Introduction

Globally, approximately 14% of adults are living with diabetes, the vast majority of whom have type 2 diabetes mellitus (T2DM)1. Impaired glucose metabolism (IGM, i.e. prediabetes), however, remains a largely hidden condition, affecting an estimated 6% (impaired fasting glucose (IFG)) to 9% (impaired glucose tolerance (IGT)) of the population, where these populations are at considerable elevated risk of developing T2DM2. Although other non-modulating (e.g. genetic) factors are involved, promoting healthy lifestyle behavior is the first-line prevention strategy to slow down the progression from normoglycemia (NG) to IGM or to T2DM3.

Engaging in healthy lifestyle behavior as prevention of disease development or as management of the disease can encompass the optimization of 24-hour movement behaviors (24 h-MBs), which include physical activity (PA), sedentary behavior (SB), and sleep4,5. These behaviors within the 24 h-MB composition collectively influence glycemic control, which in turn, might impact the development of T2DM6,7. Given their interdependence within the finite composition of a 24-hour day, achieving an increase of one behavior, e.g. moderate-to-vigorous PA (MVPA), automatically leads to a decrease in one or more of the other behaviors, i.e. light PA (LPA), SB or sleep. Depending on the ‘reallocation of time’-combination (i.e. which behavior(s) change(s) at the ‘expense’ of others), differences in magnitude of health impact have been seen4,5. For example, reallocating time from SB into LPA and/or MVPA theoretically improves glucose levels, HbA1c and waist circumference, with greater benefits in individuals with T2DM compared to those with IGM or NG7. Moreover, we recently observed that adults with T2DM spend more time in SB and less time in LPA and MVPA than those without diabetes, with no differences in sleep duration8. However, these differences in 24 h-MBs were not found when comparing adults with NG versus adults with IGM9. Therefore, it is hypothesized that the changes in 24 h-MBs over time might depend on health status of adults such as their Glucose Metabolism Status (GMS), with less optimal changes in 24 h-MBs (e.g. more SB, less PA) in adults with T2DM or adults who progress to a worse GMS. Understanding these changes could provide valuable insights in which groups should be prioritized to promote healthier 24 h-MBs as well as which behaviors are feasible to target in terms of interventions.

Currently, limited evidence exists on how 24 h-MBs evolve over time and whether such changes are associated with differences in the evolvement of GMS, i.e. either remaining in their GMS or progressing or regressing from one group to the other. Therefore, this study aimed to investigate the changes in time spent in 24 h-MBs over 9 years of time taking into account all different possible trajectories in GMS, ranging from having a stable GMS over time (i.e. NG, IGM or T2DM) to making progression or improvement in the GMS state.

Methods

Participants and procedure

This study uses longitudinal data from The Maastricht Study, an observational prospective population-based cohort study, the rationale and methodology of which have been described previously10. In brief, the study focuses on the etiology, pathophysiology, complications and comorbidities of T2DM and is characterized by an extensive phenotyping approach. All individuals aged between 40 and 75 years and living in the southern part of the Netherlands were eligible for participation. Participants were recruited through mass media campaigns and from the municipal registries and the regional Diabetes Patient Registry by mailings reaching to 9187 participants in the final cohort. Recruitment was stratified according to known T2DM status with an oversampling of individuals with T2DM for reasons of efficiency. The first phase of the data collection took place between September 2010 and October 2020 within a period of three months per participant (Timepoint 1; T1). A second phase of measurements started in November 2020 and is still ongoing. Participants were chronologically invited to participate in the second phase, reaching 2567 participants who completed the second wave of measurements between November 2020 and April 2023 (Timepoint 2; T2).

The study has been approved by the institutional medical ethical committee (NL31329.068.10) and the Minister of Health, Welfare and Sports of the Netherlands (Permit 131088-105234-PG), conducted in accordance with the Declaration of Helsinki. All participants gave written informed consent.

Included in the present study were all adults with valid available 24 h-MBs data on both T1 and T2 (n = 1957). Adults with other types of diabetes than T2DM (e.g. T1DM, MODY) were excluded (n = 23).

