Physical activity and the incidence of type 2 diabetes in the Shanghai women s health study

Int. J. Epidemiol. Advance Access published September 19, 2006 Published by Oxford University Press on behalf of the International Epidemiological Association Ó The Author 2006; all rights reserved. International Journal of Epidemiology doi:10.1093/ije/dyl209 Physical activity and the incidence of type 2 diabetes in the Shanghai women s health study Raquel Villegas, 1 Xiao-Ou Shu, 1 * Honglan Li, 2 Gong Yang, 1 Charles E Matthews, 1 Michael Leitzmann, 3 Qi Li, 2 Hui Cai, 1 Yu-Tang Gao 2 and Wei Zheng 1 Accepted 23 August 2006 Background Methods Results Conclusions Keywords Leisure-time physical activity (LPA) has been associated with a reduced risk of type 2 diabetes. However, the potential effect of other types of physical activity on type 2 diabetes is still uncertain. The aim of this study was to examine the effect of occupational, commuting, daily living, and LPA on the incidence of type 2 diabetes in a cohort of middle-aged women. We prospectively followed 70 658 women who had no prior history of diabetes at study recruitment for 4.6 years. Participants completed in-person interviews at baseline that collected information on diabetes risk factors including physical activity habits. Anthropometric measurements were taken by trained interviewers. Multivariate-adjusted hazard ratios were estimated by levels of occupational, commuting, daily living, and LPA. We documented 1973 incident cases of diabetes during 326 625 person-years of follow-up. LPA and daily living physical activity (DPA) were associated with a moderately reduced risk of type 2 diabetes. The relative risk for type 2 diabetes associated with LPA and DPA categories were 1.00, 0.89, 1.05, and 0.83, (P trend 5 0.12) and 1.00, 0.98, 0.95, and 0.88, (P trend 5 0.06) respectively. LPA was associated with lower risk of type 2 diabetes in employed participants (P trend 5 0.09) while DPA was mainly associated with a reduction in risk in non-employed participants (P trend,0.01). While occupational physical activity was not associated with type 2 diabetes risk in this population, commuting to work was associated with a reduction in risk. A combination of DPA and LPA was associated with a reduced risk of type 2 diabetes. This study suggests that physical activity, either from leisure-time exercise or daily activity reduces the risk of type 2 diabetes in women, supporting the current health promotion efforts encouraging both exercise and non-exercise activity levels. occupational, daily living leisure, commuting physical activity, type 2 diabetes The prevalence of type 2 diabetes has been increasing rapidly worldwide. 1 As there is no available cure at the moment for type 2 diabetes, primary prevention is of great importance. 2 Identification of modifiable risk factors for the development 1 Department of Medicine, Vanderbilt Epidemiology Center Vanderbilt- Ingram Cancer Center, Vanderbilt University School of Medicine, Nashville, TN 37232, USA. 2 Department of Epidemiology, Shanghai Cancer Institute, Shanghai, China. 3 Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD 20892, USA. * Corresponding author. Center for Health Services Research, 6009 Medical Center East, Vanderbilt University, Nashville TN 37232-8300, USA. E-mail: xiao-ou.shu@vanderbilt.edu of diabetes offers the potential for primary prevention of this important public health problem. 3 Physical inactivity has emerged in epidemiological studies as an independent risk factor for type 2 diabetes. Observational studies and intervention trials have shown a beneficial effect of regular exercise on both insulin resistance and glucose intolerance. 4 18 A sedentary lifestyle (defined as hours of television watching and or having a sedentary occupation) was associated with a higher risk of type 2 diabetes in the Nurses Health Study and the Health Professional Follow-up Study. 7,19 In the past, most studies that have examined the influence of physical activity on type 2 diabetes incidence have focused on leisure-time physical activity (LPA), seldom on other types 1of10

2of10 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY of activity such as occupational and household/daily-liferelated physical activity. Non-leisure time activity may play a role in disease risk particularly in countries, such as China, Denmark, The Netherlands, and Finland where activities related to commuting, daily living, or occupation comprise a considerable proportion of the total physical activity. 20 A recent report from a Finnish prospective study suggests that occupational, commuting, and LPA are independently and inversely associated with reduced risk of type 2 diabetes. 20 The British Heart Regional Study has reported inverse associations between type 2 diabetes and physical activity scores reflecting LPA, occupational physical activity (OPA), and travelling to work physical activity combined. 14,17 Data on associations between household/daily-life physical activity and the risk of diabetes are scarce. Furthermore most previous studies have been conducted in North American and European countries and limited data are available from China, the most populous country in the world. The incidence of type 2 diabetes has markedly increased over the past 2 decades in China, with the prevalence of type 2 diabetes increasing dramatically in urban areas such as Shanghai. 13,21 23 The Shanghai Women s Health Study, (SWHS), a large population-based cohort study initiated in 1997, uses an instrument that was designed specifically to assess various types of physical activities of women in an environment with high prevalence of active lifestyle. Thus, the SHWS provides an exceptional opportunity to investigate associations between different types of physical activity (daily living, occupational, leisure-time, and commuting to work) in relation to risk of type 2 diabetes in middle-aged women. In this cohort, ~53 % of the participants were employed at the time of the survey. This allowed for a comprehensive evaluation of LPA and household/ daily living activity separately on women who were employed and women who were not employed at the time of the baseline survey. Research design and methods Study population The Shanghai Women s Health Study (SWHS) is a populationbased prospective cohort study conducted in Shanghai, China. All eligible women (n 5 81 170) who were aged 40 70 years and resided in seven typical urban communities between March 1997 and May 2000 were approached for the study and 75 221 women were enrolled yielding a participation rate of 92.7%. Participants completed a detailed baseline survey including an in-person interview for assessment of dietary intake, physical activity, and other lifestyle factors. Details of the SWHS survey have been reported elsewhere. 