Longitudinal multivoxel MR spectroscopy study of pediatric diffuse brainstem gliomas treated with radiotherapy

Int. J. Radiation Oncology Biol. Phys., Vol. 62, No. 1, pp. 20 –31, 2005 Copyright © 2005 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/05/$–see front matter

doi:10.1016/j.ijrobp.2004.09.027

CLINICAL INVESTIGATION

Brain

LONGITUDINAL MULTIVOXEL MR SPECTROSCOPY STUDY OF PEDIATRIC DIFFUSE BRAINSTEM GLIOMAS TREATED WITH RADIOTHERAPY ANNE LAPRIE, M.D.,*† ANDREA PIRZKALL, M.D.,*‡ DAPHNE A. HAAS-KOGAN, M.D.,‡ SOONMEE CHA, M.D.,* ANURADHA BANERJEE, M.D.,§ TERRY P. LE, B.A.,* YING LU, PH.D.,* SARAH NELSON, DR. RER. NAT.,* AND TRACY R. MCKNIGHT, PH.D.* Departments of *Radiology, ‡Radiation Oncology, and §Neurosurgery, University of California, San Francisco, San Francisco, CA; † Department of Radiation Oncology, Claudius Regaud Institute, Toulouse, France Background and Purpose: After radiotherapy (RT), children with diffuse intrinsic pontine gliomas (DIPG) are followed with sequential magnetic resonance imaging (MRI). However, MRI changes do not necessarily reflect tumor progression, and therefore additional noninvasive tools are needed to improve the definition of progression vs. treatment-related changes. In this study, we determined the feasibility and accuracy of multivoxel proton magnetic resonance spectroscopic imaging (1H-MRSI) for monitoring pediatric patients with DIPG. Methods and Patients: Twenty-four serial examinations of MRI/MRSI (7 2D-MRSI and 17 3D-MRSI) were performed on 8 patients with DIPG who received local RT. A total of 1635 voxels were categorized as “normal” or “abnormal” based on corresponding imaging findings on contrast-enhanced T1- and T2-weighted MRI. The choline to N-acetyl-aspartate ratio (Cho:NAA) and choline to creatine ratios (Cho:Cr) within each category of MRI abnormality were compared to their counterpart in normal surrounding tissues. The changes in these ratios corresponding to each type of abnormality were evaluated before RT, at response, and at recurrence, as determined by the clinical status of the patients. The presence or absence of lactate and lipid peaks was noted for each voxel. MRI/MRSI was performed on posterior fossa and supratentorial tissue of 3 volunteer pediatric patients. Results: The Cho:NAA and Cho:Cr values within the imaging abnormalities (3.8 ⴞ 0.93 and 3.55 ⴞ 1.37, respectively) were significantly higher than the mean values in normal-appearing regions (0.93 ⴞ 0.2 and 1.13 ⴞ 0.38, respectively) (p < 0.005). Cho:NAA values decreased from studies at diagnosis to the time of response to RT (3.12 ⴞ 0.5 and 2.08 ⴞ 0.73, respectively), followed by an increase at the time of relapse (from 1.83 ⴞ 0.92 to 4.29 ⴞ 1.08). Loss of lactate and lipid peaks correlated with response, and their presence and stability with relapse. In 3 patients, increased spectral abnormalities preceded the radiological and clinical deterioration by 2–5 months. Conclusion: Multivoxel MRSI is a feasible and reproducible noninvasive tool for assessing pediatric DIPG. Longitudinal multivoxel MRSI measurements have potential value in assessing response to radiation or other therapies, because they offer more coverage than single-voxel techniques and provide reliable spectral data. © 2005 Elsevier Inc. Brainstem glioma, Magnetic resonance spectroscopy, Radiation therapy, Children.

INTRODUCTION Diffuse intrinsic pontine gliomas (DIPG) represent 80% of pediatric brainstem gliomas and approximately 10% of all pediatric brain tumors, the most frequent solid tumors in children. Clinical symptoms and magnetic resonance imaging (MRI) findings are highly specific for diagnosing DIPG and obviate the need for potentially morbid biopsies for histological confirmation before radiotherapy (RT) (1). After RT, which is the mainstay of treatment, the majority of patients improve clinically. However, DIPG carries a bleak

prognosis: The progression-free interval is usually shorter than 6 months, and survival is poor with median survival times of less than 1 year and 2-year survival rates of less than 20%. Many new therapies have been evaluated for DIPG over the last 2 decades, including radiation dose escalation with hyperfractionated radiation therapy (2, 3), accelerated radiation therapy (4), preradiation chemotherapy (5), and high-dose chemotherapy (6). Unlike other types of pediatric brain tumors, little progress has been made toward increasing the survival rates of patients with DIPG.

Reprint requests to: Tracy Richmond McKnight, Ph.D., Department of Radiology, University of California, San Francisco, 185 Berry Street, Suite 350, San Francisco, CA 94107-1739. Tel: (415) 353-9467; Fax (415) 353-9423; E-mail: [email protected] Presented at the 46th ASTRO meeting, October 6, 2004, Atlanta, Georgia.

Supported in part by a grant from La Fondation de France (A.L.) and NIH K01-CA90244 (T.M.). Acknowledgments—We acknowledge Pranathi Srinivas and Forrest Crawford for outstanding technical assistance. Received Jun 3, 2004, and in revised form Sep 10, 2004. Accepted for publication Sep 12, 2004. 20

