Age‐Specific Associations of Moderate‐to‐Vigorous Physical Activity With Relative and Absolute Risk of Type 2 Diabetes in UK Biobank Participants Authors: Qiuyu Feng, Raymond Noordam, Denghui Hu, Gali Albalak, Ko Willems van Dijk, Diana van Heemst Published: 2026 ABSTRACT ABSTRACTAimsWe assessed the age‐specific associations between moderate‐to‐vigorous physical activity (MVPA) and incident type 2 diabetes.Materials and MethodsWe included UK Biobank participants with valid accelerometry data without diabetes mellitus at baseline (N = 87 975; 62.7 years [IQR, 55.6–67.9]; 57% women). We examined whether age, sex and baseline metabolic disease status modified the associations of MVPA volume (lower versus higher) and, in exploratory analyses, MVPA timing (later‐day versus earlier‐day) with incident type 2 diabetes. Age‐specific relative and absolute associations were assessed across 5‐year age bins using hazard ratios (HRs) and incidence rate differences (IRDs), and additionally stratified by sex and baseline metabolic disease status.ResultsAssociations of MVPA volume with type 2 diabetes were most consistently modified by age, baseline prediabetes (all p‐values for interaction < 0.05), with stronger associations observed among participants younger than the median population age (< 62.7 years, HR = 2.44 [95% CI: 2.05–2.90] versus HR = 1.65 [95% CI: 1.47–1.86]) and participants without prediabetes. Across 5‐year age groups, HRs declined with age, whereas IRDs increased in the overall population. No clear effect modification was observed for the association between MVPA timing and type 2 diabetes risk.ConclusionsThe relative association between higher MVPA volume and lower type 2 diabetes risk attenuated with age and was weaker among individuals with prediabetes, whereas absolute risk reductions were larger among older and metabolically higher‐risk individuals. These findings support considering both relative and absolute measures when targeting MVPA interventions for type 2 diabetes prevention.Keywords: cohort analysis, exercise, physical activity epidemiology, triaxial accelerometer, type 2 diabetes Abbreviation moderate‐to‐vigorous physical activity Introduction Regular engagement in moderate‐to‐vigorous physical activity (MVPA) is a cornerstone of type 2 diabetes (T2D) prevention [1]. Higher total MVPA volume improves glucose homeostasis through enhanced insulin sensitivity [2], reduced adiposity [3] and favourable effects on lipid metabolism [4]. Consequently, current public health guidelines [5, 6] recommend ≥ 150 min/week of MVPA to reduce T2D risk. From a life‐course perspective, potential benefits of physical activity for T2D prevention and management may differ by age, reflecting age‐related changes in metabolism and physiological responses to physical activity. Several studies reported attenuated associations of physical activity with T2D risk [7, 8, 9, 10] and glucose control [11] at older ages. However, most used small samples or broad age categories, limiting their ability to characterize the trajectories of age‐specific associations. In addition, the association between physical activity and T2D may follow sex‐specific age patterns, potentially due to sex differences in hormonal changes with age. Age related changes in sex hormone levels occur abruptly during menopause in women [12] but more gradually in men [13], which may influence the physiological response to physical exercise [14] and T2D risk [15]. Furthermore, previous cohort studies [7, 16] observed that metabolic status (e.g., metabolic syndrome, obesity) might modify the effect of physical activity on incident T2D risk. The age‐related trajectory of this association remains unclear, and it is unknown whether it is modified by sex and metabolic status. Most studies [7, 17, 18, 19] report relative association measures, such as hazard ratios, which are useful for etiologic inference but provide limited information on absolute disease burden. Absolute risk measures, which are more informative for estimating population‐level prevention potential [20], are rarely reported [21]. Relative estimates do not capture variation of disease incidence and therefore provide limited insights into the potential preventive impact on different populations. A weaker relative association in older or higher‐risk groups does not necessarily imply a smaller preventive benefit in these populations, as higher incidence rate reductions may translate into larger absolute risk differences. Evaluating absolute risk differences alongside relative measures is essential for interpreting the public health impact of physical activity in different populations. Growing evidence [22, 23], including from our prior work [23, 24], indicates that MVPA performed later during the day, independent of total MVPA volume, was associated with lower incident T2D risk and better glucose control than MVPA performed earlier during the day. However, it remains unclear whether associations of MVPA timing with T2D risk, in terms of both relative and absolute risk, differ across strata of age or metabolic status. Therefore, the objectives of this study were to examine whether age, sex, and baseline metabolic status—including adiposity, prediabetes, dyslipidemia and hypertension—modify the associations between MVPA volume and, in exploratory analyses, MVPA timing with incident T2D. We further evaluated age‐specific relative and absolute associations within subgroups stratified by sex and baseline metabolic status. Materials and Methods Study Design and Population The UK Biobank study was approved by the North‐West Multi‐center Research Ethics Committee and included over 500 000 participants aged 