Magnetic resonance imaging reveals heterogeneous regions within high-grade gliomas, such as a contrast-enhanced rim, a necrotic core, and non–contrast-enhanced abnormalities. It is unclear which of these regions best describes tumor aggressiveness. We hypothesized that the vascular leakage volume, reflecting disorganized angiogenesis typical of glioblastoma, would be a strong predictor of clinical outcome. The FLAIR tumor volume, post-gadolinium T1 tumor volume, tumor vascular leakage volume determined by dynamic contrast-enhanced imaging, and volume of the contrast-enhanced rim seen on post-gadolinium T1-weighted images were defined for 20 patients about to undergo treatment for newly diagnosed high-grade gliomas. The potential for imaging characteristics to improve prediction of survival and time to progression over clinical variables was tested by using Cox regression analysis. Single-variable Cox regression analysis of each of the four tumor subvolumes revealed that the vascular leakage volume was the only significant predictor of survival. When the joint effect of clinical variables and the vascular leakage volume were tested for prediction of survival, only the age and the vascular leakage volume were selected as significant predictors. However, when time to progression was tested as a dependent variable, both the vascular leakage volume and the vascular permeability were selected as copredictors, along with surgical status. Our findings suggest that for patients with high-grade glioma, time to progression after radiation therapy is influenced by both underlying biological aggressiveness (vascularity) and volume of aggressive tumor. In contrast, survival depends chiefly on the volume of aggressive tumor at the time of presentation. (Cancer Res 2006; 66(17): 8912-7)

Although magnetic resonance imaging (MRI) acquisition and postprocessing can reveal a great deal of information about the physiology of high-grade gliomas, it is not clear what information is most valuable in understanding tumor physiology and in predicting tumor aggressiveness. Grade 4 gliomas are generally composed of a necrotic core with low vessel density, a contrast-enhanced rim characterized by active tumor growth and angiogenesis, and a more peripheral volume with vasogenic edema and possible tumor cell infiltration and without contrast enhancement (1, 2). Tumor volume defined on FLAIR images based on T2 abnormality encompasses all of these regions, and extends beyond the contrast-enhanced rim, whereas post-contrast T1-weighted images roughly estimate the regions of neovascularization and solid tumor. To date, it is unclear which aspect of these heterogeneous tumor volumes best describes tumor aggressiveness.

One possible aspect of tumor physiology that may represent tumor aggressiveness in high-grade gliomas is vascular leakage. The vasculature plays a very important role in glioma growth and infiltration. Malignant gliomas are able to recruit and synthesize vascular networks for further growth and proliferation (2). Vascular proliferation is also important in determining the biological aggressiveness and histopathologic grading of glioma (35). Overexpression of vascular endothelial growth factors and rapid growth of new vessels can lead to hyperpermeable vasculature in gliomas. Therefore, vascular leakage may be an indicator of tumor aggressiveness. Previous studies have shown that the vascular permeability to gadolinium (Gd)-diethylenetriaminepentaacetic acid (DTPA) estimated by dynamic contrast-enhanced (DCE) MRI was associated with histologic grade (68). However, these previous studies also revealed a wide range of vascular permeabilities within each histologic grade and a marked overlap between glioma grades. This raises the question of whether this large variation of vascular permeability reflects different stages or degrees of neovascularization in high-grade gliomas that might represent aggressiveness of the tumor.

In this study, we assessed vascular leakage in high-grade gliomas by estimating vascular permeability to Gd-DTPA from DCE T2*-weighted imaging (911). We compared the usefulness of measuring vascular leakage with other standard measurements such as the FLAIR tumor volume, the post-Gd T1 tumor volume and the volume of the contrast-enhanced rim estimated from contrast T1-weighted images for their relative ability to predict survival and time to progression. We hypothesized that the vascular leakage volume, which reflects disorganized angiogenesis typical of glioblastoma, would be a strong predictor of clinical outcome.

