Abstract
A new body adiposity index (BAI = (hip circumference)/((height)1.5) − 18) has been developed and validated in adult populations. We aimed to assess the validity of BAI in an older population. We compared the concordance correlation coefficient between BAI, body mass index (BMI), and percent body fat (fat%; by dual-energy X-ray absorptiometry) in an older population (n = 954) participating in the Baltimore Longitudinal Study of Aging. BAI was more strongly correlated with fat% than BMI (r of .7 vs .6 for BAI vs BMI and fat%, respectively, p < .01) and exhibited a smaller mean difference from fat% (−5.2 vs −7.6 for BAI vs BMI and fat%, respectively, p < .01) indicating better agreement. In men, however, BMI was in better agreement with fat% (r of .6 vs .7 for BAI vs BMI and fat%, respectively, p < .01) with a smaller mean difference from fat% (−3.0 vs −2.2 for BAI vs BMI and fat%, respectively, p < .01). Finally, BAI did not accurately predict fat% in people with a fat% below 15%. BAI provides valid estimation of body adiposity in an older adult population; however, BMI may be a better index for older men. Finally, BAI is not accurate in people with extremely low or high body fat percentages.
Key Words: Body adiposity, BMI, Older population.
The prevalence of obesity in the older population has risen over the past few decades both in the United States and worldwide (1). In the United States, nearly 70% of people over the age of 65 are overweight or obese (2). This is problematic as obesity leads to increased morbidity and mortality, and decreased function and quality of life in older individuals (3). Importantly, at advanced age, people tend to have disproportionately more fat than lean mass, and less fat is distributed subcutaneously (4). These changes, however, are not captured by simple body weight or relative body weight for height assessment.
Currently, the gold standard of assessing body adiposity is dual-energy X-ray absorptiometry (DXA) (5). Although DXA is accurate, it is not routinely used in clinical settings or large-scale research studies because it is costly and time consuming. Other methods such as bioelectrical impedance analysis or skinfold thickness are quicker and less expensive but are not as accurate in an older population (6–8). Up until recently, body mass index (BMI = (weight (kg))/(height (m2))) was the most widely used adiposity index. This index has been validated (9–11) and shown to correlate with body fatness. However, BMI has been proven inaccurate in some occasions and does not account for fat mass versus fat-free mass (12). Bedogni and colleagues (13) showed that in older women, BMI only explained 54.8% of fat percentage variance measured by DXA with a high prediction error of 15%. This poor association between BMI and body percent body fat (fat%) cautions against using BMI as an adiposity index in older individuals.
Bergman and colleagues (5) recently developed a new index for body adiposity called the body adiposity index (BAI). This index uses both hip circumference (HC) and height (BAI = (HC)/((height)1.5) − 18). They found that unlike BMI, BAI correlated well with body fat percentage in both men and women without statistical correction (5). Thus far, however, BAI has only been validated in Mexican American, white, and black adults (ages 41±13.4 years) (14). To date, no information on the validity of this index has been established in an older population. Therefore, the purpose of this study was to determine the validity of the newly developed BAI in an older population from the Baltimore Longitudinal Study of Aging, consisting of both men and women of different races. In this study, we included all individuals aged 55 years and older as the total sample, and we also did analyses on people aged 65 years and older. Our hypothesis was that BAI would be positively correlated with measured fat% in an older population (regardless of sex or race) without statistical adjustment and would correlate more strongly with measured fat% than BMI.
Methods
Study Population
The participants for this study were a subgroup from Baltimore Longitudinal Study of Aging, which is a long-term, open-panel study on normal human aging. Study design and recruitment for the Baltimore Longitudinal Study of Aging has been described in detail previously (15). Most of the participants were community-dwelling individuals from the Washington–Baltimore area. The study population consists of all participants aged 55 years and older seen between 2003 and 2012 with complete data on height, weight, HC, and body fat percentage. There were 954 participants (483 men and 471 women) for the final data analysis.
Anthropometric Measurements
Height, weight, and HC were measured as previously described. Briefly, height and weight were measured with participants wearing a light hospital gown with no shoes. HC was measured at the point of maximal protrusion of the gluteal muscle, and in the anterior place, the symphysis of the pubis. BMI was calculated as weight (kilograms) divided by the square of height (meters). HC and height were used to calculate BAI (BAI = (HC)/((height)1.5) − 18). Fat% was assessed by DXA (model DPX-L Lunar Radiation, Madison, WI) as previously described (16).
