Reference No. 1

Population Reference No. 2

Screening Individuals at Average-Risk of Breast Cancer

Clinical Context and Test Purpose

The questions addressed in this portion of the evidence review:

Does the use of MRI as an adjunct to screening for breast cancer improve the net health outcome of patients who are asymptomatic with average-risk of developing breast cancer?
Is this degree of increased accuracy likely to improve net health outcomes via the earlier diagnosis and treatment?

The following PICO was used to select literature to inform this review.

Populations

The relevant population of interest is patients at average-risk of developing breast cancer.

Interventions

The intervention of interest is MRI as an adjunct to screening with mammography.

Comparators

The following test is currently being used to make decisions about managing breast cancer: mammography alone.

Outcomes

Breast MRI is performed as an adjunct to routine screening; timing can be guided by national guidelines on breast cancer screening (see Supplemental Information section).

Study Selection Criteria

Reported on the accuracy of the marketed version of the technology (including any algorithms used to calculate scores)
Included a suitable reference standard
Patient/sample clinical characteristics were described
Patient/sample selection criteria were described.

Clinically Valid

A test must detect the presence or absence of a condition, the risk of developing a condition in the future, or treatment response (beneficial or adverse).

Review of Evidence

Systematic Reviews

In a systematic review of literature conducted by Nelson et al (2016) for the U.S. Preventive Services Task Force breast cancer screening recommendation update, no randomized controlled trials (RCTs) or nonrandomized observational studies identified evaluated adjunctive MRI for screening average-risk women for breast cancer.^8, Because the prevalence of breast cancer is extremely low in average-risk young women, screening with a test such as MRI that has lower specificity would result in a lower positive predictive value (PPV) and many more false-positive results. Compared with mammography, there would be greater numbers of workups and biopsies with increased anxiety and morbidity with adjunctive MRI screening applied to young, average-risk women.

Health Quality Ontario (2016) published a systematic review of MRI as an adjunct to mammography for women, not at high-risk of breast cancer.^9, Reviewers searched for studies evaluating screening breast MRI as an adjunct to mammography compared with mammography alone. Studies needed to use pathology results as a reference standard for positive tests and clinical follow-up as a reference standard for negative tests. In addition, studies needed to report one or more outcomes of interest, which included effectiveness outcomes (eg, mortality, health-related quality of life, screening-related harms), diagnostic outcomes (eg, sensitivity, specificity), and biopsy and recall rates. Reviewers did not find any studies that met eligibility criteria. They concluded that there was a lack of evidence to inform the questions of the diagnostic accuracy of MRI plus mammography versus MRI alone and the impact of adjunct screening MRI on health outcomes in patients at less than high-risk of breast cancer.

Section Summary: Screening of Individuals at Average-Risk of Breast Cancer

The U.S. Preventative Services Task Force systematic review and guideline concluded that because the prevalence of breast cancer is low in average risk young women, screening with MRI, which has lower specificity, would result in a lower PPV and many more false positive results. A systematic review by Health Quality Ontario concluded that there was lack of evidence on the impact of MRI on health outcomes of individuals at less than high risk of breast cancer.

Summary of Evidence

For individuals who are asymptomatic with average-risk of breast cancer who receive MRI as an adjunct to screening for breast cancer, the evidence includes systematic reviews and clinical validity studies. Relevant outcomes are OS, disease-specific survival, test accuracy and validity, and resource utilization. The systematic reviews did not identify any RCTs or nonrandomized comparative studies evaluating MRI for screening average-risk individuals. One comparative observational study has been published since the systematic reviews. The diagnostic accuracy of screening tests would likely be lower in this lower prevalence population, and there would be higher false-positive rates, morbidity, and anxiety. The evidence is insufficient to determine that the technology results in an improvement in the net health outcome.

Reference No. 2

Population Reference No. 3

Screening When Breast Characteristics Limit the Sensitivity of Mammography

Clinical Context and Test Purpose

Screening MRI has been suggested for patients who may or may not be at increased risk but who have breast tissue characteristics that limit the sensitivity of mammographic screening (these characteristics are dense breast tissue, breast implants, or scarring after breast-conserving therapy [BCT]). Use of BCT consists of breast-conserving surgery (BCS) followed by radiotherapy.

The questions addressed in this portion of the evidence review:

Does the use of MRI as an adjunct to screening for breast cancer improve the net health outcome of patients who are asymptomatic with breast characteristics that limit the sensitivity of mammography?
Is this degree of increased accuracy likely to improve net health outcomes via the earlier diagnosis and treatment?

The following PICO was used to select literature to inform this review.

Populations

The population of interest is patients with breast characteristics that limit the sensitivity of mammography. For example, individuals who have dense breasts or prior BCT.

Interventions

The intervention of interest is MRI as an adjunct to screening with mammography.

Comparators

The following test is currently being used to make decisions about managing breast cancer: mammography alone.

Outcomes

Breast MRI is performed as an adjunct to routine screening; timing can be guided by national guidelines on breast cancer screening (see the Supplemental Information section).

Study Selection Criteria

For the evaluation of the clinical validity of MRI as an adjunct to screening with mammography, studies that met the following eligibility criteria were considered:

Reported on the accuracy of the marketed version of the technology (including any algorithms used to calculate scores)
Included a suitable reference standard
Patient/sample clinical characteristics were described
Patient/sample selection criteria were described.

Clinically Valid

A test must detect the presence or absence of a condition, the risk of developing a condition in the future, or treatment response (beneficial or adverse).

Review of Evidence

Randomized Controlled and Single Arm Studies: Dense Breasts

One RCT and a prospective observational study were identified that evaluated the use of supplemental MRI in patients who received screening mammography and/or ultrasound. Characteristics of the studies are shown in Table 3.

Table 3. Characteristics of Clinical Validity Studies Assessing Supplemental Breast Magnetic Resonance Imaging for Routine Screening in Women

Study	Study Population	Design	Reference Standard	Identification of Positive MRI Test	Timing of Tests	Blinding of Assessors	Comment
Bakker et al (2019)^10,	Women aged 50 to 75 years in the Netherlands with extremely dense breast tissue with negative results on screening mammography; socioeconomic status was recorded at baseline: 36.1% of women were in the highest status quartile (quartile 1), 23.6% in quartile 3, 22.7% in quartile 2, and 17.4% were in the lowest status quartile 4	Randomized controlled trial	Incidence of interval cancers (positive MRI result that was confirmed histologically) during 2-year screening period	Assessed as BI-RADS category 4 or 5 by 1 radiologist with 5+ years of experience in breast MRI; Patients with BI-RADS category of 3 received follow-up MRI after 6 months	Mammography screening with or without MRI every 2 years	NR	Funded by the University Medical Center Utrecht, the Netherlands Organization for Health Research and Development, the Dutch Cancer Society, the Dutch Pink Ribbon-A Sister's Hope organization, Stichting Kankerpreventie Midden-West, Bayer Pharmaceuticals, and Volpara Health Technologies
Berg et al (2012)^11,	Women aged 25 years and older with heterogeneously dense or extremely dense breast tissue with at least 1 risk factor for breast cancer. Women had undergone 3 negative screenings of mammography and supplemental ultrasound. 93% of women in the study were White; the remainder of women were Hispanic or Latino, Black, Native Hawaiian or Pacific Islander, Asian, or American Indian or Alaskan Native.	Prospective trial	Most severe biopsy result within 365 days of mammographic screening and/or clinical follow-up at 1 year	Assessed as BI-RADS score of 3, 4, or 5	MRI within 8 weeks of last screening mammography	Yes (interpretation was blinded to other test results)	Funded by the Avon Foundation and National Institutes of Health grants

BI-RADS: Breast Imaging Reporting and Data System; MRI: magnetic resonance imaging; NR: not reported.

Results of the clinical validity studies are shown in Table 4. Bakker et al (2019) conducted a multicenter RCT (DENSE) with 40,373 women with extremely dense breast tissue and normal mammography results who were assigned to an optional supplemental MRI or mammography-only screening.^10, There were 8061 patients invited to undergo MRI (MRI-invitation group); however, 4783 patients participated in supplemental MRI screening and 3278 chose not to participate. There were 32,312 patients who only received mammography (mammography-only group). The interval-cancer rate was 2.5 per 1000 screenings in the MRI-invitation group compared to 5.0 per 1000 screenings in the mammography-only group (rate difference, 2.5; 95% confidence interval [CI], 1.0 to 3.7; p<.001). Of note, among the 20 interval cancers diagnosed in the MRI-invitation group, 16 were diagnosed in patients who did not accept the supplemental MRI invitation (4.9 per 1000 screenings), while 4 were diagnosed in patients who underwent MRI screening (0.8 per 1000 screenings). In the 2012 ACRIN (American College of Radiology Imaging Network) 6666 trial, mammography alone was compared with mammography plus ultrasound in women 25 years or older with at least heterogeneously dense breast tissue and at least 1 other breast cancer risk factor.^11, Half (54%) of women had a personal history of breast cancer. In a MRI subanalysis, women who completed 3 rounds of screening and did not have contraindications or renal impairment were asked to undergo contrast-enhanced MRI within 8 weeks of the last screening mammography. Six hundred twenty-seven women consented and were eligible for this subanalysis, and 612 (98%) completed the needed tests; 16 cancers were detected in these women. Sensitivity increased from 44% (95% CI, 20% to 70%) for mammography plus ultrasound to 100% (95% CI, 79% to 100%; p=.004) when MRI was added. Specificity declined from 84% (95% CI, 81% to 87%) for mammography plus ultrasound to 65% (95% CI, 61% to 69%; p<.001) for all 3 tests. Over the 3 year study period, another 9 cancers were identified between screening tests, and 2 additional cancers were identified off-study.

