Differentiation of radionecrosis from tumor recurrence

Neuro-Oncology 15(12):1732 1738, 2013. doi:10.1093/neuonc/not130 NEURO-ONCOLOGY Extent of perilesional edema differentiates radionecrosis from tumor recurrence following stereotactic radiosurgery for brain metastases Jonathan E. Leeman, David A. Clump, John C. Flickinger, Arlan H. Mintz, Steven A. Burton, and Dwight E. Heron Department of Radiation Oncology (J.E.L., D.A.C., J.C.F., S.A.B., D.E.H.) and Department of Neurological Surgery (A.H.M.), University of Pittsburgh Cancer Institute, Pittsburgh, Pennsylvania Background. Differentiation of tumor recurrence from radionecrosis is a critical step in the follow-up management of patients treated with stereotactic radiosurgery (SRS) for brain metastases. A method that can reliably differentiate tumor recurrence from radiation necrosis using standard MR sequences would be of significant value. Methods. We analyzed the records of 49 patients with 52 brain metastases treated with SRS who subsequently underwent surgical resection of the same lesion. Fortyseven of the lesions had preoperative MRI available for review (90%), including T1 postcontrast, T2, and fluid attenuated inversion recovery sequences. Pre-SRS and preoperative lesion and edema volumes were manually contoured and measured in a blinded fashion using radiation treatment planning software. A neuropathologist analyzed samples for the presence of tumor and/or radiation necrosis. Results. Longer time between SRS and resection (P,.001) and a larger edema/lesion volume ratio (high T2/ T1c, P ¼.002) were found to be predictive of radionecrosis as opposed to tumor recurrence. Using a cutoff value of 10 for the edema/lesion volume ratio, we were able to predict the presence of tumor with a positive predictive value of 92%, which increased to 100% when looking only at patients who underwent resection,18 months following SRS. Conclusions. On follow-up imaging, lesions with a high edema/lesion volume ratio and lesions that progress later after SRS are more likely to contain radionecrosis. These indices may help guide clinical decision making in the context of evolving lesions after SRS for brain metastases and thereby avoid unnecessary interventions. Keywords: brain metastases, magnetic resonance imaging, radionecrosis, stereotactic radiosurgery. Differentiation of radionecrosis from tumor recurrence is a critical step in the follow-up management of patients treated with stereotactic radiosurgery (SRS) for a brain metastasis. Whether progression of the lesion is evident from imaging or evolution of neurological symptoms, accurate diagnosis of the lesion s histology is critical for appropriate management. When tumor recurrence is suspected, surgical resection becomes an important consideration, whereas radiation effects may be managed more conservatively. Multiple studies have attempted to utilize advanced imaging techniques, including PET, single-photon emission CT (SPECT), and MR spectroscopy (MR SPECT) to evaluate progressing lesions, 1 6 but the current standard for follow-up imaging remains traditional MRI. As such, a method that can reliably differentiate tumor recurrence from radiation necrosis using standard MR sequences would be of significant value. In this study, we sought to validate the technique of using T1-T2 match as an indicator of tumor recurrence as well as present a new method for detection of recurrence utilizing the edema/lesion volume ratio. We also hypothesized that other parameters, such as time from radiosurgery until resection, tumor volume, and radioresistant histology, might help in predicting whether the resected specimens show only radiation effect, only persistent/progressing tumor, or a mixture of both. Materials and Methods Received April 29, 2013; accepted July 11, 2013. Corresponding Author: Dwight E. Heron, MD, FACRO, FACR, UPMC Cancer Pavilion, Department of Radiation Oncology, 5150 Centre Avenue, #545, Pittsburgh, PA 15232 (herond2@upmc.edu). Patient and Lesion Characteristics We retrospectively analyzed the charts of 49 patients with 52 lesions treated with SRS for a brain metastasis between May 2005 and August 2011 who subsequently # The Author(s) 2013. Published by Oxford University Press on behalf of the Society for Neuro-Oncology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

