| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Netherlands Cancer Institute, Antoni van Leeuwenhoek
Hospital, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands.
2 Department of Medical Oncology, Netherlands Cancer Institute, Antoni van
Leeuwenhoek Hospital, Amsterdam, The Netherlands.
3 Department of Pathology, Netherlands Cancer Institute and Amsterdam Medical
Centre, Amsterdam, The Netherlands.
4 Department of Surgery, Netherlands Cancer Institute, Antoni van Leeuwenhoek
Hospital, Amsterdam, The Netherlands.
Received December 19, 2007;
accepted after revision May 18, 2008.
Address correspondence to C. E. Loo
(c.loo{at}nki.nl).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. In 54 patients with breast cancer, dynamic contrast-enhanced MRI was performed before chemotherapy and after two chemotherapy cycles. Imaging was correlated with final histopathology. Multivariate analysis with cross-validation was performed on MRI features describing kinetics and morphology of contrast uptake in the early and late phases of enhancement. Receiver operating characteristic (ROC) analysis was used to develop a guideline that switches patients at high risk for incomplete remission to a different chemotherapy regimen while maintaining first-line therapy in 95% of patients who are not at risk (i.e., high specificity).
RESULTS. Change in largest diameter of late enhancement during chemotherapy was the single most predictive MRI characteristic for tumor response in multivariate analysis (Az [area under the ROC curve] = 0.73, p < 0.00001). Insufficient (< 25%) decrease in largest diameter of late enhancement during chemotherapy was most indicative of residual tumor at final pathology. Using this criterion, the fraction of unfavorable responders indicated by MRI was 41% (22/54). Approximately half (44%, 14/32) of the patients who showed favorable response at MRI achieved complete remission at pathology. Conversely, 95% (21/22) of patients who showed unfavorable response at MRI had residual tumor at pathology.
CONCLUSION. Reduction of less than 25% in largest diameter of late enhancement during neoadjuvant chemotherapy shows the potential to predict residual tumor after therapy with high specificity.
Keywords: breast carcinoma • chemotherapy response • MRI
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A potential advantage of neoadjuvant chemotherapy is reduction of the size of the primary tumor, increasing the probability that breast-conserving surgery can be performed [4]. Achieving a pathologically confirmed complete remission or minimal disease at pathology [5] is consistently associated with a favorable disease-free and overall survival [1, 3, 6]. It is reasonable to view complete or near-complete remission at pathology as a sur rogate marker of extreme chemotherapy responsiveness and as an appropriate short-term goal of treatment. Unfortunately, complete remission is achieved in only a minority of patients. Attempts to increase the rate of complete remission include the administration of multiple drugs or higher doses of drugs, but also tailoring of chemotherapy to the specific sensitivity of the individual tumor. This strategy allows an individual patient to cross over to a different chemotherapy regimen if there is insufficient response and complete remission is unlikely to occur. On the other hand, if major responsiveness is observed, complete remission may ensue from continuation of the started chemotherapy regimen.
In vitro techniques continue to be unreliable in predicting response to chemotherapy [7]. Preoperative chemotherapy provides a unique opportunity to monitor response of the tumor and to tailor the treatment to each individual patient. Monitoring of response has traditionally been based on physical examination, mammography, and sonography of the breast. However, these techniques were found to result in unsatisfactory accuracy [8, 9].
Dynamic contrast-enhanced MRI allows visualization of functional properties of the tumor such as those associated with angiogenesis [10] in addition to morphologic properties such as tumor size. Several studies have found that MRI is the most accurate technique for evaluating the extent of residual disease after systemic treatment [11–15]. MRI may even allow assessment of the pathophysiologic response to chemotherapy, which occurs before volume changes [16, 17]. Moreover, it has been suggested that application of MRI may improve the complete remission rate [17–19]. However, the clinical value of dynamic MRI for predicting the efficacy of neoadjuvant chemotherapy during treatment has not been fully assessed. Quantitative guidelines to assess response to therapy based on MRI are currently lacking.
The purpose of our study was twofold. The first aim was to identify MRI features during neoadjuvant chemotherapy that predict which tumors will achieve a complete or near-complete remission at pathology and which will not. The second aim was to establish a practical MRI test to identify which tumors will not achieve complete or near-complete remission with the given chemotherapy regimen and may therefore benefit from a switch to an alternative regimen.
|
|
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Between September 2000 and September 2004, 54 patients were included to identify MRI features predictive of incomplete remission at final pathology in response to neoadjuvant chemo therapy. On the basis of these features, we formulated an MRI response prediction test. The patients re ceived neoadjuvant chemotherapy in a single-institution randomized phase II trial at The Netherlands Cancer Institute [7]. The institutional review board approved the study protocol, and written informed consent was obtained from all patients.
