Online supplemental material

Leisure-time physical activity and risk of type 2 diabetes: a meta-analysis of prospective cohort studies

Short title: Physical activity and type 2 diabetes

Pengcheng Huai1, Huiju Han1, Kathleen Heather Reilly2, Xiaolei Guo3, Jiyu Zhang3,

Aiqiang Xu1, 3

From the 1Department of Epidemiology, School of Public Health, Shandong

University, Jinan, China; the 2Independent Consultant, New York, New York; the

3Shandong Center for Disease Control and Prevention, Jinan, China.

Total number of tables: 3

Total number of figures: 1

Correspondence to Aiqiang Xu, Shandong Center for Disease Control and Prevention,

No. 16992, Jingshi Road, Jinan 250014, China. Phone: +86-531-82679606. Fax: +86-

531-82679620. E-mail: [email protected].

1 Table S1 Characteristics of included prospective cohort studies on leisure-time physical activity (LTPA) and type 2 diabetes

First Country Follow-up Population No. of LTPA (measurement mode) Adjustments author, duration cases year Waller, Finland 28 years 20487 men and 1,082 LTPA index based on frequency, BMI 2010 women, age duration and intensity of 20-50 years activities, expressed as MET hours per day. Joseph, Norway 10.8 years 12431 men and 522 LTPA based on intensity and Age, BMI, total cholesterol, triglycerides, 2010 13737 women, average time of activities. HDL cholesterol, hypertension, family age 25-98 history, education, and smoking. years Rana, USA 16 years 68907 women, 4,030 LTPA based on average time of Age, BMI, smoking status, alcohol 2007 age 30-55 moderate-to-vigorous activities, consumption, menopause status, years expressed as MET hours per postmenopausal hormone use, and family week. history of diabetes. Hsia, USA 5.1 years 87907 women, 2,271 LTPA scores based on frequency, Age, BMI, alcohol, education, smoking, 2005 age 53-71 duration and intensity of hypertension, hypercholesterolemia, dietary years activities, expressed as MET fiber , percent energy from carbohydrate hours per week. Weinstein, USA 6.9 years 37878 1,361 LTPA scores based on average Age, family history of diabetes, alcohol use, 2004 women, age time and intensity of activities, smoking status, hormone therapy use, 47-62 years expressed as energy expenditure hypertension, high cholesterol, dietary (kcal) per week. factors, randomized Woman's Health Study treatment groups and BMI.

2 Hu G, Finland 12 years 6898 men and 373 LTPA based on average time and Age, study year, systolic blood pressure, 2003 7392 women, intensity of activities. smoking status, education, commuting and age 35-64 occupational physical activity,BMI and sex. years Hu FB, USA 10 years 37918 men, 1,058 LTPA based on average time of Age, pack-years of smoking, parental family 2001 age 40-75 activities, expressed as MET- history of diabetes, alcohol intake, vitamin years hours per week. E supplement use, and BMI. Haapanen, Finland 10 years 1340 men and 118 LTPA based on frequency and Age, smoking 1997 1500 women, duration of activities, expressed age 35-63 as energy expenditure (kcal) per years week. LTPA, leisure-time physical activity; MET, metabolic equivalent; BMI, body mass index; HDL, high density lipoprotein .

3 Table S2 Other characteristics of included prospective cohort studies on leisure-time physical activity (LTPA) and type 2 diabetes First author, HR(95%CI) or Statistic used to Type 2 diabetes Definition of PA levels year RR(95%CI) * estimate HR or measurement

