European Journal of Clinical Nutrition (2014) 68, 309–315 & 2014 Macmillan Publishers Limited All rights reserved 0954-3007/14 www.nature.com/ejcn

ORIGINAL ARTICLE

Predictors of change in weight and waist circumference: 15-year longitudinal study in Australian adults S Arabshahi1,2, PH Lahmann2,3, GM Williams3 and JC van der Pols2,3 BACKGROUND/OBJECTIVES: This study examines which socio-demographic and lifestyle characteristics are associated with weight and waist circumference (WC) change in a cohort of Australian adults over a 15-year period (1992–2007). Further, it tests the effect of period of birth (birth cohort) on mean weight and WC at two time points, 15 years apart. SUBJECTS/METHODS: Up to three repeated measures of weight (n ¼ 1437) and WC (n ¼ 1317) were used. Self-reported data on socio-demographic and lifestyle characteristics were derived from repeated questionnaires. Multivariable models, stratified by sex, were adjusted for potential confounders. RESULTS: Participants born more recently were heavier, on average, than those in the same age group 15 years earlier, but there was no such secular trend in WC. Age at baseline was associated with change in weight and WC, but the pattern was different: participants gained weight up to age 55 years, while WC gain continued to 65 years. In women, higher level of recreational physical activity was associated with lower WC gain (Po0.05). Parity was also associated with WC change in women (Po0.05), but there was no linear trend. CONCLUSIONS: Age was the most important factor associated with change in weight and WC in both sexes, apparently reducing the influence of all potential covariates. Among women, physical activity and parity were also associated with change in weight and WC. This study provides longitudinal evidence to support public health efforts that address the continuous increases in average weight and WC of many populations around the world. European Journal of Clinical Nutrition (2014) 68, 309–315; doi:10.1038/ejcn.2013.260; published online 8 January 2014 Keywords: body weight changes; waist circumference; obesity; socio-demographic factors; longitudinal studies; cohort effect

INTRODUCTION Obesity is a significant health problem worldwide.1,2 A number of social, economic and lifestyle characteristics may affect change in weight in individuals. However, findings from longitudinal studies on the overall effect of such factors are inconsistent. For example, the association between smoking and body mass is complex,3 with some studies indicating that smoking is associated with weight loss,4,5 while others show the opposite.6 Conclusions have also been inconsistent in previous studies of alcohol use and risk of obesity.7–11 Central obesity, independent of total body fat, is associated with increased risk of some chronic disease.12 However, limited evidence is available on the socio-demographic and lifestyle factors that influence change in waist circumference (WC) over time.10,13 Further, it is currently unclear whether or not the predictors of long-term change in general obesity differ from those of abdominal obesity. Such information is needed to improve the design of obesity prevention efforts. Results from repeated cross-sectional studies in different populations14–16 have indicated that the period of birth (birth cohort) has a significant independent effect on the risk of weight gain. Similar evidence is scarce for WC.17 We have addressed these research questions by analyzing longitudinal assessments of weight and WC in an Australian community.

SUBJECTS AND METHODS Study population We used data collected in the Nambour Skin Cancer Study, which involved a random, population-based sample of Australian men and women. The design and follow-up methods have previously been described in detail.18,19 In brief, participants were 1621 residents of Nambour, a semirural township in Queensland, who were originally randomly selected from the electoral roll (voting is compulsory in Australia)20 and participated in the Nambour Skin Cancer Study and were followed up between 1992–2007. Body weight was measured at a study clinic using a standard protocol at three time points (1992, 1996 and 2007), and WC was measured twice (1992 and 2007). Participants were included in these analyses if they had at least one measurement of weight or WC. Of the total 1621 participants enrolled in the Nambour Study in 1992, weight was measured for 1248 participants in 1992, 1271 participants in 1996 and 712 participants in 2007 (data missing completely for 184 individuals (11%)). The weight analyses therefore included 249 (17%) participants with one observation, 582 (41%) with two observations and 606 (42%) with three observations. These 1437 individuals (57% women) contributed a total of 3231 observations (one observation ¼ one measurement of weight of one person at one time point) in the 15-year follow-up period. WC was measured in 1221 participants in 1992 and 712 participants in 2007 (data missing completely for 304 individuals (19%)). The study population therefore included 701 (53%) participants with one observation and 616 (47%) with two observations. These 1317 participants (57% women) contributed a total of 1933 observations during the

1 Department of Medicine, Southern Clinical School, Monash Medical Centre, Monash University, Melbourne, Victoria, Australia; 2Queensland Institute of Medical Research, Cancer and Population Studies, Locked Bag 2000, Royal Brisbane Hospital, Brisbane, Australia and 3The University of Queensland, School of Population Health, Brisbane, Queensland, Australia. Correspondence: Assistant Professor JC van der Pols, School of Population Health, The University of Queensland, Brisbane, Queensland, Australia. E-mail: [email protected] Received 14 December 2012; revised 17 October 2013; accepted 18 October 2013; published online 8 January 2014

Weight and waist circumference change S Arabshahi et al

310 follow-up period. The Queensland Institute of Medical Research Ethics Committee approved the study and all participants provided informed written consent.

Covariates Details of socio-demographic, lifestyle, diet and health-related behaviors were collected through self-completed questionnaires in 1992, 1996 and 2007. Participants were considered to have a medical condition if they had reported having glaucoma, gallstones, high cholesterol, high triglycerides, diabetes/high blood sugar, hypertension, angina, heart attack, stroke or cancer. Recreational physical activity was categorized based on selfreported engagement in walking (low physical activity level), moderate activity or vigorous exercise in the past two weeks. Smoking status was ascertained based on the timing and number of pack-years smoked (a pack year is defined as twenty cigarettes smoked every day for one year), calculated from the frequency and duration of cigarettes smoked, for each period preceding each measurement year. Occupation was categorized using standard categories for Australia.21 Baseline body mass index (BMI) and WC were used as categorical variables and classified according to WHO criteria: overweight: 25p BMIo30 kg/m2; WC men 94–102 cm, women 80–88 cm; obesity: BMIX30 kg/m2; WC men X102 cm, women X88 cm. The diet assessment method was a self-administered semi-quantitative food frequency questionnaire adapted from the US Nurses’ Health Study, which had previously been validated in this study population.22 Using food frequency questionnaire data from all measurement years (1992, 1996 and 2007), the participants were divided into four equal groups according to their ranked energy intake (kJ/day), separately for men and women. Alcohol consumption (g per day) was calculated from the food frequency questionnaire data, and categorized based on the national alcohol guidelines.23

