| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiation Medicine, Division of Radiation Oncology, Arthur G.
James Hospital and Solove Research Institute, The Ohio State University,
College of Medicine, 300 W 10th Ave., Rm. 080, Columbus, OH 43210.
2 Department of Radiology, The Ohio State University, Columbus, OH.
3 Magnetic Resonance Imaging Center, Department of Radiology, The University of
Iowa, Iowa City, IA.
4 Department of Radiological Sciences, Oklahoma University Health Sciences
Center, Oklahoma City, OK.
5 Department of Radiation Oncology, M. D. Anderson Cancer Center, Orlando,
FL.
6 Department of Radiation Oncology, Baylor College of Medicine, Houston,
TX.
7 Department of Radiation Oncology, The University of Iowa College of Medicine,
Iowa City, IA.
8 Department of Obstetrics and Gynecology, University of Connecticut, Hartford,
CT.
Received January 9, 2005;
accepted after revision April 27, 2005.
Supported by NIH RO1 CA 71906.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Serial MRI examinations (T1-weighted and T2-weighted images) were performed in 60 patients (age range, 29-75 years; mean, 53.3 years) with advanced cervical cancer (stages IB2-IVB/recurrent) who were treated with RT at four time points: start of RT, during RT (at 2-2.5 and at 4-5 weeks of RT), and post-RT. Tumor configuration was classified qualitatively into oval, lobulated, and complex based on MR film review. Two methods of tumor volume measurement were compared: ellipsoid computation of three orthogonal diameters (diameter based) and ROI volumetry by delineating the entire tumor volume on the MR workstation (ROI based). Temporal changes of tumor shape and the respective tumor volumes measured by the two methods were analyzed using linear regression analysis.
RESULTS. Most tumors (70%) had a non-oval (lobulated and complex) shape before RT and became increasingly irregular during and after RT: 84% at 2-2.5 weeks of RT (p = 0.037), 86% (p = 0.025) at 4-5 weeks, and 96% post-RT (p = 0.010), compared with 70% pre-RT. Diameter-based and ROI-based measurement correlated well before RT (r = 0.89) but not during RT (r = 0.68 at 2-2.5 weeks, r = 0.67 at 4-5 weeks of RT).
CONCLUSION. Most cervical cancers are not oval in shape pretherapy, and they become increasingly irregular during and after therapy because of nonconcentric tumor shrinkage. ROI-based volumetry, which can optimally measure irregular volumes, may provide better response assessment during treatment than diameter-based measurement.
Keywords: cervical cancer • response configuration • tumor configuration • tumor response • tumor shape • tumor size measurement • tumor volume
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
For clinical practice and clinical trials, the diameter-based tumor size measurement has been the standard method to assess the initial tumor volume and tumor regression [1, 16]. This method is used under the assumption that the configuration of the tumor closely approximates an ellipsoid (oval or round) shape [17]. Tumor volume is typically estimated with one maximal diameter of the tumor, as outlined in the Response Evaluation Criteria in Solid Tumors (RECIST) [15], the most widely used tumor measurement standard, or as two [16] or three orthogonal tumor diameters, derived from imaging studies or by clinical palpation. With the advancement and availability of commercially available computer software for quantitative analysis in cross-sectional imaging, including MR and CT scanners, the entire tumor, regardless of its shape, can now be identified and traced as a region of interest (ROI) on each imaging slice, and the 3D ROI-based quantitative measurement of tumor volume can now be more readily performed in clinical and community settings [3, 4, 18].
Despite the increasing importance of imaging-based tumor volume and tumor regression in cancer care, the optimal method for adequate volume measurement for specific tumors and specific therapy regimens has not been defined and remains controversial [1, 2, 17, 19-21]. The optimal method likely depends on the overall configuration of the tumor before therapy and the configuration changes it undergoes during and after therapy.
Advanced cervical cancer is an excellent model to study this question because it is typically not amenable to surgical resection, therefore treated with cytotoxic therapy [22-25], and shows measurable tumor regression during and after therapy [3, 14, 26]. These sequential tumor volumes and serial tumor configurations can be assessed noninvasively by MRI [27-29]. MRI is an excellent imaging technique for the 3D delineation of cervical cancer [27-29]. In addition, tumor volume and tumor regression rate have prognostic significance for clinical management in cervical cancer [3, 14, 22-25].
The purpose of this study was to evaluate the patterns and distribution of tumor shape before radiation therapy (RT) in cervical cancer, the temporal changes in tumor configuration coinciding with tumor regression during and after RT, and the effect of the tumor configuration changes on the correlation of tumor volume measurement between the diameter-based and the ROI volumetry method.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Therapy consisted of a combination of external beam RT with 24 MeV photons delivering a pelvic dose ranging from 39.6 to 70 Gy including field reductions. External beam radiation was combined with a standard low-dose rate of brachytherapy. No changes in the patients' therapy were made based on the findings of the MRI examinations.
