Molecular Breast Imaging and Digital Breast Tomosynthesis for Dense Breast Screening: The Density MATTERS Trial

Abstract

Background:

Molecular breast imaging (MBI) relies on the functional uptake of a radiotracer, technetium 99m sestamibi, to reveal cancers that are occult on mammograms due to breast density.

Purpose:

To assess the performance of screening MBI as a supplement to digital breast tomosynthesis (DBT) in women with dense breasts.

Materials and Methods:

In this prospective, multiyear, multicenter trial from five sites, women with dense breasts were prospectively enrolled from 2017 to 2022 and underwent two annual screening rounds of DBT and MBI to assess the incremental cancer detection rate (CDR, reported as cancers per 1000 screenings) of supplemental MBI and to compare other performance metrics of DBT and MBI.

Results:

A total of 2978 participants were included. Participants had a mean age of 56.8 years ± 9.3 (SD) and a mean lifetime Tyrer-Cuzick risk of 12.0% ± 7.9 (SD). At year 1, the CDR was 5.0‰ (15 of 2978 participants) with DBT and 11.8‰ (35 of 2978 participants) with DBT plus prevalence screening MBI (incremental CDR, 6.7‰ [95% CI: 4.2, 10.6]; P < .001); the invasive CDR was 3.0‰ (nine of 2978 participants) with DBT and 7.7‰ (23 of 2978 participants) with DBT plus prevalence screening MBI (invasive incremental CDR, 4.7‰ [95% CI: 2.7, 8.1]; P < .001). At year 2, the CDR was 5.8‰ (15 of 2590 participants) with DBT and 9.3‰ (24 of 2590 participants) with DBT plus incidence screening MBI (incremental CDR, 3.5‰ [95% CI: 1.7, 6.8]; P = .001); the invasive CDR was 1.5‰ (four of 2590 participants) with DBT and 3.9‰ (10 of 2590 participants) with DBT plus incidence screening MBI (invasive incremental CDR, 2.3‰ [95% CI: 0.9, 5.3]; P = .048). The year 1 recall rate was 8.6% (255 of 2978 participants) with DBT and 17.9% (534 of 2978 participants) with DBT plus prevalence screening MBI (difference, 9.4% [95% CI: 8.4, 10.5]). The year 2 recall rate was 8.9% (231 of 2590 participants) with DBT and 13.8% (356 of 2590 participants) with DBT plus incidence screening MBI (difference, 4.8% [95% CI: 4.1, 5.7]). Twenty-nine participants had cancers detected only with MBI: 21 (72%) had invasive cancers (median size, 0.9 cm), 26 (90%) had node-negative cancers, and six (20%) had advanced cancers. The interval cancer rate was 0.7‰ (two of 2978 participants) in year 1 and 0.8‰ (two of 2590 participants) in year 2.

Conclusion:

The addition of MBI to DBT screening increased invasive cancer detection by 2.5-fold and modestly increased the recall rate at the second screening round.

Summary

In women with dense breasts, adding molecular breast imaging to digital breast tomosynthesis (DBT) increased both overall and invasive cancer detection rates, while modestly raising the recall rate compared with DBT alone.

Breast density not only masks cancers from mammographic detection but is also independently associated with breast cancer risk (1). Despite substantial progress in breast density awareness and mandated disclosure of density information to patients (2,3), there is still a lack of consensus among clinicians regarding recommendations for supplemental screening for women with dense breasts (4). U.S.-based organizations now universally endorse digital breast tomosynthesis (DBT) for dense breast screening, and DBT is frequently used as a primary, rather than supplemental, screening examination (4,5). However, the United States Preventive Services Task Force remains undecided on the value of other modalities, stating in 2024 that “current evidence is insufficient” to recommend supplemental screening with US or MRI for dense breasts (6).

The American College of Radiology now recommends annual MRI in addition to DBT for women with dense breasts seeking supplemental screening (7). As a supplement to DBT in dense breasts, the incremental cancer detection rate (CDR) for prevalence screening MRI is 10 per 1000 screenings, compared with rates of one to two per 1000 screenings for US (8-11). However, there are several barriers to MRI use, including lack of access, high cost, and patient intolerance, limiting its utilization for the estimated 47% of women presenting for screening who have dense breasts (8). Single-institution studies have shown that both contrast-enhanced mammography (CEM) and molecular breast imaging (MBI) significantly improve cancer detection over mammography, prompting the National Comprehensive Cancer Network to endorse CEM and MBI as suitable alternatives for individuals who qualify for but cannot undergo MRI (12-17). Although the Contrast-Enhanced Mammography Imaging Screening Trial will provide the first prospective, multiyear, multicenter comparison of CEM and DBT performance for dense breast screening (18), there are no such trials, to our knowledge, comparing MBI with DBT.

To assess the performance of screening MBI as a supplement to DBT in women with dense breasts, we conducted a prospective, multiyear, multicenter trial. We hypothesized that adding MBI to DBT would increase the invasive CDR relative to that of DBT alone.

Materials and Methods

Study Design and Participants

The Density MATTERS (or Molecular Breast Imaging and Tomosynthesis to Eliminate the Reservoir) Trial followed a prospective, intraindividual, cross-sectional design with longitudinal follow-up and was conducted at five sites in the United States, including three academic medical centers and two community hospitals. The study protocol was Health Insurance Portability and Accountability Act compliant and approved by the Mayo Clinic Institutional Review Board, which served as the institutional review board of record for all sites. Written informed consent was obtained from all participants.

Enrollment occurred from July 2017 to July 2022 through consecutive sampling. Participants underwent two annual screening rounds of MBI (prevalence screening at year 1 and incidence screening at year 2) in addition to clinically scheduled DBT screening at each round. Follow-up for breast cancer diagnoses, targeted for 1 year after each screening round, was completed in September 2024.

