Advance Publication by-J-STAGE Circulation Journal Official Journal of the Japanese Circulation Society http://www.j-circ.or.jp The Significance of Measuring Body Fat Percentage Determined by Bioelectrical Impedance Analysis for Detecting Subjects With Risk Factors Kentaro Yamashita, MD; Takahisa Kondo, MD, PhD; Shigeki Osugi, MD, PhD; Keiko Shimokata, MD, PhD; Kengo Maeda, MD, PhD; Naoki Okumura, MD; Kyoko Matsudaira, MD; Satoshi Shintani, MD, PhD; Takashi Muramatsu, MD, PhD; Kunihiro Matsushita, MD, PhD; Toyoaki Murohara, MD, PhD

Background: Body fat percentage (BF%) determined by bioelectrical impedance analysis is widely used at home and in medical check-ups. However, the clinical significance of measuring BF% has not been studied in detail.

Methods and Results: A cross-sectional study was carried out on a cohort of 10,774 middle-aged Japanese men who had undergone an annual check-up in 2008. Cut-off points were evaluated for (BMI), waist circumference (WC), and BF% for detecting participants with cardiovascular disease (CVD) risk factors ( mellitus, , dyslipidemia), and effectiveness compared for each marker’s cut-off point. Additionally, the effects of smoking on cut-off points were evaluated. The cut-off points of BMI, WC, and BF% for detecting partici- pants with 1 or more CVD risk factors were 22.7 kg/m2, 81.4 cm, and 20.3%, respectively. The cut-off points of BF% for 1 or more CVD risk factors classified 3.43% more subjects into correct categories than those of BMI (P<0.001). The cut-off points of BMI, WC, and BF% for detecting individuals with 3 CVD risk factors in current smokers were 24.9 kg/m2, 87.8 cm, and 23.7%, while those in non-smokers were 23.3 kg/m2, 83.9 cm, and 22.3%, respectively.

Conclusions: BF% could be more effective in detecting individuals with early stage CVD risk accumulation than BMI. The cut-off points for current smokers were lower than those for non-smokers in all markers.

Key Words: Body fat percentage; Body mass index; Cardiovascular disease risk factor; ; Smoking

he prevalence of and obesity has increased on a global scale. Obesity is known to increase the Editorial p ???? T risk of diabetes mellitus (DM), hypertension, dyslip- idemia, and cardiovascular disease (CVD).1–5 In Japan, the Various alternative methods of the anthropometric markers Ministry of Health, Labour and Welfare has been promoting to evaluate obesity, such as body mass index (BMI) and WC, the National Health Promotion Movement in the 21st Century have been proposed. However, controversy remains regarding (Health Japan 21), and health check-ups specifically pro- the best anthropometric marker for detecting individuals with grammed against towards the person CVD risks,8 such as high low-density lipoprotein cholesterol aged 40 or over has been offered since 2008. Waist circumfer- (LDL-C) and metabolic syndrome,2,6,9,10 and the associations ence (WC) is one of the diagnostic criteria of metabolic syn- between anthropometric markers, body fat distribution, and drome.6 Approximately 52 million people are targeted by this CVD risks differ across populations. For example, Asians in- promotion; however, the consultation rate was about 39% in cluding Japanese show lower average BMI, but at the same 2008.7 Therefore, many people missed out on an opportunity BMI, they have higher body fat percentages (BF%) compared to evaluate their CVD risks. to Caucasians.11–13 Moreover, CVD events were frequently

Received March 12, 2012; revised manuscript received May 17, 2012; accepted June 8, 2012; released online July 11, 2012 Time for primary review: 9 days Department of Cardiology, Nagoya University Graduate School of Medicine, Nagoya (K.Y., T.K., S.O., K.S., K. Maeda, N.O., K. Matsudaira, S.S., T. Muramatsu, K. Matsushita, T. Murohara); Department of Advanced Medicine in Cardiopulmonary Disease, Nagoya University Graduate School of Medicine, Nagoya (T.K.), Japan Mailing address: Takahisa Kondo, MD, PhD, Department of Cardiology, Nagoya University Graduate School of Medicine, 65 Tsurumai, Showa-ku, Nagoya 466-8550, Japan. E-mail: [email protected] ISSN-1346-9843 doi: 10.1253/circj.CJ-12-0337 All rights are reserved to the Japanese Circulation Society. For permissions, please e-mail: [email protected] Advance Publication by-J-STAGE YAMASHITA K et al.

