Respiratory and nonrespiratory symptoms before age 1 year predict school-age asthma

Abstract

Background: New research traces the origins of asthma to prenatal and early-life exposures. Targeted interventions require early identification of infants who are at increased risk of asthma.

Objective: We aimed to use a composite measure of symptoms and related signs of respiratory and nonrespiratory illness collected in the first year of life to predict childhood asthma.

Methods: The Infant Immune Study enrolled pregnant women in their third trimester. Data on parental asthma, infant symptom data, and childhood asthma were collected at enrollment and over repeated visits. We applied multiple correspondence analysis to reduce the dimensionality of 11 respiratory and nonrespiratory symptoms at ages 6 and 9 months. Latent dimensions were used in a multivariable logistic regression model of childhood asthma. Bias-corrected Brier scores, area under the receiver operating characteristic curve statistics, and predictiveness curves assessed model discrimination and calibration.

Results: Of the 393 enrolled infants with complete data (82%), 17% developed asthma. Predictive model estimates showed that a latent dimension associated with the number of symptoms reported-interpreted as the intensity of illness-was predictive of subsequent asthma. Additionally, 2 other independent symptom dimensions were predictive: one characterized by wheezing and eczema and the other associated with diarrhea and vomiting. The bias-corrected Brier score was 0.136, indicating excellent predictive performance, and the area under the receiver operating characteristic curve was 0.71 (95% CI = 0.64-0.78).

Conclusion: Intensity of respiratory signs and symptoms during the first year of life may be an early clinical expression of asthma risk. This risk could also be associated with disease mechanisms distinct from those causing wheezing and related to airway and gastrointestinal dysfunction.

Keywords: Childhood asthma; birth cohort study; eczema; gastrointestinal; multiple correspondence analysis; predictive model; respiratory symptoms; wheeze.