Thread by @AH_AlbertHenry

If you haven't seen it already, here is our paper on the relationship between sleep duration, cognition, and dementia – published in the @IJEeditorial earlier this month: academic.oup.com/ije/advance-ar… 🧬 😴 🧠

We've known for a while that sleeping too little or too much are both linked with poor health outcomes (e.g. mortality, CVD, neurodegenerative diseases) but questions remain whether these associations are results from the underlying health problems or other confounding factors.

We tried to answer some of these questions by performing both observational and Mendelian randomization (MR) analyses using data from almost 400,000 subjects of European ancestry in the @uk_biobank and from the International Genomics of Alzheimer's Project nature.com/articles/ng.28…

Put simply, we used 78 SNPs identified from GWAS by @hsdash @drsamuelejones et al. nature.com/articles/s4146… as instrumental variable for sleep duration in our MR analysis.

For those unfamiliar with MR, my favourite lay-ish summary is Davies, et al (2018) bmj.com/content/362/bm…

Subject to its assumptions, MR has some advantages over classic epidemiology methods, as genetically-determined sleep duration are relatively free of confounders and reverse causation and reflects long-term effects of sleep duration as shown here.

Technical note: MR aficionado may notice that both our gene-exposure and gene-outcome estimates come from the same UK Biobank sample (i.e. one-sample MR). This can increase type-I error, so we tried to circumvent this by splitting our sample and meta-analyse the results (Fig. 1)

Both our observational and MR analyses suggest that a linear increase in sleep duration is associated with poorer cognitive outcomes but with relatively small effect sizes, e.g. 3% more error in visual memory test per additional hr/day of (genetically-determined) sleep duration.

As linear model might not be ideal in this case, we performed non-linear MR which indeed revealed that both short (~ <7 hr/day) and long (~ >9 hr/day) hours of sleep showed poorer visual memory outcomes compared to the average ~7 hr/day, with a suggestive J-shaped association.

To summarise, both sleeping too little and too much maybe bad for certain cognitive functions to some extent. Our study adds another layer of evidence, but is not definitive. The mechanism through which sleep can affect cognition is still a subject for future investigations.

IMO healthy adults with 'healthy' sleep habits (e.g. 7-9 hrs/day, feeling well-rested after waking up, lots of sweet dreams) should not worry too much. The message here is to watch out for a persistent unhealthy sleep, which is often reflected by short / long sleep duration.

That said, I think future studies should look into:
1. Other aspects of sleep (sleep quality, chronotype, insomnia, etc)
2. More precise measure of sleep activities (e.g. actigraphy) and cognition
3. Longitudinal outcomes & dementia (we did this but perhaps underpowered)

Last but not least, just wanted to say huge thanks to my colleagues from @UCL_IHI and @UCL_ICS: Michalis, Stefano, Ghazaleh, @SpirosDenaxas, Dionisio and my superb supervisor-mentors @vgarfield87 and Caroline, who initiated this project and have guided me from start to finish! 😊

Side note:
Here is a relevant news article published in @DailyMailUK today: dailymail.co.uk/news/article-7…

Some may not appreciate it, but I think @vicky_dmail did a great job writing a careful lay summary so if you find it misleading, please read the original paper. It's open access!

@threadreaderapp uhm... unroll please?