Association between Term Equivalent Brain Magnetic Resonance Imaging and 2-Year Outcomes in Extremely Preterm Infants: A Report from the Preterm Erythropoietin Neuroprotection Trial Cohort

Abstract

Objectives

To compare the term equivalent brain magnetic resonance imaging (MRI) findings between erythropoietin (Epo) treated and placebo control groups in infants 24^0/7-27^6/7 weeks of gestational age and to assess the associations between MRI findings and neurodevelopmental outcomes at 2 years corrected age.

Study design

The association between brain abnormality scores and Bayley Scales of Infant Development, Third Edition at 2 years corrected age was explored in a subset of infants enrolled in the Preterm Erythropoietin Neuroprotection Trial. Potential risk factors for neurodevelopmental outcomes such as treatment assignment, recruitment site, gestational age, inpatient complications, and treatments were examined using generalized estimating equation models.

Results

One hundred ten infants were assigned to Epo and 110 to placebo groups. 27% of MRI scans were rated as normal, and 60%, 10%, and 2% were rated as having mild, moderate, or severe abnormality. Brain abnormality scores did not significantly differ between the treatment groups. Factors that increased the risk of higher brain injury scores included intubation; bronchopulmonary dysplasia; retinopathy of prematurity; opioid, benzodiazepine, or antibiotic treatment >7 days; and periventricular leukomalacia or severe intraventricular hemorrhage diagnosed on cranial ultrasound. Increased global brain abnormality and white matter injury scores at term equivalent were associated with reductions in cognitive, motor, and language abilities at 2 years of corrected age.

Conclusions

Evidence of brain injury on brain MRIs obtained at term equivalent correlated with adverse neurodevelopmental outcomes as assessed by the Bayley Scales of Infant and Toddler Development, Third Edition at 2 years corrected age. Early Epo treatment had no effect on the MRI brain injury scores compared with the placebo group.

Advanced technology and improvements in obstetric and neonatal care have increased the survival of extremely low birth weight (ELBW) infants.^1-3 However, neurodevelopmental impairment including cerebral palsy, cognitive deficits, hearing, or visual deficits still occurs in approximately 40% of infants born before 28 weeks of gestation.⁴ Identification of infants who are most at risk for subsequent neurodevelopmental disability in this population remains a major challenge.

One tool that may assist with this challenge is obtaining magnetic resonance imaging (MRI) of the brain during the neonatal period. Although cranial ultrasound is currently the most convenient and widely used technique in neonatal intensive care units, use of term-equivalent age brain MRI as a biomarker for neurodevelopmental outcomes among ELBW infants has been increasing. Recent studies confirm abnormal MRI findings that are not seen by cranial ultrasounds such as white matter signal abnormalities, loss of volume, cystic abnormalities, enlarged ventricles, thinning of the corpus callosum, delayed myelination, cerebellar hemorrhages, decreased gray matter volume, and delayed cortical gyration.⁵ Multiple studies using qualitative scoring systems for neonatal brain MRI have tried to determine the predictive value of MRI findings on neurodevelopmental outcomes at 2 years of age with inconsistent results.^6-8

The Preterm Erythropoietin Neuroprotection (PENUT) Trial was a phase III, randomized, placebo-controlled, double-blind trial to assess the safety and efficacy of early high dose recombinant human erythropoietin (Epo) treatment for neuroprotection in extremely preterm infants.³ A subset of enrolled subjects at eight prospectively identified study sites at trial initiation underwent brain MRI imaging between 36^0/7 and 36^6/7 weeks of postmenstrual age (PMA) as part of the study protocol.

The objective of the current study is to compare the qualitatively defined brain MRI findings between treatment and placebo control groups at 36 weeks of PMA and to assess the associations between these defined brain abnormalities and neurodevelopmental delays at 22-26 months of corrected age. We hypothesized that Epo treatment would improve brain development at 36 weeks of PMA and that these infants would have different MRI scan findings than the placebo group. We also hypothesized that a combination of structural brain measurements from the brain MRI scans would accurately predict neurodevelopment at 2 years of corrected age (22-26 months of corrected age). We sought to define which of these MRI findings best predict neurodevelopmental outcomes in cognitive, motor, and language scales.

Methods

The PENUT Trial enrolled inborn infants 24^0/7-27^6/7 weeks of gestation between December 2013 and September 2016 at 19 sites. Details of the protocol, institutional review board approvals, consent, enrollment, methods, and primary outcome have been previously described.³ Patients were excluded based on known life-threatening disorders, chromosomal anomalies, disseminated intravascular coagulopathy, twin-twin transfusion, hematocrit >65%, hydrops fetalis, or known congenital infection.³ This study was institutional review board approved at all participating sites and registered with ClinicalTrials.gov (NCT01378273) and the Food and Drug Administration (IND#12656).

Of the participating infants surviving to 36 weeks of PMA, those who had a brain MRI scan at 1 of the 8 designated subsets of PENUT Trial recruitment sites and who were evaluated for neurodevelopmental outcomes at 22-26 months of corrected age were included in this analysis.