Table 4. Results of Clinical Validity Studies Assessing Supplemental Breast Magnetic Resonance Imaging for Routine Screening in Women

Study	Initial N	Final N	Excluded Images	Cancer Rate	Clinical Validity, % (95% CI)
					Sensitivity	Specificity	PPV	NPV
Bakker et al (2019)^10,	40,373 (8061 were invited to undergo MRI screening)	40,373 (Of 8061 who were invited to undergo MRI screening, 4783 underwent screening)	11 died, 3 moved abroad	Interval Cancer Rate
MRI invitation + mammography				2.5 per 1000 screenings (95% CI, 1.6 to 3.8)	95.2 (88.1 to 98.7)	92 (NR)	Recall for additional testing: 17.4 (14.2 to 21.2) Biopsy: 26.3 (21.7 to 31.6)	NR
Mammography alone				5.0 per 1000 screenings (95% CI, 4.3 to 5.8)	NR	NR	NR	NR
Berg et al (2012)^11,	627 women were screened for the MRI substudy	612 MRI participants	15 were excluded because there was no reference standard	Cancer diagnosis
Supplemental MRI				16 (2.6%) participants	100 (79 to 100)	65 (61 to 69)	19 (11 to 29)	NR
Mammography and ultrasound				NA	44 (20 to 70)	84 (81 to 87)	18 (8 to 34)	NR

 CI: confidence interval; NA: not applicable; MRI: magnetic resonance imaging; NPV: negative predictive value; NR: not reported; PPV: positive predictive value.

Tables 5 and 6 discuss relevant limitations of the studies.

Table 5. Study Relevance Limitations of Clinical Validity Studies of Supplemental Breast Magnetic Resonance Imaging for Routine Screening in Women

Study	Population^a	Intervention^b	Comparator^c	Outcomes^d	Duration of Follow-Up^e
Bakker et al (2019)^10,	4. Enrolled populations do not reflect relevant diversity			1. Health outcomes not reported
Berg et al (2012)^11,	4. Enrolled populations not reflect relevant diversity			1. Health outcomes not reported

The study limitations stated in this table are those notable in the current review; this is not a comprehensive gaps assessment.

^aPopulation key: 1. Intended use population unclear; 2. Study population is unclear; 3. Study population not representative of intended use; 4, Enrolled populations do not reflect relevant diversity; 5. Other. ^bIntervention key: 1. Classification thresholds not defined; 2. Version used unclear; 3. Not intervention of interest.^c Comparator key: 1. Classification thresholds not defined; 2. Not compared to credible reference standard; 3. Not compared to other tests in use for same purpose.^d Outcomes key: 1. Study does not directly assess a key health outcome; 2. Evidence chain or decision model not explicated; 3. Key clinical validity outcomes not reported (sensitivity, specificity, and predictive values); 4. Reclassification of diagnostic or risk categories not reported; 5. Adverse events of the test not described (excluding minor discomforts and inconvenience of venipuncture or noninvasive tests).^e Follow-Up key: 1. Follow-up duration not sufficient with respect to natural history of disease (true-positives, true-negatives, false-positives, false-negatives cannot be determined).

Table 6. Study Design and Conduct Limitations of Clinical Validity Studies of Supplemental Breast Magnetic Resonance Imaging for Routine Screening in Women

Study	Selection^a	Blinding^b	Delivery of Test^c	Selective Reporting^d	Data Completeness^e	Statistical^f
Bakker et al (2019)^10,		1. Not blinded to test groups
Berg et al (2012)^11,			4. Expertise of evaluators not described.			2. Comparison with other tests not reported.

The study limitations stated in this table are those notable in the current review; this is not a comprehensive gaps assessment.^aSelection key: 1. Selection not described; 2. Selection not random or consecutive (ie, convenience).^bBlinding key: 1. Not blinded to results of reference or other comparator tests.^cTest Delivery key: 1. Timing of delivery of index or reference test not described; 2. Timing of index and comparator tests not same; 3. Procedure for interpreting tests not described; 4. Expertise of evaluators not described.^d Selective Reporting key: 1. Not registered; 2. Evidence of selective reporting; 3. Evidence of selective publication.^e Data Completeness key: 1. Inadequate description of indeterminate and missing samples; 2. High number of samples excluded; 3. High loss to follow-up or missing data.^f Statistical key: 1. Confidence intervals and/or p values not reported; 2. Comparison with other tests not reported.

Observational Studies: Following Breast-Conserving Therapy

Two prospective studies have reported on the performance of surveillance breast MRI following BCT.^12,^13, Study characteristics are shown in Table 7. Both studies were performed in Korea and it is unclear whether the populations overlapped.

Table 7. Characteristics of Clinical Validity Studies Assessing Surveillance Breast Magnetic Resonance Imaging After Breast-Conserving Therapy

Study	Study Population	Design	Reference Standard	Identification of Positive MRI Test	Timing of Tests	Blinding of Assessors	Comment
Kim et al (2017)^13,	Women in Korea undergoing surveillance breast MRI following BCT from 2014 to 2016	Prospective observational	Pathology for positive results Cancer not confirmed at 1-year surveillance imaging for negative results	Assessed as BI-RADS category 4 or 5 by 1 radiologist with 10+ years of experience in breast MRI	MRI within 4 wk of screening mammography and breast US	No (readers knew results of prior imaging studies)	Funded by Bayer Korea
Cho et al (2017)^12,	Women aged ≤50 years in Korea undergoing surveillance breast MRI following BCT from 2010 to 2016	Prospective observational	Pathology for positive results Cancer not confirmed at 1-year surveillance imaging for negative results	Assessed as BI-RADS category 3+ by 1 radiologist with 5+ years of experience in breast MRI	MRI within 2 mo of screening mammography and breast US	Yes	Funded by Bayer Korea Overlap with Kim (2017) unclear

BCT: breast-conserving therapy; BI-RADS: Breast Imaging Reporting and Data System; MRI: magnetic resonance imaging; US: ultrasound.

Results of the clinical validity studies for surveillance of breast MRI following BCT are shown in Table 8. The sensitivity of MRI was higher than mammography and ultrasound with overlapping CIs in both studies. Specificity of MRI was lower than mammography and ultrasound. The combination of mammography and MRI was 100% sensitive and 87% specific. The review by Cho et al (2017) reported that the recall rate was significantly higher for mammography plus MRI (13.8%; 95% CI, 12.0% to 15.5%) compared with mammography alone (4.4%; 95% CI, 3.3% to 5.5%), as was the biopsy rate (2.7% [95% CI, 2.0% to 3.4%] vs. 0.5 [95% CI, 0.2% to 0.8%]). The yield per 1000 examinations was 8.2 (95% CI, 4.3 to 12.2) for mammography plus MRI versus 4.4 (95% CI, 1.5 to 7.2) for mammography.^12,

Table 8. Results of Clinical Validity Studies Assessing Surveillance Breast Magnetic Resonance Imaging After Breast-Conserving Therapy

Study	Initial N	Final N	Excluded Images	Recurrence Rate, %	Clinical Validity (95% CI),%
					Sensitivity	Specificity	PPV	NPV
Kim et al (2017)^13,	421 women (429 breast MRIs)	414 women (422 breast MRIs)	Initial diagnosis of malignant phyllodes tumor, lobular carcinoma in situ (n=6), or developed supraclavicular lymph node metastasis within 12 mo (n=1)	2.6
MRI					82 (48 to 98)	95 (92 to 97)	31 (15 to 51)	99 (98 to 100)
US					18 (2 to 52)	98 (96 to 99)	20 (3 to 56)	98 (96 to 99)
Mammography					18 (2 to 52)	99 (98 to 100)	40 (5 to 85)	98 (96 to 99)
Cho et al (2017)^12,	801	754	Withdrew consent (n=39) or had systemic metastasis (n=7); unclear (n=1)	2.3
MRI					88 (66 to 97)	90 (88 to 91)	24 (14 to 37)	NR
US					65 (41 to 83)	90 (89 to 92)	35 (19 to 55)	NR
Mammography					53 (31 to 74)	96 (95 to 97)	73 (43 to 90)	NR
Mammography plus MRI					100 (82 to 100)	87 (85 to 89)	29 (18 to 42)	NR

 CI: confidence interval; MRI: magnetic resonance imaging; NPV: negative predictive value; NR: not reported; PPV: positive predictive value; US: ultrasound.

Tables 9 and 10 display notable limitations identified in each study.

Table 9. Study Relevance Limitations of Clinical Validity Studies of Surveillance Breast Magnetic Resonance Imaging After Breast-Conserving Therapy

Study	Population^a	Intervention^b	Comparator^c	Outcomes^d	Duration of Follow-Up^e
Kim et al (2017)^13,				1. Health outcomes not reported
Cho et al (2017)^12,				1. Health outcomes not reported

The study limitations stated in this table are those notable in the current review; this is not a comprehensive gaps assessment.^aPopulation key: 1. Intended use population unclear; 2. Clinical context is unclear; 3. Study population is unclear; 4. Study population not representative of intended use.^bIntervention key: 1. Classification thresholds not defined; 2. Version used unclear; 3. Not intervention of interest.^c Comparator key: 1. Classification thresholds not defined; 2. Not compared to credible reference standard; 3. Not compared to other tests in use for same purpose.^d Outcomes key: 1. Study does not directly assess a key health outcome; 2. Evidence chain or decision model not explicated; 3. Key clinical validity outcomes not reported (sensitivity, specificity, and predictive values); 4. Reclassification of diagnostic or risk categories not reported; 5. Adverse events of the test not described (excluding minor discomforts and inconvenience of venipuncture or noninvasive tests).^e Follow-Up key: 1. Follow-up duration not sufficient with respect to natural history of disease (true-positives, true-negatives, false-positives, false-negatives cannot be determined).

Table 10. Study Design and Conduct Limitations of Clinical Validity Studies of Surveillance Breast Magnetic Resonance Imaging After Breast-Conserving Therapy

Study	Selection^a	Blinding^b	Delivery of Test^c	Selective Reporting^d	Data Completeness^e	Statistical^f
Kim et al (2017)^13,		1. Not blinded to results of mammography, US, or PET/CT
Cho et al (2017)^12,

 CT: computed tomography; PET: positron emission tomography; US: ultrasound.The study limitations stated in this table are those notable in the current review; this is not a comprehensive gaps assessment.^aSelection key: 1. Selection not described; 2. Selection not random or consecutive (ie, convenience).^bBlinding key: 1. Not blinded to results of reference or other comparator tests.^cTest Delivery key: 1. Timing of delivery of index or reference test not described; 2. Timing of index and comparator tests not same; 3. Procedure for interpreting tests not described; 4. Expertise of evaluators not described.^d Selective Reporting key: 1. Not registered; 2. Evidence of selective reporting; 3. Evidence of selective publication.^e Data Completeness key: 1. Inadequate description of indeterminate and missing samples; 2. High number of samples excluded; 3. High loss to follow-up or missing data.^f Statistical key: 1. Confidence intervals and/or p values not reported; 2. Comparison with other tests not reported.

