Introduction

Brain arteriovenous malformations (AVMs) are the most common cause of intracranial hemorrhage in pediatric patients, accounting for 30%–50% of pediatric spontaneous cerebral hemorrhage [4, 34]. The overall mortality rate due to AVM rupture in children is as high as 21%–25% in some series [16, 19]. In general, the natural history of hemorrhage in AVMs is 1.9%–4.61% in the general population. However, the angioarchitecture of AVMs is not static and changes over time as children grow and develop into adults [6, 14], which means the annual risk of hemorrhage in children may differ from that in adults. Current treatment strategies include microsurgical resection, embolization, radiosurgery, or a combination of therapies [4]. The majority of pediatric hemorrhagic AVMs should undergo microsurgical resection because of the particularly elevated cumulative risk of recurrent hemorrhage during their lifetimes and the high rate of postoperative obliteration [20]. Generally, emergent surgery was suggested in patients with acute cerebral hernia or acute deterioration, and delayed intervention was recommended to the clinically stable ruptured AVMs for the adequate preoperative evaluation [2]. However, the high risk of recurrent hemorrhage and re-destruction of the reconstructed neurofunction might be detrimental [1, 23]. In this study, we retrospectively reviewed 111 consecutive pediatric hemorrhagic AVMs to elucidate long-term neurological outcomes and corresponding prognostic predictors, and to try to confirm the advantages of early intervention.

Materials and methods

Study design and participants

We performed a retrospective review of all patients diagnosed with AVMs resected at our neurosurgical institution from March 2010 to June 2017. Pediatric cases were isolated from the database of all AVMs with a cutoff age of 18 years at diagnosis. The study was performed according to the guideline of the Helsinki Declaration and was approved by the ethics committee of Beijing Tiantan Hospital. The inclusion criteria were as follows: (1) The diagnosis of AVM was confirmed with digital subtraction angiography (DSA) and/or magnetic resonance imaging (MRI). (2) Patients underwent microsurgical resection surgery. (3) Pediatric patients with a history of hemorrhage which could be attributed to AVM rupture. Patients with multiple AVMs, hereditary hemorrhagic telangiectasia (HHT), or unobtainable data were excluded from the study.

Data collection

Baseline demographic, clinical features and neuroimaging characteristics of the pediatric hemorrhagic AVM patients were retrospectively collected. Hemorrhage presentation was defined as hemorrhage that could be attributed to AVM rupture. The natural hemorrhage observation period was defined as the interval between birth to admission, and the re-rupture observation period was defined as the interval between the first hemorrhagic event to admission. Radiological data included review for lesion size, eloquent area, venous drainage style, and other angioarchitectural/hemodynamic characteristics (location, aneurysms, deep perforating arteries, diffuse nidus, etc.). The definition of eloquent area and deep venous drainage was consistent with the evaluation criteria in Spetzler-Martin (SM) grading system [33]. Flow-related aneurysms are operational terms describing aneurysms which lie on a pathway that carries nonnutritive blood flow (contrast) supplying the AVM shunt [17]. Diffuseness was determined from preoperative angiograms with TOF images used to identify intervening brain parenchyma within the nidus. Deep perforating arteries supplying was determined from preoperative angiograms or TOF images and included lateral lenticulostriate arteries from the M1 segment of middle cerebral artery (MCA), medial lenticulostriates from the A1 segment of anterior cerebral artery (ACA), anterior and posterior choroidal arteries, thalamoperforators from the posterior communicating artery, and P1 segment of posterior cerebral artery (PCA).

The microsurgical procedures were performed with intraoperative neuronavigation, ultrasonography, indocyanine fluorescence angiography (ICG), continuous monitoring of electroencephalogram, and somatosensory evoked potential. Circumferential dissection was carried along the borders of AVM within the perinidal gliotic tissue. The short-term neurological outcomes were evaluated 1 week after the surgery, and the long-term neurological outcomes were assessed at the last follow-up. The neurological status were assessed by modified Rankin Scale (mRS) system for children with age-specific modification as described previously [22]. The presence of new or persisting neurological deficits, or repeat hemorrhage were collected during the follow-up. A final mRS > 2 was considered disabled.