Patients. Twenty patients with newly diagnosed high-grade gliomas who underwent three-dimensional conformal radiation therapy (median dose of 70 Gy) and participated in a prospective, Institutional Review Board–approved, clinical MRI protocol were included in this study based on adequate DCE T2*-weighted images available for analysis of vascular permeability (Table 1). Ten patients had biopsy only and 10 patients had subtotal tumor resection, but all patients met the inclusion criterion of a minimum evaluable tumor volume of 4 mL after resection. Patients whose performance status was worse than Zubrod 2 (in bed >50% of the time) were also excluded. Seven patients received concurrent temozolomide treatment and two had concurrent poly-ICLC (an experimental proposed immune modulator; ref. 12). Patients underwent MRI 1 to 2 weeks prior to radiation therapy. For the 10 patients who had partial tumor resection, the pre–radiation therapy MRI was done at a median of 20 days (range, 12-26 days) after surgery. At the time of MRI, 13 patients received dexamethasone with a median prescribed dose of 8 mg per day. All 13 patients received dexamethasone for >1 week and were on a stable dose for at least 4 days prior to the MRI. Patients were followed up to 22 to 49 months after completion of treatment.

Table 1.

Patients' demographics, histology, and treatment information

Patient no.Age (y)Overall survival (mo)Clinical outcome (death, 1; alive, 0.)Surgical status (resection, R; biopsy, Bx)Dose (Gy)GradeConcurrent drug
39 11.4 Bx 72 Temozolomide 
31 19.1 Bx 72 Temozolomide 
60 4.1 75 Temozolomide 
70 13.1 60 Temozolomide 
23 29.7 66 Temozolomide 
62 22.6 75 Temozolomide 
73 8.2 Bx 75 Temozolomide 
68 16.1 60 Poly-ICLC 
62 7.4 60 Poly-ICLC 
10 44 49.3 70 NA 
11 66 11 70 NA 
12 54 7.2 Bx 70 NA 
13 44 14.3 Bx 70 NA 
14 61 6.2 Bx 50 NA 
15 50 6.7 Bx 70 NA 
16 63 6.2 Bx 70 NA 
17 58 6.7 Bx 52 NA 
18 48 37.6 Bx 70 NA 
19 76 2.9 Bx 66 NA 
20 41 18.5 70 NA 
Patient no.Age (y)Overall survival (mo)Clinical outcome (death, 1; alive, 0.)Surgical status (resection, R; biopsy, Bx)Dose (Gy)GradeConcurrent drug
39 11.4 Bx 72 Temozolomide 
31 19.1 Bx 72 Temozolomide 
60 4.1 75 Temozolomide 
70 13.1 60 Temozolomide 
23 29.7 66 Temozolomide 
62 22.6 75 Temozolomide 
73 8.2 Bx 75 Temozolomide 
68 16.1 60 Poly-ICLC 
62 7.4 60 Poly-ICLC 
10 44 49.3 70 NA 
11 66 11 70 NA 
12 54 7.2 Bx 70 NA 
13 44 14.3 Bx 70 NA 
14 61 6.2 Bx 50 NA 
15 50 6.7 Bx 70 NA 
16 63 6.2 Bx 70 NA 
17 58 6.7 Bx 52 NA 
18 48 37.6 Bx 70 NA 
19 76 2.9 Bx 66 NA 
20 41 18.5 70 NA 

MRI. The MRI protocol included T1-weighted, T2-weighted, FLAIR, DCE T2*-weighted imaging with i.v. administration of a single-dose (0.1 mL/kg) bolus of Gd-DTPA, and post-contrast T1-weighted imaging. For DCE imaging, 36 volumes of dynamic T2*-weighted images were acquired by a gradient-echo echo-planar imaging pulse sequence with TR, 2 seconds; TE, 60 ms; field-of-view, 220 × 220 mm2; matrix, 128 × 128; flip angle, 60 degrees; and 14 interleaved slices with 6 mm thickness and 0 mm gap. Gd-DTPA (0.1 mL/kg) was injected i.v. by a power injector at a rate of 2 mL/s, followed immediately by 15 mL of saline flush at the same rate. This protocol sufficiently suppressed the T1 effect on the estimate of vascular permeability. The vascular permeability to Gd-DTPA in the brain and tumor were computed from DCE T2*-weighted images as described below.

