| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, University of California, San Francisco, San
Francisco, CA 94143.
3 Department of Surgery, University of California, San Francisco, San Francisco,
CA 94143.
4 Department of Oncology, University of Texas Southwestern Medical Center,
Dallas, TX 75390.
5 Department of Oncology, University of California, San Francisco, San
Francisco, CA 94143.
Received July 12, 2004;
accepted after revision September 9, 2004.
Address correspondence to S. C. Partridge
(scp3{at}u.washington.edu).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. The study included 62 patients undergoing neoadjuvant chemotherapy. The longest diameter and volume of each tumor were measured on MRI before and after one and four cycles of treatment. Change in diameter on clinical examination, tumor size at pathology, and the number of positive nodes were determined. Each measure of tumor extent was assessed for the ability to predict RFS.
RESULTS. Univariate Cox analysis showed initial MRI volume was the strongest predictor of RFS (p = 0.002). Final change in MRI volume (p = 0.015) was more predictive than change in diameter on MRI (p = 0.077) or clinical examination (p = 0.27). Initial diameter on MRI (p = 0.003) and clinical examination (p = 0.033), tumor size at pathology (p = 0.016), and number of positive nodes (p = 0.045) were also significantly predictive of RFS. Early change in MRI volume (p = 0.071) and diameter (p = 0.081) after one chemotherapy cycle showed trends of association with RFS. Multivariate analysis showed initial MRI volume (p = 0.005) and final change in MRI volume (p = 0.003) were significant independent predictors.
CONCLUSION. MRI tumor volume was more predictive of RFS than tumor diameter, suggesting that volumetric changes measured using MRI may provide a more sensitive assessment of treatment efficacy.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
One advantage of a neoadjuvant chemotherapy regimen is that it allows observation of changes in the tumor during treatment to assess treatment efficacy. Large clinical trials have found that the degree of response of the primary tumor to preoperative chemotherapy correlates with patient survival [2-5]. This suggests tumor response may be a surrogate for evaluating the effect of chemotherapy on micrometastases and could therefore be an important prognostic indicator of treatment outcome.
Clinical examination based on palpable change in tumor size is the most common method for monitoring treatment effectiveness, and clinical response is recognized as a factor of prognostic importance [2, 4, 5]. X-ray mammography and sonography are also frequently used for the evaluation of breast lesions. However, studies have shown MRI to be superior to mammography, sonography, and clinical examination for revealing tumor extent in the breast [6-9]. Furthermore, contrast-enhanced MRI has been found to more accurately reflect the presence and size of residual disease after chemotherapy [10-17].
Most investigations of neoadjuvant treatment response have quantified tumor extent or size by diameter measurement in one or two dimensions, and several have shown no statistical difference between response assessments by unidimensional and bidimensional measurements [18-20]. Likewise, the response assessment criteria recently set forth by the international working group for Response Evaluation Criteria in Solid Tumors (RECIST) are based solely on unidimensional measurements of a tumor's longest diameter [21]. Alternatively, it has been suggested that 3D volume characterization of a tumor may be more advantageous for evaluation of change in lesion size, especially in cases of irregular tumor morphology or multifocal disease. Recent investigations using volumetric changes in tumors to monitor treatment response have reported varying sensitivities [22-25]. We hypothesize that measurements of tumor volume, calculated by automated segmentation of MRI images, characterize lesion extent more accurately than diameter measures, and may thus prove more sensitive to tumor size changes in response to treatment. Therefore, the purpose of this study was to assess the value of MRI measurements of breast tumor volume and diameter before, during, and after neoadjuvant chemotherapy treatment for predicting disease recurrence in comparison with established prognostic factors from clinical examination and pathology.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
MRI Acquisition
Imaging was performed on a 1.5-T Signa scanner (GE Healthcare) using a
bilateral phased-array breast coil (Open Breast Coil [OBC], MRI Devices).
Technical requirements of our contrast-enhanced imaging sequence for maximum
sensitivity in disease detection included full breast coverage with no gap
between slices, fat suppression, and good spatial resolution. A 3D fast
gradient-recalled echo sequence
[6] was used (TR/TE, 8/4.2;
flip angle, 20°; 2 repetitions [oversampling to remove phase wrap]). The
field of view was typically 18-20 cm, with 2-mm slice thickness and 256
x 192 acquisition matrix. The resulting in-plane resolution was
approximately 0.7 x 0.94 mm, and 60 slices were acquired in the sagittal
orientation to cover the entire symptomatic breast. The contrast agent used
was gadopentetate dimeglumine (Magnevist, Schering) at a dose of 0.1 mmol/kg
body weight. Contrast injection was followed by a 10-mL saline flush.
