Evaluation of Perfusion CT in Grading and Prognostication of High-Grade Gliomas at Diagnosis: A Pilot Study

Transcription

1 Neuroradiology/Head and Neck Imaging Original Research Shankar et al. Neuroradiology/Head and Neck Imaging Original Research Jai Jai Shiva Shankar John Woulfe 2 Vasco Da Silva 2 Thanh B. Nguyen 3 Shankar JJS, Woulfe J, Da Silva V, Nguyen TB Keywords: cerebral blood volume, high-grade glioma, perfusion CT, permeability surface area product DOI:.224/AJR Received March 2, 22; accepted after revision July 3, 22. Department of Diagnostic Imaging, QEII Hospital, 5743 Southwood Dr, 87, Halifax, NS B3H E6, Canada. Address correspondence to J. J. S. Shankar (shivajai@rediffmail.com; shivajai@gmail.com). 2 Department of Pathology, Neuropathology Division, The Ottawa Hospital, Ottawa, ON, Canada. 3 Department of Diagnostic Imaging, The Ottawa Hospital, Ottawa, ON, Canada. WEB This is a Web exclusive article. AJR 23; 2:W54 W X/3/25 W54 American Roentgen Ray Society Evaluation of Perfusion CT in Grading and Prognostication of High-Grade Gliomas at Diagnosis: A Pilot Study OBJECTIVE. Differentiation of grade 3 astrocytoma from glioblastoma multiforme can be difficult with conventional structural imaging but is important for prognosis. The purpose of this study was to assess perfusion CT in differentiating high-grade gliomas (HGGs) and their role in prognosis in the care of patients with HGG. SUBJECTS AND METHODS. Twenty patients with previously untreated HGG underwent prospective evaluation with perfusion CT. Permeability surface area product (PS) and cerebral blood volume (CBV) were calculated by the deconvolution method and were compared between HGGs with Student two-sample t tests. Receiver operating characteristic curves were generated for PS, CBV, and the conjoint factor PS + CBV. Cox regression analysis was used to correlate these parameters with patient survival over a follow-up period. Hazard ratios were calculated, and Kaplan-Meier survival curves were drawn. RESULTS. There was a significant difference between grade 3 and grade 4 gliomas for PS (p =.22) and PS + CBV (p =.9) but not for CBV alone (p =.4). Receiver operating characteristic analyses showed that PS (area under the curve [AUC],.72) and CBV + PS (AUC,.73) can be used to differentiate grade 3 from grade 4 gliomas but that CBV alone cannot be so used (AUC,.54). There was a significant relation between patient outcome and age (p =.34) and CBV + PS (p =.48). Patients with HGG and a CBV + PS greater than 9 had a poor outcome (hazard ratio, 6.). CONCLUSION. PS and CBV + PS can be used to differentiate grade 3 from grade 4 gliomas. The outcome of patients with HGG depends on age and CBV + PS. T he prognosis of cerebral glioma depends on the histologic grade of the tumor. Median survival times for patients with grade 3 and grade 4 gliomas are 22 months and 4 months []. CT and MRI by conventional techniques are not always accurate in predicting the histologic grade of gliomas, including high-grade gliomas (HGGs) [2, 3]. Necrosis and microvascular proliferation are histologic features differentiating glioblastoma multiforme, or grade 4 glioma, from anaplastic astrocytoma, or grade 3 glioma. Microvascular proliferation resulting from angiogenesis (formation of new blood vessels from existing ones) plays a central role in the growth and spread of tumors [4 6]. Quantification of tumor angiogenesis is important in predicting tumor grade, which in turn affects prognosis [7]. Because angiogenesis increases both microvascular density and vascular permeability in tumors, two perfusion parameters can be used to quantify and characterize this process: cerebral blood volume (CBV) and permeability surface area product (PS) [8 7]. CBV is defined as the volume of flowing blood present within the blood vessels per g of brain tissue at a particular point in time. CBV can be used as a prognostic marker in patients with glioma [4 24]. It may also help in defining treatment options and is being assessed as a predictive marker for antiangiogenesis drugs [25]. PS characterizes the diffusion of contrast agent from blood vessels into the interstitial space due to deficient or leaky blood-brain barrier. The diffusion flux of a contrast agent across capillary endothelium depends on its size, the diffusion coefficient, and the total surface area of the pores. Both perfusion CT and perfusion MRI can be used to assess CBV and PS. Owing to the nonlinear relation between contrast and change in signal intensity, perfusion MRI is limited in absolute quantification of perfusion parameters. Two contrast boluses are W54 AJR:2, May 23

