- Email: [email protected]
Cardiovascular and all-cause mortality in relation to various anthropometric measures of obesity in Europeans
Nutrition, Metabolism & Cardiovascular Diseases (2015) 25, 295e304
Available online at www.sciencedirect.com
Nutrition, Metabolism & Cardiovascular Diseases journal homepage: www.elsevier.com/locate/nmcd
Cardiovascular and all-cause mortality in relation to various anthropometric measures of obesity in Europeans X. Song a,b,*, P. Jousilahti b, C.D.A. Stehouwer c, S. Söderberg d,e, A. Onat f, T. Laatikainen b,g,h, J.S. Yudkin i, R. Dankner j,k, R. Morris i, J. Tuomilehto b,l,m, Q. Qiao a,b,n, for the DECODE Study Group1 a
Department of Public Health, Hjelt Institute, University of Helsinki, Helsinki, Finland Department of Chronic Disease Prevention, National Institute for Health and Welfare, Helsinki, Finland Department of Internal Medicine and Cardiovascular Research Institute Maastricht (CARIM), Maastricht University Medical Centre, Maastricht, The Netherlands d Department of Public Health and Clinical Medicine, Umeå University, Umeå, Sweden e Baker IDI Heart and Diabetes Institute, Melbourne, Australia f Department of Cardiology, Turkish Society of Cardiology Cerrahpas¸a Medical Faculty, Istanbul, Turkey g Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio, Finland h Hospital District of North Karelia, Joensuu, Finland i Department of Primary Care & Population Sciences, Royal Free and University College Medical School, London, UK j Unit for Cardiovascular Epidemiology, The Gertner Institute, Sheba Medical Center, Tel Hashomer, Israel k Division of Epidemiology and Prevention, School of Public Health, Sackler School of Medicine, Tel Aviv University, Tel Aviv, Israel l Center for Vascular Prevention, Danube University Krems, Krems, Austria m King Abdulaziz University, Jeddah, Saudi Arabia n R&D AstraZeneca AB, Mölndal, Sweden b c
Received 17 March 2014; received in revised form 14 August 2014; accepted 15 September 2014 Available online 20 September 2014
KEYWORDS Abdominal obesity; Cardiovascular mortality; All-cause mortality
Abstract Background and aims: Cardiovascular and all-cause mortality in relation to various anthropometric measures of obesity is still controversial. Methods and results: Body mass index (BMI), waist circumference (WC), waist-to-height ratio (WHtR), waist-to-hip ratio (WHR), A Body Shape Index (ABSI) and waist-to-hip-to-height ratio (WHHR) were measured at baseline in a cohort of 46,651 European men and women aged 24 e99 years. The relationship between anthropometric measures of obesity and mortality was evaluated by the Cox proportional hazards model with age as a time-scale and with threshold detected by a piecewise regression model. Over a median follow-up of 7.9 years, 2381 men and 1055 women died, 1071 men (45.0%) and 339 women (32.1%) from cardiovascular disease (CVD). BMI had a J-shaped relationship with CVD mortality, whereas anthropometric measures of abdominal obesity had positive linear relationships. BMI, WC and WHtR showed J-shaped associations with all-cause mortality, whereas WHR, ABSI and WHHR demonstrated positive linear relationships. Accordingly, a threshold value was detected at 29.29 and 30.98 kg/m2 for BMI, 96.4 and 93.3 cm for WC, 0.57 and 0.60 for WHtR, 0.0848 and 0.0813 m11/6 kg2/3 for ABSI with CVD mortality in men and women, respectively; 29.88 and 29.50 kg/m2 for BMI, 104.3 and 105.6 for WC, 0.61 and 0.67 for WHtR, 0.95 and 0.86 for WHR, 0.0807 and 0.0765 for ABSI in men and women, respectively, and 0.52 for WHHR in women with all-cause mortality. Conclusion: All anthropometric measures of abdominal obesity had positive linear associations with CVD mortality, whereas some showed linear and the others J-shaped relationships with
* Corresponding author. Department of Public Health, University of Helsinki, Mannerheimintie 172, PL41, FI-00014 Helsinki, Finland. Tel.: þ358 9 19127320. E-mail address: xin.song@helsinki.fi (X. Song). 1 See Appendix. http://dx.doi.org/10.1016/j.numecd.2014.09.004 0939-4753/ª 2014 Elsevier B.V. All rights reserved.
296
X. Song et al.
all-cause mortality. BMI had a J-shaped relationship with either CVD or all-cause mortality. Thresholds detected based on mortality may help with clinical definition of obesity in relation to mortality. ª 2014 Elsevier B.V. All rights reserved.
Introduction Obesity is a major risk factor for development of chronic diseases and an important cause of mortality [1,2]. Waist circumference (WC) and waist-to-height ratio (WHtR) appear to be better anthropometric measures of abdominal obesity than body mass index (BMI) and have stronger correlation with intra-abdominal fat content and cardiometabolic risk factors [3]. But to date, the association of anthropometric measures of obesity with all-cause mortality is still controversial: a J- or U-shaped [2,4e7], or a positive linear [8,9] relationship for BMI, a J-shaped [2,10] or a positive linear [10e12] relationship for WC, a positive linear [11,12] or a U-shaped [2] relationship for waist-tohip ratio (WHR). A positive linear [4,6], a J- or U-shaped [5,11,12] relationship for BMI and a positive linear relationship [2,11,12] for these anthropometric measures of abdominal obesity with cardiovascular disease (CVD) mortality has been reported. In addition, A Body Shape Index (ABSI) has been proposed recently to combine WC, weight and height together in one algorithm to predict allcause mortality [7], as well as waist-to-hip-to-height ratio (WHHR) was shown to be superior to BMI, WC or WHtR in predicting CVD risk in the elderly [13]. Considering the inconsistent findings we set up the study, based on the data of the Diabetes Epidemiology: Collaborative analysis Of Diagnostic criteria in Europe (DECODE) study, to investigate the relationships of mortality from CVD and all-cause with various anthropometric measures of obesity, and to detect whether there are potential thresholds existing. ABSI and WHHR that were introduced recently are evaluated together with BMI, WC, WHtR and WHR.
The current data analysis comprised of 46,651 Caucasians (47.1% women) aged 24e99 at baseline, from twelve cohorts in four European countries (Finland, Sweden, Turkey and UK). Data collection included a self-reported questionnaire on smoking status and leisure-time physical activity, and a medical examination to measure weight, height, WC, and hip circumference, as described in detail in previous DECODE publications [14]. Participants with missing data on smoking status, leisure-time physical activity, weight, height, WC and hip circumference, without exact date of emigration, or completely lost to follow-up were excluded. Subjects who had emigrated but the dates of the emigration were recorded and treated as censored cases. Thus, the follow-up was complete. Definition of covariates
Methods
Smoking status at baseline was classified based on responses to the questionnaire into three categories of never smokers, former smokers and current smokers. Reading, watching TV, house works, sewing and walking <1 km daily that do not require moving much and do not physically tax was defined as leisure-time physically inactive, and all other higher levels of physical activity were defined as physically active. Height and weight were measured without shoes and with light clothing. WC was measured midway between the lower rib margin and iliac crest. Hip circumference was measured at the level of widest circumference over the greater trochanters. BMI was calculated as weight in kilograms divided by the square of height in meters. WHtR, WHR, or WHHR was calculated as WC divided by height, hip circumference, or both in meters. The calculation of ABSI was based on WC adjusting for weight and height, which was defined as follows: ABSI Z WC height5/6 weight2/3 [7].
Study population
Definition of fatal events
The DECODE collaboration includes 40 studies and their investigators from 14 European countries who have conducted population-based or large occupational surveys for diabetes and its risk factors applying standard 75 g oral glucose tolerance tests for diagnosis of diabetes [14]. All survey participants included in the data analysis are Caucasians. Individual participant data from each cohort was sent to the National Institute for Health and Welfare in Helsinki, Finland for collaborative data analyses. Each study was approved by the local ethics committees, and the analysis plan was approved by the ethics committee of the National Institute for Health and Welfare, Helsinki, Finland.
