MRI-guided Radiotherapy Identies Early Pseudoprogression of Glioblastoma

Background The standard glioblastoma treatment paradigm consists of surgery, pre-radiotherapy MRI, six weeks of chemoradiotherapy, followed by post-radiotherapy MRI and continued adjuvant temozolomide. In a signicant proportion of patients, post-radiotherapy MRI demonstrates tumor enlargement due to either treatment failure (true progression) or treatment response (pseudoprogression). Recently introduced MRI-guided radiotherapy systems obtain daily MRI of glioblastoma patients, and we hypothesized that progression can be identied early during radiotherapy. Methods Fourteen glioblastoma patients underwent tri-Cobalt-60 MRI-guided radiotherapy in 30 fractions over 6 weeks delivered with concurrent temozolomide. The tumor target volume was delineated on MRI before each of the 30 fractions. The images obtained by the 0.35 T MRI-guided system is shown to be similar to T2-weighted images obtained by a clinical diagnostic MRI-scanner. Hyperintense volumes were measured over time through radiotherapy. Results Four of fourteen patients demonstrated increases of at least 25% and 1.5 cc in T2 hyperintense volume through radiation therapy. This volume expansion correlated with both T2/FLAIR and contrast-enhanced volume expansion on post-radiotherapy diagnostic MRIs. In three of four cases, signicant volume growth only started at week three of treatment, with the most prominent changes occurring during weeks four and ve. While patient numbers are limited, patients with growth during therapy exhibited excellent survival, consistent with the known improved survival of patients with pseudoprogression. Conclusions MRI acquisition during radiotherapy early pseudoprogression typically starting during week 3 or 4 of treatment. This and other daily MRI techniques during radiotherapy could enable early adaptation of glioblastoma


Abstract
Background The standard glioblastoma treatment paradigm consists of surgery, pre-radiotherapy MRI, six weeks of chemoradiotherapy, followed by post-radiotherapy MRI and continued adjuvant temozolomide. In a signi cant proportion of patients, post-radiotherapy MRI demonstrates tumor enlargement due to either treatment failure (true progression) or treatment response (pseudoprogression). Recently introduced MRI-guided radiotherapy systems obtain daily MRI of glioblastoma patients, and we hypothesized that progression can be identi ed early during radiotherapy.
Methods Fourteen glioblastoma patients underwent tri-Cobalt-60 MRI-guided radiotherapy in 30 fractions over 6 weeks delivered with concurrent temozolomide. The tumor target volume was delineated on MRI before each of the 30 fractions. The images obtained by the 0.35 T MRI-guided system is shown to be similar to T2-weighted images obtained by a clinical diagnostic MRI-scanner. Hyperintense volumes were measured over time through radiotherapy.
Results Four of fourteen patients demonstrated increases of at least 25% and 1.5 cc in T2 hyperintense volume through radiation therapy. This volume expansion correlated with both T2/FLAIR and contrastenhanced volume expansion on post-radiotherapy diagnostic MRIs. In three of four cases, signi cant volume growth only started at week three of treatment, with the most prominent changes occurring during weeks four and ve. While patient numbers are limited, patients with growth during therapy exhibited excellent survival, consistent with the known improved survival of patients with pseudoprogression.
Conclusions Daily MRI acquisition during radiotherapy identi es early pseudoprogression typically starting during week 3 or 4 of treatment. This and other daily MRI techniques during radiotherapy could enable early adaptation of therapy in glioblastoma patients.

