| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Hepatobiliary Imaging |
1 Department of Medical Imaging, National Taiwan University Hospital and
Department of Radiology, National Taiwan University College of Medicine, No.
7, Chung-Shan S Rd., Taipei 100, Taiwan.
2 Division of Cancer Research, National Health Research Institute, Taipei,
Taiwan.
3 Department of Medicine, Kaohsiung Medical University Hospital, Kaohsiung,
Taiwan.
Received July 31, 2003; accepted after revision April 20, 2004.
Address correspondence to T. T.-F. Shih
(ttfshih{at}ha.mc.ntu.edu.tw).
OBJECTIVE. To evaluate the perfusion changes in advanced hepatocellular carcinoma (HCC) treated with the antiangiogenic agent thalidomide, we used dynamic contrast-enhanced MRI.
SUBJECTS AND METHODS. Dynamic contrast-enhanced MRI was performed before and during thalidomide treatment in seven patients with advanced unresectable HCC that had failed to respond to prior local therapy. A turbo fast low-angle shot sequence was performed in a 1.5-T MR scanner. An operator-defined region of interest was placed in the maximal enhancement region of the tumor site and adjacent tumor-free parenchyma of all patients. A time–intensity curve was plotted and analyzed. The peak enhancement in the first-pass study, the maximal enhancement, and the initial enhancement slope percentage in the first-pass study of the tumor and parenchyma were measured. The changes in these three perfusion parameters were estimated and correlated with clinical outcomes. The seven patients were categorized into two groups on the basis of their clinical outcomes: group A patients were those who had progressive disease, whereas group B patients were those who had stable disease or partial response.
RESULTS. Four of the seven patients were classified as group A, and the other three were classified as group B patients. When comparing the MRI parameters for the tumors before and during treatment in group A and group B patients, we found a statistically significant difference for the peak enhancement in the first-pass study, the maximal enhancement, and the enhancement slope percentage in the first-pass study. When comparing the parenchymal parameters, we found a statistically significant difference in the maximal enhancement and borderline significance in the peak enhancement in the first-pass study (p = 0.057) between group A and group B patients.
CONCLUSION. The dynamic MRI parameters showed significant differences between two groups of patients with different clinical outcomes.
Hepatocellular carcinoma (HCC) is a common malignancy in Taiwan. HCC is usually a hypervascular tumor and is hyperintense on gadopentetate dimeglumine–enhanced arterial-dominant phase MRI and is isointense or hypointense relative to the liver parenchyma on delayed phase MRI because of its predominantly arterial blood supply [1–3]. Various treatment techniques have been used to treat HCC, including hepatic resection, transarterial chemoembolization, percutaneous ethanol injection, and radiofrequency interstitial thermal ablation [4–8]. Focal arterial infusion of chemotherapeutic agents can occasionally cause tumor regression in advanced HCC, but no established regimen has been confirmed to be efficacious [4, 9, 10].
Since 1999, the antiangiogenic agent thalidomide has been used in clinical trials as a treatment for advanced HCC in our country. Angiogenesis is the formation of new blood vessels from the existing vascular bed [11, 12]. Angiogenesis is essential for tumor growth, invasion, and metastasis, and it is hoped that the study of angiogenesis will help in the design of new treatment strategies for various kinds of malignancies [11–13]. Thalidomide was recently noted to exhibit antiangiogenic activity, with orally administered thalidomide inhibiting both basic fibroblast growth factor and vascular endothelial growth factor–induced angiogenesis in the rabbit cornea micropocket assay [14]. The inhibition of neoangiogenesis is, of course, of interest because it may delay or suppress tumor growth [11, 14, 15].