1

H-MRSI of pediatric pontine glioma

Therefore, renewed effort has focused on new paradigms of treatment (7). These new emerging paradigms require critical evaluation, particularly because MRI characteristics are less specific for disease status after RT. The most common dilemma arises with the appearance of a new contrastenhancing lesion, which often occurs after RT but is not specific for relapse (8, 9). Thus, MRI alone may not accurately distinguish between recurrent tumor and inflammatory or necrotic changes that can result from treatment. Moreover, the absence of biopsy for DIPG, either at diagnosis or at relapse, renders additional noninvasive methods for tissue characterization even more desirable. Proton magnetic resonance spectroscopic imaging (1HMRSI) represents a noninvasive method that provides additional metabolic diagnostic indices beyond the anatomic information and tumor volume measurements given by MRI. No longer simply a research tool, multivoxel 1HMRSI is being used clinically for adult gliomas to monitor response to therapy, particularly in supratentorial locations (10, 11). The characteristic magnetic resonance (MR) spectroscopic features of high-grade tumors are an elevated peak corresponding to choline-containing compounds (Cho) and a depressed N-acetyl-aspartate (NAA) peak. Cho measured with MRSI is associated with cellular membranes, and high Cho levels suggest rapid membrane turnover (12, 13) and/or high cellular density (14, 15). NAA is a metabolite found primarily in neurons and is, therefore, interpreted as a measure of neuronal density (16) and/or reversible changes in mitochondrial function in neurons (17). Spectral information can also be obtained for creatine compounds (Cr) that are indicators of cellular energetics (13). The ratios of Cho:NAA and Cho:Cr are therefore commonly used to quantify the metabolic abnormalities in tumor. Lactate may arise from anaerobic metabolism in hypoxic regions, and necrosis is distinguished by an increase in the lactate/lipid (LL) resonance and a reduction of all other metabolites. Because of the large field of view (80 –300 cc) typically used with multivoxel techniques, performing posterior fossa MRSI is technically difficult, because of the close proximity of the skull. Consequently, there is a potential for magnetic susceptibility artifacts from adjacent bone, fat, and air surfaces (18) that can make shimming more difficult and/or falsely inflate the size of the spectral peaks. Therefore, mainly single-voxel magnetic resonance spectroscopy (MRS) techniques have been performed on lesions of the posterior fossa in previous studies. MRSI, unlike singlevoxel MRS, provides larger coverage and therefore inspection of different parts of the tumor region, in addition to simultaneous assessment of healthy surrounding tissues (19). Greater tumor and healthy brain tissue coverage providing better metabolite evaluation, combined with the lack of substantial multivoxel MRS data focused on pediatric diffuse pontine gliomas, prompted us to analyze serial multivoxel examinations. Our goals were to evaluate the feasibility and accuracy of such techniques for monitoring DIPG in pediatric patients, to identify biologically significant

● A. LAPRIE et al.

21

MRS indices of response to therapy and relapse, and to correlate the MRSI data with clinical and MRI findings. METHODS AND PATIENTS Patients The protocol for this study was approved by the University of California at San Francisco Institutional Review Board, and informed parental consent was obtained for all patients presented in this study. We selected all pediatric patients who underwent an MRS between September 2001 and September 2003 at our institution and were diagnosed with a brain tumor. Among the 15 patients fulfilling these criteria, 7 presented with supratentorial or cerebellar tumors of various histologies (low-grade glioma, ependymoma, anaplastic astrocytoma, primitive neuroectodermal tumor), and 8 patients had DIPG treated with RT. We focused our study on these 8 patients presenting with DIPG (5 females and 3 males, mean age, 6 years; range, 4 –9 years). Diagnosis was made on clinical and radiological findings. Twentyfour MRI/MRSI were performed and analyzed in the context of clinical and MRI findings. Although no patient had surgical resections, 2 patients underwent a biopsy before RT. No patient had neurofibromatosis type 1. All patients underwent focal external beam RT. A total accumulated dose of 56 – 60 Gy was delivered in 1.5–1.8 Gy daily fractions. Seven of the patients had concomitant systemic therapy with one of the following agents: Temozolomide (Temodar), ZD1839 (Iressa), or STI571 (Gleevec). All patients were on steroids from the time of diagnosis.

MR data acquisition Magnetic resonance images were acquired on a 1.5T clinical imager from General Electric Medical Systems (Milwaukee, WI) with the standard head coil. All examinations were performed under general anesthesia. In each case, a set of sagittal scout images were obtained. Patient studies included acquisition of spinecho images with T2 and proton density weighting using echo times (TEs) of 30 and 80 ms and a repetition time (TR) of 2500 ms. T2-weighted axial fluid attenuated inversion recovery images (TR ⫽ 10002 msec, TE ⫽ 147 msec), as well as T1-weighted threedimensional spoiled gradient echo (3D-SPGR) images (TR ⫽ 27 ms, TE ⫽ 6 ms, flip angle 40°), were obtained before and after the injection of a standard dose of gadolinium contrast agent (11, 20). The spectral acquisition made use of point-resolved spectroscopy (PRESS) volume selection technique. The selected volume was chosen to exclude subcutaneous fat and regions with large variations in magnetic susceptibility, such as the sinuses and skull. The PRESS excitation region was positioned to cover as much of the imaging abnormality and presumably normal surrounding tissue as possible. Very selective suppression (VSS) pulses were used for outer volume suppression to minimize artifacts due to subcutaneous lipids within osseous regions adjacent to the tissues of interest (21). The PRESS volume was first imaged to confirm location and size, followed by shimming and water suppression parameters. Before May 2002, both 2D- and 3D-MRSI were performed; subsequently, only 3D-MRSI scans were performed. For the 2D-MRSI scans (n ⫽ 7), the parameters were as follows: TR of 1000 ms, TE of 144 ms, 24 ⫻ 24 ⫻ 1 or 18 ⫻ 18 ⫻ 1 phase-encodes, voxel volume ⫽ 1.0 cm3 or 1.78 cm3. The parameters for the 3D-MRSI technique were a TR of 1000 ms and TE of 144 ms. Either 12 ⫻ 12 ⫻ 8 or 16 ⫻ 8 ⫻ 8 phase-encode steps were employed, providing a nominal voxel size of 1.0 cm3. The

22

I. J. Radiation Oncology

● Biology ● Physics

setup and data acquisition for the spectroscopic component of the examination took approximately 20 min. The postcontrast 3D-SPGR images were acquired immediately before spectroscopy. As a result, patient movement between the 3D-SPGR and spectroscopy was assumed to be minimal; hence, the 3D-SPGR images were used as the reference images for the processed 2D or 3D arrays of spectra. The T2 images were registered to the 3D-SPGR images, by means of a rigid-body surface-matching method that has been described previously (22).