Image registration. All images were coregistered to post-contrast T1-weighted images using mutual information and simplex optimization. For T2*-weighted images that were acquired using gradient-echo echo-planar imaging, affine transformation including 12 variables (3 for translation and 9 for rotation and shearing) was optimized to compensate head motion, slice misalignment, and geometric distortion due to magnetic field susceptibility.

Permeability estimate. Vascular permeability to Gd-DTPA was estimated voxel-by-voxel from DCE T2*-weighted images, based on an iterative algorithm that is described elsewhere (11). In a brief, the contrast input function was determined from volumes of interest of normal white matter, and the time delay of the bolus arrival was adjusted for individual voxels. The recirculation of the contrast was included in the model and fitted together with the first passage of the bolus because recirculation has a great effect on the estimate of the transfer constant. Also, the efflux of the Gd-DTPA from tissue to plasma, and both intravascular and extravascular contrast were considered in the pharmacokinetic model.

Tumor volumes. We defined four tumor volumes on the multislice MR images that included the whole tumor: (a) the FLAIR tumor volume was defined as the volume of hyperintense signals noted on the FLAIR MR images. The FLAIR hyperintense abnormality included the volumes within and beyond the contrast-enhanced volume of the post-Gd T1-weighted images, including any central resection cavity or cyst. (b) The post-Gd T1 tumor volume was defined on the post-Gd T1-weighted images and included both the enhancing rim and the tumor core (or surgical cavity). (c) The enhanced rim volume was defined on the natural logarithm of the ratio of post- to pre-contrast T1-weighted (nondynamic) images by thresholding intensities 1 SD above the mean determined in normal tissue as described previously (13). This volume provides an estimate of the volume with contrast uptake. (d) The tumor vascular leakage volume was computed for voxels that were within the FLAIR tumor volume and had a vascular permeability (transfer constant of Gd-DTPA determined by DCE T2*-weighted images) of >0.005 per minute, the minimal detectable transfer constant (11), to exclude voxels with zero permeability (see Fig. 1). We also determined the mean of vascular permeability averaging over voxels having non-zero transfer constants of Gd-DTPA (>0.005 min−1) within the FLAIR tumor volume, which depicts the average severity of vascular leakage and is independent of the volume of vascular leakage (no correlation, P > 0.3).

Figure 1.

Post-Gd T1-weighted image (left), FLAIR image (middle), and vascular permeability map (right) color-coded and overlaid on the FLAIR image of a patient with grade 4 glioma. Left, the white contour represents the post-Gd T1 tumor volume that includes the contrast-enhanced rim and the tumor core, and the region between the white and black contours depicts the enhanced rim. Middle, the white contour denotes the FLAIR tumor volume. Right, the black contour encloses voxels with a transfer constant of >0.005 min−1 in the FLAIR tumor volume.

Figure 1.

Post-Gd T1-weighted image (left), FLAIR image (middle), and vascular permeability map (right) color-coded and overlaid on the FLAIR image of a patient with grade 4 glioma. Left, the white contour represents the post-Gd T1 tumor volume that includes the contrast-enhanced rim and the tumor core, and the region between the white and black contours depicts the enhanced rim. Middle, the white contour denotes the FLAIR tumor volume. Right, the black contour encloses voxels with a transfer constant of >0.005 min−1 in the FLAIR tumor volume.

Close modal

Statistical analysis. To determine which volumetric components of the tumor can best predict survival, each of the four volumes, i.e., the FLAIR tumor volume, the post-Gd T1 tumor volume, the enhanced rim volume, and the vascular leakage volume were tested independently for prediction of survival using single-variable Cox regression analysis. A similar analysis was applied to recognized clinical prognostic factors such as age, grade, status of subtotal resection versus biopsy only, and status of concurrent chemotherapy to identify the clinical predictors. After single-variable analysis, the joint effect of the tumor volume predictor and the clinical predictors was tested by multivariate Cox regression. The final model with clinical predictors was created and “frozen,” after which, the new imaging predictors were added to the model individually. Similar tests were done for the prediction of time to progression.