24,25 In person follow-up for all living cohort members was first conducted from 2000 to 2002 with a response rate of 99.8% and a second follow-up was conducted from 2002 to 2004 with a response rate of 98.7%. Only 934 participants were lost to follow up. Physical activity Detailed assessment of physical activity was obtained using a validated questionnaire. 26 The validity of the questionnaire was evaluated by comparing Spearman correlations (r) for the physical activity questionnaire with two criterion measures administered over a period of 12 months (four 7-day physical activity logs and up to 28 7-day physical activity questionnaires). Significant correlations between the physical activity questionnaire and criterion measures for exercise were observed (physical activity log, r 5 0.74; 7-day physical activity questionnaire, r 5 0.80). Significant correlations between the physical activity questionnaire lifestyle activities and the 7-day questionnaire were also observed (r 5 0.30 0.88). The reproducibility of the questionnaire (2-year test retest) was evaluated using kappa statistics and intraclass correlation coefficients (kappa 5 0.64 and the intraclass correlations coefficients were between 0.14 and 0.54). 26 The questionnaire evaluated regular exercise and sports participation during the last 5 years. Participants were asked first if they had engaged in regular exercise/sports (at least once a week for the last 3 months) during the past 5 years. Exercisers were asked to report details for up to three types of exercises/sports (i.e. type, hours/week, and years of participation in each activity). Physical activity energy expenditure was estimated using standard metabolic equivalent values (MET). 27 Exercise/sports energy expenditure was estimated by the weighted average of energy expended in all activities reported over the last 5 years preceding the (MET-hours/day/year). We also collected information on daily activities such as walking, stair climbing, cycling and household activities. We calculated the metabolic equivalents for each daily activity and the total METs for daily activity. Summary energy expenditure values (MET-hours/day) for these activities were estimated using the following MET values: housework, 2.0 METs; walking, 3.3 METs; stair climbing, 9.0 METs, and cycling 4.0 METs, using a compendium of physical activity values. 27 Details were also obtained on occupational activity of the most recent job among women who were employed at the time of the baseline survey (52.6% of the cohort). Participants were classified into high, medium, or low activity levels using job codes from the different occupations reported in their most recent job. We also collected information on the daily commuting journey to/from work. The daily commuting journey was broken into three categories: (i) using bus or vehicle; (ii) walking or cycling 1 29 min; (iii) walking or cycling more than 30 min. In addition, we calculated the METs-hours/day of total commuting activity. Therefore the study included LPA, daily living physical activity (DPA) measurements for all participants, and among the subset of employed women, we have additional OPA and commuting physical activity (CPA) measurements. Anthropometry All anthropometric measurements, including weight, height, and circumferences of waist and hips, were taken at baseline recruitment according to standard protocol by trained interviewers who were retired medical professionals. 25 From these measurements, the following variables were created: BMI: weight in kilogram divided by the square of height in meters, WHR: waist circumference divided by hip circumference. Dietary intake and other factors Usual dietary intake was assessed through an in-person interview using a validated food frequency questionnaire. 24

PHYSICAL ACTIVITY AND TYPE 2 DIABETES 3of10 The Chinese Food composition tables 28 were used to estimate intake of nutrients and the total energy intake (kcals/day). From the total number of participants that were free of diabetes at baseline (N 5 70 658), we excluded participants that had extreme values for total energy intake (,500 or.3500 kcals/day; N 5 118), 29 leaving 70 540 participants for the analysis. Socio-demographic factors such as age, level of education (none, elementary school, middle/high school, college), family income in yuan/year (,10 000, 10 000 19 999, 20 000 29 999,.30 000), occupation (professional, clerical, farmers/others, housewife/retired), having ever smoked (yes, no) and alcohol intake (ever drank beer, wine or spirits at least three times per week) and a history of hypertension (yes, no) were included in the analyses as potential confounders. We also adjusted for presence of chronic diseases at baseline: coronary heart disease (CHD), stroke, and cancer. Endpoint ascertainment At baseline we identified 70 658 participants who were free of diabetes. Incident diabetes was identified through follow-up surveys. A total of 1973 new cases of diabetes were reported. We considered a diabetes case to be confirmed if participants reported having been diagnosed with diabetes by a physician and met at least one of the following criteria: report a fasting glucose level of at least 7 mmol/l or an oral glucose tolerance test (OGTT) with a value of at least 11.1 mmol/l and/or use of hypoglycaemic medication (i.e. insulin or oral hypoglycaemic drugs). Of the self-reported cases, 1643 participants met the study criteria and are referred to in this report as confirmed cases of diabetes. We conducted analyses with all diabetic participants and among confirmed cases only and found similar results. Thus, we present here results with all cases of diabetes included (N 5 1973). Statistical analysis Person-years for each participant were calculated as the interval between the baseline recruitment to the diagnosis of type 2 diabetes, censored at death or completion of the second follow-up interview. The Cox proportional hazards model was used to estimate the risk of type 2 diabetes by levels of physical activity. Test for linear trend were performed by entering the categorical variables as continuous parameters in the models. In all models we adjusted for potential confounding variables including age, income level, education level, occupation, smoking, and alcohol intake. We adjusted for pre-existing chronic disease (coronary heart disease, stroke, and cancer) and for hypertension to minimize the possibility that changes in physical activity following diagnosis of such conditions would obscure the true association between physical activity and incidence of diabetes. In addition, we repeated the analyses after exclusion of participants with chronic diseases. We present data before adjustment for BMI and WHR as we believe that these variables represent both risk factors and mediating factors on the development of type 2 diabetes. We report results after adjustment for WHR and BMI in the text. Analyses of the association between LPA and DPA and diabetes are presented for all participants and stratified by employment status (yes/no). Interaction between employment status and LPA and DPA was