LAUSR

Background Most previous prospective studies have shown an association between lack of participation in cognitive or social activities and increased risk of dementia. However, most studies have been conducted on adults aged over 75 with short follow-up, and thus their findings may be affected by reverse causation bias whereby changes in behaviour are the result of preclinical dementia. In the Million Women Study, we describe the short-term and long-term associations between a lack of participation in cognitive or social activities in women aged 60–65 years old and subsequent risk of dementia. Methods In 2001, 851,305 UK women, without prior dementia, mean age 60 (SD=5) years, reported participation in six social activities (religious group, voluntary work, adult education, art/craft group, music/singing group and bingo); five years later, 645,967 reported hours per day spent doing cognitive activities: reading and watching TV. They were followed by electronic record-linkage to national databases for validated hospital admissions with dementia. Cox regression yielded adjusted relative risks (RRs) for dementia, with follow-up split into 4 year intervals. Results For the analysis of cognitive activities, 15,529 women had newly-diagnosed dementia over a mean follow-up of 11 (SD=2) years. During the first 4 years of follow-up, there were apparent excess risks of dementia associated with not reading and not watching TV, but associations weakened substantially with longer follow-up. For women not reading vs reading for 1+ hours/day, the RRs for dementia reduced from 3.84 (95% CI 3.19–4.62) in the first 4 years to 1.27 (1.14–1.40) after 8+ years follow-up; for not watching TV vs watching TV for 1+ hours/day, the RRs reduced from 2.18 (1.55–3.06) in the first 4 years to 1.29 (1.09–1.52) after 8+ years follow-up. For the analysis of social activities, 24,437 women had newly-diagnosed dementia over a mean follow-up of 15 years. A lack of participation in any of the six social activities was weakly associated with dementia in the first 12 years of follow-up, with RRs of 1.15 (0.96–1.38), 1.10 (1.01–1.20) and 1.06 (1.01–1.12) during <4, 4-<8 and 8-<12 years respectively, but after 12+ years follow-up there was no association (0.98, 0.94–1.01). Conclusion The associations of dementia with a lack of participation in cognitive and social activities become much weaker with longer follow-up and are likely to be largely due to changes in behaviour caused by preclinical dementia.