2 often needed for quantitative measurement of CBV and PS at perfusion MRI. Perfusion CT is a reliable technique that can be used for absolute quantification of both CBV and PS in a single acquisition with a single bolus of contrast medium [26]. Perfusion studies have been used to differentiate HGG from low-grade glioma [2,, 4 6, 8, 2, 2, 27, 28]. However, only a few studies have been conducted on differentiating the two grades (grades 3 and 4) of HGG [23, 29]. It is also not clear whether all HGGs have equally poor prognoses, because many other factors can affect prognosis in the care of patients with HGG, such as age, sex, performance status, tumor size, time of presentation, and treatment received [9, 26]. The purpose of this study was to assess the diagnostic accuracy of CBV and PS derived from perfusion CT in differentiating the grades of HGG and to assess the prognostic value of CBV and PS. Subjects and Methods Patient Population This prospective study was approved by the institutional ethics committee. Between 26 and 28, 25 consecutively registered adult patients with suspected glioma were referred from the neurosurgery department for CT of the head with contrast administration for neuronavigation and preoperative planning before treatment was started. They were approached to participate in our study. The exclusion criteria were pathologic findings other than untreated glioma and any contraindication to CT and use of CT contrast media, such as pregnancy and severe renal failure. Four patients declined to participate in the study. Informed consent was obtained from all participating patients. One patient had low-grade glioma and was subsequently excluded from the analysis. For all 2 patients, a neuropathologist confirmed the histologic grade of glioma according to World Health Organization criteria. All patients underwent follow-up during the standard therapy, which included surgery, chemotherapy with temozolomide, and radiotherapy determined by physicians at our hospital. The median follow-up period was 9 months. Perfusion CT Technique Perfusion CT was performed with 6- or 64- MDCT scanner (LightSpeed 6 or LightSpeed+, GE Healthcare) depending on availability of the scanner. Perfusion CT was performed before the neuronavigation protocol. An unenhanced head CT scan was acquired to localize the tumor before acquisition of the perfusion scan. Four slices of 5 mm each with a matrix size of A Fig. 59-year-old woman with seizures and right frontal glioblastoma multiforme. Placement of regions of interest (ovals) in tumor and normal-appearing white matter in contralateral hemisphere. A, Cerebral blood volume map. B, Permeability surface area product map. were selected at the level of the tumor. Multiple images were acquired at each level starting 5 seconds after injection of 5 ml of nonionic iodinated contrast medium (iohexol, Omnipaque 3, GE Healthcare) at a rate of 4 ml/s. The acquisition parameters were 8 kvp and ma. The images were acquired every second for a total of 99 seconds. The images were postprocessed at a separate workstation with perfusion CT software (CT Perfusion 3 FuncTool, GE Healthcare), which is based on a two-compartment model and a deconvolution method. In compartmental analysis, the tracer (contrast material in the case of CT) is modeled as entering an organ via an artery, rapidly distributing uniformly within the blood vessels and extracellular space, and after a short interval starting to leave the organ via a vein. The deconvolution method is based on modeling the tissue s impulse response function, which is the time-enhancement curve of the tissue resulting from an idealized instantaneous injection of one unit of tracer. The deconvolution method assumes that contrast material is nondiffusible. Although nondiffusibility is a reasonable assumption in the brain, it is not the case for brain tumor owing to disruption of the blood-brain barrier. In general, leakage into the interstitial space is slow with respect to the transit time of the contrast material and if diffusion is zero only leads to small errors in most organs. Quantitative permeability information can be obtained by the use of the Cenic alteration of the deconvolution method, which models perfusion in brain tumors as the sum of two impulse response functions, one due to intravascular tissue perfusion and the other modeling the flow of contrast material into the extravascular space [3]. This adaptation allows the deconvolution method to generate permeability and perfusion values but requires a longer data acquisition period to determine the outflow characteristics of the extravascular contrast material. The superior sagittal sinus was used as the venous output function in all patients. The largest and most proximal artery available at the imaging level was used for the arterial input function. TABLE : Mean Cerebral Blood Volume and Permeability Surface Area Product for Gliomas and Normal-Looking White Matter Permeability Surface Tissue No. of Patients Cerebral Blood Volume (ml/ g) Area Product (ml/ g/min) Grade 4 glioma ± ± 6. Grade 3 glioma ± ±.63 Normal-looking white matter ± ±.28 B AJR:2, May 23 W55