Extensive neoangiogenesis is a distinctive feature of neoplasms, and techniques used to assess neoangiogenesis can only detect those tumors in which the angiogenic process has been turned on [16–18]. In the evolution of angiogenesis, the early phases such as initiation and promotion cannot be detected on MRI, and dynamic contrast-enhanced MRI can only detect the later phases characterized by increased diffusion, perfusion, and new vessel formation [16]. Multiple reports regarding the evaluation of neovascularization and perfusion of normal tissues or various kinds of malignancies using dynamic contrast-enhanced MRI have been published [19–24]. The time–intensity curve first-pass study data derived from dynamic contrast-enhanced MRI or equilibrium methods from dynamic contrast-enhanced MRI have been used in many studies, either for depiction of tissue perfusion in various malignancies or the evaluation of treatment effect in various carcinomas [4, 21–24]. To date, there have been no published studies of MRI evaluation of angiogenesis and perfusion changes in HCC after antiangiogenic agent administration. The purpose of this study was to analyze the treatment effects of thalidomide in advanced HCC by monitoring perfusion changes in tumors on dynamic contrast-enhanced MRI.
Subjects and Methods
Patient Population
Between October 2001 and August 2002, seven patients with unresectable HCC
(four men and three women; age range, 26–77 years; median age, 67 years)
were prospectively studied. The diagnosis of HCC was based on elevated serum
-fetoprotein levels (
400 ng/mL; five patients), previous
biopsy-proven HCC (seven patients), and typical HCC imaging findings on CT or
MRI in cirrhotic livers (seven patients). All patients had multiple hepatic
tumors that had failed to respond to prior local therapy, including surgery,
transarterial chemoembolization, percutaneous ethanol injection, or a
combination of these therapies. According to our protocol, dynamic
contrast-enhanced MRI would be performed before and between 6 and 10 weeks
after the initiation of thalidomide treatment. All patients underwent
pretreatment MRI of the abdomen, after which 100 mg of oral thalidomide was
administered twice daily (200 mg/day). The use of thalidomide (Thado, TTY
Biopharm) had been approved for all of the patients by the ethics committee of
the hospital and our national Department of Health. Institutional review board
approval and informed consent were obtained for each patient who received MRI
examinations.
Pretreatment MRI of the seven patients was performed at a median of 8 days (range, 0–32 days) before oral thalidomide administration, and follow-up MRI was performed at a median of 55.4 days (range, 42–78 days) after the initiation of thalidomide treatment. The clinical follow-up period ranged from 18 to 19 months (mean, 18.2 months).
MR Images Acquisition
MRI was performed using a 1.5-T superconducting magnet (Magnetom Sonata,
Siemens Medical Solutions) with a phased-array body coil. The imaging protocol
was the same for the pretreatment and follow-up MRI of each patient and
included coronal fast imaging with steady progression (TR/TE, 6.3/3.0; flip
angle, 70°; matrix, 256 x 256; slice thickness/interslice gap, 6/2.4
mm; field of view, 350 x 350 mm), axial fat-suppressed turbo spin-echo
T2-weighted (TR/effective TE, 5,000/182; echo-train length, 29; matrix, 97
x 256; field of view, 247 x 330 mm; slice thickness/interslice
gap, 6/1.8), and axial fat-suppressed fast low-angle shot (FLASH) sequence
(TR/TE, 261/4.8; flip angle, 70°; matrix, 96 x 256; field of view,
247 x 330 mm; slice thickness/interslice gap, 6/1.8). The dynamic
contrast-enhanced MRI was aimed at a fixed slice where the largest tumor part
was located in each patient, with 2D T1-weighted turbo FLASH sequences
(301/1.2; flip angle, 8°; matrix, 125 x 256; field of view, 285
x 380 mm for coronal sections or 247 x 330 mm for axial sections;
slice thickness, 6.5 mm). The acquisition time for each slice was 0.6 sec.
Dynamic scanning started 5 sec after the initiation of contrast bolus
injection. A 0.15 mmol/kg bolus of gadodiamide (Omniscan, Nycomed) was rapidly
administered manually (at a rate of approximately 2.5 mL/sec) by one
investigator via a previously placed 22-gauge IV cannula in a dorsal hand
vein. Immediately afterward, a 20-mL saline flush was administered at the same
injection rate.
