European Radiology

, Volume 22, Issue 1, pp 18–25

Comparison of diffusion-weighted MR imaging and FDG PET/CT to predict pathological complete response to neoadjuvant chemotherapy in patients with breast cancer

Authors

  • Sang Hee Park
    • Department of RadiologySeoul National University Hospital
    • Department of RadiologySeoul National University Hospital
  • Nariya Cho
    • Department of RadiologySeoul National University Hospital
  • Jung Min Chang
    • Department of RadiologySeoul National University Hospital
  • Seock-Ah Im
    • Department of Internal MedicineSeoul National University Hospital
  • In Ae Park
    • Department of PathologySeoul National University Hospital
  • Keon Wook Kang
    • Department of Nuclear MedicineSeoul National University Hospital
  • Wonshik Han
    • Department of SurgerySeoul National University Hospital
  • Dong-Young Noh
    • Department of SurgerySeoul National University Hospital
Breast

DOI: 10.1007/s00330-011-2236-x

Cite this article as:
Park, S.H., Moon, W.K., Cho, N. et al. Eur Radiol (2012) 22: 18. doi:10.1007/s00330-011-2236-x

Abstract

Objective

To compare the use of diffusion-weighted MR imaging (DWI) and 18F-FDG PET/CT to predict pathological complete response (pCR) in breast cancer patients receiving neoadjuvant chemotherapy.

Methods

Thirty-four women with 34 invasive breast cancers underwent DWI and PET/CT before and after chemotherapy and before surgery. The percentage changes in the apparent diffusion coefficient (ADC) and the standardised uptake value (SUV) were calculated, and the diagnostic performances for predicting pCR were evaluated using receiver operating characteristic (ROC) curve analysis.

Results

After surgery, 7/34 patients (20.6%) were found to have pCR. Az values for DWI, PET/CT and the combined use of DWI and PET/CT were 0.910, 0.873 and 0.944, respectively. The best cut-offs for differentiating pCR from non-pCR were a 54.9% increase in the ADC and a 63.9% decrease in the SUV. DWI showed 100% (7/7) sensitivity and 70.4% (19/27) specificity and PET/CT showed 100% sensitivity and 77.8% (21/27) specificity. When DWI and PET/CT were combined, there was a trend towards improved specificity compared with DWI.

Conclusions

DWI and FDG PET/CT show similar diagnostic accuracy for predicting pCR to neoadjuvant chemotherapy in breast cancer patients. The combined use of DWI and FDG PET/CT has the potential to improve specificity in predicting pCR.

Key Points

DWI breast MR and PET/CT show similar accuracy for predicting pathological response

The combined use of DWI and PET/CT can potentially improve specificity

This can assist individualised treatment in breast cancer patients receiving neoadjvant chemotherapy

Keywords

Breast cancerChemotherapyDiffusion-weighted imagingFDG PETPET/CT

Introduction

Neoadjuvant chemotherapy has been established as the standard treatment method for locally advanced breast cancer. Patients who achieve pathological complete response (pCR) after neoadjuvant chemotherapy have longer disease-free and overall survival rates compared with non-responders [13]. Dynamic contrast-enhanced magnetic resonance (MR) imaging is known to provide the most accurate assessment of tumour response after neoadjuvant chemotherapy [4, 5].

Recent studies have investigated the use of diffusion-weighted MR imaging (DWI) and PET with fluorine-18 FDG as assessment tools for tumour response [68]. DWI can measure apparent diffusion coefficients (ADCs), which provide a quantitative measure of the diffusivity of water, and which provide information related to tumour cellularity and the integrity of cell membranes. In addition, this method is sensitive to intratumoral changes induced by chemotherapy [911]. Furthermore, DWI is helpful in predicting response to chemotherapy in patients with breast cancer [12, 13]. The few clinical studies performed in breast cancer cases have demonstrated that changes in ADCs precede changes in tumour diameter [1416].