RR (method, device used) Waller, Moderate LTPA: Cox Death certificates Low LTPA: ﹤ 0.59 MET h/day. 2010 0.50(0.32-0.78) proportional Moderate LTPA: 1.30-2.49 MET h/day. High LTPA: hazards High LTPA: ≥ 4.50 MET h/day. 0.61(0.40-0.94) regression Joseph, High LTPA: Cox Self-reported diabetes, Low LTPA: less than 3 h/wk of light activity without 2010 0.70(0.51-0.96) proportional health surveys, or hospital sweating or dyspepsia. hazards models digital patient records. High LTPA: hard activity with sweating or becoming out of breath for ≥3 h/wk. Rana, High LTPA: Cox Self-reported physician- Low LTPA: < 2.1 MET h/wk. 2007 0.60(0.55-0.67) proportional diagnosed diabetes High LTPA: ≥ 21.8 MET h/wk. hazards regression Hsia, Moderate LTPA: Multivariate Self-reported use of Low LTPA: 0-2.3 MET h/wk. 2005 0.80(0.70-0.91) Cox antidiabetic medication Moderate LTPA: 7.5-13.9 MET h/wk. High LTPA: proportional High LTPA:23.5-143.0 MET h/wk. 0.78(0.67-0.91) hazard models Weinstein, Moderate LTPA: Cox Self-reported physician- Low LTPA: 0-199 kcal/wk. 2004 0.89(0.80-0.99) proportional diagnosed diabetes Moderate LTPA: 200-1499 kcal/wk. High LTPA: hazards models High LTPA: more than 1500 kcal/wk. 0.82(0.70-0.97) Hu G, High LTPA: Cox National registries Low LTPA: almost completely inactive, e.g. reading, 2003 0.84(0.57-1.25) proportional watching TV, or doing some minor physical activity but not Moderate LTPA: hazards models of moderate or high level.

4 0.81(0.64-1.02) High LTPA: performing vigorous physical activity more than 3 h a week. Hu FB, Moderate LTPA: Pooled logistic Self-reported physician- Low LTPA: 0-5.9 MET h/wk. 2001 0.72(0.60-0.86) regression diagnosed diabetes Moderate LTPA: 13.8-24.2 MET h/wk. High LTPA: High LTPA: ≥ 40.9 MET h/wk. 0.62(0.50-0.76) Haapanen, High LTPA: Cox Self-reported diabetes For men: Low LTPA: 0-1100 kcal/wk. 1997 0.52(0.32-0.83) proportional High LTPA: more than 1900 kcal/wk. hazards models For women: Low LTPA: 0-900 kcal/wk. High LTPA: more than 1500 kcal/wk. LTPA, leisure-time physical activity; MET, metabolic equivalent. * Low level of LTPA is considered as referent.

5 Table S3 Quality assessment of included studies according to Newcastle-Ottawa Scale. First Newcastle-Ottawa Scale items for cohort studies author, Representati Selection of Ascertainment Demonstration That Comparability of Assessment Was Follow-Up Adequacy Total year -veness of the Non- of Exposure Outcome of Interest Cohorts on the of Long Enough for of Follow stars the Exposed Exposed Was Not Present at Basis of the Outcome Outcomes to Occur Up of Cohort Cohort Start of Study Design or Analysis Cohorts Waller, 1 1 1 1 2 1 1 0 8 2010 Joseph, 1 1 0 1 2 1 1 1 8 2010 Rana, 0 1 0 1 2 0 1 1 6 2007 Hsia, 0 1 1 0 2 0 1 1 6 2005 Weinstein 0 1 0 1 2 1 1 1 7 , 2004 Hu G, 1 1 1 1 2 1 1 1 9 2003 Hu FB, 0 1 1 1 2 1 1 0 7 2001 Haapanen 1 1 1 0 1 0 1 0 5 , 1997

6 Table S4 Boundary of low-level LTPA and high-level LTPA*

UNIT Low-level LTPA High-level LTPA MET h/week <6 MET h/week >21 MET h/week kcal/week <1000 kcal/week >1500 kcal/week h/week† <3 h/week >3 h/week *Boundary of low-level LTPA and high-level LTPA were summarized based on included studies. † Two included studies used h/week as LTPA unit and reported only two levels of LTPA.

7 1789 citations identified in literature search

1710 excluded according to title and abstract

78 full-text given more detailed assessment

22 excluded 3 additional from 16 reviews hand-searching 6 duplicates

59 considered for inclusion criteria

13 excluded for not 33 excluded taking low level of 7 didn’t meet (1) (4) Inclusion criteria; LTPA as reference 16 didn’t meet (2) Inclusion criteria; 9 didn’t meet (3) (6) Inclusion criteria; 5 excluded for overlap 1 didn’t meet (5) Inclusion criteria. definition of LTPA with other studies

8 articles for final meta-analysis

Figure S1 Flow chart of study selection.

8