Accuracy of MRI in Prediction of Pathologic Complete Remission in Breast Cancer After Preoperative Therapy: A Meta-Analysis
Abstract
OBJECTIVE. Prediction of pathologic complete remission in breast cancer after preoperative therapy presents difficulties. We performed a meta-analysis to determine the ability of MRI to predict pathologic complete remission in patients with breast cancer after preoperative therapy.
MATERIALS AND METHODS. Medical subject heading terms (“MRI” and “Breast Neoplasm”) and key words (“neoadjuvant” or “primary systemic” or “preoperative” or “presurgery”) were used for a literature search in the MEDLINE database. A meta-analysis of pooled data from eligible studies was performed to estimate the accuracy of MRI in predicting pathologic complete remission after preoperative therapy in patients with breast cancer.
RESULTS. Twenty-five studies were included in this meta-analysis. Pooled weighted estimates of sensitivity and specificity were 0.63 (range, 0.56–0.70) and 0.91 (range, 90.89–0.92), respectively. Heterogeneity between studies was highly influenced by the pathologic complete remission rate, with a regression coefficient of –6.160 (p = 0.020). Subgroup analysis showed that the specificity of MRI in studies with a pathologic complete remission rate of ≥ 20% was lower than that in studies with a pathologic complete remission rate of < 20% (p = 0.0003).
CONCLUSION. This meta-analysis indicates that MRI has high specificity and relatively lower sensitivity in predicting pathologic complete remission after preoperative therapy in patients with breast cancer. The pathologic complete remission rate may influence the performance of MRI accuracy in this setting, which deserves further investigation.
Introduction
The benefits of x-ray mammography and ultrasound for the screening and diagnosis of breast cancer are well established. Breast MRI, complementing mammography, and ultrasound have become an important component in the evaluation of breast malignancies during the past 10 years. The clinical use of breast MRI is increasing, especially for applications requiring injection of paramagnetic contrast agent. The strong soft-tissue contrast obtained with MRI and the enhancement of breast lesions after administration of gadolinium-based contrast material both contribute to great detail in the depiction of anatomic structures in the breast. Peters et al. [1] performed a meta-analysis and found that contrast-enhanced MRI had higher sensitivity and lower specificity in the diagnosis of patients with breast lesions.
Preoperative therapy is viewed as the standard treatment for locally advanced breast cancer and is widely accepted as an optimal option for systemic therapy in primary operable breast cancer [2]. Several prospective randomized trials have shown that preoperative therapy is apparently equivalent to adjuvant chemotherapy in terms of overall survival and distance metastasis and allows significantly more patients to receive breast-sparing treatment in the early stages of breast cancer [3]. Furthermore, these studies have shown that patients who achieve pathologic complete remission, which is most often defined as the complete disappearance of invasive breast carcinoma in the primary tumor area, after preoperative therapy have a better prognosis than those who do not [4–6], and this better prognosis can be used to quickly determine the response to treatment and guide further treatment [2]. In recent years, several studies have been performed to find markers to predict the probability of patients' achieving pathologic complete remission, including routine clinical and pathologic characteristics [7, 8], biomarkers [9], gene expression profiling [10–12], and imaging techniques [13]. Because of the shortage of consistent and reproducible results and because there have been relatively few prospective randomized trials, we still do not have an ideal marker to predict responses to preoperative therapy.
Several imaging techniques are available to predict the effect of preoperative therapy among patients with breast cancer. As already discussed, mammography and sonography are the primary diagnostic tools for the evaluation of therapy, but accuracy is still limited in some situations [14]. In preoperative therapy, the accuracy of detecting residual tumor volume is problematic and may be affected by chemotherapeutic response changes and by the presence of therapy-induced enhancing lesions and discontinuous foci. Studies have shown that MRI can determine response to chemotherapy treatment more accurately than mammography, ultrasound, and clinical examination [15], and now extensive research has been done on the performance of MRI in evaluating the efficacy of preoperative therapy in patients with breast cancer. However, the results were based on studies that differed in terms of patient characteristics, methods, chemotherapeutic regimens, and reference standards, some of which were ineffective for evaluating the predictive value because of small population sizes.
A meta-analysis thus needs to be performed. By combining relevant evidence (e.g., a series of effect sizes, such as odds ratio [OR] and risk ratio) from similar studies, statistical power is increased and more-precise estimates can be obtained. Effects of confounding factors are lessened by pooling results from all studies, making the results more applicable to the wider population. Most important, the meta-analysis provides a framework for the assessment of between-study heterogeneity—that is, the methodologic, epidemiologic, clinical, and biologic dissimilarity across the various studies [16].
MRI has been shown to be of value in predicting tumor size when there is no response or a complete response [17]. Therefore, we focused our study on the performance of MRI in prediction of pathologic complete remission after preoperative therapy. The aim of this meta-analysis was to evaluate the ability of contrast-enhanced MRI to predict pathologic complete remission after preoperative therapy in patients with breast cancer.