Data collection

Participant characteristics

The following variables were obtained by a general socio-demographic questionnaire at T1 and T2: age (years), sex (male; female), education (low (no education, primary education or lower vocational education); medium (general secondary education, general vocational education or higher secondary and pre-university education); high (higher vocational education or university)), smoking status (never; former; current), alcohol consumption (None; Low (women: ≤7 glasses/week; men: ≤14 glasses/week); high (women > 7 glasses/week; men > 14 glasses/week), marital status (single; married-domestic partnership or civil; widowed; divorced; living together; other), having a partner (yes; no)10. Onset of diabetes (year) was questioned and subtracted from the year of visit 1 to determine the duration of diabetes10.

Diet quality was assessed using a valid and reliable Food Frequency Questionnaire and evaluated based on the Dutch Health Diet Index (DHD15), which scores adherence to the Dutch dietary guidelines on a scale of 0–140, with higher scores reflecting greater adherence11,12. Health-related Quality of Life (HRQOL) was measured by using the valid and reliable 36-item short form health questionnaire13,14. The T-scores retrieved from this questionnaire are standardized against a US population with a mean of 50 and an standard deviation of 10. Higher scores reflect a better HRQOL13,14. Depressive symptoms were measured using the valid and reliable Patient Health Questionnaire (PHQ-9)15. PHQ-9 consists of 9 items with a 0 to 3 point scale. A score of 0–4 represents no to minimal depression, 5–9 is classified as mild, 10–14 is classified as moderate, 15–19 as moderately severe and between 20 and 27 as severe depression15. One question out of the M.I.N.I. Mini International Neuropsychiatric Interview (M.I.N.I.) Dutch version was used which asks about a current episode of depression (yes/no) at the study visit16.

Regarding cardiometabolic health, the following variables were measured at each study visit: Body Mass Index (BMI) in kg/m2, waist circumference in cm, waist to hip ratio, HbA1c in mmol/mol, fasting plasma glucose (FPG) in mmol/L, HDL-cholesterol in mmol/L, LDL-cholesterol in mmol/L, total cholesterol in mmol/L, triglycerides in mmol/L, 2-hour plasma glucose (2hPG) in mmol/L, and blood pressure in mmHg. Medication use was determined by requesting to bring a list of all medication to the research visit. During the medication interview, the proposed medication was registered by the staff by generic name, dose and frequency and was classified as glucose lowering (yes/no), blood pressure lowering (yes/no) and/or lipid lowering medication (yes/no)10.

Body weight and height were recorded without footwear and while wearing light clothing, using a calibrated scale and stadiometer with a precision of 0.5 kg and 0.1 cm, respectively (Seca, Hamburg, Germany)10. Waist and hip circumferences were measured twice using a measuring tape (Seca, Hamburg, Germany)10. Waist circumference was assessed at the midpoint between the lower rib margin and the iliac crest at the end of expiration, while hip circumference was measured at the widest point over the greater trochanter, both recorded to the nearest 0.5 cm10. Plasma glucose levels were assessed using a standardized enzymatic hexokinase reference technique10. Measurements of serum total cholesterol, HDL cholesterol, and triglycerides were conducted through standard enzymatic and/or colorimetric methods with an automated analyzer (Beckman Synchron LX20, Beckman Coulter Inc., Brea, USA)10. When applicable, LDL cholesterol concentrations were estimated using the Friedewald equation10. HbA1c levels were determined via ion-exchange high-performance liquid chromatography (HPLC) using the Variant™ II system (Bio-Rad, Hercules, California, USA)10. Two-hour oral glucose tolerance test (OGTT) was assessed using an oral glucose tolerance test after an overnight fast. Blood samples were taken at baseline, and 15, 30, 45, 60, 90 and 120 min after ingestion of a 75 g glucose drink10. Blood pressure readings were taken on the right arm three times after a 10-minute rest period using a non-invasive blood pressure monitor (Omron 705IT, Japan)10.

Objectively measured 24-hour movement behaviors

The Maastricht Study objectively collected 24 h-MB data using ActivPAL3 accelerometers (version 6.4.1; PAL Technologies, Glasgow, UK). The ActivPAL is a thigh-worn triaxial accelerometer, made waterproof by using a sleeve and tegaderm17. The device was directly attached on the skin of the right thigh with transparent tape and participants were asked to wear the device for eight consecutive days. Data were uploaded using the ActivPAL software and processed using customized software written in MATLAB R2018b (MathWorks, Natick, MA, USA)17.

The first day of recorded data was excluded, as participants first got the attachment of the device followed by physical examination at the research center. Data from the final wear day providing ≤ 14 waking hours of data were excluded from the analysis. Participants were included in the analysis if they provided at least a minimum of five valid days to assess the weekly pattern (Exclusion n = 210). Out of the 1934 participants, 1724 (89%) had a minimum of five valid days of data.

