Main

Birth asphyxia among term infants remains a considerable problem in perinatal medicine with high rates of mortality and handicap (1–4). The clinical grading of HIE after perinatal asphyxia has become a useful prognostic tool in the term infant (5–7). Mild HIE usually results in a normal outcome and severe HIE in either death or CP. However, moderate HIE can result in either CP or a normal outcome (8). Early prediction of future handicap is important when selecting infants for early interventions, which may be possible in the near future (9, 10).

Several techniques have been used to investigate the cerebral pathology related to perinatal asphyxia. Cerebral morphology has been evaluated with computed tomography (CT) (11) and magnetic resonance imaging (MRI) (12, 13). Methods capable of detecting functional disturbances, such as phosphorus (31P) magnetic resonance spectroscopy (14) and proton (1H) magnetic resonance spectroscopy (15), have been used to investigate changes in brain energy metabolism during the evolution of the secondary energy failure (16).

PET, measuring either cerebral blood flow (17) or CMRgl, has been used in a few studies on newborn infants. The first PET studies with FDG in newborn infants did not allow any quantification, making only relative values possible (18, 19). Phelps et al. (20) modified the Sokoloff et al. (21) method and replaced arterial blood samples with venous, which made it easier to quantify CMRgl also in children. However, CMRgl has been quantified by only a few investigators in newborn infants (22, 23) due to the required number of blood samples (20). In Finland, a semiquantitative analysis was used to measure SUV of glucose (24) and also an alternative method for quantification (25) when used for investigating CMRgl in early infancy. In the present study, we used a new method to quantify CMRgl with PET, based on the glucose metabolism of the erythrocytes (26).

The aim of the present study was to correlate cerebral glucose metabolism in the subacute period after asphyxia with the neurologic clinical scoring (HIE) of the term newborn infant after perinatal asphyxia and to quantify total and regional CMRgl in the neonatal brain. We also aimed to investigate the preliminary outcome of the infants and to correlate the CMRgl with the semiquantitative method, SUV multiplied by the glucose concentration in the blood (SUV × Bglc).

MATERIAL AND METHODS

Subjects.

Twenty term infants born January 1995–April 1997 with HIE after birth asphyxia were investigated. The infants were treated in the neonatal intensive care unit at the University Hospital in Lund, Sweden, a tertiary-level neonatal unit. The following inclusion criteria were used:1, term newborn (gestational age ≥ 36 wk);2, birth asphyxia: Apgar < 7 at 5 min;3, severe umbilical cord acidemia and/or late or prolonged decelerations on cardiotocography and/or placental abruptio; and 4, early postnatal signs of HIE (grade 1–3) (5). One infant with a 5-min Apgar 7 was included due to late decelerations on cardiotocography and moderate HIE for 7 d with seizures.

Nine infants were inborn and eleven were outborn and referred from hospitals within the southern region of Sweden during the first day of life. Clinical data on the infants are shown in Table 1. Fifteen of the 20 infants had a severe perinatal acidemia with a pH < 7.05 and/or base excess < − 12 mmol/L, measured in either umbilical cord blood (arterial or venous) or the umbilical catheters or capillary blood from the infant within 1 h after delivery. One infant was small for GA and two were large for GA (27). Eleven infants needed mechanical ventilation during the first days of life. Eight infants required dopamine treatment due to arterial hypotension (mean blood pressure < 30–35 mm Hg) on the first day of life. All infants had normal values of C-reactive protein the first day of life. Nine infants developed seizures, all within 24 h after birth. On the day of the PET scan, none of the infants had seizures, but three infants were still on treatment with antiepileptic drugs, two with phenobarbital and one with phenytoin. Two infants needed supplementary oxygen treatment on the day of the PET examination.

Table 1 Clinical characteristics of the infants*
Full size table

The present study was approved by the Research Ethics Committee and the Radiation Safety Committee at Lund University Hospital and by The Swedish Medical Products Agency. Informed consent was obtained from all parents.

Clinical scoring.