We began by evaluating the tumor volumes and characteristics of the tumor vascular leakage. The tumor volume varied from the FLAIR tumor volume, the post-Gd T1 tumor volume, and the vascular leakage volume to the enhanced rim volume in a descending order (Table 2). The vascular leakage volume was slightly smaller than the post-Gd T1 tumor volume that included the enhanced rim and tumor core (or surgical cavity), but was ∼30% greater than the enhanced rim volume. The FLAIR tumor volume was more than twice as great as the post-Gd T1 tumor volume. The mean vascular permeability to Gd-DTPA in the tumor varied from 0.015 to 0.065 min−1 among the patients, whereas the intratumor variation (reflected in the SD of each individual tumor) was as large as the mean of an individual tumor (Table 2).

Table 2.

Tumor volumes determined from different MR measures and vascular permeability

Patient no.K, mean ± SD (min−1)Vascular leakage volume (cm3)FLAIR tumor volume (cm3)Post-Gd T1 tumor volume (cm3)Enhanced rim volume (cm3)
0.026 ± 0.031 16.2 53.1 29.3 19.4 
0.056 ± 0.063 54.4 176.2 32.6 28.4 
0.027 ± 0.028 57.7 165.9 50.0 25.4 
0.018 ± 0.016 31.5 115.3 74.7 40.3 
0.029 ± 0.057 11.5 47.8 46.0 15.8 
0.020 ± 0.036 26.7 131.3 25.9 20.3 
0.035 ± 0.043 57.7 85.6 80.4 56.1 
0.015 ± 0.012 13.0 43.1 31.0 14.8 
0.033 ± 0.039 44.4 121.7 37.5 10.1 
10 0.029 ± 0.027 70.1 171.4 104.6 69.9 
11 0.054 ± 0.064 48.7 101.3 65.7 47.5 
12 0.041 ± 0.038 69.3 151.5 53.7 19.1 
13 0.053 ± 0.076 22.0 79.1 2.7 
14 0.049 ± 0.066 128.9 353.9 107.8 85.9 
15 0.032 ± 0.027 48.0 98.0 37.8 33.0 
16 0.018 ± 0.016 31.5 69.0 15.7 7.7 
17 0.033 ± 0.040 72.5 162.9 101.4 72.1 
18 0.065 ± 0.064 2.3 18.5 3.6 2.2 
19 0.046 ± 0.045 58.0 97.2 44.9 32.1 
20 0.022 ± 0.022 29.2 81.9 50.6 15.2 
Mean ± SD 0.035 ± 0.015 44.7 ± 28.9 116.2 ± 72.4 49.8 ± 31.1 30.8 ± 24.2 
Patient no.K, mean ± SD (min−1)Vascular leakage volume (cm3)FLAIR tumor volume (cm3)Post-Gd T1 tumor volume (cm3)Enhanced rim volume (cm3)
0.026 ± 0.031 16.2 53.1 29.3 19.4 
0.056 ± 0.063 54.4 176.2 32.6 28.4 
0.027 ± 0.028 57.7 165.9 50.0 25.4 
0.018 ± 0.016 31.5 115.3 74.7 40.3 
0.029 ± 0.057 11.5 47.8 46.0 15.8 
0.020 ± 0.036 26.7 131.3 25.9 20.3 
0.035 ± 0.043 57.7 85.6 80.4 56.1 
0.015 ± 0.012 13.0 43.1 31.0 14.8 
0.033 ± 0.039 44.4 121.7 37.5 10.1 
10 0.029 ± 0.027 70.1 171.4 104.6 69.9 
11 0.054 ± 0.064 48.7 101.3 65.7 47.5 
12 0.041 ± 0.038 69.3 151.5 53.7 19.1 
13 0.053 ± 0.076 22.0 79.1 2.7 
14 0.049 ± 0.066 128.9 353.9 107.8 85.9 
15 0.032 ± 0.027 48.0 98.0 37.8 33.0 
16 0.018 ± 0.016 31.5 69.0 15.7 7.7 
17 0.033 ± 0.040 72.5 162.9 101.4 72.1 
18 0.065 ± 0.064 2.3 18.5 3.6 2.2 
19 0.046 ± 0.045 58.0 97.2 44.9 32.1 
20 0.022 ± 0.022 29.2 81.9 50.6 15.2 
Mean ± SD 0.035 ± 0.015 44.7 ± 28.9 116.2 ± 72.4 49.8 ± 31.1 30.8 ± 24.2 

