Sensitization trajectories in childhood revealed by using a cluster analysis
Assessment of sensitization at a single time point during childhood provides limited clinical information. We hypothesized that sensitization develops as specific patterns with respect to age at debut, development over time, and involved allergens and that such patterns might be more biologically and clinically relevant.Objective:
We sought to explore latent patterns of sensitization during the first 6 years of life and investigate whether such patterns associate with the development of asthma, rhinitis, and eczema.Methods:
We investigated 398 children from the at-risk Copenhagen Prospective Studies on Asthma in Childhood 2000 (COPSAC2000) birth cohort with specific IgE against 13 common food and inhalant allergens at the ages of ½, 1½, 4, and 6 years. An unsupervised cluster analysis for 3-dimensional data (nonnegative sparse parallel factor analysis) was used to extract latent patterns explicitly characterizing temporal development of sensitization while clustering allergens and children. Subsequently, these patterns were investigated in relation to asthma, rhinitis, and eczema. Verification was sought in an independent unselected birth cohort (BAMSE) constituting 3051 children with specific IgE against the same allergens at 4 and 8 years of age.Results:
The nonnegative sparse parallel factor analysis indicated a complex latent structure involving 7 age- and allergen-specific patterns in the COPSAC2000 birth cohort data: (1) dog/cat/horse, (2) timothy grass/birch, (3) molds, (4) house dust mites, (5) peanut/wheat flour/mugwort, (6) peanut/soybean, and (7) egg/milk/wheat flour. Asthma was solely associated with pattern 1 (odds ratio [OR], 3.3; 95% CI, 1.5–7.2), rhinitis with patterns 1 to 4 and 6 (OR, 2.2–4.3), and eczema with patterns 1 to 3 and 5 to 7 (OR, 1.6–2.5). All 7 patterns were verified in the independent BAMSE cohort (R2> 0.89).Conclusion:
This study suggests the presence of specific sensitization patterns in early childhood differentially associated with development of clinical outcomes. Using such patterns in future research might provide more robust and clinically relevant results.