Introduction	Choose Top of pageABSTRACTIntroduction <<Subjects and MethodsResultsDiscussionReferencesCITING ARTICLES

Gliomatosis cerebri is a rare tumor, fewer than 300 cases having been reported in the medical literature [1, 2]. In the World Health Organization (WHO) classification of CNS tumors, gliomatosis cerebri is defined as a diffuse glial tumor involving at least two brain lobes. It is often bilateral and frequently extends to infratentorial structures, including the spinal cord [3]. The histogenesis is currently unknown, and this neoplasm has been included in the category of neuroepithelial tumors of uncertain origin.

Gliomatosis cerebri is essentially diagnosed when MRI studies show diffuse and contiguous infiltration of the white matter that is isointense on T1-weighted sequences and hyperintense on T2- and proton density-weighted sequences. Brain structure is preserved, frequently with bihemispheric extension through commissural pathways, such as the corpus callosum, to the basal ganglia and brainstem [4-6]. In most cases, brain biopsy is performed for antemortem diagnosis [7], although it is possible to diagnose this tumor with imaging criteria because the MRI appearance is quite distinctive. Biopsy specimens are often difficult to evaluate [8] often making it impossible to assess the degree of aggressivity of this tumor. In addition, areas of the lesion have varying degrees of cellular proliferation and differentiation, as autopsy studies have shown [9]. Stereotactic brain biopsy does not yield information useful for prognosis in cases of gliomatosis cerebri [8]. Furthermore, the surgical procedure involved in brain biopsy, however minimal, is not exempt from mortality and morbidity [7].

MR spectroscopy (MRS) is a noninvasive diagnostic method that can be used for metabolic characterization of brain tumors. Use of MRS in gliomatosis cerebri has been reported only fairly recently [10], and the largest published series to date is nine cases [8]. The most common finding in gliomatosis cerebri is a decreased level of N-acetyl aspartate (NAA). The existence of a peak in myoinositol-to-glycine ratio has also been described as characteristic of gliomatosis cerebri, especially in the absence of elevated choline (Cho) level [11-13]. In the case of choline, however, there are discrepancies in the literature. Although some authors have reported a moderate increase in this metabolite in patients with gliomatosis cerebri [10-13], others have described normal values [1, 8]. To the best of our knowledge, this discrepancy has not been explained. The purpose of our study was to study the differences, taking into account survival time as an additional possible explanatory variable for the discrepancy in choline levels.

Subjects and Methods	Choose Top of pageABSTRACTIntroductionSubjects and Methods <<ResultsDiscussionReferencesCITING ARTICLES

The subjects were all seven of the patients with gliomatosis cerebri diagnosed at our center from September 1999 through March 2003. Our institution covers an urban population of 650,000. All patients underwent MRI and proton MRS (¹H MRS) on a 1.5-T system (Gyroscan II ACS, Philips Medical Systems). Both the MRI and ¹H MRS studies were performed before treatment was initiated.

The following sequences were used in the imaging study: sagittal T1-weighted spin echo (TR/TE, 409/15), axial FLAIR (8,000/120), axial T2-weighted turbo spin echo (3,500/120), and axial T1-weighted spin echo (600/15) with magnetization transfer contrast enhancement after IV injection of gadopentetate dimeglumine, 0.1 mmol/kg (Magnevist, Schering).

Radiologic inclusion criteria were strict, including not only the WHO criteria (contiguous infiltration of at least two brain lobes) but also those postulated by Galanaud et al. [8]. The latter criteria are no history of focal glioma, no significant mass effect, lesion primarily affecting the white matter and basal ganglia, and no significant enhancement (< 1 cm³) on the MR image after IV contrast injection at the time of diagnosis.

Tumor biopsy was performed on all patients. In one case, biopsy was an open procedure, and in the others, stereotactic technique was used. The results were compatible with the tentative imaging diagnosis. In another case, the histologic specimen was not evaluable because of poor quality; however, the diagnosis of gliomatosis cerebri was confirmed on the basis of imaging and clinical progression criteria.

We collected data on age, sex, initial symptoms, time between initial symptoms and MRS, stereotactic biopsy results, treatment received (radiation therapy and dosage), and survival time from onset of the initial symptom through the date of death. Five patients were treated with whole-brain radiation therapy, receiving total doses of 40-55 Gy. Two patients received no treatment, one because of rapidly progressive deterioration and the other because of refusal by the patient's family. In all cases, the cause of death was secondary to progression of the primary disease with progressive worsening of neurologic symptoms and no contribution by other intercurrent diseases.

