AI identifies three distinct types of rare vasculitis, study says
Findings could lead to more individualized treatment of EGPA
Written by |
A type of artificial intelligence (AI) identified three main forms of eosinophilic granulomatosis with polyangiitis (EGPA), each characterized by distinct biological changes and disease manifestations, according to a new study.
Findings suggest that the three groups may also differ in how patients respond to treatment. Although further work to validate and expand on these findings is necessary, researchers are hopeful that the results could eventually help facilitate more individualized treatment for EGPA.
An early-access version of the study, “Identification of clinical-biological endophenotypes in patients with eosinophilic granulomatosis with polyangiitis using unsupervised machine learning: a real-world study,” was published in Respiratory Research. The work was supported by the National Natural Science Foundation of China.
EPGA is rarest form of ANCA-associated vasculitis
EGPA is the rarest form of ANCA-associated vasculitis (AAV), a group of autoimmune conditions marked by inflammation in small blood vessels that damages organs.
This rare form is highly variable in terms of how symptoms manifest and how it progresses. While it usually involves the lungs and respiratory system, it can affect many different parts of the body as well. EGPA is also marked by granulomas, or clumps of inflammatory cells, and high levels of immune cells called eosinophils (eosinophilia).
Notably, most EGPA patients do not test positive for self-reactive antibodies, called ANCAs, that drive most AAV cases.
“It is generally accepted that ANCA-positive patients are more likely to exhibit features of classic [body-wide] small-vessel [inflammation], whereas ANCA-negative patients more commonly present with prominent eosinophilia and tissue infiltration,” the researchers wrote.
However, “numerous clinical observations indicate substantial variations in disease progression trajectories, organ damage patterns, and responses to different therapeutic modalities … , even among patients within the same ANCA status group,” they added.
AI analyzed data from more than 200 patients
To better understand EGPA’s variability, a team of scientists in China turned to machine learning, a type of AI where a computer is given a large data set, then uses complex mathematical rules to identify patterns in the data and make predictions.
The researchers specifically hoped to identify EGPA endophenotypes, or groups of patients with distinct biomarker and clinical profiles. After feeding the AI tool with clinical and biomarker data from more than 200 EGPA patients who were seen at a clinic in Beijing between 2015 and 2023, three endophenotypes were identified.
“This study is the first to utilise unsupervised machine learning on multidimensional real-world clinical data to identify and validate three stable and distinct clinical-biological endophenotypes in EGPA,” the researchers wrote.
Three types differed in characteristics
The most common endophenotype, including 38% of patients, was dubbed the Eosinophilic-Airway Inflammation Type. This group was characterized by exceptionally high eosinophil levels, as well as high levels of a specific type of disease-associated antibody called IgE.
This type also showed the lowest rate of ANCA-positivity (3.8%). All patients had airway involvement, including asthma, and few showed disease involvement in other organs or systems.
The second group, which included 31.7% of patients, was dubbed the ANCA-Systemic Vasculitis Type. These patients were the most likely to test positive for ANCAs (26.2%) and showed the most severe body-wide disease activity and high levels of an inflammatory marker called CRP.
This group was also marked by multiorgan involvement. In addition to airway involvement, which was also present in most patients, these patients showed the highest rates of damage to nerves outside the brain and spinal cord (76.9%), and skin involvement (36.9%).
The third group, dubbed the Paucieosinophilic-Multi-Organ Damage Type, included 30.2% of patients. It was characterized by relatively low levels of eosinophils and the highest rates of digestive tract (29%) and heart (14.5%) involvement.
Preliminary classification using routine indicators can help inform early differentiation of treatment strategies, optimise patient outcomes, improve treatment efficiency, and provide more precise direction for drug development and clinical trial design in this field.
Treatment strategies and short-term responses also varied across the three groups. For example, the approved therapy Nucala (mepolizumab) was most commonly used in the Eosinophilic-Airway Inflammation Type group (70.5% vs. 19.4% to 27.7%), and this group showed the highest six-month remission rate (94.9% vs. 80.7% to 86.2%).
Because the analysis was based on real-world data, where each patient’s doctors were making individualized calls about treatment based on the person’s specific situation, the links between disease type and treatment response “should be considered exploratory and hypothesis-generating,” the team wrote.
In order to make their findings accessible to other scientists and clinicians, the researchers developed an online tool that predicts which of the three EGPA forms a person falls into based on 14 key clinical metrics. In an independent group of 30 EGPA patients, the tool showed 90% accuracy for making classifications.
“Preliminary classification using routine indicators can help inform early differentiation of treatment strategies, optimise patient outcomes, improve treatment efficiency, and provide more precise direction for drug development and clinical trial design in this field,” the researchers wrote. “We advocate for the prospective validation and refinement of this classification in multicentre, multidisciplinary [studies] before its incorporation into management guidelines.”
Leave a comment
Fill in the required fields to post. Your email address will not be published.