Eligible participants included women aged 40–75 years who were clinically asymptomatic, presented for screening DBT, and had dense breasts (American College of Radiology Breast Imaging Reporting and Data System category C or D as visually assessed by a radiologist [19]) reported on their most recent screening mammogram within the 24 months before enrollment. Women were excluded if they had undergone whole-breast US within 12 months before enrollment or had previously undergone MBI, breast MRI, or CEM at any time. Other exclusions included pregnancy or lactation, current treatment for breast cancer or high-risk breast lesions, current use of a breast cancer risk-reducing medication, or recent breast biopsy or surgery.

Self-reported race, ethnicity, and breast cancer risk factor information was collected via questionnaires to describe the study cohort. Breast cancer risk scores were calculated according to the Tyrer-Cuzick model, version 8.0b, and the Breast Cancer Surveillance Consortium model, version 2 (20,21).

Imaging and Interpretation

DBT included bilateral craniocaudal and mediolateral oblique tomosynthesis views with either a two-dimensional full-field digital mammogram or synthetic two-dimensional view, according to standard practice at each site. MBI was performed according to previously described procedures (22). Briefly, following the intravenous injection of 8 mCi technetium 99m sestamibi, bilateral craniocaudal and mediolateral oblique planar views were acquired for 10 minutes per view with a dual-head cadmium zinc telluride gamma camera (LumaGem [CMR Naviscan] at four sites and Eve Clear Scan [Smart Breast] at the remaining site).

The DBT and MBI examinations were interpreted independently by board-certified breast radiologists who were blinded to the result of the other modality and had 1–4 years (n = 3), 8–15 years (n = 5), or 16–30 years (n = 11) of experience in breast imaging. DBT examinations were interpreted according to the American College of Radiology Breast Imaging Reporting and Data System criteria (19) and assigned an assessment category of 0 (incomplete), 1 (negative), or 2 (benign); an assessment of 0 was considered positive. Test-positive DBT studies were assigned an additional, lesion-specific assessment ranging from 1 (negative) to 5 (highly suggestive of malignancy). Before interpreting MBI scans, radiologists were required to complete a training module and pass a test on MBI interpretation. MBI was interpreted according to a previously validated lexicon (23), and each examination was assigned an assessment that parallels the Breast Imaging Reporting and Data System assessment scale, ranging from 1 (negative) to 5 (highly suggestive of malignancy). MBI assessment categories of 3, 4, or 5 were considered positive.

Management of Screening Findings

If either DBT or MBI findings were positive, an integrated read was performed to determine the diagnostic work-up. DBT findings were further assessed with diagnostic mammography, targeted US, MRI, and short-interval follow-up imaging, as appropriate per standard of care. Per study protocol, MBI findings could prompt additional diagnostic evaluation in the setting of a negative DBT examination, but a negative MBI examination could not downgrade DBT work-up recommendations. Positive MBI findings were initially evaluated with a diagnostic mammogram, targeted US, or both. In the absence of a mammographic or sonographic correlation, further workup was dictated by the level of suspicion. Short-interval follow-up was recommended for MBI findings with assessments of three; MRI was recommended for MBI findings with assessments of four or five. All lesions with final assessments of four or five were recommended for biopsy with appropriate imaging guidance. MBI-guided biopsy became available at two of five sites during the study but was rarely used (six participants).

Reference Standard

For each screening round, participants were considered to have breast cancer if they had any histopathologic diagnosis of invasive breast cancer or ductal carcinoma in situ within 365 days after screening or before the next annual screening examination. Otherwise, a lack of breast cancer diagnosis could be confirmed with negative or benign findings at breast imaging, chart review, or participant contact at least 335 days but ideally 365 days after study screening.

End Points

The primary end point was the incremental invasive CDR of supplemental MBI at the initial (prevalence) screening. The secondary end points included the overall CDR, invasive CDR, sensitivity, specificity, recall rate, biopsy rate, positive predictive value of recall (PPV1), positive predictive value of biopsies performed (PPV3), and interval cancer rate. An exploratory end point was advanced CDR; advanced cancer was defined as having at least one of the following characteristics: distant metastases, positive lymph nodes, invasive disease of at least 20 mm in extent, or invasive cancer between 10 and 20 mm in size that was also triple negative or human epidermal growth factor receptor 2 positive (24).

Adverse Events

Adverse events were recorded and categorized according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.0.

Statistical Analysis

The screening performance of DBT alone and MBI alone were analyzed separately on the basis of independent readings of each modality. The performance of DBT with supplemental MBI (DBT plus MBI) was based on the interpretation of both modalities together.

The analysis set for each screening year included eligible participants who completed both DBT and MBI screenings. A target sample size of 2631 participants provided 90% power (one-sided α = .025) to detect a difference in invasive cancer detection of four per 1000 screenings between DBT plus MBI versus DBT alone at year 1 under the null hypothesis that the true difference is one per 1000 or less.

Each screening year was analyzed separately. Point estimates and 95% CIs for all performance measures were based on the Wilson score method for proportions with the Yates continuity correction. To determine differences in non–positive predictive value performance measures between DBT alone and MBI alone, a McNemar test was performed using the mid-P approach under a two-sided alternative, and point estimates and 95% CIs for differences in these performance measures were derived using the same Wilson score approach. Comparisons of CDRs between DBT alone and DBT plus MBI were evaluated using exact binomial tests under a superiority-by-a-margin testing strategy. Specifically, the one-sided null hypothesis that the difference in CDRs, θ, was less than one in 1000 (ie, null hypothesis [H₀]: θ ≤ 0.001 vs alternative hypothesis [H₁]: θ > 0.001) was tested. All comparisons of PPV1 and PPV3, along with estimation of differences and corresponding 95% CIs, were performed using generalized estimating equation regression modeling approaches (25). An α level of .05 was considered statistically significant for all hypothesis tests, and all the statistical analyses were performed using R version 4.2.2 (R Core Team).

Results

Participant Characteristics

Among the 3006 women who were eligible, 2978 completed the initial screening and 2590 completed a second annual round of screening with both DBT and MBI (Fig 1). Follow-up for breast cancer was completed in 2969 of 2978 participants (>99%) following year 1 screening and in 2586 of 2590 participants (>99%) following year 2 screening. The 14 participants with incomplete follow-up data were assumed to have negative breast cancer status.

Figure 1:

Participant flowchart. BI-RADS = Breast Imaging Reporting and Data System, DBT = digital breast tomosynthesis, DCIS = ductal carcinoma in situ, MBI = molecular breast imaging, WBUS = whole-breast US. * = Of the 2932 participants with no breast cancer in year 1, follow-up was completed with negative or benign breast imaging (n = 2769), chart review (n = 133), or participant contact (n = 30) at least 335 days after the year 1 screening. ^† = Deaths were not due to breast cancer. ^‡ = Of the 2559 participants with no breast cancer in year 2, follow-up was completed with negative or benign breast imaging (n = 2165), chart review (n = 308), or participant contact (n = 86) at least 335 days after the year 2 screening.

The 2978 women analyzed (Table 1) had a mean age of 56.8 years ± 9.3 (SD), and most were postmenopausal (n = 1890; 69%) and had category C density (n = 2446; 82%) at initial screening DBT. The lifetime breast cancer risk according to the Tyrer-Cuzick model was less than 20% in 88% of the participants; 80% had no family history of breast cancer, and 98% had no personal history of breast cancer. The median 5-year risk of invasive breast cancer according to the Breast Cancer Surveillance Consortium model was 1.8% (range, 0.4%–16.3%).

Table 1:

Baseline Demographics and Risk Characteristics of Eligible Participants (n = 2978)

Characteristic	Value
Age (y)
Mean*	56.8 ± 9.3
Median†	57 (40–75)
Race (n = 2930)‡
American Indian or Alaskan Native	7 (<1)
Asian	59 (2)
Black or African American	268 (9)
Native Hawaiian or other Pacific Islander	0
Multiple races	4 (<1)
White	2592 (88)
Hispanic or Latino (n = 2910)§	77 (3)
Menopausal status (n = 2758)ǁ
Premenopausal or perimenopausal	868 (31)
Postmenopausal	1890 (69)
ACR BI-RADS breast density from year 1 DBT
A: Almost entirely fat	0
B: Scattered fibroglandular densities	138 (5)
C: Heterogeneously dense	2446 (82)
D: Extremely dense	394 (13)
Known BRCA-1 or -2 genetic mutation#	0
History of breast cancer	62 (2)
Prior benign breast biopsy showing atypia or LCIS	46 (2)
History of at least one first-degree relative with breast cancer	587 (20)
Tyrer-Cuzick 10-year risk score (n = 2884)**
Mean*	4.7 ± 3.5
Median†	3.7 (0.7–44.1)
Tyrer-Cuzick lifetime risk score (n = 2884)**
Mean*	12.0 ± 7.9
Median†	10.1 (1.0–69.9)
Lifetime risk by Tyrer-Cuzick model (n = 2884)**
>20%	366 (13)
15%–20%	366 (13)
<15%	2152 (75)
Breast Cancer Surveillance Consortium 5-year risk score (n = 2889) (%)††
Mean*	1.9 ± 1.0
Median†	1.8 (0.4–16.3)
Breast Cancer Surveillance Consortium 10-year risk score (n = 2889) (%)††
Mean*	4.0 ± 1.8
Median†	3.8 (0.9–27.8)

Note.—Except where indicated, data in parentheses are percentages. ACR BI-RADS = American College of Radiology Breast Imaging Reporting and Data System, DBT = digital breast tomosynthesis, LCIS = lobular carcinoma in situ.

^*Data are means ± SDs.

^†Data in parentheses are ranges.

^‡Forty-eight participants did not report race

^§Sixty-eight participants did not report ethnicity.

^ǁA total of 220 participants aged younger than 55 years did not report menopausal status or were uncertain about their menopausal status due to hysterectomy, endometrial ablation, or prior chemotherapy.

^#At the time of study enrollment, 74 participants reported having been tested for BRCA genetic mutations; none reported testing positive.

^**The Tyrer-Cuzick risk score was calculated using the Ikonopedia International Breast Cancer Intervention Study (or IBIS) online calculator version 8.0b (20) for 2884 participants who provided sufficient family pedigree information and reported no history of breast cancer before enrollment.

^††The Breast Cancer Surveillance Consortium risk score was calculated using the online calculator version 2 (21) for 2889 participants who were younger than 75 years and reported no history of breast cancer before enrollment.

Cancer Detection

In the first screening round (year 1), in which 2978 participants underwent incidence screening DBT (with images from prior DBT or prior two-dimensional digital mammography available) and prevalence screening MBI, 37 participants were diagnosed with 39 breast cancer lesions; among them, 24 participants had invasive cancer and 13 had ductal carcinoma in situ only (Table 2). The overall cancer yield was 15 of 2978 (CDR, 5.0‰ [95% CI: 2.9, 8.5]) for DBT alone versus 26 of 2978 (CDR, 8.7‰ [95% CI: 5.8, 13.0]) for prevalence screening MBI alone (P = .04). DBT plus MBI resulted in an overall cancer yield of 35 of 2978 (CDR, 11.8‰ [95% CI: 8.3, 16.5]). The incremental overall CDR for supplemental MBI was 6.7 (95% CI: 4.2, 10.6). The invasive CDRs were 3.0‰ (95% CI: 1.5, 6.0) for DBT alone, 6.4‰ (95% CI: 4.0, 10.1) for prevalence screening MBI alone (P = .02 vs DBT alone), and 7.7‰ (95% CI: 5.0, 11.8) for the combination of DBT plus MBI (Table 3). The incremental invasive CDR for supplemental MBI at year 1 was 4.7‰ (95% CI: 2.7, 8.1).

