Multicenter Study on the Development of Pulmonary Arterial Hypertension Screening Models Based on Artificial Intelligence for Patients With Systemic Sclerosis

Pulmonary arterial hypertension (PAH) is a rare and serious disease, affecting fewer than 50 people per million inhabitants. Its diagnosis requires right heart catheterization, an invasive procedure. PAH is a diverse condition and is often linked to autoimmune diseases such as systemic sclerosis (SSc), which affects about 277 people per million inhabitants in Spain, meaning that over 12,000 people may have the disease in the country. PAH develops in around 10% of SSc patients and is the main cause of death in this group. Although there is no cure, pulmonary vasodilator drugs have helped patients live longer, sometimes at the cost of reduced quality of life.

In more advanced stages of PAH, continuous intravenous or subcutaneous therapies are often needed. Traditional treatments mainly focus on widening the blood vessels in the lungs to reduce heart problems. More recently, new drugs have been developed that act directly on the mechanisms causing the disease, with the goal of improving blood flow in the lungs.

Artificial intelligence (AI) and a better understanding of disease mechanisms are changing healthcare. However, it is not yet known how useful AI might be in screening, diagnosing, and predicting outcomes in patients with SSc-associated PAH (SSc-PAH). In past decades, screening programs using clinical data, lab tests, and echocardiography have been developed to detect PAH before symptoms appear. These programs have helped identify patients earlier and reduce mortality. However, their low specificity can lead to many unnecessary right heart catheterizations. This problem may have increased since the 2022 update of pulmonary hypertension diagnostic criteria, which now use less strict hemodynamic thresholds, potentially making early diagnosis more difficult.

This is an ambispective observational study, combining retrospective data from existing patient records with prospective follow-up of newly enrolled patients.

The aim is to improve early detection of PAH in SSc patients by using AI-based algorithms that integrate simple and reproducible clinical data, such as electrocardiograms and echocardiograms. It is expected that these AI models will perform better than traditional screening programs, allowing earlier detection of PAH in many patients. Earlier and more accurate screening could also reduce the number of unnecessary invasive procedures, benefiting both clinical outcomes and patients' experience of their health.

The study will also examine protein expression in SSc-PAH patients, detailed measures of right heart function using echocardiography at rest and during exercise, and patient-reported health status. This will help determine how useful these factors are for predicting outcomes and for guiding treatment, supporting more personalized care and improving both clinical results and patient-reported health.

Through the collaboration of reference centers for pulmonary hypertension and systemic autoimmune diseases, together with patient associations, this study aims to ensure that many affected patients can access earlier and better care, ultimately improving survival and quality of life.

Inclusion criteria

• Age ≥ 18 years
• Clinical diagnosis of systemic sclerosis (SSc) according to ACR/EULAR criteria
• For controls (SSc without PAH): absence of pulmonary arterial hypertension; patients with isolated or combined post-capillary pulmonary hypertension (pulmonary capillary pressure > 15 mmHg) or Group 3 pulmonary hypertension may be included, limited to 20% of this group
• For cases (SSc-associated PAH): confirmed PAH by right heart catheterization (mean pulmonary arterial pressure > 20 mmHg, pulmonary capillary pressure < 15 mmHg, pulmonary vascular resistance > 2 Wood Units)

Exclusion criteria

• Missing data in the main variables at diagnosis (clinical assessment, blood tests, electrocardiogram, transthoracic echocardiogram).
• Inability to provide informed consent