There is a high demand for affordable body composition assessments of body fat (%BF) and fat-free mass. Research demonstrates that the American College of Sport’s Medicine (ACSM) recommended Jackson and Pollock (JP) skinfold prediction equations underestimate %BF. PURPOSE: The purpose of this study was to validate an alternative equation for women created from dual energy x-ray absorptiometry (DXA). The DXA criterion (DC) equation is: %BF= -6.40665 + 0.491946(S3SF) – 0.00126(S3SF)2 + 0.12515(hip) + 0.06437(age); where S3SF = sum of triceps, suprailiac, thigh; hip = circumference in cm; age = years. METHODS: Anthropometrics (skinfolds and circumferences) and a DXA scan were completed on 78 women (mean ± SD) [age: 28.0 ± 10.1 yr., height: 165.1 ± 5.9 cm, mass: 63.5 ± 10.5 kg., BMI: 23.2 ± 3.2 kg/m2]. Three JP skinfold prediction equations (JP7, JP3a, and JP3b) and the DC equation were compared to DXA %BF. RESULTS: Two-way ANOVA with repeated measures detected significant differences (p < 0.05) in the %BF with post hoc-comparisons revealing significant differences among JP7 (21.4 ± 5.8), JP3a (22.3 ± 5.9), and JP3b (22.7 ± 5.7) as compared to the DXA (26.6 ± 5.4). No significant difference existed between DC %BF (26.6 ± 5.6) and DXA %BF (26.6 ± 5.4) (p = 1.0) and the two assessments were highly correlated (R = 0.87). The standard error of the measurement for the DC equation was low (2.92%). CONCLUSION: The DC equation more accurately predicted %BF across a general population of women than the recommended ACSM equations.
Keywords: body composition, skinfolds, DXA
The body is composed of water, protein, minerals, and adipose tissue. Excess adipose tissue (obesity) has been shown to be detrimental to human health and has been linked to multiple medical conditions such as diabetes mellitus, heart disease, and several types of cancer (Wellens et al., 1996). Maintaining a healthy percentage of body fat is essential in minimizing the occurrence of these negative conditions (Steinberger et al., 2005). According to the Centers for Disease Control and Prevention, nearly 64% of Americans are classified as overweight or obese (Ogden, Carroll, Kit, & Flegal, 2014). With the increasing prevalence of obesity, the need for accessible and accurate body composition assessments has been in high demand.
Body composition can be measured in a variety of ways. The most commonly used assessment of body composition is the two-component model (2C) which assesses fat mass (FM) and fat free mass (FFM)(Wagner & Heyward, 1999; Wellens et al., 1996). FFM erroneously assumes that water, bone, and muscle are of equal density (Van der Ploeg, Gunn, Withers, & Modra, 2003). Commonly, the 2C model involves measuring body density (Db) and then using a conversion formula, such as the Siri equation to estimate body fat (%BF) (Wagner & Heyward, 1999). Various 2C assessments include skinfolds (SF), hydrostatic weighing (HW), air displacement plethysmography (ADP), and bioelectrical impedance (BIA). With advancements in technology, three, four, and five component (3C, 4C, and 5C) models have been developed to resolve the error associated with 2C models. Multi-component models, combining body density (Db), total body water (TBW), and bone mineral density (BMD) data are frequently used to derive criterion measures of body composition (Wagner & Heyward, 1999; Withers, Laforgia, & Heymsfield, 1999).
Dual-energy x-ray absorptiometry (DXA), a 3C model, uses x-rays to scan the body and divide it into three components: FM, FFM, and BMD (Ellis, 2000; Pietrobelli, Formica, Wang, & Heymsfield, 1996). Research has shown that BMD varies among individuals contributing an additional advantage for utilizing the DXA compared to various 2C models. We have shown, in concordance with other literature, that SF equations based on only 2C underestimate %BF when compared to DXA (Ball, Altena, & Swan, 2004a; Ball, Cowan, Thyfault, & LaFontaine, 2014; Ball, Swan, & Desimone, 2004b; Sardinha, Lohman, Teixeira, Guedes, & Going, 1998; Williams, Going, Lohman, Hewitt, & Haber, 1992; Withers et al., 1998). In a group of 150 women aged 18-65 yr., %BF calculated with the original Jackson and Pollock (JP) prediction equations (Jackson & Pollock, 1985; Jackson, Pollock, & Ward, 1980) was significantly underestimated when compared to DXA (Ball et al., 2004b). Similar results have been found in men (Ball et al., 2004a). Due to the lack of accessibility to laboratory techniques such as the DXA, anthropometrics remains an important and inexpensive tool for health professionals. The need to develop an anthropometric equation developed from a 3C model will improve the accuracy of these field methods. In 2004, our laboratory published two newly developed SF prediction equations for men and women using DXA as the criterion model (Ball et al., 2004a; Ball et al., 2004b). The prediction equation for men has been cross-validated and yielded a low standard error of the estimate (2.72%) and was highly correlated (R = 0.87) with the DXA (Ball et al., 2014). However, no validation study has been completed since the female equation was published.