Table 2:

Characteristics of Individual Participants with Breast Cancer Detected at Initial Screening (Year 1) and within 1 Year of Follow-up

Participant No./Age (y)/Breast Density*	Tyrer- Cuzick Lifetime Risk Score (%)	Lesion No.	Lesion Detected at DBT	Lesion Detected at MBI	Cancer Type	Grade	Largest Tumor Dimension (cm)	ER Status	PR Status	HER2 Status	AJCC TNM Stage Group†
1/69/C	25.6	1	No	Yes	IDC	3	1.6	Positive	Positive	Negative	IA
2/57/C	10.8	2	No	Yes	IDC	2	2.6	Positive	Negative	Positive	IIA
3/71/C	5.4	3	No	Yes	IDC	1	1.0	Positive	Positive	Negative	IA
4/71/C	6.4	4	No	Yes	IDC	2	0.9	Positive	Positive	Negative	IIA
5/62/C	5.5	5	No	Yes	IDC	2	0.9	Negative	Negative	Negative	IA
6/44//D	33.4	6	No	Yes	IDC	2	0.6	Positive	Positive	Negative	IA
7/51/C	18.7	7	Yes	No	ILC	2	1.4	Positive	Positive	Negative	IA
8/64/C	17.0	8a, 8b	8a: no; 8b: no	81: yes; 8b: yes	8a: IDC; 8b: IDC	8a: 1; 8b: 1	8a: 1.7; 8b: 0.8	8a: positive; 8b: NA	8a: positive; 8b: NA	8a: negative; 8b: NA	8a: IA; 8b: IA
9/70/C	16.5	9‡	No	No	ILC	1	0.8	Positive	Positive	Negative	IA
10/58/C	36.3	10	Yes	Yes	IDC	3	7.8	Negative	Negative	Negative	IIIA
11/64/C	11.7	11	Yes	Yes	IDC	2	0.8	Positive	Positive	Negative	IA
12/66/C	6.6	12	Yes	Yes	IDC	3	2.4	Negative	Negative	Negative	IIA
13/53/B	19.7	13	No	Yes	IDC	2	0.5	Positive	Positive	Positive	IA
14/52/D	11.5	14a, 14b§	14a: no; 14b: no	14a: yes; 14b: no	14a: ILC; 14b: invasive tubular	14a: 1; 14b: 1	14a: 0.2; 14b: 0.2	14a: positive; 14b: NA	14a: positive; 14b: NA	14a: negative; 14b: NA	14a: IA; 14b: IA
15/62/C	13.2	15	No	Yes	IDC	1	0.4	Positive	Positive	Negative	IB
16/67/C	7.3	16	No	Yes	ILC	2	2.2	Positive	Positive	Negative	IIA
17/67/C	5.1	17	Yes	No	ILC	2	0.3	Positive	Negative	Negative	IA
18/53/C	27.1	18	No	Yes	ILC	3	1.0	Positive	Positive	Negative	IA
19/57/C	7.0	19	Yes	No	IDC	2	1.4	Positive	Positive	Negative	IIA
20/50/C	19.8	20	No	Yes	IDC	2	0.7	Positive	Positive	Negative	IA
21/65/C	9.4	21	No	Yes	IDC	2	0.6	Positive	Positive	Negative	IA
22/66/C	8.5	22	Yes	No	IDC	1	0.4	Positive	Positive	Negative	IA
23/66/C	NA	23	Yes	Yes	IDC	2	1.8	Positive	Positive	Negative	IA
24/47/D	20.8	24	Yes	Yes	IDC	2	2.4	Positive	Positive	Negative	IIA
25/57/C	NA	25	Yes	No	DCIS	2	1.0	Positive	Positive	NP	0
26/57/D	7.2	26‡	No	No	DCIS	3	4.0	Negative	Negative	NP	0
27/59/B	6.4	27	Yes	No	DCIS	2	0.5	Positive	Positive	NP	0
28/61/C	15.9	28	No	Yes	DCIS	3	2.4	Negative	Negative	NP	0
29/65/C	20.7	29	No	Yes	DCIS	2	0.3	Positive	Positive	NP	0
30/65/C	14.4	30	No	Yes	DCIS	2	0.3	Positive	Positive	NP	0
31/47/C	25.6	31	No	Yes	DCIS	1	1.0	Positive	Positive	NP	0
32/49/C	16.2	32	Yes	No	DCIS	3	2.5	Positive	Positive	NP	0
33/43/D	17.6	33	Yes	Yes	DCIS	3	NA	Negative	Negative	NP	0
34/70/C	10.7	34	Yes	No	DCIS	2	0.3	Positive	Positive	NP	0
35/43/C	10.4	35	No	Yes	DCIS	2	NA	Positive	Positive	NP	0
36/53/C	NA	36	Yes	No	DCIS	3	0.8	Negative	Negative	NP	0
37/48/C	18.0	37	No	Yes	DCIS	3	2.5	Negative	Negative	NP	0

Note.—Cancers include those detected at initial screening with digital breast tomosynthesis (DBT), molecular breast imaging (MBI), or both (35 participants) and those diagnosed during the follow-up period (two participants), which lasted until the second round of annual screening or, in the absence of subsequent screening, within 365 days after initial screening. AJCC = American Joint Committee on Cancer, DCIS = ductal carcinoma in situ, ER = estrogen receptor, HER2 = human epidermal growth factor receptor 2, IDC = invasive ductal carcinoma, ILC = invasive lobular carcinoma, NA = not available, NP = not performed, PR = progesterone receptor.

^*Breast density was determined with the American College of Radiology Breast Imaging Reporting and Data System at the time of the current DBT screen (19).

^†Staging group was determined with the American Joint Committee on Cancer TNM system (eighth edition), which includes extent of the tumor (T), extent of spread to the lymph nodes (N), and presence of metastasis (M) (36).

^‡Lesions 9 and 26 were interval cancers.