Table 1. Baseline Characteristics of Middle-Aged Japanese ing status, and compared the associations of the anthropomet- Male Workers ric markers and CVD risk factors. Total Age (years)* 47.4±5.7 Methods Current smoker (%) 43.3 Study Population BMI (kg/cm2)* 23.3±2.9 We performed a cross-sectional study on a cohort of 14,087 WC (cm)* 83.1±7.3 middle-aged Japanese men who underwent annual medical BF% (%)* 21.6±5.0 check-ups at their workplaces during April 2008 to December Systolic blood pressure (mmHg)* 123.8±12.9 2008. In this analysis, we excluded individuals with a history Diastolic blood pressure (mmHg)* 78.2±8.3 of CVD, cancer, or those who had been on medication for hy- Total cholesterol (mmol/L)† 5.20 (5.18–5.21) pertension, DM, or dyslipidemia whose anthropometric mea- Triglycerides (mmol/L)† 1.16 (1.15–1.17) sures might have been changed as a result of these underlying HDL-C (mmol/L)† 1.51 (1.50–1.52) diseases and treatments, leaving 10,774 men eligible for the LDL-C (mmol/L)† 3.13 (3.12–3.14) present analysis. The study protocol was approved by the Ethics Review Com- Fasting plasma glucose (mmol/L)† 5.24 (5.23–5.25) mittee of Nagoya University Graduate School of Medicine, Prevalence of CVD risk factors (%) Nagoya, Japan. Hypertension 14.2 Diabetes mellitus 1.6 Baseline Examination and Analysis of Blood Samples Dyslipidemia 44.5 Anthropometric measurements were taken in the standing po- Triglycerides ≥150 23.2 sition while the participants wore light clothing and no shoes. HDL-C <40 5.6 Height was measured to the nearest 0.1 cm using a portable LDL-C ≥140 29.1 stadiometer and weight was measured to the nearest 0.1 kg using Number of CVD risk factors (%) an electronic scale that was calibrated prior to each measure- No risk factors 48.3 ment. BMI was calculated as weight in kilograms divided by the square of the height in meters. WC was measured to the nearest 1 risk factor 43.4 0.1 cm at the umbilical level. BF% was estimated by BIA using 2 risk factors 8.0 the standard tetrapolar technique according to the manufactur- 3 risk factors 0.3 er’s instructions for distal electrode placement on both feet (a BMI, body mass index; WC, waist circumference; BF%, body fat stand-on mode, TBF-210, TANITA, Tokyo, Japan). percentage; HDL-C, high-density lipoprotein cholesterol; LDL-C, Trained nurses obtained information on smoking and past low-density lipoprotein cholesterol; CVD, cardiovascular disease; CI, confidence intervals. medical history in face-to-face 15–30 min interviews. Subjects *Mean ± SD. †Geometric mean (95% CI). responded to an item which classified them as current or non- smokers. Current smokers were defined as those who had been smoking for at least a year. observed among non-obese Asians (BMI <25 kg/m2).5,14,15 We used the blood sample data from physical check-ups of Therefore, measuring BF% would be useful to detect indi- the participants. Blood samples were drawn from the individu- viduals with CVD risks, even in the normal range of BMI or als after overnight or 8 h fasting for determination of the serum WC. It is crucial to detect the presence of early stage CVD risk lipids and plasma glucose. accumulation in the primary prevention. Bioelectrical impedance analysis (BIA) is widely accepted Definition of CVD Risk Factor Clustering as a safe, rapid, low cost, and reliable technique to estimate We used the criteria for CVD risk factors proposed by the Third BF%.16–18 BF% determined by BIA is widely used at home Report of the National Cholesterol Education Program Expert and in medical check-ups for screening measurement of an Panel on Detection, Evaluation, and Treatment of High Blood individual’s because of its simplicity. How- Cholesterol in Adults (NCEP-ATPIII).30 Patients with systolic ever, the clinical significance of measuring BF% by BIA has blood pressure ≥140 mmHg and/or diastolic BP ≥90 mmHg not been well studied in detail. The information derived from were defined as having hypertension. Abnormal levels of serum the body fat scale would add value to evaluate individual lipids were defined as follows: LDL-C ≥3.63 mmol/L, high- health condition. density lipoprotein cholesterol <1.04 mmol/L, and triglycerides Smoking is still prevalent in Japan as well as Asian coun- ≥1.70 mmol/L. Dyslipidemia was defined as abnormal levels of tries, and is one of the chief determinants of adult mortality at least one of the serum lipids. Patients with fasting plasma from CVD and other non-communicable diseases.19–21 Current glucose ≥7.0 mmol/L were defined as having DM. smokers show a high prevalence of CVD risk factors as well as high incidence of CVD even though they have lower body Statistical Analysis weight than never or ex-smokers.22–25 Smoking is associated Continuous variables are expressed as mean ± standard deviation with visceral fat accumulation,26 insulin resistance,27 and met- (SD). Lipids and fasting plasma glucose were naturally logarith- abolic syndrome.28,29 Considering these results, we need to mically transformed to approximately normalize the distribution examine the cut-off points of anthropometric markers in indi- and used in the analysis, transformed back for data presentation, viduals stratified by smoking status. and are shown as geometric means and the 95% confidence Thus, we investigated the associations of BF% with CVD intervals (95% CI). Categorical variables were expressed as risk factors, and evaluated the usefulness of BF% in compari- percentages and were compared with a chi-square test. Re- son with BMI and WC to detect participants with hyperten- ceiver-operating characteristics (ROC) curves were plotted sion, dyslipidemia, and DM in a Japanese male population. using measures of sensitivity and 1-specificity based on the cut- Additionally, we stratified subjects according to current smok- off values for various anthropometric markers. The area under Advance Publication by-J-STAGE Body Fat Percentage and Cardiovascular Risk Factor

Figure 1. Odds ratios for the presence of cardiovascular disease (CVD) risk factors by body mass index (BMI), waist circumfer- ence (WC), and body fat percentage (BF%) quintiles are shown. Values were adjusted for age and current smoking status. *P<0.05; compared with each reference (q1). BMI quintiles were classified as follows: q1, ≤21.1; q2, 21.2–22.7; q3, 22.8–24.1; q4, 24.2–26.0; q5, ≥26.1 (kg/m2). WC quintiles were classified as follows: q1, ≤78.0; q2, 78.1–82.0; q3, 82.1–85.0; q4, 85.1–90.0; q5, ≥90.1 (cm). BF% quintiles were classified as follows: q1, ≤17.8; q2, 17.9–20.6; q3, 20.7-23.1; q4, 22.6–26.4; q5, ≥26.5 (%).

