Proposal for a new risk stratification classification for meningioma based on patient age, WHO tumor grade, size, localization, and karyotype

Patients and Samples

Overall, 302 patients (91 males, 211 females; mean age of 60 ± 15y; range, 6y–87y), who had been diagnosed with meningioma at the Neurosurgery Service of the University Hospital of Salamanca, were prospectively included in this study. Participants were enrolled in the study after informed consent had been given in accordance with the guidelines of the local Ethics Committee and the Declaration of Helsinki. The great majority of participants (283/302; 94%) underwent complete tumor resection; this included 35 of 38 grade II/III tumors (n = 30 and 8 cases, respectively). Adjuvant radiotherapy was given, in addition to surgery, to 21 participants with WHO grade II/III tumors. One participant with anaplastic meningioma also received systemic chemotherapy in addition to radiotherapy.

Histological tumor diagnosis was performed according to the WHO criteria.¹⁷ At the moment of closing this study, 42 of 302 participants (14%) had relapsed after a median follow-up of 65 months. From the whole series, 41 cases were excluded from survival analyses because follow-up data were not available (n = 30 cases) or the participant had died shortly after surgery (n = 11 cases). Follow-up studies were performed according to a standard clinicobiological protocol that included MRI techniques performed 3 months after surgery and every 12 months thereafter. Additional MRI studies were performed whenever clinical signs and/or symptoms were noted and/or a relapse was suspected. Immediately after surgical removal, part of the fresh tumor tissue was frozen in liquid nitrogen and stored at −80°C until used for iFISH and/or SNP array studies. In addition, EDTA-anticoagulated peripheral blood (PB) samples were also obtained from each participant at diagnosis and processed in parallel with the tumor.

iFISH Studies

For all freshly frozen tumor samples obtained after surgery, iFISH analyses were performed to identify the numerical alterations of 11 distinct chromosomes by using a panel of probes specific for those chromosomes and chromosomal regions more frequently altered in meningiomas (1p36/1q25, 7, 9p34, 10, 14q32.3, 15q22, 17q21, 18q21, 22q11.2, X and Y). All probes were obtained from Vysis Inc. and were used in double stainings, as previously described in detail:¹⁸ LSI BCR/ABL dual-color probe for chromosomes 9 and 22; LSI PML/RAR-α dual-color probe for chromosomes 15 and 17; LSI IgH/BCL2 dual-color probe for chromosomes 14 and 18; 1p36/1q25 dual-color probe for chromosome 1; CEP 7 and 10 DNA probes conjugated with Spectrum Orange and Spectrum Green, respectively; and CEP X (Spectrum Orange) and CEP Y (Spectrum Green) for chromosomes X and Y, respectively.

Copy Number Alterations by SNP Arrays

Copy number (CN) alterations were analyzed by SNP arrays in a subset of 50 samples previously reported in the literature using the GeneChip Human Mapping 250K Nsp and 250K Sty arrays (Affymetrix) according to the manufacturer's instructions, as previously described in detail.¹⁹ DNA from paired (frozen) tumor tissue and normal PB samples was purified using the QIAamp DNA Mini Kit (Qiagen) according to the manufacturer's instructions. DNA purity and integrity were determined with a NanoDrop-1000 spectrophotometer (Nano-Drop Technologies) and by conventional electrophoretic procedures in 1% agarose gel, respectively. The SNP call rate per array was always ≥92% (range, 92%–99.8%). Overall, 200 .CEL files were obtained. In addition, another external series of 82 meningiomas²⁰ analyzed by 100K Affymetrix SNP arrays, which also had data about relevant tumor characteristics and patient survival available at the GEO public database (access code: GSE16583), was also included in this part of the study. Follow-up data were available for 108 of 132 of these cases, and 17 of them (13%) had relapsed after a median follow-up of 59 months (range, 1–115 months).

