Main

Acute cerebral hypoxia-ischemia, usually due to “birth asphyxia” (critically impaired intrapartum gas exchange), is the major cause of perinatal brain injury(1). Previous studies by phosphorus MRS of the brains of birth-asphyxiated infants have shown changes indicating reduced oxidative phosphorylation which were at their worst 2 to 4 d after birth and which were predictive of an adverse outcome(2, 3). When oxidative phosphorylation is impaired, anaerobic glycolysis and the production of lactic acid are to be expected. Proton (1H) MRS allows noninvasive observations to be made of cerebral Lac and other metabolites, particularly total Cr, Cho, and NAA, in newborn infants with and without hypoxic-ischemic brain injury.

Previous investigators have provided data suggesting that Lac is undetectable in the normal neonatal brain at term, but may be found in the brains of both preterm infants and infants who are small for gestational age(4). The presence of cerebral Lac has, however, been recorded in term infants after birth asphyxia(5) and was thought to indicate a poor prognosis, as was a reduced NAA/Cho peak-area ratio in infants studied at mean ages of 7(5) and 14(6) days. Frequently, however, Lac has remained undetected in infants with significant hypoxic-ischemic encephalopathy who subsequently had adverse outcomes(5, 6).

The first purpose of this study was to define normal changes with gestation in cerebral 1H-metabolite peak-area ratios. An asymmetric PRESS pulse sequence(7) was used, as it has been shown that such a sequence greatly enhances the detectability of Lac(8). Two regions of the brain were studied, the thalamus and the OP region, differing with respect to myelination(9) and susceptibility to hypoxic-ischemic injury(1). The hypothesis was then tested that abnormalities of peak-area ratios, particularly those indicating an increase in cerebral Lac and a fall in NAA, would be found in the brains of infants thought to have sustained perinatal hypoxia-ischemia and might be predictive of an adverse outcome. The findings were related to the results of neurodevelopmental follow-up examinations at 1 y of age.

METHODS

Infants studied. A total of 54 infants was studied; 35 control infants and 19 suspected of hypoxic-ischemic brain injury. The control infants were born at 29-42 (median 34) wk of gestation weighing 1.16-4.30 (median 2.30) kg. Gestational age was assigned from maternal dates, antenatal ultrasound, and the physical and neurologic characteristics of the infants. Their weights and head circumferences at birth and at study were between the 3rd and 97th centiles. There was no evidence of birth asphyxia. In the 21 who had umbilical cord blood taken, the base deficit ranged from -6 to +5 mmol·L-1. Cranial ultrasound scans of the 20 infants who were admitted to the neonatal unit were normal; ultrasound scans were not performed on the other 15 infants, who were recruited from the postnatal wards. Nine of the control infants had recovered from mild hyaline membrane disease, three of whom had required mechanical ventilation for 1 to 5 d and had received surfactant. Each control infant was studied on one occasion only aged 1-46(median 9) d, at a GPA age of 31-42 (median 36) wk. Both the thalamus and the OP region were studied in 15 of the infants, only the thalamus in 14, and only the OP region in 6.

The 19 asphyxiated infants had been admitted to the neonatal intensive care unit with evidence or suspicion of hypoxic-ischemic brain injury. Recruitment to the study was based upon a clinical history consistent with intrapartum or postnatal asphyxia and the presence of: 1) a base excess of less than -15 mmol·L-1 in cord blood or in the first arterial blood sample obtained after birth or after a postnatal asphyxial episode or2) abnormal neurologic signs such as hypotonia, abnormal movements, or seizures. The principal reason for entry to the study was presumed birth asphyxia in 17 infants and a postnatal asphyxial episode within 24 h of delivery in the remaining two. The clinical details of the 17 birth-asphyxiated infants are given in Table 1. The median base excess was -21 mmol·L-1. Of the remaining two infants, one had an arterial base excess of -12 mmol·L-1 after a respiratory arrest associated with a neonatal seizure, and the other had a base excess of-13 mmol·L-1 after a respiratory arrest secondary to aspiration. All except one of the 19 infants were resuscitated by endotracheal intubation and mechanical ventilation, and 10 required ventilation for over 2 d. All had abnormal neurologic signs; 5 infants were graded as Sarnat stage I, 12 stage II, and 2 stage III(10). (Both of the infants with postnatal asphyxia were graded II and had seizures.) One birth-asphyxiated infant was subsequently diagnosed at 18 d of age as also having congenital hypothyroidism [serum-free thyroxine, 8 pmol·L-1 (normal range 9-24 pmol·L-1), TSH >100 MU·L-1 (0.3-5.0 MU·L-1)]. The 19 infants were studied on a total of 34 occasions within the 1st wk of life: 9 infants were studied once only, 5 were studied twice, and 5 three times. The median age at first study was 1.3 (range 0.1-4.5) d and for all studies 1.8 (0.1-6.7) d at a GPA of 27-41 (median 39) wk. Both regions of the brain were studied in 14 infants, only the thalamus in 4, and only the OP region in 1.

