Type 2 diabetes (T2D) is a significant global health concern, contributing to substantial mortality. Physical activity (PA) is recognized as crucial for reducing this excess mortality risk. However, much of the existing research relies on self-reported PA data, which is susceptible to recall bias and limitations in capturing light-intensity activity. This study aimed to address these limitations by using objectively measured PA data from accelerometers to investigate the dose-response relationship between PA and mortality in a large cohort of individuals with T2D. The UK Biobank, with its extensive dataset and accelerometer-measured PA data, provided a unique opportunity to conduct a robust analysis, overcoming limitations of previous studies relying on self-reported data, which often lack the precision and detail needed to develop effective PA recommendations for T2D patients. The lack of accelerometer-based evidence, especially studies with large sample sizes and detailed dose-response analyses focusing on T2D, highlighted the importance of this investigation. The study aimed to clarify the relationship between PA duration and intensity and various mortality outcomes (all-cause, cancer, and cardiovascular disease) in this high-risk population. Understanding this relationship is crucial for developing tailored PA guidelines to improve the health and longevity of T2D patients. This study also planned to explore whether factors related to diabetes severity modify the observed association, offering further refinement for personalized recommendations.

Previous research has indicated a strong link between PA and reduced mortality risk in individuals with T2D. However, most studies utilized self-reported PA data, which is prone to recall bias and lacks the sensitivity to capture light-intensity PA (LPA). Meta-analyses have demonstrated a reduced mortality risk associated with higher levels of PA in T2D patients; however, the precise volume of PA required remains unclear due to inconsistencies in study methodologies. Studies employing accelerometer-based PA measurements in T2D populations have been limited, with smaller sample sizes and less comprehensive analyses. In contrast, accelerometer studies in the general population consistently found stronger associations between PA and mortality compared to self-reported data, indicating the potential for more accurate and impactful results when using objective measurements. Therefore, this study aimed to leverage accelerometer data from the UK Biobank to provide a more accurate and detailed assessment of the association between PA and mortality in a sizable cohort of T2D individuals.

This prospective cohort study utilized data from the UK Biobank, specifically focusing on 4003 participants with T2D at baseline. Participants wore Axivity AX3 accelerometers on their dominant wrist for seven days, recording triaxial acceleration data at 100 Hz. Data were processed to determine the duration of light-intensity PA (LPA), moderate-intensity PA (MPA), vigorous-intensity PA (VPA), and moderate-to-vigorous-intensity PA (MVPA). The primary outcomes were all-cause, cancer, and cardiovascular disease (CVD) mortality, ascertained through linkage with national death registries. A comprehensive set of covariates, including demographic factors, lifestyle variables (smoking, alcohol consumption, diet, sleep), anthropometric measurements (BMI, waist circumference), diabetes-related factors (duration, HbA1c, medication use), and health history (cancer, CVD, hypertension), were incorporated into the analysis. Restricted cubic splines in Cox proportional hazards models were employed to examine the dose-response relationships between PA duration and mortality risk, while accounting for potential confounders. The proportional hazard assumption was verified. Population-attributable fractions (PAFs) were calculated to quantify the potential impact of increased PA on mortality reduction. Sensitivity analyses were conducted to assess the robustness of the findings by excluding participants with poor self-rated health, early deaths, those with pre-existing conditions, and by employing different analytical approaches (e.g., multiple imputation, competing risk analysis, metabolic equivalent of task (MET) calculations). Stratified analyses explored effect modification by various demographic and clinical factors.

The study revealed significant inverse associations between the duration of PA and mortality risk, irrespective of PA intensity. L-shaped dose-response curves were observed for all-cause and cancer mortality, while a linear negative association was found for CVD mortality. The inflection points of the L-shaped curves varied across PA intensities, suggesting that the optimal PA duration varies according to intensity. Compared to the least active participants, substantial reductions in mortality risk were observed at the inflection points for each PA intensity: -60% for LPA at 1800 min/week, -70% for MPA at 300 min/week, -45% for VPA at 30 min/week, and -70% for MVPA at 300 min/week. PAFs indicated that a significant proportion of deaths could have potentially been prevented with higher PA levels: 18.8% for LPA, 28.0% for MPA, and 31.1% for VPA. These associations were largely consistent across various sensitivity analyses and strata. In particular, at least 150 min/week of MPA or 30 min/week of VPA was significantly associated with lower mortality risk in patients with T2D. The risk matrix further demonstrated that various combinations of PA intensities were associated with reduced mortality risks, indicating flexibility in achieving beneficial health outcomes through different PA regimens. Note that the results of this study were robust in various sensitivity analyses.

This study's findings strongly support the established benefits of PA in reducing mortality risk among individuals with T2D. The use of accelerometer-derived PA data provided a more precise and objective measure of PA compared to self-reported data, which may explain the larger effect sizes observed in this study compared to previous research using self-reported PA data. The L-shaped dose-response relationships highlight the importance of achieving a certain threshold of PA, beyond which further increases may yield diminishing returns. The findings suggest a need for flexible recommendations tailored to individual preferences and capabilities, as evidenced by the effectiveness of various combinations of PA intensities. These observations strongly suggest that the current recommendations for physical activity in T2D patients, while beneficial, may benefit from further optimization by considering both PA duration and intensity using objective measures.

This large-scale prospective cohort study using accelerometer-measured PA data provides robust evidence of the significant association between PA duration and reduced mortality risk in T2D individuals. The results suggest a need for updated PA guidelines that emphasize both duration and intensity and that account for the L-shaped dose-response relationships observed. Future research should investigate the underlying biological mechanisms of the observed association, further refine the optimal PA recommendations based on individual characteristics and preferences, and explore the potential impact of other PA properties such as fragmentation, diurnal pattern, and distributional pattern on mortality risk in T2D patients. Larger studies across various populations could further strengthen the generalizability of these findings.

While this study had a large sample size and used objective PA measurements, several limitations exist. The 7-day accelerometer measurement may not perfectly reflect habitual PA. Wrist-worn accelerometers might not fully capture all types of PA. Residual confounding, even after adjusting for numerous factors, cannot be entirely excluded. Selection bias is also a concern, as participants included in the analysis were healthier than the broader T2D population. The generalizability of findings to other populations needs further verification. Finally, the study was observational, preventing causal inferences. Future work should address these limitations for a more comprehensive understanding.