Materials and Methods
Identification and Eligibility of Relevant Studies
The MEDLINE database was used to identify studies on human subjects. The search was based on the combination of medical subject heading terms (“MRI” and “Breast Neoplasm”) and keywords (“neoadjuvant” or “primary systemic” or “preoperative” or “presurgery”). The abstracts were separately screened by two reviewers. For relevant abstracts, full articles were obtained and examined. References of relevant articles and reviews were then screened for additional studies. Papers citing identified studies were also examined using the Social Sciences Citation Index (January 1998 to April 2009) through the Web of Science. The following inclusion criteria were adopted to make variables comparable that might introduce bias or explain heterogeneity of the results: first, the studies must have been published between January 1998 and April 2009; second, they had to be peer-reviewed original studies published in English; third the studies must have included at least 10 female subjects; fourth, the studies must have provided sufficient data, either directly or indirectly through a 2 × 2 table (sensitivity or specificity with absolute numbers of false-positive, false-negative, true-positive, and true-negative findings), to enable calculation of point estimates and 95% CIs for the operating characteristics of MRI compared with the reference standard; fifth, the studies had to consider histopathologic findings as the valid reference standard; sixth, the largest or the most recent article was selected among reports that included overlapped patients; and finally, the articles had to have more than nine “yes” responses for the 14 questions in the Quality Assessment of Diagnostic Accuracy Assessment tool [18]. Studies with results of different diagnostic methods that were presented in combination and could not be separated were excluded.
Data Extraction
For each report, two investigators separately extracted and recorded data on authors, year of publication, number of patients analyzed, mean age, initial clinical stage, hormonal receptor and human epidermal growth factor receptor 2 status, preoperative therapy regimens, pathologic complete remission rate, reference standard, histologic subtype, magnetic field strength, contrast agent type and dose, image and data analytic methods, and a 2 × 2 contingency table. To resolve disagreement between reviewers, a third reviewer assessed all discrepant items, and the majority opinion was used for analysis.
Data Synthesis and Statistical Analysis
Stata version 8.2 (StataCorp) and Meta-DiSc [19] software were used for statistical analysis. For each study, we constructed a 2 × 2 contingency table in which all participants were classified as having positive or negative imaging results at baseline and as having pathologic complete remission after preoperative therapy or not.
To calculate the log OR (log odds of true-positive rate / log odds of false-positive rate), we added 0.5 to each cell in any 2 × 2 table that contained one or more zero values. We estimated the weighted summary sensitivity (i.e., the ability to identify patients achieving pathologic complete remission after preoperative therapy), specificity (i.e., the ability to detect residual tumor tissue after preoperative therapy), and OR and constructed a summary receiver operating characteristic (ROC) curve, which was made from true-positive rates (sensitivity) against false-positive rates (1 – specificity). In a meta-analysis, each individual study represented a unique point on a summary ROC curve. The summary ROC curve was placed over the points to form a smoothed curve, which could be achieved by using a regression model. We defined the maximum joint sensitivity and specificity as point Q* on a summary ROC curve that was intersected by a diagonal line that ran from the top left corner to the bottom right corner of the ROC diagram. This point, at which sensitivity and specificity were equal, was a global measure of test accuracy, similar to the area under the ROC curve (AUC). The maximum joint sensitivity and specificity of a perfect test is 1.0, and the maximum joint sensitivity and specificity of a test that has no diagnostic value is 0.5.
To determine whether these values were significantly affected by heterogeneity between individual studies, we performed meta-regression analysis. We considered variances to be explanatory if their regression coefficients were statistically significant (p < 0.05). Publication bias was then assessed by using funnel plots with the log OR plotted against the standard error of the log OR in each study. Furthermore, in clinical trials and practice, the effective and recommended preoperative chemotherapy regimens' pathologic complete remission rates were about 20–30% [5, 20, 21]. Then we chose 20% as the cutoff value for high versus low pathologic complete remission rate in our meta-regression analysis. Funnel plots graphically show publication bias. An asymmetric funnel plot would suggest that additional small studies may have been conducted but not published because of unfavorable results.
Results
Eligible Studies
The computer search and extensive cross-checking of references initially yielded 231 potential articles. After we ruled out the obviously irrelevant abstracts, 78 studies were left, and their full texts were obtained. After review of the full texts of the remaining 78 articles, an additional 53 trials were excluded for the following reasons: they did not use histopathologic evidence as the reference standard (n = 7); patients were treated with concurrent chemotherapy and radiotherapy (n = 3); the studies presented results from a combination of MRI and other diagnostic techniques that could not be differentiated for assessment of single test (n = 5); the studies did not present sufficient data to construct or calculate true-positive, false-positive, true-negative, and false-negative results to create a 2 × 2 contingency table (n = 27); or the studies were reviews or proceedings of a symposium (n = 8). In addition, the articles of Wasser et al. [22] and Chen et al. [23] overlapped with two other retrieved studies (Wasser et al. [17] and Chen et al. [24], respectively). The latter two reports were viewed as eligible for including larger groups of patients (see the sixth inclusion criteria). The study by Chen et al. [25] was also excluded for overlapping with another study by Demartini et al. [26]. Therefore, 25 eligible studies were included in the meta-analysis. Table 1 summarizes the main characteristics of all eligible studies.