tested by introducing the cross product term of two categorical variables in the model along with the main effect term. Analyses of CPA and OPA are restricted to employed participants. The combined effects of LPA and DPA were evaluated in all participants and stratified by employment status. All analyses were performed using SAS (version 9.1) and all tests of statistical significance were based on two-sided probability. Results During 4.6 years of follow up and a total of 326 625 personyears at risk, we identified 1973 new cases of type 2 diabetes, corresponding to an incidence rate of 6.03 per 1000 personyears. Participants with type 2 diabetes tended to be older, have a higher BMI, WHR, daily energy intake, be less educated, more likely to have a non professional job, and have a history of CHD, hypertension, and stroke than non-diabetic women (Table 1). Participants who engaged in leisure time physical activity were more likely to be older and to have a college education but less likely to be employed. Participation in leisure time activity was also associated with having coronary heart disease, cancer, stroke, and hypertension. We found that diabetes risk was lower among women with high LPA although there was no evidence for a dose response relationship over this narrow range of LPA (Table 2). The relative risks by categories of LPA hours per week were 1.00, 0.89, 1.05, and 0.83. Analysis with additional adjustment for BMI and WHR attenuated the relationship. Stair climbing was inversely associated with the risk of diabetes and this was true both before and after adjustment for BMI and WHR. Cycling was also associated with a reduced risk of type 2 diabetes (RR 5 0.81; 95%CI 0.66 1.00, P 5 0.045). Household activities and walking were unrelated to the risk of diabetes. When we examined the risk of diabetes by the quartiles of the distribution of METs from all daily activities combined we found an inverse association of marginal significance (P 5 0.06) before and after adjustment for BMI and WHR. When total METs (LPA combined with DPA) were examined, the reduced risk was only evident in the higher activity category (RR 5 0.91; 95%CI: 0.79 1.05). There was no evidence of a dose response association. When we repeated the analysis after exclusion of participants with chronic diseases at baseline, we found similar results. In this sub-analysis the association between diabetes and cycling was attenuated and the association between LPA, total METs, and walking and diabetes was accentuated. Table 3 shows the risk of type 2 diabetes by LPA and DPA among non-employed participants. The association between LPA and the risk of type 2 diabetes failed to reach statistical significance in this subgroup. We found an inverse association between daily living activities and the risk of type 2 diabetes with the exception of hours of weekly walking. Total METs from LPA and DPA were inversely associated with the risk of diabetes. The relative risk for total METS for this group were 1.00, 1.06, 1.01, and 0.88, (P for trend,0.01). Further adjustment for BMI and WHR attenuated the associations but overall the associations remained significant or marginally significant. We repeated the analysis after exclusion of

4of10 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 1 Comparison between participants with and without diabetes and leisure time exercise participation (yes/no) by sample population, regarding demographics and other diabetes risk factors a Diabetic status Leisure time exercise Non diabetics All Diabetic Non active Active Variables N 5 68 685 N 5 1973 P-value N 5 45 980 N 5 24 555 P-value Age group (%),0.001,0.001 <9 years 51.09 25.60 60.59 31.32 50 59 years 23.96 30.01 21.50 29.06 >60 years 24.95 44.40 17.91 39.73 BMI 23.85 26.84,0.001 23.77 24.23,0.001 WHR 0.81 0.85,0.001 0.80 0.81,0.001 Kcal /day 1688.19 1688.95 0.93 1678.6 1697.73 0.40 Education (%) None 19.67 37.46,0.001 16.02 27.50,0.001 Elemental 37.63 32.49 41.69 29.59 Middle/high school 28.71 20.27 29.54 26.45 College 13.99 9.78 12.73 16.05 Household Income (%),10 000 15.70 19.72,0.001 15.29 16.73,0.001 10 000 19 999 38.14 38.42 38.72 37.04 20 000 29 999 28.40 25.44 28.63 27.75.30 000 17.76 16.42 17.33 18.47 Occupation (%) Professional 19.32 12.52,0.001 20.60 16.38,0.001 Clerical 12.37 7.65 14.67 7.65 Farmers/others 21.74 13.38 26.53 12.12 Housewife/retired 46.57 66.45 38.20 63.85 Ever smoke (%) 2.63 3.85,0.001 2.24 2.31 0.51 Ever drink (%) 2.27 2.18 0.83 2.00 2.74,0.001 CHD (%) 6.52 14.45,0.001 4.48 10.30,0.001 Stroke (%) 0.87 2.94,0.001 0.66 1.42,0.001 Cancer (%) 1.85 2.13 0.34 1.9 3.10,0.001 Hypertension (%) 21.39 49.376,0.001 18.46 29.11,0.001 a P-values were calculated using chi-square test and t-test in the case of BMI, kcal/day and WHR. participants with chronic diseases at baseline and found similar results. The association between diabetes and cycling was attenuated, while the association between diabetes and walking was accentuated. Table 4 shows associations of LPA, DPA, OPA, and CPA with diabetes among employed participants. Overall we found that occupational activity and METs from DPA were not associated with diabetes risk in employed participants. We found that CPA was inversely related to risk of diabetes in this group (RR for METs categories of CPA were 1.00, 0.88, and 0.70, P for trend,0.01). We also looked at associations between minutes spent in commuting and diabetes risk. In multivariate analyses, the hazard ratios of diabetes with none, 1 29 min/day, and more than 30 min/day walking or cycling to work were 1.00, 0.95, and 0.83 (P for trend 5 0.09), (data not shown in Table 4). In addition, stair climbing was inversely and significantly associated with diabetes risk in the employed participants (P for trend 0.01). Surprisingly we found that the more hours spent in household activity the higher the risk in currently employed participants, and the association remained significant (P for trend,0.01) in fully adjusted analyses. We did cross tabulation between categories of household activities and other physical activity types. We found an inverse association between LPA and CPA and hours of household activity as well as a direct association between hours of household activity and OPA. We did not find that a combination of DPA METs and LPA METs or a combination of DPA METs, LPA METs, and CPA METs was associated with the incidence of type 2 diabetes in participants who were employed. Testing for multiplicative interaction of employment status with LPA and DPA showed a P-value of 0.13 (between employment status and LPA) and a P-value of 0.63 for interaction between work and DPA. We repeated the analysis after exclusion of participants with chronic diseases at baseline and results were unchanged. We looked at the combined effects of LPA and DPA on the risk of type 2 diabetes among all participants and by employment status. Overall we found that participants with high levels of both LPA and DPA had lower risk of diabetes.

PHYSICAL ACTIVITY AND TYPE 2 DIABETES 5of10 Table 2 Physical activity categories and diabetes incidence in all participants All participants N 5 70 540 Chronic disease at baseline excluded N 5 64 130 RR1 95%CI P trend RR2 95%CI P trend Leisure time (LPA) a Hours per week Zero 1.00 0.12 1.00 0.11,1.4 0.89 0.77 1.02 0.87 0.75 1.02 1.5 3.5 1.05 0.92 1.19 0.96 0.82 1.13.3.5 0.83 0.74 0.98 0.89 0.76 1.03 METs- h/d y Zero 1.00 0.24 1.00 0.05,0.80 0.88 0.77 1.02 0.89 0.76 1.03 0.80 1.99 1.01 0.88 1.16 0.99 0.85 1.15.1.99 0.91 0.80 1.04 0.83 0.70 0.97 Daily living (DPA) b Walking min/day,60 1.00 0.16 1.00 0.02 60 90 1.03 0.89 1.18 1.00 0.86 1.17 91 120 1.05 0.91 1.22 1.01 0.86 1.19.120 0.91 0.78 1.06 0.84 0.70 0.99 Cycling No 1.00 0.045 1.00 Yes 0.81 0.66 1.00 0.86 0.69 1.07 0.18 Stairs (flight of stairs/day),5 1.00,0.001 1.00,0.001 5 9 0.90 0.79 1.02 0.88 0.77 1.02 10 15 0.85 0.75 0.95 0.83 0.73 0.95.15 0.79 0.70 0.89 0.76 0.66 0.87 Household h/day,1 1.00 0.99 1.00 0.90 1 2.5 0.96 0.84 1.10 0.94 0.81 1.10 2.6 3 0.93 0.81 1.07 0.96 0.82 1.13.3 0.98 0.85 1.13 0.99 0.84 1.16 METs (DPA),7.85 1.00 0.06 1.00 0.02 7.85 11.26 0.98 0.86 1.12 0.99 0.85 1.15 11.27 15.2 0.95 0.83 1.08 0.92 0.79 1.07.15.2 0.88 0.77 1.01 0.86 0.73 0.99 METs (LPA1DPA),8.20 1.00 0.07 1.00 0.02 8.20 11.82 1.08 0.93 1.24 1.04 0.89 1.21 11.83 16.00 1.06 0.92 1.23 1.02 0.88 1.19.16.00 0.91 0.79 1.05 0.84 0.72 0.99 RR1 adjusted for age, kcal/day, education level, income level, occupation, smoking, alcohol, hypertension, and chronic diseases RR2 adjusted for age, kcal/day, education level, income level, occupation, smoking, alcohol, and hypertension. a LPACategories: METs-h/y d: (none, terciles of the METs-h/y d) ; hours per week: (none, less than 30 min per day, between 1.4 and 3.5 h/week and more than 3.5 h/week). b DPA categories: quartiles. Table 5 shows the combined effect of LPA and DPA in an analysis restricted to participants without chronic disease at baseline. Participants with the highest level of both LPA and DPA were at a lower risk of diabetes than those with low levels of LPA and DPA (RR 5 0.65, 95%CI: 0.49 0.85) in an analysis done including all participants (P for interaction 0.07). In non-employed participants the relative risk for type 2 diabetes for those with high levels of LPA and DPA was

6of10 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 3 Physical activity categories a and diabetes incidence in non employed participants All participants N 5 33 243 Chronic disease at baseline excluded N 5 28 283 RR1 95%CI P trend RR2 95%CI P trend Leisure time (LPA) MET h/d y zero 1.00 0.82 1.00 0.70,0.80 0.90 0.76 107 0.93 0.38 1.12 0.80 1.99 1.01 0.86 1.18 1.00 0.83 1.21.1.99 0.98 0.85 1.12 0.96 0.82 1.13 Daily living (DPA) Walking m/d,60 1.00 0.19 1.00 0.08 60 90 1.09 0.87 1.37 1.07 0.82 1.39 91 120 1.17 0.94 1.46 1.14 0.88 1.48.120 0.95 0.76 1.19 0.89 0.69 1.16 Cycling No 1.00 0.02 1.00 0.13 Yes 0.71 0.53 0.95 0.79 0.58 1.07 Stairs (flights of stairs/day),5 1.00,0.01 1.00,0.01 5 9 0.96 0.83 1.12 0.96 0.80 1.14 10 15 0.87 0.75 1.00 0.85 0.72 1.00.16 0.81 0.70 0.95 0.78 0.66 0.93 Household h/d,1 1.00 0.02 1.00 0.04 1 2.5 0.81 0.68 0.97 0.75 0.60 0.93 2.6 3 0.71 0.59 0.86 0.71 0.57 0.88.3 0.78 0.66 0.93 0.74 0.60 0.91 METs (DPA),7.85 1.00,0.01 1.00,0.01 7.85 11.26 0.88 0.73 1.07 0.89 0.71 1.11 11.27 15.2 0.88 0.74 1.05 0.85 0.69 1.05.15.2 0.79 0.66 0.94 0.76 0.62 0.94 METs (LPA1DPA),8.20 1.00,0.01 1.00,0.01 8.20 11.82 1.06 0.85 1.32 0.99 0.78 1.27 11.83 16.00 1.04 0.84 1.27 0.98 0.78 1.23.16.00 0.88 0.72 1.07 0.80 0.64 1.00 RR1 adjusted for age, kcal/day, education level, income level, smoking, alcohol, hypertension, and chronic diseases. RR2 adjusted for age, kcal/day, education level, income level, smoking, alcohol, and hypertension. a LPA Categories: METs-h/y d: (none, terciles); DPA categories: quartiles. 