40–73 years at recruitment (2006–2010). All UKB participants provided written informed consent. Between February 2013 and December 2015, a UKB subsample (n = 103 661) participated in an accelerometer sub‐study for objective assessment of physical activity over a maximum of 7 days [25]. Quality control exclusion criteria for accelerometry data were similar to those used previously [23, 26], and we additionally excluded participants who recorded zero MVPA during the entire accelerometer wear period (see exclusion criteria in Figure S1). In total, 12 747 participants were excluded. Additionally, we excluded 2939 participants who were diagnosed with diabetes before the accelerometer wear period date. Finally, 87 975 participants with valid data were included in the current study (Figure S1). Exposure In UKB, objective physical activity was assessed at 100 Hz over up to 7 days using the Axivity AX3 wrist‐worn triaxial accelerometer (Axivity Ltd., Newcastle upon Tyne, UK) with a dynamic range of ±8g [25]. Participants were instructed to wear the accelerometer continuously and to carry out their normal daily activities. MVPA was identified using a validated machine‐learning accelerometer‐based two‐level random forest classifier that categorized 30‐s epochs into four activity types (sedentary, standing utilitarian movements, walking and running/high‐energy activities), which were then mapped to corresponding intensity levels (sedentary, light, moderate (100–400 milligravity), vigorous (> 400 milligravity)) [27]. MVPA volume was categorized as lower or higher using the age‐specific median of weekly MVPA minutes, to avoid disproportionately classifying older participants as having lower MVPA. To compare our results with clinical guidelines, we additionally conducted a sensitivity analysis using the WHO‐recommended threshold of ≥ 150 min/week of MVPA [5]. MVPA timing was defined using a binary classification to facilitate interpretation of the interaction analyses: earlier‐day MVPA (05:00–12:59) and later‐day MVPA (13:00–23:59). Participants were categorized into MVPA timing groups according to the time window in which they performed at least 50% of their total MVPA. The 13:00 cut‐off was observed previously by our hourly analyses on T2D risk [23], in which associations appeared to differ before and after early afternoon, and provided more balanced group sizes than cut‐offs shifted 1–2 h earlier or later. Sensitivity analyses tested alternative cut‐offs shifted 1–2 h earlier or later using the same modelling strategy. The MVPA volume groups and MVPA timing groups were used as exposure in all statistical analyses. Potential Effect Modifiers The present study considered demographic (age and sex) and cardiometabolic risk factors (dyslipidemia, hypertension, higher adiposity and prediabetes) as potential confounders or effect modifiers, which were collected and measured at study enrolment. Information on sex and age (at the time of accelerometer wear) was collected based on a self‐reported online questionnaire. Dyslipidemia was defined as abnormal blood lipid levels or current use of lipid‐lowering drugs. Prediabetes was defined using baseline HbA1c only. Participants with HbA1c levels between 39 and 47 mmol/mol (5.7%–6.4%), in the absence of diagnosed diabetes, were classified as having prediabetes, according to clinical guidelines [28] and previous UKB research [29]. Hypertension was defined at baseline as having elevated blood pressure (systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg) and/or by the use of antihypertensive medication. Higher adiposity was defined as BMI ≥ 30 kg/m2, higher waist circumference (≥ 102 cm in men or ≥ 88 cm in women), or higher body fat percentage (> 25% in men or > 35% in women) at baseline. To assess and visualize effect modifications by above risk factors in subsequent stratified analyses, participants were categorized according to each risk factor: sex groups (women, men), age bins (younger, older based on the median age of the study population, binary: < 62.7, > = 62.7; 5 years bin: < 50, 50–55, 55–60, 60–65, 65–70, > 70 years), metabolic disease status (prediabetes, dyslipidemia, hypertension and higher adiposity; each individually coded as yes/no). Covariates The following variables were included as potential confounders: ethnicity (European/others), education (College or University degree or above/any other qualification/no qualification), Townsend Deprivation index (continuous), smoking status (never/previous/current), family history of diabetes (yes/no) and sleep duration (continuous), were measured and collected at baseline (most variables were derived from self‐reported questionnaires, except sleep duration and light physical activity, which were derived from accelerometer data collected concurrently with the exposure). Missingness for covariates and potential effect modifiers was summarized in the Table S1. Outcome Incident T2D was defined using the UK Biobank first‐occurrence field for non‐insulin‐dependent diabetes mellitus (Field 130 708) only, corresponding to the first recorded occurrence of ICD‐10 code E11. This field is derived from linked primary care, hospital inpatient, death register and self‐reported data, with the earliest recorded date used as event date. Events were included if they occurred after accelerometer assessment and before death, loss‐to‐follow‐up or the end of the study period (November 6, 2023), whichever came first. Statistical Analyses Statistical analyses were performed using R version 4.4.3 (R Foundation for Statistical Computing, Vienna, Austria). Baseline characteristics of the included study population were presented as median (interquartile