Total body potassium differs by sex and race across the adult age span1- 3

Qing He, Moonseong Heo, Stanley Heshka, Jack Wang, Richard N Pierson Jr, Jeanine Albu, Zimian Wang, Steven B Heymsfield, and Dympna Gallagher

ABSTRACT Background: Total body potassium (TBK) is an index of fat-free mass. Data describing changes in TBK in African American, Asian, or Hispanic populations have not been reported. Objective: The aim was to investigate possible sex and racial dif­ferences in TBK in adults over an age range of 70 y. Design: The study used longitudinal and cross-sectional data col­lected in a body-composition unit from 973 men and 1368 women of African American, Asian, white, and Hispanic race-ethnicity. Random coefficient models in which baseline weight and height were taken into account were applied to estimate sex-specific changes in TBK among the 4 racial-ethnic groups. Results: The ages of 30 and 31 y were identified for women and men, respectively, as the cutoffs after which TBK began to decline. Both sexes had similar racial-ethnic patterns for expected mean TBK ~t the age cutoffs: African Americans had the highest value, followed by whites, Hispanics, and Asians. After the age cutoffs, the decline in TBK differed by race and sex. In women, African Americans showed the most rapid decline, whereas Asians had the lowest. In men, Hispanics had the most rapid decline in TBK, followed by African Americans, whites, and Asians. Conclusion: Significant sex and racial differences exist in the rate of change in TBK with age. Further studies are needed to explore the associations of declining TBK with health risks. Am J Clin Nutr 2003;78:72-7.

KEY WORDS Total body potassium, race, sex, age, skeletal muscle mass, African Americans, Asians, whites, Hispanics

INTRODUCTION

Total body potassium (TBK) is an index of the body's cell mass (1) and is quantifiable by the detection of ,),-rays that emanate from the body by virtue of its content of the naturally occurring radioisotope of potassium (4oK). It is estimated that 60% of the body's 40K (84 g/140 g) is found in skeletal muscle , and the remainder is found in other organs and tissues (2). TBK has been used as an index of skeletal muscle mass in many previous stud­ies (3-8) and can also provide an estimate of fat-free mass (9-11).

During the adult years, skeletal structures including muscle and bone are dynamically changing and play important roles in influ­encing cognitive and physical functional status, nutritional and endocrine status, quality of life, and comorbidity (12). Osteo­porotic fractures may be related to the atrophy of skeletal muscle mass or its functional impairment (13). The loss of muscle mass with aging is associated with a deterioration in physical func­tion, including poorer lower extremity performance (14). Skeletal

muscle atrophy, or sarcopenia, is most prevalent in the elderly, increases with age, and is strongly associated with disability, inde­pendent of morbidity (15). Accordingly, information on changes in TBK throughout the life span may be helpful in identifying periods of high health risks for primary care settings, with impli­cations for treatment and prevention.

Sex differences in the changes in TBK during adulthood have been reported in white populations (16-18). Changes in TBK in African American, Asian, or Hispanic populations have not been previously reported. Different patterns of skeletal muscle change or loss may have different relations to outcomes, which in turn may require different or population-specific approaches to risk management. The aims of the present study were to 1) investigate race-ethnic and sex differences in changes in TBK during adult­hood and 2) identify age cutoffs when TBK values begin to decline in a cohort of African American, Asian, Hispanic, and white women and men.

SUBJECTS AND METHODS

Subjects

Subjects were 1368 women (299 African American, 197 Asian, 669 white, and 203 Hispanic) and 973 men (198 African Amer­ican, 152 Asian, 444 white, and 179 Hispanic) who participated in one or more body-composition studies at our research center from 1986 through 2000. Longitudinal data were available for 166 women (55 African American, 20 Asian, 82 white, and 9 His­panic) and 88 men (17 African American, 21 Asian, 40 white, and 10 Hispanic); the remaining subjects (89%) were measured once. Both cross-sectional and longitudinal data were used in this analysis. At the time of the first measurement, the subjects ranged in age from 20 to 90 y. The longest follow-up interval in anyone

I From the Obesity Research Center, St Luke's-Roosevelt Hospital , New

York (QH, SH, JW, RNP, JA, ZW, SBH, and DG); the Institute of Human Nutri­

tion, College of Physicians & Surgeons, Columbia University, New York (QH

and DG); and the Department of Psychiatry, Weill Cornell Medical College, White Plains, NY (MH). •

• 2 Supported in part by grants R29-AG 14715, NIH-RR00645 , RO I-DK37352,

ROI-DK4261S , and ROI-DK40414 from the National Institutes of Health.

3 Address reprint requests to D Gallagher, Obesity Research Center, 1090

Amsterdam Avenue, 14th Floor, New York, NY 10025 . E-mail: dglOS@

columbia.edu .

Received September IS , 2002.

Accepted for publication January 7, 2003 .