0.64 (95% CI 0.46 0.89) compared with those with low levels of DPA and LPA (P for interaction 0.06). Conclusions While the association between LPA and type 2 diabetes has been investigated in a number of studies, 4,6,10,11,17,18,30 the influence of other types of physical activity on type 2 diabetes risk has not been systematically described. Lack of assessment of all types of activities could result in an underestimation of the magnitude of association between total physical activity and type 2 diabetes risk. This may be of particular concern in countries, such as China, Denmark, The Netherlands, and Finland where activities related to commuting, daily living, or occupation comprise a considerable proportion of total physical activity 20 or among women for whom housework and active transport are a particular source of energy expenditure. 31 We comprehensively evaluated the association of physical activities from exercise, daily living and occupational physical

PHYSICAL ACTIVITY AND TYPE 2 DIABETES 7of10 Table 4 Physical activity categories a and diabetes incidence in employed participants All participants N 5 37 292 Chronic disease at baseline excluded N 5 35 842 RR1 95%CI P trend RR2 95%CI P trend Leisure time (LPA) MET h/d y zero 1.00 0.09 1.00 0.03,0.80 0.86 0.67 1.10 0.81 0.62 1.06 0.80 1.99 1.09 0.82 1.44 0.90 0.66 1.25.1.99 0.65 0.45 0.95 0.68 0.46 1.01 Daily living (DPA) Walking,60 h/d 1.00 0.45 1.00 0.14 60 90 1.00 0.83 1.20 0.98 0.81 1.19 91 120 0.88 0.70 1.11 0.84 0.66 1.07.120 0.96 0.74 1.25 0.86 0.65 1.15 Cycling No 1.00 0.45 1.00 0.61 Yes 0.92 0.68 1.24 0.92 0.67 1.26 Stairs (flights of stairs/day),5 1.00,0.01 1.00,0.01 5 9 0.79 0.63 0.99 0.78 0.61 0.99 10 15 0.79 0.64 0.98 0.79 0.63 0.99.16 0.72 0.59 0.89 0.70 0.56 0.87 Household,1 h/d 1.00,0.01 1.00 0.01 1 2.5 1.16 0.94 1.42 1.13 0.91 1.41 2 3 1.31 1.05 1.64 1.29 1.02 1.64.3 1.34 1.04 1.74 1.36 1.04 1.78 METs (DPA),7.85 1.00 0.91 1.00 0.60 7.85 11.26 1.07 0.88 1.29 1.06 0.87 1.30 11.27 15.2 0.97 0.78 1.20 0.92 0.73 1.15.15.2 1.05 0.82 1.34 0.97 0.75 1.27 Commuting (CPA),2.61 METs 1.00,0.001 1.00,0.001 2.6 5.5 0.88 0.90 1.29 0.80 0.65 0.98.5.5 0.70 0.50 1.41 0.67 0.55 0.82 Occupational (OPA) Light 1.00 0.58 1.00 0.86 Moderate 1.11 0.92 1.34 1.07 0.89 1.30 High 0.86 0.51 1.46 0.81 0.48 1.39 METs (LPA 1 DPA),8.20 1.00 0.96 1.00 0.57 8.20 11.82 1.08 0.89 1.31 1.06 0.87 1.29 11.83 16.00 1.07 0.87 1.32 1.02 0.82 1.28.16.00 0.96 0.74 1.23 0.88 0.67 1.17 METs (LPA 1 DPA 1 CPA ),9.18 1.00 0.73 1.00 0.24 9.18 11.89 1.09 0.89 1.33 1.11 0.90 1.36 11.90 15.49 1.03 0.83 1.28 0.99 0.78 1.24.15.50 0.96 0.75 1.23 0.91 0.91 1.18 RR1 adjusted for age, kcal/day, education level, income level, smoking, alcohol, hypertension, and chronic diseases. RR2 adjusted for age, kcal/day, education level, income level, smoking, alcohol, and hypertension. a Categories of LPA: terciles and for DPA are quartiles.