For the analysis of SNP array data, the GCOS version 1.3 (Affymetrix), the Copy Number Analysis Tool (CNAT v4.0, Affymetrix), dChip 2007 (http//www.dchip.org; Dana Farber Institute), and the GeneChip Genotyping Analysis (GTYPE 4.1; Affymetrix) software programs were used. CN values were calculated for each SNP array and plotted according to chromosomal localization. Genotypes were generated using the BRLMM algorithm included in the Genotyping Console software version 3.0.2 (Affymetrix). Based on the results obtained from normal PB samples, cutoff values of ≤1.30 and ≥2.50 (arbitrary units) were used to establish CN losses and gains, respectively.

Statistical Methods

Standard deviation values, range, median and mean were calculated for all continuous variables, and frequencies were used for categorical variables. To establish the statistical significance of differences observed between groups, the Student t test and the Mann–Whitney U test were used for continuous variables, whereas the chi-square test was used for qualitative variables. Relapse free-survival (RFS) curves were plotted according to the method of Kaplan–Meier, and the (1-sided) log-rank test was used to establish the statistical significance of differences observed between survival curves. Multivariate analysis of prognostic factors for RFS was performed using the Cox stepwise regression model. In this part of the study, only those variables showing a significant association with RFS in the univariate analysis were included. For all statistical analyses, the SPSS version 15.0 software package was used (SPSS Inc). P values <.05 were considered to be associated with statistical significance.

Histopathological Features and Cytogenetic Profile of Meningiomas

Two-hundred sixty-four meningiomas (87%) were WHO grade I tumors, while only 30 cases (10%) were grade II and 8 (3%) were grade III meningiomas. Around half of the tumors corresponded to meningothelial meningiomas (n = 147; 49%). The rest of the tumors were 57 (19%) transitional, 34 (11%) psammomatous, 24 (8%) atypical, 17 (5%) fibroblastic, 7 (2%) angioblastic, 5 (2%) anaplastic, 5 (2%) secretory, 3 (1%) chordoid, and 2 (1%) rhabdoid meningiomas; the remaining case was a papillary tumor. According to tumor localization, most meningiomas (n = 276; 91%) corresponded to intracranial tumors (35% cranial base, 21% convexity, 18% parasagittal, 12% falcine, 3% tentorial, and 2% intraventricular; only 26 (9%) were spinal meningiomas. Brain edema was found in 173 of 302 (57%) cases and evaluated according to its extension as light (smaller or equal to the volume of the tumor) in 72 cases (24%), moderate (double the volume of the tumor) in 65 cases (21%), and severe edema (more than twice the volume of the tumor) in 36 cases (12%).

From the 302 meningiomas analyzed, 90 (30%) displayed no cytogenetic alterations by iFISH for the 11 chromosomes analyzed. In contrast, the other 212 (70%) cases showed numerical alterations for ≥1 chromosome. As expected, chromosome 22 was the most frequently altered chromosome (173/302 cases; 57%), its alteration mainly consisted of monosomy 22/22q deletions (166/173 altered cases). Other recurrently altered chromosomes included chromosome Y in males (26/91 cases; 28%), chromosome 1p (67/302 cases; 22%), chromosome 14 (47/302 cases; 16%), chromosome X in females (28/211 cases; 13%), chromosome 1q (37/302 cases; 12%), and chromosome 10 (31/302 cases; 10%). Overall, chromosomal losses were more frequently observed than gains (55% vs 2% for chromosome 22, 26% vs 2% for chromosome Y, 21% vs 1% for chromosome 1p, 13% vs 3% for chromosome 14, 12% vs 1% for chromosome X, and 6% vs 4% for chromosome 10), except for chromosome 1q, which was more frequently gained (10%) than lost (2%). All other chromosomes analyzed were found to be altered at lower (≤8%) frequencies (Supplementary Table 1).

