Evaluation of the growth rates and related prognostic factors in radiation-induced meningiomas

Literature dedicated to growth patterns and growth rate influencing factors of radiation-induced meningiomas (RIMs) is limited. To deliver new insights into the topic, a volumetric growth analysis of RIMs was performed. This single-center, retrospective cohort study included patients diagnosed with intracranial meningioma who received radiation treatment at least > 5 years before the RIM diagnosis. Volumetric analysis of individual RIMs was performed using 3D volumetry at the time of RIM diagnosis and during follow-up. RIM growth was determined by calculating absolute (AGR), and relative (RGR) growth rates. Prognostic factors associated with RIM growth were evaluated. A total of 26 patients with 33 meningiomas were enrolled in the study and radiologically/clinically followed up during a median duration of 5.6 years (IQR 3.9–8.8 years). Median AGR was 0.19 cm3 per year and the median RGR was 34.5% per year. Surgically managed RIMs were more likely fast-growing compared to observed ones based on the AGR (p < 0.002). The recurrence rate after total resection was 14.3%. Younger age at RIM diagnosis was associated with higher tumor growth (RGR ≥ 30%, p = 0.040). A significant correlation was found between the length of latency period and the RGR (p = 0.005). To diagnose RIM as early as possible comprehensive MRI surveillance is required. Younger patients with shorter latency periods may profit from shortened MRI intervals, with further management being dependent on the growth rate and eventual symptomatology.


Introduction
Radiotherapy (RT) of the cranium is a major treatment modality for intracranial and extracranial tumors, pediatric patients notwithstanding. As the long-term survival of childhood cancer reaches approximately 80% [1], the likelihood of late RT side effects increases, including the development of subsequent brain tumors. Among these, meningiomas are the most common tumors following cranial RT [2][3][4]; the risk of development is assessed as one in eight at the age of 40 years [5][6][7]. To be classified as radiation-induced meningioma (RIM), the following criteria must be fulfilled: development in the radiation field, different pathology from the primary neoplasm, and at least 5 years of latency after RT treatment [8][9][10][11]. Stated improvements in childhood cancer survival along with widespread use of brain magnetic resonance imaging (MRI) in follow-up protocols have led to increased incidence of RIMs. Concurrently, optimal management is a matter of debate, since RIMs appear to be clinically more aggressive than sporadic meningiomas, tend to occur in multiple locations, and recur after surgery [12]. Small asymptomatic tumors may be amenable to a watchand-wait strategy, while those presenting with neurological symptomatology require neurosurgical resection or stereotactic radiosurgery [13]. Furthermore, the growth rate of RIMs has not been widely studied [14], particularly using volumetric methods. Volumetric analysis of RIMs growth patterns would be meaningful for both the patient and oncologist since patients treated in their childhood with RT for cancer are undergoing regular brain imaging with small, asymptomatic meningiomas being frequently diagnosed. In order to further elucidate the growth patterns of RIMs, we conducted a retrospective volumetric study of RIMs diagnosed in patients years after initial radiation treatment as part of their treatment for their primary tumor.

Study design and patient population
In this retrospective, single-center cohort study, patients diagnosed with RIM who received cranial radiation treatment during childhood (until and including 18 years of age) and at least five years before the meningioma diagnosis were included. Only tumors diagnosed between 2005 and 2020 were included since older MRI scans were not available in the digital format required for accurate measurements. Most patients were diagnosed incidentally via MRI as part of our comprehensive tumor follow-up protocol for their primary tumor. Data regarding primary tumor characteristics (tumor biology, location), performed oncological treatment (radiation dose and area, chemotherapy agents, duration of initial therapy) and therapy outcome (remission or disease recurrence) were collected. Demographic data included gender, age at diagnosis of the primary tumor, and age at RIM diagnosis. Latency period was defined as the time between the end of radiation therapy for the primary tumor and the RIM diagnosis (first appearance on MRI). RIM was considered to be symptomatic when newly developed symptoms were primarily attributable to the meningioma. Ki67 immunohistochemical analysis was performed on formol fixed and paraffin embedded tissues of all surgically resected tumors. Ki67 index was calculated as ratio of positive to all tumor cells in selected microscopic field of vision with the highest activity.

