Relation between primary tumor FDG avidity and site of first distant metastasis in patients with breast cancer: Erratum

[This corrects the article DOI: 10.1097/MD.0000000000004266.].


Introduction
Breast cancer is the most common cancer in women and the leading cause of cancer-related mortality. [1] Dissemination of breast cancer to other organs may be detected at initial presentation or may occur as metastatic relapse years after patients receive curative surgery. [1] Distant metastasis, rather than primary breast tumor, is the main cause of patient death. Accordingly, the presence and site of breast cancer metastasis is of crucial importance to the clinical course and outcome of patients. [2][3][4] In breast cancer, loco-regional spread is associated with substantially more favorable outcome compared with distant metastasis. Furthermore, even with distant metastasis, prognosis is significantly influenced by the site of first spread. Among preferential sites for breast cancer dissemination, metastases to the liver or lung have been shown to lead to poorer patient outcome compared with the bone. [5,6] Metastasis occurs by cancer cell migration to target tissue and their subsequent survival and proliferation. The "seed" and "soil" hypothesis proposes that the distribution of metastatic spread is not a matter of chance but rather related to the biological compatibility between migrating tumor cells and receiving organ. [6] Site-specific metastasis is thus critically influenced by the molecular characteristics of cancer cells originating from primary tumor and their interaction with target tissue microenvironment. In breast cancer, tumor subtype has been implicated to influence sites of metastasis, [7] and higher tumor grade has been shown to be associated with spread to sites of more adverse prognosis such as lung or liver. [8] However, tumor histology offers only weak outlooks on the pattern of metastasis. More recent studies are focusing on gene-expression signatures that are linked with risk of dissemination or with sitespecific metastasis. [9,10] However, any requirement for tissue sampling suffers from significant intratumor heterogeneity of Author's contributions: CHL designed the study, carried out the data collection and data interpretation, and drafted the manuscript. SHM and YSC participated in imaging processing and statistical analysis. JEL and YHI participated in verification of the pathology results and clinical information. YSC and BTK participated in data collection and helped with data analysis. KHL designed the study, and coordinated data interpretation and manuscript preparation. All authors read and approved the final manuscript. malignant as well as nonmalignant cells that comprise the tumor. [11] Routine gene expression analysis may also be restricted by its availability, cost, and limited reproducibility. [12] In breast cancer, there is also controversy regarding subjectivity of subtype classification [13] and optimal cutoffs for receptor expression. [14] Positron emission tomography/computed tomography (PET/ CT) is widely used for evaluating malignant tumors, [15] where 18 F-fluorodeoxyglucose (FDG) uptake serves as a marker for heightened glycolytic metabolism. In breast cancer patients, tumor FDG uptake level has been shown to correlate with tumor subtype, treatment response, and patient outcome. [16] Unlike histopathologic tests, PET/CT imaging provides quantitative measurement of FDG uptake over the entire tumor volume in a simple, noninvasive, and reproducible manner. Moreover, PET/ CT is often performed in cancer patients at initial presentation for tumor staging or evaluating treatment response. A previous study observed higher FDG uptake in breast cancers that developed distant metastasis compared with those that did not metastasize, [17] suggesting a link between primary tumor glycolytic activity and risk of metastasis. Given that target tissue microenvironment is likely to place specific metabolic demands to invading cells, the glycolytic phenotype of primary breast tumors may also influence its preferred site of metastasis. Uncovering such relations could improve our understanding of the mechanisms of sites-specific tumor spread and help identify new therapeutic targets.
In this study, we thus investigated whether primary tumor FDG avidity is associated with the first site of distant spread in advanced breast cancer. This was tested in patients who already had metastasis at the time of initial diagnosis as well as those who developed metastatic relapse following curative resection.

Study population
Study subjects were selected from 1415 patients who underwent palliative chemotherapy for advanced breast cancer at our institute. Among these, 299 patients who underwent FDG PET/ CT at the time of initial diagnosis were included as study candidates. As this study was focused on distant metastasis, we excluded 17 cases with only regional chest wall or lymph node (LN) involvement. We also excluded 8 patients who had concurrent (n = 3) or previous malignancies (n = 5). Thus, a total of 274 subjects were finally included for analysis. This retrospective study was approved by Samsung Medical Center Institutional Review Board and the requirement for written consents was waived.
The study subjects were categorized either as patients who had metastatic disease at the time of initial presentation (n = 193), or those who developed metastatic relapse during follow-up after curative resection (n = 71). There were an additional 10 cases that did not show metastasis at initial diagnosis, but displayed metastatic disease after neoadjuvant chemotherapy. These cases were not included in the analysis but were assessed separately.

