Lifetime Risk of First Symptomatic ICH or Seizure in Familial Cerebral Cavernous Malformations: A Multicenter Patient Data Analysis.

Abstract

Background and Objective:

Familial cavernous malformations (FCM) are vascular lesions that pose a lifelong risk of symptomatic hemorrhage (SH) and seizures, yet their natural history remains unclear. This study aims to determine the cumulative lifetime risk of a first SH and/or seizure and assess whether genetic mutations influence these risks.

Methods:

This international, multicenter retrospective cohort study included data from 16 tertiary referral centers and one patient advocacy group. Eligible patients had confirmed or suspected FCM, available magnetic resonance imaging (MRI) data, documented baseline clinical features, and longitudinal follow-up. Functional outcomes were assessed using the modified Rankin Scale (mRS) at last follow-up. Direct adjusted survival curves and mixed-effects Cox regression analyses were performed to estimate cumulative lifetime risk. The association between genetic mutations and SH/seizure rates was evaluated, and mixed-effects logistic regression assessed the impact of SH/seizures on mRS outcomes.

Results:

A total of 1,592 FCM patients were included, with a mean age of 37.6 years (SD 17.1) and 55.7% female. The median follow-up was 42 years (IQR: 27–55), totaling 64,146 person-years. Of these, 869 (54.6%) had confirmed FCM, 775 (48.7%) experienced at least one hemorrhage, and 447 (28.1%) had at least one seizure. Genetic testing was performed in 47.7%, identifying CCM1 (31.0%), CCM2 (4.8%), and CCM3 (1.9%) mutations. The lifetime risk of a first SH was ~80%, with an event rate that remained constant beyond age 20. The lifetime risk of a first seizure was ~45%. Patients with CCM3 mutations exhibited a more aggressive hemorrhagic course than those with CCM1 (HR 1.799, 95% CI: 1.008–3.208). SH and seizures were independently associated with worse mRS outcomes at last follow-up.

Discussion:

The event rate of symptomatic hemorrhage and seizures remained stable over time, leading to high cumulative lifetime risks. Patients with CCM3 mutations exhibited a more aggressive disease course. Limitations include the non-population-based design, selection bias from tertiary centers, retrospective data collection, and variability in data extraction across centers. However, this study represents the largest international FCM cohort to date, improving the precision of risk estimates and providing valuable insights into disease progression.

Introduction

Cerebral cavernous malformations (CCM) represent a frequent type of neurovascular malformations¹ and a substantial cause of spontaneous intracerebral hemorrhage (ICH) with an estimated prevalence of around 0.5% in the general population^2,3. CCM-related ICH is prone to detrimental symptoms such as seizures and/or focal neurological deficits able to considerably impact the patients quality of life^2,4–6. Around 15–20% (prevalence ~1:2000–3000) of CCM occur in the familial form of the disease^7–9 following an autosomal dominant inherited germline mutation in genes located on chromosomes 7q (CCM1), 7p (CCM2), and 3q (CCM3)^9–11. A known hallmark feature of familial CCM (FCM) is the presence of multiple CCM lesions distributed throughout the whole central nervous system, as well as the absence of a CCM-associated developmental venous anomaly (DVA)^9,10,12–16. In view of the well-known difficulties in obtaining a genetic confirmation of FCM disease, patients with suspected FCM can be subdivided into confirmed FCM cases (multiple CCM, absence of DVA, and at least one other member of the same family with CCM or positive genetic testing), and assumed FCM cases (multiple CCM, absence of DVA, and either no concerned relative or unknown family history, no genetic testing performed)^9,15,17. To date, only few studies have tried to assess the natural course, as well as the overall cumulative risk of symptomatic ICH and/or seizure in patients with FCM^13,15–22. However, due to the rarity of this specific subgroup of patients, such studies mainly include small uni-center cohorts with short follow-up (FU) periods.

Consequently, substantial information on the specific FCM-associated risk for experiencing ICH or seizures is limited. Treatment decisions often rely on studies reporting on the much more common sporadic form of CCM. To this extent, and in view of the unmet need to investigate larger cohorts, we aimed to assess the lifetime risk of hemorrhage or seizure in FCM patients in an individual patient-level data analysis from 17 centers located in different regions around the world.

