Ki-67 labeling index predicts tumor progression patterns and survival in patients with atypical meningiomas following stereotactic radiosurgery

The clinical data of 451 consecutive patients with meningiomas treated with single-session SRS between June 1992 and February 2022 at our institution were collected from an institutional Gamma Knife database. Among these patients, 47 patients with atypical meningiomas were identified, and patients with a follow-up period of < 3 months (n = 8) were excluded. In total, 39 atypical meningiomas were analyzed for this study. Patients who underwent external-beam radiotherapy before SRS were included. Tumor tissue from all surgically obtained specimens were reviewed, and all diagnoses were confirmed based on the 2016 WHO Classification of Tumors of the Central Nervous System criteria. Tissue from seven tumors that were resected and diagnosed as atypical meningioma before 2007 were re-evaluated and reconfirmed to meet the 2016 criteria [2, 34]. Formalin-fixed, paraffin-embedded specimens were subjected to immunohistochemistry for Ki-67, and the LI within the hotspot was calculated under a light microscope by board-certified pathologists. All participants in the present study provided written informed consent, and all components of this study were authorized by the appropriate Institutional Review Board (The Research Ethics Committee of our institution, #2231).

SRS was performed using a Leksell GammaKnife (Elekta AB, Stockholm, Sweden). After head fixation using a Leksell frame (Elekta Instruments, Stockholm, Sweden), stereotactic imaging (computed tomography [CT] before July 1996, magnetic resonance imaging [MRI] between August 1996 and January 2018, and cone-beam CT thereafter) was performed to obtain the precise three-dimensional coordinates. Thin-slice MR images (gadolinium-contrasted T1-weighted images and fast imaging employing steady-state acquisition) obtained the day before SRS were co-registered to define the tumor, surrounding cranial nerves, and vasculature. Radiosurgical treatment was planned and approved by the dedicated neurosurgeons and radiation oncologists involved with the procedure. All treatment planning was performed using commercially available software programs (KULA treatment planning system until 1998 and Leksell GammaPlan® [Elekta Instruments]).

Clinical evaluations and MRI were performed every 3–6 months after the date of SRS. Data on tumor status and SRS-related complications were prospectively collected for the institutional Gamma Knife database. Two neurosurgeons and two neuroradiologists independently assessed radiologic evidence of recurrence. Tumor progression was defined as one of three types of events: (1) “intrafield recurrence,” defined as a > 10% increase in volume inside the 50% isodose line on two or more consecutive imaging studies (Fig. 1A, B), (2) “marginal recurrence,” defined as new tumor progression between the lines of 50% and 20% isodose (Fig. 1C, D), and (3) “remote recurrence,” defined as new tumor progression outside the 20% isodose line (Fig. 1E, F) [25, 26, 32]. AREs were graded according to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. Peritumoral T2 signal change expanding beyond a 2-mm margin from the tumor after SRS was also considered to be an ARE.

Baseline patient, tumor, and SRS dosimetry characteristics were summarized. Continuous variables were presented as medians and interquartile ranges (IQRs), while categorical variables were presented as numbers and percentages. The Kaplan–Meier method was used to evaluate therapeutic effects, with outcomes comprising local tumor control rates (LCRs), progression-free rates (PFRs), and DSS rates. Local tumor control was defined as the absence of intrafield recurrence, and progression-free status was defined as the absence of any recurrence pattern including a lack of intrafield, margin, or remote recurrence. DSS was defined as the absence of any mortality associated with the treated tumors. To further analyze the association between Ki-67 LI and SRS outcomes, the patients were classified into three groups based on Ki-67 LI: low LI (< 5%), intermediate LI (5%–10%), and high LI (> 10%) groups, based on the previous studies on Ki-67 LI in meningiomas [33, 35]. The LCRs, PFRs, and DSS rates between groups were compared using the log-rank test, and multiple curves were compared using the log-rank test with the Bonferroni correction. In addition, receiver operating curve (ROC) analyses on Ki-67 LI were performed to calculate each area under the curve (AUC), sensitivity, and specificity on intrafield recurrence, any progression, and disease-specific mortality, and to indicate a cut-off Ki-67 LI increasing each risks using Youden index. Continuous variables (age, maximum diameter, target volume, and radiosurgical dose) were entered into the models after dichotomization using the median values. Factors associated with local tumor recurrence and AREs were examined using bivariate and multivariable Cox proportional hazards analyses. Statistical significance was set at p < 0.05. All statistical analyses were performed using the JMP® Pro 17.0.0 software (SAS Institute Inc., Cary, NC, USA).

Baseline patient characteristics and dosimetry data are summarized in Table 1. The median (IQR) age and follow-up period were 73 (63–77) years and 26 (18–59) months, respectively. The minimum follow-up period among alive patients at the last visit was 9 months. The median (IQR) Ki-67 LI was 9.6% (5.0–10.0%), with 5, 25, and 9 tumors in the low (< 5%), intermediate (5%–10%), and high (> 10%) LI groups, respectively. Postoperative adjuvant SRS was performed for seven (18%) patients after tumor resection before any noted progression. External-beam radiotherapy was performed for 13 (33%) patients before SRS. The median (IQR) central and prescription doses were 36 Gy (32–40 Gy) and 18 Gy (16–18 Gy), respectively.

In the entire cohort, 14 (36%) patients experienced intrafield recurrence, and 12 (31%) experienced marginal recurrence (Supplementary Table 1). One patient experienced intrafield and marginal recurrence simultaneously. Remote recurrence was observed in three (8%) patients, and all lesions were detected simultaneously with marginal recurrence. In the entire cohort, the cumulative 1-, 3-, and 5-year LCRs were 84%, 68%, and 53%, respectively (Fig. 2A); and the PFRs were 79%, 41%, and 23%, respectively (Fig. 2C).

Using the three groups stratified by Ki-67 LI, the 1-, 3-, and 5-year LCRs with low LI were all 100%, and LCRs with intermediate LI were 88%, 74%, and 53%, respectively; the 1- and 3-year LCRs with high LI were 63% and 25%, respectively (Fig. 2B; p = 0.011). Differences between low and high LI (p = 0.033) and between intermediate and high LI (p = 0.021) were marginally significant. The ROC analysis for intrafield recurrence and Ki-67 LI showed an AUC of 0.797 (Supplementary Fig. 1). The Youden index identified a cutoff at Ki-67 LI = 8%, with a sensitivity of 1.000 and specificity of 0.600. Bivariate and multivariable regression analyses were performed to determine if Ki-67 LI was associated with recurrence. In addition to Ki-67 LI, sex and tumor volume were used for multivariable analysis. Tumors with high Ki-67 LI were associated with greater intrafield recurrence risk using bivariate (hazard ratio [HR] 3.75, 95% confidence interval [CI] 1.13–12.49, p = 0.031; Table 2) and multivariable (HR 6.10, 95% CI 1.56–23.95, p = 0.010; Table 2) analyses. Larger tumor volume was associated with greater intrafield recurrence risk only in multivariable analysis (HR 1.08, 95% CI 1.01–1.16, p = 0.023; Table 2). Furthermore, 1-, 3-, and 5-year PFRs with low, and intermediate LI were 100%, 100%, and 50%; and 84%, 46%, and 26%, respectively; the 1- and 3-year PFRs with high LI were 56% and 0%, respectively (Fig. 2D p = 0.003). Differences between low and high LI (p = 0.009) and between intermediate and high LI (p = 0.004) were significant as well. The ROC analysis for any progression and Ki-67 LI showed an AUC of 0.793 (Supplementary Fig. 1). The Youden index indicated a cutoff at Ki-67 LI = 7%, with sensitivity of 0.880 and specificity of 0.714. Tumors with high Ki-67 LI were identified as risk factors of lower PFRs using bivariate (high vs. low, HR 11.97, 95% CI 1.46–98.04, p = 0.021: high vs. intermediate, HR 3.43, 95% CI 1.35–8.70, p = 0.010; Table 3) and multivariable (high vs. low, HR 12.71, 95% CI 1.51–107.23, p = 0.019: high vs. intermediate, HR 4.12, 95% CI 1.54–11.04, p = 0.005; Table 3) analyses in which tumor volume and indication of SRS in addition to Ki-67 LI were included. Tumor volume was not associated with lower PFRs.

Among the 39 patients, 29 (74%) were alive by last follow-up, 7 (18%) had died of uncontrolled tumor progression, and 3 (8%) had died from unrelated causes. The cumulative 1-, 3-, and 5-year DSS rates were 97%, 81%, and 81%, respectively (Fig. 2E). Stratified by Ki-67 LI, the 1-, 3-, 5-year DSS rates with low LI were all 100%, those with intermediate LI were 100%, 100%, and 89%, respectively; and those with high LI were 89%, 50%, and 50%, respectively (p = 0.019; Fig. 2F). The ROC analysis for disease-specific mortality and Ki-67 LI showed an AUC of 0.766 (Supplementary Fig. 1). The Youden index identified a cutoff at Ki-67 LI = 8%, with sensitivity of 1.000 and specificity of 0.469. A significant difference in DSS rates between intermediate and high LI groups was identified (p = 0.015), and high LI compared with intermediate LI was identified as a risk factor related to shorter DSS using bivariate (HR 6.27, 95% CI 1.14–34.26, p = 0.034; Supplementary Table 2) and multivariable (HR 6.55, 95% CI 1.19–35.95, p = 0.035; Supplementary Table 2) analyses in which tumor volume and Ki-67 LI were included. Larger tumor volume was also associated with shorter DSS in both bivariate (HR 1.09, 95% CI 1.00–1.19, p = 0.032; Supplementary Table 2) and multivariable (HR 1.09, 95% CI 1.00–1.20, p = 0.043; Supplementary Table 2) analyses.

AREs were observed in three (8%) patients, and all were peritumoral T2 signal changes on MRI 3–9 months after SRS. Two patients were asymptomatic (CTCAE grade 1); although one patient experienced increased convulsions likely related to expanding T2 signal changes and required escalation of antiepileptic therapy, his symptoms and AREs were transient and well-controlled after medical therapy alone (CTCAE grade 2).