^§A second cancer lesion (lesion 14b) was identified in participant 14 at the time of surgical excision of screen-detected lesion 14a.

Table 3:

Performance Characteristics of DBT, MBI, and the Combination of DBT with Supplemental MBI

Performance Metric	DBT		MBI*		P Value: DBT vs MBI	DBT plus MBI		Difference of DBT plus MBI vs DBT Alone: Estimate
	No. of Participants	Estimate	No. of Participants	Estimate		No. of Participants	Estimate
CDR, overall (no. of women per 1000 screenings)
Year 1	15/2978	5.0 (2.9, 8.5)	26/2978	8.7 (5.8, 13.0)	.04	35/2978	11.8 (8.3, 16.5)	6.7 (4.2, 10.6)†
Year 2	15/2590	5.8 (3.4, 9.8)	15/2590	5.8 (3.4, 9.8)	.91	24/2590	9.3 (6.1, 14.0)	3.5 (1.7, 6.8)‡
CDR, invasive (no. of women per 1000 screenings)
Year 1	9/2978	3.0 (1.5, 6.0)	19/2978	6.4 (4.0, 10.1)	.02	23/2978	7.7 (5.0, 11.8)	4.7 (2.7, 8.1)§
Year 2	4/2590	1.5 (0.5, 4.2)	8/2590	3.1 (1.4, 6.3)	.18	10/2590	3.9 (2.0, 7.3)	2.3 (0.9, 5.3)ǁ
Sensitivity (%)
Year 1	15/37	40.1 (25.2, 57.8)	26/37	70.3 (52.8, 83.6)	.04	35/37	94.6 (80.5, 99.1)	54.1 (37.1, 70.2)
Year 2	15/26	57.7 (37.2, 76.0)	15/26	57.7 (37.2, 76.0)	.91	24/26	92.3 (73.4, 98.7)	34.6 (17.9, 55.6)
Specificity (%)
Year 1	2703/2941	91.9 (90.9, 92.9)	2629/2941	89.4 (88.2, 90.5)	<.001	2444/2941	83.1 (81.7, 84.4)	−8.8 (−7.8, −9.9)
Year 2	2349/2564	91.6 (90.5, 92.6)	2421/2564	94.4 (93.5, 95.3)	<.001	2232/2564	87.1 (85.7, 88.3)	−4.6 (−3.8, −5.5)
Recall rate (%)
Year 1	255/2978	8.6 (7.6, 9.6)	340/2978	11.4 (10.3, 12.6)	<.001	534/2978	17.9 (16.6, 19.4)	9.4 (8.4, 10.5)
Year 2	231/2590	8.9 (7.9, 10.1)	158/2590	6.1 (5.2, 7.1)	<.001	356/2590	13.8 (12.5, 15.2)	4.8 (4.1, 5.7)
Biopsy rate (%)
Year 1	76/2978	2.5 (2.0, 3.2)	122/2978	4.1 (3.4, 4.9)	<.001	163/2978	5.5 (4.7, 6.4)	2.9 (2.4, 3.6)
Year 2	60/2590	2.3 (1.8, 3.0)	62/2590	2.4 (1.9, 3.1)	.83	105/2590	4.1 (3.3, 4.9)	1.7 (1.3, 2.3)
PPV1 (%)
Year 1	15/255	5.9 (3.5, 9.7)	26/340	7.7 (5.2, 11.1)	.32	35/534	6.6 (4.7, 9.1)	0.7 (−1.5, 2.9)
Year 2	15/231	6.5 (3.8, 10.7)	15/158	9.5 (5.6, 15.4)	.17	24/356	6.7 (4.5, 10.0)	0.3 (−1.7, 2.2)
PPV3 (%)
Year 1	15/76	19.7 (11.8, 30.8)	26/122	21.3 (14.6, 29.8)	.75	35/163	21.5 (15.6, 28.8)	1.7 (−0.05, 0.09)
Year 2	15/60	25.0 (15.1, 38.1)	15/62	24.2 (14.6, 37.0)	.90	24/105	22.9 (15.5, 32.3)	−2.1 (−9.2, 5.0)

Note.—Numbers in parentheses are 95% CIs. The exact P value thresholds for binomial testing of the one-sided hypothesis that the difference between DBT plus MBI and DBT is greater than one incremental detection per 1000 screenings versus less than or equal to one incremental detection per 1000 screenings are listed in the footnotes below. CDR = cancer detection rate, DBT = digital breast tomosynthesis, MBI = molecular breast imaging, PPV1 = positive predictive value of recall, reported at the participant level, PPV3 = positive predictive value of biopsies performed, reported at the participant level.

^*The study was not designed to evaluate MBI as a sole screening test, as test-positive findings at blinded MBI interpretation (assessment categories of 3, 4, or 5) generated an integrated interpretation of MBI with current screening DBT, and participants may not have been recalled for diagnostic workup if MBI findings were explained by benign features at DBT.

^†P < .001 for year 1 overall CDR.

^‡P < .001 for year 1 invasive CDR.

^§P = .001 for year 2 overall CDR.

^ǁP = .048 for year 2 invasive CDR.

In the second screening round (year 2), in which 2590 participants underwent incidence screening with both MBI and DBT, 26 participants were diagnosed with breast cancer; among them, 13 had invasive cancer and 13 had ductal carcinoma in situ only (Table 4). The overall cancer yield for both DBT alone and MBI alone was 15 of 2590 (CDR, 5.8‰ [95% CI: 3.4, 9.8]). The overall cancer yield for the combination of DBT plus MBI was 24 of 2590 (CDR, 9.3‰ [95% CI: 6.1, 14.0]), corresponding to an incremental overall CDR of 3.5‰ (95% CI: 1.7, 6.8) in year 2. The invasive CDRs were 1.5‰ (95% CI: 0.5, 4.2) for DBT alone, 3.1 (95% CI: 1.4, 6.3) for MBI alone (P = .18 vs DBT alone), and 3.9‰ (95% CI: 2.0, 7.3) for the combination of DBT plus MBI. The incremental invasive CDR for supplemental MBI at year 2 was 2.3‰ (95% CI: 0.9, 5.3).