Methods

Participating Centers

We carried out a systematic review (PROSPERO register CRD42021274956) to identify recent (2011–2021) international cohorts of FCM with a minimum size of 30 patients. 27 cohorts were identified and contacted. 16 different tertiary referral centers and 1 patient advocacy group responded and agreed to participate in a data pooling study (Figure 1). Data transfer agreements were reached with all centers. Local ethics committees approved the participation in this study at all sites.

Figure 1:

Flow chart of the Systematic review according to PRISMA Guidelines. FCM corresponds to Familial Cerebral cavernous malformation.

The study includes the following sixteen centers and 1 patient advocacy group: University Hospital Essen, Germany; French Institute of Health and Medical Research, France; Texas Tech University Health Sciences Center, El Paso, United States; Fondazione IRCCS-Casa Sollievo della Sofferenza, San Giovanni Rotondo (FG), Italy; Beijing Tiantan Hospital, China; Greifswald University Medical Center, Germany; Stanford Medicine Health Care, United States; University Hospital Bern, Switzerland; Institute for Neurological Research Dr. Raul Carrea (FLENI), Buenos Aires, Argentina; Mayo Clinic Rochester, United States; Royal Edinburgh Hospital, Scotland; Freiburg University Hospital, Germany; the Brain Vascular Malformation Consortium^21–23, which included the University of New Mexico, University of California-San Francisco, Barrow Neurological Institute, and the Alliance to Cure Cavernous Malformations, United States.

Standard Protocol Approvals, Registrations, and Patient Consents

This study was approved by the University Hospital Essen Institutional Review Board (approval: 14–5751-BO, 19–8662-BO) and local ethics committees at all participating centers. As a retrospective study, ethics boards waived informed consent, and all patient data were de-identified. No experiments on live vertebrates were conducted. Authorization for identifiable patient disclosures has been obtained where applicable.

Data Availability

De-identified participant data, study protocol, and statistical analysis plan will be available upon request for five years after publication. Researchers with valid scientific requests can contact the corresponding author at philipp.dammann@uk-essen.de. Access will comply with institutional and ethical regulations.

Study Design and Participants

This multicenter cohort study assessed (through patient registry) all patients in the above mentioned institutions that fulfilled the following inclusion criteria and were treated up to the year 2022: diagnosis of confirmed or assumed familial CCM, available magnetic resonance imaging (MRI) data, as well as digitalized and complete data records. Patients with an assumed or confirmed familial disease were defined as: presence of multiple CCM and/or genetically confirmed by testing and/or known affected relatives, as well as the absence of associated developmental venous anomaly (DVA)^12,24. De-identified patient-level datasets included the following characteristics: age at diagnosis, date of diagnosis, sex, genetic confirmation of CCM-related mutations, presence of the disease in other members of the same family, CCM-related ICH, date of first hemorrhage, CCM-related seizure, date of first seizure, neurological functional outcome through the modified Rankin scale (mRS) at last follow-up, date of last follow-up, surgical or radio-surgical treatment of a lesion, presence of multiple treatments.

Mode of presentation was determined according to the following reporting standards: asymptomatic, non-hemorrhagic focal neurological deficit, non-hemorrhagic seizure, symptomatic bleeding⁵. FU data were acquired through routine appointments in the specialized outpatient clinics of the above-mentioned centers. Follow-up duration was recorded based on the last follow-up year (LFU year). Follow-up data were unavailable for UMCEP patients due to institutional differences in record-keeping, but center-adjusted analyses confirmed no significant impact on overall conclusions. Endpoints of the study were the occurrence of ICH according to the ensuing reporting standards⁵: radiological evidence of an acute ICH of the CCM seen on a recent MRI accompanied by an acute or subacute onset of neurological symptoms related to the anatomical region of the lesion. Cavernoma related seizure was defined as a definite association of seizure with a circumscribed CCM according to an experienced neurologist in the above-mentioned centers. Neurological functional outcome was assessed through the mRS at the last available FU examination. A favorable outcome was defined as mRS ≤1, whereas an unfavorable outcome was defined as mRS ≥2²⁵.

