P2C244: Assessing the Utility of Cortisol Levels in Predicting Iatrogenic Adrenal Insufficiency in Neonates Exposed to Prolonged Steroid Courses

Background: Steroids are often used in premature neonates to wean respiratory support. Prolonged steroid exposure can cause secondary adrenal insufficiency (AI) due to suppression of the hypothalamic-pituitary-adrenal axis. Our current NICU protocol for evaluating for AI involves measuring cortisol levels to determine whether steroids can be stopped safely and whether an infant needs an adrenocorticotropin hormone (ACTH) stimulation test to assess for AI. There is little literature regarding whether cortisol levels in neonates are predictive of AI. The objective of this study was to evaluate whether cortisol levels obtained, per our NICU protocol, in neonates after a prolonged (≥ 10 days) steroid course are predictive of AI.

Methods: Data were collected on neonates born 2017-2022 who received ≥ 10 days of systemic steroids and had a cortisol level measured after the steroid course. For each steroid course, data were collected on whether the neonate developed signs of AI, as defined as:

1) The clinical team restarting steroids due to signs of AI,
2) Signs of AI (hypotension, hypoglycemia, or hyponatremia) that caused the clinical team to intervene (besides restarting steroids) within seven days of stopping steroids, or
3) A failed ACTH stimulation test (obtained routinely for cortisol levels of 3-12 mcg/dL).

In the specified time period, 71 cortisol levels (in 67 babies) met criteria to be included in the analysis -- measured after at least 10 days of consecutive steroids, after a single held steroid dose, and within 7 days of stopping steroids. Cortisol levels were compared using Wilcoxon rank sum test since the levels were not normally distributed. A P-value of < 0.05 was considered significant.

Results: Analyses showed no difference between cortisol levels using the indicators of AI described in Table 2. Of 27 neonates who at some point had a low cortisol level/levels ( < 3 mcg/dL), 24 received extra steroids because of the low levels, for a median of 25 extra days (interquartile range 13-43 days).

Conclusion: These findings demonstrate that our current protocol based on routinely measuring cortisol levels to determine when it is safe to stop steroids is not diagnostic since there was no difference between cortisol levels obtained on neonates who developed signs of AI and those who did not. Clinical AI was rarer than expected, despite many of these patients being “labeled” as having AI in the medical record. Only 5.5% of ACTH stimulation tests performed on neonates with moderately low cortisol levels (3-12 mcg/dL) resulted in a failed test (cortisol < 18 mcg/dL). Our results show that neonates have been receiving unnecessary prolonged steroid courses in response to low cortisol levels. We are revising our NICU protocol with the goal of reducing unnecessary steroid treatment, blood draws, and associated cost of cortisol level testing.