Estimating Visceral Fat Area by Multifrequency Bioelectrical Impedance

We recruited 84 men and 49 women, aged 20–67 years, from among the students and teaching and administrative staff at our university. All participants were fully informed of the procedures, risk, and discomfort. The study protocol was approved by the bioethics committee of Utsunomiya University.

The details of the study method have been described previously (13). Briefly, we measured anthropometry parameters. BMI was calculated as body weight (in kilograms) divided by the square of height (in meters). Waist-to-height ratio (WHtR) and waist-to-hip ratio were calculated as waist circumference (W) divided by height and by hip circumference, respectively. We took computed tomography (CT) at the umbilical level and calculated VFA and SFA using the image analysis software Fat Scan version 3.0 (N2 System, Hyogo, Japan). The methods were tested against measures of VFA by CT. We measured impedance by the tetrapolar impedance method using multifrequency BI (MFBIA-07; Tanita, Tokyo, Japan). The repeatability of this device was confirmed in the same participants (data not shown). We measured impedance more than 2 h after exercise, eating, and drinking. Impedance was measured at the frequencies of 5, 25, 50, 100, and 250 kHz in the supine posture, and the value measured the first time was used. We used two types of electrode arrangement. First, we placed sensing electrodes symmetrically about the body axis, at a separation of 10 cm. Current electrodes were placed to the right and left of the sensing electrodes. The distance between the current electrodes and sensing electrodes was 10 cm (13). From this electrode arrangement, we obtained the impedance of the whole abdomen (IP_VFA+SFA). After measuring IP_VFA+SFA, we brought the right sensing electrode closer to the right current electrode, to 3 cm. From this electrode arrangement, we obtained the impedance of SFA (IP_SFA) from just under the skin (10). In this study, we considered abdominal composition as a parallel circuit model composed of VFA and SFA. We calculated the impedance of VFA (IP_VFA) for each frequency as IP_VFA = (IP_VFA+SFA × IP_SFA)/(IP_VFA+SFA − IP_SFA).

We randomly assigned participants into two groups, one composed of 103 participants and the other of 30 participants. Stepwise multiple regression analysis was performed on the 103 participants to estimate VFA. Their mean BMI and VFA by CT were 22.67 kg/m² (range 17.19–34.07) and 51.71 cm² (5.55–212.20), respectively. Independent variables were sex, age, anthropometry parameters, and IP_VFA at each frequency. We cross-validated the regression equation against CT-measured VFA in 30 additional participants (VFA range 12.6–159.1 cm²). The Bland-Altman method was used to examine the mean difference and 1.96 SD between VFA observed by CT and that estimated by IP_VFA (14). We calculated the sensitivity and specificity at VFA ≥100 cm² by the regression equation (15). The correlation between impedance and VFA and SFA was examined by Pearson correlation coefficient. All P values were two-tailed, and P < 0.05 was accepted as statistically significant.

The weakest and strongest correlation between impedance obtained at the five frequencies and VFA and SFA were r = 0.734–0.747 (IP_VFA) and r = 0.834–0.872 (IP_SFA), respectively.

The regression equation was VFA = 3.57 × sagittal abdominal diameter + 311.97 × WHtR + 0.71 × age + 23.93 × sex + 1.57 × IP_VFA (250 kHz) − 174.35 (r = 0.904, P < 0.01) (Fig. 1).

Also, we observed a strong correlation in the cross-validation subsample (r = 0.905, P < 0.01). The Bland-Altman method showed a mean difference and 1.96 SD of 0.00 ± 40.78 cm². There was no increasing bias for heavier participants. We observed a high sensitivity and specificity (0.941 and 0.988, respectively) when we discriminated VFA ≥100 cm² or <100 cm² by the regression equation. Meanwhile, waist circumference (W) at the umbilicus level (men: W ≥85 cm, women: W ≥90 cm) is used for screening of VFA ≥100 cm² in Japan (15), thus sensitivity and specificity were 0.882 and 0.919, respectively, by W in our participants.

In this study, because subcutaneous fat layer thickness affected the impedance when electrodes were placed on the abdomen (6), we considered abdominal composition as a parallel circuit model and calculated IP_VFA using the formula for a parallel circuit. Therefore, we eliminated the effect of SFA by this model.

In Japan, waist circumference at the umbilicus level was used to screen for VFA ≥100 cm² because CT has some problems such as radiation exposure (15). However, our regression equation demonstrated higher sensitivity and specificity than waist circumference.

A major strength of our study is that the number of study participants was more than in any previous study (4,–6,13). Additionally, we cross-validated the regression equation and obtained a strong correlation (r = 0.905, P < 0.01). On the other hand, our study has several limitations. First, the study participants were young (mean age ± SD: 30.3 ± 10.8 years), and the proportion of VFA ≥100 cm² was small (16.5%), so we may not be able to adapt this regression equation for middle-aged people who have a higher proportion of VFA ≥100 cm² than young people. Second, the data are limited to the Japanese population, which may have different VFA characteristics than other populations.

Our new method using BI is a simple and convenient method for accurately estimating VFA. We can easily screen excess accumulation of VFA, which is associated with metabolic syndrome. The method may be a useful tool for primary prevention of metabolic syndrome.

This study was supported partly by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science, and Technology Japan (no. 19500592). M.N. is a recipient of a Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists.

No potential conflicts of interest relevant to this article were reported.