Statistical analysis Influence of period of birth (birth cohort) was ascertained using a twosample t-test comparing mean weight or WC in 1992 and 2007. Longitudinal associations between socio-demographic and lifestyle factors and change in weight or WC were assessed by linear regression using the generalized estimating equations approach.24 This approach takes into consideration concurrent changes in the response variable and covariates over time.25 Weight and WC were used as continuous variables; change in weight or WC per year in each category of covariates were calculated by including an interaction term between the variable and time (year of observation as a continuous variable). The analyses were stratified by sex because the pattern of associations was expected to be different for men and women. Covariates were included as time-dependent (changing over time) or time-independent (constant over time) variables. Time-dependent variables were medical condition, physical activity, alcohol consumption, smoking status, energy intake and hormone replacement therapy use in women. Time-independent variables were age, BMI, education and occupation (all assessed in 1992) and parity (assessed in 2004). The analyses were carried out in two steps: (1) univariate analyses for each covariate and weight or WC change as the outcome, and (2) multivariable adjusted analyses to identify which covariates were independently related to weight or WC change over time. In multivariable models, adjustments were made for variables for which the univariate association with weight or WC change had a P-value o0.1 (Supplementary Tables 2–3). Although the Quasi-likelihood under the Independence model Criterion is commonly used to review model fit and was considered in our analyses, our main approach was to investigate the contribution of each explanatory variable independent of other variables that were also found to be associated with the outcome in the univariate analyses. BMI at baseline was not associated with weight change over time (data not shown), thus we did not adjust for it in multivariable models. Since baseline abdominal obesity was associated with WC change in both men and women (data not shown), all multivariable modelling of this outcome included baseline abdominal obesity as a covariate. Additional adjustment for baseline height in the weight and WC models did not change the results and was therefore not retained in the model. Covariates added to the models included both the main effect and the interaction term with time for the variable. Age (cantered on the mean) and its squared value were used in the multivariable model when adjusting for this variable. A P-value for the overall association between European Journal of Clinical Nutrition (2014) 309 – 315

each covariate and the outcome was derived from the likelihood ratio test for interaction of each covariate by time. A P-value for subgroup comparisons within each covariate was derived from a Wald Test based on parameter estimates and standard errors from the generalized estimating equations model. To examine the pattern of missingness due to persons dropping out of the study, and to check whether the assumption of missing at random could be made, we compared the characteristics of the participants included in the analyses with those excluded using multiple logistic regressions applying the generalized estimating equations approach. Values of Po0.05 were considered statistically significant. All analyses were carried out using SAS statistical package version 9.1 (SAS Institute, Cary, NC, USA).

RESULTS At baseline in 1992, participants’ mean (s.d.) age was 50 (13) years; 40% were overweight and 17% were obese, while 28% were abdominally overweight and 29% were abdominally obese. Individuals excluded from the weight analyses (11%) due to missing weight data were more likely to be in the youngest or oldest age categories (Po0.01). Persons excluded from the WC analyses (19%) were also more likely to be in the youngest or oldest age categories, and tended to have a lower education level, and were more likely to have a non-professional occupation (all Po0.01) (detailed results not shown). Descriptive data of weight, height, BMI and WC by examination year are presented in Supplementary Table 1. Data to establish the influence of period of birth (birth cohort) on body weight and WC are shown in Table 1. Participants born in more recent years were heavier than their counterparts in the same age groups 15 years earlier. However, the differences were not statistically significant for those aged 34–44 years and X75 years in both sexes. There was no association between birth cohort and WC in this study population. Participants gained on average 0.28 (±s.e.: 0.02) kg per year or 4.2 (±s.e.: 0.3) kg weight in total over 15 years and 0.20 (±s.e.: 0.02) cm per year in WC or 3 (±s.e.: 0.3) cm in total over the 15-year follow-up. Gain in weight and WC differed significantly by sex. Average weight gain was 0.22 (±s.e.: 0.03) kg per year or 3.2 (±s.e.: 0.45) kg in total for men compared to 0.32 (±s.e.: 0.03) kg per year or 4.8 (±s.e.: 0.45) kg for women (P ¼ 0.008), while WC increased by 0.12 (±s.e.: 0.03) cm per year or 1.94 (±s.e.: 0.46) cm in total in men compared to 0.25 (±s.e.: 0.03) cm per year or 3.67 (±s.e.: 0.46) cm in total in women (P ¼ 0.004). The multivariable associations with socio-demographic and lifestyle factors are presented in Tables 2 (men) and 3 (women) (univariate analyses are shown in Supplementary Tables 2 and 3).

Factors associated with change in weight and WC in multivariable analysis: men Age was associated with change in both weight and WC in men, with a general increase in weight and WC over the 15 years of follow-up in the three lowest age categories (25–54 years). In men aged 55–64 years, average weight decreased slightly while WC increased. In men aged X65 year, there was a general reduction in both weight and WC (Table 2). Education and occupation were not associated with weight or WC change in men, although the average yearly change among sub-groups suggested that less educated men and those with a non-professional occupation were more likely to gain more weight during the follow-up period. Recreational physical activity was associated neither with change in weight nor with WC. While the average yearly change in weight among sub-groups suggested that men with higher physical activity gained less weight, there was no evidence for a significant effect of exercise on weight change in men. & 2014 Macmillan Publishers Limited

Weight and waist circumference change S Arabshahi et al

311 Table 1.

Weight and waist circumference by birth cohort, Nambour Study, 1992 and 2007 Weight analysis 1992

Age group (year)

Waist circumference analysis

2007

1992

b

2007

b

P-value

P-value n

a

Weight (kg)

n

Weight (kg)

a

n

Waist circumference (cm)a

n

Waist circumference (cm)a

Men 25–34 35–44 45–54 55–64 65–74 X75

81 105 133 102 112 7

82.8 80.8 84.6 80.0 78.2 76.9

(12.2) (11.5) (14.1) (9.9) (12.5) (13.6)

