Infant cortisol response after prolonged antenatal prednisolone treatment
Professor V. Glover, Wolfson and Weston Research Centre for Family Health, Institute of Reproductive and Developmental Biology, Faculty of Medicine, Imperial College London, Hammersmith Campus, Du Cane Road, London W12 0NN, UK.
Prednisolone is widely used to treat medical conditions in pregnancy, despite the lack of long term safety studies on infants. Animal studies have shown that antenatal glucocorticoid treatment can cause in utero growth restriction and up-regulation of the offsprings' hypathalamic–pituitary–adrenal axis. We recruited women treated antenatally with prednisolone, and followed 12 of the infants up to four months, using routine infant vaccinations as a stressor. Birthweights were similar to controls (n= 289, uncomplicated, singleton term pregnancies), as were infants' baseline and stress-induced cortisol levels. Mothers rated their infants as less difficult and more adaptable than controls. This study provides initial reassurance about the safety of prednisolone in pregnancy.
Despite the mounting evidence from animal studies for in utero growth restriction and hypothalamic–pituitary–adrenal axis hyperactivity in offspring of steroid-treated mothers, the possibility of long term adverse effects to infants of prolonged antenatal prednisolone in humans has received little attention.
Women with medical problems such as systemic lupus erythematosus, ulcerative colitis or severe asthma often have to take prednisolone during pregnancy in doses which can be as high as 40 mg per day.1 Compared with betamethasone and dexamethasone, which cross the placental barrier in ratios of 3:1 and 2:1, respectively, only 10% of prednisolone crosses the placenta to reach the fetus. However, as this drug is often required throughout gestation, compared with short one-day courses of the other steroids used to enhance fetal lung maturation, the total drug dose of prednisolone is considerable over a sensitive time of neurodevelopment. A literature search revealed minimal follow up data on infants from prednisolone-treated pregnancies. No studies have examined infant hypothalamic–pituitary–adrenal responses following prolonged antenatal prednisolone treatment. We recruited women treated antenatally with prednisolone, and examined infant parameters of growth at birth and their hypothalamic–pituitary–adrenal responses at two and four months.
Twenty women with singleton pregnancies treated antenatally with prednisolone were recruited consecutively through Queen Charlotte's & Chelsea Hospital's obstetric medical clinic, as approved by the institutional ethical committee, with informed consent obtained. Numbers were limited by the time frame of the concurrent control study. Details of corticosteroid treatment and delivery data, including mode of delivery, gestation, birthweight, Apgar scores, cord arterial pH and meconium passage, were extracted from the medical notes.
Routine infant vaccinations (diphtheria, tetanus, pertussis, haemophilus influenza type B and meningitis C) were administered using a standard two-injection technique by the same practitioner at two and four months. Vaccination is known to produce a cortisol response in infants at this age. Infant saliva was collected before and 20 minutes after vaccination for cortisol estimation using ‘Salivettes’ (Sarstedt, UK). Infant crying times were measured as the time to first 5-second pause, using a video recording. 12 infants returned at two months. Of these one was excluded at both time points, and two more were excluded at four months (Table 1). Thus, for cortisol analysis, n= 11 infants at two months and n= 9 at four months.
Table 1. Details of prednisolone treatment and infant cortisol levels in the 12 patients that were followed up. All were given hydrocortisone 100 mg intravenously in labour. Cortisol units (nmol/L). No infant cortisol values are given for Patient 8 due to a history of gastro-oesophageal reflux. (The assay used is known to yield artificially high results in the presence of acidity). At four months, no cortisol values are given for Patient 9 due to topical hydrocortisone treatment for widespread eczema or Patient 10 due to an adverse reaction to vaccination at two months.
|2||Wegener's granulomatosis||10|| 0–39||13.6||−5.2||2.7||17.7||No (azathioprine)|
|3||Rheumatoid arthritis||12|| 0–34||7.5||1.3||1.5||6.3||Yes|
|4||Systemic lupus erythematosus||5|| 0–38||12.4||−4.0||4.5||9.0||No (azathioprine)|
|5||Ulcerative colitis||10|| 0–40||8.1||12.2||1.0||7.0||Yes|
|6||Hyperemesis gravidarum|| 30*|| 9–23||2.3||25.1||10.6||1.1||Yes|
|7||Asthma|| 40*|| 35–39||2.4||11.3||6.3||8.0||Yes|
|8||Asthma|| 40*|| 32–35||–||–||–||–||Yes|
|9||Polmorphous eruption of pregnancy|| 40*|| 38–39||8.7||1.8||–||–||Yes|
|10||Systemic lupus erythamatosus|| 40*|| 22–35||13.6||−4.5||–||–||Yes|
|11||Pemphigoid gestationalis|| 40*|| 37–41||2.8||21.1||6.3||5.6||Yes|
|12||Rheumatoid arthritis||Not known|| 0–38||1.5||7.9||5.9||−0.1||Yes|
The most common reason for not returning for follow up was inconvenience, due to many women living far from the hospital, but having booked at Queen Charlotte's because of the specialist medical clinic.