range, IQR) for continuous variables and frequency (proportion) for categorical variables. MVPA volume and timing were categorized into lower and higher groups and earlier‐day, later‐day groups, respectively, using the higher MVPA group and early‐day MVPA as reference categories. To examine whether the association between MVPA volume, timing and T2D differed across population subgroups, we evaluated potential effect modification by established T2D risk factors, including age, sex, dyslipidemia, hypertension, higher adiposity and prediabetes, adjusted for ethnicity, education, Townsend Deprivation index, smoking status, family history of diabetes and sleep duration. For models evaluating MVPA timing, total MVPA volume was additionally adjusted for. BMI may be both a marker of the healthy‐exerciser effect and a potential mediator, so BMI was not included in the primary models but was added in sensitivity analyses to assess adiposity‐related residual confounding. Effect modification was assessed by including interaction terms between exposure and each potential modifier in Cox proportional hazards models. For each potential modifier, participants with missing data on that modifier were excluded from the corresponding effect modification analysis; therefore, the analytical sample size varied across modifier‐specific analyses. To assess the impact of missing HbA1c, we multiply imputed HbA1c and derived prediabetes status using the primary definition. To reduce potential reverse causation due to preclinical metabolic dysfunction affecting physical activity, we additionally excluded participants who developed T2D within the first 2 years after baseline. To visualize dose–response associations across key modifier strata, we modelled weekly MVPA volume and later‐day MVPA proportion as a continuous exposure using restricted cubic spline Cox models with the same covariates. To further characterize age‐specific associations, analyses were stratified using 5‐year age categories. Within each age stratum, hazard ratios (HRs) and 95% confidence intervals (CIs) for incident T2D comparing lower versus higher MVPA volume and later‐day versus early‐day MVPA were estimated using multivariable‐adjusted Cox proportional hazards models. To complement relative risk estimates, age‐specific incidence rates of T2D were calculated within each MVPA category and expressed per 1000 person‐years. Absolute incidence differences were estimated as the difference in incidence rates between MVPA groups (lower versus higher and later‐day versus early‐day) within each age stratum. Given known sex differences in T2D incidence rates, all age‐stratified analyses were repeated separately for men and women. For effect modifiers identified in the analyses above, age‐stratified analyses were stratified by the presence or absence of each modifier. Age‐specific hazard ratios, incidence rates and absolute incidence differences between MVPA groups were estimated to assess potential heterogeneity by sex and potential effect modifiers. Results based on the multivariable‐adjusted Cox proportional hazards models were expressed as hazard ratios (HR) and 95% confidence intervals (95% CI). Results Participant Characteristics A total of 87 975 UKB participants were included (57% female; median [IQR] age 63 [56, 68] years; Table 1). Overall, the two different MVPA volume groups (median based) presented similar sociodemographic and lifestyle characteristics (Table 1). However, lower MVPA group participants were more likely to be women and had a substantially higher prevalence of higher adiposity. Higher MVPA group participants accumulated 409 (312, 566) minutes/week of MVPA, and all met the recommended threshold [5] of 150 min/week of moderate physical activity. In contrast, lower MVPA group participants accumulated 121 (66, 178) minutes/week of MVPA, with only 37% meeting the recommended accumulated moderate physical activity per week. TABLE 1Baseline characteristics in UK Biobank participants having accelerometry data and without pre‐existing diabetes.CharacteristicOverall N = 87 975Higher MVPA N = 44 020Lower MVPA N = 43 955Age, years62.7 (55.6, 67.9)62.0 (54.9, 67.3)63.6 (56.3, 68.6)Female49 924 (57%)21 056 (48%)28 868 (66%)TDI−2.46 (−3.83, −0.22)−2.43 (−3.81, −0.12)−2.50 (−3.84, −0.31)European ethnicity85 001 (97%)42 725 (97%)42 276 (96%)Diabetes family history (yes)13 962 (16%)6551 (15%)7411 (17%)Sleep duration, hours7.33 (6.76, 7.86)7.31 (6.74, 7.83)7.36 (6.78, 7.89)MVPA min/week239 (121, 409)409 (312, 566)121 (66, 178)Smoking statusNever50 696 (58%)25 619 (58%)25 077 (57%)Previous31 180 (36%)15 816 (36%)15 364 (35%)Current5872 (6.7%)2467 (5.6%)3405 (7.8%)EducationDegree or above38 361 (44%)21 633 (50%)16 728 (38%)Any other qualification41 764 (48%)19 065 (44%)22 699 (52%)No qualification6994 (8.0%)2940 (6.7%)4054 (9.3%)Prediabetes (yes)9228 (11.3%)4148 (10%)5080 (12%)Hypertension (yes)40 412 (45.9%)19 373 (44.0%)21 039 (47.8%)Dyslipidemia (yes)44 405 (50%)20 199 (45.9%)24 206 (55.1%)Higher adiposity (yes)44 347 (51%)18 296 (42%)26 051 (60%) Note: Data are presented as median (with interquartile range—IQR) or n (%). There is 6334 and 880 missing values for prediabetes and higher adiposity, respectively. For disease variables, denominator for percentage is non‐missing observations. Higher adiposity was defined as BMI ≥ 30 kg/m2, elevated waist circumference (≥ 102 cm in men or ≥ 88 cm in women), or higher body fat percentage (> 25% in men or > 35% in women). The prevalence of obesity defined by BMI ≥ 30 kg/m2 alone was 17.6%. Baseline characteristics in UK Biobank participants having accelerometry data and without pre‐existing diabetes. Note: Data are presented