key: cord-1036311-m52i1sj5 authors: HL, Chung; J, Sun; GJ, Whitman; JWT, Leung; LP, Middleton; HT, Le-Petross title: Ultrasound Features to Differentiate COVID-19 Vaccine–Induced Benign Adenopathy from Breast Cancer Related Malignant Adenopathy date: 2022-02-21 journal: Acad Radiol DOI: 10.1016/j.acra.2022.02.015 sha: e344bb42658cbed86d1190fb99dcc1a46c1a56db doc_id: 1036311 cord_uid: m52i1sj5 OBJECTIVE: : To identify nodal features used to distinguish COVID-19 vaccine-Induced benign reactive adenopathy from malignant adenopathy. MATERIALS AND METHODS: : This IRB-approved, single-institution, retrospective study compared features of 77 consecutive patients with benign adenopathy secondary to a mRNA COVID-19 vaccine with 76 patients with biopsy-proven malignant adenopathy from breast cancer. Patient demographics and nodal features were compared between the two groups using univariate and multivariate logistic regression models. A receiver operating characteristic analysis with the maximum value of Youden's index was performed for the cutoff value of cortical thickness for predicting nodal status. RESULTS: : The mean cortical thickness was 5.1 mm +/- 2.8 mm among benign nodes and 8.9 mm +/- 4.5 mm among malignant nodes (p < 0.001). A cortical thickness ≥ 3.0 mm had a sensitivity of 100% and a specificity of 21% (AUC = 0.60, 95% CI: 0.52-0.68). When the cutoff for cortical thickness was increased to ≥ 5.4 mm, the sensitivity decreased to 74%, while the specificity increased to 69% (AUC = 0.77, 95% CI: 0.70-0.84). CONCLUSION: : Cortical thickness and morphology are predictive of malignancy. Cortical thickness cutoff of ≥ 5.4 mm demonstrates higher specificity and improved accuracy for detecting malignant adenopathy than a cutoff of ≥ 3.0 mm. As worldwide vaccination against the SARS-CoV-2 virus was initiated at the end of 2020, radiologists around the world have noted a marked increase in the incidence of unilateral axillary adenopathy [1, 2] . The changes in lymph node features after vaccination, reported in a study of 91 volunteer employees, include an increase in the lymph node size by 5.3 mm, increase in cortical thickness by 2.1 mm, increase in the number of visible lymph nodes by 1.8, and an overall increase in the associated Doppler signal [3] . Faermann et al reported that the mean cortical thickness of lymph nodes after vaccination was 5 mm ± 2 mm and that the longitudinal length of lymph nodes increased between the first and the second doses of a mRNA COVID-19 vaccine [4] . As reports of vaccine-induced reactive adenopathy increased, management recommendations were developed and modified regarding the need for imaging and biopsy [5] [6] [7] [8] . The differential diagnosis for unilateral axillary adenopathy is broad and ranges from benign etiologies such as vaccine-induced reactive adenopathy to metastatic adenopathy from breast carcinoma or other malignancies [9] . There is an overlap in the ultrasound appearance of benign and malignant nodes, making differentiation of benign versus malignant difficult. Objectively differentiating benign from malignant adenopathy is of clinical value and can reduce costs and anxiety related to unnecessary biopsies and/or the need for follow-up examinations. Normal-appearing lymph nodes have a reniform shape with an imperceptible or thin cortex and a visible central hilum, while suspicious or malignant lymph nodes demonstrate variable degrees of cortical thickening [10] . Currently, a cortical thickness of ≥ 3.0 mm is commonly accepted as the cutoff threshold for differentiating benign from malignant adenopathy [11] [12] [13] [14] [15] . The primary aim of this retrospective study was to identify ultrasound features that may be used to differentiate COVID-19 vaccine-induced benign reactive adenopathy from malignant adenopathy secondary to breast cancer. Accountability Act-complaint retrospective study was performed at a National Cancer Institutedesignated comprehensive cancer center. The need for informed consent was waived. The electronic health records (EHR) were queried for all patients with an axillary ultrasound (AUS) examination describing ipsilateral adenopathy after the receipt of a mRNA COVID-19 vaccine from December 2020 to September 2021. We excluded patients < 18 years old, patients without an AUS examination performed at our center, and patients with axillary adenopathy on the ipsilateral side of a newly diagnosed breast cancer. Imaging studies, pathology results, and vaccine records were reviewed. Ultrasound lymph node features were collected from the AUS examinations, including the number of visible suspicious lymph nodes, the maximum longitudinal lymph node size, the nodal cortical thickness, and visibility of the hilum. A retrospective review was performed to classify the lymph nodes into one of six morphologic types: type 1, no visible cortex; type 2, thin (< 3.0 mm) cortex; type 3, diffuse, uniform cortical thickening (≥ 3.0 mm); type 4, generalized lobulated cortical thickening; type 