Background
Glioblastoma (GBM) is the most aggressive brain tumor in adults (1). Standard clinical treatment for GBM consists of tumor resection followed by six weeks of radiation therapy (RT) with temozolomide, resulting in a median survival of about 15 months (2). Magnetic Resonance Imaging (MRI) following the initial six week treatment period reveals gadolinium-enhanced volume increase in up to 49% of patients (3). Of patients with this volume increase between pre and post radiotherapy scans, about one-third result from progressive tumor growth and two-thirds are due to therapy changes termed pseudoprogression (4)(5)(6). True progression is characterized by continuous tumor growth despite treatment with a poor median survival ~ 10 months; while pseudoprogression re ects therapy response with recruitment of blood vessels and/or necrosis and improved median survival ~ 38 months (3,7). No current clinical modality can reliably separate true and pseudoprogression at early time points. Therefore, patients are typically managed without therapy change for many additional months to determine whether changes are progressive (true progression) or spontaneously improve (pseudoprogression).
Early differentiation of true and pseudoprogression would enable modi cation of radiotherapy or chemotherapy and prevent unnecessary surgeries (8). Brain biopsies are invasive, and thus there are strong efforts towards using non-invasive methods such as MRI to predict eventual patient survival based on early changes (8,9). Typically, MRI is obtained approximately two weeks before beginning a six weeks course of chemoradiation and then again four weeks after completion, for a total time interval of three months between scans. Previous studies have assessed different modalities of MRI such as T1weighted (10), T2-weighted (11), perfusion (12) diffusion weighted imaging (DWI) (13), and radiomics (14) at these several month intervals. However, there is limited data analyzing changes that occur during the radiotherapy treatment time interval because it has not been feasible to perform frequent MRI scans due to the cost and logistics of standalone MRI scans. Therefore, the kinetics of change during radiotherapy has not been fully elucidated.
Recently, the combination of MRI and radiotherapy (MRI-RT) devices has become available, providing the opportunity to acquire daily MRI for radiotherapy patients (15,16). We observed that GBM patients treated with MRI-RT at our institution had a variable time course of changes during treatment that we hypothesize might provide additional information for predicting treatment outcome and adapting radiotherapy treatment. Here we provide our preliminary experience with 14 GBM patients treated on the MRI-RT device and describe the visualized changes during radiotherapy.

Patients
Fourteen glioblastoma patients, ten men and four women ages 24-78, underwent RT delivered by tri-

Radiation therapy planning and delivery
Two to four weeks after surgery patients underwent a pre-treatment diagnostic 1.5 T or 3 T MRI including T1-weighted after gadolinium contrast and T2-weighted scans. Simulation was performed at that time on the tri-cobalt MRIdian system (ViewRay, Mountain View, USA) in position for treatment including MRI compatible thermoplastic mask (Civco, Coralville, USA). Vendor supplied 6-channel surface array posterior torso coil was placed under the masking system and 5-channel surface array anterior head and neck coil was wrapped around the mask anteriorly. Patients received 60 Gy in 30 fractions in a single volume with clinical target volume (CTV) expansion of 2 cm on the cavity and enhancing tumor and planning target volume expansion of 0.3 cm concurrent with temozolomide 75 mg/mg 2 using MRI-RT with daily MRI (2) .

Imaging acquisition
The radiation treatment system is equipped with a 0.35 T MRI (Siemens Avanto, version syngo MR B17, IDEA version VB19, Siemens Medical Solutions, Erlangen, Germany). The scanner includes a balanced steady-state free precession (bSSFP) pulse sequence (17). The bSSFP is commonly used in relatively low eld MRI acquisition due to its capability of providing an increased signal-to-noise ratio with higher temporal resolution (18,19). The signal originated by the bSSFP sequence is also known for its mixed composition that is weighted by T1 and T2. However, the use of high ip angles produces high T2weighting as implemented on this 0.35 T system (17). The bSSFP-derived images were acquired with the

Image analysis
The volume of hyperintensity on the bSSFP images from each treatment fraction was contoured manually using MIM (version 6.7, Cleveland, USA). Tumors were reported as a signi cant change if the volume changed at least 25% (similar to Macdonald's criteria for progressive disease (20)) between any two time points (e.g. fraction 1 compared to fraction 30) and that change was greater than 1.5 cc as a threshold to ensure no changes were simply due to contouring variability.
Patients received post-treatment MRI 3-6 weeks after completion of concurrent chemoradiotherapy and continued adjuvant chemotherapy. Patients with observed T2-weighted growth on this study were all noted by post-treatment radiology report to have growth of enhancement and edema on post-treatment MRI.

