Introduction: Cavernomas are vascular lesions with a genetic heritage that can be spotted on the central nervous system. Whenever these lesions are localized in eloquent regions, surgical resection is not recommended. In this type of situation, Gamma Knife stereotactic radiosurgery (GKSRS) could be a feasible option for treating patients. Thus, we aimed to explore the outcomes associated with this procedure. Methods: We performed a systematic review and meta-analysis of reconstructed time-to-event data based on Kaplan-Meier curves. A thorough search was conducted on PubMed, Cochrane, Web of Science, and Embase databases targeting papers that provided information regarding hemorrhagic outcomes associated with GKSRS through Kaplan-Meier curves. Results: After a systematic search in the specific databases, seven studies were included in this review. Notably, a total of 1,071 patients had 1,104 cavernomas treated by GKSRS. Assessment of short-term and long-term post-procedure outcomes was performed, with the estimated overall events-free rate at 2 years being 89.8% (95% CI: 87.7–91.5), while, at 10 years, the estimated overall events-free rate was 71.3% (95% CI: 67.2–75.1). Conclusion: GKSRS seems to be a good alternative for the control of symptomatic events in early and long-term follow-up, despite the need for further investigation provided by future studies.

Highlights

  • Gamma Knife seems to be a safe and effective treatment option for cerebral cavernous malformations, when applicable.

  • Our analysis shows that 86.6% of the patients (95% CI: 84.2–88.6) are expected to be symptom free 5 years after radiosurgery with Gamma Knife.

  • Of 946 patients, 5 had related-to-treatment mortality. The identified proportion was 0% (95% CI: 0–1%; I2 = 0%). This result elucidates the safety of the procedure.

Cerebral cavernous malformations (CCMs) are vascular lesions found in the central nervous system with a genetic origin [1, 2]. The estimated prevalence of these vascular malformations can be up to 0.4% of the general population [3]. In general, CCM patients do not show any associated symptoms, staying asymptomatic throughout their lives [3]. Nevertheless, the emergence of symptoms, such as epileptic seizures, headaches, and impaired consciousness, suggests a potential occurrence of a hemorrhagic event, which can be detected through imaging and clinical assessment [3, 4]. Cavernous angioma symptomatic hemorrhage can only be diagnosed by consistent imaging findings that suggest bleeding, associated with related neurological deficits [1]. While the mechanisms of bleeding remain unclear, it is established that hemorrhage is the most significant event, as it serves as a major source of impairment [1]. Despite its severity, the likelihood of a hemorrhage occurring is a mere 0.08% [2]. Conversely, the occurrence of bleeding itself constitutes a risk factor that substantially augments the probability of new hemorrhagic events and associated complications [1, 2].

CCMs can occur in a sporadic form, with a single lesion, and in a familial form, in which multiple lesions can be found [5]. The sporadic form is known to be associated with a developmental venous anomaly, while the familial form is not [2]. The familial form is an autosomal dominant genetic mutation that affects one of three genes: CCM1, CCM2, or CCM3 [2, 6]. A family screening is recommended when finding multiple lesions in the brain, in order to identify possible family members that are also affected by this genetic mutation. Magnetic resonance imaging is, at the present date, the best and most used imaging method to evaluate the typical CCMs’ popcorn-like lesions in the brain, with a hemosiderin halo and blood in different stages of metabolization, along with signs of calcifications [2, 5]. The SWI sequence was tested and established as the best magnetic resonance imaging sequence to observe CCMs that could not be found in other sequences [5].

Despite being largely executed for a long time, surgery is still a conflicting procedure, presenting no solid scientific evidence for or against it [7]. When evaluating the surgical strategy, the lesion’s location plays a crucial role, especially in instances where the CCM is situated in eloquent areas of the brain [2, 7]. Radiosurgery could be considered as an option in these cases where lesions are situated in eloquent or deep-seated areas. However, there is ongoing debate regarding whether the observed reduction in hemorrhage risk 2 years post-stereotactic radiosurgery (SRS) is truly attributable to the procedure or is merely a reflection of the natural history of CCMs [2, 7, 8]. These debates center around the findings of Barker et al. [8], who have demonstrated a trend of untreated CCM bleeding clustering within a 2-year period after the initial hemorrhage, with a subsequent decline beyond this 2-year period [7, 8].

Gamma Knife stereotactic radiosurgery (GKSRS) is a noninvasive procedure that uses precise, high-dose gamma radiation to target and treat specific areas in the brain. It is a highly targeted ablative form of radiotherapy that can be an alternative to manage specific CCM lesions while minimizing exposure to surrounding healthy tissue [4, 9]. Regardless of the benefits and excellent outcomes reported by selected institutions or registry-based studies, its indication remains a source of debate. Thus, we conducted a systematic review and meta-analysis, using reconstructed time-to-event data, to contribute to the current scientific evidence concerning early and long-term outcomes of Gamma Knife radiosurgery for CCMs.

Eligibility Criteria

This systematic review comprised all studies that reported the use of GKSRS for treating CCMs and presented part of their data using Kaplan-Meier curves. In order to enhance the reliability and minimize the risk of bias in our survey, we excluded case reports, reviews, non-English articles, studies in which the assessed outcomes were not found, letters, and comments. In addition, due to the absence of information regarding Kaplan-Meier curves among the studies in which other types of SRS were used, we have included reports exclusively on GKSRS.

