Predicting response to neo-adjuvant chemotherapy and assessment of residual disease in breast cancer using contrast-enhanced spectral mammography: a combined qualitative and quantitative approach

Magnetic resonance imaging (MRI) is the gold standard imaging modality for evaluation of response for neo-adjuvant chemotherapy (NAC) in breast cancer as it has the advantage of providing both; morphology assessment together with providing functional information which can be obtained by contrast injection. Until the recent emergence of contrast-enhanced mammography as a promising breast imaging modality, these features were considered unique for MRI. The aim of the study is to evaluate the competence of contrast-enhanced spectral mammography (CESM) in the prediction of response to NAC and the assessment of residual disease extent, as well as the assessment of a new combined (quantitative and qualitative) evaluation approach that is proposed by the authors. The study included 81 patients with pathologically proved breast cancer scheduled for receiving NAC. They underwent 2 CESM examinations; pre- and post-NAC (maximum 10 days before surgery). All patients were assessed using the RECIST 1.1 criteria and a combined approach (RECIST+ qualitative subjective assessment). Results were in correlation to postoperative pathology using the Miller-Payne grading. For statistical evaluation, patients were classified into responders and non-responders. Postoperative histopathology showed that 60/81 lesions were responders (Miller-Payne grades 3, 4, and 5) while the combined response evaluation approach and RECIST 1.1 alone showed 57/60 (95%) patients and 46/60 patients (76.7%) as responders respectively. The combined response evaluation approach showed higher sensitivity and positive and negative predictive values compared to the evaluation based on RECIST alone (95%, 87.6%, and 81.2% compared to 76.6%, 86.7%, and 50% respectively). CESM can be readily used to assess tumor response to NAC and allows the assessment of functional changes in residual tumor cells in addition to size discrepancy. Using CESM, we could accurately assess residual tumor extent and thus enable better surgical planning.


Background
Neo-adjuvant chemotherapy (NAC) allows the physicians to monitor disease development in vivo, hence the need for a modality that can assess tumor response during treatment and that can accurately detect any residual disease left after therapy [1].
There is increasing evidence that magnetic resonance imaging (MRI) is an excellent imaging tool to monitor response to neo-adjuvant chemotherapy, for both early response monitoring and assessment of residual disease extent [1]. However, breast MRI is expensive and not widely available. Similar to breast MRI, contrastenhanced spectral mammography (CESM) could be of particular interest in the assessment of the extent of disease [2].
The aim of the study is to evaluate the competence of contrast-enhanced spectral mammography (CESM) in the prediction of response to NAC and the assessment of residual disease extent as well as the assessment of a new combined (quantitative and qualitative) evaluation approach that is proposed by the authors.

Patients
From March 2015 to June 2016, 81 patients with an age range between 27 and 77 years were enrolled in this prospective study. The study protocol was approved by the Institutional Review Board of a "Multidisciplinary Breast Cancer Institute" and informed consent was applied for the used data of the enrolled individuals. According to the decision of the "Multidisciplinary Breast Cancer Tumor Board," all non-metastatic patients with locally advanced or triple-negative breast cancer who were scheduled for surgery after completing the full course of neo-adjuvant chemotherapy were eligible to join the study.
Patients who were not candidates for NAC or those who have already started their NAC course, patients with distant metastases, pregnant females, those with a history of allergy, or those with renal impairment were excluded from the study (Fig. 1).

Imaging assessment schedule
All participants were subjected to a baseline CESM (pre-NAC) at the time of diagnosis and before NAC as well as another study after completion of the full course of NAC and before surgery (post-NAC). In the post-NAC exam, only the breast of concern was examined to minimize unnecessary exposure. The maximum interval between the post-NAC exam and surgery was 10 days.

