Anatomic Location is a Risk Factor for Atypical and Malignant Meningiomas

Abstract

Background

Grade II and III meningiomas have higher rates of tumor recurrence than grade I meningiomas after surgery and/or external irradiation. As the utility of non-invasive treatments for brain tumors increases, it is becoming increasingly important to assess the likelihood that a tumor is not benign before treatment initiation. Hence, we have reviewed a large-series of our patients to determine risk-factors for higher-grade pathology with particular interest paid towards tumor location.

Methods

We reviewed 378 patients presenting to our institution from 2000 to 2007 with: histologically confirmed meningioma, central pathology grading according to the WHO 2000 guidelines, and tumor location confirmed with preoperative imaging. We performed univariate and multivariate logistic regression on potential risk-factors for high-grade pathology.

Results

Risk-factors for grade II/III pathology included non-skull base location (twofold) and male gender (two-fold). Patients with prior surgery had a three-fold increased incidence of higher-grade meningiomas at presentation to our center. We controlled for this referral bias by performing a multivariate regression, and analysis without patients receiving prior treatment. 97% of operations were performed for tumor size and clinical symptoms, while less than 3% for interval growth or features concerning of higher-grade pathology.

Conclusion

Non-skull-base meningiomas, male gender, and prior surgery impart increased risk for grade II or III pathology. This increased risk translates to probable poorer prognosis and increased likelihood of recurrence after treatment. Thus, it is prudent to take these specific variables into consideration in conjunction with the complete clinical presentation when advising patients regarding their prognosis.

Introduction

Meningiomas are the most common extra-axial primary brain tumor ¹. Although a majority of these tumors are low grade, a significant proportion will recur after initial treatment ². Literature published since the WHO 2000 classification report higher recurrence rates at five years following surgical excision for WHO grade II (41%) ^{2, 3} and III (70–91%) ^{3, 4} than for WHO grade I lesions (3%) ³. Historically, surgical resection was the standard treatment for meningiomas, however over the past 20 years Gamma Knife radiosurgery (GKS) has provided an increasingly accepted method for non-invasive treatment. In order to minimize patient morbidity, the suggestion has been put forth that specific subsets of patients with meningiomas can be safely observed with adequate clinical and radiologic follow-up or treated with primary GKS without histologic diagnosis ^5–11. Although a conservative treatment approach minimizes or delays patient morbidity, it is not without risk of tumor progression if the interval of follow-up is too long ¹². Without a histologic diagnosis, higher-grade tumors will be missed and inappropriately observed or treated with primary GKS. The risk in treating without biopsy is evident as tumor control rates after GKS are 50% at two years for grade II meningiomas, and 17% at 15 months for grade III meningiomas ⁶.

In order to better identify higher grade tumors prior to treatment initiation, multiple studies have attempted to define risk factors that contribute to increased recurrence or predict higher grade of tumor ^{2, 13–19}. Recently, Sade and colleagues reported that skull base meningiomas have a four fold decreased risk of being atypical or malignant as compared to non-skull base tumors ²⁰. However, many analyses typically work with heterogeneous patient populations and as such it is imperative that the multiple potential confounders be identified and controlled for prior to concluding that skull base meningiomas are less likely to be malignant, especially given the possible conclusion that if a tumor is not likely to be malignant, that it is acceptable to irradiate it without a tissue diagnosis.

The possibility that skull base meningiomas are at decreased risk for high-grade behavior has been suggested for some time ^{21, 22}, however there are fewer papers on the subject attempting to establish the true risk in a statistically sound manner, raising the possibility that the risk reduction conferred by a skull base location has been overestimated by the failure to control for confounders. In this study, we utilize multivariate logistic regression to test the hypothesis that skull base meningiomas are less likely to be a grade II or III meningioma, after controlling for other confounding variables.

Methods

Patient Population

All patients undergoing neurosurgical intervention at UCSF are prospectively enrolled in a database. Using this database, we identified all patients between 2000 and 2007 who underwent evaluation and treatment for meningioma at our institution. Only specimens received since 2000 were included as the WHO classification system changed considerably with the release of the 2000 guidelines ²³. Thus, we wanted to ensure a consistent central pathology grading scheme across all our data. We excluded all patients with any other intracranial tumor history. We excluded all patients with neurofibromatosis type 2. From this cohort, we evaluated all patients undergoing craniotomy for resection of a histologically proven meningioma, which provided a total of 378 patients. This study was approved by the UCSF Committee on Human Research under the approval number H7828-29842-03.