Stepping minutes were further categorized into LPA (< 100 steps/min or standing) and MVPA (≥ 100 steps/min), using standard definitions18. Both were summed to have an estimation of time spent in total PA. The total time spent sitting, and engaging in LPA and MVPA was averaged across valid wear days to calculate daily averages. Sleep and wake periods were identified using an automated algorithm to identify time-in-bed which was used as an estimate for sleep duration17.

Glucose metabolism status trajectories

As we are interested in change over time in GMS, we classified participants into six different trajectory groups based on their evolution in GMS from T1 to T2. Diabetes status was assessed at both time-points using the WHO 2006 criteria (results from a 2 h OGTT and the use of glucose-lowering medication)19. T2DM was defined as FPG ≥ 7.0 mmol/l and 2hPG ≥ 11.1 mmol/l or the use of glucose-lowering medication. IGM was defined as IGT (FPG < 6.1 mmol/l and 2hPG between ≥ 7.8 and < 11.1 mmol/l) and/or IFG (FPG between 6.1 and 6.9 mmol/l and 2hPG < 7.8 mmol/l)19. NG was defined as FPG and 2hPG below the IGM thresholds19.

The first three groups preserved a stable GMS over time and were called the ‘Stable’ groups, i.e. ‘Stable NG’ for adults with a NG at both time-points (n = 975), ‘Stable IGM’ for those with IGM at both time-points (n = 89), and ‘Stable T2DM’ for adults with T2DM at both time-points (n = 282). The so-called ‘Progression’ groups consisted of adults who shifted towards an unhealthier GMS between T1 and T2. The ‘Progression to IGM’ included adults who shifted from NG to IGM (n = 166). The ‘Progression to T2DM’ were adults shifting from NG or IGM to T2DM (n = 111). Adults who shifted from NG to T2DM (n = 36) and from IGM to T2DM (n = 75) were grouped together due to small sample sizes. The last group included participants who reversed their GMS over time. Due to the small samples sizes of the groups with different reversing GMS status, all adults were combined into one group called the ‘Improvers’ (n = 101). This group included adults who shifted from T2DM to IGM (n = 7), from T2DM to NG (n = 10) and from IGM to NG (n = 84) between T1 and T2. Figure 1 gives an overview of the participant flowchart.

Data analysis

Participant characteristics are presented as means and standard deviations (± SD) for normally distributed variables and a median and Inter Quartile Range [IQR] for not normally distributed variables (visual histogram and Shapiro Wilk test < 0.05). Proportions (%) were used to report categorical data. To compare participants’ characteristics between the six GMS groups, a One-Way-Anova was used for continuous and normally distributed variables followed by a Tukey HSD Posthoc testing (Bonferroni correction) or a Kruskal-Wallis Test for non-normally distributed variables followed by a Dunn’s test post-hoc testing (Bonferroni correction) to identify pairwise group differences for sociodemographic data. A Chi-Square test was used for categorical variables to compare characteristics across the six groups, followed by a pairwise Chi-Square to identify pairwise differences. Changes over time for each group separately was done using a paired sample t-test for continuous and normally distributed data and Wilcoxon for continuous and not normally distributed data. A Stuart Maxwell test was used to explore changes over time for categorical variables.

Compositional Data Analysis (CoDA) is a statistical approach to account for the codependency of the 24 h-MBs, where all behaviors within one day are treated within one finite whole. Treating them as a composition is done by using the R package compositions20,21. The 24 h-MBs compositions (consisting of sleep, SB, LPA, and MVPA) were expressed as three isometric log-ratios for each timepoint22. Variation matrices were created to explore the variance covariance of the 24 h-MBs [see Additional File Table 1]22. Linear mixed effects models were used to explore differences in the changes of 24 h-MB compositions from T1 to T2 between the six groups. To use the 24 h-MB composition as the dependent variable in the models, ILRs were used in long ‘stacked’ format with a dummy variable indicating whether they were ILR1, ILR2 or ILR3. As described in Miatke and colleagues (2025), random slopes were added at the log ratio level, grouped by participant ID, and repeated within participant ID (random intercept) and ILRs (random intercept) for each time-point23. A fixed three-way interaction effect between the grouping variable and the dummy variable representing the ILR numbers (24 h-MB composition) and the time-points were used to explore bivariate associations between the different groups and time-points for the 24 h-MB composition (type III MANOVA F-test). Posthoc testing for the interaction effect between time and group was explored to investigate the differences in changes of 24 h-MB composition over time between the six groups (Posthoc testing, Bonferroni correction 0.05/15 = 0.003). The main effect of Time (T1 vs. T2) was explored to investigate differences in 24 h-MB compositions over time for each GMS group separately (Posthoc testing, Bonferroni correction 0.05/6 = 0.008). The main effect of GMS group at T1 was explored to investigate differences in 24 h-MB compositions between groups (one-by-one) at phase 1 (Posthoc testing, Bonferroni correction 0.05/15 = 0.003). Exploration of which behavior within the composition was driving the significant effect was done using mean differences and bootstrapped with 1000 replicates to generate a 95% confidence interval.