Median follow-up time varied between 2.5 and 21.8 years between different cohorts. Vital status and causes of death were recorded in all of the studies included. CVD mortality was defined according to the International Classification of Disease codes 401e448 (9th revision) or I10eI79 (10th revision). Statistical analyses The Cox proportional hazards model was used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs) of CVD or all-cause mortality with anthropometric measures of obesity. Analyses were adjusting for baseline
Study
Finland FINRISK (1987) FINRISK (1992) FINRISK (1997) FINRISK (2002) Sweden Uppsala (1991e1995) Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Turkey TARFS (1998e2002) UK Whitehall II (1991e1993) Total
(1986) (1990) (1994) (1999) (2004)
BMI (kg/m2)
WC (cm)
WHtR
WHR
ABSI (m11/6 kg2/3)
WHHR (m1)
No. of participants
Age (years)
Median follow-up (years)
No. of deaths CVD
All-cause
2541 2570 3788 3808
43.8 44.3 48.6 48.0
(11.2) (11.2) (13.5) (13.0)
26.66 26.57 26.87 27.23
(3.71) (3.87) (3.94) (4.12)
92.2 93.9 94.5 95.4
(10.7) (11.2) (11.3) (11.8)
0.53 0.53 0.54 0.54
(0.06) (0.07) (0.07) (0.07)
0.90 0.92 0.93 0.97
(0.06) (0.07) (0.07) (0.07)
0.0784 0.0796 0.0796 0.0796
(0.0041) (0.0040) (0.0040) (0.0041)
0.52 0.52 0.53 0.55
(0.04) (0.05) (0.05) (0.05)
21.8 16.8 11.8 6.8
293 152 205 69
561 340 418 162
1057 671 761 877 869 864
71.0 46.0 45.0 49.6 50.6 50.9
(0.6) (11.4) (11.3) (14.0) (14.3) (14.1)
26.19 25.42 25.84 26.22 26.67 27.19
(3.38) (3.43) (3.35) (3.65) (3.49) (3.95)
94.4 92.4 91.4 93.4 95.3 96.4
(9.4) (9.2) (9.1) (10.1) (9.7) (10.7)
0.54 0.52 0.52 0.53 0.54 0.54
(0.05) (0.06) (0.05) (0.06) (0.06) (0.06)
0.94 0.94 0.93 0.94 0.92 0.96
(0.05) (0.05) (0.06) (0.06) (0.07) (0.06)
0.0811 0.0806 0.0789 0.0799 0.0805 0.0803
(0.0037) (0.0032) (0.0036) (0.0040) (0.0039) (0.0036)
0.54 0.54 0.53 0.53 0.52 0.54
(0.04) (0.04) (0.04) (0.04) (0.05) (0.04)
10.0 20.5 16.5 12.5 7.4 2.5
126 49 30 37 14 0
276 131 79 126 56 7
1580
53.2 (12.4)
26.41 (4.01)
94.3 (11.0)
0.56 (0.06)
0.93 (0.07)
0.0819 (0.0052)
0.55 (0.05)
7.9
55
109
5300 24 686
49.3 (6.0) 49.0 (12.4)
25.10 (3.15) 26.34 (3.78)
87.4 (9.2) 92.7 (11.0)
0.50 (0.05) 0.53 (0.06)
0.90 (0.06) 0.93 (0.07)
0.0768 (0.0035) 0.0792 (0.0042)
0.51 (0.04) 0.53 (0.05)
5.9 7.9
41 1071
116 2381
Anthropometric measures of obesity and mortality
Table 1 Baseline characteristics and the data of follow-up of the survey in men.
Abbreviations: BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; WHR, waist-to-hip ratio; ABSI, A Body Shape Index; WHHR, waist-to-hip-to-height ratio; CVD, cardiovascular disease. Data are means (standard deviations) or as noted.
297
Abbreviations: BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; WHR, waist-to-hip ratio; ABSI, A Body Shape Index; WHHR, waist-to-hip-to-height ratio; CVD, cardiovascular disease. Data are means (standard deviations) or as noted.
39 1055 6 339 5.8 11.8 0.48 (0.05) 0.50 (0.05) 0.0683 (0.0052) 0.0733 (0.0054) 0.77 (0.07) 0.81 (0.07) 0.47 (0.07) 0.51 (0.08) 75.5 (11.7) 82.0 (12.6) 50.2 (6.1) 46.9 (12.3) 2346 21 965
25.66 (4.69) 26.19 (5.00)
56 35 7.9 0.54 (0.06) 0.0778 (0.0072) 0.84 (0.08) 0.58 (0.09) 90.7 (12.7) 52.7 (12.3) 1619
28.79 (5.65)
75 51 68 27 5 18 12 16 3 0 20.5 16.5 12.5 7.5 2.5 (0.05) (0.04) (0.06) (0.05) (0.05) 0.53 0.50 0.51 0.50 0.52 (0.0054) (0.0042) (0.0060) (0.0047) (0.0045) 0.0783 0.0729 0.0757 0.0753 0.0761 (0.07) (0.06) (0.08) (0.07) (0.07) 0.86 0.81 0.83 0.82 0.85 (0.08) (0.07) (0.08) (0.08) (0.08) 0.52 0.49 0.52 0.52 0.53 (12.2) (11.0) (12.4) (11.8) (12.9) 85.3 79.4 84.2 84.9 86.6 (4.37) (4.39) (4.67) (4.63) (5.12) 25.01 24.99 25.82 26.35 26.62
Finland FINRISK (1987) FINRISK (1992) FINRISK (1997) FINRISK (2002) Sweden Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Turkey TARFS (1998e2002) UK Whitehall II (1991e1993) Total
(1986) (1990) (1994) (1999) (2004)
(11.1) (11.4) (14.0) (14.1) (13.9) 45.6 44.8 49.4 50.1 49.7 685 793 902 900 909
310 190 176 58 119 55 61 14 21.8 16.9 11.8 6.8 (0.05) (0.05) (0.05) (0.05) 0.48 0.49 0.49 0.52 (0.0045) (0.0043) (0.0042) (0.0042) 0.0714 0.0724 0.0726 0.0742 (0.06) (0.07) (0.07) (0.06) 0.78 0.79 0.80 0.84 (0.07) (0.08) (0.08) (0.08) 0.49 0.49 0.50 0.52 (11.2) (11.7) (12.2) (12.6) 79.4 80.2 81.3 83.6 (4.90) (4.92) (4.95) (5.05) 26.04 25.70 26.07 26.39 (11.4) (11.5) (12.7) (13.0) 43.7 44.0 46.1 46.6 2812 2828 3788 4383
All-cause CVD
Median follow-up (years) WHHR (m1) ABSI (m11/6 kg2/3) WHR WHtR WC (cm) BMI (kg/m2) Age (years) No. of participants
Tables 1 and 2 provide the descriptive characteristics of the cohorts. Over a median follow-up of 7.9 years, 2381 men and 1055 women died, 1071 men (45.0%) and 339 women (32.1%) from CVD. Table 3 shows that age, high distribution of anthropometric measures of obesity, smoking and leisure-time physical inactivity were significantly associated with CVD and all-cause mortality. The best-fitting conventional model was conventional polynomial model for BMI with CVD and all-cause mortality as well as for WC and WHtR with all-cause mortality in both genders (P < 0.05 for LRT and for deviance difference test against conventional linear model), which suggests a nonlinear relationship (Supplementary Table 1). However, the best-fitting conventional model was observed as conventional linear model for other anthropometric measures of abdominal obesity with CVD or allcause mortality, of which model fitness was not
Study
Results
Table 2 Baseline characteristics and the data of follow-up of the survey in women.
smoking status, leisure-time physical activity and cohort using attained age as the time scale. All analyses were performed separately for men and women. The shape of relationship between anthropometric measures of obesity and mortality was explored using both parametric models (conventional linear or polynomial models) and nonparametric models (the linear or restricted cubic spline regression model). Akaike’s information criterion (AIC) was used to judge the model fitness between conventional linear model and polynomial models (including quadratic, cubic or factional polynomial model), the lower the AIC value the better the model fitness is. The reduction of AIC between nested models is evaluated by the likelihood ratio test (LRT) or a deviance difference test [15,16]. Nonparametric smooth functions using the restricted cubic spline or the linear spline corresponding to the best-fitting conventional polynomial or linear model were also fitted to model their potentially nonlinear relationship [17]. Further, AIC difference 4 was considered as considerably less supported relative to the lowest AIC value between non-nested models [18]. The existence of threshold was estimated by a piecewise regression model using nonlinear least-squares estimation incorporating the “nl” function in Stata, with the lowest value of residual sum of squares, root-mean-square error and AIC, whereas the highest value of R-squared as the criteria for selecting optimal model. Analyses of threshold detection for certain anthropometric measures of obesity were also carried out on the data set with outliers excluded. Outliers are defined to be values of anthropometric measures of obesity that are more or less than three median absolute deviations away from the median [19]. In the Cox proportional hazards model, anthropometric measure of obesity was entered as a continuous variable, and factors including leisure-time physical activity and smoking status as categorical variables, along with a BMI-leisure-time physical activity or a BMI-smoking status interaction term. Data analyses were performed in Stata Intercooled 11.2 (StataCorp, College Station, TX).
