Disentangling the heterogeneity of allergic respiratory diseases by latent class analysis reveals novel phenotypes

Refined phenotyping of allergic diseases may unravel novel phenotypes. Conjunctivitis as an independent disorder has never been approached.

To identify distinct classes of allergic respiratory diseases using latent class analysis (LCA) and distinguish each class using classification and regression tree (CART) analysis.

Seven hundred and twenty-eight adults from the Portuguese general population study ICAR had a structured medical interview combined with blood collection, skin prick tests, spirometry with bronchodilation, and exhaled nitric oxide. LCA was applied to 19 variables. The CART algorithm selected the most likely variables distinguishing LCA-classes.

A six-class model was obtained. Class 1 (25%): nonallergic participants without bronchial or ocular symptoms. Classes 2 (22%) and 3 (11%): nasal and ocular (low levels) symptoms without nasal impairment, monosensitized (Class 2) or polysensitized (Class 3). Class 4 (13%): polysensitized participants with high levels of nasal and ocular symptoms, and nasal impairment. Classes 5 (16%) and 6 (14%): high level of nasal, bronchial and ocular symptoms with nasal impairment (non-allergic or polysensitized, respectively). Participants in classes 5 and 6 had more bronchial exacerbations and unscheduled medical visits (P < 0.001). Ocular symptoms were significantly higher in classes with nasal impairment, compared to those without impairment (P < 0.001) or no nasal symptom (P < 0.001). CART highlighted ocular symptoms as the most relevant variable in distinguishing LCA-classes.

Novel severe phenotypes of participants with co-occurrence of ocular, nasal and bronchial symptoms, and exacerbation-prone were identified. The tree algorithm showed the importance of the ocular symptoms in the expression of allergic diseases phenotypes.