Cardiovascular and Metabolic Risk ORIGINAL ARTICLE

Physical Activity and Metabolic Risk in Individuals With a Family History of Type 2

ULF EKELUND, PHD NICHOLAS J. WAREHAM, MD, PHD idemiological studies because of their SIMON J. GRIFFIN, DM ON BEHALF OF THE PROACTIVE RESEARCH complex and latent nature (7). Previous GROUP researchers have primarily relied on self- reported when examin- ing associations between physical activity OBJECTIVE — We sought to examine the independent associations between different di- and clustered metabolic risk and its sub- mensions of physical activity with intermediary and clustered metabolic risk factors in over- components (8–14). We have recently weight individuals with an increased risk of type 2 diabetes to inform future preventive action. suggested that objectively measured physical activity energy expenditure RESEARCH DESIGN AND METHODS — We measured total body movement and five (PAEE) is independently associated with other subcomponents of physical activity by accelerometry in 258 adults (aged 30–50 years) clustered metabolic risk in a healthy pop- with a family history of type 2 diabetes. We estimated aerobic fitness from an incremental treadmill test. We measured body composition by bioimpedance and waist circumfer- ulation (15) and that PAEE predicts pro- ence, blood pressure, fasting triglycerides, HDL cholesterol, glucose, and insulin with standard gression to the methods. We constructed a standardized continuously distributed variable for clustered risk. independent of aerobic fitness and body fatness (16). However, it was unclear RESULTS — Total body movement (counts dayϪ1) was significantly and independently from these studies which, if any, of the associated with three of six risk factors (fasting triglycerides, insulin, and HDL) and with subdimensions of physical activity exert a clustered metabolic risk (P ϭ 0.004) after adjustment for age, sex, and . Time spent greater influence on individual and clus- at moderate- and vigorous-intensity physical activity (MPVA) was independently associated tered metabolic risk factors. Given that ϭ with clustered metabolic risk (P 0.03). Five- and 10-min bouts of MVPA, time spent strategies to promote distinct types of sedentary, time spent at light-intensity activity, and aerobic fitness were not significantly physical activity will differ, it is important related with clustered risk after adjustment for confounding factors. to understand the dimensions of activity CONCLUSIONS — Total body movement is associated with intermediary phenotypic risk most strongly associated with a particular factors for cardiovascular and metabolic disease and with clustered metabolic risk inde- outcome when one is designing preven- pendent of aerobic fitness and obesity. Increasing the total amount of physical activity in sed- tive interventions for high-risk groups. entary and individuals may have beneficial effects on metabolic risk factors. Targeting the wrong dimensions of phys- ical activity may limit the potential bene- Diabetes Care 30:337–342, 2007 fits predicted by observational studies. The purpose of the present study was he metabolic syndrome is loosely understand how intermediary pheno- to examine the magnitude and directions defined as a cluster of cardiovascular typic metabolic risk factors are influenced of associations between total body move- T disease (CVD) risk factors, includ- by potentially modifiable lifestyle behav- ment; accumulated time spent sedentary, ing disturbed insulin and glucose metab- iors, such as physical activity. at light-intensity activity, and at moderate- olism, , abdominal obesity, Total body movement (i.e., the total and vigorous-intensity physical activity and dyslipidemia. This syndrome pre- amount of physical activity) and patterns (MVPA); and bouts of MVPA with estab- dicts the development of type 2 diabetes, of activity (i.e., time spent sedentary and lished metabolic risk factors to inform CVD, and all-cause mortality (1,2). Low at various intensity levels and bouts of future preventive action. This cross- levels of physical activity are believed to sustained activity) are separate dimen- sectional study was conducted in healthy be an important determinant of this clus- sions of activity, which may be associated men and women at high risk of develop- ter of metabolic risk factors. Given the with intermediary phenotypes of meta- ing type 2 diabetes because of a family global increase in the prevalence of obe- bolic risk in different ways. However, history of the disease. sity, type 2 diabetes, and the metabolic these subdimensions of activity are inher- syndrome (3–6), there is a need to further ently difficult to measure precisely in ep- RESEARCH DESIGN AND ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● METHODS — Individuals in the From the Medical Research Council Epidemiology Unit, Cambridge, U.K. study population were all participants in Address correspondence and reprint requests to Ulf Ekelund, MRC Epidemiology Unit, Elsie Widdowson the ProActive Study, a randomized, con- Laboratory, Fulbourn Road, CB1 9NL, Cambridge, U.K. E-mail: [email protected]. trolled trial examining the efficacy of a Received for publication 8 September 2006 and accepted in revised form 31 October 2006. Abbreviations: CVD, ; FFM, fat-free mass; MVPA, moderate- and vigorous- family-based intervention to increase intensity physical activity; PAEE, physical activity energy expenditure. physical activity among individuals de- A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion fined as high-risk through having a paren- factors for many substances. tal history of type 2 diabetes. The DOI: 10.2337/dc06-1883 selection procedures, screening, recruit- © 2007 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby ment, inclusion criteria, randomization, marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. design, and methods have been described