| Study | No. of Cases | Age (y), Mean (Range) | Initial Clinical Stage | Histologic Subtype | Preoperative Therapy Regimen | Receptor Status | Magnet Strength (T) | Contrast Material (Dose) |
|---|---|---|---|---|---|---|---|---|
| Esserman et al. [27] | 33 | 46.3 (32-75) | III or large tumor | IDC, ILC, IBC | A plus C* 4 | 22 ER+, 11 PR+ | 1.5 | Gd-DTPA (0.1 mmol/kg) |
| Balu-Maestro et al. [28] | 51 (60 breast cancers) | — | — | IDC, ILC | N plus E; A plus C; A plus P; CEF | — | — | — |
| Partridge et al. [29] | 52 | 47.3 (29-72) | — | — | A plus C*4; A plus C*4→ P | — | 1.5 | Gd-DTPA (0.1 mmol/kg) |
| Rieber et al. [30] | 58 | 51.4 (27-72) | T2-4 | — | A plus C; E plus P * 3-5 | — | 1.5 | Gd-DTPA (0.15 mmol/kg) |
| Cheung et al. [31] | 33 | 44.9 (29-63) | LABC | IDC, ILC, IC | E plus P *3 | — | 1.5 | Gadolinium (0.1 mmol/kg) |
| Delille et al. [32] | 14 | 46.3 (28-61) | LABC | IDC | A plus C; H plus P; A plus C→T or P; CMF; T plus A plus C | 6 ER+, 7 PR+, 8 HER2+ | 1.5 | Gd-DTPA (0.1 mmol/kg) |
| Rosen et al. [15] | 21 | — | II—IIIB | — | A plus P*4; A plus T*4 | — | 1.5 | Gd-DTPA (0.1 mmol/kg) |
| Wasser et al. [17] | 50 | 47 (31-64) | IIA—IIIB | IDC, ILC, IC | E plus P; E plus C | — | 1.5 | Gd-DTPA (0.1 mmol/kg) |
| Bodini et al. [33] | 73 | 56 (26-71) | IIa—IIIb | IDC, ILC, IC | E * 3-4 | — | 0.5 | Gd-DTPA (0.1 mmol/kg) |
| Denis et al. [34] | 40 | 47 (32-58). | LABC (non-IBC) | — | CEF*6; T-based | — | 1.5 | Gd-DTPA (0.2 mL/kg) |
| Londero et al. [35] | 15 | 48.8 (34-64) | LABC | IDC, ILC, IC | A plus C *2-T*2 | — | 1.0 | Gd-DTPA (0.1 mmol/kg) |
| Martincich et al. [36] | 30 | 49 (36-65) | IIa—IIIb | IDC, ILC, IC | A plus P*4 | 18 ER and PR+; 12 ER and PR- | 1.5 | Gadolinium chelate (0.1 mmol/kg) |
| Warren et al. [37] | 67 (69 breast cancers) | 46.2 (28-62) | Large or IBC or high-grade breast cancer | IDC, ILC | A plus C*6; A plus T*6; E*4→CMF*4; E*4; T plus H | 36 ER+ | 1.5 | Gadolinium (0.16 mmol/kg) |
| Demartini et al. [26] | 15 (16 breast cancers) | 45 (32-56) | LABC | IDC, ILC, IC | Anthracyclin-based; CMF | — | 1.5 | Gadolinium |
| Schott et al. [38] | 43 | 48 (26-66) | I—III | IDC, ILC, IC | A plus T* 4 | 25 ER+ | 1.5 | Gadolinium (0.15 mmol/kg) |
| Yeh et al. [39] | 41 | 47 (31-65) | IIB—III | IDC, ILC, IC | ddA*4→wP*9 | — | 1.5 | Gd-DTPA (0.1 mmol/kg) |
| Akazawa et al. [40] | 38 | — | — | IDC, ILC | T; E plus C→T; P→FEC | — | 1.0 | Gd-DTPA (0.1 mol/kg) |
| Belli et al. [41] | 45 | 53.7 (30-76) | IIa—IIIb | IDC, ILC, IC | CMF-A-CMF; E plus T | 20 ER+; 17 PR+ | 1.5 | Gd-DTPA (0.2 mL/kg) |
| Garimella et al. [42] | 76 | 52.6 (36.5-72) | LABC | — | CMF; CEF; 8 E plus C | — | 1.5 | — |
| Hsiang et al. [43] | 46 | 50 (30-70) | LABC | — | A plus C*2-4 →PCb ± H | 20 HER2+; 27 ER+ | 1.5 | Gd-DTPA-BMA (1 mL/10 lb) |
| Segara et al. [44] | 68 | 49.7 (29.5-71.6) | I—III | IC | Cis; XLD; H plus N; T plus Cb plus H | 28 ER+; 25 PR+; 51 HER2+ | 1.5 | Gd-DTPA (20 mL) |
| Bhattacharyya et al. [45] | 32 | 42 (24-60) | T1-3N0-2 | — | A plus C; or E plus C *6 | — | 1.5 | Gd-DTPA (16 mmol/kg) |
| Chen et al. [24] | 51 | — | LABC or T0-2N1 | IDC, ILC | A plus C *2-4→P plus Cb; N, P, and Cb ± H; avastin | 25 HER2+ | 1.5 | Gd-DTPA-BMA (1 mL/10 lbs) |
| Nicoletto et al. [46] | 26 | 47 (30-57) | Tumor ≥ 3 cm or IIIA or IIIB | IDC, ILC, IC | N plus E *4→T* 3-4 | 11 HR+ | — | — |
| Moon et al. [47] | 195 | 45.5 | — | — | A plus T; H plus P; anthracycline-based | 101 ER+; 68 PR+; 63 HER2+ | 1.5 | Gd-DTPA (0.1 mmol/kg |
Note—A = adimycine, BMA = bis(methyamide), C = cyclophosphomide, Cb = carboplatin, CEF = cyclophosphamide, epirubicine, and 5-fluorouracil, Cis = cisplatin, CMF = cyclophosphamide, methotrexate, and 5-fluorouracil, ddA = dose-dense adimycine, E = epirubicine; ER = estrogen receptor, FEC = 5-fluorouracil, epirubicine, and cyclophosphamide, Gd-DTPA = gadopentetate dimeglumine, H = trastuzumab, HER2 = human epidermal growth factor receptor 2, HR = hormonal receptor, IBC = inflammatory breast cancer; IC = invasive carcinoma, IDC = invasive ductal carcinoma, ILC = invasive lobular carcinoma, LABC = locally advanced breast cancer, N = navelbine; P = paclitaxel, PCb = paclitaxel plus carboplatin, PR = progesterone receptor, T = docetaxel, wP = weekly paclitaxel, XLD = capecetabine.
