Growth Hormone Assay-Adjusted Standardization Reveals Distinct Clinical Phenotypes in Acromegaly

To identify distinct clinical phenotypes in acromegaly based on growth hormone (GH) assay standardization and unsupervised machine learning.

This was a multicenter cross-sectional analysis of 416 patients diagnosed with acromegaly from 2010 onward. Patients were stratified according to baseline serum GH levels standardized to the assay-specific upper limit of normal (GHxULN) using a binary classification (GH-B: <1.0×ULN vs ≥1.0×ULN) and a four-tier classification (GH-4: <0.25, 0.25-0.99, 1.0-9.9, ≥10×ULN). Unsupervised cluster analysis included age, GHxULN, insulin-like growth factor 1 (IGF-1)xULN, tumor diameter, and T2-weighted signal intensity.

Overall, 36% of patients had GH levels within the normal reference range for their assay (GH-B <1.0×ULN). Microadenomas (23.1%) were more frequent in older patients and associated with lower GH/IGF-1 levels. Across GH-4 categories, significant gradients were observed for age (z = -5.34, P < .001), tumor size (z = 8.01, P < .001), IGF-1 (z = 9.00, P < .001), and symptom duration (z = 4.34, P < .001). Higher GH categories were associated with greater odds of arthropathy (odds ratio 3.5, P = .015 for 1.0-9.9×ULN and odds ratio 6.58, P = .002 for ≥10×ULN). Cluster analysis revealed 3 phenotypes: cluster 1 (49.0%) [older age, lower GH/IGF-1, intermediate tumor size]; cluster 2 (44.4%) [intermediate age, moderate biochemical activity, smaller tumors]; cluster 3 (6.6%) [younger age, markedly elevated GH/IGF-1, large aggressive tumors].

GH standardization to assay-specific ULN reveals clinically meaningful phenotypes in acromegaly that correlate with age, tumor characteristics, and disease severity (particularly arthropathy). GHxULN complements IGF-1 by capturing tumor secretory activity, and this stratification approach may support more individualized clinical decision-making.