Preferential brain locations of low-grade gliomas

Comparison with glioblastomas and review of hypothesis




The objectives of this study were to register the brain locations in a consecutive series of low-grade gliomas (LGGs) and compare these localizations with the locations of de novo glioblastomas (GBMs) that were collected during the same period in an effort to analyze whether LGGs are situated in preferential areas and to review the pathophysiologic hypothesis of such a phenomenon.


One hundred thirty-two patients with LGG and 102 patients with GBM who were followed consecutively between 1996 and 2003 by the authors were reviewed, whatever their treatment. Using anatomic, three-dimensional magnetic resonance imaging, the location of each tumor was analyzed accurately according to a classification system based on the proximity of eloquent areas previously reported by the authors.


One hundred nine LGGs (82.6%), compared with 55 GBMs (53.9%), were situated within functional regions (P < 0.001). More specifically, 36 LGGs (27.3%), compared with 11 GBMs (10.8%), were localized in the region of the supplementary motor area (SMA) (P < 0.001); and 33 LGGs (25%), compared with 11 GBMs (10.8%), were located within the insula (P < 0.001).


The current findings suggest that LGGs are located preferentially in “secondary” functional areas (immediately near the so-called primary eloquent regions), especially within the SMA and the insular lobe. This preferential localization may be explained by developmental, cytomyeloarchitectonic, neurochemical, metabolic, and functional reasons. A better knowledge of the pathophysiologic mechanisms underlying preferential LGGs locations may improve understanding of the genesis and natural history of these tumors and, subsequently, their management. Cancer 2004. © 2004 American Cancer Society.

Although recent studies have argued in favor of an impact of surgery on the natural history of low-grade gliomas (LGGs),1 the resection of LGGs remains controversial, mainly because of a high risk of postoperative sequelae.2 This risk appears to be due essentially to the fact that infiltrative LGGs often are observed near or within eloquent regions,2 which explains the use of preoperative and intraoperative functional mapping methods to minimize morbidity.3 Although a more frequent frontal location has been mentioned occasionally4 without specifying the exact part of the lobe that was invaded, to our knowledge, there is no dedicated study that has tried to determine whether LGGs significantly involve specific (functional) brain areas.

Herein, we compare two consecutive series of patients with LGGs and de novo glioblastomas (GBMs) (i.e., no secondary anaplastic LGGs) who were followed by the same physicians during the same period, and we analyze the respective hemispheric locations of these two kinds of tumors. The objectives were to determine whether LGGs arise more frequently in preferential areas and to review the pathophysiologic hypothesis of such a phenomenon.


Between 1996 and 2003, 234 patients who had histologically confirmed supratentorial LGGs or GBMs were managed consecutively by the authors with simple stereotactic biopsy or surgical resection. Two series were discriminated in this study: the first study (S1), which included only patients who had LGGs (World Health Organization [WHO] Grade 2), and the second study (S2), which included only patients who had de novo GBMs (WHO Grade 4), with the exclusion of patients who had secondary anaplastic LGGs and Grade 3 gliomas—to avoid the possible inclusion of Grade 2 gliomas that evolved to a higher grade prior to diagnosis. For all patients in both S1 and S2, the presenting symptoms and neurologic examinations were reviewed.

Furthermore, the topography of each tumor was analyzed on a preoperative magnetic resonance image (T1-weighted and/or spoiled-gradient images before and after gadolinium enhancement in the three orthogonal planes and T2-weighted axial images). Each tumor was graded functionally relative to its location with respect to eloquent brain in two groups: 1) tumors that involved functional regions, according to the definition that we previously proposed,3 (e.g., the primary sensorimotor area, supplementary motor area, speech centers, visual cortex, and insula) and 2) tumors near or remote from these eloquent areas at the time of diagnosis.

Due to the infiltrative features of gliomas, to classify the tumors in these two groups, we took into account the location of the epicenter of the lesion. Namely, with regard to LGGs, any slight signal abnormality in any functional area did not suffice to classify the tumor into Group 1 if the epicenter was located in nonfunctional tissue. In addition, for GBMs, only the location of contrast enhancement was used to decide whether the tumor was located in an eloquent region and not the perienhancement signal abnormality (although it may have contained viable tumor cells).


The clinical and radiologic data from the 234 patients are summarized in Table 1.

Table 1. Brain Locations of Low-Grade Gliomas versus Glioblastomas
LocationFunctional areasNonfunctional areas
SitesLGGGBMChi-square P valueSitesLGGGBMChi-square P value
  1. LGG: low-grade glioma; GBM: glioblastoma; SMA: supplementary motor area; L: left; R: right; PMC: premotor cortex; FO: frontal operculum; NS: nonsignificant; MI: primary motor area; SI: primary somatosensory area; VI: primary visual area; PTO: parietotemporooccipital junction.

