MRI Screening of BRCA Mutation Carriers: Comparison of Standard Protocol and Abbreviated Protocols With and Without T2-Weighted Images

Abstract

BACKGROUND.

Increasing evidence supports the role of abbreviated MRI protocols for breast cancer detection. However, abbreviated protocols have been poorly studied in patients who are BRCA1 or BRCA2 mutation carriers. Furthermore, the need for T2-weighted sequences in abbreviated protocols remains controversial.

OBJECTIVE.

The purpose of this study was to compare, in the evaluation of patients with BRCA mutations, the diagnostic performance of a standard full breast MRI protocol with the performance of abbreviated protocols that included and did not include a T2-weighted sequence.

METHODS.

This retrospective study included 292 patients (mean age, 47.9 years) who were BRCA1 or BRCA2 mutation carriers who underwent 427 screening breast MRI examinations according to a standard full protocol who could be classified as having benign (n = 407) or malignant (n = 20) findings based on histopathology or imaging follow-up. Four readers independently assessed examinations in three separate sessions (theoretic abbreviated protocol, which included the first postcontrast acquisition; theoretic abbreviated protocol with addition of a T2-weighted sequence; and the standard full protocol) and assigned BI-RADS categories. Categories 3–5 were considered to represent positive examinations. Interreader agreement was assessed, and diagnostic performance was compared by use of pooled reader data.

RESULTS.

Interreader agreement on BI-RADS category, expressed as kappa values, was 0.55 for the standard, 0.45 for the abbreviated, and 0.57 for the abbreviated plus T2-weighted protocols. Pooled sensitivity was 94% for the standard, 92% for the abbreviated, and 90% for the abbreviated plus T2-weighted protocols (all p > .001). Pooled specificity was 80% for the standard, 71% for the abbreviated, and 83% for the abbreviated plus T2-weighted protocols (p < .001 for abbreviated plus T2-weighted compared with both standard and abbreviated). Pooled PPV was 19% for the standard, 14% for the abbreviated, and 20% for the abbreviated plus T2-weighted protocols (p < .001 for abbreviated compared with both standard and abbreviated). Pooled NPV was 100% for the standard, 99% for the abbreviated, and 99% for the abbreviated plus T2-weighted (all p > .001) protocols. Pooled accuracy was 80% for the standard, 73% for the abbreviated, and 83% for the abbreviated plus T2-weighted protocols (p < .001 for abbreviated compared with both standard and abbreviated plus T2-weighted).

CONCLUSION.

The abbreviated protocol without T2-weighted imaging had suboptimal performance. However, addition of the T2-weighted sequence yielded comparable sensitivity and accuracy and a small increase in specificity compared with the full protocol.

CLINICAL IMPACT.

The findings support implementation of abbreviated MRI with T2-weighted imaging for breast cancer screening of patients with BRCA mutations.

Women are deemed at high risk of breast cancer if they have a greater than 20% cumulative lifetime risk of a breast cancer diagnosis [1]. Among women at high risk, the lifetime risk is as high as 70% among BRCA mutation carriers [2], who often present with aggressive cancers at a young age [3]. Since identification of the BRCA1 and BRCA2 genes, routine clinical genetic risk assessment, including testing for these genes, has become widely adopted, such that the population of women considered at greater than average risk of breast cancer is expanding. For these patients, early detection, when the disease is amenable to curative treatment, is pivotal.

Breast MRI is the most sensitive imaging modality for breast cancer detection, as evidenced by data from the Dense Tissue and Early Breast Neoplasm Screening (DENSE) [4] and Eastern Cooperative Oncology Group–American College of Radiology Imaging Network (ECOG-ACRIN) EA1141 [5] trials. The recommendation for patients at high risk is to begin annual mammography at age 30 and annual screening MRI at age 25, depending on the type of genetic mutation and the youngest age at breast cancer onset among first-degree relatives [1, 6, 7]. The recommendation for annual supplemental screening MRI of patients at high risk is supported by a bias toward biologically less aggressive cancers among cancers detected with mammography than cancers detected with breast MRI, resulting in high mortality and interval cancer rates if patients at high risk undergo screening with mammography alone [8].

Despite the recommendation for patients at high risk, the widespread use of MRI for screening has limitations, including high cost, long acquisition time that impedes high-volume patient throughput, and the need for IV administration of gadolinium-based contrast agents. Abbreviated breast MRI protocols have been introduced to address these limitations. The aim of abbreviated MRI is to reduce acquisition and interpretation times by including only the sequences needed to maintain high sensitivity for breast cancer detection. Nonetheless, controversy persists regarding which sequences are warranted in abbreviated protocols to maintain the specificity of and diagnostic confidence in the examination. A particular question is the need for T2-weighted sequences. Studies [8–13] have shown the utility of abbreviated MRI for reducing cost and examination times while maintaining breast cancer detection in patients not at high risk. However, data supporting abbreviated breast MRI for patients at high risk, including data specifically regarding BRCA mutation carriers, remain scarce [14, 15]. We conducted this study to compare the diagnostic performance of a standard full breast MRI protocol and of abbreviated protocols with and without T2-weighted sequences in patients with BRCA mutations.