Search Strategy and Data Extraction

In September 2023, a comprehensive systematic search was conducted across multiple specific databases including PubMed, Cochrane, Web of Science, and Embase. The search strategy employed the following keywords: (“cerebral cavernous malformations” OR “cerebral cavernous angioma” OR “cavernous malformations” OR “cavernous malformation” OR “CCM” OR “cavernous angioma” OR “cavernoma” OR “cavernomas” OR “cavernous hemangioma” OR “cavernous haemangioma”) AND (“gamma knife” OR “radiotherapy” OR “radiosurgery”) AND (“stereotactic”).

The search scope was limited to studies published until September 2023. The data extraction process was carried out collaboratively by two authors, BLG and MDB, who adhered to predefined search criteria and conducted quality assessments. In instances of potential conflicts, a third author, GFG, assumed the responsibility of resolution. Ultimately, only studies presenting pertinent data in the form of Kaplan-Meier curves were considered for inclusion in this research project.

Outcomes

The outcomes considered for this study were cumulative risk of bleeding, cumulative risk for the patients to be symptom free, adverse radiation effects (AREs), mortality, and surgical resection performed after GKSRS. The cumulative risk of bleeding was considered the time necessary for a patient to experience a hemorrhagic event due to the CCMs. The cumulative risk for the patients to be symptom free was considered the time necessary for a patient to experience at least one symptom related to an eventual CCM bleeding (e.g., epilepsy and neurological deficits). AREs were defined as either brain-tissue lesions (symptomatic or asymptomatic new cyst formation or T2 hyperintensity by Dumot et al. [10], Dumot et al. [11] and Jacobs et al. [12]; perilesional edema developing within 12 months after treatment – transient – or more than 12 months – permanent by Nagy et al. [13]) or hair loss and anhidrosis by Wu et al. [14].

Statistical Analysis

In this meta-analysis, we used the two-stage approach as described by Liu et al. [15] based on the R package “IPDfromKM” (version 0.1.10). This is a meta-analysis strategy widely used in different evidence-based healthcare domains [16‒19].

In the first stage, raw data coordinates (time and survival probability) were extracted from each Kaplan-Meier curve of single studies previously selected. In the second stage, the data coordinates were processed based on the raw data coordinates from the first stage in conjunction with the numbers at risk at given time points, and IPD were reconstructed. Since some studies lack a lot of risk information, we estimated it based on the height of the steps in Kaplan-Meier curves. Finally, the reconstructed IPD from all studies were merged to create the study data set using GraphPad Prism 10. The primary endpoint was overall symptoms free at follow-up. The secondary endpoint was 2 years symptom free. Subgroup endpoints were (1) intracranial bleeding and (2) seizure.

Additionally, we analyzed the data concerning outcomes experienced by the patients following the Cochrane Collaboration and the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [20]. Conducting a single-arm meta-analysis, we considered the proportion of outcomes within the studied population, expressed along with 95% confidence intervals, to examine treatment effects.

To evaluate variations between studies, we employed the Cochran Q test and I2 statistics, focusing on heterogeneity. In cases of noticeable heterogeneity, indicating substantial differences between studies, we opted for a random-effects model in our analysis. This model recognizes that the true treatment effect can vary across studies due to inherent differences. Conversely, when differences were minimal, suggesting a more uniform pattern across studies, we turned to a common-effects model, assuming a shared true treatment effect for all studies. We deemed a p value <0.05 and an I2 value greater than 35% as significant indicators of notable heterogeneity in treatment effects. Statistical analysis was performed using the software R (version 4.3.1, R Foundation for Statistical Computing, Vienna, Austria).

Quality Assessment

The Risk of Bias in Non-Randomized Studies of Interventions (ROBINS-I) tool was systematically used to assess included studies for risk of bias [21]. The studies and their characteristics were classified into low, moderate, serious, and critical risk of bias. Two independent reviewers assessed the risk for bias. When there was a disagreement, a third reviewer checked the data and made the final decision.

Study Selection

A total of 935 references were assembled in our search, comprising multiple databases: 203 from PubMed, 9 from Cochrane, 329 from Web of Science, and 394 from Embase. Of these, 353 duplicates were removed and 582 references were screened. In an initial evaluation, 523 articles were excluded after the title or abstract screen. Then, 59 studies were sought for retrieval, with 6 not being retrieved. Among the 53 articles assessed, 47 were excluded after full-text screen and data extraction. One citation was manually added to our study. At last, seven studies were included in our analysis. A summary of this process is described in Figure 1 [22].

Fig. 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) flow diagram that comprehensively elicits the process of study selection. A total of 582 non-duplicate references were screened, while only seven met the inclusion criteria.

Fig. 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) flow diagram that comprehensively elicits the process of study selection. A total of 582 non-duplicate references were screened, while only seven met the inclusion criteria.

Close modal

Risk of Bias Assessment

The totality of studies included in this survey was assessed using the ROBINS-I tool for bias scrutiny [21]. Among them, 5 were classified as having a moderate risk of bias, while 2 were found to have a serious risk of bias. Online supplementary Figures 1 and 2 (for all online suppl. material, see https://doi.org/10.1159/000539079) show the qualitative assessment of the studies with the ROBINS-I tool, considering each one of the six domains evaluated [10‒14, 23, 24]. There are several concerns regarding confounding factors and selection bias in the studies due to moderate differences between groups regarding sex and age. Furthermore, bias in the measurement of outcomes was a reason for concern. There is also a selection bias due to the fact that patients who undergo SRS are the ones with CCMs located in highly complex areas that could depict small signs of symptomatic presentation.