Fat suppression was an important imaging requirement for discriminating enhancing lesions from bright fat signal on T1-weighted images and was achieved using a frequency-selective inversion recovery preparatory pulse to eliminate the fat signal before image acquisition [14]. The total scanning time was 5 min, and the low order phase-encoding data were acquired at the center of the scan, resulting in an effective time point of 2.5 min from the start of the scan. Three time points were acquired during each MRI examination: a baseline scan before contrast agent injection, followed by two sequential scans after contrast agent injection, yielding a temporal sampling of 0, 2.5, and 7.5 min.
MRI Postprocessing
Contrast enhancement in the breast was calculated on a voxel-by-voxel basis
and was used to identify tumor extent. To accurately segment tumors from
breast tissue, a threshold was applied of at least 70% enhancement from
baseline at 2.5 min after injection of a contrast agent. To account for
dampening of contrast agent uptake response after chemotherapy, the criteria
for determining malignancy on posttreatment MRI scans were relaxed to include
all tissue with enhancement above baseline levels in the previous region of
tumor. We reported previously that reduction of the enhancement threshold
criterion improved sensitivity for the identification of residual disease in
the breast and did not result in overestimation of lesion size after
chemotherapy [14]. A reduced
enhancement threshold of 40% increase from baseline was found to better
capture tumor volume in lesions after completion of chemotherapy
treatment.
The enhancement criteria were applied to segment tumors in 3D MRI data sets using a semiautomated software algorithm developed inhouse [26] using the Interactive Data Language (IDL, Research Systems) programming environment. The software algorithm first creates a set of maximum intensity projections from the images at each time point. The projections are made in the lateromedial, craniocaudal, and anteroposterior directions and allow the tumor extent to be quickly identified in all dimensions. Next, the operator defines a restricted volume to be evaluated by drawing a box around the lesion of interest on two orthogonal projections. In some cases it may be necessary for the user to omit from the volume calculations large enhancing blood vessels or other interfering regions adjacent to the tumor that are located within the restricted volume. In these cases, the operator also encircles on the projections the structure or region to be omitted. The enhancement threshold is then applied on a voxel-by-voxel basis within the restricted volume, and excluding any manually omitted regions, to automatically segment the tumors in the 3D MRI data sets and selected voxels are summed to calculate tumor volumes. The automated calculations improved the efficiency and reproducibility of serial volumetric measurements. A tumor longest diameter was also measured manually from maximum-intensity-projection images.
The MRI parameters included in the analysis were initial tumor volume and diameter measured before treatment, early change in tumor volume and diameter after one cycle of chemotherapy, and the final change in tumor volume and diameter after completion of all chemotherapy treatment before surgery.
Clinical Variables
Clinical response assessment was based on change in a tumor's longest
diameter, estimated by palpation at physical examination. Clinical tumor
diameter measurements obtained before treatment (initial) and after completion
of all chemotherapy (final) were included in the analysis. Clinical response
was calculated for each patient as the percent change from initial tumor
diameter after all chemotherapy treatment.
Patient age at the beginning of treatment was recorded, and tumor grade determined by diagnostic biopsy when available. The diameter of residual disease in the breast and number of positive lymph nodes were determined from pathology reports after surgery. In 15 cases where the extent of disease for which was not given in the pathology report, the cases were rereviewed by a pathologist and surgeon to determine pathologic size. Recurrence-free survival (RFS) was assessed for each patient based on clinical examination and mammographic imaging at 6-month or 1-year intervals after surgery. The length of RFS was defined as the time between the primary surgery and local or distant recurrence, or the time to last follow-up in patients with no evidence of recurrence.
Statistical Analysis
The predictor variables evaluated were MRI measures of tumor volume and
tumor diameter before treatment, after one cycle of chemotherapy, and at
completion of chemotherapy measures of tumor diameter by clinical examination
before and at completion of treatment, and pathology measures of residual
disease diameter and number of involved lymph nodes. The outcome variable used
for analysis was length of RFS, and all observations of patients without
disease recurrence were censored. The median and range of values are reported
to summarize the distribution of each variable in the study population.
Univariate analysis based on a log-rank test of Cox proportional hazards regression [27] was used to identify variables associated with RFS. A hazard ratio (HR) and p value are reported for each variable in comparison with length of RFS. Factors found to be significant with the univariate test were then entered in a stepwise multivariate analysis to identify significant independent predictors. A p value of 0.05 was considered statistically significant for all analyses.