Statistical Analyses
Lin’s concordance correlation was used to assess the concordance between BAI and DXA-derived fat percentage (%) to see if BAI was in agreement with the “gold standard” (DXA), and thus a valid index in predicting fat%. Lin’s concordance can be expressed as Pearson correlation coefficient multiplied by mean shift–related bias correction factor. It is used to assess the agreement of a new measurement method to established gold standard (17). To illustrate the concordance of BAI and BMI with DXA-derived fat%, we used Tukey mean–difference plot (also known as Bland–Altman plot). Pearson correlation was also performed to assess which index (BMI or BAI) correlated more strongly with fat%. To study sex and race differences, we divided the total sample into sex- and race-specific subgroups. Linear regression models were used to assess the contribution of BAI and BMI in explaining the variance of fat%. Mean differences were calculated between BMI and fat% and BAI and fat% and then compared using a t test. Statistical significance was set as p < .05.
Results
The participants were 55–96 years old and were primarily white (63.2%) and black (24.0%). Because the sample size was too small to produce a good effect size to examine any other ethnic groups, we included only white and black participants in the race analysis. Participant characteristics are shown in Table 1.
Table 1.
Participant Characteristics in the Present Sample of BLSA Study
| Total (n = 954) | Women (n = 471) | Men (n = 483) | |
|---|---|---|---|
| Age (y) | 70.4±9.5 | 68.9±9.6 | 71.9±9.2* |
| Age ≥ 65 (y) (n) | 640 | 268 | 372 |
| Ethnicity (n) | |||
| White (n) | 603 | 346 | 394 |
| Black (n) | 229 | 139 | 90 |
| Native American (n) | 7 | 3 | 4 |
| Hispanic (n) | 10 | 6 | 4 |
| Asian (n) | 40 | 22 | 18 |
| Hawaiian (n) | 28 | 8 | 20 |
| Not classified (n) | 37 | 18 | 19 |
| Weight (kg) | 77.8±16.1 | 70.8±14.7 | 84.7±14.3* |
| Height (m) | 1.7±0.1 | 1.6±0.1 | 1.8±0.1* |
| BMI (kg/m2) | 27.2±4.7 | 26.9±5.2 | 27.5±4.1* |
| BAI | 29.6±5.9 | 32.5±5.9 | 26.8±4.2* |
| Hip circumference (cm) | 104.0±10.8 | 104.2±12.1 | 103.8±9.3 |
| Body fat percentage (%) (DXA) | 34.8±8.8 | 40.0±7.3 | 29.8±7.2* |
Notes: Values are means ± SDs unless otherwise specified. BAI = body adiposity index; BLSA = Baltimore Longitudinal Study of Aging; BMI = body mass index; DXA = dual-energy X-ray absorptiometry; n = number of participants.
*Significant difference between men and women (p < .05).
Lin’s concordance correlation coefficient analysis showed that concordance (R c) between BAI and fat%, and BMI and fat% was significant in all groups (Table 2). Further, the R c between BAI and fat% was stronger than that between BMI and fat% in total sample and in most of the subgroups (women, white, and black), but not for men. Pearson correlations showed that both BAI and BMI were significantly correlated with fat% (from DXA) in the total sample and in all subgroups including women, men, white, black, and participants aged 65 years and older (p < .01). Additionally, BAI was more strongly correlated with fat% than BMI in the total sample, in the older subgroups aged 65 years and older and in both white and black groups (p < .01). However, BAI was not more strongly correlated with fat% than BMI when controlled for sex (age ≥ 55 and age ≥ 65).
Table 2.
