Diagnostic Performance of Ultrasound-Guided Fine-Needle Aspiration of Nonpalpable Breast Lesions in a Multidisciplinary Setting: The Institut Curie’s Experience
Abstract
Objectives: To assess the diagnostic performance of ultrasound-guided fine-needle aspiration (USFNA) in nonpalpable breast lesions (NPBLs) in a multidisciplinary setting.
Methods: In total, 2,601 NPBLs underwent USFNA by a radiologist-pathologist team. Gold-standard diagnosis was based on surgery, core-needle biopsy, or 1-year imaging follow-up. USFNA’s diagnostic performance was analyzed in different clinical and imaging subgroups.
Results: USFNA’s sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were, respectively, 92.6% (95% confidence interval [CI], 90.8%-94.2%), 96.8% (95% CI, 95.8%-97.6%), 94.8% (95% CI, 93.2%-96.1%), and 95.4% (95% CI, 94.3%-96.4%). The best PPV was achieved in Breast-Imaging Reporting and Data System (BI-RADS) categories 4C and 5 and the best NPV in BI-RADS categories 2, 3, and 4A and in patients younger than 50 years. The mitotic count, BI-RADS categories, associated palpable cancer, and age (<50 or ≥50 years) were statistically independent factors (P < .05) between USFNA’s false-negative and true-positive results.
Conclusions: USFNA is a robust diagnostic procedure in NPBLs. Age and the BI-RADS category of the lesion are important factors determining its performance.
Study flowchart. BI-RADS, Breast-Imaging Reporting and Data System; IFU, incomplete follow-up; LFU, lost to follow-up; USFNA, ultrasound-guided fine-needle aspiration.
Study flowchart. BI-RADS, Breast-Imaging Reporting and Data System; IFU, incomplete follow-up; LFU, lost to follow-up; USFNA, ultrasound-guided fine-needle aspiration.
Materials and Methods
Over the past three decades, detection of nonpalpable breast lesions (NPBLs) has significantly increased due to breast cancer screening programs,1 based on breast imaging. To better standardize breast imaging reports and define patient management, the Breast-Imaging Reporting and Data System (BI-RADS) lexicons of the American College of Radiology (ACR) have been introduced.2 Each category is defined by a range of positive predictive values (PPVs) for cancer. Accordingly, it is recommended that BI-RADS categories 2 and 3, with low PPVs (0% and <2%, respectively), only warrant imaging follow-up, whereas higher BI-RADS categories (from 4-5) warrant pathologic evaluation. The number of NPBLs requiring pathologic diagnosis has significantly increased as well.3 Percutaneous imaging-guided sampling techniques have been developed to sample NPBLs—namely, fine-needle aspiration (FNA), core-needle biopsy (CNB), and vacuum-assisted biopsy (VAB). Stereotactic, ultrasound (US), or magnetic resonance imaging (MRI) guidance have been successively developed to target NPBLs and coexist nowadays.4 Use of these techniques has improved patient management and avoided unnecessary surgical biopsies for benign lesions.5
Regarding FNA, stereotactic guidance has been demonstrated to show inadequate diagnostic accuracy and is therefore not recommended.4,6 US-guided FNA (USFNA) is a valuable procedure because it is rapid, economical, simple, and safe in comparison to CNB and VAB.7 Several studies have assessed its diagnostic performance and found variable results.8‐27 At this time, none have studied in detail which factors affect its performance such as patient and breast imaging characteristics (BI-RADS classification and subcategorization in categories 4A, 4B, and 4C) or histologic differences between USFNA true-positive and false-negative results (reported to be between 0.7% and 21.6%8‐27). Better understanding of these factors could help improve the technique and define which populations would benefit the most from it.
Our institution has a long experience in breast cytology dating back to its early development by Zajdela28 in the mid-20th century. Since 1992, we have introduced USFNA as a diagnostic tool for NPBLs. One particular feature of USFNA in our institution is the “four-hand procedure” where the radiologist selects the lesion and keeps track of the target with the US probe while the pathologist performs the aspiration. Improved diagnostic performance has been reported when experienced pathologists are involved in the sampling procedure,29 yet, this has not been analyzed in a large study. From our experience, we also think that this four-hand procedure further improves the diagnostic performance of USFNA.
We have previously published a preliminary study validating the use of USFNA in NPBLs.16 In the present study, we report on our experience in a larger series of NPBLs collected and diagnosed using USFNA during 4 consecutive years. Its diagnostic performance and the main clinical, imaging, cytologic, and pathologic factors influencing its performance were analyzed.
Results
Patients
The institutional review board of our institution approved this study, and the need for patients’ informed consent was waived.