Imaging and volumetric measurements
Data regarding the multiplicity of meningiomas (single vs multiple), location (individual locations, non-eloquence/ eloquence, side), presence of edema, distinct tumor border, regular shape, calcifications, signal intensity on T2 MRI (iso-/hypo-/hyperintense in relation to gray matter) were collected. Only tumors with at least two sequential MRI examinations spanning a minimum of 2 months were included. Tumors that underwent surgery immediately after diagnosis (without another MRI) were excluded.
Data on tumor volume characteristics were obtained according to two independent measurements based on 3D tumor volumetry (3DV) performed on serial axial slices using the 3D Slicer software [15]. MRI sequences using slice thickness of 1 mm or less were used for manual delineation of the tumor on each slice. These thin-slice sequences are a routine part of our MRI protocol for tumors. Dural tails and hyperostosis (none present) were not included. Measurements were performed by two clinicians (an experienced neurosurgeon not involved in the surgeries and a neuroradiologist) blinded to the clinical data of individual patients. In cases of significant differences, a third independent analysis was performed. The final tumor volume was calculated as the mean value from two (of potentially three) most similar results.
The definition of RIM growth was determined according to literature data, adjusted for radiological characteristics, including the baseline tumor volume, of RIMs analyzed in this study: fast-growing and slow-growing RIMs were defined according to the absolute growth rate (AGR) ≥ 0.2 cm 3 , and AGR < 0.2 cm 3 per year (value naturally splitting our groups into slow-and fast-growing ones), respectively, and analogously according to the relative growth rate (RGR) ≥ 30%, and RGR < 30% per year, respectively [16,17]. The final definition of growth was determined according to AGR and/or RGR.

RIM management and follow-up
Patients included in this study were either actively observed, underwent surgical or radiosurgical intervention. Intervention was recommended for fast-growing or symptomatic tumors or in accordance with patient preference after extensive discussion. Extent of surgical resection was evaluated according to the Simpson grading system [18]. Observed RIMs were monitored with serial MRIs (typically once every 6 months). Follow-up length for RIMs included in growth analysis was calculated as the time from RIM diagnosis until the last performed MRI before the intervention, end of study (October 2022), or death. Follow-up was continued after the intervention to assess for potential tumor recurrence.

Statistical analysis
The normality of the data was evaluated according to the Kolmogorov-Smirnov test. Normally distributed continuous data were compared using t-test or ANOVA for multiple comparisons. Mann-Whitney U test or Kruskal Wallis tests were performed for nonparametric testing. Differences between categorical variables were evaluated using the chisquared test, and the correlation between variables was calculated using the Pearson correlation coefficient. In terms of observing the growth patterns of RIMs, we calculated (1) the absolute growth rate (AGR), defined as the increase in tumor volume per year, along with (2) the relative growth rate (RGR), defined as the percentage increase in tumor volume per year. Data analysis was performed using STATISTICA software (TIBCO Software). OriginPro software (OriginLab Corporation) was used to conduct graphical interpretations.

