Influence of isolated low‐grade intracranial haemorrhages on the neurodevelopmental outcome of infants born very low birthweight

To determine whether isolated low‐grade germinal matrix‐intraventricular haemorrhages (LG‐GMH‐IVH) and low‐grade punctate cerebellar haemorrhages (LG‐CBH) contribute to the neurodevelopment of infants born preterm with very low birthweight (VLBW).

The epidemiology of preterm birth is changing, with more infants surviving at lower gestational ages, carrying a high vulnerability of germinal matrices (intraventricular germinal subependymal matrix and subpial external granular layer; Figure S1), with the consequent risk of developing germinal matrix-intraventricular haemorrhage (GMH-IVH) or cerebellar haemorrhage (CBH) or both. 1,2 In the past, severe brain lesions, such as IVH grade 4, massive CBH, or the concomitance of the two, made it difficult to interpret the role of each insult on the developing brains of infants born preterm and therefore on their neurological outcome. 3,4 In recent decades, the severity of both IVH and CBH has decreased because of improvements in neonatal care, with more infants who are born preterm surviving with less severe haemorrhages. In parallel, the advent of brain magnetic resonance imaging (MRI) has improved neurological screening. In particular, susceptibility-weighted imaging (SWI), a sequence based on high-resolution, three-dimensional, fully velocity-compensated gradient-echo using both magnitude and phase images, allows detection of even small-sized haemorrhages that may be missed by cranial ultrasound. 1,[5][6][7] Outcomes of children born extremely preterm carrying either GMH-IVH or CBH can vary from severe neurological sequelae (such as cerebral palsy associated with profound intellectual disability and neurosensory high impairment) in major haemorrhages [8][9][10][11] to more complex and subtle consequences in minor haemorrhages (mild motor impairment, mild intellectual disabilities, or learning issues). [11][12][13][14][15][16][17] However, the neurological outcomes of children born preterm carrying mild intracranial haemorrhages, compared with those without lesions, remain controversial. 15 This controversy may be explained by the prevalence of even large cohorts of children born extremely preterm carrying mild haemorrhagic lesions studied with less accurate diagnostic tools (cranial ultrasound or brain MRI without SWI), which can miss focal isolate bleedings, creating a bias in the comparison group. 12,15,17,18 Effects of mild intracranial bleedings on infants born preterm should exist. Recently it has been demonstrated that mild IVH detected by brain MRI could cause microstructural changes, with selective vulnerability of white matter regions, based on varying gestational age and adverse neurodevelopmental outcome at 2 years corrected age. 14 Also, a study of CBH demonstrated that the presence of focal cerebellar bleeding seems to carry a high, size-dependent 17 risk for adverse motor and behavioural outcomes at 4 years 6 months of age. 11 This study aimed to investigate the impact of isolated lowgrade GMH-IVH (LG-GMH-IVH) and isolated low-grade punctate CBH (LG-CBH), confirmed or detected at SWI, on the neurodevelopment of infants born extremely preterm.

M ET HOD
The local ethical committee (Comitato Etico Regione Liguria, Genoa, Italy) approved this prospective observational study (number 39354), conducted in accordance with the Declaration of Helsinki. Each infant's parent provided informed written consent.
We prospectively enrolled all infants admitted to the neonatal intensive care unit of Gaslini Children's Hospital between January 2012 and July 2017 weighing less than 1500 g at birth (very low birthweight [VLBW]) who underwent a routine brain scan protocol including serial cranial ultrasound and brain MRI with SWI at term-corrected age (between 39 +0 and 41 +6 weeks postmenstrual age).