Follow-up was conducted at the first 3–6 months and annually after discharge by clinical visit and telephone interview. A brain MRI at 6 months, angiogram at 1 year, and yearly brain MRI up to 5 years after surgery were routinely performed. Researchers who performed follow-up assessments were blinded to treatment modalities.

Statistical analysis and propensity-score matching

The categorical variables were presented as counts (with percentages); the continuous variables were presented as the means ± standard deviations. Two-tailed t tests were used otherwise for continuous variable with Gaussian distribution. The Mann-Whitney U (Wilcoxon) test was used to compare non-normal distribution continuous variables. For categorical variables, either the Fisher exact test or the Pearson chi-square test was used.

Propensity-score analysis methodology (PSM) was used to compare the postoperative neurological outcomes of different surgical timing before the surgery by reducing the imbalance in the baseline patient characteristics. The hemorrhagic early phase was defined as less than 30 days after bleeding. On the basis of the covariates from the logistic model, the propensity score for each patient with respect to the baseline characteristics was generated. Nearest matching algorithm with a 1:1 ratio was applied. Absolute standardized difference analysis was performed to verify the representativeness of matching results. The outcome of interest was mRS score during follow-up. Annualized rupture rate was calculated and a Poisson rate test was used to compare the differences between the annualized rupture risk and re-rupture risk (Stata 15.0, StataCorp, College Station, TX). The follow-up time of the annualized rupture risk was calculated from the time of birth. The multi-logistic regression test was used to determine prognostic predictive factors. Statistical significance was defined as P < 0.05. Statistical analysis was performed using SPSS (Version 25.0, IBM).

Results

Baseline characteristics

A total of 111 pediatric hemorrhagic AVM patients (63.1%, 60 males and 51 females) were included from our institutional microsurgical database of 176 pediatric AVMs from March 2010 to June 2017 (Fig. 1). The patients’ mean age was 11.1 ± 4.0 years (range 1.0–18.0 years). The mean hematoma volume of preoperative hemorrhagic event was 31.1 ± 12.2 ml (range 5–50 ml). The mean pre-treatment mRS scores were 1.7 ± 0.9. Most of the lesions (101 cases, 91.0%) were located in supratentorial area. The average size was 3.4 ± 1.4 cm (range 1.1–9.1 cm), and the SM grade of all AVMs were as follows: 72 grade I–II lesions (64.9%), 25 grade III lesions (22.5%), and 14 grade IV–V lesions (12.6%). Forty-one patients (36.9%) involved eloquent areas, and 28 patients (25.2%) had deep venous drainage. Twelve patients (10.8%) were accompanied with flow-related aneurysms. Twenty-five patients (22.5%) had a diffuse nidus and 19 patients (17.1%) had deep perforating arteries (Table 1).

Fig. 1.
figure 1

The flow diagram of patient screening.

Full size image
Table 1 Demographic and angiographic characteristics of the pediatric hemorrhagic AVM patients
Full size table

The annualized rupture rate for the whole pediatric AVM cohort (n = 176) was 5.5% per patient per year. For the pediatric hemorrhagic AVM cohort, the annualized rupture rate was 9.3% per patient per year, and the annualized re-rupture rate was 9.8% per patient per year. No significant difference was found between the rupture rate and the re-rupture rate in the pediatric hemorrhagic AVM cohort (P = 0.844) (Table 2).

Table 2 The annualized rupture risk of pediatric AVM patients
Full size table

Clinical outcomes

All of the 111 pediatric hemorrhagic AVM patients maintained continuous clinical or angiographic follow-up during an average 4.3 ± 2.1 years (range 1.2–8.7 years) follow-up period. All patients (100.0%) were confirmed obliteration of the AVMs after microsurgical resection by perioperative DSA or MRI. AVM recurrence was found in one patient 9 months after the resection surgery due to recurrent hemorrhage. The mean short-term mRS score was 2.0 ± 0.9, and the mean long-term mRS score was 1.1 ± 0.9. Eight patients (7.2%) of the patients had disabilities (mRS > 2) and 91 patients (82.0%) achieved neurological deficit-free (mRS < 2) at the last follow-up. No patient died during follow-up.

