Predictive Value of Dual-phase ¹⁸F-FDG PET/CT in the Assessment of Neoadjuvant Chemotherapy Response in Patients with Locally Advanced Breast Cancer: A prospective Comparative Study with Dynamic Contrast-enhanced Magnetic Resonance Imaging

Introduction

Methods

Study Cohort
We included patients diagnosed with LABC and planned to receive NAC. Patients Stage IIB, IIIA, IIIB, or IIIC diseases were included according to the American Joint Committee on Cancer 7^th edition.[] Patients were scanned with ¹⁸F-FDG PET/CT and DCE-MRI before treatment (baseline), after 2-3 cycles of NAC (interim), and after the end of treatment, before surgery.

We did not have patients with dose infiltration, suboptimal image quality, and a feature that would prevent PET/CT or MRI. Breast cancer diagnosis in all patients was confirmed histopathologically from biopsy materials. We recorded the size of the residual tumor from the pathology results of patients who underwent a mastectomy after NAC. We accepted the absence of invasive tumor in the surgical specimen as a complete pathological response, including carcinoma in situ.[-] We grouped the patients as those with a complete pathological response (PCR) or residual tumor according to the results of the histopathological evaluation (non-PCR).

Imaging Protocol
DCE MRI and ¹⁸F-FDG PET/CT have been performed sequentially within 3 days (0-6 days).

Whole-Body ¹⁸F-FDG PET/CT Imaging
A dedicated PET/CT scanner (GE Medical Systems Discovery ST PET/CT scanner, LLC 3000 N, Grandview Blvd, Waukesha, Wisconsin, USA) was used for ¹⁸F-FDG PET/CT imaging. All patients were requested to fast for at least 4-6 h before the PET/CT examination, and their blood glucose levels were ≤180 mg/dl before the ¹⁸F-FDG injection. Patients were scanned from the skull base to the mid-thigh in the supine position, at six to seven-bed positions (3 min per bed position) with a 128×128 matrix. Iterative image processing was applied to the images (2 iterations, 21 subsets). A lowdose CT scan (4-slice, 120 kV, 300 mA) was obtained for attenuation correction and anatomic localization.

Late Prone Imaging
We produced a dense sponge material coil for PET/CT prone imaging based on the breast MRI unit's breast coil. We have optimized its dimensions so that the patient is not trapped in the PET/CT gantry. Late prone images were obtained using that breast coil. In baseline PET/CT late prone images, FDG uptake time was a median of 142 min (99-191 min), and in interim late prone images, FDG uptake time was 126.5 min (88-199 min).

MRI
MRI was performed on a 1.5 Tesla (General Electric) device using an 8-channel breast coil. Pre-contrast axial T1W (3 mm), axial (3 mm), and sagittal (4 mm) T2 fat-suppressed STIR sequences were obtained in the prone position. After intravenous administration of gadolinium contrast agent (0.5 mmol/kg), 6 times T1W (3 mm) fat-suppressed gradient echo dynamic sequences in the axial plane were obtained. After the extraction images were obtained, AngioMap and 3D reconstructed images were obtained using Computer- Aided Diagnosis (CADstream) software.

Data Analysis
Data Analysis in 18F-FDG PET/CT
Visual evaluation
Two nuclear medicine physicians with over 20 years of expertise and a research assistant evaluated the images at the AW-46 workstation with a consensus. We recorded the localization and primary tumor focus and excluded patients with distant metastases.

Semi-quantitative Analysis
FDG PET/CT whole-body and late prone imaging
We calculated the tumor's SUVmax, SUVmean, and SULpeak values. Metabolic volumes (MTVs) were measured with VOI. To measure MTV, we used two different evaluation methods: The volume of the lesion measured using the threshold value of 42% of the SUVmax (MTV- 42) and the volume of the metabolically active part of the tumor visually (MTV-adp).[] We calculated total lesion glycolysis (TLG) using SUVmean (bw)×MTVs formula. The percentage change of all measured numerical parameters after 2-3 cycles was calculated according to the following formula (% change=value after 2-3 cycles of chemotherapy-baseline/baseline value×100).

Data Analysis in MRI
Two experienced breast radiologists with over 20 years of expertise performed the visual and semi-quantitative analysis of the MRI images. Subtraction image was obtained by subtracting the images obtained before and after intravenous gadolinium. Contrast areas were detected. Tumor size, volume, and time-contrast curves were obtained using AngioMap and 3D reconstructed images using CADstream software.

