Combined imaging biomarkers for therapy evaluation in glioblastoma multiforme: correlating sodium MRI and F-18 FLT PET on a voxel-wise basis

Abstract

We evaluate novel magnetic resonance imaging (MRI) and positron emission tomography (PET) quantitative imaging biomarkers and associated multimodality, serial-time-point analysis methodologies, with the ultimate aim of providing clinically feasible, predictive measures for early assessment of response to cancer therapy. A focus of this work is method development and an investigation of the relationship between the information content of the two modalities. Imaging studies were conducted on subjects who were enrolled in glioblastoma multiforme (GBM) therapeutic clinical trials. Data were acquired, analyzed and displayed using methods that could be adapted for clinical use. Subjects underwent dynamic [¹⁸F]fluorothymidine (F-18 FLT) PET, sodium (²³Na) MRI and 3-T structural MRI scans at baseline (before initiation of therapy), at an early time point after beginning therapy and at a late follow-up time point after therapy. Sodium MRI and F-18 FLT PET images were registered to the structural MRI. F-18 FLT PET tracer distribution volumes and sodium MRI concentrations were calculated on a voxel-wise basis to address the heterogeneity of tumor physiology. Changes in, and differences between, these quantities as a function of scan timing were tracked.

While both modalities independently show a change in tissue status as a function of scan time point, results illustrate that the two modalities may provide complementary information regarding tumor progression and response. Additionally, tumor status changes were found to vary in different regions of tumor. The degree to which these methods are useful for GBM therapy response assessment and particularly for differentiating true progression from pseudoprogression requires additional patient data and correlation of these imaging biomarker changes with clinical outcome.

Introduction

Although routinely used for some aspects of cancer care, minimally invasive medical imaging procedures, which can be repeated at regular intervals, have a much greater potential for guiding clinical decision-making than is currently realized. For successful clinical translation, imaging protocols must (a) yield relevant information and (b) be feasible under the time and resource constraints imposed by clinical operation. These two requirements can be at odds in that information gained typically comes at the expense of protocol and analysis complexity. Multimodality platforms, e.g., positron emission tomography (PET)/computed tomography (CT) and now magnetic resonance imaging (MRI)/PET, help to ameliorate this situation and have substantially increased the opportunity for the use of multimodality data for therapy management. However, optimal use requires the identification and evaluation of effective biomarkers and the development of efficient acquisition and analysis methods. As part of a project to address these issues, this paper examines the joint use of sodium (²³Na) MRI and PET using [¹⁸F]fluorothymidine (F-18 FLT) to evaluate therapy effect in patients with glioblastoma multiforme (GBM). As described below, F-18 FLT is expected to trace DNA synthesis, and therefore cellular proliferation, but is complicated by blood–brain barrier (BBB) effects [1]. Sodium MRI reveals tissue sodium concentration (TSC), changes in which have also been associated with cellular proliferation in general [2] and in brain tumors specifically [3]. However, its interpretation is complicated [4] since the ²³Na signal derives from a combination of intra- and extracellular contributions [3], [5]. Thus, the two modalities are expected to have both overlapping and complementary components that could be exploited to improve patient management in GBM.

Glioblastoma multiforme, a grade IV astrocytoma, accounts for 50%–60% of primary brain tumors diagnosed in adults each year in the United States with an incidence in North America of 3.0 per 100,000 population. GBM has an estimated 2-year survival rate of 8.7% in the absence of therapy [6]. Even with treatement, prognosis is poor since the recurrence rate is high; however, there has been modest improvement since 2005 when the standard-of-care treatment for newly diagnosed GBM changed from radiation therapy (RT) alone to radiation therapy plus temozolomide (TMZ) chemotherapy (RT-TMZ) [7]. Patients treated with optimal therapy, including surgical resection, radiation therapy and chemotherapy, have a median survival of approximately 12 months, with fewer than 25% surviving for 2 years and fewer than 10% surviving for 5 years [8]. The increasing array of therapy options (e.g., bevacizumab) and the rapid advances being made in the development of targeted therapeutic approaches coupled with the narrow window for effective GBM management decisions provide a strong impetus to develop noninvasive, molecular, event-specific imaging approaches for early evaluation of therapy. Additionally, such methods can play a key role in expediting clinical therapy trials.

The use of CT and MRI for monitoring of tumor structural properties is well established, and standards have been developed for therapy response characterization based on tumor size measurements, e.g., Response Evaluation Criteria In Solid Tumors (RECIST). Likewise, the use of PET with [¹⁸F]fluorodeoxyglucose (F-18 FDG) for measurement of tissue glucose metabolism can be useful for initial tumor staging, restaging and, in some cases, therapy response monitoring. However, monitoring of brain tumors presents special challenges. For example, a transient increase in the bidimensional [9] or unidimensional (RECIST) size of the contrast-enhancing lesion in structural MRI scans, sometimes known as “pseudoprogression” (as opposed to true progression), has been reported in patients undergoing MRI at the standard clinical assessment time of 4 weeks after the completion of radiotherapy (e.g., Ref. [10], [11]). For patients on the widely used Stupp regimen [12], it is precisely this time point at which a follow-up MRI scan is performed that serves as the baseline for further adjuvant therapy. Thus, determining whether an apparent radiological worsening is true tumor progression or pseudoprogression is critical to the decision whether or not to continue current chemotherapy or to change to another treatment regimen. The utility of F-18 FDG PET for this purpose is limited due to high normal brain uptake that makes assessment of low metabolic areas of treated but viable tumor or small regions of tumor recurrence difficult. This limitation is exacerbated by reduced uptake in radiation- or chemotherapy-induced vasogenic edema in the tumor region that can be incorrectly interpreted as a beneficial response [13]. At present, there are no validated imaging criteria for differentiation of pseudoprogression from true progression. Sodium MRI and F-18 FLT PET are potential tools for evaluation of therapy response in GBM.