18F-FDG PET or PET/CT enables visualisation of the increased glucose metabolism of cancer tissue. FDG PET has been shown to identify primary tumours, regional lymph nodes and distant metastases for primary and recurrent breast cancer with high diagnostic accuracy [17, 18]. PET/CT can provide different information about tumour response to chemotherapy compared with DWI. The metabolic activity within tumours has been a valuable biomarker, and the standardised uptake value (SUV) has been used as a quantitative parameter for the assessment of tumour response [8, 1921].

Recent results have suggested that both DWI and FDG PET/CT are promising methods for assessing tumour response after neoadjuvant chemotherapy; to our knowledge, however, there is no published study that compares the response assessment of DWI and FDG PET/CT in breast cancer patients. The purpose of this study was to compare DWI and FDG PET/CT in predicting pCR in patients with breast cancer receiving neoadjuvant chemotherapy.

Materials and methods

Case selection

This retrospective study was approved by our institutional review board, which waived the requirement for patient informed consent. Forty-two patients with biopsy-confirmed invasive breast cancer of at least 2 cm by dynamic contrast-enhanced MR imaging, who underwent DWI and 18F-FDG PET/CT before and after neoadjuvant chemotherapy but before surgery between April 2007 and May 2008 were included in this study. All patients underwent surgery, even if they were categorised as non-responders after chemotherapy. Eight patients were later excluded from the study including patients with distant metastases (n = 6) and patients with inadequate pre- or post-chemotherapy DWI quality due to excessive susceptibility or motion artefacts (n = 2). The final study population consisted of 34 women (mean age, 44 years; range, 27–60 years).

Of the 34 patients, 23 (67.6%) received three cycles of neoadjuvant chemotherapy with a combination of doxorubicin and docetaxel and the remaining 11 patients who had HER-2-positive tumours received six cycles of therapy with a combination of paclitaxel, gemcitabine and trastuzumab.

MR imaging technique

Patients underwent MR imaging before chemotherapy and after completing three or six cycles of chemotherapy. All pre- and post-chemotherapy breast MR imaging examinations were performed at 1.5 T (Signa; GE Medical Systems, Milwaukee, WI, USA) using a dedicated phased-array breast coil; subjects were imaged in the prone position. DWI was acquired in the transverse plane covering both breasts. A spin-echo single-shot echo-planar imaging sequence with diffusion-sensitising gradients applied along the x-, y-, and z-axes (i.e. isotopic DWI) was used before and after 180-degree pulses, and these images were used to synthesise isotopic transverse images (b values of 0 and 750 s/mm2; repetition time in ms/echo time in ms, 10,000/60.9; image matrix, 190 × 190; field of view, 240 × 240 mm; slice thickness/gap, 5 mm/0 mm; 2 NEX; acquisition time, 80 s). We set the b value to 750 s/mm2 based on expert recommendations [10] and the results of our pilot study. Fat-suppressed, T2-weighted turbo spin-echo sagittal images were also obtained. The following image parameters were used: repetition time in ms/echo time in ms, 5500/85.2; flip angle, 90 degrees; image matrix, 256 × 160; field of view, 200 × 200 mm; and slice thickness/gap, 1.5 mm/0 mm. A three-dimensional, T1-weighted, fast spoiled gradient-echo sequence was also performed, with bilateral sagittal imaging together with one unenhanced and four dynamic series after 90 s, 270 s, 360 s and 510 s after intravenous bolus enhancement. The following image parameters were used: repetition time in ms/echo time in ms, 6.5/2.5; flip angle, 0 degrees; image matrix, 320 × 160; field of view, 200 × 200 mm; and slice thickness/gap, 1.5 mm/0 mm. Gadobenate dimeglumine (Multihance; Bracco Imaging, Milan, Italy) was injected into an antecubital vein using an automated injector (Spectris MR, Medrad Europe, Maastricht, Netherlands) at a dose of 0.1 mmol/kg of body weight and a rate of 2 mL/s, and this was followed by a 20-mL saline flush.