All 20 infants were examined daily with clinical neurologic scoring for HIE according to Sarnat and Sarnat (5) for the three grades:1, mild;2, moderate; and 3, severe. Six infants developed HIE grade 1, twelve HIE grade 2, and two HIE grade 3. The clinical neurologic condition of the infants with HIE grade 1 had on average normalized by postnatal d 3. In infants with HIE grade 2 and 3, clinical signs of HIE remained for 3–15 d. The infants were video recorded during the uptake period of FDG (during the injection and for 10–15 min). The videotapes were later visually evaluated according to alertness and motor activity and the state of the infants divided into three categories, modified after Prechtl (28). The three categories were 1, quiet asleep, eyes closed;2, quiet awake, few movements, eyes open; and 3, awake, more movements, and/or crying.

PET scanning.

PET scans were performed at postnatal age 4–24 d (median, 11 d). Twelve of the infants were investigated on d 8–12. GA at the PET examination was 37–44 wk (median, 41 wk). The PET scanner (PC384-7B, Scanditronix, Uppsala, Sweden) has four detector rings, generating four main and three cross-slices, thickness approximately 10–11 mm. The sensitivity is 675 cps/kBq/mL for the main slices and 945cps/kBq/mL for the cross-slices, and the spatial resolution is 7–8 mm. The axial range is 81 mm. We used an image matrix of 128 × 128 pixels with a pixel size of 2.2 × 2.2 mm2. Corrections for random, scatter, and attenuation were done (29–31). Dead-time correction was also applied.

The 2-18FDG was synthesized by direct nucleophilic exchange on a quarternary 4-aminopyridinium resin (32). 18F was produced at the Lund electrostatic tandem accelerator with a proton beam of 6.0 MeV and produced by the 18O (p,n) 18F reaction (33). The radiochemical purity of 2-18FDG was better than 97%.

Administration of 2–3.7 MBq (50–100 μCi) of 18FDG per kilogram body weight was done during a 1-min injection into either a central or peripheral venous catheter. All infants had fasted for at least 2 h before the injection. Approximately 25 min after the injection, the infants were fed either by breast-feeding or bottle-feeding or by a nasogastric tube.

The infants were placed in a supine position on a warm water mattress (37°C), and the head was positioned in the PET camera lightly fixed on a vac-pac containing polystyrene beads.

Most infants were asleep during the measurement. No sedation was used. During the PET scan, the infants were monitored clinically by a pediatrician and a neonatal nurse and, if necessary, with pulse oximetry. The PET scanning started approximately 60 min after FDG administration and consisted of two separate 10-min scans, in between which the couch was moved half a plane separation (6.75 mm) to get a higher axial sampling. CMRgl values were calculated using a modified autoradiographic formula based on the deoxyglucose model, which was also used by Rhodes et al.(34) :MATH where MRgl indicates metabolic rate of glucose, Cgl is blood glucose, and C*i is the amount of FDG-6-PO4 in an ROI. T is the time-point postinjection for the PET scan. The LC used for correction between the difference in metabolism of deoxyglucose and glucose is set to 0.52 for brain and assumed to be constant over time (35). C*p (t)dt is the plasma concentration of FDG. To obtain the plasma integral of FDG, the erythrocytes were used as reference tissue (26). After the PET examination, one venous blood sample of 1 mL was drawn from the same catheter (possible in this method) used for the injection (26). The value of the integrated input function at the time for the PET scan to the sample time was calculated using a monoexponential function derived from a population-based venous blood curve (26). We also calculated the SUV, which has been used as a more simplified method, semiquantification. Using the SUV method, the Bglc during the PET has to be taken into account (36). MATH

Radiation dosimetry.

The absorbed dose was calculated according to the MIRD models based on the distribution data of Gallagher et al.(37), assuming a 1-h bladder-voiding interval. The bladder wall receives the highest absorbed dose, 0.8 mGy/MBq. The estimated effective whole-body dose equivalent was 0.16–0.17 mSv/MBq.

Regional cerebral glucose metabolism.

The acquired PET recordings of CMRgl were quantified by the use of ROI with a subdivision in anatomical regions based on an anatomical atlas (38). To match this anatomical atlas, the 14 brain slices from PET with 6.75-mm spacing were converted to 10 slices with ∼10-mm spacing, parallel to the orbitomeatal plane and starting at the orbitomeatal plane and continuing upwards. The ROI we chose were frontal, parietal, temporal, and occipital cortex, cerebellum, and subcortical gray matter (including basal ganglia and thalami).