Three sets of dynamic turbo FLASH sequences were performed. A total of 30 dynamic images for each set were obtained in the same anatomic section. An interval of 10 sec each for breathing was placed between the first and second sets and the second and third sets. It took a total of 79 sec from the initiation of contrast injection to the completion of the whole dynamic MRI study. A delayed contrast-enhanced axial fat-suppressed FLASH sequence was performed 2 min after the completion of the dynamic study.
Data Analysis
Signal intensity (SI) values were measured in the dynamic pretreatment and
follow-up MRI studies of all seven patients in operator-defined regions of
interest encompassing the tumor site with maximal enhancement and adjacent
tumor-free liver parenchyma. The region of maximal tumor enhancement was
determined by measurements of at least four different regions of interest
within the largest tumor part of each patient, avoiding areas that had been
treated with transarterial chemoembolization Lipiodol (iodinated oil,
Amersham) retention. A time–intensity curve was plotted for each region
of interest (Figs. 1A,
1B, and
1C).
|
|
|
Three time–intensity curve patterns were defined: pattern I was a
rapid wash-in phase followed by a washout phase in the latter portion. Pattern
II was an initially rapid-rising slope followed by a second slow-rising phase.
Pattern III was a rapid wash-in followed by a plateau after peak enhancement
(Figs. 1A,
1B, and
1C). The baseline SI value
(SIbase) in a time–intensity curve was the mean SI
value in the first three images. The SI1st was the peak SI
value in the first-pass study of contrast enhancement. The
SImax was the maximal SI value in a time–intensity
curve. The contrast rise time (
T in seconds) was the time interval
between SIbase and SI1st. The peak
enhancement in the first-pass study was abbreviated as
E1st and was measured as
SI1st–SIbase. The maximal
enhancement (Emax) was measured as
SImax–SIbase. The enhancement
slope percentage in the first-pass study was defined as follows:
 |
In this study, we used three
parameters—E1st,
Emax, and slope percentage—for perfusion
evaluation. The differences in the pretreatment MRI parameter values minus
their corresponding follow-up MRI parameter values were measured and
abbreviated as "d" values.
The seven patients' pretreatment and follow-up MR images were compared with
respect to changes in tumor number, tumor size, SI on T2-weighted imaging and
delayed contrast-enhanced axial fat-suppressed FLASH sequences, and d
values of tumors and the adjacent tumor-free liver. The seven patients were
subsequently categorized into two groups on the basis of tumor response.
Progressive disease (group A) was defined as the appearance of new lesions or
an increase of more than 25% in the areas of any original lesions (including
an increase of solid component within a necrotic lesion) or a progressive
elevation of the serum -fetoprotein level. Patients whose followup
examinations did not meet any of these criteria were considered to have stable
disease or improvement (group B).
Statistical Analysis
The differences in the E1st,
Emax, and slope percentage on pretreatment and
follow-up MR images were tested with the Wilcoxon's signed ranked test and
Wilcoxon's rank sum test. The two tests are non-parametric tests suitable for
small sample sizes.
Results
Among the seven tumor time–intensity curves on pretreatment MR images, four were pattern I, and three were pattern II. Of the seven parenchymal time–intensity curves, six were pattern II, and one was pattern III. On follow-up MR images, no change in the time–intensity curve pattern was found in any patient for each tumor and parenchymal region of interest.
The corresponding laboratory data collected at the same period as the
pretreatment and follow-up MRI and clinical outcomes at the time of the
follow-up MRI are summarized in Table
1. Four patients were classified as group A and three as group B.
On followup MR images, no increase of the necrotic components in tumors was
found in any of the patients. Of the four patients in group A, one had no
morphologic change evident in any tumors (patient 1,
Fig. 2), whereas the other
three patients showed tumor size increase (patients 2, 3, and 4). One tumor
(patient 3) showed a definite increase in solid tumor although the tumor size
did not change (Figs. 3A and
3B). Of the three patients in
group B, two showed no change in tumor size (patients 5 and 6 and
Fig. 4), and one patient showed
a reduction in the size of one of the tumors associated with a greater than
95% fall in the level of serum -fetoprotein during treatment (patient
7).
