The relationship between anthropometry and body composition assessed by dual-energy x-ray absorptiometry in women 75-80 years old : are new skinfold equations needed?

• Main Author: Hill, Andrea Dalton
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/10644

A link between age-related changes in body composition (BC) and the increased prevalence of disease and disability in old age has been well established (Chumlea & Baumgartner, 1989; Going et al., 1995; Shephard, 1997). Consequently, B C assessment is becoming increasingly important in the evaluation o f the health and functional status of the older adult. Individuals 75 years and older comprise one of the fastest growing segments of the population in North America (Canada, 1999; Donatelle & Davis, 1994), yet current B C measurement techniques may not be accurate or reliable in this older age group. The intent of this research was to develop new body fat prediction equations in elderly women based on anthropometry and the criterion method of dual energy X-ray absorptiometry (DEXA), which is considered to be more valid than conventional densitometry among the aging population (Baumgartner et al., 1995; Kohrt, 1998; Visser et al., 1998). Anthropometry, skinfold (SF) anthropometry, and DEXA (Hologic QDR-4500W) body fat data were initially collected in a sample of 43 women 75-80 years old (m = 77.4yrs) as part of a larger study investigating the effects of strength training on strength, function, bone mineral density (BMD), and BC. Eight BC prediction equations for the elderly were selected from the literature and applied to these data. The correlation, between prediction equations and DEXA ranged from 0.76-0.97. However, paired t-tests difference scores (δ) showed that all but one o f the equations overestimated DEXA body fat i n these older aged women (δ ranged from -3.3kg to 4.0kg and 4.4% to 9.0%; p<0.001 in all cases). New equations were derived for FM , %Fat, trunk fat mass (TFM) and percent trunk fat (%TF) using a coffiblnation of stepwise and all possible subsets regression procedures, as both total and regional' percent fat are important health indicators (Going et al., 1995). The following were entered as predictor variables: weight (WT), height (HT), BMI, hip circumference (HC), waist circumference (WC), SF's o f the subscapular (SS), suprailiac (SI), abdominal (ABD), and midaxillary (MA) sites, the SS to triceps skinfold ratio (SSTRI), and the sum o f triceps, biceps, SI and SS (SUM4SF); except H C and SUM4SF were not included in the trunk fat regressions. Ultimately, the measure of interest in body composition assessment is the value %Fat and thus supports using the %Fat equation over that for F M . Moreover, %Fat equation was associated with less error ( C.V.[sub Fat] = 5.9%; C.V.[sub FM] = 6.4%). The %TF equation, however, was less precise than the equation for total %Fat and therefore was not considered further in this research. Subsequent analysis showed the %Fat equation to be internally valid using the jackknife method for data splitting. Finally, %fat equations developed in this study sample were tested in two independent samples of elderly women (71.1yrs and 74.5 yrs) and one sample of younger women (33.4 yrs) shared by Baumgartner (1999) and Brodowicz (1999). Both independent studies used DEXA instruments manufactured by Lunar. New equations were derived for this application using only the variables measured in these independent studies as the predictor variables. The modified prediction equations were reasonably correlated (r = .73, .81) with %Fat from DEXA (Lunar) in the elderly women, yet paired t-tests results showed that the new equations significantly underestimated %fat by 6.6% ± 3.9 (p< 0.001)(BROD), and 5.1% ± 4.5 (p< 0.001)(BAUM). An unexpected finding was the accurate prediction of %Fat in the younger women (δ = -0.7% ± 5.4; p = 0.45). The correlation between predicted and measured %Fat was also stronger (r = .89). However, the two methods were not interchangeable as a trend in the residuals indicated that %Fat was underpredicted at low body fat and overpredicted at high body fat in the younger women. A major finding of this study was that neither existing equations nor the newly derived equations were able to accurately and reliably predict body fat in independent samples of elderly women. Some of the prediction error can be attributed to inter-method differences and differences in DEXA manufacturer, but this lack of agreement also emphasizes the problem of sample specificity with regression equations. Equations will always perform better in the sample from which they were derived and must be interpreted with caution when applied externally. A second major finding of this research was that a single "best" equation did not exist for these data, but rather, several alternative models provided similar equation statistics and regression coefficients. However, the combination of WT, HT (or BMI) and SF's was better than SF's alone. Nonetheless, this study demonstrated that a strong relationship between anthropometry and DEXA exists among elderly women and that internally valid equations can be proposed for this population. Moreover, it is reasonable to conclude that prediction equations based on DEXA have greater face validity in elderly women than those based on densitometry, as the DEXA model is associated with fewer assumptions. Due to.the relatively small sample size, the new %Fat equation cannot be recommended at this time. However, this study shows promise for future use of DEXA and anthropometry in elderly women.