Methods

Patient Selection

This HIPAA-compliant retrospective study was approved by the institutional review board at Memorial Sloan Kettering Cancer Center. The requirement for informed consent was waived. An institutional database was queried to identify women who were carriers of either the BRCA1 or BRCA2 mutation who underwent screening breast MRI with a standard full protocol from January 2014 to October 2015. The presence of BRCA1 and BRCA2 mutations was determined by commercially available genetic panels. The initial search did not identify patients who underwent MRI examinations for other reasons (e.g., extent of disease, surgical planning, chemotherapy treatment monitoring) but did identify patients with a personal history of breast cancer treated by lumpectomy or by unilateral or bilateral mastectomy.

The initial search yielded 1700 screening breast MRI examinations of BRCA1 or BRCA2 carriers. Further patient selection was based on the clinically reported BI-RADS categories. Examinations categorized BI-RADS 1 or 2 required at least 1 year of benign imaging follow-up; examinations categorized BI-RADS 3 required either 2 years of benign imaging follow-up or histopathologic diagnosis; examinations categorized BI-RADS 4 or 5 required histopathologic diagnosis.

A total of 1273 examinations that did not meet the required follow-up for the clinically assigned BI-RADS category were excluded. This process resulted in a final sample of 427 screening MRI examinations of 292 patients (all women; mean age, 47.9 years; range, 23–78 years) with BRCA mutations. The patient selection flowchart is shown in Figure 1. Among the 427 examinations, the clinically assigned category was BI-RADS 1 in 38.4% (164), BI-RADS 2 in 47.1% (201), BI-RADS 3 in 4.0% (17), BI-RADS 4 in 10.3% (44), and BI-RADS 5 in 0.2% (1). The clinically assigned BI-RADS categories were used for initial patient selection but otherwise were not incorporated in the subsequent study execution and data analysis. Prior screening MRI for comparison was available at clinical interpretation of 71.2% (304) of examinations.

Fig. 1—

Flowchart shows patient selection.

Breast MRI Examinations

All breast MRI examinations were performed with a 3-T system (Discovery MR750, GE Healthcare) with a dedicated 16-channel phased-array breast coil (Vanguard, Sentinelle). The full protocol included an axial fat-suppressed T2-weighted fast spin-echo sequence (TR/TE, 4384/102; matrix size, 288 × 224; FOV, 30 × 30 cm; slice thickness, 3 mm; number of excitations, 2; flip angle, 111°; acquisition time, ≈ 3–4 minutes), an axial non–fat-suppressed T1-weighted fast spin-echo sequence (TR/TE 8.5/3.0; matrix size, 288 × 224; FOV, 30 cm²; slice thickness, 3 mm; number of excitations, 2; flip angle, 111°; acquisition time, ≈ 3–4 minutes), and a 3D T1-weighted fat-suppressed gradient-echo sequence (volume imaging for breast assessment [VIBRANT]; TR/TE, 4.3/2.1; matrix size, 320 × 192; FOV, 30 cm²; slice thickness, 1 mm; number of excitations, 1; flip angle, 10°; acquisition time, ≈ 1–2 minutes) performed before and after IV injection of 0.1 mmol/kg gadopentetate dimeglumine (Magnevist, Bayer HealthCare) with three postcontrast acquisitions at 60-second intervals. The postcontrast sequences were further postprocessed with built-in software to generate subtraction and maximum-intensity-projection (MIP) images.

Breast MRI Protocols

For each examination, three separate image sets were constructed, representing three potential breast MRI protocols for comparison. The standard full protocol (hereafter called standard) included a fat-suppressed T2-weighted sequence, a non–fat-suppressed T1-weighted sequence, and precontrast and postcontrast fat-suppressed T1-weighted sequences (including three postcontrast acquisitions at 60-second intervals) and postprocessed subtraction and MIP reconstructions from the postcontrast sequences. In addition, two theoretic abbreviated protocols were constructed on the basis of descriptions in the literature [16]. The first abbreviated protocol (hereafter called abbreviated) included a precontrast fat-suppressed T1-weighted sequence and subtracted and MIP images from the first precontrast fat-suppressed T1-weighted acquisition. The second abbreviated protocol (hereafter called abbreviated plus T2-weighted) comprised all sequences in the abbreviated protocol and a fat-suppressed T2-weighted sequence.

Patients did not undergo abbreviated MRI for purposes of this study; rather, the acquired standard full protocol was used to generate the image sets for the two theoretic abbreviated protocols for retrospective interpretation as part of this investigation. The approximate acquisition time for the standard protocol was 16 minutes. The anticipated approximate acquisition time for the abbreviated protocol was 4 minutes and for the abbreviated plus T2-weighted protocol, 8 minutes. The total examination time for the standard protocol was 30–40 minutes. The anticipated examination time for the abbreviated protocol was 10–15 minutes and for the abbreviated plus T2-weighted protocol was 15–20 minutes. Table 1 summarizes the three breast MRI protocols compared in this study.