underwent surgical resection of the same lesion. Of these, 5 patients were excluded because preop imaging with all necessary MR sequences was unavailable, leaving 44 patients with 47 lesions included in our imaging correlation analysis. All 49 patients were included in the remainder of analyses. Patient and tumor characteristics are presented in Table 1. The median age at the time of SRS was 58 years (range, 29 83). Nineteen patients were male (39%), and 30 were female (61%). The median KPS was 80 (range, Table 1. Characteristics of patients who received SRS for a brain metastasis followed by surgical resection Characteristic Value Patients, n (F/M) 49 (19/30) Lesions, n 52 Lesions with preop MRI, n (%) 47 (90) Median age, y 58 Primary malignancy, n Lung 20 Melanoma 13 Breast 9 Colon 4 Renal 3 Other 3 Median KPS score (range) 80 (70 90) Median number of brain metastases (range) 2 (1 7) Median graded prognostic assessment score 1.5 (0.5 3.5) (range) Recursive partitioning analysis class, n I 27 II 23 III 0 Symptoms at time of surgery, n (%) Symptomatic 41 (79) Asymptomatic 11 (21) Median interval between primary diagnosis and 28.7 (0 247) SRS, mo (range) Median tumor volume, cc (range) 4.6 (0.14 20) Received WBRT, n (%) 11 (22) Prior to SRS 5 (10) After SRS 6 (12) Treatment order, n (%) SRS resection 34 (65) SRS resection SRS 8 (15) Resection SRS resection 10 (19) Repeat SRS, n (%) 20 (38) Median SRS dose, Gy (range) 20.5 (8 29) Treatment fractions, n (%) 1 44 (85) 2 1 (2) 3 3 (6) 5 1 (5) Treatment modality, n (%) Cyberknife 51 (98) Trilogy 1 (2) 70 90). The median number of brain metastases at the time of SRS was 2 (range, 1 7), and 28 patients (54%) had active systemic disease at the time of SRS. The median graded prognostic assessment was 1.5 (range, 0.5 3.5). Twenty-seven patients (54%) were class I in recursive partitioning analysis, 23 (46%) were class II, and none were class III. Primary histologies included lung (n ¼ 20), melanoma (n ¼ 13), breast (n ¼ 9), colon (n ¼ 4), renal (n ¼ 3), endometrial (n ¼ 1), ovarian (n ¼ 1), and germ cell (n ¼ 1). The median time between primary diagnosis and diagnosis of brain metastases was 28.7 months (range, 0 247). Five patients had received whole brain radiation therapy (WBRT) prior to SRS (10%), while 6 patients received WBRT after SRS (12%). Ten lesions were resected, received adjuvant SRS to the resection bed, and were subsequently resected again (19%), while 8 of the lesions received SRS and then subsequent adjuvant SRS after resection (15%). The remaining 34 lesions received SRS followed by resection (65%). A total of 20 lesions received repeat SRS (38%). The median time between SRS and resection was 4.6 months (range, 0.1 27.9). Resection was performed due to lesion progression on follow-up imaging following SRS, worsening of symptoms, or failure of a course of steroids. SRS Treatment Characteristics The median treatment volume was 4.6 cc (range, 0.14 20) treated with a median prescription dose of 20.5 Gy (range, 8 29) to the 80% isodose line. Forty-four of the lesions were treated in a single fraction (85%); 1 was treated in 2 fractions (2%); 3 were treated in 3 fractions (6%); and 1 was treated in 5 fractions (2%). The median tumor coverage was 98.4% (range, 86.6% 100%). Fifty-one lesions (98%) were treated using the Cyberknife Radiosurgery System (Accuray), and 1 (2%) was treated using the Trilogy Radiosurgery System (Varian Medical Systems). Analysis of Pre-SRS Preoperative Imaging Forty-seven (90%) of the lesions had preoperative MRI available for review, including T1 sequences pre- and postcontrast, T2 with and without fluid attenuated inversion recovery (FLAIR) as well as thin-slice contrast-enhanced spoiled gradient recalled acquisition in steady state (SPGR). Lesion volumes were measured using Eclipse radiation treatment planning software (Varian Medical Systems). For T1 and SPGR images, lesion volume was determined by contouring the contrast-enhancing lesion volume on every slice. For T2 and FLAIR images, the outline of the lesion was contoured by distinguishing the area of decreased signal surrounded by the volume of high T2 signal. T2 edema volume was determined by contouring the entire T2 enhancing area surrounding the lesion on each slice. Lesion and edema volumes were similarly contoured and measured for pre-srs images. All lesions were contoured in a blinded fashion without NEURO-ONCOLOGY DECEMBER 2013 1733

