Advances in MRI Assessment of Gliomas and Response to Anti-VEGF Therapy

The use of perfusion imaging as a biomarker for response to antiangiogenic drugs has generated significant interest. There are no validated biomarkers for antiangiogenesis that are currently available for clinical use. The two most common methods of perfusion imaging are dynamic susceptibility contrast (DSC) imaging and dynamic contrast-enhanced (DCE) imaging. DSC imaging can be used to generate maps of relative cerebral blood volume (rCBV), relative cerebral blood flow (rCBF), and mean transit time. Other DSC parameters include relative peak height, which is correlated with rCBV, and percentage of signal intensity recovery (PSR), which is a marker for capillary permeability. DCE imaging is generally used to measure K-trans, a constant for the rate at which contrast material moves from the vasculature to the extracellular space. K-trans is dependent upon the surface area of capillaries, as well as the leakiness of the capillaries. Thus, K-trans is often used as a marker for the permeability of tumor vasculature. Using these parameters, many groups have investigated the role of perfusion imaging in evaluating gliomas. Table 2 provides a summary of the key findings from recent magnetic resonance (MR) perfusion studies. Hirai et al. [8] studied 49 patients with high-grade astrocytomas prior to the initiation of radiation and chemotherapy and found that rCBV has prognostic value. Using a multivariate regression model to control for age and other clinical features, maximum rCBV was found to be an independent prognostic factor of survival: the 2-year survival rate was 67% for patients with a maximum rCBV below 2.3 and 9% for patients with a maximum rCBV above 2.3. Even when controlling for astrocytoma grade, maximum rCBV remained prognostic.

Several studies have focused on using perfusion imaging to assess response to antiangiogenic therapy (ie, as a predictive rather than prognostic indicator). Sawlani et al. [9] studied 16 patients with recurrent GBM treated with bevacizumab and evaluated mean rCBV, mean leakage coefficient, and hyperperfusion volume (HPV), which is defined as the fraction of tumor with an rCBV above a prespecified threshold, as potential biomarkers for treatment response. They found that the percent change in HPV (using an rCBV threshold of 1.00) from baseline to first follow-up had a statistically significant hazard ratio of 1.077 when correlated with time to progression. Although HPV may have potential as a predictive biomarker, the hazard ratio was only slightly greater than 1, and the sample size was small; thus, additional refinements are needed to improve the clinical utility of this technique.

Sorensen et al. [10] studied 31 patients with recurrent GBM treated with cediranib, an inhibitor of the VEGF receptor tyrosine kinases. They calculated a “vascular normalization index” by combining K-trans, microvessel volume, and circulating collagen IV. A prior study that used a mouse model to evaluate the effect of cediranib found that edema alleviation was associated with a transient increase in circulating collagen IV, and it was thought that the increase in circulating collagen IV was reflective of basement membrane thinning [11]. The authors found that this new index (measured 1 day after treatment initiation) was predictive of overall and progression-free survival. The combination of imaging data with serum assays is a novel avenue of increasing the sensitivity and specificity of markers that may prove clinically useful.

Cha et al. [12] studied 18 patients with recurrent malignant gliomas treated with both thalidomide (an antiangiogenic agent) and carboplatin and found that changes in rCBV were more strongly correlated with treatment response based on clinical status than enhancing tumor size. In a multicenter trial with 36 patients with malignant gliomas treated with cilengitide (an angiogenesis inhibitor), Akella et al. [13] found that rCBF correlated with radiographic and clinical response. The correlation between rCBV and radiographic and clinical response approached significance. Both studies, as well as that from Sorensen et al. [10] discussed above, suggest that that perfusion imaging may be useful in the early assessment of response to angiogenesis inhibitors.

Perfusion imaging has also been investigated as a potential biomarker for prognosis following standard treatment (ie, without antiangiogenic therapy). In a retrospective study with 36 GBM patients treated with radiation and temozolomide, Mangla et al. [14] found that the percentage change in rCBV from the pre- and post-treatment measurements (> 5% or ≤ 5%) was predictive of 1-year survival with a sensitivity of 90% and a specificity of 60%. Patients with increased rCBV had a median survival of 235 days, whereas those with decreased rCBV had a median survival of 529 days. In a larger study with 189 patients with low-grade and high-grade gliomas, Law et al. [15••] found that rCBV could be used to predict time to progression. Patients with rCBV less than 1.75 had a median time to progression of 3585 days, whereas patients with rCBV greater than 1.75 had a median time to progression of 265 days, indicating that survival in both high- and low-grade gliomas can be stratified by rCBV. For low-grade gliomas, increased rCBV may be a marker for degeneration into high-grade tumors, but Caseiras et al. [16] demonstrated that tumor growth on T1-weighted imaging is actually a stronger predictor of malignant transformation than change in rCBV. Cao et al. [17] found that baseline as well as changes in rCBV were predictive of survival in 23 patients with high-grade gliomas. Thus, a number of studies have shown that high rCBV or increasing rCBV portends a poorer prognosis across multiple tumor grades. How much value this adds to standard MRI remains to be definitively determined.