DWI and ADC analysis

The MR images were reviewed on a picture archiving and communication system workstation monitor (m-view; Marotech); the images were interpreted by consensus between two radiologists (N.C. and S.H.P., with 9 and 3 years’ experience interpreting breast MR images, respectively). Both reviewers were blinded to the therapeutic responses to neoadjuvant chemotherapy and to the pathological findings.

Apparent diffusion coefficient maps were calculated, and dedicated software (Interactive Data Language Virtual Machine, version 6.3; Research Systems, Boulder, CO, USA) was used to derive ADC values using a linear regression model and the following equation: \( {\text{ADC}} = - \ln \left[ {S(b)/S(0)} \right]/b \) , where b is the diffusion weighting factor, and S(b) and S(0) are signal intensities with and without diffusion-sensitising gradients, respectively [16]. After identifying a hypointense lesion on an ADC map, regions of interest (ROIs) were drawn manually to include the entire tumour in the three largest cross-sectional planes using dynamic contrast-enhanced MR imaging information for reference purposes. Care was taken to avoid including normal breast parenchyma or fat. The mean ADC of each ROI was determined, and an average ADC was calculated for each tumour. The percentage of ADCs (% ADCs) was calculated as follows: \( \% {\text{ADC}} = \left( {{\text{ADC}}2\left[ {\text{after}} \right] - {\text{ADC}}1\left[ {\text{before}} \right]} \right)/{\text{ADC}}1\left[ {\text{before}} \right] \times 100 \) [6].

PET/CT and SUV analysis

Whole-body PET/CT imaging was performed using combined PET/CT (Gemini; Philips Medical Systems, Bothell, WA, USA). The Gemini is an open PET/CT system that combines helical dual-slice CT and three-dimensional PET equipped with its own transmission source. After allowing the patients to fast for at least 4 h, the patients received an intravenous injection of 240 to 400 MBq 18F-FDG; oral contrast medium was not administered. The blood glucose levels were checked in all patients before FDG administration, and none of the patients had a blood glucose level that exceeded 200 mg/dL. The CT images were acquired from the cerebellum to the pelvis without the use of intravenous contrast media, and a whole-body emission PET was performed for the same axial coverage. The images were reconstructed with a 512 × 512 matrix and a 50-cm field of view. The PET, CT and fused PET/CT images were generated and were then reviewed on a computer workstation.

Two institutional nuclear medicine physicians prospectively interpreted all PET/CT images by consensus. They were unaware of any details on tumour response to neoadjuvant chemotherapy or the pathological findings. When a hypermetabolic lesion was detected on pre-chemotherapy and post-chemotherapy PET/CT images, the maximum standardised uptake value (SUV) was prospectively calculated. The following equation was utilised: \( {\text{SUV}} = {\text{A}}/\left( {{\text{ID}}/{\text{BW}}} \right) \), where A was the decay-corrected mean activity in tissue (measured in millicuries per millilitre), ID was the injected dose of FDG (measured in millicuries), and BW was the patient body weight (measured in grams). The percentage of SUVs (% SUVs) was calculated as follows: \( \% {\text{SUV}} = \left( {{\text{SUV}}1\left[ {\text{before}} \right] - {\text{SUV}}2\left[ {\text{after}} \right]} \right)/{\text{SUV}}1\left[ {\text{before}} \right] \times 100 \).

Pathological examination

The surgical specimens were cut into 5-mm slices, fixed in 10% neutral-buffered formalin, and processed for histological examination. If a gross tumour was evident, each paraffin block containing the tumour was sliced and stained with haematoxylin and eosin (H&E) for evaluation. If no gross tumour was found, the tissue marker or hookwire left in the breast was identified, and examined.

Residual disease after chemotherapy was classified into one of three categories:
  1. 1)

    No residual malignancy and no sign of cancer cells;

     
  2. 2)

    No residual invasive cancer and DCIS present; or

     
  3. 3)

    Residual invasive cancer and pCR including categories 1 and 2 [2224].

     

In cases of residual invasive cancer, the pathological size was determined as the longest dimension.