USFNA
All NPBLs that underwent USFNA between 2003 and 2007 at our institution were considered eligible for this study. Patients were referred to our consultation for NPBLs detected at screening, during follow-up of a previously treated breast cancer, or in the workup of a recently diagnosed breast lesion. Patients’ characteristics (age, sex, history of breast cancer) and lesion location within the breast and size at US examination were recorded during the procedure. A patient, referred for assessment of additional NPBLs discovered in the workup of a recently diagnosed breast cancer, was considered as having a history of breast cancer. Clinical, breast imaging, and cytological data of the USFNA procedures were prospectively recorded in our local database storage system and were extracted for this study. Overall, 3,865 NPBLs were consecutively diagnosed by USFNA in 2,891 patients during the study timeframe, and 1,264 (32.7%) NPBLs in 864 patients were excluded from analysis: five (0.4%) in four patients because malignancy had already been confirmed by previous sampling (BI-RADS category 6), 843 (66.8%) in 613 patients who were lost to follow-up (including 261 [20.7%] in 195 patients followed up for less than a year), 355 (28.1%) in 209 patients who were treated without histologic confirmation of the nature of the lesion, and 61 (4.7%) in 38 patients because the final diagnosis was inconclusive (in 36 due to neoadjuvant treatment, in 15 due to multiple nodules, and in 10 due to diffuse cancer).
USFNA Sensitivity, Specificity, PPV, and NPV Subgroup Analysis According to Clinical, Breast Imaging, and Aspiration Criteria
All lesions were sampled by a radiologist-pathologist team, composed by one of nine senior radiologists performing the breast US examinations and one of four pathologists to perform the aspirations, or by the radiologist alone in the absence of the pathologist.
Distribution of mitotic indexes relative to the histologic grade in false-negative (FN) and true-positive (TP) ultrasound-guided fine-needle aspiration (USFNA) results. In TP cases, the mitotic count increased with the histologic grade, whereas in FN cases, the mitotic count stayed lower overall, independently of the grade. This difference in distribution explains that, although both the histologic grade and the mitotic count were significantly different between FN and TP USFNA results at univariate analysis (P < .05), only the latter was independently associated with FN and TP rates in multivariate analysis (P = .01).
Distribution of mitotic indexes relative to the histologic grade in false-negative (FN) and true-positive (TP) ultrasound-guided fine-needle aspiration (USFNA) results. In TP cases, the mitotic count increased with the histologic grade, whereas in FN cases, the mitotic count stayed lower overall, independently of the grade. This difference in distribution explains that, although both the histologic grade and the mitotic count were significantly different between FN and TP USFNA results at univariate analysis (P < .05), only the latter was independently associated with FN and TP rates in multivariate analysis (P = .01).
Univariate Analysis Comparison of Clinical and Imaging Characteristics Between USFNA FN and TP Casesa
| Clinical and Imaging Characteristics | Cytology
|
||
|---|---|---|---|
| FN | TP | P Value | |
| Age, y | |||
| <50 | 5 (10.6) | 135 (23.6) | .04b |
| ≥50 | 42 (89.4) | 437 (76.4) | |
| Menopause | |||
| Yes | 35 (74.5) | 361 (63.1) | .27 |
| No | 8 (17.0) | 156 (27.3) | |
| NA | 4 (8.5) | 55 (9.6) | |
| HRT | |||
| Yes | 15 (31.9) | 137 (24.0) | .51 |
| No | 6 (12.8) | 95 (16.6) | |
| NA | 26 (55.3) | 340 (59.4) | |
| Personal history of breast cancer | |||
