Multi-level modelling of international variations and time trends in asthma and allergic diseases in children

Introduction: Asthma is the most common chronic disease in children, but little is understood about its underlying causes and reasons for global differences and time trends in prevalence. This thesis uses data from the International Study of Asthma and Allergies in Childhood (ISAAC) and Global Asthma Network (GAN) to explore these issues. These are multi-centre, multi-country, standardised cross-sectional symptom and risk factor surveys, at three time points over a 27-year period, in adolescents aged 13-14 and children aged 6-7. Methods: The ISAAC and GAN data form a complex hierarchy including individuals within schools, within centres, within countries, with centre-level data available at multiple time points. Mixed-effects logistic regression models were used to estimate associations between individual-level risk factors and asthma symptoms, and also between school-level risk factors and individual-level asthma symptoms, at one time-point. Mixed-effects linear regression models were used to estimate ecological associations at the centre level, between the same risk factors and both current symptom prevalence and time trends in prevalence. Findings: Prevalence of asthma symptoms is generally highest in high income countries and has remained stable, while it has been increasing in lower-middle income countries, and decreasing or remaining stable in low-income countries. Risk factors including paracetamol use, frequent truck traffic, and antibiotics in the first year of life have strong associations with asthma symptoms at the individual level, and generally also at the school level. However, these factors do not explain geographical differences in prevalence or global time trends. Conclusion: Risk factors with strong evidence of an association with asthma symptoms at the individual level do not explain global patterns and time trends. This could be due to either ecological bias, unmeasured confounding or because the determinants of asthma at a population level are actually different to the determinants at an individual level.