— 18 72 83 71 58

— 85.5 (11.1) 91.8 (16.1) 86.8 (11.5) 84.3 (15.3) 75.4 (10.8)

0.1 0.001 0.0001 0.004 0.7

81 104 131 96 108 7

93.4 92.7 98.1 97.3 98.7 100.5

(10.7) (8.5) (11.1) (9.4) (10.2) (10.8)

— 18 72 83 71 58

91.2 98.3 98.2 98.0 95.9

(8.8) (12.4) (9.3) (12.7) (10.4)

0.5 0.9 0.5 0.7 0.3

Women 25–34 35–44 45–54 55–64 65–74 X75

83 167 198 159 92 9

64.6 67.0 69.4 68.7 67.0 70.1

(12.2) (12.7) (14.0) (12.2) (11.4) (13.0)

— 30 90 142 91 57

— 68.6 (14.4) 75.1 (15.3) 74.5 (14.7) 70.8 (14.5) 68.6 (11.6)

0.5 0.003 0.003 0.05 0.7

81 166 192 154 92 9

77.8 79.6 83.2 85.7 88.0 90.2

(10.3) (10.2) (11.3) (10.1) (10.4) (14.9)

— 30 90 142 91 57

78.7 84.7 87.3 86.3 87.7 91.2

(12.5) (11.7) (12.2) (12.8) (9.8) (8.8)

0.7 0.3 0.2 0.3 0.6 0.5

a

From two-sample t-test. bMean (s.d.), all such values.

Factors associated with change in weight and WC in multivariable analysis: women Age was associated with weight change (P ¼ 0.0003) over time in women, but the association with WC change was of borderline statistical significance (P ¼ 0.07). Women aged 25–64 years on average gained in both weight and WC, whereas women aged X65 years on average lost weight and WC (Table 3). Recreational physical activity was not associated with weight change in women, but was associated with WC change. Higher levels of physical activity were related to lower increases in WC, although a test for a linear trend was not statistically significant (data not shown). Parity was not associated with weight change but was associated with WC change, though again not showing a linear trend (data not shown). Except for age at baseline and recreational physical activity and parity among women, none of the other examined variables were related to 15-year change in weight or WC in this cohort.

DISCUSSION This study is one of very few to present data on changes in both body weight and WC over an extended period of time in a community-based sample of Australian adults, while considering concurrent changes in lifestyle factors. The findings clearly demonstrate a birth cohort effect on weight in middle-aged adults, such that the later-born men and women were heavier by B8.3 kg and 5.5 kg, respectively, than their counterparts 15 years earlier. This confirms previous observations from repeated crosssectional studies in Australia14,26 and Scandinavian countries.15,16 Interestingly, period of birth was not associated with WC in our study, which is different from prior findings in other populations,17 but very few studies have been able to study this. It would be valuable to have these discrepant findings for weight and WC confirmed in other and larger data sets. Age at baseline was associated with weight change in both men and women, and with WC change in men (borderline statistical significance in women). Indeed, age was the most important predictor of gain in weight and WC. It seems that age reduced the influence of all other measured covariates on anthropometric & 2014 Macmillan Publishers Limited

change in this study. Age is generally known to be one of the main factors affecting anthropometric changes, typically resulting in increases in body weight with age up to about 60 years, and declines after that.27 Our findings are in agreement with such a pattern. However, it is noteworthy that our data suggest that women continued to gain weight and WC until an older age compared to men, a finding which corresponds with earlier longitudinal studies28,29 and may be partly explained by changes due to the menopausal transition in women in this age group.30 Age-related weight gain has been shown to occur even among physically active people.31 Our results are in accordance with previous findings32 and provide additional evidence that men and women are more likely to gain weight in their 20–40 s, but they are likely to gain WC throughout adulthood, except for those aged X65 years. Given the association between increased WC and chronic disease,12 this finding is of major concern. The possibility that overweight may lead to a healthy survival effect in older adults is the topic of much current debate.33 Weight loss in our study population aged X65 years old is in agreement with previous reports.34,35 Body shape, size and composition are affected by changes due to aging.36 Sarcopenia, a process of loss of muscle mass happens with aging.37 These changes result in weight loss in old people. The greater weight loss in older male than female participants in this cohort is supported by other reports that showed a higher prevalence of sarcopenia in older men than in older women.38 However, some studies have observed the opposite,39 which suggests that weight loss due to sarcopenia may be populationdependent. The amount of weight gain during 15 years (on average almost half a kilogram increase in weight each year for men and women in the youngest age group) in this aging population is a matter of concern. The adverse effect of obesity on cardiovascular disease morbidity40 and mortality41 has been well established. Weight gain is also adversely related to cardiovascular disease risk factors42,43 and even modest increases in weight are associated with a substantially increased risk of cardiovascular disease in middle-aged men and women.9,16 Gains in weight and WC were B30% and 50% greater, respectively, in females than in males (0.32 (±s.e.: 0.03) kg per year vs 0.22 (±s.e.: 0.03) kg per year and 0.25 (±s.e.: 0.03) cm per year vs 0.13 (±s.e.: 0.03) cm per year, European Journal of Clinical Nutrition (2014) 309 – 315

Weight and waist circumference change S Arabshahi et al

312 Table 2. Multivariable-adjusted results for associated socio-demographic and lifestyle characteristics to longitudinal change in weight and waist circumference in men, Nambour Study, 1992–2007a Covariates

n (%)

P-valuec

P-valued

(0.31, 0.61) (0.28, 0.51) (0.16, 0.34) (  0.13, 0.11) (  0.38,  0.16)

Ref.f 0.5 0.009 0.0001 0.0001

0.0001

134 181 215 147 152

(0.09, 0.31) (0.13, 0.38) (0.18, 0.38) (  0.06, 0.25) (  0.28, 0.40)

0.2 0.08 0.04 Ref.f 0.7

0.2

306 131 272 81 15

Weight change (kg per year) (95% CI)b

Waist circumference change (cm per year) (95% CI)b

P-valuec

(16%) (22%) (26%) (18%) (18%)

