Aims To evaluate the effects of habitual leisure-time physical activity (LTPA) on incident type 2 diabetes in a prospective cohort of Chinese adults with impaired fasting glucose (IFG).
Methods 44 828 Chinese adults aged 20–80 years with newly detected IFG but free from cardiovascular and cerebrovascular disease were recruited and followed up from 1996 to 2014. Incident type 2 diabetes was identified by fasting plasma glucose ≥7 mmol/L. The participants were classified into four categories based on their self-reported weekly LTPA: inactive, low, moderate, or high. Hazard ratios (HRs) and population attributable fractions (PAFs) were estimated with adjustment for established diabetic risk factor.
Results After 214 148 person-years of follow-up, we observed an inverse dose–response relationship between LTPA and diabetes risk. Compared with inactive participants, diabetes risk in individuals reporting low, moderate and high volume LTPA were reduced by 12% (HR 0.88, 95% CI 0.80 to 0.99; P=0.015), 20% (HR 0.80, 95% CI 0.71 to 0.90; P<0.001), and 25% (HR 0.75, 95% CI 0.67 to 0.83; P<0.001), respectively. At least 19.2% (PAF 19.2%, 95% CI 5.9% to 30.6%) of incident diabetes cases could be avoided if the inactive participants had engaged in WHO recommendation levels of LTPA. This would correspond to a potential reduction of at least 7 million diabetic patients in the Greater China area.
Conclusions Our results show higher levels of LTPA are associated with a lower risk of diabetes in IFG subjects. These data emphasise the urgent need for promoting physical activity as a preventive strategy against diabetes to offset the impact of population ageing and the growing obesity epidemic.
- Physical Activity
- Cohort Study
Statistics from Altmetric.com
Estimations suggest there are about 112 million diabetic patients across the Greater China area (Mainland China, Hong Kong and Taiwan) accounting for 40–60% of premature deaths before the age of 60 years and at least US$51 billion of economic burden.1 Progressive deterioration in glucose metabolism occurs many years before the clinical diagnosis of type 2 diabetes, for which impaired fasting glucose (IFG) is an early detectable pathological change. Individuals with IFG constitute a significant proportion of the Chinese population, with one in four Chinese adults meeting the American Diabetes Association (ADA) definition of IFG.2 Every year 6–9% of individuals with IFG progress to diabetes.3 More importantly, persistent dysglycaemia driven by insulin resistance leads to endothelial dysfunction and vascular complications. Compared with normoglycaemic individuals, people with IFG have a higher risk of vascular and chronic kidney disease and all-cause mortality.4 It is therefore clinically important to promote preventive strategies targeting individuals with IFG to delay the progression to type 2 diabetes, vascular complications and premature death.
Physical activity is an effective strategy to maintain cardiometabolic health.5 6 In individuals with impaired glucose tolerance (IGT), a diet and physical activity intervention programme has been shown to reduce the risk of diabetes and cardiovascular mortality by 45% and 41%, respectively.7 IFG represents hepatic insulin resistance, while IGT occurs when insulin production is impaired. IGT is therefore considered as an advanced stage of pre-diabetes, and the majority of previous prevention trials focus on individuals with IGT. Although it is believed that prevention programmes against type 2 diabetes should start as early as possible, studies demonstrating the protective effects of physical activity in IFG populations are scarce. Therefore, we assessed the association between habitual leisure-time physical activity (LTPA) and incident type 2 diabetes in a prospective cohort of Chinese adults with IFG as defined by the ADA criteria. Population attributable fractions (PAFs) were computed to estimate the proportion of preventable diabetic cases in this IFG population.