| Ipsilateral | 14 (29.8) | 148 (25.9) | .39 |
| Contralateral | 2 (4.3) | 22 (3.8) | |
| Bilateral | 0 (0) | 31 (5.4) | |
| None | 31 (66) | 371 (64.9) | |
| Family history of breast cancer | |||
| Yes | 10 (21.3) | 146 (25.5) | .68 |
| No | 13 (27.7) | 141 (24.7) | |
| NA | 24 (51.1) | 285 (49.8) | |
| BI-RADS category | |||
| 0 | 2 (4.3) | 5 (0.9) | .02b |
| 2 | 0 (0) | 0 (0) | |
| 3 | 1 (2.1) | 16 (2.8) | |
| 4A | 2 (4.3) | 25 (4.4) | |
| 4B | 13 (27.7) | 69 (12.1) | |
| 4C | 15 (31.9) | 252 (44.1) | |
| 5 | 14 (29.8) | 205 (35.8) | |
| Size, mm | |||
| Mean ± SD | 8.6 ± 4.9 | 9.7 ± 5.3 | .05 |
| ≤5 | 10 (21.3) | 87 (15.2) | .2 |
| 5-10 | 28 (59.6) | 305 (53.3) | |
| 10-20 | 7 (14.9) | 158 (27.6) | |
| >20 | 2 (4.3) | 21 (3.7) | |
| NA | 0 (0) | 1 (0.2) | |
| Associated palpable cancer | |||
| Yes | 23 (49.0) | 204 (35.7) | .02b |
| No | 19 (40.4) | 350 (61.2) | |
| NA | 5 (10.6) | 18 (3.1) | |
| Clinical and Imaging Characteristics | Cytology
|
||
|---|---|---|---|
| FN | TP | P Value | |
| Age, y | |||
| <50 | 5 (10.6) | 135 (23.6) | .04b |
| ≥50 | 42 (89.4) | 437 (76.4) | |
| Menopause | |||
| Yes | 35 (74.5) | 361 (63.1) | .27 |
| No | 8 (17.0) | 156 (27.3) | |
| NA | 4 (8.5) | 55 (9.6) | |
| HRT | |||
| Yes | 15 (31.9) | 137 (24.0) | .51 |
| No | 6 (12.8) | 95 (16.6) | |
| NA | 26 (55.3) | 340 (59.4) | |
| Personal history of breast cancer | |||
| Ipsilateral | 14 (29.8) | 148 (25.9) | .39 |
| Contralateral | 2 (4.3) | 22 (3.8) | |
| Bilateral | 0 (0) | 31 (5.4) | |
| None | 31 (66) | 371 (64.9) | |
| Family history of breast cancer | |||
| Yes | 10 (21.3) | 146 (25.5) | .68 |
| No | 13 (27.7) | 141 (24.7) | |
| NA | 24 (51.1) | 285 (49.8) | |
| BI-RADS category | |||
| 0 | 2 (4.3) | 5 (0.9) | .02b |
| 2 | 0 (0) | 0 (0) | |
| 3 | 1 (2.1) | 16 (2.8) | |
| 4A | 2 (4.3) | 25 (4.4) | |
| 4B | 13 (27.7) | 69 (12.1) | |
| 4C | 15 (31.9) | 252 (44.1) | |
| 5 | 14 (29.8) | 205 (35.8) | |
| Size, mm | |||
| Mean ± SD | 8.6 ± 4.9 | 9.7 ± 5.3 | .05 |
| ≤5 | 10 (21.3) | 87 (15.2) | .2 |
| 5-10 | 28 (59.6) | 305 (53.3) | |
| 10-20 | 7 (14.9) | 158 (27.6) | |
| >20 | 2 (4.3) | 21 (3.7) | |
| NA | 0 (0) | 1 (0.2) | |
| Associated palpable cancer | |||
| Yes | 23 (49.0) | 204 (35.7) | .02b |
| No | 19 (40.4) | 350 (61.2) | |
| NA | 5 (10.6) | 18 (3.1) | |
BI-RADS, Breast-Imaging Reporting and Data System; FP, false positive; FN, false negative; HRT, hormone replacement therapy; NA, not available; USFNA, ultrasound-guided fine-needle aspiration.
Values are presented as number (%) unless otherwise indicated.
Statistically significant difference (P < .05) using Fisher exact test or χ2 test when appropriate.
Univariate Analysis Comparison of Clinical and Imaging Characteristics Between USFNA FN and TP Casesa
| Clinical and Imaging Characteristics | Cytology
|
||
|---|---|---|---|
| FN | TP | P Value | |
| Age, y | |||
| <50 | 5 (10.6) | 135 (23.6) | .04b |
| ≥50 | 42 (89.4) | 437 (76.4) | |
| Menopause | |||
| Yes | 35 (74.5) | 361 (63.1) | .27 |
| No | 8 (17.0) | 156 (27.3) | |
| NA | 4 (8.5) | 55 (9.6) | |
| HRT | |||
| Yes | 15 (31.9) | 137 (24.0) | .51 |
| No | 6 (12.8) | 95 (16.6) | |
| NA | 26 (55.3) | 340 (59.4) | |
| Personal history of breast cancer | |||
| Ipsilateral | 14 (29.8) | 148 (25.9) | .39 |
| Contralateral | 2 (4.3) | 22 (3.8) | |
| Bilateral | 0 (0) | 31 (5.4) | |
| None | 31 (66) | 371 (64.9) | |
| Family history of breast cancer | |||
| Yes | 10 (21.3) | 146 (25.5) | .68 |
| No | 13 (27.7) | 141 (24.7) | |
| NA | 24 (51.1) | 285 (49.8) | |
| BI-RADS category | |||
| 0 | 2 (4.3) | 5 (0.9) | .02b |
| 2 | 0 (0) | 0 (0) | |
| 3 | 1 (2.1) | 16 (2.8) | |
| 4A | 2 (4.3) | 25 (4.4) | |
| 4B | 13 (27.7) | 69 (12.1) | |
