As shown in PRISMA flowchart (Fig. 1), The initial search resulted in 23 studies. Two more articles were added through other sources (reference check). After initial screening and full-text assessment with regard to inclusion and exclusion criteria, 7 studies included in the quantitative analysis(6–8, 10, 16).
Study Demographics And Characteristics
Table 1 presents the data extraction from the seven included studies. The mean age of participants with dense breasts in the studies was ranging from 35 to 83 years.
Of the seven included studies, one was prospective(6), and six of the seven studies were retrospective(7–10, 16, 17).
Risk assessment was based on BRCAPRO in Kuhl et al. study (6); in the study of Weinstein et al, patients who acknowledged having a lifetime risk of < 20% based on any risk assessment model were considered average risk (7). Comstock et al. enrolled average risk patients based on the BCSC risk calculator (10). Kennard et al and Chen et al (2022) included patients with mixed risk(9, 17), the two remaining studies did not mention anything about the risk assessment(8, 16).
Four studies (6, 8, 16, 17) included patients with normal or benign mammography results. Weinstein Comstock and Kennard did not mention negative or normal mammography as an inclusion criterion(7, 9, 10) and relied on normal clinical assessments. Six studies assessed breast density based on ACR BIRADS criteria (6–8, 10, 16, 17). Kennard did not mention the ACR level and mentioned that patients with “dense breast tissue” were included(9).
Table 1
Demographics, inclusion and exclusion details of the five full-text articles.
Study name | Study design | Risk | inclusion | Dense breast assessment | exclusion | Number of included patients with dense breast | Age | Follow-up duration |
Initial mammographic study | density |
Kuhl 2014 (6) | Prospective | Mild to moderate risk (BRCAPRO) | 1. screening 2. history of breast cancer to screen the contralateral breast; 3. a normal clinical examination 4. normal or benign digital mammogram 5. negative or benign ultrasound study for dense breast | 1.Normal or benign digital mammography and, | Heterogeneously or extremely dense breasts (ACR 3 to ACR 4) (who had normal or benign ultrasound study) | Affected breast in women with history of breast cancer | 427 | 54.2 (25–73) | 2-year validation period |
Weinstein 2020 (7) | Retrospective | Average risk | 1. asymptomatic women with heterogeneously or extremely dense breasts 2. negative or benign (BI-RADS 1 or 2) final assessment44 on their mammogram obtained within the prior 11 months. 3. women who underwent DBT imaging. 4. patients who acknowledged having lifetime risk >. 20% | negative or benign tomosynthesis | Heterogeneously dense or extremely dense (ACR) | 1. nondiagnostic clincial examinations 2. patients who did not get gadolinium; 3.patient with motion during AB-MRI 3. Patients with scattered fibroglandular density 4. Patients who had prior 2D/DM mammogram 5. Patients with prior mammogram > 11 months ago | 511 | 58 (32–83) | 1 year of follow-up |
Chen 2017 (abb) (16) | retrospective | N/S | dense breast tissue and negative mammography results | and negative mammography results | dense breast tissue (ACR) | N/S | 356 | 48.2 ± 4.7 (30–75) | followed for 2 years with MG or US |
Chen 2017 (app) (8) | retrospective | N/S | dense breast tissue and negative mammography results | Negative mammography | Dense breast tissue (ACR) | N/S | 478 | 49.3 (30–71) | N/S |