Five sites acquired data from Siemens 3T MRI (Siemens) scanners and three acquired data from Philips 3T MRI (Philips) scanners. Acquisition protocols (Table I; available at www.jpeds.com) were standardized across sites using a 3-dimensional geometric phantom provided by the PENUT Trial (Data Spectrum Corporation model MRI/3D-GEO/P) as well as a consistent human scan at the time of site initiation visit. T1 and T2 weighted and gradient images were blindly scored by 2 neuroradiologists independently utilizing the method of Kidokoro et al.⁷ Minor modifications to the scoring rubric were made for 3 domains (myelination delay, gyral maturation, extracerebral space) because of difficulty in discerning differences between rating categories with the included scans (Table II; available at www.jpeds.com). Disagreements in scores were resolved through concurrent consensus review. The Kidokoro scoring rubric classifies global brain abnormality scores of 0-3 as normal, 4-7 as mild abnormality, 8-11 as moderate abnormality, and ≤12 as indicating severe abnormality. Global brain abnormality scoring category cut points were adjusted by 1 point (0-2, 3-6, 7-10, ≥11) to account for modifications in the scoring rubric and prevalence of injury in the 3 domains presented in the original Kidokoro scoring rubric.

Table I.

MRI scanning parameters

T1 3-dimensional (3D) scan parameters for Siemens scanners – example for Siemens Syra 3T scanners, pulse sequence gradient echo, TR/TE 50/9.2 milliseconds, flip angle 30.0 degrees, acquisition matrix 138 × 192 × 120, scan resolution 1 × 1 × 1 mm, number of averages 1, scan orientation sagittal.

T2 scan parameters for Siemens scanner, pulse sequence 3D spin echo, TR/TE 2000/336 milliseconds, flip angle 120 degrees, acquisition matrix 168 × 192 × 125, scan resolution 1.04 × .04 × 1 mm, number averages 1, scan orientation sagittal.

T1 3D scan parameters for Philips scanners – example for Philips 3T Acheiva Scanners, pulse sequence gradient echo, TR/TE 50/9.21 milliseconds, flip angle 30.0 degrees, acquisition matrix, 14 4 × 144 × 190, scan resolution 0.97 × 0.97 × 1.0, averages 1, scan orientation sagittal.

T2 3D scan parameters for Philips scanners × example for Philips 3T Achieva Scanners, pulse sequence, 3D spin echo, TR/TE 2000/334 milliseconds, flip angle 90.0, acquisition matrix 288 × 288 × 130, scan resolution 0.7 × 0.7 × 1.6 mm, averages 1, scan orientation axial.

(Full details can be obtained from the authors if interested).

TE, echo time; TR, repetition time.

Table II.

PENUT MRI scoring rubric adapted from Kidokoro et al

	Score 0	Score 1	Score 2	Score 3	Score 4	Kidokoro scoring range	PENUT scoring range
Cerebral WM						0-17	0-15
Cystic lesions	None	Focal unilateral	Focal bilateral	Extensive unilateral	Extensive bilateral	0-4	0-4
Focal signal abnormality	None	Focal punctate	Extensive punctate	Linear		0-3	0-3
Myelination delay	PLIC and corona radiata	Only PLIC	Minimal – no PLIC			0-2	n/a
Thinning of corpus collosum	None	Partial genu/body <1.3 mm or splenium <2.0 mm	Global genu/body <1.3 mm and splenium <2.0 mm			0-2	0-2
Dilated lateral ventricles	Both sides VD <7.5 mm	One side 7.5 mm ≤ VD <10 mm	Both sides 7.5 mm ≤VD <10 mm or one side VD ≥10 mm	Both sides VD ≥10 mm		0-3	0-3
Volume reduction	cBPW ≥77 mm	77 mm >cBPW ≥72 mm	72 mm >cBPW ≥67 mm	67 mm >cBPW		0-3	0-3
Cortical GM						0-9	0-6
Signal abnormality	None	Focal unilateral	Focal bilateral	Extensive unilateral	Extensive bilateral	0-4	0-4
Gyral maturation	Delay <2 wk	2 ≤delay <4 wk	Delay ≥4 wk			0-2	0 or 2 (</> 4 wk)
Increased extracerebral space	IHD <4 mm	4 mm ≤IHD <5 mm	5 mm ≤IHD <6 mm	IHD ≥6 mm		0-3	n/a
Deep GM						0-7	0-4
Signal abnormality	None	Focal unilateral	Focal bilateral	Extensive unilateral	Extensive bilateral	0-4	0-4
Volume reduction	cDGMA ≥9.5	9.5 >cDGMA ≥8.5	8.5 >cDGMA >7.5	7.5 >cDGMA		0-3	n/a
Cerebellum						0-7	0-7
Signal abnormality	None	Focal unilateral	Focal bilateral	Extensive unilateral	Extensive bilateral	0-4	0-4
Volume reduction	cTCD >50 mm	50 mm >cTCD ≥ 47 mm	47 mm >cTCD ≥ 44 mm	cTCD <44 mm		0-3	0-3
Global brain abnormality score						0-40	0-32

cBPW, corrected biparietal width; cDGMA, corrected deep grey matter area; cTCD, corrected transcerebellar diameter; GM, grey matter; IHD, intrahemispheric distance; PLIC, posterior limb of the internal capsule; VD, ventricular diameter; WM, white matter.