Parameters Measured by MRI
(1) Number of tumors, (2) three dimensions, (3) spherical volume (SV=length × height × thickness×0.52), (4) perfusion volume (angiovolume [AV]), (5) time-contrast curves. a-Type 1-2-3 contrast enhancement percentages. b-Percentages of "rapid" and "medium" at the beginning of the contrast curve. c-PSI (PSI, maximum value of the contrast curve). The parameters change after 2-3 cycles of chemotherapy was calculated according to the formula used in PET/CT. The difference values were calculated since Types I, II, and III, and "rapid" and "medium" values were given as percentages.

Statistical Analysis
The conformity of the variables to the normal distribution was examined with the Kolmogorov-Smirnov test. Continuous variables were expressed as median (min-max) and mean with standard deviation. The parameters calculated in the whole-body and late studies were compared using parametric or non-parametric tests. As a non-parametric test, Kruskal-Wallis analysis (K-W) and Jonckheere-Terpstra (J-P) trend analysis were performed in multiple groups. Chi-square, Fisher, t-test, or Mann-Whitney U tests were used when examining the response to NAC with univariate analyses. The diagnostic decision-making properties of the calculated parameters in predicting the surgical response were analyzed by ROC curve analysis. In the presence of significant threshold values, the sensitivity, specificity, and positive and negative predictive values were calculated. p<0.05 was considered statistically significant. Statistical analyses were performed using SPSS 18.

Results

The clinical data of patients are given in Table 1.

Table 1: Clinical information of patients

NAC Regimen
The chemotherapy regimen included four cycles of adriamycin and cyclophosphamide every 21 days, followed by weekly paclitaxel for 12 weeks. Patients with HER2+breast cancer also received concomitant weekly trastuzumab with paclitaxel.

Surgical Response Assessment
All patients underwent modified radical mastectomy following the end of NAC. The complete pathological response was detected in the primary tumor in 12 patients (30.8%). In the remaining 27 patients (69.2%), residual tumors ranging in size from 5 to 70 mm (median: 25 mm) were observed.

Visual Evaluation
¹⁸F-FDG PET/CT
We evaluated the primary tumor"s whole-body (n=39) and late prone (n=37) images at baseline. In 19 patients, additional tumors were detected with late prone images. After NAC, while the size and metabolic activity of the lesions decreased at different levels, no primary tumor was observed in whole-body images in six patients and late images in three patients. In one patient, the metabolic activity of the tumor increased. Two unifocal and one multifocal tumor not observed in wholebody images were detected in three patients with late images. We performed whole-body imaging in 34 patients at the end of NAC. Late images were present in 32 patients. While the residual tumor was observed in 19 patients with whole-body images, the residual tumor was detected in 24 patients with late images. Compared with the surgical response, the sensitivity, specificity, and positive and negative predictive values of wholebody imaging and late prone imaging were 62.5% versus 91.3%, 80% versus 71.4%, 62.6% versus 91.3%, and 47% versus 62.5%, respectively. While the highest sensitivity and positive and negative predictive values were obtained in late images, the specificity value was high with whole-body images detecting residual tumors.

Contrast-enhanced Dynamic MRI
Baseline DCE-MRI was obtained in 39 patients. While a single tumor focus (57 mm [17-200 mm]) was observed in 60% of the patients, two tumor foci were observed in 35%, and more than 2 tumor foci were observed in 5% of the patients. A significant reduction in tumor size was observed in the interim study (32 mm [5-100 mm]) (p<0.0001), and the major focus disappeared in one of the patients with two tumor foci. After NAC, the primary lesion completely disappeared in seven patients, while the lesion size (27 mm [11-85 mm]) decreased significantly in other patients. According to the surgical outcome, the sensitivity, specificity, and positive and negative predictive values of MRI for residual tumors were 84%, 37.5%, 80.7%, and 42.8%, respectively, Figure 1.

Fig 1: Sixty-six years old, left breast invasive ductal carcinoma, Grade 2, non-luminal tumor, tumor dimensions on MRI 38×16×29 mm, spherical volume: 9.1 cc, angiovolume: 9.4 cc, peak signal intensity: 187 (upper and middle row), SUVmax: 14.4 (bottom row).
MRI: Magnetic resonance imaging.

Quantitative Evaluation 18F-FDG PET/CT
Interim images of 38 patients were evaluated. The % change values between the interim and baseline images are given in Table 2.