X. Song et al.
No. of deaths
298
Anthropometric measures of obesity and mortality
299
significantly improved by conventional polynomial model (P > 0.05 for LRT or for deviance difference test against conventional linear model), which indicates a linear relationship. Further, among the best-fitting models, most anthropometric measures of abdominal obesity had lower AIC values than BMI except for ABSI in women. The nonlinear and linear relationships detected by the parametric conventional modeling were further supported by the nonparametric modeling (AIC difference <4). The spline regression analysis showed that most simple model (K Z 3 for the restricted cubic spline regression or K Z 1 for the linear spline regression) had the lowest AIC value compared to the models with more parameters. HRs with 95% CIs for mortality estimated by the best restricted cubic or linear spline regression model were plotted in Figs. 1 and 2 against the whole distribution of an anthropometric measure as compared with the mean value. HR for BMI with CVD mortality, as well as for BMI, WC and WHtR with all-cause mortality in both genders decreased first at the lower, then increased gradually at the middle, and increased more at the upper distributions of these anthropometric measures of obesity, which
indicates a J-shaped relationship with two potential thresholds existing (Figs. 1 a and g, and 2 aec and gei). A threshold value was detected at 29.29 and 30.98 kg/m2 for BMI with CVD mortality in men and women, respectively; 29.88 and 29.50 kg/m2 for BMI, 104.3 and 105.6 cm for WC, 0.61 and 0.67 for WHtR with all-cause mortality in men and women, respectively (Supplementary Table 2). HR for CVD or all-cause mortality in both genders increased positively with increasing in anthropometric measures of abdominal obesity (Figs. 1 bef and hel, and 2 def and jel), with a threshold detected at 96.4 and 93.3 cm for WC, 0.57 and 0.60 for WHtR, 0.0848 and 0.0813 m11/ 6 kg2/3 for ABSI in men and women, respectively, for CVD mortality; 0.95 and 0.86 for WHR, 0.0807 and 0.0765 for ABSI in men and women, respectively, and 0.52 for WHHR in women for all-cause mortality (Figs. 1 b, c, e, h, i and k, and 2 d, e and jel, and Supplementary Table 2). The results were not substantially altered after excluding the first five years of follow-up (Supplementary Table 3), or additionally adjusting for one or more other CVD risk factors, including hypertension, diabetes or dyslipidemia (results not shown). There was no interaction between an
Table 3 Baseline characteristics of participants according to cardiovascular disease and all-cause mortality.a Cardiovascular disease mortality
All-cause mortality
No
Yes
No
Yes
1071 61.1 (60.1e62.0)b 27.2 (27.0e27.5)b 96.0 (95.3e96.6)b 0.55 (0.55e0.56)b 0.94 (0.94e0.94)b 0.0805 (0.0803e0.0808)b 0.54 (0.54e0.54)b
22,305 48.9 (48.7e49.0) 26.3 (26.3e26.4) 92.5 (92.4e92.7) 0.53 (0.53e0.53) 0.93 (0.93e0.93) 0.0790 (0.0790e0.0791) 0.53 (0.53e0.53)
2381 59.8 (59.0e60.7)b 26.7 (26.5e26.8)b 94.4 (94.0e94.9)b 0.54 (0.54e0.54)b 0.94 (0.93e0.94)b 0.0803 (0.0801e0.0804)b 0.54 (0.54e0.54)b
69.6 30.4b
81.7 18.3
71.9 28.1b
31.1 31.9 37.0b
45.5 31.3 23.2
32.7 30.0 37.3b
339 61.4 (59.3e63.6)b 27.7 (27.2e28.2)b 86.2 (85.0e87.5)b 0.54 (0.53e0.55)b 0.83 (0.82e0.84)b 0.0748 (0.0742e0.0753)b 0.52 (0.51e0.52)b
20,910 46.9 (46.7e47.1) 26.2 (26.1e26.2) 81.9 (81.7e82.0) 0.51 (0.51e0.51) 0.81 (0.81e0.81) 0.0732 (0.0731e0.0733) 0.50 (0.50e0.50)
1055 59.2 (57.9e60.4)b 26.8 (26.5e27.1)b 84.0 (83.3e84.7)b 0.52 (0.52e0.53)b 0.82 (0.82e0.82)b 0.0742 (0.0739e0.0746)b 0.51 (0.51e0.51)b
56.3 43.7b
76.5 23.5
63.7 36.3b
64.6 11.2 24.2b
62.9 18.3 18.8
60.9 13.6 25.5b
Men Number 23,615 Age (years) 49.5 (49.3e49.7) BMI (kg/m2) 26.3 (26.3e26.4) WC (cm) 92.6 (92.4e92.7) WHtR 0.53 (0.53e0.53) WHR 0.93 (0.93e0.93) ABSI (m11/6 kg2/3) 0.0791 (0.0790e0.0791) WHHR (m1) 0.53 (0.53e0.53) Leisure-time physical activity, % Physically active 81.3 Physically inactive 18.7 Smoking status, % Never smokers 44.9 Former smokers 31.2 Current smokers 23.9 Women Number 21,626 Age (years) 47.3 (47.1e47.5) BMI (kg/m2) 26.2 (26.1e26.2) WC (cm) 81.9 (81.8e82.1) WHtR 0.51 (0.51e0.51) WHR 0.81 (0.81e0.81) ABSI (m11/6 kg2/3) 0.0732 (0.0732e0.0733) WHHR (m1) 0.50 (0.50e0.50) Leisure-time physical activity, % Physically active 76.2 Physically inactive 23.8 Smoking status, % Never smokers 62.8 Former smokers 18.2 Current smokers 19.1
Abbreviations: BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; WHR, waist-to-hip ratio; ABSI, A Body Shape Index; WHHR, waist-to-hip-to-height ratio. a Data are age-adjusted means (95% confidence intervals) for obesity indicators or as noted. b P < 0.001 for the difference between yes and no.
300
X. Song et al.
Anthropometric measures of obesity and mortality
anthropometric measure and either smoking status or leisure-time physical activity observed.
Discussion We found that BMI had a J-shaped relationship with CVD mortality, whereas anthropometric measures of abdominal obesity had positive linear relationships. BMI, WC and WHtR showed J-shaped associations with all-cause mortality, whereas WHR, ABSI and WHHR demonstrated positive linear relationships. Thresholds detected based on mortality may help with clinical definition of obesity in relation to mortality. Our findings are in line with previous studies that reported a J- or U-shaped relationship for BMI [2,4e7], a Jshaped relationship for WC [2,7,10] but a positive linear relationship for WHR [9,11,12] with all-cause mortality, and, a positive linear association between CVD mortality and all anthropometric measures of abdominal obesity [2,11,12]. Our previous DECODE study has shown that most anthropometric measures of abdominal obesity are superior to BMI in predicting CVD mortality except for ABSI [20]. The possible mechanisms for this more direct relation between anthropometric measures of abdominal obesity with CVD mortality might be that the intra-abdominal adipocytes are prone to empty their free fatty acids directly into the portal vein, exposing the liver to high concentrations of free fatty acids, which might lead to hyperinsulinaemia, dyslipidemia and hypertension [21]. In addition, adipose tissue, especially tissue from abdominal fat deposits, secretes factors (adipokines) that may promote development of chronic diseases [22]. The thresholds detected may have important clinical implications in the context of definition of obesity in relation to serious clinical outcomes, i.e., mortality. Current existing definitions for abdominal obesity among Caucasians were mostly focused on the WC measurement, at WC of 102 and 88 cm [23], or 94 and 80 cm [24], or at WHR of 0.95 and 0.80 [23], in men and women, respectively. These existing cut-off values have, however, been determined arbitrarily based on analysis of the trade-offs between sensitivity and specificity for discrimination of diabetes or metabolic syndrome [25]. Most of the previous studies were cross-sectional [23]. The WC or the WHR cut-off values proposed by the previous studies were, however, close to thresholds detected in our study for CVD or all-cause mortality, or included in the values between the first and the second thresholds for all-cause mortality. To apply the first threshold will give a high sensitivity and the second one a
301
high specificity depending on the purpose of the application of the thresholds. Considering our study comprising cohorts recruited most in 1990s, and the WC increasing with time, it might be speculated that the thresholds identified were lower than they are now. On the other hand, it can be argued that relationships between anthropometric measures of obesity and mortality may not be affected by changes in the population mean, which needs to be further investigated based on data collected recently. The thresholds detected for all-cause mortality reflects a mixed relationship between obesity and different causes of death. This might be affected by differences in constitutions of the causes of death across countries. Smokers tend to be lean but have high mortality rates [26]. Smoking has been suggested to be associated with visceral adipose fat accumulation, perhaps through simultaneously affecting lipoprotein lipase activity and increasing cortisol levels [27]. In our study, the relationship between anthropometric measures of obesity and mortality was not substantially altered by smoking status. In addition, limited evidence indicates that leisure-time physical inactivity might play an intermediate role in the relationship between obesity and mortality [28]. Our study did not detect significant interaction between obesity and leisure-time physical activity in relation to CVD and allcause mortality. Moreover, the obesity-CVD mortality relationships were not substantially altered after additional adjustment for one or more other CVD risk factors, including hypertension, diabetes or dyslipidemia, which could mediate a larger proportion of excess CVD risk [29]. Our study was based on several European populationor occupational-based studies, with sufficient power to investigate the association between anthropometric measures of obesity and mortality. In addition, the relationships between anthropometric measures of obesity and mortality were examined in the same cohort using both parametric and nonparametric approaches in parallel, which might better capture the true relationships. Yet, limitations of the study exist. Data are based on Caucasian European surveys, the results may not be applicable to other ethnic groups or races due to difference in percentage of body fat across ethnic groups [30]. The study does not have data on changes in anthropometric measurements before the baseline and during the follow-up which makes it impossible to exclude the possibility of ‘reverse causation’ [31], a consequence of underlying diseases before enrollment rather than a cause of death. The potential influence of reverse causality was checked by excluding the first five years of follow-up of which less than 7% of the study population and 25% of the mortality events were excluded, and the results were not altered.
Figure 1 Hazard ratio and 95% confidence intervals for anthropometric measures of obesity with cardiovascular disease (CVD) mortality among men and women. Dashed lines indicate hazard ratios and 95% confidence intervals (CIs) derived from the restricted cubic spline regression with three interior knots placed at the 10th, 50th, and 90th percentiles of body mass index (BMI, a and g), or from linear spline regression with one interior knot placed at the 50th percentiles of anthropometric measures of abdominal obesity (bef and hel), using age as the underlying time-scale in the Cox proportional hazards model and adjusting for smoking status, leisure-time physical activity and cohort, and mean of anthropometric measures of obesity was set as the reference value. Vertical dot line indicates the location which represent threshold (with 95% CIs denoted in subtitle) derived from the piecewise regression model followed by restricted cubic spline regression or linear spline regression. Analyses of thresholds for BMI in both genders were carried out on the data set with outliers excluded. Solid lines indicate slopes before and after thresholds.
Figure 2 Hazard ratios and 95% confidence intervals for anthropometric measures of obesity with all-cause mortality among men and women. Dashed lines indicate hazard ratios and their 95% confidence intervals derived from the restricted cubic spline regression with three interior knots placed at the 10th, 50th, and 90th percentiles of body mass index (BMI), waist circumference, and waist-to-height ratio (aec and gei), or from the linear spline regression with one interior knot placed at the 50th percentiles of waist-to-hip ratio, A Body Shape Index, and waist-to-hip-to-height ratio (def and jel), using the same method as described in Figure 1. Analyses of thresholds for BMI in women were carried out on the data set with outliers excluded.