Frontal (65 LGG, 33 GBM)SMA36 (12R, 24L) (27.3%)11 (5R, 6L) (10.8%)< 0.001Frontopolar/prefrontal areas6 (4R, 2L) (4.5%)9 (3R, 6L) (8.8%)NS
 L PMC/L FO (Broca)12 (9%)3 (2.9%)NSR PMC/R FO5 (3.8%)5 (4.9%)NS
 Precentral gyrus (MI)6 (4R, 2L) (4.5%)5 (3R, 2L) (4.9%)NS    
Parietal (12 LGG; 9 GBM)Retrocentral gyrus (SI)8 (5R, 3L) (6%)5 (3R, 2L) (4.9%)NSR superior parietal lobule4 (1R, 3L) (3%)4 (2R, 2L) (3.9%)NS
     Anterior and midtemporal7 (3R, 4L) (5.3%)16 (12R, 4L) (15.7%)0.008
Temporal (13 LGG; 20 GBM)L posterior temporal6 (4.5%)2 (1.97%)NSR posterior temporal2 (1.97%)NS
PTO (8 LGG; 28 GBM)L PTO7 (5.3%)17 (16.7%)0.004R PTO1 (0.75%)11 (10.8%)< 0.001
Occipital (1 LGG; 1 GBM)VIIL (0.75%)IR (0.98%)NS    
Insula (paralimbic) (33 LGG; 11 GBM)Insula33 (25R, 8L) (25%)11 (5R, 6L) (10.8%)< 0.001    
Total (132 LGG; 102 GBM)109 LGG (82.6%)55 GBM (53.9%)< 0.00123 LGG (17.4%)47 GBM (46.1%)< 0.001

S1: LGGs

Clinical presentation

S1 included 67 males and 65 females (mean age, 36 years; range, 17–63 years). The presenting symptoms were seizures in 95% of patients. All patients had a Karnofsky performance status between 80 and 100 with no or only a slight deficit.

Tumor location

In S1, 109 tumors were located within eloquent corticosubcortical regions (82.6%). Thirty-six LGGs (27.3%) involved the precentral frontomesial structures (12 right and 24 left) (i.e., the region of the supplementary motor area [SMA], including the SMA proper and the pre-SMA with or without the anterior cingulum [which was invaded in 28 tumors]) (Fig. 1). Thirty-three LGGs (25%) invaded the insula (25 right and 8 left) with or without the orbitofrontal cortex (7 tumors), the temporopolar cortex (7 tumors), or both (15 tumors) (i.e., the so-called “paralimbic system”) (Fig. 2). Moreover, 8 LGGs (6%) involved the primary somatosensory area (5 right and 3 left), and 6 LGGs (4.5%) involved the primary motor area (4 right and 2 left). Twenty-five LGGs (18.9%) invaded the language centers, including 12 LGGs (9%) in the premotor/Broca area, 6 LGGs (4.5%) in the left posterior temporal areas, and 7 LGGs (5.3%) in the left parietotemporooccipital junctions. Finally, 1 LGG (0.75%) involved the left primary visual area. Conversely, 23 LGGs (17.4%) invaded “noneloquent” regions, including 6 LGGs (4.5%) in the frontopolar/prefrontal areas (4 right and 2 left), 5 LGGs (3.8%) in the right premotor cortex/frontal operculum, 4 LGGs (3.9%) in the right superior parietal lobules, 7 LGGs (5.3%) in the anterior/midtemporal regions (3 right and 4 left), and 1 LGG (0.75%) in the right parietotemporooccipital junction.

Figure 1.

(A) An axial, T2-weighted magnetic resonance image (MRI) and (B) a sagittal, T1-weighted MRI show a “typical” left frontomesial, precentral, low-grade glioma involving the supplementary motor area.

Figure 2.

(A) An axial, T2-weighted magnetic resonance image (MRI) and (B) a coronal, T1-weighted MRI show a “typical” right, low-grade glioma involving the insular lobe.

S2: De Novo GBMs

Clinical presentation

S2 included 58 males and 44 females, (mean age, 68 years; range, 56–90 years). The presenting symptoms were a neurologic deficit in 50% of patients, intracranial hypertension in 30% of patients, and seizures in 20% of patients.

Tumor location

In S2, 55 GBMs (53.9%) involved eloquent corticosubcortical regions, including 11 GBMs (10.8%) in the SMA regions (5 right and 6 left), 11 GBMs (10.8%) in the insula regions/paralimbic system (5 right and 6 left), 5 GBMs (4.9%) in the primary somatosensory areas (3 right and 2 left), 5 GBMs (4.9%) in the primary motor areas (3 right, 2 left); 22 GBMs (21.57%) in the language centers, including 3 GBMs (2.9%) in the premotor/Broca area, 2 GBMs (1.97%) in the left posterior temporal areas, and 17 GBMs (16.7%) in the left parietotemporooccipital junction; and 1 GBM (0.98%) in the right primary visual area. Conversely, 47 GBMs (46.1%) invaded “noneloquent” regions, including 9 GBMs (8.8%) in the frontopolar/prefrontal areas (3 right and 6 left), 5 GBMs (4.9%) in the right premotor cortex/frontal operculum, and 4 GBMs (3.9%) in the superior parietal lobules (2 right and 2 left); 16 GBMs (15.7%) in the anterior/midtemporal regions (12 right and 4 left), 2 GBMs (1.97%) in the right posterior temporal regions, and 11 GBMs (10.8%) in the right parietotemporooccipital junction.

