Study Finds ML Can Aid Blood Test for Rare Adrenal Tumors

Research presented at Association for Diagnostics & Laboratory Medicine (ADLM) 2026 found that ML models could improve interpretation of plasma free metanephrine blood tests for rare adrenal and related tumors, ADLM stated in a press release. The study analyzed data from 20,516 adults tested at Samsung Medical Center from 2011 to 2024 for pheochromocytomas and paragangliomas, known as PPGL.

Plasma free metanephrines are the recommended initial test for PPGL. Among 19,797 patients who did not have PPGL, 25.2 percent had elevated metanephrine results that could have led to false positive findings.

Researchers tested several ML algorithms that combined metanephrine results with structured clinical information from electronic health records. The added data included kidney and urine biomarkers, medications, and other diseases.

Further analysis found that much of the apparent improvement came from shortcut learning. The models learned patterns in which follow up tests were ordered, which can reflect clinician suspicion of PPGL rather than independent biochemical signals. The ADLM 2026 presentation covers the testing process and a revised ML algorithm developed after that analysis.