325 infants were recruited to the TOBY trial between 2002 and 2006. MRI scans were obtained from 151 infants from the 22 hospitals with facilities for neonatal MRI. shows the study profile. 12 (seven cooled, five non-cooled) of the 151 sets of images were unsuitable for analysis. One infant who had an overt pontocerebellar hypoplasia seen on MRI was excluded from further analysis. MRIs from three cooled and four non-cooled infants were taken post mortem, and these images were not included. Therefore, images were available for analysis for 131 infants: 64 in the group allocated to cooling and 67 infants in the non-cooling group. There was no difference in gestational age or birthweight between the two groups. The cooled infants had similar 10-min Apgar scores and there was no difference in the severity of amplitude integrated EEG between the two groups (p=0·32; ). The clinical characteristics of these 131 infants were not different from the rest of the infants in the TOBY trial, although there were more severely abnormal amplitude integrated EEG findings in infants that were not included in the substudy (webappendix
Clinical characteristics of the infants
The median age at postnatal scan was 8 days (range 2–30) in both groups and all infants were older than 37 weeks gestational age at the time of scan. Observer agreement for classification of MRI scans was greater than 99%. None of the patterns of injury were deemed to be unusual for patients with hypoxic–ischaemic encephalopathy.
On univariate analysis, treatment allocation was significantly associated with abnormal MRI in the basal ganglia and thalami (p=0·01), white matter (p=0·01), and the posterior limb of the internal capsule (p=0·03); there was an improvement of trend across categories of severity of MRI abnormalities in the basal ganglia and thalami (p=0·02) and white matter (p=0·02) in the cooled group compared with the non-cooled group. Postnatal age at scan was significantly associated with abnormalities in the basal ganglia and thalami when examined as a continuous variable (p=0·02) or categorised into less than 8 days or 8 or more days (p=0·01), but did not correlate with abnormalities in the white matter or cortex. The amplitude integrated EEG grade correlated with abnormalities in the basal ganglia and thalami (p=0·001), white matter (p=0·003), and posterior limb of the internal capsule (p<0·0001). Postnatal age at scan and grade of amplitude integrated EEG were included as covariates in all the binomial logistic regression analyses of treatment allocation and MRI findings ().
Grades of cerebral lesions seen on MRI in cooled and non-cooled infants
Lesions in the basal ganglia and thalami were detected in 38 of 64 (59%) cooled infants and 53 of 67 (79%) non-cooled infants (adjusted OR 0·36, 95% CI 0·15–0·84; p=0·02, ). The posterior limb of the internal capsule was normal in 34 of 64 (53%) cooled infants and 23 of 67 (34%) non-cooled infants, equivocal in two of 64 (3%) cooled infants and five of 67 (7%) non-cooled infants, and abnormal in 28 of 64 (44%) cooled infants and 39 of 67 (58%) non-cooled infants (0·38, 0·17–0·85; p=0·02). White matter abnormalities were seen in 41 of 64 (64%) cooled infants and 56 of 67 (84%) non-cooled infants (0·30, 0·12–0·77; p=0·01). Cortical abnormalities were seen in 30 of 64 (47%) cooled infants and 42 of 66 (64%) non-cooled infants (0·62, 0·27–1·41; p=0·25).
Lesions that predicted abnormal neurodevelopmental outcome (defined as at least one of: moderate or severe lesions in the basal ganglia and thalami [grade 2 or 3], an abnormal posterior limb of the internal capsule, or severe white matter abnormalities [grade 3]) were seen in 29 of 64 (45%) cooled infants and 42 of 67 (63%) non-cooled infants (adjusted OR 0·42, 95% CI 0·20–0·92; p=0·03). 22 of 64 (34%) cooled infants and 11 of 67 (16%) non-cooled infants had normal scans (2·81, 1·13–6·93; p=0·03).
Logistic regression analysis showed that, in addition to treatment allocation, grade of amplitude integrated EEG was associated with abnormalities in the basal ganglia and thalami, white matter, and posterior limb of the internal capsule (all p<0·0001); major abnormalities that predicted abnormal outcome (p<0·0001); and normal scans (p=0·001). 37 of 59 (63%) infants scanned at less than 8 days of age had major abnormalities on MRI that were predictive of abnormal outcome compared with 34 of 72 (47%) infants scanned later (p=0·08; ). After adjustment for grade of amplitude integrated EEG and treatment allocation in infants scanned at less than 8 days of age compared with those scanned later, the OR for the detection of major abnormalities seen on MRI was 2·35 (1·07–5·14; p=0·03).
Postnatal age at scan, MRI abnormalities, and outcome up to 18 months of age in cooled and non-cooled infants
Three infants (two cooled and one non-cooled) had evidence of sinus thrombosis and 47 (36%) had evidence of haemorrhage: 25 from the cooled group and 22 from the non-cooled group. The haemorrhage was subdural in 39 infants, moderate in ten infants, and mild in 29 infants.
Assessment of the primary outcome was available for all but one infant. Of 130 infants, six cooled and 11 non-cooled infants died (p=0·23), whereas 19 cooled and 22 non-cooled infants had severe disability at 18 months of age (p=0·70; ). In the main TOBY trial,12
the rates of death were 42 of 163 (0·26) in the cooled group and 44 of 162 (0·27) in the non-cooled group. The rates for severe disability in survivors were 32 of 120 (0·27) in the cooled group and 42 of 117 (0·36) in the non-cooled group.
In this subset, all 17 infants who died and 36 of 41 (88%) infants with severe disability had moderate or severe lesions in the basal ganglia and thalami and an abnormal posterior limb of the internal capsule. Four of the remaining five infants who were classified as having severe disability had mild lesions in the basal ganglia and thalami and mild or normal white matter; the other infant had normal basal ganglia and thalamus but moderate lesions in the white matter. These five infants had an MDI less than 70 but a psychomotor developmental index (PDI) greater than 70. Four of the five infants had a GMFCS of 1 (the mildest level of impairment).
11 infants with moderate lesions and two infants with severe lesions in the basal ganglia and thalami did not have severe disability at 18 months. An abnormal posterior limb of the internal capsule was seen in 11 of 13 infants. These 13 infants had a PDI and MDI greater than 70, but five had a GMFCS of 1. All 34 infants (22 cooled, 12 non-cooled) with normal basal ganglia and thalami and normal or mild lesions in the white matter had a normal neuromotor outcome at 18 months.
The predictive accuracy of moderate or severe lesions in the basal ganglia and thalami, severe white matter lesions, or an abnormal posterior limb of the internal capsule for death or severe disability at 18 months of age was 0·84 (95% CI 0·74–0·94 [53 of 63]) in the cooled group and 0·81 (0·71–0·91 [54 of 67]) in the non-cooled group ().
Predictive ability of major MRI abnormalities during first 4 weeks after birth for death or severe disability at 18 months