Around half of all cytogenetically altered cases (n = 106; 35%) showed isolated alterations of a single chromosome, whereas the other half (n = 106; 35%) displayed cytogenetic alterations (losses and/or gains) involving ≥2 chromosomes (Table 1). Among those cases carrying isolated chromosomal alterations, the most frequent pattern consisted of loss of one chromosome 22 or del(22q) (83/106 cases; 78%), while isolated involvement of other chromosomes was restricted to a few cases. Isolated loss of chromosome 1p was found in 9 tumors, loss of chromosome Y in 4, and loss of chromosome 10 in 2. Losses of chromosomes X and 9 were found in one case each, and; isolated gains of chromosomes 1q and 14 were detected in 5 and one cases, respectively.

Except for chromosome 10 in the whole series and chromosomes Y and X in males, alterations of all other individual chromosomes were significantly more frequent among WHO grade II/III versus grade I tumors (P ≤ .001; Supplementary Table 1). In line with these findings, diploid tumors and cases with single chromosomal alterations were more frequently (P ≤ .001) observed among grade I meningiomas. Eighty-two of 90 (91%) diploid cases and 101 of 106 (95%) tumors with only one altered chromosome corresponded to WHO grade I tumors, while more complex cytogenetic patterns predominated among grade II/III meningiomas; 25 of 38 (66%) grade II/III tumors showed a complex karyotype. Of note, tumors with complex karyotypes represented more than half of the male cases (51/91 cases; 56%), whereas they only accounted for 26% of female cases (55/211 cases; P ≤ .001).

Prognostic Impact of Tumor Cytogenetics and Other Relevant Clinical and Histopathological Features of the Disease

Tumor cytogenetics, as assessed by iFISH, showed a significant association with the incidence of relapses and participants' RFS (Supplementary Table 1). Accordingly, alterations of chromosomes 1p, 1q, 7, 9, 10, 14, 18, and 22 in the whole series and chromosome X in females were associated with both a higher incidence of relapses (P < .05; Supplementary Table 1) and/or a significantly shorter RFS (P < .03; Supplementary Table 1 and Supplementary Fig. 1). Multivariate analysis, including all individual chromosomes with a significant impact on RFS in the univariate study, showed that only the alteration of chromosome 14 (P = .001), gains of chromosome 7 (P = .047) and losses of chromosome 18 (P < .001) had an independent predictive value for a poorer outcome (Supplementary Table 1). In turn, when the 3 major iFISH cytogenetic subgroups of meningiomas were considered, participants with tumors carrying a complex karyotype showed a significantly shorter RFS than those showing a diploid karyotype and presence of isolated chromosomal alterations (75% RFS of 78 months vs not reached, respectively; P = .001) (Table 1; Fig. 1H). Of note, individual chromosomes lost their independent prognostic value once these 3 cytogenetic profiles were included in the analysis (data not shown). Participants carrying tumors with complex karyotypes showed lower 5-, 10-, and 15-year RFS rates than those with diploid karyotypes or isolated chromosomal alterations (79% ± 5% vs 94% ± 3% and 91% ± 3%; 63% ± 7% vs 90% ± 6% and 88% ± 4%; and 55% ± 8% vs 78% ± 12% and 80% ± 7%, respectively) (Table 1). Of note, within those cases with an isolated chromosomal alteration, participants who had isolated monosomy 22/del(22q) showed a slightly better RFS than those with isolated alterations of other chromosomes: 75% RFS not reached versus 82 months (P > .05); 5-, 10- and 15-years RFS rates of 92% ± 4% versus 89% ± 8%; 92% ± 4% versus 74% ± 15%; and 82% ± 8% versus 74% ± 15%, respectively (Table 1). Of note, combined assessment of chromosomes 1p, 7, 14, 18, and 22 represented the minimal chromosomal panel that could be applied for identification of the 3 most common subtypes of genetic patterns used in the prognostic scoring system.

