The study was designed as an Italian retrospective multicenter survey, realized on behalf of the Italian Society of Pediatric Neurology (SINP). An ad-hoc questionnaire to collect clinical information was emailed to each SINP member. Other specialists, including members of the Italian Societies of Neurosurgery and of Neuroradiology, and pediatric hematologists of the Italian Association of Pediatric Hematology Oncology, were involved through personal contacts by the study coordinators. Data collection was from December 2015 to August 2018.

Inclusion criteria were diagnosis of MMD or MMS according to the international consensus diagnostic criteria (2, 13); and symptom onset or incidental diagnosis before age 18 years.

Clinical presentation of moyamoya was categorized according to the international consensus “disease types” (2): TIA, recurrent TIA, ischemic stroke (IS), hemorrhagic stroke, headache, epilepsy, and others. Asymptomatic patients diagnosed incidentally during investigations performed for other medical conditions were considered as a distinct group (incidental MMS, “incMMS”), with the aim of identifying possible differences from patients with MMD or MMS who had moyamoya-related symptoms at the time of diagnosis (“symptMMD,” “symptMMS,” collectively denominated “symptMMD/MMS”). Moreover, the categories of “single” and “multiple phenotypes” were introduced to differentiate patients with history of one or more than one disease type at the time of diagnosis. Outcome was investigated with open fields about “cognitive,” “motor,” “memory,” and “behavior” problems and by using the modified Rankin Scale (mRS), a scale initially created for stroke but now widely used in pediatric neurology, despite its limitations in detecting non-motor issues (14). We operatively defined poor outcome as mRS > 2.

Available neuroimaging scans were reviewed by an experienced pediatric neuroradiologist in order to evaluate unilateral vs. bilateral involvement, anterior vs. posterior circulation distribution, presence of ischemic or hemorrhagic lesions, and micro-bleedings.

The study complied with the general ethical requirements for retrospective observational studies. In particular, no experimental interventions were performed, and patient identity cannot be retrieved from the manuscript.

Descriptive statistic was used to organize and present data.

When comparing the three groups of patients (symptMMS, symptMMD, and incMMS) with respect to quantitative variables, the Mann–Whitney test followed by the Dwass–Steel–Critchlow–Fligner method for multiple pairwise comparisons were used. For categorical variables, the groups were compared with Fisher's exact test followed by pairwise comparisons correcting the p-values with Bonferroni's method.

Incidence of events in the follow-up was estimated along with 95% confidence interval, with a Poisson model considering the natural logarithm of follow-up time as offset parameter. The statistical analysis was performed with SAS 9.4 (SAS Institute Inc., Cary, NC, USA) for Windows.

Sixty-five patients were included (34/65, 52% women). Mean age at moyamoya diagnosis was 7.4 years (median 7, range 0.4–24.5). A total of 27/65 (42%) patients had MMD and 38/65 (58%) patients had MMS. Among patients with MMS, 22/38 (58%) patients had underlying genetic conditions (most frequent being neurofibromatosis type 1 in 6/22, sickle cell disease in 5/22, and other suspected genetic multisystemic syndromes in 8/22; Table 1A).

Twelve of the 65 patients (18%), all patients with MMS, were diagnosed incidentally during investigations performed for an underlying medical conditions (incMMS), whereas the remaining 53 patients (82%) with MMD or MMS were diagnosed due to the presence of neurological symptoms (symptMMD/MMS; Table 1A and Figure 1A).

Age at diagnosis was significantly lower in the 53 symptomatic patients (symptMMD/MMS) than in the 12 patients diagnosed incidentally (incMMS) (p = 0.0467) (Table 2).

The first-ever clinical manifestation was IS in 19/65 (29%), TIA in 17/65 (26%), seizures in 9/65 (14%), and headache in 8/65 (12%) patients (Table 1A and Figure 1B; all symptMMD/MMS).