72 Am J Clin NutI' 2003 ;78 :72-7 . Printed in USA. © 2003 American Society for Clinical Nutrition

by guest on February 5, 2016

ajcn.nutrition.orgD

ownloaded from

TBK CHANGES IN ADULTS DIFFER BY SEX AND RACE 73

TABLE 1 Subject characteristics I

Women Men

African American Asian White Hispanic African American Asian White Hispanic

(n = 299) (n = 197) (n = 669) (n = 203) (n= 198) (n = 152) (n = 444) (n = 179)

Age (y) 51.5 ± 18.8b 46.8 ± 18.5" 45.5 ± 18.7" 47.9 ± 15.7",b 46.7 ± 18.8" 46.8 ± 18.4" 44.4 ± 18.7" 44.3 ± 15.6" Weight (kg) 75 .3 ± 17.3d 54.3 ± 8.3" 63 .9 ± 13.2b 67 .7 ± 13.2c 80.7 ± 14.2b 67 .9 ± 9.3" 78.9 ± 13.3b 78.2 ± 15.3b

Height (cm) 161.9 ± 7.2b 157.7 ± 6.5" 163.1 ±6.9b 156.4 ± 6.5" 175.6±7.2b 170.8 ± 6.3" 176.1 ± 7.5b 169.9 ± 7.6" BMI (kg/m2) 28 .7 ± 5.9c 21.8±3.1" 24.1 ± 4.9b 27.7 ± 5.SC 26.1 ± 4.1b,c 23.3 ± 2.8" 25.4 ± 3.8b 27.0 ± 4.SC TBK (mmol) 2506 ± 446d 2131 ± 315" 2426 ± 372c 2257 ± 327b 3899 ± 684c 3362 ± 476" 3867 ± 621 c 3592 ± 630b

IX ± SD. TBK, total body potassium. Means in a row within each sex with different superscript letters are significantly different, two-sided P < 0.05 (one-way ANOYA with Bonferroni correction).

subject was 14 y, and the shortest was several hours . Recruitment of subjects occurred through advertisements in local newspapers, on radio stations, and in flyers posted in the local community. Inclusion criteria required that the subjects be ambulatory, not exercise vigorously, and have no orthopedic problems that could affect any of the variables under investigation . Only subjects who denied any major current health problems were enrolled in the study. Race was determined by self-report. Subjects were asked to choose from the categories Asian, Black (African American), and Caucasian. All parents and grandparents of the African American and white subjects were required to be non-Hispanic African American and non-Hispanic white, respectively. Each subject com­pleted a medical examination that included screening blood tests completed after the subject had fasted overnight. The study was approved bX the institution's ethics committee, and all subjects gave written consent to participate.

Body composition

Body weight was measured to the nearest 0.1 kg (Weight Tronix, New York), and height was measured to the nearest 0.5 cm by using a stadiometer (Holtain, Crosswell, United Kingdom). The 4-pi whole body counter was used to measure 40K (19). The 40K net counts accumulated over 9 min were corrected for self­absorption on the basis of an earlier 42K calibration study (20). The within-subject CV in our laboratory for 40K counting is 4.7% (21). TBK was calculated from measured 40K (19). The TBK counter was calibrated daily by counting a standard of known activity (each standard bottle contains 5000 decays/min of 4oK) and adjust­ing all subsequent patient readings on that day by the deviation from that standard and for background counts.

Statistical analysis

The mean values for subject characteristics were compared among racial-ethnic groups by sex with the use of a one-way analysis of variance with Bonferroni correction, and statistical significance was set at two-sided 0.05 . A locally weighted regres­sion line (22) was fitted to the data to visually display the trends in TBK over the age range for women and men separately. Ran­dom coefficient regression models adjusted for weight and height were applied to estimate sex-specific changes in TBK among the 4 racial -ethnic groups . The models incorporated both cross ­sectional and longitudinal data . To estimate peak TBK, we applied random coefficient piecewise linear regression modeling, allowing different intercepts and slopes before and after a range of age cutoffs (23). An optimal age cutoff was chosen as the max­imum age that satisfied the following 2 conditions: 1) significant difference in slope before compared with after the specified age

and 2) reversed directions of slope after the specified age . Age cutoffs between 30 and 60 y were tested in 5-y increments and were further explored at l-y increments between the ages of 25 and 40 y. Pairwise differences in mean TBK values and slopes among the racial-ethnic and sex groups were tested by using Wald t tests in the random coefficient models. All statistics were computed by using SAS software, version 8 (24), and two-tailed P values < 0.05 were considered significant. The SAS code used for fitting the random coefficient piecewise linear regression model is provided in Appendix A.

RESULTS

Subject characteristics

The sample sizes and mean (± SD) age, weight, height, body mass index, and TBK of the subjects are summarized in Table 1. Among the women, Asians, whites, and Hispanics did not differ significantly in age, whereas African Americans were older than the other groups. Mean weight and TBK were significantly differ­ent among the 4 racial-ethnic groups, and TBK values were lowest in Asians . Among the men, mean age did not differ significantly among the 4 racial-ethnic groups. Asians had the lowest mean weight and TBK. Despite the significant differences, the distribu­tions by race within each sex overlapped to a considerable degree.