MRI Protocol
The imaging protocol included serial MRI studies at four well-defined time
points: at the beginning of the RT, early during RT (at an RT dose of 20-25 Gy
at 2-2.5 weeks of RT), midway during RT (at 45-50 Gy at 4-5 weeks of RT), and
at the follow-up visit (1-2 months after the completion of all therapy). Four
of the 60 patients did not have the early RT MRI at 2-2.5 weeks, and four
patients did not have the post-RT follow-up MRI because of intervening
morbidity, refusal, or scheduling conflicts. Because all patients with fewer
than four MRI studies had three studies, it was decided that the information
gained from the inclusion of these patients outweighed the disadvantage of not
having completely uniform data in all cases. This resulted in a total of 232
MRI studies in the 60 patients.
The MRI examinations were obtained using a standard body coil with a 1.5-T superconductive scanner including Signa (GE Healthcare) and Vision (Siemens Medical Solutions) scanners. The change in platforms was related to a change in imaging systems at our institution. Imaging included sagittal 5-mm (4-mm thickness with 1-mm gap) conventional fast spin-echo T2-weighted images (TEeff/TR, 104/4,000; echo-train length, 10; number of excitations [NEX], 2) and axial 7-mm (5-mm thickness with 2-mm gap) T2-weighted and T1-weighted images (TE/TR, 16/600; NEX, 2).
Qualitative analysis of tumor configuration and its temporal changes—In this retrospective review, the hard copy films of the serial MR studies were qualitatively evaluated by three reviewers, two MRI radiologists (14 and 10 years of experience, respectively) and one radiation oncologist (10 years of experience in MRI for imaging-based radiation therapy planning and MRI). The review occurred over approximately 5 months in multiple sessions.
The tumor was defined as an abnormal area with intermediate to high signal intensity on T2-weighted images with respect to the surrounding cervical stroma and uterus, and lower than the fluid signal in the urinary bladder [13, 27, 29] (Figs. 1A, 1B, and 1C). The reviewers reviewed all serial MR studies of each of the 60 patients and qualitatively classified the tumor configuration into three categories: oval, lobulated, and complex shape (Figs. 1A, 1B, and 1C). The oval category was defined as a smooth configuration with a broad well-defined border without lobulations, closely approximating a round or oval shape. The lobulated category was defined as a smooth lobulated (single or multiple) well-defined border and without infiltrating strands. The complex category was defined as an irregular configuration with infiltrating borders or strands extending into surrounding healthy tissue. Discrepancies among reviewers were resolved by consensus or majority of opinions. The results of the qualitative evaluation of tumor shape at the four different time points were tabulated and analyzed for the distribution pattern of the three tumor configuration categories and the temporal changes of tumor shape before, during, and after RT.
|
|
|
/ 6). For the
ROI-based 3D volumetry, the tumor area or ROI on each imaging slice on the
sagittal T2-weighted image of each MR study was first delineated, and the 3D
ROI-based volumes were then calculated by the summation of all tumor areas and
multiplication by the slice profile. Statistical analysis—The correlation between the serial tumor volumes derived from the diameter-based and ROI-based measurement methods at each of the four measurement time points was performed for the 60 patients using linear regression analysis. The correlation coefficients between the two measurement methods were analyzed with respect to the distribution of the tumor configurations in each of the four measurement time points (before, early during, mid-way during, and after RT).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
When each of the three morphologic patterns was examined individually, the lobulated configuration increased in number early during the course of RT (at 2-2.5 weeks) and later (at 4-5 weeks) declined in number in favor of the complex configuration, indicating that most tumors became increasingly irregular in shape during and after therapy. The complex configuration increased in number rapidly and persistently during and after RT (to 43% at 2-2.5 weeks, 56% at 4-5 weeks, and 64% post-RT) (Fig. 2). The oval shape continuously declined in number during and after RT in favor of the more irregular configurations.
