Population

We conducted a population-based study designed to capture all neonates born at term with perinatal stroke born between 1st January 2007 and 31st December 2017 and cared for in the central Hungarian region, including the capital, Budapest. Findings on neonates with perinatal arterial ischaemic stroke (PAIS) have been published elsewhere.² In the present study, we describe the findings of neonates diagnosed with PHS.

Magnetic resonance imaging (MRI) scans were performed at a single centre: the MR Research Center of Semmelweis University, Budapest, Hungary. We reviewed the imaging studies of 1400 term and near-term neonates (≥36wks) who had a brain MRI performed before 28 days of age and identified 96 infants with evidence of a focal haemorrhagic event. Patient selection is shown in Figure S1. Inclusion criteria were (a) neonates born at not less than 36 weeks up to 28 days of age and (b) a brain MRI confirming the diagnosis of PHS without any other concurrent abnormality.

Like in previous studies,⁷ we also excluded neonates with encephalitis, tumour, non-accidental brain injury, and birth asphyxia from the study. We also excluded neonates with exclusively intraventricular, subarachnoidal, subdural, or epidural haemorrhage without intraparenchymal bleeding, as well as neonates with periventricular haemorrhagic leukomalacia. Infants with congenital syndromes with known neurodevelopmental sequelae or other significant brain lesions including neonates with concomitant PAIS or cerebral sinovenous thrombosis were not addressed in this study.

Ethical approval for the study was obtained from the Hungarian Medical Research Council (19934-4/2018/EKU). Informed written parental consent was obtained to recruit patients into the study and perform the neurodevelopmental assessments.

MRI

As described previously,² brain imaging was performed on 3T Philips Achieva and 3T Philips Ingenia MR scanners (Philips Medical Systems) at the MR Research Center of Semmelweis University, Budapest, Hungary. The scanning protocol included diffusion-weighted imaging, conventional T1- and T2-weighted imaging, T2*-weighted imaging, and susceptibility-weighted imaging. In selected cases magnetic resonance angiography and/or magnetic resonance spectroscopy was added, as appropriate. Radiologists trained and experienced in neonatal brain MRI evaluated the MRI. Classification was performed according to the affected brain compartment as described by Govaert et al.⁷ (Figure 1), except that exclusively extra-axial bleeding without an intraparenchymal haemorrhage was not included in this study. We also noted the involvement of the thalamus and/or basal ganglia since previous reports have described an association between involvement of the thalamus and/or basal ganglia and chronic neurodevelopmental outcomes.

As an additional approach, we measured the size of the PHS by drawing a region of interest covering the area judged to have abnormally high signal intensity on T2*-weighted images using the three-dimensional image analysis package of the MRIcron program.⁸ Lesion volumes were expressed as total volume (cubic centimetres) and as a percentage of supratentorial brain volume after excluding ventricular volume. The ratio of PHS to the supratentorial brain volume was further subcategorized as small (<5%), moderate (5%–10%), or large (>10%) on the basis of the relative stroke volume.

Clinical data

Clinical characteristics, maternal and neonatal risk factors, and presenting symptoms of PHS were collected from chart reviews. The collected data are described in detail in Appendix S1.⁹ Bleeding diathesis was investigated in 15 neonates in the present study. Echocardiography was used as indicated to confirm/rule out congenital heart disease (CHD).

Neurodevelopmental outcome

In addition to routine follow-up visits between the age of 1 year 6 months to 12 years, we also prospectively performed a systematic neurodevelopmental follow-up assessment of children who were diagnosed with PHS and were available for follow-up in 2018. As described previously,² a child developing according to population norms was defined as one with symptom-free survival. An infant with chronic neurodevelopmental sequelae was documented if they presented with any one or more of the following outcomes: CP, cognitive impairment, behavioural problems, epilepsy, language disorder, visual field defect, or hearing loss. Developmental tests used during the follow-up visits and the classification of CP and epileptic syndromes^10-14 are described in detail in Appendix S2.

Statistical analysis

Descriptive statistics were expressed as frequencies and percentages in the population studied. Mean and standard deviation (SD) or median and interquartile range (IQR) were determined as appropriate. Univariate logistic regression models were fitted to describe relationships between clinical predictors of interest and neurodevelopmental outcome. A multivariable logistic regression model was fitted to ascertain the effect of brain territorial involvement of PHS on the basis of the notion that a patient would probably have a chronic neurodevelopmental outcome while controlling for other clinically relevant factors identified by the univariate analysis, and for emergency Caesarean section on the basis of previous studies.^{5, 15} Factors found to be clinically relevant by the univariate analysis included CHD and clinical seizure on admission. Models were checked according to the Hosmer and Lemeshow's goodness of fit tests.¹⁶ Odds ratios (OR) with 95% confidence intervals (CI) were calculated for each clinical variable. All statistical tests were two-sided, where p-values of <0.05 were considered to indicate statistical significance. Statistical analyses were performed using IBM SPSS Statistics (version 25, IBM Corp.).

Patient characteristics and general findings

A total of 294 000 live births of ≥36 weeks' gestation were registered in the central Hungarian region over the 11-year study period. Acute PHS was diagnosed in 55 neonates, yielding a disease incidence of one per 5300 live births (prevalence 1.87, 95% CI: 1.38–2.36, per 10 000 live births for newborn infants of ≥36 weeks' gestation).