Using complete case modelling, participants were excluded if they had missing data on the following covariates at T1, i.e. sex, age, education, smoking, diet. The following models adjusted for different sets of covariates were explored, with the total number of participants having complete cases in between brackets for each model. Model 1: crude (n = 1724); Model 2: model 1 + sex + age + education category (n = 1699), Model 3: model 2 + smoking status + diet (n = 1659). As overweight and obesity are contributors to the risk of diabetes, an additional analysis was performed showing no strong confounding or mediating effect of delta BMI (BMI follow-up – BMI baseline) on the associations. See Additional File 2 for a detailed explanation and interpretation. All analyses were performed in R version 4.1.124. Statistical significance was set at 0.05, unless stated otherwise.

Fig. 1
figure 1

Participants flow chart. T1: Timepoint 1, T2: Timepoint 2, NG: normoglycemia, IGM: Impaired glucose metabolism, T2DM: Type 2 diabetes mellitus. Stable groups refer to individuals who retain a stable GMS from T1 to T2, Progression groups include adults whose GMS condition worsened, so from normoglycemia at T1 to impaired glucose metabolism at T2, or from normoglycemia/impaired glucose metabolism at T1 to T2DM at T2. The Improvers include the adults whose GMS condition improved over time so adults with T2DM/impaired glucose metabolism at T1 to normoglycemia/impaired glucose metabolism at T2.

Full size image

Results

Descriptive results of GMS groups

A total of 1724 participants (50.8% female, median age of 60 [54.0, 64.0] years) were included in this study. Participants in the Stable NG group were significantly younger compared to the other five groups (p < 0.003). Sex was significantly different across the six groups with the Stable T2DM group having the lowest percentage of females (23.8%) and Stable NG group the highest (59.5%) (p < 0.003).

Additionally, body composition at T1 was significantly different across groups with the Stable NG group having the lowest and the Stable T2DM group having the highest data for BMI (24.70 kg/m2 [22.80, 26.80] vs. 28.70 kg/m2 [25.90, 31.48]), WC (88.20 cm [80.45, 95.40] vs. 104.55 cm [95.32, 112.22]), and WHR (0.89 [0.83, 0.96] vs. 1.02 [0.96, 1.07]) (all p < 0.003). The Stable T2DM group showed a borderline elevated systolic BP (141.26 mmHg (17.15)), which was significantly higher than that in the Stable NG group (128.00 mmHg [117.00, 140.00]), the Progression to IGM (134.20 mmHg (17.56)), and the Improvers group (133.90 mmHg (15.85), (all p < 0.003). LDL-cholesterol was the lowest for the Stable T2DM group (2.30 mmol/L 2.30 [1.80, 2.90]) compared to all other groups (p < 0.003). The FPG (7.40 [6.80, 8.38] mmol/l), 2hPG (14.53 (3.56) mmol/l), and HbA1c (49.00 [45.00, 54.00] mmol/mol) of the Stable T2DM group were significantly higher compared to all other groups (p < 0.003).

Smoking status was different between the Stable T2DM group and the Stable NG group with higher percentages as former (59.6% versus 48.3%) and current (11.2 versus 8.4%) smokers and lower percentages in the no smokers groups (31.9% versus 43.3%) regarding the Stable T2DM group (p < 0.003). The DHD15 was significantly lower for the Stable T2DM group (82.20 [70.70, 90.42]) compared to the Stable NG group (85.60 [75.95, 96.65] (p < 0.003). HRQOL Physical component showed the lowest scoring for the Stable T2DM group and was significantly different compared to the Stable NG group, the Progression to IGM group and the Improvers group (p < 0.003). HRQOL Mental component scored significantly higher for the Stable T2DM group compared to the Stable NG group (p < 0.003). However all groups remained within the mean ± 1 SD, indicating a normal range of HRQOL. No significant differences between groups were found for depression.