Data Collection and Analysis

Clinical information was retrospectively reconstructed using patient medical records, radiologic data and pathologic specimens from both UCSF and outside medical facilities. All clinical assessments were performed by a neurosurgeon. Patient age was defined by age at the time of surgery. Both the pre-operative, post-contrast T1 magnetic resonance (MR) imaging and/or the surgeon’s operative note were reviewed in order to confirm tumor location. Extensive or multifocal non-skull base tumors, as well as tumors arising from the convexity, falx, and ventricles were considered non-skull base, while all other tumor locations were classified as skull base (Table 1). No spinal meningiomas were included in our data. Central pathology review was performed on the basis of the World Health Organization (WHO II) guidelines from 2000 ²³. For patients who presented with recurrent tumors, primary-operative pathology was performed at an outside institution.

Table 1.

Distribution of meningiomas in our patients by location and WHO grade according to the WHO 2000 classification.

Location	Grade I % (N)	Grade II % (N)	Grade III % (N)	Grade II/III % (N)	N Total	% of Total
Cavernous sinus	67% (2)	33% (1)	0% (0)	33% (1)	3	0.8%
Clinoid	100% (1)	0% (0)	0% (0)	0% (0)	1	0.3%
Clivus	100% (2)	0% (0)	0% (0)	0% (0)	2	0.5%
Convexity	68% (57)	27% (23)	5% (4)	32% (27)	85	22.5%
CPA	93% (25)	7% (2)	0% (0)	7% (2)	27	7.1%
Extensive Convexity	67% (6)	22% (2)	11% (1)	33% (3)	9	2.4%
Falx	80% (49)	18% (11)	2% (1)	20% (13)	61	16.1%
Foramen magnum	92% (11)	8% (1)	0% (0)	8% (1)	12	3.2%
Intraventricular	71% (5)	29% (2)	0% (0)	29% (2)	7	1.9%
Jugular foramen	100% (3)	0% (0)	0% (0)	0% (0)	3	0.8%
Middle fossa	75% (6)	25% (2)	0% (0)	25% (2)	8	2.1%
Multifocal	83% (10)	8% (1)	8% (1)	16% (1)	12	3.2%
Olfactory groove	71% (10)	29% (4)	0% (0)	29% (4)	14	3.7%
Orbit	100% (7)	0% (0)	0% (0)	0% (0)	7	1.9%
Parasellar	100% (4)	0% (0)	0% (0)	0% (0)	4	1.1%
Petroclival	100% (5)	0% (0)	0% (0)	0% (0)	5	1.3%
Petrous	100% (2)	0% (0)	0% (0)	0% (0)	2	0.5%
Planum	92% (12)	8% (1)	0% (0)	8% (1)	13	3.4%
Posterior Fossa	100% (21)	0% (0)	0% (0)	0% (0)	21	5.6%
Sphenoid wing	85% (33)	14% (6)	0% (0)	14% (6)	39	10.3%
Tentorium	100% (2)	0% (0)	0% (0)	0% (0)	2	0.5%
Tuberculum	85% (35)	15% (6)	0% (0)	15% (6)	41	10.8%
Total	82% (309)	16% (62)	2% (7)	18% (69)	378	100%

Abbreviations: CPA=cerebellopontine angle

Statistical Analysis

Univariate analysis was used to identify covariates that might affect the combined rate of Grade II and III meningiomas in these patients. Risk factors for WHO Grade II and III tumors were selected for analysis based on a priori hypotheses from previously published literature as well as theoretical concerns. Binary and categorical variables were compared using Pearson’s χ² test or the χ² test for trend, respectively. Continuous variables were compared using an independent samples t-test, after statistical demonstration of the normality of the data.