The most common presenting symptom of PHS was seizure activity, occurring in 61% of patients, at a median age of 2 days (IQR: 1–3). The rate of respiratory distress was also high, detected in 39% of the neonates.

Among the 55 infants with PHS, 36 (65%) had a maternal and/or a neonatal risk factor that might have contributed to the development of the PHS. Underlying aetiologies among others included alloimmune thrombocytopenia (n=3), CHD (n=5), and cerebral vascular abnormalities (n=6) including arteriovenous malformation (n=4), cavernoma (n=1), and venous angioma (n=1). In 10 patients (18%) more than one possible primary etiological factor was present, suggesting that the aetiology of PHS might have been multifactorial. Clinical characteristics of the study population are summarized in Table 1.

Affected brain regions

Diagnostic brain MRI was performed at a median age of 5 (IQR: 3–7) days. PHS was typically unifocal (80%) and unilateral (89%). Stroke occurred in the frontal (24%), temporal (24%), parietal (18%), and occipital lobes (18%), as well as in the basal ganglia and/or thalamus (22%). Eight patients (15%) developed hydrocephalus. The median stroke volume was 10.6cm³ (IQR: 2.2–22.8) with a median stroke percentage of the supratentorial brain volume of 2.7% (IQR: 0.7–6.8). Most of the strokes were subcategorized as small (n=34, 62%), with the minority as moderate (n=15, 27%) or large (n=6%–11%). Detailed frequencies of stroke subtypes are shown in Table 2.

Neurodevelopmental outcome

Long-term neurodevelopmental outcome data were available in 50 (91%) of the 55 children with the last follow-up visit occurring at a median age of 60 months (IQR: 35–88). Fifteen children were assessed by the Brunet–Lézine test (n=13) or the Bayley Scales of Infant Development, Second Edition (n=2) at a median age of 24 months (IQR: 19–34), while 30 children were evaluated by the Stanford–Binet Intelligence Scales, Fifth Edition at a median age of 74 months (IQR: 60–108). Finally, five children were followed by a paediatric neurologist without using a formal neurodevelopmental assessment owing to their severe neurodevelopmental delay. Of the five infants lost to follow-up, one died from complications of stroke and CHD in the neonatal period, two children with neurodevelopmental findings meeting population norms between 12 months and 18 months of age were lost to further follow-up thereafter, and two patients were lost to follow-up after the neonatal period. All infants lost to follow-up had unifocal, unilateral territorial involvement on MRI not involving the basal ganglia and/or thalami.

Twenty of the 50 infants (40%) with longitudinal follow-up had been developing according to the population norms. In individuals with chronic neurodevelopmental sequelae, the most common outcomes were behavioural problems (24%), epilepsy (22%), and language disorders (18%). As out of the 11 patients (22%) initially diagnosed with epilepsy two subsequently remained seizure free for more than a year without antiepileptic treatment, the rate of active epilepsy was 18% among the individuals in the long-term. Unexpectedly, CP was not as frequent a finding as in neonates with PAIS² since it was only documented in eight patients (16%); yet in those, CP was more severe as 75% of the patients (six out of eight) presented with tetraparesis. Children developed visual field defect (8%) and hearing loss (6%) infrequently. Thirteen individuals (26%) were recorded with more than one type of developmental sequelae.

Risk factors associated with adverse neurodevelopmental outcome

None of the neonates with frontal lobe PHS developed later CP, cognitive impairment, or epilepsy during the follow-up period, and the overall rate of impaired neurodevelopmental outcome was also less (38% vs 50–66% in other lobes).

Univariate analysis of stroke territory subtypes and their relation to neurodevelopmental outcome domains revealed several associations between brain territorial involvement and neurodevelopmental outcome domains (Table S1). Among the clinical risk factors, CHD conferred significantly higher odds for cognitive impairment and clinical seizure on admission was associated with developing epilepsy beyond the neonatal period (Table S1).

Multivariable logistic regression analysis revealed that parietal lobe PHS is a significant independent predictor of CP (OR 6.7; 95% CI 1.1–41.4) and cognitive deficit (OR 23.6; 95% CI 2.9–194.9) while infants with basal ganglia and/or thalamus involvement had seven times higher odds (95% CI 1.3–37.7) for epilepsy. Strokes involving multiple lobes were significant independent predictors of impairment in several neurodevelopmental domains including CP (OR 6.4; 95% CI 1.0–40.5), epilepsy (OR 10.8; 95% CI 1.8–64.3), and the need for ventriculoperitoneal shunt placement (OR 5.7; 95% CI 1.0–30.7; Tables 3–5).

When a logistic regression model was fitted, there was no relationship between infarct size category and neurodevelopmental outcomes. This finding suggests that the location of the stroke is more important in predicting the neurodevelopmental outcome than the volume of the lesion itself.

Finally, among the clinical risk factors, clinical seizure on admission was associated with the risk of later developing epilepsy (OR 8.8; 95% CI 1.0–81.7). Indeed, neonates with clinical seizure on admission developed epilepsy more frequently than those without clinical seizure on admission (30% vs 5%, p=0.03).