the ROC curves (AUC) is a measure of the diagnostic power of Association of Quintiles of Anthropometric Markers With test. A perfect test will have an AUC of 1.0, and an AUC equal CVD Risk Factors Figure 1 shows the adjusted odds ratios to 0.5 means the test performs no better than chance. To obtain for detecting subjects with CVD risk factors according to an- the optimal cut-off point for anthropometric markers in detect- thropometric markers, categorized by quintiles. All anthropo- ing individuals with CVD risk factors, we chose the point on the metric markers showed significantly positive correlation with ROC curves that showed the largest Youden index (sensitivity+ the prevalence of CVD risk factor(s). Adjusted OR of 1 or specificity–1).31 We considered subjects whose anthropometric more and 2 or more CVD risk factors increased steeply with markers were higher than cut-off points as high risk group. Odds increasing BF% quintiles and to a lesser extent with increasing ratios (OR) were calculated using multiple logistic regression BMI and WC quintiles. However, WC showed higher OR for analysis with adjustments for age and current smoking status. having 3 risk factors than BMI and BF%. The effects of reclassification using anthropometric markers Cut-Off Points of Anthropometric Markers and CVD Risk were assessed using recently published methods that estimated Factors Table 2 shows the results of ROC curve analyses. the net reclassification improvement (NRI).32 Improvement in AUC for BF% was higher for having 1 or more CVD risk fac- reclassification was estimated by taking the sum of differences tors (AUC=0.689) compared with those for BMI or WC. The in proportions of participants reclassified high-risk categories cut-off points of BMI, WC, and BF% detecting subjects with minus the proportion reclassified low-risk categories for people 1 or more, 2 or more, and 3 CVD risk factors ranged from 22.7 who had CVD risks and the proportion of subjects moving low- to 24.7 kg/m2, from 81.4 to 88.0 cm and from 20.3 to 23.0%, risk categories minus the proportion moving high-risk categories respectively. When detecting participants with 1 or more and for people who did not have CVD risks. Using this method, the 2 or more CVD risk factors, sensitivity of BF% cut-off points overall sum is defined as the NRI. A P-value of less than 0.05 showed higher values than those of BMI and WC, whereas was considered statistically significant. All statistical analyses specificity of BMI showed higher values than those of BF% were conducted using SPSS statistical package for Windows and WC. According to the analyses of components of CVD version 19.0 J (SPSS Inc, Chicago, IL, USA). risk factors, AUC for BF% was the highest for dyslipidemia (AUC=0.682) among the 3 anthropometric markers. Dyslip- idemia emerged at lower cut-off points of all these anthropo- Results metric markers, whereas DM emerged at higher cut-off points. Baseline Characteristics Shown in Table 3 are the OR of each anthropometric marker The baseline characteristics of the study subjects and the prev- categorized by each cut-off value for the presence of CVD risk alence of CVD risk factors are shown in Table 1. The mean factors. All anthropometric markers were significantly associ- age of the study subjects was 47.4±5.7 years and the proportion ated with CVD risk factors even after adjustment for age and of current smoker was 43.3%. Mean BMI, WC, and BF% were current smoking status. Then we adjusted for anthropometric 23.3±2.9 kg/m2, 83.1±7.3 cm, and 21.6±5.0%, respectively. markers simultaneously. BF% cut-off points showed higher The prevalence of dyslipidemia (44.5%) was higher than other OR for detecting individuals with CVD risk factors except for risk factors. Approximately half of the participants did not hypertension than BMI and WC cut-off points. have a CVD risk factor. We then assessed the NRI of WC and BF% compared with Advance Publication by-J-STAGE YAMASHITA K et al.

Table 2. AUC and Cut-Off Point of Anthropometric Markers AUC (95% CI) Cut-off Sensitivity (%) Specificity (%) CVD risk factor ≥1 (n=5,569) BMI (kg/m2) 0.670 (0.660–0.680) 22.7 67.3 57.5 WC (cm) 0.665 (0.655–0.675) 81.4 71.1 53.5 BF% (%) 0.689 (0.680–0.699) 20.3 72.1 56.1 CVD risk factor ≥2 (n=895) BMI (kg/m2) 0.691 (0.673–0.709) 24.7 54.6 74.1 WC (cm) 0.685 (0.668–0.703) 84.1 62.7 62.8 BF% (%) 0.698 (0.682–0.715) 22.0 71.4 57.8 CVD risk factor 3 (n=37) BMI (kg/m2) 0.727 (0.640–0.815) 23.5 81.1 56.4 WC (cm) 0.775 (0.703–0.846) 88.0 67.6 76.8 BF% (%) 0.763 (0.686–0.839) 23.0 83.8 63.5 Hypertension (n=1,533) BMI (kg/m2) 0.641 (0.626–0.656) 24.7 46.0 74.6 WC (cm) 0.634 (0.619–0.649) 84.2 55.5 63.6 BF% (%) 0.635 (0.620–0.650) 20.8 71.0 48.8 Diabetes mellitus (n=177) BMI (kg/m2) 0.603 (0.557–0.648) 25.6 38.4 81.1 WC (cm) 0.617 (0.574–0.659) 87.0 45.8 73.1 BF% (%) 0.619 (0.574–0.664) 23.2 57.1 65.3 Dyslipidemia (n=4,791) BMI (kg/m2) 0.657 (0.647–0.668) 22.7 68.1 55.0 WC (cm) 0.654 (0.643–0.664) 81.4 72.1 51.1 BF% (%) 0.682 (0.672–0.692) 20.3 73.6 53.6 AUC, area under the receiver-operating characteristics curve. Other abbreviations as in Table 1.