|
|
|
|
|
The tumor time–intensity curve parameters on pretreatment and
follow-up MRI studies of the seven patients are summarized in
Table 2. The decrease in slope
percentage parameter after thalidomide treatment was of statistical
significance (p < 0.05), with the E1st
and Emax parameters revealing no definite
statistical difference (p > 0.05). The liver parenchyma
time–intensity curve parameters on pretreatment and follow-up MRI
examinations of the seven patients are listed in
Table 3. Considering all seven
patients as a whole, none of the three parenchymal parameters exhibited
significant difference during treatment
(Table 3).
|
|
On further analysis of the difference values from pretreatment MRI
parameters minus follow-up MRI parameters (d values) in tumors
between group A and B patients, the d values of
E1st, Emax, and slope
percentage in group A patients were lower than those in group B patients, with
all three parameters showing statistically significant difference
(Table 4). The data suggested a
more profound reduction of the three parameters in the tumors of group B
patients than those in the tumors of group A during treatment. When the
parenchymal parameters of group A and B patients were compared, the d
values of Emax in group A patients were
significantly lower than those of group B patients, the d values of
E1st were lower but with borderline significance
(p = 0.057), and the d values of slope percentage were lower
but without statistical significance (Table
4). The perfusion parameters of liver parenchyma of group B
patients were more profoundly reduced during treatment as compared with those
of group A patients.
|
Discussion
Tumor angiogenesis is induced by a shift in the balance maintained between angiogenic inducers and angiogenic inhibitors [14]. Thalidomide is known to inhibit angiogenesis induced by basic fibroblast growth factor, vascular endothelial growth factor, and endothelial cell activity directly [11, 14]; however, the exact mechanism by which thalidomide can inhibit angiogenesis in cancer therapy remains largely unclear [14]. Thalidomide is now commercially available and has been tested in phase II trials for treatment of multiple myeloma, head and neck cancer, Kaposi's sarcoma, renal cell carcinoma, hormone-refractory prostate cancer, breast cancer, ovarian cancer, and high-grade brain tumor and as adjuvant therapy in recurrent or metastatic colorectal cancer [11]. Thalidomide has been found to provide a partial clinical response in 5%, stable disease in 40%, and progressive disease in 55% of patients with far advanced, heavily pretreated HCC [15].
Many techniques exist for the evaluation of tumor angiogenesis [12, 16–18, 23, 25]. Evaluation of human tumor angiogenesis using contrast-enhanced CT [20] and dynamic contrast-enhanced MRI [13, 16–18, 21–24] is based on the observation that angiogenesis can increase the perfusion and permeability of tumors [12, 13, 16]. Many researchers have used dynamic contrast-enhanced MRI to determine tumor perfusion in breast, cervical, bladder, and prostate cancers and in bone sarcomas [20–24]. Boss et al. [21] and Mayr et al. [22, 24] analyzed the perfusion changes in cervical carcinoma after radiotherapy using parameters obtained from dynamic contrast-enhanced MRI time–intensity curves. Researchers have contended that the appearance of rapid enhancement of HCC during the arterial-dominant phase of dynamic contrast-enhanced MRI is evidence of residual or recurrent disease after percutaneous ethanol injection or transarterial chemoembolization [1, 2, 26, 27]. In our study, time–intensity curve parameters from dynamic contrast-enhanced MRI were used in an attempt to evaluate perfusion changes in HCC during thalidomide treatment.
In previous studies, time–intensity curve patterns were used to analyze the probable nature of breast, bone marrow, and musculoskeletal malignancies [16, 18, 20] and the correlation with cellular grades of HCC [3]. Yamashita et al. [3] in 1994 stated that the time–intensity curve pattern of "rapid wash-in and rapid signal intensity decrease" of hepatic tumors was most likely seen in poorly differentiated HCC, whereas the time–intensity curve pattern of "slight or minimal enhancement" was best seen in well-differentiated HCC. However, the time–intensity curve patterns in our study could not be assumed to reliably correspond to cellular grades of HCC for two reasons. First, all patients in our study had already-known advanced terminal HCC. Second, all the patients had undergone several rounds of transarterial chemoembolization before thalidomide treatment. Accordingly, the residual or recurrent tumors shown on MRI in this study may not reveal the original time-intensity curve patterns before transarterial chemoembolization.
