A study was conducted to evaluate the validity of bioelectrical impedance as an accurate assessment of body composition. Four hundred and eighteen male and female volunteers were used as subjects. Each subject reported to the Exercise Physiology Lab at the University of Southern California in a normally hydrated condition. Body composition evaluation was made by hydrostatic weighing (H2O) and bioelectrical impedance (Imp).
H2O was performed in a seated position in a 1000-gallon tank. A minimum of 5 trials was performed on each subject. Residual lung volume was measured, immediately prior to hydrostatic weighing, utilizing the oxygen dilution method and employing a Hewlett-Packard nitrogen analyzer. Body fat was calculated using the formula of Brozek et al. (4.57/Dd-4.142×100%). Bioelectrical impedance was measured (BIO|ANALOGICS’ ELG System) on each subject (prior to H20 weighing) employing recently developed algorithmic equations. The method involves the use of a tetrapolar lead system where an 800uA constant current source at 50khz was applied to the subject and a measurement of impedance (resistance and reactance) was obtained.
Average body fat for the males (N=208) was 15.2% measured by H20 and 14.4% as measured by Imp. The range of body fat for this group was 3-35%. The validity coefficient for this group was r=0.76 and the standard error of estimate (SEE) was 3.24%. The corresponding value was for the female group (N=211) were as follows: Average body fat by H20=23.9%; by Imp=23.4%; range=11-39%; validity coefficient, r=-0.83; SEE=3.15%. Validity for the combined sample was r=-0.88; SEE=3.30%. The results of the present study support the use of this bioelectrical impedance technique as simple, reliable and yet accurate method of assessing the percent of body fat in males and female in the clinical setting.
INTRODUCTION
The relationship between human obesity and long-term health problems has been well-established 2, 25, 27. Obesity and dietary excess have been strongly associated with diseases such as coronary heart disease, stroke, hypertension, diabetes, cancers of the digestive and reproductive systems, and osteoporosis 31. For several decades the use of height/weight tables and Body Mass Index (BMI) has been employed for health assessment and definition of overweight. However, the overweight person is not necessarily the obese person 12. Obesity is precisely defined as an excess of body fat frequently resulting in a significant impairment of health 30. Twenty-five percent of the United States population is reported to be obese. It is clear that the assessment of body composition (body fat and fat-free weight) in humans has become increasingly important not only for research but for clinical applications as well.
Methods for assessing body composition have been available since the 1950’s. Most of these methods involve fairly sophisticated instrumentation and/or techniques, and are considered research laboratory procedures. The NIH has stated that the precise determination of the amount of body fat requires technically sophisticated methods that are available only in research laboratories 30. Excellent reviews of these procedures are available in reviews by Buskirk 5 and Lukaski 21. Recently another method for the prediction of body composition has been developed. This technique has been termed bioelectrical impedance. Basically this method makes use of the fact that electrical conductivity of the human body occurs primarily through the water and electrolytes of the fluid spaces (Total Body Water, TBW) 5, 21. The relationship between impedance and body composition was first shown by Thomasset 26. The technique has been shown to be simple to administer, involving minimal technician training. However, controversy has arisen regarding the accuracy of the bioimpedance technique. Investigators have found the technique to be reliable 19, 20, 21, with reasonable (R=0.70-0.80) 7, 11, 13, 24, 28 to excellent validation against densitometry (R=0.88-0.93) 20. On the other hand, more recent investigators 14, 24 have criticized the prediction accuracy of the technique, reporting SEE’s of between 4.6-6.4% fat, concluding that the equations (RJL Systems) provided for the technique gave unsatisfactory results.
In light of the increasing clinical significance of human body composition and the controversy regarding the impedance methodology, the purpose of the present study was to evaluate the validity of a specific bioelectrical impedance analyzer, which incorporates a recently developed algorithmic method to predict human body composition.
METHODS
A total of 602 subjects (317 females and 285 males) volunteered for body composition evaluation. Subject characteristics appear in Table 1. Subjects reported to the laboratory in a fasted (4-5 hours) and normally hydrated condition. A series of circumference measurements were taken using a cloth metric tape. A complete description of the sites and method appear in a text by Behnke and Wilmore 1. Bioelectrical impedance (BIO|ANALOGICS) was measured in the supine position, with sensor electrodes placed at standard locations on the right wrist and right ankle. An excitation current of 800 uA at 50 kHz was introduced into the subject and the voltage drop was measured. Residual lung volume was measured utilizing the oxygen dilution method described by Wilmore 29. Height was measured by a standard stadiometer and weight was measured to the nearest 50 grn. using a Horns balance beam scale. Hydrostatic weighing was done in a 1000-gallon tank with subjects in a seated position and suspended from a calibrated Chatillon autopsy scale. Underwater trials were continued until weight stabilized (avg. 4-8 trials). Density was calculated according to the method of Goldman and Buskirk 10 and % fat calculated according to the formula of Brozek, et al 4. These methods have been well standardized in our laboratory 8, 9.