It is hypothesized that the new female DC equation will more accurately predict %BF than previously recommended anthropometric equations created by Jackson and Pollock. The purpose of this study was to validate the DC SF equation in an independent sample of women. The secondary purpose was to compare the currently recommended JP SF equations to the DXA.
Eighty women were recruited for the study. All subjects signed an informed consent reviewed and approved by University of Missouri Human Subjects Institutional Review Board. Subjects were between the ages of 18-57 years old and completed the study with 100% compliance rate.
Subjects reported to the laboratory having fasted 4 hours, avoided caffeine for 12 hours, avoided alcohol for 24 hours, and avoided exercising 6 hours. Participants were instructed to wear a sports bra and spandex shorts or a swimsuit. All body composition (anthropometrics and DXA) measurements were completed within one hour of each other on the same day in the morning.
Height was taken to the nearest 0.1 centimeter (cm) using a Seca216 stadiometer (Seca, Chino, CA) and weight was taken to the nearest 0.10 kilogram utilizing a calibrated scale (Cosmed, Concard, CA). Hip circumference (largest protrusion of the buttocks) were measured to the nearest 0.1 cm using a measuring tape. Body mass index (BMI) (kg/m2) was calculated.
Seven SF sites (subscapular, tricep, midaxillary, chest, abdomen, suprailiac, and thigh) were measured in millimeters (mm) using a Lange caliper (Cambridge Scientific Industries, Cambridge, MD). Each site was measured twice and averaged. If the first two measurements differed by more than 2 mm apart, a third measurement was taken with the average of the two closest measurements calculated. Four SF equations were used to predict %BF (table 1). The first three equations were developed by Jackson and Pollock (JP7, JP3a, and JP3b) and are currently recommended by ACSM. JP7, JP3a, an dJP3b were used to find Db and the Siri equation was then used to calculate %BF from Db (Siri, 1956). The Female DC equation was the fourth equation used to predict %BF (Ball et al., 2004b).
Participants completed one whole body DXA scan on with model QDR Discovery-A (Hologic, Inc., Bedford, MA) and wore minimal clothing with no metal. The subjects were placed supine on the DXA table and were positioned according to manufacturer instructions. Body composition was found using QDR computer software and results for fat, lean, and bone mineral mass was recorded in grams.
Data were analyzed using SPSS v24 (SPSS Inc., Chicago, IL) and Microsoft Excel for windows (Microsoft Corp., Seattle, WA). Values are expressed as means ± SD. A two-way ANOVA with repeated measures was performed to detect significant differences in % BF among the five testing methods (JP7, JP3a, JP3b, DC, and DXA); an alpha level of 0.05 was used for significance testing. A Bonferroni Post-hoc analysis and Bland-Altman plots were used to compare each SF method to the DXA. The standard error of the estimate (SEE) was found to examine the accuracy of the DC equation. A Pearson’s Product Moment correlation was computed to examine the relationship between the %BF estimates from assessments that were significantly different from DXA.
Descriptive statistics for 78 women who participated in the study can be found in table 2. Mean age was 28.0 ± 10.1 yr., and mean BMI was 23.2 ± 3.2 kg/m2. Two subjects completed the study but were excluded from analysis. Exclusion criteria included a BMI > 35 kg/m2 and was based on the consistent finding that SFs are inaccurate on obese populations (Ball et al., 2014; Gray et al., 1990).
Two-way ANOVA with repeated measures detected significant differences (p < 0.05) in the %BF with Bonferonni post hoc-comparisons revealing significant differences among JP7 (21.4 ± 5.8), JP3a (22.3 ± 5.9), and JP3b (22.7 ± 5.7) as compared to the DXA (26.6 ± 5.4) (table 3). No significant difference existed between DC %BF (26.6 ± 5.6) and DXA %BF (26.6 ± 5.4) (p = 1.0). All prediction equations were highly correlated with the DXA (table 3). The standard error of the measurement for the DC equation was low (2.92%) when using the DXA as a criterion.
