Title: Accuracy of generalized body composition equations with diverse men and women Authors
|Jackson, Andrew - UNIV. OF HOUSTON|
|Mcfarlin, Brian - UNIV. OF HOUSTON|
Submitted to: Medicine and Science in Sports and Exercise
Publication Type: Abstract Only
Publication Acceptance Date: March 19, 2007
Publication Date: May 7, 2007
Citation: Jackson, A.S., McFarlin, B., Ellis, K., Bray, M.S. 2007. Accuracy of generalized body composition equations with diverse men and women [abstract]. Medicine and Science in Sports and Exercise. 39(5 Suppl):S35. Technical Abstract: The subjects used to develop the Jackson-Pollock (JP) body composition equations published in 1978 and 1980 were primarily white men and women. Current public health and demographic data show that the American population has grown more diverse and obese. To examine the accuracy of the JP equations applied to a diverse sample of men and women. The sample included 817 females (65%) and males (35%) who ranged in age from 17 to 63 y. Self-identified ethnic group was: Caucasian, 33%; Hispanic, 23%; African-American, 27%; Asian-Indian, 6%; Asian, 7%; others, 4%. Dual energy X-ray absorptiometry (DXA) was used to measure % fat. The anthropometric data of these subjects were compared with the 1978 and 1980 JP data. ANOVA showed these men and women were shorter, heavier, and had higher skinfold fat than the JP men and women. These differences produced BMI and %fat mean differences of 3.3 kg/m2 and 4.0% fat for men and 5.7 kg/m2 and 9.4% fat in women. While the correlations between JP and DXA %fat were high (men R=0.95, women r=0.87), graphic and regression analysis of the residuals (DXA %fat - JP %fat) identified systematic bias. The JP equations significantly (p<0.01) under-estimated DXA %fat of men and women, and the error was a function of body fatness. DXA %fat was underestimated in men and women with JP %fat<28% and <45%, respectively. An analysis of the ethnic group residuals showed that the Hispanic, African-American, and Asian-Indian men and women means differed significantly (p<0.05) from 0. This documented an ethnic group bias. JP %fat is highly correlated with DXA %fat, but yields bias estimates of DXA %fat. The likely source of bias is the increase in body fatness and ethnicity of the contemporary men and women, but measurement differences between DXA %fat and %fat derived with the Siri two-component, densitometry model is another possible source of bias.