Technique, protocols and adverse reactions for contrast-enhanced spectral mammography (CESM): a systematic review

We reviewed technical parameters, acquisition protocols and adverse reactions (ARs) for contrast-enhanced spectral mammography (CESM). A systematic search in databases, including MEDLINE/EMBASE, was performed to extract publication year, country of origin, study design; patients; mammography unit/vendor, radiation dose, low-/high-energy tube voltage; contrast molecule, concentration and dose; injection modality, ARs and acquisition delay; order of views; examination time. Of 120 retrieved articles, 84 were included from 22 countries (September 2003–January 2019), totalling 14012 patients. Design was prospective in 44/84 studies (52%); in 70/84 articles (83%), a General Electric unit with factory-set kVp was used. Per-view average glandular dose, reported in 12/84 studies (14%), ranged 0.43–2.65 mGy. Contrast type/concentration was reported in 79/84 studies (94%), with Iohexol 350 mgI/mL mostly used (25/79, 32%), dose and flow rate in 72/84 (86%), with 1.5 mL/kg dose at 3 mL/s in 62/72 studies (86%). Injection was described in 69/84 articles (82%), automated in 59/69 (85%), manual in 10/69 (15%) and flush in 35/84 (42%), with 10–30 mL dose in 19/35 (54%). An examination time < 10 min was reported in 65/84 studies (77%), 120 s acquisition delay in 65/84 (77%) and order of views in 42/84 (50%) studies, beginning with the craniocaudal view of the non-suspected breast in 7/42 (17%). Thirty ARs were reported by 14/84 (17%) studies (26 mild, 3 moderate, 1 severe non-fatal) with a pooled rate of 0.82% (fixed-effect model). Only half of CESM studies were prospective; factory-set kVp, contrast 1.5 mL/kg at 3 mL/s and 120 s acquisition delay were mostly used; only 1 severe AR was reported. CESM protocol standardisation is advisable.


Background
During the 1960s and 1970s, randomised controlled trials proved that screen-film mammography for breast cancer screening yields a reduction in breast cancer mortality [1]. Since the early 2000s, screen-film mammography was progressively replaced by digital mammography (DM), which improved performance especially in women under 50 years of age and in case of dense breasts, even though providing an intrinsically inferior spatial resolution [2]. In the last two decades, digital breast tomosynthesis brought substantial further improvements [3,4], increasing cancer detection rate and reducing the recall rate [5].
Contrast-enhanced mammography is the combination of X-ray mammography with intravenous administration of iodinated contrast agent (ICA) [6]. It was first attempted using a digital subtraction technique [7][8][9], but this approach was soon abandoned due to difficulties in co-registration of unenhanced and contrast-enhanced images [10,11]. In the last two decades, contrast-enhanced spectral mammography (CESM) has been introduced, based on dualenergy breast exposure (about 26-33 kVp and 44-50 kVp) after contrast administration, so that the precontrast exposure was no longer needed [10,12]. CESM allows for the visualisation of enhancing findings over the normal unenhancing breast tissue, exploiting the increased contrast uptake of malignancies [6,10,13].
A time delay between the first appearance of new imaging techniques and their implementation in diagnostic routine is expected for many reasons, including not only the definition of indications but also the reproducibility of results. The latter is strongly influenced by technique details, such as contrast agent concentration, dose and injection rate, breast compression and positioning, exposure parameters and acquisition protocol. Indeed, the fact that CESM is variably performed across different centres, without an agreed and standardised technique, does not come as a surprise: this circumstance echoes the one observed for contrast-enhanced breast MRI in the 1990s, now settled by the publication of detailed international guidelines [43][44][45][46]. Therefore, the aim of this work was to review CESM studies focusing on adopted technique, contrast agent issues and acquisition workflow. This effort is crucial for future CESM investigations to be reproducible and comparable.

Study protocol
No ethics committee approval was needed for this systematic review. The study protocol was registered on PROSPERO (protocol CRD42018118554), the international prospective register of systematic reviews [47]. This systematic review was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [48].

