Institutional review board approval was obtained for this single-center study and informed consent was waived due to its retrospective nature (No. 159/2020 BO2). The study followed the regulations of the declaration of Helsinki.

From January 2016 to September 2019, a total of n = 129 female patients who underwent clinically indicated CESM were retrospectively included in our study. All subjects were at least 18 years old and not pregnant or lactating. Sixty-seven patients had suspicious findings (BI-RADS 4 or 5), identified by mammography, B-mode ultrasound, or both. The other patients had unclear findings that needed further evaluation with contrast enhanced imaging and breast MRI could not be performed, or received the examination as part of a high-risk screening, when there were contraindications to obtain breast MRI.

CESM images were acquired by using Senographe EssentialTM (GE Healthcare), corresponding to machine A and Selenia ® Dimensions ® (Hologic), corresponding to machine B. Both mammography machines were used simultaneously in this time period in our radiology department. All patients received 1.5 ml/kg body weight iodine-based contrast agents, injected intravenously with a flow rate of 3ml per second. One hundred five patients received iopromid (Ultravist® 300, Bayer), 22 patients iomeprol (Imeron® 350, Bracco). For two patients, the applied contrast agent was not documented. The first image was obtained 2 min after the application of the contrast agents; all images were acquired within less than 5 min. Images of machine A were acquired with 22–49 kV while images with image B were acquired with 20–49 kV. In premenopausal patients, the time of examination was adapted to the menstrual cycle, and images were acquired between the 7th and 14th day of the menstrual cycle. If a timing was not possible (e.g., due to hysterectomy, contraceptives or irregular menstrual cycle), no progesterone levels were determined and the examination was performed without an exact timing. In case of a suspicion of cancer or biopsy-proven breast cancer, a timing of the examination to the menstrual cycle was not performed.

For image evaluation, a dedicated workstation (Centricity RA1000, General Electric) was used. Firstly, all images were analyzed qualitatively by two radiologists with a minimum of 2 years of experience in breast imaging and afterwards verified by one senior radiologist with 10 years of experience in breast imaging. Images were categorized by an overall background enhancement by using a numbered scale of 1 = minimal, 2 = mild, 3 = moderate, and 4 = marked [25]. In event of a deviation, a consensus reading had to be performed. The qualitative analysis also considered the extent to which the suspicious findings could be delimited as such on the basis of the contrast medium uptake. In the case of poor assessability, a higher level of background enhancement was chosen in cases of doubt. Secondly, quantitative image analysis was performed by placing an oval shaped ROI into the image, inspired by the methodology of a recent published study [7]. The ROI was set to include the most applicable extent of the breast. The correct assessment of the ROI was confirmed by a senior radiologist. The ROI results show a dimensionless value that measures and thus quantifies the intensity enhancement within the acquired image averaged over the surface. Each ROI had a minimum size of 3 cm2. The assessment of the ROI was performed under supervision of a senior radiologist with 10 years of experience in breast imaging to guarantee the exclusion of pathological findings and artifacts. The artifacts in CESM images can create a brighter, but also a darker image impression. Air trapping artifacts, motion artifacts, and negative contrast enhancement lead to a dark image impression, for example because in air trapping artifacts the contact between the breast tissue and the detector is incomplete, whereas rim artifacts produce an arc-shaped subcutaneous area with a brighter image impression [26, 27]. These differences in brightness are measurable within the ROI and would thus distort the measured mean values. Therefore, known image artifacts and suspicious findings were not included in the ROI. Each ROI was placed three times and the average values of maximum, minimum, mean, and standard deviation were determined within the ROI, as described previously [7, 28].

Statistical analysis was performed using statistical programs (MedCalc Statistical Software version 18.10 (MedCalc Software bvba, http://www.medcalc.org; 2018) and jmp15, MP®, Version 15 SAS Institute Inc1989–2019.). Quantitative data were tested for normal distribution using the Kolmogorov-Smirnov test.

A Mann-Whitney U test was performed to compare non-normally distributed data. C Interrater-reliability of ordinal data was tested using Cohen’s kappa with values ≤ 0 as indicating no agreement and 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial, and 0.81–1.00 as almost perfect agreement [29].

Interrater agreement for quantitative data was tested using the intraclass correlation coefficient (ICC) with values ≤ 0.5 indicating a poor, 0.5–0.75 a moderate, 0.75–0.90 a good, and > 0.9 an excellent agreement [30]. Hodges Lehmann estimation was performed to test for average differences.

We performed a multivariate regression analysis to test for potential biases influencing background enhancement. The factors machine, contrast agent, breast density (ACR A-D), age, and menopausal status were included in the analysis. The menopausal status was analyzed using a Likert scale from 1 to 5 (1, premenopausal; 2, postmenopausal; 3, examination not adapted to the menstrual cycle; 4, irregular menstrual cycle or no menstruation due to contraception).

To test whether the results could have a potential impact on diagnostic accuracy, all images with histologically confirmed pathological findings were assessed for lesion conspicuity. CESM images were rated using a Likert scale from 1 to 5 (1 pathology not reliably distinguishable from background enhancement; 2, poor lesion conspicuity; 3, moderate lesion conspicuity; 4, good lesion conspicuity; 5 excellent lesion conspicuity). The Mann-Whitney U test was used to test the results between both machines for differences.