NOTE: K, Gd-DTPA transfer constant, a metric of vessel permeability to Gd-DTPA.

In 20 patients with high-grade gliomas (grades 3 and 4), single-variable Cox regression analysis of each of the four tumor subvolumes and the tumor vascular permeability revealed that only the tumor vascular leakage volume was a significant predictor of survival (P = 0.02), but the FLAIR tumor volume, the post-Gd T1 tumor volume, the enhanced rim volume, and the vascular permeability were not (P = 0.2, 0.4, 0.4, and 0.9, respectively). Single-variant Cox regression analysis was then applied to the commonly observed clinical prognostic factors, which showed that only age was a significant predictor of survival (P = 0.03), and that grade, status of surgery, and status of concurrent chemotherapy were not (P = 0.9, 0.1, and 0.4, respectively). Finally, the joint effect of age and vascular leakage volume for prediction of survival was found to be stronger (P = 0.009) than by either the age (P = 0.03) or the vascular leakage volume alone (P = 0.02).

Next, we tested if the two predictors for survival (the age and the vascular leakage volume) obtained from all the high-grade glioma patients were valid for the subgroup of the patients with grade 4 (n = 15). Only the vascular leakage volume was identified as a significant predictor (P = 0.04) and age was not (P > 0.1).

To understand why the vascular leakage volume determined from DCE imaging predicted survival, but why the enhancing rim volume determined from nondynamic T1-weighted images did not, we investigated the spatial relationship between these two volumes. We found that DCE images and Gd-enhanced T1-weighted images did not provide consistent estimates on vascular leakage. The DCE estimates of the vascular leakage volume were within 6% of the enhancing rim volume obtained from the post-Gd T1-weighted images in only 5 of the 20 patients. In 11 patients, the enhancing rim volumes were smaller than the tumor vascular leakage volume estimated by DCE imaging by a median of 48% (Fig. 2), suggesting that the T1 contrast-enhancement underestimated the tumor vascular leakage. In the remaining four patients, the opposite finding was observed—the enhancing rim volume was greater than the vascular leakage volume by a median of 24%, suggesting that the vascular leakage was overestimated by the T1 contrast enhancement (Fig. 3). These numerical results were supported by visual inspection regardless of the thresholds used to determine the two volumes (Figs. 2 and 3).

Figure 2.

Post-contrast T1-weighted image (left) and the color-coded permeability map (right) overlaid on the post-contrast T1-weighted image. Green contours enclose the contrast-enhanced rim seen on the T1-weighted images, whereas the volume between the inner and outer pink contours has vascular permeability to Gd-DTPA of >0.005 min−1. Arrow, area in which vascular leakage was detected by quantitative analysis of DCE data but not by contrast-enhanced T1-weighted images.

Figure 2.

Post-contrast T1-weighted image (left) and the color-coded permeability map (right) overlaid on the post-contrast T1-weighted image. Green contours enclose the contrast-enhanced rim seen on the T1-weighted images, whereas the volume between the inner and outer pink contours has vascular permeability to Gd-DTPA of >0.005 min−1. Arrow, area in which vascular leakage was detected by quantitative analysis of DCE data but not by contrast-enhanced T1-weighted images.

Close modal
Figure 3.