3 Shankar et al. In most cases this vessel was the internal carotid artery or one of its major branches. CBV and PS maps were generated. PS is the rate at which contrast agent flows into the extravascular tissues. It is related to another commonly stated parameter of vascular leakage, the transfer constant (K trans ) as follows: K trans = EF, where F is the flow and E is the extraction fraction. In high-flow states (PS / F < ), K trans is equivalent to the PS [25]. One of the authors placed multiple regions of interest (ROIs) in areas of tumor with the highest CBV and PS values (Fig. ). The highest value obtained represented the most malignant part of the tumor. The reader was blinded to the pathologic results. ROIs were placed in the solid portion of the tumors, and the area of necrosis or cystic degeneration was avoided. Another ROI was placed in contralateral normal-appearing white matter (Fig. ). Statistics Student two-sample t tests were used to compare PS and CBV between glioma and normalappearing white matter and between grade 3 and grade 4 gliomas. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the perfusion parameters in the diagnosis of different grades of HGG were calculated. The diagnostic accuracy of CBV, PS, and CBV + PS in differentiating grade 3 from grade 4 gliomas was assessed with ROC analysis and calculation of the area under the curve (AUC) for each parameter. AUC can be interpreted as the probability that the test result from a randomly chosen individual with disease is more indicative of disease than that from a randomly chosen individual without disease. A value closer to is regarded as a higher true-positive rate. Cox regression analysis with proportional hazards was used to determine the relation between overall survival time and CBV, PS, CBV + PS, and age. In our study, a direct relation between overall survival time and grade could not be tested because all patients with grade 3 glioma survived the follow-up period. Kaplan-Meier survival curves were drawn to assess the relation between CBV + PS and overall survival time for patients with HGG. A value of p <.5 was considered statistically significant. Permeability Surface Area Product (ml/ g/min) Grade 3 Grade 4 Fig. 2 Scatter diagram shows permeability surface area product values for grade 3 and grade 4 gliomas. Results Fifteen of the 2 patients (nine men, six women; mean age, 63 ± 4 [SD] years) had glioblastoma multiforme, or grade 4 glioma, and five (four men, one woman; mean age, 57 ± years) had grade 3 glioma. Three of the grade 3 gliomas were anaplastic astrocytoma, and two were anaplastic oligodendrogliomas. Table summarizes CBV and PS by grade of glioma. The mean CBV and PS for gliomas were significantly higher (p <.) than those for normal-appearing white matter. Although there was some overlap in PS between grade 3 and grade 4 gliomas (Fig. 2), the mean PS was significantly higher (p =.22) for grade 4 gliomas than for grade 3 gliomas. At a cutoff PS value of more than 3.5 ml/ g/min, the sensitivity, specificity, PPV, and NPV were 87%, 6%, 87%, and 6%. There was overlap in CBV between grade 3 and 4 gliomas (Fig. 3). The mean CBV of grade 4 gliomas was higher than that of grade 3 gliomas, but the difference was not statistically significant (p =.4). At a cutoff CBV of more than 3.5 ml/ g, the sensitivity, specificity, PPV, and NPV were 53%, 6%, 8%, and 3%. Post hoc analysis showed that the empirical conjoined parameter CBV + PS was significantly different between grade 3 and grade 4 gliomas (p =.9). At a cutoff of CBV + PS 9, this conjoined parameter had sensitivity, specificity, PPV, and NPV of 67%, Cerebral Blood Volume (ml/ g) Grade 3 Grade 4 Fig. 3 Scatter diagram shows cerebral blood volume values for grade 3 and grade 4 gliomas. 