0.22 0.20 0.19 0.01  0.14

(0.03, 0.41) (0.09, 0.32) (0.08, 0.29) (  0.16, 0.19) (  0.31, 0.04)

Ref.f 0.9 0.8 0.1 0.005

0.02

(38%) (16%) (34%) (10%) (2%)

0.16 0.17 0.18 0.14  0.02

(0.03, 0.29) (0.03, 0.30) (0.05, 0.30) (  0.01, 0.31) (  0.25, 0.21)

0.9 0.9 0.8 Ref.f 0.2

0.8

n (%)

P-valued

Age in 1992e (year) 25–34 35–44 45–54 55–64 X 65

230 296 362 252 261

(16%) (21%) (26%) (18%) (19%)

0.46 0.39 0.25  0.01  0.27

Educatione Grade 12 or less Technical/diploma Trade/apprenticeship Bachelor or higher Other

505 211 453 127 27

(38%) (16%) (34%) (10%) (2%)

0.20 0.25 0.28 0.09 0.06

Occupatione Professional Para-professional Non-professional

346 (26%) 87 (7%) 890 (67%)

0.15 (0.05, 0.24) 0.16 (  0.02, 0.34) 0.27 (0.19, 0.35)

Ref.f 0.9 0.02

0.07

204 (25%) 52 (7%) 548 (68%)

0.08 (  0.04, 0.21) 0.12 (  0.06, 0.30) 0.19 (0.10, 0.29)

Ref.f 0.8 0.1

0.2

Medical condition No Yes

987 (71%) 408 (29%)

0.24 (0.16, 0.32) 0.21 (0.11, 0.31)

Ref.f 0.6

0.6

305 (37%) 519 (63%)

0.16 (0.06, 0.26) 0.16 (0.02, 0.29)

Ref.f 1.0

1.0

0.34 0.26 0.19 0.12

(0.21, 0.46) (0.14, 0.38) (0.09, 0.28) (  0.01, 0.25)

Ref.f 0.3 0.04 0.01

0.06

258 184 237 117

(32%) (23%) (30%) (15%)

0.19 0.22 0.14 0.16

(0.00, (0.05, (0.03, (0.01,

Ref.f 0.8 0.6 0.7

0.8

0.23 0.21 0.16  0.01

(0.15, 0.32) (0.12, 0.31) (  0.10, 0.42) (  0.31, 0.28)

Ref.f 0.7 0.6 0.1

0.5

356 352 37 84

(43%) (43%) (4%) (10%)

0.09 0.19 0.23 0.29

(  0.01, 0.19) (0.08, 0.30) (  0.14, 0.59) (0.10, 0.47)

Ref.f 0.1 0.5 0.04

0.2

0.01 (  0.20, 0.23) 0.13 (0.04, 0.22) 0.32 (0.10, 0.54)

Ref.f 0.3 0.04

0.1

Recreational physical activity Sedentary 377 (28%) Low 340 (25%) Moderate 391 (29%) High 250 (18%) Smoking status Life-long non smoker Ex-smoker Smoker, 1–7 pkyrsg Smoker,47 pkyrsg

592 610 106 93

(42%) (43%) (8%) (7%)

0.38) 0.39) 0.25) 0.30)

Alcohol consumptionh (g per day) None 197 (15%) Moderate 1007 (77%) Heavy 98 (8%)

0.27 (0.08, 0.45) 0.19 (0.11, 0.26) 0.28 (0.05, 0.50)

Ref.f 0.4 0.9

0.5

110 (15%) 569 (76%) 67 (9%)

Frequency of alcoholic beverages (per week) None 197 (15%) o1 200 (15%) 1 80 (6%) 2–4 177 (14%) 5–6 101 (8%) Daily 547 (42%)

0.27 0.19 0.32 0.18 0.21 0.19

(0.08, (0.04, (0.08, (0.04, (0.04, (0.09,

0.45) 0.33) 0.56) 0.32) 0.37) 0.29)

Ref.f 0.5 0.8 0.4 0.7 0.4

0.9

110 110 46 98 58 324

(15%) (15%) (6%) (13%) (7%) (45%)

0.01 0.07 0.27  0.04 0.21 0.19

(  0.20, 0.22) (  0.11, 0.26) (  0.03, 0.58) (  0.17, 0.10) (0.02, 0.39) (0.08, 0.29)

Ref.f 0.6 0.1 0.7 0.2 0.1

0.06

Energy intake (kJ) 3459–7954 7965–9572 9573–1 1387 1 1387–1 6731

0.26 0.24 0.14 0.19

(0.13, (0.13, (0.02, (0.07,

0.39) 0.35) 0.26) 0.31)

Ref.f 0.8 0.1 0.4

0.4

187 186 187 186

(25%) (25%) (25%) (25%)

0.08 0.12 0.16 0.18

(  0.06, 0.23) (  0.03, 0.28) (0.03, 0.29) (0.03, 0.32)

Ref.f 0.7 0.4 0.3

0.7

326 325 326 325

(25%) (25%) (25%) (25%)

n ¼ 1401 observations across 627 individuals for weight analysis and 829 observations across 568 individuals for waist circumference analysis. aAdjusted for age, medical condition (both weight and waist circumference analyses), and baseline waist circumference (where appropriate). bValues are means±95% CI, derived from the interaction between covariates and time using generalized estimating equation (GEE) and data collected in 1992, 1996 and 2007. c Comparison between categories. P-values from Wald test based on parameter estimate and standard error from GEE model. dP-value from likelihood ratio test for interaction of variable by time. eTime-independent variable. fReference. gPack-years smoked. hModerate: p40 g per day and heavy: 440 g per day in men.

respectively), which corresponds to results of national28 and international studies,13,44,45 and places women at particular risk of developing overweight and/or obesity. Higher levels of recreational physical activity were associated with lower levels of WC gain in women. Unexpectedly, physical European Journal of Clinical Nutrition (2014) 309 – 315

activity levels were not associated with changes in weight, although among men it appeared that weight gain was lowest with higher activity levels. It is notoriously difficult to obtain valid measures of physical activity46 and some misclassifications may have occurred. Also, the nature of total physical activity in this & 2014 Macmillan Publishers Limited