Post-contrast T1-weighted image (top row, left) and the color-coded permeability map (top row, right). Tissue volumes enclosed by both white and black contours were enhanced by Gd-DTPA on T1-weighted images, but only the volume enclosed by the black contour was detected to have vascular leakage by DCE analysis. DCE signal intensities from both regions show a large fraction of blood volume, 6.6 and 4.7 times (white and black regions, respectively) that of normal white matter (NWM). However, only signals from the black contour region manifest high permeability (0.11 min−1). The DCE curve in the white contour region after being normalized to NWM at the peak is matched with the NWM curve (bottom), suggesting that gadolinium enhancement in the white contour region largely represents blood volume. The normalized curve in the black contour region shows an elevation after the peak compared with the NWM curve, indicating that there is vascular leakage.

Figure 3.

Post-contrast T1-weighted image (top row, left) and the color-coded permeability map (top row, right). Tissue volumes enclosed by both white and black contours were enhanced by Gd-DTPA on T1-weighted images, but only the volume enclosed by the black contour was detected to have vascular leakage by DCE analysis. DCE signal intensities from both regions show a large fraction of blood volume, 6.6 and 4.7 times (white and black regions, respectively) that of normal white matter (NWM). However, only signals from the black contour region manifest high permeability (0.11 min−1). The DCE curve in the white contour region after being normalized to NWM at the peak is matched with the NWM curve (bottom), suggesting that gadolinium enhancement in the white contour region largely represents blood volume. The normalized curve in the black contour region shows an elevation after the peak compared with the NWM curve, indicating that there is vascular leakage.

Close modal

Finally, analyses similar to those used for the prediction of survival were applied to time to progression. Single-variable Cox regression revealed that the status of surgery was a significant predictor (P = 0.04), the vascular leakage volume was a marginal predictor (P = 0.08), and other variables were not (P > 0.1). The joint effects of the status of surgery, the tumor vascular leakage volume, and the vascular permeability were stronger than any single effect (P = 0.003 for overall, P = 0.004 for the status of surgery, P = 0.03 for the vascular leakage volume, and P = 0.04 for vascular permeability).

In this study, we found that the vascular leakage volume of the tumor was a better predictor of survival than other tumor subvolumes such as the FLAIR tumor volume, post-Gd T1 tumor volume, and the enhancing rim volume in patients with high-grade gliomas. Age was the only clinical factor that was found to predict survival in this patient sample. The joint effect of vascular leakage volume of the tumor and age was stronger than either predictor alone. When the same variables were tested for prediction of time to progression, both the vascular leakage volume of the tumor and the mean vascular permeability were identified as the copredictors, as well as surgical status. Again, the joint effect of the three predictors was stronger than either variable alone. The mean vascular permeability, which depicts the severity of vascular leakage, was selected as a significant copredictor for time to progression but not for survival.

Although it is logical that the spread of tumor should be related both to time to progression and survival, it is important to consider why the mean vascular permeability predicted only time to progression. The vascular permeability estimated by the transfer constant of Gd-DTPA from plasma to tumor or brain tissue is a measure of the degree of blood-tumor (or blood-brain) barrier opening. The Gd-DTPA with a molecular size of ∼500 Da does not pass the intact tight endothelium junctions that make up the blood-brain barrier. In glioma, a number of angiogenic growth factors stimulate the growth of new, immature blood vessels that are highly permeable and tortuous. A high value of vascular permeability indicates abnormality of the blood-brain (or blood-tumor) barrier and presumably reflects a larger size pore or disruption of the endothelial junctions. Thus, the magnitude of blood-tumor barrier opening seems to be a surrogate marker for aggressiveness of the tumor, and affects time to progression. However, within this group of patients with high-grade glioma, the differences in aggressiveness seem to have less of an effect on survival than the overall tumor burden as manifested by postoperative vascular leakage volume.