Table 2: Percent change values of 18F-FDG PET/CT parameters between interim and baseline imaging in PCR and non-pCR patients (n=38)

Response to NAC
When the % changes were compared to surgical response, %TLG-adp and % MTV-adp were significantly different according to surgical response in the wholebody and late images. These values showed more variation in the group with the complete surgical response. We found no difference in whole-body and late prone images for %MTV-42.

DCE MRI
Interim MRI was performed on 38 patients. Baseline and interim MRI parameters were not significantly different between receptor subgroups and grades. When the PCR and non-PCR groups were compared, rapid (p=0.044), medium (p=0.044), and peak (p=0.034) values were statistically different and changed on baseline MRI. Long diameter (p=0.035) and volume (p=0.02) in interim MRI were significantly lower in the PCR group. MRI parameters changes are given in Table 3. When the % change values calculated in 34 patients were compared according to the surgical response, long diameter (p=0.041), volume (0.001), and curve peak (0.03) showed more changes in the PCR group compared to the non-PCR group, Figure 2.

Table 3: Percent change values of MRI parameters between interim and baseline imaging in PCR and non-pCR patients (n=38)

Fig 2: (a-c) Fifty-seven years old, left breast invasive ductal carcinoma, Grade 3, TN, baseline PET/CT (bottom row) tumor size 65×62×55 mm, SUVmax: 18.6. After two cycles of chemotherapy, the tumor size decreased to 38×43×42 mm and SUVmax to 11.2 on PET/CT (upper row) (a), 43 years old, left invasive ductal carcinoma, Grade 3, TN, baseline PET/CT (bottom row) tumor size 40×33×38 mm, SUVmax: 21.2, tumor size regressed to 15×14×9 mm and SUVmax: 2 after two cycles of chemotherapy (upper row) (b), 54 years old, right breast invasive ductal carcinoma, Grade 3, ER-, PR-, HER2+, tumor size in baseline PET/CT (bottom row) 47×35×41 mm, SUVmax: 16.7, tumor completely disappeared after two cures of chemotherapy (upper row). Surgical outcome: No residual tumor was observed (c).
PET/CT: Positron emission tomography/Computed tomography; TN: Triple negative; ER: Estrogen receptor; PR: Progesterone receptor; HER2: Human epidermal growth factor receptor-2.

Predictive Value of Parameters
Between the interim and baseline studies, percent change values of PSI and SV from MRI, SULpeak, MTV-adp, and TLG-adp from PET/CT could predict PCR with high accuracy. The list of parameters for which ROC analysis was performed to evaluate NAC response and whose p-value was significant is given in Table 4. In addition, the sensitivity, specificity, positive and negative predictive values, and accuracy values calculated for the determined threshold values are given in the same table (Table 4).

Table 4: Table showing the sensitivity, specificity, positive-negative predictive, and accuracy values calculated for the ¹⁸FFDG PET/CT and MRI parameters and the threshold value, whose p-value was significant in the ROC analysis

Discussion

When comparing standard whole-body imaging with late prone imaging in the visual evaluation of ¹⁸FFDG PET/CT, we found that the assessment of primary tumor was most successfully performed with late prone images. The breast was evaluated more easily in the prone position using a breast coil. We detected additional primary lesions in late imaging due to the increase in ¹⁸F-FDG uptake in the tumor with time, the decrease in the level of ¹⁸F-FDG in the normal breast tissue, and the increase in the tumor/ground activity contrast. In primary tumor evaluation, prone imaging is recommended to increase ¹⁸F-FDG PET/ CT sensitivity. Other authors have also described the use of breast coils which are also used to fuse MRI/ PET images to increase the specificity of MRI images. [,] Late prone ¹⁸F-FDG PET/CT images were more compatible with MRI. It is known that the late component of dual imaging increases not only specificity but also sensitivity, and our finding is consistent with the literature.[,]

While different tumor metabolic activity reduction levels were observed with interim ¹⁸F-FDG PET/ CT, we detected a reduction in lesion size with MRI. When the post-NAC, pre-surgery, ¹⁸F-FDG PET/CT, and MRI images were compared with the histopathological results, the sensitivity of MRI was higher than that of whole-body supine ¹⁸F-FDG PET/CT images. However, late images were more successful than MRI. Positive predictive values were higher than negative predictive values.