8of10 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 5 Combined effect of DPA and leisure time physical activity in participants with no chronic disease at baseline and stratified by employment status (yes/no) Leisure time physical activity Daily living physical activity None Medium,2.39 High >2.39 All participants a Low,8.85 1.00 0.90 (0.72 1.12) 1.22 (0.90 1.66) Medium 8.85 13.5 0.96 (0.80 1.11) 0.92 (0.75 1.13) 0.79 (0.60 1.05) High.13.5 0.94 (0.72 1.12) 0.87 (0.71 1.06) 0.65 (0.49 0.85) P interaction 0.07 Non-employed participants b Low,8.85 1.00 0.87 (0.61 1.24) 1.36 (0.92 1.99) Medium 8.85 13.5 0.94 (0.73 1.23) 0.94 (0.71 1.24) 0.88 (0.63 1.22) High.13.5 0.87 (0.68 1.11) 0.89 (0.69 1.16) 0.64 (0.46 0.89) P interaction 0.06 Employed participants b Low,8.85 1.00 0.91 (0.68 1.21) 0.94 (0.53 1.69) Medium 8.85 13.5 0.92 (0.77 1.18) 0.86 (0.60 1.23) 0.30 (0.10 0.95) High.13.5 1.10 (0.87 1.39) 0.63 (0.38 1.05) 0.75 (0.31 1.83) P interaction 0.21 a Adjusted for age, kcal/day, smoking, alcohol intake, household income, education level, occupation status and hypertension. b Adjusted for age, kcal/day, smoking, alcohol intake, household income, education level and hypertension. activities in a population-based cohort study. We found that DPA was associated with lower diabetes risk only in nonemployed participants, while LPA was associated with a reduced diabetes risk only in currently employed participants. Our results highlight the need to study middle-aged women stratified by employment status, because activity patterns may vary between employed and non-employed women. Data on associations between OPA and type 2 diabetes risk are scarce. 20 OPA was associated with a lower risk of type 2 diabetes in Pima Indians 32 and in a Finnish population. 20 We did not find that OPA was associated with the risk of type 2 diabetes. Physical activity has declined among urban residents of China, 33 and changes in activity patterns have been towards increasing sedentary occupations and lifestyles. 34 In our population only 3.11% of women reported jobs that require high activity, which may explain, at least in part, the null association found in our study. Walking and cycling (other than to/from work) was only modestly associated with the risk of diabetes in our study. Walking pace was inversely associated with the risk of type 2 diabetes in a large cohort of women in the USA. 35 Most walking in the US is a leisure type of activity while in China most walking is part of daily living. Thus our study may lack sufficient variability on walking to detect a possible association between walking and diabetes risk. Similarly, cycling (other than to/from work) was only modestly associated with a lower risk of type 2 diabetes. Stair climbing was inversely associated with a lower risk of type 2 diabetes and there was a dose response relationship. Household activity was associated with a lower risk of diabetes in non-employed participants and with a higher risk of in currently employed participants. Participants in the higher household activity category were less likely to participate in LPA or commute to work, most likely reflecting time constrains. We repeated the analysis adjusted for LPA and CPA and the association was attenuated and no longer significant. The relative risk of housework hours categories were 1.00, 1.26, 1.22, and 0.97 (P trend 5 0.86) and 1.00, 1.15, 1.00, and 0.82 (P trend 5 0.60) before and after exclusion of participants with chronic disease at baseline respectively. CPA was inversely associated in this population. The fully adjusted relative risk of diabetes with none, less than 30 and over 30 min were 1.00, 0.97, and 0.79, P for trend 5 0.02 (data not shown in table). Our results are similar to those found in another study from Finland (the hazard ratios for type 2 diabetes with none, 1 29 min/day and more than 30 min/day walking or cycling to work were 1.00, 0.96, and 0.64, respectively; P for trend 5 0.001). 20 These results are of great importance for this population in which active transportation for work is an important component of energy expenditure. A recent report stated that acquiring motor vehicles in China was accompanied by an increase in obesity. 36 Obesity and overweight has become a major public health concern in China. 37 The prevalence of overweight and obesity doubled in females and tripled in males from 1989 to 1997, with the rate of increase for overweight on an annualized basis being over 1% for men and women, 34 accelerated from 0.4% per year increase during the early 1980 s. 38 Both, an increase in obesity and a more sedentary lifestyle would be followed by an increase in type 2 diabetes. Therefore it will be important to encourage people to continue to walk or cycle to work, regardless of whether they have a motor vehicle or no. A protective effect of physical activity is biologically plausible. 39 Skeletal muscle is the predominant site for insulin resistance 40 and exercise training has been shown to improve insulin sensitivity in these tissues. 39 Exercise training improves insulin sensitivity via increased oxidative enzymes, glucose transporters (GLUT4), and capillarity in muscle, as well as by reducing abdominal fat. 41 Studies evaluating the effect of