Search strategy and eligibility criteria
In February 2019, a systematic search was performed on MEDLINE (PubMed, https://www.ncbi.nlm.nih.gov/ pubmed/), EMBASE (Elsevier), the Cochrane Library (Cochrane Database of Systematic Reviews) and the Cochrane Central Register of Controlled Trials for articles that reported or may have reported CESM technique. A controlled vocabulary (medical subject headings in PubMed and EMBASE thesaurus keywords in EMBASE) was used. The search string was (cesm OR 'contrast enhanced spectral mammography'/exp. OR 'dual energy mammography' OR 'contrast enhanced digital mammography'/exp. OR 'contrast-enhanced mammography' OR 'dual-energy subtraction mammography' OR cedm OR cedsm OR 'contrast enhanced spectral imaging' OR 'high energy and low energy digital mammography') AND ('procedures'/exp. OR 'method' OR 'methods' OR 'procedure' OR 'procedures' OR 'technique' OR 'acquisition'/exp. OR 'contrast medium'/exp. OR 'contrast agent' OR 'contrast dye' OR 'contrast material' OR 'contrast media' OR 'contrast medium' OR 'radiocontrast medium' OR 'radiography contrast medium' OR 'roentgen contrast medium' OR 'image processing'/exp. OR 'image processing' OR 'image processing, computer-assisted' OR 'processing, image').
The search was limited to original studies on humans published in English, French and Spanish on peerreviewed journals, with an available abstract. No publication date limits were applied. First article screening was performed by two independent readers (A.C. and M.Z., with 1-and 3-year experience in breast imaging, respectively) considering only title and abstract. Eligible articles were those that reported in the title or in the abstract the use of CESM technique or that could have contained these data in the manuscript. After downloading eligible articles, the full text was read for a complete assessment. Finally, references of included articles were handsearched to check for further eligible studies.

Data extraction
Data extraction was performed independently by the same two readers who performed the literature search. Disagreements were settled by consensus. For each analysed article, year of publication, institution (such as hospitals, imaging facilities, breast units including radiology sections or any other type of centre in which CESM is performed) and country origin as well as research groups, design, number of patients and demographics were retrieved. Mammography unit, vendor, radiation dose and technical features such as low-and high-energy peak kilovoltage (kVp), anode/filter combinations and exposure parameters were also extracted. Moreover, contrast agent type, dose and concentration were retrieved, as well as injection modality, if manual or automated, flow rate and additional post-contrast saline flush or "bolus chaser" if present. Furthermore, mild, moderate or severe adverse reactions to ICAs were extracted alongside strategies for their prevention. Regarding the acquisition protocol, time between contrast injection and first image acquisition and maximum examination duration were extracted. Regarding the order of views, we reported the acquisition sequence of the standard mammographic projections considering the craniocaudal (CC) and the mediolateral oblique (MLO) views, including the first side acquired. Missing data were requested to authors.

Evidence synthesis
To avoid risk of data duplication bias, in case of articles published by the same research group, we considered the possibility of performing subgroup analysis: therefore, before delving into further analysis of protocol description, we chose to change our viewpoint from the number of articles reporting a specific protocol to the minimum number of times a protocol was reported by a single research group.
Regarding the pooled rate of adverse reactions related to ICA administration across studies, statistical analysis was performed using Comprehensive Meta-Analysis v2.2.057 (Biostat, Englewood, NJ, USA) using the metaanalysis model "Number of events and study population". I 2 statistics was first calculated to assess heterogeneity and the fixed-effect model was used to provide the rate of adverse reactions and 95% of confidence intervals (CI). The risk of publication bias was assessed by visually inspecting funnel plot and performing the Egger test [49].

Studies
A flowchart of study selection is shown in Fig. 1 focused on technical features. The geographic distribution of research groups is depicted in Fig. 2.

Populations and settings
Data synthesis is reported in Table 1. The number of patients ranged from 5 [63] to 2303 [13], for a total of 14,012 patients, with mean or median age ranging from 45 years [40] to 66 years [23]. In 29/84 studies (35%), CESM was performed on patients from comprehensive databases of heterogeneous settings, such as pre-or post-operative evaluation, adjuvant or neoadjuvant chemotherapy response monitoring and equivocal findings at conventional imaging. The remaining 55 studies (65%) were individually centred on a unique setting. Twenty-seven studies (32%) performed CESM on suspicious cases from conventional imaging and screening recalls, 11 studies (13%) in a firstline screening setting, 7 (8%) performed CESM exclusively for known cancer staging, 4 (5%) in a pre-operative setting, 4 (5%) to assess and monitor the response to adjuvant chemotherapy and 2 (2%) in a post-operative setting.
Timing of CESM examination with menstrual cycle was reported only in 18/84 studies (21%). In 10/18 (56%) articles, it was mentioned but not applied; in 6/18 (33%), it was applied with a feasibility window between the 5th and 14th day of menstrual cycle; in 2/18 (11%), CESM was synchronously performed with MRI in different phases of menstrual cycle to evaluate and compare background parenchymal enhancement.