Data and statistical analysis

Age and pre-chemotherapy tumour size were compared between the pCR and non-pCR groups using the independent-samples t test. The best cut-offs for % ADC and % SUV for differentiating CR and non-CR were determined from the receiver operating characteristic (ROC) analysis. The best cut-off was defined as the value corresponding to the highest average of sensitivity and specificity. The resulting best cut-off values were then used to assess the complete responsiveness of the tumour by DWI and PET/CT. CR for the combined use of DWI and PET/CT was defined as the tumour showing CR on both imaging techniques, and non-CR was defined as presenting non-CR on at least one of the imaging techniques. An ROC curve analysis was performed to compare the diagnostic performance of DWI, PET/CT and the combined use of DWI and PET/CT to predict pCR using the final histopathology as the reference standard. The diagnostic accuracy was estimated by calculating the area under the ROC curve (Az value). The best cut-off values for % ADC and % SUV were then used to calculate the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy for predicting pCR. Sensitivity, specificity and accuracy were compared between DWI and PET/CT, between DWI and the combined use of DWI and PET/CT and between PET/CT and the combined use of DWI and PET/CT using McNemar’s test. The agreement in the response assessment between DWI and histopathology, between PET/CT and histopathology and between DWI and PET/CT was calculated using weighted kappa statistics. A kappa value of less than 0.20 indicated poor agreement, 0.21–0.40 indicated fair agreement, 0.41–0.60 indicated moderate agreement, 0.61–0.80 indicated good agreement, and 0.81 or greater indicated very good agreement [25]. The correlation between % ADC and % SUV before and after chemotherapy was evaluated using the Pearson’s correlation coefficient. All statistical analyses were performed using the MedCalc software (version 11.4.4.0; MedCalc Software, Mariakerke, Belgium). A p value of less than 0.05 was considered to indicate a statistically significant difference.

Results

The median time interval between the completion of chemotherapy and the second DWI was 17 days (range, 6–31 days), and that between chemotherapy and the second PET/CT was 18 days (8–25 days). All patients underwent either modified radical mastectomy (n = 22) or breast-conserving surgery (n = 12) as the definitive surgery after chemotherapy. The histological types were invasive ductal carcinoma (n = 32), mucinous carcinoma (n = 1) and mixed ductal and lobular carcinoma (n = 1).

After chemotherapy, definitive surgery revealed pCR in 7 (20.6%) of the 34 patients. Four patients showed no evidence of malignant cells, and three patients showed ductal carcinoma in situ only. The remaining 27 patients showed residual invasive tumours (scattered cells, small foci, or bulk tumour); despite different residual lesions, they were all categorised into the non-pCR group. In our study, trastuzumab was given to all patients with HER-2-positive tumours and no patients with HER-2-negative tumours. The patients with HER-2-positive tumours showed significantly higher pCR rates (5 of the 11 [45.5%] patients with HER-2-positive tumours and 2 of the 23 [8.7%] patients with HER-2-negative tumours, p = 0.043).

The mean tumour diameter of the 34 breast cancers was 5.6 cm (median size, 5.1 cm; range, 2.3–10.0 cm) according to pre-chemotherapy dynamic contrast-enhanced MR imaging, and no significant difference in mean pretreatment tumour size was observed between patients with pCR (6.2 cm ± 1.1 [standard errors]) and patients with residual tumours (5.4 cm ± 0.4) (p = 0.440). There was no significant difference in mean age between patients with pCR (44.4 years) and patients with residual tumours (42.3 years) (p = 0.538).