regard to the lesion histology. The median number of days between imaging and resection was 5 (range, 0 50). Histological Determination Resected tissue was analyzed by a neuropathologist. Samples were determined to consist of pure recurrent tumor, pure radiation effect, or a mixture of both entities and were reported per our institutional standards. Statistics A 2-sided P-value of.05 was accepted as statistically significant for all tests. Coding for pathology outcome was scored as 0 for pure radiation effect (necrosis) with no viable tumor and as 1.0 for pure tumor recurrence, with 0.5 for mixed tumor plus radiation effect. Nonparametric ordinal univariate correlations between those pathology outcome values and preop variables were assessed with the Spearman Rho rank correlation test. Multivariate analysis of these correlations was performed with forward stepwise multivariate linear regression. Bivariate logistic regression was used to analyze outcome coded as pure radiation effect versus tumor (combining the pure tumor and mixed tumor/radiation effect categories into one). Survival time was computed from the time of SRS and median survival was calculated using the Kaplan Meier method. 7 All statistical tests were carried out using SPSS v15.0. This project was reviewed and approved by the University of Pittsburgh Institutional Review Board. Clinical Outcomes Results The median length of follow-up was 10 months (range, 0.7 63). Seven patients (13%) suffered postoperative complications, including pulmonary emboli (n ¼ 3), postsurgical meningitis (n ¼ 1), cerebrospinal fluid leak (n ¼ 1), third nerve palsy (n ¼ 1), and pseudomeningocele (n ¼ 1). Forty-three of the lesions (83%) had follow-up imaging available for review postoperatively. Of these, 23 were found to have a distant brain failure (53%). The median time from SRS to distant brain failure was 8 months (range, 1.8 39.3). At the time of analysis, 33 patients (67%) were dead, 13 patients (27%) were alive, and 3 patients (6%) were lost to follow-up. Ten patients (30%) died from neurological causes, 3 (9%) died from systemic causes, and 20 (61%) died from unknown causes. The median survival from the time of SRS was 11.1 months. Pathology Results After analysis by a neuropathologist, 27 of the resected lesions (52%) were reported as containing tumor recurrence alone, 14 (27%) were classified as containing radionecrosis alone, and 11 (21%) contained a mixture of both tumor and radionecrosis. Imaging/Pathology Correlates Table 2 presents results of univariate and multivariate analyses for correlation of preoperative imaging indices with pathology. No pre-srs imaging variables were found to predict lesion pathology following resection. Analysis of preoperative imaging demonstrated that the ratio of edema (signified by high T2 signal) to T1 enhancing volume was highly predictive of lesion pathology on univariate (P ¼.001) and multivariate analysis (P ¼.002). Specifically, lesions that presented with a greater degree of surrounding edema were more likely to consist of radionecrosis (Fig. 1). No other preop imaging indices were found to predict lesion pathology, including T1-T2 match (Fig. 2). Examples of preoperative images of patients with tumor recurrence or radionecrosis identified following resection are displayed in Fig. 3. Time to Resection As shown in Table 2, the time between SRS and resection was found to be significantly correlated with lesion pathology on both univariate (P,.01) and multivariate (P,.01) analyses. Specifically, a longer latent period between SRS and progression was associated with a higher likelihood of radionecrosis. All lesions that were resected.12 months following SRS demonstrated radiation effect without evidence of tumor on pathological analysis (Figs. 2 and 3). Table 2. Correlation of pathology outcome with preop imaging and treatment parameters Variable P (uni Spearman) P (multi LinearRegr) Time SRS to resection.00004.00007 Preop edema/lesion