Table 4:

Characteristics of Individual Participants with Breast Cancer Detected at the Second Round of Screening (Year 2) and within 1 Year of Follow-up

Participant No./Age (y)/Breast Density*	Tyrer- Cuzick Lifetime Risk Score (%)	Lesion No.	Lesion Detected at DBT	Lesion Detected at MBI	Cancer Type	Grade	Largest Tumor Dimension (cm)	ER Status	PR Status	HER2 Status	AJCC TNM Stage Group†
38/74/C	5.4	38	No	Yes	IDC	2	1.0	Positive	Positive	Negative	IA
39/61/C	15.5	39a, 39b	39a: no; 39b: no	39a: no; 39b: yes	39a: IDC; 39b: DCIS	39a: 2; 39b: 1	39a: 0.3; 39b: 0.7	39a: positive; 39b: positive	39a: positive; 39b: positive	39a: negative; 39b: NP	39a: IA; 39b: IA
40/58/C	21.0	40	Yes	Yes	IDC	2	1.4	Positive	Positive	Negative	IA
41/73/C	4.1	41‡	No	No	IDC	3	3.4	Positive	Positive	Negative	IIA
42/73/C	5.9	42a, 42b	42a: no; 42b: no	42a: yes; 42b: yes	42a: IDC; 42b: IDC	42a: 2; 42b: 2	42a: 1.1; 42b: 1.1	42a: positive; 42b: positive	42a: positive; 42b: positive	42a: positive; 42b: positive	42a: IA; 42b: IA
43/69/C	10.5	43	No	Yes	ILC	1	9.1	Positive	Positive	Negative	IIB
44/67/C	13.3	44	Yes	No	IDC	1	0.4	Positive	Positive	Negative	IA
45/64/C	6.2	45	No	Yes	IDC and ILC	2	0.9	Positive	Negative	Negative	IA
46/61/C	11.0	46	Yes	Yes	IDC	3	3.6	Positive	Positive	Negative	IIB
47/66/C	NA	47	Yes	No	IDC	1	0.6	Positive	Positive	Negative	IA
48/59/C	5.6	48	No	Yes	IDC and ILC	1	0.9	Positive	Positive	Negative	IIA
49/71/C	11.5	49	No	Yes	IDC	2	0.9	Positive	Positive	Negative	IA
50/65/C	6.4	50‡	No	No	IDC	3	3.9	Negative	Negative	Negative	IIB
51/67/C	7.5	51	Yes	No	DCIS	2	0.3	Positive	Positive	NP	0
52/73/C	1.3	52	Yes	No	DCIS	3	4.2	Positive	Positive	NP	0
53/58/C	NA	53	Yes	Yes	DCIS	3	2.2	Positive	Positive	NP	0
54/46/C	17.5	54	Yes	Yes	DCIS	3	5.8	Positive	Positive	NP	0
55/51/C	11.1	55	Yes	No	DCIS	2	0.7	Positive	Positive	NP	0
56/50/B	10.3	56	Yes	No	DCIS	3	4.0	Negative	Negative	NP	0
57/60/C	8.7	57	Yes	Yes	DCIS	3	1.7	Negative	Negative	NP	0
58/72/C	12.6	58	No	Yes	DCIS	1	1.4	Positive	Positive	NP	0
59/54/C	7.3	59	Yes	No	DCIS	3	2.0	Positive	Positive	NP	0
60/53/C	8.1	60	Yes	Yes	DCIS	2	2.2	Positive	Positive	NP	0
61/47/C	15.9	61	Yes	No	DCIS	2	0.3	Positive	Positive	NP	0
62/51/C	27.3	62	No	Yes	DCIS	2	2.0	Positive	Positive	NP	0
63/65/C	5.2	63	Yes	No	DCIS	2	1.5	Positive	Positive	NP	0

Note.—Cancers include those detected at the second screening with digital breast tomosynthesis (DBT), molecular breast imaging (MBI), or both (24 participants) and those diagnosed during the follow-up period (two participants), which lasted until the next round of annual screening or, in the absence of subsequent screening, within 365 days after the second screening. AJCC = American Joint Committee on Cancer, DCIS = ductal carcinoma in situ, ER = estrogen receptor, HER2 = human epidermal growth factor receptor 2, IDC = invasive ductal carcinoma, ILC = invasive lobular carcinoma, NA = not available, NP = not performed, PR = progesterone receptor.

^*Breast density was classified with the American College of Radiology Breast Imaging Reporting and Data System at the time of the current DBT screen (19).

^†Staging group was determined with the American Joint Committee on Cancer TNM system (eighth edition), which includes extent of the tumor (T), extent of spread to the lymph nodes (N), and presence of metastasis (M) (36).

^‡Lesions 41 and 50 were interval cancers.

Incremental Cancers

Across both screening rounds, 30 breast cancer lesions were detected in 29 participants with MBI only (Tables 2, 4). Most of these incremental breast cancers were invasive (22 of 30 [71%] lesions), and they had a median invasive lesion size of 0.9 cm (range, 0.2–9.1 cm). Among the participants with breast cancer detected only with MBI, 26 of 29 (90%) had node-negative cancers, and six of 29 (20%) had advanced cancer. Examples of incremental cancers are shown in Figures 2-5.

Figure 2:

Images in a 62-year-old woman who presented for screening. (A) Image from digital breast tomosynthesis screening (the synthesized two-dimensional mediolateral oblique view of the left breast is shown) at year 1 was interpreted as negative and showing heterogeneously dense breast. (B) Image from molecular breast imaging screening (mediolateral oblique view) at year 1 reveals a 0.9-cm nonmass focal area of uptake in the left breast (arrow). (C) Targeted US scan (transverse image) shows a suspicious mass in the 3-o’clock position of the left breast, 4 cm from the nipple. US-guided core biopsy and lumpectomy revealed a 0.9-cm grade 2 invasive ductal carcinoma, an estrogen receptor–negative, progesterone receptor–negative, human epidermal growth factor receptor 2–negative lesion, with two sentinel nodes negative (N0).