Concordance Correlation Coefficients and Pearson Correlations for BAI and BMI vs Fat% (from DXA) in the Total Sample and Subgroups
| Number of Participants | Concordance Correlation Coefficient (R c) (95% CI) | Pearson Correlation R (95% CI) | |||
|---|---|---|---|---|---|
| BAI and Fat%* | BMI and Fat%* | BAI and Fat%* | BMI and Fat%* | ||
| Total (age ≥ 55) | 954 | .55 (0.51–0.58) | .30 (0.27–0.33) | .74 (0.71–0.77)† | .58 (0.53–0.62) |
| Women | 471 | .43 (0.38–0.48) | .24 (0.21–0.27) | .72 (0.67–0.76) | .80 (0.77–0.83) |
| Men | 483 | .42 (0.37–0.48) | .57 (0.52–0.61) | .55 (0.49–0.61) | .71 (0.66–0.75) |
| White | 603 | .52 (0.47–0.56) | .28 (0.24–0.32) | .70 (0.66–0.74)† | .54 (0.48–0.59) |
| Black | 229 | .61(0.54–0.66) | .31 (0.25–0.37) | .81 (0.76–0.85)† | .62 (0.53–0.69) |
| Age ≥ 65 | 640 | .54 (0.50–0.58) | .30 (0.26–0.34) | .71 (0.67–0.75)† | .57 (0.52–0.62) |
| Women | 268 | .45 (0.39–0.51) | .24 (0.20–0.27) | .71 (0.65–0.77) | .81 (0.77–0.85) |
| Men | 372 | .40 (0.33–0.46) | .54 (0.49–0.59) | .52 (0.44–0.59) | .71 (0.65–0.76) |
| White | 409 | .52 (0.47–0.57) | .31 (0.26–0.36) | .68 (0.63–0.73)† | .56 (0.49–0.62) |
| Black | 137 | .59 (0.51–0.70) | .29 (0.21–0.37) | .78 (0.70–0.84)† | .60 (0.48–0.70) |
Notes: BAI was calculated as hip circumference in centimeters divided by height in meters to the power of 1.5 minus 18. BMI was calculated as weight in kilograms divided by height in meters squared. BAI = body adiposity index; BMI = body mass index; CI = confidence interval; Fat% = body fat percentage.
*Significant correlation with fat% (p < .01).
†Correlation between BAI and fat% was significantly stronger than correlation between BMI and fat% (p < .01).
We calculated the mean differences between BAI and fat% as well as between BMI and fat% (Table 3), and further demonstrated the agreement with Bland–Altman plots in which the difference (BAI − fat% or BMI − fat%) was plotted against the mean (BAI and fat% or BMI and fat%; Figure 1). Mean difference between BAI and fat% was significantly smaller (indicating better agreement) than that between BMI and fat% (indicating worse agreement) in the total sample and all subgroups (p < .01), except for the men subgroup (Table 3, Figure 1). In the men subgroup (both age ≥ 55 and age ≥ 65 years), BMI and fat% actually had significantly lower mean differences than BAI and fat% did (p < .01). Bland-–Altman plots showed that both BMI and BAI underestimated fat%, but BAI had a smaller mean difference than BMI did (Figure 1) except for in the men subgroup. In addition, at low levels of fat%, BAI was more likely to overestimate fat% for men (by 10% or more).
Table 3.
Mean Difference Between BAI and Fat% (from DXA) vs BMI and Fat% (from DXA)
| Mean Difference (Mean ± SD) | ||
|---|---|---|
| BAI and Fat% | BMI and Fat% | |
| Age ≥ 55 | −5.2±6.0* | −7.6±7.2 |
| Women | −7.4±5.1* | −13.1±4.4 |
| Men | −3.0±6.0† | −2.2±5.2 |
| White | −5.0±6.1* | −7.4±7.2 |
| Black | −5.9±5.6* | −8.6±7.4 |
| Age ≥ 65 | −4.7±6.2* | −7.1±7.2 |
| Women | −7.0±5.3* | −13.2±4.5 |
| Men | −3.0±6.3† | −2.7±5.3 |
| White | −4.5±6.3* | −6.7±7.1 |
| Black | −5.7±5.8* | −8.6±7.5 |
Notes: BAI = body adiposity index; BMI = body mass index; DXA = dual energy X-ray absorptiometry; Fat% = body fat percentage.
*Significantly smaller mean difference for BAI and fat% than the mean difference between BMI and fat % (p < .01).
†Significantly greater mean difference for BAI and fat% than the mean difference between BMI and fat% (p < .01).
Figure 1.
Bland–Altman limits-of-agreement plots between body adiposity index (BAI) or body mass index (BMI) and percent body fat (from dual energy X-ray absorptiometry [DXA]). (A) Bland–Altman plots between BAI and percent body fat (from DXA) in total sample and age ≥ 65 years; (B) Bland–Altman plots between BMI and percent body fat (from DXA) in total sample and age ≥ 65 years. BAI showed better agreement with percent body fat than BMI did both in the total sample and in age ≥ 65 years.