volume.001.0019 ratio a Preop T1/T2 mismatch b.483.386 Preop Low T2/T1c c.049.856 Pre-SRS Low T2/T1c.498.678 Pre-SRS edema/lesion volume.252.816 ratio Pre-SRS dose.499.366 Pre-SRS treatment volume.096.979 Pre-SRS fractions.383.569 Pre-SRS whole brain radiation.343.562 Repeat SRS.737.107 Radioresistant d.069.896 Pure radiation necrosis was assigned a value of 0, pure tumor was assigned a value of 1, and mixed lesions were assigned a value of 0.5. Univariate (uni) P-values are nonparametric Spearman Rho rank correlations. Multivariate (multi) values are from forward stepwise linear regression (LinearRegr) with only significant variables (P,.05) entered into the final model. a Edema/lesion volume ratio ¼ ratio of the edema or high T2 MR signal volume divided by T1 contrast-enhancing volume. b T1/T2 mismatch ¼ qualitative assessment of lesion size matching between T1 contrast-enhancing and T2 sequences. c Low T2/T1c ¼ ratio of low T2 MR signal volume in the region of the tumor divided by the T1 contrast-enhancing tumor volume. d Radioresistant ¼ melanoma, renal cell, or gastrointestinal primary. 1734 NEURO-ONCOLOGY DECEMBER 2013

Fig. 1. Scatter plot of pathological outcomes classified by time since radiosurgery and ratios of the edema (high T2 signal) volume in the region of the tumor divided by the volume of the T1 contrast-enhancing volume. All of the lesions resected,18 months from SRS with a ratio of edema to enhancing tumor,10 showed the presence of tumor with or without radiation effect. Fig. 2. Scatter plot of pathological outcomes classified by time since radiosurgery and ratios of the low T2 signal volume divided by the volume of the T1 contrast-enhancing volume. T1-T2 lesion volume match was not predictive of histology. Prediction of Persistent Tumor Table 3 shows the results of logistic regression univariate and multivariate forward stepwise analysis of factors predicting the presence of tumor being present (pure tumor or mixed tumor and radiation effect) versus pure radiation effect (with no viable tumor) at the time of resection. Using a cutoff value of 10 for the edema/lesion volume ratio (volume of high T2 edema signal divided by the contrast-enhancing tumor volume), we found that tumor was present in 22/24 resected tumors with edema/lesion volume ratio,10 and 13/24 tumors with edema/lesion volume ratio 10. An edema/lesion volume ratio,10 predicted persistent tumor with a sensitivity of 63% ¼ 22/(22 + 13), a specificity of 85% ¼ 11/(11 + 2), a positive predictive value (PPV) of 92% ¼ 22/(22 + 2), and a negative predictive value (NPV) of 46% ¼ 11/24. If an edema/lesion volume ratio,10 was evaluated as a predictive parameter for residual tumor only in patients undergoing surgery,18 months after radiosurgery, 22/22 resected tumors with edema/lesion volume ratio,10 had tumor present (PPV ¼ 100%) and 10/23 with edema/lesion volume ratio 10 had tumor present (NPV ¼ 43%) for a sensitivity of 63% ¼ 22/35 and specificity of 100% (10/10). Discussion SRS is becoming a more widespread treatment for initial management of brain metastases. In cases where lesions progress following treatment, the importance of distinguishing the recurrence of tumor from radionecrosis is paramount. Patients with recurrent cancer may benefit more from surgical resection or repeat SRS, while patients without remaining tumor may be spared craniotomy and effectively be managed more conservatively with the use of steroids and other medical treatments, including vitamin E, pentoxyphylline, or bevacizumab. As MRI remains the current standard for follow-up evaluation of brain metastases, the ability to differentiate these two entities with standard sequences is desirable. Kano et al 8 correlated preoperative MRI with histopathological findings in 68 patients treated with gamma knife radiosurgery for a brain metastasis. They found that a correspondence between the contrast-enhancing T1 and low-signal T2 volumes was associated with tumor recurrence, while the lack of a clearly defined T2 lesion was associated with radiation necrosis. The T1/ T2 mismatch was found to have a sensitivity of 83% and a specificity of 91% for detecting radionecrosis. Additionally, they found that a shorter time interval between SRS and resection was