Table 1 provides an overview of all characteristics of the participants at T1. See Additional File Table 3 for all descriptives at both T1 and T2.

Table 1 Descriptives of GMS groups at timepoint 1.
Full size table

Differences in 24-hour movement behavior compositions between GMS groups at T1

The 24 h-MB composition at T1 was different between the six groups (p < 0.001). Posthoc testing showed significant differences in the 24 h-MB composition between the Stable NG group, the Stable T2DM group, the Progression to T2DM group, and the Progression to IGM group. The Stable T2DM group spent more time in SB (53.3 min/day; [33.3; 73.8]), less time in LPA (− 41.5 min/day; [− 21.9; − 61.8]) and MVPA (− 16.1 min/day [− 10.9; − 20.3]) compared to the Stable NG group. The Progression to T2DM group also spent more time in SB (34.5 min/day, [6.7; 61.3]), less time in LPA (− 29.6 min/day, [− 8.2; − 52.4]) and less time in MVPA (− 10.7 min/day, [− 4.9; − 15.8]) compared to the Stable NG group. The Progression to IGM group had a significantly different 24 h-MB composition compared to the Stable T2DM group with significant less time spent in SB (− 51.8 min/day [− 29.4; − 77.9]), more time in LPA (46.3 min/day [24.7; 72.0]) and more time in MVPA (11.8 min/day [4.7; 19.5]). See Tables 2 and 3 for all pairwise comparison between groups at T1 representing differences in 24 h-MBs. Additional file Table 4 shows all posthoc effects.

Changes in 24-hour movement behavior compositions over 9 years of follow-up between GMS groups

Overall similar trends across groups were noticed, where all groups significantly changed their 24 h-MB compositions over 9.0 ± 0.6 years of follow-up (time effect p < 0.008 for all groups). Results showed a significant but marginal (about 10 min) increase in time spent sleeping for the Stable NG group, the Stable T2DM group and the Progression to IGM group, a significant increase in SB for all groups, a significant decrease in LPA for all groups, and a significant decrease in MVPA except for the Improvers group (Tables 2 and 4 and Additional file Table 4).

Despite this overall trend, differences were found between groups over time. Changes in 24 h-MB compositions from T1 to T2 were significantly different between the Stable T2DM and Stable NG group, between the Progression to T2DM and Stable NG group and between the Stable T2DM and the Improvers group (group*time effect p < 0.003) (Fig. 2). Unfavorable changes in 24 h-MB compositions were found over time for the Stable T2DM group compared to Stable NG with a bigger increase in SB over time (Stable T2DM T1: 598.1 min/day T2: 645.9 min/day; Stable NG T1: 544.7 min/day; T2 566.2 min/day, i.e. difference of 26.4 min/day [5.3; 42.4]), a bigger decrease over time in LPA (Stable T2DM T1: 297.2 min/day T2: 248 min/day; Stable NG T1: 338.7 min/day; T2 308.7 min/day, i.e. difference of − 19.2 min/day [− 3.3; − 32.4]) and an bigger decrease over time in MVPA (Stable T2DM T1: 41.7 min/day T2: 29.8 min/day; Stable NG T1: 57.8 min/day; T2 54.6 min/day, i.e. difference of − 8.7 min/day [− 4.1; − 12.4]) for the Stable T2DM group. The Progression to T2DM group showed a bigger increase over time in SB (Progression to T2DM T1: 579.2 min/day T2: 631.7 min/day; Stable NG T1: 544.7 min/day; T2 566.2 min/day i.e. difference of 31.1 min/day [8.3; 55.6]), and a bigger decrease over time in LPA (Progression to T2DM T1: 309.1 min/day T2: 258.4 min/day; Stable NG T1: 338.7 min/day; T2 308.7 min/day i.e. difference of − 20.7 min/day [− 2.0; − 40.8]), compared to the Stable NG group (p = 0.003). Last, the 24 h-MB composition of the Improvers was more favorable compared to the Stable T2DM group (p < 0.001) with a bigger increase over time in MVPA (Improvers T1: 50.3 min/day; T2 48.6 min/day; Stable T2DM T1: 41.7 min/day T2: 29.8 min/day i.e. difference of 10.2 min/day [3.1; 18.2]). An additional sensitivity analysis confirmed that a change in BMI over time (delta BMI) was neither a confounder or mediator of these relationships (see Additional File 2). See Table 2 for the mean time spent in 24 h-MBs at each timepoint for each group. See Table 5 for all pairwise comparison between groups representing the additional increases or decreases of behaviors within the 24 h-MB compositions. Additional file Table 4 shows all posthoc interaction effects.