Table 3. Crude and Adjusted ORs Using the Cut-Off Values for the Presence of CVD Risk Factors Crude Model 1* Model 1+all markers OR (95% CI) OR (95% CI) OR (95% CI) CVD risk factor ≥1 BMI 2.79 (2.58–3.02)‡ 2.84 (2.62–3.07)‡ 1.40 (1.25–1.57)‡ WC 2.82 (2.61–3.06)‡ 2.80 (2.59–3.03)‡ 1.45 (1.30–1.61)‡ BF% 3.31 (3.06–3.59)‡ 3.41 (3.14–3.69)‡ 2.26 (2.03–2.51)‡ CVD risk factor ≥2 BMI 3.45 (3.00–3.96)‡ 3.56 (3.09–4.09)‡ 2.05 (1.71–2.46)‡ WC 2.84 (2.46–3.28)‡ 2.82 (2.44–3.24)‡ 1.25 (1.04–1.50)† BF% 3.41 (2.94–3.97)‡ 3.55 (3.05–4.12)‡ 2.14 (1.77–2.58)‡ CVD risk factor=3 BMI 5.54 (2.43–12.63)‡ 5.83 (2.55–13.30)‡ 1.20 (0.41–3.54) WC 6.91 (3.47–13.77)‡ 6.97 (3.49–13.89)‡ 2.95 (1.28–6.79)† BF% 8.99 (3.75–21.57)‡ 9.60 (3.99–23.07)‡ 4.95 (1.70–14.47)† Hypertension BMI 2.51 (2.24–2.80)‡ 2.58 (2.31–2.88)‡ 1.78 (1.54–2.06)‡ WC 2.18 (1.95–2.43)‡ 2.16 (1.94–2.41)‡ 1.21 (1.05–1.40)† BF% 2.33 (2.07–2.62)‡ 2.41 (2.14–2.72)‡ 1.65 (1.43–1.90)‡ Diabetes mellitus BMI 2.67 (1.97–3.63)‡ 2.97 (2.18–4.04)‡ 1.85 (1.22–2.82)† WC 2.29 (1.70–3.09)‡ 2.29 (1.70–3.09)‡ 1.16 (0.78–1.74) BF% 2.50 (1.85–3.38)‡ 2.70 (1.99–3.65)‡ 1.86 (1.27–2.72)† Dyslipidemia BMI 2.61 (2.41–2.83)‡ 2.65 (2.45–2.87)‡ 1.31 (1.17–1.46)‡ WC 2.71 (2.50–2.94)‡ 2.71 (2.50–2.94)‡ 1.47 (1.32–1.64)‡ BF% 3.22 (2.97–3.49)‡ 3.26 (3.01–3.55)‡ 2.24 (2.01–2.49)‡ OR, odds ratio. Other abbreviations as in Table 1. *Values were adjusted for age and current smoking status (yes/no) in Model 1. †P<0.05; ‡P<0.001. Advance Publication by-J-STAGE Body Fat Percentage and Cardiovascular Risk Factor

Figure 2. The net reclassification improvement of waist circumference (WC) or body fat percentage (BF%) compared with body mass index (BMI) by using each cut-off point for detecting the presence of cardiovascular disease risk factors.

Table 4. Cut-Off Points of Anthropometoric Markers in the Sensitivity Range of 80% or More Cut-off Sensitivity (%) Specificity (%) OR (95% CI)* CVD risk factor ≥1 (n=5,569) BMI (kg/m2) 21.7 80.4 41.3 2.95 (2.70–3.21) WC (cm) 79.5 80.5 41.8 2.96 (2.71–3.23) BF% (%) 19.1 80.6 44.5 3.44 (3.15–3.75) CVD risk factor ≥2 (n=895) BMI (kg/m2) 22.4 81.1 41.8 3.14 (2.65–3.74) WC (cm) 82.0 81.3 44.0 3.37 (2.83–4.00) BF% (%) 20.8 80.0 48.3 3.87 (3.27–4.59) CVD risk factor=3 (n=37) BMI (kg/m2) 23.5 81.1 56.4 5.83 (2.55–13.30) WC (cm) 84.0 83.8 55.3 6.29 (2.6–15.08) BF% (%) 23.0 83.8 63.5 9.60 (3.99–23.07) Abbreviations as in Tables 1,3. *Values adjusted for age and current smoking status (yes/no) are shown.

BMI by using each cut-off point (Figure 2). BF% cut-off 24.7, and 24.9 kg/m2, 83.1, 87.5, and 87.8 cm, and 20.4, 22.0, point detecting subjects with 1 or more CVD risk factors clas- and 23.7%, while those in non-smokers were 22.9, 23.9, and sified 3.43% more subjects into the correct categories than 23.3 kg/m2, 81.4, 84.2, and 83.9 cm, and 20.3, 21.2, and 22.3%, BMI cut-off points (P<0.001). BF% cut-off point detecting respectively. As the number of CVD risk increased, the differ- subjects with dyslipidemia classified 4.06% more subjects into ences between the cut-off points for smokers and those for the correct categories than the BMI cut-off points (P<0.001). non-smokers tended to increase. NRI analysis demonstrated As for detecting participants with 2 or more CVD risk factors, that BF% cut-off points for detecting subjects with 1 or more BMI cut-off point classified 3.23% more subjects into the cor- CVD risk factors classified more subjects into the correct cat- rect categories than the WC cut-off point. egories than BMI cut-off points even in both current smokers We additionally calculated cut-off points in the sensitivity and non-smokers. range of 80% or more (Table 4). Under this condition, BF% cut-off points classified 3.45% and 5.33% more participants into the correct categories than BMI cut-off points did when Discussion detecting subjects with 1 or more, and 2 or more CVD risk Our results clarified that in the middle-aged Japanese male factors, respectively (Figure 3). workers, BF% measured by BIA could detect subjects who Cut-Off Points of Anthropometric Markers Stratified by have 1 or 2 CVD risk factors more accurately than BMI. We Smoking Status The results of ROC analyses, OR, and NRI could not find out the superiority of WC over BMI among stratified by current smoking status are shown inTable 5. The subjects having 1 or 2 CVD risk factors. Current smokers cut-off points of anthropometric markers for current smokers showed smaller cut-off points for detecting subjects with CVD were lower than those for non-smokers. The cut-off points of risk factors than non-smokers. BMI, WC, and BF% for detecting subjects with 1 or more, 2 Obesity is a medical condition in which excess body fat has or more, and 3 CVD risk factors in current smokers were 23.7, accumulated to the extent that it causes adverse effects on Advance Publication by-J-STAGE YAMASHITA K et al.