Anthropometric measures of obesity and mortality
The relationship between certain anthropometric measures of abdominal obesity and mortality might be influenced by the potential correlation between some anthropometrical measures, such as hip and waist circumference [32]. Since this is a collaborative data analysis, certain lifestyle variables such as physical activity has been recorded differently in different studies. In spite of the efforts have been made to “harmonize” the variable and to adjust for studies in data analysis, discrepancies exist. A recent systematic review showed, however, that variations in anatomic locations of measurement of WC did not influence clinical outcomes regarding mortality from all-cause and CVD [33]. The discrepancies in study design and methodologies have been taken into account in our data analysis by fitting the study cohort into the models as a co-variable. In summary, all anthropometric measures of abdominal obesity had positive linear associations with CVD mortality, whereas some showed linear and the others J-shaped relationships with all-cause mortality. BMI had a J-shaped relationship with either CVD or all-cause mortality. Thresholds detected based on mortality may help with clinical definition of obesity in relation to mortality. Conflict of interest The authors declare no conflict of interest. Acknowledgments This work was supported by the Academy of Finland (Suomen Akatemia) [1129197, 136895 and 141005]. Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.numecd.2014.09.004. Appendix Studies and investigators in this collaborative study are: Finland National FINRISK 1987, 1992 and 1997 Cohorts: J. Tuomilehto1,2,3, P. Jousilahti1, J. Lindström1, 1. Department of Chronic Disease Prevention, National Institute for Health and Welfare, Helsinki; 2. Center for Vascular Prevention, Danube University Krems, Krems, Austria; 3 King Abdulaziz University, Jeddah, Saudi Arabia. National FINRISK 2002 Study: J. Tuomilehto1,2,3, T. Laatikainen1,4,5, M. Peltonen1, J. Lindström1, 1. Department of Chronic Disease Prevention, National Institute for Health and Welfare, Helsinki; 2. Center for Vascular Prevention, Danube University Krems, Krems, Austria; 3 King Abdulaziz University, Jeddah, Saudi Arabia; 4. Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio, Finalnd; 5. Hospital Distric of North Karelia, Joensuu, Finland. Sweden Northern Sweden MONICA Survey: S. Söderberg1,2, M. Eliasson1, 1. Department of Public Health and
303
Clinical Medicine, Umeå University, Umeå, Sweden; 2. Baker IDI Heart and Diabetes Institute, Melbourne, Australia. The Uppsala Longitudinal Study of Adult Men (ULSAM): B. Zethelius, Department of Public Health/Geriatrics, Uppsala University Hospital, Uppsala. Turkey Turkish Adult Risk Factor Study (TARFS): A Onat1,2, 1. Turkish Society of Cardiology, Istanbul; 2. Department of Cardiology, Cerrahpasa Medical Faculty, Istanbul University, Istanbul. United Kingdom Whitehall II Study: M.G.Marmot1, A.G. Tabák1,2, M. Kivimäki1,3, E.J. Brunner1, D.R. Witte1,4, 1. Department of Epidemiology and Public Health, University College London, London, UK; 2. Semmelweis University Faculty of Medicine, 1st Department of Medicine, Budapest, Hungary; 3. Finnish Institute of Occupational Health, Helsinki, Finland; 4. Steno Diabetes Center, Gentofte, Denmark. References [1] Obesity: preventing and managing the global epidemic. Report of a WHO consultation. World Health Organ Tech Rep Ser 2000;894: iexii. 1e253. [2] Pischon T, Boeing H, Hoffmann K, Bergmann M, Schulze MB, Overvad K, et al. General and abdominal adiposity and risk of death in Europe. N Engl J Med 2008;359:2105e20. [3] Browning LM, Hsieh SD, Ashwell M. A systematic review of waistto-height ratio as a screening tool for the prediction of cardiovascular disease and diabetes: 0.5 could be a suitable global boundary value. Nutr Res Rev 2010;23:247e69. [4] Kivimäki M, Ferrie JE, Batty GD, Davey Smith G, Elovainio M, Marmot MG, et al. Optimal form of operationalizing BMI in relation to all-cause and cause-specific mortality: the original Whitehall study. Obesity (Silver Spring) 2008;16:1926e32. [5] Klenk J, Nagel G, Ulmer H, Strasak A, Concin H, Diem G, et al. Body mass index and mortality: results of a cohort of 184,697 adults in Austria. Eur J Epidemiol 2009;24:83e91. [6] Prospective Studies Collaboration, Whitlock G, Lewington S, Sherliker P, Clarke R, Emberson J, Halsey J, et al. Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet 2009;373:1083e96. [7] Krakauer NY, Krakauer JC. A new body shape index predicts mortality hazard independently of body mass index. PLoS One 2012;7:e39504. [8] Katzmarzyk PT, Craig CL, Bouchard C. Adiposity, adipose tissue distribution and mortality rates in the Canada Fitness Survey follow-up study. Int J Obes Relat Metab Disord 2002;26:1054e9. [9] Lahmann PH, Lissner L, Gullberg B, Berglund G. A prospective study of adiposity and all-cause mortality: the Malmo Diet and Cancer Study. Obes Res 2002;10:361e9. [10] Koster A, Leitzmann MF, Schatzkin A, Mouw T, Adams KF, van Eijk JT, et al. Waist circumference and mortality. Am J Epidemiol 2008;167:1465e75. [11] Petursson H, Sigurdsson JA, Bengtsson C, Nilsen TI, Getz L. Body configuration as a predictor of mortality: comparison of five anthropometric measures in a 12 year follow-up of the Norwegian HUNT 2 study. PLoS One 2011;6:e26621. [12] Czernichow S, Kengne AP, Stamatakis E, Hamer M, Batty GD. Body mass index, waist circumference and waist-hip ratio: which is the better discriminator of cardiovascular disease mortality risk?: evidence from an individual-participant meta-analysis of 82 864 participants from nine cohort studies. Obes Rev 2011;12:680e7. [13] Carlsson AC, Risérus U, Engström G, Ärnlöv J, Melander O, Leander K, et al. Novel and established anthropometric measures and the prediction of incident cardiovascular disease: a cohort study. Int J Obes (Lond) 2013;37:1579e85. [14] Will new diagnostic criteria for diabetes mellitus change phenotype of patients with diabetes? Reanalysis of European
304
[15] [16] [17] [18]
[19]
[20]
[21]
[22] [23]
[24]
[25]
X. Song et al. epidemiological data. DECODE Study Group on behalf of the European Diabetes Epidemiology Study Group. BMJ 1998;317:371e5. Greenland S. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology 1995;6:356e65. Huelsenbeck JP, Hillis DM, Nielsen R. A likelihood-ratio test of monophyly. Syst Biol 1996;45:546e58. Patrick R, Willi S. Multivariable modelling with cubic regression splines: a principled approach. Stata J 2007;7:45e70. Burnham KP, Anderson DR. Model selection and multimodel inference: a practical information-theoretic approach. New York: Springer-Verlag; 2003. Leys C, Ley C, Klein O, Bernard P, Licata L. Detecting outliers: do not use standard deviation around the mean, use absolute deviation around the median. J Exp Soc Psychol 2013;49:764e6. Song X, Jousilahti P, Stehouwer CD, Söderberg S, Onat A, Laatikainen T, et al. Comparison of various surrogate obesity indicators as predictors of cardiovascular mortality in four European populations. Eur J Clin Nutr 2013;67:1298e302. Cameron AJ, Zimmet PZ. Expanding evidence for the multiple dangers of epidemic abdominal obesity. Circulation 2008;117: 1624e6. Haslam DW, James WP. Obesity. Lancet 2005;366:1197e209. Lean ME, Han TS, Morrison CE. Waist circumference as a measure for indicating need for weight management. BMJ 1995;311: 158e61. International Diabetes Federation (IDF). Worldwide definition of the metabolic syndrome. Brussels, Belgium: International Diabetes Federation; 2006. Waist circumference and waist-hip ratio: report of a WHO Expert Consultation., Geneva, 8e11 December 2008. Geneva: WHO; 2011.
[26] Koster A, Leitzmann MF, Schatzkin A, Adams KF, van Eijk JT, Hollenbeck AR, et al. The combined relations of adiposity and smoking on mortality. Am J Clin Nutr 2008;88:1206e12. [27] 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;87:801e9. [28] Hu G, Tuomilehto J, Silventoinen K, Barengo NC, Peltonen M, Jousilahti P. The effects of physical activity and body mass index on cardiovascular, cancer and all-cause mortality among 47 212 middle-aged Finnish men and women. Int J Obes (Lond) 2005;29: 894e902. [29] Global Burden of Metabolic Risk Factors for Chronic Diseases Collaboration (BMI Mediated Effects), Lu Y, Hajifathalian K, Ezzati M, Woodward M, Rimm EB, et al. Metabolic mediators of the effects of body-mass index, overweight, and obesity on coronary heart disease and stroke: a pooled analysis of 97 prospective cohorts with 1.8 million participants. Lancet 2014;383:970e83. [30] WHO Expert Consultation. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet 2004;363:157e63. [31] Willett WC, Dietz WH, Colditz GA. Guidelines for healthy weight. N Engl J Med 1999;341:427e34. [32] Cameron AJ, Magliano DJ, Shaw JE, Zimmet PZ, Carstensen B, Alberti KG, et al. The influence of hip circumference on the relationship between abdominal obesity and mortality. Int J Epidemiol 2012;41:484e94. [33] Ross R, Berentzen T, Bradshaw AJ, Janssen I, Kahn HS, Katzmarzyk PT, et al. Does the relationship between waist circumference, morbidity and mortality depend on measurement protocol for waist circumference? Obes Rev 2008;9:312e25.