Section Summary: Screening When Breast Characteristics Limit the Sensitivity of Mammography

The RCT from the Netherlands (Bakker 2019) found that among women with dense breasts, the use of MRI increased the cancer detection rate and decreased the interval cancer rate compared to mammography. However, the false positive rate was 79.8 per 1000 screenings. The trial is continuing in order to assess the effects over time of adjunctive screening with MRI. The prospective cohort trial by the American College of Radiology Imaging Network (ACRIN 6666; Berg 2012) found that the addition of MRI resulted in high cancer detection, but with increased false positive findings. The evidence is insufficient to show that the use of adjunctive MRI to screen average risk individuals who have dense breasts improves the net health outcome.

Two studies assessed the addition of MRI to mammography for surveillance of women who had been treated for cancer with BCT. The sensitivity of adjunct MRI was greater than mammography alone, but with overlapping confidence intervals. The companion study of women under 50 years showed higher cancer detection rates with adjunct MRI but lower specificity than mammography alone; the authors suggested that adjunctive mammography improves detection of early stage but biologically aggressive cancer in the population of younger women. However, to the extent that younger women may constitute a higher risk population, the delineation of MRI for screening high risk individuals is addressed in high risk screening section of this policy. The evidence is insufficient to demonstrate that adjunctive MRI for screening improves the net health outcome when breast characteristics limit the sensitivity of mammography.

Summary of Evidence

For individuals with characteristics limiting the accuracy of mammography (eg, dense breasts) who receive MRI as an adjunct to screening for breast cancer, the evidence includes a systematic review (TEC Assessment), RCT, and diagnostic accuracy studies. Relevant outcomes are OS, disease-specific survival, test accuracy and validity, and resource utilization. There are limited data on the diagnostic accuracy of MRI versus mammography in patients who have had BCT or who have dense breasts. The evidence is insufficient to determine that the technology results in an improvement in the net health outcome.

Reference No. 3

Population Reference No. 4

Detection Uses

Detecting Suspected Occult Breast Primary Tumor With Axillary Nodal Adenocarcinoma With a Negative Mammography and Physical Exam

Clinical Context and Test Purpose

Breast MRI has been advocated to help detect suspected occult primary breast cancer in patients with adenocarcinoma in the axillary lymph nodes after mammography and physical exam have failed to reveal a breast tumor. Localization of a primary breast tumor might permit BCT instead of presumptive mastectomy.

The questions addressed in this portion of the evidence review:

Does the use of MRI as an adjunct to detect breast cancer eligible for BCT improve the net health outcome compared to standard techniques in individuals with suspected occult breast primary tumor with axillary nodal adenocarcinoma and negative mammography?
Is this degree of increased accuracy likely to improve net health outcomes via the earlier diagnosis, better patient management decisions, and more appropriate treatment?

The following PICO was used to select literature to inform this review.

Populations

The population of interest is individuals with suspected occult breast primary tumor with axillary nodal adenocarcinoma and negative mammography.

Interventions

The intervention of interest is MRI examination as an adjunct to detect breast cancer eligible for BCT.

Comparators

The comparator of interest is a preemptive mastectomy.

Outcomes

The outcomes of interest for diagnostic accuracy include test accuracy and test validity (ie, sensitivity, specificity). Primary outcomes of interest for clinical utility are the avoidance of invasive procedures (eg, biopsy, mastectomy), the ability to detect cancer that would require additional or earlier treatment, and overall mortality and breast cancer-specific mortality rates.

Breast MRI is performed after a positive breast cancer screening or diagnostic examination.

Study Selection Criteria

For the evaluation of the clinical validity of MRI as an adjunct to detect breast cancer eligible for BCT, studies that met the following eligibility criteria were considered:

Reported on the accuracy of the marketed version of the technology (including any algorithms used to calculate scores)
Included a suitable reference standard
Patient/sample clinical characteristics were described
Patient/sample selection criteria were described.

Clinically Valid

A test must detect the presence or absence of a condition, the risk of developing a condition in the future, or treatment response (beneficial or adverse).

Review of Evidence

Systematic Reviews

De Besser et al (2010) evaluated 8 retrospective studies in a systematic review of studies on the use of MRI in patients (N=220) with mammographically occult breast cancer and an axillary metastasis.^14, In 7 studies, a potential primary lesion was detected in a mean of 72% of cases (range, 36% to 86%). Pooling individual patient data yielded a sensitivity of 90% (range, 85% to 100%) in detecting an actual malignant tumor. Specificity, however, was 31% (range, 22% to 50%).

Clinically Useful

Direct Evidence

Direct evidence of clinical utility is provided by studies that have compared health outcomes for patients managed with and without the test. Because these are intervention studies, the preferred evidence would be from RCTs.

Evidence on detection of suspected occult breast cancer is based on a TEC Assessment (2004)^15, and a subsequent meta-analysis, which appear to be the only direct evidence available for this indication. The Assessment concluded that, in this small subgroup of patients, adjunctive use of breast MRI allowed a substantial portion of patients (25% to 61%) to avoid the morbidity of mastectomy; risk of the unnecessary biopsy was estimated to be 8%.

Section Summary: Detecting Suspected Occult Breast Primary Tumor With Axillary Nodal Adenocarcinoma With a Negative Mammography and Physical Exam

The use of MRI to guide BCS rather than presumptive mastectomy appears to offer the substantial benefit of breast conservation for those patients in whom MRI detects the primary tumor.

Summary of Evidence

For individuals who have suspected occult breast primary tumor with axillary nodal adenocarcinoma with negative mammography who receive MRI as an adjunct to detect breast cancer eligible for BCT , the evidence includes a systematic review (TEC Assessment) and meta-analysis. Relevant outcomes are OS, disease-specific survival, test accuracy and validity, and resource utilization. The studies found that adjunctive use of breast MRI to guide BCS rather than preemptive mastectomy allowed a substantial portion of patients to avoid the morbidity of mastectomy. The evidence is sufficient to determine that the technology results in an improvement in the net health outcome.

Reference No. 4

Population Reference No. 5

Detecting Contralateral Breast Cancer After Established Breast Cancer

Clinical Context and Test Purpose

Patients with a diagnosed breast cancer are at higher risk for a synchronous or subsequent breast cancer in the contralateral breast, and breast MRI has been suggested as a more sensitive screening test compared to mammography.

The questions addressed in this portion of the evidence review:

Does the use of MRI as an adjunct to detect breast cancer in the contralateral breast improve the net health outcome compared to standard techniques in individuals with suspected occult breast primary tumor with axillary nodal adenocarcinoma and negative mammography?
Is this degree of increased accuracy likely to improve net health outcomes via the earlier diagnosis, better patient management decisions, and more appropriate treatment?

The following PICO was used to select literature to inform this review.

Populations

The population of interest is individuals with breast cancer.

Interventions

The intervention of interest is MRI examination as an adjunct to detect breast cancer in the contralateral breast.

Comparators

The comparator of interest is mammography and clinical assessment alone.

Outcomes

The outcomes of interest for diagnostic accuracy include test accuracy and test validity (ie, sensitivity, specificity). Primary outcomes of interest for clinical utility are the avoidance of invasive procedures (eg, biopsy, mastectomy), the ability to detect cancer that would require additional or earlier treatment, and overall mortality and breast cancer-specific mortality rates.

Breast MRI is performed after a positive breast cancer screening or diagnostic examination.

Study Selection Criteria

For the evaluation of the clinical validity of MRI examination as an adjunct to detect breast cancer in the contralateral breast, studies that met the following eligibility criteria were considered:

Reported on the accuracy of the marketed version of the technology (including any algorithms used to calculate scores)
Included a suitable reference standard
Patient/sample clinical characteristics were described
Patient/sample selection criteria were described.

Clinically Valid

A test must detect the presence or absence of a condition, the risk of developing a condition in the future, or treatment response (beneficial or adverse).

Review of Evidence

Single Arm Studies

Lehman et al (2007) reported on the results of the ACRIN-A6667 trial.^16, They found that 30 (3%) of 969 women with a recent diagnosis of unilateral breast cancer had contralateral cancer at the time of initial diagnosis using MRI. Contralateral lesions were not detected by mammography or physical exam. Eighteen (60%) of the 30 cancers were invasive and 12 (40%) were ductal carcinoma in situ (DCIS). In this study, 121 (12.5%) patients had biopsies, with a positive biopsy rate of 24.8%. With 1-year follow-up, the sensitivity of MRI was 91% and specificity was 88%. Results of this trial in a diverse group of patients were similar to the findings of others.

Liberman et al (2003) reported on 212 women who had negative mammograms of the asymptomatic contralateral breast and found 12 cancers (prevalence, 5%) on MRI, including 6 DCIS and 6 infiltrating carcinomas.^17, However, the PPV of these findings was only 20%, with a specificity of 76%. Lehman et al (2005) found 4 contralateral cancers in 103 patients; in this study, 10 biopsies were done.^18,

Clinically Useful

Direct Evidence

No RCTs assessing diagnostic breast MRI in individuals with suspected contralateral breast cancer after established breast cancer were identified.

Chain of Evidence

Indirect evidence on clinical utility rests on clinical validity. If the evidence is insufficient to demonstrate test performance, no inferences can be made about clinical utility.

A trial with nearly 1000 women found that MRI had high sensitivity and reasonably high specificity for identifying contralateral lesions not detected by mammography or physical examination. Although long-term outcomes of contralateral breast cancers are not fully known, important management changes will occur based on such findings, and these management changes should lead to improved outcomes.

Section Summary: Detecting Contralateral Breast Cancer After Established Breast Cancer

The available evidence suggests that adjunctive MRI can identify contralateral breast cancers in women with negative mammograms. A trial with nearly 1000 women found that MRI had high sensitivity and reasonably high specificity for identifying contralateral lesions not detected by mammography or physical examination. Although long-term outcomes of contralateral breast cancers are not fully known, important changes in management will occur as a result of the findings, and these management changes should lead to improved outcomes. That is, in addition to the presumed benefits of early detection, simultaneous treatment of synchronous cancers can occur rather than multiple treatments on separate occasions.

Summary of Evidence

Reference No. 5

Population Reference No. 6

Detecting Breast Cancer in the Case of Low-Suspicion Findings on Conventional Mammography

Clinical Context and Test Purpose

Patients with abnormal findings on mammography are categorized according to the level of suspicion of the findings. Patients with low-suspicion findings are often recommended to undergo short-interval follow-up after 3 to 6 months (instead of immediate biopsy). This follow-up may continue for 2 years to demonstrate the stability of benign findings or to detect progression; progression would indicate the need for biopsy. Breast MRI has been investigated as a more sensitive technique to further characterize low-suspicion breast lesions, so that patients with MRI-negative lesions may be reassured and avoid prolonged follow-up and those with MRI-positive lesions may be referred for early biopsy, possibly leading to earlier diagnosis and treatment.

The question addressed in this portion of the evidence review is: In patients with a diagnosis of low-suspicion findings on conventional testing not indicated for immediate biopsy, are net health outcomes improved by the use of MRI as an adjunct compared to standard care with short-interval mammographic follow-up to detect breast cancer?

The following PICO was used to select literature to inform this review.

Populations

The population of interest is individuals with low-suspicion findings on conventional mammography.

Interventions

The intervention of interest is MRI examination as an adjunct to standard care with short-interval mammographic follow-up.

Comparators

The comparator of interest is standard care and short-interval mammographic follow-up.

Outcomes

The outcomes of interest for diagnostic accuracy include test accuracy and test validity (ie, sensitivity, specificity). Primary outcomes of interest for clinical utility are the avoidance of invasive procedures (eg, biopsy, mastectomy), the ability to detect cancer that would require additional or earlier treatment, and overall mortality and breast cancer-specific mortality rates.

Breast MRI is performed after a positive breast cancer screening or diagnostic examination.

Study Selection Criteria

For the evaluation of the clinical validity of MRI examination as an adjunct to standard care with short-interval mammographic follow-up, studies that met the following eligibility criteria were considered:

Reported on the accuracy of the marketed version of the technology (including any algorithms used to calculate scores)
Included a suitable reference standard
Patient/sample clinical characteristics were described
Patient/sample selection criteria were described.

Clinically Valid

A test must detect the presence or absence of a condition, the risk of developing a condition in the future, or treatment response (beneficial or adverse).

See the Clinically Useful section for discussion.

Clinically Useful

Review of Evidence

Direct Evidence

Chain of Evidence

Indirect evidence on clinical utility rests on clinical validity. If the evidence is insufficient to demonstrate test performance, no inferences can be made about clinical utility.

Because the clinical validity of adjunctive MRI has not been established, a chain of evidence supporting the clinical utility of this modality cannot be constructed.

Section Summary: Detecting Breast Cancer in the Case of Low-Suspicion Findings on Mammography

Currently, there is a lack of direct evidence supporting use for this indication. Well-designed prospective confirmatory studies would be necessary to permit conclusions about the effect of this adjunctive use of breast MRI on health outcomes.

Summary of Evidence

Reference No. 6

Population Reference No. 7

Detecting Breast Cancer by Further Characterizing Suspicious Breast Lesions

Clinical Context and Test Purpose

Breast lesions detected by clinical exam or mammography that are considered suspicious are frequently referred for biopsy; however, only a minority of such biopsies reveal breast cancer due to the relatively low specificity of clinical and radiologic exams. Breast MRI has been investigated as a technique to further characterize suspicious breast lesions so that patients with benign lesions may be spared a biopsy procedure. One infrequent situation (niche use) in which MRI of the breast may be helpful and improve health outcomes is in the management of patients who have a suspicious lesion that can only be seen on one mammographic view (ie, the lesion cannot be seen in other views or on an ultrasound). Patients who fall under this category have a lesion that is not palpable, and therefore, percutaneous biopsy localization cannot be performed. Instead, MRI would be used to localize the suspicious lesion and permit biopsy (this technique would presumably lead to earlier diagnosis of breast cancer as opposed to waiting until the lesion was visible on 2 mammographic views or on ultrasound). The previously described scenario is an infrequent occurrence, so the evidence base addressing this use is mainly anecdotal, but the clinical rationale supporting this use is good.

The question addressed in this portion of the evidence review is: Does MRI as an adjunct to further characterize breast lesions lead to better net health outcomes than biopsy based on mammography and clinical assessment?

The following PICO was used to select literature to inform this review.

Populations

The population of interest is individuals with suspicious breast lesions.

Interventions

The intervention of interest is MRI examination as an adjunct to mammography and clinical assessment.

Comparators

The comparator of interest is biopsy based on mammography and clinical assessment.

Outcomes

The outcomes of interest for diagnostic accuracy include test accuracy and test validity (ie, sensitivity, specificity). Primary outcomes of interest for clinical utility are the avoidance of invasive procedures (eg, biopsy, mastectomy), the ability to detect cancer that would require additional or earlier treatment, and overall mortality and breast cancer-specific mortality rates.

Use of MRI is performed after a positive breast cancer screening or diagnostic examination.

Study Selection Criteria

For the evaluation of the clinical validity of MRI examination as an adjunct to mammography and clinical assessment, studies that met the following eligibility criteria were considered:

Reported on the accuracy of the marketed version of the technology (including any algorithms used to calculate scores)
Included a suitable reference standard
Patient/sample clinical characteristics were described
Patient/sample selection criteria were described.

Clinically Valid

A test must detect the presence or absence of a condition, the risk of developing a condition in the future, or treatment response (beneficial or adverse).

Review of Evidence

Systematic Reviews

A systematic review published by Medeiros et al (2011) analyzed 69 studies including 9298 women.^19, Pooled sensitivity was 90% (95% CI, 88% to 92%), and pooled specificity was 75% (95% CI, 70% to 79%). The pooled positive likelihood ratio of an abnormal MRI for malignancy was 3.6 (95% CI, 3.0 to 4.2) and the pooled negative likelihood ratio was 0.12 (95% CI, 0.09 to 0.15). For breast cancer or high-risk lesions versus benign lesions, the area under the curve for MRI was 0.91.

A systematic review published by Zhang et al (2022) included 29 studies with 2976 patients and 3365 suspicious breast lesions.^20, The sensitivity and specificity of MRI features in differentiating malignant from benign breast lesions ranged from 73.8% to 91.9% and from 33.9% to 85.4%, respectively. The enrolled studies showed high heterogeneity. For differentiating malignant from benign breast lesions, the area under the curve values of MRI features; irregular shape, noncircumscribed margin, mass enhancement, heterogeneous internal enhancement, and type II or III time intensity curve patterns were 0.79, 0.87, 0.63, 0.82, and 0.89, respectively.

Single Arm Studies

Two single-institution, prospective cohort studies examined the diagnostic accuracy of breast MRI for lesions identified by mammography or ultrasound. Strobel et al (2015) in Germany included lesions characterized as Breast Imaging Reporting and Data System (BI-RADS) category 4 by conventional workup in 340 women.^21, Most women were postmenopausal (61%), had no previous breast biopsy (64%), or family history of breast cancer (62%), and underwent initial evaluation for routine screening (88%). Of 353 lesions, 135 (38%) were biopsied; lesions down-graded to BI-RADS categories 1, 2, or 3 on MRI were followed with imaging for 18 months, except for pure clustered microcalcifications (without accompanying mass), which were biopsied or followed with imaging for 24 months at patient discretion; none of the lesions monitored progressed during follow-up. The overall incidence of malignancy including DCIS was 20% (n=69). The MRI down-graded 256 (28%) of 353 lesions, confirmed 37 (11%) lesions, and upgraded 50 (14%) lesions. The PPV of MRI was 73% compared with 19% for conventional imaging. The negative predictive value (NPV) of MRI was 99% (and could not be calculated for conventional imaging). For pure clustered microcalcifications, sensitivity was 89% (25/28 lesions) and the false-negative rate was 12% (3/28 lesions). False-positive MRI findings resulted in a biopsy for 5 (1.5%) of 340 women.

In a similar study, Li et al (2014) in China included 84 women with BI-RADS categories 3, 4, or 5 microcalcifications on mammography.^22, Most patients were premenopausal (81%), had no family history of breast cancer (83%), and underwent initial evaluation for routine screening (56%). All lesions were biopsied surgically (n=91). The incidence of malignancy including DCIS was 46%. The PPV of MRI was 87% compared with 60% for mammography. The NPV of the MRI was 91%.

de Oliveira Pereira et al (2020) performed a cross-sectional study in Brazil of 32 women with suspected breast tumor based on findings from mammography, ultrasonography, or MRI.^23, The mean age of patients was 54.6 years, and the mean breast lump size was 1.6 cm. The sensitivity, specificity, PPV, and NPV were 100%, 50%, 66.7%, and 100%, respectively, for MRI; 56.2%, 87.5%, 81.8%, and 66.7% for mammography; and 75%, 18.8%, 48%, and 42.8% for ultrasonography.

Clinically Useful

Review of Evidence

Direct Evidence

No RCTs assessing diagnostic breast MRI in individuals to further characterize suspicious breast lesions were identified.

Chain of Evidence

Indirect evidence on clinical utility rests on clinical validity. If the evidence is insufficient to demonstrate test performance, no inferences can be made about clinical utility.

Available evidence has not shown this use of breast MRI would improve health outcomes. Considering the relative ease of breast biopsy, the sensitivity of breast MRI would have to be virtually 100% to confidently avoid biopsy. Although MRI performs well, it is clear that the sensitivity is not 100%. False-negative results tend to occur, particularly in certain subcategories, such as DCIS, but invasive carcinomas may not be detected on MRI, also leading to false-negative results. The potential harm to health outcomes of failing to diagnose breast cancer or at least of delaying the diagnosis of breast cancer is of significant concern.

Section Summary: Detecting Breast Cancer by Further Characterizing Suspicious Breast Lesions

Use of MRI for evaluation of suspicious breast lesions has relatively high sensitivity and a moderately high specificity. However, it has not yet been established whether the NPV is sufficient to preclude the need for biopsy. Although 3 more recent studies have reported NPVs greater than 90% in certain types of breast lesions, these studies were conducted in single, non-U.S. institutions that require replication in larger, multicenter trials. Therefore, the use of MRI to further characterize suspicious lesions is currently unlikely to alter clinical management. In addition, the fairly high rate of false-positives will lead to substantial numbers of unnecessary biopsies.

Summary of Evidence

Reference No. 7

Population Reference No. 8

Treatment-Related Uses

Treatment-related uses addressed here are surgical planning, evaluating tumor response to neoadjuvant therapy, and evaluating residual tumor after BCT. Preoperative planning includes identification of multicentric disease in clinically localized breast cancer; surgical decisions after neoadjuvant chemotherapy; evaluation of suspected chest wall involvement; and localizing lesions prior to biopsy.