5, focal cortical lobulation; or type 6, totally hypoechoic node with no visible hilum [10] . A comparable number of breast cancer patients with malignant adenopathy from the same institution who had available pre-treatment AUS examinations, percutaneous lymph node biopsy results, and subsequent axillary lymph node dissection confirmation were randomly selected as the comparison malignant cohort. No special selection criteria was used to select the malignant cases based on other factors such as age, histology, or racial background, though we excluded patients with N3 nodal stage or distant metastatic disease. Retrospective data collection of the breast cancer cases with malignant adenopathy was performed to identify the characteristics of the malignant group according to histology, molecular subtype, tumor (T) and nodal (N) classification and is shown in Table 1 .The benign cohort was independently reviewed by one breast imaging fellowship trained radiologist (HC) with 20 years of practice experience who was blinded to the AUS findings of the malignant cases. The AUS findings of the malignant cohort were independently reviewed by a second breast imaging fellowship trained radiologist (HL) with 21 years of practice experience who was blinded to the AUS findings of the vaccineinduced reactive adenopathy cases. The primary reviewer for each cohort obtained consensus from two additional radiologists (GW and JL) with 23 plus years of practice experience in a small percentage of cases. Patient demographics and lymph node features were summarized using frequencies, percentages, means, standard deviations, medians, minima, and maxima. These findings were compared between the benign and malignant groups using the Wilcoxon Rank Sum test, Fisher's exact test, and Pearson Chi-squared test. The distributions of cortical thickness among the nodal morphologic types were compared between benign and malignant nodes using box plots, and the associations were tested using the Kruskal-Wallis test. A receiver operating characteristics analysis with the maximum value of Youden's index was performed to select the cutoff value for cortical thickness in predicting benign versus malignant nodal status. Univariate analyses were used to select unadjusted significant predictors, which then underwent multivariate logistic regression model analyses to assess associations, adjusted for the effects of covariates. Backward model selection by the Akaike information criterion was used to select the final model. Statistical significance was defined as p < 0.05. Statistical analysis was carried out using R (version 3.6.3, R Development Core Team). A total of 153 patients with axillary adenopathy were identified for analysis, with 77 patients with benign reactive adenopathy after the receipt of a mRNA COVID-19 vaccine and 76 patients with biopsy-proven malignant adenopathy from breast cancer. The median age of the benign reactive adenopathy cohort was 59 years (range, 30-84 years) while the median age for the malignant adenopathy group was 55 years (range, 28-87 years). After the final multivariate analysis, age was not a significant difference between the two groups (p = 0.9). With regard to race, the pairwise comparisons of the different racial backgrounds and the p-values adjusted using false discovery rate method showed no statistically significant racial differences (p=0.163). The demographic characteristics of both groups are shown in Table 2 The number of visible suspicious lymph nodes, the lymph node size, the nodal cortical thickness, and the morphologic classification were compared between the vaccine and the breast cancer groups. Patients with fewer than three suspicious nodes detected on AUS had a higher probability of malignancy compared to patients with three or more suspicious nodes (OR = 2.7, 95% CI: 1.1-6.7, p = 0.029). For lymph node size, the mean maximum longitudinal length was 19.7 mm (standard deviation (SD), 8.0) for the benign reactive nodes and 21.1 mm (SD, 8 .0) for the malignant nodes. There was no statistically significant difference in the size of the lymph nodes between the two groups (p = 0.221). When the nodal cortical thickness was compared between the two groups, the mean cortical thickness was 5.1 mm (SD, 2.8) for the benign nodes and 8.9 mm (SD, 4.5) for the malignant nodes (p < 0.001). The commonly accepted threshold for abnormal cortical thickening is 3.0 mm. In our study, when a threshold cortical thickness ≥ 3.0 mm was selected as abnormal or suspicious for malignancy, the sensitivity for malignant nodal involvement was 100% and the specificity was 21% (AUC = 0.60, 95% CI: 0.52-0.68). When the cortical thickness threshold was increased to ≥ 5.4 mm, the sensitivity decreased to 74% while the specificity increased to 69% (AUC = 0.77, 95% CI: 0.70-0.84; Figure 1 ). The degree of cortical thickening was directly correlated with the nodal morphologic type (r2 = 0.57, OR = 6.2, 95% CI: 3.1-12.7, p < 0.001, Figure 2 ). When the cortical thickness