Technical features and parameters
In 70 out of 84 studies (83%), different systems from General Electric Healthcare (Chicago, IL, USA) were used, all with a prototype or a commercial release of the SenoBright upgrade which is required to perform dualenergy contrast-enhanced imaging. Twelve out of 84 articles (14%) reported the adoption of Selenia Dimensions Mammography Unit (Hologic Inc., Marlborough, MA, USA), while the remaining 2/84 (3%) studies were conducted with a Siemens Healthineers (Erlangen, Germany) Mammography System (Mammomat or Mammomat Inspiration).
Of 69 studies detailing the tube voltage of both low-and high-energy acquisitions, all but one (99%) acquired lowenergy images between 26 and 33.2 kVp, which is the peak kilovoltage threshold of iodine, while all 69 acquired highenergy images well above this threshold, i.e. between 44 and 50 kVp. The anode/filter combination was reported Fig. 1 Flowchart of the study selection and exclusion for articles on contrast-enhanced spectral mammography by 42/84 studies. Exposure parameters were unambiguously reported only in one study [10], whereas in 5 early studies [7,8,32,59,85], they were manually adjusted according to breast thickness and density; thirty-five other studies declared an automatic regulation of these parameters performed by the mammography unit.
The outline of the image acquisition sequence remains more variable. Ten out of 84 studies (12%), accounting for 2734 patients (19%) did not clearly describe it and did not provide a reference to other protocols, while 3/84 (4%, 103/14012 patients, 1%) employed a curtailed and sideinsensitive acquisition sequence. Adherence to standard but unspecified digital mammography protocols was declared by 29/84 (34%) studies, for total 3741/14012 patients (27%). The other half of the articles analysed (42/84, accounting for 7434/14012 patients, 53%) unequivocally detailed an acquisition sequence. Of these 42 studies, 14 (34%, 2048/7434 patients, 28%) adopted a projection order that was conventionally agreed upon, while the other 28 (66%, accounting for 5386/7434 patients, 72%) based their acquisition sequence on the presence of previous suspect or clearly pathologic findings.
Eighty-four articles came from 38 different research groups. Subgroup analysis according to research groups showed that 17 acquisition sequences based on a conventionally agreed projection order were executed in 15 research groups. As described in Fig. 3, the most common sequence description, reported by 6/17 (35%) institutions, was MLO -MLO -CC -CC (in order of acquisition), without any further indication about the first side to be examined (right or left or side with/without suspicious lesion or already diagnosed cancer). The second most common sequence (4/17, 24%) was CC -CC -MLO -MLO with the first projection standardised on the right side (independently of pathology or with suspected pathology).
Among the 22 acquisition sequences (coming from 20 institutions) centred on the presence of previous suspect or clearly pathologic findings, we found substantial variability between different orders of acquisition, as shown in Fig. 4. However, the most common sequence, adopted by 4/22 (19%) research groups, was 1) CC, suspected side; 2) CC, non-suspected side; 3) MLO, suspected side; and 4) MLO, non-suspected side.

Contrast agent adverse reaction rate meta-analysis
Regarding side effects from ICA administration, 48/84 studies (57%) declared a preventive anamnestic screening for previous adverse reactions or general contraindications to ICA administration. Pre-examination tests of renal function was mentioned in 39/84 studies (46%). Of note, 14/84 studies (29%) reported 30 adverse reactions out of 14012 patients, of which 26/30 (87%) were mild reactions limited to pruritus, hives, "scratchy throat" or other minor skin flushing that resolved promptly even when antihistamines or corticosteroids were not administered. In 3/30 (10%) cases [54,58,87], side effects were of moderate importance with nausea and vomiting, widespread urticaria resolved only after antihistamines and corticosteroids per os, and dyspnea that equally responded to oral antihistamine administration. Only 1/30 (3%) severe adverse reaction, requiring "intensive care" but resolved after short time, occurred in 14012 patients (0.007%) [61].
Therefore, the number of adverse reactions related to ICA administration ranged from 0, reported by 70 (88%) studies, to a maximum of 6 adverse reactions [14] with a total of 30 adverse reactions, showing no heterogeneity (Q = 64, degree of freedom 83, τ = 2.0972, I 2 = 0%, p = 0.931). As shown in the forest plot of Fig. 5, using fixedeffect model, the pooled rate of adverse reactions across studies was 0.82%, with 0.64% and 1.05% as 95% CI.
Visually inspecting the funnel plot in Fig. 6, risk of publication bias was found, as confirmed by the Egger test (p = 0.00028).