PHYSICAL ACTIVITY AND TYPE 2 DIABETES 9of10 physical activity on insulin sensitivity in general have reported a favourable effect of physical activity. 12,15,42 This study has several strengths, including the high quality of the data collected and the large sample size available for analysis that has allowed us to precisely assess risk with adjustment for cofounders for many demographic and socioeconomic factors as well as to evaluate potentially important interactions. In addition, this cohort has very high follow up rate. The physical activity assessment instrument was specifically designed to a assess wide range of activities and was validated in the study population. 26 Reliance on self-reported diabetes is the major limitation of the study. Misclassification of diabetes could weaken the association between physical activity and the risk of type 2 diabetes. To address the possibility of surveillance bias we conducted analyses restricted to women with at least one recorded symptom of diabetes and found similar results (data not shown). Measurement error in the physical activity questionnaire may introduce biases in the study. Our validation study with the Shanghai Women s Health Study suggests that the questionnaire that we used in the study has good reliability and validity, although the validity coefficients for household and daily living activities were lower than that for commuting and exercise/leisure-time activities. 26 Measurement errors in physical activity that are not systematically related to our outcome would be expected to have attenuated the relative risks of diabetes in this study. In conclusion our study has provided evidence that not only LPA but also other types of activity such as DPA and CPA are associated with a reduction of type 2 diabetes risk. The principle implications of our findings are in highlighting the importance KEY MESSAGES of physical activity for the prevention of type 2 diabetes in China and possibly in other developing countries. In the past 2 decades, China has experienced a rapid and substantial decline of physical activity levels as a result of economic development, enhanced food supply, expansion of television, computerization, and mechanization, more prevalent car ownership, and increased public transportation. 36 In parallel with decreasing levels of physical activity, the prevalence of overweight and obesity has increased significantly in China. 43 As a consequence, diabetes mellitus has become a major public health issue in China. The prevalence of type 2 diabetes increased from 1.9 to 5.6% between 1993 and 2003 in China. 44 The observed temporal trends for physical inactivity and adiposity are expected to further increase the incidence rates of diabetes in China in the years to come. 44 Thus, promoting an active lifestyle or regular exercise should receive the highest public health priority. Acknowledgements This work was supported by USPHS Grant RO1 CA 70867 from the National Cancer Institute. R.V., G.Y., C.M., M.L., W.Z., and X.O.S. drafted and provided critical revision of the manuscript; R.V., G.Y., H.C., and X.O.S. analyzed and interpreted the data; G.Y., H.L., Q.L., and Y.G., X.O.S., and W.Z. were responsible for implementation of the study and acquisition of the data; W.Z. and X.O.S. conceived of and designed the study. X.O.S. will act as guarantor for the paper. R.V. has checked the references for accuracy and completeness. Conflict of Interest: The authors have no conflicts of interest to declare. Our study shown that increased participation in a broad range of activities provides protection against type 2 diabetes, providing strong support for current promotion campaigns that encourage increasing activity levels through exercise, non-exercise, or a combination of both. References 1 Zimmet P. Globalization, coca-colonization and the chronic disease epidemic: can the Doomsday scenario be averted? J Intern Med 2000;247:301 10. 2 Hu FB, Manson JE, Stampfer MJ et al. Diet, lifestyle, and the risk of type 2 diabetes mellitus in women. N Engl J Med 2001;345:790 97. 3 Marshall JA, Bessesen DH, Hamman RF. High saturated fat and low starch and fibre are associated with hyperinsulinaemia in a non-diabetic population: the San Luis Valley Diabetes Study. Diabetologia 1997;40:430 38. 4 Burchfiel CM, Sharp DS, Curb JD et al. Physical activity and incidence of diabetes: the Honolulu Heart Program. Am J Epidemiol 1995;141:360 68. 5 Eriksson KF, Lindgarde F. Prevention of type 2 (non-insulindependent) diabetes mellitus by diet and physical exercise. The 6-year Malmo feasibility study. Diabetologia 1991;34:891 98. 6 Helmrich SP, Ragland DR, Leung RW, Paffenbarger RS Jr. Physical activity and reduced occurrence of non-insulin-dependent diabetes mellitus. N Engl J Med 1991;325:147 52. 7 Hu FB, Leitzmann MF, Stampfer MJ, Colditz GA, Willett WC, Rimm EB. Physical activity and television watching in relation to risk for type 2 diabetes mellitus in men. Arch Intern Med 2001;161:1542 48. 8 Irwin ML, Mayer-Davis EJ, Addy CL et al. Moderate-intensity physical activity and fasting insulin levels in women: the Cross- Cultural Activity Participation Study. Diabetes Care 2000;23:449 54. 9 Kriska AM, Hanley AJ, Harris SB, Zinman B. Physical activity, physical fitness, and insulin and glucose concentrations in an isolated Native Canadian population experiencing rapid lifestyle change. Diabetes Care 2001;24:1787 92. 10 Manson JE, Rimm EB, Stampfer MJ et al. Physical activity and incidence of non-insulin-dependent diabetes mellitus in women. Lancet 1991;338:774 78. 11 Manson JE, Nathan DM, Krolewski AS, Stampfer MJ, Willett WC, Hennekens CH. A prospective study of exercise and incidence of diabetes among US male physicians. JAMA 1992;268:63 67. 12 Mayer-Davis EJ, D Agostino R, Jr, Karter AJ et al. Intensity and amount of physical activity in relation to insulin sensitivity: the Insulin Resistance Atherosclerosis Study. JAMA 1998;279:669 74.