Statistical analysis.

The correlation between total mean CMRgl and the Sarnat and Sarnat (5) score was tested by the Kendall rank correlation and Kruskal-Wallis test. The results of the ROI analysis of the regional CMRgl distribution were submitted to a factor analysis (Varimax, oblique rotation) to redefine the ROI into a smaller number of functionally correlated factors to decrease the risk of mass significance at testing. The factor scores (relative CMRgl values) were then tested against the Sarnat and Sarnat (5) score by Kendall rank correlation test. Correlation between CMRgl and SUV × Bglc was tested by the Pearson correlation. Correlation of CMRgl with GA and outcome was evaluated with nonparametric tests (Spearman and Wilcoxon).

RESULTS

Total CMRgl decreased with increasing severity of HIE and was inversely correlated with the degree of HIE (p = 0.005) (Fig. 1 and Table 2). In all infants, the PET scan showed that the most metabolically active brain areas were the deep subcortical parts, thalamus, basal ganglia, and sensorimotor cortex, whereas the cortical areas (frontal, parietal, and occipital regions) were less metabolically active (Fig. 2). There was a significant inverse correlation between the Sarnat score and the CMRgl also in the cerebellum, thalamus, and basal ganglia and the frontal cortex (p < 0.01) (Table 2). The regional CMRgl distribution obtained by the factor analysis contained two parts representing factor 1, the subcortical parts and the temporal cortex, and factor 2, the cortical parts of the brain. The regional CMRgl distribution did not correlate significantly with the clinical prognostic parameter (HIE). No region correlated significantly more than any other did in the method we used.

Figure 1
figure 1

Plot of CMRgl in 19 term infants according to the degree of HIE and outcome at 2 y of age. 1, mild HIE;2, moderate HIE;3, severe HIE.

Full size image
Table 2 Total and regional CMRgl and SUV multiplied by Bglc (SUV × Bglc) in 20 infants with HIE
Full size table
Figure 2
figure 2

CMRgl measured in the subacute period (10–11 d) after birth asphyxia in three infants with different degrees of HIE. Regional CMRglc in cerebellum (top), thalamus (middle), and sensorimotor cortex (bottom). The infant to the right (HIE-3) has developed CP (tetraplegia) with complicated seizures. The infant to the left (HIE-1) is healthy (2-y follow-up). The scale of the different images is identical.

Full size image

The SUV multiplied by the Bglc correlated significantly with the HIE grade (p < 0.05). There was a significant correlation between CMRgl and SUV × Bglc (r = 0.73, p < 0.01).

CMRgl increased with GA in all groups of HIE. Correcting for GA gave an even higher significance in the correlation between CMRgl and the HIE (p = 0.004). The global CMRgl correlated significantly with GA (weeks) at PET (p < 0.05) but not with the postnatal age (days) at PET (p > 0.2).

There was a significant correlation also between CMRgl and the outcome at 2 y (p < 0.05). The correlation of CMRgl with GA (weeks) and outcome is presented in Figure 3. At follow-up (2 y), all six infants with mild HIE were neurologically normal, but four infants with moderate HIE and the two with severe HIE developed CP. In five of the six infants with CP, quantitative values of CMRgl were obtained, with mean/median (range) 18.1/15.2 (10.2–31.4) μmol·min-1·100 g-1, to compare with mean/median (range) 41.5/37.7 (13.0–100.8) μmol· min-1·100 g-1 in the 13 infants with no neurologic sequela at 2 y of age. In one infant with HIE-2 who developed CP, we failed to quantify CMRgl. One infant with moderate HIE who did not develop CP but epilepsy and visual impairment had a CMRgl of 20.2 μmol·min-1·100 g-1. In the group of infants with moderate HIE, the three infants who developed CP had CMRgl, mean/median (range) 21.8/18.9 (15.2–31.4), to compare with the seven infants who by 2 y of age have not developed CP or epilepsy, mean/median (range) 29.5/24.0 (13–65) μmol· min-1·100 g-1.

