions, the 2 groups are not comparable directly. The methods of calculating the performance indices have also varied. Following the recommendations of the NHSBSP,3 we and others have grouped cases in the atypical category with positive and suspicious results,6, 10 whereas others have combined them with benign imprints.8, 9, 12, 13 Our results indicate that atypical CIC is a truly equivocal category, with almost equal odds of being malignant or benign.
|Reference||No.||Patients||Sensitivity, %||Specificity, %||PPV, %||NPV, %||Source|
|Jacobs et al., 19999||52||“Routine breast evaluation”||94.0||95.0||100.0||100.0||U.S.|
|March et al., 19997||102||Not specified||75.0||93.0||85.7||93.2||U.S.|
|Sneige et al., 20008||172||Impalpable||83.0||95.0||100.0||74.3||U.S.|
|Albert et al., 200014||173||Symptomatic and screen detected||95.9||96.5||97.8||84.3||Germany|
|Newman et al., 20016||203||Screen-detected microcalcifications||94.7||87.7||100.0||95.0||Western Australia|
|Green and Mathew, 200111||100||Screen detected||92.3||98.0||96.0||96.3||U.S.|
|Anlauf 200315||563||Not stated in abstract||89.0||88.0||96.0||67.0||Germany|
|Kass et al., 200312||199||“Presenting to outpatients”||74.0||97.0||93.0||86.0||U.S.|
|Jones et al., 200410||111||Screen detected||97.0||78.0||100.0||98.5||U.K.|
|Carmichael et al., 200413||41||Symptomatic||91.0||89.0||97.0||73.0||U.K.|
|Current study||567||Screen detected||86.4||83.7||98.2||77.8||South Australia|
Despite the differences in methodology, the published performance indicators of imprint cytology have been consistently good, particularly for a preliminary test. The sensitivity has exceeded 85% and the specificity has been higher that 80% in most series. The slightly lower sensitivity in the current series is due to the fact that cases with difficult cytology are overrepresented, because lesions with positive FNAB usually do not go on to core biopsy in our program.
Core imprint cytology provides reliable and timely preliminary results of breast core biopsies. Whereas the advantage of a definitive positive result is available in only 40% of malignant cores, in experienced hands these results are highly reliable. This technique is associated with negligible additional cost and may be incorporated into the existing design of multidisciplinary breast clinics to offer same-day counseling for more patients undergoing core biopsy.
The authors thank the radiologists, pathologists, and surgeons associated with BreastScreen SA and the program radiographers for help in preparing the core imprints. They also thank the Medical Executive Committee for comments on an earlier draft of the article.
- 3National Health Service Breast Screening Programme. Guidelines for non-operative diagnostic procedures and reporting in breast cancer. London: National Health Service Breast Screening Programme, 2001.
- 4Teaching atlas of mammography. New York: Thieme; 1983., .
- 5Medical decision making. Stoneham, MA: Butterworth; 1988., , , .