In this trial, patients were randomized to receive either doxorubicin 60 mg/m2 and cyclophospha mide 600 mg/m2 or doxorubicin 50 mg/m2 and docetaxel 75 mg/m2. Chemotherapy was admini stered every 3 weeks in a total of six cycles.
Patients who declined randomization for the chemotherapy regimen received the standard doxorubicin–cyclophosphamide chemotherapy and were also included in the MRI study. MRI was performed before the first course of chemotherapy and before the third course (in the fifth or sixth week of treatment).
Surgery
After the last course of chemotherapy, all selected patients underwent
mastectomy or breast-conserving therapy according to the standard protocols at
our institute.
MRI Technique
MRI was performed on a 1.5-T Magnetom Vision scanner (Siemens Medical
Solutions) with a dedicated bilateral phased-array breast coil. Images were
acquired with the patient in the prone position and with both breasts imaged
simultaneously. A comprehensive dynamic protocol consisted of fast dynamic
imaging in the first 45 seconds followed by slow dynamic imaging in five
consecutive series at 90-second intervals.
The protocol started with an unenhanced coronal 3D T1-weighted low-angle shot sequence. Subsequently, 30 fast dynamic series were initiated; after the third series, a bolus of gadoteridol (Prohance, Bracco-Byk Gulden; 14 mL, 0.1 mmol/kg) was administered IV at 2–4 mL/s by a power injector (Spectris, Medrad).
The fast dynamic series (low resolution) were transverse 2D gradient-echo series (low-angle shot) with seven slices of 8 mm (distance factor, 0.5) and a pixel size of 6.67 x 5.0 mm (acquisition time, 1,160 milliseconds; TR/TE, 29.19/1.4; flip angle, 30°; 1 acquisition; field of view, 320 mm). The slow dynamic series (high resolution) consisted of a coronal low-angle shot 3D acquisition equivalent to the unenhanced coronal 3D series, with a voxel size of 1.21 x 1.21 x 1.69 mm and the following parameters: acquisition time, 90 sec onds; 1 acquisition; 8.1/4.0; flip angle, 20°; and field of view, 310 mm. Acquisition time was 7.5 minutes for the entire dynamic series.
Radiologic Assessment and Clinical Reading
Interpretation of breast MR images was done using a viewing station that
permitted simultaneous viewing of two series reformatted and linked in three
orthogonal directions [20].
Dynamic curves of contrast enhancement in the first 45 seconds were also
available. The viewing station displays all image series (unenhanced and
contrast-enhanced), subtraction images at 90-second intervals and maximum
intensity projections (MIPs) of both breasts (Fig.
1A,
1B,
1C,
1D,
1E,
1F).
Using these tools, two dedicated radiologists ex perienced in evaluating MRI of the breast, and working in consensus, interpreted the cases. The assessment of all features took approximately 15 minutes per scan. The radiologists were unaware of pathology outcome. Temporal and morphologic char acteristics of contrast uptake were scored in four groups (Table 1): bolus enhancement, initial enhance ment, late enhancement, and over all judgment.
|
The characteristics in the first group (Table 1, Bolus enhancement) all involve temporal features of contrast uptake in the first 45 seconds after injection of the contrast agent (fast dynamic, low-resolution series). The time between contrast enhancement in the aorta and in the tumor was determined in regions of interest (ROIs) manually selected in the tumor and in the aorta. Relative increase in enhancement of the tumor 3 seconds after the start of enhancement was calculated by [(SI3 – SIstart) / SIstart] x 100%, where SIstart denotes the signal intensity at the first sign of enhancement (abrupt increase of signal intensity of the tumor on the curve), and SI3, the intensity after 3 seconds. Comparable equations were used for the time points at 12 and 17 seconds.