TABLE 1:

Comparison of Breast MRI Protocols

Characteristic	Standard	Abbreviated	Abbreviated Plus T2-Weighted
Sequences and image sets	Fat-suppressed T2-weighted fast spin-echo (3–4 min) Non-fat-suppressed T1-weighted fast spin-echo (3–4 min) Precontrast fat-suppressed T1-weighted gradient-echo VIBRANT (1–2 min) Postcontrast fat-suppressed T1-weighted gradient-echo VIBRANT (three acquisitions at 60-s intervals; 1–2 min per acquisition) Subtraction images MIP images	Fat-suppressed T1-weighted gradient-echo VIBRANT (1–2 min) First postcontrast fat-suppressed T1-weighted gradient-echo VIBRANT (1–2 min) Subtraction images MIP images	Fat-suppressed T2-weighted fast spin-echo (3–4 min) Fat-suppressed T1-weighted gradient-echo VIBRANT (1–2 min) First postcontrast fat-suppressed T1-weighted gradient-echo VIBRANT (1–2 min) Subtraction images MIP images
Plane of all sequences	Axial	Axial	Axial
Approximate image acquisition time (min)	16	4	8
Total examination time (min)	30–40	10–15	15–20

Note—VIBRANT = volume imaging for breast assessment, MIP = maximum intensity projection.

Breast MRI Analysis

Four breast imaging fellowship-trained radiologists (K.P. with 15, I.D.N. with 5, C.S. with 5, and J.S. with 2 years of experience) independently reviewed the image sets acquired with the three protocols; the reviews were performed in separate sessions separated by at least 21 days. One protocol was reviewed per session (abbreviated protocol in session 1, abbreviated plus T2-weighted protocol in session 2, standard protocol in session 3); examinations were allocated randomly in each section. Examinations were interpreted with a Centricity Viewer system (GE Healthcare). The readers were instructed to view only the images relevant to the given protocol in each session. The readers were aware that all patients were BRCA mutation carriers but otherwise were blinded to clinical information (including clinically assigned BI-RADS categories) and prior imaging examinations. Before the first session, the readers jointly reviewed several cases not included in the study to establish a standard reading workflow to be used by all readers.

In interpreting each examination, the readers first evaluated background parenchymal enhancement using the subtracted and MIP postcontrast images for all three protocols. They also used the amount of fibroglandular tissue on non–fat-suppressed T1-weighted images for the standard protocol and fat-suppressed T1-weighted precontrast images for the two abbreviated protocols. The readers then reviewed all of the images available for the given protocol to assign a BI-RADS category and a probability of malignancy (1–100%). When assigning probabilities of malignancy, the readers were guided by probabilities in the BI-RADS descriptors but were permitted to assign any percentage for the individual case at their discretion (i.e., a reader could assign different percentages for two lesions assigned the same BI-RADS category). When assigning BI-RADS category 3, 4 or 5, the readers also described the location, laterality, clock-face position (1–12 o’clock), and depth (anterior, middle, or posterior) of the lesion. Readers were not permitted to assign BI-RADS category 0.

The BI-RADS assessments were based on a single lesion, the largest and most suspicious enhancing lesion detected on subtraction images. Readers were not permitted to report multiple lesions. All assessments were based on the American College of Radiology BI-RADS atlas, 5th edition [17]. The readers manually recorded interpretation time as the time from initial display of the images on the screen to when the reader closed the display of the examination. The interpretation time thus reflected the time required to detect and characterize enhancing lesions. Because reports were not generated, the interpretation times did not include potential time for dictation or report completion.

Statistical Analysis

Characteristics of the study sample were summarized descriptively. Variables were summarized with medians and interquartile ranges or means and ranges. Continuous variables were compared by Wilcoxon rank sum test. Imaging and histopathologic follow-up were used for the reference standard of benign or malignant for all examinations. Probabilities of malignancy between examinations were compared with benign and malignant results according to the reference standard. Examinations assigned BI-RADS categories 1 and 2 were considered negative, and those assigned BI-RADS categories 3, 4, and 5 were considered positive.

The sensitivity, specificity, PPV, NPV, and accuracy of each protocol for breast cancer detection were computed on the basis of the classifications of examinations as positive or negative. Associated 95% CIs were estimated by the Clopper-Pearson method for binomial proportions [18]. Sensitivities, specificities, and accuracies were compared between protocols by exact McNemar test, and PPV and NPV were compared by the generalized score test proposed by Leisenring et al. [19]. These comparisons accounted for correlations among multiple examinations of individual patients. By use of the assigned probabilities of malignancy, ROC analysis was performed to determine the AUC of the three protocols, which were compared by DeLong test for correlated ROCs. Diagnostic performance was also assessed for pooled data among all four readers. Given the large number of comparisons, comparisons were considered statistically significant at p < .001.

Interreader agreement on BI-RADS category (categories 1 and 2 considered negative and categories 3, 4, and 5 considered positive) was calculated with Fleiss kappa coefficients. Kappa coefficients were interpreted with the definitions of Landis and Koch [20]: 0.00–0.20, slight agreement; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, substantial; 0.81–1.00, almost perfect. Research Electronic Data Capture Software (REDCap, REDCap Consortium) [21] was used to create a research database with information on all examinations. Statistical analysis was performed with SAS software (version 9.4, SAS Institute).