The spectroscopic study consisted of single-voxel acquisition localized in the bulk of the gliomatosis infiltration and avoiding unaffected areas. The acquisition protocol entailed a single-voxel point-resolved spectroscopic sequence (1,500/136; number of excitations, 128) [14] with a sample volume of 10.32 ± 7.7 cm³ (mean ± SD). This protocol was selected in an attempt to achieve accurate quantification of the NAA, creatine, and choline peaks.

The control group consisted of 16 recruited healthy volunteers. These subjects underwent MRI with T1-weighted (gradient echo; 15/4.5) and T2-weighted (5,809/120) sequences. The spectroscopic study was performed with a point-resolved spectroscopic sequence with the same parameters as for the patients, localized in the frontal lobe (because this lobe is considered the one with the largest proportion of white matter) with a mean sample volume of 6.59 ± 1.9 cm³.

Processing of the spectroscopic images consisted of decreasing exponential weighting of the data in the time domain with a decay constant of 2 Hz. Before the signal was quantified, the residual water peak was removed by use of a Hankel total least squares algorithm [15]. The spectra were quantified with a nonlinear time-domain analysis and an advanced method for accurate, robust, and efficient spectral fitting algorithm [16]. The NAA (2.01 ppm), creatine (3.02 ppm), and choline (3.22 ppm) peaks were analyzed. These resonances were adjusted to a Lorentzian (decreasing complex exponential) model through acquisition of the signal amplitude for each peak in relation to the T2^* relaxation effects. The NAA and choline amplitude values were normalized to the creatine values.

We can accept that if they correspond to a relatively short period of survival time, the data can be modeled as a linear regression. In this linear model, the age and sex of each patient can be included as independent variables. Unless otherwise stated, all the values are mean ± SD. All subjects or their legal representatives (for two patients, one because of incapacity and one because of minor age) signed an informed consent form before MRI was performed. The entire study had received previous approval by the independent ethics committee of our hospital.

Results	Choose Top of pageABSTRACTIntroductionSubjects and MethodsResults <<DiscussionReferencesCITING ARTICLES

The mean patient age at diagnosis was 40 years (range, 10-64 years). The time between onset of symptoms and performance of ¹H MRS varied from less than 1 week (four cases) to 2 months. The most frequent clinical finding at diagnosis was motor deficits of the corticospinal tract (three patients) followed by headaches, epileptic seizures, and neurosensorial deficits (two patients in each case). At histologic examination the tumor was classified as low-grade astrocytoma in five cases. In one case, tumor aggressivity was not assessed because of the small tissue sample volume. No patient had areas of enhancement after IV contrast injection.

Fitting the data to a linear model with survival time, sex, and age as independent variables (r = -0.92) showed that survival time was statistically significant (t = -3.187, p =0.05) but that age and sex were not (t = -0.841, -0.395; p = 0.462, 0.206, respectively). Although no association was observed between NAA-to-creatine (NAA/Cr) ratio and survival time, comparison between the mean NAA/Cr ratios of control subjects (1.93 ± 0.4) and of patients (1.07 ± 0.7) showed that the average NAA/Cr ratio was significantly lower (t = 3.763, p = 0.001) in patients than in control subjects. In the patient with the shortest survival time, the spectrum showed a characteristic lactate signal doublet at 1.33 ppm (Fig. 1A), presumably associated with anaerobic glycolysis.

Discussion	Choose Top of pageABSTRACTIntroductionSubjects and MethodsResultsDiscussion <<ReferencesCITING ARTICLES

From the epidemiologic perspective, gliomatosis cerebri is a neoplasm that has a peak incidence at the age of 40-50 years [3]. It can, however, affect persons of a broad range of ages, cases having been found both in neonates [17] and persons of advanced age. The natural history of this tumor is variable. Reports in the literature show survival times ranging from 25 days to 22 years [9, 18]. The histologic characteristics of the tumor do not usually reflect the clinical behavior.

