Preventing Diabetes: Timing Your Exercise Matters More Than You Think

Over the past few decades, sports medicine has focused almost entirely on the health impacts of exercise volume and intensity, while paying insufficient attention to the timing of exercise. However, a growing body of basic and epidemiological evidence shows that human metabolic regulation exhibits distinct circadian rhythms. The same stimulus applied at different times of day can trigger drastically different physiological responses.
Previous studies have linked the temporal distribution of overall physical activity to diabetes risk and glycemic markers. However, overall physical activity largely reflects fixed daily routines (e.g., work or household chores) and has low adjustability. In contrast, moderate‑to‑vigorous physical activity (MVPA) mainly refers to purposeful exercise, which is voluntary and more flexible in scheduling, making it a more suitable basis for exercise recommendations.
Several studies have observed that the timing of MVPA/exercise is associated with glucose metabolism, with most finding greater benefits from performing such activity in the afternoon or evening than in the morning. Nevertheless, these studies generally had small sample sizes and none explored the association between MVPA timing and the incidence risk of type 2 diabetes.
The present study hypothesized that performing moderate‑to‑vigorous physical activity in the afternoon and evening confers the greatest metabolic benefits for diabetes prevention.
To test this hypothesis, the study integrated two major authoritative databases:

The UK Biobank included 84,528 participants without a baseline history of diabetes who wore accelerometers. They were followed for an average of 7.5 years, with incident type 2 diabetes recorded.
The U.S. National Health and Nutrition Examination Survey (NHANES) included 6,998 participants from the 2011–2014 cycle, used to analyze diabetes prevalence and multiple glycemic markers.

Both datasets used accelerometers to accurately record the minute‑by‑minute distribution of MVPA throughout the day, thus avoiding recall bias from questionnaires.
Waking time was divided into three periods:

Morning: 05:00–09:59
Mid‑morning to noon: 10:00–14:59
Afternoon to night: 15:00–24:00

Participants were classified into morning‑dominant, mid‑morning‑to‑noon‑dominant, afternoon‑to‑night‑dominant, and unevenly timed groups, based on which period contained more than 50% of their MVPA.
Cox proportional hazards models were used to assess incident diabetes risk in the UK Biobank. Multiple regression was applied in NHANES to analyze diabetes prevalence, 2‑hour OGTT, fasting glucose, insulin, HOMA‑IR, and other indicators. Models were adjusted stepwise for demographics, lifestyle, diet, sleep, socioeconomic status, and total exercise volume to examine the independent effect of exercise timing.
In the UK Biobank, 2,015 new cases of type 2 diabetes occurred during follow‑up.

When adjusting only for demographic variables, the unevenly timed MVPA group had the lowest diabetes risk.
After further controlling for total exercise volume, the afternoon‑to‑night MVPA group showed the lowest risk.

In the fully adjusted model:

Compared with the morning group, the mid‑morning‑to‑noon group had a 20% lower diabetes risk (HR = 0.80),
The afternoon‑to‑night group had a 29% lower risk (HR = 0.71),
The unevenly timed group had a 25% lower risk (HR = 0.75) (Figure 1).

This indicates that with the same total MVPA volume, performing exercise in the afternoon or night is associated with the lowest diabetes risk.
Further substitution models showed that shifting 20% of exercise volume from the morning to the afternoon‑night period was linked to an additional 9% reduction in diabetes risk. This result remained robust after adjusting for body weight, waist circumference, and shift work. Although the effect was slightly attenuated after adding obesity indicators, suggesting that part of the benefit may be mediated through body fat, exercise timing itself still made an independent contribution. Highly consistent results were obtained across multiple sensitivity analyses (different time grouping methods, sleep adjustment windows).
The NHANES data provided external validation for these findings. Despite its cross‑sectional design, it also showed a consistent association between afternoon‑to‑evening MVPA and lower diabetes prevalence, better 2‑hour OGTT results, and lower HOMA‑IR, fasting glucose, and insulin levels.
The consistent direction of results across two countries, with different study designs and populations, greatly strengthens the reliability of the conclusions.