Risk factors that impacted combined frequency of WHO Grade II and III tumors with a significance of p = 0.2 or less on univariate analysis were included in a stepwise binary logistic regression. All odds ratios on multivariate analysis reflect the odds of having a WHO Grade II or III meningioma at the time of surgery. The goodness of fit of the regression model was confirmed by demonstrating a non-significant p-value on the Hosmer-Lemeshow test.

We tested interaction terms betweeneach of the three variables that significantly impacted risk for WHO Grade II/III tumors on multivariate analysis. The statistical significance of the interactions was assessed with the use of backward stepwise regression, inwhich statistical significance was estimated by means of thelikelihood-ratio test to assess the effect of removing interaction terms for all strata of the given variable. After finding that none of the interaction terms would significantly (unadjusted p > 0.2 for all terms) alter the log likelihood of the regression model if removed, we calculated the adjusted risk ratios without adjusting for interactions.

Continuous variables are presented as mean ± standard error. Statistical tests were considered significant when the two-sided p < 0.05 after correcting for multiple comparisons using the Bonferroni method. All statistical tests were performed using SPSS version 16.

Results

Patient Population

378 patients underwent craniotomy for removal of a histologically proven meningioma. The mean age was 54 (±0.94) and the range was 15 to 90. 73% of the patients were female. The frequencies of WHO Grade I, II, III, and II or III lesions were 82%, 16%, 2%, and 18% respectively. Of note, 8% of our patients had undergone prior surgery. Additional demographic data is presented in Table 2.

Table 2.

Clinical characteristics of patients in this series.

N = 378	Range	Mean
Age	15–90	54.0
	Frequency	Percent
Race
Asian	34	9%
Black	17	5%
Caucasian	271	72%
Hispanic	28	7%
Pacific Islander	5	1%
Other	16	4%
Unknown	7	2%
Female	278	73%
Male	100	27%
>65	93	25%
<65	285	75%
Total
WHO Grade I	309	82%
WHO Grade II	62	16%
WHO Grade III	7	2%
WHO Grade II +
III	69	18%
Non-Skull Base	166	44%
Skull Base	212	56%
Preop Deficit	27	7%
Prior Surgery	30	8%
Priorx XRT	27	7%

Abbreviations: XRT=radiation therapy

Risk Factors Associated With Tumor Grade

Table 3a and 4a demonstrate the results of univariate and multivariate logistic regression analysis respectively for our entire patient population. A wide variety of tumor locations were represented in our sample. 56% of tumors arose from the skull base while 44% were non-skull base lesions. Patients with non-skull base lesions were significantly more likely to have WHO Grade II or III tumors (27% vs. 12%, p < 0.001). This relationship upheld in a multivariate analysis revealing a two-fold risk increase for non-skull base locations (OR 2.13, 95% CI 1.2 – 3.8, p < 0.01). Convexity meningiomas are frequently observed and surgical intervention may only occur for large tumors (>3cm), symptom development, interval growth, or radiographic findings concerning for aggressive disease course. We analyzed for any potential selection bias by reviewing the reason for operation in all cases of convexity meningiomas. We found less than 3% of patients with a convexity meningioma underwent surgery for interval growth or atypical MRI features, while over 97% of these patients underwent surgery because of tumor size greater than 3 cm or for significant clinically debilitating symptoms.

Table 3.

Univariate analysis of potential risk factors for atypical and malignant pathology for all patients (A) and those without any prior treatment (B).

Table 3a
	Mean	SE	Significance
Age WHO I	54	0.8	p = 0.82
Age WHO II + III	54	2.0
WHO Grade II + III	Frequency	Percent	Significance
Age >65	22	24%	p = 0.12
Age < 65	47	17%
Male	30	30%	p < 0.001
Female	39	14%
Skull Base	25	12%	p < 0.001
Non-Skull Base	44	27%
Preop Deficit	9	33%	p < 0.05
No Preop Deficit	59	18%
Prior Surgery	15	50%	p < 0.001
No Prior Surgery	54	16%
Prior XRT	10	37%	p < 0.05
No Prior XRT	59	17%

Table 3b
	Mean	SE	Significance
Age WHO I	54	0.8	p = 0.31
Age WHO II + III	56	2.3
WHO Grade II + III	Frequency	Percent	Significance
Age >65	16	20%	p = 0.14
Age < 65	33	13%
Male	20	26%	p < 0.005
Female	29	12%
Skull Base	20	10%	p < 0.005
Non-Skull Base	29	22%
Preop Deficit	4	24%	p < 0.35
No Preop Deficit	44	15%

Abbreviations: XRT=radiation therapy

Table 4.