was reviewed along with the nodal morphologic classification for all 153 patients, the mean cortical thickness was 4.8 mm among type 3 nodes, 7.1 mm among type 4 nodes, 10.3 mm among type 5 nodes, and 14.5 mm among type 6 nodes. Univariate analysis indicated younger age, greater cortical thickness, and the absence of a visible hilum as predictors of malignancy. After final multivariate analysis, higher nodal morphologic type classification (types 4, 5 or 6) and fewer than three suspicious lymph nodes detected on AUS remained as significant predictive features of malignant adenopathy. For example, a morphologic type 4 node with generalized lobulated cortical thickening had a 7.5 times higher probability of being malignant compared to a type 3 node with diffuse, uniform cortical thickening (OR = 7.5, 95% CI: 3.1-19.4, p < 0.001). A morphologic type 5 node with focal cortical lobulation had a 123.0 times higher probability of being malignant compared to a type 3 node (OR = 123.0, 95% CI: 21.7-2,340, p < 0.001) ( Table 3) . Patients with fewer than three suspicious nodes had a higher probability of malignancy, compared to patients with three or more nodes (OR = 2.7, 95% CI: 1.1-6.7, p = 0.029). The number of suspicious lymph nodes and the nodal morphologic type were unrelated. In the current environment of primary and booster COVID-19 vaccine recommendations, radiologists' encounters with vaccine-induced, benign reactive adenopathy are expected to continue. While the current accepted threshold for abnormal cortical thickening is ≥ 3.0 mm, our data shows that a cortical thickness cutoff of ≥ 5.4 mm would allow for an improvement in AUS specificity (69% versus 21%), albeit with some loss in sensitivity (74% versus 100%). The thicker cortical cutoff threshold of 5.4 mm also demonstrates improved accuracy compared to 3.0 mm (77% versus 60%) in differentiating benign from malignant lymph nodes. In the absence of suspicious ipsilateral breast imaging findings and confirmation of vaccination on the side of adenopathy, a cortical thickness < 5.4 mm may be assessed as benign, especially if the nodal morphology shows only diffuse, uniform cortical thickening. The morphologic classification previously described by Bedi et al is not universally used in clinical practice [10] . This classification allows for standardized reporting and categorization of lymph nodes. In our study population, the cortical thickness and its effect on the morphology was the single most important differentiating feature of benign versus malignant lymph nodes. While benign reactive lymph nodes were most often morphologic type 3 ( Figure 3 ) and some type 4 (Figure 4 ), our results showed that morphologic type 4 and type 5 ( Figure 5 ) nodes had significantly higher probabilities of being malignant. In fact, a morphologic type 5 node was only observed in a single case of vaccine-induced, benign reactive adenopathy ( Figure 6 ). There were no patients with a morphologic type 6 node among the benign cohort but 15 of 76 (20%) patients in the malignant group had morphologic type 6 nodes. The effacement of the hilum of a node was only observed among the malignant nodes and not in the benign nodes. While a lymph node with cortical thickening ≥ 3.0 mm has indeterminate potential for malignancy, there is a significantly higher likelihood of malignancy if the cortical thickening is great enough to cause the lymph node to develop lobulated cortical thickening (Figure 7) . Our data suggests that the higher morphologic types are more likely to be seen with malignant adenopathy rather than with benign reactive adenopathy from COVID-19 vaccines. The observation that patients with fewer (< 3) suspicious nodes had a higher probability of malignant adenopathy compared to patients with more (≥ 3) suspicious nodes should be interpreted with caution. The malignant group was comprised of clinical N1 and N2 breast cancer patients and excluded locally advanced and inflammatory breast cancer patients, which may account for the fewer number of suspicious lymph nodes reported among the malignant group. Since ultrasound is an operator dependent imaging modality, it is possible that not all suspicious lymph nodes were documented at the time of image acquisition. Therefore, rather than interpreting this observation to mean that fewer suspicious lymph nodes implies an association with malignancy, our findings suggest that vaccine-induced benign adenopathy may manifest with many suspicious nodes. While the precise mechanism of this manifestation is unknown at this time, it is possible that the immune response triggered by a mRNA vaccine may be more systemic, rather than localized to the injection site alone. Potential inter-reader differences between the two radiologists who independently reviewed either the benign or malignant adenopathy group are expected, however we suspect these differences to be minimal without invalidation of our findings as cortical measurements and its effect on the hilum have been previously