Baseline biochemical evaluation for etiologic differentiation in ACTH-dependent Cushing's syndrome: machine learning and composite score analysis in a multicenter cohort of 566 patients
Authors' conclusion
Combining screening tests in composite scoring and machine learning differentiates subtypes of ACTH-dependent CS and offers a potential diagnostic alternative to the CRH stimulation test.
Study context
- Population
- Patients with confirmed ACTH-dependent Cushing's syndrome (CS) and available screening tests at diagnosis; 509 (90%) with Cushing's disease (CD)509 [90%] with CD
- Participants
- 566 patients
- Comparison
- Screening tests alone and combined for differentiating CD from ectopic CS (ECS)
- Outcome
- Diagnostic performance in differentiating CD from ECS
Methodological detailsretrospective, multicenter, composite score models, machine learning (balanced random forest), cut-offs by Youden's Index
- Retrospective
- Multicenter
- Composite score models
- Machine learning (balanced random forest)
- Cut-offs by Youden's Index
Performance
As reported in the abstract
Differentiation of CD from ECS
| Test or model | AUC | AUC 95% CI | AUC CI | Sensitivity | Specificity |
|---|---|---|---|---|---|
| Morning ACTH | 0.776 | .688-.853 | — not stated in the abstract | 74% | 77% |
| 24h-UFC | 0.854 | — not stated in the abstract | .816-.923 | 72% | 83% |
| 1 mg dexamethasone suppression test | .828 | — not stated in the abstract | .740-.906 | 70% | 86% |
| Composite score | 0.865 | — not stated in the abstract | .798-.933 | 78% | 92% |
| Composite score + pituitary magnetic resonance imaging (MRI) | 0.931 | — not stated in the abstract | .887-.974 | 88% | 85% |
| Machine learning | 0.853 | — not stated in the abstract | .741-.966 | 71% | 84% |
| Machine learning + MRI | 0.902 | — not stated in the abstract | .805-.999 | 76% | 91% |
AUC
Morning ACTH0.776
24h-UFC0.854
1 mg dexamethasone suppression test.828
Composite score0.865
Composite score + pituitary magnetic resonance imaging (MRI)0.931
Machine learning0.853
Machine learning + MRI0.902
AUC 95% CI
Morning ACTH.688-.853
24h-UFC— not stated in the abstract
1 mg dexamethasone suppression test— not stated in the abstract
Composite score— not stated in the abstract
Composite score + pituitary magnetic resonance imaging (MRI)— not stated in the abstract
Machine learning— not stated in the abstract
Machine learning + MRI— not stated in the abstract
AUC CI
Morning ACTH— not stated in the abstract
24h-UFC.816-.923
1 mg dexamethasone suppression test.740-.906
Composite score.798-.933
Composite score + pituitary magnetic resonance imaging (MRI).887-.974
Machine learning.741-.966
Machine learning + MRI.805-.999
Sensitivity
Morning ACTH74%
24h-UFC72%
1 mg dexamethasone suppression test70%
Composite score78%
Composite score + pituitary magnetic resonance imaging (MRI)88%
Machine learning71%
Machine learning + MRI76%
Specificity
Morning ACTH77%
24h-UFC83%
1 mg dexamethasone suppression test86%
Composite score92%
Composite score + pituitary magnetic resonance imaging (MRI)85%
Machine learning84%
Machine learning + MRI91%
— not stated in the abstract
Original publication
Baseline biochemical evaluation for etiologic differentiation in ACTH-dependent Cushing's syndrome: machine learning and composite score analysis in a multicenter cohort of 566 patients. Eur J Endocrinol. Published 31 July 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →