669 Am J Epidemiol 2003;158:669–675American Journal of EpidemiologyCopyright © 2003 by the Johns Hopkins Bloomberg School of Public HealthAll rights reservedVol. 158, No. 7Printed in U.S.A.DOI: 10.1093/aje/kwg191Physical Activity, Obesity, and the Incidence of Type 2 Diabetes in a High-Risk PopulationAndrea M. Kriska1, Aramesh Saremi2, Robert L. Hanson2, Peter H. Bennett2, Sayuko Kobes2, Desmond E. Williams2, and William C. Knowler21 Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, PA. 2 Phoenix Epidemiology and Clinical Research Branch, National Institute of Diabetes and Digestive and Kidney Diseases, Phoenix, AZ.Received for publication October 23, 2002; accepted for publication April 7, 2003.This study, examining the longitudinal relation among physical activity, body mass index, and development oftype 2 diabetes in a high-risk population, is unique because diabetes was determined by oral glucose tolerancetesting rather than by self-report. A physical activity questionnaire assessing past year leisure and occupationalactivity was administered to 1,728 nondiabetic Pima individuals aged 15–59 years as part of a series of clinicexaminations in the Gila River Indian Community from 1987 to 2000. During an average follow-up period of 6years, 346 subjects developed diabetes. Using time-dependent Cox proportional hazards modeling adjusting forage, the authors found that total activity was related to diabetes incidence in women and men (p < 0.05 in womenonly). After additional adjustment for body mass index, the relation between activity and diabetes incidence wasweakened in both men and women. When the age-adjusted diabetes incidence rates were examined by levelsof activity stratified by tertile of body mass index, the diabetes incidence rate remained lower in more active thanin less active men and women from all body mass index groups, with the exception of the middle body mass indextertile in men (p < 0.05 in women only). These results suggest that the adoption and maintenance of a physicallyactive lifestyle can play a significant role in preventing type 2 diabetes.diabetes mellitus; exercise; incidence; motor activity; obesity Abbreviation: MET, metabolic equivalent.It has been suggested in a variety of observational andexperimental epidemiologic studies that physical activitymay play a significant role in the prevention of type 2diabetes mellitus. Recent findings of the clinical trials in thisarea provide the most convincing evidence that physicalactivity, in conjunction with diet and weight loss, canprevent diabetes in a variety of populations and age groups(1–3). Specifically, studies in men and women with impairedglucose tolerance at baseline from a variety of racial andethnic backgrounds in the United States, along with men andwomen from China and Finland, demonstrated a decrease inthe incidence of type 2 diabetes as the result of interventionsthat included physical activity (1–3). However, with theexception of the Chinese study, which was randomized byclinic, all of these intervention trials combined physicalactivity with weight loss and diet in their interventionscheme. In other words, the independent effect of physicalactivity intervention was not tested directly.It is likely that physical activity can play an independentrole in the prevention of type 2 diabetes separately from itseffect on weight loss and body composition. Exercisetraining studies have supported the contention that physicalactivity improves insulin sensitivity independently of anyeffect of activity on weight loss and fat distribution (4). Like-wise, in a recent cross-sectional population study, physicalactivity was shown to be negatively associated with insulinconcentrations in two populations at high risk for diabetesthat differed greatly by body mass index (5). The fact that therelation between physical activity and insulin sensitivity wassimilar in both the Pima Indians, among whom the preva-lence of obesity is quite high, and the Mauritians, who areleaner, suggests a beneficial role of physical activity onCorrespondence to Dr. Andrea M. Kriska, Room 505B, Parran Hall, Graduate School of Public Health, University of Pittsburgh, 130 DeSoto Street, Pittsburgh, PA 15261 (e-mail: [email protected]).670 Kriska et al. Am J Epidemiol 2003;158:669–675insulin sensitivity, and therefore perhaps even on the preven-tion of diabetes, that is separate from any influence of phys-ical activity on body composition (5).Pima Indians of the Gila River Indian Community ofArizona have one of the world’s highest documented inci-dence rates of type 2 diabetes mellitus, and they have a highprevalence of obesity (6, 7). Physical activity levels havebeen assessed in this population using the original version ofthe same activity questionnaire that is being used in anumber of epidemiologic studies of diabetes worldwide (8,9). The present study examined the role of physical activityand obesity in the development of type 2 diabetes in PimaIndians. The information gained from examining the longitu-dinal relation among physical activity, body mass index, anddevelopment of diabetes in this high-risk population isunique, because diabetes is determined objectively by oralglucose tolerance testing rather than by subjective reportingof the clinical diagnosis of diabetes.MATERIALS AND METHODSStudy descriptionPima Indians of the Gila River Indian Community ofArizona have participated in a longitudinal population-baseddiabetes research study conducted by the National Instituteof Diabetes and Digestive and Kidney Diseases since 1965.Individuals over the age of 5 years who currently live in adesignated part of the community are invited to participate.At intervals of approximately 2 years, each subject isinvited for a comprehensive examination conducted at thestudy clinic located in the community (6, 7). At each exami-nation, a 75-g oral glucose tolerance test is performed, inwhich venous serum insulin and plasma glucose concentra-tions are determined after an overnight fast and 2 hours post-load. Diabetes is diagnosed if the 2-hour postload plasmaglucose concentration is at least 11.1 mmol/liter (200 mg/dl)at this examination or if a diagnosis is made during thecourse of routine medical care (7, 10). The examination alsoincludes a medical history, physical examination, andmeasurement of height and weight (with light indoorclothing but without shoes). Obesity is estimated by the