Comstock 2020 (10) | retrospective study with longitudinal follow-up | Average risk (Breast Cancer Surveillance Consortium (BCSC) risk calculator) | 1. clinically asymptomatic women 2. dense breasts on last mammogram. 3. a history of benign breast biopsy 4. remote history of treated breast cancer 5. family history of breast cancer | N/S | Dense breast (type of criteria is not reported) | 1. screening breast ultrasound within the past 12 Months 2. ever had a breast-MRI, a molecular breast imaging study, or a contrast-enhanced Mammogram 3. qualify for full-protocol breast-MRI based on American Cancer Society guidelines | 1327 | 54.9 ± 8.5 (40–75) | 11–13 months |
Chen 2022 | retrospective | consecutive women with mixed risk | women with dense breast and negative or benign fndings on mammography | Negative or benign mammography | Dense breast tissue (ACR) | 1. patients did not undergo mammography 2.women with breasts consisting of scattered fbroglandular tissue and almost entirely fat observed by Mammography 3. women who had been received neoadjuvant chemotherapy 4.women without reference standard 5. BI-RADS categories 4 and 5 on mammography | 102 | 48.2(30–73) | 2 years |
Kennard 2022 | retrospective | Any risk | 1. be up to date with screening mammography per United States Preventative Services Task Force Screening recommendations, 2. greater than 6 months from any prior MRI 3. greater than 2 years from any breast cancer diagnosis. | Negative, benign and positive mammography | Dense breast (type of criteria is not reported) | 1. pregnancy 2. current diagnosis of breast cancer | 83 | 52.2 ± 10.1 | At least 6 months |
N/S = not stated |
IMAGING SPECIFICATION AND INTERPRETATION
As shown in Table 2, magnetic strength for Kuhl et al. and Weinstein et al. study was 1.5 T with axial orientation. (6, 7) Chen et al. 2017 studies and Chen 2022 performed the MRIs with a 3T magnetic strength and axial orientation(8, 16). Comstock et al. and Kennard et al. did not report the strength of their MRIs(9, 10). Of the seven studies, 4 mentioned the Time to perform. Kuhl et al. performed the MRIs in 184 seconds, Chen (2022) 3 minutes, Kennard et al. <20 minutes and Comstock et al. performed the MRIs in less than 10 minutes. Kuhl et al. and Chen et al. (app) interpreted the AB-MRI in 28 sec (20–68 seconds) and 42 ± 18 seconds, respectively(6, 8). Chen et al (abb) read the AP-1 in 37 seconds and AP-2 in 54 seconds(16). Chen et al. (2022) interpreted the ABB-MRI results in 33.4 seconds (17.8–53.9 s).
Abbreviated MRI protocol in Kuhl, Chen 2017(app) and Chen 2022 studies was assessed in FAST + MIP phases(6, 8). Weinstein et al. study assessed the AP in Short-TI Inversion Recovery (STIR), pre-contrast T1 fat-suppressed, and one postcontrast sequence(7). In Chen 2017(abb) study(16), AP was assessed in two part; AP-1 ( FAST and MPI), AP-2 ( FAST + MIP + DWI ). Chen 2017 (app) and Chen 2022 used the AP (FAST + MIP)(8, 17). Comstock and Kennard study used a T2-weighted acquisition, and a T1-weighted acquisition before and after bolus injection of contrast(9, 10).
In the six studies, at least two expert radiologists with a minimum of 6 years of experience in breast cancer MRI had interpreted the abbreviated MRI based on FULL BIRADS criteria (6–8, 10, 16, 17). Kennard did not mention the experience of radiologists(9). Except for Weinstein and Kennard studies (did not perform full protocol MRI), all the other studies mentioned that the readers were blind to full protocol results(6–10, 16, 17). Since in all studies the biopsies are performed after the AB-MRI studies, the MRI readers all were unaware of reference standard results.
