Early cortisol values and long-term outcomes in extremely low birth weight infants

Abstract

Objective

Both excess and insufficient levels of glucocorticoid in extremely low birth weight (ELBW) infants have been associated with adverse hospital outcomes, whereas excess glucocorticoid exposure has been associated with long-term adverse neurodevelopment. Our objective was to evaluate the relationship between neonatal cortisol concentrations and long-term outcomes of growth and neurodevelopment.

Study Design

As part of a multicenter randomized trial of hydrocortisone treatment for prophylaxis of relative adrenal insufficiency, cortisol concentrations were obtained at 12 to 48 h of postnatal age and at days 5 to 7 on 350 intubated ELBW infants, of whom 252 survived and returned for neurodevelopmental follow-up at 18 to 22 months corrected age. Cortisol values from each time point were divided into quartiles. Growth and neurodevelopmental outcome were compared for each quartile.

Result

Median cortisol value was 16.0 μg per 100 ml at baseline for all infants, and 13.1 μg per 100 ml on days 5 to 7 in the placebo group. Outcomes did not differ in each quartile between treatment and placebo groups. Low cortisol values at baseline or at days 5 to 7 were not associated with impaired growth or neurodevelopment at 18 to 22 months corrected age. High cortisol values were associated with an increase in cerebral palsy, related to the increased incidence of severe intraventricular hemorrhage (IVH) and periventricular leukomalacia.

Conclusion

Low cortisol concentrations were not predictive of adverse long-term outcomes. High cortisol concentrations, although predictive of short-term adverse outcomes such as IVH and periventricular leukomalacia, did not additionally predict adverse outcome. Further analysis into identifying factors that modulate cortisol concentrations shortly after birth could improve our ability to identify those infants who are most likely to benefit from treatment with hydrocortisone.

Introduction

Glucocorticoid treatment in extremely low birth weight (ELBW) infants has been surrounded by controversy, with questions of when to treat, how much to use and which of the many choices of steroid is the most appropriate. Mounting evidence shows that either insufficient or excess glucocorticoid can affect short- and long-term outcomes in this very vulnerable population. Use of high-dose dexamethasone facilitates extubation,^1,2 but is associated with acute complications such as hypertension, gastrointestinal perforation and hyperglycemia. Early high-dose dexamethasone is associated with long-term disability, with an increased incidence of cerebral palsy.^3,4 In contrast, there is evidence that inadequate exposure to glucocorticoid, such as with relative adrenal insufficiency, may be associated with hypotension, the development of bronchopulmonary dysplasia (BPD)^5-9 and death.¹⁰ Results of randomized trials of prophylactic treatment with hydrocortisone have shown conflicting results regarding the difference in the incidence of BPD or need for blood pressure support.^11-13

Ascertaining appropriate use of steroids is likely confounded by individual susceptibility to inflammation, with some infants benefiting from supplementation, whereas others suffer the adverse effects of excess exposure. We previously examined cortisol concentrations shortly after birth (first 48 h of life) and at days 5 to 7 of life to determine whether infants at high risk for short-term adverse outcomes could be identified and thus selected for treatment. Low cortisol concentrations at either time point did not identify a high-risk population; however, high cortisol concentrations were associated with increased mortality and severe (grades 3 to 4) intraventricular hemorrhage (IVH).¹⁴ High cortisol concentrations have additionally been associated with gastrointestinal perforations.^11,12

Because extremes in steroid exposure may have adverse effects on long-term outcomes despite short-term benefits, we endeavored to evaluate long-term outcome of ELBW infants based on their early cortisol concentration. Our aim was to determine whether infants who may benefit or be harmed by supplementation with hydrocortisone could be identified in the neonatal period from a group at risk for adverse long-term outcomes.