Correlation of image and spectral data The visual analyses of MR images, spectra,, and metabolite images were performed via an interactive display package that was developed in our laboratory using the Interactive Data Language (IDL, Research Systems, Boulder, CO) image processing application. The software displays an overlay of the MRSI array of spectra on the corresponding image slices (Fig. 1). With the aid of the software, we were able to interactively label voxels using the point-and-click method, identifying correspondences to the different imaging characteristics. Voxels were categorized as “normal appearing” or “abnormal” based on corresponding imaging findings on contrast-enhanced T1-weighted and T2-weighted MRI sequences. The categorization of voxels was performed by one operator (A.L.) and reviewed by an experienced neuroradiologist (S.C.). With the T2-weighted MRI sequences as a guide, voxels were categorized as either (a) normal-appearing infratentorial tissue, (b) normal-appearing supratentorial tissue, or (c) hyperintensity. With respect to the contrast-enhanced T1-weighted images, voxels were categorized as either (a) normal-appearing infratentorial tissue, (b) normal-appearing supratentorial tissue, (c) gadolinium enhancement, (d) hypointense signal except (e), or (e) hypointense signal inside rim enhancement indicating macroscopic necrosis. To avoid any equivocal data, voxels involving CSF, as well as voxels that did not correspond to any of the categories described above, were not labeled. An example of 3D-MRSI and of the categorization of voxels is displayed in Fig. 1c. The choline:NAA ratio (Cho:NAA) and choline:creatine ratio (Cho:Cr) were calculated for every voxel included in the study. A difficulty of using ratios to quantify the metabolic abnormality is that ratios with very small denominators (i.e., low NAA or Cr levels) are very sensitive to the noise in the spectrum. To remedy this problem, all of the ratios for each patient were stabilized by adding a correction factor to the NAA or Cr level. In each case, the NAA and Cr levels were nominally ordered, and the value at the fifth percentile was chosen as the correction factor. The stabilized ratios were compared between the T2 normal-appearing voxels from the patients and the control subjects to determine whether the voxels outside of the MRI lesion had normal metabolism. We then compared stabilized ratios in the voxels from both T2 categories (normal-appearing and hyperintense) for all patient examinations. The T2 categories were used because the T2-weighted hyperintense lesion better reflects the entire tumor load, i.e., solid tumor and infiltrative margin, as opposed to the T1-weighted contrast-enhancing lesion, which is simply an indicator of blood– brain barrier disruption. Lastly, we evaluated the changes in Cho:NAA and Cho:Cr and the presence of LL peaks within each category of MRI abnormality and at three specific time points: before RT, at clinically determined response, and at clinically determined recurrence.

Statistical methods The mean, standard deviation, and percentages of variation were calculated to characterize the patients, MRI features, Cho:NAA

Volume 62, Number 1, 2005

ratios, and Cho:Cr ratios. The statistical significance of the differences of Cho:NAA and Cho:Cr ratios within MRI abnormalities and normal tissues was quantified using a two-sided t test, because it involved 20 studies. Two-sided t tests were also used to describe the variations of Cho:NAA and Cho:Cr ratios at presentation, response, and relapse and the differences of Cho:NAA and Cho:Cr ratios between normal supratentorial and normal infratentorial tissues.

RESULTS Subjects Median follow-up was 18 months (range, 13–35 months). All 8 patients presented with clinical symptoms related to brainstem involvement within 6 months before diagnosis. They showed initial clinical and MRI tumor response or stabilization within 2 months of completing RT but relapsed within a median time of 4 months after RT (range, 2–25 months), leading to death in 5 of the 5 patients in a median time of 13 months after diagnosis (range, 10 –19 months). Six patients had MRSI before treatment; 5 of these also had posttreatment MRSI. The 2 remaining patients underwent serial MRSI after RT only. Table 1 summarizes the clinical information, serial MR–related data, and serial MR spectroscopy findings within the T2 abnormality. MRSI Good-quality spectra were obtained in all examinations. For each 2D and 3D examination, the mean number of voxels covered by the PRESS box was 10 (range: 4 –36) and 87 (range: 12–320), respectively. A total of 1635 voxels for the patients were obtained and were categorized on both T1-postcontrast and T2 images. Cho:NAA and Cho:Cr ratios regarding T1 abnormalities Sixteen of the MRSI examinations covered both contrast enhancement (CE) and T1 hypointensity surrounding contrast enhancement. Cho:NAA and Cho:Cr ratios were always higher in both regions of contrast enhancement and hypointensity compared to normal tissue. For hypointensity and contrast-enhancement regions, the Cho:NAA ratios were 2.26 ⫾ 1.39 and. 2.98 ⫾ 1.34, respectively, and the Cho:Cr ratios were 2.7 ⫾ 1.25 and 3.31 ⫾ 1.98, respectively. In cases where voxels were situated within the hypointensity inside the rim of contrast enhancement, the values were of 3.9 ⫾ 2.16 for Cho:NAA, and of 4.66 ⫾ 3.84 for Cho:Cr. Cho:NAA and Cho:Cr ratios with respect to MRI T2 lesions vs. normal-appearing tissue Infratentorial normal-appearing tissue voxels were obtained in 20 examinations. The mean values for Cho:NAA and Cho:Cr ratios in abnormal and normal voxels were calculated in each examination. Cho:NAA and Cho:Cr ratios were higher in every examination within the abnormalities compared to normal-appearing regions (Fig. 2). The difference was statistically significant (two-tailed t test) for Cho:NAA ratios (3.8 ⫾ 0.93 in abnormal regions vs. 0.93 ⫾

23

Fig. 1. The 4 slices (a) of the preradiation fluid attenuated inversion recovery images of Patient 1 are displayed with PRESS box laid on tumor and surrounding normal supratentorial and infratentorial tissue. Voxels were categorized using A for infratentorial normal tissue, B for supratentorial normal tissue, and C for voxels completely involved by tumor. (b) Corresponding PRESS box with spectra. (c) Example of voxel categorization and 3 corresponding voxels showing typical supratentorial normal, infratentorial normal, and tumor spectra.

