Prediction of Advanced Axillary Lymph Node Metastases (ypN2-3) Using Breast MR imaging and PET/CT after Neoadjuvant Chemotherapy in Invasive Ductal Carcinoma Patients

We aimed to investigate the value of breast magnetic resonance (MR) imaging and positron emission tomography-computed tomography (PET/CT) in predicting advanced axillary lymph node (ALN) metastases (ypN2-3) after neoadjuvant chemotherapy (NAC) in invasive ductal carcinoma patients. A total of 108 patients with invasive ductal carcinoma underwent breast MR imaging and PET/CT both before and after NAC (termed initial staging and restaging, respectively). The number of positive ALNs and the short diameter (SD) of the largest ALN on breast MR imaging and maximal standardized uptake value (SUVmax) in the ALNs on PET/CT were evaluated. Odds ratio (OR) for prediction of advanced ALN metastases was calculated. The negative predictive value (NPV) of restaging imaging for exclusion of advanced ALN metastases was also calculated. Patients with advanced ALN metastases were more likely to have a higher number (≥2) of positive LNs (OR, 8.06; P = 0.015) on restaging MR imaging. No clinico-pathological factors were significantly associated with advanced ALN metastases. With restaging MR imaging, PET/CT, and MR imaging plus PET/CT, the NPV for excluding advanced ALN metastases was 97.3%, 94.4%, and 100.0%. A higher number of positive ALNs on restaging MR imaging was an independent predictor for advanced ALN metastases after NAC.

PET/CT for whole-body scanning was performed after patients had fasted for at least 6 hours using the following PET/CT systems: Discovery 600 and Discovery 690 (GE Healthcare, Milwaukee, WI, USA). Approximately 3.7-5.6 MBq of 18 F-fluorodeoxyglucose (18F-FDG) per kilogram of body weight was injected intravenously (IV), and the patients were advised to rest for 1 hour before image acquisition. For attenuation correction, a low-dose CT scan was obtained before the PET scan without contrast enhancement from the base of the skull vertex to the knee. The PET scan was performed with a maximal space resolution of 5.1 mm (Discovery 600) and 4.9 mm (Discovery 690) at 1.5 minute per bed position.
Image Analysis. For breast MR imaging, two breast-imaging radiologists (H.J.K. and W.H.K., with 18 years and 10 years of experience, respectively) determined the number of positive ALNs and measured the short diameter (SD) of the largest ALN from the ipsilateral breast in the images. The ALN was considered to be positive when one or more findings were noted as follows: cortical thickness >3 mm, existence of eccentric cortical thickening, loss of fatty hilum, and round or lobulated nodal shape 10,[13][14][15][16] . The ALNs ipsilateral to the breast cancer were compared to the contralateral axillary LNs. If the ALNs did not show any significant difference in the above characteristics, the ALNs were considered negative, which is in line with previous studies 15,17 .
For PET/CT, one nuclear medicine physician (S.W.L. with 13 years of experience) and one breast-imaging radiologist (W.H.K.) reviewed the images in consensus. Regions of interest (ROI) were manually placed over the area of highest activity on slices of the ALNs, and the maximal standardized uptake value (SUVmax) within the ROI was obtained. Especially in the ALNs of low 18 F-FDG uptake (SUVmax <1), the ROI was carefully drawn over the corresponding ALNs on CT images. The SUVmax was calculated using the volume viewer software on a GE Advantage Workstation 4.6 (GE Healthcare) with the following formulas: SUVmax = maximum activity in ROI (MBq/g)/[injected dose (MBq)/body weight (g)]. For both breast MR imaging and PET/CT, all reviewers knew that the patients had invasive breast cancer; however, no further information, including pathological nodal stages, was given to the reviewers. Pathological Evaluation. All patients underwent axillary surgery with SLNB and/or ALND, and the final pathological nodal stage was determined based on the histopathological results of surgical specimens. ALNs were examined using hematoxylin and eosin staining. Each ALN was classified as negative or positive for metastases; SCieNtifiC REPoRTs | (2018) 8:3181 | DOI:10.1038/s41598-018-21554-z the total numbers of ALNs sampled and the number with metastases were recorded. All histopathological evaluation was performed by a pathologist (J.Y.P. with 18 years of experience in breast pathology).

