For this application of IRT, SPT results at age two from 537 children enrolled in the CCAAPS cohort were utilized. Most (81%) of the children were Caucasian. Few mothers reported smoking (12%) and a majority of the mothers breastfed (71%). Maternal and paternal asthma were common in this cohort, 24% and 13%, respectively, as this was a high risk cohort. The frequency of sensitization to specific allergens at age two is presented in . The prevalence of sensitization to any allergen was 39.5%, and the most common was Timothy grass (7.8%) followed by egg (7.6%), while the least common allergen was milk (2%). A majority, 60.5%, were SPT- to all 17 allergens, 19.6% were positive to only one allergen, and 19.9% were positive to two or more allergens. Of those who were sensitive to at least two allergens (n=107), the most common cosensitization occurred between Timothy grass and maple (12%).
Frequency of Sensitization to Individual Allergens at Age Two
The IRT model was fit using all 17 allergen responses at age two. Difficulty (δ) is indicated by the location on the latent trait scale in which there is a 50% probability of being SPT+ to that allergen. The difficulty parameters (δ) are shown in . In the results of the ICC curves for Timothy and cladosporium are shown. The difficulty parameter for Timothy is 1.8, indicating an atopy predisposition score of 1.8 is necessary for a child to have a 50% probability of being SPT+ to Timothy. For cladosporium a child needs to have an atopy score of 2.48 to have a 50% probability of being SPT+ to cladosporium.
Model Parameter Estimates of Skin Prick Test Positive for Each Allergen Conditional on Predisposition for Atopy
The slope of the ICC indicates how well each allergen is able to discriminate or differentiate individuals with high versus low atopy predisposition scores. The discrimination parameters (α) are shown in . Meadow fescue was the most discriminating allergen, indicating that this allergen was highly associated with predisposition for atopy with a slope of 2.76 which translates into a log odds of 2.76 (i.e., for each unit increase in the atopy score, the odds of testing SPT+ to meadow fescue increases by 15.8). Timothy allergen and maple mix were the second and third most discriminating allergens for predisposition for atopy with log odds of 2.27 and 2.25, respectively. Short ragweed and Alternaria allergens had the lowest association between testing SPT positive and predisposition for atopy (log odds 0.69). Model fit was assessed using Pearson’s chi-square statistic and indicated that the two parameter logistic model fit the data well (p-value=0.13). In addition, the likelihood ratio test comparing the two parameter logistic model with the one parameter logistic model verified that the two parameter logistic model provided a better fit (p-value = 0.001).
Each child’s atopy predisposition score was estimated by the model using the posterior distribution of the latent trait. Atopy scores are built on allergen characteristics as determined by the IRT model parameters and the child’s SPT response pattern. The atopy score incorporates the complete panel of 17 SPT responses, and is unique for that panel. The SPT response patterns for the 18 children with the three highest and three lowest scores are shown in . As indicated there were 325 (60.5%) of children who were negative for all SPTs and their atopy predisposition score was −0.31. For two subjects with eight SPTs+ the score ranged from 2.23 to 2.45 depending on which allergens were positive as they only shared positivity on four of the eight allergens ().
SPT Response Patterns for High and Low Atopy Scores at Age Two
In order to determine if an atopy score truly reflects a child’s tendency to develop allergic symptoms at age four, unadjusted odds ratios were calculated. At age four, 194 (36%) children had at least one allergic symptom reported. All three allergic disease outcomes were significantly associated with IRT atopy scores (). To compare IRT model atopy scores with the binary SPT variable and with the categorical SPT variable, unadjusted associations between the binary and categorical SPT variables and age four allergic symptoms were also calculated. For the binary SPT variable, rhino-conjunctivitis was the only significantly associated allergic symptom. Whereas, all three allergic symptoms, persistent wheeze, rhino-conjunctivitis, and eczema were significantly associated with the IRT atopy predisposition score. For the categorical SPT variable, all three allergic symptoms were significantly associated with the SPT+ to two or more allergens compared to negative to all, but no difference was observed between those with just one positive compared to negative to all (). This finding supports the need to incorporate the number of SPT+.
Unadjusted Associations between Age Four Allergic Symptoms and Measures of Allergic Sensitization at Age Two
To further illustrate the importance of the type of allergen and how this is accounted for in the IRT model, unadjusted associations between SPT+ to short ragweed (lowest discriminating allergen) and SPT+ to meadow fescue (highest discriminating allergen) with age four allergic symptoms were computed. No significant associations were observed between being SPT+ to short ragweed at age two and allergic symptoms at age four. Whereas, SPT + to meadow fescue was very discriminating with over a three fold increase in persistent wheezing at age four (OR = 3.15, 95% CI: 1.15, 8.66). This finding supports that the IRT model is accurately identifying the types of allergens that are most associated with the underlying trait, predisposition for atopy. The IRT model incorporates both the number of SPT+ as well as the type of allergen that is positive.