Data Synthesis
Individual study estimates and the weighted summary of sensitivity, specificity, and ORs and their 95% CIs; p value for heterogeneity; and I2 values are summarized in Figures 1, 2, 3. The pooled sensitivity and specificity were 63% and 91%, respectively. In seven studies, the 95% CIs of OR included 1; however, the pooled OR and corresponding 95% CIs did not, which, on the whole, confirmed the diagnostic value of MRI in predicting pathologic complete remission after preoperative therapy in breast cancer. The summary ROC curves, Q* index, and AUC values for MRI are shown in Figure 4. The Q* index estimate and AUC values were 0.8115 and 0.8810, respectively.
I2 is an index for heterogeneity: I2 = [Q – (k – 1)] / Q × 100%, where Q is the chi-square value of heterogeneity, and k is the number of studies included. Along with p < 0.05 for heterogeneity, I2 > 50% further indicates heterogeneity between studies. Although no heterogeneity was identified in the OR test (p = 0.665), the heterogeneity in sensitivity test and specificity test was highly significant (p < 0.0001 and I2 > 60%), confirming that there was strong evidence of between-study heterogeneity (Figs. 1, 2, 3).
To assess possible explanations for the heterogeneity, we applied single-factor meta-regression analysis by adding the publication year, the mean age of patients, the pathologic complete remission rate, and the ratios of patients positive for estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 separately as variants. Estrogen receptor status and progesterone receptor status were combined in one study and could not be separated [36]; thus, we did not include the data in meta-regression analysis. Statistical significance was found between pathologic complete remission and log OR (p = 0.0204). The coefficient and p value are listed in Table 2. The morphology of the lesions (mass vs nonmass enhancement) at baseline was considered as another variable. Nonmass enhancement lesions often fragmented after the therapy and fragmentation might affect the MRI performance in evaluating the tumoral response at the end of chemotherapy. However, because few studies reported such information, a meta-regression analysis could not be performed.
| Receptor Status | ||||||
|---|---|---|---|---|---|---|
| Variable | Pathologic Complete Remission | Mean Age | Publication Year | Progesterone Receptor Positive | Estrogen Receptor Positive | HER2 Positive |
| No. of studies | 25 | 21 | 25 | 5 | 8 | 5 |
| Coefficient | -6.160 | -0.090 | -0.065 | -5.433 | -3.340 | 1.072 |
| Standard error | 2.465 | 0.083 | 0.0983 | 12.652 | 5.263 | 2.587 |
| p | 0.020 | 0.292 | 0.5151 | 0.710 | 0.554 | 0.719 |
Note—HER2 = human epidermal growth factor receptor 2.
It has been established that magnets with an intensity field of ≥ 1 T are optimal for performing contrast-enhanced MRI of the breast and that ≥ 1.5-T magnets are even better [48]; however, we still enrolled one study [33] using a 0.5-T magnet and two studies using 1.0-T magnets [35, 40]. Subgroup analyses were thus executed to evaluate the contribution of the three studies to the heterogeneity (magnetic field strength < 1.5 T vs magnetic field strength of 1.5 T). We further conducted subgroup analysis of pathologic complete remission rates (< 20% vs ≥ 20%). Table 3 summarizes the results of the subgroup analyses of magnetic field strength and pathologic complete remission rate. No statistically significant differences in sensitivity, specificity, or OR (p > 0.05) were found among studies that used differing magnetic field strengths. The specificity of MRI in studies with a pathologic complete remission rate ≥ 20% was lower than that in studies with a pathologic complete remission rate < 20% (p = 0.0003); however, the sensitivity and OR did not show a statistically significant difference. Most studies included patients who received different preoperative therapy regimens, who were at different clinical stages, and who had different histologic subtypes and cancer grade, which made it impossible to conduct subgroup analyses for these variants. As shown in Figure 5, the funnel plots with studies missing from the bottom left quadrant suggested a publication bias. Additional studies with an OR of approximately 1 might have been conducted but not published because of unfavorable results.