The technique of CESM examination
Examinations were performed on a commercial CESM device (SenoBright CESM, GE Medical Systems, Milwaukee WI).
A bolus of an iodinated contrast agent (iohexol, 300 mg I/ml) at a dose of 1.5 ml/kg is injected into the antecubital vein contralateral to the side of concern. After a 2-min wait, the 2 standard mammographic views were taken for each breast. As there was no need for a "pre-contrast" image, compression was applied after the injection to reduce the potential for movement artifacts and to retain the degree of image detail associated with standard mammography. In each mammographic position, a pair of low-and high-energy images was acquired. The low-energy images (similar to mammography) were acquired at peak kilovoltage values ranging from 26 to 31 kVp, which is below the K-edge of iodine. High-energy images (sensitive to iodine) were acquired at 45-49 kVp, which is above the K-edge of iodine.
Iodine-enhanced images are obtained by subtraction of the two images through appropriate image processing in a way to reduce the background parenchymal enhancement so that enhancing lesions are seen [3,4].

Interpretation of CESM findings
The analysis was performed using independent double reading by two different radiologists (10, 19 years' experience in mammography and 4, 7 years' experience in CESM). Readers were blinded for each other analysis, the pathology reports, and follow up the outcome. In case of disagreement between the two radiologists, the mammograms were re-evaluated by a third reader (the Chief Breast Imaging Consultant of the Department) and an agreement was achieved. The patterns of contrast uptake and lesion size were reported in both the preand post-NAC studies. Response to NAC was assessed in the post-NAC CESM once applying the Response evaluation Criteria in Solid Tumors (RECIST 1.1) and once using a combined quantitative and qualitative approach, which was proposed by the authors (see "assessment of response to NAC" detailed below).

Patterns of contrast enhancement pre and post-NAC
The pattern of contrast was described in both pre-and post-NAC studies as follows: A. Baseline pattern of contrast enhancement (Fig. 2) The initial contrast enhancement pattern of breast cancer was classified into four categories described by Tozaki  B. Combined quantitative and qualitative assessment (Table 1) The combined assessment was proposed by the involved researchers depending on a combination of assessment of change in the largest diameter of the target lesion together with the evaluation of the difference in intensity of contrast uptake between the pre-and post-NAC studies.
The largest dimension of the lesions was measured in both the pre-and post-NAC CESM studies and the percentage of change in size was calculated. According to the percent of change in lesion size, we classified the response into 6 patterns: progressive, stable, lesions Evaluation of the change in lesion intensity was subjective. Readers had to report whether there was no change, mild to moderate decrease in intensity, marked decrease in intensity with only faint residual enhancement, and no residual enhancement.
Assessment of response patterns was based on combinations between size and intensity changes resulting in 6 response patterns: -Progressive disease: at least a 20% increase in tumor size with/without a change in the intensity/pattern of enhancement.

Histopathology and immuno-histochemistry
A. Initial histopathology and immuno-histochemistry Initial histopathology was performed from pretreatment formalin-fixed, paraffin-embedded core biopsy samples. Immunohistochemistry (ER, PR, HER2/neu, and KI67) was performed on the same core biopsy samples. ER and PR statuses were scored from 0 to 8. Total scores of 0-2 are considered negative, a score of 3 is equivocal, and scores of 4-8 are considered positive. Testing for HER2/neu gene amplification was also performed by using the silver enhanced in situ hybridization technique. A ratio of 2.0 or more and/or Her2 gene copy number of 6 or more is considered positive for amplification of this gene, a ratio of less than 2.0 and a gene coy number < 4 is considered negative, and a ratio of less than 2 and a gene copy number ≥ 4 and < 6 is scored as equivocal.

B. Post NAC assessment
Assessment of response was according to the percent of change of tumor size and cellularity. Negative margins of the resected specimen should be confirmed to guarantee accurate size assessment. The largest diameter of both the original tumor bed and residual area with invasive cancer were then measured in the gross specimen using a standard ruler and were confirmed microscopically by submitting the specimen to the histologic mapping of serial specimen sections under the microscope. In non-concentric shrinkage, the residual tumor size was considered as the diameter of the crosssectional area containing residual tumor cells. To assess for the response, the percentage of the area with residual carcinoma within the original tumor bed is calculated and the average cellularity is estimated by calculating the average cellularity in the microscopic fields of the tumor bed. The abnormal (provisional tumor) area is measured in the gross specimen using a standard ruler. Consecutive blocks of the whole abnormal area (including adjacent fibrotic tissue) are submitted for microscopic examination to confirm gross tumor measurement. The total tumor size is estimated by measuring the maximum dimension of the viable invasive cancer on the examined slides fitted together. The percent of the area with residual viable invasive carcinoma within the tumor area is then calculated for assessment of therapy response.
Tumor regression was semi-quantitatively graded using surgical biopsy specimens by two pathologists individually based on the Miller-Payne grading system which compares tumor cellularity in pre-neo-adjuvant therapy core biopsy specimens and the surgical specimens as follows [8]:

Statistical analysis
Data were coded and entered using the statistical package SPSS (Statistical Package for the Social Sciences) version 24. Data were summarized using mean, standard deviation, median, minimum, and maximum in quantitative data and using frequency (count) and relative frequency (percentage) for categorical data. Standard diagnostic indices including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated. For comparing categorical data, Chi-square (χ 2 ) test was performed. The exact test was used instead when the expected frequency is less than 5. Correlations between quantitative variables were done using Spearman correlation coefficient. A P value less than 0.05 was considered as statistically significant. The correlation between maximum tumor diameter based on CESM and histopathology was expressed using scatter plots. In addition, the agreement between these measurements was expressed using Bland-Altman plots.
Cohen's kappa coefficient was used to compare the inter-observer agreement. The inter-observer % of agreement was 96%.

Results
The study included 81 patients who performed a preand post-NAC CESM examination.
On assessing the subjective change in intensity of contrast between the pre-and post-NAC studies, the two readers agreed in their initial independent readings on 80/81 cases (kappa 0.96).

Baseline patterns of contrast enhancement and shrinkage patterns
The baseline patterns of contrast uptake and the corresponding shrinkage pattern of the 81 lesions are shown in Tables 2 and 3.

CESM assessment of response to neo-adjuvant chemotherapy
A. RECIST 1.1 evaluation In reference to the (RECIST 1.1) criteria, progressive disease (PD) was not identified in any of the cases. Stable disease (SD) was seen in 28/81 (34.6%) of lesions, partial

B. Combined quantitative and qualitative evaluation
In reference to the combined evaluation that was proposed by the involved researchers, progressive disease was not seen in any of the lesions. Stable disease was seen in 4/81 (4.9%) of the lesions, poor response was seen in 12/81 (14.8%) of the lesions, moderate response was seen in 21/81 (26%) of the lesions, marked response was seen in 24/81 (29.6%) of the lesions, and complete response was seen in 20/81 (24.7%) of the lesions.
Correlation between the radiological response (using the RECIST criteria and the combined response evaluation) and the pathological response Evaluation of response using the combined approach showed significantly better agreement with histopathology results than the evaluation based on RECIST 1.1 alone (Table 6).

The overall accuracy of CESM in the evaluation of response to neo-adjuvant chemotherapy
The combined response evaluation approach showed higher diagnostic indices as compared to the evaluation based on RECIST 1.1 alone (Table 7, Fig. 7).

Correlation between the residual tumor size in CESM and histopathology specimens
A strong correlation (R: 0.921, P value < 0.001) was found between the size of residual mass lesions on CESM (mean size: 3.12 cm ± 2.95SD) and histopathology specimens (mean size 3.33 cm =/− 3.22SD) in the total 81 lesions. The overall mean size discrepancy was 0.85 cm ± 1.04SD (Table 8, Figs. 8 and 9).
The size of the lesions showing types I, II, and IV shrinkage patterns showed a significant correlation with the histopathology size. The mean CESM-pathologic size discrepancy was least in types I, II, and IV shrinkage patterns and greatest in type III ( Table 8).
The mean CESM-pathologic size discrepancy was least with the triple-negative tumors and HER2-positive tumors as shown in Table 8.

