Contrast-enhanced spectral mammography in neoadjuvant chemotherapy monitoring: a comparison with breast magnetic resonance imaging

Preoperative neoadjuvant chemotherapy (NAC) is considered the standard treatment for locally advanced breast carcinomas to observe chemosensitivity, reduce micrometastatic diseases and for down-staging of primary tumors [1, 2].

Response to NAC is currently assessed by combining clinical examination and conventional imaging techniques such as mammography and ultrasound (US). These methods, however, often fail to predict the final pathologic response [3]. Breast contrast-enhanced magnetic resonance imaging (MRI) has proven to be superior to mammography and US in the assessment of tumor extent and presence of additional foci (i.e. multifocality and/or multicentricity), and highly sensitive in identifying residual disease following NAC [4‐6]. Contrast-enhanced spectral mammography (CESM), a promising dual-energy breast imaging technique acquired after intravenous iodine-based contrast agent, allows the evaluation of vascularized lesions in the breast, which has, up to now, been a prerogative mainly of breast MRI [7]. Compared with mammography and US, CESM improves the sensitivity for breast cancer detection without decreasing specificity [8‐10]. CESM digital detector has higher spatial resolution than MRI (i.e. typically pixel sizes 100 μm and 1 mm, respectively), revealing details that are approximately 10 times better [11].

CESM is currently used for the detection of primary breast cancer, for the assessment of the extent of disease, as a problem-solving tool and as a replacement for MRI where the latter is contraindicated [12]. In terms of patient preferences and tolerance, significantly higher overall preference towards CESM has been demonstrated, due to faster procedure time, greater comfort and significantly lower rate of anxiety [13].

MRI and CESM assess treatment response by evaluating tumor size reduction and type of vascularization. In the assessment of the size of the tumor no relevant discrepancies between CESM and MRI have been observed [11, 14, 15]. However MRI involves high costs and limited availability also in industrialized countries [16]. Thus there is the rationale for substituting MRI with CESM in NAC monitoring if the performance of CESM is similar to that of MRI.

To the best of our knowledge, there are no data in the literature reporting CESM versus MRI for NAC monitoring. Therefore, the purpose of our prospective study was to compare CESM and breast MRI in the assessment of tumor response, using postoperative histologic assessment as the gold standard.

This prospective study was performed in accordance with the Declaration of Helsinki and approved by the local ethics committee. All eligible patients were willing to undergo all study procedures and provided written informed consent prior to enrolment.

Of the initial 54 patients enrolled, 46 completed the study (8/54 were excluded because of interruption in NAC before any imaging monitoring could be performed). Table 1 shows detailed patient characteristics. No multifocal or multicentric tumors were detected.

Among the 46 patients considered, as decided in the Breast Unit Multidisciplinary Group, mainly because of poor tolerance to NAC, 5 women interrupted the chemotherapy 3 months after it was started, after the second step of the imaging evaluation, and directly underwent surgery. In these patients, the second CESM and MRI have been analyzed as “post NAC” and compared to the results of assessments of the surgical specimens. Of these, according to the RECIST 1.1 criteria, two patients achieved pCR, one pPR, one pSD and one pPD. No adverse reaction was observed either to iodinated contrast medium or to gadolinium contrast agent.

In a comparison of the post-NAC measurements with the histopathological response in surgical specimens, CESM had stronger agreement with histology than did MRI (Lin's coefficient 0.81 and 0.59, respectively; CESM-MRI concordance difference 0.22, CI 0.07–0.58; PCC 0.85 and 0.67, respectively). Mean tumor sizes measured on CESM and MRI are shown in Fig. 2. Both methods tended to underestimate the real extent of the residual tumor (mean underestimation of 4.1 mm in CESM and of 7.5 mm in MRI).

Measurements on CESM image sets agreed strongly with MRI, especially before (Lin’s coefficient 0.96, CI 0.94–0.98) and during (Lin’s coefficient 0.94, CI 0.89–0.97) NAC (Fig. 3). The post-NAC agreement coefficient was 0.76 (Lin’s coefficient, CI 0.61–0.86), revealing considerable divergence, which reflects the different correlation with histopathological assessment.

