Can childhood asthma be predicted at birth?

Abstract

Background: Early life appears optimal for prevention of asthma, but interventions require a relevant target population, to date not clearly identified at birth.

Objective: We therefore aimed to identify the predicting capacity of factors known around birth for asthma and rhinitis at 10 years.

Methods: The included 614 healthy term babies with lung function measured at birth in the 1992/1993 Environment and Childhood Asthma study in Oslo attended a 10-year follow-up visit including a structured interview and skin prick test (SPT) for allergies. The logistic regression analyses included 37 general variables from an extensive birth questionnaire; lung function; cord blood total immunoglobulin E and soluble CD14. A history of asthma, current asthma, history of rhinitis and 'healthy' (no history of asthma, rhinitis and negative SPT) was predicted on a group level and individual predicted probabilities were calculated.

Results: The predictability of the models [area under the curve (95% confidence intervals)] was 0.74 (0.69, 0.79), 0.72 (0.64, 0.78), 0.69 (0.54, 0.72) and 0.67 (0.62, 0.71) for a history of asthma, current asthma, rhinitis and 'healthy', respectively. The best model predicted a history of asthma correctly in 93/124 (75%), and incorrectly in 176/490 (36%) children without asthma. The positive predictive values for all outcomes were low (19-61), the highest predicting healthy.

Conclusion: Although at best 75% of children with a history of asthma could be predicted at birth, an intervention applied to our predicted high-risk children would be started more often in children without than with future disease. Parental allergic disease alone appears insufficient to identify high-risk populations in future studies of asthma and allergic disease.