In our series, even though all of the evaluable histologic specimens were classified as low-grade tumors, the survival time ranged from 3 to 25 months with a clear lack of association with histologic diagnosis. In our work, we found a negative correlation between survival time and choline level. Because this linear relation was obtained for a relatively short range of survival times, extrapolation of results may not be straightforward. This result suggests that choline value may influence survival time, perhaps explaining the discrepancies in the scientific literature regarding gliomatosis cerebri. Some authors have found a significant increase in choline level compared with the value among control subjects [10-13], whereas others have reported normal choline levels [1, 8]. Because it is well established that the aggressiveness of a tumor is directly related to higher choline level [19], we hypothesize that the histologic malignancy of gliomatosis cerebri may be indirectly reflected as reduction in survival time. Previous reports [1] have described the relation between grade of gliomatosis cerebri and spectroscopic findings. Thus high-grade lesions (WHO grades III and IV) have markedly higher maximum Cho/Cr and Cho/NAA ratios and lower NAA/Cr ratios, and the differences are most pronounced for the Cho/NAA relation. In addition, the presence of lipids or lactate indicates the presence of high-grade gliomatosis cerebri. One of our patients had a spectrum with lactate doublet at 1.33 ppm, and the disease was the most aggressive from the clinical perspective, the survival time being 3 months.

View larger version (26K)

Fig. 1A —Patient spectra. 50-year-old man with gliomatosis cerebri. Survival period was 3 months; sample volume, 8.0 cm3.

View larger version (25K)

Fig. 1B —Patient spectra. 44-year-old man with gliomatosis cerebri. Survival period was 6 months; sample volume, 6.0 cm3.

View larger version (24K)

Fig. 1C —Patient spectra. 23-year-old man with gliomatosis cerebri. Survival period was 9 months; sample volume, 6.75 cm3

View larger version (29K)

Fig. 1D —Patient spectra. 54-year-old woman with gliomatosis cerebri. Survival period was 12 months; sample volume, 27.0 cm3.

View larger version (23K)

Fig. 1E —Patient spectra. 64-year-old man with gliomatosis cerebri. Survival period was 16 months; sample volume, 8.0 cm3.

The existence of a relation between histologic aggressiveness and choline level is difficult to visualize in the clinical setting because gliomatosis cerebri often is diagnosed with biopsy. In most cases it is impossible to sample the area of greatest malignancy within the tumor. In addition, postmortem histologic analysis of the entire tumor is performed on very few patients with gliomatosis cerebri.

View larger version (33K)

Fig. 1F —Patient spectra. 10-year-old girl with gliomatosis cerebri. Survival period was 20 months; sample volume, 4.5 cm3.

View larger version (32K)

Fig. 1G —Patient spectra. 34-year-old man with gliomatosis cerebri. Survival period was 5 months; sample volume, 12.0 cm3.

Although MRS results have correlated with survival time in cases of glioma [20], to our knowledge similar studies have not been conducted in cases of gliomatosis cerebri. In our series, no relation was found between NAA/Cr ratio and survival time. On the contrary, this parameter seems to be associated with Cho/Cr ratio. This phenomenon may be explained by the histopathologic particularities of gliomatosis cerebri, which is characterized by accumulation of tumor cells around the neurons or their axons with little or no invasion of these tumor cells [3, 9, 21]. This finding suggests that cellular turnover, as indicated by the choline peak (membrane generation and destruction), may be related more to tumor proliferation and aggressivity than to the NAA level, which is commonly considered a neuronal marker, in normally functioning neurons. This situation contrasts to the findings for glioma, which does not preserve neuronal structures and in which there is a relation between shorter survival time and decreased NAA level and increased choline level [20, 22]. It would have been interesting to obtain a histologic proliferation index (for example, Ki-67 antigen) to examine its potential correlation with choline level. Unfortunately, it is difficult to obtain a tissue sample from the MRS voxel position within the tumor.

One of the limitations of our study was the use of single-voxel MRS, in which only a volume element is analyzed and spatial diversity is not shown [1]. Areas in which images showed maximum tumoral involvement were selected for the spectroscopic study because they presumably corresponded to the areas of highest histologic aggressivity. Two patients in our series did not undergo radiation therapy. It is not likely, however, that this lack of treatment would have been a confounding factor because it is unclear whether radiation therapy improves outcome among patients with gliomatosis cerebri [23].

To conclude, we suggest that in patients with gliomatosis cerebri, choline level may be associated with survival time. This relation may explain discrepancies in quantification of this metabolite in the absence of other variables, such as histologic grade determined with biopsy, that may clarify the discrepancies.