DIABETES CARE, VOLUME 30, NUMBER 2, FEBRUARY 2007 337 Activity and metabolic risk in diabetes previously (17). Briefly, potential partici- measured using the hexokinase method, merly known as the CSA activity monitor) pants were identified via diabetes regis- and plasma triglycerides and HDL choles- model WAM 7164 (Manufacturing Tech- ters and medical records of family history terol were measured with standard enzy- nology, Fort Walton Beach, FL) acceler- in 20 general practices in East Anglia. To matic methods. Plasma-specific insulin ometer over 4 consecutive days. The exclude very active individuals, partici- was determined by a 1235 AutoDELFIA participants wore the accelerometer in an pants completed a screening activity automatic immunoassay system using a elastic waistband at the lower back during questionnaire describing occupational two-step time-resolved fluorometric as- daytime, except while bathing and during and leisure activity, based on published say (DAKO, Ely, Cambridgeshire, U.K.). other aquatic activities. Participants who questionnaires (18,19). Cross-reactivity with intact proinsulin is did not manage to record at least 500 min/ Of the 465 eligible individuals, 399 Ͻ0.5% at 2,736 pmol/l, with 32–33 split day of activity for at least 3 days were ex- were recruited for baseline measure- proinsulin is 1% at 2,800 pmol/l, and cluded from further analyses (n ϭ 7). The ments. Complete data on aerobic fitness, with C-peptide is Ͻ0.1% at 5,280 pmol/l. outcome variables were total body move- Ϫ anthropometrics, and biochemistry at Typical interassay coefficients of variance ment (counts day 1), which is an indi- baseline were available for 365 individu- are 3.1% at 29.0 pmol/l, 2.1% at 79.4 cator of the total volume of physical Ϫ als. In a subsample (n ϭ 258), patterns of pmol/l, 1.9% at 277 pmol/l, and 2.0% at activity, and time (minutes day 1) spent physical activity were also measured ob- 705 pmol/l, respectively (n ϭ 174 in each at different activity intensity categories jectively by means of accelerometry. This case). averaged per day over the measurement constitutes the sample for the present period. These were calculated to deter- analyses. All data presented in this report Clustered metabolic risk mine which subcomponents of activity, if are cross-sectional data from the baseline We constructed a standardized continu- any, are associated with individual and data collection conducted before ran- ously distributed variable (zMS) for clus- clustered metabolic risk and to establish domization. All participants provided tered metabolic risk, which we have whether these associations were related in written informed consent, and ethical described in detail previously (15,16). a dose-response manner. Intensity permission for the study was granted by This variable was derived by standardiz- thresholds for moderate (1,952–5,724 Ϫ the Eastern England Multicenter Research ing and then summing the following con- counts min 1) and vigorous intensity ac- Ϫ Ethics Committee. tinuously distributed indexes of obesity tivity (Ͼ5,725 counts min 1) were de- (waist circumference), hypertension fined according to Freedson et al. (24). Anthropometric and metabolic tests ([systolic blood pressure ϩ diastolic Because Ͼ60% of participants did not ac- Height and weight were measured using a blood pressure]/2), hyperglycemia (fast- cumulate any time in vigorous intensity rigid stadiometer and calibrated scales ing plasma glucose), physical activity, we constructed a single with the subjects wearing light clothing. (fasting insulin), inverted fasting HDL variable by combining accumulated time Waist circumference was measured in du- cholesterol, and hypertriglyceridemia to in MVPA. Sedentary behavior was defined Ϫ plicate using a metal tape. Resistance create a Z score. The purpose of using a as Ͻ100 counts min 1 and light inten- Ϫ (ohms) was assessed using a standard bio- continuously distributed variable was to sity activity as 101–1,951 counts min 1. impedance technique (Bodystat, Isle of maximize statistical power (23). The cutoff for sedentary behavior is an Man, U.K.). This method has previously arbitrary threshold, which we have used been shown to be valid (20) and reliable Assessment of aerobic fitness and previously (25). We also calculated the (21). Total body water and fat-free mass physical activity average number of 5- and 10-min bouts of (FFM) were calculated using the imped- Aerobic fitness (VO2max) was predicted as sustained physical activity at the MVPA ance index (square of height divided by oxygen uptake at maximal heart rate (220 level. In these analyses we allowed 1 min resistance), and body weight and resis- minus age) by extrapolation of the regres- to drop below the threshold for MVPA in tance were calculated according to pub- sion line established during the individual each 5-min bout. We calculated the per- lished equations (22). Fat mass was calibration for the relationship between centage of participants accumulating 30 calculated as body weight minus FFM. oxygen consumption and heart rate dur- min or more per day at MVPA according Percent body fat was calculated as fat ing a submaximal graded treadmill ex- to the recommendations from the Centers ϫ mass/body weight 100. ercise test. Oxygen uptake and CO2 for Disease Control and Prevention and Blood pressure was measured with production were continuously measured the American College of Sports Medicine subjects in the seated position using an by indirect calorimetry throughout the (26). Data reduction, cleaning, and anal- Accutorr automatic sphygmomanometer test (Vista XT metabolic system; yses of accelerometer data were per- (Datascope, Cambridge, U.K.). Systolic Vacumed, Ventura, CA). Participants formed using a specially written program and diastolic blood pressures were mea- breathed through a face mask (Hans Ru- (MAHUffe; see www.mrc-epid.cam. sured in triplicate at minute intervals, and dolph, Kansas City, MO), and expired air ac.uk). the mean of these measurements was used was measured with a turbine flowmeter, in analyses. carbon dioxide concentration with an in- Statistics Fasting blood samples were taken, frared sensor, and oxygen concentration Descriptive summary statistics were cal- centrifuged and aliquoted on site, and im- with a fast differential paramagnetic sen- culated using means Ϯ SD. Fasting insu- mediately placed on ice and transferred to sor. Gas analyzers were calibrated with lin and triglycerides were logarithmically the laboratory where they were stored at gases of known composition, and the tur- transformed owing to their skewed distri- Ϫ70°C within 4 h. Blood samples were bine flow meter was calibrated with a 3-l bution. Geometric mean and reference in- measured in the routine National syringe before each measurement. tervals (1.96 ϫ SD) are presented in the Service laboratory at Addenbrooke’s Hos- Data on free-living physical activity RESULTS. pital in Cambridge. Plasma glucose was was assessed with an MTI ActiGraph (for- We modeled the associations be-

