The added value of whole-body magnetic resonance imaging in the management of patients with advanced breast cancer

This study investigates the impact of whole-body MRI (WB-MRI) in addition to CT of chest-abdomen-pelvis (CT-CAP) and 18F-FDG PET/CT (PET/CT) on systemic treatment decisions in standard clinical practice for patients with advanced breast cancer (ABC). WB-MRI examinations in ABC patients were extracted from our WB-MRI registry (2009–2017). Patients under systemic treatment who underwent WB-MRI and a control examination (CT-CAP or PET/CT) were included. Data regarding progressive disease (PD) reported either on WB-MRI or on the control examinations were collected. Data regarding eventual change in treatment after the imaging evaluation were collected. It was finally evaluated whether the detection of PD by any of the two modalities had induced a change in treatment. Among 910 WB-MRI examinations in ABC patients, 58 had a paired control examination (16 CT-CAP and 42 PET/CT) and were analysed. In 23/58 paired examinations, additional sites of disease were reported only on WB-MRI and not on the control examination. In 17/28 paired examinations, PD was reported only on WB-MRI and not on the control examination. In 14 out of the 28 pairs of examinations that were followed by a change in treatment, PD had been reported only on WBMRI (14/28; 50%), while stable disease had been reported on the control examination. In conclusion, WB-MRI disclosed PD earlier than the control examination (CT-CAP or PET/CT), and it was responsible alone for 50% of all changes in treatment.


Introduction
Advanced breast cancer (ABC) encompasses metastatic (MBC) and locally advanced breast cancer (LABC). MBC is still an incurable disease with a 5-year survival rate between 15% and 27% [1,2], and progression to MBC occurs in about 20-30% of non-metastatic patients [3]. LABC is diagnosed at presentation in 8.5% of American and 4% of European patients with breast cancer, and despite aggressive treatment, many of these patients eventually progress to a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 monitoring of MBC patients [20,21], which rely on DWI as indirect but solid biomarker of tissue cellularity [22]. DWI enables the radiologist to rapidly recognize the presence and evolution of metastatic disease, providing early signs of disease response or progression [21]. Apparent Diffusion Coefficient (ADC) maps derive from a quantitative analysis of diffusionweighted images; ADC values show an inverse correlation with cellularity in many tumour histotypes, including breast cancer [23]. Evaluating the findings highlighted by DWI according to their corresponding ADC values helps differentiating benign from malignant lesions [24]. Notably, DWI alone is not always accurate this task [25], therefore correlation with morphologic T1 and T2 images is mandatory for avoiding false-positive and false-negative results [26]. One of the strengths of WB-MRI is that it can be performed in clinically acceptable examination times (30-40 min). Two other great advantages include the lack of contrast injection and ionising radiation. Due to the superior sensitivity for initial bone infiltration, WB-MRI has become the gold standard in the diagnosis and assessment of multiple myeloma [27,28]. Guidelines for acquisition, interpretation and reporting of WB-MRI in advanced prostate cancer [29] have recently been published, as a support for the increasing use of WB-MRI in metastatic and high-risk prostate cancer. Due to high performance in bone-metastatic tumours, WB-MRI has been naturally suggested for the evaluation of MBC. At our centre, an increasing number of patients with bone-predominant or bone-only MBC is being evaluated using WB-MRI, alongside other whole-body imaging modalities such as CT-CAP and PET/CT. Two imaging modalities may occasionally be performed at the same time point for the characterization of complex metastatic patterns, unclear tumour behaviour, or when the hypothesis of disease progression is affected by discordant data.
This retrospective study of our practice in the evaluation of ABC patients under systemic treatment investigates whether the use of WB-MRI in addition to chest-abdomen-pelvis CT (CT-CAP) or PET/CT brings an additional benefit in patient management, allowing for earlier changes of ineffective treatments.

Population
We interrogated an ongoing registry including all WB-MRI performed in our institute from February 1 st 2009 to April 31 st 2017 (2444 examinations). We included in the study all WB-MRI performed on ABC patients undergoing systemic treatment; all examinations performed for baseline disease staging were excluded. All WB-MRI performed within 8 weeks from a CT-CAP or an 18 F-FDG PET/CT were included in the study. Each WB-MRI included in the study was therefore paired to a CT-CAP or a PET/CT (defined as "control examinations") performed under the same systemic treatment. Patient's medical records were made available for all included examinations; more specifically, we collected the medical records compiled at the start of the ongoing treatments, and those compiled immediately after each pair of examinations (within one month). The ethics committee of our institution (European Institute of Oncology, Milan) approved this study. The ethics committee waived the requirement for informed consent, as this retrospective study did not alter or influence the diagnostic and therapeutic paths of the patients, involving summary data from imaging reports and medical records.