Weight and waist circumference change S Arabshahi et al

313 Table 3. Multivariable-adjusted results for associated socio-demographic and lifestyle characteristics to longitudinal change in weight and waist circumference in women, Nambour Study, 1992–2007a Covariates

Age in 1992e (year) 25–34 35–44 45–54 55–64 X 65

n (%)

Weight change (kg per year) (95% CI)b

P-valuec

P-valued

n (%)

Waist circumference change (cm per year) (95% CI)b

P-valuec

P-valued

0.17 (  0.02, 0.36) 0.29 (0.15, 0.44) 0.35(0.17, 0.53) 0.25 (0.06, 0.44)  0.02 (  0.28, 0.24)

Ref.f 0.2 0.2 0.9 0.07

0.07

0.2 0.2 0.2 Ref.f 0.8

0.6

247 443 525 386 229

(13%) (24%) (29%) (21%) (13%)

0.49 0.35 0.28 0.18  0.14

(0.29, 0.68) (0.19, 0.51) (0.11, 0.46) (0.00, 0.36) (  0.35, 0.07)

Ref.f 0.2 0.03 0.002 0.0001

0.0003

149 278 318 228 131

(13%) (25%) (29%) (21%) (12%)

Educatione Grade 12 or less Technical/diploma Trade/apprenticeship Bachelor or higher Other

1106 433 62 79 42

(64%) (25%) (4%) (5%) (2%)

0.27 0.29 0.28 0.30 0.19

(0.11, 0.43) (0.11, 0.47) (  0.04, 0.61) (0.10, 0.50) (  0.23, 0.60)

0.7 0.9 0.9 Ref.f 0.6

1.0

688 267 39 51 25

(64%) (25%) (4%) (5%) (2%)

Occupatione Professionals Para-professionals Non-professionals

333 (20%) 96 (6%) 1262 (74%)

0.26 (0.07, 0.45) 0.28 (0.04, 0.51) 0.25 (0.09, 0.41)

Ref.f 0.8 0.9

1.0

203 (19%) 56 (5%) 791 (75%)

0.29 (0.12, 0.47) 0.31 (0.09, 0.52) 0.34 (0.20, 0.47)

Ref.f 0.8 0.6

0.8

Medical condition No Yes

1278 (70%) 541 (30%)

0.24 (0.10, 0.39) 0.29 (0.12, 0.46)

Ref.f 0.4

0.4

653 (60%) 443 (40%)

0.34 (0.21, 0.46) 0.33 (0.16, 0.49)

Ref.f 0.9

0.9

0.35 0.34 0.23 0.25

Ref.f 0.9 0.2 0.3

0.4

325 356 270 105

(31%) (34%) (25%) (10%)

0.51 0.43 0.28 0.19

(0.34, 0.67) (0.25, 0.61) (0.13, 0.42) (  0.04, 0.43)

Ref.f 0.4 0.007 0.01

0.01

(0.20, 0.47) (0.25, 0.59) (  0.03, 0.69) (  0.32, 0.26)

Ref.f 0.3 1.0 0.02

0.09

735 270 25 72

(67%) (24%) (2%) (7%)

0.33 0.39 0.26 0.24

(0.19, 0.46) (0.21, 0.57) (  0.12, 0.64) (  0.02, 0.49)

Ref.f 0.4 0.7 0.5

0.7

Alcohol consumptionh (g per day) None 460 (27%) Moderate 1157 (68%) Heavy 79 (5%)

0.24 (0.04, 0.43) 0.33 (0.17, 0.50) 0.23 (  0.01, 0.46)

Ref.f 0.2 0.9

0.4

259 (26%) 687 (69%) 51 (5%)

0.29 (0.11, 0.48) 0.32 (0.15, 0.50) 0.20 (  0.18, 0.58)

Ref.f 0.8 0.5

Frequency of alcoholic beverages (per week) None 460 (27%) o1 536 (32%) 1 115 (7%) 2–4 201 (12%) 5–6 110 (6%) Daily 274 (16%)

0.24 0.33 0.13 0.38 0.48 0.27

(0.04, 0.43) (0.14, 0.52) (  0.10, 0.37) (0.18, 0.57) (0.23, 0.73) (0.09, 0.45)

Ref.f 0.3 0.3 0.2 0.06 0.8

0.1

259 323 60 117 78 160

(26%) (32%) (6%) (12%) (8%) (16%)

0.38 0.36 0.38 0.35 0.33 0.34

(0.21, (0.15, (0.15, (0.15, (0.05, (0.13,

Ref.f 0.8 1.0 0.8 0.7 0.7

Energy intake (kJ) 3459–7973 7974–9616 9620–1 1423 1 1440–1 6731

0.33 0.21 0.28 0.24

(0.13, (0.05, (0.10, (0.07,

Ref.f 0.2 0.6 0.4

0.6

250 249 249 249

(25%) (25%) (25%) (25%)

0.20 0.29 0.19 0.15

(  0.10, 0.49) (0.03, 0.55) (  0.07, 0.45) (  0.11, 0.40)

0.38 (0.22, 0.54) 0.16 (  0.05, 0.36) 0.26 (0.07, 0.44)

Ref.f 0.03 0.3

0.08

545 (49%) 111 (10%) 448 (41%)

0.25 0.28 0.33 0.32 0.17 0.24

Ref.f 0.8 0.5 0.6 0.5 0.9

0.5

62 65 313 253 145 114

Recreational physical activity Sedentary 483 (27%) Low 649 (37%) Moderate 447 (25%) High 191 (111%) Smoking status Life-long non smoker Ex-smoker Smoker, 1–7 pkyrsg Smoker,47 pkyrsg

Current use of HRT No Yes Unknown Paritye Nulliparous 1 child 2 children 3 children 4 children 5–7 children

1218 450 82 75

424 424 424 424

(67%) (25%) (4%) (4%)

(25%) (25%) (25%) (25%)

1047 (57%) 223 (12%) 560 (31%) 108 112 513 422 236 184

(7%) (7%) (32%) (27%) (15%) (12%)