Patients’ Characteristics

After excluding duplicates and non-eligible studies, 7 articles met our eligibility criteria (Table 1) [10‒14, 23, 24]. All the studies were non-randomized and observational studies, while only 2 studies were multicentric cohorts. A total of 1,071 patients were included in the original study. One cohort (Jacobs et al. [12]) reported only brainstem lesions (n = 76), and two studies (Dumot et al. [10]; Nagy et al. [11]) reported outcomes for familial CCM form [10, 12, 13]. Four studies reported AREs with a pooled mean percentage of 11.97%, being 6.1% with transient radiation effect and 4.2% with permanent radiation morbidity. These characteristics are better described in Table 1 [10‒14, 23, 24].

Table 1.

Patients’ baseline characteristics

StudyDumot [10] (2023)Dumot [11] (2023)Jacobs [12] (2018)Karaaslan [23] (2021)Nagy [13] (2018)Régis [24] (2000)Wu [14] (2022)
Study type 
Brainstem only No No Yes No No No No 
Patients, n 109 381 76 195 210 49 51 
Sex, males, n (%) 63 (57.8) 170 (44.6) 40 (52.6) 76 (39.0) 96 (45.7) 26 (53.1) 25 (49.0) 
Age, years (range) 28.9 median 37.5 median 41.60 (5–79) 41.94 (16–70) 37.0 median (0.5–77) 36.0 44.1 (11–84) 
Follow-up time, years 3.5 median 5.7 5.5 median 7.92 
Familial form included No Yes No No Yes No No 
Number of CCMs treated 109 414 76 195 210 49 51 
50% marginal isodose 14.00 Gy 12.00 Gy 15.00 Gy 13.70 Gy mean 12–13 Gy (depending on the location) 19.17 Gy mean 10–12 Gy 
Mean lesion volume, cm3 (range) 1.3 0.6 0.66 (0.05–6.8) 0.68 (0.01–4.37) 1.70 (0.04–9.10) 
Prior surgery, n 19 14 21 
Pre-GKSRS annual hemorrhage rate, % 43.35 31.3 44.00 2.60 14.52 
Pre-GKSRS experienced hemorrhages, n 
 0 47 77 
 1 55 324 64 
 At least 2 90 69 54 
Clinical symptoms pre-GKSRS, n 
 Headache 11 72 95 27 
 Neurological deficits 10 148 18 13 
 True seizures without bleeding signs 68 
StudyDumot [10] (2023)Dumot [11] (2023)Jacobs [12] (2018)Karaaslan [23] (2021)Nagy [13] (2018)Régis [24] (2000)Wu [14] (2022)
Study type 
Brainstem only No No Yes No No No No 
Patients, n 109 381 76 195 210 49 51 
Sex, males, n (%) 63 (57.8) 170 (44.6) 40 (52.6) 76 (39.0) 96 (45.7) 26 (53.1) 25 (49.0) 
Age, years (range) 28.9 median 37.5 median 41.60 (5–79) 41.94 (16–70) 37.0 median (0.5–77) 36.0 44.1 (11–84) 
Follow-up time, years 3.5 median 5.7 5.5 median 7.92 
Familial form included No Yes No No Yes No No 
Number of CCMs treated 109 414 76 195 210 49 51 
50% marginal isodose 14.00 Gy 12.00 Gy 15.00 Gy 13.70 Gy mean 12–13 Gy (depending on the location) 19.17 Gy mean 10–12 Gy 
Mean lesion volume, cm3 (range) 1.3 0.6 0.66 (0.05–6.8) 0.68 (0.01–4.37) 1.70 (0.04–9.10) 
Prior surgery, n 19 14 21 
Pre-GKSRS annual hemorrhage rate, % 43.35 31.3 44.00 2.60 14.52 
Pre-GKSRS experienced hemorrhages, n 
 0 47 77 
 1 55 324 64 
 At least 2 90 69 54 
Clinical symptoms pre-GKSRS, n 
 Headache 11 72 95 27 
 Neurological deficits 10 148 18 13 
 True seizures without bleeding signs 68 

R, retrospective; n, number; CCMs, cerebral cavernous malformations; cm3, cubic centimeters; GKSRS, Gamma Knife stereotactic radiosurgery.

Outcomes from GKSRS

The clinical and radiosurgical aspects of the patients were not expressed in an equal manner in all of the studies. In our survey, only a subset of the included studies provided data due to a lack of information. The mean or median treatment dose was reported in all studies, except for Wu et al. [14]. The post-GKSRS annual hemorrhage rate >2 years and post-GKSRS annual hemorrhage rate <2 years percentages were available in all studies except for Dumot et al. [11] and Régis et al. [24]. A total of 108 (10.1%) patients presented with AREs. Data concerning transient and permanent AREs were available in 2 studies. A total of 15 (1.4%) patients had lesions removed after SRS. Although the total mortality of 15 patients encompassed our scrutiny, only 5 of them were procedure related. Of these, 2 were due to hemorrhage and 3 were from unknown causes.

Analysis of Being Symptom Free in the Overall Population

Seven studies reported complete Kaplan-Meier analysis for symptom-free post-Gamma Knife treatment, while only 1,062 patients could be analyzed due to missing data from the remaining 9 individuals. Figure 2a b depicts the full pooled Kaplan-Meier curve for cumulative risk and the Kaplan-Meier curve for the specific time frame of 2 years. The estimated probabilities of being symptom free at 6 months, one, two, five, ten, and 15 years were 95.9% (95% CI: 94.5–99.9), 91.7% (95% CI: 89.8–93.2), 89.8% (95% CI: 87.7–91.5), 86.6% (95% CI: 84.2–88.6), 71.3% (95% CI: 67.2–75.1), and 58.9% (95% CI: 53.7–63.7).

