ECE2022 Poster Presentations Late-Breaking (41 abstracts)
Recalculating renin and aldosterone to improve specificity in the diagnosis of primary aldosteronism
Luc Doyle 1 , Julie Okiro 1 , Aysha Sarwani 1 , Michael Troy 2 , Yousef Ansari 1 , Darragh O′Donoghue 1 , David Lappin 3 , John Mcevoy 4 , Paula O’Shea 2,5 , John Ferguson 6 & Michael Conall Dennedy 7
1University Hospital Galway, Centre for Diabetes, Endocrinology and Metabolism, Galway, Ireland; 2School of Medicine, National University of Ireland Galway (NUIG); 3University Hospital Galway, Department of Nephrology, Galway, Ireland; 4National University of Ireland Galway (NUIG), School of Medicine, Galway, Ireland; 5University Hospital Galway, Department of Clinical Biochemistry, Saolta University Health Care Group (SUHCG), Galway, Ireland; 6National University of Ireland Galway (NUIG), Biostatistics Unit, HRB Clinical Research Facility, NUI Galway, Galway, Ireland; 7National University of Ireland Galway (NUIG), Discipline of Pharmacology and Therapeutics, Lambe Institute for Translational Research, NUI Galway, Galway, Ireland
Rationale: The Aldosterone:Renin ratio (ARR) is commonly used for patients fulfilling screening criteria for primary hyperaldosteronism (PA), followed by confirmatory testing. Reference intervals for interpretation of the ARR vary in accordance with local population and assay 1. While ARR provides high sensitivity for PA, this is compromised by low specificity, further compounded by medication interference. However, additional variables may improve the specificity of ARR as a diagnostic test, potentially mitigating the need for confirmatory testing in all patients. In the current study, we investigated the relationship between aldosterone and renin as single diagnostic measure for PA.
Methodology: A population-based study of individuals attending a specialist hypertension clinic was performed. PA was screened using the ARR, and confirmed using the saline infusion test. 82 patients with matched ARR and Saline Infusion Tests were investigated. Logistic Regression was used to estimate the relationships between renin, aldosterone, ARR, clinical variables and the probability of a diagnosis of PA. Predictive capacity of each model was measured using Area Under the Curve using “leave one out” cross validation to avoid overfitting.
Results: The AUC for the model using ARR on its own was estimated to be 0.68. The model, excluding ARR, but including renin, aldosterone and their interaction on the log-scale: log(Renin) + log(Aldosterone) + log(Renin)*log(Aldosterone), improved the AUC to 0.73. Using this log-structure, as opposed to including aldosterone and renin as linear effects in a logistic model, makes sense since the model involving ARR alone is nested within the log-structure model, which would not be true under the linear model. Covariates including eGFR, serum potassium and the presence of an adrenal nodule were then individually tested for statistical significance, conditional on the choice of this log-scale interactive model. The model Log(Renin) + log(Aldosterone) + log(Renin)*log(Aldosterone) + Adrenal_Nodule provided the highest performance with an AUC of 0.782. For the ARR and the log-model, at a sensitivity of 80%, specificity was 37.5% and 64% respectively, and for a sensitivity of 98%, specificity was 12.5% and 24% respectively.
Conclusion: This log model incorporating the adrenal nodule as a variable improved the AUC from 0.68 (model with ARR alone) to 0.782. This study highlights the importance of statistically re-visiting well-established calculations to better inform clinical practice. Ongoing validation of our findings is proceeding in other clinical samples.
Volume 81
Article tools
My recent searches
My recently viewed abstracts
Authors
Endocrine Abstracts
ISSN 1470-3947 (print) | ISSN 1479-6848 (online)
© Bioscientifica 2024 |
Privacy policy |
Cookie settings
BiosciAbstracts
Bioscientifica Abstracts is the gateway to a series of products that provide a permanent, citable record of abstracts for biomedical and life science conferences.
Find out more