Post-NAC, 28% of patients had no residual pathological enhancement on CESM as compared to 46% on MRI.

According to histopathological assessment, 87% (40/46) of patients responded to NAC (pCR + pPR), with pCR achieved in 17% (8/46). CESM defined 40 patients as responders (CR + PR) while MRI defined 41. In this single discordance, the shrinkage of tumor was 25% on CESM and 32% on MRI, close to the threshold (30%). The main differences between CESM and MRI were observed in the distinction between CR and PR. Referring to histopathological response on surgical specimens, diagnostic performance indexes in pCR prediction are reported for both methods in Table 2.

On histopathological assessment 20 tumors had a DCIS component, mostly in the form of small foci in the tumor bed (Fig. 4). Of these, only three cases were classified as pCR since no concomitant residual invasive component was found: two were considered as CR both on CESM and MRI, whereas in just one case CESM revealed 5 mm of faint enhancement in the tumor bed, confirmed by histopathological assessment corresponding to a 3 mm DCIS. No surgical re-intervention was needed.

Table 3 describes the tumor characteristics at diagnosis and after NAC, according to the BI-RADS lexicon. During and after chemotherapy, we mainly observed progressive tumor shrinkage and loss of enhancement among the responders (Fig. 5). Focusing on CESM, the shrinkage of the lesion was concentric in 60% of patients (Fig. 6), while in 40% there were multiple small foci of faint enhancement included in the tumor bed (Fig. 7).

As none of the study patients had evidence of bilateral neoplasms at the pre-NAC evaluation, the CESM examinations were always mono-lateral. The total mean AGD of a complete mono-lateral CESM, consisting of low-energy and high-energy exposures on two mammographic views (CC and MLO), was 7.15 mGy (range 3.88–23.34 mGy), depending on breast thickness (mean of 59 mm, range 40–85 mm) and glandularity, considered as the percentage of peak breast density measured automatically by the mammographic unit (mean of 50%, range 9–95%).

In this prospective study we hypothesized that CESM could provide an accurate assessment of tumor response after NAC, comparable to that provided by MRI, which is currently considered the most reliable diagnostic method in treatment monitoring. In this study, the correlation between measurements using these imaging tools and histopathological assessment, the gold standard, was stronger for CESM (Lin’s coefficient 0.81; PCC 0.85) than for MRI (Lin’s coefficient 0.59; PCC 0.67), and both imaging tools involved slight underestimation.

In women who did not undergo NAC, Lobbes et al. [13] identified PCC > 0.9 with both CESM and MRI compared to histopathological assessment, with a small systematic overestimation of the tumor diameter measured on MRI, but not on CESM. In a similar setting, Łuczyńska et al. [11] described overestimation of lesion size both in CESM (1.7 mm) and in MRI (1.8 mm). Fallenberg et al. [16] identified underestimation of the tumor size using MRI compared to CESM and pathological assessment, describing an increasing discrepancy in the size of the pathological lesion in larger breast tumors measured by the imaging methods. This evaluation may also be relevant to the patients enrolled in our study, as NAC is recommended for locally advanced breast carcinomas, and at diagnosis the mean size of the lesions was 43 mm.

Several variables influence tumor shrinkage and its evaluation during NAC, including primary tumor size, edema or necrosis. Also, the various chemotherapeutic drugs contribute differently to the variation in the primitive lesion, with significant consequences on imaging. Taxanes have been reported to have greater antiangiogenic activity compared to anthracyclines [29], and this antivascular effect, with the lack of surrounding inflammatory response, lowers the enhancement on tumor tissue, leading to underestimation of tumor size by MRI [5]. These effects were similar in our patients as the NAC protocol was the same, including both anthracyclines and taxanes.