Multivariate analysis of risk factors for WHO grade II or III pathology for all patients (A) and those without any prior treatment (B).

Table 4a
WHO Grade II+ III	OR	95% Confidence Interval		Wald	Significance
		−	+
Non-Skull Base	2.13	1.20	3.79
Male	2.10	1.17	3.76	6.30	p < 0.01
Prior Surgery	3.47	1.50	8.04	8.46	p < 0.01
Age > 65	1.50	0.82	2.75	1.74	p = 0.19
Preop Deficit	1.60	0.60	4.26	0.89	p = 0.35
Prior XRT	1.40	0.55	3.60	0.49	p = 0.49

Table 4b
WHO Grade II + III	OR	95% Confidence Interval		Wald	Significance
		−	+
Non-Skull Base	2.18	1.06	4.48
Male	2.42	1.26	4.65	6.99	p < 0.01
Age > 65	1.14	0.27	4.77	0.73	p = 0.86

Abbreviations: XRT=radiation therapy, HR=hazard ratio

Prior surgery for tumor excision was also found to be strongly associated with increased risk for high-grade pathology on univariate analysis (50% vs. 16%, p < 0.001) and was the strongest risk factor in multivariate analysis as well (OR 3.5, 95% CI 1.5 – 8.0, p < 0.01). Seventy percent of patients who had received prior surgery presented to our institution with non-skull base meningiomas as compared to 42% of those who presented to our institution for a first operation (p < 0.01).

Although 70% of the patients were female, we found that a disproportionate amount of males had higher-grade pathology. Univariate analysis demonstrated increased risk for male gender (30% vs. 14%, p < 0.001) for grade II and III tumors. On multivariate analysis, males had a two-fold increased risk for higher-grade tumors (OR 2.10, 95% CI 1.2 – 3.8, p < 0.01). Patients who had received prior radiation therapy had a significantly increased risk for atypical or malignant meningiomas (37% vs. 17%, p < 0.05) in univariate analysis, although this relationship was no longer significant on multivariate regression. Similarly, the presence of a pre-operative deficit conferred increased risk for grade II and III pathology on univariate analysis (33% vs. 18%, p < 0.05), but was not significant in multivariate regression. Lastly, on univariate analysis there was a trend for patients over the age of 65 to have increased risk of grade II/III tumors (24% vs. 17%, p = 0.12). We therefore included age over 65 as a covariate in our multivariate regression, however it was not found to be a significant risk factor for atypical or malignant pathology.

Because prior treatment may have introduced an element of bias into our methods, even after attempting to control for this factor by performing a multivariate analysis, we reanalyzed our patient population after excluding those patients with any prior treatment (57 patients excluded). The univariate and multivariate results for this analysis are presented in table 3b and 4b respectively. This subgroup analysis further confirmed our findings.

Discussion

Determining factors that can accurately predict meningioma grade prior to tissue diagnosis will help guide clinicians towards optimal treatment by helping to appropriately balance the risk of surgical morbidity with the need for tissue diagnosis. In this study, we demonstrated that non-skull base meningiomas are twice as likely to be grade II or III, after controlling for other factors that contribute to high-grade tumor risk. Additionally, our analysis demonstrated that patients who have had prior surgery for meningioma resection are over three times as likely to have a grade II or III meningioma at recurrence than those tumors at initial presentation. Further, we found that male gender also imparts a two-fold risk for a patient to have an atypical or malignant meningioma.