The present study was based on an ongoing population-based cohort of 570 414 Chinese adults aged 20–80 years (up to December 2014) who participated in a standard medical screening programme run by the MJ Health Management Institution in Taiwan. Details of the cohort have been published elsewhere.8 Briefly, this is an open cohort which was started in 1996. All the participants were of Chinese descent residing in Taiwan. They joined the MJ Health Screening Programme through a paid membership and were encouraged to visit the institution periodically to receive comprehensive medical assessments, including measurement of fasting blood glucose and lipids, physical and biomedical examinations, and a self-administered questionnaire documenting sociodemographic parameters, lifestyle information (physical activity, diet, smoking, drinking and sleep) and detailed history of physician-diagnosed disease and treatments. From 1996 to 2014, a total of 248 481 participants visited the institution at least twice. Among them 56 451 non-diabetic participants whose fasting plasma glucose (FPG) values ranged from 5.6–6.9 mmol/L in their first visit were identified as IFG. Their glycaemic status was monitored in subsequent visits. The subsequent development of type 2 diabetes in these individuals was used to determine the incidence. Yearly visits were made by 98.7% of the participants, and the total number of visits ranged from 2–19. To reduce the risk of reverse causality, we further excluded 6905 participants who reported pre-existing physician-diagnosed cardiovascular and cerebrovascular disease, such as hypertension, coronary heart disease and stroke, from analyses; 4718 participants missing physical activity information were also excluded. The final study population comprised 44 828 participants with newly detected IFG. Selection of the participants is presented in the online supplementary figure 1. Compared with those IFG participants with only one visit, the 44 828 participants in the present study had similar baseline distributions of sex (male: 59.6% vs 63.6%), age (mean: 43.7 vs 42.6 years), LTPA (mean: 5.3 vs 5.9 MET (metabolic equivalent)-hours/week), FPG (mean: 5.9 vs 5.9 mmol/L), and body mass index (BMI) (mean: 24.4 vs 24.2 kg/m2).
All participants gave informed consent to authorise the MJ Health Management Institution to analyse data generated from the MJ Programme. Personal identification details were removed and remained anonymous when the data were released for research purposes.
Assessment of physical activity
Physical activity was assessed by three questions in a self-administered questionnaire at every visit. First, participants were asked to report the intensity of weekly LTPA performed in the past month with examples of activities under four intensity categories: light (eg, walking), moderate (eg, brisk walking), medium vigorous (eg, jogging), and high vigorous (eg, running). According to Ainsworth’s compendium of physical activities,9 a MET value (1 MET=1 kcal/hour/kg) was assigned to each intensity category as follows: 2.5 METs for light, 4.5 METs for moderate, 6.5 METs for medium vigorous, and 8.5 METs for high vigorous.8 Second, participants were asked to estimate the duration usually spent on LTPA every week in the past month. The participants who did not do any leisure-time exercise or exercised less than 1 hour a week were classified as inactive.8 9 The volume of LTPA (MET-hours/week) was calculated by multiplying the intensity (METs) by duration (hours/week). It requires 7.5 MET-hours/week to achieve the minimum level of the WHO recommended LTPA.10 The participants were therefore classified into one of the following categories using cut-points of 3.75 MET-hours/week (half of the recommended level), 7.5 MET-hours/week (recommended level), and 15.0 MET-hours/week (double the recommended level). This resulted in the designation of the following categories: inactive (no LTPA or LTPA <3.75 MET-hours/week), low (LTPA 3.75 to <7.5 MET-hours/week), moderate (LTPA 7.5 to <15.0 MET-hours/week), and high (LTPA ≥15.0 MET-hours/week). Thirdly, the participants were asked to categorise their physical labour intensity at work with various examples: ‘Mostly sedentary (eg, clerk)’, ‘Sedentary with occasional walking (eg, seamstress)’, ‘Mostly standing or walking (eg, retail salesperson)’, ‘Hard labour (porter)’. Details of conversion, content validity and reliability of the physical activity questions have been published previously.8
Outcome and covariate measurements
Incident type 2 diabetes was defined as FPG ≥7 mmol/L measured after an overnight fast for 12 hours and/or self-reported physician-diagnosed type 2 diabetes. Body height and weight were measured in participants with light clothing and barefoot using an auto-anthropometer (Nakamura KN-5000A, Tokyo, Japan). Waist circumference was measured at the midway between the top of the hip bone and the bottom of the ribs. Systolic (SBP) and diastolic (DBP) blood pressures and heart rate were measured on the right arm by an auto-sphygmomanometer (Citizen CH-5000, Tokyo, Japan). FPG, total cholesterol (TC), high- (HDL-C) and low- (LDL-C) density lipoprotein-cholesterol, and triglycerides were measured in plasma enzymatically with a validated auto-analyser (Hitachi 7150, Tokyo, Japan). Complete blood count was measured by a haematology analyser (Abbott Cell-Dyn 3500/3700, USA).