In this study, the cumulative impact of these variables is reflected in the progressively weaker correlation between measurements using CESM and MRI, comparing the pre-NAC (Lin’s coefficient 0.96) and during-NAC (Lin’s coefficient 0.94) evaluations to the final post-NAC (Lin’s coefficient 0.76). Apart from the changes in tumor micro-vessel functionality, it must be considered that the overall loss of cellularity is not always reflected by a decrease in tumor size, because the fibrous stroma remains [30]. Furthermore, when the residual cancer cells appear as small foci or scattered cells, they receive nutrients via diffusion and not from vascular perfusion [30]. Similarly, the contrast agent also moves to the ducts by diffusion, so the amount of contrast reaching the tissue is time dependent. Fallenberg et al. [16] observed that longer time delays between contrast injection and CESM exposure could result in stronger enhancement and hence better visibility compared to MRI. In contrast to the rapid washout in MRI, enhancement on CESM in fact persists for at least 10 min after contrast agent infusion [15]. Further studies are encouraged to improve a protocol for CESM dedicated to NAC monitoring, especially focusing on the optimal timing in CESM acquisition after NAC to stress the via-diffusion enhancement. Other possible variables include differences in the molecular structure of gadolinium-based and iodine-based contrast medium and the effects of breast compression. And last of all, the 10 times higher spatial resolution of CESM versus MRI [11] may help reveal tiny foci of residual enhanced lesion on the tumor bed.

According to the RECIST 1.1 criteria, the main issue considered was the distinction between CR and PR, with a significant divergent post-NAC assessment of no residual pathological enhancement on CESM (30%) and on MRI (48%). In the limited population of this study, CESM showed promising results in the definition of CR, correctly assessing each pCR, whereas one was missed on MRI. Otherwise, each patient with residual disease enhanced in CESM was confirmed on assessment of the surgical specimen.

The reliable assessment of both the extent of residual disease and pCR after NAC is crucial to decision making, surgical planning and prediction of final outcomes. Any overestimation of lesion sizes would have less severe consequences on surgical treatment than would underestimation, especially regarding safety margins. However the underestimation of residual tumor in this study had no consequences on the surgical planning since it mainly affected the discrimination between PR and CR in responding patients, in whom the tumor bed had to be completely excised anyway.

The advantages of CESM over breast MRI are the short examination time and lower costs [14]. Furthermore, CESM permits evaluation of both microcalcifications (on the low-energy image) and of enhanced structure (on the recombined image) in a single examination, overcoming this limitation of MRI and its variable sensitivity to in situ cancers [11, 31]. Even if the persistence of residual DCIS is considered as pCR, we believe that its extent must be considered for correct surgical planning, and CESM may be a precious tool for this purpose. Further evaluations are needed. CESM might also be a good alternative if MRI is not available or for patients with contraindications to MRI (e.g. pacemakers) or claustrophobia. The tolerance and overall preference of CESM versus MRI [13] must be taken into serious consideration in women who are facing the psychological burden of chemotherapy and have to repeat the examination three times in approximately 6 months. Disadvantages are the use of an iodine-based contrast agent instead of a gadolinium-based agent and the additional radiation dose. In this study the mean AGD of a complete mono-lateral CESM, consisting of low-energy and high-energy exposures for two mammographic views (CC and MLO), was 7.15 mGy. These doses are close to the ranges described in the literature [9, 16, 32, 33]. The AGD for the low-energy image was equivalent to that of one conventional mammogram, while for the high-energy image it was approximately 20% of the dose of one conventional mammogram [9]. However, this added dose was not considered as restricting, since all patients enrolled also had indications for radiotherapy treatment, which includes a total dose of 45 Gy on the target, i.e. about 10,000 times the dose of a CESM. In this study our ethical preference was to limit CESM to the affected breast to avoid any additional dose.

This study had several limitations. First of all, the small number of patients enrolled who were not randomized, which limits the statistical significance of the results. Nevertheless, the promising findings encourage further studies. Second, MRI was performed using different units, which may have led to heterogeneous findings despite following the same protocol. Also, the participation of seven independent radiologists might be considered a limitation because of a supposed non-homogeneous interpretation of images. However, this reflects the routine diagnostic workflow. Finally, no multicentric or multifocal breast cancers were included in the study.

CESM and MRI lesion size measurements were highly correlated. CESM seems at least as reliable as MRI in assessing the response to NAC, with encouraging results, and may be considered as an alternative if MRI is contraindicated or its availability is limited.