described in many prior publications. The AUS nodal features used in this study has previously undergone meticulous comparison of AUS findings with pathology from surgery [10] . Other limitations include the small sample size and the lack of follow-up imaging in some patients within the vaccine cohort due to the changes and evolving recommendations on the management of COVID-19 vaccine-induced axillary adenopathy. Earlier patients were more likely to receive a follow-up or biopsy recommendation, however, later during the pandemic period, publications outlined scenarios in which it may be appropriate to do less [6] [7] [8] . In February 2021, Lehman et al [6] suggested that unilateral axillary adenopathy on the ipsilateral side of recent vaccination may be assessed as benign so long as there were no other imaging findings. Later in June 2021, the same group suggested that isolated unilateral axillary lymphadenopathy on the ipsilateral side of the vaccination was akin to the ACR BI-RADS recommendations for unilateral axillary adenopathy due to an inflammatory cause. In the context of a benign or negative clinical follow-up, the authors suggested that imaging follow-up was not necessary and follow-up imaging should be reserved for cases of contralateral adenopathy or persistence of adenopathy beyond 6 weeks. Should adenopathy persist at follow-up imaging, tissue sampling may then be indicated [7] . Similar to these recommendations, a multidisciplinary panel of experts from three leading tertiary cancer centers proposed additional guidelines for post vaccination imaging. This multi-institutional group suggested that for patients in whom adenopathy was overwhelmingly most likely due to vaccination, expectant management without default follow-up imaging was appropriate with allowance for follow up imaging and/or obtainment of tissue diagnosis based on individual risk, particularly in patients with an ipsilateral breast or other malignancies such as head and neck, melanoma, or lymphoma that may present with malignant adenopathy [8] . Due to the novelty and frequency of vaccine-induced adenopathy, radiologists were likely self-regulating and In this study comparing the ultrasound lymph node features of benign reactive adenopathy from COVID-19 vaccines to biopsy proven malignant adenopathy from breast cancer, cortical thickness and its resultant morphology were the most predictive features helpful in differentiating a node as benign or malignant. We further quantified cortical thickness Longitudinal AUS demonstrates a node with generalized lobulated cortical thickening, with a maximum cortical thickness of 4.5 mm (solid arrows). Fine needle aspirate biopsy confirmed a benign reactive node. A 41-year-old woman with malignant metastatic adenopathy due to breast cancer. Longitudinal AUS demonstrates a type 5 node measuring 37 mm in long axis, with focal cortical lobulation. At the focal cortical thickening, the cortex measures 15 mm (arrows) while the remainder of the node shows a thin cortex (arrowheads). Fine needle aspiration biopsy performed on the focal hypoechoic lobulation confirmed metastatic adenopathy from breast cancer. Unilateral axillary Adenopathy in the setting of COVID-19 vaccine Management of Unilateral Axillary Lymphadenopathy Detected on breast MRI in the era of coronavirus disease (COVID-19) vaccination Unilateral axillary adenopathy induced by COVID-19 vaccine: US follow-up evaluation COVID-19 vaccination induced lymphadenopathy in a specialized breast imaging clinic in Israel: Analysis of 163 cases SBI Recommendations for the Management of Axillary Adenopathy in Patients with Recent COVID-19 Vaccination Mitigating the impact of coronavirus disease (COVID-19) vaccinations on patients undergoing breast imaging examinations: A pragmatic approach Unilateral Lymphadenopathy After COVID-19 Vaccination: A Practical Management Plan for Radiologists Across Specialties Multidisciplinary recommendations regarding post-vaccine adenopathy and radiologic imaging: Radiology scientific expert panel Imaging findings of variable axillary mass and axillary lymphadenopathy Cortical morphologic features of axillary lymph nodes as a predictor of metastasis in breast cancer: in vitro sonographic study Assessment of Ultrasound Features Predicting Axillary Nodal Metastasis in Breast Cancer: The Impact of Cortical Thickness Axillary ultrasound for breast cancer staging: an attempt to identify clinical/histopathological factors impacting diagnostic performance Accuracy of individual descriptors and grading of nodal involvement by axillary ultrasound in patients of breast cancer Axillary ultrasound in patients with clinically nodenegative breast cancer: which features are predictive of disease? Preoperative sonographic classification of axillary lymph nodes in patients with breast cancer: node-to-node correlation with surgical histology and sentinel node biopsy results Editorial support was provided by Bryan Tutt, Scientific Editor, Research Medical Library.