0.34 0.42 0.33  0.03

(0.16, (0.16, (0.07, (0.02,

0.55) 0.52) 0.40) 0.47)

0.53) 0.38) 0.46) 0.42)

(0.01, 0.49) (0.02, 0.55) (0.17, 0.50) (0.16, 0.47) (  0.01, 0.35) (0.06, 0.42)

(7%) (7%) (33%) (26%) (15%) (12%)

0.32 0.31 0.22 0.47 0.42

(0.18, 0.47) (0.16, 0.45) (  0.06, 0.51) (0.24, 0.69) (0.01, 0.84)

0.55) 0.56) 0.62) 0.56) 0.61) 0.54)

0.8

1.0

Ref.f 0.4 0.9 0.7

0.5

0.41 (0.25, 0.57) 0.22 (0.01, 0.44) 0.36 (0.23, 0.50)

Ref.f 0.1 0.6

0.3

0.30 0.47 0.34 0.45 0.15 0.29

Ref.f 0.4 0.7 0.3 0.2 0.9

0.04

(0.04, 0.56) (0.16, 0.78) (0.21, 0.47) (0.30, 0.60) (  0.01, 0.30) (0.10, 0.47)

Abbreviation: HRT, hormone replacement therapy. n ¼ 1830 observations across 810 individuals for weight analysis and 1104 observations across 749 individuals for waist circumference analysis. aFor weight analysis: adjusted for age, medical condition, alcohol consumption, smoking status, current use of HRT and parity; For waist circumference analysis: adjusted for age, baseline waist circumference, medical condition, current use of HRT and parity. bValues are means±95% CI, derived from the interaction between covariates and time using generalized estimating equation (GEE) and data collected in 1992 and 2007. c Comparison between categories. P-values from Wald test based on parameter estimate and standard error from GEE model. dP-value from likelihood ratio test for interaction of covariate by time. eTime-independent variable. fReference. gPack-years smoked. hModerate p20 g per day and heavy 420 g per day in women.

& 2014 Macmillan Publishers Limited

European Journal of Clinical Nutrition (2014) 309 – 315

Weight and waist circumference change S Arabshahi et al

314 semi-rural community may have varied between men and women, though we cannot confirm that from our data. Nonetheless, our data are in agreement with the notion that increased physical activity is one of the main strategies to improve anthropometric characteristics of young and middle-aged adults. Gain in WC was on average 2.5 times lower in highly physically active women compared to sedentary women. Given the known strong association between increased WC and risk of cardio-metabolic disease (including heart disease, stroke, diabetes, hypertension and some types of cancer),12 our results show that while age plays a driving role in increasing the obesity epidemic, involvement in physical activities can significantly reduce the age-related WC gain and thus the risk of cardio-metabolic disease, especially in women. Our results of the association between parity and WC change but not with weight change provide longitudinal evidence on earlier reports from cross-sectional studies, using skinfolds measurements47 and body scans,48 suggesting that parity is associated with a shift in fat distribution independent of total weight in women. However, our study also showed significant age-associated WC gain in both younger and older women up to 64 years of age. In agreement with previous findings,48 gain in WC was also evident in nulliparous women, thus they cannot be assigned to pregnancy physiology alone and suggest a shift of fat tissue from the lower to the upper body that happens during the adult life-span and accelerates while aging.48 It appears that parity can accelerate the reallocation of adipose tissue. None of the other variables were associated with change in weight or WC at the overall group level. Factors such as socioeconomic status, smoking and drinking behaviors vary between populations and some heterogeneity in findings is expected. In contrast to others, a major strength of our investigation is that we were able to allow for change in predictor variables over time. This was possible due to the repeated assessment of many of the variables (e.g., changes in the amount of alcohol consumed over time). This would have removed some of the biases that may arise if only a one-off assessment of health behaviors is used to ascertain its association with anthropometric changes over time, and hence is a superior method of modelling.49 Unexpectedly, total energy intake was not related to change in the anthropometric measures. The problem of obesity-related misreporting of energy intake in dietary assessment has been well documented.50 Results did not appreciably change (data not shown) when we excluded under-reporters (16%) based on the method by Black and Goldberg,50 suggesting that underreporting of energy intake is an unlikely explanation for the lack of energyweight change association as previously shown in our study population.51 Nambour is a semi-rural community but we expect that this has had a minimal effect on the generalizability of our findings to the wider Australian population and beyond. The majority of Australians live in urban environments, and in terms of work, housing and mobility, the majority of our study participants would have resembled an urban pattern of behaviors more than those in a rural setting. Our findings are in agreement with national crosssectional data on body weight over the past 15 years, showing an average weight gain of 3.9 kg in men and 4.1 kg in women between 1995 and 2011–1226 compared to 3.2 kg in men and 4.8 kg in women over the 15 years in our study. The relative distribution of physical activity levels in our study population was also comparable to that reported in national surveys.41 This study has several strengths. Internationally it is one of few to measure longitudinal change in general and abdominal obesity in a population-based sample. The analytical design allowed us to investigate concurrent changes in predictive factors and body measures over time. Only measured anthropometric data was used, thus eliminating systematic bias due to self-reported body size. European Journal of Clinical Nutrition (2014) 309 – 315

It is important to interpret these results in light of some limitations. In longitudinal analyses, loss to follow-up is a matter of concern. However, a main advantage of the applied generalized estimating equations statistical model is that all available data over the follow-up period could be used. Our findings showed that the small number of study participants who were excluded from the analyses were more likely to be in the youngest or oldest age-categories. The effect of age on weight change may have been different between those included and those not included. However, inclusion of age as a covariate in all our models is expected to have avoided some of this possible bias. To further address the issue of possible bias due to missing data, we repeated the analyses in the subgroup of participants who had complete data on body measures over time. The results and the magnitude of the overall associations from these additional analyses were not materially different from those presented in the tables (data not shown). To further explore our data, we investigated possible effect modification by assessing whether the association between age and change in weight (or WC) was modified by baseline BMI (or baseline abdominal obesity), and whether the association between energy intake and these outcomes was modified by physical activity. No significant associations were observed (results not shown). It remains possible that change in weight or WC may have caused change in some participants’ characteristics. Such reverse causation remains a possibility in these longitudinal analyses and needs to be considered in interpreting results. In conclusion, this study confirms a birth cohort effect on body weight, with later-born cohorts being heavier. Our findings add longitudinal evidence to the knowledge that gain in anthropometric measures is a widespread problem affecting both sexes. Age is the most important factor affecting change in weight and WC in both men and women. However, engaging in recreational physical activity can prevent WC gain, especially in women.