8%, 9%, and 44%. ROC analysis showed that CBV + PS (AUC,.73) and PS (AUC,.72) had the highest accuracy in differentiating grade 3 from grade 4 gliomas. CBV alone was not accurate (AUC,.54) (Fig. 4). The mean follow-up period in this study was.2 months (range, 44 months; median, 9 months) (Fig. 5). Eight of the 5 patients with grade 4 gliomas died during the follow-up period. The patients who died had an average follow-up period of 5 months. Cox regression analysis showed a significant relation between the overall survival time of the patients and both age of the patients (p =.34) and CBV + PS (p =.48) independently. CBV alone (p =.37) and sex (p =.647) did not appear to be significant factors in the outcome of our patients. PS did not reach statistical significance (p =.79). The hazard ratio for age was.7 (95% CI,..3) and for CBV + PS greater than 9 was 6. (95% CI, ). Figure 6 shows the Kaplan-Meier survival curves drawn for patients with HGG and CBV + PS values greater than and less than 9. W56 AJR:2, May 23

4 Sensitivity A Specificity B Specificity C Fig. 4 Graphs show areas under receiver operating characteristic curves. A, Cerebral blood volume (CBV). B, Permeability surface area product (PS). C, CBV + PS. Discussion HGG is characterized by neovascularity and increased angiogenic activity with a higher proportion of immature and hyperpermeable vessels. Because microvascular proliferation is an important characteristic in the grading of gliomas, imaging techniques that yield hemodynamic information about the tumor may help in their characterization. This may help overcome some of the diagnostic limitations due to conventional imaging and inaccurate histopathologic grading due to sampling bias. Perfusion imaging has been useful in grading of cerebral neoplasms. Most perfusion studies comparing histologic features have been conducted with perfusion MRI techniques [8 24]. Perfusion CT, however, has been used as an alternative method for assessment of cerebral hemodynamics of brain tumors [26 28]. Relation Between Tumor Grade and Perfusion Parameters Both tumor blood volume and PS have been found useful in differentiating lowgrade glioma from HGG [2,, 4 6, 8, 2, 2, 27, 28], but there have been few studies on assessment of perfusion parameters for differentiation of grade 3 from grade 4 gliomas [23, 29]. These two grades are difficult to differentiate with routine CT and MRI, and most of the time the two grades are lumped together for prognostication. In our study, grade 4 gliomas had higher CBV, but this finding was not significant in differentiating the two grades. The PS was significantly higher in the grade 4 gliomas than in the grade 3 gliomas. This result is in agreement with findings of Jain et al. [28], who found that PS had a stronger association with Fig. 5 Bar diagram shows follow-up period for patients with high-grade gliomas. Red bar indicates patient died during follow-up. Sensitivity glioma grade than did CBV, cerebral blood flow, or mean transit time. This suggests that vascular permeability is a more important factor than microvascular density for differentiating grade 3 from grade 4 gliomas. CBV is influenced by the presence of large vessels other than capillaries. The overall increase in CBV in grade 4 gliomas might thus be less apparent than changes in capillary leakage measured with PS. This finding may reflect that the neoangiogenic vessels of HGG are highly immature and therefore have increased permeability. This phenomenon may be secondary to the higher concentration of vascular endothelial growth factor in these tumors [3]. Relation Between Patient Prognosis and Perfusion Parameters The purpose of grading of gliomas has been to assess prognosis. In our study, because none of the patients with grade 3 glioma died during Patient No Sensitivity Specificity the follow-up period, a hazard ratio for grade 3 versus grade 4 could not be calculated. Patient age was a significant factor in outcome of HGG (hazard ratio,.7). This is in keeping with the published results showing worse prognosis among older patients with HGG [32]. Sex, CBV and PS individually were not found to be statistically significant factors in the outcome of HGG. PS, however, had a value close to statistical significance (p =.79). Various studies have assessed the correlation between perfusion parameters and prognosis of glioma. The MRI-derived K trans, which is a function of cerebral blood flow and PS, has been found to be related to the prognosis of the glioma [2]. Mills et al. [2] found that patients with lowgrade glioma have lower K trans values and better prognosis than patients with HGG. However, HGG patients with high K trans values had improved survival rates. Mills et al. suggested this finding was counterintuitive and did not Follow-Up Period (mo) AJR:2, May 23 W57

