Premium
Looking at competing events through a different lens in dementia research: Examples from the Rotterdam Study
Author(s) -
RojasSaunero Liliana Paloma,
Ikram M Arfan,
Swanson Sonja
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.045204
Subject(s) - dementia , rotterdam study , medicine , gerontology , demography , logistic regression , population , cohort , cohort study , disease , environmental health , sociology
Background Our ability to both ask clear questions and obtain clear answers in dementia research is always challenged by the competing risk of death. For example, we might ask at least two questions: (1) what is the risk of dementia, regardless of risk of death, and (2) what is the risk of dementia if we were able to prevent death during the study period? We aim to demonstrate estimating these quantities in three comparisons: higher vs. lower education, women vs. men, and apoe4 carriers vs non‐carriers. Method We used data from individuals over 55 years with no prior history of dementia or cognitive impairment within the Rotterdam Study, a population‐based cohort followed since 1990. To estimate risk of dementia we used inverse probability weighting (IPW) for lost to follow‐up (LTFU) to answer question (1), and for the risk of death, and IPW for death and LTFU for question (2), with our models including time‐updating covariates (systolic blood pressure, body max index, smoking status, development of heart disease, diabetes and cancer). We calculated the risks at each time‐point using pooled logistic regression. Result Over 20 years of follow up, 4869 individuals where followed, 657 developed dementia, 1540 died and 1029 were lost to follow up. Regardless of whether we estimated risks for question (1) or (2), individuals with higher education had a lower risk of dementia compared to lower education. In contrast, women had an increased risk of dementia compared to men only when we approached question (2) rather than (1). Apoe4 carriers had an increased risk of dementia compared to the non‐carriers, though risks were lower for question (2) rather than (1). Conclusion Estimates for the risk of dementia will always be affected by how we include death in our research question, as demonstrated in our three examples. We will discuss the implications of these research questions, including whether they represent realistic scenarios of clinical or public health interest, and the considerations when conducting such analyses in the presence of competing events.