Discussion
With the development of new lines of cancer treatment options, various diagnostic imaging modalities are emerging accompanied by new strategies and guidelines to predict early tumor response to NAC and to assess the burden of residual disease [9][10][11][12]. The current study included 81 patients who have completed the full NAC course. Assessment of response was performed in reference to the previously described MRI shrinkage patterns by Kim et al. [6].
The combined response evaluation proposed by the authors showed significantly better agreement with histopathology than the evaluation based on RECIST 1.1 alone. Contrast-enhanced mammography is less expensive and less time consuming and may even have a chief advantage over MRI as it can assess the true extent of residual tumors associated with microcalcifications [1,13,14].
After NAC, some histopathological changes may occur in the tumor bed with an overall reduction in size and the amount of viable tumor cellularity giving the different post-NAC shrinkage patterns described by Kim et al. [6,15]. Some tumors may show concentric shrinkage patterns, while others may disintegrate or fragment into discrete foci which are difficult to measure. In some other tumors, a fibrous stroma may persist in the original tumor bed. This fibrous stroma may show subtle enhancement giving a false impression of poor response to NAC as there will be no change in tumor diameter [16]. Knowing the shrinkage pattern helps in the accurate assessment of residual disease before conservative surgery [6,17].
Goorts et al. stated that a possible explanation for the better prediction of pathological response by MRI halfway through NAC than after NAC is that taxanes might suppress MRI enhancement irrespective of the cytotoxic activity, as reported by Schrading et al. And since in Goorts et al. patient's cohort, taxanes were also only given during the second half of treatment, this could be a plausible explanation; however, their study also showed that this might be false negative, especially when lobular carcinomas are classified as complete responders. Furthermore, the two triple-negative tumors classified as type 0 were both true negatives [16].
Kim et al. stated that types III and I shrinkage patterns were more frequently observed in the pathological responder group, and type IV was more frequently noted in the non-responder group [6]. Our results were in agreement with Kim et al., where 17/24 (70.8%) of type 1 shrinkage pattern and 8/9 (88.8%) type 3 shrinkage pattern were responders. On the other hand, solitary lesions showed the most favorable response to NAC where 6/26 (23%) lesions showed a complete radiological response and even the remaining 20/26 (77%) with the residual disease showed type 1 concentric shrinkage pattern making conservative surgery more amenable. On the other hand, replaced lesions showed the least favorable response where 7/10 (70%) lesions showed persistent diffuse enhancement ending up with a modified radical mastectomy (MRM).
We then assessed the accuracy of CESM in the evaluation of residual disease; once using the RECIST 1.1 and another time using the combined response evaluation approach (Table 4). This was performed in correlation with the postoperative histopathology results.  Morphologic measurement of change in tumor size in response NAC helps assess therapeutic effectiveness through the use of the RECIST 1.1 [7]. Since the publication of the RECIST, several reports have been published regarding its low reliability in evaluating responses. The RECIST 1.1 mainly depend on tumor size alterations and they do not express other morphologic (tumor necrosis, hemorrhage, and cavitation), functional, or metabolic changes that may occur with NAC [10]. Forner et al. stated that when applying RECIST alone in the assessment of response to NAC, tumor necrosis and partial tumor response can be missed [18]. Magnetic resonance imaging has long been considered the most promising imaging modality in assessing response NAC, as it not only provides accurate tumor size assessment but also gives us information about the change in tumor enhancement pattern reflecting the functional change in tumor cells [19]. However, since breast MRI and CESM show similar performance and are based on similar principles, we proposed a new assessment approach based on both size assessment and enhancement pattern using CESM. In reference to the postoperative histopathology results of the 81 patients included in the current study, 60/81 lesions were responders (Miller-Payne grades 3, 4, and 5). Using the combined response evaluation approach, 57/60 patients (95%) were classified as responders, while by using RECIST 1.1 alone, only 46/60 patients (76.7%) were classified as responders. The combined response evaluation approach showed higher sensitivity and positive and negative predictive values compared to the evaluation based on RECIST alone.