Cortisol assays were performed in duplicate in batches using a standard solid-phase RIA for plasma (DPC, Los Angeles, California), adjusted for salivary concentrations. Intra- and inter-assay coefficients of variation were 4.1% and 9.9%. Delta cortisol was defined as the post-vaccination minus the pre-vaccination level. Statistical analysis used SPSS 11 and GraphPad Prism 3.0. Non-parametric statistics were used throughout. All statistical analyses were two-tailed. P < 0.05 was taken to be statistically significant.
At four weeks, two months and four months, mothers were asked to complete questionnaires on infant feeding, the Edinburgh Postnatal Depression Scale (EPDS), Spielberger State and Trait anxiety questionnaires and the Bates questionnaire of negative infant behaviour. Infants were described as ‘predominantly’ breast-fed if they were breast-fed more than 50% of the time. Mother–infant interaction was examined at four months by video recording mother–infant pairs playing together in a standard well-validated setup. These tapes were reviewed and ranked by a trained observer, blind to patient details.
A concurrent cohort of infants unexposed in utero to corticosteroids (uncomplicated, term, singleton pregnancies, n= 289), which were recruited sequentially to another study with identical follow up procedures, was used as controls. Exclusion criteria were delivery <36 weeks, birthweight <2500 g, known oligohydramnios or abnormal umbilical Doppler waveforms, antenatal corticosteroid treatment, moderate to severe pre-eclampsia or other maternal conditions that compromise fetal wellbeing.
The prednisolone dose varied between women and over time during any individual pregnancy, from 5 mg to a maximum of 40 mg. Length of treatment also varied; the median (range) was 24.5 (1 to 40) weeks (Table 1).
The prednisolone-treated (steroid cohort) (n= 20) and control cohorts (n= 289), were largely comparable. Median (range) age was 31.5 (21 to 44) years and 32 (18 to 45) years, respectively. Median parity was zero for both groups. All the controls were primiparous. However, although the steroid cohort was mainly primiparous, due to the small number of women taking prednisolone in pregnancy, multiparae were included. Therefore, in the steroid cohort, the range of parity was from 0 to 5. The steroid cohort was delivered earlier, median (range) 38 (32 to 41) weeks, compared with the control group, 40 (36 to 42) weeks (P= 0.0001, Mann–Whitney). Birthweight was lower in the steroid group, 3068 (2188 to 4180) g compared with 3412 (2074 to 4602) g in the control cohort (P < 0.02, Mann–Whitney). However, after correction for gestational age, there was no difference in birthweight between the two groups. A higher proportion of the steroid cohort were delivered by caesarean section (65%), compared with the control group (41%) (P= 0.04, Fisher's exact test). There were no differences between the two cohorts in any other indices examined. For the steroid cohort and control cohort, respectively, these were: sex of the baby (35% female and 50% female), head circumference [34 (32 to 39) cm and 34.5 (30 to 39) cm], arterial cord pH [7.29 (7.17 to 7.36) and 7.26 (6.96 to 7.40)], 1-minute Apgar score [9 (6 to 10) and 9 (4 to 10)] and meconium-stained amniotic fluid (25% and 21%).
Baseline cortisol level and delta cortisol at two months and at four months were similar to those of the control group. All values from the steroid group were within the normal range of the control cohort. [For the control group at two months, median (range) baseline cortisol was 3.8 (0.1 to 34.6) nmol/L, and delta cortisol 8.1 (−10.5 to 47) nmol/L. At four months, the comparable values were 4.9 (0 to 27.2) nmol/L and 5.7 (−8.2 to 43) nmol/L].
In the steroid group, there was a trend towards an inverse relationship between baseline cortisol level at two months and baseline level at four months (rs=−0.65, P= 0.07, n= 9). A similar trend was observed in delta cortisol levels at the two time points in the steroid group (rs=−0.65, P= 0.07, n= 9), those with a larger rise in cortisol response to vaccination at two months, having a smaller rise in cortisol response at four months, and vice versa. This pattern was not seen in controls, in whom no relation was found between baseline or delta cortisol levels at the two time points.
The steroid group was significantly less likely to be predominantly breast-fed at four weeks (50%vs 80%, P < 0.05, Fisher's exact test) and at four months (8%vs 47%, P < 0.05, Fisher's exact test). There was no statistical difference in EPDS scores at any time point between the steroid and control groups. At four weeks EPDS median (range) scores were 6 (0 to 25) in the control group and 10 (0 to 19) in the steroid group (P= 0.09, Mann–Whitney). These values fell to 5 (0 to 19) and 6.5 (0 to 14), respectively at two months (P= 0.3), and 4 (0 to 24) and 5 (0 to 9), respectively at four months (P= 0.6). Spielberger anxiety scores did not vary between groups at any time point (data not supplied).