Imaging data
Serial cranial ultrasound examinations were performed by two ultrasonographers according to the routine protocol in our neonatal intensive care unit, which includes cranial ultrasound studies at days 1, 2, 3, and 7 after birth and subsequently weekly until term-corrected age. All cranial ultrasound examinations were performed through the anterior and the mastoid fontanel using an 8Mhz convex probe (Aloka Prosound Alpha 7, Aloka Co., Ltd., Tokyo, Japan). All ultrasound images of standard coronal and sagittal views were collected in an electronic database.
GMH-IVH was classified according to Inder et al., 19 while CBH was classified according to previous literature. 6,10 MRI studies including SWI sequence were performed on a 1.5T scanner (InteraAchieva 2.6; Philips, Best, the Netherlands) using a dedicated neonatal head array coil. 14 The parameters for SWI were as follows: repetition time/echo time 33 ms/48 ms; flip angle 15°; field of view 120 mm × 120 mm; matrix size 152 × 137; total acquisition time 2 minutes 57 seconds.
Foci of signal loss detected by SWI inside the germinal matrix and/or along the ependyma of lateral ventricles and/ or in the caudothalamic notch, without continuity suggestive of veins, were interpreted as GMH-IVH; single or multiple small (≤4 mm) foci of signal loss on SWI within the cerebellum 6,10 were classified as LG-CBH.
Cranial ultrasound and MRI findings were blindly reviewed by two ultrasonographers (AP and LAR) and by two

What this paper adds
• Low-grade haemorrhages seem to have an independent role in the adverse neurodevelopment of infants born extremely preterm. • Infants born preterm with low-grade germinal matrix-intraventricular haemorrhage (LG-GMH-IVH) are indicated for physical therapy at a later age than those with low-grade punctate cerebellar haemorrhage (LG-CBH). • Compared with LG-CBH, LG-GMH-IVH is associated with a greater extent of attention problems at 3 years corrected age.
neuroradiologists blinded to cranial ultrasound results (DT and MS) respectively. Consensus was sought in cases of disagreement ( Figure 1). Isolated LG-GMH-IVHs were defined as subependymal or intraventricular hemosiderin depositions detected by MRI at term-corrected age in patients with normal serial cranial ultrasound examinations or with a cranial ultrasound diagnosis of grade 1 or 2 GMH-IVH, in the absence of other abnormalities. Similarly, isolated LG-CBHs were defined as LG-CBH in the absence of any other brain abnormalities detected by MRI (Figure 1).
According to imaging findings, infants were divided into three groups: isolated LG-GMH-IVH, isolated LG-CBH, and no-lesion (absence of detectable congenital or pretermbirth-related brain abnormalities).

Clinical characteristics of included infants
Prenatal, intrapartum, and postnatal characteristics of included infants were collected according to known risk factors for GMH-IVH 19,20 and CBH, 11 whose definition was reported in previous studies. [21][22][23]

Parental characteristics
Parental medical history of neurological, psychiatric, immunological, and malformative diseases was collected as it could have been a potential risk associated with neurodevelopmental disorders. 24 Other factors that could affect neurodevelopmental outcome were considered such as F I G U R E 1 Examples of isolated low-grade germinal matrix-intraventricular haemorrhage (LG-GMH-IVH) and isolated low-grade punctate cerebellar haemorrhage (LG-CBH) detected by susceptibility-weighted imaging (SWI) after positive or negative cranial ultrasound. Patient 1: LG-GMH-IVH detected by (a, b) cranial ultrasound and confirmed by (c) SWI at term-corrected age. Patient 2: LG-GMH-IVH missed by (d, e) cranial ultrasound and detected by (f) SWI at term-corrected age. Patient 3: LG-CBH missed by (g, h) cranial ultrasound through mastoid fontanel and detected by (i) SWI at term-corrected age. the educational levels of both parents (years of education) and potential fragility factors (i.e. adoption or single-parent family). 25

