select search filters
briefings
roundups & rapid reactions
Fiona fox's blog
A study published in eClinical Medicine looks at improvements in sleep, physical activity and diet being linked to a longer life.
Prof Martyn Standage, Professor of Psychology, Health & Applied Science, and Director for the Centre for Motivation and Behaviour Change at the University of Bath said:
“This is a very impressive study showing that people with very poor sleep, activity, and diet patterns could, in theory, gain meaningful extra modelled years of life by improving all three together. Yet it is important to remember that the headline idea that ‘five minutes more sleep and two minutes of brisk activity’ buys an extra year is a product of statistical modelling of data from a self‑selected volunteer cohort, not a formula that clinicians should prescribe.
“In practice, behaviour change is more than bolting on a few extra minutes of sleep or movement, or simply cutting back on portions; it is about supporting people to develop more regular, restorative sleep, find enjoyable ways to move more, and shift towards a higher‑quality overall diet – for example, more vegetables, whole grains, and fish, and fewer sugary drinks and processed meats – that they can sustain.
“The real value of this work is that it reinforces a pragmatic, multi‑behaviour approach in which small, realistic upgrades to sleep, movement, and diet, made together, are a sensible starting point in practice, even if the precise ‘minutes‑for‑years’ figures are best viewed as illustrative rather than literal.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, Open University, said:
“This study does do something new, though that’s arguably shown more in the details than the general approach. Commonly, in studies of the effect of lifestyle on risk of early death, the research looks at associations between just one aspect of lifestyle (say diet, or physical activity) and health. Fewer studies have looked at associations between several lifestyle aspects taken together, and health, though there have been some.
“The new paper claims that to be the first study that uses measurements of sleep time and physical activity that are derived from body-worn devices (accelerometers worn on the wrist), together with a diet score, to investigate the association of all three, taken together, with lifespan and healthy lifespan. They don’t usually use the term ‘healthy lifespan, but call it ‘healthspan’ – the average time from birth (in a particular sense) that people survive without serious chronic disease. So, in principle this research can tell us something that wasn’t previously known.
“Overall, then, the paper does do something new in broad terms. But it is confusingly written in places, uses inappropriate and misleading terminology in places, and arguably puts too much confidence in its statistical modelling approach.
“It (correctly) mentions that, because it is an observational study, one can’t be confident that the associations that it finds between lifestyle and health are ones of cause and effect. The researchers do make adjustments and additional data analyses that aim deal with several other possible causes and effects, but it’s not possible to be sure that everything relevant has been considered.
“The study provides some evidence, even though it’s far from decisive evidence, that the lifestyle changes in sleep time, physical activity and diet, might not need to be as large as people sometimes think to achieve worthwhile reductions in the chance of early death or increases in healthy lifespan. But in my view the way some of those conclusions are expressed is very misleading – I’ll explain below. The paper rightly says that more research will be needed – though I do ask myself how that will be done, given the limiting issues that the researchers themselves point out in using the very large UK Biobank data set. Where will better data actually come from?
“A snag is that the paper uses complicated statistical methods that are not always described clearly. It’s therefore difficult to tell the extent to which the findings have emerged from the researchers’ choice of statistical analyses, rather than things that are clearer from the data. Unfortunately, I can’t check all the methodology details because many are given only in supplementary material for the paper, which I haven’t seen. The researchers acknowledge this issue to some extent, writing that some of their findings are “theoretical extrapolations from [the particular statistical model they fitted] rather than direct observation”.*
“More importantly, it’s not clear to what extent the choice of statistical methods is driven by aspects of the data that they found when they started the data analysis, rather than having been planned in advance. There is no sign in the paper that the analysis methods were registered in advance, which would have been good practice.
“There is an important confusion in the way the results are reported. The researchers refer to “meaningful improvements” in lifespan and healthspan. This phrasing makes it look as if the improvements are definitely worth having, in the sense of making a real difference to people’s lives and health. But it turns out that they actually mean something like “improvements which are just large enough so that they can be detected statistically, with a reasonable chance of actually existing”, or for short “smallest detectable improvements”. (In the jargon, they are the smallest improvements in lifespan or in healthspan that were statistically significant. It’s a classic mistake in reporting statistics to write as if “statistically significant” and “has importance in the real world” mean the same thing. They absolutely don’t.)