| Subgroup | Sensitivity, Mean (95% CI) | Specificity, Mean (95% CI) | Diagnostic Odds Ratio, Mean (95% CI) | Q* Index (Mean) | Area Under the Curve (Mean) |
|---|---|---|---|---|---|
| Magnet strength (T) | |||||
| < 1.5 | 0.429 (0.099-0.816) | 0.908 (0.841-0.953) | 7.509 (1.286-43.840) | 0.6945 | 0.7514 |
| 1.5 | 0.627 (0.548-0.702) | 0.914 (0.892-0.932) | 17.648 (10.678-29.167) | 0.8098 | 0.8794 |
| Analytic results (p)a | 0.0742 | 0.3364 | 0.1709 | — | — |
| Pathologic complete remission (%) | |||||
| < 20 | 0.622 (0.543-0.696) | 0.907 (0.885-0.926) | 20.593 (10.877-38.991) | 0.8595 | 0.9253 |
| ≥ 20 | 0.793 (0.696-0.871) | 0.817 (0.770-0.857) | 14.155 (5.921-33.843) | 0.7967 | 0.8662 |
| Analytic results (p) | 0.1231 | 0.0003 | 0.6104 | — | — |
a
Results of subgroup analysis evaluating differences in sensitivity, specificity, and odds ratio between subgroups.
Discussion
In this meta-analysis, we explored the ability of breast MRI to predict pathologic complete remission of breast cancer after preoperative therapy, because accurate prediction of pathologic complete remission might define a subset of patients who can undergo successful breast-conserving surgery or define a subset of patients with a better prognosis who would not benefit from additional systemic chemotherapy. Our meta-analysis, which included data from 1,212 patients, has shown that contrast-enhanced MRI has high specificity (90.7%) and relative lower sensitivity (63.1%) in predicting pathologic complete remission after preoperative therapy in patients with breast cancer. Our results, together with those of previous reports, show that breast MRI is a highly accurate method of predicting residual tumor extent after preoperative therapy. However, a recent meta-analysis showed that MRI has high sensitivity (90%) and lower specificity (72%) in patients referred for biopsy of a breast lesion [1]. This difference can be partly explained by the various study populations. Our study focused on patients with a relatively later stage of breast cancer treated with preoperative therapy. Another reason may be associated with the effect of preoperative therapy, which reduces the enhancement of tumor tissue. When antiangiogenic drugs are used, the damaged tumor vessels may affect the delivery of MR contrast agents, thus leading to underestimation of residual disease [24, 49]. This finding warrants further investigation. Furthermore, preoperative therapy integrating antiangiogenesis drugs can break the tumor down into cells or cell clusters, which may not need the vascular supply to survive. In contrast to the concentric shrinkage model of tumor response, this cellular tumor breakdown can cause difficulty in the detection of residual disease with contrast-enhanced MRI [49]. On the basis of the relatively low sensitivity of MRI in predicting pathologic complete remission after preoperative therapy, breast radiologists should be more cautious in reporting complete remission at MRI and should give more importance to “minimal signs” after preoperative therapy.
Furthermore, the heterogeneity in sensitivity and specificity was highly significant (p < 0.0001 and I2 > 60%), confirming that there was strong evidence for between-study heterogeneity. In meta-regression analysis, statistical significance was found only between pathologic complete remission and log OR, with a correlation coefficient of –6.160. Several studies have indicated that the superiority of MRI in measuring responses to chemotherapy is expected, because of its high sensitivity and low specificity for residual disease in a patient population unlikely to achieve pathologic complete remission, and have suggested that MRI would be less accurate in a data set with a higher pathologic complete remission rate; the biologic mechanisms explaining the association between MRI accuracy and pathologic complete remission rate are not clear. One possible explanation can be the differential degree of histologic grade of breast cancer. Dynamic enhancement patterns and morphologic patterns of breast MRI have been reported to be associated with tumor grade [50, 51]. Higher histologic grade tumors are more likely to achieve pathologic complete remission, thereby affecting the MRI accuracy [52, 53]. The study by Moon et al. [47] showed that, for younger patients with human epidermal growth factor receptor 2–positive breast cancer, MRI accuracy in evaluating the residual disease was lower, which may be explained by the high prevalence of high-grade tumor. Another possible explanation can be the therapeutically induced alterations of breast cancer lesions, such as fibrosis, necrosis, and inflammation, the presence of discontinuous foci, and a decrease in contrast enhancement by the tumor after preoperative therapy [15]. Another reason may be associated with the antiangiogenesis effect of preoperative chemotherapy drugs, which can damage vessels in breast tissue, thereby impairing the delivery of MR contrast agents [49]. All of these changes may make it difficult for MRI to accurately predict pathologic complete remission, especially in a cohort with a high rate of pathologic complete remission; further study is needed to validate this assumption.
In a subgroup analysis of studies with pathologic complete remission rates higher or lower than 20%, a statistically significant difference was found in specificity only, suggesting that MRI has higher specificity (90.7%) in groups with pathologic complete remission rates lower than 20%. Subgroup analysis of magnetic field strength suggested no difference between studies adopting field strength less than 1.5 T and those with field strength equal to 1.5 T. However, it should be recognized that such estimates of effects might be imprecise because only three studies were available in the below 1.5 T subgroup. Formal meta-regression analyses showed no significant associations between log OR and the publication year, mean age of patients, and the ratios of patients positive for progesterone receptor, estrogen receptor, and human epidermal growth factor receptor 2 separately, which might be interpreted by the insufficient variability in explanatory variables between studies. We noticed that certain variables, such as progesterone receptor–positive and human epidermal growth factor receptor 2–positive ratios, were available in less than half of the eligible articles, which might impair the reliability of the result.
