Main

Preterm infants who have immature lungs and require intensive care are exposed to more reactive oxidants than term infants. These can originate from the high concentrations of oxygen they require, xanthine oxidase activation during episodes of hypoxemia and reoxygenation (1), and neutrophil activation associated with inflammation (2–6). Free radical generation and oxidative injury are strongly implicated in CLD (also referred to as bronchopulmonary dysplasia), ROP, and IVH, the major complications of prematurity (1, 2, 7–9). These conditions have been termed the oxygen radical diseases of prematurity to emphasize their probable common pathogenesis (1).

Evidence for oxidative injury comes predominantly from measurements of biochemical markers of lipid peroxidation and protein oxidation. These studies have shown that oxidation products can be measured in plasma (10–14), urine (14–16), lung aspirates (17, 18), or breath (19–23) of premature infants. In a number of these studies, concentrations were higher in the lowest birth weight, high-risk infants (20–22), and were elevated either after hypoxic insult (12, 23–26) or in association with high Fio2 (15–18). However, direct evidence that oxidation plays a causative role in the diseases of prematurity is limited. Several groups have reported higher concentrations of oxidant markers in association with CLD (10, 13, 16, 26). Others found a positive association with mortality, ROP, and IVH, but not with CLD (23). However, it is not always clear whether these associations just reflect a higher disease incidence in the lowest birth weight infants who receive more oxygen. There is also, in some cases, uncertainty about assay specificity. For example, it is not clear whether the TBA assay we used previously to show an association with CLD that persisted after correction for birth weight (10) is a true measure of lipid peroxidation (see below). Further work is needed, therefore, before we can be certain whether oxidants contribute to the pathogenesis of these neonatal diseases.

If oxidants are important, then antioxidant supplementation has the potential to decrease oxidative injury and improve outcome. There is some evidence that vitamin E may decrease the severity of both IVH and ROP, although no clear benefit has been shown in the case of CLD (27, 28). There have been few other trials of antioxidant supplementation, and the question of whether biochemical oxidant markers are affected by antioxidants has not been investigated. Selenium, as a component of the glutathione peroxidases, is an important antioxidant (29). Very low levels have been found in premature infants (30–32), potentially putting them at risk of oxidative injury. We have recently carried out a multicenter controlled trial of selenium supplementation of preterm infants during the first month of life (33). Rather surprisingly, we found that supplementation gave no significant protection against CLD or ROP. We have now examined whether supplementation has an effect on oxidant markers.

The present study was undertaken to establish whether there is evidence for systemic lipid or protein oxidation in premature infants by measuring oxidation products in plasma. We have determined whether there are associations between concentrations of these markers and CLD or ROP and also whether they are affected by selenium supplementation. The study population was a subgroup of the infants involved in the selenium trial and included all infants cared for in two centers. Lipid peroxidation was assessed by an HPLC-based TBA assay that is considered to measure predominantly protein-bound MDA (34). As an index of protein oxidation, we measured protein carbonyls using a sensitive ELISA (35). Carbonyl groups are produced on proteins when they are oxidized, but can also represent covalently bound aldehyde products of lipid peroxidation (36, 37).

METHODS

Patient details.

All infants with birth weight <1500 g and admitted within 48 h of birth to the Neonatal Intensive Care Unit at Christchurch Women's or Dunedin Hospital between 1994 and 1996 were eligible for study after their parents had signed informed consent. The study received approval from the Southern Regional Health Authority (Canterbury and Otago) Ethics Committees. There were 173 infants enrolled (136 in Christchurch and 37 in Dunedin), this being 78% of eligible infants. All infants were treated in accordance with standard protocols in the unit. Oxygenation was monitored by continuous pulse oximetry (Nelcor N100 & N200; initial target oxygen saturation was 87–95%) and frequent blood gas analysis. Arterial oxygen tension was maintained at 55–75 mm Hg. The hospital protocols called for antenatal glucocorticoids (three 8-mg doses of dexamethasone at 8-h intervals) to be administered to women in preterm labor at <34 wk gestation when possible. Postnatal steroids were administered at the neonatologist's discretion to infants of >7 d of age with Fio2 >50% and high mean airway pressure to aid weaning from positive-pressure ventilatory support. In these infants, dexamethasone was started at 0.5 mg/kg/d, then given at reducing doses during a 15- or 42-d course (mean dose, 0.08 mg/kg/d).

For infants unable to tolerate oral feeds, total parenteral nutrition was commenced on d 2 to 3 with 1 g/kg/d Vaminolact (Kabi Pharmacia, Auckland, New Zealand) and 10–15% dextrose. Intralipid 20% (Kabi Pharmacia) was commenced on d 3 to 4 at 0.5 g/kg/d. Both were increased gradually as tolerated to a maximum of 3 g/kg/d. Supplementary trace elements (Ped-El, Kabi Pharmacia, Stockholm, Sweden; 4 mL/kg/d) and vitamins (MVI-Paediatric, Rhone-Polenc Rorer, Lower Hutt, New Zealand; 2 mL/kg/d) were routinely added to total parenteral nutrition. In a double blinded fashion, infants were randomized to receive either selenium or a placebo, as described in detail elsewhere (33). Parenteral supplementation was with 7 μg/kg/d sodium selenate, and oral supplementation was with 5 μg/kg/d sodium selenite. Supplementation was continued until 36 wk PMA or hospital discharge if earlier.

Biochemical analyses.

Blood samples (0.6 mL) were collected into heparin at the time of routine sampling, and often from an indwelling arterial or venous catheter, before randomization for selenium supplementation (mean, 2.6 d), at 7 and 28 d after birth, and at discharge or 36 wk PMA. Samples were placed immediately in a dark refrigerator, and plasma was separated within 12 h and stored at −80°C until analysis. This procedure is acceptable for MDA analysis (38), and we have also found protein carbonyls to be stable for years at −80°C and at least a day at room temperature.

Plasma samples were heated with TBA, and the TBA-MDA adduct was analyzed by HPLC with fluorescence detection (39). Protein carbonyls were determined by an ELISA involving reaction of the protein with dinitrophenylhydrazine and detection with an anti-dinitrophenylhydrazine antibody (35). Total protein concentrations were measured for reference using the Bio-Rad assay (Bio-Rad Laboratories, Richmond, CA, U.S.A.). Plasma selenium was analyzed using an atomic absorption spectrometer (Varian Spectra AA40, Varian Industries, Sunnyvale, CA, U.S.A.) with Zeeman background correction. Glutathione peroxidase was measured with t-butyl hydroperoxide as substrate on a centrifugal analyzer (40). To assess relationships between different TBA assays, a selection of 2- and 7-d plasmas were also analyzed by the HPLC method of Wade and van Rij (41), in which lipids are extracted and heated with TBA in the presence of iron to generate the MDA-TBA complex. Plasma fatty acid profiles were also determined by gas chromatography after methylation.

Outcome measures.

Primary outcome measures, as for the randomized controlled trial of selenium supplementation (33), were defined as oxygen dependency at 28 d of age, and total number of days of oxygen dependency. Secondary outcome measures included death, death from d 7 to d 28 or oxygen dependency at 28 d of age, oxygen dependency at 36 wk PMA, death from d 7 to 36 wk PMA or oxygen dependency at 36 wk PMA, and ROP. ROP was assessed according to our routine screening protocol (42), with infants having an initial examination at 6 wk of age by an ophthalmologist skilled in indirect ophthalmoscopy. Retinopathic changes were recorded using the International Classification of Retinopathy of Prematurity (43).

Statistical analysis.

Statistical analyses were performed using SigmaStat (Jandel Scientific, San Rafael, CA, U.S.A.). Differences between groups were assessed using the Mann-Whitney or Kruskal-Wallis test. Relationships between variables were assessed using linear regression when the data were normally distributed, or using Spearman's rank order correlation.

RESULTS

MDA and protein carbonyls in infant plasma.