Three parameters were used as a semi-quantitative analysis to assess
perfusion changes in this study. Of these three,
E1st and slope percentage were first-pass data and
could be considered as indicators for tissue microcirculation and permeability
to contrast material [23,
24]. The
E1st parameter may not be as direct an indicator as
the slope percentage parameter for evaluating tissue perfusion, since absolute
values of SI enhancement of tissues were used, and SI values differ greatly
with various pulse sequences. Nevertheless, E1st
parameter may have been used as a reference value if all MRI examinations were
performed with the same pulse sequence and machine setting, and the SI of
enhancement is related to the extent of vascularization and the wash-in and
wash-out properties [16]. The
slope percentage parameter has been used in assessment of tissue
vascularization and perfusion evaluation of bone marrow, musculoskeletal,
breast, and cervical malignancies
[18–20,
28], with accuracies ranging
from 71% to 97% [18].
In this study, we also analyzed Emax in addition
to the first-pass data. Advanced HCC is predominantly supplied by hepatic
arteries, whereas the hepatic parenchyma is mostly nourished by portal veins.
Maximal enhancement of HCC in some patients in this study was found to occur
during the portal phase, not the peak of the first-pass phase, and the maximal
enhancement of parenchyma in most patients appeared during the portal phase.
On the basis of these findings, measurement of Emax
was important for the perfusion assessment of the tumor and tumor-free
parenchyma.
When considering the paired parameters of the tumors of all seven patients
on pretreatment and follow-up MR images, we found that
E1st and Emax showed no
statistical difference on follow-up MR images when compared with those
parameters on pretreatment images and that during treatment only the slope
percentage in tumor decreased with statistical difference. The results failed
to correlate well with the disease progression in four of the seven patients.
However, for group A patients (clinical progression group), the corresponding
E1st, Emax, and slope
percentage in tumors tended to appear to increase on follow-up MR images or to
decrease to a lesser degree than those in group B patients (stable or improved
group). In examining the detailed parameter values of
E1st and Emax of all
patients in Tables 2 and
4, we found that both the
negative and positive d values existed and that they were scattered
in a wide range. The resultant offsetting of these data led to the conclusion
that no significant difference among the patients existed. Because most of the
negative d values occurred in group A and most of the positive
d values occurred in group B patients, significant differences
appeared when two groups were separated for comparison. In our study, we
supposed that a positive d value indicated a decrease in tumor
perfusion during treatment and subsequent clinical improvement or stability.
It could be said that changes in these parameters may parallel clinical
outcomes in the two groups of patients and that the hepatic tumor vasculature
of group B patients might be more susceptible to the angiogenesis inhibition
process than that of group A patients.
On the other hand, although tumor-free parenchyma showed relatively greater perfusion increases or lesser perfusion decreases in group A patients than in group B patients, the tumor part still revealed more statistically significant change. The antiangiogenic activity of thalidomide in two clinical groups was more evident in the tumor part than in the parenchyma.
Methodologies used to analyze perfusion on dynamic contrast-enhanced MRI vary considerably among the different series [13, 16–18, 21–24, 28, 29]. The quantitative technique, or pharmacokinetic method, was more complicated but may make direct comparisons of the pharmacokinetic parameters in a given patient or in different patients imaged at the same or at different imaging centers possible, if the method of contrast administration is kept constant [18, 23]. The semi-quantitative method used in our study has been used in many reported studies [19–21, 23, 28, 29]. Semiquantitative parameters were relatively straightforward to calculate and have been shown to have acceptable sensitivity, specificity, and accuracy in various reports [18–21, 23, 28]. The drawbacks are that these parameters did not accurately reflect contrast medium concentration in the tissue of interest and were subject to variations in MR scanner settings [23]. However, the slope percentage showed an almost significant correlation with the microvascular density (i.e., related with degree of vascularization and perfusion of tumor) rather than malignancy or benignity [28, 29]. Therefore, the semiquantitative analysis could also be applied to assess the perfusion changes of tumor after treatment. Despite its simplicity, the semiquantitative method used in this study has achieved satisfactory results.