24

Table 1. Summary of clinical, MRI, and MRSI findings

Gender

1

F

4

2

M

4

Value of Cho:Cr

Change in Cho:Cr (%)

LL peaks

Minimal Decreased

3.6 1.9

⫺47%

2.2 1.3

⫺39%

Yes No

3.2 1.4

⫺56%

1.5 1.0

⫺33%

No No

Regression Stable PD PD

Minimal Modified : faint Rounded area Unchanged Unchanged Increased Rim enhancement

1.7 1.3 2.4 6.1

⫹23% ⫺26% ⫹84% ⫹154%

1.4 1.3 1.3 2.0

⫹46% ⫺7% ⫺3% ⫹54%

No No No Yes

Stable

Rim enhancement Modified : patchy

3.0 6.8

⫹128%

2.3 4.7

⫹103%

Yes Yes

Rim enhancement Stable Better defined Increased Surrounding Enhancement New focus

2.4 2.9 2.9 4.0

⫹24% ⫺3% ⫹40%

2.6 2.6 1.9 4.6

⫹1% ⫺27% ⫹141%

Yes No No No

4.0

0%

3.7

⫺19%

No

Clinical outcome

3D 3D

Pre-RT 2m

2D 2D

Pre-RT 7-d

Response

3D 3D 3D 3D

2-m 4-m 7-m 8-m

3D 3D

Pre-RT 7-d

2D 2D 2D 3D

Pre-RT 7-d 2-m 8-m

Response Stable TRC

3D

12-m

PD or TRC*?

Response

Contrast enhancement

Time to progression after RT

Time to death after RT

4-m

15-m

7-m 3

M

9

2-m 4

F

6

17-m

12 m? 5

M

9

3D

Pre-RT

Rim enhancement

5.5

5.0

No

6

F

4

2D 3D 3D 2D

Pre-RT 7d 1-m 2-m

PD Stable PD

Minimal, nodular Increased Decreased Increased

3.2 3.5 1.9 4.4

⫹10% ⫺46% ⫹131%

3.4 3.2 4.3 4.4

⫺8% ⫹37% ⫹2%

No No Yes Yes

⫺16%

4.7 6.9

⫹47%

No No

⫹400%

1.6 1.9

⫺15%

No No

7

F

5

3D 3D

2-m 4-m

Relapse Stable

2 rims Stable

2.3 1.9

8

F

7

3D 3D

4-m 20-m

Response Stable

Slightly increased Slightly increased

1.0 5.0

3-m

8-m

2- or 3-m

9-m

2-m

11-m

25-m

*Time of MRS is either pre-RT or expressed in months (m) or days (d) after completion of RT.Abbreviations: LL ⫽ lactate/lipid; PD ⫽ progressive disease; TRC ⫽ treatment-related changes.

Volume 62, Number 1, 2005

Change in Cho:NAA (%)

Time of MRS*

● Biology ● Physics

Value of Cho:NAA

Type of MRS

I. J. Radiation Oncology

Patient

Age at diagnosis

1

H-MRSI of pediatric pontine glioma

● A. LAPRIE et al.

25

Fig. 2. Differences in values for Cho:NAA and Cho:Cr ratios between normal and abnormal tissues as defined on T2-weighted sequences.

0.2 in normal regions, p ⬍ 0.001) and for Cho:Cr ratios (3.55 ⫾ 1.37 in abnormal regions vs. 1.13 ⫾ 0.38 in normal regions, p ⬍ 0.001). Infratentorial vs. supratentorial normal tissue Comparison of normal supratentorial and infratentorial Cho:NAA and Cho:Cr ratios was performed based on 10 examinations in 6 patients. There was a strong trend toward higher Cho:NAA ratios in infratentorial tissue than in supratentorial tissue (p ⫽ 0.08) with the mean difference being 0.17 ⫾ 0.35. Such differences were not observed in Cho:Cr ratios.

Variations of metabolic ratios longitudinally Among the 5 patients with MRSI studies before and after RT, 4 responded to RT within a median time of 2 months. In those patients, Cho:NAA values decreased from diagnosis to response to RT (3.12 ⫾ 0.5 and 2.08 ⫾ 0.73, respectively, p ⫽ 0.06). A subsequent increase in Cho:NAA was observed at the time of relapse (median 4 months after RT), from 1.83 ⫾ 0.92 to 4.29 ⫾ 1.08 (p ⫽ 0.01) (Figs. 3 and 4). A similar analysis of Cho:Cr ratios did not show much difference between diagnosis and relapse (2.41 ⫾ 0.82 vs. 2.31 ⫾ 1.5, p ⫽ 0.91). However, the difference in Cho:Cr ratio seemed more pronounced between response, or stable

Fig. 3. Box plots illustrating the choline:N-acetyl-aspartate (Cho:NAA) and choline:creatine (Cho:Cr) ratios at diagnosis and clinical response to RT.

26

I. J. Radiation Oncology

● Biology ● Physics

Volume 62, Number 1, 2005

Fig. 4. Box plots illustrating the choline:N-acetyl-aspartate (Cho:NAA) and choline:creatine (Cho:Cr) at response/ stabilization disease and progression.