Available online at www.sciencedirect.com
Nutrition, Metabolism & Cardiovascular Diseases journal homepage: www.elsevier.com/locate/nmcd
Cardiovascular and all-cause mortality in relation to various anthropometric measures of obesity in Europeans X. Song a,b,*, P. Jousilahti b, C.D.A. Stehouwer c, S. Söderberg d,e, A. Onat f, T. Laatikainen b,g,h, J.S. Yudkin i, R. Dankner j,k, R. Morris i, J. Tuomilehto b,l,m, Q. Qiao a,b,n, for the DECODE Study Group1 a
Department of Public Health, Hjelt Institute, University of Helsinki, Helsinki, Finland Department of Chronic Disease Prevention, National Institute for Health and Welfare, Helsinki, Finland Department of Internal Medicine and Cardiovascular Research Institute Maastricht (CARIM), Maastricht University Medical Centre, Maastricht, The Netherlands d Department of Public Health and Clinical Medicine, Umeå University, Umeå, Sweden e Baker IDI Heart and Diabetes Institute, Melbourne, Australia f Department of Cardiology, Turkish Society of Cardiology Cerrahpas¸a Medical Faculty, Istanbul, Turkey g Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio, Finland h Hospital District of North Karelia, Joensuu, Finland i Department of Primary Care & Population Sciences, Royal Free and University College Medical School, London, UK j Unit for Cardiovascular Epidemiology, The Gertner Institute, Sheba Medical Center, Tel Hashomer, Israel k Division of Epidemiology and Prevention, School of Public Health, Sackler School of Medicine, Tel Aviv University, Tel Aviv, Israel l Center for Vascular Prevention, Danube University Krems, Krems, Austria m King Abdulaziz University, Jeddah, Saudi Arabia n R&D AstraZeneca AB, Mölndal, Sweden b c
Received 17 March 2014; received in revised form 14 August 2014; accepted 15 September 2014 Available online 20 September 2014
KEYWORDS Abdominal obesity; Cardiovascular mortality; All-cause mortality
Abstract Background and aims: Cardiovascular and all-cause mortality in relation to various anthropometric measures of obesity is still controversial. Methods and results: Body mass index (BMI), waist circumference (WC), waist-to-height ratio (WHtR), waist-to-hip ratio (WHR), A Body Shape Index (ABSI) and waist-to-hip-to-height ratio (WHHR) were measured at baseline in a cohort of 46,651 European men and women aged 24 e99 years. The relationship between anthropometric measures of obesity and mortality was evaluated by the Cox proportional hazards model with age as a time-scale and with threshold detected by a piecewise regression model. Over a median follow-up of 7.9 years, 2381 men and 1055 women died, 1071 men (45.0%) and 339 women (32.1%) from cardiovascular disease (CVD). BMI had a J-shaped relationship with CVD mortality, whereas anthropometric measures of abdominal obesity had positive linear relationships. BMI, WC and WHtR showed J-shaped associations with all-cause mortality, whereas WHR, ABSI and WHHR demonstrated positive linear relationships. Accordingly, a threshold value was detected at 29.29 and 30.98 kg/m2 for BMI, 96.4 and 93.3 cm for WC, 0.57 and 0.60 for WHtR, 0.0848 and 0.0813 m11/6 kg2/3 for ABSI with CVD mortality in men and women, respectively; 29.88 and 29.50 kg/m2 for BMI, 104.3 and 105.6 for WC, 0.61 and 0.67 for WHtR, 0.95 and 0.86 for WHR, 0.0807 and 0.0765 for ABSI in men and women, respectively, and 0.52 for WHHR in women with all-cause mortality. Conclusion: All anthropometric measures of abdominal obesity had positive linear associations with CVD mortality, whereas some showed linear and the others J-shaped relationships with
* Corresponding author. Department of Public Health, University of Helsinki, Mannerheimintie 172, PL41, FI-00014 Helsinki, Finland. Tel.: þ358 9 19127320. E-mail address: xin.song@helsinki.fi (X. Song). 1 See Appendix. http://dx.doi.org/10.1016/j.numecd.2014.09.004 0939-4753/ª 2014 Elsevier B.V. All rights reserved.
296
X. Song et al.
all-cause mortality. BMI had a J-shaped relationship with either CVD or all-cause mortality. Thresholds detected based on mortality may help with clinical definition of obesity in relation to mortality. ª 2014 Elsevier B.V. All rights reserved.
Introduction Obesity is a major risk factor for development of chronic diseases and an important cause of mortality [1,2]. Waist circumference (WC) and waist-to-height ratio (WHtR) appear to be better anthropometric measures of abdominal obesity than body mass index (BMI) and have stronger correlation with intra-abdominal fat content and cardiometabolic risk factors [3]. But to date, the association of anthropometric measures of obesity with all-cause mortality is still controversial: a J- or U-shaped [2,4e7], or a positive linear [8,9] relationship for BMI, a J-shaped [2,10] or a positive linear [10e12] relationship for WC, a positive linear [11,12] or a U-shaped [2] relationship for waist-tohip ratio (WHR). A positive linear [4,6], a J- or U-shaped [5,11,12] relationship for BMI and a positive linear relationship [2,11,12] for these anthropometric measures of abdominal obesity with cardiovascular disease (CVD) mortality has been reported. In addition, A Body Shape Index (ABSI) has been proposed recently to combine WC, weight and height together in one algorithm to predict allcause mortality [7], as well as waist-to-hip-to-height ratio (WHHR) was shown to be superior to BMI, WC or WHtR in predicting CVD risk in the elderly [13]. Considering the inconsistent findings we set up the study, based on the data of the Diabetes Epidemiology: Collaborative analysis Of Diagnostic criteria in Europe (DECODE) study, to investigate the relationships of mortality from CVD and all-cause with various anthropometric measures of obesity, and to detect whether there are potential thresholds existing. ABSI and WHHR that were introduced recently are evaluated together with BMI, WC, WHtR and WHR.
The current data analysis comprised of 46,651 Caucasians (47.1% women) aged 24e99 at baseline, from twelve cohorts in four European countries (Finland, Sweden, Turkey and UK). Data collection included a self-reported questionnaire on smoking status and leisure-time physical activity, and a medical examination to measure weight, height, WC, and hip circumference, as described in detail in previous DECODE publications [14]. Participants with missing data on smoking status, leisure-time physical activity, weight, height, WC and hip circumference, without exact date of emigration, or completely lost to follow-up were excluded. Subjects who had emigrated but the dates of the emigration were recorded and treated as censored cases. Thus, the follow-up was complete. Definition of covariates
Methods
Smoking status at baseline was classified based on responses to the questionnaire into three categories of never smokers, former smokers and current smokers. Reading, watching TV, house works, sewing and walking <1 km daily that do not require moving much and do not physically tax was defined as leisure-time physically inactive, and all other higher levels of physical activity were defined as physically active. Height and weight were measured without shoes and with light clothing. WC was measured midway between the lower rib margin and iliac crest. Hip circumference was measured at the level of widest circumference over the greater trochanters. BMI was calculated as weight in kilograms divided by the square of height in meters. WHtR, WHR, or WHHR was calculated as WC divided by height, hip circumference, or both in meters. The calculation of ABSI was based on WC adjusting for weight and height, which was defined as follows: ABSI Z WC height5/6 weight2/3 [7].
Study population
Definition of fatal events
The DECODE collaboration includes 40 studies and their investigators from 14 European countries who have conducted population-based or large occupational surveys for diabetes and its risk factors applying standard 75 g oral glucose tolerance tests for diagnosis of diabetes [14]. All survey participants included in the data analysis are Caucasians. Individual participant data from each cohort was sent to the National Institute for Health and Welfare in Helsinki, Finland for collaborative data analyses. Each study was approved by the local ethics committees, and the analysis plan was approved by the ethics committee of the National Institute for Health and Welfare, Helsinki, Finland.