Methods

Population and procedures

A secondary analysis was performed with data collected in a multicenter study of low-dose hydrocortisone therapy used as prophylaxis of early adrenal insufficiency (PROPHET). The study protocol was approved by institutional review boards at all participating institutions and parental consent was obtained before enrollment. A full description of the sample and details of the study protocol have been previously reported.¹¹ In brief, cortisol concentrations were obtained for ELBW infants (birth weight 500 to 999 g) who were requiring mechanical ventilation at 12 to 48 h of life. Infants were randomized to receive normal saline placebo or hydrocortisone sodium succinate, 1 mg kg⁻¹ per day, divided twice daily for 12 days, followed by 0.5 mg kg⁻¹ per day for 3 days. Baseline cortisol concentrations were obtained for all infants at study entry, between 12 and 48 h postnatal age. An additional cortisol concentration was obtained on days 5 to 7 of life. Data were collected on hospital outcome. Growth and neurodevelopmental outcome of survivors were assessed at 18 to 22 months corrected age. Weight, height and head circumference were recorded, and neurologic examination and developmental assessment through the Bayley Scales of Infant Development II were performed. The full follow-up procedure has previously been described.¹⁵

As we previously reported, the median cortisol value was 16.0 mcg per 100 ml at baseline for all infants, and 13.1 mcg per 100 ml on days 5 to 7 in the placebo group.¹⁴ To determine physiologic ranges, quartiles were determined at both baseline (using 332 values) and at days 5 to 7 (using values from 153 infants in the placebo group). Infants treated with hydrocortisone before study entry were excluded. At baseline, infants in the lower quartile had cortisol concentrations of <8.9 mcg per 100 ml, and those in the upper quartile had concentrations of >31 mcg per 100 ml. At days 5 to 7, the lower quartile was <8.7 μg per 100 ml and the upper quartile was >18.1 mcg per 100 ml.

For this analysis, outcomes of infants in each quartile were first examined by treatment groups to evaluate the effect of treatment within the quartiles. Outcomes did not differ by randomization assignment (placebo or hydrocortisone), and hence data were combined. Therefore, outcomes based on days 5 to 7 of cortisol levels represent the full cohort, including both who were treated with hydrocortisone and those who received placebo.

Next, we established that birth weight, gestational age and demographic characteristics for those infants who returned for follow-up did not differ between the quartile groups for cortisol at baseline or at days 5 to 7 of life. Growth parameters at 36 weeks of corrected age and incidence of BPD were similar across the baseline quartile groups, but there was an increased incidence of spontaneous gastrointestinal perforation, severe IVH and periventricular leukomalacia in infants with cortisol levels in the upper quartile at both baseline and days 5 to 7 of life. These findings are consistent with those found in the full cohort of survivors, suggesting that the children who returned for follow-up are representative of the full cohort, with similar distributions across the quartile groups. Exposure to antenatal steroids was also similar across all groups.

We then compared growth parameters, neurologic and developmental outcomes at 18 to 22 months corrected age with cortisol concentration quartiles at baseline and at 5 to 7 days of life to determine the impact of early cortisol values on long-term outcome.

Statistical analysis

Chi-square tests, or Fisher’s exact tests when appropriate, were used for bivariate analysis of dichotomous outcomes. These analyses are summarized by frequencies and percentages within cortisol quartiles. Continuous outcomes are summarized by means and s.d.. Logistic regression was used for multivariable analysis of dichotomous outcomes, and linear regression was used for multivariable analysis of continuous outcomes, adjusting for factors known to effect long-term developmental outcomes: maternal education, birth weight, gestational age, severe IVH, BPD and periventricular leukomalacia. All hypothesis testing is two sided with P-values of <0.05 considered statistically significant. All analyses were conducted in SAS Version 9.1 (SAS Institute Inc., Cary, NC, USA).

Results

A total of 360 infants were enrolled in the original study, and 294 infants survived to discharge. An additional three infants died before follow-up. Of the 291 survivors, 252 (87%) were evaluated for growth and neurodevelopmental outcomes at 18 to 22 months corrected age. Cortisol determinations were available on 246 of the infants who returned for follow-up evaluation.

Outcomes at 18 to 22 months corrected age are shown in Table 1, grouped by cortisol concentration quartile at baseline. Growth parameters did not differ between the groups. The incidence of cerebral palsy was significantly higher (20%) in those with the highest cortisol values shortly after birth compared with the lower and middle quartile groups (9 and 13%, respectively). The Mental Developmental Index (MDI) and the Psychomotor Developmental Index trended downward with increasing cortisol quartile group, although the trends were not significant. Infants had a significantly higher rate of Psychomotor Developmental Index <70 with increasing cortisol quartile group (P = 0.03). The rate of major neurologic impairment also increased with increasing cortisol quartile group, but not significantly so.

