MSACL 2025 Abstract

Interlaboratory Transferability of a Multi-Steroid LC-MS/MS Method Coupled With Machine Learning for Primary Aldosteronism Subtype Classification

Kevin Mantik (1), Renee Sahertian(1), Christopher Hodgkins(1), Daniel Yagoub(1), Laura K Buckton(1), Andrea R Horvath(1), Graeme Eisenhofer (2), Mirko Peitzsch (3), Catleen Conrad (3), Manuel Schulze (4), David Marshall (5), Brian Keevil (5).
(1) New South Wales Health Pathology, Department of Chemical Pathology, Prince of Wales Hospital, Sydney, Australia; (2) Department of Medicine III, University Hospital Carl Gustav Carus, Technische Universität Dresden, Germany; (3) Institute of Clinical Chemistry and Laboratory Medicine, University Hospital and Faculty of Medicine Carl Gustav Carus, Technische Universität Dresden, Germany; (4) Center for Interdisciplinary Digital Sciences, Department of Information Services and High Performance Computing, Technische Universität Dresden, Germany; (5) Department of Biochemistry, Manchester University NHS Foundation Trust, Manchester, United Kingdom.

Renee Sahertian, B. Sc. (Presenter) NSW Health Pathology

Presenter Bio: I am a hospital scientist with 12 years’ experience in clinical mass spectrometry, based in Sydney, Australia. My clinical mass spectrometry experience covers the development, validation, and operation of assays within Endocrinology and Clinical Chemistry, with particular emphasis on steroid analysis and the use of integrated systems, including robotics.

Relevant Financial Disclosures (within past 24 months, reported on Aug 15, 2025) No relevant financial relationship(s) to disclose.

INTRODUCTION:
The diagnosis of primary aldosteronism (PA) typically involves a combination of biochemical screening and confirmatory testing, imaging, and invasive procedures such as adrenal venous sampling (AVS) 1. The emergence of peripheral steroid profiling combined with machine learning (ML) shows promise to simplify this diagnostic pathway2,3. A collaboration with the international multicenter PROspective study on the diagnostic value of Steroid PROfiling in primary ALDOsteronism (PROSALDO) has led to local optimisation and validation of a multi-steroid panel by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Our aim was to assess the reproducibility of steroid data and ML probability scores for PA subtype classification between laboratories.

METHODS:
In Sydney, Australia, an LC-MS/MS method was developed and validated to efficiently measure 15 steroids in serum and plasma. Using patient samples from the PROSALDO study, steroid concentrations and associated machine learning-generated probability scores were compared with the Technische Universität Dresden, Germany. Steroid profiling with ML probability was compared to conventional diagnostic tests including the aldosterone-to-renin ratio (ARR), saline infusion testing (SIT), and AVS. Pooled patient samples with and without PA were also analysed in Sydney, Dresden, and Manchester UK for steroid concentrations and ML probability imprecision and agreement.

RESULTS:
The average total imprecision of our method ranged from 3.0% to 12.9% for all steroids. A comparison of 77 patient samples from 25 PROSALDO patients showed Passing-Bablok regression slopes between 0.93 and 1.11, and Bland-Altman relative differences ranging from -10.9% to +8.7% compared to the Dresden reference LC-MS/MS method. Implementation of ML algorithms also yielded comparable diagnostic probability scores with improved imprecision between Sydney, Manchester, and Dresden. In two local patient cases, steroid profiling and ML correctly identified unilateral PA subtypes despite an unsuccessful AVS procedure and a false-negative ARR screen, respectively.

CONCLUSION:
Our LC-MS/MS steroid profiling method demonstrates acceptable analytical performance and good agreement with independent methods at Dresden and Manchester. Inter-laboratory comparisons of machine learning-derived probability scores for discriminating PA subtypes and primary hypertension were also effectively reproducible.

REFERENCES
1. Funder JW, Carey RM, Mantero F, Murad MH, Reincke M, Shibata H, et al. The Management of Primary Aldosteronism: Case Detection, Diagnosis, and Treatment: An Endocrine Society Clinical Practice Guideline. J Clin Endocrinol Metab. 2016;101(5):1889-916.
2. Mirko Peitzsch, Tanja Dekkers, Matthias Haase, Fred C.G.J. Sweep, Ivo Quack, Gerald Antoch, Gabriele Siegert, Jacques W.M. Lenders, Jaap Deinum, Holger S. Willenberg, Graeme Eisenhofer, An LC–MS/MS method for steroid profiling during adrenal venous sampling for investigation of primary aldosteronism, The Journal of Steroid Biochemistry and Molecular Biology, Volume 145, 2015, Pages 75-84.
3. Eisenhofer G, Durán C, Cannistraci CV, et al. Use of Steroid Profiling Combined with Machine Learning for Identification and Subtype Classification in Primary Aldosteronism. JAMA Netw Open. 2020;3(9):e2016209.