Neurodevelopmental assessments were done at 2 years corrected age by centrally certified examiners using the Bayley Scales of Infant and Toddler Development, 3rd Edition (BSID-III). All examiners were blinded to the patient’s treatment arm, medical history, and imaging results.

Detailed data were collected prospectively by centrally trained research coordinators and included maternal and infant characteristics. Maternal data included maternal education, race, multiple births, delivery complications, acute chorioamnionitis, antenatal steroid administration, and mode of delivery. Infant data included treatment group, gestational age, delayed cord clamping, sex, birth weight, small for gestational age status, Apgar score at 5 minutes, head circumference, baseline Epo levels and complications of prematurity such as sepsis, severe necrotizing enterocolitis, surgical closure of the patent ductus arteriosus, severe bronchopulmonary dysplasia (BPD), and spontaneous intestinal perforation.

Gestational age was determined by the best obstetrical estimate. Severe intraventricular hemorrhage (IVH) was defined as unilateral or bilateral grade III or IV IVH on cranial ultrasound according to the Papile grading system.⁹ Periventricular leukomalacia was defined as periventricular hyperechogenicity or cystic changes on cranial ultrasound. Severe retinopathy of prematurity (ROP) was defined as treatment with laser photocoagulation, cryotherapy, or bevacizumab. Severe sepsis was defined as culture-proven bacterial or fungal sepsis resulting in blood-pressure support or substantive new respiratory support. Severe necrotizing enterocolitis was defined as stage 2b or 3 according to the Modified Bell Staging Criteria. Severe BPD was defined as requiring high flow nasal cannula, nasal continuous positive airway pressure, noninvasive positive pressure ventilation, or invasive mechanical ventilation at 36 weeks of PMA.

Statistical Analyses

A modified intent-to-treat approach was used in all analyses, with all randomized infants who received the first dose of study treatment included in the analyses. For all statistical comparisons between groups, generalized estimating equation (GEE) models with robust SEs and a working independence correlation structure were used to account for inclusion of infants from a multiple gestation.¹⁰ To assess potential enrolment bias, the clinical and demographic characteristics of infants surviving to 36 weeks of PMA that had an eligible MRI scan were compared with infants that similarly survived but were not scanned.

Global and regional brain abnormality scores were compared between randomized treatment groups (placebo vs Epo) and gestational age at birth using a Poisson-family GEE-based Wald test. Poisson GEE models with adjustment for gestational age, treatment, and enrolling hospital were used to compare differences in abnormality scores between baseline demographic and clinical characteristics, and inpatient complications and treatments that occurred prior to the 36-week PMA MRI.

GEE models (working Gaussian distribution) were used to estimate the association between global and regional brain abnormality scores and BSID-III cognitive outcomes (motor, cognitive, and language) at age 2 years of corrected age with adjustment for treatment assignment, gestational age, and study recruitment site. All statistical analyses were performed using the R statistical software package (v 3.3.0).

Results

Among the 422 infants enrolled at the 8 designated PENUT MRI sites, 59% (n = 250) underwent brain MRI scanning between 36^0/7 and 36^6/7 weeks of PMA per study protocol. Scans from 220 infants met standards for clinical assessment and were included in the analyses. Follow-up BSID-III assessments were conducted at 2 years of corrected age on 81% (n = 178) of the infants with usable MRI scans (Figure 1; available at www.jpeds.com).

Figure 1.

CONSORT diagram of PENUT MRI cohort.

To assess the representativeness of enrolled infants, demographic and clinical characteristics and neonatal complications of the MRI cohort were compared with infants who did not receive an MRI at the designated MRI site who were alive at 36 weeks of PMA (Table III; available at www.jpeds.com). Infants enrolled in the MRI cohort were heavier at birth (827g vs 782 g; P = .01), had a lower rate of severe sepsis (4% vs 12%; P = .003), but had a higher rate of severe BPD (42% vs 32%; P = .02) compared with infants that were not scanned.

Table III.

Baseline demographics and inpatient complications and treatments by MRI cohort enrollment status