Quantitative Analysis of Volume Measurement Methods at Four Time Points
Figures 4A,
4B,
4C,
4D, and
5 summarize the tumor volumes
derived from diameter-based and ROI-based measurements at the four time
points. The tumor volume showed a trend to decrease with time
(Fig. 5). The volume derived
by the diameter-based method was overall larger than that by the ROI-based
method. Compared with the ROI-based method, the median tumor volume measured
by the diameter-based method exceeded the ROI-based volume by 28% (69
cm3 vs 54 cm3) pre-RT, by 61% (50 cm3 vs 31
cm3) at 2-2.5 weeks of RT, and by 29% (9 cm3 vs 7
cm3) at 4-5 weeks of RT. This difference was greatest at 2-2.5
weeks during RT and decreased at 4-5 weeks and post-RT, when the overall tumor
volume rapidly declined.
|
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
However, the ideal method for the measurement of tumor volume and tumor shrinkage remains controversial [1, 2, 17, 19-21]. Different measurement techniques have been used in many tumors, including cervical cancer. For imaging-based tumor volume assessment, both the initial tumor volume and the regression rate of tumors have been traditionally estimated from measurements of one, two, or three diameters of the tumor. These reference standard diameter-based tumor measurements have been reported to be equivalent to the more complex 3D ROI volumetry that delineates the entire tumor region three dimensionally [7, 13, 21, 30]. However, this concept has been challenged by studies showing discordance between diameter-based and 3D volumetry measurements [17, 20, 31]. Others have reported that the 3D ROI-based tumor volume regression measurement is superior as a predictor of tumor control or survival [3, 4, 18].
The advanced cervical cancer chosen for this project is an excellent model to study the pattern of initial tumor shape and its temporal configuration changes with MRI for a number of reasons: First, advanced cervical cancer is typically not amenable to surgical resection, and therefore tumor shape can be sequentially studied during and after RT. Second, cervical cancer shows measurable tumor regression during or after RT on MRI, ideal for assessing temporal changes in tumor shape induced by cytotoxic therapy [3, 4, 13, 32]. Third, initial tumor volume and tumor regression rate are clinically significant prognostic and predictive factors in cervical cancer [3, 14, 22-25]. Fourth, there are four well-defined time points for MRI based on prognostic significance and clinical management [3, 13, 14, 26]. Fifth, MRI is an excellent imaging technique for the delineation of cervical cancer [27-29]. Finally, MRI is noninvasive and easily available in the common clinical setting.
Although the diameter-based tumor volume measurement has been formally adopted by the World Health Organization (WHO) criteria [16] and RECIST guidelines [15] as an end point for the assessment of therapy response in cancer, its value and interpretation have to be considered carefully in the context of individual tumors. RECIST, the most widely used current guideline, relies on a one-dimensional measurement using the longest tumor diameter.
The estimation of tumor volume with the diameter-based method relies on the assumption that the configuration of the tumor is ellipsoid (oval or round) and remains ellipsoid throughout and after therapy. It is thus assumed that the shrinkage during and after therapy occurs in a concentric fashion along all dimensions of the tumor.
The ideal measurement method likely depends on the overall configuration and geometric properties of the tumor before, during, and after therapy. However, to our knowledge, very little information is available regarding the shape of tumors at pretherapy baseline and temporal changes in tumor configuration induced by the cytotoxic effects during and after therapy. Furthermore, it is unknown how these factors influence the accuracy of a specific method for measurement of tumor volume and tumor regression for therapy response assessment. Our study was therefore intended to address these issues.
Our results show that, contrary to the general perception, most of the cervical cancers (70%) did not have a shape closely approximating the ellipsoid (oval or round) configuration before treatment. Furthermore, during treatment the tumor shape became increasingly irregular, and the likelihood of the tumor to maintain an oval shape declined rapidly during and after therapy (Fig. 2). As the tumors gradually underwent radiation-induced cell killing and volume reduction, the tumor volume appeared to shrink in an irregular nonconcentric fashion. This explains why the tumor shape deteriorated from the initial compact well-circumscribed shapes to irregular fragmented configurations (Figs. 2, 3A, 3B, and 3C). Because only the maximal diameter in three orthogonal planes is measured, tumor measurements derived from the diameter-based method are expected, as shown in our results, to be larger than those measured by the ROI-based volumetry method.
As a result of the increasingly irregular tumor shape, the correlation between diameter-based and ROI-based measurements worsened as the tumor configurations further deviated from the ideal oval or ellipsoid shape (Figs. 4B and 4C). This observation was particularly true during treatment, when the tumor size became smaller and simultaneously configuration became more irregular (i.e., when the degree of deviation from the ideal ellipsoid shape increased as the ratio of irregularity per unit of tumor volume increased). This degree of deviation from ellipsoid explains why, at the second and third measurement time points during RT (at 2-2.5 weeks and at 4-5 weeks), the correlation of diameter-based with ROI-based volume was poorest (r = 0.68 and r = 0.67, respectively). Unfortunately, these early tumor response measurements during therapy, which can provide most critical predictive information for therapy response, prognosis, and early decision making, are those most adversely affected by the compromised accuracy of diameter-based tumor measurement.