Baseline patient and tumor characteristics
A total of 26 patients (46.2% males, 53.8% females) diagnosed with overall 33 RIMs between 2005 and 2020 that met our inclusion criteria were analyzed. Additional 7 RIMs were treated surgically immediately after the initial MRI due to size and symptoms and were therefore not included in the study. The most frequent diagnosis and the primary indication for radiotherapy was medulloblastoma in 7 patients (26.9%), followed by astrocytoma in 4 (15.4%) and ependymoma in 3 patients (11.5%). Most of the primary tumors were supratentorial (n = 14/26, 53.8%). The mean age at initial radiation treatment was 7.2 ± 5.3 years.
Of these 26 patients, the majority were diagnosed with solitary RIM (n = 21/26, 80.8%), while the remaining five patients developed multiple RIMs (four patients harbored two RIMs, and one patient developed four RIMs). In six patients (23.1%) new symptomatology led to MRI examination, although in none of the patients the symptoms could be directly attributable to the RIM. The remaining patients were truly asymptomatic (n = 20/26, 76.9%).
The median total follow-up (time interval from the initial diagnosis of the primary tumor to the final radiological and clinical examination) was 27 years (IQR = 22-33 years). The median follow-up from the RIM diagnosis to the last recorded radiological examination was 5.6 years (IQR = 3.9-8.8 years). The demographic data of included patients and characteristics of analyzed RIMs are presented in Table 1 and in Online Resources 1 and 2 in further detail.

Primary tumor therapy
Cranial radiation doses were available for all 26 patients and ranged from 12 to 55.8 Gy, with a median of 50 Gy (IQR = 40-53 Gy). The range of used fractions was 1.5-1.8 Gy/fraction/day. We did not find a statistically significant correlation between the radiation dose and subsequent tumor growth rate. Most patients received chemotherapy (96.2%) as a part of the primary tumor management. Data regarding the chemotherapy are available in Online Resource 1.

RIM management and treatment outcome
Of overall 33 diagnosed RIMs included in the study (Fig. 1), 16 RIMs (48.5%) were eventually surgically resected with a median time from the RIM discovery to surgical intervention of 30 months (IQR = 25-46 months), and 5 tumors (15.2%) were managed radiosurgically with a median time to radiosurgery of 46 months (IQR = 24-81 months). The differences between the times to a particular intervention modality were not statistically significant (p = 0.345). Twelve RIMs remain under observation (n = 12/33, 36.4%). Total resection was achieved in 14/16 tumors (87.5%). Subtotal resection was achieved in two patients (12.5%, Fig. 1). One of these patients underwent two repeated radiosurgical interventions due to progression of residual tumor (9 and 61 months after surgery). The residual tumor remained stable in the other patient during 98 months of follow-up. In addition, 2 patients recurred after total resection; one required additional surgery after 58 months, the other remains under observation. Total recurrence rate was 18.8% (n = 3/16). The time to recurrence was 4, 12 and 100 months. Remaining patients from the group of treated meningiomas are under observation with no recurrence diagnosed (median follow-up of 68 months (IQR = 45-104 months). Two patients died during the follow-up period (n = 2/26, 7.7%), one due to glioblastoma (65 months after RIM diagnosis), and one due to progressive encephalopathy (46 months after RIM diagnosis).
Histopathological analysis was performed in all surgically resected RIMs, with an average Ki67 of 5.1% (range 1.5-10%) among 16 RIMs. Only one of the analyzed RIMs was classified WHO Grade II (6.3%), the remaining were Grade I.