Data Collection and Statistical Analysis.
Clinical data included age, as well as clinical tumor (T), and clinical lymph nodes (N) stages. The following pathological information from percutaneous biopsy results obtained before NAC were included: histological grade, estrogen receptor (ER), progesterone receptor (PR), and HER2 status. Clinical T and N staging was performed at initial diagnosis on the basis of the 7th American Joint Committee on Cancer 18 . The expression of ER, PR, and HER2 was assessed by immunohistochemical staining. The expression of ER and PR was quantified using the Allred score, considering a total Allred score >2 as positive for ER or PR 19 . A HER2 score of 0 or 1 was considered negative (HER2-negative), a value of 3 was considered positive (HER2-positive), and value of 2 was considered equivocal. For equivocal cases, silver-enhanced in situ hybridization (SISH) was performed and an HER2/CEP17 ratio ≥2 or an HER2/CEP17 ratio <2 with an average HER2 copy number ≥6 were considered positive (HER2-positive) 20 . Hormone receptor (HR)-positive status was defined as tumors expressing ER and/or PR.
The optimal cut-off values of the number of positive ALNs or SD of the largest ALN on breast MR imaging, and the nodal SUVmax on PET/CT were independently calculated from receiver operating characteristic (ROC) analysis, using the best Youden index (sensitivity + specificity − 1) for the prediction of residual or advanced ALN metastases. Then, patients were divided into binary groups according to each cut-off value. Diagnostic performances of restaging MR imaging or PET/CT, including overall diagnostic accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), were estimated for each modality, and a combination of both modalities (restaging MR imaging plus PET/CT) for the prediction of advanced ALN metastases. For restaging MR imaging, positive was defined either as having either the number of positive ALNs or as having the SD of the largest ALN above the cut-off value. For restaging MR imaging plus PET/CT, positive was defined when either modality was above the cut-off value. The overall diagnostic accuracy was estimated by calculating the area under the ROC curve (AUC) with 95% confidence interval (CI). Sensitivity was compared using the McNemar's test, and the comparison of the AUC was performed with the method of Delong et al. 21 .
The clinico-pathological and imaging findings of patients with and without residual or advanced ALN metastases were compared using the independent t test or chi-square test as appropriate. The OR (odds ratio) and 95% CI for residual or advanced ALN metastases were calculated with univariate logistic regression analysis, and variables with P < 0.10 were selected for the final multivariate model. All statistical analyses were performed with the statistical software SPSS version 24.0 (Chicago, IL, USA) and MedCalc version 17.1 (Mariakerke, Belgium). Two-tailed P values of less than 0.05 were considered statistically significant.
Ethical Standards. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Due to our retrospective review of the data, and the requirement of an informed consent was waived after approval of institutional review board.

Discussion
There has been a growing interest in the prediction of the axillary tumor burden after NAC in recent years. Previous studies have suggested the axillary ultrasound as a modality for evaluating the nodal response after NAC 10,22 . Breast MR imaging is considered as the most accurate imaging modality for providing anatomical information and allows the inspection of the axilla and breast 17,23 . PET/CT as a modality to image tumor viability based on increased glycolysis also enables the localization of abnormal ALNs showing metabolic changes with anatomical information from CT 24,25 . Both modalities offer advantages over ultrasound for the visualization of the whole area of the axilla irrespective of the characteristics of the patient (e.g., obesity) or the experience of breast imager. Our study demonstrated that, in addition to evaluation of the primary tumor, the axillary findings of breast MR imaging and PET/CT can be helpful in predicting nodal status after NAC treatment. For the diagnosis of residual ALN metastases, including both low and high burden of axillary tumor, our study demonstrated that positive HR status and higher (>0.9) nodal SUVmax on restaging PET/CT were significant predictors after NAC, but not the findings on breast MR imaging. It is well known that HR-positive breast cancers are less chemosensitive than HR-negative breast cancers, resulting in a low burden of axillary tumor even with negative imaging [26][27][28] . Using univariate analysis, You et al. also reported that HR-positive breast cancers were associated with a higher rate of incorrect diagnosis by ultrasound and breast MR imaging for nodal response after NAC 14 . Taken together, these results indicate that restaging MR imaging is limited in predicting the nodal response in patients with low burden of axillary tumor or HR-positive breast cancer after NAC.