Although meta-regression analyses confirmed the contribution of the pathologic complete remission rate to heterogeneity, the analyses we performed to detect heterogeneity were still insufficient. Most studies differed according to preoperative therapy regimens and reported different histologic subtypes and cancer grades, which made it impossible to conduct subgroup analyses for these three variants. Other possible factors related to scanning method and acquisition protocol, such as TR, TE, field of view, flip angle, section thickness, and matrix size, were not taken into account because there were too many variables to include. Another difference lay in patients' menopausal status. MRI of the premenopausal breast might have reduced specificity because of the presence of hormone-reactive parts of the breast. Moreover, the detection of pseudo-lesions that enhance after administration of gadolinium-based contrast agents was variable throughout the menstrual cycle [54]. The use of hormone replacement therapy will also alter enhancement of the breast. The effect does not disappear completely immediately after stopping hormone replacement therapy once the diagnosis of breast cancer is made [55]. Unfortunately, we failed to assess whether these variants contributed to heterogeneity because few studies have reported relevant data (Table 1).
Our study has several limitations. First, gaps in the bottom left quadrant suggest the presence of publication bias, which may have led to an overestimation of the true diagnostic performance. It is possible that publication bias in this field restricted the publication of studies with less-promising results, because the funnel plots suggested a lack of studies with an OR value of approximately 1. In addition, we only included studies published in English, which might invoke the so-called “Tower of Babel” bias, which refers to the fact that investigators working in a language other than English could be sending only studies with positive results to international journals. Although certain less-qualified studies would be neglected by limiting the publication language to English, the Tower of Babel bias would make it possible that studies with negative results could have been left out. Another limitation stemmed from the fact that there were no reference standards for the interpretation of MRI results, though high rates of interobserver concordance were ensured.
In conclusion, our meta-analysis shows that MRI has high specificity (90.7%) and relative lower sensitivity (63.1%) in predicting pathologic complete remission after preoperative therapy in patients with breast cancer. The performance of breast MRI is influenced by the pathologic complete remission rate of preoperative therapy regimens, so it will be worthwhile to determine the associated biologic mechanisms and to perform further studies to validate the predictive power of breast MRI in patients with breast cancer who received preoperative therapy.
Footnotes
Address correspondence to K. W. Shen ([email protected]).
This research was supported in part by grants from the Leading Academic Discipline Project of Shanghai Municipal Education Commission (project J50208).
Y. Yuan and X. S. Chen contributed equally to this work.
References
1.
Peters NH, Borel Rinkes IH, Zuithoff NP, Mali WP, Moons KG, Peeters PH. Meta-analysis of MR imaging in the diagnosis of breast lesions. Radiology 2008; 246:116–124
2.
Kaufmann M, von Minckwitz G, Bear HD, et al. Recommendations from an international expert panel on the use of neoadjuvant (primary) systemic treatment of operable breast cancer: new perspectives 2006. Ann Oncol 2007; 18:1927 –1934
3.
Mauri D, Pavlidis N, Ioannidis JP. Neoadjuvant versus adjuvant systemic treatment in breast cancer: a meta-analysis. J Natl Cancer Inst 2005; 97:188 –194
4.
Wolmark N, Wang J, Mamounas E, Bryant J, Fisher B. Preoperative chemotherapy in patients with operable breast cancer: nine-year results from National Surgical Adjuvant Breast and Bowel Project B-18. J Natl Cancer Inst Monogr 2001; 30:96 –102
5.
Bear HD, Anderson S, Brown A, et al. The effect on tumor response of adding sequential preoperative docetaxel to preoperative doxorubicin and cyclophosphamide: preliminary results from National Surgical Adjuvant Breast and Bowel Project Protocol B-27. J Clin Oncol 2003; 21:4165 –4174
6.
Rastogi P, Anderson S, Bear HD, et al. Preoperative chemotherapy: updates of National Surgical Adjuvant Breast and Bowel Project Protocols B-18 and B-27. J Clin Oncol 2008; 26:778–785
7.
Colleoni M, Viale G, Zahrieh D, et al. Expression of ER, PgR, HER1, HER2, and response: a study of preoperative chemotherapy. Ann Oncol 2008; 19:465 –472
8.
Tiezzi DG, Andrade JM, Ribeiro-Silva A, Zola FE, Maran HR, Tiezzi MG. HER-2, p53, p21 and hormonal receptors proteins expression as predictive factors of response and prognosis in locally advanced breast cancer treated with neoadjuvant docetaxel plus epirubicin combination. BMC Cancer 2007; 7:36
9.
Goldstein NS, Decker D, Severson D, et al. Molecular classification system identifies invasive breast carcinoma patients who are most likely and those who are least likely to achieve a complete pathologic response after neoadjuvant chemotherapy. Cancer 2007; 110:1687 –1696
10.
Gianni L, Zambetti M, Clark K, et al. Gene expression profiles in paraffin-embedded core biopsy tissue predict response to chemotherapy in women with locally advanced breast cancer. J Clin Oncol 2005; 23:7265 –7277
11.