For each of these indications, study selection prioritized systematic reviews focusing on the relevant patient population and purpose. Systematic reviews were supplemented by studies of clinical validity. For the evaluation of clinical validity of MRI examination for the proposed purpose, studies that met the following eligibility criteria were considered:

Reported on the accuracy of the marketed version of the technology (including any algorithms used to calculate scores)
Included a suitable reference standard
Patient/sample clinical characteristics were described
Patient/sample selection criteria were described.

In addition, we sought studies of clinical usefulness. These are studies that report the outcomes of using MRI for the proposed purpose, with preference for RCTs.

Objective: Surgical Planning

The question addressed in this portion of the evidence review is whether the use of MRI evaluation as an adjunct to guide treatment planning (eg, surgical approach) for patients with known or suspected breast cancer improves the net health outcome compared with standard techniques.

The sections on surgical planning address 4 specific indications (1) identification of multicentric disease in clinically localized breast cancer; (2) surgical decisions after neoadjuvant chemotherapy; (3) evaluation of suspected chest wall involvement; and (4) localizing lesions prior to biopsy.

Preoperative Mapping to Identify Multicentric Disease With Clinically Localized Breast Cancer

Clinical Context and Test Purpose

Patients with clinically localized breast cancer are considered candidates for BCS followed by radiotherapy. However, mastectomy may be considered in patients with multicentric disease (in a separate quadrant of the breast). Breast MRI has been investigated as a technique to assess the extent of the tumor in the breast, specifically to detect multicentric disease as an aid to surgical planning.

The question addressed in this evidence review is: Does the use of MRI for preoperative mapping to identify multicentric disease improve net health outcomes?

The following PICO was used to select literature to inform this review.

Populations

The populations of interest is individuals with clinically localized breast cancer.

Interventions

The intervention of interest is MRI as an adjunct to standard evaluation methods.

Comparators

The following tests and practices are currently being used to make decisions about managing breast cancer: standard workup without MRI.

Outcomes

Relevant outcomes of interest for diagnostic accuracy include test accuracy and test validity (ie, sensitivity, specificity). Primary outcomes of interest for clinical utility include avoidance of invasive procedures (eg, biopsy, mastectomy), the ability to detect cancer requiring additional or earlier treatment, and overall mortality and breast cancer-specific mortality rates.

Breast MRI is performed after identification of suspicious breast lesions, or before or after treatment for breast cancer.

Study Selection Criteria

Methodologically credible studies were selected using the following principles:

To assess efficacy outcomes, comparative controlled prospective trials were sought, with a preference for RCTs;
In the absence of such trials, comparative observational studies were sought, with a preference for prospective studies.
To assess long-term outcomes and adverse events, single-arm studies that capture longer periods of follow-up and/or larger populations were sought.
Consistent with a 'best available evidence approach,' within each category of study design, studies with larger sample sizes and longer duration were preferred.
Studies with duplicative or overlapping populations were excluded.

Review of Evidence

Systematic Reviews

Several meta-analyses have evaluated evidence on additional disease detected by MRI and changes in clinical management, most of which were by the same research group.^24,^25,^26,^27,^28,^29,

Li et al (2022) conducted a systematic review of 19 studies (4 RCTs, 15 observational) that evaluated the efficacy of preoperative MRI in patients with invasive breast cancer.^29, All breast cancer types were included but patients had to be undergoing curative surgery (eg, excision or BCS). All studies included a control group. The primary outcome, mastectomy rate, was significantly increased with preoperative MRI (odds ratio [OR], 1.36; 95% CI, 1.13 to 1.64; p=.001; I²=91%) based on data from 16 studies (n=86,075). Preoperative MRI significantly reduced the rate of reoperation (OR, 0.77; 95% CI, 0.62 to 0.97; p=.02; I²=71%). Other outcomes, including primary BCS, secondary mastectomy, and the rate of positive margins, were not significantly different between groups. An analysis of 3 studies in patients with invasive lobular carcinoma found similar results for all outcomes among patients who did and did not receive preoperative MRI.

The most recent meta-analysis published by Houssami et al was in 2017.^26, Studies included in the review were comparative (randomized or nonrandomized), evaluated preoperative MRI versus an alternative approach that did not include MRI, and reported quantitative data on surgical outcomes. The primary endpoint for the meta-analysis was whether patients underwent mastectomy as surgical treatment. Secondary endpoints were re-excision rates after BCS, positive margins after BCS, and receipt of contralateral prophylactic mastectomy. Nineteen studies met the inclusion criteria-3 RCTs and 16 nonrandomized comparative studies. For the primary study endpoint, a pooled analysis of 15 studies (N=85,975) found significantly greater odds of receiving a mastectomy after preoperative MRI than after no MRI OR , 1.39; 95% CI, 1.23 to 1.57; p<.001). Findings were the same in analyses stratified by publication dates, suggesting that the higher mastectomy rates were not limited to older studies conducted when the MRI-guided biopsy was less common. In an analysis limited to patients with invasive lobular cancer, there was no significant difference in the odds of mastectomy (6 studies: pooled OR ; 1.00; 95% CI, 0.75 to 1.33; p=.988) or the odds of re-excision (5 studies: OR, 0.65; 95% CI, 0.35 to 1.24; p=.192). Among the secondary outcomes, a pooled analysis of 3 studies found a significantly higher odds of contralateral prophylactic mastectomy after MRI (OR, 1.91; 95% CI, 1.25 to 2.91). There were no significant differences between groups on other secondary outcomes (ie, re-excision rates, positive margins, reoperation rates).

One meta-analysis has addressed breast cancer recurrence rates. This meta-analysis, by Houssami et al (2014), analyzed individual patient data from 4 studies-1 RCT and 3 nonrandomized comparative studies (N=3180).^28, Most patients (62% to 93%) had localized, invasive disease and received BCT and systemic chemotherapy. After a median follow-up of 2.9 years (interquartile range [IQR], 1.6 to 4.5 years), there was no difference in estimated 8-year ipsilateral local (adjusted hazard ratio [HR], 0.88; 95% CI, 0.52 to 1.51; p=.65) or distant (adjusted HR, 1.18; 95% CI, 0.76 to 2.27; p=.48) recurrence-free survival overall or in patients who received BCT only.

Randomized Controlled Trials

Since the publication of the Houssami et al (2017) meta-analysis, Bruck et al (2018) reported on the results of an RCT to evaluate the diagnostic value of preoperative MRI in 100 patients with newly diagnosed unifocal stage I invasive ductal carcinoma.^30, Patients were randomized in a 1:1 ratio to preoperative breast MRI or surgery without MRI. Breast MRI detected an additional finding in 14 patients (28%) and MRI detected lesions in 7 (14%) patients, that were confirmed to be malignant. Seven (14%) patients underwent breast reoperation in the MRI group compared with 12 (24%) patients in the control group (p=.20). Definitive mastectomy was performed in 6 (12%) patients in the MRI group compared with 2 (4%) in the control group (p=.14).

Mota et al (2023) conducted a single-center, open-label RCT (BREAST-MRI) in patients with breast cancer undergoing breast conserving surgery.^31, Two hundred fifty seven patients received preoperative MRI and 267 patients served as controls. Local relapse-free survival (p=.7), overall survival (p=.8), and reoperation rates (p=.85) were similar between groups; however, 21 patients underwent mastectomy in the MRI group compared to 1 patient in the control group.

A discussion of the 3 RCTs included in the Houssami et al (2017) meta-analysis (described above) is as follows.

The RCT by Gonzalez et al (2014) in Sweden assessed 440 women who underwent surgical treatment of invasive breast cancer with or without presurgical breast MRI.^32, Breast MRI provided incremental information that altered the treatment plan in 40 (18%) of 220 patients in the MRI group. Conversion from planned BCS to mastectomy occurred more often in the MRI group (20%) than in the control group (10%; p=.024). However, more patients in the MRI group had planned BCS at baseline (70%) than in the control group (60%; p=.036). The ipsilateral reoperation rate was 5% in the MRI group versus 15% in the control group (p<.001). Reoperation rates among those initially planned for BCS were 5% and 22%, respectively (p<.001).

A second RCT, the preoperative MRI and surgical management in patients with nonpalpable breast cancer trial, was reported by Peters et al (2011).^33, It randomized 463 patients with suspicious, nonpalpable breast lesions identified by mammography or ultrasound to prebiopsy MRI or usual care. Of 207 evaluable patients in the MRI group, 11 additional suspicious lesions were identified on MRI and were occult on other imaging studies. All 11 additional lesions underwent biopsy, with 2 (18%) positive for malignancy. The incidence of mastectomy was similar between groups (32% vs. 34% ; p=.776), as was the incidence of BCS (68% vs. 66%). The incidence of re-excisions due to positive tumor margins was significantly greater in the MRI group (34%) than in the control group (12%; p=.008).

A multicenter RCT from the U.K., Comparative effectiveness of MRI in breast cancer trial, reported by Turnbull et al (2010), examined the impact of presurgical MRI on the need for additional treatment within 6 months.^34, This study was an open, parallel-group trial conducted at 45 centers in the U.K. and enrolled 1623 women with biopsy-proven breast cancer who were scheduled for wide local excision BCT. Of 816 patients in the MRI group, 58 (7%) underwent mastectomy as a result of MRI findings and/or patient choice, compared with 10 (1%) patients in the no-MRI group who underwent mastectomy by patient choice. There was no statistically significant reduction in reoperation rates in those who received MRI scans (19% in both groups; OR, 0.96; 95% CI, 0.75 to 1.24; p=.77). In the MRI group, 19 (2%) patients had a "pathologically avoidable" mastectomy, defined as a mastectomy based on MRI results showing more extensive disease but histopathology showing only localized disease. Twelve months after surgery, there was no statistically significant difference in the quality of life between groups.

Observational Studies

In addition to the RCTs, Onega et al (2018) reported on the association between preoperative MRI and all-cause mortality in 5 registries (N=4454) of the National Cancer Institute-sponsored Breast Cancer Surveillance Consortium.^35, Data from the Breast Cancer Surveillance Consortium registries were linked to Medicare claims data or electronic health records; women ages 66 years and older with initial nonmetastatic breast cancer (stage I to III) diagnosed from 2005 to 2010 were included with follow-up continuing through 2014. Nine hundred seventeen (21%) women underwent preoperative MRI. The unadjusted 5-year cumulative probability of death was 0.12 for women with MRI and 0.17 for those without (HR, 0.67; 95% CI, 0.54 to 0.82). However, after adjustment for age, sociodemographic, and clinical factors, the association was attenuated (HR, 0.90; 95% CI, 0.72 to 1.12).