FDG PET/CT imaging
All patients fasted for at least 6 hours and had blood glucose levels of <200 mg at the time of PET/CT. Imaging was performed 60 minutes after injection of 5 MBq/kg FDG without intravenous or oral contrast on a Discovery LS or a Discovery STe PET/CT scanner. Whole-body CT was performed with a continuous spiral technique with an 8-slice helical CT (140 keV; 40-120 mA; Discovery LS) or a continuous spiral technique with 16-slice helical CT (140 keV; 30-170 mA; Discovery STe). After the CT scan, an emission scan was obtained from head to thigh for 4 minutes per frame in 2D mode with reconstruction of attenuation-corrected PET images (4.3 Â 4.3 Â 3.9 mm) using an ordered-subset expectation maximization algorithm (28 subsets, 2 iterations; Discovery LS), or for 2.5 minutes per frame in 3D mode with reconstruction of attenuation-corrected PET images (3.9 Â 3.9 Â 3.3 mm) using a 3D ordered-subset expectation maximization algorithm (20 subsets, 2 iterations; Discovery STe).

PET image analysis
On transaxial PET images, a circular region of interest was placed on breast tumors, and maximum standardized uptake values (SUVmax) of FDG uptake were measured. The threshold for FDG avidity was based on the mean SUVmax among subjects with metastatic relapse (SUVmax = 10.1). This was because any influence of metastatic tumor on primary tumor FDG uptake could be excluded in this group. This threshold for FDG avidity was also applied to patients with metastasis at presentation.

Characterization of primary tumors and metastatic lesions
Clinical information including tumor histology and metastatic sites was obtained from medical records. Tumor subtype was histologically determined. Tumors expressing both estrogen and progesterone receptors were considered hormone receptor (HR) positive. Tumor size was based on MRI, but CT, ultrasound or mammographic measurements were used if MRI was not available. Whole breast involvement was treated as a 10 cm-sized tumor.
Metastatic sites were determined by pathology or conclusive imaging findings. In 1 patient, liver metastasis was based on strong clinical evidence that led to additional treatment. Involvement of LNs other than axillary, internal mammary, or supraclavicular were considered distant metastasis. When more than 1 metastatic lesion was detected at the same time, it was not possible to determine which lesion occurred first, and all of the lesions were treated as the first metastatic site.

Statistical analysis
Significance of difference in characteristics according to tumor FDG avidity was assessed by Student's t tests, Pearson x 2 tests, or Fisher's exact tests. Difference in frequency of metastasis to the specific sites was analyzed by x 2 or Fisher's exact tests. SPSS software for Windows was used for statistical analysis. P values < 0.05 were considered statistically significant.

Characteristics of study subjects
The clinical characteristics of the "metastasis at presentation" group (n = 193) are summarized in Table 1. These patients had a mean age of 52.7 ± 11.0 years, primary tumor size of 6.3 ± 3.4 cm, and HR-positive rate of 59.6%. The primary tumor had an average SUVmax of 10.6 ± 5.4, and 98/193 tumors (50.8%) were FDG avid (SUVmax ≥10.1). Patients with FDG-avid primary tumors were slightly younger than those with nonavid tumors. FDG-avid primary tumors were slightly larger than nonavid tumors, but there was no significant difference in tumor subtype or histology according to FDG avidity. The clinical characteristics of the "metastatic relapse" group (n = 71) are summarized in Table 2. These patients had a mean age of 48.9 ± 9.7 years; primary tumor size of 5.9 ± 2.9 cm; and HR positive rate of 38.0%. The primary tumor had an average SUVmax of 10.1 ± 4.7, and 31/71 tumors (43.7%) were FDG avid (SUVmax ≥10.1). Patients with FDG-avid and nonavid primary tumors showed no significant difference in age, tumor size, histology, stage, or premetastasis treatment.
In addition to the 2 groups described above, there were 10 cases that showed no evidence of metastasis at presentation but revealed metastatic disease after neoadjuvant chemotherapy. The primary tumor was FDG-avid in 8 cases and nonavid in 2 cases. Only 1 tumor was HR positive. The metastatic site was visceral organ in 7 cases, distant LNs in 2 cases, and bone in 1 case.