| 4C | 15 (31.9) | 252 (44.1) | |
| 5 | 14 (29.8) | 205 (35.8) | |
| Size, mm | |||
| Mean ± SD | 8.6 ± 4.9 | 9.7 ± 5.3 | .05 |
| ≤5 | 10 (21.3) | 87 (15.2) | .2 |
| 5-10 | 28 (59.6) | 305 (53.3) | |
| 10-20 | 7 (14.9) | 158 (27.6) | |
| >20 | 2 (4.3) | 21 (3.7) | |
| NA | 0 (0) | 1 (0.2) | |
| Associated palpable cancer | |||
| Yes | 23 (49.0) | 204 (35.7) | .02b |
| No | 19 (40.4) | 350 (61.2) | |
| NA | 5 (10.6) | 18 (3.1) | |
| Clinical and Imaging Characteristics | Cytology
|
||
|---|---|---|---|
| FN | TP | P Value | |
| Age, y | |||
| <50 | 5 (10.6) | 135 (23.6) | .04b |
| ≥50 | 42 (89.4) | 437 (76.4) | |
| Menopause | |||
| Yes | 35 (74.5) | 361 (63.1) | .27 |
| No | 8 (17.0) | 156 (27.3) | |
| NA | 4 (8.5) | 55 (9.6) | |
| HRT | |||
| Yes | 15 (31.9) | 137 (24.0) | .51 |
| No | 6 (12.8) | 95 (16.6) | |
| NA | 26 (55.3) | 340 (59.4) | |
| Personal history of breast cancer | |||
| Ipsilateral | 14 (29.8) | 148 (25.9) | .39 |
| Contralateral | 2 (4.3) | 22 (3.8) | |
| Bilateral | 0 (0) | 31 (5.4) | |
| None | 31 (66) | 371 (64.9) | |
| Family history of breast cancer | |||
| Yes | 10 (21.3) | 146 (25.5) | .68 |
| No | 13 (27.7) | 141 (24.7) | |
| NA | 24 (51.1) | 285 (49.8) | |
| BI-RADS category | |||
| 0 | 2 (4.3) | 5 (0.9) | .02b |
| 2 | 0 (0) | 0 (0) | |
| 3 | 1 (2.1) | 16 (2.8) | |
| 4A | 2 (4.3) | 25 (4.4) | |
| 4B | 13 (27.7) | 69 (12.1) | |
| 4C | 15 (31.9) | 252 (44.1) | |
| 5 | 14 (29.8) | 205 (35.8) | |
| Size, mm | |||
| Mean ± SD | 8.6 ± 4.9 | 9.7 ± 5.3 | .05 |
| ≤5 | 10 (21.3) | 87 (15.2) | .2 |
| 5-10 | 28 (59.6) | 305 (53.3) | |
| 10-20 | 7 (14.9) | 158 (27.6) | |
| >20 | 2 (4.3) | 21 (3.7) | |
| NA | 0 (0) | 1 (0.2) | |
| Associated palpable cancer | |||
| Yes | 23 (49.0) | 204 (35.7) | .02b |
| No | 19 (40.4) | 350 (61.2) | |
| NA | 5 (10.6) | 18 (3.1) | |
BI-RADS, Breast-Imaging Reporting and Data System; FP, false positive; FN, false negative; HRT, hormone replacement therapy; NA, not available; USFNA, ultrasound-guided fine-needle aspiration.
Values are presented as number (%) unless otherwise indicated.
Statistically significant difference (P < .05) using Fisher exact test or χ2 test when appropriate.
Univariate Analysis Comparison of Pathologic Characteristics Between USFNA FN and TP Casesa
| Cytology
|
|||
|---|---|---|---|
| Tumor Characteristics | FN | TP | P Value |
| Type | |||
| Ductal invasive | 33 (70.2) | 451 (78.8) | .32 |
| Lobular invasive | 12 (25.5) | 81 (14.2) | |
| Mucinous | 0 (0) | 3 (0.5) | |
| Other | 2 (4.3) | 37 (6.5) | |
| Lymphovascular invasion | |||
| Yes | 5 (10.6) | 339 (59.3) | .06 |
| No | 35 (74.5) | 90 (15.7) | |
| NA | 7 (14.9) | 143 (25.0) | |
| Grade (EEb) | |||
| I | 29 (61.7) | 228 (39.9) | .01c |
| II | 14 (29.8) | 209 (36.5) | |
| III | 3 (6.4) | 118 (20.6) | |
| NA | 1 (2.1) | 17 (3.0) | |
| Hormone receptors | |||
| Positive | 41 (87.2) | 484 (84.6) | .85 |
| Negative | 4 (8.5) | 60 (10.5) | |
| NA | 2 (4.3) | 28 (4.9) | |
| HER2 status | |||
| Amplified | 1 (2.1) | 228 (39.9) | .33 |
| Not amplified | 21 (44.7) | 36 (6.3) | |
| NA | 25 (53.2) | 308 (53.8) | |
| Tumor size (infiltrative), mm | |||
| Mean ± SD | 14.9 ± 13.1 | 16.2 ± 10.9 | .09 |
| Mitotic count | |||
| Mean (mitosis/field) ± SD | 3.5 ± 2.34 | 10.6 ± 5.16 | <.001c |
| ≤11 | 43 (91.5) | 406 (70.9) | .01c |
| 11-22 | 2 (4.3) | 64 (11.1) | |
| >22 | 2 (4.2) | 97 (17.0) | |
| NA | 0 (0) | 5 (1.0) | |
| Cytology
|
|||
|---|---|---|---|
| Tumor Characteristics | FN | TP | P Value |
| Type | |||
| Ductal invasive | 33 (70.2) | 451 (78.8) | .32 |
| Lobular invasive | 12 (25.5) | 81 (14.2) | |
| Mucinous | 0 (0) | 3 (0.5) | |
| Other | 2 (4.3) | 37 (6.5) | |
| Lymphovascular invasion | |||
| Yes | 5 (10.6) | 339 (59.3) | .06 |