The use of perfusion imaging to differentiate true progression from pseudoprogression has also been an area of interest. Barajas et al. [18••] investigated DSC parameters as a means of differentiating recurrent GBM from radiation necrosis (verified by histopathology) in 57 patients. They found that mean, maximum, and minimum relative peak height and rCBV were significantly greater in patients with recurrent tumor than in patients with radiation necrosis. In addition, mean, maximum, and minimum PSR values were significantly lower in patients with recurrent tumor than in those with radiation necrosis. However, there was substantial overlap in rCBV values, which presents challenges when applying a threshold to an individual patient. Hu et al. [19] examined the use of rCBV in differentiating high-grade glioma recurrence from post-treatment radiation effect in 13 patients. From these patients, they obtained 40 tissue specimens. Based on histopathology, 16 were identified as post-treatment radiation effect, and 24 were identified as recurrent glioma. In the post-treatment radiation effect group, rCBV values ranged from 0.21 to 0.71. In the glioma recurrence group, rCBV values ranged from 0.55 to 4.64. Hu et al. [19] found that an rCBV threshold value of 0.71 optimized the differentiation of the two groups. This threshold had a sensitivity of 91.7% and a specificity of 100%. Sugahara et al. [20] also examined the ability of rCBV measurements to discriminate tumor recurrence from radiation treatment effect in mostly low- and high-grade gliomas, examining new enhancing lesions that developed within irradiated regions. rCBV ratios were greater than 2.6 in cases of recurrent tumor and less than 0.6 in cases of post-treatment radiation effect. Thus, differentiating tumors from radiation necrosis likely is dependent on typical rCBV values of various tumor types. Tumors with higher rCBV (eg, grade IV tumors and oligodendrogliomas [21]) may be more easily distinguishable from radiation necrosis than tumors with lower rCBV.

Although the reports from Barajas et al. [18••] and Hu et al. [19] both used DSC perfusion protocols, the ranges are lower for both tumor and radiation necrosis in the Hu et al. paper. Part of this difference may be due to the inclusion of both grade III and IV tumors in the report by Hu et al. [19], but it also is possible that differences in acquisition, processing, and MRI platforms may result in different thresholds across different sites even when using similar protocols. Variability in postprocessing methods can make comparisons across sites difficult [22]. Therefore, it will be important to establish standardized protocols to allow for more meaningful comparisons, particularly for the use in multicenter trials.

Most MR perfusion protocols use gadolinium-based contrast agents. However, Gahramanov et al. [23] argue that due to its tendency to extravasate into the extracellular space, gadolinium is a suboptimal contrast agent and may actually underestimate rCBV in some patients with progressive disease. They support the use of an iron oxide nanoparticle called ferumoxytol, which has chemical characteristics that are much different than gadolinium-based agents. It is thought that ferumoxytol, due to its larger molecular size, will exhibit significantly less extravasation from leaky vessels into the interstitium and thus allow for a more accurate measurement of rCBV. In their study, 14 GBM patients with new enhancing lesions after radiation therapy underwent DSC imaging with both gadoteridol and ferumoxytol. In all seven cases of clinically presumptive tumor recurrence, ferumoxytol-DSC revealed high rCBV values. In contrast, gadoteridol-DSC (a gadolinium-based agent) showed high rCBV values for only 4 of the 7 cases. Based upon these results, it appears that ferumoxytol-DSC is more sensitive in detecting true progression than gadoteridol-DSC. However, this small study relied upon clinical status and MRI changes to identify the enhancing lesion, and thus histopathologic confirmation of these findings would be desirable.

Diffusion-weighting imaging (DWI) is another MRI sequence under investigation in the assessment of gliomas. DWI is based on the movement of water molecules, and the values from these sequences are reported as the apparent diffusion coefficient (ADC). Lower ADC values reflect lower (more restricted) diffusion. Several physiologic properties of tumors may influence ADC values. Water molecules are generally more restricted in their movement within cells and less restricted in the extracellular space. Because necrosis involves degradation of cellular integrity, it is thought that necrosis increases ADC. In a similar way, edema increases interstitial fluid, thereby increasing ADC. Conversely, increased cellular density lowers ADC by restricting diffusion. Because of these relationships, DWI has been studied as a means of evaluating the effects of therapy on malignant gliomas.