Figure 3. The net reclassification improvement of waist circumference (WC) or body fat percentage (BF%) compared with body mass index (BMI) by using each cut-off point in the sensitivity range of 80% or more for detecting the presence of cardiovas- cular disease risk factors.

Table 5. AUC and Cut-Off Points of Anthropometric Markers Stratified by Current Smoking Status Sensitivity Specificity AUC (95% CI) Cut-off OR (95% CI)* NRI (%) (%) (%) CVD risk factor ≥1 Non-current smoker (n=3,039) BMI (kg/m2) 0.663 (0.649–0.676) 23.7 54.1 69.7 2.74 (2.47–3.05) Reference WC (cm) 0.657 (0.644–0.671) 83.1 58.2 64.7 2.54 (2.29–2.82) –0.84 (–3.03, 1.35) BF% (%) 0.680 (0.666–0.693) 20.4 70.5 55.8 3.11 (2.79–3.46) 2.57 (0.12, 5.01)† Current smoker (n=2,530) BMI (kg/m2) 0.682 (0.667–0.697) 22.9 63.3 63.9 3.10 (2.75–3.50) Reference WC (cm) 0.675 (0.660–0.691) 81.4 71.7 55.2 3.10 (2.74–3.50) –0.32 (–2.85, 2.20) BF% (%) 0.702 (0.687–0.717) 20.3 73.2 57.9 3.86 (3.41–4.38) 3.98 (1.44, 6.51)† CVD risk factor ≥2 Non-current smoker (n=494) BMI (kg/m2) 0.693 (0.670–0.717) 24.7 55.7 73.3 3.53 (2.92–4.26) Reference WC (cm) 0.680 (0.656–0.704) 87.5 46.4 78.3 3.11 (2.58–3.75) –4.32 (–8.30, –0.35)† BF% (%) 0.702 (0.679–0.725) 22 71.3 58.5 3.64 (2.97–4.45) 0.76 (–3.52, 5.05) Current smoker (n=401) BMI (kg/m2) 0.690 (0.663–0.717) 23.9 63.8 65.6 3.53 (2.84–4.37) Reference WC (cm) 0.692 (0.666–0.717) 84.2 66.6 63.2 3.43 (2.76–4.26) 0.31 (–3.78, 4.39) BF% (%) 0.693 (0.668–0.718) 21.2 78.3 51.1 3.91 (3.06–5.00) –0.09 (–4.84, 4.66) CVD risk factor=3 Non-current smoker (n=19) BMI (kg/m2) 0.787 (0.678–0.896) 24.9 73.7 73.8 8.28 (2.97–23.07) Reference WC (cm) 0.845 (0.757–0.932) 87.8 84.2 77.1 17.79 (5.17–61.14) 13.78 (–0.84, 28.40) BF% (%) 0.845 (0.757–0.932) 23.7 89.5 70.0 21.09 (4.86–91.55) 11.96 (–5.94, 29.86) Current smoker (n=18) BMI (kg/m2) 0.669 (0.539–0.799) 23.3 83.3 54.9 6.59 (1.90–22.89) Reference WC (cm) 0.699 (0.596–0.803) 83.9 88.9 47.9 7.23 (1.66–31.50) –1.39 (–12.36, 9.57) BF% (%) 0.674 (0.559–0.788) 22.3 83.3 56.9 7.10 (2.05–24.62) 2.04 (–13.40, 17.49) NRI, net reclassification improvement. Other abbreviations as in Tables 1–3. *Values were adjusted for age. †P<0.05. Non-current smoker: n=6,102; current smoker: n=4,666. health. In the long-term prospective cohort studies, not only as risk factors, because information about systolic blood pres- obese but also overweight individuals have an increased risk for sure, history of diabetes, and lipids is more informative than metabolic disorders, cardiovascular events, and total death.2,5,14 BMI and WC when predicting the first-onset CVD.35 However, However, several often-used CVD risk scores (eg, Framingham to identify individuals who have obesity-related CVD risks and risk score,33 SCORE34) do not contain anthropometric markers to develop awareness of individual health conditions, establish- Advance Publication by-J-STAGE Body Fat Percentage and Cardiovascular Risk Factor ment of anthropometric markers to evaluate more precise CVD petite-suppressing action of nicotine.50 However, a previous risk in each population group is also important. study reported that smokers show higher prevalence of hyper- Several epidemiologic studies have shown that Asians have tension, hyperglycemia, and dyslipidemia than never smokers higher amounts of body fat at lower BMI and WC than Western among the non-obese subjects, although WC were not signifi- populations.11,13 The Regional Office for Western Pacific -Re cantly different.25 In our study, the cut-off points of anthropo- gion of WHO proposed a separate classification of obesity for metric markers for current smokers were lower than those for Asia that adult overweight be defined as a BMI≥ 23.0 kg/m2, and non-smokers. In current smokers, the metabolic disorders that obesity be defined as a BMI≥ 25.0 kg/m2 (WPRO criteria).15 would progress before starts to change. Therefore, In the present study, the cut-off points of BMI for detecting in- it is important to obtain information of smoking status when dividuals with 1 or more and 2 or more CVD risk factors were we measure the anthropometric markers. in the overweight range of the WPRO criteria. Overweight There are some limitations that should be taken into consid- subjects proposed by the WPRO criteria have higher risks for eration. First, BF% by BIA is affected by individual and envi- metabolic disorders in a Japanese population,5 so, cut-off points ronmental factors such as age, sex, time of measurement car- of BMI described in this study would be reasonable. ried out, consumption of food or beverage, sweating, and Recently, WC has been proposed as more sensitive mea- physical activity. The BIA prediction equations vary depend- surements of visceral obesity and more indicative of cardiovas- ing on populations and equipments used. However, all partici- cular risk. WC criterion for suggested by pants in the present study were Japanese males and we used the NCEP-ATPIII and the International Diabetes Federation (IDF) same type of body composition analyzer validated for Japanese. is 90 cm for Asian male36 and that suggested by Japan Society Moreover, we measured BF% under the controlled conditions: for the Study of Obesity (JASSO) is 85 cm.6 Some studies in in the fasting state and