Results

Characteristics of Patients and MRI Examinations

The characteristics of the patients and MRI examinations are summarized in Table 2. Of the 299 patients, 47% (140) had a BRCA1 mutation, 49% (146) had a BRCA2 mutation, and 2% (6) had both mutations. A total of 5% (20/427) of MRI examinations yielded a histopathologically confirmed cancer. Of the 20 cancers, 55% (11) were masses (mean size, 8.7 mm; range, 4–15 mm), and 45% (9) were nonmass lesions (mean size, 21.4 mm; range, 10–50 mm). A total of 60% (12) of cancers were invasive (nine poorly and three moderately differentiated; nine estrogen receptor positive, nine progesterone receptor positive, one HER2 positive), and 40% (8) were intraductal (five poorly and three moderately differentiated). Among the 20 breast cancers, 15 were diagnosed with MRI-guided biopsy and five with ultrasound-guided biopsy after an imaging correlate was identified on MRI-directed ultrasound. A total of 95% (407) of examinations were normal (164) or revealed benign findings (243). Of the 243 examinations with benign findings, benignity was established by histopathologic analysis in 11% (26) and by 2-year MRI follow-up in 89% (217). All histologically confirmed benign lesions were diagnosed with MRI-guided biopsy.

TABLE 2:

Characteristics of 299 Patients and 427 MRI Examinations

Feature	Value
Patient
Age at MRI examination (y)	48.8 (23–78)
Carrier of BRCA1 mutation	140/299 (47)
Carrier of BRCA2 mutation	146/299 (49)
Carrier of both BRCA1 and BRCA2 mutations	6/299 (2)
MRI examinations
Normal (i.e., no finding)	263/427 (62)
Benign finding	243/427 (57)
Malignant finding	20/427 (5)
Examinations with benign finding
Histopathology reference standard	26/243 (11)
Benign breast parenchyma and sclerosing adenosis	14/26 (54)
Fibrocystic changes	1/26 (4)
Fibroadenomatoid change or fibroadenoma	3/26 (12)
Pseudoangiomatous hyperplasia	2/26 (8)
Lymph node	1/26 (4)
Ductal hyperplasia	1/26 (4)
Granulomatous changes	1/26 (4)
Atypical ductal hyperplasia	1/26 (4)
Lobular carcinoma in situ	1/26 (4)
Papilloma	1/26 (4)
Imaging follow-up reference standard	217/243 (89)
Examinations with malignant finding (histopathology reference standard)	20/20 (100)
Invasive ductal carcinoma	11/20 (55)
Ductal carcinoma in situ	8/20 (40)
Mixed, invasive ductal carcinoma and ductal carcinoma in situ	1/20 (5)

Note—Except for age (mean with range in parentheses), values are counts with percentage in parentheses.

Summary of Reader Findings and Examination Times by Protocol

Table 3 summarizes findings by reader. With positive examinations defined as those assigned BI-RADS categories 3–5, for all readers the number of positive examinations was greatest for the abbreviated protocol and lowest (or tied for lowest) for the abbreviated plus T2-weighted protocol. The numbers of positive examinations for the standard, abbreviated, and abbreviated plus T2-weighted protocol were 94, 106, and 94 for reader 1; 105, 121, and 96 for reader 2; 127, 221, and 95 for reader 3; and 74, 85, and 66 for reader 4. The median probability of malignancy for the standard, abbreviated, and abbreviated plus T2-weighted protocols was 26%, 40%, and 26% for reader 1; 60%, 55%, and 62% for reader 2; 31%, 26%, and 31% for reader 3; and 40%, 40%, and 40% for reader 4. For all readers, the median probability of malignancy was significantly greater among the 20 malignant cases (range, 50–80%) than among the 407 benign cases (range, 2–50%) (all p < .001) (Table S1, which can be viewed in the AJR electronic supplement to this article, available at doi.org/10.2214/AJR.21.27022).

TABLE 3:

Summary of Reader Results

	Reader 1			Reader 2			Reader 3			Reader 4
Finding	Standard	Abbreviated	Abbreviated Plus T2-Weighted	Standard	Abbreviated	Abbreviated Plus T2-Weighted	Standard	Abbreviated	Abbreviated Plus T2-Weighted	Standard	Abbreviated	Abbreviated Plus T2-Weighted
Background parenchymal enhancement
Minimal	267 (63)	234 (55)	281 (66)	202 (47)	235 (55)	206 (48)	112 (26)	127 (30)	89 (21)	220 (52)	207 (49)	234 (55)
Mild	87 (20)	102 (24)	81 (19)	114 (27)	104 (24)	111 (26)	233 (55)	227 (53)	248 (58)	133 (31)	142 (33)	120 (28)
Moderate	39 (9)	47 (11)	35 (8)	71 (17)	61 (14)	70 (16)	64 (15)	59 (14)	68 (16)	58 (14)	65 (15)	59 (14)
Marked	33 (8)	43 (10)	29 (7)	39 (9)	26 (6)	39 (9)	17 (4)	13 (3)	21 (5)	15 (4)	12 (3)	13 (3)
Nonapplicablea	1	1	1	1	1	1	1	1	1	1	1	1
Amount of fibroglandular tissue
Almost entirely fatty	51 (12)	54 (13)	54 (13)	48 (11)	59 (14)	58 (14)	40 (9)	47 (11)	31 (7)	18 (4)	19 (4)	18 (4)
Scattered	146 (34)	130 (31)	140 (33)	159 (37)	187 (44)	168 (39)	165 (39)	159 (37)	179 (42)	147 (35)	137 (32)	147 (35)
Heterogeneously dense	137 (32)	129 (30)	145 (34)	154 (36)	124 (29)	132 (31)	175 (41)	177 (42)	170 (40)	199 (47)	205 (48)	186 (44)
Extreme	92 (22)	113 (27)	87 (20)	65 (15)	56 (13)	68 (16)	46 (11)	43 (10)	46 (11)	62 (15)	65 (15)	75 (18)
Nonapplicablea	1	1	1	1	1	1	1	1	1	1	1	1
BI-RADS
1	214 (50)	227 (53)	238 (56)	129 (30)	169 (40)	185 (43)	169 (40)	142 (33)	190 (44)	288 (67)	342 (80)	291 (68)
2	119 (28)	94 (22)	95 (22)	193 (45)	137 (32)	146 (34)	131 (31)	64 (15)	142 (33)	65 (15)	0 (0)	68 (16)
3	44 (10)	27 (6)	43 (10)	42 (10)	52 (12)	31 (7)	74 (17)	124 (29)	46 (11)	26 (6)	33 (8)	26 (6)
4	49 (11)	76 (18)	50 (12)	61 (14)	67 (16)	63 (15)	52 (12)	97 (23)	49 (11)	48 (11)	52 (12)	42 (10)
5	1 (0)	3 (1)	1 (0)	2 (1)	2 (1)	2 (1)	1 (0)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)
Positive resultb	94 (22)	106 (25)	94 (22)	105 (25)	121 (28)	96 (22)	127 (30)	221 (52)	95 (22)	74 (17)	85 (20)	66 (15)
Probability of malignancyc	26 (26–43)	40 (23–52)	26 (26–43)	60 (33–77)	55 (31–72)	62 (36–72)	31 (16–55)	26 (16–40)	31 (30–60)	40 (26–55)	40 (27–54)	40 (18–50)