Table 1 Clinical details of the 17 birth-asphyxiated infants

Throughout each study, monitoring was continued of breathing (Graseby MR10 apnoea monitor, Watford, UK), skin temperature (Thermistor probe 400 series, Portex, Hythe, UK), ECG, and arterial oxygen saturation by pulse oximeter(both Corometrics 556 monitor, Vickers, Sidcup, UK). Mechanical ventilation(modified Vickers transport incubator ventilator) and infusions of i.v. fluids were continued. In some babies requiring more intensive surveillance, transcutaneous Po2 and Pco2 were monitored (Corometrics 556 monitor) and also blood pressure via an indwelling arterial catheter and pressure transducer (Medex Medical, Rossendale, UK). A pediatrician was present inside the Faraday cage of the spectrometer for the duration of each study. None of the control infants was sedated, and the asphyxiated infants were not given any medication in addition to that prescribed as part of their clinical care. Five asphyxiated infants were receiving anticonvulsant medication and five others analgesic or sedative treatment at the time of study. The mean base excess in the last blood sample before study was -1.6 mmol·L-1 (SD 3.0), at a median time of 6 h before the scan. The study was approved by the University College London Committee on the Ethics of Human Research, and informed consent was obtained from parents.

1H MRS. A 2.4-tesla Bruker Biospec equipped with actively shielded gradients was used with series-tuned, inductively coupled, Helmholtz head coils of either 15- or 18-cm diameter(11). A transverse scout image (3-mm slice thickness) was first obtained with a TE of 31 ms and recovery time of 400 ms. This was used to position a sagittal image coplanar with the midline. A coronal image passing through the center of the cerebellum was then produced from the sagittal scout. For spectroscopy, a (2 cm)3 voxel was centered on the thalamus using the sagittal image. The coronal image was used to position the OP voxel [also (2 cm)3] in the center of the left cerebral hemisphere such that contact with the skull and consequent spectral fat contamination was avoided (Fig. 1).

Figure 1
figure 1

Positions of (2 cm)3 voxels for spectroscopy.(A) Sagittal cerebral image from an infant at 38 wk of GPA showing the thalamic voxel. (B) Coronal image of an infant at 40 wk of GPA showing the OP voxel.

Spectra were acquired using an asymmetric PRESS pulse sequence in which the 90 ° and first 180 ° pulses were 7.6 ms apart, with a TE of 270 ms, recovery time of 1730 ms, 2048 simultaneously sampled-quadrature data points, a spectral width of 1250 Hz, and 128 echoes averaged. Water suppression was achieved by three chemical-shift-selective pulses and accompanying spoiler gradients. PRESS acquisition-pulse amplitudes were adjusted by maximizing, and water-suppression pulses by minimizing, the water signal from each voxel. Magnetic field homogeneity was optimized by monitoring the unsuppressed water echo and adjusting the X, Y, and Z shim coil currents. Echoes were baseline-corrected, exponentially filtered to give 1 Hz line broadening, and then zero-filled to 2048 points before fast-Fourier transformation, manual (0 and 1st order) phasing, and cubic-spline baseline corrections. Peak areas on the spectra were determined by Lorentzianχ2 minimization(12) with the following prior knowledge: starting values for chemical shifts, peak widths and amplitudes, and multiplicity (for the Lac methyl doublet) including J-coupling and relative component amplitudes. Cho, Cr, and NAA were fitted as singlets.