Table 1.

Infant characteristics at 18-month follow-up by baseline cortisol concentration

	Lower quartile (<8.7 μg per 100 ml) N=66	Middle quartiles (8.7–18.1 μg per 100 ml) N=124	Upper quartile (>18.1 μg per 100 ml) N=56
Weight (kg)	10.8±2.6	10.6±1.3	11.2±1.7
Height (cm)	81.1±4.3	80.7±3.4	82.2±3.8
Head circumference (cm)	47.2±1.9	47.0±1.7	47.8±2.1
Cerebral palsy	6 (9%)	16 (13%)	11 (20%)a
MDI	80±20	79±19	77±19
PDI	86±18	84±18	80±24
MDI <70	20/64 (31%)	39/120 (33%)	17/53 (32%)
PDI <70	12/65 (18%)	28/120 (23%)	19/53 (36%)
Major neurologic impairment	26/66 (39%)	50/123 (41%)	24/55 (44%)
Normal gross motor	51 (77%)	92 (74%)	43 (77%)

Abbreviations: MDI, Mental Developmental Index; PDI, Psychomotor Developmental Index.

^aP < 0.03.

On the basis of the linear regression models, severe IVH, maternal education and BPD were independent predictors of MDI. In logistic regression models, severe IVH increased the odds of an MDI <70 (odds ratio 8.41 (2.39 to 29.6)), whereas higher maternal education was protective (odds ratio 0.76 (0.58 to 0.99)). Only severe IVH predicted Psychomotor Developmental Index <70 (odds ratio 3.71 (1.14 to 12.13)). Linear regression models revealed that cortisol quartiles at this time point were not predictive of Bayley scores.

A similar analysis was performed examining the outcomes at 18 to 22 months with cortisol concentration quartiles at days 5 to 7 of life (Table 2). Growth parameters were similar across the quartiles, but cerebral palsy was increased in the infants in the upper quartile (21%) compared with the lower and middle quartiles (11 and 9%, respectively). Again, only severe IVH, maternal education and BPD were found to be independent predictors of MDI, whereas only severe IVH predicted MDI <70 (odds ratio 7.75 (2.15 to 28.0)). There were no independent predictors of Psychomotor Developmental Index <70. Having a cortisol value in the middle quartile compared with the upper quartile was associated with a normal gross motor exam (odds ratio 2.48 (1.01 to 6.09)); otherwise, cortisol concentrations at days 5 to 7 of life were not predictive of outcome.

Table 2.

Infant characteristics at follow-up by days 5 to 7 of cortisol concentration

	Lower quartile (<8.7 μg per 100 ml) N=45	Middle quartiles (8.7–18.1 μg per 100 ml) N=106	Upper quartile (>18.1 μg per 100 ml) N=92
Weight (kg)	10.5±1.3	10.8±1.4	10.9±2.3
Height (cm)	80.6±3.3	81.5±4.0	81.0±3.5
Head circumference (cm)	47.5±1.7	47.2±1.9	47.2±1.9
Cerebral palsy	5 (11%)	10 (9%)	19 (21%)a
MDI	80±18	80±18	75±19
PDI	85±20	85±17	81±22
MDI <70	12/42 (29%)	34/103 (33%)	31/89 (35%)
PDI <70	11/43 (26%)	20/103 (19%)	28/89 (31%)a
Major neurologic impairment	17/43 (40%)	46 (43%)	39 (42%)
Normal gross motor	34 (76%)	87 (82%)	61 (66%)

Abbreviations: MDI, Mental Developmental Index; PDI, Psychomotor Developmental Index.

^aP < 0.03.

Discussion

Both excess and insufficient endogenous concentrations of glucocorticoid in ELBW infants have been associated with adverse hospital outcomes,^{2,5,7,9,16-18} whereas exposure to excess exogenous glucocorticoid has been associated with long-term adverse neurodevelopment.^3,4 This was most recently shown with the use of dexamethasone for bronchopulmonary dysplasia, resulting in a strong statement by the American Academy of Pediatrics cautioning against the use of steroids in preterm infants.¹⁹ Although postnatal dexamethasone use decreased after this statement, it is still a common therapy in intensive care nurseries. In addition, hydrocortisone use has become more common.²⁰ Clinical evidence regarding long-term outcome after exposure to hydrocortisone is beginning to emerge. Two studies have documented the lack of adverse outcome observed in the 18- to 22-month follow-up of ELBW infants treated for early prophylaxis with hydrocortisone.^15,21 Additional studies have examined preterm infants treated later with hydrocortisone for treatment of BPD, comparing them with infants who were not exposed.^22,23 Follow-up at age 8 years found no adverse effects of hydrocortisone therapy on magnetic resonance imaging findings or neurocognitive outcome.^22,23