as median (with interquartile range—IQR) or n (%). There is 6334 and 880 missing values for prediabetes and higher adiposity, respectively. For disease variables, denominator for percentage is non‐missing observations. Higher adiposity was defined as BMI ≥ 30 kg/m2, elevated waist circumference (≥ 102 cm in men or ≥ 88 cm in women), or higher body fat percentage (> 25% in men or > 35% in women). The prevalence of obesity defined by BMI ≥ 30 kg/m2 alone was 17.6%. Associations of MVPA With T2D Risk During a median follow‐up of 7.7 years (IQR: 7.1–8.2), 2140 participants (2.4%) developed T2D. After adjustment for considered confounders, the lower MVPA group had a higher risk of developing T2D than the higher MVPA group (HR = 2.07, 95% CI: 1.89–2.27). Compared with the early‐day MVPA group, the later‐day MVPA group had a lower risk of T2D (HR = 0.90; 95% CI: 0.82–0.98). Restricted cubic spline analyses showed same results (Figures S2A and S3A). Effect Modification by Age and Sex As shown in Table 2, of the factors investigated, age modified the association between MVPA volume and timing and T2D risk (Volume: p‐value for interaction < 0.001; Timing: p‐value for interaction = 0.02). Specifically, for MVPA volume, the association was strongest in the younger group (younger: HR = 2.44 [95% CI: 2.05–2.90]; older: HR = 1.65 [95% CI: 1.47–1.86]). For MVPA timing, the beneficial effect of later‐day MVPA was strongest in the younger group, while being negligible in the older group (younger: HR = 0.77 [95% CI: 0.66–0.91]; older: HR = 0.96 [95% CI: 0.86–1.08]). No evidence for interactions with sex was observed (Table S3). A similar pattern was observed when we considered MVPA as a continuous exposure (Figure 1A). TABLE 2Association between MVPA volume and type 2 diabetes risk by age, sex and metabolic disease.Higher volumeLower volumeAbsolute differenceHR (95% CI) lower vs. higher P for interactionN with/without T2DIncidence ratesN with/without T2DIncidence ratesAgeYounger220/21 7751.30534/21 4583.181.89 (1.57, 2.21)2.44 (2.05, 2.90)< 0.001Older528/21 4973.21858/21 1055.302.09 (1.65, 2.54)1.65 (1.47, 1.86)SexFemale207/20 8491.29613/28 2552.801.52 (1.24, 1.80)1.79 (1.51, 2.12)0.638Male541/22 4233.13779/14 3087.033.90 (3.34, 4.46)1.91 (1.70, 2.15)PrediabetesNo234/36 5730.83474/35 1321.750.92 (0.73, 1.11)2.11 (1.78, 2.50)0.001Yes359/378911.97642/443817.986.01 (4.14, 7.87)1.54 (1.34, 1.78)Higher adiposityNo207/25 0731.07205/17 2631.540.47 (0.22, 0.73)1.57 (1.27, 1.93)0.748Yes535/17 7613.901176/24 8756.092.19 (1.71, 2.67)1.66 (1.48, 1.85)DyslipidemiaNo279/23 4311.54460/19 4003.071.53 (1.20, 1.86)2.18 (1.86, 2.57)0.017Yes469/19 8413.06932/23 1635.182.12 (1.69, 2.56)1.63 (1.44, 1.83)HypertensionNo267/23 9631.44505/22 8282.861.42 (1.11, 1.72)1.72 (1.46, 2.02)0.293Yes481/19 3093.23887/19 7355.802.57 (2.09, 3.05)1.60 (1.42, 1.80) Note: Models were adjusted for age, sex, ethnicity, education, Townsend deprivation index, smoking status, sleep duration, family history of diabetes and total daily light physical activity. Incidence rates and absolute differences are expressed per 1000 person‐years. P for interaction was calculated on the multiplicative scale (hazard ratio). There is 6334 and 880 missing values for prediabetes and higher adiposity, respectively.Abbreviations: HR, hazard ratio; MVPA, moderate to vigorous physical activity. Association between MVPA volume and type 2 diabetes risk by age, sex and metabolic disease. Note: Models were adjusted for age, sex, ethnicity, education, Townsend deprivation index, smoking status, sleep duration, family history of diabetes and total daily light physical activity. Incidence rates and absolute differences are expressed per 1000 person‐years. P for interaction was calculated on the multiplicative scale (hazard ratio). There is 6334 and 880 missing values for prediabetes and higher adiposity, respectively. Abbreviations: HR, hazard ratio; MVPA, moderate to vigorous physical activity. Dose–response association between weekly MVPA volume and incident type 2 diabetes by age group and prediabetes status. Models were adjusted for age, sex, ethnicity, education, Townsend Deprivation Index, smoking status, sleep duration, family history of diabetes and total daily light physical activity. (A) shows associations stratified by age group; (B) shows associations stratified by prediabetes status. Solid lines indicate hazard ratios and shaded areas indicate 95% confidence intervals. Hazard ratios were estimated relative to 0 min/week of MVPA. The x‐axis was truncated at 800 min/week, approximately the 95th percentile of weekly MVPA, to reduce the influence of sparse data at extreme MVPA values. MVPA moderate‐to‐vigorous physical activity. When stratified by 5‐year age groups for visualization, the relative risk of T2D comparing lower versus higher MVPA volume decreased with age (Figure 2A). In contrast, T2D incidence rates increased substantially with age in both MVPA volume groups (Figure 2B), resulting in larger absolute incidence rate difference between groups at older ages (Figure 2C). Age‐specific pattern of relative and absolute type 2 diabetes risk associated with MVPA volume in the overall population and by sex. MVPA volume groups (lower and higher) were defined using the age‐specific median of MVPA. (A), (D) and (G) showed the age‐specific relative risk estimates among whole population, women, men, respectively; (B), (E) and (H) showed the age‐specific incidence rates among whole population, women, men, respectively; (C), (F) and (I) showed the age‐specific absolute incidence rates difference among whole population, women, men, respectively. Y‐axis scales vary across panels in Figures 1 