The functional diffusion map (fDM) was developed to take advantage of the relationship between ADC and cell density by examining voxel-wise changes in ADC measured in the same patient over time [24‐27]. By examining voxel-wise change in ADC, the sensitivity of detecting subtle changes in tumor cell density is dramatically increased. This technique has primarily been applied as a tool to predict response to cytotoxic chemotherapy and radiotherapy within the contrast-enhancing tumor bed [24‐26]; however, recent studies have demonstrated its utility outside regions of contrast enhancement [27‐29] and as a tool for studying the effects of anti-VEGF therapy [30, 31]. Specifically, Ellingson et al. [31] found a significant decrease in ADC immediately following treatment with bevacizumab; however, the rate of change of tissue showing abnormally low ADC (fDM-classified “hypercellular” tissue) within regions of T2 signal abnormality was shown to be an early predictor of tumor progression, time to progression, and overall survival. In other words, if the fDM-classified hypercellular regions (low ADC) are expanding rapidly there is a higher likelihood of tumor recurrence and death, regardless of the specific treatment paradigm. Figure 1 demonstrates a patient with a recurrent GBM treated with bevacizumab and temozolomide illustrating substantial growth of hypercellular tumor regions prior to standard radiographic tumor progression.

Bevacizumab treatment has been noted to produce persistent diffusion restriction. Rieger et al. [32, 33] found that 13 out of 18 patients with recurrent malignant gliomas who received bevacizumab therapy exhibited diffusion-restricted areas, along with decreased ADC values. Areas of diffusion restriction had decreased rCBV. A biopsy of one of the diffusion-restricted areas revealed atypical necrosis with an upregulation of nuclear hypoxia-inducible factor-1α. These findings suggest that prolonged antiangiogenic therapy can result in chronic hypoxia, which, in turn, can lead to diffusion restriction. Gerstner et al. [34] presented a case report in which a patient with GBM experienced persistent diffusion restriction after bevacizumab treatment, and a biopsy revealed infiltrative tumor. Gerstner et al. [35] also observed areas of persistent diffusion restriction after the initiation of antiangiogenic therapy. They studied 30 patients with recurrent GBM treated with cediranib and found that this treatment resulted in lower ADC values. The percent volume of FLAIR hyperintensity comprised by low ADC increased significantly from baseline (2.3%) to day 28 (2.9%), day 56 (5.0%), and day 112 (6.3%) of cediranib therapy. They suggest that these findings are indicative of increasing infiltrative tumor. However, histopathologic correlation was not available. Because the studies by Gerstner et al. [34, 35] and Rieger et al. [32, 33] point to two different pathophysiologic processes that may explain persistent diffusion restriction, further investigation is needed to determine whether diffusion restriction is due to antiangiogenic effects of bevacizumab or whether it is due to tumor recurrence. Potentially, the degree of diffusion restriction could differentiate infiltrative tumor from hypoxic change.

Because ADC is influenced by cellular density, necrosis, and edema, ADC values can potentially be used as a noninvasive surrogate for VEGF expression and thus susceptibility to bevacizumab. We evaluated ADC histogram analysis as a predictive biomarker of response to bevacizumab in patients with recurrent GBM [36••]. Regions of enhancing tumor were segmented and mapped onto ADC images for ADC histogram generation. Tumors with low ADC values prior to initiation of bevacizumab were more likely to progress by 6 months compared to those with high ADC values. This histogram analysis was 72.5% accurate in predicting 6-month progression-free survival, and it was more accurate than the Macdonald criteria at first follow-up. In the non–bevacizumab-treated cohort, ADC values were not predictive of median survival. If these findings are confirmed in a larger study, this would be the first validated MRI-based predictive biomarker for response to bevacizumab in patients with GBM.

MR spectroscopy is a technique that allows for the characterization of the chemical makeup of a particular region of interest. Much of the research surrounding MR spectroscopy has focused on its potential ability to differentiate true progression from pseudoprogression. Smith et al. [37] studied 33 patients with primary intracranial neoplasms and found that ratios of brain metabolites can assist in differentiating recurrent tumor from post-treatment radiation effect. They found that patients with recurrent tumor had elevations in the choline (Cho)/creatine (Cr) and Cho/N-acetylaspartate (NAA) ratios compared to those with post-treatment radiation effect. Patients with recurrent tumor also experienced a decrease in the NAA/Cr ratio compared to those with post-treatment radiation effect. The Cho/NAA ratio had a sensitivity of 85% and a specificity of 69% in detecting true progression. In another study, Weybright et al. [38] examined 29 patients with new contrast-enhancing brain lesions in the vicinity of a previously treated brain neoplasm. Histopathology was used to confirm the identity of the lesion in 16 patients, and imaging and clinical follow-up were used to determine the identity of the lesion in the remaining 13 patients. Consistent with the study conducted by Smith et al. [37], Weybright et al. [38] found that the Cho/Cr and Cho/NAA ratios were significantly greater in tumor recurrence compared with post-treatment radiation effect. The NAA/Cr ratio was significantly lower in recurrent tumor compared with post-treatment radiation effect. Some have suggested that creatine should not be used as a denominator in metabolite ratios because creatine may be anticorrelated with progression-free survival [39, 40]. Although these studies were small, they suggest that MR spectroscopy may be helpful in differentiating true progression from pseudoprogression.