Fortune-Greeley et al (2014) retrospectively examined case records of 20,332 women with invasive breast cancer in the Surveillance Epidemiology and End Results-Medicare-linked dataset.^36, Twelve percent of patients had a preoperative MRI. Among patients with invasive lobular carcinoma, but no other histologic types, preoperative breast MRI was associated with lower odds of reoperation after initial partial mastectomy (adjusted OR, 0.59; 95% CI, 0.40 to 0.86).

Zeng et al (2020) performed a retrospective analysis of 512 women age ≤50 years undergoing BCT.^37, Preoperative MRI was performed in 64.5% of women. In patients who did versus did not receive preoperative MRI, mean age was 43.4 and 43.6 years, and tumor size was 1.64 and 1.80 cm, respectively. In those who received MRI versus no MRI, local recurrence occurred in 7.9% versus 8.2% of patients, respectively (adjusted HR with MRI vs. no MRI, 1.03; 95% CI, 0.53 to 1.99), and was associated with distant recurrence in 6.4% versus 6.6% of patients (adjusted HR with MRI vs. no MRI, 0.89; 95% CI, 0.43 to 1.84).

Section Summary: Preoperative Mapping to Identify Multicentric Disease With Clinically Localized Breast Cancer

Preoperative MRI as an adjunct to mammography and clinical assessment identifies additional foci of ipsilateral breast cancer and results in a higher rate of mastectomy. For example, a 2017 meta-analysis of 17 studies found significantly higher odds of receiving a mastectomy after preoperative MRI versus no MRI in women with breast cancer. Follow-up studies have reported mixed results, including no significant reduction in reoperation rates after MRI while other studies have reported lower odds of reoperation in patients with invasive lobular carcinoma. No significant differences in ipsilateral local or distant recurrence-free survival after MRI-guided treatment were found in meta-analyses. While there is limited evidence that use of MRI to identify multicentric disease improves recurrence free survival or reduces operations in the overall population, benefit might accrue to sub populations, particularly high risk individuals.

Summary of Evidence

Reference No. 8

Population Reference No. 9

Guiding Surgical Decisions After Neoadjuvant Chemotherapy

Clinical Context and Test Purpose

Patients with locally advanced breast cancer are usually offered neoadjuvant chemotherapy to reduce tumor size and permit BCT. Evaluation of tumor size and extent using conventional techniques (ie, mammography, clinical examination, ultrasonography) is suboptimal, and breast MRI has been proposed as a means to more accurately determine tumor size for surgical planning. Breast MRI before chemotherapy is used to document tumor location so that the tumor can be optimally evaluated after chemotherapy, especially if the size and degree of contrast enhancement are greatly reduced. Tumors that respond to chemotherapy get smaller and may even disappear; however, the actual reduction in size is a delayed finding, and earlier changes in tumor vascularity have been observed in chemotherapy-responsive tumors. A decline in contrast enhancement on MRI has been noted in tumors relatively early in the course of chemotherapy. This MRI finding as an early predictor of tumor response has been explored as a means to optimize the choice of the chemotherapeutic agent (eg, to alter chemotherapy regimen if the tumor appears unresponsive).

The question addressed in this evidence review is: Does the use of MRI for guiding surgical decisions after neoadjuvant chemotherapy in patients with locally advanced breast cancer improve net health outcomes?

The following PICO was used to select literature to inform this review.

Populations

The population of interest is individuals with locally advanced breast cancer undergoing neoadjuvant chemotherapy.

Interventions

The intervention of interest is MRI to guide surgical decisions after neoadjuvant chemotherapy.

Comparators

The following tests and practices are currently being used to make decisions about managing breast cancer: mammography and clinical assessment.

Outcomes

Breast MRI is performed after identification of suspicious breast lesions, or before or after treatment for breast cancer.

Study Selection Criteria

Methodologically credible studies were selected using the following principles:

To assess efficacy outcomes, comparative controlled prospective trials were sought, with a preference for RCTs;
In the absence of such trials, comparative observational studies were sought, with a preference for prospective studies.
To assess long-term outcomes and adverse events, single-arm studies that capture longer periods of follow-up and/or larger populations were sought.
Consistent with a 'best available evidence approach,' within each category of study design, studies with larger sample sizes and longer duration were preferred.
Studies with duplicative or overlapping populations were excluded.

Review of Evidence

Systematic Reviews

Compared with conventional methods of evaluating tumor size and extent (ie, mammography, clinical exam, ultrasound), MRI of the breast provides an estimation of tumor size and extent that is at least as good as or better than that based on alternatives. Drew et al (2001) found MRI to be 100% sensitive and specific for defining residual tumor after chemotherapy.^38, Conversely, mammography achieved 90% sensitivity and 57% specificity (mammography results considered equivocal), and the clinical exam was only 50% sensitive and 86% specific. Similarly, Partridge et al (2002) reported on correlations of residual tumor size by histopathology of 0.89 with MRI and 0.60 with a clinical exam.^39, The MRI results were well-correlated with results of the histopathologic assessment (criterion standard) with correlation coefficients ranging from 0.72 to 0.98; however, MRI is not intended as a replacement for histopathologic assessment.

Marinovich et al (2015) published an individual patient data meta-analysis of agreement between MRI and pathologic tumor size and other evaluation methods after neoadjuvant chemotherapy.^40, To be eligible for inclusion, studies had to evaluate at least 15 patients undergoing neoadjuvant chemotherapy who were evaluated with MRI and at least 1 other test (ie, mammography, ultrasound, clinical examination) after surgery. Studies also had to report residual tumor size (ie, longest diameter). Twenty-four studies met inclusion criteria, and individual patient data were available for 8 of these studies (N=300). The pooled mean difference (MD) in size estimates between MRI and pathology (8 studies, n=243) was 0.0 cm (95% CI, -0.1 to 0.2 cm). In 4 studies comparing size estimates of mammography and pathology, the MD was 0.0 cm, but the 95% CI was wider (-0.3 to 0.4 cm). In 5 studies (n=123) reporting on the MD between ultrasound and pathology, the pooled estimate was -0.3 cm (95% CI, -0.6 to 0.1 cm). The largest size variance was for studies (3 studies, n=107) comparing clinical examination with pathology (pooled MD , -0.8 cm; 95% CI, -1.5 to -0.1 cm).

Previously, Lobbes et al (2013) reported on a systematic review of 35 studies (N=2359) reporting on the ability of MRI to predict tumor size after neoadjuvant chemotherapy.^41, Literature was searched to July 2012. Median correlation coefficient was 0.70 (range, 0.21 to 0.98). Variation in size between MRI and pathology ranged from -1.4 to +2.0 cm.

Section Summary: Guiding Surgical Decisions After Neoadjuvant Chemotherapy

Studies, including a 2015 meta-analysis, have found that MRI results are well-correlated with pathologic assessment for measuring residual tumor size after neoadjuvant chemotherapy and that MRI performed better than conventional methods. Using breast MRI instead of conventional methods to guide surgical decisions regarding BCT versus mastectomy after neoadjuvant chemotherapy would be at least as beneficial and might lead more frequently to appropriate surgical treatment.

Summary of Evidence

Reference No. 9

Population Reference No. 10

Evaluating Suspected Chest Wall Involvement

Clinical Context and Test Purpose

Tumors located near the chest wall may invade the pectoralis major muscle or extend deeper into chest wall tissues. Typically, modified radical mastectomy removes only the fascia of the pectoralis muscle; however, tumor involvement of the muscle would also necessitate the removal of the muscle (or a portion of it). In smaller tumors, it is necessary to determine how closely the tumor abuts the pectoralis muscle and whether it invades the muscle to determine whether there is an adequate margin of normal breast tissue to permit BCT. Breast MRI has been suggested as a means of determining pectoralis muscle/chest wall involvement for surgical planning and to assist in the decision whether to use neoadjuvant chemotherapy.

The question addressed in this evidence review is: Does the use of MRI to diagnose chest wall involvement of posteriorly located breast tumors improve net health outcomes?

The following PICO was used to select literature to inform this review.

Populations

The population of interest is individuals with posteriorly located breast tumors.

Interventions

The intervention of interest is MRI to diagnose chest wall involvement.

Comparators

The following tests and practices are currently being used to make decisions about managing breast cancer: mammography.

Outcomes

Breast MRI is performed after identification of suspicious breast lesions, or before or after treatment for breast cancer.

Study Selection Criteria

Methodologically credible studies were selected using the following principles:

To assess efficacy outcomes, comparative controlled prospective trials were sought, with a preference for RCTs;
In the absence of such trials, comparative observational studies were sought, with a preference for prospective studies.
To assess long-term outcomes and adverse events, single-arm studies that capture longer periods of follow-up and/or larger populations were sought.
Consistent with a 'best available evidence approach,' within each category of study design, studies with larger sample sizes and longer durations were sought.
Studies with duplicative or overlapping populations were excluded.

Review of Evidence

Observational Studies

Morris et al (2000) prospectively studied 19 patients with posteriorly located breast tumors suspected to involve the pectoralis major muscle based on either mammography or clinical exam.^42, Thirteen tumors were thought to be fixed to the chest wall on clinical exam, and 12 appeared to have pectoral muscle involvement on mammography. The MRI results were compared with surgical and pathologic findings. The presence of abnormal enhancement within the pectoralis major muscle on MRI was 100% sensitive and 100% specific for identifying 5 tumors that actually involved the pectoralis major muscle.

Two other retrospective studies have reported on 4 cases in which MRI was able to determine the involvement of the chest wall with 100% accuracy.^43,^44,

Section Summary: Evaluating Suspected Chest Wall Involvement

Evidence on MRI for evaluating suspected chest wall involvement with posteriorly located tumors is based on prospective and retrospective observational studies. All studies found that MRI was able to detect chest wall involvement with 100% accuracy. Given the high level of diagnostic accuracy for MRI compared with criterion standard and conventional alternative techniques, the evidence is considered sufficient to conclude that breast MRI improves net health outcome.

Summary of Evidence

Reference No. 10

Population Reference No. 11

Evaluating and Localizing Lesions Prior to Biopsy

Clinical Context and Test Purpose

The question addressed in this evidence review is: Does the use of MRI to evaluate and localize breast lesions prior to biopsy improve net health outcomes?