Nine participants [age: 26.7 ± 5.3 yr., height: 162.6 ± 4.2 cm, mass: 60.4 ± 7.4 kg., BMI: 22.8 ± 2.4 kg/m2] underwent two 7-site SF assessments and two DXA scans. The mean %BF of the two DXA trials was 25.6 ± 5.4 and 26.0 ± 5.6. No statistical difference was found between the two trials (p < 0.05). The mean difference between trials was 0.34% and the trials were highly correlated (r = 0.99). Using the same nine subjects, the sum of the seven SF site for each trial had means of 112.1 ± 36.8 and 110.6 ± 35.7. No significant difference was found between the two trials (p < 0.05) and were highly correlated (r = 0.99).
Differences Between Methods
A Bland Altman plot (Figure 1) was used to compare the DC equation to DXA. Each data point represents the difference between %BF derived from the DC equation and DXA for each participant. The vertical axis represents the difference of the two measurements and the horizontal axis represents the average of the two measurements. If two methods are comparable, more data points will be closer to zero meaning the differences are small (Bland & Altman, 1986). In figure 1, the data points are clustered within 2 SDs suggesting there is little difference between the two measurements. Bland Altman plots were also used to investigate proportional bias. Proportional bias would indicate that the methods do not agree equally through the range of measurements and that a change in x is directly related to the change in y (Ludbrook, 1997). To evaluate this relationship, the difference between the methods were regressed resulting in a non-significant beta coefficient (p = 0.552) indicating no proportional bias. Bland Altman plots were used as a visual representation of the differences seen between JP7, JP3a, and JP3b compared to the DXA (Figure 2).
The purpose of this study was to validate the DC SF equation in an independent sample of women. The secondary purpose was to compare the currently recommended Jackson and Pollack SF equations to the DXA. The recommended JP SF equations were developed in the 1980s using 2C models (Jackson et al., 1980). Our data coincides with other reports that the current ACSM equations underestimate %BF and these equations need to be re-evaluated with newer laboratory methods.
The sample of this study is similar to that of Ball et al. (2004b) from which the DC equation was developed and Jackson et al. (1980) from which the professionally recommended SF equations were developed.Age, height, and weight for this validation sample [age: 28.0 ± 10.1 yr., height: 165.1 ± 5.9 cm, mass: 63.5 ± 10.5 kg.] as compared to Ball et al. (2004b) [age: 28.7 ± 8.6 yr., height: 165.1 ± 6.7 cm, mass: 63.3 ± 13.3 kg.] and to the Jackson (1980) [age: 31.4 ± 10.8 yr., height: 165.0 ± 6.0 cm, mass: 57.1 ± 7.6 kg.] All women who participated in these experiments were mostly Caucasian and body fat was normally distributed across a wide range of fatness (16.3 – 44.0%).
According to Lohman (1992), a SF prediction equation will have a low SEE, high correlation to the criterion method, and a small mean difference. The DC equation had a SEE of 2.92%, was highly correlated (r = 0.85) with the DXA, and the mean difference between the two was the absolute minimum (0%). JP7, JP3a, an dJP3b had a SEE of 3.07%, 3.13%, and 3.30% respectfully, and as previously seen in Ball et al. (2004b), high correlations were seen between the three JP SF equations (table 3). However, there was a systemic underestimation in %BF by 3-5.5%. The high correlations and consistent underestimation suggests the difference between these equations may underlie with the criterion method utilized when developing these equations. The three JP equations used Db determined by HW as the criterion method, a 2C model and not a multi-component model, such as DXA.
HW has been viewed to be the criterion method, historically, for measuring body composition (Brožek, Grande, Anderson, & Keys, 1963). However, 2C models make the incorrect assumption that FFM has a consistent density of 1.1000g/cc despite age, gender, fatness, or activity status (Brožek et al., 1963; Withers et al., 1999). Withers et al. (1998) utilizing a 4C model of composition, demonstrated the FFM density to be significantly greater (1.1075 g/cc) than the assumed density. Therefore, HW significantly underestimated %BF by 2.3-2.8% which is similar to the mean difference between DXA and the three prediction equations observed in this study. Current research suggests DXA is the criterion method in modern day. Despite the limitation that the DXA assumes the hydration of soft lean tissue (non-bone and nonfat) is 73 g/mL, research indicates there are better agreement and no significant differences between the DXA and 4C models leading to acceptance of the DXA as a criterion method (Prior et al., 1997).