It is still a challenge to determine the margin or the extent of the tumor due to the diffuse or infiltrative nature of glioma. The tumor volume defined on the FLAIR hyperintense abnormalities includes the necrotic core, the contrast-enhanced rim, and non–contrast-enhanced abnormalities. In the latter region, we believe that there may exist both tumor cell infiltration and vasogenic edema. Therefore, FLAIR tumor volume may result in an overestimate of tumor volume, especially in high-grade (grade 4) gliomas, due to the uncertainty of edema and tumor cell infiltration in the nonenhanced region. On the other hand, the tumor volume defined on the post-Gd T1-weighted images excludes the region of non–contrast-enhanced abnormality, and may, therefore, underestimate the tumor volume. Other imaging approaches such as choline-containing compounds and methionine uptake have been investigated for definition of the radiation target (1416). The vascular leakage volume estimated by DCE MRI describes the spread of increased vascularity, which includes both focal areas with severe vascular abnormalities and regions with mild vascular leakage that often extends into the nonenhanced abnormal region. Furthermore, the vascular leakage volume is a better predictor for survival than the FLAIR and post-Gd T1 tumor volume, and is a copredictor for time to progression. Our findings suggest that the extent of vascular leakage could be a surrogate marker for the extent of high-grade gliomas.

Finally, we need to consider the difference between vascular leakage estimated by post-Gd T1-weighted imaging and by the transfer constant of Gd-DTPA from plasma to tissue by DCE imaging. There are a few factors that are not fully controlled in the post-Gd T1-weighted images and can influence the appearance of enhancement. First, the post-Gd T1 enhancement does not differentiate intravascular from extravascular contrast. Thus, a region that has a large fractional blood volume or a long mean transit time of vasculature can appear as enhanced in the post-Gd T1-weighted images. Second, a rapid contrast washout that can result from a great rate of the efflux of Gd-DTPA from tissue to plasma, a small extravascular extracellular space, or both, could lead to a nonenhanced or less enhanced appearance on the post-Gd T1-weighted images. On the other hand, DCE imaging can minimize these factors' influence on the estimate of the transfer constant of Gd-DTPA using a pharmacokinetic model. Our data showed that there was a discrepancy in the extent of vascular leakage estimated by the transfer constant of Gd-DTPA from one by post-Gd T1 enhancement. Consistent spatial overlaps of vascular leakage between the two estimates were only observed in 25% of patients (5 of 20). In 55% of the patients, the vascular leakage regions that were determined by DCE imaging extended beyond the boundaries of the T1 contrast-enhanced rims. In the remaining 20% of the patients, there existed regions that were enhanced on the post-contrast T1-weighted images but were not determined to have vascular leakage by DCE imaging. Although the volume of vascular leakage and the volume of enhanced rim depend on the thresholds of the variables (see Materials and Methods), we believe that the sensitivity and specificity of the transfer constant of Gd-DTPA to vascular leakage compared with post-Gd T1 enhancement result in the first variable being a better predictor for survival than the latter one.

In this study, we excluded patients who had <4 mL of residual tumor volumes after resection in order to have a sufficient volume of residual tumor to be evaluated on MR images. In addition, we excluded patients with performance status worse than Zubrod 2 (in bed >50% of the time), as we were concerned that they would not survive long enough to complete all of the imaging studies. Compared with unselected patients, the former feature would tend to select for patients with a worse prognosis (patients with a complete resection tend to do better than those with residual disease), whereas the latter would tend to select for a better prognosis (patients with poor performance status do worse). Although it is likely that these opposite effects cancel each other out, it is possible that the median survival of this group might differ from one observed in the unselected patients. However, it is unlikely that the relationship between the tumor vascular leakage volume and survival will be altered by the patient selection criteria. Finally, the prognostic value of the vascular leakage volume that was found in this pilot study of 20 patients provides a scientific base to form a hypothesis that can be further tested in a trial with a large sample.

In this study, we found that the vascular leakage volume of a glioma was a stronger predictor for survival than the other tumor subvolumes determined from FLAIR and post-Gd T1-weighted images in high-grade gliomas. The mean vascular permeability, although not a predictor for survival, was a predictor for time to progression. This suggests that the leaky vasculature may represent the fast-growing and aggressive subvolume of the tumor. Therefore, both the extent and severity of vascular leakage seem to have prognostic values for different clinical end points.