We examined the association of LTPA and incident type 2 diabetes using the Cox proportional hazards method. Model 1 was adjusted for sex and age (continuous). Model 2 was further adjusted for marital status (single, married or cohabiting, divorced or widowed), education (primary school or less, secondary school, tertiary level or higher), physical labour at work (mostly sedentary, sedentary with occasional walking, mostly standing or walking, hard labour), smoking (never, ever), alcohol drinking (frequency: <1/week, ≥1/week), sleep duration (<6 hours/day, 6–8 hours/day, >8 hours/day), vegetable intake (<1 bowl/day, ≥1 bowl/day), SBP (continuous), heart rate (continuous), and TC (continuous). Model 3 attempted to identify possible factors mediating the association by further adjusting for BMI, waist circumference, FPG, triglycerides and white blood cell count (WBC) to determine whether physical activity exerts an anti-diabetic effect through its impact on these factors.
The population attributable fraction (PAF)11 12 associated with LTPA (ie, how many incident type 2 diabetes can be prevented if the participants in the inactive group engaged in more physical activity) was estimated by the PUNAFCC module in STATA13 using the adjusted hazard ratios from Model 2. As the prevalence of physical activity and its hazard ratios for type 2 diabetes in the general population were similar between Mainland China, Hong Kong and Taiwan,8 14 15 we assumed the prevalence of LTPA and risk for diabetes found in the current population to be similar to those found in those settings. The PAFs calculated from this study (a Taiwanese IFG population) were used to estimate the number of preventable diabetic cases in the total IFG population for the Greater China area.
The proportional hazard assumption was examined by plotting the Kaplan-Meier survival curves and by Schoenfeld residuals. There was no evidence against the proportionality assumption. Because no interaction with sex was found in any models of analyses (all P for interaction terms >0.05), combined results for both sexes were reported. Sensitivity analysis was conducted by excluding incident type 2 diabetes identified in the first 2 years of follow-up to address potential reverse causality. All statistical analyses were performed using STATA version 14.0 (Stata Corporation, College Station, TX, USA). A two-tailed P value <0.05 was considered statistically significant.
Table 1 shows the baseline characteristics of the study population across the four categories of LTPA. More than half (54.6%) of the participants were inactive. Participants engaging in low and moderate volumes of physical activity constituted 18.9% (n=8450) and 11.9% (n=5328) of the study population, with 14.7% (n=8450) performing a high volume of LTPA. There was no significant difference in baseline characteristics across the four groups. After 214 148 person-years of follow-up, we identified 2535, 731, 542, and 612 new cases of type 2 diabetes in the groups with inactive, low, moderate, and high physical activity volume, respectively. The incidence rate of type 2 diabetes was 2.15 per 100 person-years in inactive individuals and 1.95 per 100 person-years in individuals doing low to high volume LTPA.
A dose–response relationship was observed between LTPA and the risk of type 2 diabetes (table 2 and online supplementary table 1). The coefficient between the volume of LTPA and incident diabetes was −0.66 (95% CI −1.04 to −0.28) (P=0.001) (see online supplementary table 1). A higher volume of LTPA was associated with lower hazard ratios of type 2 diabetes (table 2). Compared with inactive participants, diabetes risk in individuals reporting low, moderate and high volume LTPA were reduced by 12% (HR 0.88, 95% CI 0.80 to 0.99), 20% (HR 0.80, 95% CI 0.71 to 0.90), and 25% (HR 0.75, 95% CI 0.67 to 0.83), respectively (Model 2, P for trend <0.001). Repeated analyses excluding SBP, heart rate and TC showed similar results (see online supplementary table 2). Further adjustment for BMI, waist circumference, FPG, triglycerides and WBC (Model 3) attenuated the estimates in low, moderate and high LTPA groups by 67%, 65% and 60%, respectively, suggesting physical activity may reduce the risk of diabetes via reducing central obesity, improving glucose and lipid metabolism, and decreasing systemic inflammation.