(95%, 87.6%, and 81.2 % compared to 76.6%, 86.7%, and 50% respectively). The specificity of the RECI ST evaluation was slightly higher than the combined approach (66.9% as compared to 61.6%). Both methods of radiological evaluation coincided with postoperative pathology in predicting PCR in 20/21 (95.2%) lesions, with an additional 1 false-positive and 1 false-negative case. Pathology of the false-positive responder showed areas of fibrosis and hyalinosis entangling breast lobules leading to a decreased contrast uptake, yet with residual viable tumor cells. False-negative responders were significantly more in the RECIST 1.1 evaluation than in the combined approach (14 compared to 3 non-responders respectively) as the overall loss of tumor cells was not always expressed as a change in tumor size. Another explanation to the false-positive cases was also suggested in previous studies stating that the residual tumor may shrink when fixed in formalin after surgical removal giving a false impression of size reduction [20,21].  The overall reported diagnostic indices of the combined approach matched some reported results for MRI in the same context and mismatched others. For example, in a meta-analysis performed by Yuan et al., MRI showed a high specificity (90.7%) and relative lower sensitivity (63.1%) in predicting pathologic complete remission after NAC while in another study, MRI showed a sensitivity of 100%, a specificity of 50%, a PPV of 83.3%, and an NPV of 100% [22,23]. Thus, both overestimation and underestimation were previously observed and reported [24].
Many studies have presented correlation coefficients between size measurements assessed by MRI compared to pathological tumor size measurement [25][26][27]. In an analysis of these studies by Lobbes et al., the median correlation coefficient was calculated and found to be 0.698, with a range of 0.21-0.982 [24]. In the current study, the overall correlation between size measurements assessed by CESM compared to pathological tumor size was 0.921 (P < 0.001). The correlation coefficient achieved by CESM in our study was significantly higher  than that recorded for MRI in the meta-analysis by Lobbes et al [24]. In recent studies, it was observed that the accuracy of breast MRI to assess residual disease and predict pCR depended on breast cancer subtypes. For example, it was observed that MRI was able to predict pCR more accurately in patients having HER2-positive tumors when compared to HER2-negative tumors. In the latter case, a higher false-negative rate was observed [24].
HER2-negative and hormone receptor-positive cancers and lesions showing non-mass-like enhancement are more likely to show residual disease as small foci or scattered cells after NAC, leading to underestimation of residual disease extent at MR imaging, and the diagnostic results of MR imaging should be used with caution in surgical planning [28].
Loo et al. also concluded that response monitoring using breast MRI is accurate in patients having triplenegative or HER2-positive tumors. However, they found it was inaccurate in ER-positive/HER2-negative subtypes [29].
In the current study, the overall accuracy of CESM in assessing the 41 patients with hormone receptorpositive/HER2-negative tumors was 88.9%. Five out of these 41 patients achieved pCR; CESM was able to predict pCR in 4 out of 5 patients with only one false positive case.
In a study by Chen et al., the results indicated that the diagnostic accuracy of MR imaging is better in HER2positive than in HER2-negative cancers, with a size discrepancy of 0.5 ± 0.9 cm versus 2.3 ± 3.5 cm (P = .009) [28]. These results suggest that patients with HER2negative hormone receptor-positive disease are the poorest candidates for MR imaging evaluation [28].
In the current study, the mean size discrepancy for the HER2-positive tumors was 0.5 ± 0.7 cm versus 1.1 ± 1.2 cm for HER2-negative tumors (P < 0.001). Note the smaller size discrepancy achieved by using CESM in evaluating the residual disease in Her2-negative tumors compared to the MR evaluation in the Chen et al. study [28].
The present study had some limitations. The combined response evaluation approach carries the limitation of being subjective, which raises the need for the introduction of non-subjective methods in the evaluation of the change in the intensity of enhancement by CESM. However, the inter-observer variability between the different readers was excellent (kappa 0.960). Patients were only evaluated by CESM before and after completion of NAC. During the course, the evaluation was based on clinical and ultrasound assessments. The authors did not want to subject the patients to unnecessary irradiation before validating the efficiency of CESM to assess response. In a future application, it is necessary to do a limited CESM examination of the affected breast to predict early response.

Conclusion
Contrast-enhanced spectral mammography can be readily used to assess tumor response to NAC. It showed high sensitivity and positive and negative predictive values in this study. One main advantage of CESM is that it allows the assessment of functional changes in residual tumor cells in addition to size discrepancy. Assessment based on size discrepancy alone as stated in the RECIST criteria does not reflect morphologic, functional, or metabolic changes that may occur with NAC especially in non-mass forming tumors. Using the combined assessment that was proposed by the authors, CESM can accurately assess residual disease after NAC.