Note—Unless otherwise indicated, values are number of examinations with percentage in parentheses. Percentages computed from 427 examinations, except for background parenchymal enhancement and amount of fibroglandular tissue, for which percentages were computed from 426 examinations because in one examination these two characteristics were nonapplicable because there was a case of bilateral radical mastectomy. Not all percentages total 100 owing to rounding.

^aPercentage not calculated.

^bCategory 1 or 2 considered negative, 3–5 considered positive.

^cMedian with interquartile range in parentheses.

Interreader agreement on BI-RADS category was moderate for all three protocols (standard, κ = 0.55; abbreviated, κ = 0.45; abbreviated plus T2-weighted, κ = 0.57). For all readers, median interpretation time was highest for the standard and lowest for the abbreviated protocol (all p < .001); for example, for pooled reader data, the median interpretation time was 68 seconds for the standard, 38 seconds for the abbreviated, and 40 seconds for the abbreviated plus T2-weighted protocol (Table 4).

TABLE 4:

Interpretation Time (s) by Reader and Protocol

Reader	Standard	Abbreviated	Abbreviated Plus T2-Weighted	p
1	43 (37–48)	20 (16–25)	28 (25–33)	< .001
2	60 (44–82)	42 (34–54)	46 (40–57)	< .001
3	100 (65–146)	46 (36–60)	53 (36–78)	< .001
4	64 (52–84)	28 (20–39)	34 (27–41)	< .001
Pooled	68 (55–88)	38 (29–48)	40 (34–47)	< .001

Note—Values are median with interquartile ranges in parentheses.

Screening Performance

Table 5 summarizes the screening performance of the three protocols for each reader and for pooled reader data. Table S2 (which can be viewed in the AJR electronic supplement to this article, available at doi.org/10.2214/AJR.21.27022) shows the corresponding p values. Pooled sensitivity was 94% for the standard, 92% for the abbreviated, and 90% for the abbreviated plus T2-weighted protocol. For readers 3 and 4, sensitivity was 95% for the standard and ranged from 80% to 90% for the abbreviated and the abbreviated plus T2-weighted protocols. However, no pairwise comparison of sensitivity among protocols for any reader or for pooled readers was statistically significant (all p > .001). Figure 2 shows an example of an invasive ductal carcinoma that three of four readers detected with both the abbreviated protocol and the abbreviated plus T2-weighted protocol.

TABLE 5:

Screening Performance, Stratified by Reader and Protocol

Protocol	Sensitivity	Specificity	PPV	NPV	Accuracy	AUC
Reader 1
Standard	90 (18/20) [68–99]	81 (331/407) [77–85]	19 (18/94) [12–29]	99 (331/333) [99–100]	82 (349/427) [78–85]	0.94 [0.91–0.97]
Abbreviated	90 (18/20) [68–90]	78 (319/407) [74–82]	17 (18/106) [10–26]	99 (319/321) [98–100]	79 (337/427) [75–83]	0.89 [0.81–0.95]
Abbreviated plusT2-weighted	95 (19/20) [75–100]	82 (332/407) [77–85]	20 (19/94) [13–30]	100 (332/333) [98–100]	82 (351/427) [78–86]	0.91 [0.83–0.98]
Reader 2
Standard	95 (19/20) [75–100]	79 (320/407) [75–83]	18 (19/106) [11–27]	100 (320/321) [98–100]	79 (339/427) [75.5–83.4]	0.93 [0.89–0.96]
Abbreviated	100 (20/20) [80–100]	75 (306/407) [71–79]	20 (20/101) [10–24]	100 (306/306) [80–100]	76 (326/427) [72–80]	0.92 [0.89–0.94]
Abbreviated plusT2-weighted	95 (19/20) [75–100]	81 (330/407) [77–85]	20 (19/96) [12–29]	100 (330/331) [98–100]	82 (349/427) [78–85]	0.92 [0.87–0.96]
Reader 3
Standard	95 (19/20) [75–100]	73 (299/407) [69–78]	15 (19/127) [9–22]	100 (299/300) [98–100]	74 (318/427) [70–79]	0.85 [0.74–0.95]
Abbreviated	90 (18/20) [68–99]	50 (204/407) [45–55]	8 (18/221) [5–12]	99 (204/206) [97–100]	52 (222/427) [47–57]	0.80 [0.72–0.87]
Abbreviated plusT2-weighted	80 (16/20) [56–94]	81 (328/407) [76–84]	17 (16/95) [10–26]	99 (328/332) [97–100]	80 (344/427) [76–84]	0.90 [0.85–0.94]
Reader 4
Standard	95 (19/20) [75–100]	86 (352/407) [83–90]	26 (19/74) [16–37]	100 (352/353) [98–100]	87 (371/427) [83–90]	0.95 [0.92–0.97]
Abbreviated	90 (18/20) [68–99]	84 (340/407) [80–87]	21 (18/85) [13–31]	99 (340/342) [98–100]	84 (358/427) [80–87]	0.89 [0.82–0.97]
Abbreviated plusT2-weighted	90 (18/20) [68–99]	88 (357/407) [84–91]	27 (18/67) [16–38]	99 (357/359) [98–100]	88 (375/427) [84–91]	0.90 [0.81–0.98]
Pooled
Standard	94 (75/80) [81–99]	80 (1303/1628) [76–83]	19 (75/399) [12–27]	100 (1303/1308) [99–100]	81 (1378/1708) [79–82]	0.92 [0.89–0.94]
Abbreviated	92 (74/80) [80–98]	72 (1167/1628) [68–75]	14 (74/534) [9–21]	99 (1167/1173) [98–100]	73 (1241/1708) [71–75]	0.88 [0.85–0.91]
Abbreviated plusT2-weighted	90 (72/80) [76–97]	83 (1346/1628) [79–86]	20 (72/353) [13–30]	99 (1346/1354) [98–100]	83 (1418/1708) [81–85]	0.90 [0.87–0.94]

Note—Except for AUC, values are percentage with number of examinations in parentheses. Values in brackets are 95% CI.

Fig. 2—

66-year-old female BRCA1 mutation carrier with 8-mm mass (arrow) in right breast. Comparison of abbreviated (A–C) and abbreviated plus T2-weighted (D) protocols constructed from standard full protocol screening breast MRI examination. For abbreviated, abbreviated plus T2-weighted, and standard protocols, assigned BI-RADS categories were as follows: reader 1 (15 years of experience) 4, 4, and 4; reader 2 (5 years) 4, 2, and 2; reader 3 (5 years) 3, 4, and 4; and reader 4 (2 years) 4, 4, and 4. Histopathologic analysis of mass yielded invasive ductal carcinoma.

A, Axial precontrast fat-suppressed T1-weighted image constructed with abbreviated protocol.

B, Axial postcontrast fat-suppressed T1-weighted subtracted image constructed with abbreviated protocol.

C, Axial postcontrast maximum intensity postcontrast T1-weighted image constructed with abbreviated protocol.

D, Axial fast-suppressed T2-weighted image constructed with abbreviated plus T2-weighted protocol.

Pooled specificity was significantly higher for the abbreviated plus T2-weighted protocol (83%) than for either the standard (80%) or the abbreviated protocol (72%) (both p < .001). For readers 2, 3, and 4, specificity was significantly higher for the abbreviated plus T2-weighted protocol (81–88%) than for the abbreviated protocol (50–84%) (all p < .001). For reader 3, specificity was also significantly higher for the abbreviated plus T2-weighted protocol (81%) than for the standard (73%) (p < .001). Figure 3 shows an example of benign asymmetric nonmass enhancement for which the addition of T2-weighted imaging to the abbreviated protocol resulted in a lower BI-RADS category assignment for some readers, contributing to the increased specificity of this protocol.

Fig. 3—

45-year-old female BRCA1 mutation carrier who underwent screening breast MRI.

A and B, Axial postcontrast T1-weighted (A) and fat-suppressed T2-weighted (B) MR images show asymmetric nonmass enhancement (arrow) in right breast. For abbreviated, abbreviated plus T2-weighted, and standard protocols, assigned BI-RADS categories were as follows: reader 1 (15 years of experience) 1, 1, and 2; reader 2 (5 years) 3, 3, and 2; reader 3 (5 years) 3, 1, and 1; and reader 4 (2 years) 1, 1, and 1. Mass was stable on follow-up imaging after 2 years (not shown). Thus, examination would be deemed positive by two of four readers using standard protocol and by none of four readers using abbreviated plus T2-weighted protocol. Negative result with abbreviated plus T2-weighted protocol would avoid unnecessary biopsy.