disease, vs. progression (2.46 ⫾ 1.28 vs. 3.51 ⫾ 1.53, p ⫽ 0.33) (Figs. 3 and 4). Presence of lactate/lipid (LL) before and after RT In 7 examinations on 5 of the patients, LL peaks were present, situated within both T1 and T2 abnormalities. On T1 postcontrast images, the LL peaks were situated: on contrast enhancement in 3 examinations; on T1 hypointensity (outside the contrast enhancement) in 2 examinations; on both the rim enhancement and inside the rim in 1 examination and on all 3 types of T1 abnormalities in 1 examination. LL peaks were detected on the tumor in 3 of the 6 pretreatment MRS scans. LL peaks resolved after RT in 2 of these 4 cases: 1 patient relapsed 4 months after RT, and the other is still alive with stable disease 14 months after completion of RT. The 9-year-old boy (Patient 3), who did not show clinical or radiological response to radiation, displayed LL peaks both at diagnosis and at the end of RT. In one patient who did not exhibit initial LL peaks, LL peaks appeared 8 weeks before clinical and MRI recurrence, along with elevations of Cho:NAA and Cho:Cr. Clinical value of MRS In 3 patients, spectral abnormalities preceded the clinical and MRI deterioration by 2–5 months. In a 4-yearold girl (Patient 6, Fig. 5), LL appeared 1 month after RT, despite stable findings on clinical examinations and MRI and constant Cho:NAA and Cho:Cr ratios. Relapse occurred 1 month later with concomitant worsening of clinical symptoms, MRI indications of recurrence, and increases in choline ratios. For Patient 3, LL was present before and after treatment, and choline ratios increased twofold at the end of RT, 2 months before clinical and MRI-determined relapse. The third patient (Patient 8) underwent her first MRSI 4 months after RT and was determined to have tumor response to radiation based on improved clinical status and stable size and contrast enhancement of tumor on MRI. Twenty months after RT,

although the clinical findings and MRI size on T2 were identical, CE was slightly increased, and spectra had tumor-like appearances, reflected in a sixfold increase in the Cho:NAA ratios. She presented with clinical and MRI evidence of relapse 5 months later. For Patient 4, MRS provided critical additional information, because clinical and imaging findings were discordant at the 12-month examination. Specifically, the patient presented with a suspected clinical relapse, but the longitudinal stability of both Cho:NAA and symptoms confirmed the MRI findings of treatment-related changes rather than tumor recurrence, and the patient is still alive 14 months after diagnosis. DISCUSSION It is important to improve the management of patients with DIPG, because, unlike other pediatric brain tumors, diffuse intrinsic pontine gliomas carry a very poor prognosis that has not improved in the past decade. Furthermore, the fact that biopsy is recommended neither at diagnosis nor at relapse highlights the need for noninvasive, reliable tools complementary to MRI to improve the management of such tumors. Because DIPG is an unresected high-grade type of tumor always treated with one main treatment modality, its longitudinal study by MRS is particularly reproducible. All these arguments prompted us to perform this study. Predictive value of MRS Magnetic resonance spectroscopy is clinically used in adult high-grade brain tumors to guide biopsy and monitor treatment response (23, 24). It has also been proposed as a tool for radiosurgical planning, particularly when elevated Cho:NAA ratios appear outside of contrast-enhancing MRI lesions (25, 26). It is a promising tool that could help define the target volume for conformal external beam radiotherapy of supratentorial high-grade and low-grade glioma (27–30).

1

H-MRSI of pediatric pontine glioma

● A. LAPRIE et al.

27

Fig. 5. Conventional gadolinium-enhanced T1-weighted images and proton MRSI at 1 month post-RT and at 2 months post-RT of a girl age 4 years (See Table 1, Patient 6). (a) At 1 month follow-up, Cho is low, as well as NAA, but an LL peak appears inside the rim enhancement. (b) At 2-month follow-up, LL peaks persist, and high Cho peaks appear, along with clinical progression and progression of MRI abnormalities (including increased contrast enhancement and increased size on T2-weighted images).

In performing serial examinations of patients with focal lesions, the combination of 3D-MRI and 3D-MRSI is expected to be valuable for assessing response to therapy,

radionecrosis, or disease progression in adult brain gliomas (10, 11, 31), and recent publications have suggested its role in predicting tumor outcome (32, 33).

28

I. J. Radiation Oncology

● Biology ● Physics

Numerous pediatric studies, using mainly single-voxel MRS, have reported the usefulness of MRS for managing patients with brain tumors (34 –36. With respect to brainstem tumors, two recent single-voxel studies including DIPG but also less deadly adult and pediatric brainstem gliomas concluded that the higher the Cho:NAA ratio, the higher the grade of tumor (37, 38). The predictive value of MRS was confirmed in 2D-MRSI studies on pediatric brain tumors (39, 40) of different grades and of supratentorial or infratentorial origins; the technique used was that of the “worst voxel” analysis. In our study, we chose rather to describe the results for all voxels corresponding to each abnormality in each examination. Our technique resulted in more moderate differences in global values between normal and abnormal tissues but gave, in our opinion, a broader report of the range of values that can be found. Feasible and reliable The results of this study provide evidence that multivoxel MRSI is clinically feasible and provides good spectral data of both tumor and surrounding healthy tissue in pediatric posterior fossa. This is, to our knowledge, a novel observation concerning three-dimensional MRSI and DIPG. 1H-MRS of the brainstem, even single voxel, has been considered difficult because of the limitations of 1H-MRS procedures in the posterior fossa (41, 42). The main challenges of 1H-MRS of the central nervous system structures in the posterior fossa are the proximity of bone and subcutaneous fat, causing heterogeneity of the main magnetic field, and difficult shimming. These difficulties were overcome in this study with the use of very selective suppression radiofrequency bands that were placed over portions of the PRESS excited region and/or on the osseous and lipid-rich anatomy surrounding the PRESS region. The pulses were used also to conform the selected volume to the selected anatomy and to more accurately define the edge of the PRESS region using oblique spatial saturation (43). Because reliability of single-voxel MR spectroscopy findings are dependent on voxel position, multivoxel spectroscopy offers a larger coverage of the tumor, which enhances the ability to accurately obtain serial data from the tumor (19). Population studied Approximately 500 children present with DIPG every year in the United States (7). Because of the low incidence of this disease, the population studied here is small. However, the patients’ characteristics were homogenous regarding diagnosis, age, duration of follow-up, and outcome. Moreover, the total number of examinations and voxels analyzed is high. The age of patients in this study ranged from 4 to 9, giving a homogeneous set of data. Indeed, myelination, which induces great variation in spectra, is over by 24 months. Furthermore, because of the bleak prognosis of DIPG, the interval of time between first and last 1H-MRSI in each patient was short, obviating potential variations due to aging. Therefore the variations of the metabolite values are expected to be due only to tumor outcome.