RESULTS AND DISCUSSION
The results of the body composition evaluation for the 602 subjects appear in Table 1. The mean values for the males and females are typical for the age profile of the subjects. As can be seen, the average values for densitometry and bioelectrical impedance are very similar, with only 0.5 % fat separating the mean values for the males. Table 3 illustrates the correlation coefficient between % fat, calculated by densitometry (criterion method), and % fat predicted by impedance. The values obtained in the present study were similar in magnitude to the values reported by several investigators 7, 11, 13, 24. We and other investigators 7, 11, 13, 14, 24 have not been able to replicate the extremely high validity coefficients reported by Lukaski, et al 20.
In the present study, the SEE’s for the females and males, respectively, were 3.42% and 3.77% fat (R=0.80, 0.71). These values are similar to those reported by Guo, et al 11, but considerably lower than those reported by other investigators 13, 14, 24. The discrepancies of various bioimpedance techniques validated against densitometry found between investigators may have been caused by use of different formulas in the specific impedance analyzers. Colvin, et al 7 utilized three different analyzers on the same subjects and reported R-values ranging from 0.72-0.81 and SEE’s from 2.4-4.9%.
Jackson, et all 14 have criticized the impedance method due to the low validity coefficients and high SEE’s they found (R=0.71-.76 and SEE=4.6-6.4% fat) (RJL Systems). Furthermore, they reported that BMI is at least as good a predictor of % fat as the impedance device they tested with correlation of R=0.74-0.75 and SEE’s of 4.3-4.8%. In the present study the R-values for BMI were 0.69, SEE’s were 4.1 and 4.6%. Our data would indicate that the bioelectrical impedance system tested is certainly a better predictor of body composition than BMI.
The use of anthropometric methods, such as skinfolds and circumferences have long been advocated as convenient, inexpensive, and simple procedures to predict body composition 1, 15. Unfortunately, many of the prediction formulas generated by these methods (especially skinfolds) tend to be population specific. Jackson, et al 14 have reported validity coefficients of R=0.92 (males) and R=0.88 (females) between the sum of 7 skinfolds and % fat measured by densitometry. The SEE’s were 2.6-3.6% fat. However, some investigators 6, 18 have indicated that the prediction accuracy of the skinfold method is approximately ±5% body fat. Pollock and Jackson 23 cautioned that the validity of estimating body density in the obese may be suspect when using skinfold fat measures. Similarly, Bray, et al 3 found that skinfold measurements had high variability, and considerable interobserver variability, and concluded that the use of skinfold calipers is not to be recommended in assessing or following obese patients.
Several investigators 13 have evaluated the combination of bioelectrical impedance and anthropometric variables in predicting body composition. Guo, et al, used a combination of skinfolds and circumferences together with impedance. The best of skinfolds and circumference measurements for the female group yielded a validity coefficient of R=0.85 with a SEE of 3.83% fat, when compared with densitometry. The addition of impedance to the regression formula increased the R to 0.90 and decreased the SEE to 3.22%. For the male group a similar analysis resulted in a smaller SEE (from 3.32% to 3.28%), but no change in the R-value. Hodgdon and Fitzgerald 13 added a combination of skinfolds; and circumferences to impedance in an attempt to improve the prediction accuracy. With the anthropometric values the R-values increased from 0.79 and 0.82 to 0.87 and 0.87, for the male and female groups respectively. Similarly, the SEE’s were reduced from 5.01 and 4.25% to 3.90 and 3.61%, for males and females.
In the present study circumference measures were obtained on 419 (211 females and 208 males) of the original 602 subjects. Utilizing prediction equations (BioAnalogics) which incorporated a series of circumference measures (including circumference ratios) in combination with bioelectrical impedance, percent body fat was determined. For the female group the multiple correlation was R=0.89 (R2=79%) and the SEE=2.68%. The values (R=0.83; R2=68%; SEE=3.03%) for the males, though not as good as the female group, still demonstrated a substantial improvement in predictive ability with the addition of the circumference equations. Our data is in agreement with that of Hodgdon and Fitzgerald 13 who have shown an improvement in both correlation and SEE with a combination of anthropometric and impedance measurements. The SEE’s found in the present study are less than those reported by Hodgdon and Fitzgerald and the present technique did not require the use of skinfolds.
CONCLUSIONS
The evaluation of body composition is an important component of a preventive health examination, especially considering the incidence of obesity in the United States. Most of the accurate methods for measuring body fat are limited to the research laboratory or are extremely expensive. The bioelectrical impedance system tested has been shown to be a simple, reliable and yet accurate method of assessing the percent of body fat in males and females in the clinical setting. The present study also indicates that the inclusion of circumference measurement equations added substantially to the accuracy of the impedance method and makes it an ideal technique that can be implemented in research laboratories as well as the non-research clinical setting.