Neurological and neurodevelopmental outcome
Two child neurologists with at least 10 years of experience performed annual neurodevelopmental evaluations at 1 year, 2 years, and 3 years corrected age, blinded to cranial ultrasound and MRI results.
Developmental assessment was performed using the Griffiths Mental Development Scales, Extended and Revised version (GMDS-ER). This tool consists of two different protocols: 26,27 one for children aged 0 to 2 years (performed in this cohort at 1 year and 2 years corrected age); the other for the age group 2 to 8 years (performed in this cohort at 3 years corrected age). Raw numbers achieved on the test were converted to standardized development scores, covering a global developmental score (total) and five subdomains: locomotor, personal-social, hearingspeech, eye-hand coordination, and performance. A sixth subdomain (practical reasoning), which is the precursor of executive functions and learning skills, was tested only at 3 years, as foreseen by the 2 to 8 years GMDS-ER protocol. 27 We decided to keep the annual time points of evaluation until 3 years corrected age, to improve the sensibility of the test in predicting adverse outcomes, as suggested by experts in the field. 28 The age of development of motor and linguistic skills was also recorded (developmental stages) as well as the indication to rehabilitation programme (physical and speech therapy), which was assigned by clinicians on the basis of their clinical examination and national guidelines, owing to the observational nature of this study.
Clinical diagnosis of cerebral palsy, severe vision or hearing impairment, and developmental or language delay were also reported.

Behavioural outcome
Behavioural variables 17,29,30 were recorded at 3 years corrected age on the basis of clinical evaluation by the child neurologists, according to the International Classification of Diseases, 10th Revision. 31 The following data were recorded: self-injurious behaviours, attention problems, hyperactivity, oppositional behaviour, and relational/behavioural atypias.

Statistical analysis
We used the Statistical Package for the Social Sciences for Windows (SPSS v.23 Inc., Chicago, IL, USA) for statistical analysis. Statistical significance was set at p < 0.05.
Descriptive statistics were generated for the whole cohort; data were expressed as mean and standard deviation for continuous variables, and absolute and relative frequencies for categorical variables.
Demographic and clinical characteristics were compared using Fisher's exact test or a χ 2 test, and a Mann-Whitney U test for categorical and continuous variables.
Univariable analysis determined the potential risk factors, which were significantly associated with unsatisfactory scores in the GMDS-ER total development scores (<85) at 1 year, 2 years, and 3 years corrected age. Logistic regression analyses were used for each variable, and odds ratios (ORs) were calculated with 95% confidence intervals (CIs). The absence of exposure to the factor/variable less likely to be associated with the risk was used as the reference for each analysis.
Multivariable analysis was then performed, and only the variables that proved to be statistically significant (p < 0.05) or with p < 0.08 in the univariable analysis along with other factors we considered potentially influent on the final neurodevelopmental outcome were assessed with the GMDS-ER. The directed acyclic graph methodology served this purpose 32 ( Figure S2). The model that best fitted was based on the backward stepwise selection procedures, and each variable was removed if it did not contribute significantly. In the final model, p < 0.05 was considered statistically significant, and all p-values were based on two-tailed tests. Figure S3 shows the initial sample (420 infants with VLBW admitted to our neonatal intensive care unit between January 2012 and July 2017) and the selection of the study cohort. Table S1 summarizes basic demographics of the infants lost at follow-up.

