# Integrating ultrasound and clinical risk factors to predict carotid plaque vulnerability in gout patients: a machine learning approach

## TL;DR

This study uses machine learning to predict carotid plaque vulnerability in gout patients by combining ultrasound data and clinical factors.

## Contribution

A novel machine learning model integrating gout-specific and cardiovascular factors for predicting carotid plaque vulnerability.

## Key findings

- Tophi, power Doppler signal grades, and gout flare frequency are independent risk factors for carotid plaque vulnerability.
- The random forest model achieved excellent predictive performance with a C-index of 0.997.
- The model bridges rheumatic inflammation and cardiovascular risk assessment in gout patients.

## Abstract

This study aimed to identify independent risk factors for carotid plaque (CP) vulnerability in patients with gout and to develop a predictive model incorporating both gout-specific and cardiovascular factors.

This study was designed as a retrospective cohort analysis that enrolled patients with newly diagnosed gout. These patients were retrospectively followed for a period of 1 to 2 years to evaluate the incidence of CP vulnerability. CP vulnerability was assessed using standardized ultrasound examinations and graded according to the Plaque Reporting and Data System (Plaque-RADS). Multivariate ordinal logistic regression analysis was employed to identify independent risk factors associated with CP vulnerability, with a particular focus on the impact of gout-related variables. Based on these results, a random forest prediction model was developed by integrating ultrasound imaging features and clinical variables to predict CP vulnerability.

Tophi (OR = 1.760, p = 0.009), power Doppler (PD) signal grades (Grade 2: OR = 1.540, p = 0.002; Grade 3: OR = 1.890, p = 0.001), and the number of gout flares in the last year (OR = 1.524, p = 0.001) were identified as independent risk factors for CP vulnerability. The random forest model showed excellent predictive performance (C-index = 0.997) and highlighted tophi, PD signal grades, and gout flare frequency as key gout-specific contributors to CP risk.

The presence of tophi, positive PD signals, and increased number of gout flares are significantly associated with CP vulnerability in patients with gout. The proposed machine learning model, integrating gout-specific and cardiovascular factors, provides a novel and effective approach for personalized risk stratification and management in gout patients, bridging the gap between rheumatic inflammation and cardiovascular risk assessment.