associated with tumor recurrence. In their study, T1/T2 mismatch was determined by qualitative assessment of the match between lesion borders on the different sequences. Dequesada et al 9 reviewedpreoperative MRI and pathological specimens from 32 patients who underwent radiosurgery for a brain metastasis and defined a novel radiographic feature, the lesion quotient, as the ratio of the maximum cross-sectional area of a lesion on axial T1- and T2-weighted sequences. A lesion quotient of.0.3 had a NPV for radiation necrosis of 96% (sensitivity 80%, specificity 96%), while a lesion quotient of,0.6 had a NPV for recurrent tumor of 100% (sensitivity 15%, specificity 100%). The lesion quotient NEURO-ONCOLOGY DECEMBER 2013 1735

Fig. 3. A 42-year-old female with breast cancer was treated with 21 Gy in a single fraction to a right parietal brain metastasis. She received repeat SRS with 20 Gy in a single fraction 7 months later following progression. She subsequently developed left-sided motor weakness 5 months following repeat SRS. (A C) MRI demonstrated enlargement of the lesion with a large volume of high T2 signal. Upon resection, pathology demonstrated radionecrosis with no viable tumor. A 45-year-old female with colon cancer was treated for a left frontal brain metastasis with 18 Gy in a single fraction. (D F) Six weeks following treatment, speech deficits developed and MRI demonstrated enlargement of the T1 enhancing lesion with minimal surrounding edema. Following resection, pathology demonstrated adenocarcinoma. demonstrated superior predictive value to other preoperative imaging findings, including heterogeneous enhancement, marginal enhancement, and cyst formation. Stockham et al 10 attempted to confirm the validity of the lesion quotient for prediction of lesion pathology in 51 patients treated with gamma knife radiosurgery followed by either biopsy or resection using the same method as Dequesada et al. Their analysis demonstrated a PPV and NPV of only 25% and 73%, respectively, for the prediction of radionecrosis and only 62% and 39% for recurrent tumor. Importantly, we found a significant correlation between the ratio of high T2 signal and T1 contrast-enhancing volume on preoperative imaging and lesion pathology. Specifically, lesions with more edema identified on T2-weighted sequences were more likely to demonstrate necrotic pathology without evidence of residual or recurrent tumor. Using a cutoff value of 10 for the edema/ lesion volume ratio, we were able to predict the presence of tumor with a PPV of 92%, which increased to 100% when looking only at patients who underwent resection,18 months following SRS. To our knowledge, this is the first report of quantitative measurement of edema for prediction of lesion pathology following SRS for brain metastases. The extent of edema surrounding a lesion is readily assessable by clinicians and may provide an additional tool for accurate diagnosis of growing lesions. Our data also reveal longer time between SRS and resection to be predictive of radionecrosis. This is in keeping with previous reports that also show that recurrent tumor is more likely to result in faster tumor progression. 8 Furthermore, combining the time to resection with the edema/lesion volume ratio on preoperative imaging allowed for improved prediction, where all of the 22 lesions that were resected,18 months following SRS with an edema/lesion ratio,10 demonstrated the presence of tumor with or without necrosis. Detecting tumor, whether in the presence of necrosis or not, is critical for management of such lesions, as it indicates the need for further surgical or radiotherapeutic intervention. However, because tumor cells, whether viable or not, may be present for months following SRS, it remains unclear whether tumor in this time period is residual or recurrent. A prospective study utilizing MRI indices for determination of lesion pathology and management would clarify the utility of these techniques. Patel et al 11 have demonstrated that patients with a brain metastasis that increases in size following treatment may actually benefit from improved survival compared with patients whose lesions decrease in size or remain stable. Increase in lesion size, as well as increase in surrounding edema, may be indicative of a therapeutic immune response that portends improved tumor control and improved survival. Biologically, necrotic areas may be expected to induce more edema given the inflammatory response that accompanies radionecrosis. Furthermore, in their study, 10 patients with enlarging lesions and surrounding edema post-srs underwent salvage surgery 1736 NEURO-ONCOLOGY DECEMBER 2013