During the perioperative period, 38 patients (34.2%) appeared improved mRS than pre-operation, 48 patients (43.3%) were unchanged, and 25 patients (22.5%) were worsened. Postoperative ICH occurred in three patients (2.7%) and hematoma evacuation was necessary in two patients (1.8%). In terms of long-term neurological outcomes, most of the patients (75 patients, 67.6%) were improved, whereas 22.5% (n = 25) unchanged and 9.9% (n = 11) worsened (Fig. 2).

Fig. 2.
figure 2

The change of short-term and long-term neurological outcomes in the whole pediatric hemorrhagic AVM cohort.

Full size image

Predictors of long-term disability (mRS > 2)

In the univariable logistic regression analyses, hematoma volume (25.0 ± 16.7 vs 17.2 ± 10.2 ml, P = 0.050) and pre-treatment mRS (2.6 ± 1.1 vs 1.6 ± 0.9, P = 0.002) had an unadjusted association with long-term disabilities (n = 8). Accompanied flow-related aneurysms showed a trend toward significance (37.5% vs 8.7%, P = 0.053). The multivariable logistic regression analyses demonstrated that pre-treatment mRS (OR 2.393, 95%CI 1.031–5.555, P = 0.042) accompanied with flow-related aneurysms (OR 6.174, 95%CI 1.100–34.649, P = 0.039) were independent risk factors for long-term disabilities (Table 3).

Table 3 Risk factor analysis of long-term postoperative disability in the pediatric hemorrhagic AVMs
Full size table

We further explored the relationship between surgical timing and clinical prognosis. To reduce the selectivity error caused by inconsistent baseline characteristics, we matched 27 hemorrhagic early phase cases to 27 hemorrhagic non-early phase cases according to propensity score. The two groups were compared with each other to verify that no significant differences were present between these two groups after the propensity-score matching (PSM). Finally, the short-term mRS of the early intervention group was found better than the delayed intervention group (P = 0.033). However, the long-term neurological outcomes were similar between these two groups (P = 0.367, follow-up period: average 4.4 ± 2.1 years). Interestingly, one patient in the early intervention group experienced a repeat hemorrhage from a recurrent AVM 9 months after the lesion resection+hematoma evacuation (perioperative DSA confirmed occlusion). The recurrent lesion was close to the original surgical cavity, and then the patient underwent reoperation (perioperative DSA and follow-up DSA 1 year after the surgery confirmed occlusion) (Table 4).

Table 4 Comparison of the neurological outcomes between different surgical timing (early intervention vs delayed intervention) in the pediatric hemorrhagic AVMs after propensity-score matching (PSM)
Full size table

Discussion

Studies aimed at identifying clinical features, neurofunctional outcomes, and predictors of poor outcomes in pediatric hemorrhagic AVMs were relatively rare [7, 29], especially in the selection of surgical timing [2, 12]. Previous studies have propagated that hemorrhage in pediatric AVM is associated with 25% mortality rate whereas it is 6–10% in adults [9]. Given the long life expectancy of pediatric patients and high mortality and morbidity associated with AVM-related hemorrhage, the surgical outcomes and related risk factors should be identified and the pediatric hemorrhagic AVM patients should undergo definitive treatment for obliteration of the lesion to reduce the risk of recurrent hemorrhage. In this study, we found that the annualized rupture risk for the whole pediatric cohort was 5.5%. For the pediatric hemorrhagic cohort, the annualized rupture risk was 9.3%, and the annualized re-rupture risk was 9.8%. After long-term follow-up, 7.2% of the patients had disabilities (mRS > 2) and 82.0% achieved neurological deficit-free (mRS < 2). Pre-treatment mRS score accompanied with flow-related aneurysms was independent risk factors for long-term disability. In terms of surgical timing, the short-term outcomes of the early intervention group might be better than the delayed intervention group, the long-term outcomes were similar between the two groups.

