Theoretically, the extent of resection and the dose and extent of radiation should be determined according to the pathologic tumor grade of the glioma, in addition to prognostic factors such as age, Karnofsky scores, et cetera. Unfortunately, an innate and perhaps unavoidable limitation of the current tumor grading approaches lies in the vagueness of the descriptions used to define each grade. Our study findings indicate that the relative CBV measurement can be used to predict time to progression, clinical outcome, and glioma biology so that, with further optimization and improvements to automate postprocessing and quantitation techniques, the relative CBV measurement may provide an important imaging biomarker of glioma malignancy that may affect therapeutic choices and patient outcome.
In our study, patients with complete response and stable disease had a lower relative CBV than did patients with progressive disease and death. This suggests that the baseline relative CBV value can be used to predict whether a patient is likely to have a poor prognosis. This finding is supported by our finding that Kaplan-Meier estimates of median time to progression indicated that patients with a relative CBV of less than 1.75 had a median time to progression of 3585 days, whereas patients with a relative CBV of more than 1.75 had a time to progression of 265 days irrespective of histopathologic tumor type. When we reviewed both HGGs and LGGs, we found that relative CBV measurements can further aid in the classification of the time to progression. This observation has preoperative implications for the neurosurgeon in that a glioma with the imaging features of an LGG but with a higher relative CBV should perhaps be excised more aggressively, as more extensive resection has been shown to improve survival in glioma patients (29
). Similarly, after a stereotactic biopsy, subtotal resection, or even gross total resection, an HGG or LGG with a high relative CBV may be treated more aggressively with postoperative radiation therapy and chemo-therapeutic regimens.
Researchers in a number of studies have demonstrated the utility of relative CBV in the prediction of patient outcome. In a study of 73 patients with glioma, Schmainda et al (31
) were able to correctly classify 96% of HGGs but only 69% of LGGs by using dynamic contrast-enhanced MR imaging. Similarly, Law et al (21
) found that relative CBV can help to identify LGGs (with high relative CBV) that will progress more rapidly than will LGGs with low relative CBV and that this subset of LGGs has a propensity for malignant transformation. In view of the findings in these studies, it seems highly likely that if the clinical end point rather than histologic type is considered as the reference standard for tumor grading, relative CBV measurements would demonstrate much greater specificity in the prediction of tumor biology than histo-logic type alone. That is, tumors that were classified as LGG histologically but as HGG by using dynamic susceptibility-weighted contrast-enhanced MR imaging were likely to be HGG at the time of the study or had undergone malignant transformation after the histologic assessment was performed. This not only has implications at the time of surgery but, as Cao et al (32
) showed recently, early temporal changes during radiation therapy in heterogeneous regions of high and low perfusion in gliomas might be used to predict different physiologic responses to radiation therapy. This observation also may provide the opportunity to identify tumor subvolumes that are radioresistant and might benefit from intensity-modulated radiation therapy or dose painting.
A potential limitation in our study was the possible effect of different treatment protocols on the time to progression for HGGs and LGGs. We were not able to control for the effects of radiation dose and extent of radiation or a combination of radiation therapy and chemotherapy. Chemotherapy consisted of temozolomide (Temodar; Schering, Kenilworth, NJ), carboplatin (Paraplatin; Bristol-Myers Squibb, New York, NY), procarbazine hydrochloride (Mutulane; Hoffmann-La Roche, Nutley, NJ), irinotecan (Camptosar; Pfizer, New York, NY), or a combination thereof. Currently, there are also no clear data that indicate that any particular treatment regimen produces radically better outcomes than any other, so that even though we have not controlled for it in this study, differences in the treatment protocol should not result in significant differences in patient outcome. It is well known that some lesions respond to certain therapies, whereas other lesions respond to different therapies and still other lesions respond to no therapy at all, even though they are classified as the same grade of tumor (31
A second potential limitation was the possible subjectivity and reproducibility of a region-of-interest–based technique of relative CBV measurements. Although Wetzel et al (28
) have shown acceptable reproducibility with this method, we and others have explored more objective and semiautomated methods, such as histogram analysis, to further improve the inter- and intraobserver reproducibility of relative CBV measurements. A further confounding factor is the inclusion of gliomas with oligodendroglial components in the analysis. Researchers (15
) have well described that oligodendrogliomas generally have increased relative CBV, and our observations also confirm this finding; 50% of LGOs had a relative CBV of more than 1.75 and 36% of LGAs had a relative CBV of more than 1.75. We also analyzed gliomas with and without oligodendroglial components and found that gliomas with oligodendroglial components (LGO, LGOA, AOA) and a high relative CBV of more than 1.75 had a longer time to progression than did gliomas without oligodendroglial components (LGA, AA, GBM) irrespective of tumor grade.
Finally, the definition of progression used in our study was based partly on the use of imaging criteria. With the assumption that some patients underwent radiation therapy, the occurrence of radiation necrosis, misinterpreted as disease progression, may have biased the results. Fortunately, at review of our results, the number of patients that may have biased the results is small, as most patients with progression shown at imaging also had clinical progression.
Current methods for determination of glioma grade, such as the World Health Organization classification, have considerable limitations. A comparison of the reproducibility of the World Health Organization classification versus an automated method for relative CBV measurement, as well as the use of these two methods for the prediction of outcome and survival in patients with gliomas, is under way at our institution. It is likely that improvements in the current World Health Organization classification and incorporation of molecular markers, as well as imaging biomarkers of outcome, in combination will provide the optimal means of predicting glioma biology and, in turn, of determining direct therapy in the future. This is likely to influence the extent of neurosurgical resection, as well as the role of postoperative radiation therapy and chemotherapy, adding relative CBV to current known prognostic factors: age, histologic tumor type, Karnofsky score, and extent of resection (22