Fig. 2.

Reconstructed Kaplan-Meier curve for being symptom free. a Full pooled Kaplan-Meier curve for cumulative risk of symptoms. b Kaplan-Meier curve for risk of symptoms in a time frame of 2 years.

Fig. 2.

Reconstructed Kaplan-Meier curve for being symptom free. a Full pooled Kaplan-Meier curve for cumulative risk of symptoms. b Kaplan-Meier curve for risk of symptoms in a time frame of 2 years.

Close modal

Analysis of Being Specific Symptoms Free in Secondary Endpoints

Four studies (n = 941) reported complete Kaplan-Meier analysis for bleeding-free post-Gamma Knife treatment. Figure 3a, b represents the pooled Kaplan-Meier curve for this specific cumulative risk and the Kaplan-Meier curve for the specific time frame of 2 years. The estimated probabilities of being symptom free at 6 months, 1, 2, 5, and 10 years were 98.2% (95% CI: 97.1–98.9), 94.7% (95% CI: 93.0–96.0), 93.0% (95% CI: 91.2–94.5), 89.9% (95% CI: 87.6–91.8), and 73.9% (95% CI: 0.696–0.77).

Fig. 3.

Reconstructed Kaplan-Meier curve for bleeding-free and seizure-free post-Gamma Knife treatment. a Pooled Kaplan-Meier curve for the cumulative risk of bleeding. b Kaplan-Meier curve for bleeding risk within a time frame of 2 years. c Pooled Kaplan-Meier curve for cumulative risk of presenting seizures. d Kaplan-Meier curve for seizure risk in a time span of 2 years.

Fig. 3.

Reconstructed Kaplan-Meier curve for bleeding-free and seizure-free post-Gamma Knife treatment. a Pooled Kaplan-Meier curve for the cumulative risk of bleeding. b Kaplan-Meier curve for bleeding risk within a time frame of 2 years. c Pooled Kaplan-Meier curve for cumulative risk of presenting seizures. d Kaplan-Meier curve for seizure risk in a time span of 2 years.

Close modal

In the same way, two studies (n = 121) reported complete Kaplan-Meier analysis for seizure-free post-Gamma Knife treatment in pure epileptic individuals. Figure 3c, d) outlines the pooled Kaplan-Meier curve. The estimated probabilities of being seizure free at 6 months, 1, 2, 5, and 10 years were 77.8% (95% CI: 69.1–83.3), 68.0% (95% CI: 58.6–75.7), 64.1% (95% CI: 54.5–72.2), 60.5% (95% CI: 50.6–69.0), and 55.0% (95% CI: 40.9–67.1).

Comparing outcomes after GKSRS, patients with intracranial bleeding had better overall outcome control than seizure individuals. In the specific 2-year follow-up analysis, again patients with intracranial bleeding had better outcome control than seizure individuals. The reconstructed Kaplan-Meier curve can be seen in online supplemental Figure 3.

Radiation Isodose Sensitivity Analysis

Because Régis et al. [24] used a mean higher isodose (19.17 Gy) normally prescribed to treat CCMs with SRS, we decided to conduct a sensitivity analysis excluding this study. This left the data of 1,013 patients from the remaining studies available for analysis. As depicted in Figure 4a, there was no statistically significant difference between cohorts (p = 0.270). The estimated probabilities of being symptom free at 6 months, 1, 2, 5, and 10 years were 98.0% (95% CI: 97.0–98.7), 93.9% (95% CI: 92.2–95.3), 91.9% (95% CI: 90.0–93.5), 88.6% (95% CI: 86.2–90.5), and 73.0% (95% CI: 68.7–76.7).

Fig. 4.

Reconstructed Kaplan-Meier curve of special cases. a Curve excluding Régis et al. [24] due to usage of a mean higher isodose (19.17 Gy) than what is normally prescribed to treat CCMs. b Curve excluding two studies due to the usage of SRS to treat familial CCM forms [14, 17].

Fig. 4.

Reconstructed Kaplan-Meier curve of special cases. a Curve excluding Régis et al. [24] due to usage of a mean higher isodose (19.17 Gy) than what is normally prescribed to treat CCMs. b Curve excluding two studies due to the usage of SRS to treat familial CCM forms [14, 17].

Close modal

Familial Form Sensitivity Analysis

Because the last guideline published by the Angioma Alliance formally stated against the usage of SRS to treat familial CCM forms, we decided to conduct a sensitivity analysis excluding the two studies that included familial types of the disease [7, 10, 13]. This left the data of 443 patients from the remaining studies available for analysis. As depicted in Figure 4b, there was no statistically significant difference between cohorts (p = 0.270). The estimated probabilities of being symptom free at 6 months, 1, 2, 5, and 10 years were 91.5% (95% CI: 88.5–93.8), 87.1% (95% CI: 83.6–89.9), 85.3% (95% CI: 81.6–88.3), 82.2% (95% CI: 78.2–85.5), and 63.1% (95% CI: 44.0–56.1).

Adverse Radiation Effects

From the 827 patients encompassed in 5 studies, 108 reported AREs. After random-effects model analysis due to high heterogeneity, the identified proportion of AREs was calculated to be 12% (95% CI: 6–19%; I2 = 87%). The most heterogeneous study was Wu et al. [14], in which only 1 out of 51 patients presented AREs. The forest plot is expressed in Figure 5a.

Fig. 5.

