CAI Score for the Diagnosis of Central Adrenal Insufficiency

Diagnosing central adrenal insufficiency (CAI) is challenging in patients with inconclusive morning cortisol (4-18 µg/dL). Dynamic tests like the short Synacthen test (SST) are imperfect, necessitating reliable tools to stratify CAI risk and reduce diagnostic delays.

To develop and validate a predictive scoring system integrating clinical, biochemical, and imaging variables for CAI diagnosis in patients with indeterminate morning cortisol levels.

This is a retrospective study of 341 adults with suspected CAI and indeterminate morning cortisol. CAI was confirmed via SST (peak cortisol <18 µg/dL). Multivariate analysis identified key predictors for the CAI score, including morning cortisol, pituitary hormone deficits, tumor size, and treatment history. A machine learning model was also developed to enhance the prediction accuracy.

Lower morning cortisol (6.27 vs 10.29 µg/dL, P < .0001), male sex [odds ratio (OR) 1.77, P = .011], larger pituitary tumors (2.46 vs 1.89 cm, P = .044), and ≥3 pituitary hormone deficits (OR 35.38, P = .001) independently predicted CAI. The CAI score (range 0-13.5 points) stratified risk, with scores ≥4.5 indicating high CAI likelihood (70.1% vs 10.5% at 0 points). Combining serum cortisol and pituitary hormone deficits improved diagnostic accuracy [area under the curve (AUC) 0.745] over cortisol alone (AUC 0.680). A web-based tool (https://cai-predictor.streamlit.app/) was created for convenient clinical application.

The CAI score improves diagnostic accuracy in ambiguous cases of suspected CAI by integrating morning serum cortisol with key clinical parameters. As morning serum cortisol and SST have limitations in CAI, the CAI score may act as a supportive rather than a standalone tool in the evaluation of patients for CAI.