Follow-up. At 12 mo of age (corrected for preterm birth if necessary) a structured neurologic examination(13) was carried out by a pediatrician unaware of the MRS results, and in addition the Griffiths Developmental Scales(14) were administered by a psychologist blind to both the infant's previous history and MRS results. From the results of these assessments, the neurodevelopmental status of the infants was assigned by one of us (A.S.), also without knowledge of the MRS results, as normal, minor impairment (without disability), or major impairment(with disability). A neurologic examination only was carried out on all of the 23 control infants who had reached 15 mo of corrected age.

Data analysis. Comparisons between data from the two voxel locations in the 15 control babies in whom both regions were studied were made by paired t test, as the data were normally distributed. As multiple comparisons were made, only values of p < 0.01 were considered significant. Linear regression was used to examine relationships between peak-area ratios and GPA, or postnatal age, in the 35 control infants. The 95% confidence intervals for a single point were also calculated for peak-area ratios in the normal infants. SD scores were determined from the difference between data points and the regressions for the control data. Values obtained for the asphyxiated infants were often nonnormally distributed; they were therefore expressed as medians with interquartile ranges, and compared with values from the control infants by the Mann-Whitney test, also using a significance level of p < 0.01. Fisher's exact test was used to determine whether values from the asphyxiated infants lying outside normal 95% confidence limits were related to poor outcome, which was defined as either death or disabling major impairment at 1 y of age.

RESULTS

Control infants. 1H spectra from the thalamus(A) and OP region (B) of a control infant are illustrated in Figure 2. Good spectral resolution allowed the clear assignment of peaks due to Cho at ≈3.2 ppm, Cr at ≈3.0 ppm, NAA at≈2.0 ppm, and Lac at ≈1.3 ppm. Lac was detected in all the control spectra and identified by its chemical shift and inversion at TE 135 ms(Fig. 3) (see “Discussion”). Smaller peaks that could be assigned to other metabolites were also frequently detected, such as glutamate plus glutamine and myoinositol plus glycine. InFigure 2, relative to the Cho, Cr, and NAA peaks, the Lac resonance is clearly larger in the OP region than in the thalamus. Peak-area ratios for the two regions, from the 15 infants in whom both were studied, are given in Table 2. Lac/Cho, Lac/Cr, and Lac/NAA were all highly significantly greater in the OP region than in the thalamus.

Figure 2
figure 2

1H spectra obtained during the same examination from the thalamus (A) and the OP region (B) of a control infant at 35 wk of GPA (born at 33 wk of gestation and studied at 15 d of age). This infant was neurologically normal when assessed at 15 mo. Resonance identifications: 1, glutamate/glutamine; 2, myoinositol/glycine; 3, Cho; 4, Cr; 5, glutamate/glutamine; 6, NAA; 7, Lac. The dashed line is the Lorentzian fit to the spectra.

Figure 3
figure 3

1H spectra from the OP region of a control infant at 34 wk of GPA (born at 30 wk of gestation, studied aged 30 d) with TE 270 ms (A), and TE 135 ms (B), the NAA to Lac region of the spectrum illustrating inversion of the Lac doublet. Resonance identifications are as for Figure 2.

Table 2 1 H metabolite peak-area ratios in the normal newborn brain

Regression equations for peak-area ratios versus GPA are given inTable 3. Significant falls in Lac relative to the other metabolites were found, particularly in the OP region, and NAA/Cho increased. The changes in Lac/NAA and NAA/Cho with GPA are illustrated by the spectra inFigure 4, and the values obtained from the control infants are shown Figures 5 and6. No significant changes in peak-area ratios were found in either region in relation to postnatal age, assessed both before and after subtraction of the GPA trend.

Table 3 Relation of cerebral1H metabolite peak-area ratios to gestational plus postnatal age
Figure 4
figure 4

Thalamic 1H spectra from infants of(A) 32 wk of GPA (born at 30 wk of gestation and studied aged 10 d) and (B) 39 wk of GPA (aged 2 d). Resonance identifications are as for Figure 2. The Lac/NAA peak-area ratio is higher(0.30), and NAA/Cho lower (0.42), in the spectrum from the preterm infant(A) than in that from the term infant (B), where Lac/NAA = 0.19 and NAA/Cho = 0.68.