Ayers M, Symmans WF, Stec J, et al. Gene expression profiles predict complete pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 2004; 22:2284 –2293
12.
Cleator S, Tsimelzon A, Ashworth A, et al. Gene expression patterns for doxorubicin (Adriamycin) and cyclophosphamide (cytoxan) (AC) response and resistance. Breast Cancer Res Treat 2006; 95:229–233
13.
Chagpar AB, Middleton LP, Sahin AA, et al. Accuracy of physical examination, ultrasonography, and mammography in predicting residual pathologic tumor size in patients treated with neoadjuvant chemotherapy. Ann Surg 2006; 243:257–264
14.
Skaane P, Young K, Skjennald A. Population-based mammography screening: comparison of screen-film and full-field digital mammography with soft-copy reading—Oslo I study. Radiology 2003; 229:877 –884
15.
Rosen EL, Blackwell KL, Baker JA, et al. Accuracy of MRI in the detection of residual breast cancer after neoadjuvant chemotherapy. AJR 2003; 181:1275 –1282
16.
Trikalinos TA, Salanti G, Zintzaras E, Ioannidis JP. Meta-analysis methods. Adv Genet 2008; 60:311–334
17.
Wasser K, Sinn HP, Fink C, et al. Accuracy of tumor size measurement in breast cancer using MRI is influenced by histological regression induced by neoadjuvant chemotherapy. Eur Radiol 2003; 13:1213 –1223
18.
Whiting P, Rutjes AW, Reitsma JB, Bossuyt PM, Kleijnen J. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003; 3:25
19.
Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Med Res Methodol 2006; 6:31
20.
Green MC, Buzdar AU, Smith T, et al. Weekly paclitaxel improves pathologic complete remission in operable breast cancer when compared with paclitaxel once every 3 weeks. J Clin Oncol 2005; 23:5983 –5992
21.
von Minckwitz G, Kummel S, Vogel P, et al. Intensified neoadjuvant chemotherapy in early-responding breast cancer: phase III randomized GeparTrio study. J Natl Cancer Inst 2008; 100:552–562
22.
Wasser K, Klein SK, Fink C, et al. Evaluation of neoadjuvant chemotherapeutic response of breast cancer using dynamic MRI with high temporal resolution. Eur Radiol 2003; 13:80–87
23.
Chen JH, Mehta RS, Malcioglu O, Su MY. Inflammatory breast cancer following neoadjuvant chemotherapy: can magnetic resonance imaging precisely predict the pathological response? Ann Surg Oncol 2008; 15:3609 –3613
24.
Chen JH, Feig B, Agrawal G, et al. MRI evaluation of pathologically complete response and residual tumors in breast cancer after neoadjuvant chemotherapy. Cancer 2008; 112:17–26
25.
Chen X, Moore MO, Lehman CD, et al. Combined use of MRI and PET to monitor response and assess residual disease for locally advanced breast cancer treated with neoadjuvant chemotherapy. Acad Radiol 2004; 11:1115 –1124
26.
Demartini WB, Lehman CD, Peacock S, Russell MT. Computer-aided detection applied to breast MRI: assessment of CAD-generated enhancement and tumor sizes in breast cancers before and after neoadjuvant chemotherapy. Acad Radiol 2005; 12:806–814
27.
Esserman L, Kaplan E, Partridge S, et al. MRI phenotype is associated with response to doxorubicin and cyclophosphamide neoadjuvant chemotherapy in stage III breast cancer. Ann Surg Oncol 2001; 8:549 –559
28.
Balu-Maestro C, Chapellier C, Bleuse A, Chanalet I, Chauvel C, Largillier R. Imaging in evaluation of response to neoadjuvant breast cancer treatment benefits of MRI. Breast Cancer Res Treat 2002; 72:145 –152
29.
Partridge SC, Gibbs JE, Lu Y, Esserman LJ, Sudilovsky D, Hylton NM. Accuracy of MR imaging for revealing residual breast cancer in patients who have undergone neoadjuvant chemotherapy. AJR 2002; 179:1193 –1199
30.
Rieber A, Brambs HJ, Gabelmann A, Heilmann V, Kreienberg R, Kühn T. Breast MRI for monitoring response of primary breast cancer to neoadjuvant chemotherapy. Eur Radiol 2002; 12:1711 –1719
31.
Cheung YC, Chen SC, Su MY, et al. Monitoring the size and response of locally advanced breast cancers to neoadjuvant chemotherapy (weekly paclitaxel and epirubicin) with serial enhanced MRI. Breast Cancer Res Treat 2003; 78:51 –58
32.
Delille JP, Slanetz PJ, Yeh ED, Halpern EF, Kopans DB, Garrido L. Invasive ductal breast carcinoma response to neoadjuvant chemotherapy: noninvasive monitoring with functional MR imaging pilot study. Radiology 2003; 228:63–69
33.
Bodini M, Berruti A, Bottini A, et al. Magnetic resonance imaging in comparison to clinical palpation in assessing the response of breast cancer to epirubicin primary chemotherapy. Breast Cancer Res Treat 2004; 85:211 –218
34.
Denis F, Desbiez-Bourcier AV, Chapiron C, Arbion F, Body G, Brunereau L. Contrast enhanced magnetic resonance imaging underestimates residual disease following neoadjuvant docetaxel based chemotherapy for breast cancer. Eur J Surg Oncol 2004; 30:1069 –1076
35.
