Study Cohort
Among a pool of 135 axillary FNAs performed on LNs with cortical thickness ³3 mm, 84 nodes in 84 patients were metastatic at surgical pathology, while 51 nodes in 50 patients (including the patient with bilateral synchronous breast cancer) were negative (Figure 1). Patients with metastatic LNs were younger (mean age 52.8 years ± 13.9) compared to patients with negative LNs (59.9 years ± 12.6; p=0.012) (Table 1). Patients with metastatic LNs were significantly more likely to receive neoadjuvant therapy (85.7%) compared to patients with negative LNs (52.9%). There was no significant difference in self-reported race, laterality, or breast tumor histopathology. FNAs performed on primary tumor histopathology of ILC and IDLC were more likely to be metastatic, while FNAs performed on primary tumor histopathology of DCIS/microinvasion were more likely to be negative, although this difference was not significant (Table 1).
FNA accuracy
Out of 135 FNAs included in our study, 133 (98.5%) were diagnostic and 2 (1.5%) were non-diagnostic at cytological analysis, despite 3 FNA passes performed (2 passes with a 25G needle, and 1 pass with a 20G needle) in both non-diagnostic cases. Of the two patients with non-diagnostic FNAs, one subsequently underwent a negative core needle biopsy of the questioned LN, and the other did not undergo additional percutaneous sampling prior to surgery. Both patients had a negative sentinel lymph node biopsy at the time of surgery.
In the metastatic node pool, 12 out of 84 FNAs (14.3%) returned false negative cytology with subsequent positive surgical pathology at the time of excision (Figure 1). Imaging features and FNA technical factors in the false-negative cytology cases are summarized in Table A.1.
Clinical and histologic characteristics
Among the clinical characteristics studied, post-menopausal status (p=0.005) and clinically detected palpable LNs (p=0.015) were significantly associated with axillary nodal metastasis on FNA or final surgical pathology at univariable analysis (Table 2). Other factors, including prior history of breast cancer, genetic mutation, breast tumor size and location, did not show a statistical difference between metastatic and negative nodes at univariable analysis. Histologic characteristics, including breast tumor grade (p=0.08) and receptor status (p=0.15), were not significant predictors of nodal metastasis at univariable analysis.
At multivariable analysis, age was removed for being highly colinear with menopausal status. While postmenopausal status and clinically detected LNs were no longer significant in this analysis, higher breast tumor grade (grade 2, when compared to grade 1) had an increased likelihood for axillary LN metastasis (odds ratio 4.71, p=0.05) (Table 3).
Cortical thickness
Axillary LN cortical thickness was significantly higher in metastatic compared to negative nodes (8.7±6.4 mm vs. 4.9±1.9 mm; p<0.001), as displayed in Table 4. A common language effect size statistical analysis revealed a 0.78 probability (95% CI 0.68, 0.85) that any random patient with axillary metastasis would have a larger cortical thickness than a random draw from patients without axillary metastasis.
The PPV and corresponding 95% CI for axillary nodal metastasis were calculated for varying cutoffs of LN cortical thickness and displayed in a step-plot (Figure 2). The PPV of axillary LN cortical thickness ³3 mm and ³3.5 mm was similar at 0.62 [95% CI 0.53, 0.70] and 0.63 [0.54, 0.72] respectively. Compared to LN cortical thickness ³3 mm, the PPV increased to 0.67 [0.57, 0.76] for cortical thickness ³4 mm and 0.74 [0.64, 0.83] for cortical thickness ³4.25 mm.
Other imaging characteristics
At univariable analysis, significant predictors of nodal metastasis on imaging included diffuse, rather than eccentric, cortical thickening (p<0.001), abnormal hilum (p<0.001), round nodal shape (p<0.001) and indistinct nodal margin (p=0.003) (Table 4). At multivariable analysis, only diffuse cortical thickening (p=0.038) and abnormal hilum (p=0.016) remained significant predictors of metastasis (Table 3).