Fig. 2
figure 2

Differences in 24 h-MBs over time between GMS groups. NG normoglycemia, IGM impaired glucose metabolism, T2DM type 2 diabetes mellitus, SB sedentary behavior, LPA light physical activity, MVPA moderate to vigorous physical activity.

Full size image
Table 2 Differences in 24-hour movement behaviors (min/day) between groups over time (Group*Time), between timepoints for each group separately (Time), between groups at T1 (group T1).
Full size table
Table 3 Within 24-hour movement behaviors (min/day) changes between groups at T1 (posthoc group T1).
Full size table
Table 4 Within 24-hour movement behaviors (min/day) changes between timepoint 1 and timepoint 2 for each group separately (posthoc time).
Full size table
Table 5 Within 24-hour movement behaviors (min/day) changes between groups over time (posthoc Group*Time).
Full size table

Discussion

This paper explored the associations between changes in 24 h-MBs among adults with different GMS trajectories. Over approximately 9 years, 24 h-MB compositions deteriorated in all groups into the same direction, in that SB increased at the expense of LPA (all groups), while their sleep duration remained relatively stable. Also, MVPA decreased in all groups except in the Improvers group. However, the magnitude of these changes differed between the groups with the stable T2DM and Progression to T2DM groups showing the greatest deteriorations and stable NG and Improvers the least.

Less optimal 24 h-MB compositions in the Stable T2DM and Progression to T2DM groups were already present at T1, which aligns with other research showing that adults with T2DM tend to have less favorable 24 h-MB compositions than those without T2DM8,25. Our results suggest that an already less optimal 24 h-MB composition may be associated with a greater deterioration over time. Furthermore, the less favorable composition observed in the Progression to T2DM group at T1 suggests that these adults could benefit from early identification and lifestyle interventions aimed at optimizing their 24 h-MBs, potentially preventing further GMS progression.

Next, group characteristics might also explain differences in changes in 24 h-MB compositions over time. The Stable NG group demonstrated more favorable body composition characteristics (i.e. lower BMI, WC, WHR) when compared to the Stable T2DM group and the Progression to T2DM group, who had the least favorable adiposity markers. Since adiposity is associated with higher risks of IGM and T2DM26, as well as with less optimal 24 h-MB composition27, these factors may have contributed to the observed differences in changes in 24 h-MB compositions over time across those groups. Finally, a better dietary quality was found for the Stable NG group and the Improvers group compared to the stable T2DM group and the Progression to T2DM group. This may suggest that adults with healthier eating patterns may be more likely to engage in positive lifestyle behaviors, such as more PA and less SB. However, no research has specifically examined these associations among adults with T2DM. Existing studies among adults with overweight and obesity show inconclusive results regarding the correlation between a healthy diet and PA or SB28.

Compared to Stable NG group, adults who progressed from NG or IGM to T2DM (Progression to T2DM) exhibited a similar and less optimal change in 24 h-MB composition over time as the Stable T2DM group. This is clinically important, as it suggests that early lifestyle intervention can be targeted at those with the worst movement behaviors, so as to prevent or reverse glycemic deterioration or diabetes. Whether this occurred in the present study is unknown, as it was not specified whether participants were diagnosed with T2DM between the two visits, making it challenging to assess the impact of receiving such a diagnosis on movement behaviors. Although receiving such a diagnosis may serve as a wake-up call, prompting lifestyle changes, existing literature is inconclusive on this issue. Some studies report an increase in PA following a T2DM diagnosis29,30, but others found no significant changes in PA or SB31,32. Nevertheless, based on the findings in the Improvers group, it could be argued that 24-MBs deteriorated less in those reverting back after T2DM diagnosis, although it is also likely that they were able to adopt dietary changes.