338 DIABETES CARE, VOLUME 30, NUMBER 2, FEBRUARY 2007 Ekelund and Associates

Table 1—Descriptive characteristics of participants

Men Women P value by ANOVA n 103 155 Age (years) 40.9 Ϯ 6.4 40.7 Ϯ 6.4 NS Weight (kg) 90.4 Ϯ 16.4 73.7 Ϯ 14.4 Ͻ0.0001 Height (m) 1.78 Ϯ 0.07 1.64 Ϯ 0.06 Ͻ0.0001 BMI (kg/m2) 28.4 Ϯ 4.6 27.4 Ϯ 5.1 NS Fat mass (kg) 22.5 Ϯ 9.4 26.6 Ϯ 10.0 0.01 FFM (kg) 67.9 Ϯ 8.7 47.1 Ϯ 6.2 Ͻ0.0001 Waist (cm) 101.4 Ϯ 12.0 88.1 Ϯ 11.8 Ͻ0.0001 Systolic blood pressure (mmHg) 127.3 Ϯ 11.2 120.8 Ϯ 13.6 Ͻ0.0001 Diastolic blood pressure (mmHg) 81.9 Ϯ 8.8 76.2 Ϯ 9.3 Ͻ0.0001 Fasting insulin (mmol lϪ1) 55.7 (50.2–61.8) 46.5 (43.4–49.8) 0.02 Fasting glucose (mmol lϪ1) 5.10 Ϯ 0.86 4.78 Ϯ 0.52 Ͻ0.0001 Triglycerides (mmol lϪ1) 1.45 (1.31–1.61) 1.05 (0.98–1.11) Ͻ0.0001 HDL (mmol lϪ1) 1.22 Ϯ 0.32 1.57 Ϯ 0.38 Ͻ0.0001 . Ϫ1 Ϯ Ϯ VO2MAX (ml FFM min ) 59.6 11.1 57.8 10.4 NS Time sedentary (min dayϪ1) 442 Ϯ 97 419 Ϯ 77 0.03 Time light (min dayϪ1) 309 Ϯ 80 320 Ϯ 68 NS Time moderate and vigorous (min dayϪ1) 30 Ϯ 18 24 Ϯ 16 0.003 No. of 5-min bouts of MVPA 0.9 Ϯ 0.9 0.8 Ϯ 0.9 NS No. of 10-min bouts of MVPA 0.3 Ϯ 0.5 0.2 Ϯ 0.4 NS Total counts (ϫ 1,000 dayϪ1) 283 Ϯ 104 264 Ϯ 98 NS Data are means Ϯ SD or geometric mean (95% CI). n ϭ 258. tween all physical activity subcompo- RESULTS — Table 1 shows the de- tween accumulated time spent sedentary, nents (total counts, time spent sedentary, scriptive characteristics of study partici- at light intensity activity, and at MVPA at light intensity activity, and at MVPA, pants. Thirty-two percent of participants and total body movement (counts Ϫ and bouts of MVPA) and the phenotypes were classified as normal weight, 40% day 1), with the intermediary phenotypic of the metabolic syndrome and clustered were overweight, and an additional 27% metabolic risk factors and clustered met- metabolic risk in separate models. These were obese. Age, weight, height, BMI, abolic risk. Because all outcomes are ex- analyses were adjusted for age and sex. waist circumference, fat mass, FFM, pressed in the same unit (SD), it is When obesity was not the outcome of in- VO2max, fasting HDL cholesterol, triglyc- possible to directly compare the magni- terest, we assessed whether the subcom- erides, glucose, and the summary score of tude of associations between the different ponents of physical activity were clustered metabolic risk did not differ sig- components of activity to each of the out- associated with each intermediary pheno- nificantly (all P Ͼ 0.15) between those comes, assuming they are measured with type per se after adjustment for waist included in this report (n ϭ 258) and the the same degree of measurement error. Ϫ circumference, age, and sex. We then ex- rest of the study participants (n ϭ 105). Total body movement (counts day 1) amined whether the different physical ac- However, fasting insulin (log trans- was significantly and independently asso- tivity subcomponents were associated formed) was slightly higher in those in- ciated with three (fasting insulin, triglyc- with clustered metabolic risk (zMS)in cluded in the present report (P ϭ 0.041). erides, and HDL cholesterol) of the six separate models. Finally, stepwise multi- Sex differences were observed for individual risk