Variables	Enrolled in MRI cohort	Not enrolled in MRI cohort*	P value†
n	220	202
Gestational age at birth, n (%)			.23
24 wk	45 (20%)	52 (26%)
25 wk	60 (27%)	43 (21%)
26 wk	62 (28%)	49 (24%)
27 wk	53 (24%)	58 (29%)
Mean (SD) wk	26.0 (1.1)	25.9 (1.2)	.93
Birth weight, mean (SD) g	827 (180)	782 (183)	.01
Weight <10th percentile	29 (13%)	32 (16%)	.43
OFC <10th percentile	42 (20%)	30 (15%)	.23
Multiple gestation	53 (24%)	50 (25%)	.88
Sex	110 (50%)	90 (45%)	.26
Female
Male	110 (50%)	112 (55%)
Prenatal steroids, n (%)	205 (93%)	177 (88%)	.05
Apgar at 5 min <5, n (%)	39 (18%)	45 (22%)	.23
Treatment assignment	110 (50%)	102 (51%)	.92
Baseline Epo level, median (IQR)	8.7 (4.5, 22.6)	7.5 (4.1,24.7)	.71
Inpatient complications
Grade III/IV IVH	20 (9.1%)	22 (10.9%)	.57
Treated PDA	99 (45%)	78 (39%)	.18
PVL	15 (7.2%)	13 (6.7%)	.76
Sepsis	9 (4.1%)	20 (12.4%)	.003
Severe BPD	93 (42%)	63 (32%)	.02
Severe NEC	10 (4.5%)	13 (6.4%)	.46
Severe ROP	17 (7.7%)	16 (7.9%)	.89
SIP	5 (2.3%)	9 (4.5%)	.24
Inpatient treatments
Intubation >1 wk	133 (60%)	130 (64%)	.37
Postnatal steroids	64 (29%)	72 (36%)	.17
Vasopressor use	102 (46%)	90 (45%)	.72
Opioids >1 wk	70 (32%)	81 (40%)	.08
Benzodiazepines >1 wk	58 (26%)	41 (20%)	.13
Antibiotics >1 wk	188 (85%)	176 (87%)	.54

NEC, necrotizing enterocolitis; PDA, patent ductus arteriosus; SIP, spontaneous intestinal perforation.

^*Among patients surviving to 37 weeks of PMA that would have been eligible for MRI.

^†Adjusted for treatment assignment and gestational age at birth.

Table IV summarizes the demographic and clinical data for the 220 infants in the MRI cohort, separated by brain injury scores (normal, mild moderate, or severe); 27% of infant scans (n = 60) were considered normal, and 60% (n = 133), 10% (n = 23), and 2% (n = 4) were rated as having evidence of mild, moderate, or severe brain abnormality respectively. One-half of the infants were assigned to each of the Epo (n = 110) and placebo (n = 110) treatment groups. Infant birth weight was significantly different by global brain abnormality category (P = .01), higher in infants with a normal MRI score (mean of 905 g vs 799 g, 773 g, and 878 g for mild/moderate/severe infants, respectively). There were no other statistically significant differences in baseline infant characteristics between infants with normal, mild, moderate, or severe scores.

Table IV.

Baseline infant characteristics by 36-week brain MRI injury score

Characteristic	Overall*	36-wk MRI brain abnormality score				P value
		Normal (0-2)	Mild (3-6)	Moderate (7-10)	Severe (11+)
n	220	60 (27%)	133 (60%)	23 (10%)	4 (1.8%)
Gestational age at birth, n (%)						.09
24 wk	45 (20%)	7 (16%)	31 (69%)	5 (11%)	2 (4.4%)
25 wk	60 (27%)	16 (27%)	38 (63%)	5 (8.3%)	1 (1.7%)
26 wk	62 (28%)	18 (29%)	37 (60%)	7 (11%)	0 (0%)
27 wk	53 (24%)	19 (36%)	27 (51%)	6 (11%)	1 (1.9%)
Mean (SD) wk	26.0 (1.1)	26.2 (1.0)	25.9 (1.1)	25.9 (1.1)	25.5 (1.5)	.27
Birth weight, mean (SD) g	827 (180)	905 (186)	799 (166)	773 (198)	878 (148)	.01
Multiple gestation						.72
Yes	53 (24%)	15 (28%)	32 (60%)	5 (9.4%)	1 (1.9%)
No	167 (76%)	45 (27%)	101 (60%)	18 (11%)	3 (1.8%)
Sex						.42
Female	110 (50%)	36 (33%)	56 (51%)	15 (14%)	3 (2.7%)
Male	110 (50%)	24 (22%)	77 (70%)	8 (7.3%)	1 (0.9%)
Head circumference <10th percentile						.79
Yes	42 (20%)	13 (31%)	24 (57%)	4 (9.5%)	1 (0%)
No	173 (80%)	46 (27%)	106 (61%)	18 (10%)	3 (2.4%)
Weight <10th percentile						.53
Yes	29 (13%)	10 (34%)	16 (55%)	3 (10%)	0 (0%)
No	191 (87%)	50 (26%)	117 (61%)	20 (10%)	4 (2.1%)
Antenatal steroids, n (%)						.32
Yes	205 (93%)	54 (26%)	126 (61%)	22 (11%)	3 (1.5%)
No	15 (6.8%)	6 (40%)	7 (47%)	1 (6.7%)	1 (6.7%)
Apgar at 5 min <5, n (%)						.10
Yes	39 (18%)	54 (30%)	107 (59%)	18 (10%)	1 (0.6%)
No	180 (82%)	6 (15%)	25 (64%)	5 (13%)	3 (7.7%)
Treatment assignment						.31
Epo	110 (50%)	31 (28%)	67 (61%)	12 (11%)	0 (0%)
Placebo	110 (50%)	29 (26%)	66 (60%)	11 (10%)	4 (3.6%)
Baseline Epo level, median (IQR)	8.7 (4.5, 22.6)	11.3 (4.9,18.4)	7.4 (4.3,19.0)	8.9 (4.9, 39.2)	44.9 (30.6, 56.3)	.28

^*The values for percentages in this column refer to the totals in that column while the percentages in the corresponding rows are based on the numbers of the MRI scores from normal to severe in the row.