“For lifespan, the smallest detectable improvement is one year added to life expectancy. One year on life expectancy is actually not much. And because of the inevitable statistical margin of error, and also the consequences of the particular statistical modelling choices that the researchers made, it could in fact be considerably less than one year. The researchers report that this could be achieved by sleeping on average for 5 minutes more a day, being moderately or vigorously physically active for an extra 2 minutes a day, and having a slightly better diet score.
“It’s true that that doesn’t sound much (though one would have to achieve all three of these, on an ongoing basis). The researchers can’t establish that this extra year of lifespan, that they found in their data and statistical modelling, is all caused by the lifestyle differences anyway (because the study is observational).
“But to get what I’d consider to be a really meaningful increase in lifespan, say 4 years, the differences would have to be a lot bigger, according to the researchers’ statistical model. That would require a whole 45 minutes of extra sleep (on average, every day), about 7 minutes extra moderate or vigorous exercise daily, and an increase in the diet score of double what’s included for the on-year lifespan increase. Again, one would have to achieve all of these (or some combination of them that’s equivalent in terms of the scoring method that the researchers developed).
“The fact that what the researchers call “meaningful improvements” are actually “smallest detectable improvements”, in a statistical sense, is the explanation of why their “meaningful improvement” in healthspan is much longer than that for lifespan, four years rather than one. There is much more statistical variability in the measure of the association between the three lifestyle factors (sleep, physical activity, diet) and healthspan than there is in the association between the lifestyle factors and lifespan. This seems to be because a statistical model for healthspan has to take into account uncertainties in the rates of diagnosing several chronic diseases rather than just death rates (though again I can’t be sure of the details because they are in supplementary material that I have not seen).
“So, the researchers can’t be statistically confident that differences in the lifestyle factors can increase healthspan at all until the improvement in healthspan is at least four years. The changes in the lifestyle factors that, according to the statistical model, might achieve this four-year increase in healthspan would be 24 extra minutes’ sleep a day on average, about 4 minutes extra physical activity, and rather a big increase (23 points on a scale that effectively goes from about 30 to about 80 across everyone in the study) in the diet score.
“The estimates for extra sleep and extra physical activity are actually smaller than the extra times estimated to be associated with a four-year increase in lifespan, though the increase in the diet score is greater. The fact that the required increases in moderate to vigorous physical activity time sound rather less demanding that the other lifestyle differences, for a four-year increase in either lifespan or healthspan, is in line with the researcher’s’ overall finding that changes in physical activity are associated with larger differences in lifespan or healthspan than are changes in sleep or diet – though all three lifestyle factors do contribute, they find.
“I’ve already mentioned that, although the researchers often write in terms of cause and effect, implying that the differences in the lifestyle factors are what cause the differences in lifespan and healthspan, the results are all based on observational data. People that have differences between these lifestyle factors will also differ in many other ways, and despite the statistical adjustments that the researchers, we can’t be completely sure that these other differences aren’t what cause the differences in lifespan or healthspan, to some extent at least. The researchers do acknowledge this, towards the end of their paper.
“Also, although the paper is generally worded in terms of possible effects on health of changing the lifestyle factors, the data from UK Biobank do not in fact look directly at how people’s health and life might have changed after they changed their lifestyle. For each person, sleep, physical activity and diet were recorded only once. So, what’s being compared is a measure of the lifespan and healthspan for different people, who had different measures of the lifestyle factors as recorded in the UK Biobank data.
“But if someone is currently scoring low on the three lifestyle factors and then they change their lifestyles in an attempt to increase their lifespan or healthspan, that might or might not improve their health to the extent of someone who was already at that higher level on the lifestyle factors. That’s one of many reasons why the researchers point out that their research isn’t the final word on these matters, and more needs to be done, including research that looks directly at what happens when individuals do change their lifestyle.
“And finally, some nitpicking about writing style. The paper is not always well written in correct English, which doesn’t help. Two egregious examples. On the fifth page, they refer to the ‘nadir’ of a curve, meaning its highest point, but ‘nadir’ actually means the lowest point, If they really wanted to use fancy language, the term should have been ‘zenith’, but ‘highest point’ is simpler and clearer. And on the eighth page, they say one number is “exponentially higher” than another, meaning simply that it is a lot larger, rather than the mathematical and scientific meaning of ‘exponential’. This use of ‘exponential’ is becoming very common in everyday use, but a scientific paper full of other mathematical and statistical technical terms is just not the place to use ‘exponential’ in this sense.