Volume 62, Number 1, 2005

The uniformity of the population and the large data sets enhance the accuracy of our findings. Changes in Cho:NAA and Cho:Cr ratios are consistent with high-grade glioma In this study, Cho:NAA and Cho:Cr ratios on tumor were significantly higher than on normal infratentorial tissue, reflecting a reliable indicator of altered metabolism in the tumor that is consistent with published studies on pediatric tumors (39, 44) and on adult supratentorial gliomas (45, 46). Although Cho:NAA ratios were higher on normal infratentorial than normal supratentorial voxels, the difference between normal infratentorial and tumor voxels was still significant. With regard to the longitudinal changes in Cho:NAA that were reported for patients that received RT, our findings of decreases at response and increases at relapse must be tempered by the small number of patients used in the analysis (n ⫽ 4). In addition to the t tests that were performed to illustrate the strength of the trends, we also calculated the percent change in the Cho:NAA ratio at each examination and found that it was consistently correlated with the evolution of the tumor, either response, stabilization, or regression. Changes in Cho:NAA reflected the clinical and imaging findings better than changes in Cho:Cr. This suggests that longitudinal changes in Cho:NAA ratio, rather than the value of Cho:NAA ratio change at a single time point, better predict dynamic changes in tumor undergoing therapy. Lactate/lipid peaks Lactate/lipid peaks were detected in 5 patients at various time points of their clinical course, and their evolution seemed to parallel, or even precede in one case, the evolution of the disease. In one case, the vanishing of the LL peaks after RT on the 5 follow-up examinations supported the MRI finding of treatment-related changes rather than tumor recurrence. Accurate measurement of lactate is challenging, because lactate and lipid peaks are overlapping in frequency. These peaks are typically associated with high-grade lesions and have been considered as an indicator of tumor malignancy (47– 49). Lactate is thought to be an indicator of altered metabolism in brain tumors. Increased lactate may occur when the anaerobic glycolytic pathway exceeds the capacity of the lactate-catabolizing respiratory pathways (13) or when clearance of lactate in the necrotic regions is reduced (50). Although several studies reported that the lactate resonances present more frequently in high-grade gliomas than low-grade lesions (51, 52), the correlation between lactate presence and tumor malignancy remains uncertain (47, 52). Another possibility is that the lactate resonances are indicators of hypoxia, which is a factor in poor response to treatment of radiotherapy owing to the absence of oxygen, which is required for fixation of radiation-induced damage in DNA (53). Tarnawski et al. published recently a study concluding that the ratio of lactate:NAA from pretreatment MR data was found to be the strongest prognostic factor for

1

H-MRSI of pediatric pontine glioma

● A. LAPRIE et al.

29

malignant glioma patients with radiotherapy (54). A study of lactate using bioluminescence methods showed similar results (55). As for lipids, they are a byproduct of membrane destruction and are related to apoptosis (56) and necrosis in high-grade brain tumors (48, 49). Future MRSI studies at our institution will employ lactate editing to investigate the possibility that lactate and lipid play separate roles in the differentiation of radionecrosis from recurrence and in the prediction of survival. The potential value of MRS in differentiating tumor necrosis from recurrence (31, 57) is of particular interest in this disease where biopsy can be very harmful and where development of enhancement after radiation is common and does not imply necessarily a change in tumor character (9).

Interest of differentiating radiation necrosis from relapse Relying solely on clinical symptoms and MRI changes, which can be either tumor evolution or treatment-related changes, may lead to erroneous management of patients with brainstem gliomas. Even though lesion growth caused by either radiation necrosis or by relapse results in death in most cases, knowing the nature of the growing abnormalities is important in assessing the effect of new therapies under evaluation. Moreover, early diagnosis of relapse can allow the physicians and the parents to consider resuming the post-RT treatment and start new Phase I/II therapies while the child is still in reasonable condition.

Comparison of MRI and MRSI features The imaging features of the tumors in this study were typical of DIPG. Unlike supratentorial high-grade gliomas, which are usually highly enhancing lesions on T1-weighted MRI at diagnosis, the DIPG studied presented either as nonenhancing, minimally enhancing, or enhancing lesions with increased contrast uptake after RT. Increased CE after RT can be observed in lesions that eventually stabilize as well as in those that rapidly progress (9), which limits its usefulness as a prognostic marker. A particularly interesting result of this study was that Cho:NAA ratios were elevated on both CE and on surrounding T1 hyposignal, which suggests that CE observed in this tumor reflects the blood– brain barrier damage rather than increased tumoral metabolic activity compared to surrounding hyposignal. A recent study on 8 patients (adults and children) with various types of brainstem tumor with either 2D or single-voxel spectroscopy concluded that Cho:NAA ratios and Cho:Cr ratios were higher on recurrent tumor than on treatment-related lesions (58).

CONCLUSION In conclusion, our results indicate that three-dimensional MR spectroscopy of the brainstem is clinically feasible in children and offers good quality spectral data. Longitudinal three-dimensional MRSI could be a reliable clinical tool to help the neuroradiologist and oncologist in differentiating relapse from treatment-related changes and improve the determination of efficacy of RT and new therapies. It could then help decide whether to continue or resume primary post-RT treatment and possibly start Phase I/II experimental therapies. The analysis of the anatomic and metabolic properties of DIPG may further assist in the identification of new prognostic factors and therefore has the potential to improve outcome. These encouraging data need to be confirmed on more patients with precise follow-up times, and a further study would ideally include lactate-edited sequences. Therefore, a three-dimensional lactate-edited spectroscopy protocol is ongoing at our institution to monitor the evolution after RT of pediatric diffuse intrinsic pontine gliomas.