At the time of diagnosis, 36/65 (55%) patients had a history of a single clinical phenotype, whereas 17/65 (26%) patients had a history of multiple phenotypes (all symptMMD/MMS) (Table 1B and Figure 1A), but none of the patients had hemorrhagic stroke.

Overall, by the time of diagnosis, cerebrovascular events (TIA and IS), whether isolated or associated with other symptoms, occurred or had occurred in 43/65 patients (66%) (all symptMMD/MMS) (Figure 1C): TIA/TIAs in 21/65 (32%) patients (estimated mean age at TIA presentation: 6 years), and IS in 27/65 (42%) patients (mean age at stroke presentation: 4.1 years; median 2.75, range 0.4–14). Among patients with symptMMD/MMS, stroke before diagnosis was more frequent among patients with symptom onset before age 24 months (7/10, 70%) than in those with later onset (20/43, 47%) (Table 3 and Supplementary Material S1).

The presence or history of headache before diagnosis was reported in 12/65 (18%) patients (tension-type headache in 1/12 patients, and the remaining patients had migraine associated with tension-type headache) (all symptMMD/MMS). Of these 12 patients, 4/12 patients had isolated headache (mean 32 months between headache onset and moyamoya diagnosis; median 24, range 0–72), 7/12 patients had headache followed by other clinical manifestations (mean 42 months between headache onset and moyamoya diagnosis; median 16, range 1–120), and 1/12 patients had headache occurring after an undiagnosed IS (25 years between stroke and moyamoya diagnosis).

Movement disorder (dyskinetic, choreiform, and dystonic pattern) was reported in 3/65 (5%) patients before diagnosis (age range 1.25–13.3 years) (all symptMMD/MMS), associated with different clinical phenotypes (mainly IS).

47/64 (73%) patients underwent neurosurgery, between 1995 and 2017: 42/47 (89%) indirect revascularization, 4/47 (9%) combined techniques, and 1/47 (2%) direct revascularization.

Mean age at surgery was 6.1 years (median 4.8, range 0.4–17.7) and at mean 15.7 months after diagnosis (median 3, range few days to 8 years; data available in 39/47 patients). Surgery was carried out within 1 month after diagnosis in 10/47 operated patients (21%).

Overall, stroke at diagnosis had been reported in 23/47 operated patients (49%), and in 7/10 (70%) of the patients operated within 1 month after diagnosis (mean age at diagnosis in this latter subgroup: 3.5 years).

Surgery complications occurred in 3/47 patients (6%): one patient had a severe intra-operatory ischemic damage; one patient had a mild frontal blood suffusion; and one patient had subdural hematoma.

A second intervention was carried out in 22/47 patients (47%) (21/22 patients to complete the first surgery, and 1/22 patients to evacuate a subdural hematoma). Complications of the second surgery occurred in 4/22 (18%) patients: removal of a iatrogenic foreign body in one patient; liquor fistula requiring lumbar puncture in one patient; cerebral edema due to revascularization in one patient; and asymptomatic parenchymal hematoma with spontaneous resolution in one patient.

Acetylsalicylic acid (ASA) was used in 45/65 patients (69%) overall, including 38/47 (80.9%) operated patients. Transient adverse events occurred in 3/45 patients overall (petechiae, cutaneous reactions). ASA was started within 1 month after diagnosis in 26/45 (58%) patients, between 4 and 12 months in 4/45 (9%) patients, and more than 1 year later in 5/45 (11%; data available for 35/45 patients).

Unfractioned/low-molecular-weight heparin (LMWH) was used in 5/65 (8%) patients (4 patients were rapidly shifted to ASA; and 1 patient continued LMWH for 6 months).

Flunarizine was used in 6/65 patients (9%) (adverse events in 1/6 patients: weight gain), and calcium blockers were used in 3/65 (5%) patients (no adverse events).

Antiseizure medications were used in 30/65 patients (46%).

Among the 5 patients with sickle cell disease, 3 patients received hematopoietic stem cell transplantation, and 1 patient started red cell exchange transfusion after an IS.