Forest plots of outcomes of post-GKSRS. a This image provides random and common effects-weighted statistical information concerning the proportion of AREs related to the GKSRS intervention in a 95% confidence interval. b This image provides random and common effects-weighted statistical information concerning the proportion of mortality related to the GKSRS intervention in a 95% confidence interval. c This image provides random and common effects-weighted statistical information regarding the proportion of resected cavernomas post-GKSRS performance in a 95% confidence interval.

Fig. 5.

Forest plots of outcomes of post-GKSRS. a This image provides random and common effects-weighted statistical information concerning the proportion of AREs related to the GKSRS intervention in a 95% confidence interval. b This image provides random and common effects-weighted statistical information concerning the proportion of mortality related to the GKSRS intervention in a 95% confidence interval. c This image provides random and common effects-weighted statistical information regarding the proportion of resected cavernomas post-GKSRS performance in a 95% confidence interval.

Close modal

Related Mortality

Of the 946 patients encompassed in 5 studies, 5 had related-to-treatment mortality. After common and random-effects model analysis, the identified proportion was 0% (95% CI: 0–1%; I2 = 0%). The forest plot is expressed in Figure 5b.

Lesions Surgically Removed after GKSRS

From the 751 patients encompassed in 4 studies, 15 CCMs were surgically removed after GKSRS. Due to high heterogeneity, in the random-effects model analysis, the identified proportion was calculated to be 2% (95% CI: 0–3%; I2 = 51%). The forest plot is expressed in Figure 5c.

Currently, the evidence-based management of CCMs remains limited. Despite many publications, there remains controversy regarding optimal management strategies. High-level evidence in the form of a randomized controlled trial does not exist, which leads to insufficient equipoise on treatment among specialists. The population-based study revealed poor outcomes over 5 years following surgery for deeper CCMs located in the insula, basal ganglia, thalamus, or brainstem. For very experienced teams, the postoperative morbidity rate associated with these CCMs is 5–18%, with the rate of mortality approaching 2% [25, 26]. SRS has been suggested as an alternative treatment for symptomatic CCM located in eloquent areas, with most series identifying a decline in the hemorrhage rate more than 2 years after SRS treatment. The mechanism of action involves inducing vascular injury, thrombosis, and fibrosis within the malformation, thereby promoting the stabilization or reduction in the size of the lesion over time [27, 28].

However, causality is hard to be inferred as most natural history studies demonstrate decreasing rebleeding rates after 2 years post-bleeding. By now, radiosurgery may be considered for solitary CCM lesions with previous symptomatic hemorrhage if the lesion lies in eloquent areas that carry an unacceptable high surgical risk (class IIb, level B) [7]. In this scenario, the challenges arise from relatively small and heterogeneous study populations, coupled with significant variability in outcome reporting, which hinders the meaningful conduct of meta-analysis. The present study aimed to overcome these issues by collecting raw pooled patient data from published survival curves [29]. In this respect, it has been successful in pooling thousands of patients’ overall symptoms-free (n = 1,062), bleeding-free (n = 941), and seizure-free data (n = 121), delivering robust estimates for these outcomes up to the furthest point of data maturity.

To the best of our knowledge, this is the first pooled meta-analysis of reconstructed time-to-event data of early and long-term outcomes of GKSRS for CCMs. The estimated overall event-free rate at 2 years was 89.8% (95% CI: 87.7–91.5), while at 10 years, the estimated overall event-free rate was 71.3% (95% CI: 67.2–75.1). It is known that in population studies and in many case series without clear selection criteria, the risk of the first bleeding is extremely low (0.08% per patient-year). However, once a symptomatic bleeding has occurred, the annual risk of a subsequent episode rises dramatically (10-fold by some estimates). This elevated risk is greatest soon after a hemorrhage but persists thereafter with a 5-year risk estimated at 42% (95% CI: 27–58%) [30]. In our study, the 5-year overall event-free rate after GKSRS was 71.3% (95% CI: 67.2–75.1). Of note, we depicted an overall 15-year event-free rate after GKSRS of 55.0% (95% CI: 40.9–67.1). Despite some limitations, this meta-analysis suggests that GKSRS is a useful and durable option for the treatment of patients with complex CCM when compared with the natural history of symptomatic CCM, with a somewhat small chance of radiation morbidity (11.9%).

In matters of the secondary endpoints, the estimated probabilities of being bleeding free at two and 10 years were 93.0% (95% CI: 91.2–94.5) and 73.9% (95% CI: 0.696–0.77), while the estimated probabilities of being seizure free at 2 and 10 years were 64.1% (95% CI: 54.5–72.2) and 55.0% (95% CI: 40.9–67.1). Although it is hard to compare both outcomes due to differences in the cohorts, this analysis is of paramount importance, since it can tailor the indication of SRS depending on the symptoms of the patient. This tendency is in accordance with the literature published in previous meta-analyses by Wen et al. and Gao et al. [9, 31].

Turning to the complications and post-procedure aspects, we performed a pooled analysis of AREs, procedure-related mortality, and the number of CCMs surgically removed after the performance of GKSRS. Some previous meta-analyses were conducted regarding either SRS of brainstem cavernous angiomas or the use of GKSRS for treating cavernomas [4, 32, 33]. Despite their results, none thoroughly analyzed the same variables as we did, which makes our study unique. The proportion of AREs in our review is 12% (95% CI: 6–19%; I2 = 87%). Although Lu et al. [32], have discussed AREs, they did not provide any statistical information on this subject. Conversely, Kim et al. [33] showed that 7.3% of the patients presented with symptomatic AREs. Addressing the mortality issue, we noticed 15 deaths among the patients, of which only five were related to radiosurgery, including two from hemorrhagic events and three from unknown causes. This represented a proportion of 0% (95% CI: 0–1%; I2 = 0%). Ultimately, 15 CCMs were surgically removed after GKSRS, with a proportion of 2% (95% CI: 0–3%; I2 = 51%). This can lead to major confounding outcomes because we will not be able to associate correctly some of them with the performed procedure. Notably, it means that some positive or negative outcomes cannot be related to GKSRS or surgical resection once both of them were registered in those patients. The fact that 60 individuals had previous surgical treatment also corroborates this rationale, and further studies analyzing this profile of patients are necessary.