Figure 5
figure 5

Relation between Lac/NAA peak-area ratio in the thalamus (A), the OP region (B), and GPA. = control infants; • = asphyxiated infants; ▪ = an asphyxiated infant subsequently discovered to have congenital hypothyroidism. The maximum Lac/NAA is given for asphyxiated infants studied more than once. The regression line and 95% confidence limits for normal values are shown.

Figure 6
figure 6

Relation between NAA/Cho peak-area ratio in the thalamus (A), the OP region (B), and GPA. = control infants; • = asphyxiated infants; ▪ = an asphyxiated infant subsequently discovered to have congenital hypothyroidism. The minimum NAA/Cho is given for asphyxiated infants studied more than once. The regression line and 95% confidence limits for normal values are shown.

Perinatal asphyxia. Spectra from the thalamus of a birth-asphyxiated 41-wk gestation infant at 2, 22, and 46 h of age are shown in Figure 7A. For comparison, a spectrum from the thalamus of a normal 41-wk gestation infant obtained at 84 h of age is shown inFigure 7B. The Lac resonance at ≈1.3 ppm is clearly larger relative to NAA in the asphyxiated baby. The prominent doublet (at≈1.1 ppm) adjacent to Lac is due to propan-1,2-diol, the injection medium of phenobarbitone with which the infant was treated for seizures(15).

Figure 7
figure 7

1H spectra from (A) the thalamus of a birth-asphyxiated infant of 41 wk of gestation, aged 2, 22, and 46 h, and(B) the thalamus of a normal infant at 41 wk of gestation, aged 84 h. Resonance identifications as for Figure 2, but also 8, propan-1,2-diol. In the asphyxiated infant, Lac/NAA was 0.52, 0.44, and 0.65, and NAA/Cho was 0.72, 0.71, and 0.70, at 2, 22, and 46 h, respectively. In the normal infant, Lac/NAA was 0.17 and NAA/Cho 0.70. The birth-asphyxiated infant had major neurodevelopmental disability at 1 y of age.

The maximum values for Lac/NAA observed between 2 and 160 h of age are shown in Figure 5, together with data from the control infants. Lac/NAA in the thalamus was above 95% confidence limits for controls in 10 of the 18 asphyxiated babies who had thalamic studies, and in two of these this ratio was also above 95% confidence limits in the OP region. One other infant, at 27 wk of GPA, had a high OP Lac/NAA. In total, therefore, values for Lac/NAA above 95% confidence limits were found in 11 of the 19 asphyxiated infants, in either the thalamus or the OP region, or both.Table 4 gives the median and interquartile ranges for SD scores of peak-area ratios from the asphyxiated infants. Lac in the thalamus, relative to NAA, Cho, and Cr, was significantly greater in the asphyxiated infants compared with the controls. In the OP region, only Lac/Cho was significantly elevated, although there was a trend for Lac/NAA and Lac/Cr also to be raised.

Table 4 GPA-dependent SD scores for1H metabolite peak-area ratios in the 19 asphyxiated infants

By 1 y of age, 2 asphyxiated infants had died and all 13 survivors who had reached this age had been examined. Five had major neurodevelopmental impairments and were seriously disabled, with one or more of: seizures requiring anti-convulsant treatment, axial imbalance and limb hypertonicity causing inability even to sit at 1 y of age, or a Griffiths general quotient more than 3 SD below the mean (<55). Four other survivors were assessed as normal at 1 y and four had minor neurologic impairments only, with normal Griffiths general quotients. Table 5 shows the relation between Lac/NAA values either above or within normal 95% confidence limits and outcome at 1 y in the 15 asphyxiated infants in whom this was known. Lac/NAA values above 95% confidence intervals were significantly associated with death or major disabling impairment (p < 0.002). The 23 control infants so far followed up at 15 mo of corrected age were neurologically normal.

Table 5 Relationship between Lac/NAA values and neurodevelopmental outcome at 1 y in 15 of the asphyxiated infants

Figure 6, which gives data for minimum recorded values for NAA/Cho in the asphyxiated infants, provides no evidence for reduced values compared with control infants, although in the thalamus of the infant later diagnosed as having congenital hypothyroidism, NAA/Cho was below the 95% confidence limits. The OP region was not studied in this infant. There was a trend toward a reduced thalamic NAA/Cr in the asphyxiated infants (seeTable 3).