At least one neuroimaging scan was available for 58/65 patients (89%). In 45/58 patients with available information on neuroimaging timing with respect to diagnosis, 19/45 patients were with an available scan at diagnosis; and the remaining patients were performed later in the follow-up. Moyamoya diagnosis received confirmation in 58/58 patients (100%) of neuroimaging after revision by our senior pediatric neuroradiologist. Bilateral steno-occlusions and networks were observed in 47/58 patients (81%). Among patients with MMD with available neuroimaging scans, 22/24 patients (92%) had bilateral steno-occlusions and networks; and 2/24 patients (8%) had bilateral steno-occlusions and left unilateral network. Among patients with MMS with available neuroimaging scans, 25/34 patients (74%) had bilateral steno-occlusions and networks, 8/34 patients (23%) had unilateral steno-occlusion and network (2 left-sided; 6 right-sided), and 1/34 patients (3%) had bilateral steno-occlusions and left unilateral network.

Posterior circulation involvement was documented in 30/58 patients (51%), and at least one ischemic lesion in 39/58 patients (67%). Neuroimaging data are reported in Figure 2 and Supplementary Material S2.

Data on disease course were available for 61/65 patients.

Overall, 24/61 patients experienced cerebrovascular events (TIA and IS) after diagnosis (39%): 12/61 patients (20%) had IS (3 of them had their first stroke after diagnosis), and 12/61 patients (20%) had TIAs. The mean age at disease onset in children who experienced IS after diagnosis was lower than in children with no stroke after diagnosis (3 years vs. 6 years).

Only 1 child experienced TIA in the follow-up without having had any cerebrovascular event before diagnosis.

Among 47 patients who underwent surgery, 11/47 patients suffered from cerebrovascular events after surgery (23%): 5/47 patients had IS (11%) (5/5 patients already had stroke at the time of diagnosis), and 6/47 patients had TIAs (13%).

Based on these data, it was possible to calculate, during the entire period from symptom onset to the last follow-up, a patient/year incidence of 10.26% (IC 95% 7.58–13.88%) for stroke, 9.28% (IC 95%: 6.75–12.75%) for TIA, and a combined incidence of 19.54% (IC 15.69–24.32%) for cerebrovascular events (TIA or IS).

The involvement of posterior circulation at neuroimaging increased the incidence patients/year to 15.51% (10.78–22.32%), 9.63% (6.07–15.28%), and 25.14% (18.89–33.45%), respectively, for stroke, TIA, and cerebrovascular events.

The occurrence of cerebrovascular events after diagnosis was more frequent among patients with symptom onset before 24 months of age (7/10, 70%) than in those with symptom onset after 24 months of age (17/39, 44%) (Table 3 and Supplementary Material S1).

Headache after diagnosis was reported in 25/61 patients (41%) (migraine in 10/25 patients, tension headache in 3/25 patients, more than one headache type in 2/25 patients, and not specified in the others). Headache frequency was not uniformly reported; generally, for most patients, intensity and frequency of episodes reduced in time.

Seizures occurred in 9/61 patients after diagnosis (15%) (6/9 patients had new onset of seizures after diagnosis; 4/9 patients experienced seizure after surgery).

Movement disorders were reported in 5/61 patients after diagnosis (8%; in 1/5 patients, the symptom had started before diagnosis) (dyskinetic movements in 2/5 patients, dystonic movements in 2/5 patients, and myoclonic movements in 1/5 patients).

At last follow-up, 20/52 patients (39%) had mRS 0, 10/52 patients (37%) had mRS 1–2, and 13/52 patients (25%) had mRS > 2 (Table 1B and Figure 3); no deaths occurred.

The proportion of patients with mRS > 2 at last follow-up was significantly higher in symptomatic patients (symptMMD/MMS) than in patients diagnosed incidentally (incMMS) (p = 0.021) (Table 2).