REFERENCES 1. Albright AL, Packer RJ, Zimmerman R, et al. Magnetic resonance scans should replace biopsies for the diagnosis of diffuse brain stem gliomas: A report from the Children’s Cancer Group. Neurosurgery 1993;33:1026 –1029; discussion 1029 –1030. 2. Prados MD, Wara WM, Edwards MS, et al. The treatment of brain stem and thalamic gliomas with 78 Gy of hyperfractionated radiation therapy. Int J Radiat Oncol Biol Phys 1995;32: 85–91. 3. Packer RJ, Boyett JM, Zimmerman RA, et al. Outcome of children with brain stem gliomas after treatment with 7800 cGy of hyperfractionated radiotherapy. A Children’s Cancer Group Phase I/II Trial. Cancer 1994;74:1827–1834. 4. Lewis J, Lucraft H, Gholkar A. UKCCSG study of accelerated radiotherapy for pediatric brain stem gliomas. United Kingdom Childhood Cancer Study Group. Int J Radiat Oncol Biol Phys 1997;38:925–929. 5. Kalifa C, Valteau D, Pizer B, et al. High-dose chemotherapy in childhood brain tumours. Childs Nerv Syst 1999;15:498 – 505. 6. Bouffet E, Raquin M, Doz F, et al. Radiotherapy followed by high dose busulfan and thiotepa: A prospective assessment of

7.

8.

9.

10.

11.

12.

high dose chemotherapy in children with diffuse pontine gliomas. Cancer 2000;88:685– 692. Broniscer A, Gajjar A. Supratentorial high-grade astrocytoma and diffuse brainstem glioma: Two challenges for the pediatric oncologist. Oncologist 2004;9:197–206. Barkovich AJ, Krischer J, Kun LE, et al. Brain stem gliomas: A classification system based on magnetic resonance imaging. Pediatr Neurosurg 1990;16:73– 83. Barkovich A. Intracranial, orbital and neck tumors of childhood. In: Wilkins LW, editor. Pediatric neuroimaging. Philadelphia: Lippincott, Williams, and Wilkins; 2000. p. 466 – 468. Nelson SJ, Vigneron DB, Dillon WP. Serial evaluation of patients with brain tumors using volume MRI and 3D 1H MRSI. NMR Biomed 1999;12:123–138. Nelson SJ, Huhn S, Vigneron DB, et al. Volume MRI and MRSI techniques for the quantitation of treatment response in brain tumors: Presentation of a detailed case study. J Magn Reson Imaging 1997;7:1146 –1152. Usenius JP, Kauppinen RA, Vainio PA, et al. Quantitative metabolite patterns of human brain tumors: Detection by 1H NMR spectroscopy in vivo and in vitro. J Comput Assist Tomogr 1994;18:705–713.

30

I. J. Radiation Oncology

● Biology ● Physics

13. Kugel H, Heindel W, Ernestus RI, et al. Human brain tumors: Spectral patterns detected with localized H-1 MR spectroscopy. Radiology 1992;183:701–709. 14. Miller BL, Chang L, Booth R, et al. In vivo 1H MRS choline: Correlation with in vitro chemistry/histology. Life Sci 1996; 58:1929 –1935. 15. Gupta RK, Cloughesy TF, Sinha U, et al. Relationships between choline magnetic resonance spectroscopy, apparent diffusion coefficient and quantitative histopathology in human glioma. J Neurooncol 2000;50:215–226. 16. Simmons ML, Frondoza CG, Coyle JT. Immunocytochemical localization of N-acetyl-aspartate with monoclonal antibodies. Neuroscience 1991;45:37– 45. 17. Bates TE, Strangward M, Keelan J, et al. Inhibition of Nacetylaspartate production: Implications for 1H MRS studies in vivo. Neuroreport 1996;7:1397–1400. 18. Smith JK, Londono A, Castillo M, et al. Proton magnetic resonance spectroscopy of brain-stem lesions. Neuroradiology 2002;44:825– 829. 19. Ricci PE, Pitt A, Keller PJ, et al. Effect of voxel position on single-voxel MR spectroscopy findings. AJNR Am J Neuroradiol 2000;21:367–374. 20. Nelson SJ. Analysis of volume MRI and MR spectroscopic imaging data for the evaluation of patients with brain tumors. Magn Reson Med 2001;46:228 –239. 21. Star-Lack J, Vigneron DB, Pauly J, et al. Improved solvent suppression and increased spatial excitation bandwidths for three-dimensional PRESS CSI using phase-compensating spectral/spatial spin-echo pulses. J Magn Reson Imaging 1997;7:745–757. 22. Nelson SJ, Nalbandian AB, Proctor E, et al. Registration of images from sequential MR studies of the brain. J Magn Reson Imaging 1994;4:877– 883. 23. Lee PL, Gonzalez RG. Magnetic resonance spectroscopy of brain tumors. Curr Opin Oncol 2000;12:199 –204. 24. Leclerc X, Huisman TA, Sorensen AG. The potential of proton magnetic resonance spectroscopy ((1)H-MRS) in the diagnosis and management of patients with brain tumors. Curr Opin Oncol 2002;14:292–298. 25. Graves EE, Nelson SJ, Vigneron DB, et al. A preliminary study of the prognostic value of proton magnetic resonance spectroscopic imaging in gamma knife radiosurgery of recurrent malignant gliomas. Neurosurgery 2000;46:319 –328. 26. Pirzkall A. MR-spectroscopy: Potential for brain tumor diagnosis and guiding therapy. In: Black P, Loeffler JS, editors. Cancer of the nervous system. Philadelphia: Lippincott Williams & Wilkins; 2005; in press. 27. Pirzkall A, Li X, Oh J, et al. 3D MRSI for resected high-grade gliomas before RT: Tumor extent according to metabolic activity in relation to MRI. Int J Radiat Oncol Biol Phys 2004;59:126 –137. 28. Pirzkall A, McKnight TR, Graves EE, et al. MR-spectroscopy guided target delineation for high-grade gliomas. Int J Radiat Oncol Biol Phys 2001;50:915–928. 29. Pirzkall A, Nelson SJ, McKnight TR, et al. Metabolic imaging of low-grade gliomas with three-dimensional magnetic resonance spectroscopy. Int J Radiat Oncol Biol Phys 2002;53: 1254 –1264. 30. Graves EE, Pirzkall A, Nelson SJ, et al. Registration of magnetic resonance spectroscopic imaging to computed tomography for radiotherapy treatment planning. Med Phys 2001;28:2489 –2496. 31. Rock JP, Hearshen D, Scarpace L, et al. Correlations between magnetic resonance spectroscopy and image-guided histopathology, with special attention to radiation necrosis. Neurosurgery 2002;51:912–920. 32. Oh J, Henry RG, Pirzkall A, et al. Survival analysis in patients with glioblastoma multiforme: Predictive value of choline-to-

Volume 62, Number 1, 2005

33. 34.

