Effect of osteoporotic conditions on the development of peritumoral brain edema after LINAC-based radiation treatment in patients with intracranial meningioma

This study included patients from the NOVALIS registry. The NOVALIS registry was designed for prospective research of patients who received radiation treatment in our hospital [8]. We investigated all consecutive patients with intracranial meningioma from the registry who underwent linear accelerator (LINAC)-based radiation treatment for the first time at our hospital from July 7, 2014 to September 30, 2020.

All meningiomas were diagnosed by radiologic findings alone or pathological confirmation after surgical tumor resection. All radiologic findings were confirmed by experienced neuro-radiologists. We defined PTBE as newly developed PTBE or the progression of preexisting PTBE in follow-up imaging with newly developed neurologic symptoms after radiation treatment [3]. To identify PTBE after radiation treatment, we only included patients with meningioma in the study who met all following conditions: (1) follow-up for at least 6 months, (2) at least one follow-up imaging (CT/magnetic resonance imaging [MRI]), (3) no preexisting PTBE except for patients who underwent surgery for the meningioma before radiation treatment, and (4) measurable intercortical space of the frontal skull on brain CT scan. The last imaging follow-up period after radiation treatment was examined in all patients.

The detailed radiation technique in our hospital was previously described [8]. The NOVALIS Tx system (Varian Medical Systems, CA, USA; Brainlab, Feldkirchen, Germany) was used to treat all meningioma patients. We used noninvasive thermoplastic masks for the simulation-CT for radiation treatment and during the radiation treatment in all patients. To improve the precision of radiation treatment, the Novalis ExacTrac image system and robotic couch of the NOVALIS Tx system were used to adjust the patients’ positions based on the information from the real-time image acquisition.

We used the 3D treatment/planning systems of the NOVALIS Tx, including iPlan (Brainlab, Feldkirchen, Germany) and Eclipse (Varian, CA, USA), for the radiation planning using MRI/CT-fusion images of the patients. The gross tumor volume (GTV), clinical target volume (CTV), and planning target volume (PTV) were automatically calculated by the 3D treatment/planning system of the NOVALIS Tx. We defined the GTV as the exact enhanced area of the meningioma on contrast-enhanced T1-weighted MRI images. For the operated patients, the GTV was defined as the postoperative resection cavity (if available) plus the enhanced area of residual tumor without inclusion of the PTBE area. In patients without surgery and for Grade I benign meningioma, the CTV was identical to the GTV. For Grade II and III meningiomas, the CTV was usually defined as 1–2 cm margin added to the GTV [9]. When the tumor was located near an organ at risk, we reduced the expansion of the CTV margin near the area of the tumor that was close to the organ at risk. The PTV was defined as a symmetrical 0- to 2-mm expansion from the CTV.

We defined the fractionated stereotactic radiotherapy (FSRT) as > 10 sessions (1.8–2.0 Gy/fraction), hypofractionated stereotactic radiotherapy (hFSRT) as 6 to 10 fractions, hypofractionated SRS (hf-SRS) as 2 to 5 fractions, and stereotactic radiosurgery (SRS) as a single session treatment [10, 11]. To compare treatment doses between patients who received radiation with various fractionations, we calculated the biologically effective dose (BED) based on the following equation: BED = nd × (1 + d/3), where n is the fraction, d is the dose of one fraction, and α/β = 3 [12].

We used the simulation-CT images (Philips Brilliance Big Bore CT Simulators) for radiation planning to measure the frontal skull HU values in all patients of the study cohort. We previously described the detailed methods for measuring HU values on frontal cancellous bone on brain CT [6‐8, 13]. The frontal skull HU values were measured at each of the four lines on the frontal cancellous bone between the right and left coronal sutures at the CT slice that the lateral ventricles disappear on the brain CT (Additional file 1: Fig. 1). To avoid including cortical bone, all CT images were magnified for the HU value measurement.

We previously reported the Skull HU and BMD (SHUB) registry in our hospital [6]. In addition to the previous registry (from January 1, 2010 to December 31, 2016), we further enrolled patients (> 18 years old) who had both procedure codes for dual-energy X-ray absorptiometry (DXA) (NMF03) and brain CT (RCG01A and B) in our hospital between January 1, 2017 and December 31, 2019. Following the same protocol as before, the lowest T-score value for patients who underwent multiple DXA scans was used for the analysis. All CT images were obtained using a Siemens CT scanner in our hospital with continuous slices, no gap, and 4.0–5.0-mm slice thickness [6]. When patients received multiple brain CT scans, the brain CT image closest to the date of the selected DXA scan was used. To reduce the time interval heterogeneity, we excluded patients with more than 3 years between DXA and brain CT. Based on the study showing a slow progression to osteoporosis in postmenopausal women, we believe that our within 3-year time interval between the DXA and brain CT scans may be suitable for investigation of the relationship between frontal skull HU values and BMD [14]. In addition, patients with excessively narrow intercortical space of the frontal skull on brain CT were excluded. Therefore, 2025 patients were finally included in the updated SHUB registry.

The BMD was assessed in the lumbar spine (L1–L4) and femoral neck using a Discovery Wi DXA system (Hologic, Bedford, MA, USA) in all patients of the SHUB registry. The lower T-score between the lumbar spine and femoral neck was used as the T-score for the registry. Based on the World Health Organization T-score classification, we defined osteoporosis as a T-score ≤ − 2.5, osteopenia as a T-score > − 2.5 and ≤ − 1.0, and a normal BMD as a T-score > − 1.0.