10 of 10 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY 13 Pan X, Li G, Hu Y. Effect of dietary and/or exercise intervention on incidence of diabetes in 530 subjects with impaired glucose tolerance from 1986-1992. Zhonghua Nei Ke Za Zhi 1995;34:108 12. 14 Perry IJ, Wannamethee SG, Walker MK, Thomson AG, Whincup PH, Shaper AG. Prospective study of risk factors for development of non-insulin dependent diabetes in middle aged British men. BMJ 1995;310:560 64. 15 Regensteiner JG, Mayer EJ, Shetterly SM et al. Relationship between habitual physical activity and insulin levels among nondiabetic men and women. San Luis Valley Diabetes Study. Diabetes Care 1991;14:1066 74. 16 Sigal RJ, Kenny GP, Wasserman DH, Castaneda-Sceppa C. Physical activity/exercise and type 2 diabetes. Diabetes Care 2004;27:2518 39. 17 Wannamethee SG, Shaper AG, Alberti KG. Physical activity, metabolic factors, and the incidence of coronary heart disease and type 2 diabetes. Arch Intern Med 2000;160:2108 16. 18 Wei M, Gibbons LW, Mitchell TL, Kampert JB, Lee CD, Blair SN. The association between cardiorespiratory fitness and impaired fasting glucose and type 2 diabetes mellitus in men. Ann Intern Med 1999;130:89 96. 19 Hu FB, Li TY, Colditz GA, Willett WC, Manson JE. Television watching and other sedentary behaviors in relation to risk of obesity and type 2 diabetes mellitus in women. JAMA 2003;289:1785 1791. 20 Hu G, Qiao Q, Silventoinen K et al. Occupational, commuting, and leisure-time physical activity in relation to risk for Type 2 diabetes in middle-aged Finnish men and women. Diabetologia 2003;46:322 329. 21 Cockram CS. Diabetes mellitus: perspective from the Asia-Pacific region. Diabetes Res Clin Pract 2000;50(Suppl 2):3 7. 22 Gu D, Reynolds K, Duan X et al. Prevalence of diabetes and impaired fasting glucose in the Chinese adult population: International Collaborative Study of Cardiovascular Disease in Asia (InterASIA). Diabetologia 2003;46:1190 98. 23 Zhong XL. Diabetes mellitus survey in China. Chin Med J (Engl) 1982;95:423 30. 24 Shu XO, Yang G, Jin F et al. Validity and reproducibility of the food frequency questionnaire used in the Shanghai Women s Health Study. Eur J Clin Nutr 2004;58:17 23. 25 Zhang X, Shu XO, Gao YT et al. Anthropometric predictors of coronary heart disease in Chinese women. Int J Obes Relat Metab Disord 2004;28:734 40. 26 Matthews CE, Shu XO, Yang G et al. Reproducibility and validity of the Shanghai Women s Health Study physical activity questionnaire. Am J Epidemiol 2003;158:1114 22. 27 Ainsworth BE, Haskell WL, Whitt MC et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc 2000;32(9 Suppl):S498 S504. 28 Wang GY, Shen ZP. Chinese Food Composition Tables. Beijing: People s Health Publishing House 1991. 29 Joshipura KJ, Hu FB, Manson JE et al. The effect of fruit and vegetable intake on risk for coronary heart disease. Ann Intern Med 2001;134:1106 14. 30 Todoroki I, Shinchi K, Kono S, Imanishi K. Lifestyle and glucose tolerance: a cross-sectional study of Japanese men. Ann Epidemiol 1994;4:363 68. 31 He XZ, Baker DW. Differences in leisure-time, household, and work-related physical activity by race, ethnicity, and education. JGen Intern Med 2005;20:259 66. 32 Kriska AM, Saremi A, Hanson RL et al. Physical activity, obesity, and the incidence of type 2 diabetes in a high-risk population. Am J Epidemiol 2003;158:669 75. 33 Popkin BM. The nutrition transition in low-income countries: an emerging crisis. Nutr Rev 1994;52:285 98. 34 Bell AC, Ge K, Popkin BM. Weight gain and its predictors in Chinese adults. Int J Obes Relat Metab Disord 2001;25:1079 86. 35 Hu FB, Sigal RJ, Rich-Edwards JW et al. Walking compared with vigorous physical activity and risk of type 2 diabetes in women: a prospective study. JAMA 1999;282:1433 39. 36 Bell AC, Ge K, Popkin BM. The road to obesity or the path to prevention: motorized transportation and obesity in China. Obes Res 2002;10:277 83. 37 Popkin BM. The nutrition transition and its health implications in lower-income countries. Public Health Nutr 1998;1:5 21. 38 Ge K, Weisell R, Guo X et al. The body mass index of Chinese adults in the 1980s. Eur J Clin Nutr 1994;48(Suppl 3):S148 S154. 39 Spelberg A MJE. Towards prevention of non-insulin dependent diabetes mellitus. In: Leslie RDG (ed). Causes of Diabetes. Chichester: John Wiley & Sons, 1993, pp. 319 45. 40 DeFronzo RA, Bonadonna RC, Ferrannini E. Pathogenesis of NIDDM. A balanced overview. Diabetes Care 1992;15:318 68. 41 Chisholm DJ, Campbell LV, Kraegen EW. Pathogenesis of the insulin resistance syndrome (syndrome X). Clin Exp Pharmacol Physiol 1997;24:782 84. 42 Kriska AM, Pereira MA, Hanson RL et al. Association of physical activity and serum insulin concentrations in two populations at high risk for type 2 diabetes but differing by BMI. Diabetes Care 2001;24:1175 80. 43 Yu Z, Song G, Guo Z et al. Changes in blood pressure, body mass index, and salt consumption in a Chinese population. Prev Med 1999;29:165 72. 44 Wang Y, Mi J, Shan XY, Wang QJ, Ge KY. Is China facing an obesity epidemic and the consequences? Epub ahead of print May 2, 2006. The trends in obesity and chronic disease in China. Int J Obes (Lond) 2006.