Grant support: NIH grants 2 PO1 CA59827, PO1 CA85878, and R21 CA11369901.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1
Holash J, Maisonpierre PC, Compton D, et al. Vessel cooption, regression, and growth in tumors mediated by angiopoietins and VEGF.
Science
1999
;
284
:
1994
–8.
2
Holash J, Wiegand SJ, Yancopoulos GD. New model of tumor angiogenesis: dynamic balance between vessel regression and growth mediated by angiopoietins and VEGF.
Oncogene
1999
;
18
:
5356
–62.
3
Zagzag D, Friedlander DR, Dosik J, et al. Tenascin-C expression by angiogenic vessels in human astrocytomas and by human brain endothelial cells in vitro.
Cancer Res
1996
;
56
:
182
–9.
4
Ausprunk DH, Folkman J. Migration and proliferation of endothelial cells in preformed and newly formed blood vessels during tumor angiogenesis.
Microvasc Res
1977
;
14
:
53
–65.
5
Brem S. The role of vascular proliferation in the growth of brain tumors.
Clin Neurosurg
1976
;
23
:
440
–53.
6
Roberts HC, Roberts TP, Brasch RC, et al. Quantitative measurement of microvascular permeability in human brain tumors achieved using dynamic contrast-enhanced MR imaging: correlation with histologic grade.
AJNR Am J Neuroradiol
2000
;
21
:
891
–9.
7
Provenzale JM, Wang GR, Brenner T, et al. Comparison of permeability in high-grade and low-grade brain tumors using dynamic susceptibility contrast MR imaging.
AJR Am J Roentgenol
2002
;
178
:
711
–6.
8
Law M, Yang S, Babb JS, et al. Comparison of cerebral blood volume and vascular permeability from dynamic susceptibility contrast-enhanced perfusion MR imaging with glioma grade.
AJNR Am J Neuroradiol
2004
;
25
:
746
–55.
9
Weisskoff RW, Boxeman JL, Sorensen AG, et al. Simultaneous blood volume and permeability mapping using a single Gd-based contrast injection. The 2nd Annual Meeting of International Society for Magnetic Resonance in Medicine. vol. 1. San Francisco; 1994. p. 279.
10
Johnson G, Wetzel SG, Cha S, et al. Measuring blood volume and vascular transfer constant from dynamic, T(2)*-weighted contrast-enhanced MRI.
Magn Reson Med
2004
;
51
:
961
–8.
11
Cao Y, Shen Z, Chenevert TL, et al. An estimate of vascular permeability and cerebral blood volume using Gd-DTPA contrast enhancement and dynamic T2* weighted MRI.
J Magn Reson Imaging
2006
;
24
:
288
–96.
12
Salazar AM, Levy HB, Ondra S, et al. Long-term treatment of malignant gliomas with intramuscularly administered polyinosinic-polycytidylic acid stabilized with polylysine and carboxymethylcellulose: an open pilot study.
Neurosurgery
1996
;
38
:
1096
–103; discussion 1103–4.
13
Cao Y, Tsien CI, Shen Z, et al. Use of magnetic resonance imaging to assess blood-brain/blood-glioma barrier opening during conformal radiotherapy.
J Clin Oncol
2005
;
23
:
4127
–36.
14
Graves EE, Nelson SJ, Vigneron DB, et al. Serial proton MR spectroscopic imaging of recurrent malignant gliomas after gamma knife radiosurgery.
AJNR Am J Neuroradiol
2001
;
22
:
613
–24.
15
Graves EE, Nelson SJ, Vigneron DB, et al. A preliminary study of the prognostic value of proton magnetic resonance spectroscopic imaging in gamma knife radiosurgery of recurrent malignant gliomas.
Neurosurgery
2000
;
46
:
319
–26; discussion 326–8.
16
Grosu AL, Weber WA, Riedel E, et al. L-(methyl-11C) methionine positron emission tomography for target delineation in resected high-grade gliomas before radiotherapy.
Int J Radiat Oncol Biol Phys
2005
;
63
:
64
–74.