Study cohort
The overall incidence among all infants with VLBW admitted to our unit for all IVH of any grade was 25.5% (107 out of 420) and 13.5% (57 out of 420) for all CBH of any severity. Among the surviving infants, the incidence of IVH of any grade was 20.4% (75 out of 367), and CBH of any severity was 9.5% (35 out of 367).
Consensus was reached about the presence and grade of GMH-IVH and LG-CBH in all cases, both between ultrasonographers and between neuroradiologists. The sensitivity of cranial ultrasound in detecting GMH-IVH was 62% (21 out of 34), while none of the LG-CBH diagnosed by SWI was identified by cranial ultrasound, despite the use of mastoid fontanel.
Two-hundred and forty infants were included in the cohort. Among the included infants who completed the follow-up, 34 had isolated LG-GMH-IVH, 17 had isolated LG-CBH, and 189 had no brain lesions on MRI. Table 1 shows comparisons of parental, prenatal, intrapartum, and postnatal characteristics between the following groups: (1) LG-GMH-IVH versus no-lesion; (2) LG-CBH versus no-lesion; and (3) LG-GMH-IVH versus LG-CBH.
At 3 years corrected age, no children were diagnosed with cerebral palsy. Two infants were diagnosed with severe hearing impairment (requiring hearing aids): 1 out of 189 among the no-lesion group, 1 out of 17 among the LG-CBH group, and none among the LG-GMH-IVH group (p = 0.156). There were five children with severe visual impairment due to retinopathy of prematurity: 2 out of 189 in the no-lesion group, 1 out of 34 in the LG-GMH-IVH group, and 2 out of 17 in the LG-CBH group (p = 0.167).  Table S2 summarize the neurodevelopmental and behavioural outcomes at 1 year, 2 years, and 3 years corrected age between groups: LG-GMH-IVH versus nolesion, LG-CBH versus no-lesion, and LG-GMH-IVH versus LG-CBH.

Neurodevelopmental and behavioural outcomes
On the GMDS-ER scores, the LG-GMH-IVH and LG-CBH groups had worse neurodevelopmental performance at 1 year, 2 years, and 3 years corrected age than the no-lesion group, with the greatest impairment seen in hearing-speech development scores at 3 years corrected age ( Figure 2c and Table S2).
The LG-CBH group had lower GMDS-ER scores than the LG-GMH-IVH group at 1 year corrected age in total and locomotor development scores (Figure 2a), and at 2 years corrected age in total, personal-social, and hearing-speech development scores (Figure 2b). These differences between the two groups disappeared at 3 years ( Figure 2c).
Regarding developmental milestones (Figure 2d), the LG-CBH group achieved autonomous walking and spoke two-word sentences at significantly older ages than the no-lesion group. They also showed a significantly higher need for physical and speech therapy. At 3 years corrected age, developmental delay was more frequently found in the LG-CBH group than in the no-lesion and LG-GMH-IVH groups.
The LG-GMH-IVH group had better motor outcomes in achieving autonomous walking than the LG-CBH group, with a lower frequency of indication to physical therapy (20.6% vs 70.6%).
At 3 years corrected age, language delay was equally distributed among three groups. Even though age at first twoword sentences was achieved significantly later for both the LG-GMH-IVH and LG-CBH groups compared with the no-lesion group, the indication for speech therapy was given more frequently among the LG-CBH group than both the no-lesion and LG-GMH-IVH groups.
Regarding behavioural features at 3 years corrected age (Figure 2e), children with LG-GMH-IVH presented more frequently with hyperactivity, relational, and behavioural atypias than the no-lesion group (Figure 2e). Attention problems were more frequent in the LG-GMH-IVH group compared not only with the no-lesion group but also with the LG-CBH group.
We therefore performed the univariable analysis at 1 year, 2 years, and 3 years corrected age (Table 2) to evaluate potential prenatal, intrapartum, and postnatal factors that could have affected the presence of satisfactory or unsatisfactory scores on the GMDS-ER at the difference stages of follow-up. At 1 year corrected age, birth by Caesarean section, maternal years of education, a lower paternal age, and a higher level of education served as protective factors. At 2 years corrected age, while protective factors remain important, it emerged that the presence of LG-CBH had an adverse impact on the neurodevelopmental outcome as measured by GMDS-ER total scores. At 3 years of age, antecedent protective factors remained (except for Apgar score at the first minute and father's age at the infant's birth), and higher birthweight seemed related to a satisfactory outcome, while the presence of isolated low-grade haemorrhages (both LG-GMH-IVH and LG-CBH) as well as the presence of patent ductus arteriosus seemed associated with an unsatisfactory GMDS-ER score. In addition, having a non-native speaking parent was associated with an unsatisfactory GMDS-ER score.