The "metastasis at presentation" group
In this group of 193 patients, 114 had a single site of metastasis at presentation (59.1%), 68 had 2 sites (35.2%), and 11 had ≥3 sites (5.7%). In subjects with ≥2 metastatic lesions at presentation, it could not be determined which occurred first, and all were treated as first sites of metastasis. The most common first site of metastasis was the bone, which was present in 121 subjects (62.7%). This was followed in frequency by the lung (30.1%), liver (25.4%), and distant LNs (18.7%).
In this group, patients with FDG-avid primary tumors were significantly more likely to have first dissemination to the lung compared with those with nonavid tumors (Table 3). There was no difference in frequency of first metastasis to the bone according to primary tumor FDG avidity (Table 3). FDG PET/CT findings of a representative case with FDG-avid primary breast cancer and lung metastasis at initial presentation are illustrated in Fig. 1.
Unlike FDG avidity, patients with primary tumors who were HR positive were significantly more likely to have first metastasis to the bone (68.7% vs 53.9%) and significantly less likely to have Table 1 Characteristics of breast cancer patients with "metastasis at presentation" categorized according to primary tumor FDG avidity.  Patients among this group with FDG-avid primary tumors were significantly more likely to have first metastasis to the lung or liver compared with those with nonavid tumors (Table 4). There was no difference in frequency of first metastasis to the bone according to primary tumor FDG avidity (Table 4). FDG PET/CT findings of a representative case with FDG nonavid primary breast cancer that later recurred as bone metastasis, and a case with FDG-avid primary breast cancer that later recurred as liver metastasis are illustrated in Figs. 2 and 3, respectively. Table 3 First organ of metastasis in the "metastasis at presentation" group according to primary tumor FDG avidity or hormone receptor status.  Patients among this group with HR-positive primary tumors were significantly more likely to have first metastasis to the bone (55.6% vs 27.3%), compared with those with HR-negative tumors (Table 4). First metastasis to distant LNs had a trend for higher frequency in patients with HR-negative tumors (43.2% vs 22.2%, P = 0.07; Table 4). When we evaluated the risk of metastasis to specific organs according to a combination of FDG avidity and HR status, those with FDG-avid but HR-negative primary tumors where significantly more likely to have lung metastasis (11/19; 57.9%) compared with the remaining patients (14/52; 26.9%; P = 0.016).