and 2 to improve readability; visual comparisons of incidence‐rate slopes between panels should therefore be made with caution. MVPA moderate‐to‐vigorous physical activity. For sex, the decrease in relative risk estimates comparing lower versus higher MVPA volume with increasing age was more pronounced in women (Figure 2D) than in men (Figure 2G). Among women, T2D incidence rates increased slowly and in parallel with age in both MVPA volume groups (Figure 2E), resulting in relatively stable absolute incidence rate differences across age strata (Figure 2F). Among men, T2D incidence rates increased more steeply with increasing age, particularly in the lower MVPA group (Figure 2H), leading to progressively larger absolute incidence rate differences at older ages (Figure 2I). Age‐stratified spline analyses showed broadly similar dose–response patterns, with substantial overlap in confidence intervals between younger and older participants (Figure S3B). The hazard ratios and absolute incidence rate difference for later‐day versus earlier‐day MVPA did not show a consistent age‐related trend (Figure S4). Effect Modification by Baseline Metabolic Disease At baseline, 9228 participants (11.3%) had prediabetes, 44 405 (50%) had dyslipidemia, 44 347 (51%) had higher adiposity and 40 412 (45.9%) had hypertension. As shown in Table 2, prediabetes and dyslipidemia modified the association between MVPA volume and T2D risk (Prediabetes: P for interaction = 0.001; Dyslipidemia: P for interaction = 0.017). Specifically, for prediabetes, the association was stronger in people without prediabetes (without prediabetes: HR = 2.11 [95% CI: 1.78–2.50]; with prediabetes: HR = 1.54 [95% CI: 1.34–1.78]). For dyslipidemia, the association was stronger in people without dyslipidemia (without dyslipidemia: HR = 2.18 [95% CI: 1.86–2.57]; with dyslipidemia: HR = 1.63 [95% CI: 1.44–1.83]). A similar pattern was observed when we considered MVPA as a continuous exposure (Figure 1B). In contrast, dose–response patterns were broadly similar among participants with and without dyslipidemia (Figure S2D). As shown in Table S2, we observed suggestive evidence for effect modification by prediabetes for the association between MVPA timing and T2D risk (p‐value for interaction is 0.054). Specifically, the beneficial effect of MVPA later during the day was stronger in people without prediabetes, while negligible in people with prediabetes. No evidence for effect modifications was observed for higher adiposity and hypertension. At baseline, among participants with prediabetes compared to participants without prediabetes, we observed stronger attenuation of relative risk estimates of lower versus higher MVPA volume with increasing age (Figure 3A,D). Among participants without prediabetes, T2D incidence rates in both low and high MVPA volume groups were low and relatively stable across age strata, resulting in small and stable absolute incidence rate differences (Figure 3B,C). Among participants with prediabetes, incidence rates in both low and high MVPA volume groups were substantially higher and increased with increasing age, most rapidly between ages 50–55 years, leading to the largest absolute incidence rate differences at ages 50–55 years (Figure 3E,F). Similar patterns were observed for dyslipidemia (Figure S5), although the absolute incidence rate differences were smaller than those observed for prediabetes. Age‐specific pattern of relative and absolute type 2 diabetes risk associated with MVPA volume by baseline prediabetes status. (A) and (D) showed the age‐specific relative risk estimates among participants with and without baseline prediabetes status, respectively; (B) and (E) showed the age‐specific incidence rates among participants with and without baseline prediabetes status, respectively; (C) and (F) showed the age‐specific absolute incidence rate differences among participants with and without baseline prediabetes status, respectively. Y‐axis scales vary across panels in Figures 1 and 2 to improve readability; visual comparisons of incidence‐rate slopes between panels should therefore be made with caution. MVPA moderate‐to‐vigorous physical activity. Sensitivity Analysis Effect modification analyses using the WHO‐recommended threshold of ≥ 150 min/week provided broadly similar results (Table S3), with evidence of interaction with hypertension (P for interaction = 0.014). Analyses using alternative timing cut‐offs showed generally similar inverse associations between later‐day MVPA and T2D risk. However, no evidence of effect modification by age was observed using these alternative timing‐window definitions (Table S5). The results, with the exception of dyslipidemia, remained similar when T2D cases within the first 2 years of follow‐up were excluded (Table S6). Multiple‐imputation sensitivity analysis for missing HbA1c produced results nearly identical to the complete‐case analysis. Additional adjustment for BMI attenuated the association between MVPA volume and T2D risk, but the association remained evident (Table S7). Discussion In this prospective cohort study, we observed that age and prediabetes modified associations of MVPA volume and T2D risk, with lower relative risk estimates but higher absolute incidence rate differences at older ages and among individuals with prediabetes. Age‐specific patterns of relative and absolute risk differed by sex and prediabetes status. No clear and consistent evidence for effect modification was observed for the association between MVPA timing and T2D risk. Higher MVPA volume was associated with stronger relative risk reductions in T2D among younger participants and those without prediabetes. As results