Median follow-up time varied between 2.5 and 21.8 years between different cohorts. Vital status and causes of death were recorded in all of the studies included. CVD mortality was defined according to the International Classification of Disease codes 401e448 (9th revision) or I10eI79 (10th revision). Statistical analyses The Cox proportional hazards model was used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs) of CVD or all-cause mortality with anthropometric measures of obesity. Analyses were adjusting for baseline
Study
Finland FINRISK (1987) FINRISK (1992) FINRISK (1997) FINRISK (2002) Sweden Uppsala (1991e1995) Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Turkey TARFS (1998e2002) UK Whitehall II (1991e1993) Total
(1986) (1990) (1994) (1999) (2004)
BMI (kg/m2)
WC (cm)
WHtR
WHR
ABSI (m11/6 kg2/3)
WHHR (m1)
No. of participants
Age (years)
Median follow-up (years)
No. of deaths CVD
All-cause
2541 2570 3788 3808
43.8 44.3 48.6 48.0
(11.2) (11.2) (13.5) (13.0)
26.66 26.57 26.87 27.23
(3.71) (3.87) (3.94) (4.12)
92.2 93.9 94.5 95.4
(10.7) (11.2) (11.3) (11.8)
0.53 0.53 0.54 0.54
(0.06) (0.07) (0.07) (0.07)
0.90 0.92 0.93 0.97
(0.06) (0.07) (0.07) (0.07)
0.0784 0.0796 0.0796 0.0796
(0.0041) (0.0040) (0.0040) (0.0041)
0.52 0.52 0.53 0.55
(0.04) (0.05) (0.05) (0.05)
21.8 16.8 11.8 6.8
293 152 205 69
561 340 418 162
1057 671 761 877 869 864
71.0 46.0 45.0 49.6 50.6 50.9
(0.6) (11.4) (11.3) (14.0) (14.3) (14.1)
26.19 25.42 25.84 26.22 26.67 27.19
(3.38) (3.43) (3.35) (3.65) (3.49) (3.95)
94.4 92.4 91.4 93.4 95.3 96.4
(9.4) (9.2) (9.1) (10.1) (9.7) (10.7)
0.54 0.52 0.52 0.53 0.54 0.54
(0.05) (0.06) (0.05) (0.06) (0.06) (0.06)
0.94 0.94 0.93 0.94 0.92 0.96
(0.05) (0.05) (0.06) (0.06) (0.07) (0.06)
0.0811 0.0806 0.0789 0.0799 0.0805 0.0803
(0.0037) (0.0032) (0.0036) (0.0040) (0.0039) (0.0036)
0.54 0.54 0.53 0.53 0.52 0.54
(0.04) (0.04) (0.04) (0.04) (0.05) (0.04)
10.0 20.5 16.5 12.5 7.4 2.5
126 49 30 37 14 0
276 131 79 126 56 7
1580
53.2 (12.4)
26.41 (4.01)
94.3 (11.0)
0.56 (0.06)
0.93 (0.07)
0.0819 (0.0052)
0.55 (0.05)
7.9
55
109
5300 24 686
49.3 (6.0) 49.0 (12.4)
25.10 (3.15) 26.34 (3.78)
87.4 (9.2) 92.7 (11.0)
0.50 (0.05) 0.53 (0.06)
0.90 (0.06) 0.93 (0.07)
0.0768 (0.0035) 0.0792 (0.0042)
0.51 (0.04) 0.53 (0.05)
5.9 7.9
41 1071
116 2381
Anthropometric measures of obesity and mortality
Table 1 Baseline characteristics and the data of follow-up of the survey in men.
Abbreviations: BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; WHR, waist-to-hip ratio; ABSI, A Body Shape Index; WHHR, waist-to-hip-to-height ratio; CVD, cardiovascular disease. Data are means (standard deviations) or as noted.
297
Abbreviations: BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; WHR, waist-to-hip ratio; ABSI, A Body Shape Index; WHHR, waist-to-hip-to-height ratio; CVD, cardiovascular disease. Data are means (standard deviations) or as noted.
39 1055 6 339 5.8 11.8 0.48 (0.05) 0.50 (0.05) 0.0683 (0.0052) 0.0733 (0.0054) 0.77 (0.07) 0.81 (0.07) 0.47 (0.07) 0.51 (0.08) 75.5 (11.7) 82.0 (12.6) 50.2 (6.1) 46.9 (12.3) 2346 21 965
25.66 (4.69) 26.19 (5.00)
56 35 7.9 0.54 (0.06) 0.0778 (0.0072) 0.84 (0.08) 0.58 (0.09) 90.7 (12.7) 52.7 (12.3) 1619
28.79 (5.65)
75 51 68 27 5 18 12 16 3 0 20.5 16.5 12.5 7.5 2.5 (0.05) (0.04) (0.06) (0.05) (0.05) 0.53 0.50 0.51 0.50 0.52 (0.0054) (0.0042) (0.0060) (0.0047) (0.0045) 0.0783 0.0729 0.0757 0.0753 0.0761 (0.07) (0.06) (0.08) (0.07) (0.07) 0.86 0.81 0.83 0.82 0.85 (0.08) (0.07) (0.08) (0.08) (0.08) 0.52 0.49 0.52 0.52 0.53 (12.2) (11.0) (12.4) (11.8) (12.9) 85.3 79.4 84.2 84.9 86.6 (4.37) (4.39) (4.67) (4.63) (5.12) 25.01 24.99 25.82 26.35 26.62
Finland FINRISK (1987) FINRISK (1992) FINRISK (1997) FINRISK (2002) Sweden Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Northern Sweden MONICA Turkey TARFS (1998e2002) UK Whitehall II (1991e1993) Total
(1986) (1990) (1994) (1999) (2004)
(11.1) (11.4) (14.0) (14.1) (13.9) 45.6 44.8 49.4 50.1 49.7 685 793 902 900 909
310 190 176 58 119 55 61 14 21.8 16.9 11.8 6.8 (0.05) (0.05) (0.05) (0.05) 0.48 0.49 0.49 0.52 (0.0045) (0.0043) (0.0042) (0.0042) 0.0714 0.0724 0.0726 0.0742 (0.06) (0.07) (0.07) (0.06) 0.78 0.79 0.80 0.84 (0.07) (0.08) (0.08) (0.08) 0.49 0.49 0.50 0.52 (11.2) (11.7) (12.2) (12.6) 79.4 80.2 81.3 83.6 (4.90) (4.92) (4.95) (5.05) 26.04 25.70 26.07 26.39 (11.4) (11.5) (12.7) (13.0) 43.7 44.0 46.1 46.6 2812 2828 3788 4383
All-cause CVD
Median follow-up (years) WHHR (m1) ABSI (m11/6 kg2/3) WHR WHtR WC (cm) BMI (kg/m2) Age (years) No. of participants
Tables 1 and 2 provide the descriptive characteristics of the cohorts. Over a median follow-up of 7.9 years, 2381 men and 1055 women died, 1071 men (45.0%) and 339 women (32.1%) from CVD. Table 3 shows that age, high distribution of anthropometric measures of obesity, smoking and leisure-time physical inactivity were significantly associated with CVD and all-cause mortality. The best-fitting conventional model was conventional polynomial model for BMI with CVD and all-cause mortality as well as for WC and WHtR with all-cause mortality in both genders (P < 0.05 for LRT and for deviance difference test against conventional linear model), which suggests a nonlinear relationship (Supplementary Table 1). However, the best-fitting conventional model was observed as conventional linear model for other anthropometric measures of abdominal obesity with CVD or allcause mortality, of which model fitness was not
Study
Results
Table 2 Baseline characteristics and the data of follow-up of the survey in women.
smoking status, leisure-time physical activity and cohort using attained age as the time scale. All analyses were performed separately for men and women. The shape of relationship between anthropometric measures of obesity and mortality was explored using both parametric models (conventional linear or polynomial models) and nonparametric models (the linear or restricted cubic spline regression model). Akaike’s information criterion (AIC) was used to judge the model fitness between conventional linear model and polynomial models (including quadratic, cubic or factional polynomial model), the lower the AIC value the better the model fitness is. The reduction of AIC between nested models is evaluated by the likelihood ratio test (LRT) or a deviance difference test [15,16]. Nonparametric smooth functions using the restricted cubic spline or the linear spline corresponding to the best-fitting conventional polynomial or linear model were also fitted to model their potentially nonlinear relationship [17]. Further, AIC difference 4 was considered as considerably less supported relative to the lowest AIC value between non-nested models [18]. The existence of threshold was estimated by a piecewise regression model using nonlinear least-squares estimation incorporating the “nl” function in Stata, with the lowest value of residual sum of squares, root-mean-square error and AIC, whereas the highest value of R-squared as the criteria for selecting optimal model. Analyses of threshold detection for certain anthropometric measures of obesity were also carried out on the data set with outliers excluded. Outliers are defined to be values of anthropometric measures of obesity that are more or less than three median absolute deviations away from the median [19]. In the Cox proportional hazards model, anthropometric measure of obesity was entered as a continuous variable, and factors including leisure-time physical activity and smoking status as categorical variables, along with a BMI-leisure-time physical activity or a BMI-smoking status interaction term. Data analyses were performed in Stata Intercooled 11.2 (StataCorp, College Station, TX).