We were somewhat surprised to find two studies that included familial forms of the disease in their cohort of SRS, once the guideline stated that radiosurgery is not recommended for familial CCM disease with concern about de novo lesion genesis (class III, level C) [7, 10, 13]. However, one should highlight that this recommendation has a low level of evidence, and more studies should try to address this concern. Nonetheless, the performance of symptom control of SRS seems to be the same in this special type of disease.

It is noteworthy that alternative forms of SRS may also be considered based on specific clinical scenarios and institutional expertise. Other SRS modalities, such as CyberKnife or linear accelerator-based systems, could offer comparable precision in targeting CCM lesions [34‒36].

The choice among these modalities may be influenced by factors such as lesion characteristics, location, and patient-specific considerations. It is essential to note that while alternative forms of SRS could offer comparable precision in targeting CCM lesions, this manuscript specifically focused on GKSRS. The decision to narrow the discussion to this type of SRS was driven by the primary consideration of utilizing Kaplan-Meier curves as a data strategy, and unfortunately, sufficient KM information for other SRS modalities was not available. The selection of the SRS modality should be guided by a comprehensive evaluation of individual patient profiles, lesion attributes, and the expertise available at the treatment center, ensuring a tailored and effective therapeutic approach for CCMs. We therefore recognize the importance of other SRS modalities in treating CCMs, with probably the same outcomes.

Limitations

The limitations of this study are related to the features of the studies meta-analyzed, namely an absoluteness of retrospective study designs and a lack of standardized definitions for symptomatic events and intracranial bleedings. The literature on CCMs has been trying to systematize the definition of intracranial hemorrhage as clinical and radiological evidence of intracerebral bleeding [37]. It is certain that the lack of standardized criteria for patient inclusion and reporting outcomes significantly impacted the statistical measures calculated in this meta-analysis. Further, five studies were considered to have a moderate risk of bias and two were considered to have a serious risk by the ROBINS-I analysis [21]. This is a significant issue; the most common domain identified as serious potential sources of bias was “bias in measured outcomes,” while “bias due to confounding,” “bias due to selection of participants,” and “bias in classification of interventions” were all moderate risks of bias. With the aim of including as many studies as possible, a decision was made to include studies with small cohorts. This may have fallen victim to selection bias if patients were deemed eligible for more than one type of procedure (i.e., surgery vs. SRS). Also, some studies lost a significant proportion of patients to follow-up. It is believed that the meta-analysis method used to pool time-to-event probabilities will have corrected this issue.

In patients with selected CCM locations that are not deemed to be submitted to surgical resection, GKSRS appears to be a good alternative for short- and long-term control of symptomatic disease. Despite these findings, there is a demand for additional evidence, preferably derived from a pragmatically randomized controlled trial involving patients with deep-seated CCM lesions, in which surgical intervention would pose significant morbidity.

We thank Aliança Cavernoma Brazil and Casa Hunter for their structural support for the development of the research of our group.

A statement of ethics is not applicable because this study is based exclusively on published literature.

The authors declare that they have no conflict of interest.

This paper had no funding support.

Material preparation and analysis were performed by Gustavo da Fontoura Galvão and Gabriel Verly. Data collection was performed by Gabriel Verly, Matheus Duque Bessa, and Bernardo Lisboa Galvão Santos. Revision of the data was performed by Gustavo da Fontoura Galvão. The first draft of the manuscript was written by Gustavo da Fontoura Galvão with the support of Gabriel Verly. Review and editing for critical intellectual content were performed by Pablo Valença, Flávio Sampaio Domingues, and Marcello Reis da Silva. Supervision was performed by Marcello Reis da Silva and Jorge Marcondes. All authors commented on1 the previous versions of the manuscript. All authors read and approved the final manuscript.

The data used to perform this meta-analysis can be found within the body of the article and within its supplementary material. Further inquiries can be directed to the corresponding author. Declaration of generative AI and AI-assisted technologies were used in the writing process.

During the preparation of this work, the authors used ChatGPT in order to improve grammar and readability. After using this tool/service, the authors reviewed and edited the content as needed and took full responsibility for the content of the publication.