Figure 5:

Images in a 69-year-old woman who presented for screening. (A) Image from digital breast tomosynthesis (DBT) screening (synthesized two-dimensional craniocaudal view of the right breast) at year 2 was negative and showing heterogeneously dense breast. (B) Molecular breast imaging (MBI) screening (right craniocaudal view) at year 2 shows a 4.7-cm nonmass focal area of uptake (arrow). US-guided biopsy and mastectomy revealed a grade 2 invasive lobular carcinoma, 9.1 cm in greatest extent, that was estrogen receptor positive, progesterone receptor positive, and human epidermal growth factor receptor 2 negative, with two sentinel nodes and 10 axillary nodes negative (N0). (C) Prior year 1 DBT was interpreted as negative. (D) Prior year 1 MBI was interpreted as negative with bilateral moderate background parenchymal uptake; a retrospective review revealed that mild uptake correlated with the location of the lesion detected at year 2.

Advanced Cancers

In the first screening round, seven of 2978 participants (2.4 per 1000 screened) were diagnosed with advanced cancers; cancer was detected with DBT alone in four of the seven (57%) participants and with DBT plus MBI in seven (100%). In year 2, six of 2590 participants (2.3 per 1000 screened) had advanced cancers; cancer was detected with DBT alone in one of the six (16%) participants and with DBT plus MBI in four (67%). In two of the six (33%) participants, cancer was not detected with either modality (interval cancers).

Interval Cancers

Interval breast cancer (eg, cancer undetected at screening and diagnosed before the next screen) was found in two of 2978 participants who underwent year 1 screening, for an interval cancer rate per 1000 screenings of 0.7 (95% CI: 0.1, 2.7). One interval cancer was a 0.8-cm grade 1 invasive lobular carcinoma detected by means of short-interval follow-up MRI, which was performed to evaluate a separate screen-detected lesion in the contralateral breast (determined to be benign). The other interval cancer was a 4-cm high-grade ductal carcinoma in situ identified by means of participant-reported symptoms 6 months after the year 1 screening (Table 2).

Among the 2590 participants who underwent year 2 screening, two had interval cancers diagnosed within 365 days of follow-up (interval cancer rate of 0.8‰ [95% CI: 0.1, 2.8]), including a 3.4-cm grade 3 invasive ductal carcinoma and a 3.9-cm grade 3 invasive ductal carcinoma. Both were initially identified as patient-reported lumps that were not included in the DBT or MBI field of view (one located in the chest wall and one in the axillary tail) but were later visualized by means of diagnostic evaluation with MRI.

Recall and Biopsy

At year 1, 255 of 2978 (8.6%) participants were recalled for a diagnostic workup due to the results of DBT alone, and 340 of 2978 (11.4%) were recalled due to the results of MBI alone (P < .001). The combination of DBT plus MBI resulted in a recall rate of 17.9% (534 of 2978 participants), an increase of 9.4% (95% CI: 8.4, 10.5) relative to DBT alone. The biopsy rate increased with the addition of MBI, from 2.5% (76 of 2978 participants) for DBT alone to 5.5% (163 of 2978 participants) for the combination of DBT plus MBI (increase of 2.9% [95% CI: 2.4, 3.6]). There was no evidence of a difference in PPV1 between DBT alone (5.9%) and the combination of DBT plus MBI (6.6%, difference of 0.7% [95% CI: −1.5, 2.9]). Likewise, there was no evidence of a difference in PPV3 between DBT alone (19.7%) and DBT plus MBI (21.5%, difference of 1.7% [95% CI: −0.05, 0.09]) (Table 3).

At year 2, the recall rate was 8.9% (231 of 2590 participants) for DBT alone versus 6.1% (158 of 2590 participants) for MBI alone (P < .001). The combination of DBT plus MBI in year 2 resulted in a recall rate of 13.8% (356 of 2590), an increase of 4.8% (95% CI: 4.1, 5.7) relative to DBT alone. The addition of MBI increased the biopsy rate from 2.3% for DBT alone to 4.1% for the combination of DBT plus MBI (increase of 1.7% [95% CI: 1.3, 2.3]). As in year 1, the positive predictive values at year 2 did not significantly change with the addition of MBI; the combination of DBT plus incidence MBI resulted in a PPV1 of 6.7% and a PPV3 of 22.9%.

Adverse Events

Among the 5568 screenings conducted with DBT plus MBI, 10 (0.18%) adverse events in nine participants were reported, including mild to moderate lightheadedness or nausea related to sestamibi injection (n = 6), mild pain in the breast or ribs after imaging (n = 3), and mild rash associated with sestamibi injection (n = 2). No serious adverse events were reported.

Discussion

To our knowledge, this is the first multicenter prospective evaluation of molecular breast imaging (MBI) as a supplement to digital breast tomosynthesis in women with dense breasts and average risk. In this trial, MBI depicted an additional 6.7 cancers per 1000 screenings at year 1 (prevalence round) and an additional 3.5 cancers per 1000 screenings at year 2 (incidence round). Among the incremental cancers detected only with MBI, 70% were invasive. Ninety percent of cancers detected only with MBI were node negative, indicating that MBI screening provides early detection. Additionally, 20% of those cancers were advanced, suggesting that MBI can reveal a reservoir of mammographically occult, clinically important disease with lethal potential, possibly having escaped detection over multiple mammographic screens.