The adverse impact of intracranial low-grade haemorrhages on VLBW neurodevelopmental outcomes
On the final multivariable model, we evaluated, at each year, the impact of these significant (p < 0.05) or with p < 0.08 protective or negative factors on determining satisfactory or unsatisfactory neurodevelopmental outcome. The model was adjusted for male sex, gestational age, birthweight, intrauterine growth restriction (expressed as foetal size measures less than the third centiles associated with other abnormal functional parameters seen at prenatal ultrasound 33 ), and parental years of education. The model was also adjusted for some variables suggested by the creation of the directed acyclic graph ( Figure S2) (such as bronchopulmonary dysplasia and mother's age at the infant's birth), considered potentially influent in determining adverse neurodevelopmental outcome defined as an unsatisfactory GMDS-ER score. Table 3 shows the main results, while in Table S3

LG-CBH vs no-lesion p a
LG-

DISCUS SION
In this prospective observational study, we found that the presence of isolated minimal low-grade haemorrhages (both intraventricular and cerebellar), identified by SWI, contributed to adverse neurodevelopmental outcome of infants born extremely preterm up to 3 years corrected age. Other factors potentially influencing the neurodevelopmental outcome of our cohort were the presence of intrauterine growth restriction, male sex, and level of paternal education.
Although the literature provides increasing evidence about the adverse effects of mild intracranial haemorrhages (both IVH and CBH), 8,11,12,14,16 the role of these minor lesions in contributing to poor neurodevelopmental outcomes remains to be further investigated. For instance, Wy et al. published their study of 462 children born extremely

