ANALYSIS OF RAAS BIOMARKERS TO CLASSIFY ANTI-HYPERTENSIVE DRUG TREATMENTS IN THE GENERAL POPULATION: THE CHRIS STUDY

Abstract

Objective: Hypertension is a leading cause of death worldwide. Large-scale population-based epidemiological studies are an opportunity to assess the effectiveness of anti-hypertensive drug (AHD) treatment. However, these studies often lack treatment documentation, especially in situations where linkage to electronic health records is not established, with consequent recall and classification biases. When relevant biomarkers are measured, it should however be possible to attribute individuals to the most likely AHD treatment, limiting information loss. We investigate whether unsupervised cluster analysis applied to measured RAAS biomarkers may help to classify different AHD treatments in a general population sample. Design and method: In the Cooperative Health Research in South Tyrol (CHRIS) study, automated drug barcode scans allowed classification of AHD specific treatment among participants. We measured angiotensin I, angiotensin II and aldosterone levels on 800 participants using liquid chromatography combined with tandem mass spectrometry (LC-MS/MS) analysis, as implemented in the RAAS Triple-A assay. Age and sex matched participants (age: 43–90 years; 54% females) were split into 8 balanced groups: normotensive (n = 101); untreated hypertensive (n = 100); non-AHD (n = 100); ACEi (n = 99); ARB (n = 98); ACEi+diuretic (n = 100); ARB+diuretic (n = 102); and beta-blockers (n = 100). We performed unsupervised cluster analysis based on the three biomarkers ignoring the known AHD treatment. We evaluated the extent of agreement between the automated drug recognition system and RAAS based classification. Results: We identified three clinically heterogeneous clusters. Cluster 1 (n = 444) was characterized by participants not receiving any AHD and having high systolic blood pressure. Cluster 2 (n = 235) identified the use of ARB with or without diuretics. These individuals were more likely to present diabetes. Cluster 3 (n = 121) identified ACEi with or without diuretics and included older and overweight participants. The inter-classification agreement between the automated recognition system and RAAS clusters was highest for ACEi and ARB. Conclusions: In the absence of specific information, unsupervised cluster analysis applied to RAAS biomarkers can reliably identify individuals under ACEi and ARB treatment.