Table 2
Abbreviated magnetic resonance imaging and interpretation specifications
Study name | Magnet strength (T) | Orientation | Time to perform | Time to read for AB-MRI | Protocol phases | Readers blind to full protocol results | interpretation | expertise |
Kuhl 2014 (6) | 1.5 | Axial | 184 seconds | 28 seconds (range, 20 to 68 seconds) | FAST + MIP | Yes | Full BI-RADS assesment | Two readers (18 and 6 years) |
Weinstein 2020 (7) | 1.5 | Axial | N/S | N/S | Short-TI Inversion Recovery (STIR), precontrast T1 fat-suppressed, and one postcontrast sequence | N/S | BI-RADS 5th edition and a negative or benign (BI-RADS 1 or 2) final assessment on their mammogram obtained within the prior 11 months | Three readers ( ranging from 11–30 years (mean 21 year)) |
Chen 2017 (abb) (16) | 3 | Axial | N/S | AP-1 = 37 [22–56] seconds AP-2 = 54[39–77] seconds | AP-1 (FAST and MIP AP-2 (FAST + MIP + DWI) | Yes | BI-RADS MRI lexicon, and BI-RADS final assessment categories 1, 2, or 3 were considered MRI negativity, and categories 4 or 5 were considered MRI positivity | three radiologists’ (third senior radiologist with 15 years of breast MRI experience) |
Chen 2017 (app) (8) | 3 | Axial | N/S | 42 ± 18 seconds | AP (FAST + MIP) | Yes | BIRADS MRI CATEGORIES | Two senior radiologists > 10 years |
Comstock 2020 (10) | N/S | N/S | < 10 minutes | N/S | T2-weighted acquisition, and a T1-weighted acquisition before and after bolus injection of contrast | Yes | BIRADS CATEGORIES | two different board certified breast radiologists (experience is not mentioned) |
Chen 2022 | 3 | Axial | 3 mins | 33.4 s (17.8–53.9 s) | AP (FAST + MIP) | Yes | BI-RADS CATEGORIES | two breast radiologists with 9 and 20 years of experience in breast MRI |
Kennard 2022 | N/S | N/S | < 20 minutes | N/S | T2-weighted acquisition and a T1-weighted acquisition before and after bolus injection of contrast | Full protocol was not performed | BI-RADS CATEGORIES | multiple radiologists (70% of whom are breast radiology fellowship trained) |
N/S = not stated |
DIAGNOSTIC ACCURACY OF AB-MRI
Table 3 provides the diagnostic accuracy parameters derived from included studies. The sensitivity of AB-MRI ranged from 92.9% (16) to 100% (6, 7, 9). The specificity ranged from 86.5% (17) to 94.4% (6). The biopsy was the main reference standard in all studies. The results from Kuhl study are based on only first round of screening(6).
Table 3
Diagnostic accuracy of abbreviated breast magnetic resonance imaging in participants with dense breast in patients entered the analysis
Study name | Number of Cases with dense breast | Number of AB-MRI scans | Reference standard | TP | FP | TN | FN | Sensitivity% | Specificity | PPV(BIRADS 4–5) | NPV |
Kuhl 2014 (6) (FAST) | 443** | 606 | Biopsy/ short-term MRI/two years follow-up | 11 | 34 | 560 | 0 | 100 | 94.29 | 31.4 | 100 |
Weinstein 2020 (7) | 475 | N/S | Biopsy/1 year follow-up | 12 | 27 | 270 | 0 | 100 | 90.9 | 30.8 | 100 |
Chen 2017 (abb) (16) | 356 | N/S | Biopsy/ two year follow-up | 13 | 46 | 296 | 1 | 92.9 | 86.5 | 22.0 | 99.7 |
Chen 2017 (app) (8) | 478 | N/S | Biopsy | 15 | 24 | 181 | 1 | 93.8 | 88.3 | 21.7 | 99.8 |
Comstock 2020 (10) | 1444 | N/S | Biopsy/11–13 months follow-up | 22 | 187 | 1220 | 1 | 95.7 | 86.7 | 31.4 | 99.7 |
Chen 2022 reader 1 | 102 | N/S | Biopsy/ two year follow-up | 65 | 8 | 32 | 3 | 96 | 80 | 89 | 91 |
Chen 2022 reader 2 | 64 | 9 | 31 | 4 | 94 | 78 | 87 | 88 |
Kennard 2022 | 83 | 93 | Biopsy/ follow-up | 1 | 12 | 80 | 0 | 100 | 87 | 87.1 | 100 |
*, number of cancers divided by the total number of biopsies performed (BI-RADS 4 and 5). **first round of screening with no recall N/S = not stated |
QUALITY ASSESSMENT
Patients' selection, the index test, the reference standard, and the study's flow and timing were identified as the primary sources of bias.
All studies had low-risk regarding all causes of concerns in bias and applicability.
META-ANALYSIS
As shown in Figs. 2 and 3, a meta-analysis was conducted to assess the accuracy of abbreviated MRI protocol (index test) screening of women with dense breasts using histology or follow-up as the reference standard. The results show the forest plot and SROC for sensitivity and specificity of included studies besides summary points with 95% interval. This meta-analysis included 7 studies (plus one reader) that provided Sensitivity, Specificity, and Confidence Interval data. The pooled overall accuracy measures for all included studies were sensitivity = 0.96 [0.91– 0.98] and specificity = 0.88 [0.85–0.91].