For a specific range of fat% (from DXA), the percent difference between the mean estimates for the Baltimore Longitudinal Study of Aging study subpopulations was calculated. As shown in Table 4, BAI predicts fat% best at a range of 20%–30% of body fat. BAI tended to overestimate fat% at the lower fat% range (<15% body fat as assessed by DXA) and underestimate fat% at the higher fat% range (40% body fat and greater as assessed by DXA). This was true for the total sample as well as each subgroup. Finally, a linear regression model was used to assess the contribution of BAI (or BMI) to the variance of fat% as well as the interaction of sex and race. The model that included BAI explained 56% of the variance of fat%, whereas the model that included BMI only explained 33%. BAI for men and women presented similar linear regression slopes (interaction terms of sex × BAI were not significant; Figure 2A), whereas BMI for men and women had different regression slopes (interaction terms of sex × BMI were significant; Figure 2B).
Table 4.
Ability of BAI to Predict Fat% for the BLSA Study in Subpopulations
| BAI: Mean ± SD (n) | ||||||
|---|---|---|---|---|---|---|
| Fat% Groups (DXA) | Total Sample | Age ≥ 65 | Women | Men | Black | White |
| 0–10 | 23.0±2.3 (4) | 23.0±2.3 (4) | N/A (0) | 23.0±2.3 (4) | 26 (1) | 22.0±1.3 (3) |
| 10–15 | 24.8±2.8 (14) | 25.5±2.2 (12) | 28.7 (1) | 24.5±2.7 (13) | 22.0±3.1 (3) | 25.7±2.0 (8) |
| 15–20 | 23.0±2.4 (23) | 23.4±2.2 (16) | N/A (0) | 23.0±2.4 (23) | 22.5±1.8 (4) | 23.3±2.7 (16) |
| 20–25 | 24.8±2.5 (85) | 24.9±2.6 (65) | 25.3±2.2 (6) | 24.7±2.6 (79) | 24.7±2.6 (15) | 24.9±2.7 (54) |
| 25–30 | 25.7±3.1 (154) | 25.6±3.3 (112) | 26.8±3.2 (39) | 25.4±3.0 (115) | 25.3±2.7 (25) | 25.9±2.8 (100) |
| 30–35 | 28.0±3.6 (213) | 27.9±3.8 (143) | 28.7±4.0 (81) | 27.5±3.3 (132) | 28.0±4.7 (45) | 28.0±3.2 (144) |
| 35–40 | 29.3±4.1 (181) | 29.3±4.5 (127) | 29.9±3.8 (102) | 28.4±4.4 (79) | 29.6±2.9 (36) | 29.3±4.6 (117) |
| 40–45 | 32.9±4.3 (153) | 32.8±4.2 (95) | 33.0±4.1 (116) | 32.5±5.1 (37) | 33.8±4.0 (41) | 32.7±4.1 (98) |
| 45–50 | 37.8±4.0 (87) | 38.1±4.3 (43) | 37.7±3.9 (86) | 45.0 (1) | 38.3±3.6 (41) | 37.3±4.3 (44) |
| 50–55 | 40.9±5.4 (36) | 42.0±5.1 (21) | 40.9±5.4 (36) | N/A (0) | 41.4±4.2 (17) | 40.6±6.7 (17) |
Notes: Data are presented at the mean ± SD of BAI values and the “n” in parentheses; n: the number of participants of each sub-population that falls in each fat% range group. BAI = body adiposity index; BLSA = Baltimore Longitudinal Study of Aging; DXA = dual energy X-ray absorptiometry; Fat% = body fat percentage.
Figure 2.
(A) Regression between body adiposity index (BAI) and percent body fat. (B) Regression between body mass index (BMI) and percent body fat for both men and women. The interaction terms for men and women were not significant in the regression of percent body fat vs BAI (thus we used one line to represent the linear regression), but significantly different for percent body fat vs BMI (thus we used two lines).
Discussion
The purpose of this study was to determine the validity of the BAI equation so that this adiposity measure can be used in a clinical and research setting as a quicker and less expensive alternative to DXA. We found that the newly developed BAI was valid in an older population. BAI correlated significantly with fat% in the total study sample as well as in all subgroups. Moreover, BAI showed a better agreement with fat% than BMI did in white and black race groups and women. Conversely, in men, BMI showed better agreement with fat% than BAI did. Finally, BAI explained a larger percentage of the variation in fat% than BMI based on the regression model (BAI: 56% vs BMI: 33%). Therefore, BAI may be able to better predict fat% than BMI in the older population.