Despite the differences between groups over time, a similar trend was noticed: all groups significantly altered their 24 h-MB compositions over time, increasing their SB while decreasing their LPA and MVPA. However, the Improvers group maintained their MVPA levels over time but this group also exhibited an increase in SB and a decrease in LPA. Despite sustaining their MVPA levels, their overall 24 h-MB composition did not improve, suggesting that their GMS improvement was likely driven by other lifestyle factors, such as diet. Notably, their dietary quality (DHD15) improved from T1 to T2, a trend not observed in other GMS groups except for the Stable NG group.

The changes in 24 h-MBs into less optimal compositions over time might be partly attributed to aging, since decreased levels of PA33 and increased levels of SB34 are often reported in older adults. Sleep duration typically decreases with age and sleep fragmentation increases, however, studies suggest that sleep duration stabilizes in later adulthood (≥ 60 years), which could explain the lack of significant difference in sleep duration over time in the present study35,36. Our findings are in agreement with a paper of older adults showing lower PA, especially MVPA, and higher SB after 9 years of follow-up37. They additionally conducted semi-structured interviews to explore anything that might have happened to interrupt or change their daily routines in the follow-up period37. The most commonly reported reasons were health problems such as injuries and diseases, motor decline, psychosocial factors such as death of someone close, and poor relationships37. Moreover, during middle (30–60 years) and older adulthood (over 60 years), the most commonly reported life events affecting PA fell into four key areas: employment (e.g. job changes, retirement), relationships (e.g. marriage, divorce, widowhood), residence changes (e.g. moving), and health-related issues (e.g. illness or symptoms)38. A paper on 24 h-MB compositions around the transition to retirement showed that differences in 24 h-MBs were dependent on the type of occupation the adults retired from. When retiring from manual labor, there was an increase in SB at the expense of PA, whereas those retiring from non-manual occupations were found to increase sleep at the expense of SB39.

All groups experienced a significant 10 min decrease in MVPA, except for the Stable NG group and the Improvers group. Considering that five minutes of brisk walking (1 mg) is often considered the minimal clinically important difference for accelerometer data, focusing on maintaining the MVPA levels at T1 might be already clinically relevant40. Regarding SB, there do not exist clear evidence on minimal clinically important difference. Last, sleep remained stable across groups, however, no information was available on the sleep quality of the adults.

This is the first study to examine associations of changes in 24 h-MBs across different GSM trajectories over 9 years of follow-up. Other strengths of this study include the use of objective 24 h-MBs measured by ActivPAL and a large cohort of adults. However, the study also has limitations. First, 215 adults were excluded due to not having enough valid 24 h-MBs accelerometer data. These adults were slightly younger, more likely to be male and had a higher BMI. Second, the sleep-wake algorithm used to detect sleep may have introduced some misclassification, such as inaccurately categorizing time spent lying in bed as sleep17. Third, the classification into trajectories of GMS resulted in varying group sizes, ranging from 975 (Stable NG) to 89 (Stable IGM) and the necessity to combine groups due to small sample sizes, e.g. Improvers including all individuals improving over GMS trajectories. Combining these groups may have reduced specificity in analyzing different GMS trajectories, as individuals progressing from NG to T2DM may exhibit less optimal 24 h-MBs compared to those transitioning from IG to T2DM. Fourth, there was a notable sex imbalance in that only 20% of participants in the Stable T2DM group were female, whereas the other groups contained on average 50% females. It is well established that T2DM is more prevalent in middle-aged men, while its prevalence tends to increase in women as they age beyond 60 years41. Therefore, the imbalance of only 20% females in this group limits its generalizability. Next, adults with T2DM in this study had relatively well-controlled diabetes, which may limit the generalizability of the findings to individuals with poorly controlled T2DM. Last, life events potentially impacting 24 h-MBs were not reported in the present study, but could have affected their 24 h-MB composition to a large extent.

Conclusion

The groups with stable T2DM and those who progressed to T2DM experienced the most unfavorable trends in 24 h-MBs over 9 years of time, showing a greater increase in SB and a more pronounced decline in LPA and MVPA. Moreover, less optimal 24 h-MB compositions in the Stable T2DM and Progression to T2DM groups were already present at T1, which suggest that an already less optimal 24 h-MB composition may be associated with a greater deterioration over time. Despite differences across groups, all GMS trajectory groups exhibited a similar trend of an increase in SB and a decline in LPA and MVPA, except for those improving in GMS who maintained a stable MVPA. These findings highlight the need to raise awareness about the natural progression of lifestyle behaviors, which tend to become less favorable over time, particularly for individuals with poorer GMS profiles.