The characteristics in the second group
(Table 1, Initial enhancement)
were assessed in the series 90 seconds after contrast injection (slow dynamic,
high resolution) minus the unenhanced series. Change in extent of the
enhancing tumor was evaluated in both quantitative and qualitative terms. In
quantitative terms, the largest diameter of the tumor was assessed and
measured in three reformatted planes (sagittal, axial, and coronal), and the
percentage of difference in the largest diameters between scans 1 and 2 was
denoted (Table 1, Initial
enhancement, footnote c). In cases of rim enhancement, the necrotic core was
included in measurement of the largest diameter (Figs.
1A,
1B,
1C). In cases of multinodular
or dif fuse tumor growth, measurement of the complete enhancing area,
including intermediate (nonenhancing) tissue occupied by tumor, was assessed.
In qualitative terms, the differences in extent of the enhancing tumor between
scan 2 and scan 1 were assessed in four categories: progression, no change,
decrease < 50%, or decrease 
50%. Morphology of enhance ment was scored
in three categories: mass, multinodular, and diffuse. In addition to the
patterns at scan 1 and scan 2, the change in pattern between scans 1 and 2 was
denoted in three categories: shrinking mass, diffuse reduction, and reduction
to small residual foci in the original tumor. Relative enhancement was given
by [(SI90 – SI0) / SI0] x 100%,
where SI0 denotes the signal intensity at time point 0 (before
contrast administration) and SI90, the time point after 90 seconds.
Each signal intensity was measured in an ROI that covered the most enhancing
part of the tumor.
The characteristics in the third group (Table 1, Late enhancement) were assessed in the series 450 seconds after contrast injection minus the 90-second series. Washout was defined as an area of decreased signal intensity at 450 seconds relative to 90 seconds, represented as an area of dark—i.e., black—signal intensity at subtracted images. Plateau was defined as an area of unchanged signal intensity at 450 seconds relative to 90 seconds, represented as an area of dark—i.e., gray—signal intensity on subtracted images. The largest diameter of the total area occupied by washout or plateau kinetics in the tumor, including intermediate nonenhancing tissue (e.g., necrotic core), was measured and defined as LDlate (Figs. 1D, 1E, 1F), where LD is largest diameter. Differences in the extent of late enhancement in the tumor between scan 1 and scan 2 were described in quantitative (Table 1, Late enhancement, footnote g) and in qualitative terms comparable to those described in the preceding text for initial enhancement. Relative late enhancement was given by [(SI450 – SI90) / SI90] x 100%, where SI450 denotes the signal intensity at 450 seconds. Measurement of signal intensity was done in the part of the tumor that showed the strongest washout. These parts were examined by interactively moving the cursor in the three formatted planes, producing underlying relative late enhancement values (in percentages).
Overall assessment in the fourth group (Table 1, Overall judgment) was provided by the radiologists who considered all of the afore mentioned characteristics to produce a qualitative assessment of response in four categories: com plete remission (no enhancement), partial re mission, no response, and progression.
Evaluation of Response
Response of the primary tumor was determined by evaluating the surgical
specimen at pathology. Complete remission was defined as the complete absence
of residual tumor cells at microscopy. When a small number of scattered tumor
cells were seen, the samples were classified as near-complete pathologic
remission. Because the aim of this study was to identify MRI features
associated with a high specificity of the primary tumor to chemotherapy,
tumors with near-complete pathologic remission were included in the group of
complete remission for analytic purposes
[7,
21].
All other responses involving residual vital tumor in the surgical specimen were defined as incomplete remission at final pathology. The presence or absence of tumor-positive axillary lymph nodes was not taken into account for this analysis.
Statistical Analysis
The software programs SPSS 10 (SPSS) and Matlab R12 (TheMathWorks) were
used for all analyses. A p value of 0.002 was defined as a
significant test result (p = 0.05 ad justed for the number of tested
characteristics using the Bonferroni correction). Univariate analyses were
performed using Student's t tests for continu ous normally
distributed characteristics, Mann-Whitney U exact tests for
abnormally dis tributed continuous characteristics, and Fisher's exact tests
for binomial categoric characteristics.
The features that yielded significant separation after Bonferroni correction were included in the multivariate analysis. For multivariate analysis, logistic regression with feature selection by double cross-validation was used [22, 23]. The aim of this approach is to obtain a combination of characteristics that is unlikely to be predictive by chance. The analysis consists of validation in an inner loop and in an outer loop. In the outer loop, each case was left out consecutively. In the inner loop, 100 10-foldcross-validations were performed for each feature combination. In each cross-validation, a logistic model was built from the training set, and the area (Az) under the receiver operating characteristic (ROC) curve was used to quantify the performance of the cross-validated model [24]. Feature select ion was restricted to three simultaneous covariates at most. The feature combination that yielded the best performance was retrained on all cases in the inner loop and subsequently applied to the omitted case in the outer loop to produce a posterior probability of unfavorable response. This pro cedure was repeated until all cases were tested in the outer loop. The feature combination that was selected most frequently was chosen to be the predictive model, and the performance of this model was quantified by the area under the ROC curve obtained from the posterior probabilities assigned to each case.