“* They don’t actually describe in the research paper the specific statistical modelling approach that they refer to in this sentence – they say they used a multivariable restricted cubic spline model, which might well be appropriate, but they don’t actually say in their paper where this model was used. (It’s probably again in supplementary material that I haven’t had access to.)”
Prof Eef Hogervorst, Professor of Biological Psychology, Director Dementia Research, National Centre for Sports and Exercise Medicine, Loughborough University said:
“This large UK-based study of 59,000 people followed people over 8 years (with an average age of 64) and confirms earlier findings of lifestyles being associated with disease (including dementia)-related longevity in other cohorts. The authors stated that even very minor changes, with 5 minutes more sleep per night, 2 minutes more of engaging in moderate to vigorous activity (both objectively assessed with accelerometers) and some small healthy dietary changes was potentially associated with one more healthy year of life. While almost half an hour of sleep extra, combined with 4 minutes of exercise and more dietary changes might provide 4 more healthy years of life.
“This is important as while women live longer on average than men, this is often spent in poor health, which has high human and economic costs. The same lifestyles investigated in the same cohort were published last year in Nature with regards to mortality, finding a 10% reduction over an 8-year follow-up of engaging in these lifestyles (15 min more sleep, 2 min of moderate-vigorous activity and a good diet combined). A 64% reduction in risk of dying was seen with 42-103 min of moderate-vigorous activity + sleeping between 7-8 hours and a good diet. That study did not find associations of diet by itself, only when investigated together with sleep and activity.
“What was new is that a) minimum amount of these healthy behaviours already was associated with better longer health and that b) objective assessment using accelerometers was used rather than self-report for sleep and physical activity.
“However, this objective assessment was only for 3-7 days which may not reflect long-term engagement in activities. Using wrist accelerometers may not necessarily reflect actual sleep and activity so well and other assessments (thigh sensors, sensors in the mattress to detect movement) may be better. However, these may be better than self-report usually used for such assessment.
“Dietary intake was self-reported and collected 3-9 years before data were collected on sleep and activity. Diets are known to change, especially in response to morbidity (e.g. cardiovascular disease) so this may be a form of systematic bias.
“Other limitations of systematic bias are that several other studies have reported this, but with clustering of healthy behaviours closely linked to people having high socioeconomic status. The cohort investigated here (UK Biobank) is healthier than the UK population (as the authors discussed) and we frequently see more healthy, well educated with good financial status participants (being more able) to be included in these studies.
“People who have more financial means are also less exposed to pollution, have less (financial) stress which can affect sleep and engaging in activities/diet and many other issues related to multi-morbidity and poor health over the lifespan.
“While statistically this can be controlled for, these differences are so systematic and closely linked, it is difficult to disentangle them. With increasing differences in socioeconomic status, the rich getting richer and many now living in abject poverty in the UK and the US, differences in experiencing healthy longer life will increase, as Prof Marmot’s group predicted many years ago. “
‘Minimum combined sleep, physical activity, and nutrition variations associated with lifeSPAN and healthSPAN improvements: a population cohort study by Nicholas A. Koemel et al. was published in eClinical Medicine at 23:30 UK time on Tuesday the 13th of January 2026.
DOI: https://doi.org/10.1016/j.eclinm.2025.103741
Declared interests:
Prof Martyn Standage :
1) Industry funding for any of your own research
Declared: The 2011-2014 research study ‘International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE)’ was indirectly funded by Coca-Cola (i.e., a team that I led at the University of Bath was subcontracted by Pennington Biomedical Center to collect data from schoolchildren aged 9 to 11 years).
2) Industry support to attend scientific meetings – Nothing to declare.
3) Current or previous advisory roles or committee membership involving industry – Nothing to declare.
4) Previous employment in companies – Nothing to declare.
5) Other industry funding/links (even if not directly linked to the particular story)
Declared: Consultancy provided to Decathlon, ESPN, Intel, NNB Genco
Prof Kevin McConway: “I don’t have any conflicts of interest on this.”
Prof Eef Hogervorst: “Eef Hogervorst was funded by Proctor to investigate diet on memory performance and has obtained funding from various Governmental and Charity funders to do research in diet and dementia risk.”