on dyslipidemia were not consistent throughout sensitivity analyses, the main observations should be interpreted cautiously. Similar age‐related attenuations have been reported in previous longitudinal studies using different age group definitions [7, 9, 10]. Although evidence regarding metabolic health as an effect modifier remains limited, prior studies reported stronger inverse associations among individuals without metabolic syndrome [7] or obesity [30], consistent with our observations among individuals without metabolic dysfunction. Importantly, higher MVPA volume remained associated with lower T2D risk among individuals with prediabetes or dyslipidemia. Given that metabolic dysfunction may be partly reversible through lifestyle modification, including increased physical activity and weight loss [31], these observations support promoting MVPA before substantial metabolic dysfunction develops, while reinforcing its importance across all ages and metabolic risk groups. Speculatively, reduced metabolic responsiveness to physical activity with increasing age and metabolic dysfunction may partly explain the observed interaction with age on T2D. Ageing is associated with declines in insulin sensitivity [32], β‐cell function [33] and metabolic flexibility [34], while prediabetes involves impairment in mitochondrial substrate utilization, insulin resistance and lipid‐induced cellular toxicity [35], all of which could plausibly attenuate the metabolic responsiveness to physical activity [36]. However, attenuated relative associations should not be interpreted as limited preventive relevance of MVPA in older adults or individuals with metabolic dysfunction, given their higher baseline risk. We observed greater absolute T2D risk reductions associated with higher MVPA among older adults and participants with prediabetes, despite weaker relative associations in these groups. Age‐specific patterns across 5‐year age groups differed by sex and prediabetes status. Among women, despite stronger attenuation of relative risk with age, absolute risk reduction remained relatively low and stable. In contrast, among men, relative associations attenuated modestly, but T2D incidence increased steeply with age, particularly in the lower MVPA group, resulting in larger absolute risk reductions at older ages. This steeper age‐related increase in incidence among men is consistent with previous epidemiological evidence [37]. These observations suggest stronger relative associations with higher MVPA at younger ages, particularly among women, but larger absolute risk reductions at older ages, especially among men. Thus, MVPA promotion among older adults, as a population, may yield greater absolute reductions in T2D incidence, particularly when intervention resources are limited. Among participants with prediabetes, absolute risk reductions were substantially larger at younger ages and peaked around ages 50–55, although this pattern should be interpreted cautiously because metabolic status was assessed several years before accelerometry and may have changed over time. Although later‐day MVPA was associated with lower T2D risk in the overall population, we found no consistent evidence that this association differed by age, sex, or metabolic status. This may reflect greater sensitivity of timing analyses to time‐window definitions, population heterogeneity and residual confounding. Thus, our findings do not support subgroup‐specific MVPA timing recommendations. Given that the association with T2D risk was more modest for MVPA timing than for MVPA volume, our results primarily support prioritizing higher overall MVPA volume for T2D prevention. This study has several notable strengths. First, the large sample size allowed investigation of age‐specific associations of MVPA with T2D risk across 5‐year age bins, as well as stratified analyses by sex and baseline metabolic disease. Second, examining both relative and absolute measures of T2D risk provided comprehensive insights on potential preventive impact of MVPA. Several limitations should also be acknowledged. First, prediabetes was defined at baseline (3–9 years before accelerometry), so this subgroup likely includes individuals who remained prediabetic as well as those who reverted to normoglycemia (participants with diagnosed diabetes before accelerometry measurement were excluded). This misclassification would likely result in an underestimation of the observed risk reductions. Second, glucose‐lowering medication use was not included in the prediabetes definition, but only 13 of 72 413 participants without prediabetes reported such use, suggesting minimal misclassification; if present, this would likely attenuate differences by prediabetes status. Third, although widely used in UK Biobank studies, a single 7‐day accelerometer assessment may not fully capture habitual MVPA and could introduce some level of regression dilution bias [38], likely attenuating associations toward the null, particularly for timing‐specific analyses. Fourth, complete‐case interaction analyses may have introduced selection bias, particularly for prediabetes due to missing HbA1c data. Finally, findings from UKB may not generalize to populations with different demographic or lifestyle profiles. In conclusion, relative risk reduction associated with higher MVPA volume decreased with age and less favourable metabolic health status, whereas absolute risk reductions were greater at older ages and among individuals with poorer metabolic profiles. These findings highlight the importance of considering both relative and absolute measures when evaluating preventive strategies. MVPA interventions should support physical