Age <4 years at symptom onset and IS (previous or recent) at diagnosis were both significantly associated with increased risk of intellectual disability (p = 0.0010 and p = 0.0071, respectively) and of mRS > 2 at last follow-up (p = 0.0106 and p = 0.0009, respectively) (Table 4).

Symptom onset in pre-school age was frequent in our cohort, similar to other Western European and extra-European cohorts (15, 27–29) without significant differences between MMS and MMD (Table 2).

Although our study confirms some degree of diagnostic delay as in the literature (18, 29–31), this is fairly reduced compared to the previous comparable Italian report published 25 years ago (16.8 vs. 25 months) (32) (Supplementary Material S4), possibly due to higher levels of suspicion and greater availability of noninvasive cerebrovascular studies.

Cerebrovascular events (IS and/or TIA) were by far the most frequent clinical manifestation of moyamoya in our cohort (Figure 1): 66% (43/65) of our patients experienced at least one cerebrovascular event (IS in 42%, 27/65) before diagnosis, consistent with other East Asian and Western studies (8, 15, 28, 29), while there were no hemorrhagic strokes.

The mean age at stroke presentation in our cohort was about 4 years, with a rate of stroke presentation before 4 years old of about 60%. The age at presentation of stroke due to any cause is generally older, ranging from 5.6 years old in the Swiss registry (33) to about 7–8 years in the German and the International Pediatric Stroke Study (34, 35); moreover, in the Canadian Pediatric Stroke Registry, only 33% of patients were <4 years old at diagnosis (6). These data suggest that probably moyamoya could cause IS earlier than other stroke etiologies.

In our study, about 1 out of 5 patients experienced TIA in the weeks before stroke, without receiving a formal diagnosis of moyamoya. The occurrence of one or multiple TIAs in the months prior to stroke has been previously described (29, 36). Furthermore, as already reported in another pediatric MMD cohort, in our study the age at TIA onset, about 6 years old, was lower, compared to usual onset of TIA in pediatrics (37). This highlights the importance that every child with a TIA—especially in early school age or before—receive prompt and extensive studies to rule out cerebral angiopathies.

Previous data have shown that cerebrovascular events are not the only manifestation of pediatric moyamoya, with headache at presentation, isolated, or associated with other clinical manifestations, ranging between 6.3% (8), 22% (38), and up to about 50% (29). Similarly, in our cohort, headache (isolated or with other symptoms) was reported in 18% of patients (12/65) before diagnosis and in 41% of patients (25/61) during the disease course, highlighting the importance of not overlooking headache as a neurological symptom, especially when associated with other phenomena. We could not identify a specific moyamoya headache pattern, differently to other secondary headaches (39), although the most represented was migraine-like.

Movement disorder was experienced by a subgroup of our patients, at presentation or during the disease course, but in none at last follow-up: disappearance of movement disorder during the clinical history has been previously described, even if the reasons are unclear (40). A relationship between movement disorder disappearance and surgery is likely also in our patients, independently from the mechanism (16, 40).

Moyamoya is a heterogeneous disease, with possible occurrence of multiple symptoms and disease types by the time of diagnosis (8). In our study, 26% of patients had “multiple phenotypes” at the time of diagnosis, defined as the presence of more than one disease type (TIA/TIAs, IS, hemorrhagic stroke, headache, epilepsy, and others) (2), emphasizing the importance of suspecting moyamoya in children presenting with multiple neurological manifestations.

Interestingly, a higher number of children in our cohort, compared with previous series, received diagnosis of moyamoya in the context of a genetic condition, such as sickle cell disease or neurofibromatosis type 1, despite being asymptomatic. This is possibly due to the changing epidemiology and the increased adherence to disease-specific guidelines. In particular, national guidelines for pediatric sickle cell disease recommend screening transcranial doppler from age 2 and magnetic resonance imaging from age 6 years. They also encourage considering hematopoietic stem cell transplantation in case of cerebral vasculopathy (41, 42).