LG-GMH-IVH vs LG-CBH p a
T A B L E 2 Pairwise comparisons between groups according to the satisfactory and unsatisfactory (score < 85) GMDS-ER total developmental scores at 1 year, 2 years, and 3 years corrected age.   preterm under 2500 g of birthweight who were tested with cranial ultrasound; they found that LG-GMH-IVH does not influence behavioural and cognitive outcomes at either 36 months or at 8 or 18 years. 18 A potential bias on the results of this and other studies on minor intracranial bleedings typical of preterm birth is the lack of accurate methodology in detecting them. 5,6 Studies conducted on large cohorts using cranial ultrasound 3,12,18,34,35 or others conducted on smaller groups using MRI but without SWI 16,17,36 present the high possibility of missing the diagnosis of intracranial haemorrhages in the comparison samples. To increase the sensibility of our study, we excluded all those of VLBW with any associated brain lesion, even with minor entities such as GMH-IVH associated with LG-CBH or to punctate white matter lesions (frequently associated with GMH-IVH 37 ). We therefore concentrated on defining whether the presence of isolated low-grade haemorrhages (either LG-GMH-IVH or LG-CBH) affects the neurodevelopmental outcome of infants born preterm. Moreover, many cohorts include infants born at term with a birthweight greater than 1500 g and of greater mean gestational ages, with a lower risk of bleedings coming from the germinal matrices, but with different risk for the preterm developing brain. 2 The adoption of SWI sequences on brain MRI at termcorrected age is an extremely precise diagnostic tool for detecting low-grade haemorrhage; it allows targeting in future surveillance programmes which may include early rehabilitative approaches ideally tapered to diagnoses of specific brain lesions. Indeed, the SWI sequence also has the advantage of being extremely easy to interpret in any neuroradiology unit.
The possible pathophysiology of adverse neurodevelopment of LG-GMH-IVH at 2 years corrected age in children carrying isolated LG-GMH-IVH can be related to white matter microstructural changes, previously demonstrated with diffusion tensor imaging. 14 Inflammatory processes may affect the maturational-dependent vulnerability of oligodendroglial precursors in the white matter adjacent to the blood remaining on the ependymal surface or the possible direct transependymal passage of blood in infants with lower gestational age, as shown in experimental models. 14,38 A previous study has suggested that suboptimal brain development can occur after early reduction of blood perfusion due to LG-GMH-IVH. This has been observed soon after LG-GMH-IVH by near-infrared spectroscopy, 39,40 and confirmed at term-corrected age by arterial spin labelling MRI, 41 which identifies the reduced brain perfusion even weeks after the occurrence of bleeding. On the other hand, the pathophysiological effects of cerebellar lesions typical of preterm birth (and, thereafter, their consequences on neurodevelopment of infants born extremely preterm) have been less investigated than those triggered by IVH. This is also because CBH was believed to be less frequent since the mastoid fontanel for cranial ultrasound or MRI was routinely adopted in neonatal neurology care. Volpe has masterfully detailed several sequelae following a cerebellar insult, encompassing cerebellar disorders secondary to a haemorrhagic insult or primarily related to preterm birth. 4,42 It is fascinating to note the behaviour of cerebral and cerebellar neuronal migration and to speculate on the different effects of the two types of haemorrhage. Cerebral neuronal migration centrifugally reaches the cortex but is limited to weeks 20 to 22 of gestation. On the other hand, cerebellar migration is centripetal but lasts up to 6 to 8 months postnatally. The subependymal germinal matrix develops at around the 8th week of gestation, reaches its maximum volume at 25 weeks, then involutes until its disappearance by the end of the 34th to 36th week. 43 The cerebellar external granular layer appears at around the 9th week and persists until the end of the first postnatal year, still producing neurons that migrate internally into the inner granular layer for months after the birth of an infant born at term ( Figure S4). 1,44 Hence, an insult at the external granular layer, which continues developing after birth, could represent a longer-lasting change in the impairment of cerebellar neuronal organization. 45 Studies of mouse models reinforce the hypothesis that even preterm birth alone can impair cerebellar development, thus suggesting a worse possibility in the case of cerebellar haemorrhage. 46,47 We found an adverse impact of both low-grade lesion types on neurodevelopmental outcome of extremely-lowbirthweight children compared with those not carrying brain lesions, with a bigger impairment in the LG-CBH group, which also had lower total GMDS-ER scores than the LG-GMH-IVH group at 1 year and 2 years.
At 3 years corrected age, the hearing-speech subdomain of the GMDS-ER was the most affected among the LG-CBH and LG-GMH-IVH groups (Figure 2c and Table S2), and language delay was equally distributed among the three groups ( Figure 2d and Table S2). Language delay in infants born extremely preterm is a common finding among extremelylow-birthweight children. The most common risk factors associated with this outcome are gestational age, postnatal steroids (and the development of bronchopulmonary dysplasia), intrauterine growth restriction, and brain injuries, particularly those involving language centres. 48 Observing our sample, we can speculate that the presence of a minimal isolated bleeding (such as LG-GMH-IVH or LG-CBH) may have an adverse impact on language development of extremely-low-birthweight children.
Second, at 3 years, the difference in GMDS-ER scores between the LG-GMH-IVH and LG-CBH groups disappeared (Figure 2a,b). The presence of lower GMDS-ER scores at 2 years in the LG-CBH group could have contributed to the relatively earlier clinical indication to rehabilitation and the greater number of children born extremely preterm treated in the LG-CBH group compared with the LG-GMH-IVH group (with both physical and speech therapy) (Figure 2d). We may speculate that this could be an encouraging signal for the effectiveness of early rehabilitation in changing neurodevelopmental trajectories. 49,50 Future individualized rehabilitation may consider the presence of minor brain lesions detected with SWI. 49,51 In our study, we considered several known prenatal, intrapartum, and postnatal risk factors associated with the development of LG-GMH-IVH and LG-CBH occurring in the first days of life 11,19,23,25 or associated with potential adverse neurodevelopmental outcome (Table 1 and Figure S2). The final statistical multivariable model of our study showed that experiencing intrauterine growth restriction has an independent contribution to adverse neurodevelopmental outcome at 3 years corrected age (Table 3); this has been already highlighted in the literature, especially in association with small birthweight (under the third centile). 52,53 Also, male sex was found to have a moderate effect on adverse neurodevelopmental outcome of those with extremely low birthweight in our cohort. The role of male sexual hormones in predicting unfavourable neurological and behavioural outcomes also from an epigenetic point of view (e.g. by regulating the cortical growth and scaling, and mediating the inflammatory responses) has been described. 54 Understanding the mechanisms that guide sexual differentiation/dimorphism in predicting epigenetic changes and brain development could be useful for very early identification of children at risk.
We also analysed the effect of parental education in predicting the developmental outcome of the cohorts studied; we observed that the LG-GMH-IVH group had higher educational attainment of both parents (Table 1). We are uncertain of this result: the role of parental factors (both maternal and paternal) in predicting outcome of infants with extremely low birthweight has only been partly investigated and needs more research. 24 The generalizability of our study is limited by the low sample number. However, our initial cohort seems representative (a total incidence of 25.5% for IVH and 13.5% for CBH), 11,15 and isolated low-grade haemorrhages are very rare (as mentioned, they can easily escape diagnosis with cranial ultrasound or anatomical brain MRI sequences). Considering that this is the first study, to our knowledge, to adopt the SWI technique on brain MRI at term-corrected age in a cohort only of infants of VLBW carrying exclusively isolated low-grade haemorrhages, and presenting a 3-year follow-up, 6,8,10,12,14,15 it should be considered as a pilot investigation; we did not perform any power calculation analysis. Another issue derives from the possible lack of independence of data about parental and prenatal variables of infants with VLBW from multiple pregnancies. We decided to keep the data taken from these infants, as being a twin from a multiple gestation is a risk factor for neonatal and neurological/neurodevelopmental outcome (including IVH and other lesions associated with preterm birth). 55,56 For this reason, and because multiple pregnancies have a similar distribution across studied samples, we decided to include them.
Moreover, given the large number of prenatal, intrapartum, and neonatal risk factors that can affect neurodevelopmental outcome of extremely-low-birthweight infants, final assumptions may have been also affected by a multiple comparison bias, but the strength of the p-values in the final multivariable regression model seems reliable.
We believe that a more detailed assessment of behavioural features at later ages on larger cohorts of infants with VLBW, studied with SWI, and carrying low-grade intracranial haemorrhages could be useful for understanding further the complex contribution of each specific brain lesion, even minor ones, to the neurodevelopmental outcomes of these infants.