Some researchers used 2D dynamic contrast-enhanced MRI to depict regions of interest [19, 20, 28], whereas others used 3D dynamic contrast-enhanced MRI for analysis of regions of interest [16–18, 22, 23, 29]. In our study, 2D dynamic contrast-enhanced MRI was used to depict regions of interest. Therefore, the dynamic images were obtained at only one level, which could have caused sampling error because not all tumors were included in the dynamic images [28]. However, we obtained unenhanced T1-weighted and T2-weighted imaging to cover the whole liver and chose an imaging plane that included most tumor components for dynamic studies to minimize the sampling error. The most evident merit of the 2D method was that the acquisition time of dynamic images could be as minimal as possible to obtain a more continuous time–intensity curve than that of 3D imaging. This difference is important in dynamic MRI of the liver because patients undergoing abdominal dynamic contrast-enhanced MRI should be breath-holding during dynamic scanning and at least two breathing intervals should be taken between dynamic sets. A more continuous time–intensity curve can minimize the possibility of missing the timing of SI1st and SImax.
In our study, we depicted regions of interest from one representative tumor and parenchymal parts for each patient. These regions might not reflect perfusion status of all tumors and hepatic tissues, because every patient in our study had multiple hepatic tumors. We minimized this problem by selecting the region of maximal enhancement determined by comparison from at least four measurements in each tumor from each selected dynamic slice, a technique that was also applied in the other studies reported in the literature regarding MRI of multiple hepatic metastases [30].
Controversy exists as to the region of interest in tumors, with some researchers placing the region of interest in the area with maximal enhancement [18, 21, 29]. Some authors make the region of interest encompass the whole tumor so that an average enhancement curve can be drawn [23, 24]. Neither method considers the heterogeneity of tumor blood supply [17, 22, 24]. Mayr et al. [22] analyzed the parameters characterizing the pixel-histogram distribution of the dynamic enhancement pattern of cervical carcinoma and found a wide range of dynamic enhancement values within the tumpor, supporting the concept of tumor heterogeneity [22]. They also concluded that quantification of the poorly vascularized regions was more important for the prediction of tumor control in cervical carcinoma treated with radiation therapy [22]. However, Boss et al. [21] made the region of interest at the site of maximal cervical cancer enhancement and concluded that onset of enhancement and time to peak enhancement could provide useful information for determining the effectiveness of radiotherapy treatment.
We placed the region of interest to encompass the site with maximal enhancement for three reasons: First, we hypothesized that thalidomide acted the most on the typical tumor regions, the hypervascular zone. Second, all of the patients had heterogeneous Lipiodol retention in the central necrotic portions of tumors due to previous transarterial chemoembolization, making selection of a region of interest in these sites difficult because the Lipiodol may undergo continuous phagocytosis and then show a different SI on serial MRI. Third, placement of region of interest at the site with maximal enhancement is well documented in some series [18, 21, 29]. After reviewing the various MRI techniques and angiogenesis data analysis, we have concluded that complete consensus is yet to be reached regarding the optimal way to define either the correct parameters for the evaluation of angiogenesis and tissue perfusion or the best MRI protocols [18].
In conclusion, our study shows that the changes in the peak enhancement in the tumor during the first-pass phase, during maximal enhancement, and in the enhancement slope percentage during the first pass on follow-up dynamic contrast-enhanced MRI parallel the clinical outcomes of the patients when they are divided into clinically progressive group and a stable disease or improved group. The MRI techniques and methods used in the study were noninvasive, easy to standardize, and reproducible before and during treatment and made attempts to assess the effect of antiangiogenic therapy seem promising. Naturally, study in larger patient groups is required to fully validate our method. The outlook of this research is to better understand and to improve the parameters for evaluating HCC treatment effects and to extend their use in the assessment of malignancy in other solid organs.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