Although these findings suggest that hydrocortisone use to correct relative adrenal insufficiency is not associated with adverse neurodevelopmental outcomes, an additional concern is whether adrenal insufficiency itself may have an adverse effect on neurodevelopment. Animal models of adrenal deficiency through adrenalectomy have shown a degeneration of neurons in the hippocampus.²⁴ This raises the question of whether a deficiency in steroid hormones can be equally as deleterious as steroid excess in infants.

In our study, we evaluated the follow-up of infants at 18 to 22 months based on their cortisol concentrations after birth (12 to 48 h of life) and at 5 to 7 days of life. Those infants with the lowest cortisol concentrations did not have more adverse neurodevelopmental outcome, as evaluated by the Bayley Scales of Infant Development and neurologic examination for cerebral palsy. Adverse outcomes were more common in the infants with the highest cortisol values at each time point, reflecting the increased incidence of severe IVH in this group. The elevated cortisol concentrations suggest a marker of vulnerability in early life, with long-term adverse outcomes related with the immediate complications rather than the cortisol concentration itself. These relationships held true for infants regardless of assignment to treatment or placebo group.

The interpretation of cortisol levels in preterm infants presents a significant challenge for neonatologists. In utero, fetal cortisol levels are low, rising slowly in the third trimester.²⁵ Low cortisol levels are common in well preterm infants, and are not a reflection of adrenal insufficiency.^26,27 When a low cortisol level is obtained on a preterm infant requiring respiratory support in the neonatal intensive care unit, it is difficult to determine whether the level is appropriately low, reflecting an appropriate in utero state, or whether it represents an inappropriate response to the extrauterine stress and thus relative adrenal insufficiency. A possible lowering of early cortisol levels by the administration of maternal antenatal steroids for fetal lung maturation further confounds this, perhaps contributing to the lack of association of low cortisol concentrations with adverse short-term hospital outcomes. In contrast, some aspects of stress, such as pain or surgery causing elevation of cortisol levels, are detrimental,²⁸ which may contribute to the association of increased mortality and IVH with elevated cortisol concentrations.

Both deficiency and excess of glucocorticoid exposure can lead to adverse outcomes in ELBW infants. Although the potential for long-term adverse consequences associated with adrenal insufficiency is suggested from animal data, our study did not find evidence for this, likely because random cortisol concentrations are not sufficient to distinguish infants with relative adrenal insufficiency acutely, and thus not helpful in the prediction of long-term outcomes. Cortisol concentrations are highly variable and difficult to interpret. Preterm infants lack the normal diurnal variation of cortisol levels, further confounding their interpretation.²⁹ Low values at these time points are not adequate evidence for relative adrenal insufficiency in isolation, and must be interpreted in clinical context. Low cortisol concentrations may reflect in utero concentrations rather than indicating relative adrenal insufficiency. Distinguishing infants with relative adrenal insufficiency from those with appropriately low cortisol values remains a challenge for neonatologists. Further analysis into identifying factors that modulate cortisol concentration is needed to improve our understanding of this complex system.

Appendix

The Prophet Study Group Principal Investigators and sites: Kristi L Watterberg, University of New Mexico; Jeffrey S Gerdes, University of Pennsylvania; Cynthia H Cole, Tufts University; Susan W Aucott, Johns Hopkins University; Elizabeth H Thilo, University of Colorado; Mark C Mammel, and Robert J Couser, Children’s Hospital and Clinics of Minnesota, University of Minnesota; Jeffery S Garland, Wheaton Franciscan Health Care-St Joseph’s, Medical College of Wisconsin; Henry J Rozycki, Virginia Commonwealth University; Corinne L Leach, State University of New York, Buffalo; Conra Backstrom, University of New Mexico; and Michele L Shaffer, Pennsylvania State University.