There have been varying reports in the literature regarding the association of meningioma grade and anatomic location ^{20–22, 24}. Sade and colleagues ²⁰ found a larger risk reduction for skull base meningiomas (four-fold) compared to our study (two-fold). It is possible that our 95% confidence intervals may overlap, however they did not report this statistic. Thus, the differences seen could be related to either different sampling or the fact that we controlled for confounders such as prior treatment, age, and gender. In our experience, patients with skull base meningiomas are more likely to receive their initial surgery at tertiary academic centers, whereas community based neurosurgeons are somewhat more comfortable treating straightforward convexity or falcine meningiomas. Our data confirmed this hypothesis, as patients who had undergone prior surgery were more likely to have a non-skull base meningioma, and had a greater than three-fold increased risk for a grade II or III tumor at the time of second surgery. This trend likely reduces the fraction of grade I convexity and falcine tumors seen at our institution, and the patients with higher grade tumors are more likely to recur and be referred to an a tertiary care center for treatment of their tumor. Regardless, non-skull base lesions appear to have an increased risk for grade II or III pathology. The mechanism underlying this risk difference may result from the distinct embryologic origin of skull-base and non-skull base dura. Discrete tissue origin may lead towards a propensity to develop different histologic neoplastic subtypes ^{20, 25, 26}. These different subtypes have been associated with different genetic mutations which could theoretically lead to varying degrees of aggressive behavior ^{27, 28}.

It appears that non-skull base and skull base meningiomas have distinct risk-benefit profiles based on both differential rates of atypia/malignancy and the increased risk of morbidity associated with operation on skull base pathology. The high rates of atypia and malignancy (27%) amongst the convexity and parasagittal cohorts and the relatively poor tumor control rates published for grade II and III meningiomas treated with GKS ^{2, 4, 12–14, 29, 30}, further support the current practice of surgery for most of these tumors for the establishment of tissue diagnosis, relief of mass effect, and durable tumor control. This is additionally supported by the good morbidity profile associated with resection of meningiomas in non-skull base locations using modern techniques ³¹. After establishment of a histologic diagnosis and tumor resection, GKS could potentially be used in the management of residual or recurrent disease. In addition, given that the risk of grade II or III pathology is not extremely low (12%) it could be argued that surgery should still play a role in establishment of tissue diagnosis and treatment of skull base meningiomas.

We found that male gender also conferred a greater than two-fold risk of having a grade II or III lesion, after controlling for other variables. Although females have an overall increased incidence of meningiomas, other studies have also reported an increased risk for atypical and malignant meningiomas associated with male gender ^{21, 32}. The mechanism by which male gender increases risk is currently unclear. However, some insight into the biology of this difference can be gained from the scientific literature which points towards differences in hormone levels, hormone receptor status, and sex chromosome genetic variation. The concept of distinct tumor biology in association with tumor grade and a connection to gender is supported by evidence demonstrating that benign meningiomas have a high level of progesterone receptor expression relative to atypical and malignant meningiomas ^33–35. Clinical and histopathologic studies have shown an inverse relationship between progesterone receptor expression level and both WHO grade and recurrence ^{35, 36}. Genetic studies have alluded to differential gene expression restricted to the sex chromosomes in meningioma cells ³⁷. Atypical and malignant meningiomas, lacking the hormone receptor expression profile, may be a different entity in their molecular pathogenesis and genetic studies are already pointing to candidate oncogenes and molecular markers that separate grade I from grade II and III meningiomas ^38–40.

There are limitations to our study that should be acknowledged. Notably, this is a retrospective study, and subject to all the limitations of data collection inherent to this study design. Further, our whole patient population was derived from a tertiary care center enriched to a degree with higher-grade tumors, and a tissue diagnosis was a requisite inclusion criterion for this study. Thus, despite our attempts to consciously control for this referral bias in our patient data, we cannot avoid the fact that our population may not be completely representative of the entire meningioma population. The odds ratio therefore, may not accurately reflect a more typical distribution of pathologic subtype that might be seen in the community setting.

In conclusion, we have found that non-skull base location, male gender, and prior surgery are all independent pre-operative risk factors for WHO grade II or III meningiomas. This increased risk translates to probable poorer prognosis and increased likelihood of recurrence after treatment. Thus, it is prudent to take these variables into consideration in conjunction with the complete clinical presentation when advising patients regarding their prognosis. Additionally, when compiled with other risk factors in a more complex prediction model, our findings might be used as an adjunct to help guide treatment decisions. Our findings also provide impetus to stimulate and, in part, guide further treatment innovation and understanding of the biology underlying meningiomas.