Endotyping allergic rhinitis in children: A machine learning approach

Introduction The diversity of allergic rhinitis (AR) phenotypes is particularly evident in childhood, suggesting the need to analyze and identify new approaches to capture such clinical heterogeneity. Nasal cytology (NC) is a very useful diagnostic tool for identifying and quantifying nasal inflammation. Data-driven approaches such as latent class analysis (LCA) assign subjects to classes based on their characteristics. We hypothesized that LCA based on NC, including the assessment of neutrophils, eosinophils, and mast cells, may be helpful for identifying AR endotypes in children. Methods A total of 168 children were enrolled. Sociodemographic characteristics and detailed medical history were obtained from their parents. All children performed NC and skin prick tests. LCA was applied for identifying AR endotypes based on NC, using the R package poLCA. All the statistical analyses were performed using R 4.0.5 software. Statistical significance was set at p <= .05. Results LCA identified two classes: Class 1 (n = 126, 75%): higher frequency of children with moderate/large number of neutrophils (31.45%); almost all the children in this class had no mast cells (91.27%) and Class 2 (n = 42, 25%): higher frequency of children with moderate/large number of eosinophils (45.24%) and moderate/large number of mast cells (50%). Conclusions The present study used a machine learning approach for endotyping childhood AR, which may contribute to improve the diagnostic accuracy and to deliver personalized health care in the context of precision medicine.

Automated summary

This study used a machine learning approach to identify different endotypes of childhood allergic rhinitis (AR) based on nasal cytology (NC). The results showed that LCA can help identify different AR phenotypes, supporting personalized healthcare in the context of precision medicine.