Figure 2 shows the distribution of the global brain abnormality scores for Epo intervention compared with placebo groups (Figure 2, A), by gestational age (Figure 2, B), and further stratified by white matter (Figure 2, C), gray matter (cortical and deep) (Figure 2, D and E), and cerebellar injury (Figure 2, F). Global brain abnormality scores did not significantly differ between the Epo and placebo treatment groups (Figure 2, A: difference = −0.3; 95% CI −1.0, 0.03; P = .33), but tended to be lower with increased gestational age at birth (Figure 2, B: difference per week: −0.3; 95% CI −0.6, 0.0; P = .06). None of the regional abnormality scores significantly differed between treatment groups (Figure 2, C-F).

Figure 2.

MRI injury scores by treatment group and gestational age at birth. Gray bars and whisker plots – Epo-treated subjects. Black bars and whisker plots – Placebo-treated subjects. No differences in MRI brain injury scores were noted between treatment groups at any gestational age. A, Global brain abnormality scores between treatment groups. B, Global brain abnormality scores by treatment group and gestational age. C, White matter brain abnormality scores by treatment group. D, Cortical gray matter abnormality scores by treatment group. E, Deep gray matter abnormality scores bt treatment group. F, Cerebellum abnormality scores by treatment group.

Figure 3 (available at www.jpeds.com) displays the distribution of global brain injury scores by prior inpatient complications, and for treatments or therapeutics used prior to 36 weeks of PMA. Having a prior diagnosis of severe IVH (P < .001), periventricular leukomalacia (PVL) (P < .001), severe BPD (P = .01), or severe ROP (P = .01) were all associated with increased brain MRI injury scores at 36 weeks of PMA. Intubation for >7 days (P = .008), and prolonged treatment (>7 days) with opioids (P < .001), benzodiazepines (P = .02), or antibiotics (P = .002) were also associated with increased global brain injury scores.

Figure 3.

Brain MRI injury severity on 36-week MRI by presence of inpatient complications, use of therapeutics, and treatments. Poisson GEE models were utilized to compare MRI global abnormality scores by inpatient complication, therapeutic, or treatment status. Yes indicates that inpatient complication or treatment was present. No indicates that inpatient complication or treatment was not present. Infants with grade III/IV intracranial hemorrhage, PVL, severe ROP, and severe BPD had higher brain MRI injury scores than those infants without such complications. Interventions for greater than 7 days such as intubation, opioid, benzodiazapine, and antibiotic treatment were associated with higher brain MRI injury scores.

BSID-III cognitive, motor, and language scores at age 2 years of corrected age are plotted against global and regional brain abnormality scores at week 36 of PMA in Figure 4, A-I, and associations are summarized in Figure 5. Increased global brain abnormality and white matter injury scores at 36 weeks of PMA were associated with reductions in cognitive, motor, and language abilities at 2 years of corrected age. After adjustment for treatment with Epo, gestational age, and recruitment site, each point of increased global abnormality score was associated with a 2.5-point reduction in motor function (95% CI −3.5, −1.5; P < .001), a 2.2-point reduction in cognitive ability (95% CI −3.3, − 1.1; P < .001), and a 1.5-point reduction in language skills (95% CI −2.6, −0.4; P < .001). Increasing white matter injury score was associated with larger reductions in motor function (effect = −4.8; 95% CI −6.5, −3.2; P < .001), cognitive ability (effect = −4.3; 95% CI −6.1, −2.6; P < .001), and language skills (effect = −2.7; 95% CI −4.5, −0.8; P < .001). An association between cerebellum injury scores and BSID-III motor scores (effect = −2.1; 95% CI —3.8, 0.2; P = .04) was observed, but was attenuated for cognitive ability and language skills.

Figure 4.

BSID-III cognitive, motor, language scores at 22-26 months by global brain injury score, white matter score, and cerebellum score. A-C, demonstrate relationship of the BSID-III subscale scores (Motor; Cognitive; Language) to the global brain MRI injury score. D-F, demonstrate relationship of the BSID-III subscale scores (Motor; Cognitive; Language) to the white matter MRI injury score. G-I, demonstrate relationship of the BSID-III subscale scores (Motor; Cognitive; Language) to the cerebellum brain MRI injury score. All BSID-III scores were significantly related to higher brain injury scores except for cognitive and language scores for the cerebellum.

Figure 5.

Forest plot of associations between BSID-III cognitive, motor, language scores at 22-26 months of corrected age and global brain injury, white matter injury, and cerebellum injury scores. Significant associations were demonstrated between higher MRI injury scores and lower BSID-III scores except for cerebellum MRI injury and BSID-III cognitive and language scores.