Pooled PPV was significantly higher for both the standard (19%) and the abbreviated plus T2-weighted (20%) protocols than for the abbreviated protocol (14%) (both p < .001). For reader 3, PPV was also significantly higher for both the standard (15%) and the abbreviated plus T2-weighted protocols (17%) than for the abbreviated protocol (8%) (both p < .001). Other pairwise comparisons of PPV among protocols were not statistically significant (all p > .001). Pooled NPV was 100% for the standard, 99% for the abbreviated, and 99% for the abbreviated plus T2-weighted protocols. No pairwise comparison of NPV among protocols for any reader or for pooled readers was statistically significant (all p > .001).

Pooled accuracy was significantly higher for the standard (81%) and abbreviated plus T2 -weighted (83%) protocols than for the abbreviated protocol (73%) (both p < .001). For readers 2, 3, and 4, accuracy was significantly higher for the abbreviated plus T2-weighted protocol (80–88%) than for the abbreviated protocol (52–84%); for reader 3, accuracy was also significantly higher for the abbreviated protocol (74%) than for the standard protocol (52%) (all p < .001).

Pooled AUC, based on probabilities of malignancy, was 0.92 for the standard, 0.88 for the abbreviated, and 0.90 for the abbreviated plus T2-weighted protocols. For reader 3, AUC was significantly higher for the standard (0.85) than for the abbreviated protocol (0.80) (p < .001). Otherwise, no pairwise comparison of AUC among protocols for any reader or for pooled readers was statistically significant (all p > .001). Figure 4 shows ROC curves by reader and protocol.

Fig. 4—

Comparison of ROC curves of three protocols. Abbrev = abbreviated protocol, Abbrev + T2 = abbreviated plus T2-weighted protocol.

A, Graph shows AUCs for reader 1: standard, 0.94; abbreviated, 0.89; abbreviated plus T2-weighted, 0.91.

B, Graph shows AUCs for reader 2: standard, 0.93; abbreviated, 0.92; abbreviated plus T2-weighted, 0.92.

C, Graph shows AUCs for reader 3: standard, 0.85; abbreviated, 0.80; abbreviated plus T2-weighted, 0.90.

D, Graph shows AUCs for reader 4: standard, 0.95; abbreviated, 0.89; abbreviated plus T2-weighted, 0.90.

Discussion

In this multireader study, we compared a standard full breast MRI protocol with abbreviated protocols with and without T2-weighted imaging for breast cancer detection in a sample of patients with BRCA mutations. The abbreviated protocol had the poorest performance, and on the basis of our data we discourage its use. However, the accuracy and sensitivity of the abbreviated protocol that included T2-weighted imaging were similar to those of the standard protocol; there were no significant difference in these measures among any of the four readers. This abbreviated protocol including T2-weighted imaging is anticipated to require approximately one-half the acquisition time and total examination time of the full protocol. The findings support the use of an abbreviated protocol that includes T2-weighted imaging in this patient population.

Our results are in line with those of previous studies showing abbreviated MRI to be a promising technique for breast cancer detection. Kuhl et al. [8] were the first to establish that abbreviated MRI is as effective as standard MRI for breast cancer detection. They prospectively tested a 3-minute abbreviated protocol in women at mild to moderately increased breast cancer risk. The protocol consisted of precontrast and postcontrast T1-weighted sequences including subtraction and MIP images. The diagnostic performance of the abbreviated protocol was comparable to that of a full diagnostic protocol; all cancers present in the study group were detected. Mango et al. [9] investigated use of the earliest postcontrast T1-weighted acquisition, along with use of the corresponding subtraction and MIP images of that acquisition, in patients with known breast cancer. They found mean sensitivity of 96% in a mean interpretation time of 44 seconds.

Other studies have also investigated the impact of inclusion of a T2-weighted sequence in abbreviated protocols. Heacock et al. [22] evaluated several abbreviated protocols in imaging of patients with biopsy-proven breast cancer, including combinations of fat-suppressed precontrast, postcontrast, and subtracted T1-weighted imaging and T2-weighted imaging. They did not find additional value of including T2-weighted imaging. In a study by Dialani et al. [14], the addition of T2-weighted imaging to an abbreviated protocol changed the interpreting radiologist’s assessments in only 3% of cases. In comparison, in our study, addition of T2-weighted imaging significantly improved specificity and accuracy over the abbreviated protocol without T2-weighted imaging for three of four readers. It also significantly improved specificity and accuracy for pooled reader data. For all readers, the number of positive examinations was substantially higher for the abbreviated protocol than for the other two protocols. Furthermore, diagnostic performance of the abbreviated protocol that included T2-weighted imaging was comparable to that of the standard protocol, and interpretation times decreased significantly.

The abbreviated protocol without T2-weighted imaging yields diagnostic information solely regarding contrast enhancement. On the basis of our findings, we believe that the additional information gained by inclusion of T2-weighted imaging in the abbreviated protocol (e.g., T2 hyperintensity within a mass versus perilesional edema) gives readers increased confidence in classifying areas of enhancement as benign or malignant. In a study by Grimm et al. [23], the performance of an abbreviated protocol that included a fat-saturated T2-weighted sequence and precontrast and early postcontrast T1-weighted sequences without corresponding MIP images was similar to the performance of a standard protocol in breast cancer detection. However, images obtained with the abbreviated protocol without the T2-weighted sequence were not interpreted in the study, precluding assessment of the added value of the sequence.

