*2.4. Patient-Based Analysis*

The patient-based outcomes consisted of determining the stage of disease at baseline and at the end of follow-up. The stage of disease was determined at clinical presentation together with conventional imaging and subsequently by FDG PET together with pathological confirmation.

#### *2.5. Lesional Analysis: Qualitative and Semi-Quantitative FDG PET Readings*

Conventional imaging was performed ≤5 weeks before or after FDG PET. Clinically relevant lesions suspicious for malignancy on any imaging modality were included in this analysis, with a maximum of 5 largest lesions per tissue type in the case of distant metastases. The included lesions were either pathologically confirmed as benign or malignant (group A) or, in the case of absent/inconclusive pathology, verified by additional imaging and/or follow up for 18 months after primary diagnosis (group B). Based on these data, we classified lesions as true positives (=malignant lesions suspect on FDG PET), true negatives (=benign lesions not suspect on FDG PET), false positives (=benign lesions suspect on FDG PET) and false negatives (=malignant lesions not suspect on FDG PET). In the case of multiple axillary lymph nodes on FDG PET, only those which were pathologically proven to be malignant were included in group A.

Quantitative analysis was performed, using in-house developed software (version 04092018, Accurate tool, R. Boellaard, Amsterdam, The Netherlands) [27]. This analysis only included lesions visible on FDG PET. Volumes of interest (VOIs) were semiautomatically defined using 50% thresholds of peak standardized uptake values (SUVpeak) adapted for local background<sup>26</sup> and verified by radiologists. For each VOI, we determined the SUVmax, SUVmean, SUVpeak and total lesion glycolysis (TLG). In addition, for primary breast lesions, VOIs were manually defined on the low-dose CT scans to calculate anatomical volumes. The correlation between these FDG PET parameters and various histopathological features of the primary tumor was assessed to investigate whether histopathological features predicted the accuracy of the FDG PET.

#### *2.6. Clinical Implications on Treatment Plan*

We investigated the impact of incorrect lesion identification by FDG PET. The pathological outcome of the surgically resected axillary lymph nodes was retrospectively compared to lymph nodes identified on FDG PET, excluding patients that had progressive disease during neo-adjuvant therapy. We postulated that without progression, any additional pathologically verified malignant node in the resection specimen, compared to baseline FDG PET, should be classified as false negative on FDG PET. In the case that this would lead to stage migration of the N-stage, from N1 to N2, such patients would require an axillary lymph node dissection, according to current guidelines instead of sentinel node/marked node resection [3].

Similarly, the number of distant metastases is relevant for the treatment plan. In the case of oligometastatic disease (<4 lesions), local treatment with curative intent can be considered, whereas in the case of extensive disease ( ≥4 distant metastases) a palliative option will prevail. We compared the distant lesions on FDG PET to the number of distant metastases confirmed through pathological verification and/or additional imaging at baseline and during the follow-up period.