Statistical Analysis

We used SAS software, version 9.4 (SAS Institute Inc., Cary, NC, USA.) for all statistical analyses. We investigated the association of clinical data with the outcome (occurrence of ICH or seizure) during a lifetime follow-up period (i.e., follow-up time starting at birth) by performing mixed effects Cox regression analysis with direct adjusted survival curves for averaging the estimated survival curves and the study centers as random effect to fit a shared frailty model for clustered data. The approach used has the advantage of computing survival curves that reflect imbalance between treatment groups by averaging the predicted survival functions for covariates such as study center. Hazard ratios (HR) and 95% Cls were reported^26–28. Model assumptions such as validity of the proportional hazards was confirmed prior to analysis. Data were censored if patients experienced a (re)bleeding (or seizure) or were lost to follow up. Overall survival curves were also computed for each study center separately. Direct center adjusted survival curves were also computed while excluding one center each from the analysis to check how strong the impact of singles study centers was on the overall shape of the survival curve. For the association of the occurrence of ICH and seizures with the mRS at last follow up, boxplots were presented. Furthermore, mixed effects logistic regression models adjusted for age at last follow up and sex with study center as random effect were calculated with dichotomous mRS as the dependent variable and ICH/seizures as the independent variable.

Results

Patient Demographics

A total of 1592 patients were included in the analysis. The mean age was 45.5 ± 18.4 years and 886 patients (55.7%) were female. 775 patients (48.7%) suffered from a SH and 447 (28.1%) revealed at least one seizure during FU. The mean age at outcome occurrence was 36.3 ± 18.2, and 29.3 ± 17.6 years respectively. 29.3% (466 patients) received at least one surgical resection of a lesion, whereas only 2.8% (45 patients) underwent radiosurgery. Most patients with available genetic testing had a CCM1 mutation (494 patients; 31%), while 76 (4.8%) and 30 (1.9%) patients had a CCM2 or CCM3 mutation, respectively. Out of all patients included in our analysis, 869 patients (54.6%) revealed a confirmed FCM diagnosis, and the other 723 patients (45.4%) had an assumed FCM diagnosis. Although most patients (775 patients; 74%) demonstrated a favorable functional outcome at the last FU, 271 patients (26%) suffered from an unfavorable mRS at the last FU. (Table 1)

Table 1:

Demographic, anatomic and clinical characteristics.

Characteristic	Frequency
Total number of patients, n	1592
Age at last FU, years, mean ± SD*	45.5 ± 18.4
Female sex, n (%)	886 (55.7 %)
Bleeding, n (%)	775 (48.7 %)
Age at bleeding, years, mean ± SD†	36.3 ± 18.2
Seizure, n (%)	447 (28.1 %)
Age at seizure, years, mean ± SD‡	29.3 ± 17.6
Treatment modality, n (%):	511 (32.1%)
- Neurosurgery, n (%)	466 (29.3 %)
- Radiosurgery, n (%)	45 (2.8 %)
- Multiple treatments, n (%)	50 (3.1 %)
Genetic mutation1:
- CCM1, n (%)	494 (31.0 %)
- CCM2, n (%)	76 (4.8 %)
- CCM3, n (%)	30 (1.9 %)
Familial Disease:
- Confirmed, n (%)	869 (54.6 %)
- Assumed, n (%)	723 (45.4%)
Functional Outcome at last FU (mRS), n (%)§
- 0	334 (31.9 %)
−1	441 (42.1 %)
−2	138 (13.2 %)
−3	77 (7.4 %)
- 4	38 (3.6 %)
- 5	8 (0.8 %)
- 6	11 (1.1 %)

CCM indicates cerebral cavernous malformation; SD, standard deviation; n, number of patients; FU, follow-up; mRS, modified ranking scale.

¹not tested for genetic mutation or no mutation detected in 992 patients.

^*38 patients missing;

^†837 patients missing;

^‡1149 patients missing;

^§545 patients missing.

Cumulative lifetime probability of hemorrhage

A total of 755 first SH events occurred during a median follow up time of 42 years (IQR: 27–55) across all centers, leading to 64146 person-years of follow up with an incidence rate of 11.8 events per 1000 person-years. Until ~20 years of age the event rate was slightly lower compared to the overall risk later in life, while after 20 years of age the event rate remained constant across the life course with an overall lifetime risk of first bleeding of ~80% (Figure 2a).

Figure 2:

(A) Bleeding events through Kaplan-Meier curves of the complete cohort, (B) stratified by center (associated center is marked on the right end of the figure with a personal color and form), and (C) stratified by genetic mutation (CCM1 shown in blue; CCM2 shown in green; CCM3 shown in red) The numbers of patients at risk are shown in a corner of each graph. HR of ICH between CCM1 VS CCM2 = 0.840/ HR of ICH between CCM3 VS CCM1 = 1.799

Survival curves for each center separately can be found in Figure 2b. Although this figure shows some heterogeneous results between centers, the exclusion of one center each from the analysis had no strong impact on the overall shape of the survival curve, meaning that the overall shape of the pooled survival curve did not seem to be mainly driven by a single center.