The overall incremental CDR of supplemental MBI decreased from the prevalence to the incidence screening rounds (from 6.7‰ to 3.5‰), likely due to elimination of the reservoir of previously masked cancers after the prevalence MBI screening; this decrease in the CDR at subsequent screening rounds has also been observed with MRI and CEM (13,26,27). However, MBI still depicted more than twice as many invasive cancers and four times as many advanced cancers as DBT did at year 2. We found that the MBI-associated recall rate decreased from 11.4% at year 1 to 6.1% at year 2, which is likely attributed to the benefit of comparing imaging findings of year 2 to prior findings of year 1.

The strengths of the study were the inclusion of a mix of academic medical centers and community hospitals and the enrollment of 12% of patients belonging to minority groups, which extends the generalizability of our findings. The rigorous assessment of participant risk with the Tyrer-Cuzick and Breast Cancer Surveillance Consortium models in this study confirms that our cohort was primarily average risk, which strengthens the applicability of the findings to the average-risk population with dense breasts.

The Density MATTERS trial was modeled after the EA1141 trial to allow indirect comparisons between MBI and abbreviated MRI performance in women at average risk with dense breasts (9). At the prevalence screening in the Density MATTERS trial, the invasive CDR per 1000 screenings was 3.0 with DBT and 6.4 for MBI (P = .02); in the EA1141 trial, the invasive CDR per 1000 prevalence screenings was 4.8 for DBT versus 11.8 with abbreviated MRI (P = .002). For incidence screening, the invasive CDR was 1.5‰ for DBT versus 3.1‰ for MBI in the Density MATTERS trial and 1.5‰ for DBT versus 4.6‰ for abbreviated MRI in the EA1141 trial (28). Interval cancer rates at the first and second screening rounds were 0.7 and 0.8 per 1000 screenings for DBT plus MBI, respectively, and 0 and 0.8 per 1000 screenings for DBT plus MRI, respectively, (28). Both studies demonstrated a doubling of advanced cancer detection relative to DBT alone (28). In both trials, more than half of invasive cancers in women at average risk with dense breasts were undetected with DBT. Thus, although the Food and Drug Administration–mandated density notification letter language (“dense tissue makes it harder to find breast cancer on a mammogram”) is technically correct, this statement fails to convey the magnitude of the detection gap revealed when more sensitive functional modalities are performed.

Fourteen years after the first prospective, single-center MBI screening trial (29), MBI adoption in practice has been sporadic. Concerns about additional radiation from MBI have been a barrier to its adoption; however, the benefit-to-risk ratio of MBI is within the range of mammography and is considered safe for routine screening (30). MBI is well tolerated by patients, is relatively inexpensive, and has improving biopsy capability (31-34). Importantly, technetium 99m has an excellent safety profile with no contraindications (except pregnancy) and an exceedingly low reaction risk, whereas the iodinated contrast material used for CEM carries a risk of anaphylaxis, and the long-term retention of gadolinium from MRI carries unknown risk (12,22,35). Thus, MBI remains an important option for women at average risk with dense breasts who seek a well-tolerated and sensitive supplement to DBT.

A potential limitation of the Density MATTERS trial was that the order of MBI and DBT examinations was not randomized, as was done in the EA1141 trial. However, this limitation was mitigated by requiring independent interpretations of MBI and DBT with blinding to the result of the other modality. A second limitation was that, by design, the study evaluated the utility of MBI as a supplement to DBT and not as a sole screening test, although independent interpretation of the modalities allowed estimation of the performance of MBI alone.

In conclusion, this study provides multicenter evidence that supplemental molecular breast imaging in women with dense breasts increases invasive cancer detection by more than 2.5 times that of digital breast tomosynthesis alone at both prevalence and incidence screening rounds and only modestly increases the recall rate.

Figure 3:

Images in a 57-year-old woman who presented for screening. (A) Image from digital breast tomosynthesis screening (synthesized two-dimensional mediolateral oblique view of the left breast) at year 1 was interpreted as negative and showing heterogeneously dense breast. (B) Image from molecular breast imaging screening (mediolateral oblique view) at year 1 reveals a 2-cm mass in the left breast (arrow). (C) Targeted US scan (transverse image) reveals a corresponding irregular hypoechoic mass in the 1-o’clock position, 5 cm from the nipple. US-guided core biopsy and lumpectomy revealed a 2.6-cm grade 2 invasive ductal carcinoma, an estrogen receptor–positive, progesterone receptor–negative, human epidermal growth factor receptor 2–positive lesion, with two sentinel nodes negative (N0).

Figure 4:

Images in a 48-year-old woman who presented for screening. (A) Image from digital breast tomosynthesis screening (synthesized two-dimensional mediolateral oblique view of the right breast) at year 1 was interpreted as negative and showing heterogeneously dense breast. (B) Image from molecular breast imaging screening (mediolateral oblique view) at year 1 reveals a nonmass segmental area of uptake in the right breast (arrows). (C) Contrast-enhanced MRI scan (subtraction maximum intensity projection) shows corresponding segmental nonmass enhancement measuring 9.8 cm to the greatest extent (arrow). MRI-guided core biopsy and lumpectomy revealed high-grade ductal carcinoma in situ, 2.5 cm in extent, estrogen receptor–negative, and progesterone receptor–negative lesions.

Key Results.

• In this multicenter, prospective study of 2978 women with dense breasts, adding molecular breast imaging (MBI) to digital breast tomosynthesis (DBT) screening provided incremental cancer detection rates of 6.7 and 3.5 cancers per 1000 screenings in the first and second screening rounds (both P < .001), respectively.

• Relative to DBT alone, supplemental MBI increased the recall rate by 9.4% at the first screening but only 4.8% at the second screening.

• Most incremental cancers detected only with MBI were invasive (21 of 29 [72%]) and node negative (26 of 29 [90%]).

Abbreviations

CDR

cancer detection rate

CEM

contrast-enhanced mammography

DBT

digital breast tomosynthesis

MBI

molecular breast imaging

PPV1

positive predictive value of recall

PPV3

positive predictive value of biopsies performed