It is known that quantitative parameters are more successful than visual evaluation. The changes in SUVmax values were examined most frequently in the studies.[-] TLG and MTVs changes have been used. [,] In some of the studies, the predictive value of TLG was reported to be higher than SUVmax values. [] On the other hand, while a study reports that SUV max change is a more significant predictor than TLG change, a study also says that both MTVs changes are equally successful.[,] The % TLG-adp and %MTV-adp values showed significant differences between the PCR and non-PCR groups in whole-body and late images. No statistically significant difference was found with %MTV-42. We found the "adapted" method superior to the method in which 42% was used as the threshold value. In the literature, parameters are generally used to predict NAC response. In this study, we also examined TLG and MTV differently. In a meta-analysis of 19 articles and 920 patients, the sensitivity of PET/CT was 84%, the specificity was 66%, the positive predictive value was 50%, and the negative predictive value was 91%.[] In the articles in this meta-analysis, the sensitivity ranged from 33% to 100%, while the specificity values were reported between 30% and 100%. In another metaanalysis, which included 745 patients and evaluated 15 studies, the values were 80.5%, 78.8%, 79.8%, and 79.5%, respectively.[] Values ranging from 40% to 88% in SUVmax have been reported to predict NAC in these studies. One of the reasons why different values were detected is that imaging timing for the prediction of NAC response is not standard. Some authors imaged after one cycle, while others imaged after 2 or 3 cycles. For ease of comparison, we studied after 2-3 cycles, which is the most frequently used time in our study. Contrary to this, Humbert et al.[] suggested that a SUVmax <2.1 after the first cycle had an accuracy of 76%. The heterogeneity of the groups in the studies, the differences in PET/CT imaging times, and the lack of many patients lead to different results.

The volume change measured in standard MRI has been defined as a sensitive parameter in predicting NAC response. The contrast enhancement curve and dynamic parameters obtained by dynamic MRI with contrast provide information about the angiogenesis of the tumor. A study comparing MRI and ¹⁸F-FDG PET/CT in TN and non-TN breast tumors found that MRI enhancement kinetics and SUVmax change were correlated. It was emphasized that MRI measured angiogenesis/ perfusion and PET measured metabolism were correlated, which was more robust in the more aggressive TN group.[] In a study comparing MRI and PET/CT, Pengel et al.[] found the % SUVmax and tumor diameter change to be equally successful in predicting NAC. However, when analyzed according to receptor subgroups, the diameter change was significant only in the TN group, while the SUVmax change was significant in the TN and luminal groups. Using SUVmax and MRI enhancement parameters together, Lim et al.[29] showed that the combined enhancement curve (at least 6%) and SUVmax (at least 41%) change after one cure is a successful predictor of disease-free survival. A study comparing PET and MRI reported that both methods were not sensitive enough but had high specificity. SUVmax, SULmax, SULpeak, and TLG parameters of FDG PET/CT, and enhancement parameters of MRI were compared. While TLG did not differ between groups according to surgical response, SUV values differed; only SUVmax change was determined as an independent predictor in multivariate analysis.[] In our study, the baseline PSI, % PSI change, and % SV change from MRI parameters differed in PCR and non-pCR groups. The highest predictive value among these parameters is % SV. Volume change was correlated with metabolic parameters, and the highest correlation was found between % MTV-adp and % TLG-adp. In addition, it has been reported that hybrid ¹⁸F-FDG PET/ MRI imaging may provide more satisfactory results than ¹⁸F-FDG PET and MRI for the early assessment of NAC response in patients with breast cancer.[30]

Limitations
Although the study started with 46 patients, we had to exclude some patients due to a lack of data. Our study group was heterogeneous as the response differed according to the receptor groups. Separate statistical evaluation according to receptor subgroups could not be made due to the low number of patients.

Conclusion

Peer-review: Externally peer-reviewed.

Conflict of Interest: All authors declared no conflict of interest.

Ethics Committee Approval: The study was approved by the Hacettepe University Non-interventional Clinical Research Ethics Committee (no: GO 13/45-29, date: 23/01/2013).

Financial Support: None declared.

Authorship contributions: Concept - M.T., M.G.A., K.A., F.B.D., B.E.; Design - O.K., M.T., P.Ö.K., M.G.A., K.A., F.B.D., B.E.; Supervision - M.T., F.B.D., B.E.; Funding - None; Materials - None; Data collection and/or processing - M.T., P.Ö.K., M.G.A., K.A., F.B.D., B.E.; Data analysis and/or interpretation - O.K., M.T., B.E.; Literature search - O.K., M.T., K.A., B.E.; Writing - O.K., M.T., K.A., F.B.D., B.E.; Critical review - O.K., M.T., P.Ö.K., M.G.A., K.A., F.B.D., B.E.

References