Median infant crying times post-vaccination did not vary between groups at either time point. Median (range) crying times were 61 (7 to 265) seconds in the control group and 60 (22 to 130) seconds in the steroid group at two months, and 52 (2 to 240) seconds and 55 (9 to 89) seconds, respectively at four months. Women in the steroid group were, however, significantly more likely to perceive their babies as being less difficult and more adaptable than the non-steroid group. High Bates scores indicate a more negative infant behavioural characteristic. Median (range) Bates scores for the parameter ‘difficult’ were 25 (12 to 45) for the control group (n= 153) and 19 (13 to 31) for the steroid group (n= 11) (P= 0.01). For ‘unadaptable’ respective scores were 6 (2 to 13), n= 153 and 2 (2 to 6), n= 11 (P= 0.0006). Subanalysis revealed that this difference was not due to the influence of multiparous women in the steroid group, because the lowest scores were all in primiparous women.
Mother–infant interaction is rated according to six dimensions, with scores ranging from low to high as follows: maternal (poor–good), maternal (intrusive–remote), maternal (depressed–happy), infant (poor–good), infant (inert–fretful) and overall mother–infant interaction (poor–good). Analysis of maternal–infant interaction from video recordings found no difference in any of these dimensions between the steroid treatment and control groups.
No effect of steroid dose or duration was discernable in any of the parameters examined.
This study is the first to examine hypothalamic–pituitary–adrenal responses in infants born at or near term after prolonged maternal glucocorticoid therapy. It also addresses other concerns arising from animal glucocorticoid studies such as possible reduction in birthweight and head circumference, and altered behavioural characteristics subsequent to prednisolone treatment.
Some possible fetal effects of prolonged antenatal prednisone treatment have been examined previously.2,3 However, despite its widespread use in obstetric medicine, the long term effects of prednisolone on infants' hypothalamic–pituitary–adrenal responses have been largely ignored. Mogadam et al.3 assessed a large cohort of prednisolone-treated pregnancies. Their complication rates were reassuring, but as the study was conducted retrospectively via gastroenterologists, with low pregnancy complication rates overall, it seems likely that obstetric complications were under-reported. A small prospective study conducted in the 1960s was reassuring with respect to fetal growth.2 A literature search found only two published studies examining hypothalamic–pituitary–adrenal responses in infants treated with steroids (both short courses in pregnancies threatened by preterm labour). In one of these, infants were delivered <32 weeks of gestation and then investigated only at 7 and 14 postnatal days after treatment with dexamethasone.4 The authors found no alteration in hypothalamic–pituitary–adrenal responses in exposed infants, although it was difficult to exclude confounding factors given the nature of the cohort studied. In the other study, treatment was with betamethasone and gestation at delivery varied from 24 to 34 weeks.5 With the complications of prematurity implicit within this gestation range, as well as 32/79 (41%) of the infants being from multiple pregnancies, the impact of potential confounders is vast and potentially insurmountable. Their results were reassuring, but both these studies exemplify the problems faced in studying this subject.
Unlike betamethasone and dexamethasone, which are only administered fairly late in development after viability, prednisolone is often administered throughout gestation (in this study, 6 of the 12 infants that were followed up were exposed to prednisolone throughout gestation). Animal studies have shown that antenatal glucocorticoid treatment exerts a permanent programming effect on offspring, partly by a reduction in the number of hippocampal glucocorticoid receptors, thus attenuating negative feedback and causing increased and prolonged cortisol responses to any given stressor. Our study in humans suggests that at least until four months of age prednisolone treatment did not alter infants' cortisol levels or response to vaccination. The lack of difference in parameters of infant behaviour is reassuring. Glucocorticoid treatment is known to accelerate organ maturation at the expense of organ size, so it was also reassuring to find no difference in birthweight and head circumference.
It is of some concern that fewer of the steroid group women breast-fed. However, two of the three women who did not breastfeed were on cytotoxic drugs (azathioprine) during this time. At present, women are advised not to breastfeed on azathioprine due to the knowledge that small amounts of drug do get into breast milk, and because there are insufficient safety data reported. Women in the steroid group were, however, significantly more likely to perceive their babies as being less difficult and more adaptable than the non-steroid group. The babies could be potentially viewed as ‘more precious’ by mothers who had perhaps doubted their ability to conceive and carry a child to term.
The trend observed in the prednisolone-exposed group, of an inverse relationship between cortisol levels at two and four months, is intriguing. Ramsay and Lewis6 demonstrated this developmental shift in a previous study, both in infants born in what they defined as optimal and in those born in non-optimal conditions. It may be that a high cortisol response is optimal in the first few months of life, but not later on, and vice versa. However, the numbers studied are small; this observation should be viewed with caution and investigated in further studies with larger numbers.
This study is the first to provide reassuring data about the effect of in utero exposure to prednisolone on the neonatal hypothalamic–pituitary–adrenal axis and behaviour, and will be of value to clinicians managing women who require prednisolone treatment during pregnancy. Longer term follow up of larger numbers of exposed infants will be required for complete reassurance, as well as enabling subanalyses of the effect of the dose, duration and timing of treatment.
The authors would like to thank SPARKS (Sports Aiding Medical Research for Children) who funded this study. The authors are grateful to Diana Adams for practical help in the clinics, and to the mothers and babies, without whom this study would not have been possible. The authors would like to acknowledge infrastructural support from the Henry Smith Charity, The Wellcome Trust and the Institute of Obstetrics and Gynaecology Trust.