TSC can be expressed as:

$$\text{TSC}=\left[{\text{Na}}^{\text{i}}\times \text{CVF}\right]+\left[{\text{Na}}^{\text{e}}\times \left(1-\text{CVF}\right)\right]\text{,}$$

where Naⁱ and Na^e are the concentrations of intracellular and extracellular sodium, respectively, and CVF is the tissue cell volume fraction. For typical values of the above parameters (e.g., [Naⁱ] approximately 40 mM, [Na^e] approximately 140 mM and CVF approximately 70%), a typical tumor TSC is 70 mM, which is in agreement with reported values in the literature.

In neoplastic tissue, sustained depolarization of the cell membrane precedes the high rate of mitotic activity that characterizes abnormal cell growth [14], [15] and leads to an increase in Naⁱ that has been demonstrated in a number of human neoplasms [2], [16], [17]. Further characterization of Naⁱ rise in several types of human carcinomas [16], [17] and glial cell lines [18], [19] has established a positive correlation between proliferative activity and increased intracellular Na+:K + ratio (mostly due to an increase in Naⁱ). Once neoplastic changes are set in the tissue, a decrease in CVF is observed due to the different morphology of neoplastic cells. The combined effect of decreasing CVF and increasing intracellular sodium content is a net increase in TSC.

However, other changes in physiology affect measured TSC. Specifically, once a cell's membrane becomes compromised, e.g., in cell death or near cell death, its ability to maintain an intra- to extracellular sodium gradient diminishes or disappears, and an increase in TSC should be observed. A rise in TSC to Na^e levels is an indication of a region of lysed cells.

Thus, it is clear that changes in TSC indicate important oncologic/therapy-related changes in tissue status. However, a precise attribution of physiologic meaning of changing TSC in tumor is not always possible without assumptions, a priori knowledge or additional data. For example, a plausible assumption is that large changes in TSC observed within days of therapy are likely due to therapy-induced cellular membrane compromise. A goal of this work is to develop methods for effective use of ²³Na MRI in conjunction with F-18 FLT PET and to demonstrate correlations between combined results of both modalities and patient outcome.

F-18 FLT is a pyrimidine analogue that reflects the activity of a thymidine kinase-1 during S phase of DNA synthesis [20]. F-18 FLT was introduced by Shields and colleagues for PET imaging of tumor proliferation in animals and humans [21]. Several studies have investigated the role of F-18 FLT in detecting and staging of different malignancies, such as lung cancer, colorectal cancer, melanoma, soft tissue sarcoma, breast cancer and brain tumors [22], [23], [24], [25], [26], [27], [28], [29].

As a marker of proliferation, F-18 FLT is considered a potential tool for monitoring anticancer treatment in solid tumors [30], [31], [32], [33], [34], [35]. In the case of brain tumors, significant correlations have been observed between F-18 FLT uptake and Ki-67 proliferation index [13], [36], [37], [38], [39], [40], [41]. Moreover, a 2007 study performed by Chen et al. [42] demonstrated that F-18 FLT PET scans of GBM patients receiving bevacizumab/irinotecan treatment showed a 25% reduction in uptake at 6 weeks compared to 2-week early therapy response assessment scans. Furthermore, the same study found that responders, as measured by F-18 FLT, survived three times as long as nonresponders (10.8 vs. 3.4 months) and tended to have a prolonged progression-free survival [42]. Finally, F-18 FLT PET was found to be a more significant predictor of overall survival than contrast-enhanced MRI alone [42].

An important consideration is that F-18 FLT does not readily cross an intact BBB, and it is necessary to account for this in the interpretation of results [1]. Nevertheless, F-18 FLT may be useful in cases where MRI contrast enhancement, typically associated with BBB breakdown, is observed [1]. This is exactly the situation present when the task is to distinguish pseudoprogression from progression. Nevertheless, for such use, it is necessary to verify that the tracer has access to the brain tissue in question. A possible method for achieving this, using scan data acquired during the initial tracer wash-in phase (e.g., within the first 2 min after tracer injection), is examined in this paper.

The purpose of this work is to investigate correlations between ²³Na MRI and F-18 FLT PET biomarkers using clinically feasible data analysis and visualization methods necessary for assessing GBM therapeutic response. We evaluate these correlations through voxel-wise comparisons, necessary to account for GBM heterogeneity, both at baseline and in response to therapy.