Results

Between August 1996 and January 1998, 99 infants were enrolled in the study. Forty seven fulfilled our criteria for good health. Table 1shows the demographic characteristics. Only one infant was being mechanically ventilated at the time of blood sampling. No infant was treated with phenobarbital or phenytoin before or at the time of blood sampling.

Before further analysis of the cortisol pattern two outliers (1233 and 2050 nmol/l), both measured on the first day of life, were excluded by visual inspection.10 SGA infants had significantly higher cortisol values than appropriate for gestational age (AGA) infants (table 1). The distribution of cortisol values among the AGA infants was non-Gaussian and skewed to the right. They had a nearly normal distribution however, when log₁₀transformed.11 The points determined by mean (2 SD) on the logarithmic scale were transformed back to the original units to provide a reference range: 73–562 nmol/l. Taking into account, gestational age, birthweight, and postnatal age as independent variables, and cortisol (log₁₀) as the dependent variable in a stepwise multiple regression model, gestational age and birthweight had a significant effect (p = 0.0041 and p = 0.032, respectively). Birthweight alone was not significantly correlated with cortisol values (regression coefficient = −0.0001, p = 0.363), but after including birthweight in the model, r² increased from 0.123 to 0.236. In this model the regression coefficient (standard error) of birth weight was 0.00066 (0.00030) and the regression coefficient of gestational age was 0.125 (0.041). However, the combination of birthweight and gestational age to predict cortisol values seemed to be cumbersome in clinical practice and birthweight was omitted on grounds of simplicity. The regression coefficient (standard error) of gestation alone was −0.045 (0.020) (p = 0.033) and the regression equation was log (cortisol) = 3.5373–0.045 × gestational age. Based on this equation and the residual standard deviation about the regression line, a reference range for each week of gestational was derived (table 2). Again, the points determined by mean (2 SD) on the logarithmic scale were transformed back to the original units to provide a reference range. Furthermore, in AGA infants, no significant differences in cortisol values were found between girls and boys (p = 0.82) as well as between those infants whose mothers had or had not received a complete course of prenatal steroids (p = 0.94) (Mann–Whitney U test).

Discussion

Table 3 summarises the published data on well preterm infants, most of whom were 30 weeks or more of gestational age.1 12 13 14 The definition of “well” varied from a single criterion to very detailed criteria.1 15 In only one study was serum glucose concentration included.14Finally, none of the studies stipulated the time interval before blood sampling during which the patient remained undisturbed. This is mandatory in intensive care patients, who have a busy programme of nursery care and medical activity. Healthy term infants show an up to 11-fold rise in cortisol half an hour after a painful venepuncture.16 The 48 hour interval was based on the observation that cortisol values in term neonates returned to baseline within 48 hours of surgery.17 Our strict criteria meant that only 47% of all eligible infants could be included in the study.

Compared with previously published cortisol values in a more mature population (median cortisol 137, 140, and 104 nmol/l for < 30, 30, and 31 weeks postconceptional age, respectively), our values are higher (median cortisol 199 nmol/l).15 This agrees with our finding of a significantly negative correlation between gestational age and cortisol concentrations and published data.2 In this study, the effect of gestational age on cortisol values is more pronounced, if birthweight is included together with gestational age in a multiple regression model, whereas birthweight alone did not have a significant effect on cortisol values. In a recent report cortisol values decreased significantly with advancing postnatal age between 1 and 8 weeks,18 but we found only a weak negative correlation between cortisol concentration and postnatal age in the first two weeks of life. This may be explained by the high rate of administration of prenatal steroids in our study group (table 1).

However, the data on adrenal suppression after prenatal administration of corticoids for fetal lung maturation are contradictory. Two groups found postnatal normalisation of suppressed cortisol values within two hours or two days, respectively.19 20 Others found that serum cortisol concentrations were maximally suppressed by 55% within the first four to seven days of life18 21 22 but not after the first week.18 23 In our study the cortisol values, which were obtained during the first week, were solely derived from infants whose mothers received prenatal steroids. Only six of 37 (16%) AGA infants, whose cortisol values were measured on days 7 to 13, were not under the influence of prenatal steroids. Thus the presented reference range applies to preterm infants after prenatal administrations of betamethasone. This is important, because the generally accepted benefit of prenatal steroids leads to an increasing rate of prenatal administration of steroids in at risk pregnancies.24 Among the published findings on serum cortisol, the administration of prenatal steroids was taken into consideration by only one author for the analysis of the data.18

Little is known about cortisol concentrations in SGA infants. The concentrations of several glucocorticoids, including cortisol, were lower in term SGA infants than in AGA infants at 12 hours of age, but the cortisol values at 24 hours of age were higher, and after the first day of life no differences were found in cortisol values between term SGA and AGA infants.25 These findings suggest a different adrenocortical response in SGA infants and we therefore separated out the SGA infants before further analysis. In this study preterm SGA infants had significantly higher concentrations compared with preterm AGA infants of the same gestational age.

Ten of 47 (21%) of our subjects had cortisol concentrations below 138 nmol/l, the threshold of adrenal insufficiency.26 However, they had no clinical signs of adrenal insufficiency. We do not know whether preterm infants under 30 weeks of gestational age with cortisol concentrations < 138 nmol/l are at greater risk of developing clinical signs of adrenal insufficiency in stressful situations. Our study does not give any information on the function of the hypothalamic–pituitary–adrenal axis in these patients.

Although newborn infants lack a circadian rhythm for cortisol,27 there is pulsatility of cortisol secretion.28 However, only small changes in serum cortisol within ill extremely low birthweight infants were shown by Jettet al.29 The standard deviations ranged between 10 and 114 nmol/l. The authors concluded that a single random cortisol measurement adequately reflected the adrenal status of an extremely low birthweight infant. We can therefore fill a gap by supplying a strictly defined reference range for serum cortisol concentrations in well AGA preterm infants with under 30 weeks of gestational age during the first two weeks of life, a period of highest critical morbidity in this population.