Comparison of Glioma Locations Between S1 and S2

LGGs were located significantly more frequently within eloquent regions compared with GBMs (P < 0.001); whereas GBMs more often invaded “nonfunctional” regions, in particular the anterior/midtemporal regions (P = 0.008). More specifically, LGGs involved two structures significantly more often than GBMs: the SMA (P < 0.001) and insula (P < 0.001) regions; whereas GBMs more often invaded the parietotemporooccipital junction (P < 0.004).


Brain Locations of LGGs

Although LGGs often were observed in eloquent areas,2 explaining the risk of their resection and, thus, the use of functional mapping methods to decrease surgical morbidity,3 to our knowledge, no specific study has attempted to evaluate whether these tumors have preferential brain locations. In the current study, we observed 1) that LGGs are situated more frequently in eloquent corticosubcortical areas compared with de novo GBMs and, more specifically 2) that LGGs are located significantly more often within the SMA and insular regions.

Although we collected a consecutive series that included more LGGs than high-grade gliomas (although the literature reports that high-grade gliomas are substantially more common than LGGs), thus suggesting a possible referral bias due to the expertise of our department in surgery for LGGs,3 the exclusion of Grade 3 gliomas may explain this ratio. Moreover, the current study data appear to be in accordance with numerous surgical series, which reported resection of LGGs specifically involving the SMA5 and the insula.6 These results also may indicate that environmental factors favoring the genesis of LGGs and GBMs differ, at least in part, complementing recent molecular biology data that also argue in favor of different intrinsic mechanisms between these two kinds of tumors.7

Pathophysiologic Hypothesis

Several hypotheses may be considered to interpret the reason for such preferential locations of LGGs in the SMA and insular regions. First, there are some cytoarchitectonic and chemoarchitectonic similarities. Indeed, the insular cortex is divided into three belts, from anterior to posterior, on the basis of a gradual cytoarchitectonic change. These include 1) an agranular belt on the anterior one-third of the insula; 2) a transitional, dysgranular belt in layers without complete laminar differentiation; and 3) a posterior granular belt with a well defined granule cell layer that occupies the posterior one-third of the insula.8 In parallel, the precentral frontomesial structures also are divided into an agranular cortex (SMA-proper, pre-SMA, and anterior cingulum) and a dysgranular-to-granular posterior cingulum.9 Thus, the so-called “SMA region” appears to be a transitional architectonic area between the agranular primary motor cortex and the granular, “multimodal” cortex. In the same way, transmitter receptor studies using cytochrome oxidase, acetylcholinesterase, and nicotinamide adenine dinucleotide phosphate-diaphorase staining demonstrated a lightly stained region on the anteroinferior part of the insula, in which neuronal somata predominate, with an intermediate profile between that of a primary area and a high-order association area.10

Second, SMA and insula have a close functional role. Indeed, both represent a functional interface between a multimodal area (prefrontal cortex) and a primary area (sensorimotor area for both SMA and insula plus the auditory cortex for insula). More specifically, both structures play a role in planning: the SMA in planning movements11 and the insula in planning speech.12

Third, because these two areas appear to represent an architectonic and functional interface, it could be hypothesized that particular interactions may exist between neurons and glia in these regions. Indeed, it is well known that glial cells play a role 1) in neuronal migration,13 which may explain the existence of migration disorders in some cortical epilepsies,14 including the extratemporal epilepsy that often originates from the SMA15 and insula16; 2) in the regulation of synaptic transmission17; 3) in the control of synapse numbers18; and 4) in the energy metabolism of the neuron, explaining the neurovascular and metabolic decoupling in gliomas.19 Consequently, if we consider that both the SMA and the insula have partly similar, particular structural and functional profiles, some repercussions with regard to the biology of the local glial cells are likely. Thus, it might be suggested that such possible changes in the local glial properties may favor the development of LGGs in these preferential corticosubcortical locations.

Although they are speculative, these nonexhaustive hypotheses are based on strong, significant, epidemiologic results concerning the rate of LGG locations, and they seem plausible considering the literature currently available in the fields of cytochemoarchitectonics, structural-functional relations, and neuron-glia communication; however, they certainly do not represent the sole explanations. If these findings are confirmed by future studies, which need to be coupled with recent progress in the fields of molecular genetics of gliomas20 and molecular manipulation of stem cells in situ,21 then they may provide a better understanding of the genesis and natural history of LGGs (possibly concerning the patterns of growth), and may improve the management of these tumors.