DISCUSSION

The purposes of this study were to define normal changes during the perinatal period in cerebral 1H-metabolite ratios and then to test whether abnormalities in these ratios were present after presumed hypoxic-ischemic injury, usually due to birth asphyxia. The studies could be performed in under 1 h(11), and it was possible, if necessary, to continue full intensive care within the bore of the magnet.

Two regions of the brain were studied, varying in their degree of maturation and differing in their susceptibility to hypoxia-ischemia. The thalamic voxel was largely occupied by both thalami but also contained adjacent structures such as part of the corpus callosum, the pineal, midbrain tectum, and hypothalamic nuclei. Overall this voxel contained predominantly gray matter. Estimates of the proportion of third ventricle cerebrospinal fluid present, derived from water relaxation measurements in eight control infants, gave a mean value of 5.9 (SD 6.0)%. The OP voxel contained mainly white matter, including that of the periventricular region and the posterior corona radiata. The mean proportion of lateral ventricle cerebrospinal fluid was estimated from six control infants to be 1.4 (2.1)%.

Methods for spectroscopy. The identification of Lac in spectra from both normal preterm and term infants was supported by its chemical shifts in the thalamic [mean 1.32 (SD 0.01) ppm; n = 29] and OP [1.31(0.02) ppm; n = 21] regions(16). The small SD indicate that these signals were from Lac rather than due to fat contamination or noise, for which greater chemical-shift variability would be anticipated. Previous investigators have not detected Lac in the brains of normal term infants. It is necessary therefore to consider the methods used in the present study in some detail. Lac detectability was optimized by the asymmetry of the PRESS sequence(8) and use of the optimum TE. At TE 270 ms the Lac doublet has a phase similar to a singlet resonance(17). At TE 135 ms, due to inhomogeneous flip angles throughout the voxel, the phase-modulated inverted Lac signal is smaller than at TE 270 ms(18, 19) (see Fig. 3). The use of a TE less than 135 ms was not contemplated; such spectra are complicated to analyze owing to overlapping multiplet resonances and broad underlying signals from macromolecules(20).

Failure to detect cerebral Lac in previous studies of term infants was probably due to inadequate sensitivity of the methods used. Several groups have used STEAM localization to study the neonatal human brain(2125); when compared with this method, PRESS has an intrinsic 2-fold increase in SNR(19, 26), providing a significant advantage for detecting low molarity metabolites. Although phase modulation is different for the two techniques, detection of the Lac doublet by STEAM is also handicapped by this effect at short TE. In STEAM studies with TE of 20 ms(22) and 7 ms(24) Lac detection was made more difficult by fat contamination, which was also a significant factor in early studies utilizing spin-echo localization with TE 270 ms(6, 27). In one of these studies, a binomial pulse, optimized for NAA, was used for water suppression, further compromising Lac detectability(6). In a recent STEAM study with TE 272 ms, the SNR was less than in the present report and cerebral Lac remained undetected(25). Lac was also frequently undetected in term infants with birth asphyxia in a PRESS study with TE 270 ms despite the absence of fat contamination(5).

The regular detection of Lac in our spectra was probably due to the use of an asymmetric PRESS pulse sequence, along with the intrinsic 2-to-1 SNR advantage of PRESS compared with STEAM, and a head coil designed specifically for spectroscopy of newborn infants(11), further maximizing SNR and spectroscopic resolution, while minimizing sources of contamination. The 270-ms TE used in the present study, although optimal for the detection of Lac, was suboptimal for the detection of glutamate/glutamine and myoinositol/glycine. Shorter TE would be more suitable for defining these metabolites.

The observations reported here are of peak-area ratios, which reflect concentration ratios, but can under some circumstances also be affected by changes in metabolite relaxation times. Absolute concentrations were not measured in this study as, until recently, the methods have been uncertain, controversial, and time-consuming.