Following chemotherapy, the mean ADC value of the 34 patients increased from 1.069 × 10−3 mm2/s to 1.562 × 10−3 mm2/s (p < 0.001), and the maximum SUV decreased from 6.83 to 2.22 (p < 0.001). The post-treatment mean ADC values and histopathological tumour diameters were inversely correlated (rho = −0.418; 95% confidence interval [CI] = −0.663 to −0.093; p = 0.014). The post-treatment maximum SUVs and histopathological tumour diameters were moderately correlated (rho = 0.673; 95% CI = 0.433 to 0.823; p < 0.001). The percentage of ADC was highly correlated with % SUV (rho = 0.701; 95% CI = 0.476 to 0.840; p < 0.001) (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-011-2236-x/MediaObjects/330_2011_2236_Fig1_HTML.gif
Fig. 1

Scatter-plot showing the relationship between changes in apparent diffusion coefficients (% ADC) from DWI and standardised uptake values (% SUV) from 18F-FDG PET/CT after neoadjuvant chemotherapy. % ADC correlated highly with % SUV (rho = 0.701, p < 0.001). pCR pathological complete response, Non-pCR non-pathological complete response

The area under the ROC curve of DWI (Az = 0.910; 95% CI = 0.761 to 0.981) was higher than that of PET/CT (Az = 0.873; 95% CI = 0.714 to 0.962), but no significant difference was found between them (p = 0.601). The best cut-offs for differentiating pCR from non-pCR in the ROC curve analysis were a 54.9% increase in ADC after chemotherapy and a 63.9% decrease in SUV after chemotherapy (Figs. 2 and 3). With this cut-off, DWI showed 100% (7/7) sensitivity, 70.4% (19/27) specificity, 46.7% (7/15) PPV, 100% (19/19) NPV, and 76.5% (26/34) accuracy, and PET/CT showed 100% (7/7) sensitivity, 77.8% (21/27) specificity, 53.8% (7/13) PPV, 100% (21/21) NPV, and 82.4% (28/34) accuracy in pCR prediction (Table 1). No significant differences in specificity or accuracy were observed between the techniques (p > 0.05). The strength of agreement was moderate between PET/CT and histology (κ = 0.590), between DWI and histology (κ = 0.494) and between DWI and PET/CT (κ = 0.516).
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-011-2236-x/MediaObjects/330_2011_2236_Fig2_HTML.gif
Fig. 2

A 45-year-old woman with invasive ductal carcinoma of the right breast who had residual invasive cancer after completing neoadjuvant chemotherapy; however, the case was assessed as a complete response by 18F-FDG PET/CT. a, d In the baseline transverse DWI (a), the mean ADC of the high-signal-intensity tumour (arrow) was 1.136 × 10−3 mm2/s. The mean ADC of the tumour (arrow) (d) increased to 1.647 × 10−3 mm2/s, and % ADC was 45.0% after chemotherapy, indicating non-complete response. b, c, e, f The baseline transverse PET/CT image (b) and PET maximal intensity image (c) showed an abnormal FDG uptake with a maximum SUV of 5.9 (arrows). The post-chemotherapy PET/CT image (e) and PET maximal intensity image (f) showed no FDG uptake (arrow), which was suggestive of complete response. The surgical histological analysis revealed a 1-cm residual invasive cancer

https://static-content.springer.com/image/art%3A10.1007%2Fs00330-011-2236-x/MediaObjects/330_2011_2236_Fig3_HTML.gif
Fig. 3

A 46-year-old woman with invasive ductal carcinoma of the left breast who had residual invasive cancer after completing neoadjuvant chemotherapy; however, the case was assessed as a complete response by DWI. a, d On the baseline transverse DWI (a), the mean ADC of the high-signal-intensity tumour (arrow) was 1.038 × 10−3 mm2/s. The mean ADC of the tumour (arrow) (d) increased to 1.642 × 10−3 mm2/s, and % ADC was 58.2% after chemotherapy, which was classified as a complete response by DWI. b, c, e, f The baseline transverse PET/CT image (b) and PET maximal intensity image (c) showed an abnormal FDG uptake with a maximum SUV of 17.4 (arrows). The post-chemotherapy PET/CT image (e) and PET maximal intensity image (f) showed residual abnormal FDG uptake with a maximum SUV of 9.4 (arrows) and % SUV of 46.0%, indicating a non-complete response. The surgical histological analysis revealed a 3.5-cm residual invasive cancer

Table 1

The diagnostic performance of DWI, PET/CT and the combined use of DWI and PET/CT for the prediction of pathological complete response