Among the infants who had a normal (0-2) global abnormality score at 36 weeks of PMA, 78% (40 of 51) subjects had no neurodevelopmental impairment at 2 years of corrected age. Among the 66 infants with any neurodevelopmental impairment at age 2 years of corrected age, 55 (83%) had a global brain abnormality score (mild, moderate, or severe) at 36 weeks of PMA. When the global brain abnormality score was severe range (≥11), 2 of 3 subjects (67%) had severe neurodevelopmental impairment at 2 years of corrected age.

Discussion

Brain injury scores assessed on 36-week PMA brain MRI scans correlated inversely with BSID-III motor, cognitive, and language scores at 2 years of corrected age in our cohort. Both global injury scores and white matter-specific injury scores correlated inversely with motor, cognitive, and language outcomes at 2 years of corrected age.

Global brain injury scores were also associated with complications of prematurity, including a prior diagnosis of severe IVH and PVL on cranial ultrasound, severe BPD, and severe ROP. Each of these factors is known to have a significant correlation to brain injury and abnormal brain growth, with corresponding negative effects on long-term neurodevelopmental outcomes.^11-15 Injury scores were also higher in infants intubated for >7 days, and those treated with opioids, benzodiazepines, or antibiotics >7 days. Although it has been shown that neonatal infection is related to neurodevelopmental and growth impairment by Stoll et al,¹⁶ to our knowledge, the use of antibiotic treatment >7 days and findings of prolonged treatment with opioids and benzodiazepines affecting brain injury scores at term equivalent MRI scans in extremely premature infants have not been previously described.

Infant birth weight, but not gestational age, was statistically correlated to global brain abnormality category in our study; infants with higher birth weight were significantly more likely to have normal MRI scans.

Although there is consensus that brain MRI at term equivalent age in extremely premature infants can detect more subtle abnormalities when compared with cranial ultrasound, its role as a predictive tool for subsequent neurodevelopmental outcomes is controversial. Brain MRI has been claimed to be an important tool to prognosticate neurodevelopmental outcomes in extremely low birth infants and has been an area of active research.^6,7-25 The Neuroimaging and Neurodevelopmental Outcomes (NEURO) study was a secondary study to the Surfactant Positive Airway Pressure and Pulse Oximetry Randomized Trial (SUPPORT NCT00233324) which enrolled infants born at 24-27^6/7 weeks of gestational age from 20 centers²⁵. The goals of this study were to investigate the relationship between early and late cranial ultrasound adverse findings and near-term brain MRI findings of white matter abnormalities (WMAs) and cerebellar lesions with neurodevelopmental outcomes. This study also assessed the relative value of early cranial ultrasound, late cranial ultrasound, and MRI to predict outcomes. Four hundred eighty infants had complete neuroimaging with late cranial ultrasound (defined as cranial ultrasound at 35-42 weeks of PMA) and brain MRI within 2 weeks of each other. A comprehensive neurodevelopmental assessment including BSID-III cognitive composite scores, Gross Motor Function Classification System levels, visual and hearing impairments were obtained at 18-22 months of corrected age in 441 children. The NEURO study concluded that adverse near-term brain MRI findings of both moderate to severe WMA and significant cerebellar lesions and adverse late cranial ultrasound findings among extremely preterm infants were associated with adverse neurodevelopmental outcomes at 18-22 months. Increasing severity of WMA and presence of cerebellar lesions were associated with significantly lower mean BSID-III cognitive scores and moderate to severe cerebral palsy, which is consistent with our study findings. Ibrahim et al reviewed the ability of term equivalent cranial ultrasound and MRI to predict neurodevelopmental impairment and concluded that MRI was more sensitive to white matter and cerebellar injury then ultrasound.²⁶ They recommended term-equivalent MRI scans in infants born at 29 weeks of estimated gestational age or less. This approach would allow for initiation of early intervention services in infants at risk for neurodevelopmental impairment and cerebral palsy.

Several MRI scoring systems have been developed for preterm infants at term equivalent age to predict neurodevelopmental outcomes,^7,13,27 however, the prognostic precision of term equivalent brain MRI abnormalities in ELBW infants for long-term neurodevelopmental outcomes are unclear and its use as a standard care is not yet recommended by American Academy of Pediatrics or American Academy of Neurology. However, a commentary by Inder et al recommends reconsideration of this stance, as cranial ultrasound provides information on overt brain injury whereas MRI provides detailed information on less overt brain injury, brain growth, and maturation.²⁸

MRI scans in our contemporary ELBW cohort demonstrated lower brain injury scores on average compared with the original Kidokoro data.⁷ Brouwer et al published similarly low injury scores in a more recent cohort of ELBW infants when compared with the original Kidokoro cohort.¹⁵ To confirm that our scores were consistent with scores determined by other experienced users of the Kidokoro scoring system, 10 selected MRI scans of varying degrees of abnormality were read by Floris Groenendaal and Linda de Vries. Figure 6 (available at www.jpeds.com) shows that our scoring and that of the external reviewers were quite similar with R = 0.88 (Spearman correlation coefficient). The Bland-Altman plot demonstrated good agreement between the different scorers.