RIM volume and growth characteristics
The median baseline volume of all RIMs was 0.2 cm 3 (IQR = 0.08-0.39). There was a significant difference between initial baseline tumor volume when considering age at RIM diagnosis. Patients over ≥ 30 years were more likely to be diagnosed with larger RIM when compared to younger patients (p = 0.013, Table 2).
The median AGR per year was 0.19 cm 3 (IQR = 0.04-0.43 cm 3 ) while the median RGR per year was 34.5% (IQR = 22.3-82.9%). The median tumor doubling time was 20.4 months (IQR = 10.0-37.9 months). Of the 33 RIMs, 15 were slow-growing and 18 were considered to be fastgrowing when defining the growth according to both AGR and RGR.
Testing prognostic factors for RGR, there was a significant difference for the age at the point of the RIM diagnosis in relation to fast-growing RIMs (p = 0.040). Patients harboring tumors with RGR ≥ 30% were younger when compared to patients having slow-growing tumors (RGR < 30%, Table 2). A significant difference was seen in the lengths of latency periods, when fast-growing tumors with RGR ≥ 30% had shorter time intervals of latency periods (median of 216 months (IQR = 169-246 months) compared to slow-growing RIMs (median of 270 months (IQR = 259-306 months), p-value 0.005). Additional tested variables such as the radiation dose, initial baseline tumor volume, and performed treatment did not render any significance in relation to the RGR. Testing prognostic factors for AGR yielded no significant difference on any of the above-mentioned variables besides the fact that patients who underwent surgery more likely had fast-growing tumors compared to observed patients (p < 0.002). A separate analysis using AGR and RGR as a combined criterion revealed the same results as the analysis with RGR alone.
Growth patterns observed based on RIM growth tendencies are presented in graphs in Figs. 2 and 3 and a summary of observed correlations and prognostic factors of tumor growth is presented in Table 2.

Discussion
In this single-institution study of 26 patients with 33 RIMs, we found an absolute growth rate of 0.19 cm 3 per year and a relative growth rate of 34.5% per year. Unsurprisingly, surgically treated tumors demonstrated a higher growth rate in comparison to patients undergoing observation or radiosurgery. In addition, younger age at RIM diagnosis and shorter latency period were correlated with increased RGR although not AGR.
These findings are in sharp contrast to a recent study by Gillespie and Islim, where a mean growth rate of 0.62 cm 3 per year and RGR of 72% per year were described for RIM [14]. Notably, the mean baseline volume of 4.9 cm 3 found in their study was also significantly higher (almost 25 times) when compared to ours. There are several explanations for this discrepancy. First, we are potentially dealing with different cohorts. For their group, most patients treated were referrals from other centers and potential self-selection for more aggressive RIM is possible as noted by the authors. Such selection bias was minimized since the pediatric oncology department in our institution served the whole country until two decades ago, and patients are regularly followed into adulthood (median total follow-up time of 27 years). Therefore, the tumor was detected in most cases at a very early stage. Second, small tumors can considerably increase in size which may be only reflected in the RGR and not the AGR. A small tumor can double or triple in size and be indicated for surgery, without necessarily displaying a high AGR [17]. Despite, given that the RGR was more than twice higher compared to our study, it suggests that their cohort was indeed faster growing. Third, it is not well known what type of growth pattern RIMs follow. A more quiescent phase preceding a phase of exponential growth, as proposed for sporadic meningiomas [19][20][21], is possible. As the tumor volume is considerably higher in the study of Gillespie and Islim, it suggests that those tumors were growing for a significant period before being diagnosed. This is supported by their long latency period (34.4 years) contrary to ours (20.4 years), the number of symptomatic cases (52%) contrary to ours, and the already mentioned greater baseline volume.
Currently, there is no consensus on what constitutes growth, fast/slow growth, or significant growth for meningiomas, much less for RIM. Therefore, even in recently published papers, different authors use different criteria to report on the same endpoints [14,17,19,22]. Adopting the growth criteria used by Gillespie and Islim (≥ 2 cm 3 AGR or ≥ 1 cm 3 AGR + ≥ RGR 30%) only 3 out of 33 (9%) meningiomas fulfilled the criterion for fast growth in contrast to 39.7% in their study [14]. Given that 21/33 of RIM (63%) underwent intervention during the observed study period, this criterion seemed to underappreciate the potential for small tumors to become problematic. Taking an increase of RGR alone by more than 15% per year as suggested by others [19,22], this number increases to 28 out of 33 (84.8%) Fig. 1 Flowchart showing treatment algorithm