factors and with clustered ple linear regression analysis was used to most of the body composition and meta- metabolic risk. Time spent sedentary was examine which of the accelerometry- bolic variables. Aerobic fitness and total marginally significantly associated with derived time estimates of physical activity body movement (total counts per day) fasting insulin (P ϭ 0.05). Time spent at variables contributed to the explained were similar for men and women. How- light-intensity activity was inversely asso- variation in clustered metabolic risk after ever, they allocated their physical activity ciated with fasting triglycerides (P ϭ adjustment for sex, age, and measure- differently between the different time es- 0.03). Time spent at MVPA was signifi- ment time. We controlled for multicol- timates of activity. Men spent significantly cantly associated with fasting insulin (P ϭ linearity by calculating the correlation more time sedentary (P ϭ 0.03) and at 0.02) and clustered metabolic risk (P ϭ coefficients between the different time- MVPA (P ϭ 0.002) than women. Two- 0.03). As indicated by the ␤-coefficients, derived variables and by calculating the thirds (66.2%) of participants accumu- the magnitude of association between to- Ϫ tolerance and variance inflation factors. lated Ͻ30 min of MPVA per day tal body movement (counts day 1) with All data were analyzed in their continuous according to the accelerometry measure- clustered metabolic risk was about 25% form although data are stratified by quar- ments. Of participants, 61 and 25% did stronger than that between MVPA and tiles of total body movement (counts not record any single 10- or 5-min bouts, clustered risk. Bouts of PA were not sig- Ϫ day 1) for illustrative purposes. All anal- respectively, of MVPA during the mea- nificantly associated with any of the inter- yses were conducted using SPSS for Win- surement period. mediary risk factors or with clustered dows (version 11; SPSS, Chicago, IL). P Ͻ Table 2 shows the separate associa- metabolic risk (data not shown). Further- 0.05 denotes statistical significance. tions (standardized ␤-coefficients) be- more, substituting waist circumference

DIABETES CARE, VOLUME 30, NUMBER 2, FEBRUARY 2007 339 Activity and metabolic risk in diabetes

with fat mass or BMI as a confounding controlled for several potential confound-

value variable did not change the results (data ing factors (sex, age, aerobic fitness, and P not shown). Similarly, when we excluded obesity), it is possible that other unmea- waist circumference for the clustered risk sured confounders, such as genetic varia- y adjusted for

0.004) 0.003 score and adjusted for waist circumfer- tion, socioeconomic variables, and early Ϫ ence as a confounder, the results were life programming, could explain our find- unchanged (data not shown).We reexam- ings. Second, our results may be restricted 0.01 to 0.005) 0.50 0.005 to 0.009) 0.56 0.003 to 0.013) 0.21 0.01 to 0.004) 0.30 0.007 to 0.001) 0.09 0.1 to 0.002) 0.17 0.02 to ined our data for the associations between to sedentary, overweight, middle-aged, Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ total body movement with individual and Caucasian individuals with a family his-