Surgery was extensively used in our patients, with indirect revascularization being the preferred technique according to patients' age in our cohort, and similar adverse events rate at first intervention to other international data (43–45). Interestingly, the frequency of surgery significantly increased in our cohort (surgical interventions performed from 1995 to 2017) compared with a previous Italian cohort published in 1997 (Supplementary Material S4) (32).

Regarding pharmacological management, a recent study demonstrated that antiplatelet therapy does not seem to prevent recurrent cerebral infarction, which recognizes a hemodynamic (vs. embolic) pathogenesis (46).

Nevertheless, according to different treatment recommendations for pediatric stroke, aspirin is suggested as initial therapy for children with acute ischemic stroke secondary to moyamoya, over no treatment (Grade 2C) (11) and may be considered in individuals with moyamoya after revascularization surgery or in asymptomatic individuals for whom surgery is not anticipated (Class IIb, Level of Evidence C) (10), and overall it is widely used as a long-term treatment in pediatric asymptomatic or symptomatic ischemic type MMD or MMS, especially by non-Asian physicians (47, 48). Accordingly, aspirin use was widespread in our cohort, exception made for asymptomatic patients (incMMS) (Table 2). Although this latter subgroup had a favorable course even with less frequent aspirin use, data are too limited to draw definite conclusions.

Both MMD and MMS are significantly associated with risk of poor neurological outcome (8, 45, 49, 50), for several reasons. First, it is striking that the risk of stroke remains also after diagnosis: in our cohort, 20% of patients experienced one or more strokes after diagnosis, including not only patients with stroke recurrence (33% of patients who suffered a first stroke before diagnosis) but also patients who presented their first stroke after diagnosis (5% of the entire study population). Arteriopathy is a known risk factor for stroke recurrence (6, 7, 35), and moyamoya vasculopathy could predispose to cerebrovascular events also in patients who presented at diagnosis with a different “disease type”.

Surgical treatment did not completely abate the risk of stroke during the disease course: among our subgroup of surgically treated patients, 11% of patients had stroke after surgery. This figure is similar to other published data (7–13%) (15, 42), but higher than other pediatric cohorts (0–3%) (22, 51, 52) (possible heterogeneity in study design and populations).

Moyamoya is also associated with variable degrees of intellectual, motor, and functional impairment at follow-up (53). The proportion of children with mRS > 2 (conventionally considered an estimate of needed support in daily life) is about 10–35% in the literature (22, 26, 36, 51, 52). In the recent UK cohort (15), 37.6% of children had mRS > 2, and 52% of them had special needs at final follow-up. Similarly, 25% of our patients had mRS > 2, and 31% of them had intellectual impairment. Early-onset disease was the factor with the most stringent significant relationship with poor outcome, as previously reported (43). As expected, also the occurrence of stroke before diagnosis was significantly correlated with cognitive and motor outcome, but the actual contribution of stroke could not be separated from the effect of age: children with early-onset disease had higher probability of suffering from stroke before the diagnosis in our cohort. A larger sample might allow separating the effect of these two variables (age at disease onset and stroke before diagnosis). Nevertheless, it is also likely that genetic factors, still partially unknown, could determine both an early-onset and a more severe disease (54, 55).

Posterior circulation involvement, although not statistically associated with outcomes, seemed to be associated with a higher rate of cerebrovascular events during the entire disease course, as reported in other cohorts (56).

Statistical comparison between symptomatic MMD (symtpMMD) and MMS (symptMMS) populations showed only one statistically significant difference about functional disability at follow-up (Table 2). Nevertheless, other parameters indicate a trend toward a more severe disease in patients with MMS (higher proportion of patients experiencing stroke after diagnosis and with intellectual disability at follow-up). This could also partially be due to their associated conditions.

Surprisingly, we registered an unexpectedly benign course in the small group of asymptomatic patients (incMMS). In spite of this, international data suggest that moyamoya is a progressive disorder even in initially asymptomatic cases, supporting the rationale of early surgical intervention to minimize morbidity from stroke, according to some authors (57). Larger cohorts are warranted to conclude about this topic.