Parameter

DWI

PET/CT

DWI + PET/CT

Sensitivity (%)

100 (7/7)

100 (7/7)

100 (7/7)

Specificity (%)

70.4 (19/27)

77.8 (21/27)

88.9 (24/27)

Positive predictive value (%)

46.7 (7/15)

53.8 (7/13)

70.0 (7/10)

Negative predictive value (%)

100 (19/19)

100 (21/21)

100 (24/24)

Accuracy (%)

76.5 (26/34)

82.4 (28/34)

91.2 (31/34)

The Az value for the combined use of DWI and FDG PET/CT (Az = 0.944; 95% CI = 0.808 to 0.994) was improved compared with either technique used alone. Although no significant difference was demonstrated compared with DWI (p = 0.515) or PET/CT, there was a trend towards an improvement in Az with the combined use of DWI and PET/CT compared with PET/CT alone (p = 0.063) (Fig. 4). The results of the response assessment by each imaging technique and by the combined use of DWI and PET/CT and their correlations with the pathological response are listed in Table 2. The sensitivity and NPV were the same (100% for both) for each imaging technique. The specificity, PPV and accuracy in pCR prediction increased to 88.9% (24/27), 70% (7/10) and 91.2% (31/34), respectively. There was a trend towards improved specificity and accuracy with the combined use of DWI and PET/CT compared with DWI (p = 0.063 for both). No significant differences in specificity and accuracy were observed between PET/CT and the combined use of DWI and PET/CT (p = 0.25 for both). All three of the patients who were falsely categorised as complete responders by both DWI and PET/CT were members of the non-pCR group, and the maximal diameters of the residual invasive tumours in their pathological specimens were 0.5, 0.3 and 0.1 cm.
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-011-2236-x/MediaObjects/330_2011_2236_Fig4_HTML.gif
Fig. 4

The ROC curves used to evaluate CR to neoadjuvant chemotherapy with DWI, PET/CT and the combined use of DWI and PET/CT. The Az value from DWI (0.910) was higher than that from PET/CT (0.873), but no significant difference was found (p = 0.601). The Az value was improved with the combined use of DWI and PET/CT (0.944), but there was no significant difference

Table 2

The assessment of tumour response by DWI, PET/CT and the combined use of DWI and PET/CT correlated with the pathological response

 

No. of patients

DWI

PET/CT

DWI + PET/CT

CR

Non-CR

CR

Non-CR

CR

Non-CR

Pathological response

CR

7 (20.6)

7

0

7

0

7

0

Non-CR

27 (79.4)

8

19

6

21

3

24

Total

34 (100.0)

15 (44.1)

19 (55.9)

13 (38.2)

21 (61.8)

10 (29.4)

24 (70.6)

The numbers in parentheses were used to calculate the percentages. CR complete response, Non-CR non-complete response.

Discussion

In this comparison between DWI and FDG PET/CT for predicting pCR to neoadjuvant chemotherapy in breast cancer patients, the percentage of ADC correlated highly with the percentage of SUV (rho = 0.701, p < 0.001). Previous studies have shown significant negative correlations between ADC and SUV in brain tumours, primary rectal cancers and pelvic lymph nodes in prostate cancer [2628]. ADC maps reflect water diffusion, while 18F-FDG uptake in tumours relies on the augmented number of functional glucose transporters and glycolytic enzymes present in metabolically active and proliferating cells [29]. Both processes may be related to tumour cellularity. Areas of high cellularity have more structures and more cell membranes, resulting in greater impedance of the motion/diffusion of water molecules and low ADC values; they also have increased cellular proliferation and higher FDG uptake. Conversely, areas with low tumour cellularity and necrotic parts are less resistant to the diffusion of water molecules, which corresponds to a higher ADC, scarce proliferative activity and a lower uptake of FDG. A decrease in tumour FDG uptake post-treatment may indicate downregulation of glucose transporters [30] or a loss of viable cells. An increase in ADC post-treatment is believed to be a consequence of cellular damage and necrosis. This connection is a likely explanation for the high correlation between ADC and SUV changes in our study.

