OP2-4 - Effects of Antihypertensive Medications on the Risk of Alzheimer’s Disease and Other Dementias: A Machine Learning Approach

Monday, April 24, 2023

4:15 PM – 5:45 PM GMT

Author(s)

Xiaomo Xiong, MS

Research Assistant
College of Pharmacy, University of South Carolina
West Columbia, South Carolina, United States
Sam Li, PhD

Assistant Professor
University of Tennessee Health Science Center, United States
Xiaoxia Wang

PharmD Student
University of South Carolina
Columbia, South Carolina, United States
GL

Gang Lv, MD

Deputy Chief Physician
Chinese PLA General Hospital, United States

Presenting Author(s)

KL

Kevin Lu, PhD, FISPE

Associate Professor
University of South Carolina
Columbia, South Carolina, United States

ABSTRACT:

Background: Clinical trials are time and resource-consuming. With very few effective antidementia drugs available for ADRD patients, repurposing drugs already approved by the Food and Drug Administration (FDA) for another indication among ADRD patients may be a more rapid and less expensive option. The literature has documented that a sizable proportion of patients with Alzheimer’s disease and related dementias (ADRD) have comorbid hypertension. However, no large-scale studies have yet been conducted to determine what factors are associated with the use of anti-hypertensive drugs among ADRD patients, whether certain antihypertensive medications are effective in reducing the risk of Alzheimer’s disease, and/or if there are associated racial/ethnic disparities.

Objectives: To identify factors associated with use of antihypertensive drugs, the effects of antihypertensive medications on the risk of ADRD, and the potential racial/ethnic disparities. This study aimed to develop unique algorithms for factors of antihypertensive drug use and identify candidate antihypertensive drugs that can lower the risks of ADR based on real-world big data and machine-learning approaches.

Methods: We will leverage the unique nationally representative survey data of the Medicare Current Beneficiary Survey (MCBS). The MCBS has been collecting data on Medicare beneficiaries for over 30 years and has conducted over 1 million interviews. These interviews have been linked to the Medicare Part D data by the Centers for Medicare & Medicaid Services (CMS) and have been proven to be highly validated over the past 30 years. Because of the unique national survey data that contain rich information, we were able to investigate large clusters of vectors at personal, institutional, and environmental levels, which include but were not limited to demographics, behavioral factors, and environmental factors (e.g., rural-urban residence, census regions, health segregation, and criminal records). The eXtreme Gradient Boosting (XGboost) approach was used to examine the effects of antihypertensive medications on the risk of ADRD. In addition, SHapley Additive exPlanations (SHAP) were used to identify the contribution and association of features to the risk of ADRD. All analyses were conducted using python 3.8.

Results: 33,590 individuals were included, representing 43,506,401 Medicare beneficiaries. One-third of the included people (10,077) were assigned to the training group, and two-thirds (23,513) were assigned to the test group. The top factors associated with the risk of ADRD were identified out of over a hundred factors. For instance, the top 5 personal factors were comorbidities, age, general health status, BMI, and marital status. Based on the area under the Receiver Operating Characteristic (ROC) curve, the prediction performance was 0.51. According to the confusion matrix, the accuracy and precision of the model were 94.5% and 48.5%, respectively. The mean absolute SHAP value illustrated that the average effect of antihypertensive drugs on model output were ARBs (0.074), ACE-Is (0.058), calcium channel blockers (0.055), diuretics (0.044), and beta-blockers (0.036).

Conclusion: In this study, we developed a unique framework and machine-learning algorithms for evaluating the potential relationships between antihypertensive drugs and medication access among Medicare beneficiaries with ADRD. Candidates for repurposing antihypertensive drugs included ARBs, ACE-Is, and calcium channel blockers.