1.
Snellings
DA
,
Hong
CC
,
Ren
AA
,
Lopez-Ramirez
MA
,
Girard
R
,
Srinath
A
, et al
.
Cerebral cavernous malformation: from mechanism to therapy
.
Circ Res
.
2021
;
129
(
1
):
195
215
.
2.
Awad
IA
,
Polster
SP
.
Cavernous angiomas: deconstructing a neurosurgical disease
.
J Neurosurg
.
2019
;
131
(
1
):
1
13
.
3.
Al-Shahi Salman
R
,
Berg
MJ
,
Morrison
L
,
Awad
IA
;
Angioma Alliance Scientific Advisory Board
.
Hemorrhage from cavernous malformations of the brain: definition and reporting standards. Angioma Alliance Scientific Advisory Board
.
Stroke Vasc Interv Neurol
.
2008
;
39
(
12
):
3222
30
.
4.
Shanker
MD
,
Webber
R
,
Pinkham
MB
,
Huo
M
,
Olson
S
,
Hall
B
, et al
.
Gamma Knife® stereotactic radiosurgery for intracranial cavernous malformations
.
J Clin Neurosci
.
2022
;
106
:
96
102
.
5.
de Souza
JM
,
Domingues
RC
,
Cruz
LCH
Jr
,
Domingues
FS
,
Iasbeck
T
,
Gasparetto
EL
.
Susceptibility-weighted imaging for the evaluation of patients with familial cerebral cavernous malformations: a comparison with t2-weighted fast spin-echo and gradient-echo sequences
.
AJNR Am J Neuroradiol
.
2008
;
29
(
1
):
154
8
.
6.
da Fontoura Galvão
G
,
da Silva
EV
,
Trefilio
LM
,
Alves-Leon
SV
,
Fontes-Dantas
FL
,
de Souza
JM
.
Comprehensive CCM3 mutational analysis in two patients with syndromic cerebral cavernous malformation
.
Transl Stroke Res
.
2024
;
15
(
2
):
411
21
.
7.
Akers
A
,
Al-Shahi Salman
R
,
A Awad
I
,
Dahlem
K
,
Flemming
K
,
Hart
B
, et al
.
Synopsis of guidelines for the clinical management of cerebral cavernous malformations: consensus recommendations based on systematic literature review by the angioma alliance scientific advisory board clinical experts panel
.
Neurosurgery
.
2017
;
80
(
5
):
665
80
.
8.
Barker
FG
2nd
,
Amin-Hanjani
S
,
Butler
WE
,
Lyons
S
,
Ojemann
RG
,
Chapman
PH
, et al
.
Temporal clustering of hemorrhages from untreated cavernous malformations of the central nervous system
.
Neurosurgery
.
2001
;
49
(
1
):
15
25
; discussion 24–5.
9.
Wen
R
,
Shi
Y
,
Gao
Y
,
Xu
Y
,
Xiong
B
,
Li
D
, et al
.
The efficacy of gamma knife radiosurgery for cavernous malformations: a meta-analysis and review
.
World Neurosurg
.
2019
;
123
:
371
7
.
10.
Dumot
C
,
Mantziaris
G
,
Dayawansa
S
,
Xu
Z
,
Pikis
S
,
Peker
S
, et al
.
Stereotactic radiosurgery for haemorrhagic cerebral cavernous malformation: a multi-institutional, retrospective study
.
Stroke Vasc Neurol
.
2023
;
2023-002380
.
11.
Dumot
C
,
Mantziaris
G
,
Pikis
S
,
Dayawansa
S
,
Xu
Z
,
Samanci
Y
, et al
.
Epilepsy associated with cerebral cavernous malformations managed with stereotactic radiosurgery: an international, multicenter study
.
J Neurol
.
2023
;
270
(
10
):
5048
56
.
12.
Jacobs
R
,
Kano
H
,
Gross
BA
,
Niranjan
A
,
Monaco
EA
3rd
,
Lunsford
LD
.
Defining long-term clinical outcomes and risks of stereotactic radiosurgery for brainstem cavernous malformations
.
World Neurosurg
.
2019
;
124
:
e58
64
.
13.
Nagy
G
,
Burkitt
W
,
Stokes
SS
,
Bhattacharyya
D
,
Yianni
J
,
Rowe
JG
, et al
.
Contemporary radiosurgery of cerebral cavernous malformations: Part 1. Treatment outcome for critically located hemorrhagic lesions
.
J Neurosurg
.
2019
;
130
(
6
):
1817
25
.
14.
Wu
X
,
Chen
W
,
Lin
Y
,
Liang
R
.
The impact of volume factor on the long-term outcome of gamma knife radiosurgery for sporadic cerebral cavernous malformations
.
World Neurosurg
.
2022
;
158
:
e627
35
.
15.
Liu
N
,
Zhou
Y
,
Lee
JJ
.
IPDfromKM: reconstruct individual patient data from published Kaplan-Meier survival curves
.
BMC Med Res Methodol
.
2021
;
21
(
1
):
111
.
16.
MP
,
Jacquemyn
X
,
Tasoudis
P
,
Dufendach
K
,
Singh
MJ
,
de la Cruz
KI
, et al
.
Five year results of endovascular versus medical therapy in acute type B aortic intramural haematoma: meta-analysis of reconstructed time to event data
.
Eur J Vasc Endovasc Surg
.
2024
;
67
(
4
):
584
92
.
17.
MP
,
Jacquemyn
X
,
Serna-Gallegos
D
,
Makani
A
,
Kliner
D
,
Toma
C
, et al
.
Long-term outcomes of valve-in-valve transcatheter aortic valve implantation versus redo surgical aortic valve replacement: meta-analysis of kaplan-meier-derived data
.
Am J Cardiol
.
2024
;
212
:
30
9
.
18.
Mylonas
KS
,
Zoupas
I
,
Tasoudis
PT
,
Vitkos
E
,
Stavridis
GT
,
Avgerinos
DV
.