Mixed effects Cox regression analysis in patients with non-missing information on gene status indicated that patients with a CCM3 mutation had a higher risk of bleeding across their life course compared to patients suffering from a CCM1 (p = 0.047; HR: 1.799; 95% CI: 1.008 – 3.208). The risk of bleeding in patients with a CCM2 mutation was slightly lower compared to patients with CCM1 (p = 0.426; HR: 0.840; 95% CI: 0.548 – 1.290) mutation. (Figure 2c).

Cumulative lifetime probability of Cavernoma related seizure

A total of 443 first seizure events occurred during a median follow up time of 43 years (IQR: 26–57 years), leading to 64927 person-years of follow up with an incidence rate of 6.8 events per 1000 person-years. The event rate of seizure remained constant across the life course with a lifetime seizure risk of ~45% in the study population as seen in Figure 3a. After the exclusion of one center that did not assess seizure events in their patients, survival curves were performed for each remaining center separately (Figure 3b). As with bleeding risk, this figure shows some heterogeneity between centers. The exclusion of one center each from the pooled survival curve had no strong impact on its overall shape, meaning the pooled results did not seem to be mainly driven by a single center. Results of mixed effects Cox regression analysis gave no strong indication of a higher seizure risk in patients with a CCM3 mutation compared to patients suffering from a CCM1 (p = 0.400; HR: 1.373; 95% CI: 0.657 – 2.869). Patients with the CCM2 mutation showed no strong indication for a lower risk of seizure compared to patients with CCM1 (p = 0.287; HR: 0.764; CI: 0.466 – 1.254) mutation.

Figure 3:

(A) Seizure events through Kaplan-Meier curves of the complete cohort, (B) stratified by center (associated center is marked on the right end of the figure with a personal color and form), and (C) stratified by genetic mutation (CCM1 shown in blue; CCM2 shown in green; CCM3 shown in red) The numbers of patients at risk are shown in a corner of each graph. HR of ICH between CCM1 VS CCM2 = 0.764/ HR of ICH between CCM3 VS CCM1 = 1.373

Functional Outcome

Each center assessed the functional outcome of their patients at last available FU. Overall, we found a higher proportion of impaired functional outcome in patients after a bleeding event compared to patients without a bleeding event (figure 4a). Mixed effects logistic regression models adjusted for sex and age at last follow up showed an OR of 2.7 (95%Cl 1.98–3.73) for having an mRS of ≥2 after bleeding. Furthermore, we assessed the association of seizure and functional outcome in our cohort and equally found some indication for unfavorable functional outcomes in patients after a seizure event compared to patients without (Figure 4b). Mixed effects logistic regression models adjusted for sex and age at the last follow up showed an OR of 1.27 (95%Cl 0.93–1.75) for having an mRS of ≥2 after seizure.

Figure 4:

Box plot graphics showing the association between functional outcome through modified Rankin Scale (mRS) and siezure events (left graphic), as well as with ICH events (right graphic). mRS score schowed on the Y-axis, presence of seizure or ICH seen in the X-axis.

Discussion

Symptomatic hemorrhage in CCM is not rare and often prone to significant morbidity affecting quality of life^2,4–6. Although patients suffering from FCM represent an important subgroup of this disease, studies investigating their natural course are missing^{9,10,13,15–17,29–31}. Our study represents an unprecedented international collaboration, as well as the largest observational cohort of FCM studied so far. We analyzed data from FCM patients, assessed the lifetime risk of hemorrhage and/or seizure, and identified risk factors for unfavorable events. Finally, we evaluated the functional neurological outcome of patients in our cohort, as well as possible risk factors influencing their outcome.

Lifetime risk of Hemorrhage

We observed a constant event rate for ICH after ~20 years of age, as well as a lifetime ICH risk of ~80%. Previous large studies assessing bleeding risk in sporadic cases, as well as small cohorts assessing the latter in FCM show a 5-year cumulative risk of 20%^{2,9,15,21,32,33}. Although not assessing the lifetime risk, such studies show an increasing risk for ICH throughout time, with studies even showing an increased risk going up to 30% in a 10-year FU period³⁴. Their findings correlate with our data indicating a constant event rate for bleeding throughout time. Notably, we further segregated our cohort into groups according to their specific gene mutations. Here, we discovered a significantly more aggressive course of the disease with higher rates of ICH in patients suffering from CCM3 mutation. Such findings are in accordance with previous literature that assessed the latter and also found that patients with CCM3 mutation have a more aggressive course in comparison with CCM1 and CCM2¹⁷.