The current study is similar to previous research conducted in our laboratory. The DC equation for men was validated on a separate sample of men from which it was created (Ball et al., 2014). Results showed excellent agreement between the DXA (18.0 + 5.9%) and the DC equation for men (19.1 + 6.3%), in addition to a low SEE (2.72%) and high correlation (R = 0.937). The newer DC SF equation for men has been utilized in various anthropometry studies and has been demonstrated to have the highest level of agreement when compared to other newly developed SF equations (Knechtle et al., 2011). Knechtle et al. (2011) utilized various equations in ultra-endurance athletes, a population not assessed in our study, and the DC equation performed better than the other tested SF equations. The current results show similar promise for the DC equation for women.
Limitations to this study include a fairly young (28.0 ± 10.1 yr.) sample with limited ethnicity. ACSM recommends using different density equations based on age and ethnicity. This equation was developed and now validated with a predominately white population. We do not know how our equation would fair with other groups when compared to the DXA. A strength to this research study is that the equation was developed with DXA model QDR 4500a and then validated with model QDR Discovery-A. Research suggests there is little difference between the two models (Economos et al., 1997; Kolta, Ravaud, Fechtenbaum, Dougados, & Roux, 1999; Tataranni, Pettitt, & Ravussin, 1996) and strengthens the validity and reliability of the DC equation. Not only was the equation developed with two different DXA models, 14 years has elapsed between the development of the DC equation and its validation. While this can be construed as a weakness, we see this as a strength to our study – after 14 years since its development, it is still the more accurate equation of choice compared to the JP SF equations.
The purpose of this study was to validate the DC SF equation developed for women in 2004. The data presented confirms that the DC equation more accurate predicts %BF with less error across a general population of women than the three recommended equations developed by JP compared to the DXA. Field-based professionals should consider using the DC SF equation to predict %BF in women and yield caution to using previous recommended SF equations. This is the first study to validate this equation and further validation studies are needed to further convince practitioners of its accuracy.
What does this article add?
This article demonstrates the need for change in the current recommended equations used for SF assessment. There is, without question, an international obesity epidemic and the prevalence of obesity continues to grow. With that, practitioners need to be using methods that will give individuals the most accurate information about their body composition and risk for developing cardio-metabolic diseases. SF equations overestimate Db and consequently underestimate %BF are outdated and need to be retired. While these equations were the first developed to allow field-based body composition assessments, the advancements in technology has superseded the use of SF equations created from 2C models. There is a universal goal to combat the obesity epidemic, however, many practitioners do not have access to clinical methods to assess body composition. Therefore, SF equations for assessing body composition need to be more precise.
Ball, S., Altena, T., & Swan, P. (2004a). Comparison of anthropometry to DXA: a new prediction equation for men. European Journal of Clinical Nutrition, 58(11), 1525-1531.
Ball, S., Cowan, C., Thyfault, J., & LaFontaine, T. (2014). Validation of a New Skinfold Prediction Equation Based on Dual-Energy X-Ray Absorptiometry. Measurement in Physical Education and Exercise Science, 18(3), 198-208.
Ball, S., Swan, P., & Desimone, R. (2004b). Comparison of anthropometry to dual energy X-ray absorptiometry: a new prediction equation for women. Research quarterly for exercise and sport, 75(3), 248-258.
Bland, J. M., & Altman, D. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. The lancet, 327(8476), 307-310.
Brožek, J., Grande, F., Anderson, J. T., & Keys, A. (1963). Densitometric analysis of body composition: revision of some quantitative assumptions. Annals of the New York Academy of Sciences, 110(1), 113-140.
Economos, C., Nelson, M., Fiatarone, M., Dallal, G., Heymsfield, S., Wang, J., . . . Russell-Aulet, M. (1997). A multi-center comparison of dual energy X-ray absorptiometers: in vivo and in vitro soft tissue measurement. European Journal of Clinical Nutrition, 51(5), 312-317.
Ellis, K. J. (2000). Human body composition: in vivo methods. Physiol Rev, 80(2), 649-680.
Gray, D. S., Bray, G. A., Bauer, M., Kaplan, K., Gemayel, N., Wood, R., . . . Kirk, S. (1990). Skinfold thickness measurements in obese subjects. The American journal of clinical nutrition, 51(4), 571-577.