| No | 35 (74.5) | 90 (15.7) | |
| NA | 7 (14.9) | 143 (25.0) | |
| Grade (EEb) | |||
| I | 29 (61.7) | 228 (39.9) | .01c |
| II | 14 (29.8) | 209 (36.5) | |
| III | 3 (6.4) | 118 (20.6) | |
| NA | 1 (2.1) | 17 (3.0) | |
| Hormone receptors | |||
| Positive | 41 (87.2) | 484 (84.6) | .85 |
| Negative | 4 (8.5) | 60 (10.5) | |
| NA | 2 (4.3) | 28 (4.9) | |
| HER2 status | |||
| Amplified | 1 (2.1) | 228 (39.9) | .33 |
| Not amplified | 21 (44.7) | 36 (6.3) | |
| NA | 25 (53.2) | 308 (53.8) | |
| Tumor size (infiltrative), mm | |||
| Mean ± SD | 14.9 ± 13.1 | 16.2 ± 10.9 | .09 |
| Mitotic count | |||
| Mean (mitosis/field) ± SD | 3.5 ± 2.34 | 10.6 ± 5.16 | <.001c |
| ≤11 | 43 (91.5) | 406 (70.9) | .01c |
| 11-22 | 2 (4.3) | 64 (11.1) | |
| >22 | 2 (4.2) | 97 (17.0) | |
| NA | 0 (0) | 5 (1.0) | |
FP, false positive; FN, false negative; HER2, human epidermal growth factor receptor 2; NA, not available; USFNA, ultrasound-guided fine-needle aspiration.
Values are presented as number (%) unless otherwise indicated.
Elston-Ellis modification of Scarff-Bloom-Richardson grading system.
Statistically significant difference (P < .05) using Fisher exact test or χ2 test when appropriate.
Univariate Analysis Comparison of Pathologic Characteristics Between USFNA FN and TP Casesa
| Cytology
|
|||
|---|---|---|---|
| Tumor Characteristics | FN | TP | P Value |
| Type | |||
| Ductal invasive | 33 (70.2) | 451 (78.8) | .32 |
| Lobular invasive | 12 (25.5) | 81 (14.2) | |
| Mucinous | 0 (0) | 3 (0.5) | |
| Other | 2 (4.3) | 37 (6.5) | |
| Lymphovascular invasion | |||
| Yes | 5 (10.6) | 339 (59.3) | .06 |
| No | 35 (74.5) | 90 (15.7) | |
| NA | 7 (14.9) | 143 (25.0) | |
| Grade (EEb) | |||
| I | 29 (61.7) | 228 (39.9) | .01c |
| II | 14 (29.8) | 209 (36.5) | |
| III | 3 (6.4) | 118 (20.6) | |
| NA | 1 (2.1) | 17 (3.0) | |
| Hormone receptors | |||
| Positive | 41 (87.2) | 484 (84.6) | .85 |
| Negative | 4 (8.5) | 60 (10.5) | |
| NA | 2 (4.3) | 28 (4.9) | |
| HER2 status | |||
| Amplified | 1 (2.1) | 228 (39.9) | .33 |
| Not amplified | 21 (44.7) | 36 (6.3) | |
| NA | 25 (53.2) | 308 (53.8) | |
| Tumor size (infiltrative), mm | |||
| Mean ± SD | 14.9 ± 13.1 | 16.2 ± 10.9 | .09 |
| Mitotic count | |||
| Mean (mitosis/field) ± SD | 3.5 ± 2.34 | 10.6 ± 5.16 | <.001c |
| ≤11 | 43 (91.5) | 406 (70.9) | .01c |
| 11-22 | 2 (4.3) | 64 (11.1) | |
| >22 | 2 (4.2) | 97 (17.0) | |
| NA | 0 (0) | 5 (1.0) | |
| Cytology
|
|||
|---|---|---|---|
| Tumor Characteristics | FN | TP | P Value |
| Type | |||
| Ductal invasive | 33 (70.2) | 451 (78.8) | .32 |
| Lobular invasive | 12 (25.5) | 81 (14.2) | |
| Mucinous | 0 (0) | 3 (0.5) | |
| Other | 2 (4.3) | 37 (6.5) | |
| Lymphovascular invasion | |||
| Yes | 5 (10.6) | 339 (59.3) | .06 |
| No | 35 (74.5) | 90 (15.7) | |
| NA | 7 (14.9) | 143 (25.0) | |
| Grade (EEb) | |||
| I | 29 (61.7) | 228 (39.9) | .01c |
| II | 14 (29.8) | 209 (36.5) | |
| III | 3 (6.4) | 118 (20.6) | |
| NA | 1 (2.1) | 17 (3.0) | |
| Hormone receptors | |||
| Positive | 41 (87.2) | 484 (84.6) | .85 |
| Negative | 4 (8.5) | 60 (10.5) | |
| NA | 2 (4.3) | 28 (4.9) | |
| HER2 status | |||
| Amplified | 1 (2.1) | 228 (39.9) | .33 |
| Not amplified | 21 (44.7) | 36 (6.3) | |
| NA | 25 (53.2) | 308 (53.8) | |
| Tumor size (infiltrative), mm | |||
| Mean ± SD | 14.9 ± 13.1 | 16.2 ± 10.9 | .09 |
| Mitotic count | |||
| Mean (mitosis/field) ± SD | 3.5 ± 2.34 | 10.6 ± 5.16 | <.001c |
| ≤11 | 43 (91.5) | 406 (70.9) | .01c |
| 11-22 | 2 (4.3) | 64 (11.1) | |
| >22 | 2 (4.2) | 97 (17.0) | |
| NA | 0 (0) | 5 (1.0) | |
FP, false positive; FN, false negative; HER2, human epidermal growth factor receptor 2; NA, not available; USFNA, ultrasound-guided fine-needle aspiration.