The characteristics of the <1500-g (VLBW) infants from whom plasma was collected are shown in Table 1. We considered them in two groups: those with birth weight <1000 g and those with birth weight 1000–1500 g (subsequently referred to as ≥1000 g). Concentrations of the lipid peroxidation product, MDA, in plasma in the period after birth overlapped the ranges for healthy, nonsmoking adults and cord blood from term infants (Fig. 1A). Median values for ≥1000-g infants were significantly lower (p < 0.05) than for cord blood from term infants but no different from the adult mean. They did not change significantly during the period of the study. The median for <1000-g infants was no different from that for ≥1000-g infants at 2 or 7 d but increased 1.4-fold (p < 0.05) at 28 d. This is reflected by negative correlations between MDA concentration and birth weight and gestational age at 28 d, but not at earlier times (Table 2).

Table 1 Profile of infants from whom plasma samples were taken * Commenced on mean d 11.8 (range, 2–80). † Related to live infants. ‡ Any stage, n = number examined. §n = 74. ¶n = 78.
Full size table
Figure 1
figure 1

Plasma MDA (A) and protein carbonyl (B) concentrations at different times for 1000- to 1499-g (clear boxes) and <1000-g (hatched boxes) infants. Box plots show medians and interquartile ranges, with error bars marking the 10th and 90th percentiles. Sample numbers (n) in each group are given. Results for healthy adults and cord blood samples from term infants are also shown.

Full size image
Table 2 Correlation between plasma MDA concentrations, measured at different times after birth, and birth weight and gestational age Data were analyzed using Spearman's rank order correlation. The 2-d samples were collected a mean of 2.6 d after birth.
Full size table

Protein carbonyl concentrations in cord plasma from term infants were substantially higher than in adults (Fig. 1B). Values for both the <1000-g and ≥1000-g groups, at all times, varied over a wider range and were significantly higher than for healthy adults (p < 0.01) but were less than that for cord blood from term infants (p < 0.05). There were no significant differences between the two groups and no correlation with gestational age or birth weight. There was a significant decrease of approximately 30% between 7 and 28 d (p < 0.05) in both groups.

Effects of selenium.

There were no significant differences in birth weight or gestational age between the selenium-supplemented and -unsupplemented infant populations. Of the supplemented infants, 76% received antenatal steroids and 27% postnatal steroids compared with 87% and 32%, respectively, for the unsupplemented infants. Supplementation resulted in a significant increase in plasma selenium, to almost double the pretreatment concentration (Table 3). Without supplementation, there was a decrease as a function of time. Most of the increase caused by selenium had occurred by 1 wk of age, which is within an average of 4–5 d of commencing supplementation. Supplementation prevented the decline after birth in activity of the selenium-containing enzyme, glutathione peroxidase. This response was less pronounced than for selenium concentration, but it was also rapid, with differences between the two groups already significant at 1 wk (Table 3). The correlation between the two factors was weak for the prerandomization samples (correlation coefficient, 0.17;p = 0.04) and increased to between 0.33 and 0.49 (p < 0.001) at the other times. There were no significant differences in initial selenium and glutathione peroxidase between the <1000-g and ≥1000-g populations, and both responded similarly to supplementation.

Table 3 Effects of selenium supplementation on plasma selenium and glutathione peroxidase concentrations Results are medians plus interquartile ranges, followed by the number of samples analyzed. * indicates a significant difference between the two groups (p < 0.0001), **p < 0.05. Samples at 2 d were before randomization and were collected at a mean of 2.6 d after birth.
Full size table

Plasma protein carbonyl and MDA concentrations did not differ significantly between the supplemented and unsupplemented groups at any time, either when data for the whole population were analyzed (Table 4) or when the <1000-g and ≥1000-g groups were examined separately (not shown). Carbonyl concentrations did not correlate with selenium or glutathione peroxidase activity at any time. Apart from a weak negative correlation with glutathione peroxidase at 36 wk (correlation coefficient, −0.23;p = 0.01), MDA was not correlated with selenium or glutathione peroxidase at any time.

Table 4 Protein carbonyls and lipid peroxidation products (MDA) in plasma from selenium-supplemented and -unsupplemented infants Results are medians plus interquartile ranges, followed by the number of samples analyzed. The two groups were not significantly different at any time.
Full size table

Relationship between oxidant markers and outcome.