Lifetime risk of Seizure

The lifetime risk of seizures in our cohort was ~45%. Although overall risk was lower than the lifetime risk for ICH, we also found constant seizure rates across the life course. So far, Fox and colleagues assessed the largest cohort analyzing the risk of seizure in patients with FCM²². The cumulative incidence of seizures was 60% by age 80 years, and risk of seizure was significantly higher in those with a genetically confirmed CCM3 mutation vs. CCM1 or CCM2 mutation. In our multi-center cohort (including the data of Fox et al. (n=479)) we found a slightly lower risk of a first seizure with constant event rates of seizure throughout time. We similarly observed a potentially more aggressive course of the disease in terms of seizure rates in patients with a CCM3 mutation.

Functional Outcome

Large studies regarding sporadic CCM, as well as some FCM studies assessing possible risk factors leading to an unfavorable functional outcome in patients with a sporadic CCM have been conducted^{2,4,9,17,35–37}. These studies validated a possible association between the occurrence of events (SH/seizure) and unfavorable functional outcome.

Accordingly, we found a higher proportion of impaired functional outcomes in patients after SH and/or seizure events, independent of age, reflecting the impact of these events on the patients’ functional performance. This effect was significantly stronger for patients with SH compared to seizures without SH.

Strengths and Limitations

The prevalence of CCM in the general population is around 0.5%^2,7 and FCM represents 15–20% of these patients³⁸, rendering large cohorts difficult. Our data is not population-based as we obtained it from multiple tertiary referral center, which can lead to selection biases. Moreover, part of the data was obtained in a retrospective study design and some data was already published in other studies of the respective centers. Another important factor is the potential overestimation of event rates due to the reduced penetrance of FCM. Prior studies indicate that up to 50% of individuals with pathogenic variants remain asymptomatic throughout their lifetime despite harboring multiple lesions^9,39. This could suggest that the true risk of ICH and seizures in FCM may be lower than reported here. Nevertheless, our results outline a life course FU study on patients with a familial disease and contributes novel data in an international collaboration of the largest dataset of FCM patients to date, improving the precision of event rate estimates and risk factors substantially. Due to the multicenter nature of our study, discrepancies in terms of data extraction between centers were possible. Compared to other studies regarding FCM, our combined cohort seems representative in terms of patient characteristics^21,22,24,40. Moreover, the substantial number of centers involved in our study showing pooled results in terms of lifetime risk of ICH and/or seizure were not strongly affected by any single center.

Future Directions

Despite the strengths of our study, several important questions remain unanswered, paving the way for future research. One of the key limitations of the current analysis is the lack of detailed lesion characterization, particularly regarding size, location, and growth dynamics. Given the increasing availability of advanced imaging modalities and automated volumetric analysis, future studies should aim to integrate standardized lesion measurements to assess their impact on ICH and seizure risk. Additionally, while our study provides crucial lifetime risk estimates for FCM, further work is needed to explore longitudinal lesion evolution and potential modifiers of disease severity, including genetic, environmental, and treatment-related factors. Prospective natural history studies incorporating systematic imaging follow-ups, genetic profiling, and functional assessments will be essential in refining risk stratification models and guiding personalized treatment approaches for FCM patients. Finally, given the challenges associated with studying rare diseases, establishing international collaborative registries with harmonized imaging and clinical data collection protocols will be critical in further advancing our understanding of FCM and optimizing patient management.

Conclusions

During a lifelong follow-up, the event rate for a first symptomatic hemorrhage or seizure remained constant over time reaching a cumulative risk of around 80% and 45%, respectively. Our findings indicate a more aggressive course in patients with CCM3 mutation. Occurrence of symptomatic hemorrhage and/or seizures were associated with a more unfavorable mRS outcome at last follow up, independent of age.

Non-standard Abbreviations and Acronyms

FCM

familial CCM

CCM

cerebral cavernous malformations

DVA

developmental venous anomaly

ICH

intracerebral hemorrhage

FU

Follow-up