Results
Four out of fourteen patients (1, 2, 3 and 8) demonstrated an increase of at least 25% and 1.5 cc in hyperintense volume during radiation therapy. All of these patients were con rmed to have possible pseudoprogession or true progression by inspection and radiology report on post-treatment postgadolinium MRI.
Patient 1 (Fig. 2), a 65 year old man, presented to radiation oncology with multifocal glioblastoma (IDH-1 mutation negative, MGMT nonhypermethylated) in the right temporal lobe after biopsy of a discrete lesion about 1 cm inferior to the axial slice shown in Fig. 2A. A tiny (~ 2 mm) area of equivocal T2 change was observed as shown, and this area was included in the 60 Gy CTV. Little change was observed in this area until about fraction 18 (late week 4), at which time rapid growth was seen until fraction 26 (week 6) when the size plateaued. No steroids were used at any time during the radiotherapy treatment. The observed ndings during treatment were consistent with one-month post treatment diagnostic imaging ( Fig. 2A, bottom two panels). Bevacizumab was added to temozolomide after the diagnostic MRI with dramatic improvement in MRI ndings. The patient died of multi-focal glioblastoma, mostly out of the initial radiation therapy eld, 24 months after biopsy. Patient 2 (Fig. 3), a 47 year old man, underwent gross total resection of a right temporal glioblastoma (IDH-1 mutation negative, MGMT nonhypermethylated). On treatment planning MRI two weeks later a tiny area of enhancement without surrounding edema was observed in the right frontal lobe suspicious for multifocal disease and added to the 60 Gy CTV. Post-operative dexamethasone taper was stopped at fraction 1 and no further steroids were given during treatment. Similar to the pre-treatment T2 images, no change in the multi-focal area was visualized on the MRI-RT scans initially. Growth was identi ed on MRI-RT early in week 3 which accelerated through week 4 and 5 and plateaued in week 6. This area continued to progress and underwent resection about 6 months after completion of radiation. Pathology demonstrated only reactive changes and in ammation with no clear tumor. The patient went to hospice 20 months after initial resection due to additional foci of progression outside of the radiation therapy eld.
Patient 8 (Fig. 4), a 59 year old man, underwent gross total resection of a right parietal glioblastoma (IDH-1 mutation negative, MGMT nonhypermethylated). Patient was tapered from dexamethasone 4 mg twice daily and maintained at 2 mg twice daily after the start of treatment due to left hand weakness.
Signi cant growth of T2 hyperintensity was identi ed starting fraction 17 and progressing nearly linearly through fraction 30. Temozolomide chemotherapy was continued through currently 9 months of followup.
Among the four patients with 25% or more growth during treatment, we observed non-linear growth kinetics ( Nine of the fourteen patients are alive after nishing the chemoradiation treatment. Median follow-up was 19.1 months (minimum 6 months) with estimated survival of 23.9 months (Fig. 6). A non-signi cant difference in estimated survival is seen at 18 months (100% for patients with growth during treatment vs. 67% without, p = 0.45), which supports the likely pseudoprogression observed in these patients.
Regarding patients with more minor changes, one patient (patient 7 shown in Fig. 1) had a 23% decrease in bSSFP hyperintense volume beginning at fraction 18 (35.7 cc) trending nearly linearly to fraction 30 (29.0 cc). Patient 5 had a 15% increase in T2 hyperintense volume almost linearly starting at fraction 12 (23.5 cc) and plateauing after fraction 19 (26.8 cc). Patients 11 and 13 underwent gross total resections and the resection cavities were observed to shrink from 1.9 to 0.6 cc and 82.6 to 72 cc, respectively. All of these changes were also reported on diagnostic radiology report post-treatment. The remaining patients had less than 10% or 1.5 cc change of hyperintense volume throughout the treatment period.