0.01 ( clustered metabolic risk by further adjust- tory of type 2 diabetes. However, given 0.003 ( 0.004 ( 0.003 ( 0.005 ( Ϫ sk in middle-aged adults with a Ϫ Ϫ Ϫ Ϫ ment for aerobic fitness, and the results the epidemic increase in overweight and were unchanged (data not shown). obesity (3) and the large proportion of

value Fitness Finally, we examined which of the adults who are not sufficiently active (27), P time estimates (i.e., time spent sedentary, it is likely that our results are generaliz- at light-intensity activity, and at MVPA) able to a large part of the general popula- 0.001) 0.03 0.002 ( 0.04) 0.005 0.003) 0.04 0.04) 0.004 contributed to the explained variance in tion. Third, the low frequency of Ϫ Ϫ Ϫ Ϫ clustered metabolic risk after adjusting individuals participating in vigorous- for age, sex, and aerobic fitness. Time intensity physical activity may reduce the 0.11 to 0.11) 0.97 0.20 to 0.20 to 0.01) 0.09 0.005 ( 0.23 to 0.23 to 0.20 to 0.14 to 0.08) 0.59 spent at MVPA (standardized ␤ϭϪ0.13 power to observe associations between Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ [95% CI Ϫ0.17 to Ϫ0.01], P ϭ 0.03) and time spent at this intensity level and met- 0.11 ( 0.09 ( 0.14 ( 0.12 ( 0.12 ( 0.03 ( sex remained as significant explanatory abolic risk. Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ variables in the final model (adjusted R2 This study also has unique strengths. ϭ 19.1%). Aerobic fitness was signifi- We reduced the potential for recall bias

value Total counts cantly and inversely associated with fast- and differential measurement error, P ing insulin (P ϭ 0.003) but not with any which is an unavoidable component of of the other individual risk factors or with self-reported physical activity, by measur- clustered metabolic risk (Table 2). ing the total volume and other subcom- 0.02) 0.02 0.008) 0.03 Ϫ Ϫ Figure 1 shows the associations be- ponents of activity with accelerometry. tween total body movement (counts These methods have been extensively val- Ϫ1 0.06 to 0.16) 0.37 0.002 ( 0.17 to 0.03) 0.17 0.21 to 0.01) 0.07 0.21 to 0.18 to 0.05) 0.27 0.17 to 0.14 to 0.07) 0.53 day ) and clustered metabolic risk. A idated in the laboratory and during free- Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ significant inverse dose-response associa- living conditions (24,28–30). Aerobic 0.07 ( 0.10 ( 0.12 ( 0.06 ( 0.09 ( 0.03 ( tion was observed between quartiles of to- fitness was derived from a submaximal Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ tal activity with clustered metabolic risk exercise test. This measure is less precise (P for trend ϭ 0.02). than a true maximal exercise test. How-

value MVPA ever, it is unlikely that this fact will bias P CONCLUSIONS — We show here the results of the present study, as the er- the cross-sectional associations between ror in predicted maximal heat rate is likely different subcomponents of objectively to be random across the cohort. It is there- 0.001) 0.03

Ϫ measured physical activity with interme- fore also unlikely that the observed asso- diary phenotypic risk factors and clus- ciations between activity variables, 0.08 to 0.14) 0.60 0.05 ( 0.17 to 0.05) 0.28 0.08 to 0.13) 0.63 0.18 to 0.07)0.23 to 0.40 0.12 to 0.05) 0.40 0.12 to 0.09) 0.75 tered metabolic risk in a group of aerobic fitness, and metabolic outcomes Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ sedentary, middle-aged individuals with are due to chance, bias, or measurement 0.06 ( 0.06 ( 0.12 ( 0.03 ( 0.02 ( a family history of type 2 diabetes. Our error. Ϫ Ϫ Ϫ Ϫ Ϫ results suggest that total body movement The association between total body was significantly associated with insulin movement and clustered metabolic risk ϳ value Light resistance, dyslipidemia, and clustered was 25% stronger than the association P

258. Data on insulin and triglycerides are log transformed. Data are adjusted for sex and age.metabolic Data on total counts from accelerometry are additionall risk independent of sex, age, between time spent at MVPA and clus-

ϭ obesity, and aerobic fitness. Time spent at tered risk. This result supports the argu- n MVPA was weakly associated with clus- ment that all body movements that tered risk, whereas bouts of MVPA and contribute to elevated energy expenditure 0.13 to 0.09) 0.73 0.03 ( 0.01 to 0.21) 0.08 0.07 to 0.14) 0.51 0.03 ( 0.04 to 0.20) 0.20 0.01 to 0.14) 0.10 0.09 to 0.12) 0.77 accumulated time spent sedentary and at contribute to a favorable metabolic risk Ϫ Ϫ Ϫ Ϫ Ϫ Ϫ light-intensity activity were unrelated to profile. This observation is important clustered risk. Promoting overall body when one is translating epidemiological 0.02 ( 0.10 ( 0.10 (0.001–0.19)0.08 0.049 (

