# Exploration of factors influencing pulmonary injury in patients with anti-neutrophil cytoplasmic antibody-associated vasculitis and construction of a predictive model

## TL;DR

This study identifies risk factors and builds a model to predict pulmonary involvement in AAV patients, aiming to enable early detection and better monitoring.

## Contribution

A new predictive model for pulmonary involvement in AAV using clinical markers and validated internally.

## Key findings

- Four independent predictors were identified: intermittent symptoms, limb numbness, older age, and lower hemoglobin.
- The model showed good discrimination with an AUC of 0.88 in training and 0.83 in validation.
- Calibration and decision curve analysis confirmed the model's clinical utility.

## Abstract

A bad prognosis results from pulmonary involvement of anti-neutrophil cytoplasmic antibody-associated vasculitis (AAV). A method for early detection of AAV patients at risk of developing pulmonary involvement is still required, despite the existence of predictive models for death. The purpose of this study was to determine risk factors and develop a predictive model for pulmonary involvement in AAV.

Seventy-one treatment-naïve AAV patients (38 with pulmonary involvement) participated in a retrospective cross-sectional investigation. To choose predictors and create a nomogram, multivariate logistic regression and the least absolute shrinkage and selection operator (LASSO) were employed. Receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA) with a 7:3 training-to-validation split were used to assess the model.

Four independent predictors were identified: intermittent symptom pattern (OR = 8.52, 95% CI: 2.17–33.46), limb numbness/paresthesia (OR = 12.21, 95% CI: 1.98–75.50), older age at onset (OR = 1.05 per year, 95% CI: 1.01–1.10), and lower hemoglobin level (OR = 0.97 per g/L, 95% CI: 0.95–0.99). A nomogram incorporating these factors demonstrated good discrimination, with an area under the curve (AUC) of 0.88 (95% CI: 0.77–0.98) in the training set and 0.83 (95% CI: 0.64–1.00) in the validation set. Calibration and decision curve analysis confirmed the model’s clinical utility.

Using standard clinical markers, we created and internally verified a pragmatic prediction model for pulmonary involvement in AAV. Personalized monitoring and early risk categorization may benefit from this approach. To verify its generalizability, external validation in prospective, multi-center cohorts is advised.