MRI Response Prediction Test
After identifying multivariate MRI features predictive for residual tumor,
we chose a point on the ROC curve corresponding to a maximum of 5%
false-positive interpretations of complete remission (which would lead to
switch of therapy in excellent responders). This operating point was
translated into an MRI response prediction test to guide treatment decisions
during chemotherapy.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Pathology Results
Thirty-nine of 54 (72%) patients had residual tumor, and 15 (28%) had
complete remission at final pathology.
MRI Characteristics
In univariate analysis, six MRI characteristics were found to be
significantly associated with residual tumor at pathology
(Table 3). In multivariate
analysis, only a single MRI characteristic remained: change in largest
diameter of late enhancement in the tumor on scan 2 relative to scan 1:
LDlate2–1 = [(LDlatescan2 –
LDlatescan1) / LDlatescan1] x 100%. The area under
the ROC curve is 0.73 after double validation. Because of the prior condition
that therapy of the excellent responders (complete remission at pathology) be
switched to alternative therapy in fewer than 5% cases, the operating point on
the ROC curve is at 25% reduction in the largest diameter of late enhancement
in the tumor on scan 2 relative to scan 1 (LDlate2–1).
|
|
When LDlate2–1 was at least 25%, 14 of the 15 patients with complete remission at pathology were correctly identified. The true-positive fraction was 93.3% (14/15) (Fig. 2, upper arm). Thirteen patients showed complete disappearance of washout or plateau in the tumor during treatment (LDlate2–1 = 100%). Nine (9/15, 60%) had complete remission at final pathology.
The MRI response prediction test was formulated as follows: Less than 25% reduction of the largest diameter of late enhancement on MRI (scan 2 relative to scan 1) during chemotherapy is indicative of residual tumor (Figs. 3A, 3B, 3C, 3D, 3E, 3F and 4A, 4B, 4C, 4D, 4E, 4F). In the remainder of this article, we define this outcome as an unfavorable response predicted with MRI.
|
|
|
|
|
|
|
|
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We found that the change of largest diameter of the late enhancement region (LDlate2–1) is the best predictor of final pathologic response on MRI during neoadjuvant chemotherapy among a large set of MRI features. This characteristic is a combination of morphology (extent) and kinetics (late enhancement) associated with tumor vascularity. Reduction less than 25% during therapy was associated with residual disease at pathology after therapy.
Others also studied the relation between morphologic features of a tumor and treatment outcome. Initial tumor size has long been known to predict patient survival and is the basis of most disease-staging systems (TNM classification). In a study by Cheung et al. [25], who evaluated 33 patients with serial MRI, it was found that all complete responders had a marked early size reduction of more than 45%.
A study performed by Partridge et al. [26] showed that initial tumor volume measured on MRI is the most significant predictive variable of recurrence-free survival, which is entirely in accordance with the rationale of the TNM classification. Their multivariate analysis showed that the initial MRI tumor volume and the change in MRI volume after completion of chemotherapy were significant independent predictors for recurrence-free survival. Early change in tumor volume at MRI after one chemotherapy cycle showed a trend of association with recurrence-free survival.
Almost all studies focused on tumor enhancement in the very early (0–60 seconds) phase or initial enhancement (60–120 seconds) rather than delayed enhancement. One of the first published studies by Gilles et al. [15] showed correlation between intensity of enhancement and the presence of residual tumor after treatment. Our current study evaluated the potential value of a combined MRI protocol involving very early, initial, and late enhancement. Although we found the change of largest diameter of initial enhancement on MRI scan 2 relative to that in MRI scan 1 important to predict pathology outcome (Table 2), the change in largest diameter of late enhancement after multivariate analysis was found to be the strongest predictive characteristic. Although more limited than assessment of volume, the largest diameter measurements (in three planes) are easily applicable to sites that have no access to dedicated volumetric analysis tools. We were not able to compare our results with recurrence-free survival or survival because of short follow-up time.