Figure 3
figure 3

CMRgl measured by PET in term infants with HIE after birth asphyxia, correlated with GA at PET, and outcome. Plots represent the infants and fit lines, the outcome groups.

Full size image

The relative glucose metabolism in the frontal motor area (% in the frontal motor area of total CMRgl) was significantly correlated with the motor activity scored by video registration in 14 of the infants during the uptake period of FDG (p < 0.02). The volume of the frontal motor area was 6% (SD 3) of the total brain volume and did not significantly influence the global CMRgl. The absolute regional glucose metabolism in this frontal motor area was not significantly correlated with the independent motor-activity score (p = 0.19).

DISCUSSION

The main result of this study was the highly significant correlation of CMRgl, measured during the subacute phase after perinatal asphyxia in term infants, with the degree of HIE. The more severe the degree of HIE, the lower was the total CMRgl (Fig. 1). Our findings (Fig. 2) that the deep subcortical parts, thalamus and basal ganglia, and the sensorimotor cortex, all phylogenetically old structures of the brain, were the most metabolically active in newborn infants are consistent with previous observations by Chugani and Phelps (39).

We found a global decrease of CMRgl in infants with moderate and severe HIE compared with infants with mild HIE. However, the subcortical regions still had the relatively highest metabolism also in the sickest infants (see Fig. 2). The ROI set we used was developed for adults and accordingly was not optimal for infants. The main difference in proportion between infant and adult brain is the location of the border between the frontal and parietal lobes. Because the CMRgl in the frontal and parietal areas were uniform and approximately equal with the exception of the Rolandic areas, which were analyzed separately, the error in location of the border between lobes presumably had little effect on the results. Due to the limited precision of the ROI analysis, we do not attempt to draw any conclusion of the CMRgl in subdivision of the main brain lobes such as the vascular border areas, as has been done for cerebral blood flow, using PET with 15O-H2O (17).

All infants with mild HIE had high values of CMRgl and did not develop CP. Infants who developed CP had moderate or severe HIE and decreased CMRgl. One of the three infants with moderate HIE who had low CMRgl (20.2 μmol·min-1·100 g-1) but no CP developed seizures and visual impairment. The combination of relatively early gestation and severe illness with need of medication and ventilator due to pulmonary complications could explain the very low metabolism in the two infants with moderate HIE who have not developed CP by 2 y of age (CMRgl, 13.0 and 13.6 μmol·min-1·100 g-1, respectively).

Three infants, all in HIE grade 2 and 3, were on antiepileptic drugs on the day of PET, and all developed CP. One with the highest value of CMRgl (31 μmol·min-1·100 g-1) among infants with CP had phenytoin. Two infants with CMRgl 10 and 15 μmol·min-1·100 g-1, respectively, had phenobarbital. The doses of phenobarbital given in the infants were much lower than in the animal studies in which they were shown to decrease CMRgl and, therefore, are unlikely to have caused a major depression of CMRgl, although studies using PET on patients with phenobarbital as epilepsy treatment (40) showed an increase in brain glucose metabolism after withdrawal of barbiturates.

Most investigators have not been able to quantify CMRgl in newborn infants due to the large number of blood samples the method requires (20, 21). The common way to describe CMRgl in infants has, therefore, been to look at relative values. Ratios to the cerebellum and thalamus have been used (39). A Canadian group, who made the first FDG-PET in newborns, used visual inspection (18) and ratios to average cortical activity and cortex/white matter ratios (19). Our quantitative results with decreased metabolic rate of glucose in the whole brain, even in the deep subcortical parts, thalamus, basal ganglia, and cerebellum after asphyxia, suggest that some regions could falsely be interpreted as hypermetabolic when using a method employing relative values only. Consequently, to be able to discuss hypo- and hypermetabolism in different regions of the brain, the absolute values for CMRgl have to be known.

Only a few investigators have used the Sokoloff-Phelps quantification method in children (22, 23). Other investigators have used the Patlak method combined with measuring the FDG content over the heart (25), which requires only a few blood samples but necessitates a body scanner.