Values are presented as number (%) unless otherwise indicated.
Elston-Ellis modification of Scarff-Bloom-Richardson grading system.
Statistically significant difference (P < .05) using Fisher exact test or χ2 test when appropriate.
Gold-Standard Diagnosis Criteria
Discussion
Lesion Characteristics
Histologic diagnoses were classified into four categories for CNB (ie, benign, suspicious, malignant, or inadequate) and into two categories (ie, benign or malignant) for surgical excision specimens. Atypical epithelial hyperplasia was considered suspicious on CNB and benign on surgical excision specimens. A definitive diagnosis was established on lesions with complete excision or with a CNB positive for cancer. Benign findings on CNB were not considered a reference standard because of potential missed sampling of the lesion.
Lesions with complex radiopathologic correlation (diffuse cancer, multiple nodules, histologic analysis of the specimen after neoadjuvant treatment) were reviewed by two of the authors (J.A.F.R. and J.K.) to establish the definitive diagnosis by consensus (benign, malignant, or inconclusive when the nature of the lesion could not be established).
In patients with neither complete excision of the lesion nor malignant CNB, benignity was confirmed by breast imaging follow-up at 12 months in accordance with ACR guidelines.31
NPBLs were classified according to the BI-RADS lexicon by the radiologist performing the breast US examination before any sampling was performed. Classification was established by review of the primary breast imaging workup (including mammography and breast MRI) available at the time of USFNA. Complete collapse of the lesion after aspiration (defined by the complete disappearance of the lesion at US after aspiration) was also recorded, as well as cytologic characteristics evoking cystic lesions. The final histologic type of cancers sampled by USFNA was noted (ductal invasive and in situ carcinomas, lobular invasive and in situ carcinomas, other types of primary invasive carcinoma, sarcoma, lymphoma).
The subset of patients diagnosed with breast cancer was cross-checked with our institute’s breast cancer registry, a database of all patients with breast cancer treated at our institution. All available data on patient and tumor characteristics and evolution are prospectively collected and updated in this database. The following personal data and cancer’s histologic characteristics were retrieved: menopausal status, hormone replacement therapy, family history of breast cancer, associated palpable cancer to the targeted NPBL, TNM status, Elston-Ellis histologic grade, mitotic index, hormone receptor status, human epidermal growth factor receptor 2 status, and tumor size. Only patients with primary invasive breast carcinomas, initially diagnosed with a full workup and treated at our institution, were considered for retrieving these additional data. Patients who were managed outside our institution after USFNA, presented in situ carcinomas or sarcomas/lymphomas of the breast, or were diagnosed for a recurrence during treatment were not considered patients with primary invasive breast carcinomas.
True-negative (TN) cases were defined as NPBLs with inadequate or benign USFNA and a definitive diagnosis of benignity. True-positive (TP) cases were defined as NPBLs with suspicious or malignant USFNA and a definitive diagnosis of malignancy. False-negative (FN) cases included NBPLs with inadequate or benign USFNA and a definitive diagnosis of malignancy. False-suspicious (FS) cases included NBPLs with suspicious USFNA and a definitive diagnosis of benignity. False-positive (FP) cases included NBPLs with malignant USFNA and a definitive diagnosis of benignity.
To assess the diagnostic performance, diagnostic accuracy (= TN + TP, excluding inadequate and suspicious lesions, over all NPBLs), sensitivity (= TP/(TP + FN)), specificity (= TN/(TN + FP + FS)), and PPVs and negative predictive values (NPVs) (PPV = TP/(TP + FP + FS) and NPV = TN/(TN + FN)) and their respective 95% confidence intervals (CIs) were calculated. These were also calculated and compared using the McNemar χ2 test in different subgroups according to clinical and imaging criteria: age (<50, ≥50 years), history of breast cancer (absent, contralateral, ipsilateral, bilateral), location in the breast, BI-RADS classification, size (≤5 mm, 5-10 mm, 10-20 mm, >20 mm), complete collapse of the lesion at aspiration, aspiration operator (radiologists were considered a single group because of a low number of sampled NPBLs per radiologist), and cystic features on cytologic smears. The NPVs for benign cytologic diagnosis and the PPVs for suspicious and malignant diagnoses were calculated as well.