CONFLICT OF INTEREST The authors declare no conflict of interest.

ACKNOWLEDGEMENTS This work was supported by National Health and Medical Research Council of Australia (data collection). We thank the Nambour Skin Cancer Study participants for their long-term participation in this study. We also thank A/Prof Geoff Marks, Mrs Maria Celia Hughes, Mr Bob Hughes, and Dr Torukiri Ibiebele for their contributions to collection of dietary and anthropometric data and data preparation, and Professor Ade`le Green for making the Nambour Study data available for these analyses.

REFERENCES 1 WHO. Obesity and overweight World Health Organisation. 2012 [23 October 2012]. Available from: http://www.who.int/mediacentre/factsheets/fs311/en/. 2 Abdullah A, Wolfe R, Stoelwinder JU, de Courten M, Stevenson C, Walls HL et al. The number of years lived with obesity and the risk of all-cause and cause-specific mortality. Int J Epidemiol 2011; 40: 985–996. 3 Chiolero A, Faeh D, Paccaud F, Cornuz J. Consequences of smoking for body weight, body fat distribution, and insulin resistance. Am J Clin Nutr 2008; 87: 801–809. 4 Kahn HS, Tatham LM, Rodriguez C, Calle EE, Thun MJ, Heath CW Jr. Stable behaviors associated with adults’ 10-year change in body mass index and likelihood of gain at the waist. Am J Public Health 1997; 87: 747–754. 5 Bernaards CM, Twisk JW, Snel J, van Mechelen W, Kemper HC. In a prospective study in young people, associations between changes in smoking behavior and risk factors for cardiovascular disease were complex. J Clin Epidemiol 2005; 58: 1165–1171. 6 Colditz GA, Segal MR, Myers AH, Stampfer M, Willett WC, Speizer FE. Weight change in relation to smoking cessation in women. J Smoking Relat Dis 1992; 3: 145–153.

& 2014 Macmillan Publishers Limited

Weight and waist circumference change S Arabshahi et al

315 7 Wannamethee SG, Shaper AG. Alcohol, body weight, and weight gain in middle-aged men. Am J Clin Nutr 2003; 77: 1312–1317. 8 Wannamethee SG, Field AE, Colditz GA, Rimm EB. Alcohol intake and 8-year weight gain in women: a prospective study. Obes Res 2004; 12: 1386–1396. 9 Liu S, Serdula MK, Williamson DF, Mokdad AH, Byers T. A prospective study of alcohol intake and change in body weight among US adults. Am J Epidemiol 1994; 140: 912–920. 10 Koh-Banerjee P, Chu NF, Spiegelman D, Rosner B, Colditz G, Willett W et al. Prospective study of the association of changes in dietary intake, physical activity, alcohol consumption, and smoking with 9-y gain in waist circumference among 16 587 US men. Am J Clin Nutr 2003; 78: 719–727. 11 Halkjaer J, Tjonneland A, Thomsen BL, Overvad K, Sorensen TI. Intake of macronutrients as predictors of 5-y changes in waist circumference. Am J Clin Nutr 2006; 84: 789–797. 12 Klein S, Allison DB, Heymsfield SB, Kelley DE, Leibel RL, Nonas C et al. Waist circumference and cardiometabolic risk: a consensus statement from shaping America’s health: association for weight management and obesity prevention; NAASO, the obesity society; the American Society for Nutrition; and the American Diabetes Association. Obesity (Silver Spring) 2007; 15: 1061–1067. 13 Tolstrup JS, Halkjaer J, Heitmann BL, Tjonneland AM, Overvad K, Sorensen TI et al. Alcohol drinking frequency in relation to subsequent changes in waist circumference. Am J Clin Nutr 2008; 87: 957–963. 14 Allman-Farinelli MA, Chey T, Bauman AE, Gill T, James WP. Age, period and birth cohort effects on prevalence of overweight and obesity in Australian adults from 1990 to 2000. Eur J Clin Nutr 2007; 62: 898–907. 15 Olsen LW, Baker JL, Holst C. Sorensen TIA. Birth cohort effect on the obesity epidemic in Denmark. Epidemiology 2006; 17: 292–295. 16 Jacobsen BK, Njolstad I, Thune I, Wilsgaard T, Lochen ML, Schirmer H. Increase in weight in all birth cohorts in a general population: The Tromso Study, 1974-1994. Arch Intern Med 2001; 161: 466–472. 17 Schroder H, Elosua R, Vila J, Marti H, Covas MI, Marrugat J. Secular trends of obesity and cardiovascular risk factors in a Mediterranean population. Obesity 2007; 15: 557–562. 18 Green A, Williams G, Neale R, Hart V, Leslie D, Parsons P et al. Daily sunscreen application and betacarotene supplementation in prevention of basal-cell and squamous-cell carcinomas of the skin: a randomised controlled trial. Lancet 1999; 354: 723–729. 19 van der Pols JC, Heinen MM, Hughes MC, Ibiebele TI, Marks GC, Green AC. Serum antioxidants and skin cancer risk: an 8-year community-based follow-up study. Cancer Epidemiol Biomarkers Prev 2009; 18: 1167–1173. 20 Commonwealth Electoral Act 1918. Part XVI. The polling, Section 245 Compulsory voting. 19 April 2011. Report No.: 1918. Contract No.: Act No. 27. 21 Australian Bureau of Statistics. Australian Standard Classification of Occupations (ASCO). Second Edition ABS: Canberra, 1997, 1220.0. 22 Marks GC, Hughes MC, van der Pols JC. Relative validity of food intake estimates using a food frequency questionnaire is associated with sex, age, and other personal characteristics. J Nutr 2006; 136: 459–465. 23 National Health and Medical Research Council. Australian Alcohol Guidelines. Health, Risks and Benefits. NHMRC, Commonwealth of Australia: Canberra, 2001, 1864961546. 24 Zeger SL, Liang KY, Albert PS. Models for longitudinal data: a generalized estimating equation approach. Biometrics 1988; 44: 1049–1060. 25 Fitzmaurice GM, Laird NM, Ware JH. Applied Longitudinal Analysis. Wiley-Interscience: Hoboken, N.J., 2004, Xix, p 506. 26 ABS. Australian Health Survey: First Results, 2011–12. Contract No.: 4364.0.55.001. Australian Bureau of Statistics: Canberra, 2012. 27 Elia M. Obesity in the elderly. Obes Res 2001; 9(Suppl 4): 244S–248SS. 28 Walls HL, Magliano DJ, McNeil JJ, Stevenson C, Ademi Z, Shaw J et al. Predictors of increasing waist circumference in an Australian population. Public Health Nutr 2011; 14: 870–881. 29 Gordon-Larsen P, Hou N, Sidney S, Sternfeld B, Lewis CE, Jacobs Jr DR et al. Fifteen-year longitudinal trends in walking patterns and their impact on weight change. Am J Clin Nutr 2009; 89: 19–26.