Ϫ movement and an increase in everyday findings into preventive action. Thus, al- ) -coefficients (95% CI). total physical activity may therefore have though structured exercise and moderate 1 ␤ Ϫ l beneficial effects on metabolic risk factors to vigorous activity contribute to the pre- ) 1

Ϫ in overweight, sedentary adults. vention of cardiovascular and metabolic ) l 1 Ϫ

l There are several limitations that need disease risk, particularly among those at Independent associations of patterns of physical activity from accelerometry and intermediary phenotypic risk factors and clustered metabolic ri to be considered when interpreting the high risk (31), an exclusive focus on in-

scoresfindings 0.07 ( from this study. First, this is a creasing vigorous activity and structured ) Z 1 Ϫ

l cross-sectional study, limiting inferences exercise could have counterintuitive con- Triglycerides (mmol OutcomeWaist (cm) 0.04 ( Sedentary Sum Blood pressure 0.02 ( Inverted HDL (mmol Insulin (mol Glucose (mmol Table 2— family history of type 2 diabetes Data are standardized monitoring time. All subcomponents except waist circumference are additionallyof adjusted for waist circumference. causality and its direction. Although we sequences, especially in populations who

340 DIABETES CARE, VOLUME 30, NUMBER 2, FEBRUARY 2007 Ekelund and Associates

terns of physical activity because of reli- 4. Zimmet P, Alberti KG, Shaw J: Global and ance on self-reported measures of activity societal implications of the diabetes epi- and lack of controls for aerobic fitness. demic. Nature 414:782–787, 2001 From a public health perspective, this fact 5. Stumvoll M, Goldstein BJ, van Haften is important because if the association be- TW: Type 2 diabetes: principles of patho- genesis and therapy. Lancet 365:1333– tween physical activity and CVD and met- 1346, 2005 abolic risk factors is mediated through 6. Eckel RH, Grundy SM, Zimmet PZ: The aerobic fitness, it is likely that more vig- metabolic syndrome. Lancet 365:1415– orous exercise is needed to prevent CVD 1428, 2005 morbidity and mortality. Individual and 7. Wareham NJ, Rennie K: The assessment population-based interventions would of physical activity in individuals and need to reflect this possibility. populations: why try to be more precise In summary, total body movement is about how physical activity is assessed? Figure 1—Clustered metabolic risk score associated with intermediary phenotypic Int J Obes Relat Metab Disord 22:S30–S38, stratified by quartiles of the total amount of risk factors for CVD and metabolic dis- 1998 Ϫ1 8. Carroll S, Cooke CB, Butterly RJ: Meta- activity (counts day ϫ 1,000) by acceler- ease and with clustered metabolic risk in- ometry in sedentary, overweight, middle-aged bolic clustering, physical activity and fit- dependent of aerobic fitness and obesity. ness in non-smoking, middle-aged men. men and women at high risk of developing di- Increasing the total amount of physical abetes (n ϭ 258). Data are adjusted for sex, Med Sci Sports Exerc 32:2079–2086, 2000 age, and monitoring time (P for trend ϭ 0.02). activity in sedentary and overweight indi- 9. Fung TT, Hu FB, Yu J, Chu NF, Spiegel- viduals may have beneficial effects on man D, Tofler GH, Willett WC, Rimm EB: these metabolic risk factors. Leisure-time physical activity, television find these activities unattractive and these watching, and plasma biomarkers of obe- targets hard to reach. Our results corrob- sity and cardiovascular disease risk. Am J orate our previous findings suggesting Acknowledgments— We thank the Medical Epidemiol 152:1171–1178, 2000 that higher levels of PAEE measured by Research Council, National Health Service 10. Rainwater DL, Mitchell BD, Comuzzie individual calibrated heart rate monitor- R&D, Royal College of General Practitioners AG, VandeBerg JL, Stern MP, MacCluer Scientific Foundation, and Diabetes U.K., JW: Association among 5-year changes in ing are favorably associated with features which funded the development and execution weight, physical activity, and cardiovas- of the