X. Song et al.
No. of deaths
298
Anthropometric measures of obesity and mortality
299
significantly improved by conventional polynomial model (P > 0.05 for LRT or for deviance difference test against conventional linear model), which indicates a linear relationship. Further, among the best-fitting models, most anthropometric measures of abdominal obesity had lower AIC values than BMI except for ABSI in women. The nonlinear and linear relationships detected by the parametric conventional modeling were further supported by the nonparametric modeling (AIC difference <4). The spline regression analysis showed that most simple model (K Z 3 for the restricted cubic spline regression or K Z 1 for the linear spline regression) had the lowest AIC value compared to the models with more parameters. HRs with 95% CIs for mortality estimated by the best restricted cubic or linear spline regression model were plotted in Figs. 1 and 2 against the whole distribution of an anthropometric measure as compared with the mean value. HR for BMI with CVD mortality, as well as for BMI, WC and WHtR with all-cause mortality in both genders decreased first at the lower, then increased gradually at the middle, and increased more at the upper distributions of these anthropometric measures of obesity, which
indicates a J-shaped relationship with two potential thresholds existing (Figs. 1 a and g, and 2 aec and gei). A threshold value was detected at 29.29 and 30.98 kg/m2 for BMI with CVD mortality in men and women, respectively; 29.88 and 29.50 kg/m2 for BMI, 104.3 and 105.6 cm for WC, 0.61 and 0.67 for WHtR with all-cause mortality in men and women, respectively (Supplementary Table 2). HR for CVD or all-cause mortality in both genders increased positively with increasing in anthropometric measures of abdominal obesity (Figs. 1 bef and hel, and 2 def and jel), with a threshold detected at 96.4 and 93.3 cm for WC, 0.57 and 0.60 for WHtR, 0.0848 and 0.0813 m11/ 6 kg2/3 for ABSI in men and women, respectively, for CVD mortality; 0.95 and 0.86 for WHR, 0.0807 and 0.0765 for ABSI in men and women, respectively, and 0.52 for WHHR in women for all-cause mortality (Figs. 1 b, c, e, h, i and k, and 2 d, e and jel, and Supplementary Table 2). The results were not substantially altered after excluding the first five years of follow-up (Supplementary Table 3), or additionally adjusting for one or more other CVD risk factors, including hypertension, diabetes or dyslipidemia (results not shown). There was no interaction between an
Table 3 Baseline characteristics of participants according to cardiovascular disease and all-cause mortality.a Cardiovascular disease mortality
All-cause mortality
No
Yes
No
Yes
1071 61.1 (60.1e62.0)b 27.2 (27.0e27.5)b 96.0 (95.3e96.6)b 0.55 (0.55e0.56)b 0.94 (0.94e0.94)b 0.0805 (0.0803e0.0808)b 0.54 (0.54e0.54)b
22,305 48.9 (48.7e49.0) 26.3 (26.3e26.4) 92.5 (92.4e92.7) 0.53 (0.53e0.53) 0.93 (0.93e0.93) 0.0790 (0.0790e0.0791) 0.53 (0.53e0.53)
2381 59.8 (59.0e60.7)b 26.7 (26.5e26.8)b 94.4 (94.0e94.9)b 0.54 (0.54e0.54)b 0.94 (0.93e0.94)b 0.0803 (0.0801e0.0804)b 0.54 (0.54e0.54)b
69.6 30.4b
81.7 18.3
71.9 28.1b
31.1 31.9 37.0b
45.5 31.3 23.2
32.7 30.0 37.3b
339 61.4 (59.3e63.6)b 27.7 (27.2e28.2)b 86.2 (85.0e87.5)b 0.54 (0.53e0.55)b 0.83 (0.82e0.84)b 0.0748 (0.0742e0.0753)b 0.52 (0.51e0.52)b
20,910 46.9 (46.7e47.1) 26.2 (26.1e26.2) 81.9 (81.7e82.0) 0.51 (0.51e0.51) 0.81 (0.81e0.81) 0.0732 (0.0731e0.0733) 0.50 (0.50e0.50)
1055 59.2 (57.9e60.4)b 26.8 (26.5e27.1)b 84.0 (83.3e84.7)b 0.52 (0.52e0.53)b 0.82 (0.82e0.82)b 0.0742 (0.0739e0.0746)b 0.51 (0.51e0.51)b
56.3 43.7b
76.5 23.5
63.7 36.3b
64.6 11.2 24.2b
62.9 18.3 18.8
60.9 13.6 25.5b
Men Number 23,615 Age (years) 49.5 (49.3e49.7) BMI (kg/m2) 26.3 (26.3e26.4) WC (cm) 92.6 (92.4e92.7) WHtR 0.53 (0.53e0.53) WHR 0.93 (0.93e0.93) ABSI (m11/6 kg2/3) 0.0791 (0.0790e0.0791) WHHR (m1) 0.53 (0.53e0.53) Leisure-time physical activity, % Physically active 81.3 Physically inactive 18.7 Smoking status, % Never smokers 44.9 Former smokers 31.2 Current smokers 23.9 Women Number 21,626 Age (years) 47.3 (47.1e47.5) BMI (kg/m2) 26.2 (26.1e26.2) WC (cm) 81.9 (81.8e82.1) WHtR 0.51 (0.51e0.51) WHR 0.81 (0.81e0.81) ABSI (m11/6 kg2/3) 0.0732 (0.0732e0.0733) WHHR (m1) 0.50 (0.50e0.50) Leisure-time physical activity, % Physically active 76.2 Physically inactive 23.8 Smoking status, % Never smokers 62.8 Former smokers 18.2 Current smokers 19.1
Abbreviations: BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; WHR, waist-to-hip ratio; ABSI, A Body Shape Index; WHHR, waist-to-hip-to-height ratio. a Data are age-adjusted means (95% confidence intervals) for obesity indicators or as noted. b P < 0.001 for the difference between yes and no.
300
X. Song et al.
Anthropometric measures of obesity and mortality
anthropometric measure and either smoking status or leisure-time physical activity observed.
Discussion We found that BMI had a J-shaped relationship with CVD mortality, whereas anthropometric measures of abdominal obesity had positive linear relationships. BMI, WC and WHtR showed J-shaped associations with all-cause mortality, whereas WHR, ABSI and WHHR demonstrated positive linear relationships. Thresholds detected based on mortality may help with clinical definition of obesity in relation to mortality. Our findings are in line with previous studies that reported a J- or U-shaped relationship for BMI [2,4e7], a Jshaped relationship for WC [2,7,10] but a positive linear relationship for WHR [9,11,12] with all-cause mortality, and, a positive linear association between CVD mortality and all anthropometric measures of abdominal obesity [2,11,12]. Our previous DECODE study has shown that most anthropometric measures of abdominal obesity are superior to BMI in predicting CVD mortality except for ABSI [20]. The possible mechanisms for this more direct relation between anthropometric measures of abdominal obesity with CVD mortality might be that the intra-abdominal adipocytes are prone to empty their free fatty acids directly into the portal vein, exposing the liver to high concentrations of free fatty acids, which might lead to hyperinsulinaemia, dyslipidemia and hypertension [21]. In addition, adipose tissue, especially tissue from abdominal fat deposits, secretes factors (adipokines) that may promote development of chronic diseases [22]. The thresholds detected may have important clinical implications in the context of definition of obesity in relation to serious clinical outcomes, i.e., mortality. Current existing definitions for abdominal obesity among Caucasians were mostly focused on the WC measurement, at WC of 102 and 88 cm [23], or 94 and 80 cm [24], or at WHR of 0.95 and 0.80 [23], in men and women, respectively. These existing cut-off values have, however, been determined arbitrarily based on analysis of the trade-offs between sensitivity and specificity for discrimination of diabetes or metabolic syndrome [25]. Most of the previous studies were cross-sectional [23]. The WC or the WHR cut-off values proposed by the previous studies were, however, close to thresholds detected in our study for CVD or all-cause mortality, or included in the values between the first and the second thresholds for all-cause mortality. To apply the first threshold will give a high sensitivity and the second one a
301
high specificity depending on the purpose of the application of the thresholds. Considering our study comprising cohorts recruited most in 1990s, and the WC increasing with time, it might be speculated that the thresholds identified were lower than they are now. On the other hand, it can be argued that relationships between anthropometric measures of obesity and mortality may not be affected by changes in the population mean, which needs to be further investigated based on data collected recently. The thresholds detected for all-cause mortality reflects a mixed relationship between obesity and different causes of death. This might be affected by differences in constitutions of the causes of death across countries. Smokers tend to be lean but have high mortality rates [26]. Smoking has been suggested to be associated with visceral adipose fat accumulation, perhaps through simultaneously affecting lipoprotein lipase activity and increasing cortisol levels [27]. In our study, the relationship between anthropometric measures of obesity and mortality was not substantially altered by smoking status. In addition, limited evidence indicates that leisure-time physical inactivity might play an intermediate role in the relationship between obesity and mortality [28]. Our study did not detect significant interaction between obesity and leisure-time physical activity in relation to CVD and allcause mortality. Moreover, the obesity-CVD mortality relationships were not substantially altered after additional adjustment for one or more other CVD risk factors, including hypertension, diabetes or dyslipidemia, which could mediate a larger proportion of excess CVD risk [29]. Our study was based on several European populationor occupational-based studies, with sufficient power to investigate the association between anthropometric measures of obesity and mortality. In addition, the relationships between anthropometric measures of obesity and mortality were examined in the same cohort using both parametric and nonparametric approaches in parallel, which might better capture the true relationships. Yet, limitations of the study exist. Data are based on Caucasian European surveys, the results may not be applicable to other ethnic groups or races due to difference in percentage of body fat across ethnic groups [30]. The study does not have data on changes in anthropometric measurements before the baseline and during the follow-up which makes it impossible to exclude the possibility of ‘reverse causation’ [31], a consequence of underlying diseases before enrollment rather than a cause of death. The potential influence of reverse causality was checked by excluding the first five years of follow-up of which less than 7% of the study population and 25% of the mortality events were excluded, and the results were not altered.
Figure 1 Hazard ratio and 95% confidence intervals for anthropometric measures of obesity with cardiovascular disease (CVD) mortality among men and women. Dashed lines indicate hazard ratios and 95% confidence intervals (CIs) derived from the restricted cubic spline regression with three interior knots placed at the 10th, 50th, and 90th percentiles of body mass index (BMI, a and g), or from linear spline regression with one interior knot placed at the 50th percentiles of anthropometric measures of abdominal obesity (bef and hel), using age as the underlying time-scale in the Cox proportional hazards model and adjusting for smoking status, leisure-time physical activity and cohort, and mean of anthropometric measures of obesity was set as the reference value. Vertical dot line indicates the location which represent threshold (with 95% CIs denoted in subtitle) derived from the piecewise regression model followed by restricted cubic spline regression or linear spline regression. Analyses of thresholds for BMI in both genders were carried out on the data set with outliers excluded. Solid lines indicate slopes before and after thresholds.
Figure 2 Hazard ratios and 95% confidence intervals for anthropometric measures of obesity with all-cause mortality among men and women. Dashed lines indicate hazard ratios and their 95% confidence intervals derived from the restricted cubic spline regression with three interior knots placed at the 10th, 50th, and 90th percentiles of body mass index (BMI), waist circumference, and waist-to-height ratio (aec and gei), or from the linear spline regression with one interior knot placed at the 50th percentiles of waist-to-hip ratio, A Body Shape Index, and waist-to-hip-to-height ratio (def and jel), using the same method as described in Figure 1. Analyses of thresholds for BMI in women were carried out on the data set with outliers excluded.