activity across population groups, while giving particular attention to older and metabolically higher‐risk populations who may derive the greatest absolute population benefit. Author Contributions Q.F. contributed to the conception and design of the study, analysis and interpretation of the results, acquisition of funding and drafting of the manuscript. K.W.v.D. contributed to the conception and design of the study, and critical revision of the manuscript. R.N. contributed to the conception and design of the study and interpretation of the results, critical revision of the manuscript and served as guarantor of the work, with full access to all data in the study and responsibility for the integrity of the data and the accuracy of the data analysis. D.v.H. contributed to the conception and design of the study and interpretation of the results, acquisition of funding and critical revision of the manuscript. G.A. and D.H. contributed to critical revision of the manuscript. All authors read and approved the final version of the manuscript. Funding This work was supported by a scholarship from the China Scholarship Counsil (202306240049); a grant from the Dutch Research Council (NWO, Dutch National Research Agenda, Research along routes by consortia, 2021–2026, BioClock: the circadian clock in modern society). Ethics Statement The UK Biobank study was approved by the North‐West Multi‐center Research Ethics Committee. Consent Informed consent from the participants was obtained before inclusion of the study. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1: Flow chart of study sample selection in UK biobank. Detail exclusion criteria for accelerometry data: Dataset: (1) unexpectedly small or large size of the uncompressed returned by the accelerometer device (Field ID: 90002); (2) data from less than 72 h or lack of data for all 1‐h periods within a 24‐h cycle during the 7‐day data collection (Field ID: 90015); (3) data was not well calibrated (Field ID: 90016); (4) participant data was recalibrated using the accelerometer record from the same device worn by a different participant (Field ID: 90017); (5) data with non‐zero count of interrupted recording periods (Field ID: 90180); (6) data with more than 768 (Q3 + 1.5 × IQR) data recording errors (Field ID: 90182); (7) The overall average proportion of time spent doing moderate‐vigorous activity across the monitoring period is 0 (Field ID: 40049). Figure S2: Dose–response association between weekly MVPA volume and incident type 2 diabetes in whole population and by age group, prediabetes status and dyslipidemia. Models were adjusted for age, sex, ethnicity, education, Townsend Deprivation Index, smoking status, sleep duration, family history of diabetes and total daily light physical activity. (A) shows associations in overall population; (B) shows associations stratified by age; (C) shows associations stratified by prediabetes status; (D) shows associations stratified by dyslipidemia status. Solid lines indicate hazard ratios and shaded areas indicate 95% confidence intervals. Hazard ratios were estimated relative to 0 min/week of MVPA. The x‐axis was truncated at 800 min/week, approximately the 95th percentile of weekly MVPA, to reduce the influence of sparse data at extreme MVPA values. Abbreviations: MVPA moderate to vigorous physical activity. Figure S3: Dose–response association between later‐day MVPA proportion and incident type 2 diabetes in whole population and by age group. Models were adjusted for age, sex, ethnicity, education, Townsend Deprivation Index, smoking status, sleep duration, family history of diabetes, total MVPA volume and total daily light physical activity. (A) shows associations in overall population; (B) shows associations stratified by age; Solid lines indicate hazard ratios and shaded areas indicate 95% confidence intervals. Hazard ratios were estimated relative to 0% of later‐day MVPA. Abbreviations: MVPA moderate to vigorous physical activity. Figure S4: Age‐specific pattern of relative and absolute type 2 diabetes risk associated with MVPA timing in the overall population and by sex. (A), (D) and (G) showed the age‐specific relative risk estimates among whole population, women, men, respectively; (B), (E) and (H) showed the age‐specific incidence rates among whole population, women, men, respectively; (C), (F) and (I) showed the age‐specific absolute incidence rates difference among whole population, women, men, respectively. Abbreviations: MVPA moderate to vigorous physical activity. Figure S5: Age‐specific pattern of relative and absolute type 2 diabetes risk associated with MVPA volume by baseline dyslipidemia. (A) and (D) showed the age‐specific relative risk estimates among participants without and with baseline dyslipidemia, respectively; (B) and (E) showed the age‐specific incidence rates among participants without and with baseline dyslipidemia, respectively; (C) and (F) showed the age‐specific absolute incidence rates difference among participants without and with baseline dyslipidemia, respectively. Abbreviations: MVPA moderate to vigorous physical activity. Table S1: Missingness of covariates and potential effect modifiers. Table S2: Median (IQR) of weekly MVPA volume by age group and sex. Table S3: Association between MVPA timing and type 2 diabetes risk by age, sex and metabolic disease. Table S4: Sensitivity analysis of the association between volume of MVPA and type 2 diabetes risk by modifiers using ≥ 150 min/week threshold. Table S5: Sensitivity analysis of the association between MVPA timing and type 2 diabetes risk and its effect modification by age using alternative timing cut‐offs. Table S6: Sensitivity