Discussion
In this study, we tested the hypothesis that primary tumor FDG avidity is associated with the preferred site of first distant spread in advanced breast cancer. Results in patients with metastatic disease at presentation revealed that first spread to the lung was Table 4 First organ of metastasis in the "metastatic relapse" group according to primary tumor FDG avidity or hormone receptor status.  www.md-journal.com significantly more frequent in patients with FDG-avid compared with nonavid primary tumors. Furthermore, in subjects with metastatic relapse following curative surgery, first spread to the lung and liver was significantly more frequent in patients with FDG-avid primary tumors. In both of our study populations, the most frequent distant site of first metastasis was the bone. When the 2 populations were compared, patients with metastatic disease at presentation had higher incidence of first metastasis to the bone, as well as the liver, compared with patients with metastatic relapse. This may be because there were more involved sites at the time of metastasis detection in the former group compared with the latter group. The strong preference of breast cancer to spread to the skeleton is well recognized. [3][4][5] As site-specific metastasis occurs through interaction of tumor cells with tissue microenvironment, dominant dissemination to the bone indicates a favorable microenvironment for breast cancer cell colonization and growth. We found that the preference for bone metastasis was not influenced by primary tumor FDG avidity in either study groups. The bone marrow provides a rich micro-metastatic niche that is composed of myeloid-derived cells, adaptive immune cells, chemokines, and various stromal components. [18] Our finding thus suggests that the favorable microenvironment of the bone allows both FDG-avid and nonavid breast cancer cells to spread and thrive equally well.
As with metastasis to the bone, risk of first metastasis to distant LNs and brain was also uninfluenced by primary tumor FDG avidity for both study groups. Distant LNs as the first site of metastasis were relatively common in our subjects, particularly considering their relatively small volume compared with major organs. Lymphatic spread is the most common route of metastasis for carcinomas, and subsequently, LNs are a frequent site for tumor cell transport and growth. Furthermore, lymphatic drainage into LNs of tumor-derived factors such as growth factors and cytokines may help make this niche receptive of migrating tumor cells. [19] The brain may also provide an environment rich in cytokines and growth factors derived by astrocytes, once tumor cells manage to pass the blood-brain barrier. [20] Taken together, our findings suggest that favorable microenvironments of the bone, distant LNs, and brain may allow successful survival and proliferation of breast cancer cells with low as well as high glycolytic activity.
In contrast to organs with favorable local milieu, lung and liver tissues may be considered less favorable for tumor cells to evolve into metastatic colonies. Indeed, we found that first metastasis to the lung occurred significantly less often in both groups of patients when primary tumors were FDG nonavid compared with when they were FDG avid. This suggests the possibility that the inherent metabolic property of breast cancer cells may be a significant factor in determining the likelihood for metastasis to the lung. One possible link between glucose metabolism and metastatic characteristics of tumor cells is hypoxia-inducible factor 1 (HIF-1). [21,22] High expression of HIF-1 in breast cancers has previously been shown to promote malignant progression and pulmonary metastasis. [23,24] Cancer cells that migrate to the lung, where oxygen is relatively plentiful, are not likely to activate HIF-1 expression unless it is already driven by tumor phenotype. Given the strong association between HIF-1 and heightened glycolytic flux, the greater incidence of lung metastasis in patients with FDG-avid primary tumors in our study could reflect higher HIF-1 expression. A previous clinical study demonstrating higher FDG uptake in metastatic pulmonary cancers with high HIF-1 expression supports this possibility. [25] However, such an association is only speculative, and further investigations will be required to clarify this issue.
In our study, first metastasis to the liver in patients with metastatic relapse was also significantly less frequent when primary tumor was FDG nonavid. The liver is the third most common site for metastasis of breast cancer, and major breakthroughs have recently been achieved in understanding this process. [26] As with pulmonary metastasis, HIF-1-regulated genes have been implicated in hepatic metastasis of breast cancers, supporting the possibility that the liver milieu may also favor growth of breast cancer cells with HIF-1 activation. Increased tumor FDG avidity may thus reflect cancer cell phenotypes that are better fit to thrive and proliferate in less favorable microenvironments. Although the reason why the association was weaker in patients with metastasis at presentation is not clear, it may be partly attributed to influence of metastatic tumor masses on primary tumor FDG uptake for this group. In addition, because 33 of the 49 liver metastases detected at diagnosis were accompanied by other metastatic sites, it is possible that at least in some of these cases, spread to the liver actually followed that of a different first metastatic site in this group.
We further compared the influence of primary tumor HR status on first metastatic sites. As a result, patients with HRpositive primary tumors had significantly greater frequency of first metastasis to the bone in both study groups. These results are consistent with previous reports that HR status can affect pattern of spread of breast cancer. [27,28] For instance, patients with HR-positive breast cancer are known to be more likely to have bone metastasis. [7,[29][30][31][32] In our results, patients with HRpositive primary tumors had lower frequency of metastasis to distant LNs. Although previous studies have indicated a significant correlation between HR-negative status and the presence of positive axillary LNs, [33] the relation between HR status and metastasis to distant LNs remains unclear. Divergent from bone or LN metastasis, the frequency of metastasis to other visceral organs in our study was not significantly influenced by HR status. Hence, we next evaluated whether combining the information of FDG avidity and HR status may help further stratify risk for site-specific metastasis. The results in patients with metastasis at presentation showed greater risk of first metastasis to the lung in subjects with primary tumors that are FDG avid but HR negative. This suggests the possibility that combining primary tumor FDG avidity and HR status may provide additional useful information.
In our study, preoperative PET/CT in breast cancer patients undergoing curative surgery discriminated FDG-avid from nonavid primary tumors. Among patients who later developed metastatic recurrence, those who had FDG-avid primary tumors turned out to have significantly greater likelihood of first relapse to the lung and liver compared to their counterparts. Therefore, although the level of breast cancer FDG uptake is not likely to guide the best mode of treatment, the presence of high uptake could guide clinicians to be more guarded against possible recurrence to the lung or liver following surgery. Furthermore, in patients with metastatic disease at presentation, recognition of the preferred sites for first metastasis according to primary tumor FDG avidity may help detect distant metastasis that might otherwise be missed. Hence, our findings suggest that assessment of primary breast cancer FDG avidity with pretreatment PET/CT offers information that may be helpful for patient management.
Limitations of our study include its retrospective design, which could have caused bias during selection and exclusion of study subjects. Also, a larger number of study subjects might have increased the statistic power to identify additional sites with significant difference between groups. Furthermore, all metastatic lesions present at the time of diagnosis were treated as "first metastatic sites" even though some may have occurred later. However, this problem appears unavoidable in retrospective studies as our own. Finally, the optimum SUVmax threshold for primary tumor FDG avidity may require further confirmation in larger study populations.

Conclusions
In patients with breast cancer, the level of primary tumor FDG uptake is associated with different preferences for site of first dissemination either at initial presentation or during post-surgery follow-up. The favorable milieu of the bone appears to allow seeding of tumor cells regardless of their FDG avidity. In contrast, the favored spread of FDG-avid tumors to the lung and liver indicate that breast tumor cells with heightened glycolytic activity may be better fit to survive and grow in these microenvironments. Further investigations will be required to assess the precise role of tumor FDG avidity as a surrogate marker of breast cancer cell biology that is relevant to the site of first metastasis.