We also calculated the semiquantitative method SUV, which previously has been used for evaluation of newborn infants (24). When using SUV, it is important to also take the Bglc into account (36). SUV multiplied by the Bglc gave a good accuracy of CMRgl in our study.

There are indications that early after hypoxia-ischemia (HI), there may be an increase in CMRgl. Experimental studies using the Vannucci et al. (41) model of neonatal rats have shown an increase in CMRgl 3–12 h after HI (42). In five term infants with HIE investigated by PET 2–3 d after asphyxia, CMRgl was increased, especially in the basal ganglia, in infants with a poor neurologic outcome (23). Hypermetabolism can also be caused by seizures when measured during the ictal phase (43). None of the infants in our study had seizures on the day of the PET scan.

The absolute values of CMRgl, especially in the less-affected infants with mild HIE, were higher in our study than in previously published studies (22, 23, 44). Infants with mild HIE, who usually develop normally (8), had a median value of total CMRgl of 48 μmol·min-1·100 g-1 (range, 38–101) (Fig. 1). These infants are the closest we can get to normal controls as, for ethical reasons, normal infants could not be investigated. The new method we used for quantification is only validated in adults (26). We cannot exclude the possibility of some instability in the method due to the small amount of blood (1 mL) we used. Another factor that could influence the results is the LC used for correction between glucose and deoxyglucose. The LC we used is set to be 0.52 for adult normal brain tissue (35, 45). Developing brains of infants may have another LC and, in addition, LC for ischemic and partly necrotic tissue can also be different (46). The Bglc may also have some influence on the LC. Also, minor changes in the LC could affect the finally calculated CMRgl.

There are, however, reasons to believe that the high values of CMRgl in the normally developing infants are correct. All infants were examined without any sedation, which is in contrast with other studies (22, 23, 47). Furthermore, at GA of 37–44 wk, when our study was performed, there is a very rapid speed in brain growth and myelination, which requires high metabolism (48). Infants have, as in our study, the highest metabolism in subcortical structures, in the thalamus, basal ganglia, and in the white matter, in contrast with adults in whom the cortical areas have the highest metabolism (22). Recently, an abrupt 5-fold increase in the volume of myelinated white matter in the brain between 35 and 41 wk of gestation has been documented with three-dimensional MRI (49).

A study from Finland showed that total CMRgl in infants (GA, 35–60 wk) increases with maturation and correlates with postconceptional age (44). The only study of CMRgl in preterm infants showed very low values (7.4–11.4 μmol·min-1·100 g-1) at 4–7 d in five infants of GA 25–34 wk (47). Even within the short interval over which our study was performed (37–44 wk), we could see an increase of CMRgl with GA at PET. The increase of CMRgl with GA was apparent in both groups of outcome, the impaired and the unimpaired (Fig. 3). These findings show that CMRgl in newborn infants increases with maturation and that the GA has to be taken into account when interpreting the results.

We analyzed the state of alertness and motor activity of the infants during the uptake period of FDG to determine whether there was any influence of the activity on the glucose uptake in the brain. We found no significant correlation of the activity score with the absolute glucose utilization in the Rolandic areas but with the relative glucose uptake. However, this part constitutes only 6% of the total brain volume and could possibly only marginally influence the total CMRgl.

In the future, specific therapy to reduce the cerebral injury of HI may be possible (9, 10). To prognosticate the impairment after perinatal asphyxia and estimate the cerebral lesion (1), early investigation of the brain is essential. It is, therefore, very important to identify the infants who are at the highest risk of being neurologically impaired. The PET technique gives further insight into the cerebral metabolism after asphyxia and could perhaps even be a complement to other methods (50) in the prognostication of neurologic handicap in neonates with HIE after perinatal asphyxia. However, PET is a rather expensive technique and not easily available. The absorbed dose of radiation the infant receives from FDG is rather small and comparable with the radiation from a CT scan of the brain.

The great advantage of the PET technique is the acquisition of functional information, in contrast with MRI (12, 13), which only gives morphologic information. PET gives information on the active brain, and the results are not blurred by any sedation. In combination with magnetic resonance spectroscopy (14, 15) and diffusion-weighted MRI (51), other functional modalities available at some hospitals today, PET can be an important complement in investigating the function of the brain after HI.