Associations of mortality with own height using son's height as an instrumental variable

• Published in: Economics and human biology 2013-07, Vol.11 (3), p.351-359
• Main Authors: Carslake, David, Fraser, Abigail, Davey Smith, George, May, Margaret, Palmer, Tom, Sterne, Jonathan, Silventoinen, Karri, Tynelius, Per, Lawlor, Debbie A., Rasmussen, Finn
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Body Height; Causality; Cause of Death - trends; Cohort studies; Confidence Intervals; Confounding factor; Confounding Factors, Epidemiologic; Female; Folkhälsovetenskap, global hälsa, socialmedicin och epidemiologi; Humans; Hälsovetenskap; Lantbruksvetenskap och veterinärmedicin; Longitudinal Studies; Male; Medicin och hälsovetenskap; Mortality; Nuclear Family; Patobiologi; Proportional Hazards Models; Sweden - epidemiology; Veterinärmedicin; Young Adult
• ISSN: 1570-677X; 1873-6130; 1873-6130
• DOI: 10.1016/j.ehb.2012.04.003

► Associations of exposures with mortality may be confounded by existing ill health. ► We used a son's height as an instrument for parents’ height to avoid confounding. ► Parents of taller sons had lower cardiovascular and respiratory disease mortality. ► Parents of taller sons had higher mortality from cancer. ► Previous studies of height are not substantially confounded by existing ill health. Height is associated with mortality from many diseases, but it remains unclear whether the association is causal or due to confounding by social factors, genetic pleiotropy,11Pleiotropy is the influence of one gene on multiple phenotypic traits. or existing ill-health. The authors investigated whether the association of height with mortality is causal by using a son's height as an instrumental variable (IV) for parents’ height among the parents of a cohort of 1,036,963 Swedish men born between 1951 and 1980 who had their height measured at military conscription, aged around 18, between 1969 and 2001. In a two-sample IV analysis adjusting for son's age at examination and secular trends in height, as well as parental age, and socioeconomic position, the hazard ratio (HR) for all-cause paternal mortality per standard deviation (SD, 6.49cm) of height was 0.96 (95% confidence interval (CI): 0.95, 0.96). The results of IV analyses of mortality from all causes, cardiovascular disease (CVD), respiratory disease, cancer, external causes and suicide were comparable to those obtained using son's height as a simple proxy for own height and to conventional analyses of own height in the present data and elsewhere, suggesting that such conventional analyses are not substantially confounded by existing ill-health.