Londero V, Bazzocchi M, Del Frate C, et al. Locally advanced breast cancer: comparison of mammography, sonography and MR imaging in evaluation of residual disease in women receiving neoadjuvant chemotherapy. Eur Radiol 2004; 14:1371 –1379
36.
Martincich L, Montemurro F, De Rosa G, et al. Monitoring response to primary chemotherapy in breast cancer using dynamic contrast-enhanced magnetic resonance imaging. Breast Cancer Res Treat 2004; 83:67 –76
37.
Warren RM, Bobrow LG, Earl HM, et al. Can breast MRI help in the management of women with breast cancer treated by neoadjuvant chemotherapy? Br J Cancer 2004; 90:1349 –1360
38.
Schott AF, Roubidoux MA, Helvie MA, et al. Clinical and radiologic assessments to predict breast cancer pathologic complete response to neoadjuvant chemotherapy. Breast Cancer Res Treat 2005; 92:231 –238
39.
Yeh E, Slanetz P, Kopans DB, et al. Prospective comparison of mammography, sonography, and MRI in patients undergoing neoadjuvant chemotherapy for palpable breast cancer. AJR 2005; 184:868 –877
40.
Akazawa K, Tamaki Y, Taguchi T, et al. Preoperative evaluation of residual tumor extent by three-dimensional magnetic resonance imaging in breast cancer patients treated with neoadjuvant chemotherapy. Breast J 2006; 12:130 –137
41.
Belli P, Costantini M, Malaspina C, Magistrelli A, Latoree G, Bonomo L. MRI accuracy in residual disease evaluation in breast cancer patients treated with neoadjuvant chemotherapy. Clin Radiol 2006; 61:946 –953
42.
Garimella V, Qutob O, Fox JN, et al. Recurrence rates after DCE-MRI image guided planning for breast-conserving surgery following neoadjuvant chemotherapy for locally advanced breast cancer patients. Eur J Surg Oncol 2007; 33:157 –161
43.
Hsiang DJ, Yamamoto M, Mehta RS, et al. Predicting nodal status using dynamic contrast-enhanced magnetic resonance imaging in patients with locally advanced breast cancer undergoing neoadjuvant chemotherapy with and without sequential trastuzumab. Arch Surg 2007; 142:855 –861
44.
Segara D, Krop IE, Garber JE, et al. Does MRI predict pathologic tumor response in women with breast cancer undergoing preoperative chemotherapy? J Surg Oncol 2007; 96:474–480
45.
Bhattacharyya M, Ryan D, Carpenter R, Vinnicombe S, Gallagher CJ. Using MRI to plan breast-conserving surgery following neoadjuvant chemotherapy for early breast cancer. Br J Cancer 2008; 98:289 –293
46.
Nicoletto MO, Nitti D, Pescarini L, et al. Correlation between magnetic resonance imaging and histopathological tumor response after neoadjuvant chemotherapy in breast cancer. Tumori 2008; 94:481 –488
47.
Moon HG, Han W, Lee JW, et al. Age and HER2 expression status affect MRI accuracy in predicting residual tumor extent after neo-adjuvant systemic treatment. Ann Oncol 2009; 20:636–641
48.
Sardanelli F, Giuseppetti GM, Canavese G, et al. Indications for breast magnetic resonance imaging. Consensus document “Attualità in senologia”, Florence 2007. Radiol Med 2008; 113:1085 –1095
49.
Bahri S, Chen JH, Mehta RS, et al. Residual breast cancer diagnosed by MRI in patients receiving neoadjuvant chemotherapy with and without bevacizumab. Ann Surg Oncol 2009; 16:1619 –1628
50.
Neubauer H, Li M, Kuehne-Heid R, Schneider A, Kaiser WA. High grade and non-high grade ductal carcinoma in situ on dynamic MR mammography: characteristic findings for signal increase and morphological pattern of enhancement. Br J Radiol 2003; 76:3–12
51.
Daldrup H, Shames DM, Wendland M, et al. Correlation of dynamic contrast-enhanced MR imaging with histologic tumor grade: comparison of macromolecular and small-molecular contrast media. AJR 1998; 171:941 –949
52.
Axelrod D, Smith J, Kornreich D, et al. Breast cancer in young women. J Am Coll Surg 2008; 206:1193 –1203
53.
Kollias J, Elston CW, Ellis IO, Robertson JF, Blamey RW. Early-onset breast cancer—histopathological and prognostic considerations. Br J Cancer 1997; 75:1318 –1323
54.
Kuhl CK, Bieling H, Gieseke J, et al. Breast neoplasms: T2* susceptibility-contrast, first-pass perfusion MR imaging. Radiology 1997; 202:87–95
55.
Reichenbach JR, Przetak C, Klinger G, Kaiser WA. Assessment of breast tissue changes on hormonal replacement therapy using MRI: a pilot study. J Comput Assist Tomogr 1999; 23:407–413
Information & Authors
Information
Published In
Copyright
© American Roentgen Ray Society.
History
Submitted: November 2, 2009
Accepted: December 29, 2009
First published: November 23, 2012
Keywords
Authors
Metrics & Citations
Metrics
Citations
Export Citations
To download the citation to this article, select your reference manager software.