before working in the morning. Second, Asia also showed that the optimal WC cut-off points for men the present findings might not apply directly to other popula- are smaller than IDF criterion.37,38 In line with those data, the tions. Third, the number of subjects with 3 CVD risk factors cut-off points of WC for detecting subjects with 1 or more and was small. Further investigation in higher risk group is needed 2 or more CVD risk factors were smaller in this study than to clarify more reliable cut-off points of anthropometric mark- those of NCEP-ATPIII, IDF, and JASSO criteria. ers for detecting the presence of 3 risk factors. WC showed stronger association with the presence of the In conclusion, anthropometric markers are useful in evalu- 3 CVD risk factors. However, the ability of BF% to detect ating CVD risk factors in that the abnormal values are well patients with CVD risk factors was superior to BMI, and espe- corresponded with the presence of dyslipidemia, hypertension, cially useful when detecting individuals with 1 or more CVD and/or DM. In particular, BF% is a useful marker to identify risk factors. Given that anthropometric markers are used for subjects of early stage of developing CVD who have 1 or 2 the first screening of subjects as methods of detecting the in- CVD risk factors. Furthermore, our study suggests that the dividuals who have CVD risk factors, the cut-off point should cut-off values of the anthropometric markers for detecting have a high sensitivity of at least 80% despite the decreasing individuals with CVD risk factors should be lower in current of specificity. In this sense, BF% was superior to BMI when smokers than in non-smokers. To confirm generalizability of detecting subjects with 1 or more, and 2 or more CVD risk the present findings, it is necessary to investigate the signifi- factors, whereas there was no significant difference between cance of using BF% by BIA to predict the prevalence of hy- WC and BMI. Additionally, the population attributable risk is pertension, DM, and dyslipidemia in other cohorts. larger in non-obese individuals with 1 or 2 risk factors than those with metabolic syndrome in Japan.39,40 Therefore, as- Acknowledgment sessing the presence of not only metabolic syndrome but clas- We are extremely grateful to the healthcare providers who worked with sical risk factors such as hypertension, DM, and dyslipidemia us for the benefit of the subjects. is important to prevent the incidence of CVD.39–42 It is crucial to predict the presence of early stage CVD risk accumulation Disclosures in the primary prevention. Conflict of Interest: None. BF% showed stronger association with the prevalence of dyslipidemia than BMI and WC. Dyslipidemia has been re- References ported to precede the appearances of blood pressure elevation, 1. Ammar KA, Redfield MM, Mahoney DW, Johnson M, Jacobsen SJ, fasting glucose elevation, and WC enlargement,43 and BF% Rodeheffer RJ. Central obesity: Association with left ventricular has been suggested to be superior to BMI for detecting sub- dysfunction and mortality in the community. Am Heart J 2008; 156: jects with dyslipidemia in a previous study.44 In the presence 975 – 981. 2. Arnlöv J, Ingelsson E, Sundström J, Lind L. Impact of body mass of a positive net energy balance, there is increased accumula- index and the metabolic syndrome on the risk of cardiovascular disease tion of triglycerides in both adipose tissues and non-adipose and death in middle-aged men. Circulation 2010; 121: 230 – 236. tissues, such as muscle and liver.45 Intramyocellular lipid, a 3. Dagenais GR, Yi Q, Mann JF, Bosch J, Pogue J, Yusuf S. Prognostic triglycerides compartment in myocyte, is positively correlated impact of body weight and abdominal obesity in women and men 45,46 with cardiovascular disease. Am Heart J 2005; 149: 54 – 60. with insulin resistance, and that in tibialis anterior muscle 4. Hubert HB, Feinleib M, McNamara PM, Castelli WP. Obesity as an was reported to be positively correlated with BF%.47 Addi- independent risk factor for cardiovascular disease: A 26-year follow- tionally BF% by BIA is a simple way to measure by oneself. up of participants in the Framingham Heart Study. Circulation 1983; Therefore, BF% can be a good method to detect individuals 67: 968 – 977. 5. Shiwaku K, Anuurad E, Enkhmaa B, Nogi A, Kitajima K, Shimono with early stage metabolic disorder. K, et al. Overweight Japanese with body mass indexes of 23.0–24.9 There is increasing evidence that smoking affects body fat have higher risks for obesity-associated disorders: A comparison of distribution and is associated with central obesity and insulin Japanese and Mongolians. Int J Obes Relat Metab Disord 2004; 28: resistance.48,49 Mean BMI of smokers is lower than that of 152 – 158. 6. Examination Committee of Criteria for ‘Obesity Disease’ in Japan, non-smokers, which is suggested to be caused by increased Japan Society for the Study of Obesity. New criteria for ‘obesity energy expenditure resulting from the sympathetic nervous disease’ in Japan. Circ J 2002; 66: 987 – 992. system activation,23 and/or lower calorie intake caused by ap- 7. The Ministry of Health Labour and Welfare. Implementation status Advance Publication by-J-STAGE YAMASHITA K et al.