5 Shankar et al. Survival Distribution Function Survival Period (mo) Fig. 6 Kaplan-Meier curves show pattern of survival over 44 months among patients with high-grade gliomas and conjoint cerebral blood volume and permeability surface area product greater than 9 (black) or less than 9 (red). Circle indicates patient died with high-grade glioma during follow-up period. have a satisfactory explanation for it. Because CBV and K trans appear to be two independent prognostic markers, we explored the usefulness of a simple conjoined marker, CBV + PS, by adding the CBV and PS values empirically. An important finding in our study was the conjoint parameter CBV + PS in differentiating the two grades of HGG and in predicting patient outcome. CBV and PS reflect vascular density and vascular permeability, respectively, and therefore the two components of tumor neovascularity. Our study results suggest that the two components of neovascularity have an additive and not an exclusive effect on the prognosis of HGG. In other words, malignancy of HGG depends on both increased microvascular density and the immaturity of tumor vessels in terms of their leakiness. It will be of interest to see whether patients with HGG and higher CBV + PS values also have higher vascular endothelial growth factor expression and production [33]. This needs to be further studied in a larger series of patients with HGG. The average survival period after imaging assessment in our study was.6 months. Patients who died during follow-up had an average survival period of 5 months. These findings were comparable to those of other studies [32]. Perfusion CT has the advantage of wider availability, faster scanning time, low cost, and ease of quantification of perfusion parameters compared with perfusion MRI. In our study the estimated average radiation dose associated with perfusion CT was 4.4 msv. This dose is slightly higher than that for a CT examination of the head, but these patients often receive a much larger dose during radiation therapy. Several limiting factors in our study might have biased the assessment of glioma grade and perfusion parameters. These include possible sampling bias because the histologic specimen might not have corresponded to the area of perfusion measurement and incomplete coverage (only 2 cm) of the tumors, particularly large tumors. The presence of oligodendrogliomas and astrocytomas in our group of grade 3 gliomas might also have caused selection bias because oligodendrogliomas have been described as having higher CBV than astrocytomas. However, we observed a higher CBV for grade 3 astrocytoma than for grade 3 oligodendroglioma. The small number of patients with grade 3 glioma (n = 5) might have limited the statistical inference and the conclusion that can be drawn from the results. Steroids can also influence the perfusion parameters, particularly PS. Other factors can bias evaluation of perfusion parameters in predicting patient outcome. The small sample size, the short follow-up period, and the existence of multiple covariates (tumor size, tumor location, type of surgery, Karnofsky performance score) also might have influenced the survival of the patients. Finally, a post hoc comparison of the conjoint parameter CBV + PS with a Bonferroni correction would decrease the probability of type error. However, the stringent nature of this correction would have made it difficult to detect any significant difference between grade 3 and grade 4 gliomas in this pilot study. The addition of PS and CBV is also experimental and may indicate a future possibility of a new vascular parameter for studying glioma. Our findings are preliminary, and a larger study is necessary to confirm them. Conclusion The results of this pilot study suggest that the perfusion CT derived conjoint parameter CBV + PS may be a better parameter than CBV and PS individually for differentiating grade 3 and grade 4 HGGs. This finding may suggest that blood volume and permeability have an additive effect on outcome of HGG. References. Salman M. Glioblastoma and malignant astrocytoma. In: Kaye AH, Laws ER Jr, eds. Brain tumors: an encyclopedic approach. New York, NY: Churchill Livingstone, 995: 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 24; 25: Law M, Cha S, Knopp EA, et al. High-grade gliomas and solitary metastases: differentiation by using perfusion and proton spectroscopic MR imaging. Radiology 22; 222: Folkman J. The role of angiogenesis in tumor growth. Semin Cancer Biol 992; 3: Nugent LJ, Jain RK. Extravascular diffusion in normal and neoplastic tissues. Cancer Res 984; 44: Jain RK, Gerlowski LE. Extravascular transport in