Endovascular treatment of type A aortic dissection: a systematic review and meta-analysis using reconstructed time-to-event data
.
J Clin Med
.
2023
;
12
(
22
):
7051
.
19.
Tan
DJH
,
Tang
ASP
,
Lim
WH
,
Ng
CH
,
Nah
B
,
Fu
C
, et al
.
Survival trends in sorafenib for advanced hepatocellular carcinoma: a reconstructed individual patient data meta-analysis of randomized trials
.
Liver Cancer
.
2023
;
12
(
5
):
445
56
.
20.
Page
MJ
,
McKenzie
JE
,
Bossuyt
PM
,
Boutron
I
,
Hoffmann
TC
,
Mulrow
CD
, et al
.
The PRISMA 2020 statement: an updated guideline for reporting systematic reviews
.
BMJ
.
2021
;
372
:
n71
.
21.
Sterne
JA
,
Hernán
MA
,
Reeves
BC
,
Savović
J
,
Berkman
ND
,
Viswanathan
M
, et al
.
ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions
.
BMJ
.
2016
;
355
:
i4919
.
22.
Haddaway
NR
,
Page
MJ
,
Pritchard
CC
,
McGuinness
LA
.
PRISMA2020: an R package and Shiny app for producing PRISMA 2020-compliant flow diagrams, with interactivity for optimised digital transparency and Open Synthesis
.
Campbell Syst Rev
.
2022
;
18
(
2
):
e1230
.
23.
Karaaslan
B
,
Gülsuna
B
,
Erol
G
,
Dağli
Ö
,
Emmez
H
,
Kurt
G
, et al
.
Stereotactic radiosurgery for cerebral cavernous malformation: comparison of hemorrhage rates before and after stereotactic radiosurgery
.
J Neurosurg
.
2022
;
136
(
3
):
655
61
.
24.
Régis
J
,
Bartolomei
F
,
Kida
Y
,
Kobayashi
T
,
Vladyka
V
,
Liscàk
R
, et al
.
Radiosurgery for epilepsy associated with cavernous malformation: retrospective study in 49 patients
.
Neurosurgery
.
2000
;
47
(
5
):
1091
7
.
25.
Gross
BA
,
Batjer
HH
,
Awad
IA
,
Bendok
BR
.
Cavernous malformations of the basal ganglia and thalamus
.
Neurosurgery
.
2009
;
65
(
1
):
7
19
. discussion 18–9.
26.
Pasqualin
A
,
Meneghelli
P
,
Giammarusti
A
,
Turazzi
S
.
Results of surgery for cavernomas in critical supratentorial areas
.
Acta Neurochir Suppl
.
2014
;
119
:
117
23
.
27.
Patel
A
,
Valle
D
,
Nguyen
A
,
Molina
E
,
Lucke-Wold
B
.
Role of genetics and surgical interventions for the management of Cerebral Cavernous Malformations (CMM)
.
Curr Chin Sci
.
2023
;
3
(
5
):
386
95
. Available from:
28.
Batra
S
,
Lin
D
,
Recinos
PF
,
Zhang
J
,
Rigamonti
D
.
Cavernous malformations: natural history, diagnosis and treatment
.
Nat Rev Neurol
.
2009
;
5
(
12
):
659
70
.
29.
Combescure
C
,
Foucher
Y
,
Jackson
D
.
Meta-analysis of single-arm survival studies: a distribution-free approach for estimating summary survival curves with random effects
.
Stat Med
.
2014
;
33
(
15
):
2521
37
.
30.
Moore
SA
,
Brown
RD
Jr
,
Christianson
TJH
,
Flemming
KD
.
Long-term natural history of incidentally discovered cavernous malformations in a single-center cohort
.
J Neurosurg
.
2014
;
120
(
5
):
1188
92
.
31.
Gao
X
,
Yue
K
,
Sun
J
,
Cao
Y
,
Zhao
B
,
Zhang
H
, et al
.
Treatment of cerebral cavernous malformations presenting with seizures: a systematic review and meta-analysis
.
Front Neurol
.
2020
;
11
:
590589
.
32.
Lu
XY
,
Sun
H
,
Xu
JG
,
Li
QY
.
Stereotactic radiosurgery of brainstem cavernous malformations: a systematic review and meta-analysis
.
J Neurosurg
.
2014
;
120
(
4
):
982
7
.
33.
Kim
BS
,
Kim
KH
,
Lee
MH
,
Lee
JI
.
Stereotactic radiosurgery for brainstem cavernous malformations: an updated systematic review and meta-analysis
.
World Neurosurg
.
2019
;
130
:
e648
59
.
34.
Kim
DG
,
Choe
WJ
,
Paek
SH
,
Chung
HT
,
Kim
IH
,
Han
DH
.
Radiosurgery of intracranial cavernous malformations
.
Acta Neurochir
.
2002
;
144
(
9
):
869
78
. ; discussion 878.
35.
Huang
YC
,
Hsu
PW
.
Linear accelerator radiosurgery for cavernous malformation
. In:
Hayat
MA
, editor.
Tumors of the central nervous system, volume 9: lymphoma, supratentorial tumors, glioneuronal tumors, gangliogliomas, neuroblastoma in adults, astrocytomas, ependymomas, hemangiomas, and craniopharyngiomas
.
Dordrecht
:
Springer Netherlands
;
2012
. p.
225
30
. Available from:
36.
Berber
T
,
Celik
SE
,
Aksaray
F
,
Yoney
A
,
Harmanci
K
,
Tambas
M
, et al
.
Radiosurgery effects and adverse effects in symptomatic eloquent brain-located Cavernomas
.
J Radiat Res
.
2023
;
64
(
1
):
133
41
.
37.
Polster
SP
,
Cao
Y
,
Carroll
T
,
Flemming
K
,
Girard
R
,
Hanley
D
, et al
.
Trial readiness in cavernous angiomas with symptomatic hemorrhage (CASH)
.
Neurosurgery
.
2019
;
84
(
4
):
954
64
.