of specific health checkup, 2008. http://www.mhlw.go.jp/bunya/ 29. Nakanishi N, Takatorige T, Suzuki K. Cigarette smoking and the risk shakaihosho/iryouseido01/info03n.html/ (accessed December 2, 2011). of the metabolic syndrome in middle-aged Japanese male office 8. Huxley R, Mendis S, Zheleznyakov E, Reddy S, Chan J. Body mass workers. Ind Health 2005; 43: 295 – 301. index, waist circumference and waist:hip ratio as predictors of car- 30. National Cholesterol Education Program (NCEP) Expert Panel on diovascular risk--a review of the literature. Eur J Clin Nutr 2010; 64: Detection Evaluation, and Treatment of High Blood Cholesterol in 16 – 22. Adults (Adult Treatment Panel III). Third Report of the National 9. Imamura T, Doi Y, Arima H, Yonemoto K, Hata J, Kubo M, et al. Cholesterol Education Program (NCEP) Expert Panel on Detection, LDL cholesterol and the development of stroke subtypes and coro- Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult nary heart disease in a general Japanese population: The Hisayama Treatment Panel III) final report.Circulation 2002; 106: 3143 – 3421. study. Stroke 2009; 40: 382 – 388. 31. Youden WJ. Index for rating diagnostic tests. Cancer 1950; 3: 32 – 35. 10. Okamura T, Kokubo Y, Watanabe M, Higashiyama A, Miyamoto Y, 32. Pencina MJ, D’Agostino RB, Vasan RS. Evaluating the added pre- Yoshimasa Y, et al. Low-density lipoprotein cholesterol and non- dictive ability of a new marker: From area under the ROC curve to high-density lipoprotein cholesterol and the incidence of cardiovascu- reclassification and beyond. Stat Med 2008; 27: 157 – 172; discus- lar disease in an urban Japanese cohort study: The Suita study. Ath- sion 207 – 212. erosclerosis 2009; 203: 587 – 592. 33. Wilson PW, D’Agostino RB, Levy D, Belanger AM, Silbershatz H, 11. Chang CJ, Wu CH, Chang CS, Yao WJ, Yang YC, Wu JS, et al. Low Kannel WB. Prediction of coronary heart disease using risk factor body mass index but high percent body fat in Taiwanese subjects: categories. Circulation 1998; 97: 1837 – 1847. Implications of obesity cutoffs. Int J Obes Relat Metab Disord 2003; 34. Conroy RM, Pyörälä K, Fitzgerald AP, Sans S, Menotti A, De Backer 27: 253 – 259. G, et al. Estimation of ten-year risk of fatal cardiovascular disease in 12. WHO Expert Consultation. Appropriate body-mass index for Asian Europe: The SCORE project. Eur Heart J 2003; 24: 987 – 1003. populations and its implications for policy and intervention strate- 35. Wormser D, Kaptoge S, Di Angelantonio E, Wood AM, Pennells L, gies. Lancet 2004; 363: 157 – 163. Thompson A, et al. Separate and combined associations of body-mass 13. Deurenberg P, Yap M, van Staveren WA. Body mass index and per- index and abdominal adiposity with cardiovascular disease: Collabora- cent body fat: A meta analysis among different ethnic groups. Int J tive analysis of 58 prospective studies. Lancet 2011; 377: 1085 – 1095. Obes Relat Metab Disord 1998; 22: 1164 – 1171. 36. International Diabetes Federation (IDF). The IDF consensus world- 14. Kondo T, Osugi S, Shimokata K, Honjo H, Morita Y, Yamashita K, wide definition of the metabolic syndrome; 2006. http://www.idf. et al. Metabolic syndrome and all-cause mortality, cardiac events, and org/webdata/docs/MetS_def_update2006.pdf/ (accessed December cardiovascular events: A follow-up study in 25,471 young- and mid- 2, 2011). dle-aged Japanese men. Eur J Cardiovasc Prev Rehabil 2011; 18: 37. Park SH, Choi SJ, Lee KS, Park HY. Waist circumference and waist- 574 – 580. to-height ratio as predictors of cardiovascular disease risk in Korean 15. World Health Organization, International Association for the Study adults. Circ J 2009; 73: 1643 – 1650. of Obesity, International Obesity Task Force. The Asia-Pacific per- 38. Wang F, Wu S, Song Y, Tang X, Marshall R, Liang M, et al. Waist spective: Redefining obesity and its treatment. Australia: Health Com- circumference, body mass index and waist to hip ratio for prediction munications Australia Pty Limited: Sidney, 2000. of the metabolic syndrome in Chinese. Nutr Metab Cardiovasc Dis 16. Lukaski HC, Johnson PE, Bolonchuk WW, Lykken GI. Assessment 2009; 19: 542 – 547. of fat-free mass using bioelectrical impedance measurements of the 39. Iso H. A Japanese health success story: Trends in cardiovascular human body. Am J Clin Nutr 1985; 41: 810 – 817. diseases, their risk factors, and the contribution of public health and 17. Tsui EY, Gao XJ, Zinman B. Bioelectrical impedance analysis (BIA) personalized approaches. EPMA J 2011; 2: 49 – 57. using bipolar foot electrodes in the assessment of body composition 40. Yatsuya H, Yamagishi K, Iso H. Adiposity and risk of cardiovascu- in mellitus. Diabet Med 1998; 15: 125 – 128. lar diseases in Japan: Secular trend, individual level associations and 18. Xu L, Cheng X, Wang J, Cao Q, Sato T, Wang M, et al. Comparisons causal pathway - implications for the prevention of cardiovascular of Body-Composition Prediction Accuracy: A Study of 2 Bioelectric diseases in societies with rapid economic development. EPMA J Impedance Consumer Devices in Healthy Chinese Persons Using DXA 2011; 2: 65 – 73. and MRI as Criteria Methods. J Clin Densitom 2011; 14: 458 – 464. 