The following PICOs were used to select literature to inform this review.

Populations

The populations of interest is individuals with a suspicious breast lesion recommended for biopsy but not localizable by mammography or ultrasonography.

Interventions

The intervention of interest is MRI to evaluate and localize breast lesion prior to biopsy.

Comparators

The following tests and practices are currently being used to make decisions about managing breast cancer: waiting until lesion becomes palpable or visible on mammography or ultrasonography.

Outcomes

Breast MRI is performed after identification of suspicious breast lesions recommended for biopsy.

Study Selection Criteria

Methodologically credible studies were selected using the following principles:

To assess efficacy outcomes, comparative controlled prospective trials were sought, with a preference for RCTs;
In the absence of such trials, comparative observational studies were sought, with a preference for prospective studies.
To assess long-term outcomes and adverse events, single-arm studies that capture longer periods of follow-up and/or larger populations were sought.
Consistent with a 'best available evidence approach,' within each category of study design, studies with larger sample sizes and longer duration were preferred.
Studies with duplicative or overlapping populations were excluded.

Review of Evidence

Observational Studies

Use of MRI to evaluate lesions prior to biopsy is infrequent. An MRI is used in this situation to permit biopsy and breast cancer diagnosis sooner than waiting until the lesion is visible on 2 mammographic views or on ultrasound or becomes palpable. The evidence base addressing this use is mainly anecdotal.

Xie et al (2023) retrospectively evaluated the value of breast MRI to downgrade suspicious lesions (BI-RADS 4A or 4B) found on ultrasound in 167 patients with 186 lesions.^45, Compared to pathology and imaging findings over the subsequent 12 months, MRI had 100% sensitivity, 92.6% specificity, 87.8% PPV, and 100% NPV. Four additional suspicious lesions were detected by MRI, of which 3 (75%) were malignant. Survival was not mentioned. The authors concluded that MRI could allow suspicious lesions to be downgraded and prevent unneeded biopsies.

De Lima Docema et al (2014) used contrast-enhanced MRI to locate occult tumors in 25 patients selected from a group who had undergone breast MRI for suspicious incidental MRI findings at a single-institution in Brazil.^46, Sentinel lymph node mapping and tumor resection were done simultaneously. Malignant tumors were confirmed in 15 (60%) patients, including 4 patients with DCIS. Survival outcomes were not reported.

Section Summary: Evaluating and Localizing Lesions Prior to Biopsy

A small cohort study in Brazil identified malignant tumors in 60% of patients with MRI-detected occult lesions using contrast-enhanced MRI. A retrospective study of patients with suspicious lesions on ultrasound reported high sensitivity, specificity, PPV, and NPV of MRI to downgrade lesion status and prevent biopsies.

Summary of Evidence

Reference No. 11

Population Reference No. 12

Evaluating Response to Neoadjuvant Chemotherapy With Locally Advanced Breast Cancer

Clinical Context and Test Purpose

The question addressed in this evidence review is: Does the use of MRI to evaluate response to chemotherapy for locally advanced breast cancer improve net health outcomes?

The following PICO was used to select literature to inform this review.

Populations

The population of interest is individuals with locally advanced breast cancer undergoing neoadjuvant chemotherapy.

Interventions

The intervention of interest is MRI to evaluate the response to chemotherapy.

Comparators

The comparator of interest is clinical assessment alone.

Outcomes

Breast MRI is performed after a period of undergoing adjuvant chemotherapy.

Study Selection Criteria

Methodologically credible studies were selected using the following principles:

To assess efficacy outcomes, comparative controlled prospective trials were sought, with a preference for RCTs;
In the absence of such trials, comparative observational studies were sought, with a preference for prospective studies.
To assess long-term outcomes and adverse events, single-arm studies that capture longer periods of follow-up and/or larger populations were sought.
Consistent with a 'best available evidence approach,' within each category of study design, studies with larger sample sizes and longer duration were preferred.
Studies with duplicative or overlapping populations were excluded.

Review of Evidence

Systematic Reviews

Four systematic reviews of MRI to evaluate response to neoadjuvant chemotherapy have been published.^41,^47,^48,^49, Characteristics of the reviews are shown in Table 11 and described briefly in the following paragraphs. Li et al (2018) compared the performance of MRI with positron emission tomography (PET) plus computed tomography (CT).^48,

Table 11. Characteristics of Systematic Reviews Assessing Magnetic Resonance Imaging to Evaluate Response to Neoadjuvant Chemotherapy

Study	Dates	Studies	Participants	N (Range)	Design	Reference Standard
Janssen et al (2022)^49,	2000 to 2019	26	Patients with early-stage breast cancer who received MRI after NAC	4497 (NR)	Observational (prospective, retrospective)	Pathologic response
Li et al (2018)^48,	Up to 2017	13	Had both PET/CT and MRI after preoperative NAC with at least 10 patients	MRI: 575 (16 to 142);PET/CT: 618 (16 to 142)	Observational (prospective, retrospective)	Postoperative pathologic result (pCR vs. non-pCR)
Marinovich et al (2013)^47,	Up to 2011	44	Newly diagnosed breast cancer undergoing NAC, with MRI undertaken after NAC	2949 (14 to 869)	Observational (prospective, retrospective)	Pathologic response based on surgical excision preferred; other references standards allowed
Lobbes et al (2013)^41,	Up to 2012	8	Newly diagnosed breast cancer for whom breast MRI was not performed at baseline or prior to surgery but after completion of NAC with at least 25 patients	560 (31 to 195)	Observational (prospective, retrospective)	NR

CT: computed tomography; MRI: magnetic resonance imaging; NAC: neoadjuvant chemotherapy; NR: not reported; pCR: pathologic complete response; PET: positron emission tomography.

Results of the systematic reviews are shown in Table 12. Janssen et al (2022) reported the results of a systematic review that evaluated the accuracy of MRI for detecting pCR after neoadjuvant chemotherapy.^49, Risk of bias was assessed using the QUADAS-2 tool. Sensitivity was highest for hormone receptor (HR)-negative/HER2-negative cancer (0.67), followed by HR-negative/HER2-positive (0.65), HR-positive/HER2-positive (0.60), and HR-positive/HER2-negative (0.55). None of the differences in sensitivity were significant between groups. Specificity results were 0.85, 0.81, 0.74, and 0.88, respectively. Specificity was significantly different between the HR-negative/HER2-positive and R-positive/HER2-negative groups (p=.046).

Li et al (2018) reported on a systematic review comparing MRI with PET/CT to evaluate pathologic response to neoadjuvant chemotherapy and included studies in which patients underwent both PET/CT and MRI after preoperative neoadjuvant chemotherapy; postoperative pathologic complete response (pCR vs. non-pCR) was used as the reference standard; and the study included at least 10 patients.^48, Methodologic quality was assessed using QUADAS-2. Most domains were rated as low-risk of bias in all studies; however, only 2 studies enrolled consecutive or random samples and in only 3 studies were the reference standard results interpreted without knowledge of the results of the index tests. There was a high level of heterogeneity in the pooled estimate of both sensitivity (88%; 95% CI, 78 to 94; I²=83%) and specificity (69%; 95% CI, 51 to 83; I²=72%) for MRI.

Marinovich et al (2013) conducted a systematic review with meta-analysis.^47, Forty-four studies (N=2949) assessing the ability of MRI to discriminate residual breast tumor after neoadjuvant chemotherapy from pCR were identified. Studies were heterogeneous in MRI parameters used, thresholds for identifying a response, and definitions of pathologic response. Median MRI sensitivity, defined as the proportion of patients with residual tumor correctly classified by MRI, and specificity, defined as the proportion of patients with pCR classified by MRI as the absence of residual tumor was 0.92 (IQR, 0.85 to 0.97) and 0.60 (IQR, 0.39 to 0.96), respectively. Specificity increased when a relative threshold for defining negative MRI (ie, contrast enhancement was less than or equal to normal breast tissue) was used rather than an absolute threshold (complete absence of MRI enhancement) with little decrement to sensitivity. The pooled area under the receiver operating characteristic curve was 0.88, and the diagnostic OR was 17.9 (95% CI, 11.5 to 28.0). A diagnostic OR of 1 indicates no discriminatory ability; higher values indicate better test performance. Accuracy decreased when residual DCIS was included in the definition of pCR. Statistical measures of between-study heterogeneity were not reported. A subset of studies compared MRI with other imaging modalities (mammography, ultrasound) and clinical exam; however, 95% CIs for pooled analyses were very large, rendering conclusions uncertain.

In the systematic review by Lobbes et al (2013), 8 studies reported on measures of diagnostic accuracy.^41, Median sensitivity, defined as the proportion of patients with pCR correctly classified by MRI, was 42% (range, 25% to 92%). Median specificity, defined as the proportion of patients without pCR correctly classified by MRI, was 89% (range, 50% to 97%). Median (range) PPV and NPV were 64% (50% to 73%) and 87% (71% to 96%), respectively.

Table 12. Results of Systematic Reviews Assessing Magnetic Resonance Imaging to Evaluate Response to Neoadjuvant Chemotherapy

Study	MRI		Mammography		PET/CT
	Sensitivity, %	Specificity, %	Sensitivity, %	Specificity, %	Sensitivity, %	Specificity, %
Janssen et al (2022)^49,
HR-/HER2- (n=1646), PE (95% CI)	0.67 (0.58 to 0.74)	0.85 (0.81 to 0.88)	NR	NR	NR	NR
HR-/HER2+ (n=1013), PE (95% CI)	0.65 (0.56 to 0.73)	0.81 (0.74 to 0.86)	NR	NR	NR	NR
HR+/HER2- (n=2273), PE (95% CI)	0.55 (0.45 to 0.64)	0.88 (0.84 to 0.91)	NR	NR	NR	NR
HR+/HER2+ (n=1144), PE (95% CI)	0.60 (0.50 to 0.70)	0.74 (0.63 to 0.83)	NR	NR	NR	NR
Li et al (2018)^48,
Total N	575	575			618	618
PE (95% CI)	88 (78 to 94)	69 (51 to 83)	NR	NR	77 (58 to 90)	78 (63 to 88)
Marinovich et al (2013)^47,
Total N	2949	2949
Median (IQR)	92 (85 to 97)	60 (39 to 96)	NR	NR	NR	NR
Lobbes et al (2013)^41,
Total N	560	560
Median (range)	42 (25 to 92)	89 (50 to 97)	NR	NR	NR	NR

CI: confidence interval; CT: computed tomography; HR; hormone receptor; IQR: interquartile range; MRI: magnetic resonance imaging; NR: not reported; PE: pooled estimate; PET: positron emission tomography.