The PAFs were calculated to evaluate the incidence reduction that could be achieved in the study participants if the inactive IFG subjects were to engage in a higher volume of LTPA (table 3). In general, the onset of type 2 diabetes in the study population could have been reduced by 9.09% (95% CI 2.15% to 15.54%) if the inactive subjects had taken part in a low volume of LTPA (3.75 to <7.5 MET-hours/week), by 15.83% (95% CI 7.47% to 23.43%) if the inactive subjects had performed a moderate volume of LTPA (7.5 to <15.0 MET-hours/week), and by 18.75% (95% CI 11.80% to 25.15%) if the inactive subjects had engaged in a high volume of LTPA (≥15 MET-hours/week). The reduction of type 2 diabetes incidence by LTPA was also estimated for different combinations of intensity and duration of physical activity (table 4). The incident diabetic cases would have been reduced by 19.2% if the inactive subjects had performed at least 150 min of moderate-intensity physical activity a week as recommended by the WHO.
These findings remained unchanged when incident type 2 diabetes diagnosed within the first 2 years were excluded in sensitivity analyses (see online supplementary table 3–5).
The incidence rate of type 2 diabetes in adults with IFG is approximately 10 times that in normoglycaemic adults according to our observations of 570 414 participants of the MJ Programme (2.06 vs 0.24 per 100 person-years, unpublished data). However, compared with those with IGT, individuals with IFG have been largely overlooked in prevention programme trials for type 2 diabetes. To our knowledge, this is the largest prospective cohort to investigate the protective effects of physical activity against type 2 diabetes among IFG subjects. We observed an inverse dose–response relationship between LTPA and the risk of diabetes in IFG subjects. The risk reduction associated with low to high volume of LTPA (≥3.75 MET-hours/week) ranged from 12–25% after adjusting for physical labour at work and other confounding factors. Our findings were consistent with a small prospective study of 1318 Chinese adults with IFG in which the risk of type 2 diabetes was found to be 35% lower in physically active subjects.15 Additionally, we estimated the PAF of LTPA which takes into account both the relative risk and prevalence of an exposure (ie, physical activity) in the study population. The advantage of PAF is that it reveals the total effect size of an exposure (ie, physical activity) for the whole study population—for instance, an exposure may have a weak relative risk but large total effect size if it is widely prevalent and affects a large proportion of the population. In this study, about one fifth of the observed incident diabetes cases could have been avoided if the inactive individuals had engaged in WHO recommended levels of LTPA. In the approximately 370 million Chinese adults with IFG in the Greater China area,2 16 increasing LTPA by one category would correspond to a potential reduction of at least 7 million cases of diabetes under the assumption that the prevalence of LTPA in the whole IFG population is similar to that observed in the Taiwanese IFG population.
Potential mechanisms underlying the protective effects of physical activity against type 2 diabetes include favourable changes in body weight, adiposity, insulin sensitivity, lipids profile and systemic inflammation.17–20 In this study, further adjustments for the potential mediating factors BMI, waist circumference, FPG, triglycerides and WBC attenuated the effects of LTPA, which suggested the risk reduction associated with physical activity was largely (60–67%) explained by adiposity and associated factors. This is consistent with previous studies in other populations.20–22 Experimental studies have demonstrated that physical activity helps to build muscle mass and stimulates glucose uptake in skeletal muscle by activating the AMP-activated protein kinase (AMPK) pathway and deactivating the Rab-GTPase-activating protein TBC1D1.23 24 Physical activity also improves insulin sensitivity by reducing intramuscular triglycerides and ceramides.25
It should be noted that LTPA was assessed via a self-administered questionnaire rather than by direct objective measures. Currently, no direct objective method can capture both energy expenditure and information documenting physical activity. Motion sensors may record the intensity and duration of physical activity but the estimation of energy expenditure varies according to the algorithms used. Current sensor technology is expensive and logistically inconvenient to use in large population-based studies. Therefore, we estimated LTPA using a questionnaire whose content validity and reliability had been examined previously.8 In the present study, participants who exercised not more than 1 hour/week were classified as inactive; however, some individuals may have exercised <60 min a week but satisfied the minimum requirement of the WHO recommendations and might have been misclassified. Since these ‘inactive’ individuals carry less disease risk than those who do no exercise at all, the risk for diabetes in the inactive group of our study might be underestimated.