analysis of the association between volume of MVPA and type 2 diabetes risk and its effect modification after excluding incident cases occurring within the first 2 years of follow‐up. Table S7: Sensitivity analysis of the association between volume of MVPA and type 2 diabetes risk and its effect modification with additionally adjusting BMI. Acknowledgements The authors acknowledge and appreciate the effort and involvement of the UK Biobank participants and staff in this study. Application no. 81423 was used to conduct the research with the UK Biobank resource. Feng Q., Noordam R., Hu D., Albalak G., Willems van Dijk K., and van Heemst D., “Age‐Specific Associations of Moderate‐to‐Vigorous Physical Activity With Relative and Absolute Risk of Type 2 Diabetes in UK Biobank Participants,” Diabetes, Obesity and Metabolism 28, no. 11 (2026): 10099–10109, 10.1111/dom.71214. Handling Editor: Shan Luo Data Availability Statement UK Biobank data are available through a procedure described at http://www.ukbiobank.ac.uk/using‐the‐resource/. References Associated Data Supplementary Materials Figure S1: Flow chart of study sample selection in UK biobank. Detail exclusion criteria for accelerometry data: Dataset: (1) unexpectedly small or large size of the uncompressed returned by the accelerometer device (Field ID: 90002); (2) data from less than 72 h or lack of data for all 1‐h periods within a 24‐h cycle during the 7‐day data collection (Field ID: 90015); (3) data was not well calibrated (Field ID: 90016); (4) participant data was recalibrated using the accelerometer record from the same device worn by a different participant (Field ID: 90017); (5) data with non‐zero count of interrupted recording periods (Field ID: 90180); (6) data with more than 768 (Q3 + 1.5 × IQR) data recording errors (Field ID: 90182); (7) The overall average proportion of time spent doing moderate‐vigorous activity across the monitoring period is 0 (Field ID: 40049). Figure S2: Dose–response association between weekly MVPA volume and incident type 2 diabetes in whole population and by age group, prediabetes status and dyslipidemia. Models were adjusted for age, sex, ethnicity, education, Townsend Deprivation Index, smoking status, sleep duration, family history of diabetes and total daily light physical activity. (A) shows associations in overall population; (B) shows associations stratified by age; (C) shows associations stratified by prediabetes status; (D) shows associations stratified by dyslipidemia status. Solid lines indicate hazard ratios and shaded areas indicate 95% confidence intervals. Hazard ratios were estimated relative to 0 min/week of MVPA. The x‐axis was truncated at 800 min/week, approximately the 95th percentile of weekly MVPA, to reduce the influence of sparse data at extreme MVPA values. Abbreviations: MVPA moderate to vigorous physical activity. Figure S3: Dose–response association between later‐day MVPA proportion and incident type 2 diabetes in whole population and by age group. Models were adjusted for age, sex, ethnicity, education, Townsend Deprivation Index, smoking status, sleep duration, family history of diabetes, total MVPA volume and total daily light physical activity. (A) shows associations in overall population; (B) shows associations stratified by age; Solid lines indicate hazard ratios and shaded areas indicate 95% confidence intervals. Hazard ratios were estimated relative to 0% of later‐day MVPA. Abbreviations: MVPA moderate to vigorous physical activity. Figure S4: Age‐specific pattern of relative and absolute type 2 diabetes risk associated with MVPA timing in the overall population and by sex. (A), (D) and (G) showed the age‐specific relative risk estimates among whole population, women, men, respectively; (B), (E) and (H) showed the age‐specific incidence rates among whole population, women, men, respectively; (C), (F) and (I) showed the age‐specific absolute incidence rates difference among whole population, women, men, respectively. Abbreviations: MVPA moderate to vigorous physical activity. Figure S5: Age‐specific pattern of relative and absolute type 2 diabetes risk associated with MVPA volume by baseline dyslipidemia. (A) and (D) showed the age‐specific relative risk estimates among participants without and with baseline dyslipidemia, respectively; (B) and (E) showed the age‐specific incidence rates among participants without and with baseline dyslipidemia, respectively; (C) and (F) showed the age‐specific absolute incidence rates difference among participants without and with baseline dyslipidemia, respectively. Abbreviations: MVPA moderate to vigorous physical activity. Table S1: Missingness of covariates and potential effect modifiers. Table S2: Median (IQR) of weekly MVPA volume by age group and sex. Table S3: Association between MVPA timing and type 2 diabetes risk by age, sex and metabolic disease. Table S4: Sensitivity analysis of the association between volume of MVPA and type 2 diabetes risk by modifiers using ≥ 150 min/week threshold. Table S5: Sensitivity analysis of the association between MVPA timing and type 2 diabetes risk and its effect modification by age using alternative timing cut‐offs. Table S6: Sensitivity analysis of the association between volume of MVPA and type 2 diabetes risk and its effect modification after excluding incident cases occurring within the first 2 years of follow‐up. Table S7: Sensitivity analysis of the association between volume of MVPA and type 2 diabetes risk and its effect modification with additionally adjusting BMI. Data Availability Statement UK Biobank data are available through a procedure described at http://www.ukbiobank.ac.uk/using‐the‐resource/.