Given the fact that human breast cancer depends on HR signaling in regards to response to endocrine therapies, breast cancers have traditionally been sub-classified into HR-positive (or “luminal”) and HR-negative diseases. As identified, even the HR-positive or luminal cancers comprise a spectrum of tumors with varying degrees of proliferation and levels of genetic aberrations. Thus, “luminal type” of HR positive tumors can be further divided into subclass A and B with luminal B being higher grade, having higher proliferation index and a poorer prognosis independent on hormonal therapy. Since a significant correlation between FDG uptake in breast cancer by PET scan and proliferation index has been observed , and the tumor proliferation, as defined by microarray-based gene signatures or Oncotype DX testing (the 21 gene assay, Genomic Health), has been shown to be one of the strongest predictors of outcome for patients with HR-positive disease , it was important to establish whether SUVmax as identified by PET scanning could noninvasively differentiate luminal A from luminal B tumors, and whether SUVmax could predict the outcome of MBC patients with luminal subtypes.
The evaluation of Baseline SUVmax of metastatic sites after relapse may provide important information about tumor proliferation and metabolism that could be of prognostic significance. In this regard, a study of the association of Baseline SUVmax with other well-established prognostic factors could be an important first step towards establishing the relevance of FDG-PET in the prognostic characterization of MBC. Our study found that Baseline SUVmax of MBC was significantly associated only with number of metastatic sites and presence of visceral metastasis. This finding could be the result of accelerated glucose metabolism and related increased metabolic activity of those more aggressive metastatic tumor phenotypes. However, the location of metastatic lesion could also influence the SUVmax. For example, bone metastasis of breast cancer is often osteoblastic and osteoblastic bone metastatic foci usually show low FDG uptake, regardless of biologic behavior of tumor cells. In patients with visceral metastases (± non-visceral metastases) of this study, almost all the SUVmax were obtained from visceral lesions. Only in patients without visceral metastases, the assessment of influence on SUVmax by metastatic locations such as bone, lymph node, skin or soft tissue might be important. However, we did not find any significant differences among these locations (P = 0.235), even between patients with bone involved only and the others (P = 0.063) in terms of median SUVmax.
Our further investigation using the ROC curve indicated that the Baseline SUVmax in the metastatic setting could not differentiate luminal A from luminal B subtypes effectively. This might be because the difference between luminal A and B tumors in proliferation and metabolism was not substantial enough to influence SUVmax. Therefore, the SUVmax cannot be used as a noninvasive indicator to differentiate luminal A from luminal B subtypes. However, it should be noted that the association between Baseline SUVmax and luminal subtypes may be confounded by the fact that relapsed or metastatic lesions may have a different HR or HER2 status from that of the primary tumor [19–24] and that core biopsies in our study were performed only in a small proportion of patients at the time of relapse. Thus, in order to clarify whether SUVmax will have a role in luminal subtypes differentiation, it may be necessary to conduct another large-sample study to correlate SUVmax with core biopsies of the PET scan tested metastatic lesions. However, from 28 patients with core biopsies in our study, only 14.3% had discordant molecular subtypes with the primary lesions (Additional file 1: Figure 1). When Baseline SUVmax was used to differentiate the luminal subtypes after relapse in these patients, the area under ROC curve was 0.551 (SE 0.122) and still indicated no help to differentiate luminal A from luminal B (Additional file 2: Figure 2).
Sorlie et al. demonstrated that breast cancer can be classified into 5 different subtypes according to cDNA molecular profiles, and that these molecular subtypes significantly influence patient’s prognosis . Subsequently, the study of Munoz et al.  showed a significantly unfavorable prognosis for luminal B patients in comparison to those with luminal A subtype in terms of RFI and OS1. Our data confirmed these results. However, in our study, we did not find different luminal subtypes predicting the PFS or OS2 after relapse, a phenomenon which could also be result of transformation of the original HR or HER2 status. Another reason for this could be a suboptimal accuracy in the measurement of our luminal subtypes, with a sensitivity of 77% (95% CI, 0.64-0.87) and a specificity of 78% (95% CI, 0.68-0.87) . Hence, it was necessary to explore new indicator to determine prognosis of the MBC patients with luminal subtypes, and PET scan was of particular attraction due to its non-invasive nature.
Our study is one of the first to show that the Baseline SUVmax of PET scan has significant association with prognosis and outcome of MBC patients with luminal subtypes in terms of PFS and OS2, with the multivariate COX regression analysis confirming the SUVmax an independent prognostic factor.
Although some studies have examined PET/CT imaging as a predictor of treatment response in the primary breast cancer lesion [26–33], significantly less is known about how baseline PET/CT imaging can be used as a prognostic tool by quantifying radiotracer accumulation in metastases. A very recently published study showed that only in patients with newly diagnosed MBC to bone was Baseline SUVmax tertile significantly associated with OS on both univariate analysis (HR = 3.13) and multivariate analysis (HR = 3.19) . However, this study was a retrospective one and did not focus on the patients with luminal subtypes.
Our study was prospectively performed with important findings for luminal type breast cancer patients with newly diagnosed metastases. Baseline SUVmax was found significantly related to the number of metastatic sites and presence of visceral metastasis but could not effectively differentiate patients with luminal A from luminal B subtype. Most importantly, although luminal subtype diagnosed according to the initial tumor sample from the primary surgery could predict the RFI, it could not predict PFS or OS2 after developing relapse or metastases. In contrast, Baseline SUVmax as determined on PET scan was predictive of both PFS and OS. In multivariate analysis using COX regression model, the Baseline SUVmax, RFI, and number of metastatic sites were three independent prognostic factors for PFS. For OS, the significant predictors were only Baseline SUVmax and RFI.
SUV has many drawbacks as it is dependent on parameters such as the delay between injection and measurement, plasma glucose concentration, body weight, instrumental factors and partial volume effect (PVE) . SUVmax is defined as the SUV derived from the single voxel showing the highest uptake within a defined region of interest (ROI) or VOI. In the absence of noise, this SUVmax is indeed the least affected by PVE and so is often considered the best measure of tumor uptake. However, in any real imaging situation, noise is always present, making SUVmax variable. Another drawback of SUVmax is that because it is derived from a single voxel, it may not be an adequate surrogate marker for true tumor biology and it can be heavily influenced by voxel size . Use of the maximum pixel value in a tumor to characterize tumor uptake, however, does make the measurement independent of the observer. This is why, despite its sensitivity to noise and voxel size, the use of SUVmax is still popular.
Several limitations of our study should be addressed. Firstly, as not all MBC patients in our center underwent PET/CT imaging, a selection bias may have played a role in our patients not being representative of general population of MBC cases with luminal disease. Secondly, more than a half of lesions with positive 18 F-FDG uptake were not biopsied, and thus metastatic disease was diagnosed only with imaging and long-time clinical follow-up. Thirdly, not all HER2-positive luminal B patients received trastuzumab, which may partly influence applicability of our results to the HER2-positive cases who will have trastuzumab therapy. Lastly, we examined PET/CT imaging from only 1 time-point and, thus, are unable to comment on the predictive effect of PET/CT imaging with regard to treatment effect.
In spite of these limitations, our study remains the first to establish the role of PET scanning as a noninvasive outcome indicator of luminal A versus luminal B MBC subtypes.