metabolic syndrome (15,16). of the ProActive trial. cular disease risk factors in Mexican Study participants recorded, on aver- We thank Jian’an Luan for statistical advice, Americans. Am J Epidemiol 152:974–982, age, fewer than one 5-min bout of MVPA Mark Hennings for helping develop the 2000 per day, and bouts of MVPA were not as- MAHUffe software, the study participants, the 11. Laaksonen DE, Lakka HM, Salonen JT, sociated with any of the intermediary practice teams for their collaboration and Niskanen LK, Rauramaa R, Lakka TA: phenotypic risk factors. It may be that the work in helping with recruitment, and the trial Low levels of leisure-time physical activity accumulated time at MVPA and particu- coordination, measurement, and intervention and cardiorespiratory fitness predict de- larly total body movement is more impor- teams. velopment of the metabolic syndrome. tant in relation to these risk factors in The ProActive research team includes, be- Diabetes Care 25:1612–1618, 2002 sides the authors, Kate Williams, Julie Grant, 12. Lakka TA, Laaksonen DE, Lakka HM, sedentary individuals. However, we can- Tom Fanshawe, A. Toby Prevost (principal in- Ma¨nniko¨ N, Niskanen LK, Rauramaa R, not exclude the possibility that bouts of vestigator), Wendy Hardeman (principal in- Salonen JT: Sedentary lifestyle, poor car- activity are more strongly associated with vestigator), William Hollingworth, David diorespiratory fitness, and the metabolic intermediary phenotypic risk factors in a Spiegelhalter (principal investigator), Stephen syndrome. Med Sci Sports Exerc 35:1279– more heterogeneous population. Our Sutton (principal investigator), and Ann Lou- 1286, 2003 data also suggest a linear dose-response ise Kinmonth (principal investigator). 13. Jakes RW, Day NE, Khaw KT, Luben R, association between the total amount of Oakes S, Welch A, Bingham S, Wareham activity with clustered metabolic risk. NJ: Television viewing and low participa- Thus, sedentary individuals may benefit References tion in vigorous recreation are indepen- from accumulating physical activity 1. Laaksonen DE, Lakka HM, Niskanen LK, dently associated with obesity and markers of cardiovascular disease risk: throughout the day or in shorter bouts Kaplan GA, Salonen JT, Lakka TA: Meta- Ͻ bolic syndrome and development of dia- EPIC-Norfolk population-based study. (i.e., 5 min), and more activity accumu- betes mellitus: application and validation Eur J Clin Nutr 57:1089–1096, 2003 lated leads to greater metabolic health of recently suggested definitions of the 14. O’Donovan G, Owen A, Kearney EM, benefits. metabolic syndrome in a prospective co- Jones DW, Nevill AM, Woolf-May K, Bird Previous epidemiological studies hort study. Am J Epidemiol 156:1070– SR: Cardiovascular disease risk factors in have clearly demonstrated that self- 1077, 2002 habitual exercisers, lean sedentary men reported physical activity predicts cardio- 2. Katzmarzyk P, Church TS, Blair SN: Car- and abdominally obese sedentary men. vascular morbidity and mortality end diorespiratory fitness attenuates the effects Int J Obes Relat Metab Disord 29:1063– points (32–34) and that both walking and of the metabolic syndrome on all-cause 1069, 2005 vigorous exercise are associated with risk and cardiovascular disease mortality in 15. Franks PW, Ekelund U, Brage S, Wong reduction in the incidence of cardiovas- men. Arch Intern Med 164:10926–1097, MJ, Wareham NJ: Does the association of 2004 habitual physical activity with the meta- cular events (35–37). Further, both phys- 3. Mokdad AH, Ford ES, Bowman BA, Dietz bolic syndrome differ by level of fitness? ical activity and obesity are independent WH, Vinicor F, Bales VS, Marks JS: Prev- Diabetes Care 27:1187–1193, 2004 predictors of all-cause mortality (38). alence of obesity, diabetes, and obesity- 16. Ekelund U, Brage S, Franks PW, Hen- However, these studies have not been able related health risk factors, 2001. JAMA nings S, Emms S, Wareham NJ: Physical to characterize minute-by-minute pat- 289:76–79, 2003 activity energy expenditure predicts pro-