A receiver operating characteristic (ROC) curve analysis was performed to identify the optimal cut-off values of mean skull HU for predicting osteopenia and osteoporosis in the patients of the SHUB registry. Mean frontal skull HU values were used as the test variable, and the individual BMD classification was entered as the state variable (dependent variable) in the ROC curve analysis. When we identified the cut-off skull HU value for predicting osteopenia, we coded the normal BMD (T-score > − 1.0) as 0 and the osteopenia and osteoporosis BMD (T-score ≤ − 1.0) as 1 and input the state variable. In the osteoporosis model, we coded the normal and osteopenia BMD (T-score > − 2.5) as 0 and osteoporosis (T-score ≤ − 2.5) as 1 and input the state variable.

The cumulative hazard for development of PTBE was calculated by Kaplan–Meier analysis classified by hypothetical BMD classification, with censoring of patients who had no PTBE on their last brain CT/MRI after radiation treatment start. Hazard ratios (HRs) with 95% confidence intervals (CIs) were then calculated using a multivariate Cox regression analysis to determine whether the possible osteoporotic condition independently predicted PTBE occurrence in patients with meningioma after radiation treatment. Sex, age (continuous variable), BMI (continuous variable), classification of mean skull HU, location (continuous variable), GTV (continuous variable), BED (continuous variable), fractionation (continuous variable), hypertension, and diabetes were entered into the multivariate model. To balance heterogeneities in patient sex, patient age distribution, tumor location, tumor volume, and radiation dose between the PTBE (−) and PTBE (+) groups, additional propensity score-matched analysis was performed using a multivariable logistic regression model. The model was based on patient sex, patient age, tumor location, GTV, and BED. We matched the PTBE (+) group to the controls in a 1:2 ratio using the greedy nearest-neighbor method using R software [15].

A p value < 0.05 was considered statistically significant. All statistical analyses were performed using R software version 3.6.3 and SPSS for Windows, version 24.0 (IBM, Chicago, IL).

There were significant linear associations between T-scores and skull HU values at four different sites on the frontal cancellous bone among the patients in the SHUB registry (Fig. 1A, B). The data in Fig. 1C show an increase of approximately 105.6 mean skull HU value per 1 T-score increase (B = 105.64; p < 0.001). The optimal cut-off value for mean skull HU to predict osteopenia was 712.419 (area under the curve [AUC] = 0.798; p < 0.001), and the HU value to predict osteoporosis was 611.704 (AUC = 0.761; p < 0.001) in the patients from the SHUB registry (Fig. 1D). According to those cut-off values, the study patients were categorized as hypothetical normal (above the cut-off HU value for osteopenia [> 712.419]), hypothetical osteopenia (between the cut-off HU values for osteopenia and osteoporosis [> 611.704 and ≤ 712.419]), and hypothetical osteoporosis (below the cut-off HU value for osteoporosis [≤ 611.704]).

A total of 99 patients with 106 intracranial meningiomas were included for the study. All patients received LINAC-based radiation treatments in our hospital over an approximate 6-year period. A total of 15 patients (14.2%) had PTBE after radiation treatment (Table 1). The mean patient age was 63.3 years and 77.4% of patients were female. A total of 42 patients (39.6%) were categorized as hypothetical osteoporosis. The mean GTV of meningioma and BED were 8.2 cc and 89.5 Gy (Gy), respectively. Patient characteristics are presented in Table 1.

Detailed information about the skull HU values according to PTBE in the study cohort and the skull HU values with additional BMD information for the SHUB registry patients is presented in Table 2. The SHUB registry showed relatively a high proportion of women and patients with an older age distribution than the study patients. The overall average mean frontal skull HU value was 716.2 in the study patients and 653.0 among the SHUB registry patients. In the study cohort, there were significant differences in skull HU values between the PTBE (−) and PTBE (+) groups. The median time between brain CT and BMD measurement was 151 days, and 36.6% of patients in the SHUB registry were diagnosed with osteoporosis.

We observed a significant negative correlation between age and mean skull HU value in the PTBE (−) group (p < 0.001). Although not statistically significant, a similar tendency of a negative association was also observed in the PTBE (+) group (p = 0.103) (Fig. 2A). We found that the PTBE (+) group showed relatively lower mean skull HU values across the age ranges compared with the PTBE (−) group. The PTBE (+) group showed marginally significant lower mean skull HU values compared with the PTBE (−) group among the older age group (p = 0.050) (Fig. 2B).

The overall cumulative hazard for PTBE development in patients with intracranial meningioma after radiation treatment is shown in Fig. 2C. When the patients were categorized as the hypothetical BMD classification, the hypothetical osteoporosis group showed a significantly higher rate of PTBE occurrence (p = 0.013) (Fig. 2D). Multivariate Cox regression analysis determined that hypothetical osteoporosis was an independent predictor for the development of PTBE in patients with meningioma after LINAC-based radiation treatment (HR 5.20; 95% CI 1.11–24.46; p = 0.037) (Table 3). The additional independent predictive factors of PTBE were older age and larger GTV.

We also performed additional propensity score-matched analyses based on patient sex, patient age, tumor location, GTV, and BED (Table 4). We found that hypothetical osteoporosis remained an independent predictive factor for the development of PTBE after LINAC-based radiation treatment in the propensity score-matched patients (HR 5.36; 95% CI 1.06–27.12; p = 0.042) (Table 5).