Imaging technique
Our WB-MRI protocol Table 1, consisting of sagittal T1-weighted and T2-weighted sequences on the whole spine, axial T1 weighted, T2-weighted and DWI from head to mid-thigh, was performed on a 1.5 T scanner (Magnetom Avanto, Siemens Healthcare Sector, Erlangen, Germany). Anatomy-specific phased-array surface coils were used for all body regions. The typical cumulative WB-MRI data acquisition time was 40 minutes. Post-processing included in-line (Water e Fat images, ADC maps) and off-line reconstruction (Radial maximum intensity projections of high b-value images, multi-plane reconstructions and relative fat-fraction maps). CT-CAP and 18 FDG PET-CT were performed using each imaging department's standard imaging protocols. CT-CAP always included intravenous contrast administration. All the original WB-MRI scans were reported by a pool of three radiologists (one senior radiologist and two junior radiologists who were supervised by the senior one, with an experience of 8, 5 and 4 years respectively in oncological WB-MRI). As to the occurrence of stable or progressive disease, particularly in the skeleton, images were interpreted according to the criteria proposed by Padhani et al. [21].

Data collection
Two residents in radiology with 3 and 1 years' experience in WB-MRI reporting independently extracted the relevant data from the original reports of WB-MRI and the control examinations. Firstly, data regarding the reported extent of disease were collected, annotating the presence or absence of metastases categorized by 7 sites: bone, primary site, lymph nodes (including regional and distant), liver, lung/pleura, peritoneum/retroperitoneum (including stomach, bowel, intra and retro-peritoneal fat), other.
Sites of metastatic disease reported by only one of the two modalities (either WB-MRI or the control examination) were annotated and termed additional sites (AS). Discordance on the assessment of disease extent was defined as the presence of one or more AS reported by one of the two modalities.
Secondly, reports were reviewed to determine whether WB-MRI and/or the control examination had described progression of disease (PD), defined as the increased extent of the disease compared to the patient's known baseline at the start of the ongoing treatment. Discordance on reported PD was defined as occurring when WB-MRI and the control examination reports concluded with different assessments (i.e. PD opposed to stable disease). When only one modality described PD, the sites of disease in which PD was reported were annotated. Thirdly, data were collected regarding all cases in which the ongoing treatment was changed as result of the imaging response. In all cases of discordance between the two modalities, data were collected as to whether the PD reported by one of the two examinations had motivated a change in treatment, as stated in each patient's medical record.

Results
A total of 910 WB-MRI examinations were performed on ABC patients in the selected period. Fifty-eight WB-MRI were paired to a control examination: 16 (28%) to CT-CAP and 42 (72%) to PET/CT. The median age of the included patients was 56 years (range, 36-80 years). Information on the characteristics of the breast cancers in these patients is summarized in Table 2. The median time distance between WB-MRI and the control examination was 27 days (range, 1-54 days). In 40 pairs of examinations (69%) the status of the patient at the time of the evaluation was M1, while in 18 pairs (31%) it was M0 (no prior history of metastases).

Extent of disease
Out of the 58 paired examinations, metastases were reported in 49 WB-MRI and in 39 control examinations. Bone metastases were the most frequent, reported by 37 WB-MRI and by 30 control examinations. The second most frequent site of metastases was lymph nodes, reported in 15 WB-MRI and in 18 control examinations. Discordant assessments on disease extent were  Table 3.

Assessment of progressive disease
Of the 58 pairs of examinations, SD was reported in 18 pairs and PD in a total of 40 pairs. PD was reported by both examinations in 23 out of 40 pairs, while in the other 17 it was only reported by WB-MRI, with the control examination reporting SD. Table 4 summarizes the anatomical distribution of the sites of PD detected only by WB-MRI and of those detected by both examinations. Twelve out of 18 non-metastatic (M0) patients were up-staged to M1, in seven of these, PD was reported only on WB/MRI.   for this paper include detailed information on the extent of disease, sites of PD and impact on systemic treatment for all pairs of examinations, S1 Table.