Control infants. The range of GPA studied (31-42 wk) was relatively narrow. Nevertheless, maturational changes in peak-area ratios were found. Previously, Lac has been found in the brains of some preterm and small for gestational age infants, but not in normal term infants(4). In our study, Lac/NAA and Lac/Cr fell with increasing gestation in both regions studied, and Lac/Cho also fell in the OP region. Metabolite relaxation times are thought not to change significantly with gestation(21), and the most likely explanation is that there was a real fall in Lac concentration. Our preliminary estimates of absolute quantities of Lac, using brain water as an internal reference, indicated both thalamic and OP concentrations of ≈3 mmol·kg-1 wet weight in 12 normal infants of 34-42 wk of GPA(11). Others have estimated 0.9-2.4 mM in healthy preterm and small for gestational age infants(4). Lac is present in the normal adult brain but, for example in the visual cortex, has a concentration less than 1 mmol·L-1 even during photic stimulation(28).

One explanation for the increased Lac in the immature brain is that it is being produced in peripheral tissues and used as a cerebral metabolic fuel(4). We did not measure blood Lac in our control infants, but mean levels in healthy infants beyond the first day of life, both preterm and term, are less than 2 mmol·L-1(29), consistent with the view that the cerebral Lac observed in this study resulted from local production in the brain, rather than diffusion from peripheral sources. The implication is that with decreasing gestation the brain becomes more dependent for energy generation on glycolysis relative to mitochondrial respiration. The cerebral [phosphocreatine]/[inorganic phosphate] ratio, which is an index of phosphorylation potential, falls with decreasing gestation(30), a finding which fits well with the concomitant increase in Lac found in the present study. Studies of animals have shown that the immature brain contains relatively fewer mitochondria(31), and has a lower concentration of cytochrome oxidase(32) and other respiratory chain enzymes(33), than the adult brain, whereas the enzymes of glycolysis develop before those of mitochondrial respiration(33). The cerebral metabolic rate for oxygen of the human neonatal brain has been found to be below the threshold for viability in adults and was lower in preterm than term infants(34). The energy requirements of the immature brain appear low, and glycolysis may be the principal mechanism by which they are met.

When data obtained simultaneously from the thalamus and OP region were compared (Table 2), the finding that Lac/Cho, Lac/Cr, and Lac/NAA were all much higher in the OP region might be taken to indicate that mitochondrial oxygen consumption was reduced there. This would be a particularly interesting conclusion, because this region is a major site for periventricular leukomalacia. However, our preliminary quantitative data indicated that the Lac concentration was similar in both regions(11), whereas the concentrations of the other metabolites were much lower in the OP voxel than in the thalamus (our unpublished observations). This could also explain why the SNR was consistently greater in the thalamus than in the more peripheral voxel (Fig 2). The OP region is relatively immature; myelination does not start there until a corrected age of at least 2 mo, but is well established in the thalamus at term(9, 35). Further investigations are required to resolve these issues.

The increase in NAA/Cho with GPA, which took place more rapidly in the OP region than in the thalamus (Fig. 6), is consistent with the results of previous neonatal studies, which, however, included abnormal as well as normal infants(21, 23). More recently, Hüppi et al.(24) found increases in the concentration of NAA with postconceptional age in the precentral area of the cerebrum in healthy infants using both 1H MRS with creatine from autopsy tissue as an internal standard, and chromatographic data from autopsy tissue. Although the two methods agreed in the age-dependent change, they yielded different concentrations; the 1H MRS quantitation method gave significantly higher values. Apart from its occurrence in oligodendrocyte type 2A progenitors(36), NAA is exclusively neuronal(37). The increase in NAA/Cho found in the present and other(21, 27, 38) studies probably largely reflects neuronal development.

No significant changes in peak-area ratios were found with postnatal age. Our control infants were studied at a median age of 9 d, and only 7 were over 3 wk of age. This sample would therefore be more sensitive for detecting postnatal changes occurring within the first few days of life, and no such changes were found, but less sensitive for detecting maturational differences emerging over several weeks resulting from extrauterine as opposed to intrauterine development.