of diagnosed RIM in time, in relation to the treatment modality used and the treatment outcome meningiomas in comparison to 95.9% in their study, which on the other hand seemed to overestimate the risk. Given that there should be relevance associated with used criteria, particularly for clinicians to identify those patients at "risk" for early intervention, we found neither criterion above to be particularly useful for our cohort. Thus, we decided to use values for AGR and RGR that would naturally split our cohort into a fast and slow growing cohort and used an AGR of ≥ 0.2 cm 3 /year and a change in RGR by ≥ 30% as a cutoffs to define fast growth (see Fig. 3). Based on this, 54.5% of our cohort fulfilled the criteria and we have used these to further investigate prognostic factors for growth.
We confirm the findings of Gillespie and Islim, that show younger age at diagnosis and surgically treated tumors to be correlated with tumor growth rate [14]. While it is established that the latency period is correlated with the dose of radiotherapy dose [4,[23][24][25], we support previous findings [14] that the dose is not correlated to the tumor growth rate. Another correlation was found between the length of the latency period with the RGR (p = 0.005), a finding not highlighted before. This suggests that patients diagnosed earlier may profit from more frequent follow-up examinations in comparison to older patients. As illustrated in Table 2, the correlation of RGR with age and latency period did not reach statistical significance when doing the same analysis with the AGR. This lack of correlation can best be explained by the minimal tumor volume at the time of diagnosis as discussed above. Unlike Gillespie and Islim, we did not find a statistically significant correlation between the baseline volume and the subsequent RIM growth rate, which could be again due to the differences in baseline volumes in both of our studies and the likelihood of larger tumors to be diagnosed in a point in time, where they exhibit exponential growth [21,26].
Our study poses new important questions regarding the growth pattern of these tumors and the correct criteria for growth measurement to identify RIM at risk for intervention. We also suggest that based on our data younger patients with . The second group consists of RIMs with a slow-growth pattern determined as AGR < 0.2 cm 3 per year or RGR 30% per year. Of note is, that based on the prediction model from Fig. 2, even apparently slow-growth tumors (RGR < 30%) will show more aggressive growth patterns after a longer observation time, which are seen in the fast-growth tumors (RGR ≥ 30%). By using RGRs of individual tumors, predictions of further RIM growth patterns are demonstrated. This model is based on exponential growth patterns commonly seen in meningiomas [21,26], when taking into account the classification of (1) no growth, (2) linear growth, and (3) exponential growth. Considering the previous graph shown in Fig. 2, most of the tumors follow an exponential growth pattern. The model presented in this Figure demonstrates "the worst-case scenario" and serves as an illustration and further suggestion for close MRI surveillance as an important approach to efficient decision-making regarding the subsequent treatment or further observation. Of note is, that based on the prediction model from Fig. 2, even apparently slow-growth tumors (RGR < 30%) will show more aggressive growth patterns after a longer observation time, which are seen in the fast-growth tumors (RGR ≥ 30%) short latency periods profit from "tighter" screening intervals after diagnosis. Ultimately, more studies that observe the natural history of RIMs over a longer period will be needed to fully elucidate their growth pattern and examine which growth criteria are most relevant.

Study strengths and limitations
Most patients were treated for their initial tumor at our hospital having the only pediatric oncology center in the whole Czech Republic until 1998, thus ensuring minimal selection bias and oncological treatment consistency. Additionally, MRI screening programs at our institution allowed us to evaluate MRI scans from the moment of diagnosis and before, giving us an insight into the early tumor dynamic, which is missing in most studies.
Our results were predominantly limited by the sample size and the retrospective nature of this study. Additionally, heterogeneous follow-up periods for some tumors, including some short time intervals after resection might underestimate the true recurrence rate that could be observed in a long-term perspective. Another factor is the small number of radiosurgically managed patients, consequently not allowing more in-depth comparison.

Conclusion
Our results suggest that young patients who underwent RT as part of cancer treatment during childhood require frequent MRI surveillance to diagnose RIM as early as possible. MRI follow-up intervals should be shortened in patients with shorter latency periods. Further management can then be adjusted based on growth rate and/or eventual symptomatology.