DIABETES CARE, VOLUME 30, NUMBER 2, FEBRUARY 2007 341 Activity and metabolic risk in diabetes

gression toward the metabolic syndrome miologic surveys. Am J Clin Nutr 77:331– P, Keina¨nen-Kiukaanniemi S, Laakso M, independently of aerobic fitness in mid- 340, 2003 Louheranta A, Rastas M, Salminen V, dle-aged healthy Caucasians: the Medical 23. Armstrong BK, White E, Saracci R. Princi- Uusitupa M, the Finnish Diabetes Preven- Research Council Ely Study. Diabetes Care ples of Exposure Measurement in Epidemi- tion Study Group: Prevention of type 2 28:1195–2000, 2005 ology. Oxford, U.K., Oxford University diabetes mellitus by changes in lifestyle 17. Williams K, Prevost AT, Griffin S, Harde- Press, 1994 among subjects with impaired glucose in- man W, Hollingworth W, Spiegelhalter 24. Freedson PS, Melanson E, Sirard J: Cali- tolerance. N Engl J Med 344:1343–1350, D, Sutton S, Ekelund U, Wareham N, bration of the Computer Science and Ap- 2001 Kinmonth AL: The ProActive trial proto- plications, Inc. accelerometer. Med Sci 32. Berlin JA, Colditz GA: A meta-analysis of col—a randomised controlled trial of the Sports Exerc 30:777–781, 1998 physical activity in the prevention of cor- efficacy of a family-based, domiciliary in- 25. Ekelund U, Åman J, Yngve A, Renman C, onary heart disease. Am J Epidemiol 132: tervention programme to increase physi- Westerterp K, Sjo¨stro¨m M: Physical activ- 612–628, 1990 cal activity among individuals at high risk ity but not energy expenditure is reduced of diabetes. BMC Public Health 4:48, 2004 in obese adolescents: a case-control study. 33. Lee IM, Paffenbarger RS Jr, Hennekens 18. Wareham NJ, Jakes RW, Rennie KL, Am J Clin Nutr 76:935–941, 2002 CH: Physical activity, physical fitness and Schuit J, Mitchell S, Hennings S, Day NE: 26. Pate RR, Pratt M, Blair SN, Haskell WL, longevity. Aging 9:2–11, 1997 Validity and repeatability of a simple in- Macera CA, Bouchard C, Buchner D, Et- 34. Barengo NC, Hu G, Lakka TA, Pekkarinen dex derived from the short physical activ- tinger W, Heath GW, King AC, et al: H, Nissinen A, Tuomilehto J: Low physi- ity questionnaire used in the European Physical activity and public health: a rec- cal activity as a predictor for total and Prospective Investigation into and ommendation from the Centers for Dis- cardiovascular disease mortality in mid- Nutrition (EPIC) study. Public Health Nutr ease Control and Prevention and the dle-aged men and women in Finland. Eur 6:407–413, 2003 American College of Sports Medicine. Heart J 25:2204–2211, 2004 19. Godin G, Shephard RJ: A simple method JAMA 273:402–407, 1995 35. Hakim AA, Petrovitch H, Burchfiel CM, to assess exercise behavior in the commu- 27. At Least Five a Week: Evidence on the Impact Ross GW, Rodriguez BL, White LR, Yano nity. Can J Appl Sport Sci 10:141–146, of Physical Activity and Its Relationship to K, Curb JD, Abbott RD: Effects of walking 1985 Health: A Report from the Chief Medical Of- on mortality among non-smoking retired 20. Simpson JA, Lobo DN, Anderson JA, Mac- ficer. London, Department of Health, men. N Engl J Med 338:94–99, 1998 donald IA, Perkins AC, Neal KR, Allison 2004 36. Manson JE, Greenland P, LaCroix AZ, Ste- SP, Rowlands BJ: Body water compart- 28. Brage S, Wedderkopp N, Franks PW, fanick ML, Mouton CP, Oberman A, Perri ment measurements: a comparison of bio- Andersen LB, Froberg K: Re-examination MG, Sheps DS, Pettinger MB, Siscovick electrical impedance analysis with tritium of validity and reliability of the CSA activ- DS: Walking compared with vigorous ex- and sodium bromide dilution techniques. ity monitor during walking and running. ercise for the prevention of cardiovascular Clin Nutr 20:339–343, 2001 Med Sci Sports Exerc 35:1447–1454, 2003 events in women. N Engl J Med 347:716– 21. Shanholtzer BA, Patterson SM: Use of bio- 29. Ekelund U, Åman J, Westerterp K: Is the 725, 2002 electrical impedance in hydration status ArteACC index a valid indicator of free- assessment: reliability of a new tool in living physical activity in adolescents? 37. Lee IM, Rexrode KM, Cook NR, Manson psychophysiology research. Int J Psycho- Obes Res 11:793–801, 2003 JE, Buring JE: Physical activity and coro- physiol 49:217–226, 2003 30. Ekelund U, Sjo¨stro¨m M, Yngve A, Poortv- nary heart disease in women: is “no pain, 22. Sun SS, Chumlea WC, Heymsfield SB, liet E, Nilsson A, Froberg K, Wedderkopp no gain” passe? JAMA 285:1447–1454, Lukaski HC, Schoeller D, Friedl K, Kucz- N, Westerterp K: Physical activity as- 2001 marski RJ, Flegal KM, Johnson CL, Hub- sessed by activity monitor and doubly la- 38. Hu FB, Willett WC, Li T, Stampfer MJ, bard VS: Development of bioelectrical beled water in children. Med Sci Sports Colditz GA, Manson JE: Adiposity as impedance analysis prediction equations Exerc 33:275–281, 2001 compared with physical activity in pre- for body composition with the use of a 31. Toumiletho J, Lindstro¨m J, Eriksson JE, dicting mortality among women. N Engl multicomponent model for use in epide- Valle TT, Ha¨ma¨la¨inen H, Ilanne-Parrika J Med 351:2694–703, 2004

342 DIABETES CARE, VOLUME 30, NUMBER 2, FEBRUARY 2007