Asphyxiated infants. Lac/NAA was selected for particular study in the asphyxiated infants because earlier investigations had given evidence that Lac was sometimes raised(5, 39) and NAA sometimes reduced(5, 6) after birth asphyxia. It was therefore hypothesized that the ratio of the two metabolites would be a particularly good marker of injury. It is also for this reason that, in asphyxiated infants studied more than once, peak-area ratios in which Lac might be maximal, or NAA minimal, were selected for analysis. The data summarized in Figure 5 show that Lac/NAA was indeed often increased in the thalamus, but much less often in the OP region. The thalamus is known to be at particular risk from acute asphyxia(1), which may explain this finding. The lack of any striking changes in the OP region may be partly because the voxel did not include the parasagittal and subcortical regions, which are especially susceptible to injury in term infants. Another possibility is that the lower SNR in this region may have accounted for a lack of sensitivity.

Previous investigators have concluded that the detection of Lac(5) or an apparently increased amount(39) often indicated a poor outcome. A failure to detect Lac was not, however, an indicator of a good prognosis(5), and interpretation of the findings was made difficult by lack of normal controls. In our study a significant association was found between Lac/NAA values above the 95% confidence limits and death or serious disabling neurodevelopmental impairments at 1 y of age. If substantiated by follow-up of further infants, these data may provide information with adequate sensitivity and specificity on which to base estimates of prognosis.

Elevated cerebral Lac/NAA was detectable early after hypoxia-ischemia; it was present at 2 h of age in the infant whose spectra are illustrated inFigure 7. In two further birth-asphyxiated infants who were studied early, one aged 6 h and the other 12 h, Lac/NAA values were above 95% confidence limits in the thalamus, but the infants' blood Lac levels immediately before study were raised at 3.0 and 3.4 mmol·L-1, respectively, evidence that Lac generated during the acute asphyxial episode had not been eliminated. Both were graded Sarnat stage I. At further study 2 d later, both had normal Lac/NAA, and both were discharged home within 10 d of birth, feeding normally. They were still less than 1 y old when last seen, so their outcome had not yet been formally assessed. Studies of acute reversed cerebral hypoxia-ischemia in the newborn piglet have shown that cerebral Lac increased sharply during the insult and then fell back close to baseline over the next few hours. Subsequently, a second increase took place by 24-48 h which was associated with the development of delayed or“secondary” energy failure and was not accompanied by a fall in intracellular pH(40, 41), nor a high blood Lac level (our unpublished observations). Secondary energy failure is known to carry a grave prognosis in birth-asphyxiated human infants(3). It remains to be shown whether elevated cerebral Lac in the first few hours of life carries the same adverse prognosis as later elevation which appears to indicate neuronal compromise.

Other investigators have suggested that reduced NAA/Cho, attributed to neuronal loss, also carries a poor prognosis(5, 6). The studies on which this conclusion was based were performed at mean ages of 7(5) and 14(6) d. Our results, obtained on infants generally studied earlier, and with the benefit of age-matched controls, showed no convincing reduction in NAA/Cho, although there was a tendency for NAA/Cr to be low in the thalamus. Elevated values for NAA/Cho were found in two infants (Fig. 6), possibly due to the marked increase in the T2 of NAA, which takes place during secondary energy failure(42). Lack of an early fall in NAA/Cho might be taken as an indication that neuronal loss has not yet occurred, thus suggesting an opportunity for cerebroprotective therapy. However, as peak-area ratios depend not only on substrate concentrations but also on metabolite T2, a reduced NAA concentration may to some extent have been masked by an increased NAA T2.

The one infant in whom a low NAA/Cho was found was later diagnosed as suffering from congenital hypothyroidism (Fig. 6). He also had clear evidence of birth asphyxia and a raised cerebral Lac relative to NAA, Cho, and Cr. The significance of the low NAA/Cho is unclear, but might indicate antenatal neuronal compromise.

We conclude that, in control infants, Lac/Cho, Lac/Cr, and Lac/NAA were all significantly greater in the OP region than in the thalamus. Significant falls in Lac, relative to the other metabolites, were found with increasing maturity, whereas NAA/Cho increased; these changes were more pronounced in the OP region. Raised Lac/NAA, but not decreased NAA/Cho, was detectable soon after perinatal hypoxia-ischemia and was associated with an adverse outcome at 1 y of age. Longer follow-up of a larger group of asphyxiated infants will be required before the clinical role of 1H MRS as a prognostic indicator can be fully defined.