Extent of disease
As expected, in our study the most frequent site of metastatic findings in all modalities was bone, being bone the most common site of metastases in breast cancer [30]. The second and third most common sites of metastases in both modalities were lymph nodes and lung/pleura, as well mirroring the common metastatic pattern reported in literature. Bone was also the most frequent additional site (AS) of disease reported by WB-MRI and not by the control examination (Figs 2-4). A meta-analysis on the detection of bone metastases [31]  The second most frequent AS on WB-MRI was peritoneum/retroperitoneum (Figs 5 and 6). MRI could be able to reveal peritoneal or retroperitoneal disease more confidently by means DWI, which can highlight small metastatic foci within normal tissues. In a study including 32 patients from Michielsen et al. WB-MRI (with DWI), PET/CT and CT were compared in the detection of peritoneal and retroperitoneal metastases of ovarian primary, with surgery as standard of reference. In this study WB-MRI was excellent in the detection of intraperitoneal disease, with sensitivity/specificity of 91% and 91%, compared to 65% and 82% for CT, and 52% and 85% for PET/CT, and was equivalent in the detection of retroperitoneal disease compared to PET/CT [32]. In another study including 26 patients from Fujii et al., the sensitivity and specificity of DWI for peritoneal carcinosis were 90% and 95.5%, respectively [33]. In our study, in 4 out of 6 WB-MRI with AS in peritoneum/retroperitoneum, the histology of the primary was ILC. This observation suggests that WB-MRI might be particularly suited for the assessment of ILC patients compared to CT and PET/CT. In ILC, metastatic spread to the gastrointestinal tract and to peritoneum/retroperitoneum is particularly frequent [34], and CT might be sometimes unable to detect these localizations, also because of its reduced sensitivity for small nodules (below 10 mm size) in the abdominal cavity [35] and in the retroperitoneum [36]. ILC, due to characteristic loss of E-cadherin adhesion protein [37], presents an un-aggregated growth [38], making metastases less measurable. PET/CT as well has a pitfall for small metastatic lesions below its spatial resolution, especially in the colon or in the small bowel where non-specific FDG uptake is often reported [39,40]. Moreover, the low FDG avidity of ILC [41] reduces the value of PET/CT for the detection of bone metastases in ILC [42].
Most of the AS reported by the control examination were in lymph nodes, being found in 3 PET/CT (sub-clavicular, hilar or mediastinal lymph nodes). This result might be related to the superior sensitivity and specificity of PET/CT (SE 98%,SP 83%) for lymph nodes compared to MRI in mediastinum (SE 80%, SP 75%) [43,44].

Assessment of progressive disease and change in treatment
There were 17 episodes of PD reported only on WB-MRI and not reported on the control examination, accounting for 43% (17/40) of all cases of PD. Among these, 14 determined a change in treatment, accounting for 50% (14/28) of all changes after imaging evaluation. Evaluating these patients by means of WB-MRI resulted in a detection of PD that would have otherwise been recognized weeks or months later. This allowed the oncologists to interrupt ineffective treatments earlier, before the potential onset of acute complications of PD, such as SREs. The scope and methodology of our work are in line with a previously published work by Kosmin et.al. [45] that evaluates the impact of WB-MRI alongside CT-CAP on the management of MBC patients under systemic treatment. The study by Kosmin et.al. [45] showed that WB-MRI had an impact on treatment management in nearly 35% of all cases (where 16/46 of all changes in treatment were due to PD reported only on WB-MRI and not on CT-CAP). In the subgroup of WB-MRI paired to CT-CAP within our study, the impact of WB-MRI on treatment management was 70%. The greater impact of WB-MRI observed in our study could be due to different factors. In the study by Kosmin the time between WB-MRI and the control examination was below two weeks, therefore the greater interval between the examinations in our study could have allowed variations of the actual disease extent. A second reason for the different result could be the difference in population: 33% (19/58) of our examinations were performed on ILC patients while these formed only 12% of the population studied by Kosmin. We believe that the higher frequency of ILC patients in our study compared to the normal rate in breast cancer population (approximately 15%) [46] might have amplified the impact of WB-MRI on the detection of metastases. Thirdly, the presence of M0 patients in our cohort might have amplified the impact of WB-MRI on patient management, since the detection of metastases in these patients has a greater chance of leading to a change in treatment. Interestingly, there were no cases of PD reported only on the control examination, since all the AS reported on PET/CT occurred in episodes of PD involving also other anatomical sites, which were reported on both modalities. There are some limitations to this study. Firstly, it was performed retrospectively without a new image analysis. This implies a certain degree of inhomogeneity between the original reports in terms of communication structure and level of detail, especially among the control examinations.We believe however that this kind of analysis is more relatable to true clinical practice. Secondly, the distance (median 27 days) between WB-MRI and the control examinations is quite high. It is important to note, however, that ours is not a perspective study, in which the schedule of imaging controls is fixed in the most convenient way. When oncologists in our institute program WB-MRI for an ABC patient, other imaging modalities such as CT or PET/CT are dismissed or sometimes interleaved at 3-6 months. This implies a subjective choice by the oncologist: indeed ESMO guidelines for ABC do not recommend a specific modality for follow-up [8]. These factors, combined with the waiting list for WB-MRI slots, motivated the quite high distance between WB-MRI and control examinations. Based on the findings of this study, we are currently planning a perspective study with a distance between WB-MRI and the control examination within seven days. Thirdly, our study does not address an eventual impact of WB-MRI on patient survival. This limitation is common to other imaging studies. Nevertheless, the earlier detection of PD and the subsequent changes in treatment might help delaying the onset of SREs and of other acute complications in metastatic patients, with a potential impact on quality of life and disability.
In conclusion, WB-MRI disclosed PD earlier than the control examination (CT-CAP or PET/CT) and it was responsible alone for 50% of all changes in treatment in our cohort. Larger and prospective studies are therefore encouraged to validate our observations in ABC patients.
Supporting information S1 Table. Detailed results for all paired evaluations. Abbreviations: G = grade; ER = oestrogen receptor status; PG = progesterone receptor status; HER2 = expression of