As there was no significant effect of selenium supplementation on plasma concentrations of the oxidant markers or, as shown elsewhere, on respiratory outcome or development of retinopathy (33), relationships between oxidant markers and outcome were assessed for the whole population, regardless of selenium supplementation. MDA concentrations at 28 d, but not at other times, were significantly higher in those infants who required oxygen at 28 d or 36 wk, and there was a similar trend for ROP (Table 5). These differences were not seen when only <1000-g infants were considered. There was also a positive correlation between MDA measured at 28 d and days of oxygen requirement, but this did not remain significant after correcting for gestational age or birth weight by multiple regression analysis. Protein carbonyls at any time were no different for infants who did and did not acquire CLD or retinopathy, and did not correlate with days on oxygen. Including the infants who died with those who had CLD or ROP did not change these conclusions. At 28 d, but not other times, carbonyls were lower in those infants who received antenatal steroids (median, 0.114;n = 90) than in those who did not (median, 0.146;n = 13;p = 0.05). Antenatal or postnatal steroid administration did not affect other variables. MDA concentrations were 14% lower in samples (n = 139) collected when the infants were receiving any supplemental oxygen compared with those breathing air (n = 122;p < 0.05). Protein carbonyls were not related to Fio2 at the time of sampling.

Table 5 Associations between plasma MDA concentration at 4 wk and outcome measures Associations were assessed using the Mann-Whitney test, for all of <1500-g infants and for the <1000-g subgroup.
Full size table

Comparison of different TBA assays as indices of lipidperoxidation.

In the present study, we did not see higher MDA values at 7 d in infants who acquired CLD, as we observed previously (10). We investigated whether this could be related to the different assays used. Both assays use HPLC to separate the TBA adduct of the lipid peroxidation product MDA, and are assumed to be an index of lipid peroxidation. However, the procedure we used previously (10, 11) measures the MDA-TBA generated when lipid extracts are heated with iron as described by Wade and van Rij (41) and Wade et al. (44), whereas the current assay measures predominantly protein-bound MDA (34, 39). To compare the two assays, we analyzed 32 infant plasma samples. MDA-TBA concentrations ranged from 0.9 to 2.8 μM in the current assay compared with 1.0 to 4.1 μM in the previous assay. Correlation between the two sets of results was poor (r2 = 0.07). Consistent with earlier findings, concentrations measured with the assay of Wade and van Rij (41) and Wade et al. (44) were higher at 7 d than at 2 d. To assess whether this assay reflects the polyunsaturated fatty acid content of the sample and therefore a susceptibility to peroxidation when heated with iron, we measured arachidonic acid (the major source of MDA) and linoleic acid concentrations in the plasmas. Their concentrations ranged from 0.30 to 0.75 mM and 0.53 to 2.67 mM, respectively. Neither arachidonic acid nor the sum of the two correlated positively with MDA-TBA measurements (slopes, −1.62 μM/mM, r2 = 0.27 and −0.16 μM/mM, r2 = 0.06, respectively). The assay, therefore, is not a measure of the amount of substrate capable of undergoing peroxidation.

DISCUSSION

It is widely accepted that free radical generation and oxidative injury contribute to the diseases of prematurity (2, 9, 21), although few direct associations between oxidation and disease severity have been established. If there were generalized oxidative stress in VLBW infants, then concentrations of oxidant markers in plasma would be expected to be higher than in healthy populations. Furthermore, if oxidation plays a causative role in these diseases, higher concentrations would be expected in the sicker infants. We have examined plasma from a large group of <1500-g infants. Concentrations of MDA, measured as an index of lipid peroxidation, overlapped the range for healthy adults and were on average lower than for cord blood from term infants. Protein carbonyl concentrations, although higher and wider ranging than for adults, were also on average less than for cord plasma from term infants. These data provide no strong evidence for generalized oxidative stress in VLBW infants.