Table 3. Correlation of pathology outcome determination of pure radiation effect with preop imaging and treatment parameters Variable P (uni LogRegr) P (multi LogRegr) Time SRS to resection.007 (.005 a ),.021 Preop edema/lesion volume.002.034 ratio b Preop edema/lesion 10.008 (.046 a ) Preop T1/T2 mismatch c.720.489 Preop Low T2/T1c d.037.430 Pre-SRS Low T2/T1c.911.575 Pre-SRS edema/lesion volume.173.746 ratio Pre-SRS dose.856.080 Pre-SRS treatment volume.096.241 Pre-SRS fractions.999.776 Pre-SRS whole brain radiation.128.211 Repeat SRS.805.228 Radioresistant e.076.703 Univariate (uni) and multivariate (multi) forward stepwise logistic regression (LogRegr) were used with only significant variables (P,.05) entered into the full final model. a.005 and.046 are the multivariate P-values using edema/lesion 10 in a separate model instead of the variable preop edema/t1c with continuous values. b Edema/lesion volume ratio ¼ ratio of the edema or high T2 MR signal volume divided by T1 contrast-enhancing volume. c T1/T2 mismatch ¼ qualitative assessment of lesion size matching between T1 contrast-enhancing and T2 sequences. d Low T2/T1c ¼ ratio of low T2 MR signal volume in the region of the tumor divided by the T1 contrast-enhancing tumor volume. e Radioresistant ¼ melanoma, renal cell, or gastrointestinal primary. where histopathology demonstrated radiation-induced necrosis in all cases. Taken together, these findings suggest that increases in perilesional edema posttreatment, particularly when they occur later after treatment, may be indicative of a beneficial inflammatory response rather than tumor progression. Such lesions, therefore, may be better managed medically rather than surgically, especially given the limited survival of patients with brain metastases who might be spared a craniotomy in their final months of life. It is unknown what the effect of steroids and other agents aimed at reducing swelling may have on a potentially favorable immune response. Therefore, the optimal management of brain metastases with progressive edema post-srs remains unknown and requires further clinical study. Importantly, our study puts forth a method that may aid clinicians in reliably differentiating radionecrosis from tumor recurrence using standard MR sequences that are commonly used for follow-up imaging. Some centers may utilize advanced imaging methods (PET, SPECT, MR SPECT) in an attempt to improve specificity, but these methods are expensive and not widely available, and their use may prolong the time before the patient receives appropriate treatment. As such, the ability to make an accurate diagnosis with traditional MR sequences and tools available to radiation oncologists and neurosurgeons in all settings is highly advantageous. Similar to Stockham et al, 10 our analysis did not find T1-T2 match a correlation between the T1 enhancing volume and low T2 signal to be predictive of recurrence. Prior studies have demonstrated T1-T2 match assessed qualitatively 8 or by measurement of lesion cross-sectional area 9 to be significantly associated with tumor recurrence. To our knowledge, this is the first study directly and quantitatively measuring tumor volumes to assess T1-T2 match. Because not all brain metastases are of spherical or ellipsoid morphology, accurate volume measurement with the use of contouring software may be a more accurate and reliable measurement for matching lesions between MR sequences. Further studies may be necessary to compare various techniques of measuring T1-T2 match, their agreement, and their predictive values. The identification of factors that may predict increased likelihood of radionecrosis versus tumor recurrence at the time of SRS would be of significant value for following patients with progression after SRS. Unfortu-nately, our study did not find any significant