To our knowledge, the only other study that reported measurements of late enhancement in the tumor was performed by El Khoury et al. [27]. Those authors examined the MRI quantification of the washout changes in breast tumors. They reported that quantification of washout variation in breast tumors is feasible and reproducible, and a significant reduction of washout volume was noted during chemotherapy (after two courses). Limitations of that study were the small number of patients (n = 19) and that only one patient achieved a complete response at pathology. Nonetheless, the results of our study using largest-diameter measurements are in agreement with these findings.
In our study, complete disappearance of washout or plateau in the tumor was observed in 13 of 54 tumors. In nine (69%) patients, this was associated with complete remission at pathology. In only four (10%) patients with residual tumor at final pathology, complete disappearance of washout or plateau in the tumor was noted. Balu-Maestro et al. [28] reported the disappearance of early abnormal enhancement in five cases in which complete histologic response occurred. Rieber et al. [29] found that flattening of the time–intensity curve (i.e., disappearance of washout) of breast cancer after one course of chemotherapy and the absence of contrast uptake after four courses indicate responders.
The combination of morphologic and kinetic characteristics to have predictive value during neoadjuvant chemotherapy has also been reported by others. In the 30 patients studied by Martincich et al. [30], a reduction in the early enhancement ratio and > 65% reduction in tumor volume after two cycles of chemotherapy were associated with a major histologic response (univariate analysis); these parameters did not retain statistical significance in multivariate analysis. Rather than focusing on initial and late enhancement, others applied pharmacokinetic models of contrast uptake using the entire time–signal intensity curve to study the response to neoadjuvant chemotherapy [16, 17, 30]. Padhani et al. [17] studied the changes in contrast agent kinetics during chemotherapy to determine whether kinetic measures can be used to predict final pathologic response. Their initial clinical results in 25 patients showed that size (morphology) and transfer constant range (kinetics) were equally accurate in predicting the absence of pathologic response after two cycles of treatment.
Although pharmacokinetic models have the potential advantage of giving insight into the underlying physiologic processes of contrast uptake, it is generally difficult to compare results among different studies because of the current lack of standardization in these models.
As in any diagnostic test, we had to choose a cutoff value that balances the true-positive fraction of the MRI test against the false-positive fraction. Switching therapy away from an already effective regimen (false-positive test result) may harm the outcome of patients with excellent response. A multidisciplinary team of radiologists, surgeons, and medical oncologists decided that the MRI test—a diagnostic tool under investigation—should not cause potential harm to more than 5% of the patients in the current clinical setting. Consequently, only 54% of patients with residual tumor at pathology could be correctly identified. Another way of considering this, however, is that almost half the number of patients may benefit from switching therapy with negligible risk, compared with the current clinical practice of not switching. Another advantage of this conservative threshold is that it allows fine-tuning of the trade-off in subsequent studies.
Our study also has some limitations. The study results were assessed in a single-institution study in a relatively small set of patients. The suggested MRI response prediction test needs to be evaluated in prospectively included patients from multiple institutions in whom chemotherapy will be switched as a result of the MRI test. Another aspect of validation concerns assessment of the impact of different chemotherapeutic agents on the performance of the MRI test. The ultimate outcome of any successful regimen is destruction of the tumor, including its neovascular system, which is evaluated by the MRI response prediction test. However, the timing of the second MRI or the 25% operating point may be fine-tuned to various chemotherapy regimens. Another limitation is that experienced radiologists evaluated the MRI in consensus, so we were not able to assess intra- and interobserver variability.
In summary, our study shows that change of largest late enhancement diameter in the tumor on serial MRI has the potential to assess early breast cancer response to neoadjuvant chemotherapy. In clinical practice, the MRI response prediction test may offer the oncologist an objective tool of high specificity to tailor the chemotherapy for each individual patient. However, additional assessment of the proposed MRI test in larger numbers of patients will be required to validate these findings.
Acknowledgments
We thank Angelique Schlief of data management and Eline Deurloo of the
radiology department for their dedication and contributions.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. Rodenhuis, I. A. M. Mandjes, J. Wesseling, M. J. van de Vijver, M.-J. T. D. F. Vrancken Peeters, G. S. Sonke, and S. C. Linn A simple system for grading the response of breast cancer to neoadjuvant chemotherapy Ann. Onc., August 28, 2009; (2009) mdp348v1. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