AC K NOW L E D G M E N T S
Open Access Funding provided by Universita degli Studi di Genova within the CRUI-CARE Agreement. Work supported by #NEXTGENERATIONEU (NGEU) and funded by the Ministry of University and Research (MUR), National Recovery and Resilience Plan (NRRP), project MNESYS (PE0000006) -A Multiscale integrated approach to the study of the nervous system in health and disease (DN. 1553 11.10.2022). We acknowledge Eu-Brain, which spent time in understanding the nature and the problem of the developing brain of infants born preterm, the families who participated in the study, the nurses and technical personnel working at the neonatology unit, and at the Child Neuropsychiatry Unit of the IRCCS Gaslini Children's Hospital.

SU PP ORT I NG I N FOR M AT ION
The following additional material may be found online: Figure S1: Representation of foetal germinal matrices at 27 weeks of gestation. Figure S2: Direct acyclic graph of study variables. Figure S3: Initial study cohort. Figure S4: Foetal appearance of germinal matrices and neurodevelopment trajectories in an early foetal brain at 9 weeks of gestation. Table S1: Comparisons of demographics and risk factors between VLBW infants included in the final sample and those who were not included in the final sample because they were lost at follow-up. Table S2: Developmental and behavioral profile of the samples. Table S3: Final multivariable analysis of risks of adverse clinical outcomes using the Griffiths scales at 1 year, 2 years, and 3 years.