Figure 6.

Inter-rater scoring between the Netherlands team and University of Washington team for 10 selected MRIs. Independent MRI injury scoring interpretations were compared between the Netherlands team and the University of Washington team. The Spearman correlation value was 0.88. The Bland-Altman plot demonstrates good agreement between the scoring team.

In a subset of infants enrolled in the Swiss EPO Neuroprotection Trial Group study, Leuchter et al showed fewer infants treated with Epo had abnormal scores for white matter injury, white matter signal intensity, periventricular white matter loss, and gray matter injury.²⁹ They concluded that high-dose Epo treatment within 42 hours after birth was associated with reduced risk of brain injury on MRI.²⁹ In our PENUT cohort, which included infants of lower gestational age (24-27 weeks included in PENUT compared with 26-32 weeks enrolled in the Swiss study), the overall brain injury scores based on MRI scans of infants who received Epo were not different from the placebo group’s scores, and no regional differences in the brain MRI scans were noted based on treatment group.

Although the Swiss group reported better brain injury scores in the Epo treatment group based on MRI scans at 36 weeks of PMA, no statistically significant differences in neurodevelopmental outcomes were found at 2 and 5 years of age.³⁰ This finding was consistent with PENUT Trial neurodevelopmental outcomes at 2 years corrected age.³

We found that white matter injury was associated with reductions in motor function, but also in cognitive development and language skills, which is consistent with previous studies.^11,18,19,31 Martinez-Biarge et al studied MRI scans of 82 preterm infants and concluded that white matter injury severity was strongly associated with the presence and severity of cerebral palsy and other neurodevelopmental impairments.²⁷

The importance of cerebellar injury in preterm infants has become increasingly recognized and has been associated with poorer cognition and motor performance.^15,32-34 In our cohort, cerebellar injury scores only correlated with motor scores and not with cognition or language scores at 2 years corrected age.

Strengths of our study include the standardization of MRI protocols at all included sites, and use of a standardized MRI scoring system at term equivalent age in a large contemporary cohort of extremely preterm infants with a high rate of retention of survivors at follow up at 2 years of corrected age. The MRI scans were assessed by neuroradiologists who were unaware of the treatment group and their neonatal intensive care unit course, and similarly, neurodevelopmental assessments were carried out by blinded centrally trained examiners at 2 years of corrected age.

It is a limitation of this study that we cannot account for differences between infants who received an MRI vs those who did not at the MRI study sites. It is possible that the more stable infants were chosen to go to MRI, and therefore the MRIs analyzed do not reflect the full extent of neuropathology seen in this preterm group.

In conclusion, evidence of global brain injury and white matter-specific injury on brain MRIs obtained at 36 weeks of PMA correlated with adverse neurodevelopmental outcomes as assessed by the BSID-III at 22-26 months of corrected age. Early Epo treatment had no effect on the MRI brain injury scores compared with the placebo group. The prognostic accuracy of term equivalent brain MRI findings in ELBW infants for long-term neurodevelopmental outcomes needs to be determined for recommendation of its routine use as a screening tool in this population.

Data statement

Data sharing statement available at www.jpeds.com.

Glossary

BPD

Bronchopulmonary dysplasia

BSID-III

Bayley Scales of Infant and Toddler Development, Third Edition

ELBW

Extremely low birth weight

Epo

Erythropoietin

GEE

Generalized estimating equations

IVH

Intraventricular hemorrhage

MRI

Magnetic resonance imaging

NEURO

Neuroimaging and Neurodevelopmental Outcomes

PENUT

Preterm Erythropoietin Neuroprotection

PMA

Postmenstrual age

PVL

Periventricular leukomalacia

ROP

Retinopathy of prematurity

WMA

White matter abnormality

Appendix List of Additional Members of the PENUT Trial Consortium

PENUT Primary Investigators and Co-Authors

Rajan Wadhawan, MD,¹ Sherry E. Courtney, MD,² Tonya Robinson, MD,³ Kaashif A. Ahmad, MBBS, MSc,⁴ Ellen Bendel-Stenzel, MD,⁵ Mariana Baserga, MD,⁶ Edmund F. La-Gamma, MD,⁷ L. Corbin Downey, MD,⁸ Raghavendra Rao, MD,⁹ Nancy Fahim, MD,⁹ Andrea Lampland, MD,¹⁰ Ivan D. Frantz, III, MD,¹¹ Janine Khan, MD,¹² Michael Weiss, MD,¹³ Maureen M. Gilmore, MD,¹⁴ Robin K. Ohls, MD,¹⁵ Jean Lowe, PhD,¹⁵ Nishant Srinivasan, MD,¹⁶ Jorge E. Perez, MD,¹⁷ Victor McKay, MD¹⁸

Nonauthors

PENUT Co-Investigators.