An unexpected finding was the significantly higher pooled specificity of the abbreviated plus T2-weighted protocol than of the standard protocol. In comparison, two recent meta-analyses [24, 25] showed no difference in sensitivity or specificity between abbreviated and standard protocols. The meta-analysis by Baxter et al. [24] (n = 2588 patients from five screening studies) showed that abbreviated MRI had pooled sensitivity of 90%, pooled specificity of 92%, and pooled AUC of 0.94, which were not significantly different from values of 92%, 95%, and 0.97 for standard MRI. The meta-analysis by Geach et al. [25] (n = 2763 women from five studies, only one study including T2-weighted sequences in the abbreviated protocol) showed that abbreviated MRI had overall sensitivity of 94.8% and overall specificity of 94.6% with standard MRI as the reference standard. Sensitivity and specificity did not differ between the two protocols among three studies with follow-up data. The specificities of both the standard and the abbreviated protocols in our study are somewhat lower than previously reported for various screening breast MRI protocols. This may have occurred because readers were not provided earlier studies for comparison and readers were aware that all patients were BRCA carriers. For all readers, the number of positive examinations was lowest with the abbreviated plus T2-weighted protocol (aside from one reader who had the same number of positive examinations for the abbreviated plus T2-weighted and the standard protocols). It is possible that once certain essential sequences for achieving high diagnostic performance are accessed, the availability of additional sequences leads to more false-positive findings and hence reduced specificity, without improving cancer detection or overall accuracy.

We observed significantly reduced interpretation times for the abbreviated protocols than for the standard protocol, consistent with earlier reports. Harvey et al. [26] reported mean interpretation times of 1.55 minutes for the abbreviated protocol and 6.43 minutes for the full protocol; they found no difference in the number of cancers detected in a screening sample of patients at high risk. Osei et al. [27] reported mean interpretation times of 60.7 seconds for the abbreviated protocol and 99.4 seconds for the full protocol with comparable cancer detection rates and recall rates. The interpretation times for abbreviated and standard protocols in our study are shorter than in the earlier studies. This finding may be related to differences in reader experience, the amount of background parenchymal enhancement, or the method of recording interpretation time (e.g., exclusion in our study of the time needed to initially load and display the images).

Breast MRI is currently recommended for screening of patients at high risk. A standard full MRI protocol, including the time needed for patient positioning in the MRI unit, requires 30–40 minutes [28]. Therefore, international guidelines recommend breast MRI screening only for women with greater than a 20% cumulative lifetime risk of development of breast cancer [1, 29, 30]. Breast MRI screening has been found to be cost-effective for mutation carriers, who account for more than 60% (BRCA1) and 40% (BRCA2) of individuals with high cumulative lifetime risk of development of breast cancer by the age of 60 years [31]. Abbreviated protocols could potentially be used to extend the use of MRI to wider screening populations (e.g., women with other genetic mutations, such as TP53, PTEN, or CHEK2 mutations) [32, 33].

Our study had limitations. First, it was a single-institution retrospective study in which all readers had dedicated training in breast MRI interpretation. Therefore, the results might not be generalizable to other environments or to general radiologists. Second, readers were permitted to report only a single lesion per examination, so the impact of the various protocols on detection of multicentric carcinomas is unknown. Third, readers did not have access to patients’ prior examinations, which does not reflect typical workflow and is expected to have decreased the readers’ performance. For example, in the study by Dialani et al. [14], incorporation of prior examinations during review of abbreviated MRI in a second session resulted in an increase in specificity from 76.2% to 88.5%. However, our study design provides insight into the potential use of abbreviated protocols in a first-round screening setting. A total of 304 of 427 examinations had a prior MRI screening examination, which may have contributed to the small number of cancers diagnosed in the sample. Fourth, though the readers did not have access to prior studies, they were aware that all patients were BRCA mutation carriers, which could have also influenced their readings. Finally, the readers may have had recall bias between sessions despite the 21-day intervals between sessions. We attempted to mitigate this possibility by randomizing allocation of the images for review in each session. Furthermore, we believe that even with possible recall bias, our approach remains less biased than methods used in some earlier studies, in which review of abbreviated and standard protocols was not separated into distinct sessions. For example, in the study by Panigrahi et al. [34], the readers evaluated the full protocol immediately after evaluating the abbreviated protocol.

In conclusion, we explored the role of abbreviated breast MRI of BRCA1 and BRCA2 mutation carriers. The abbreviated protocol without T2-weighted imaging had suboptimal performance. However, inclusion of the T2-weighted sequence in the abbreviated protocol resulted in diagnostic performance comparable to that of the full protocol with, in fact, a small increase in specificity. The findings support the use of an abbreviated MRI protocol that includes a T2-weighted sequence for breast cancer screening of patients with BRCA mutations.

HIGHLIGHTS.

Key Finding

• In BRCA mutation carriers undergoing MRI screening, a theoretical abbreviated protocol with T2-weighted imaging had comparable (all p > .001) pooled sensitivity (90%), PPV (20%), NPV (99%), and AUC (0.90) and significantly higher (p < .001) pooled specificity (83%) compared with the standard protocol (94%, 19%, 100%, 0.92, and 80%).

Importance

• Abbreviated MRI with T2-weighted imaging should be considered for screening of patients with BRCA mutations.