In addition to tumor cytogenetics, several other clinical and histopathological features of the disease also showed an adverse impact on participants' RFS, namely younger age (<55 years; P = .005), male sex (P = .02), WHO grade II/III tumor (P < .001), atypical or anaplastic tumor histopathology (P < .001), intraventricular or anterior cranial base tumor localization (P < .001), tumor size >50 mm (P < .001), and presence of brain edema (P = .01) (Table 1 and Fig. 1A–G). Multivariate analysis of prognostic factors showed that tumor cytogenetics (P = .002), together with the WHO grade (P < .001), localization (P < .001), size (P = .001), and participant age (P = .01), represented the best combination of independent prognostic factors for predicting RFS of meningioma cases (Table 1). Of note, the Simpson grade of resection did not show a significant (P > .05) impact on participant RFS.

Based on the above 5 variables (patient age, tumor cytogenetics, WHO grade, localization, and size), a prognostic score was established for each individual participant using the criteria described in Table 2. Once this score was applied, participants were divided into 4 risk groups: (i) low-risk cases with a score ≤1 (n = 62), (ii) intermediate-1 (int-1) cases with a score between 2 and 4 (n = 170), (iii) intermediate-2 (int-2) cases with a score between 5 and 6 (n = 59), and (iv) high-risk cases with a score ≥7 (n = 11). Participants included in these 4 risk categories showed progressively shorter 75% RFS rates (P < .001) from low (not reached) to int-1 (not reached), int-2 (38 months), and high-risk cases (15 months). Similarly, progressively lower 5-, 10- and 15-year RFS rates were also found for low (100% ± 0%); int-1 (93% ± 2%, 85% ± 5%, and 75% ± 7%, respectively); int-2 (70% ± 7%, 59% ± 16%, and 45% ± 11%, respectively), and high-risk cases (50% ± 16%, 0% ± 0%, and 0% ± 0%, respectively) (Table 2; Fig. 1I). For comparison purposes, RFS curves for the same series of cases, as defined according to the scoring system previously proposed by Maillo et al,¹⁵ are shown in Fig. 1J.

Most interestingly, when WHO grade I tumors were exclusively considered, tumor cytogenetics retained their prognostic significance (P = .01; Table 3; Fig. 2D) together with patient age, tumor localization, and size. Thus, complex karyotypes, age < 55 years, intraventricular and anterior cranial base localization, and a tumor size ≥50 mm were all associated with shorter 75% RFS rates both in the univariate (P = .01, P = .02, P < .001, and P = .002, respectively; Table 3; Fig. 2A–C) and in the multivariate analyses (P < .001, P = .007, P = .002 and P = .03, respectively; Table 3). When the proposed score was specifically adjusted for WHO grade I tumors (Table 2), the significantly different outcome of low-, int-1, int-2, and high-risk cases was retained, with 75% RFS rates of not reached, not reached, 51 months, and 15 months, respectively (P < .001; Table 2; Fig. 2E). This clearly improved the prediction obtained with the previously proposed Maillo et al score, particularly for the low-risk group (Fig. 2F).

Validation of iFISH Profiles by High-density Copy Number Arrays

To validate the iFISH cytogenetic profiles, the prognostic value of CN profiles obtained for the 24 human chromosomes by SNP arrays was evaluated in an additional group of 132 cases pooled from our institution (n = 50)¹⁹ and another series from the literature (n = 82)²⁰ with publicly available data. Based on SNP arrays, 41 (31%) of these 132 cases showed no CN alterations for any of the 24 chromosomes evaluated, 40 cases (30%) showed isolated alterations of a single chromosome (including monosomy 22/del[22q] in 35/40 cases), and 51 participants (39%) displayed complex karyotypes. Once again, the majority of WHO grade II/III tumors showed a CN pattern compatible with a complex karyotype (27/34 cases; 79%), whereas low grade meningiomas most frequently had a diploid karyotype or isolated alterations of a single chromosome (74/98 cases; 76%). In addition, meningioma cases carrying complex karyotypes associated with CN alterations of ≥2 chromosomes also showed a poorer outcome than cases with a diploid karyotype or isolated chromosomal alterations (P = .005; Fig. 3A). Similarly, we could also confirm in this validation series the improved predictive value of the new prognostic scoring system defined above (Fig. 3B) over the Maillo et al score (Fig. 3C), particularly for the low-risk category.