Although only 30% of the tumors had an oval shape before treatment, the pretherapy time point had the best correlation between the diameter- and ROI-based measurements. We believe the reason for the better correlation is related to the relatively larger tumor size with relatively smoother border and therefore less complexity of shape per unit of tumor volume as compared with the configurations during therapy. The overall concept that temporal tumor geometry changes influence the accuracy of response measurement, and our observations, are supported by a recent study by Mazumdar et al. [17]. This theoretic simulation study showed that simulated sequential variations in tumor shape result in statistically significant variability of the response assessment when diameter-based measurements are used.
At the posttherapy measurement time point, the overall discrepancy between the diameter-based and ROI-based methods was improved despite the predominance of irregular shapes. This is likely explained by the fact that most tumor volumes are very small and close to zero at that time, and discrepancies of volume between the measurement methods are therefore lower in magnitude.
Although the ROI-based quantitative volume measurement is a more involved and time-consuming method than the diameter-based measurement, and its practical application in a busy clinical setting has been a challenge, this technology is now becoming more readily available and user-friendly for general practice and ideally will become near-completely automated in the future. Regardless of the type or complexity in tumor shape, the ROI-based volumetry measurement includes all tumor components identified on all images throughout the lesion. Therefore, tumor-specific or therapy-induced deviations from the ideal ellipsoid shape do not compromise its accuracy.
Our study has several limitations. The patient population was accrued over a relatively long time span in cervical cancer patients, a relatively uncommon cancer. At the time of the study, new pulse sequences with improved spatial resolution and volume acquisition techniques as well as phase array coils were not available. If this study were repeated today, the results might be different because of the improved lesion delineation provided by phase array coils and newer advanced imaging techniques.
The review was a simultaneous one by three observers with discrepancies resolved in consensus, and the MR studies of each patient were reviewed consecutively. Thus, we do not have interobserver variability data for this study, and this subject would be important to evaluate in future research. The studies for each patient were reviewed consecutively because this is similar to the actual clinical situation in which the prior MR examination is usually available on review of a subsequent study. Although this consecutive review is similar to the typical clinical situation, the fact that the studies were not randomly distributed, without knowledge of which serial MRI examination belonged to which patient, may represent a limitation of our study.
Our study does not have histologic validation of the imaging findings. This is a challenge we are facing in the sequential imaging of cervical cancer and in many other unresectable cancers that are imaged serially while undergoing cytotoxic therapy. Tumors, which are amenable to serial imaging, are typically not resectable; therefore, no tumor specimen is available for precise imaging-pathology correlation. Conversely, tumors, which can be resected and are thus amenable to imaging-pathology correlation, cannot be imaged serially because the tumor is resected.
Despite this dilemma, a European study by Burghardt et al. [27] provides strong evidence of a high correlation between MRI- and pathology-based tumor volumetry in cervical cancer (stages I and IIB tumors, which are all treated surgically in Europe). Although this surgical-pathologic validation of pretherapy MRI with histologic tumor volume is strong [27], no absolute pathologic proof indicates this is also the case for tumor margin delineation and tumor volume assessment during therapy. However, a recent biopsy study by Hatano et al. [13] in patients with posttherapy MRI has shown excellent correlation between areas of increased signal intensity and residual tumor and between areas of hypointensity and absence of tumor.
For cancer care, ultimately more important than pathology correlation will be the clinical validity of the MRI-based measurement methodology for patient outcome. The methodology of measurement that is most sensitive in assessing response, predicting outcome, or both, and is the most effective in providing a window of opportunity to adjust treatment regimens in those with poor response to optimize care, will be the most useful.
In conclusion, our limited data suggest that most cervical cancers are not oval or ellipsoid in shape before, during, and after RT. Diameter-based tumor measurement therefore may not be the most ideal method to provide the critical information needed for prognosis and response assessment in cancer management and in the development of new cancer therapeutics. The measurement of the initial tumor volume and its shrinkage must account for the deviation from the presumed ellipsoid configuration before treatment, and particularly for the increasing irregularity induced by therapy. The ROI-based method has the advantage of not being limited to oval-shaped tumors and can compensate for therapy-induced irregularities in tumor configuration, providing superior volumetric measurement.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. Movva, M. Gold, P. Grigsby, and C. Verschraegen Challenges in the Management of Invasive Cervical Cancer ASCO Educational Book, January 1, 2009; 2009(1): 295 - 300. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. J. Bissler, F. X. McCormack, L. R. Young, J. M. Elwing, G. Chuck, J. M. Leonard, V. J. Schmithorst, T. Laor, A. S. Brody, J. Bean, et al. Sirolimus for Angiomyolipoma in Tuberous Sclerosis Complex or Lymphangioleiomyomatosis N. Engl. J. Med., January 10, 2008; 358(2): 140 - 151. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