Changes in the maximal diameter of a tumour (i.e. the RECIST criteria) have been widely accepted for assessing tumour response after neoadjuvant chemotherapy using conventional imaging techniques [31]. Although DWI and FDG PET/CT have no standard criteria for response assessment, the change in SUV has been proven to be a good marker. Several studies have suggested a 50% to 79% decrease in SUV as the cut-off value for responders and a 100% decrease in FDG uptake within the tumour volume (so that it is indistinguishable from the surrounding normal tissue) as the criteria for complete responders [7, 32, 33]. In our study, we used the thresholds of the changes in ADC and SUV from the ROC curve analysis for response assessment. The strength of agreement between each imaging technique and the histology was moderate.

The significance of pCR has been a matter of primary concern in neoadjuvant chemotherapy because it has exhibited a strong relationship with postoperative prognosis in several trials [2, 34]. Approximately 70% of patients demonstrate clinical response to neoadjuvant chemotherapy, but only approximately 20% achieve pCR [3, 35], similar to 20.6% found in our study. Therefore, non-invasive imaging that allows for the prediction of pCR to chemotherapy could help individualise treatment and avoid unnecessary systemic toxicity, costs and treatment delays. On the basis of the ROC curve analysis, this study revealed that DWI (Az = 0.910; 95% CI = 0.761 to 0.981) showed slightly higher diagnostic performance than PET/CT (Az = 0.873; 95% CI = 0.714 to 0.962), but the difference was not statistically significant (p = 0.601).

The combined use of DWI and FDG PET/CT showed a higher Az value (0.944; 95% CI = 0.808 to 0.994). Although no significant difference was found compared with DWI (p = 0.515) or PET/CT, there was a trend towards an improved Az value compared with PET/CT (p = 0.063). Also, specificity and accuracy in pCR prediction increased to 88.9% (24/27) and 91.2% (31/34), respectively. There was a trend towards improved specificity and accuracy with the combined use of DWI and PET/CT compared with DWI (p = 0.063 for both). No significant differences in specificity or accuracy were observed between PET/CT and the combined use of DWI and PET/CT (p = 0.25 for both). All three of the patients who were falsely categorised as complete responders by both DWI and PET/CT were members of the non-pCR group, and the tumour sizes of the residual invasive tumours in their pathological specimens were small at 0.5, 0.3 and 0.1 cm. If the residual invasive tumour size is small, the predictive accuracy for pCR by both DWI and FDG PET/CT can be decreased. Previous studies have shown that the combined use of FDG PET/CT and MRI or the use of fused PET and MRI increases the specificity of preoperative MRI in breast cancer patients [36, 37]. Additional data are needed to confirm the statistical significance of our preliminary findings.

Our study has several limitations. First, interobserver variability and the reproducibility of the ADC and SUV measurements were not evaluated. We selected three ROIs per lesion to reduce interobserver variability for ADC. However, a volumetric histogram analysis [16] and voxel-wise analysis [38] would have been more accurate. Second, our complete response criteria for % ADC and % SUV were obtained from a cut-off value determined using our study population. The use of these thresholds in different populations would likely yield different diagnostic accuracy. Third, we only included the tumour response of primary breast lesions. PET/CT can monitor entire levels of regional tumour extent including axillary and supraclavicular lymph nodes. However, because the field of view for DWI is limited to the breast and occasionally the lower axillary fossa, we included only the breast area. The small sample size of the present study is also a limitation.

In conclusion, DWI and FDG PET/CT show similar diagnostic accuracy in the prediction of pCR to neoadjuvant chemotherapy in patients with breast cancer. The combined use of DWI and FDG PET/CT can potentially improve specificity in pCR prediction.

Acknowledgements

This study was supported by a grant from the Korea Healthcare Technology R&D Project, Ministry for Health, Welfare & Family Affairs, Republic of Korea (A070001) and by a grant from the Innovative Research Institute for Cell Therapy, Republic of Korea (A062260).

Copyright information

© European Society of Radiology 2011