Follow-up evaluation

All patients were followed up from the day of surgery. Recurrence was defined as any unequivocal occurrence of new cancer foci in a disease-free patient. The site of the first cancer recurrence and the interval between the surgery and recurrence were determined. The duration of DFS was calculated as the interval between the date of surgery and that of the first confirmation of cancer recurrence, death without evidence of recurrence, or the last clinical contact attesting to disease-free status.

Statistical analysis

Data are presented as numbers (%) or means unless otherwise stated. Frequencies were compared using the χ² test for categorical variables, and Fisher's exact test was applied to small samples. Continuous variables were compared using the Mann–Whitney U test. Receiver operating characteristic (ROC) curves of SUV_max for the prediction of recurrence were generated to determine the cut-off value that yielded optimal sensitivity and specificity from Youden Index. The patient population was subdivided on the basis of a cut-off value for SUV_max derived from ROC curves, and the duration of DFS was analyzed using the Kaplan–Meier method. Differences in DFS were assessed using the log-rank test. The potential independent effects of SUV_max on DFS were assessed by multivariate analyses using the Cox proportional hazards model (backward stepwise selection) and p < 0.05 was considered statistically significant. Data were statistically analyzed using SPSS software (version 10.5, SPSS Inc., Chicago, IL, USA).

Results

Patient and tumor characteristics

Table 1 shows the clinicopathological characteristics of the 311 patients. A total of 261 (83.9%) tumors were ER-positive and 236 (75.9%) were PgR-positive according to IHC, and 50 (16.1%) were HER-2-positive according to IHC or FISH. Patients in whom enhanced computed tomography or FDG-PET/CT uncovered no evidence of distant metastatic spread were eligible for primary breast cancer surgery. Sixty (19.3%) patients had received neoadjuvant chemotherapy and 111 (35.7%) patients had received adjuvant chemotherapy. 213 (68.5%) patients had received radiotherapy to the breast or chest wall. All patients with hormone receptor-positive breast cancer had received adjuvant hormonal therapy.

The ROC curve showed an optimal SUV_max cut-off value of 3.8 for predicting recurrence (n = 311, area under the curve [AUC], 0.789; sensitivity, 83.3%; specificity, 68.2%; Fig. 1).

At a cut-off value of 3.8, a high SUV_max significantly associated with large tumor, lymph node positivity, nuclear grade III, lymphovascular invasion positivity, hormone receptor negativity and HER2 positivity (Table 2).

Fig. 2 shows a significant difference in DFS between patients with SUV_max ≤ 3.8 and >3.8 (3-year DFS rates: 98.8% vs. 91.6%; p < 0.001). The variables included in the univariate analysis of DFS in the 311 patients were SUV_max, tumor size, nodal status, nuclear grade, lymphovascular invasion status, hormone receptor status, HER2 status, and adjuvant treatment (Table 3). High SUV_max, large tumors and negative hormone receptor status were significantly associated with a short DFS. Moreover, multivariate analysis that included SUV_max, tumor size, nodal status, nuclear grade, as well as the status of lymphovascular invasion, hormone receptors, HER2, and adjuvant treatment showed that SUV_max (hazard ratio: 1.18, 95% CI: 1.07–1.30; p = 0.001) and negative hormone receptor status (hazard ratio: 6.02, 95% CI: 1.72–21.1; p = 0.005) were independent prognostic factors for DFS (Table 3).

The patients were categorized into four subtypes based upon hormone receptor and HER2 status. We identified 12 recurrent events in all patients, of which 6 (50%) occurred in patients with the triple-negative breast cancer subtype (hormone receptor-negative and HER2-negative). Patients with this subtype had early relapse and a poor prognosis. We then evaluated the SUV_max value as a prognostic factor in this subtype. The SUV_max ranged from 0.8 to 22.1, and the mean SUV_max in the triple-negative subtype was 6.32, which was considerably higher than of the value of 3.75 for all patients (Fig. 3). An analysis of ROC curves revealed an optimal SUV_max cut-off value of 8.6 for predicting recurrence of the triple-negative subtype (n = 30; AUC, 0.767; sensitivity, 66.7%; specificity, 80.0%; Fig. 4).

At a cut-off value of 8.6, a high SUV_max significantly associated with large tumors (Table 4). Fig. 5 shows a significant difference in DFS between patients who had SUV_max ≤ 8.6 and >8.6 (3-year DFS rates: 90.9% vs. 42.9%; p = 0.002).