Anthropometric measures of obesity and mortality
The relationship between certain anthropometric measures of abdominal obesity and mortality might be influenced by the potential correlation between some anthropometrical measures, such as hip and waist circumference [32]. Since this is a collaborative data analysis, certain lifestyle variables such as physical activity has been recorded differently in different studies. In spite of the efforts have been made to “harmonize” the variable and to adjust for studies in data analysis, discrepancies exist. A recent systematic review showed, however, that variations in anatomic locations of measurement of WC did not influence clinical outcomes regarding mortality from all-cause and CVD [33]. The discrepancies in study design and methodologies have been taken into account in our data analysis by fitting the study cohort into the models as a co-variable. In summary, all anthropometric measures of abdominal obesity had positive linear associations with CVD mortality, whereas some showed linear and the others J-shaped relationships with all-cause mortality. BMI had a J-shaped relationship with either CVD or all-cause mortality. Thresholds detected based on mortality may help with clinical definition of obesity in relation to mortality. Conflict of interest The authors declare no conflict of interest. Acknowledgments This work was supported by the Academy of Finland (Suomen Akatemia) [1129197, 136895 and 141005]. Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.numecd.2014.09.004. Appendix Studies and investigators in this collaborative study are: Finland National FINRISK 1987, 1992 and 1997 Cohorts: J. Tuomilehto1,2,3, P. Jousilahti1, J. Lindström1, 1. Department of Chronic Disease Prevention, National Institute for Health and Welfare, Helsinki; 2. Center for Vascular Prevention, Danube University Krems, Krems, Austria; 3 King Abdulaziz University, Jeddah, Saudi Arabia. National FINRISK 2002 Study: J. Tuomilehto1,2,3, T. Laatikainen1,4,5, M. Peltonen1, J. Lindström1, 1. Department of Chronic Disease Prevention, National Institute for Health and Welfare, Helsinki; 2. Center for Vascular Prevention, Danube University Krems, Krems, Austria; 3 King Abdulaziz University, Jeddah, Saudi Arabia; 4. Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio, Finalnd; 5. Hospital Distric of North Karelia, Joensuu, Finland. Sweden Northern Sweden MONICA Survey: S. Söderberg1,2, M. Eliasson1, 1. Department of Public Health and
303
Clinical Medicine, Umeå University, Umeå, Sweden; 2. Baker IDI Heart and Diabetes Institute, Melbourne, Australia. The Uppsala Longitudinal Study of Adult Men (ULSAM): B. Zethelius, Department of Public Health/Geriatrics, Uppsala University Hospital, Uppsala. Turkey Turkish Adult Risk Factor Study (TARFS): A Onat1,2, 1. Turkish Society of Cardiology, Istanbul; 2. Department of Cardiology, Cerrahpasa Medical Faculty, Istanbul University, Istanbul. United Kingdom Whitehall II Study: M.G.Marmot1, A.G. Tabák1,2, M. Kivimäki1,3, E.J. Brunner1, D.R. Witte1,4, 1. Department of Epidemiology and Public Health, University College London, London, UK; 2. Semmelweis University Faculty of Medicine, 1st Department of Medicine, Budapest, Hungary; 3. Finnish Institute of Occupational Health, Helsinki, Finland; 4. Steno Diabetes Center, Gentofte, Denmark. References [1] Obesity: preventing and managing the global epidemic. Report of a WHO consultation. World Health Organ Tech Rep Ser 2000;894: iexii. 1e253. [2] Pischon T, Boeing H, Hoffmann K, Bergmann M, Schulze MB, Overvad K, et al. General and abdominal adiposity and risk of death in Europe. N Engl J Med 2008;359:2105e20. [3] Browning LM, Hsieh SD, Ashwell M. A systematic review of waistto-height ratio as a screening tool for the prediction of cardiovascular disease and diabetes: 0.5 could be a suitable global boundary value. Nutr Res Rev 2010;23:247e69. [4] Kivimäki M, Ferrie JE, Batty GD, Davey Smith G, Elovainio M, Marmot MG, et al. Optimal form of operationalizing BMI in relation to all-cause and cause-specific mortality: the original Whitehall study. Obesity (Silver Spring) 2008;16:1926e32. [5] Klenk J, Nagel G, Ulmer H, Strasak A, Concin H, Diem G, et al. Body mass index and mortality: results of a cohort of 184,697 adults in Austria. Eur J Epidemiol 2009;24:83e91. [6] Prospective Studies Collaboration, Whitlock G, Lewington S, Sherliker P, Clarke R, Emberson J, Halsey J, et al. Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet 2009;373:1083e96. [7] Krakauer NY, Krakauer JC. A new body shape index predicts mortality hazard independently of body mass index. PLoS One 2012;7:e39504. [8] Katzmarzyk PT, Craig CL, Bouchard C. Adiposity, adipose tissue distribution and mortality rates in the Canada Fitness Survey follow-up study. Int J Obes Relat Metab Disord 2002;26:1054e9. [9] Lahmann PH, Lissner L, Gullberg B, Berglund G. A prospective study of adiposity and all-cause mortality: the Malmo Diet and Cancer Study. Obes Res 2002;10:361e9. [10] Koster A, Leitzmann MF, Schatzkin A, Mouw T, Adams KF, van Eijk JT, et al. Waist circumference and mortality. Am J Epidemiol 2008;167:1465e75. [11] Petursson H, Sigurdsson JA, Bengtsson C, Nilsen TI, Getz L. Body configuration as a predictor of mortality: comparison of five anthropometric measures in a 12 year follow-up of the Norwegian HUNT 2 study. PLoS One 2011;6:e26621. [12] Czernichow S, Kengne AP, Stamatakis E, Hamer M, Batty GD. Body mass index, waist circumference and waist-hip ratio: which is the better discriminator of cardiovascular disease mortality risk?: evidence from an individual-participant meta-analysis of 82 864 participants from nine cohort studies. Obes Rev 2011;12:680e7. [13] Carlsson AC, Risérus U, Engström G, Ärnlöv J, Melander O, Leander K, et al. Novel and established anthropometric measures and the prediction of incident cardiovascular disease: a cohort study. Int J Obes (Lond) 2013;37:1579e85. [14] Will new diagnostic criteria for diabetes mellitus change phenotype of patients with diabetes? Reanalysis of European
304
[15] [16] [17] [18]
[19]
[20]
[21]
[22] [23]
[24]
[25]
X. Song et al. epidemiological data. DECODE Study Group on behalf of the European Diabetes Epidemiology Study Group. BMJ 1998;317:371e5. Greenland S. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology 1995;6:356e65. Huelsenbeck JP, Hillis DM, Nielsen R. A likelihood-ratio test of monophyly. Syst Biol 1996;45:546e58. Patrick R, Willi S. Multivariable modelling with cubic regression splines: a principled approach. Stata J 2007;7:45e70. Burnham KP, Anderson DR. Model selection and multimodel inference: a practical information-theoretic approach. New York: Springer-Verlag; 2003. Leys C, Ley C, Klein O, Bernard P, Licata L. Detecting outliers: do not use standard deviation around the mean, use absolute deviation around the median. J Exp Soc Psychol 2013;49:764e6. Song X, Jousilahti P, Stehouwer CD, Söderberg S, Onat A, Laatikainen T, et al. Comparison of various surrogate obesity indicators as predictors of cardiovascular mortality in four European populations. Eur J Clin Nutr 2013;67:1298e302. Cameron AJ, Zimmet PZ. Expanding evidence for the multiple dangers of epidemic abdominal obesity. Circulation 2008;117: 1624e6. Haslam DW, James WP. Obesity. Lancet 2005;366:1197e209. Lean ME, Han TS, Morrison CE. Waist circumference as a measure for indicating need for weight management. BMJ 1995;311: 158e61. International Diabetes Federation (IDF). Worldwide definition of the metabolic syndrome. Brussels, Belgium: International Diabetes Federation; 2006. Waist circumference and waist-hip ratio: report of a WHO Expert Consultation., Geneva, 8e11 December 2008. Geneva: WHO; 2011.
[26] Koster A, Leitzmann MF, Schatzkin A, Adams KF, van Eijk JT, Hollenbeck AR, et al. The combined relations of adiposity and smoking on mortality. Am J Clin Nutr 2008;88:1206e12. [27] 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;87:801e9. [28] Hu G, Tuomilehto J, Silventoinen K, Barengo NC, Peltonen M, Jousilahti P. The effects of physical activity and body mass index on cardiovascular, cancer and all-cause mortality among 47 212 middle-aged Finnish men and women. Int J Obes (Lond) 2005;29: 894e902. [29] Global Burden of Metabolic Risk Factors for Chronic Diseases Collaboration (BMI Mediated Effects), Lu Y, Hajifathalian K, Ezzati M, Woodward M, Rimm EB, et al. Metabolic mediators of the effects of body-mass index, overweight, and obesity on coronary heart disease and stroke: a pooled analysis of 97 prospective cohorts with 1.8 million participants. Lancet 2014;383:970e83. [30] WHO Expert Consultation. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet 2004;363:157e63. [31] Willett WC, Dietz WH, Colditz GA. Guidelines for healthy weight. N Engl J Med 1999;341:427e34. [32] Cameron AJ, Magliano DJ, Shaw JE, Zimmet PZ, Carstensen B, Alberti KG, et al. The influence of hip circumference on the relationship between abdominal obesity and mortality. Int J Epidemiol 2012;41:484e94. [33] Ross R, Berentzen T, Bradshaw AJ, Janssen I, Kahn HS, Katzmarzyk PT, et al. Does the relationship between waist circumference, morbidity and mortality depend on measurement protocol for waist circumference? Obes Rev 2008;9:312e25.