We did find that BMI had better agreement with DXA than BAI did for the men, whereas in women, BAI showed better agreement with DXA. This better agreement of BMI with DXA for men might be caused by the fact that women tend to store more fat in the gluteal–femoral region, whereas men tend to store more fat in the abdominal area. The BAI equation tends to overlook the abdominal fat depot because it uses HC and height to predict body fatness. It is true that in very old age, the traditional fat depots in men and women may change, which could affect the HC difference between men and women (18); however, 91% of the participants in our study population were younger than 85 years. This could possibly explain our findings that BMI was in better agreement with DXA-derived fat% compared with the BAI equation for men only.
Alternatively, our regression analysis revealed that the linear relationship representations between BAI and fat% were better than BMI and fat% in both men and women. This was similar to findings by Bergman and colleagues (5) that BAI could predict fat% for both men and women without statistical correction. These findings were interesting and it raised the question as to which adiposity index was better for men: BAI or BMI? Depending on which data analysis was completed (mean differences vs linear regression), the interpretation of that data could lead to different conclusions about which adiposity index is better in men. With validity established (significant Lin’s concordance), general linear model with interaction reveals the need of statistical adjustment for different populations. Mean difference, on the other hand, shows how close the index is to the gold standard. Although BAI did not need any adjustment for men and women (based on the linear regression modeling approach), the mean differences between BAI and BMI with fat% would indicate that in men, BMI may actually be a better indicator of adiposity.
We were somewhat surprised that the BAI did not explain more of the variance in fat%. Fifty-six percent of the variance was accounted for in the linear regression model, which is similar to what has been reported previously between BMI and fat% (13), which has been considered a poor association by some. Because BAI did not seem to be accurate at the lowest fat% (<15% fat as measured by DXA), we computed another linear regression model that removed all participants with a body fat% below 15%. Unfortunately, the R 2 did not improve much (new R 2 of 55.7%), so removing the lowest fat% did not improve the regression model.
Similar to the findings of Bergman and colleagues (5), we discovered that the smallest percentage difference between BAI values and DXA occurred at a range of 20%–55% body fat (19,20) Additionally, in men, we found that BAI tended to overestimate fat% by more than 10% at low levels of fat%, which was in agreement with the findings by Freedman and colleagues (21) and Johnson and colleagues (22). Therefore, BAI should not be used as a direct replacement measure of fat% at the extremely low and high ends of fat%.
Because the BAI equation was not as accurate for those participants with body fat percentages outside of the middle ranges, we developed a modified BAI (BAI-m) equation. We did not have enough statistical power to develop an equation for those with very low body fat percentages (only 18 participants with fat% <15%). However, 276 participants (29% of our study sample) had a body fat percentage of 40% or greater. We randomly divided those 276 participants into two groups: “development group” and “validation group.” Using stepwise multiple regression, the BAI-m equation for this group was BAI-m = 0.4*hip/height1.5 + 24. We then compared this BAI-m with DXA in the validation group. The Pearson correlation with DXA was r = .642, p < .01. Mean difference of BAI-m from DXA was 0.3 ± 3.1. The average BAI-m versus DXA measurements for different ranges of fat% were BAI of 44.3 ± 1.5 for fat% ranging from 40% to 44.9%, BAI of 46.8 ± 1.6 for fat% ranging from 45% to 49.9%, and BAI of 47.2 ± 2.2 for fat% ranging from 50% to 55%. These BAI-m values are much closer to the actual measured body fat percentages via DXA than those derived with the original equation (Table 4). Thus, BAI-m may work better as an adiposity index for those with a higher body fat percentage (40% or greater) in an older population. Because the reason behind having a BAI equation is so clinicians do not need to rely on body fat percentage data, the decision to use either the BAI or BAI-m could be difficult. The average BAI and BMI values for people with fat% of 40%–41% in this population are 31.3 ± 0.9 and 28.5 ± 0.8, respectively. Therefore, we suggest that for people with calculated BAI greater than 31.3, or BMI greater than 28.5, BAI-m should be used to calculate predicted fat%.
