**4. Discussion**

To our knowledge, this is the first study assessing the diagnostic performance of FDG PET in patients with stage IIB/III or LRR, grade 1–2, ER+ breast cancer. In this study, the sensitivity of FDG PET for disease staging was 77.1%. Previous studies have reported a sensitivity of up to 100% for primary breast cancer [15,28] and 81–97% for restaging of LRR [15], for all types of breast cancer combined. In a meta-analysis performed by Han et al. [29], it could be seen that FDG PET outcomes led to changes in staging in 25% of patients. In addition, various studies have shown that FDG PET outcomes affected the

treatment plan in 6.5–18% of patients with primary breast cancer [11,15,29]. These data reinforce the importance of additional imaging modalities next to the conventional imaging to obtain the correct stage, which is essential for an adequate treatment plan. In our case, the treatment plan was correctly adapted by FDG PET in 7/70 (10%) patients, but in 16/70 (23%) patients FDG PET would have led to an incorrect treatment plan (Table 2). Thus, our results support the hypothesis that FDG PET is insufficient for (re)staging of grade 1-2 ER+ breast cancer.

#### *4.1. TNM Lesion Detection*

When looking into more detail to detection of individual lesions, this study shows that the sensitivity and specificity of FDG PET for lesion detection (pre-operatively) was 94.3% and 48.8% (group A)/83.6% and 20.5% (group B), respectively in patients with grade 1–2, ER+ BC. Differentiation of lesion detection based on the type of lesion (i.e., primary breast lesions, locoregional lymph nodes and distant metastases), showed that FDG PET accurately detects primary breast tumors (83/87 (95.4%). Our data are in line with a prospective study that showed similar detection rate of BC lesions when comparing FDG PET/CT with MRI (95% vs. 100%, *p* = 1.0) [30]. However, compared to other conventional imaging techniques, such as MRI, it is known that FDG PET has less sensitivity and less accuracy for determining the size of the tumor and to assess the presence of multifocal disease [15]. For locoregional lymph nodes, previous studies have shown that micrometastases are suboptimally detected with FDG PET (CT) [31,32]. However, in current clinical practice, it is essential to identify all affected nodes before neo-adjuvant treatment, as only extensively affected axillary lymph nodes (i.e., ≥N2-disease/'bulky' disease) remaining after neoadjuvant systemic treatment will, in general, require axillary lymph node dissection. In the case of N1-disease (1– ≤ 3 affected lymph nodes) at diagnosis and response to neo-adjuvant treatment, resection of the sentinel node(s) and marked node is deemed sufficient when followed by locoregional radiotherapy [3]. In our study, 26/96 (27.1%) axillary lymph nodes were incorrectly identified: 3.1% of the axillary lymph nodes were identified as false negatives and 24.0% as false positive nodes (Table S3). These incorrect identified nodes could potentially change the N-stage and eventually the locoregional treatment, making it even more important that these nodes are correctly identified.

In the case of distant metastases, FDG PET(CT) is known to have a high yield as shown in inflammatory and stage II/III BC [6,15,33–35]. In this study, distant metastases were identified in 7/70 patients (10%), which is at the lower end of what would be expected from the literature for stages IIB/III/LRR [5–7]. In 4 patients, FDG PET confirmed the suspicion of metastases as seen on conventional imaging, and in 3 patients, metastatic lesions were correctly identified on FDG PET alone. However, FDG PET also missed lung and bone metastases in 7 patients. Distant metastases were mainly located in extra-axillary lymph nodes, the lungs and bone. FDG PET lacks sensitivity for the detection of (small) lung nodules (due to partial volume effect and respiratory movement) and identifies osteoblastic lesions suboptimally (often showing low or no FDG uptake in these lesions) [15,36]. In our study, most of the lung lesions were small (range: 4–11 mm) and therefore correct identification of these might have been hampered by the partial volume effect; however, the low grade, ER+ breast cancer subtype might also have played a role. Most of the bone lesions included in this study were osteolytic and also for these lesions, applied that the specified low grade, ER+ breast cancer subtype might have affected its identification on PET.

#### *4.2. Association between FDG PET Parameters and Histopathology*

Quantification of FDG uptake only showed a trend for higher SUVmax and TLG values in malignant (true positive) lesions, compared to false positives and false negatives. However, no specific threshold for malignancy could be determined, as was described in other studies [37,38]. The histological subtype, however, correlates with FDG uptake, with ductal BC having higher FDG uptake, compared to lobular BC. This is in accordance with

other studies and can probably be explained by a lower tumor cell density, a low level of GLUT1 expression, diffuse infiltration of surrounding tissue and a decreased proliferation rate in lobular BC, eventually resulting in lower FDG uptake [17,19,20].

We did not observe a difference in FDG uptake between grade 1 and grade 2 tumors. In the literature, it is known that grade 3 tumors have significantly higher FDG uptake than grade 1–2 tumors, but no information is available regarding the correlation between FDG uptake and grade 1 and 2 tumors, separately [12–18]. Regarding the receptor status, we found that the % ER positivity and HER2 status correlated with TLG. No correlations could be found between the PR status and FDG uptake. Previous studies are somewhat contradictory about this: a few studies have shown that there is no correlation between the hormone receptor status (positive or negative) and FDG uptake [14,39], whereas others have shown that FDG uptake is affected by hormone receptor status [13,15–17,20–22]. These studies do not take the expression levels of ER and PR, separately and in combination, into account, which may be essential to identify the relation with FDG uptake. For HER2, no correlation could be found between its status (positive/negative) and FDG SUV, which is consistent with other studies [16,17,22]. However, the HER2 status did correlate with TLG, but we could not confirm this from other studies, as they did not include TLG in their analyses.

We did not find a correlation between the FDG uptake and the mitotic activity index (mean ± standard deviation: 3.1 ( ±4.1); range: 0–19), probably as we only included tumors, which are expected to be less metabolically active than other subtypes of breast cancer. Studies including more metabolically active tumors, identified by a higher Ki-67 expression, have reported higher FDG uptake [17,40].