In addition to differentiating true progression from pseudoprogression, the Cho/NAA ratio may also be helpful as a biomarker for response to radiation therapy. Oh et al. [41] studied the predictive value of the Cho/NAA ratio in 28 patients diagnosed with GBM after surgical resection but prior to radiation therapy. They found that patients with a large volume of metabolic abnormality (defined as a Cho/NAA ratio > 2) had a significantly shorter survival time than those with a small volume of metabolic abnormality (12.0 and 17.1 months, respectively). Thus, the Cho/NAA ratio may have predictive value in assessing response to radiation.

Radiotracers used in conjunction with positron emission tomography (PET) scans represent another group of potential biomarkers in assessing gliomas. By coupling radioactive tracers to molecules that are taken up by metabolically active cells, PET scans can detect and localize the presence of tumors and other cells with increased metabolism. A newer radiotracer is the nucleic acid analogue 18 F-fluorothymidine (¹⁸F-FLT), which is used as an indicator of cellular proliferation. Chen et al. [42] studied 25 patients with gliomas and found that ¹⁸F-FLT visualized all high-grade gliomas. In contrast, 18 F-fluorodeoxyglucose (¹⁸F-FDG) did not detect tumor in five patients with recurrent glioma, demonstrating that ¹⁸F-FLT had greater sensitivity than ¹⁸F-FDG. The lower sensitivity of ¹⁸F-FDG-PET in this study may be due to the high background uptake of FDG in the cortex and other metabolically active regions, whereas many other available tracers, including FLT, have little or no uptake in normal brain. In this study, ¹⁸F-FLT-PET was not only predictive of time to progression and survival, but it was also more strongly correlated with the Ki-67 staining (a histopathologic marker of proliferation that has been correlated with clinical outcomes) than ¹⁸F-FDG. In another study, Chen et al. [43] prospectively examined 19 patients with recurrent malignant gliomas treated with bevacizumab and irinotecan. ¹⁸F-FLT scans were performed before the initiation of treatment, after 1 to 2 weeks, and after 6 weeks; response was defined as a reduction in ¹⁸F-FLT uptake by more than 25% after 6 weeks. Those who responded based on ¹⁸F-FLT uptake survived three times as long as those who did not respond (10.8 vs 3.4 months). Both early and late ¹⁸F-FLT responses were significant predictors of overall survival. These studies suggest that ¹⁸F-FLT has significant promise as a biomarker of outcome in patients with GBM, which can be acquired relatively early in the treatment cycle.

Amino acid tracers have also been studied. Potzi et al. [44] studied methyl-¹¹C-l-methionine (¹¹C-methionine) as a means of assessing GBM in 28 patients. Although ¹¹C-methionine uptake detected tumor in 24 out of 28 patients, there was no correlation between uptake and survival outcomes. Because it does not correlate with clinical outcomes, ¹¹C-methionine may not be suitable as a response marker like ¹⁸F-FLT, but may be more suitable for detecting disease recurrence, including nonenhancing tumor. For instance, Galldiks et al. [45] studied 12 patients with recurrent GBM and found that metabolically active tumor volume (defined as ¹¹C-methionine uptake ≥ 1.3) was significantly larger than the volume of gadolinium-enhanced tumor (30.17 vs 13.7 cm³). One drawback of ¹¹C-methionine is its half-life of 20 min. Due to its short half-life, ¹¹C-methionine cannot be used in PET centers without a cyclotron, limiting its clinical utility. ¹⁸F-labeled amino acid analogues with longer half-lives (eg, 3,4-dihydroxy-6-[18]F-fluoro-l-phenylalanine [¹⁸F-FDOPA]) have also been used to assess gliomas. Becherer et al. [46] studied 20 patients with supratentorial brain lesions (18 of which were primary brain tumors) and found that mean standardized uptake ratios were similar for ¹⁸F-FDOPA and ¹¹C-methionine (2.05 and 2.04, respectively). Chen et al. [47] examined the diagnostic accuracy of ¹⁸F-FDOPA in assessing gliomas. They studied 81 patients with brain tumors and found that ¹⁸F-FDOPA had a sensitivity of 98%, a specificity of 86%, a positive predictive value of 95%, and a negative predictive value of 95%. These studies indicate that amino acid tracers can reliably detect gliomas, and their ability to discriminate gliosis from nonenhancing tumor merits additional investigation.