Table 3.

MRI in the Evaluation of Residual Disease

StudyType of StudyStudy PopulationNMajor Findings/Comments
Yeh et al.25ProspectivePatients with stage IIb/III breast cancer undergoing neoadjuvant chemotherapy (doxorubicin and paclitaxel)31Correlation with pathology
Equal to pathology/underestimate/overestimate
    MRI: 71%/23%/6%
    XRM: 26%/52%/23%
    US: 35%/52%/13%
    MRI was best correlated with pathology
    (P < .002)
Chen et al.26ProspectivePatients with locally advanced breast cancer receiving neoadjuvant therapy (adriamycin, cytoxan, and paclitaxel)15MRI tended to overestimate tumor response; for partial non-responders MRI correlated with pathology 0% of the time compared with 17% for CBE and 83% for PET; however, for responders, MRI correlated with pathology 90% of the time, with 70% correlation for CBE and 90% for PET; MRI size measurements of residual tumors did correlate with pathology (coefficient 0.7 compared with −0.06 for CBE)
Denis et al.27ProspectivePatients with locally advanced breast cancer receiving neoadjuvant therapy with either 5–5-fluorouracil/epirubicin/ cyclophosphamide; docetaxel only, or docetaxel with epirubicin40Correlation coefficient of MRI measurements with pathology: 5-fluorouracil/E/C: 0.89, DXL: 0.64, DXL/E: 0.16 MRI overestimated tumor response in DXL-based groups, which was believed to be due to antiangiogenic effect of DXL, which would impair MRI contrast enhancement; this was supported by the fact that DXL-based groups had residual disease of microscopic nests of tumor cells on pathology as opposed to single nodular lesions
Tozaki et al.28ProspectivePatients with locally advanced breast cancer (IIb/III) undergoing neoadjuvant chemotherapy19Accuracy (deviation from pathology of less than 2 mm) of late phase CT and MRI scans (scan 4 minutes after contrast injection) compared with pathology was up to 90% for DCIS and replaced (diffuse pre-NAC contrast enhancement) lesions and 88% for nonreplaced (localized CE) lesions; early phase scans were 0% accurate for DCIS/replaced and 75% accurate for nonreplaced lesions
Bodini et al.29ProspectivePatients with T2 to 4, N0 M0 breast cancer treated with 3 to 4 cycles of epirubicin79Clinical response correlation with pathology correlation coefficients were 0.72 for MRI and 0.68 for CBE
Warren et al.30RetrospectivePatients undergoing neoadjuvant therapy67MRI was more sensitive and specific for assessment of complete or partial response (100% and 80%) than conventional assessment methods (CAM), including XRM, US, and CBE (98% and 50%); agreement with pathology was marginally higher in MRI compared with CAM (81% compared to 68%, P = .09); MRI increased diagnostic knowledge in 70% of patients, and increased diagnostic confidence in 52%; however, MRI did not change treatment plan, decreased confidence in 20% of patients, and decreased knowledge in 17%; MRI tended to overestimate response
Lee et al.31RetrospectivePatients who had excisional biopsy who required definitive surgery, eg, because of positive margins/residual disease and were candidates for breast conservation80MRI sensitivity and specificity for residual disease was 61.2% and 69.7%, respectively; additional lesions were detected by imaging of which 10 were only seen on MRI; of these, 5 were benign, 5 were malignant; MRI changed management of patients in 29% of the cases, because of additional lesions found; whether this led to improved patient outcomes is not known
Rosen et al.32RetrospectivePatients with locally advanced breast cancer treated with neoadjuvant therapy (paclitaxel, doxorubicin, and breast hyperthermia)21Correlation with pathology equal to pathology/underestimate/overestimate MRI: 57%/10%/33% correlation coefficients were 0.74 for MRI and 0.65 (statistically nonsignificant trend); in contrast to some of the other studies, MRI tended to overestimate the residual tumor (ie, underestimate response)
Hwang et al.33RetrospectivePatients with histologically confirmed diagnosis of DCIS51MRI was 97% sensitive and 58% specific for detecting residual disease compared with histology; the sensitivity was significantly better than the sensitivity of XRM for residual disease; in addition MRI had a significantly higher NPV compared to XRM
Belli et al.34ProspectivePatients with locally advanced breast cancer being treated with neoadjuvant chemotherapy45MRI had a 90.2% sensitivity and a 100% specificity for residual disease
Cheung et al.35ProspectivePatients with locally advanced breast cancer being treated with neoadjuvant chemotherapy33Residual tumor on MRI correlated with microscopic findings (correlation coefficient = 0.982, P < .001)
Partridge et al.36ProspectivePatients with invasive breast cancer undergoing neoadjuvant therapy with doxorubicin and cyclophosphamide52Decreased tumor enhancement prechemotherapy and postchemotherapy (210% vs 166%, P < .001); MRI had correlation coefficient of 0.89 compared with pathology whereas clinical examination had coefficient of 0.60
Kawashima et al.37ProspectivePatients who underwent excisional biopsy of breast lesion50MRI in detection of residual disease: sensitivity 66%; specificity 81%; PPV 72%; NPV 83%; accuracy 63%
Drew et al.38ProspectivePatients with locally advanced breast cancer receiving neoadjuvant therapy17Dynamic CE MRI was 100% accurate in assessing residual disease; CBE and XRM were not sensitivity/specificity/PPV/NPV CBE: 50%/86%/83%/55% XRM: 90%/57%/75%/80%
Weatherall et al.39RetrospectivePatients with breast cancer with chemotherapy prior to surgery20MRI demonstrated a correlation with pathology (coefficient = 0.93); coefficients for CBE and XRM were 0.72 and 0.63, respectively
Frei et al.40RetrospectivePatients with excisional biopsy68MRI sensitivity/specificity/PPV/NPV >7 days postbiopsy: 89%/52%/81%/69% >14 days postbiopsy: 88%/58%/82%/69% >21 days postbiopsy: 91%/69%/88%/75% >28 days postbiopsy: 92%/75%/92%/75% >35 days postbiopsy: 95%/75%/91%/86% >42 days postbiopsy: 94%/75%/89%/86%The peak values for PPV and plateau for specificity occurs at >28 days, after which the improvement is not as much; therefore, this may be the best time to perform the MRI as the PPV of positive margins in this study was 69% compared to 92% for MRI at day 28, which may lead to breast-conserving surgery
Trecate et al.41ProspectivePatients with locally advanced breast cancer receiving chemotherapy30MRI: sensitivity: 96%; specificity: 75%; PPV: 92.5%; NPV 66%; accuracy 90%
Orel et al.42ProspectivePatients who underwent excisional biopsy47MRI had PPV of 82% and NPV of 61%; false negatives possibly secondary to postsurgical changes
Soderstrom et al.43)ProspectivePatients who underwent excisional biopsies19MRI had an 84% accuracy in determining whether residual tumor was present in patients postexcisional biopsy