HR values indicate the increased hazard, or risk of recurrence, for each incremental increase in the particular variable. In the literature, predictive variables have typically been categorized, thereby producing large HR values per step increase from one category to the next. Rather than imposing artificial cutoffs, we retained the continuous values for statistical analysis. For example, tumor diameter and volume were expressed in cm and cm3, respectively, rather than classifying them into general response categories. Therefore, the HR for these variables indicate the incremental hazard corresponding to a 1-cm3 increase in tumor volume or 1-cm increase in tumor diameter, respectively. However, cutoffs were used to produce Kaplan-Meier survival curves for illustrative and comparison purposes. Estimates of 2-year RFS were determined from survival curves and differences between groups were evaluated by the Wilcoxon's chisquare test.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
The median initial tumor size on MRI was 5.2 cm (range, 1.1-11.4 cm) in diameter before neoadjuvant chemotherapy. At the time of surgery, seven of 58 patients had pathologic complete response. The diameter of residual disease at pathology in the remaining patients was a median of 2.8 cm (range, 0.3-15 cm), and 36 of 58 patients had lymph node involvement on pathology.
MRI was used to characterize increases and decreases in tumor volume during neoadjuvant treatment (Figs. 2A, 2B, 2C, 3A, 3B, and 3C). One patient with a very focal mass displayed a notable reduction in tumor volume in response to treatment and continued to be disease-free 20 months after treatment (Figs. 2A, 2B, and 2C). In contrast, a second patient with a more diffuse lesion showed an increase in tumor volume on MRI during treatment and experienced disease recurrence only 8 months after surgery (Figs. 3A, 3B, and 3C). No change in tumor size was observed on clinical examination for this patient.
|
|
|
|
|
|
|
The significant variables were then combined into a forward stepwise multivariate Cox regression analysis. The resulting model for predicting RFS consisted of only two significant independent parameters: initial MRI tumor volume and final change in MRI tumor volume (Table 2).
|
To illustrate the association of initial tumor volume with RFS, we used the median initial tumor volume of 33 cm3 as a cutoff for survival curves (Fig. 4A). The group with initial MRI tumor volumes less than 33 cm3 showed significantly longer RFS than those with tumors larger than 33 cm3. Initial tumor diameter on MRI was a close alternative to initial MRI volume in the Cox model. Initial clinical diameter, pathologic tumor size, and lymph node involvement also provided less significant alternatives (Table 2).
|
|
The combination of initial MRI tumor volume and final change in MRI volume was most predictive of RFS. Longer RFS was observed in the group of patients who had both small initial tumor volume and at least partial volumetric response to chemotherapy (Fig. 4C).
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Initial tumor size has long been known to predict patient survival [28-32] and is the basis of most disease-staging systems [21, 33]. Indeed, the most significant predictive variable in our study was initial tumor volume (p = 0.002). This strong correlation between initial tumor volume and survival validates the need for improved screening and detection methods. However, it would be valuable to identify other prognostic factors that can potentially be controlled through treatment to alter outcome. Change in tumor volume is one such factor and was identified through multivariate analysis in our study to be the most significant treatment-controlled predictor of RFS (p = 0.003). Clinical assessment of tumor regression by palpation is the most common method of determining response during treatment. However, clinical examinations are inherently subjective and lack precision, and did not adequately predict treatment outcome in this study (p = 0.27). Eleven of the patients were judged to have complete response to chemotherapy by clinical examination (no residual disease remaining), but in fact, eight of these patients were found to have residual disease on pathology. In contrast, no residual cancers were missed on MRI analysis, producing no false-negatives. However, the remaining tumor volume for one patient was significantly underestimated by the semiautomated algorithm, due to a very low level of contrast enhancement in the tumor after treatment. The reason for the reduced enhancement in this case was not known, but at least one study has reported similar findings of suppressed enhancement on postchemotherapy MRI [34]. Nevertheless, in our study, this problem was observed only in this single case, and we have previously found MRI to be accurate for identification of residual disease after neoadjuvant chemotherapy [14].
Pathologic response is also known to be predictive of patient survival, as previously reported [3, 4], but is strictly an endpoint measure and cannot be used to improve treatment planning. Our results showed that although pathologic tumor size and lymph node involvement were significantly associated with RFS in univariate analysis, MRI tumor volume measures (initial and final change) were more strongly correlated with RFS in the multivariate analysis. This may be explained by the fact that the two pathologic predictors were in fact linearly correlated with each other (r = 0.57, p < 0.0001) and with initial MRI tumor volume (r = 0.5, p = 0.0001 for pathologic tumor size; r = 0.42, p = 0.001 for number of positive lymph nodes), and therefore provided only redundant and less predictive information when combined with initial MRI tumor volume in the model. On the other hand, initial tumor volume and the final change in MRI tumor volume were not correlated (p = 0.93) and were both independently predictive of treatment outcome.