Univariate analysis of DFS in the 30 patients with the triple-negative subtype included SUV_max, tumor size, nodal status, nuclear grade, and lymphovascular invasion status as variables (Table 5). A high SUV_max was significantly associated with a short DFS. Moreover, multivariate analysis showed that SUV_max (hazard ratio: 1.27, 95% CI: 1.00–1.27; p = 0.047) was an independent prognostic factor for DFS (Table 5).

Discussion

A high SUV_max significantly associated with poor prognostic features in the present study and thus could serve as an independent prognostic factor for patients with operable breast cancer, especially those with the triple-negative subtype.

Some studies have suggested that SUV_max in the primary tumor might be associated with several pathological prognostic factors^{19, 20, 21, 22, 23 and 24}; however, most of these studies did not analyze survival. Oshida and colleagues identified FDG uptake as an independent prognostic factor for the relapse-free survival of patients with breast cancer, along with the number of positive lymph nodes and histological grade.²⁷ They used the differential absorption ratio (DAR) instead of the SUV_max, as an index of FDG uptake. Without using ROC curves, they calculated statistically significant differences at a cut-off point that was set to 1.0 and then increased the value in stepwise increments of 0.5 up to 5.0 between the two groups using the log-rank test. The difference in both overall survival and relapse-free survival rates was the most significant at a cut-off of 3.0. The present findings that poor prognostic features (large tumor size, positive lymph node metastasis, tumor grade III, positive lymphovascular invasion, hormone receptor negativity, and HER2 positivity) were closely associated with our cut-off SUV_max of 3.8 were similar to those of Oshida and colleagues. A high SUV_max was an independent and poor prognostic factor in the multivariate analysis. Further validation studies are required to evaluate the appropriateness of our cut-off values.

We discovered that SUV_max can discriminate two prognoses for patients with the triple-negative subtypes. The mean SUV_max was higher in these patients than in the total number of patients, and a high SUV_max was an independent poor prognostic factor in multivariate analysis. Although triple-negative breast cancer comprises only 15–20% of all breast cancers, the lack of targeted therapy for this subgroup poses therapeutic challenges.²⁸ The ability of platinum chemotherapy alone or in combination with PARP-1 inhibitors to treat triple-negative breast cancer has been investigated in new trials.^29 and 30 Triple-negative patients with a high SUV_max were predicted to have early relapse and to require new aggressive treatment strategies. Julia Tchou et al. found an association between SUV_max and Ki67 in patients with triple-negative breast cancer and suggested that PET could be used to monitor the treatment response of such patients.³¹

The threshold for recommending chemotherapy for patients with hormone receptor-positive and the HER2-negative subtype is particularly difficult to define. The St. Gallen international expert consensus on the primary therapy of early breast cancer 2009 adopted Ki67 as a proliferation index to indirectly support the rationale for augmenting chemotherapy with endocrine therapy in such patients.³² The main disadvantage of Ki67 is the high degree of interobserver variability in assessments.³³ Ki67 LI values can vary as a function of several critical factors, including human error, Ki67 selection of the tumor areas to be counted and the specific antibody used.³⁴ SUV_max has similar issues associated with assessment uniformity. However, Nakayama et al. suggested that phantom studies of lung adenocarcinoma can overcome the difficulties in multicenter studies using PET for lung adenocarcinoma.³⁵ SUV_max could substitute for Ki67 as an index of adding chemotherapy. The limitations of our study include a small sample cohort, its retrospective design at a single institution, and the short follow-up period that did not allow adequate assessment of the hormone receptor-positive and HER2 negative-subtype. In spite of the relatively small sample and the small number of events, SUV_max was found to be a strong independent prognostic factor for DFS. Further studies are required to evaluate whether SUV_max has prognostic value in each breast cancer subtype. We are currently investigating in the multi-institutional prospective study.

In conclusion, we demonstrated that high SUV_max in a primary site was an independent prognostic factor for operable breast cancer. PET/CT is a valuable tool to provide information on initial staging and on determining the need for more aggressive therapy to prevent early relapse especially in patients with breast cancer who have triple-negative subtype. SUV_max in a primary site might be used for the new clinical trial about adjuvant chemotherapy such as dose dense chemotherapy or adding new drugs. Conversely, patients with low SUV_max had a good prognosis and might be omitted adjuvant chemotherapy, even with the triple-negative subtype. However, using PET/CT for the purpose stated in our research has not been investigated in cost-effectiveness studies.

Conflict of interest statement

None of the authors have any conflict of interest.

References