Jackson, A. S., & Pollock, M. L. (1985). Practical assessment of body composition. The Physician and Sportsmedicine, 13(5), 76-90.
Jackson, A. S., Pollock, M. L., & Ward, A. (1980). Generalized equations for predicting body density of women. Med Sci Sports Exerc, 12(3), 175-181.
Knechtle, B., Wirth, A., Knechtle, P., Rosemann, T., Rust, C. A., & Bescos, R. (2011). A comparison of fat mass and skeletal muscle mass estimation in male ultra-endurance athletes using bioelectrical impedance analysis and different anthropometric methods. Nutr Hosp, 26(6), 1420-1427. doi:10.1590/s0212-16112011000600032
Kolta, S., Ravaud, P., Fechtenbaum, J., Dougados, M., & Roux, C. (1999). Accuracy and precision of 62 bone densitometers using a European Spine Phantom. Osteoporosis international, 10(1), 14-19.
Lohman, T. G. (1992). Advances in body composition assessment: Human Kinetics Publishers.
Ludbrook, J. (1997). Comparing methods of measurements. Clin Exp Pharmacol Physiol, 24(2), 193-203.
Ogden, C. L., Carroll, M. D., Kit, B. K., & Flegal, K. M. (2014). Prevalence of childhood and adult obesity in the United States, 2011-2012. Jama, 311(8), 806-814.
Pietrobelli, A., Formica, C., Wang, Z., & Heymsfield, S. B. (1996). Dual-energy X-ray absorptiometry body composition model: review of physical concepts. American Journal of Physiology-Endocrinology And Metabolism, 271(6), E941-E951.
Prior, B. M., Cureton, K. J., Modlesky, C. M., Evans, E. M., Sloniger, M. A., Saunders, M., & Lewis, R. D. (1997). In vivo validation of whole body composition estimates from dual-energy X-ray absorptiometry. Journal of Applied Physiology, 83(2), 623-630.
Sardinha, L. B., Lohman, T. G., Teixeira, P. J., Guedes, D. P., & Going, S. B. (1998). Comparison of air displacement plethysmography with dual-energy X-ray absorptiometry and 3 field methods for estimating body composition in middle-aged men. The American journal of clinical nutrition, 68(4), 786-793.
Siri, W. E. (1956). The gross composition of the body. Adv Biol Med Phys, 4(239-279), 513.
Steinberger, J., Jacobs, D. R., Raatz, S., Moran, A., Hong, C. P., & Sinaiko, A. R. (2005). Comparison of body fatness measurements by BMI and skinfolds vs dual energy X-ray absorptiometry and their relation to cardiovascular risk factors in adolescents. Int J Obes (Lond), 29(11), 1346-1352. doi:10.1038/sj.ijo.0803026
Tataranni, P., Pettitt, D., & Ravussin, E. (1996). Dual energy X-ray absorptiometry: inter-machine variability. International journal of obesity and related metabolic disorders: journal of the International Association for the Study of Obesity, 20(11), 1048-1050.
Van der Ploeg, G., Gunn, S. M., Withers, R., & Modra, A. (2003). Use of anthropometric variables to predict relative body fat determined by a four-compartment body composition model. European Journal of Clinical Nutrition, 57(8), 1009-1016.
Wagner, D. R., & Heyward, V. H. (1999). Techniques of body composition assessment: a review of laboratory and field methods. Research quarterly for exercise and sport, 70(2), 135-149.
Wellens, R. I., Roche, A. F., Khamis, H. J., Jackson, A. S., Pollock, M. L., & Siervogel, R. M. (1996). Relationships between the Body Mass Index and body composition. Obes Res, 4(1), 35-44.
Williams, D. P., Going, S. B., Lohman, T. G., Hewitt, M. J., & Haber, A. E. (1992). Estimation of body fat from skinfold thicknesses in middle‐aged and older men and women: A multiple component approach. American journal of human biology, 4(5), 595-605.
Withers, R. T., Laforgia, J., & Heymsfield, S. (1999). Critical appraisal of the estimation of body composition via two‐, three‐, and four‐compartment models. American journal of human biology, 11(2), 175-185.
Withers, R. T., LaForgia, J., Pillans, R., Shipp, N., Chatterton, B., Schultz, C., & Leaney, F. (1998). Comparisons of two-, three-, and four-compartment models of body composition analysis in men and women. Journal of Applied Physiology, 85(1), 238-245.