Billy Thomas, MS, MPH,² Nahed Elhassan, MD, MPH,² Sarah Mulkey, MD, PhD,² Vivek K. Vijayamadhavan, MD,⁴ Neil Mulrooney, MD,⁵ Bradley Yoder, MD,⁶ Jordan S. Kase, MD,⁷ Jennifer Check, MD, MS,⁸ Erin Osterholm, MD,⁹ Thomas George, MD,⁹ Michael Georgieff, MD,⁹ Camilia R. Martin, MS,¹¹ Deirdre O’Reilly, MD, MPH,¹¹ Raye-Ann deRegnier, MD¹² Nicolas Porta, MD,¹² Catalina Bazacliu, MD,¹³ Frances Northington, MD,¹⁴ Raul Chavez Valdez, MD,¹⁴ Patel Saurabhkumar, MD, MPH,¹⁶ Magaly Diaz-Barbosa, MD,¹⁷ Todd Richards, PhD¹⁹

PENUT Research Coordinators.

John B. Feltner,¹⁹ Isabella Esposito,¹⁹ Stephanie Hauge,¹⁹ Samantha Nikirk,¹⁹ Amy Silvia,¹⁹ Bailey Clopp,¹⁹ Debbie Ott,¹ Ariana Franco Mora,¹ Pamela Hedrick,¹ Vicki Flynn,¹ Andrea Wyatt,² Emilie Loy,² Natalie Sikes,² Melanie Mason,² Jana McConnell,² Tiffany Brown,² Henry Harrison,² Denise Pearson,² Tammy Drake,² Jocelyn Wright,² Debra Walden,² Annette Guy,² Jennifer Nason,³ Morgan Talbot,³ Kristen Lee,³ Sarah Penny,³ Terri Boles,³ Melanie Drummond,⁴ Katy Kohlleppel,⁴ Charmaine Kathen,⁴ Brian Kaletka,^5,10 Shania Gonzales,^5,10 Cathy Worwa,^5,10 Molly Fisher¹⁰, Tyler Richter,^5,10 Alexander Ginder,^5,10 Brixen Reich,⁶ Carrie Rau,⁶ Manndi Loertscher,⁶ Laura Cole,⁶ Kandace McGrath,⁶ Kimberlee Weaver Lewis,⁶ Jill Burnett,⁶ Susan Schaefer,⁶ Karie Bird,⁶ Clare Giblin,⁷ Rita Daly,⁷ Kristi Lanier,⁸ Kelly Warden,⁸ Jenna Wassenaar,⁹ Jensina Ericksen,⁹ Bridget Davern,⁹ Mary Pat Osborne,⁹ Neha Talele,¹¹ Evelyn Obregon,¹¹ Tiglath Ziyeh,¹¹ Molly Clarke,¹¹ Rachel E Wegner,¹¹ Palak Patel,¹¹ Molly Schau,¹² Annamarie Russow,¹² Kelly Curry,¹³ Lisa Barnhart,¹³ Charlamaine Parkinson,¹⁴ Sandra Beauman,¹⁵ Mary Hanson,¹⁵ Elizabeth Kuan,¹⁵ Conra Backstrom Lacy,¹⁵ Edshelee M. Galvis,¹⁷ Susana Bombino,¹⁷ Denise Martinez,¹⁸ Suzi Bell,¹⁸ Corrie Long¹⁸

University of Washington Data Coordinating Center.

Christopher Nefcy,¹⁹ Mark A. Konodi, MS,¹⁹ Phuong T. Vu, PhD¹⁹

PENUT Executive Committee.

Adam Hartman MD,²⁰ T. Michael O’Shea MD,²¹ Roberta Ballard MD²²

Follow Up Committee.

Mike O’Shea, MD,²¹ Karl Kuban, MD,²³ Jean Lowe, PhD¹⁵

Independent Medical Monitor.

John Widness, MD²⁴

Sites

1. Advent Health for Children, Orlando, FL

2. University of Arkansas for Medical Sciences, Little Rock, AK

3. University of Louisville, Louisville, KY

4. Methodist Children’s Hospital, San Antonio, TX

5. Children’s Minnesota, Minneapolis, MN

6. University of Utah, Salt Lake City, UT

7. Maria Fareri Children’s Hospital at Westchester, Valhalla, NY

8. Wake Forest School of Medicine, Winston-Salem, NC

9. University of Minnesota Masonic Children’s Hospital, Minneapolis, MN

10. Children’s Minnesota, St. Paul, MN

11. Beth Israel Deaconess Medical Center, Boston, MA

12. Prentice Women’s Hospital, Chicago, IL

13. University of Florida, Gainesville, FL

14. Johns Hopkins University, Baltimore, MD

15. University of New Mexico, Albuquerque, NM

16. Children’s Hospital of the University of Illinois, Chicago, IL

17. South Miami Hospital, South Miami, FL

18. Johns Hopkins All Children’s Hospital, St. Petersburg, FL

19. University of Washington, Seattle, WA

20. National Institute of Neurological Disorders and Stroke, Rockville, MD

21. University of North Carolina School of Medicine, Chapel Hill, NC

22. University of California, San Francisco, CA

23. Boston University School of Medicine, Boston, MA

24. University of Iowa, Iowa City, IA