35. 36. 37. 38.

39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51.

52. 53.

n-acetylaspartate index, apparent diffusion coefficient, and relative cerebral blood volume. J Magn Reson Imaging 2004;19:546 –554. Li X, Jin H, Lu Y, et al. Identification of MRI and 1H MRSI parameters that may predict survival for patients with malignant gliomas. NMR Biomed 2004;17:10 –20. Chang YW, Yoon HK, Shin HJ, et al. MR imaging of glioblastoma in children: Usefulness of diffusion/perfusionweighted MRI and MR spectroscopy. Pediatr Radiol 2003; 33:836 – 842. Girard N, Wang ZJ, Erbetta A, et al. Prognostic value of proton MR spectroscopy of cerebral hemisphere tumors in children. Neuroradiology 1998;40:121–125. Lazareff JA, Gupta RK, Alger J. Variation of post-treatment H-MRSI choline signal intensity in pediatric gliomas. J Neurooncol 1999;41:291–298. Curless RG, Bowen BC, Pattany PM, et al. Magnetic resonance spectroscopy in childhood brainstem tumors. Pediatr Neurol 2002;26:374 –378. Broniscer A, Gajjar A, Bhargava R, et al. Brain stem involvement in children with neurofibromatosis type 1: Role of magnetic resonance imaging and spectroscopy in the distinction from diffuse pontine glioma. Neurosurgery 1997;40:331–337; discussion 337–338. Tzika AA, Astrakas LG, Zarifi MK, et al. Spectroscopic and perfusion magnetic resonance imaging predictors of progression in pediatric brain tumors. Cancer 2004;100:1246 –1256. Warren KE, Frank JA, Black JL, et al. Proton magnetic resonance spectroscopic imaging in children with recurrent primary brain tumors. J Clin Oncol 2000;18:1020 –1026. Mascalchi M, Brugnoli R, Guerrini L, et al. Single-voxel long TE 1H-MR spectroscopy of the normal brainstem and cerebellum. J Magn Reson Imaging 2002;16:532–537. Galanaud D, Le Fur Y, Nicoli F, et al. Regional metabolite levels of the normal posterior fossa studied by proton chemical shift imaging. Magma 2001;13:127–133. Tran TK, Vigneron DB, Sailasuta N, et al. Very selective suppression pulses for clinical MRSI studies of brain and prostate cancer. Magn Reson Med 2000;43:23–33. Sutton LN, Wang Z, Gusnard D, et al. Proton magnetic resonance spectroscopy of pediatric brain tumors. Neurosurgery 1992;31:195–202. Nelson SJ, McKnight TR, Henry RG. Characterization of untreated gliomas by magnetic resonance spectroscopic imaging. Neuroimaging Clin N Am 2002;12:599 – 613. Nelson SJ. Imaging of brain tumors after therapy. Neuroimaging Clin N Am 1999;9:801– 819. Negendank WG, Sauter R, Brown TR, et al. Proton magnetic resonance spectroscopy in patients with glial tumors: A multicenter study. J Neurosurg 1996;84:449 – 458. Kuesel AC, Sutherland GR, Halliday W, et al. 1H MRS of high grade astrocytomas: Mobile lipid accumulation in necrotic tissue. NMR Biomed 1994;7:149 –155. Remy C, Fouilhe N, Barba I, et al. Evidence that mobile lipids detected in rat brain glioma by 1H nuclear magnetic resonance correspond to lipid droplets. Cancer Res 1997;57:407– 414. Prichard JW. What the clinician can learn from MRS lactate measurements. NMR Biomed 1991;4:99 –102. Go KG, Kamman RL, Mooyaart EL, et al. Localised proton spectroscopy and spectroscopic imaging in cerebral gliomas, with comparison to positron emission tomography. Neuroradiology 1995;37:198 –206. Tien RD, Lai PH, Smith JS, et al. Single-voxel proton brain spectroscopy exam (PROBE/SV) in patients with primary brain tumors. AJR Am J Roentgenol 1996;167:201–209. Aboagye EO, Maxwell RJ, Kelson AB, et al. Preclinical evaluation of the fluorinated 2-nitroimidazole N-(2-hydroxy-3,3,3trifluoropropyl)-2-(2-nitro-1-imidazolyl) acetamide (SR-4554) as

1

H-MRSI of pediatric pontine glioma

a probe for the measurement of tumor hypoxia. Cancer Res 1997;57:3314 –3318. 54. Tarnawski R, Sokol M, Pieniazek P, et al. 1H-MRS in vivo predicts the early treatment outcome of postoperative radiotherapy for malignant gliomas. Int J Radiat Oncol Biol Phys 2002;52:1271–1276. 55. Walenta S, Schroeder T, Mueller-Klieser W. Metabolic mapping with bioluminescence: Basic and clinical relevance. Biomol Eng 2002;18:249 –262. 56. Bhakoo KK, Bell JD. The application of NMR spectroscopy to

● A. LAPRIE et al.

31

the study of apoptosis. Cell Mol Biol (Noisy-le-Grand) 1997;43:621– 629. 57. Schlemmer HP, Bachert P, Henze M, et al. Differentiation of radiation necrosis from tumor progression using proton magnetic resonance spectroscopy. Neuroradiology 2002;44:216 – 222. 58. Weybright P, Maly P, Gomez-Hassan D, et al. MR spectroscopy in the evaluation of recurrent contrast-enhancing lesions in the posterior fossa after tumor treatment. Neuroradiology 2004;46:541–549.