Annualized rupture risk in pediatric hemorrhagic AVMs

Previous studies have established the natural history of hemorrhage in AVMs is 0.9%-4.61% in the general population [3, 8, 35]. However, the natural history of AVMs in pediatric patients is not entirely understood in the existing literature. Although hemorrhagic AVMs are more common in children than adults, it does not imply that children’s AVMs are more prone to rupture [28]. In fact, pediatric AVMs do not come to clinical attention so often unless they bleed. Current natural-history studies of AVMs report unruptured AVMs have an annual hemorrhage rate of 2.2% while ruptured lesions have an annual hemorrhage rate of 4.5% [3]. Yang et al. reported a 0.9% annual risk of hemorrhage in a cohort of 90 pediatric patients with AVMs after excluding the treatment selective bias [35]. However, the risk was generated using the time interval from presentation to initial treatment. A recent study has identified activating KRAS mutations in the majority of tissue samples of AVMs, providing evidence that the brain AVM is a congenital disorder [24]. In this study, the observative interval of natural hemorrhage was defined as from birth to admission, the annualized rupture risk for the whole pediatric cohort was 5.5%, which was similar to previous studies. For pediatric hemorrhagic AVMs, the annualized hemorrhagic risk was 9.3%. To the best of our knowledge, this is the first study attempting to calculate the annualized hemorrhagic rate in a single cohort of pediatric hemorrhagic AVMs. For hemorrhagic AVMs, the hemorrhage rate could change over time, with the recurrent hemorrhage rate ranging from 2 to 18% in the first year after rupture across several studies [5, 27]. The present study suggested that the annualized re-rupture risk in pediatric hemorrhagic AVMs was 9.8%.

Outcomes of pediatric hemorrhagic AVMs

The interventional strategies for cerebral AVMs include microsurgical resection, embolization, radiosurgery, or a combination of these. Microsurgical resection has the advantages of immediate therapeutic cure and high obliteration rates [11], but the high incidence of postoperative complications hinders the choice of this strategy in clinical practice. Embolization as a primary and sole mode of therapy still remains questionable due to its low cure rate [25]. Radiosurgery is more favorable for deep lesions and lesions less than 3 cm in diameter [26]. However, long-term obliteration rate and control of recurrent hemorrhage during observation period remains unsatisfactory [32]. Recently, single-stage combined embolization and resection strategy using hybrid angio-surgical suite was proposed to solve complex cerebrovascular disease [30], but the applicable population, intraoperative embolization strategy, and long-term outcomes were still unclear.

Considering the long potential life span and good neurological plasticity of children [31], the goal of treatment in the pediatric hemorrhagic AVMs must be complete obliteration. Kiri et al. reported an obliteration rate of 89% in pediatric patients with low-grade AVMs after microsurgical resection [18]. Nair et al. demonstrated that 86.1% of 36 pediatric patients with SM grade I–III AVMs achieve a good functional outcome after microsurgical resection [25]. Darsaut et al. and Yang et al. included all major modalities, and the overall proportion of patients achieving good functional outcomes was 74.2% and 68%, respectively [4, 35]. In this study, all patients (100.0%) were confirmed obliteration by perioperative DSA or MRI. After long-term follow-up (mean 4.3 ± 2.1 years), 7.2% patients had disabilities (mRS > 2) and 82% of the patients achieve good neurological deficit-free. Interestingly, one patient was found to have relapsed AVM 9 months after the surgery due to recurrent hemorrhage. The underlying mechanism might be the pseudo-occlusion of residual lesions due to the hematoma mass compression or vasospasm [29].