Discussion
This is the rst analysis of daily MRI changes during radiotherapy of glioblastoma. Four (29%) of fourteen patients were observed with at least 25% and 1.5 cc of growth of the T2-weighted volume during radiotherapy, consistent with known rates of progression on MRI comparing pre-treatment MRI and MRIs obtained in the 1 month post-treatment period (21). Only one patient was observed with reduction in volume. In these patients where changes were present, the majority of patients exhibited the bulk of changes late during the treatment course-starting in week 3-4 and becoming most prominent in weeks  (24). This is encouraging that midpoint MRI or spectroscopy may be useful to determine prognosis and potentially adapt treatment. Nevertheless, our results suggest that acquiring MRI early in the treatment course may be sub-optimal for correlations with eventual outcome. For example, Patients 1, 2 and 8 ( Fig. 5A, B and D, respectively) only showed T2-weighted volume increase after the fourth week of treatment, indicating that studies analyzing changes during week three or earlier may be premature.
Additionally, there is signi cant heterogeneity in treatment planning margins used for treatment of glioblastoma (25). While we have planned these cases based on EORTC margins (single 2 cm CTV expansion from resection cavity and enhancement to 60 Gy) (26), the Adult Brian Tumor Consortium (ABTC) uses the typically much smaller margins of 5 mm CTV expansion from edema to 46 Gy and 5 mm CTV expansion on cavity and enhancement to 60 Gy (27). Groups using the ABTC margins have reported no difference in treatment failure rates (28)(29)(30), though this is not without controversy (31).
Interestingly, in our study we found that in three of fourteen patients by the end of week 4 (fraction 20) the edema had expanded at least 5 mm from its initial extent (+ 5 mm, + 1 cm, + 1.1 cm). Although the clinical signi cance of these ndings with regards to treatment margin or dose alterations is unclear, it can now be investigated prospectively since such growth during radiotherapy is now identi able.
Our study is limited by single pulse sequence acquisition during radiotherapy. Studies are underway at multiple institutions to obtain multiparametric MRI longitudinally on the MRI-RT system to address this question (32,33). In particular, most criteria for growth and progression are at least partly based on gadolinium enhancement (34). We did not give extra gadolinium to patients during radiotherapy in this preliminary study given recent concerns about gadolinium toxicity (35). However, given these ndings we do plan to dose gadolinium for 1-2 additional timepoints during RT in future patients for response to therapy and radiotherapy treatment adaptation as well as investigate non-invasive perfusion methods during RT (36,37). Nevertheless, the association of edema during radiotherapy with post-treatment contrast volume increase suggests that the two are related in this study. Another possible confounding is that both enhancement and edema volume can be sensitive to changes in steroid doses (38). Our default is not to use steroids in patients during radiotherapy unless the patient is symptomatic, and most did not receive steroids after the start of treatment. Thus, we have found no clear correlation between steroids and volume as noted in the case descriptions.

Conclusions
We identi ed probable pseudoprogression in four of fourteen glioblastomas during MRI-guided primary six week chemoradiotherapy by daily T2-weighted MRI volumes. In three of four cases, signi cant volume growth was not observed until at least week three of treatment, with the most prominent changes occurring during weeks 4 and 5. Further studies are underway to identify whether the kinetics of volume growth, changes in multiparametric MRI, or radiomics changes of daily MRI during radiotherapy may correlate with overall survival in an increased number of patients towards early adaptation of therapy in glioblastoma patients. The studies herein were approved by the University of Miami Institutional Review Board 20160817. All patients gave informed consent.

Consent for publication
Consent for publication is not applicable. Included data is not identi able to any individual person.

Availability of supporting data
The datasets used during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
The project described was supported by the National Cancer Institute and National Center For Advancing Translational Sciences of the National Institutes of Health under Award Numbers K12CA226330 and UL1TR002736. The authors would also like to thank the Sylvester Comprehensive Cancer Center for their support of this work. These funders provided salary support for some of the authors and for study coordination, but did not contribute to study design, collection, analysis, or interpretation of data. The content is solely the responsibility of the authors and does not necessarily represent the o cial views of the National Institutes of Health or Sylvester Comprehensive Cancer Center. Publication fees were supported by Viewray, Inc. Viewray, Inc. did not contribute to study design, collection, analysis, or interpretation of data.
Authors' contributions EAM conceived, designed, and substantially revised this study.
KJ, SD, and DM analyzed the data and co-wrote the manuscript.
JCF and TD substantially contributed to acquisition of the data.
RS, MG, and TD substantially contributed to data analysis and interpretation Contrast and resolution are similar between the two scans. The yellow arrows indicate the MRI nding.