Volume Versus Diameter Measures
In clinical practice, lesion size is typically characterized by longest
diameter on palpation, imaging, and pathology assessments. We have previously
shown that MRI measures of residual tumor diameter agree well with those on
pathology [14]. Moreover,
current RECIST response assessment criteria recommendations are based on
changes measured in a tumor's longest diameter. However, our study indicates
that one-dimensional characterization of tumor response may be less sensitive
than 3D volumetric measurements. In the multivariate model, initial MRI
diameter provided a close alternative to initial MRI volume, indicating that
the two measures may be substituted in the model with little loss of
predictive value. However, the final change in MRI diameter after chemotherapy
was not significantly predictive of RFS in this study. In contrast, the final
change in MRI volume was a significant independent predictor of RFS in the
final model. This may indicate that 3D volume measurements more accurately
capture the extent of irregularly shaped tumors, multifocality, and diffuse
shrinkage of lesions during treatment.
Early MRI Tumor Changes
Unlike pathologic response, early changes in tumor volume measured by MRI
can be assessed at a stage when the treatment plan can still be modified.
Trends of association between early changes in MRI tumor measures (volume and
diameter) and RFS were observed after one cycle of chemotherapy, although
neither was statistically significant in this study. One factor to consider
was the reduced sample size for the analysis of early measurements (n
= 32) resulting from the exclusion of patients who did not receive MRI scans
after one cycle of treatment or who received taxane treatment. In addition,
this group contained only four patients with disease recurrence.
Study Limitations and Future Directions
A limitation of this study is the relatively short follow-up for the group.
The median follow-up time was 33 months in patients who were disease-free
(n = 45). We expect the data to continue to evolve as follow-up times
increase and the time to recurrence is determined for more of the patients in
the study.
MRI monitoring provided a sensitive evaluation of change in tumor size associated with treatment response and was more significantly predictive of disease recurrence than clinical or pathologic assessments. Although final change in MRI tumor volume was a stronger predictor of RFS in this study, the results of early MRI volume change after only one cycle of chemotherapy were encouraging. Early MRI volume change was significantly correlated with final volume change and may thus provide valuable response information at a time early enough to intervene. A multicenter trial is under way to further assess the predictive value of early changes in tumor volume and changes in tumor vascularity as measured by MRI (American College of Radiology Imaging Network (ACRIN) protocol 6657: Contrast-Enhanced Breast MRI for Evaluation of Patients Undergoing Neoadjuvant Treatment for Locally-Advanced Breast Cancer) [35].
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  K.-L. Li, S. C. Partridge, B. N. Joe, J. E. Gibbs, Y. Lu, L. J. Esserman, and N. M. Hylton Invasive Breast Cancer: Predicting Disease Recurrence by Using High-Spatial-Resolution Signal Enhancement Ratio Imaging Radiology, July 1, 2008; 248(1): 79 - 87. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Hattangadi, C. Park, J. Rembert, C. Klifa, J. Hwang, J. Gibbs, and N. Hylton Breast Stromal Enhancement on MRI Is Associated with Response to Neoadjuvant Chemotherapy Am. J. Roentgenol., June 1, 2008; 190(6): 1630 - 1636. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. C. Wolff, D. Berry, L. A. Carey, M. Colleoni, M. Dowsett, M. Ellis, J. E. Garber, D. Mankoff, S. Paik, L. Pusztai, et al. Research Issues Affecting Preoperative Systemic Therapy for Operable Breast Cancer J. Clin. Oncol., February 10, 2008; 26(5): 806 - 813. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. J. Solin, S. G. Orel, W.-T. Hwang, E. E. Harris, and M. D. Schnall Relationship of Breast Magnetic Resonance Imaging to Outcome After Breast-Conservation Treatment With Radiation for Women With Early-Stage Invasive Breast Carcinoma or Ductal Carcinoma in Situ J. Clin. Oncol., January 20, 2008; 26(3): 386 - 391. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. Sachelarie, M. L. Grossbard, M. Chadha, S. Feldman, M. Ghesani, and R. H. Blum Primary Systemic Therapy of Breast Cancer Oncologist, June 1, 2006; 11(6): 574 - 589. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Vol = volume decrease. Patients divided based on
initial MRI volume of their tumors showed significant differences in RFS
(p = 0.042, Wilcoxon's test). The 2-year RFS rate was 93% for
patients with smaller tumor volumes of 33 cm3 or less (n =
30) compared with 70% for those with larger tumors (n = 28).