Risk factors for long-term outcomes

High grade, large AVM size, poor baseline mRS score [13], seizure presentation [35], unruptured lesions [11], high-grade lesions, and located in eloquent cortex were considered as significant predictors of postoperative neurofunctional deficit (NFD) in pediatric AVM patients [29]. The correlation between preoperative hemorrhage and long-term postoperative outcomes was still controversial. Several studies suggested that microsurgical resection of hemorrhagic AVMs was associated with significantly lower postoperative neurological complication rates and lower intraoperative blood loss [10, 11, 21]. There was a new published AVM grading scale that includes an older patient age, non-hemorrhagic presentation, and diffuse nidus morphology as prediction model for surgical outcomes [21]. However, some studies disagreed with this view, their data showed that the prior AVM hemorrhage showing no significant association with better functional outcomes [7, 13]. In this study, 92.8% of the pediatric hemorrhagic AVMs had an mRS score < = 2 during long-term follow-up, which was consistent with previous studies [11, 29]. We found poor pre-treatment mRS scores and accompanied with flow-related aneurysms were significantly associated with long-term disability. The presence of flow-related aneurysms might make the resection surgery more challenging, especially in the ruptured AVMs. The disorganized anatomy structure and glial cell proliferation might lead to difficulty in separating lesions and accidental rupture of aneurysms. 10.8% of the pediatric hemorrhagic AVMs were accompanied with flow-related aneurysms in the present study. Many previous studies have reported adults having more flow-related aneurysms than children [14, 15], and the differences might be attributed to chronic hemodynamic stress caused by blood shunting through the AVM [7].

Timing of surgery in pediatric hemorrhagic AVMs

Microsurgical resection was considered a preferred strategy for the treatment of ruptured cerebral AVM, especially in emergency patients [2]. However, there is still no consensus on the timing of surgery. Ahmad et al. retrospectively investigated 59 patients for surgical treatment of ruptured supratentorial AVMs, and they found the time interval between AVM bleeding and surgery did not influence early or late outcomes [23]. Martinez et al. and Beecher et al. recommended a delayed intervention for at least 4 weeks after the initial hemorrhage might benefit the clinical outcomes [1, 23]. And emergency hematoma evacuation with delayed AVM excision was considered as a safe strategy in certain situations [23]. However, Bir et al. suggested that surgical intervention after 48 h resulted in poor outcomes for ruptured AVMs [2]. In the present cohort, we compared these two intervention strategies in 54 patients according to propensity score. Finally, we found that the short-term outcomes of the early intervention group (< = 30 days) were better than the delayed intervention group (> 30 days), but the long-term outcomes were similar. The causes of better short-term outcomes in the early intervention group might be the rapid release of hematoma compression effect and avoiding the occurrence of recurrent hemorrhage. And the early surgical intervention was indicated that may decrease the likelihood of focal neurological deficit deterioration upon surgery, as the delayed intervention might lead to re-destruction of the reconstructed neurofunction [12]. However, early surgery was prone to residual lesions due to incomplete preoperative examination and poor intraoperative visual field exposure, the intraoperative DSA in the Hybrid Angio-Surgical suite was recommended to verify complete eradication of the lesion [30]. In addition, the long-term outcomes between these two groups were similar, a contributing factor was probably children’s better neural plasticity [31]. Thus, early surgical intervention was recommended for pediatric hemorrhagic AVMs after the initial hemorrhage in this cohort. And further multi-center randomized controlled trials with larger sample size are needed for pediatric hemorrhagic AVMs.

Limitation

Several potential limitations of this study need to be clarified to avoid misinterpretation of our data. Firstly, this is a single-center retrospective study and selection bias existed. The operative indication and intraoperative strategies may vary according to institutional philosophy and experience. However, to our knowledge, this study cohort remains one of the largest studies in the existing literature concerning a pediatric hemorrhagic population with AVMs. Secondly, the national policy of graded diagnosis and treatment might lead to selection bias of patients by promoting more complex pediatric AVM patients to our tertiary neurosurgical center, which will lead us to underestimate the outcomes of the pediatric hemorrhagic AVM patients.

Conclusions

The surgical intervention treatment for pediatric hemorrhagic AVMs is recommended in that the cumulative lifetime hemorrhage risk is substantial. In this patient series, the annualized rupture risk for the whole pediatric AVM cohort was 5.5%. For the pediatric hemorrhagic AVM cohort, the annualized rupture risk was 9.3%, and the annualized re-rupture risk was 9.8%. Based on the better neurological plasticity, most pediatric hemorrhagic AVM patients can achieve neurological deficit-free after surgical resection. Pre-treatment mRS score and accompanied with flow-related aneurysms were independent predictors for long-term disability (mRS > 2). In addition, early surgical intervention is recommended for pediatric hemorrhagic AVMs after the initial hemorrhage.