There were some limitations in our study. We had a relatively small number of participants with either extremely low or extremely high body fat percentages. Therefore, the lack of accuracy of BAI at these extremes should be noted with some caution. Also, we are aware that frailty is often associated with lower percentage of fat-free mass (23); however, frailty was not assessed in this study. If frail individuals do have less fat-free mass, they may have a higher body fat percentage. We found that BAI was not as accurate when body fat percentages were 40% or greater (as measured by DXA). Because 276 of our participants had a body fat percentage of 40% or greater (29% of our participant population), the use of BAI as an accurate adiposity index in those individuals is more questionable. Rather, the BAI-m is probably more accurate to use. Finally, there may be a limitation in using BAI in an older population if frailty or other conditions prevent accurate measures of height, weight, or HC. Although height and weight may not be too challenging to obtain in this population, HC may be more challenging to measure in frail individuals. In order for BAI to be used as an adiposity index, those three measurements must be accurate.
Because frailty can affect body fat percentage, and frailty tends to increase with age, we also wanted to see if there were any differences in the accuracy of BAI between people over the age of 85 compared with the rest of our participant population. Sixty-nine participants in our study were older than 85 years of age. Interestingly, we found that body fat percentage was significantly lower in people older than 85 (35.1% ± 8.7% vs 31.2% ± 9.6% for age < 85 years vs age ≥ 85 years, respectively, p < .01). However, BAI was significantly correlated with fat% in both groups (r of .75 vs .73 for age ≤ 85 years vs age > 85 years, respectively, p < .01). Thus, in our study sample, among people older than 85 years, BAI worked well in predicting fat%. Further, because their body fat percentage falls within the middle range (31.2%), there is close agreement between BAI versus DXA (Table 4).
Determining body adiposity can be challenging in older individuals, especially in large population studies. Current gold standard of body adiposity assessment (DXA) is costly and lacks the portability that makes access to them inconvenient for older people. Conversely, measuring height, weight, and HC would be fairly easy, quick, and inexpensive to do. The adverse health implications of obesity are associated more with fat mass rather than just body weight, which is a limitation to using BMI in a clinical setting. Thus, BAI may be used as a more economical, convenient assessment of body adiposity in the older population than DXA or BMI in both clinical and research settings.
This study is the first to show that BAI is a valid prediction method of fat% in the older population without statistical adjustment for men and women. We also compared BAI versus BMI in the ability to predict fat%, and we found that BAI was a more accurate measure of adiposity than BMI in older women and in white and black populations, but not in men. Finally, BAI may not be an appropriate adiposity index for extremely low or extremely high body fat percentages. We developed and validated a BAI-m equation for older individuals with a body fat percentage of 40% or greater. Whether this BAI-m equation would be valid in younger populations is yet to be determined. Studies are also needed to assess the validity of the BAI in different races such as Asians, Hawaiians, and Indian Americans, as well as in children and adolescents. Finally, future research on the relationship between this new adiposity index and long-term health outcomes is warranted.
Funding
Baltimore Longitudinal Study of Aging (BLSA) is supported by the Intramural Research Program of the National Institute on Aging of the National Institutes of Health.
Acknowledgments
We are extremely grateful for the support from Baltimore Longitudinal Study of Aging (BLSA) research group of the Intramural Research Program of the National Institute on Aging. We acknowledge the life-long contributions of the BLSA study participants and the hard work of study research group members. The authors declared no conflict of interest.
