This study is just awful. Can we stop funding epidemiological studies with little to no reproducibility? At least use more frequent objective sleep measurements, and account for confounding variables.

- They openly acknowledge they "can't say that poor sleep causes dementia" aka they established a correlation but no direction of causality

- There's self-reporting bias where participants were "asked about" their sleep patterns

- There's selection and survival bias: the participants were around 40 y/o, the ones that remained for 10 years might be different than those who dropped

- The sample was for a cardiovascular disease study

- They didn't control for unmeasured confounders other than "age, sex, education, health and lifestyle factors"

- The 1.6-2.6 years of brain aging were measured using ML algorithms and this is a small enough effect to be within a margin of error...