However, we found that patients had a higher likelihood of advanced ALN metastases when they had higher (≥2) numbers of positive ALNs on restaging MR imaging after NAC. This finding suggests the potential utility of breast MR imaging at restaging after NAC for predicting advanced ALN metastases in patients with breast cancer. In previous studies, the authors have demonstrated several clinico-pathological factors associated with the nodal  Table 3. Imaging and Clinico-Pathological Factors between Patients with and without Advanced ALN metastases after Neoadjuvant Chemotherapy. HR = hormone receptor; HER2 = human epidermal growth factor receptor 2; ALN = axillary lymph node; SUVmax = maximal standardized uptake value. Data are numbers of patients, with percentages in parentheses unless otherwise indicated by an asterisk. *Data are mean ± standard deviation. **The positive axillary LN was considered when one or more findings were noted as follows: cortical thickness>3 mm, eccentric cortical thickening, loss of fatty hilum, and round or lobulated nodal shape at MR imaging.   should be interpreted with caution in light of the intrinsic dependency of NPV on the prevalence of advanced ALN metastases. For the determination of optimum cut-off values, we used ROC analysis and the current study found that the cut-off number (≥2) of positive ALNs on restaging MR for diagnosis of advanced ALN metastases is fewer than previous studies (≥4, analogous to the number of pathological nodal staging) 14,15 . It may be due to the fact that many metastatic ALNs can be normal-looking, especially when NAC treatment has been applied and therefore, the tumor burden of each LN is decreased 12 . Likewise, we found the optimum cut-off value for SUVmax to be 1.1, which is relatively low, perhaps due to the suppressed tumor glycolysis related to NAC treatment, occurring even in the presence of viable tumor cell 31 . We believe that cut-off values obtained from our ROC analyses can be used in predicting nodal status, thereby guiding less invasive axillary surgery.
Even though statistical significance was not obtained in the improvement of the diagnostic performance of combined restaging MR imaging and PET/CT compared with the use of either restaging MR imaging or PET/ CT alone, we found that restaging MR imaging plus PET/CT improved the sensitivity compared with the use of either modality alone. Despite the slight decrease in specificity with restaging MR imaging plus PET/CT together, the overall diagnostic performance was increased. These observations may advocate the use of combined PET/ MR imaging to evaluate axillary tumor burden. Taneja et al. reported a higher diagnostic accuracy of PET/MR in axillary nodal assessment compared with either MR imaging or PET alone 32 . In patients with cervical cancer and non-small cell lung cancer, improved detection of nodal metastases with PET/MR imaging was also reported 33,34 .
Several limitations of this study are worth noting. This was a retrospective study performed at a single institution. Although breast MR imaging and PET/CT for initial staging and restaging in consecutive patients with breast cancer undergoing NAC were routinely performed during this period, we did not control for a possible selection bias of patients who underwent breast MR imaging and PET/CT. Second, complete coverage of whole axillary region (level I-III) might not be possible with breast MR imaging, as the previous study have reported 5 . In our study, complete coverage of the field of view (FOV) for level I and level II area was generally achieved in all patients; however, artifacts or incomplete coverage was present for level III axillary area in some patients. Thus, a further study with a prospective design and a larger number of patients is needed. Third, as the cortex of ALNs usually became thinner after NAC, a partial volume effect that can cause underestimation in the observed SUVmax might have been introduced. Fourth, due to our retrospective design, the impact on decision for the type of axillary surgery was not direct or clear. Another caveat for this study is that not all patients underwent ALND for exact determination of nodal status. Although approximately 80% of the patients underwent ALND and 96% of patients with SLNB had ALNs negative for metastases, some patients still had advanced ALN metastases even with only one or two metastatic sentinel ALNs or with skipped metastases with negative sentinel ALNs.
To conclude, a higher (≥2) number of positive ALNs on restaging MR imaging was an independent predictor for advanced ALN metastases after NAC. The NPV for excluding advanced ALN metastases was high as 97.6%, 94.4%, and 100.0% for restaging MR imaging, PET/CT, and MR imaging plus PET/CT, respectively. Therefore, breast MR imaging and PET/CT at restaging have the potential to serve as tools to guide axillary surgery for a less-invasive approach in patients with invasive ductal carcinoma who are undergoing NAC.