30 Toth MJ, Tchernof A, Sites CK, Poehlman ET. Menopause-related changes in body fat distribution. Ann N Y Acad Sci 2000; 904: 502–506. 31 Williams PT, Wood PD. The effects of changing exercise levels on weight and age-related weight gain. Int J Obes 2006; 30: 543–551. 32 Barr ELM, Magliano DJ, Zimmet PZ, Polkinghorne KR, Atkins RC, Dunstan DW et al. AusDiab 2005, the Australian Diabetes, Obesity and Lifestyle Study. Tracking the Accelerating Epidemic: Its Causes and Outcomes. International Diabetes Institute: Melbourne, 2006. 33 Crombie IK, Irvine L, Elliott L, Wallace H. Targets to tackle the obesity epidemic: a review of twelve developed countries. Public Health Nutr 2009; 12: 406–413. 34 Droyvold WB, Nilsen TI, Kruger O, Holmen TL, Krokstad S, Midthjell K et al. Change in height, weight and body mass index: Longitudinal data from the HUNT Study in Norway. Int J Obes 2006; 30: 935–939. 35 Shatenstein B, Kergoat MJ, Nadon S. Anthropometric changes over 5 years in elderly Canadians by age, gender, and cognitive status. J Gerontol A Biol Sci Med Sci 2001; 56: M483–M488. 36 Forbes GB. Longitudinal changes in adult fat-free mass: influence of body weight. Am J Clin Nutr 1999; 70: 1025–1031. 37 Baumgartner RN. Body composition in healthy aging. Ann N Y Acad Sci 2000; 904: 437–448. 38 Iannuzzi-Sucich M, Prestwood KM, Kenny AM. Prevalence of sarcopenia and predictors of skeletal muscle mass in healthy, older men and women. J Gerontol A Biol Sci Med Sci 2002; 57: M772–M777. 39 Patel HP, Syddall HE, Jameson K, Robinson S, Denison H, Roberts HC et al. Prevalence of sarcopenia in community-dwelling older people in the UK using the European Working Group on Sarcopenia in Older People (EWGSOP) definition: findings from the Hertfordshire Cohort Study (HCS). Age Ageing 2013; 42: 378–384. 40 Ni Mhurchu C, Rodgers A, Pan WH, Gu DF, Woodward M. Body mass index and cardiovascular disease in the Asia-Pacific Region: an overview of 33 cohorts involving 310 000 participants. Int J Epidemiol 2004; 33: 751–758. 41 Berrington de Gonzalez A, Hartge P, Cerhan JR, Flint AJ, Hannan L, MacInnis RJ et al. Body-mass index and mortality among 1.46 million white adults. N Engl J Med 2010; 363: 2211–2219. 42 Walker M, Wannamethee G, Whincup PH, Shaper AG. Weight change and risk of heart attack in middle-aged British men. Int J Epidemiol 1995; 24: 694–703. 43 Galanis DJ, Harris T, Sharp DS, Petrovitch H. Relative weight, weight change, and risk of coronary heart disease in the Honolulu Heart Program. Am J Epidemiol 1998; 147: 379–386. 44 Vergnaud AC, Estaquio C, Czernichow S, Peneau S, Hercberg S, Galan P et al. Energy density and 6-year anthropometric changes in a middle-aged adult cohort. Br J Nutr 2009; 102: 302–309. 45 Wardle J, Boniface D. Changes in the distributions of body mass index and waist circumference in English adults, 1993/1994 to 2002/2003. Int J Obes 2008; 32: 527–532. 46 Wareham N. Physical activity and obesity prevention. Obes Rev 2007; 8: 109–114. 47 Lassek WD, Gaulin SJ. Changes in body fat distribution in relation to parity in American women: a covert form of maternal depletion. Am J Phys Anthropol 2006; 131: 295–302. 48 Wells JC, Griffin L, Treleaven P. Independent changes in female body shape with parity and age: A life-history approach to female adiposity. Am J Hum Biol 2010; 22: 456–462. 49 Twisk JWR. Applied Longitudinal Data Analysis for Epidemiology: A Practical Guide. Cambridge University Press: Cambridge, UK; New York, 2003, xvi, p 301. 50 Black AE. Critical evaluation of energy intake using the Goldberg cut-off for energy intake:basal metabolic rate. A practical guide to its calculation, use and limitations. Int J Obes Relat Metab Disord 2000; 24: 1119–1130. 51 Arabshahi S, van der Pols JC, Williams GM, Marks GC, Lahmann PH. Diet quality and change in anthropometric measures: 15-year longitudinal study in Australian adults. Br J Nutr 2011; 26: 1–10.

Supplementary Information accompanies this paper on the European Journal of Clinical Nutrition website (http://www.nature.com/ejcn)

& 2014 Macmillan Publishers Limited

European Journal of Clinical Nutrition (2014) 309 – 315