Statistical Analysis
In primary invasive breast carcinomas, USFNA’s FN and TP rates were compared along with the previously cited clinical, breast imaging, and pathologic characteristics. For standard univariate analyses, continuous variables were evaluated by the Student t test or Wilcoxon-Mann-Whitney test, as appropriate. For categorical variables, χ2 or Fisher exact test was used. Multivariate analysis was performed using a logistic regression model to control for potential confounding variables. Variables with a P value less than .2 were entered into a stepwise selection procedure in a multivariate logistic model.
Study Population Characteristics
For all statistical comparisons, differences were considered statistically significant when the two-tailed P value was below 5%.
The definitive diagnosis was obtained by surgical excision in 1,367 (52.6%) cases, by malignant CNB results in 37 (1.4%) cases, and after 12 months of follow-up in 1,197 (46%) cases (mean [SD] follow-up time, 5.5 [1.9] years; range, 1-10 years). In 49 (3%) cases, the diagnosis was established by consensus: the lesion was determined as malignant in 41 (84%) and as benign in eight (16%). According to the diagnostic reference standard, 1,599 (61.5%) NPBLs were benign and 1,002 (38.5%) were malignant. Among benign lesions, 24 (1.5%) showed atypical epithelial hyperplasia (14 ductal and 10 lobular). Malignant lesions were represented by 797 (79.5%) ductal invasive carcinomas, 136 (13.6%) lobular invasive carcinomas, 11 (1.1%) mixed lobular and ductal carcinomas, eight (0.8%) other (nonductal and nonlobular) invasive carcinomas, 40 (4%) ductal in situ carcinomas, eight (0.8%) lobular in situ carcinomas, and two (0.2%) sarcomas. No lymphomas were found in our series. A total of 619 (61.8%) primary invasive breast carcinomas were identified in our institution’s breast cancer registry.
Among 1,002 definitively malignant NPBLs, cytology yielded concordant malignant results in 780 (77.8%), suspicious results in 148 (14.7%), benign results in 67 (6.7%), and inadequate results in seven (0.7%). Among 1,599 definitively benign lesions, cytology showed concordant benign results in 1,533 (95.8%), suspicious results in 37 (2.3%), malignant results in 14 (0.9%), and inadequate results in 15 (0.9%) cases. The diagnostic accuracy of USFNA was therefore 88.9% (2,313/2,601).
References
Diagnostic Performance of USFNA
Diagnostic performance analysis along the different subgroups is shown in Table 1. The best diagnostic performance in terms of PPV was achieved in BI-RADS categories 4C and 5, whereas the best NPV was achieved in BI-RADS categories 2, 3, and 4A, as well as in patients younger than 50 years and in lesions showing cytologic cystic features or with collapse after aspiration.
Analysis of FN and FP Cases
All 14 FPs (0.9%; 14/1,599 NPBLs) were confirmed by surgical excision and comprised atypical epithelial hyperplasia (two ductal, one lobular), epithelial hyperplasia without atypia (n = 1), pseudo-tumoral epithelial lesions (n = 4, including radial scars and adenosis), papillary lesions (n = 3), fat necrosis (n = 1) and fibroadenomas (n = 2).
FNA was initially developed in the diagnostic workup of palpable breast masses.7 It has been adapted and performed under US guidance for the past 30 years,14 to become an established technique in the workup of breast lesions. Currently, USFNA shows great promise in the molecular workup of breast cancer,32,33 while its role in the morphologic diagnosis of breast lesions is losing ground to CNB.6 Our results show that, contrary to this trend, USFNA can still be a valuable procedure for morphologic analysis of breast lesions: we found sensitivity to be at 92.6% and specificity at 96.8%, both within the upper boundaries of currently published ranges in the literature, from 78.4% to 99.3% and 67% to 100%, respectively Table 4. The improvement over our pervious study published in 199816 highlights the role of experience through learning and a continued, large-volume activity in achieving optimal results. Inadequate rates were significantly lower than in our previous study as well. The rate of inadequate specimens is strongly correlated to the experience of the operator,34 but we think that our multidisciplinary approach lowered these rates further: in situ correlation of imaging features and macroscopic quality of smeared material allowed us to avoid misclassifying as inadequate smears of NPBLs with adipose, fibrous, or cystic features at imaging and expected to yield low-cellularity material.35
Thanks to these consistent and satisfactory results, we use USFNA as a definitive diagnostic procedure in NPBLs to determine further management instead of a first-line screening tool11,36 or only in specific circumstances (axillary lymph node staging, cystic lesions)37 as proposed by other authors. Accordingly, we use the presented four-tier classification system (inadequate, benign, suspicious, and malignant) to communicate clear instructions regarding further management: inadequate results warrant resampling, benign results need no further intervention, suspicious results warrant further sampling by CNB or other techniques, and malignant results can justify excision without further sampling. This classification is also used in other breast cancer centers in France,30 and although it differs from the classification proposed by international guidelines,38,39 our results satisfied the guideline’s recommended diagnostic accuracy criteria.