We only included the IFG participants who had two or more health assessments in this study. These individuals had similar distributions of sex, age, BMI, LTPA and glucose to those who underwent the health assessment only once. There is no evidence of selection bias. The annual rate of progression from IFG to diabetes in this study is lower than data from previous studies (6–9%).3 This is probably because our study used a higher cut-off of IFG (FPG ≥5.6 mmol/L) than those previous studies (FPG ≥6.1 mmol/L), and thus included a larger denominator. We did not undertake oral glucose tolerance tests (OGTT) or measure glycated haemoglobin (HbA1c) that could also be used in identifying diabetes , which might underestimate the incidence of diabetes and potentially provides a conservative estimate of the true risk. As vegetables are considered a key indicator of a healthy diet,26 27 we asked the participants to report their intake of vegetables in the past week and adjusted our estimates for this factor. However, our estimates did not include information describing the effects of dietary components other than vegetable intake, family history of diabetes, or time-varying covariates. Further studies investigating the interactions between genetic makeup and physical activity and diet and longitudinal changes of these factors may provide more insights into the pathophysiological pathways of physical activity and help to tailor physical activity and diet intervention programmes to the needs of individuals.
In summary, we found that higher levels of LTPA are associated with a lower risk of diabetes in a large population of Chinese adults with IFG. The beneficial associations of physical activity can be observed even at low intensity/volume of LTPA which are attainable by an ageing population. If the prevalence of LTPA in the whole IFG population of the Greater China area is similar to that in our study (a Taiwanese IFG population), the risk reduced by LTPA corresponds to a potential decrease of at least 7 million cases of Chinese diabetic patients and may offset the rapid increases resulting from population ageing and the ongoing obesity epidemic. However, physical inactivity is still highly prevalent in the Greater China area. More than three quarters of Chinese adults are not able to perform sufficient physical activity to reap such health benefits.8 14 Our findings emphasise the urgent need to promote physical activity as a strategy for diabetes prevention.
What are the findings?
Leisure-time physical activity (LTPA) is negatively associated with the risk of diabetes in Chinese adults with impaired fasting glucose. The risk reduction associated with low to high volume of LTPA (≥3.75 MET-hours/week) ranged from 12–25% after adjusting for physical labour at work and other confounding factors.
About one fifth of the observed incident diabetes cases could have been avoided if the inactive individuals had engaged in WHO recommended levels of LTPA. In the approximately 370 million Chinese adults with IFG, increasing LTPA by one category would correspond to a potential reduction of at least 7 million cases of diabetes.
The risk reduction associated with physical activity can largely (60–67%) be explained by adiposity and associated factors.
How might it impact on clinical practice in the future?
Our findings emphasise the urgent need to promote physical activity as a strategy for diabetes prevention.
Contributors XQL designed the study, reviewed and revised the manuscript. HBD analysed and interpreted the data and drafted the manuscript. GNT interpreted the data, reviewed the manuscript and contributed to critical revision of the manuscript for important intellectual content. XL, TCC, ZZ, LC, EKY, TT and MCSW interpreted the data, reviewed and revised the manuscript. All authors can take responsibility for the integrity of the data and the accuracy of the data analysis and have read and approved the final manuscript.
Funding This study is partially supported by Environmental Health Research Fund (7104946). HBD is partially supported by the Faculty Postdoctoral Fellowship Scheme of Faculty of Medicine of the Chinese University of Hong Kong.
Competing interests None declared.
Ethics approval The ethics approval of the present study was obtained from the Joint Chinese University of Hong Kong and New Territories East Cluster Clinical Research Ethics Committee in Hong Kong (No. 2015.672).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.