References
- 1. Salihu HM, Bonnema SM, Alio AP. Obesity: what is an elderly population growing into? Maturitas. 2009;63:7–12 [DOI] [PubMed] [Google Scholar]
- 2. Salahi L. Staying fit: majority of older adults struggle with weight. ABC News Medical Unit, New York; June 23, 2010. http://abcnews.go.com/Health/OnCall/elderly-obesity-older-adults-face-weight-issues/story?id=10985159. Accessed Feburary 3, 2013 [Google Scholar]
- 3. Kennedy RL, Chokkalingham K, Srinivasan R. Obesity in the elderly: who should we be treating, and why, and how? Curr Opin Clin Nutr Metab Care. 2004;7:3–9 [DOI] [PubMed] [Google Scholar]
- 4. Chapman IM. Obesity paradox during aging. Interdiscip Top Gerontol. 2010;37:20–36 [DOI] [PubMed] [Google Scholar]
- 5. Bergman RN, Stefanovski D, Buchanan TA, et al. A better index of body adiposity. Obesity (Silver Spring). 2011;19:1083–1089 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Houtkooper LB, Lohman TG, Going SB, Howell WH. Why bioelectrical impedance analysis should be used for estimating adiposity. Am J Clin Nutr. 1996;64(3 suppl):436S–448S [DOI] [PubMed] [Google Scholar]
- 7. Rodríguez G, Moreno LA, Blay MG, et al. Body fat measurement in adolescents: comparison of skinfold thickness equations with dual-energy X-ray absorptiometry. Eur J Clin Nutr. 2005;59:1158–1166 [DOI] [PubMed] [Google Scholar]
- 8. Skerlj B, Brozek J, Hunt EE, Jr. Subcutaneous fat and age changes in body build and body form in women. Am J Phys Anthropol. 1953;11:577–600 [DOI] [PubMed] [Google Scholar]
- 9. Florey CV. The use and interpretation of ponderal index and other weight-height ratios in epidemiological studies. J Chronic Dis. 1970;23:93–103 [DOI] [PubMed] [Google Scholar]
- 10. Evans JG, Prior IA. Indices of obesity derived from height and weight in two Polynesian populations. Br J Prev Soc Med. 1969;23:56–59 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Keys A, Fidanza F, Karvonen MJ, Kimura N, Taylor HL. Indices of relative weight and obesity. J Chronic Dis. 1972;25:329–343 [DOI] [PubMed] [Google Scholar]
- 12. McCarthy HD. Body fat measurements in children as predictors for the metabolic syndrome: focus on waist circumference. Proc Nutr Soc. 2006;65:385–392 [DOI] [PubMed] [Google Scholar]
- 13. Bedogni G, Pietrobelli A, Heymsfield SB, et al. Is body mass index a measure of adiposity in elderly women? Obes Res. 2001;9:17–20 [DOI] [PubMed] [Google Scholar]
- 14. Barreira TV, Harrington DM, Staiano AE, Heymsfield SB, Katzmarzyk PT. Body adiposity index, body mass index, and body fat in white and black adults. JAMA. 2011;306:828–830 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Shock NW, Greulich RC, Andres R, et al. Normal Human Aging: the Baltimore Longitudinal Study of Aging. Washington, DC: US Government Printing Office; 1984 [Google Scholar]
- 16. Lindle RS, Metter EJ, Lynch NA, et al. Age and gender comparisons of muscle strength in 654 women and men aged 20-93 yr. J Appl Physiol. 1997;83:1581–1587 [DOI] [PubMed] [Google Scholar]
- 17. Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics. 1989;45:255–268 [PubMed] [Google Scholar]
- 18. Lisko I, Tiainen K, Stenholm S, Luukkaala T, Hervonen A, Jylhä M. Body mass index, waist circumference, and waist-to-hip ratio as predictors of mortality in nonagenarians: the Vitality 90+ Study. J Gerontol A Biol Sci Med Sci. 2011;66:1244–1250 [DOI] [PubMed] [Google Scholar]
- 19. Buchanan TA, Xiang A, Kjos SL, et al. Gestational diabetes: antepartum characteristics that predict postpartum glucose intolerance and type 2 diabetes in Latino women. Diabetes. 1998;47:1302–1310 [DOI] [PubMed] [Google Scholar]
- 20. Sumner AE, Farmer NM, Tulloch-Reid MK, et al. Sex differences in visceral adipose tissue volume among African Americans. Am J Clin Nutr. 2002;76:975–979 [DOI] [PubMed] [Google Scholar]
- 21. Freedman DS, Thornton JC, Pi-Sunyer FX, et al. The body adiposity index (hip circumference ÷ height(1.5)) is not a more accurate measure of adiposity than is BMI, waist circumference, or hip circumference. Obesity (Silver Spring). 2012;20:2438–2444 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Johnson W, Chumlea WC, Czerwinski SA, Demerath EW. Concordance of the recently published body adiposity index with measured body fat percent in European-American adults. Obesity (Silver Spring). 2012;20:900–903 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Villareal DT, Banks M, Siener C, Sinacore DR, Klein S. Physical frailty and body composition in obese elderly men and women. Obes Res. 2004;12:913–920 [DOI] [PubMed] [Google Scholar]