113 - Adrenal Insufficiency of Prematurity and Early Cortisol Levels: Adjusted Associations and Clinical Implications

Publication Number: 2108.113

Jessica Munck, University of Connecticut School of Medicine, Farmington, CT, United States; Madison L. Fitzgerald, University of Connecticut School of Medicine, Farmington, CT, United States; Mariann Pappagallo, University of Connecticut School of Medicine, Farmington, CT, United States; Ted S. Rosenkrantz, University of Connecticut School of Medicine, Farmington, CT, United States; Naveed Hussain, Connecticut Children's Medical Center, Farmington, CT, United States

Background: Low random cortisol in the first postnatal week may signal developmental adrenal insufficiency or inadequate stress response. Adrenal insufficiency of prematurity (AIP) presents with oliguria, hypotension, creatinine rise, and escalating vasoactive support, but early cortisol level relation to AIP in practice is unclear.

Objective: To identify perinatal/postnatal factors associated with low cortisol (first level) and the clinical association with AIP.

Design/Methods: In an IRB-exempt, single-center, retrospective cohort of VLBW infants < 32 weeks GA born 2012-2024 with random cortisol obtained ≤7 days of life, infants were grouped as LC (low cortisol; < 5 µg/dl) or Non-LC (≥ 5 µg/dl). Potential perinatal risk factors and postnatal outcomes including gestational age (GA-wk), birthweight (BW-g), day‑of‑life at cortisol sampling, ANS, Apgar 1/5, cord pH, SNAP/SNAPPE, sepsis, hypotension/CV insufficiency, oliguria, pressors/inotropes, and serum creatinine were analyzed for hypothesized correlations. Multivariable logistic regression models were generated; the LC model included AIP as a covariate, while the AIP model included LC plus potential clinical covariates with GA/BW adjustment. We report adjusted odds ratios (aORs, 95% CIs) and diagnostics.

Results: Of 884 VLBW infants < 32-wk GA, 87 had random cortisol levels in the first 7 days of life (LC=30, Non-LC=57). In the LC model, adjusted effects were: BW aOR 1.14, p=0.241; Apgar1 aOR 0, p=0.257; creatinine aOR 0, p=0.258; SNAP1 aOR 0, p=0.264; Apgar5 aOR .39, p=0.266; dol aOR 0.11, p=0.272; oliguria aOR 0, p=0.275; GA aOR 0, p=0.287; hypotension aOR 0, p=0.287; SNAP2 aOR 0.94, p=0.595; pressors aOR 0.06, p=0.886; AIP aOR 0, p=0.994; sepsis aOR 0, p=1.000. In the AIP model, adjustments showed no statistically significant independent predictor. Markers of high illness severity tended to associate with lower odds of LC class, consistent with stress‑response rise (wide CI). Figure 1 shows higher AIP among LC infants but is imprecise. Ridge estimates were directionally similar. Point estimates suggested lower odds of LC among infants with sepsis/hypotension, consistent with stress-response (wide CI).

Conclusion(s): In this single-center cohort, no independent predictors of LC or AIP were identified after multivariable adjustment. Trends aligned with expected physiology but were imprecise. These data underscore need for larger, prospective, and phenotype‑refined studies that standardize sampling, use dynamic testing (ACTH stimulation), and clarify indication for steroid testing/treatment.

Figure 1
AIP prevalence by cortisol category (95% CI).

Figure 2
Predictors of low cortisol, with AIP included, using adjusted odds ratio (log scale).

Table 1
Demographics and clinical characteristics of groups.