Discussion
Our systematic review included 84 articles, accounting for 14012 patients, reporting the use of CESM in various settings. The sheer number of studies and, as depicted in A number of narrative reviews [6,42,[102][103][104][105][106] favourably outlined CESM future perspectives in several clinical settings (e.g. recall work-up, pre-operative staging, and monitoring the effect of neoadjuvant therapy) as a potential alternative to MRI.
In the first phase of CESM development, some nonfixed parameters regarding contrast agent administration (i.e. contrast agent molecule, concentration, dose, flow rate, and injection modality) and some acquisition features (i.e. time between contrast injection and first acquisition, kVp ranges for low-and high-energy acquisitions) gained an international agreement. However, in the framework of comprehensive optimisation and standardisation of CESM, large-scale studies are undoubtedly needed to address the knowledge gap concerning the choice of technical parameters.
Our data show a consensus among studies (93%) on the choice of 1.5 mL/kg contrast dose administered with a 3 mL/s flow rate (74%) and a less extensive agreement on the use of Iohexol (53% of all studies) at a concentration of 350 mg iodine/mL (30% of all studies). However, these parameters have probably been empirically adopted from CT protocols, as the first investigators plainly stated [7], without any other particular explication or justification. No dose-finding studies have been published yet.
Similarly, the common use of a power contrast injector (87% of all studies, with the remaining 13% coming from a single research group) is assumed from CT and MRI protocols in which it has been demonstrated to be effective in obtaining a stable contrast inflow and bolus shape [107][108][109]. Moreover, the use of a power injector allows for the administration of a bolus chaser, reported only in 42% of all articles, a technical refinement that has shown good results in CT [110,111].
Two other points need to be mentioned. The first one is the correlation between menstrual cycle phase and background parenchymal enhancement, explored in a few studies [10,75,80] and/or fluctuations of lesion contrast uptake. Secondly, since CESM is based on a dual X-ray exposure, of which the low-energy one has been demonstrated to be equal to standard DM [66], an increase in radiation dose is expected. However, while preliminary studies estimated a negligible [7] or curtailed   AGD increase, studies specifically devised to ascertain CESM effective AGD found a substantial AGD increment ranging 42-80% [56,60,82]. While CESM AGDs remain under the threshold stated by European guidelines for screening mammography [112], further studies are needed to investigate CESM AGD [56,82]. Furthermore, we remark the absence of standardised protocols. This methodological void, especially regarding the acquisition workflow, represents a threat to reproducibility and comparison of imaging results. While 98% of all studies reporting the total examination time completed the examination before 10 min from contrast administration, and while some studies presented evidence on the irrelevance of the acquisition order [55,64], there are no studies comparing different approaches.
The pooled rate of adverse reactions to ICA administration was 0.82% (0.64-1.05% 95% CI) with a total of 30 adverse reactions in 14012 patients, a rate similar to that reported for CT 0.6% [113] in 84928 adult patients or 0.7% [114] in 29508 patients (given Iopromide, which is also used for CESM). Particularly, considering only severe adverse reactions in CT, Wang et al. [113] reported 11/84928 (0.0129%) reactions, as well as Mortelé et al. [114] 4/29508 (0.0135%). These rates seem to be higher than that found in our meta-analysis 1/14012 (0.007%), a comparison to consider with caution due to the nature of rare events such as severe reactions to ICA. One aspect to consider is the different profile of patients undergoing CESM compared to those requiring contrastenhanced CT, the former being that of basically "healthy" subjects, the latter implying the possibility of relevant disease, including also serious emergency conditions. This review has limitations. Patient data are probably shared and duplicate among some studies from the same research group. This has been shown to negatively impact on review quality [115,116] and could only be prevented via individual patient data sharing [117]. However, for technical aspects of this systematic review, our choice to evaluate study groups rather than single articles should have mitigated this bias. Conversely, our pooled rate of adverse reactions could be underestimated.
In conclusion, our review shows that CESM is unevenly performed across different centres, in terms of contrast agent type and concentration and order of view acquisition. However, most research groups performed CESM using a contrast dose of 1.5 mL/kg, factory-set kVp ranges for low-and high-energy acquisitions, beginning image acquisition after 120 s from contrast agent injection and completing the examination within 10 min. Further studies are needed to investigate the role of background parenchymal enhancement and to harvest data that can firmly back up subsequent technical guidelines and consensus statements for standardised CESM protocols. Authors' contributions Each author has participated sufficiently in this work to take public responsibility for its content. The manuscript is approved by all authors and by the responsible authorities. All authors read and approved the final manuscript.

Funding
This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.

Availability of data and materials
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.
Ethics approval and consent to participate Not applicable

Consent for publication Not applicable
Competing interests FS declares to have received grants from or to be member of speakers' bureau/advisory board for Bayer, Bracco, and General Electric. All other authors declare that they have no competing interests.