correlation of pre-srs imaging indices, SRS treatment volume, SRS fractionation, SRS dose, prior WBRT, repeat SRS, or radioresistant histology with lesion pathology at the time of resection. Our study is limited by its retrospective nature and associated biases as well as its sample size. The cohort that we analyzed was also somewhat heterogeneous and comprised patients who underwent resection and SRS or SRS alone prior to their final resection. Furthermore, in this study, we have not analyzed the relationship between the amount of edema surrounding a lesion and intracranial location. The extent of measured high T2 signal may be governed in part by the location of the lesion or the presence of adjacent structures. For instance, the edema surrounding a tumor that abuts the cranium may be spatially limited. This potentially confounding factor will need to be explored further in future studies. Reliable methods for distinguishing tumor recurrence from radionecrosis noninvasively have been elusive. This difficulty is one of the most common and most challenging issues in the care of patients with brain metastases or other tumors treated with radiation therapy. Our study presents a new method for quantitative determination of lesion pathology following radiosurgery by measuring the volume of edema surrounding the tumor. This index is easily measurable and utilizes imaging techniques that are widely used. This measurement may allow improved prediction of lesion pathology when combined with time between SRS and resection. Further studies are in order to validate this technique and its applicability to other tumors treated with radiotherapy. Conflict of interest statement. None declared. Funding This work was supported by internal institutional funding only. NEURO-ONCOLOGY DECEMBER 2013 1737

References 1. Alexiou GA, Fotopoulos AD, Papadopoulos A, Kyritsis AP, Polyzoidis KS, Tsiouris S. Evaluation of brain tumor recurrence by (99 m)tctetrofosmin SPECT: a prospective pilot study. Ann Nucl Med. 2007;21(5):293 298. 2. BelohlavekO, SimonovaG, KantorovaI, Novotny J, Jr., Liscak R. Brainmetastases after stereotactic radiosurgery using the Leksell gamma knife: can FDG PET help to differentiate radionecrosis from tumour progression? Eur J Nucl Med Mol Imaging. 2003;30(1):96 100. 3. Bobek-Billewicz B, Stasik-Pres G, Majchrzak H, Zarudzki L. Differentiation between brain tumor recurrence and radiation injury using perfusion, diffusion-weighted imaging and MR spectroscopy. Folia Neuropathol. 2010;48(2):81 92. 4. Gomez-Rio M, Martinez Del Valle Torres D, Rodriguez-Fernandez A, et al. (201)Tl-SPECT in low-grade gliomas: diagnostic accuracy in differential diagnosis between tumour recurrence and radionecrosis. Eur J Nucl Med Mol Imaging. 2004;31(9):1237 1243. 5. Ortega-Lozano SJ, del Valle-Torres DM, Gomez-Rio M, Llamas-Elvira JM. Thallium-201 SPECTin brain gliomas: quantitativeassessment in differential diagnosis between tumor recurrence and radionecrosis. Clin Nucl Med. 2009;34(8):503 505. 6. Tan H, Chen L, Guan Y, Lin X. Comparison of MRI, F-18 FDG, and 11C-choline PET/CT for their potentials in differentiating brain tumor recurrence from brain tumor necrosis following radiotherapy. Clin Nucl Med. 2011;36(11):978 981. 7. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:457 481. 8. Kano H, Kondziolka D, Lobato-Polo J, ZorroO, Flickinger JC, Lunsford LD. T1/T2 matching to differentiate tumor growth from radiation effects after stereotactic radiosurgery. Neurosurgery. 2010;66(3):486 491. Discussion 482 491. 9. Dequesada IM, Quisling RG, Yachnis A, Friedman WA. Can standard magnetic resonance imaging reliably distinguish recurrent tumor from radiation necrosis after radiosurgery for brain metastases? A radiographicpathological study. Neurosurgery. 2008;63(5):898 903. Discussion 904. 10. Stockham AL, Tievsky AL, Koyfman SA, et al. Conventional MRI does not reliably distinguish radiation necrosis from tumor recurrence after stereotactic radiosurgery. J Neurooncol. 2012;109(1):149 158. 11. Patel TR, McHugh BJ, Bi WL, Minja FJ, Knisely JP, Chiang VL. A comprehensive review of MR imaging changes following radiosurgery to 500 brain metastases. AJNR Am J Neuroradiol. 2011;32(10):1885 1892. 1738 NEURO-ONCOLOGY DECEMBER 2013