A difference of 2.8 percentage points separated two groups of older adults in a long-running British cohort. Ten years into the survival analysis, 89.5% of people whose eating schedules followed the earlier pattern were alive, compared with 86.7% of those in the later pattern.
The figures come from a 2025 study in Communications Medicine led by Hassan S. Dashti of Massachusetts General Hospital and Harvard Medical School. The team repeatedly recorded the times people said they ate, then related those trajectories to health and mortality. It did not assign anyone to an early or late breakfast.
This is reporting on an observational study, not medical advice. The finding does not show that setting an earlier alarm or moving breakfast forward will extend anyone’s life. The authors’ own interpretation was more cautious: a breakfast that drifts later in older age may be one visible sign that health, sleep or the practical work of preparing food has begun to change.
The 89.5% and 86.7% figures compare modeled patterns
The analysis included 2,945 community-dwelling adults from the University of Manchester Longitudinal Study of Cognition in Normal Healthy Old Age. They lived around Manchester and Newcastle. Their average age at enrollment was 64, with a range from 42 to 94, and 71.5% were women.
Participants answered a direct question: “What time do you eat breakfast, lunch and dinner?” The questionnaire was administered up to five times between 1983 and 2010, with mortality follow-up continuing to 2017. Most people provided only two meal-time assessments. Just 176 completed all five, a useful reminder that “followed for decades” does not mean every participant logged meals continuously.
At baseline, the average reported times were 8:22 a.m. for breakfast, 12:38 p.m. for lunch and 5:51 p.m. for dinner. The researchers used latent class analysis to identify two broad trajectories. The earlier group contained 1,391 people and the later group 1,554. These were patterns detected by a model, not groups separated by a universal cutoff such as breakfast before or after 8 a.m.
During an average 22 years of follow-up, England’s NHS death registry recorded 2,361 deaths. The 10-year survival percentages summarize the difference between the two modeled trajectories. The absolute gap was 2.8 percentage points. Because no meals were assigned, that gap cannot be read as the number of lives that an earlier breakfast would save.
Later breakfast traveled with changes in health
Meal timing did move as participants aged, although not dramatically. In the more fully adjusted model, each decade of age was associated with breakfast moving about 2.9 minutes later. Across the cohort’s five-decade age range, small changes could accumulate, but the average per-decade shift was modest.
The associations around breakfast were more revealing than the clock change alone. Fatigue, depression and multimorbidity were linked with later breakfast. Worse sleep and difficulty preparing meals were also associated with eating breakfast later in the adjusted analyses. Oral-health problems accompanied changes in dinner timing and a shorter daily eating window.
There are practical routes through each of those findings. Fatigue can delay getting out of bed. Depression can disturb sleep, appetite and routine. Dental pain or swallowing difficulty can make food less appealing. Reduced mobility can turn preparing a simple meal into a longer task. Medications and age-related appetite loss may contribute without appearing fully in a questionnaire.
A genetic score associated with evening chronotype also predicted later meals. Each standard-deviation increase in the eveningness score corresponded to breakfast being about seven minutes later. That finding suggests meal timing partly reflects a person’s underlying sleep-wake preference, rather than a single behavior chosen in isolation.
The mortality estimate weakened after adjustment
In the first mortality model, each additional hour of later breakfast was associated with an 11% higher mortality hazard. In the more extensively adjusted model, which included socioeconomic status, education, sleep duration, employment, smoking, marital status, alcohol use and subjective health, the estimate fell to 8%.
The fully adjusted hazard ratio was 1.08, with a 95% confidence interval from 1.00 to 1.17. The lower edge is important. Within the model’s assumptions, the data were compatible with an effect close to zero as well as a more substantial increase. The attenuation after adjustment also shows that some of the initial association was accounted for by differences between the people who ate earlier and later.
No mortality association was evident for lunch time, dinner time, the length of the eating window, or how close meals fell to sleep. Breakfast stood out, but that does not make breakfast a proven lever. It may instead be the meal most sensitive to wake time, sleep quality, appetite and the ability to begin the day’s routine.
The study could adjust only for what it measured. Physical activity was not included in the mortality models, illnesses were largely captured through self-report, and residual confounding remains possible. People in poorer health may also have been less likely to return later questionnaires, creating attrition bias.
Circadian biology offers a mechanism, not a verdict
There is a biologically plausible case for meal timing to matter. Food intake is one of the daily cues that helps synchronize metabolic tissues with the body’s circadian system. Eating during an organism’s usual inactive phase can produce different metabolic responses from eating the same food during its active phase.
Controlled human work supports some short-term effects. In a small randomized crossover experiment, later isocaloric eating increased hunger and reduced waking energy expenditure in adults with overweight or obesity. Because calories, activity and sleep were tightly controlled, the study could isolate timing more cleanly than an observational cohort.
It still cannot supply the missing causal link here. The crossover experiment involved a different population over a short period. It did not test whether shifting breakfast prevents mortality in older adults. Earlier ScienceBlog coverage of first-meal timing and cardiovascular risk described another observational cohort. Repetition across cohorts can strengthen a signal, but it does not turn an association into a randomized intervention.
The Manchester study’s authors were unusually direct about directionality. They wrote that, although the design could not establish which came first, disease leading to later meal timing was more likely than later meals causing disease onset. That judgment fits the observed links with fatigue, depression, sleep and meal preparation.
The useful signal may be change from a person’s norm
A universal breakfast deadline is not what this study found. The more grounded possibility is that a change within one person carries information. If someone who has eaten at a stable time for years begins taking breakfast progressively later, the clock may be recording an earlier change in sleep, mood, appetite, oral health, mobility or independence.
In a Mass General Brigham account published by the Harvard Gazette, Dashti described changes in meal routines as a potentially easy-to-monitor marker of overall health. A marker is not necessarily a cause. A fever can signal infection without the time on the thermometer causing the illness; a delayed breakfast could work in the same warning-sign sense.
The boundaries of the cohort matter too. Participants were healthy enough to live in the community when recruited, were predominantly women and were of British European ancestry. Meal times were self-reported, meal skipping and snacking were not measured reliably, and the study did not record the timing or nutritional content of everything participants consumed.
The 89.5% versus 86.7% comparison is worth reporting because it points to an ordinary routine that may reveal broader change. It is not a prescription to force breakfast earlier, and it does not establish that the clock itself controls longevity. The humbler conclusion is also the more useful one: when a familiar daily rhythm begins to move, the reason for the movement may matter more than the minute shown on the clock.