Trials

The ACRIN 6657/I-SPY trial (2012) enrolled 206 women aged 26 to 68 years with invasive breast cancer 3 cm or larger who were receiving anthracycline-based neoadjuvant chemotherapy, with or without a taxane.^50, Of the patients included in the study, 74.4% were White, 19.2% were Black, 4% were Asian, and 2.4% were more than one race or unknown race; 4.2% of patients were Hispanic or Latino. The MRI was performed at 4 time points: before chemotherapy, after 1 cycle of chemotherapy, between the anthracycline-based regimen and the taxane, and after all chemotherapy but before surgery. Various MRI parameters were evaluated for their ability to predict the pathologic outcome. Results were reported as the difference in the predictive ability for residual cancer burden, a composite pathologic index, between MRI parameters and clinical size predictors at the same time points. The MRI findings were a stronger predictor of pathologic outcomes than clinical assessment, with the largest difference being tumor volume after the first chemotherapy cycle and a difference in the area under the receiver operating characteristic curve of 0.09; the corresponding area under the receiver operating characteristic curve values after the third and fourth MRIs were 0.07 and 0.05. Similar findings were reported for predicting pCR.

Section Summary: Evaluating Response to Neoadjuvant Chemotherapy With Locally Advanced Breast Cancer

Studies, including systematic reviews, have not found sufficient evidence to determine whether breast MRI can reliably predict lack of response to neoadjuvant chemotherapy. There is a large amount of variability in reported performance characteristics of MRI in published studies, leaving uncertain the true accuracy of MRI for this purpose. Furthermore, evidence would need to show that any resulting change in patient management (eg, discontinuation of chemotherapy or change to a different regimen) would improve outcomes.

Summary of Evidence

Reference No. 12

Population Reference No. 13

Evaluating Residual Tumor After Lumpectomy or Breast Conservation Surgery

Clinical Context and Test Purpose

In BCT there is complete removal of the primary tumor along with a rim of normal surrounding tissue. Pathologic assessment of surgical margins is performed on excisional specimens to determine whether the tumor extends to the margins of resection. Surgical specimens are oriented and marked to direct re-excision if margins are shown to contain tumor; however, when the tumor is not grossly visible, the extent of a residual tumor within the breast can only be determined through repeat excision and pathologic assessment. Use of MRI has been proposed to evaluate the presence and extent of the residual tumor as a guide to re-excision when surgical margins are positive for tumor.

The question addressed in this evidence review is: Does the use of MRI to evaluate residual tumor after lumpectomy or BCT improve net health outcomes?

The following PICO was used to select literature to inform this review.

Populations

The population of interest are individuals with positive surgical margins after lumpectomy or BCT.

Interventions

The intervention of interest is MRI to evaluate the residual tumor.

Comparators

The comparator of interest is pathologic inspection.

Outcomes

Breast MRI is performed after lumpectomy or BCT.

Study Selection Criteria

Methodologically credible studies were selected using the following principles:

To assess efficacy outcomes, comparative controlled prospective trials were sought, with a preference for RCTs;
In the absence of such trials, comparative observational studies were sought, with a preference for prospective studies.
To assess long-term outcomes and adverse events, single-arm studies that capture longer periods of follow-up and/or larger populations were sought.
Consistent with a 'best available evidence approach,' within each category of study design, studies with larger sample sizes and longer duration were preferred.
Studies with duplicative or overlapping populations were excluded.

Review of Evidence

Observational Studies

Evidence on evaluating residual tumor includes several observational studies, most of which are retrospective.^51,^52,^53,^54,^55,^56,^57,^58,^59, Histopathologic examination on re-excision was used as the criterion standard. Three studies were conducted at the same institution and accrued patients during similar time periods, so overlap reporting may exist.^52,^54,^55, Most of the studies were published before 2005 and are not discussed further. Characteristics of studies published since 2015 are shown in Table 13 and described briefly in the following paragraphs.^56,^57,

Table 13. Characteristics of Clinical Validity Studies Assessing Magnetic Resonance Imaging to Evaluate Residual Tumor After Surgery

Study	Study Population	Design	Reference Standard	Threshold for Positive Index Test	Timing of Reference and Index Tests	Blinding of Assessors	Comment
Lee et al (2018)^57,	Patients in Taiwan with LCIS who had initial excision from 2011 to 2015; race or ethnicity were not described	Unclear	Histopathology	NR	NR	NR	Few details on study design or conduct provided
Krammer et al (2017)^56,	Women with positive margins after initial surgery for breast cancer from 2004 to 2013; race or ethnicity were not described	Retrospective	Histopathology	Read independently by 2 radiologists Criteria for suspected residual disease: asymmetric thickening or nodular enhancement with irregular or spiculated margins or extensive focal non-mass enhancement	NR	Radiologists had access to other imaging results, when available

LCIS: lobular carcinoma in situ; NR: not reported.

Results of the clinical validity studies published after 2015 are shown in Table 14. Lee et al (2018) reported on the results of a study comparing breast MRI with ultrasonography for detecting remnant lobular carcinoma in situ lesions after initial excision.^57, Twenty-nine patients with lobular carcinoma in situ were enrolled between 2011 and 2015. Methods are poorly described. Residual lesions were identified by pathology in 12 (41%) cases. The sensitivity of ultrasonography was 58% compared with 83% for breast MRI; precision estimates were not reported. Specificity was 100% for both modalities.

Krammer et al (2017) published a retrospective study evaluating breast MRI to assess residual disease in 175 patients who had been candidates for BCS and had positive surgical margins.^56, The MRIs were read independently by 2 radiologists, both of whom had access to the pathology report from the initial surgery and any prior breast imaging. Pathology findings served as the criterion standard. For reader 1, the sensitivity and specificity of detecting residual disease was 63% and 75%, respectively. For reader 2, sensitivity and specificity were 83% and 64%, respectively. The inter-observer agreement was moderate (k=0.56).

Table 14. Results of Clinical Validity Studies Assessing Magnetic Resonance Imaging to Evaluate Residual Tumor After Surgery

Study	Initial N	Final N	Excluded Samples	Prevalence of Condition, %	Clinical Validity (95% Confidence Interval), %
					Sensitivity	Specificity	PPV	NPV
Lee et al (2018)^57,	NR	29	Any invasive focus or other malignancy	41
MRI					83% (NR)	100% (NR)	NR	NR
Ultrasonography					58% (NR)	100% (NR)	NR	NR
Krammer et al (2017)^56,	180	175	Received chemotherapy prior to postoperative MRI (n=4), poor MRI image quality (n=1)	79
MRI					73% (NR)	72% (NR)	91% (NR)	45% (NR)

 MRI: magnetic resonance imaging; NPV: negative predictive value; NR: not reported; PPV: positive predictive value.

Tables 15 and 16 display notable limitations identified in each study.

Table 15. Study Relevance Limitations of Clinical Validity Studies of Magnetic Resonance Imaging to Evaluate Residual Tumor After Surgery

Study	Population^a	Intervention^b	Comparator^c	Outcomes^d	Duration of Follow-Up^e
Lee et al (2018)^57,	2. Study population is unclear	1,2. No description provided	1. No description provided	1. Health outcomes not reported
Krammer et al (2017)^56,	2. Study population is unclear		3. No comparator	1. Health outcomes not reported

The study limitations stated in this table are those notable in the current review; this is not a comprehensive gaps assessment.^aPopulation key: 1. Intended use population unclear; 2. Study population is unclear; 3. Study population not representative of intended use; 4, Enrolled populations do not reflect relevant diversity; 5. Other. ^bIntervention key: 1. Classification thresholds not defined; 2. Version used unclear; 3. Not intervention of interest.^c Comparator key: 1. Classification thresholds not defined; 2. Not compared to credible reference standard; 3. Not compared to other tests in use for same purpose.^d Outcomes key: 1. Study does not directly assess a key health outcome; 2. Evidence chain or decision model not explicated; 3. Key clinical validity outcomes not reported (sensitivity, specificity and predictive values); 4. Reclassification of diagnostic or risk categories not reported; 5. Adverse events of the test not described (excluding minor discomforts and inconvenience of venipuncture or noninvasive tests).^e Follow-Up key: 1. Follow-up duration not sufficient with respect tonatural history of disease (true-positives, true-negatives, false-positives, false-negatives cannot be determined).

Table 16. Study Design and Conduct Limitations of Clinical Validity Studies Assessing Magnetic Resonance Imaging to Evaluate Residual Tumor After Surgery

Study	Selection^a	Blinding^b	Delivery of Test^c	Selective Reporting^d	Data Completeness^e	Statistical^f
Lee et al (2018)^57,		1. Not described	1,3,4. Not described			1. No precision estimates provided 2. No statistical comparison to other methods
Krammer et al (2017)^56,		1. Not blinded to other imaging results

The study limitations stated in this table are those notable in the current review; this is not a comprehensive gaps assessment.^aSelection key: 1. Selection not described; 2. Selection not random or consecutive (ie, convenience).^bBlinding key: 1. Not blinded to results of reference or other comparator tests.^cTest Delivery key: 1. Timing of delivery of index or reference test not described; 2. Timing of index and comparator tests not same; 3. Procedure for interpreting tests not described; 4. Expertise of evaluators not described.^d Selective Reporting key: 1. Not registered; 2. Evidence of selective reporting; 3. Evidence of selective publication.^e Data Completeness key: 1. Inadequate description of indeterminate and missing samples; 2. High number of samples excluded; 3. High loss to follow-up or missing data.^f Statistical key: 1. Confidence intervals and/or p values not reported; 2. Comparison with other tests not reported.

Section Summary: Evaluating Residual Tumor After Lumpectomy or Breast Conservation Surgery

The available evidence is not sufficient to permit conclusions whether the use of MRI identifies the presence and/or extent of residual disease after lumpectomy or BCS and before re-excision. Most studies were retrospective, and most reported moderate sensitivity and specificity of MRI for detection of residual disease. One study published after 2015 reported the sensitivity and specificity of MRI to be over 70%. The other study published after 2015 reported a sensitivity of 83% and a specificity of 100% but offered very few details on methods, so study quality cannot be assessed.

Summary of Evidence

Reference No. 13