Rapid report on methodology: does loss to follow-up in a cohort study bias associations between early life factors and lifestyle-related health outcomes?

Research output: Contribution to journalJournal articlepeer-review

PURPOSE: To examine the consequences of non-response in a follow-up survey for associations between early-life factors and lifestyle-related health outcomes in adulthood. METHODS: In a cohort of 11532 Danish men born in 1953 we had nearly complete follow-up in the National Patient Register, but only 66% of 9507 eligible cohort members participated in a follow-up survey, in 2004. We examined whether characteristics measured early in life and discharge from hospital for alcohol abuse or tobacco-related lung diseases, were associated with survey response. Associations between the early-life characteristics and these two health outcomes were calculated in the entire cohort and among responders, and the effect of non-response was described by a Relative Odd Ratio (ROR=OR(responders)/OR(entire cohort)). RESULTS: A low response rate at age 50 years was related to having a single mother at birth, low educational attainment at age 18, and low cognitive function at ages 12 and 18. The risk of alcohol overuse and tobacco-related diseases was also highest among non-responders. However, the associations between early-life characteristics and the outcomes were nearly the same in responders as in the entire cohort. CONCLUSIONS: Although non-responders differed from responders in terms of early-life exposures and incidence of the lifestyle-related outcomes, we found no overt effects of this on the exposure-risk associations.
Original languageEnglish
JournalAnnals of Epidemiology
Volume18
Issue number5
Pages (from-to)422-424
Number of pages2
ISSN1047-2797
DOIs
Publication statusPublished - 2008

Bibliographical note

Keywords: Cognition; Cohort Studies; Denmark; Epidemiologic Methods; Family Characteristics; Follow-Up Studies; Humans; Life Style; Male; Middle Aged; Odds Ratio; Patient Dropouts; Research Subjects; Selection Bias

ID: 5398423