41. Arai K, Ishii H, Amano T, Uetani T, Nanki M, Marui N, et al. Volu- 19. Ikeda N, Saito E, Kondo N, Inoue M, Ikeda S, Satoh T, et al. What metric analysis of coronary plaque characterization in patients with has made the population of Japan healthy? Lancet 2011; 378: 1094 – metabolic syndrome using 64-slice multi-detector computed tomog- 1105. raphy. Circ J 2010; 74: 2146 – 2151. 20. Iso H, Date C, Yamamoto A, Toyoshima H, Watanabe Y, Kikuchi S, 42. Kurebayashi N, Yoshikawa D, Ishii H, Sato B, Ando H, Okada T, et et al. Smoking cessation and mortality from cardiovascular disease al. Impact of the low- to high-density lipoprotein cholesterol ratio on among Japanese men and women: The JACC Study. Am J Epidemiol composition of angiographically ambiguous left main coronary ar- 2005; 161: 170 – 179. tery plaque. Circ J 2011; 75: 1960 – 1967. 21. Kondo T, Osugi S, Shimokata K, Honjo H, Morita Y, Maeda K, et 43. Franco OH, Massaro JM, Civil J, Cobain MR, O’Malley B, D’Agostino al. Smoking and smoking cessation in relation to all-cause mortality RB. Trajectories of entering the metabolic syndrome: The Framing- and cardiovascular events in 25,464 healthy male Japanese workers. ham heart study. Circulation 2009; 120: 1943 – 1950. Circ J 2011; 75: 2885 – 2892. 44. Kobayashi J, Murano S, Kawamura I, Nakamura F, Murase Y, 22. Chelland Campbell S, Moffatt RJ, Stamford BA. Smoking and smoking Kawashiri MA, et al. The relationship of percent body fat by bioelec- cessation -- the relationship between cardiovascular disease and lipo- trical impedance analysis with blood pressure, and glucose and lipid protein metabolism: A review. 2008; 201: 225 – 235. parameters. J Atheroscler Thromb 2006; 13: 221 – 226. 23. Hofstetter A, Schutz Y, Jéquier E, Wahren J. Increased 24-hour energy 45. Szendroedi J, Roden M. Ectopic lipids and organ function. Curr expenditure in cigarette smokers. N Engl J Med 1986; 314: 79 – 82. Opin Lipidol 2009; 20: 50 – 56. 24. Liu L, Choudhury SR, Okayama A, Hayakawa T, Kita Y, Ueshima 46. Krssak M, Falk Petersen K, Dresner A, DiPietro L, Vogel SM, H. Changes in body mass index and its relationships to other cardio- Rothman DL, et al. Intramyocellular lipid concentrations are corre- vascular risk factors among Japanese population: Results from the lated with insulin sensitivity in humans: A 1H NMR spectroscopy 1980 and 1990 national cardiovascular surveys in Japan. J Epide- study. Diabetologia 1999; 42: 113 – 116. miol 1999; 9: 163 – 174. 47. Thamer C, Machann J, Bachmann O, Haap M, Dahl D, Wietek B, et 25. Mizuno O, Okamoto K, Sawada M, Mimura M, Watanabe T, Morishita al. Intramyocellular lipids: Anthropometric determinants and rela- T. Obesity and smoking: Relationship with waist circumference and tionships with maximal aerobic capacity and insulin sensitivity. J Clin obesity-related disorders in men undergoing a health screening. J Endocrinol Metab 2003; 88: 1785 – 1791. Atheroscler Thromb 2005; 12: 199 – 204. 48. Eliasson B. Cigarette smoking and diabetes. Prog Cardiovasc Dis 26. Canoy D, Wareham N, Luben R, Welch A, Bingham S, Day N, et al. 2003; 45: 405 – 413. Cigarette smoking and fat distribution in 21,828 British men and 49. Houston TK, Person SD, Pletcher MJ, Liu K, Iribarren C, Kiefe CI. women: A population-based study. Obes Res 2005; 13: 1466 – 1475. Active and passive smoking and development of glucose intolerance 27. Eliasson B, Attvall S, Taskinen MR, Smith U. Smoking cessation among young adults in a prospective cohort: CARDIA study. BMJ improves insulin sensitivity in healthy middle-aged men. Eur J Clin 2006; 332: 1064 – 1069. Invest 1997; 27: 450 – 456. 50. Perkins KA, Epstein LH, Stiller RL, Fernstrom MH, Sexton JE, 28. Chiolero A, Faeh D, Paccaud F, Cornuz J. Consequences of smoking Jacob RG, et al. Acute effects of nicotine on hunger and caloric in- for body weight, body fat distribution, and insulin resistance. Am J take in smokers and nonsmokers. Psychopharmacology (Berl) 1991; Clin Nutr 2008; 87: 801 – 809. 103: 103 – 109.