The analysis of clinical and imaging characteristics influencing USFNA’s diagnostic performance identified patient’s age and the BI-RADS classification as important factors to be considered, both in subgroup analysis and in multivariate comparison of FN and TP results. This information should be useful to individually tailor diagnostic strategies for US sampling of NPBLs, particularly in BI-RADS categories 3 and 4A, where the NPV was highest: USFNA can be an alternative to additional follow-up examinations recommended in BI-RADS category 3 NBPLs or a less invasive alternative to CNB in BI-RADS category 4A NBPLs.2 We also found very low mitotic indexes to be an independent factor in underdiagnosis of primary invasive breast carcinomas, regardless of cancer grade or other histologic tumor characteristics. Although low-grade carcinomas are a known cause of FNA’s FN results,40 the role of very low mitotic indexes in underdiagnosis has not been previously described in the literature. Our results should encourage further research to explain USFNA’s underdiagnosis in very low mitotic index invasive carcinomas and confirm our results. Better understanding of this novel finding might help improve the diagnostic performance of USFNA even further.
In light of these results, clinical and breast imaging data should be integrated and taken into account along with cytologic results to establish a final USFNA diagnosis. Nowadays, the BI-RADS classification is the mainstay of breast imaging diagnosis and the preferred means to report the results. Yet, it conveys limited diagnostic information regarding the whole procedure such as specific breast imaging features (cystic, fibrotic, adipose, high cellularity) and different characteristics at puncture and aspiration (ie, elasticity of the lesion at needling, ease in obtaining cellular material for diagnosis, morphologic changes after aspiration, including partial or complete collapse). Since pathologists performed most procedures, our final USFNA diagnoses take these into account directly. We think integrating these ancillary parameters into the diagnosis partly explains our improved performance of USFNA in the same way as the triple test (integrating clinical, breast imaging, and cytologic data) does for palpable breast lesions41: for example, we found the complete collapse of the lesion to be a very specific finding for benignity. Other authors found improved FNA performance when accounting for lesion elasticity at needling.42 In our opinion, the best way to achieve optimal radio-cytologic correlation is for the pathologist to perform the USFNA procedure, as we do in our institution with the four-hand procedure.
Our results should be taken with caution with regard to some limitations in the study design, since it was a retrospective analysis, with almost one-third of NPBLs excluded from analysis due to the lack of a definitive diagnosis and based on our daily clinical practice. Nevertheless, this high exclusion rate needs to be compared with the large volume of analyzed NPBLs in our series as well. Since we included all patients, particularly some with several lesions, and performed retrospective radio-pathologic correlations from established reports, we could not study if USFNA’s FN cases were due to inappropriate identification of the target at US. Indeed, in a few cases (1.5%), no correlation could be established between the targeted lesion at US and lesions found at pathologic examination. In other cases, radiopathologic correlation might overestimate the FN rate of USFNA. If a benign and a malignant lesion were located nearby, the former might be the only one identified whereas the latter could be undetectable at US. Therefore, even if sampling and smear interpretation were correct, the lesion would be counted as a FN because pathologic examination revealed a cancer. Such cancers would correspond to a limitation of US in the detection of breast cancer rather than USFNA. Yet, not differentiating between US and USFNA’s FN cases is closer to real-life practice, where cancer detection relies on both, and the clinical consequences for the patient of a missed diagnosis of cancer are the same, whether they are due to one or the other technique’s limitations.
As we have shown, our results are in part attributable to a highly trained and dedicated multidisciplinary team, so our conclusions might not be applicable in other settings where breast USFNA might be performed, such as general hospitals or private practices. Nevertheless, we think our study should encourage closer collaboration and improved communication of results between radiologists and pathologists in these other settings to improve the overall performance of USFNA.
In conclusion, our study showed that, given a dedicated, experienced multidisciplinary team, USFNA is a valid procedure in the workup of NPBLs. Under these conditions, USFNA is a particularly interesting procedure in young women, in BI-RADS category 3 lesions when the recommended follow-up is not applicable, or in BI-RADS category 4A lesions as an alternative to CNB. We think that taking into account all parameters involving the procedure (clinical data, imaging presentation, characteristics at aspiration) when establishing a definitive diagnosis improves the diagnostic yield. Studies evaluating the effect of each of these ancillary characteristics on the diagnostic accuracy of USFNA should be encouraged.
In conclusion, our study showed that, given a dedicated, experienced multidisciplinary team, USFNA is a valid procedure in the workup of NPBLs. Under these conditions, USFNA is a particularly interesting procedure in young women, in BI-RADS category 3 lesions when the recommended follow-up is not applicable, or in BI-RADS category 4A lesions as an alternative to CNB. We think that taking into account all parameters involving the procedure (clinical data, imaging presentation, characteristics at aspiration) when establishing a definitive diagnosis improves the diagnostic yield. Studies evaluating the effect of each of these ancillary characteristics on the diagnostic accuracy of USFNA should be encouraged.