normal and tumor tissues. Crit Rev Oncol Hematol 986; 5: Burger PC, Vogel FS. The brain tumors. In: Burger PC, Vogel FS, eds. Surgical pathology of the central nervous system and its coverings. New York: Wiley, 982: Knopp EA, Cha S, Johnson G, et al. Glial neoplasms: dynamic contrast-enhanced T2*-weighted MR imaging. Radiology 999; 2: Wong JC, Provenzale JM, Petrella JR. Perfusion MR imaging of brain neoplasms. AJR 2; 74: Cha S, Knopp EA, Johnson G, et al. Intracranial mass lesions: dynamic contrast-enhanced susceptibility-weighted echo-planar perfusion MR imaging. Radiology 22; 223: 29. Lev MH, Rosen BR. Clinical applications of intracranial perfusion MR imaging. Neuroimaging Clin N Am 999; 9: Shin JH, Lee HK, Kwun BD, et al. Using relative cerebral blood flow and volume to evaluate the histopathologic grade of cerebral gliomas: preliminary results. AJR 22; 79: Petrella JR, Provenzale JM. MR perfusion imaging of the brain: techniques and applications. AJR 2; 75: Law M, Yang S, Wang H, et al. Glioma grading: specificity and predictive values of perfusion MR imaging and proton MR spectroscopic imaging W58 AJR:2, May 23

6 compared with conventional MR imaging. AJNR 23; 24: Ding B, Ling HW, Chen KM, et al. Comparison of cerebral blood volume and permeability in preoperative grading of intracranial glioma using CT perfusion imaging. Neuroradiology 26; 48: Di Nallo AM, Vidiri A, Marzi S, et al. Quantitative analysis of CT-perfusion parameters in the evaluation of brain gliomas and metastases. J Exp Clin Cancer Res 29; 28:38 7. Jain R, Gutierrez J, Narang J, et al. In vivo correlation of tumor blood volume and permeability with histologic and molecular angiogenic markers in gliomas. AJNR 2; 32: Lev MH, Ozsunar Y, Henson JW, et al. Glial tumor grading and outcome prediction using dynamic spin-echo MR susceptibility mapping compared with conventional contrast-enhanced MR: confounding effect of elevated rcbv of oligodendrogliomas. AJNR 24; 25: Oh J, Henry RG, Pirzkall A, et al. Survival analysis in patients with glioblastoma multiforme: predictive value of choline-to-n-acetylaspartate index, apparent diffusion coefficient, and relative cerebral blood volume. J Magn Reson Imaging 24; 9: Patankar TF, Haroon HA, Mills SJ, et al. Is volume transfer coefficient (K trans ) related to histologic grade FOR YOUR INFORMATION See the AJR s monthly Journal Club article in this issue on page 957. in human gliomas? AJNR 25; 26: Mills SJ, Patankar TA, Haroon HA, et al. Do cerebral blood volume and contrast transfer coefficient predict prognosis in human glioma? AJNR 26; 27: Law M, Oh S, Babb JS, et al. Low-grade gliomas: dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging prediction of patient clinical response. Radiology 26; 238: Hirai T, Murakami R, Nakamura H, et al. Prognostic value of perfusion MR imaging of highgrade astrocytomas: long-term follow-up study. AJNR 28; 29: Caseiras GB, Chheang S, Babb J, et al. Relative cerebral blood volume measurements of lowgrade gliomas predict patient outcome in a multiinstitution setting. Eur J Radiol 2; 73: Mitchell CL, O Connor JP, Jackson A, et al. Identification of early predictive imaging biomarkers and their relationship to serological angiogenic markers in patients with ovarian cancer with residual disease following cytotoxic therapy. Ann Oncol 2; 2: Miles KA, Charnsangavej C, Lee F, et al. Application of CT in the investigation of angiogenesis in oncology. Acad Radiol 2; 7: Ellika SK, Jain R, Patel SC, et al. Role of perfusion CT in glioma grading and comparison with conventional MR imaging features. AJNR 27; 28: Jain R, Ellika SK, Scarpace L, et al. Quantitative estimation of permeability surface-area product in astroglial brain tumors using perfusion CT and correlation with histopathologic grade. AJNR 28; 29: Lupo JM, Cha S, Chang SM, Nelson SJ. Analysis of metabolic indices in regions of abnormal perfusion in patients with high-grade glioma. AJNR 27; 28: Cenic A, Nabavi DG, Craen RA, Gelb AW, Lee TY. A CT method to measure hemodynamics in brain tumors: validation and application of cerebral blood flow maps. AJNR 2; 2: Roberts WG, Palade GE. Neovasculature induced by vascular endothelial growth factor is fenestrated. Cancer Res 997; 57: Caloglu M, Yurut-Caloglu V, Karagol H, et al. Prognostic factors other than the performance status and age for glioblastoma multiforme: a singleinstitution experience. J BUON 29; 4: Roberts WG, Delaat J, Nagane M, Huang S, Cavenee WK, Palade GE. Host microvasculature influence on tumor vascular morphology and endothelial gene expression. Am J Pathol 998; 53: AJR:2, May 23 W59