Several groups have measured elevated MDA concentrations in cord blood and suggested that this may reflect hypoxia and the stresses associated with birth (23, 25, 26). Others have observed an increase in cord plasma MDA with gestational age (12). Whether the higher cord carbonyl concentrations reflect stresses at birth or a difference between term and VLBW infants needs further investigation.

There were also no detectable associations between MDA or carbonyls and either CLD or ROP. In contrast to our previous studies (10, 11), we saw no positive association between MDA at 1 wk and poor respiratory outcome. We believe the explanation for this difference lies with the assay procedures. The assay used in this study is thought to measure predominantly protein-bound MDA as well as any peroxides present in plasma as a result of lipid peroxidation. It is unclear what the assay used previously measures. The lack of correlation with the other method, as well as the low peroxide levels in plasma (45), make it unlikely that it measures lipid peroxides as initially proposed (41), and we have now established that it does not reflect the amount of lipid in the samples that is capable of undergoing peroxidation. So, although the association with poor outcome is interesting, we cannot say with certainty that it involves lipid peroxidation.

The reason why our plasma indices provided little evidence of oxidative injury in VLBW infants could be that they are not specific enough to detect systemic oxidative stress. Alternatively, oxidation could be confined to particular sites such as the lung or retina and not be detectable in plasma. Lack of specificity is probably unlikely. Protein carbonyl measurements are frequently used to assess protein oxidation. Although carbonyls can arise through processes other than oxidative ones, elevations have been detected in other instances of oxidative stress (46, 47), with particularly high concentrations measured by ELISA in plasma and lung fluid from critically ill patients with adult respiratory distress syndrome (48). Although the TBA assay has been criticized as a measure of lipid peroxidation (34, 49), the version of the HPLC method we used is accepted as the least open to interference.

It seems more likely that oxidative stress is localized, and elevations of protein and lipid markers become lost when diluted in plasma. This is suggested by our findings of considerably higher protein carbonyl and MDA concentrations in tracheal aspirates from a similar premature infant population (50). Others have also obtained evidence for enhanced lung carbonyl formation (17, 18). Unfortunately, large population studies involving measurements on lung aspirates are difficult. Newer, more sensitive tests may be more discriminatory. Higher plasma concentrations of 8-isoprostane, a specific end product of lipid peroxidation (51), have been measured in premature infants than in adults (52), and elevated plasma levels of nitric oxide-derived nitrotyrosine have been found in infants who had CLD (13). These markers are potentially useful probes for further investigation.

As the selenium status of premature infants is low at birth and declines further without nutritional supplementation (30–32), we reasoned that this could compromise their antioxidant capacity and allow more protein and lipid oxidation. As expected for a New Zealand population, we found that plasma selenium concentrations at birth were low by world standards. These almost doubled within less than a week of starting supplementation. A significant difference in glutathione peroxidase concentration between the supplemented and unsupplemented infants was also evident, even though this requires new enzyme synthesis. Any difference in antioxidant protection between the two groups would therefore be expected to be evident from that time. As we have reported elsewhere, selenium supplementation did not result in significant clinical benefit (33). Given that we have now also shown that plasma markers showed little evidence of oxidative stress, it is perhaps not surprising that supplementation resulted in no significant differences in plasma carbonyls or MDA concentrations.

Failure to show benefit from selenium supplementation may have been because the amount given was insufficient, or because the low selenium status of the infants was already adequate. Alternatively, if oxidative injury is an early critical event, selenium supplementation may have been commenced too late to be beneficial. This possibility is perhaps supported by our finding that lower selenium concentrations in mothers, or in infants before randomization, were associated with increased risk of adverse outcome (33). In addition, there is increasing evidence that the inflammatory changes that are likely precursors of neonatal CLD occur within a few hours of birth (53) or even before delivery in some infants (54).

In conclusion, we found that protein carbonyls and MDA concentrations in plasma of VLBW infants were not indicative of systemic oxidative stress and did not correlate with respiratory outcome or ROP. They were also not affected by selenium supplementation. These findings do not necessarily discount the involvement of oxygen radicals in the diseases of prematurity, but suggest it is more likely to be a localized and subtle phenomenon.