Changes in TBK during adulthood

Effect of age

The relation between TBK and age is shown in Figure 1, A and B, for women and men, respectively. The connected data points rep­resent baseline and follow-up measures in subjects in whom lon­gitudinal measures were acquired. Generally, TBK declined with age, as indicated by the locally weighted regression lines shown in the figure. For women and men, the ages of 30 and 31 y, respec­tively, were identified as the cutoffs after which the TBK regres­sion line began to decline.

Effects of sex and race: before the age cutoff

Separate regression models wf1re developed for women and men before and after the age cutoff, and the regression coefficients are presented in Table 2. In general, the men had higher expected mean TBK than did the women . The SEMs of the residuals within groups were 132.6 and 166.5 for women and men, respectively.

In women (Figure 2A), African Americans had the highest expected mean TBK values, followed by whites, Hispanics, and Asians . A similar pattern was seen in men (Figure 2B), in whom

by guest on February 5, 2016

ajcn.nutrition.orgD

ownloaded from

74 HEETAL

A 0 0 0 <D

0 0 0 l{)

0

0 0 a

E 'I'

.s .. :.: CD f- a a a

(')

a a ~

20 40 60 80 100

Age (y)

B 0 0 a <D

. . a : a a l{)

a 0

a a E 'I'

.s :.: CD

a f- a a (')

a a a N

20 40 60 80 100

Age (y)

FIGURE 1. Changes in total body potassium (TBK) with age in 1368 women (A: 1684 observations) and 973 men (B: 1200 observations). In both panels of the figure , connected data points represent baseline and follow -up measures in subjects for whom longitudinal measures were acquired. A locally weighted regression line was fitted to the data to visu­ally display the trends in TBK over the entire age range.

African American men had the highest values and Asians the lowest.

Effects of sex and race: after the age cutoff

When the race-by-slope interactions were excluded from the models to estimate overall slopes after the age cutoffs, the . rates of decline were 87 mmolldecade (P < 0.0001) for women after the age of 30 y and 176 mmolldecade (P < 0.0001) for men after the age of 31 y. Additional analyses by race with a common age cutoff of 31 y for both men and women showed that the linear declines in TBK were significantly different between men and women for all races (all P values < 0.0001), whereas the slopes before the age of 31 y were not significantly different between men and women within any race (all P values> 0.05).

Overall, after the age cutoff, the decline in TBK in men was at twice the rate of the decline in women for Asians , African Amer­icans, and whites. The decline in TBK in male Hispanics was 3 times that of female Hispanics. Asians had the lowest rate of decline for both women and men. .

TABLE 2 Expected mean total body potassium (TBK) values and slopes estimated from the random coefficent regression model for women and men in 4

racial-ethnic groups'

Value

Women (n = 1368) Expected mean TBK at age 30 y (mmol)2

African American 2617b

Asian 2422a

White 2S63b

Hispanic 2442a

Common slope before age 30 y (mmol/y) 0.843

Slope after age 30 y (mmol/y)4 African American -9.7b

Asian -6.Sa

White -9.1 b

Hispanic -6.9,,·b

Men (n = 973) Expected mean TBK at age 31 y (mmol)2

African American 4146b

Asian 383Sa

White 406Sb

Hispanic 3969a

Common slope before age 31 y (mmol/y) 1.553

Slope after age 31 y (mmol/y)4 African American - 19.8b

Asian - 13.l" White -17.6b

Hispanic -20.3b

'Means among the racial-ethnic groups by sex with different superscript letters are significantly different, two-tailed P < 0.05 (pairwise Wald t test).

2 All expected mean TBK values were evaluated after adjustment for baseline weight and height.

3 P value for the test of the hypothesis that slope is different from zero was not significant.

4 All slopes after the ages of 30 and 31 y for women and men, respec­tively, were significantly different from 0, two-tailed P < 0.05 (Wald t test).

In women, TBK began to decline at the age of =30 y. Asian women had significantly lower rates of decline (P < 0.05) than did African Americans and whites and had the lowest TBK values throughout the age range studied. Conversely, African American women had the highest expected TBK values before the age cut­off and displayed the most rapid rate of decline throughout the age range studied, although it was not significantly different from that in whites and Hispanics . The rate of TBK decline in women was greatest in African Americans (97 mmolldecade) compared with Asians (65 mmolldecade), whites (91 mmolldecade), and His­panics (69 mmolldecade) .

In men, TBK began to decrease after the age of 31 y, with His­panics having the most rapid rate of decline. Asian men showed the slowest rate of decline compared with the other ethnic groups (P < 0.05) and had the highest TBK estimate at the age of90 y. The rate of decline in TBK in men was greatest if). Hispanics (203 mrnol/decade) compared with Asians (131 mmolldecade), African Americans

. (198 mmolldecade), and whites (176 mmol/decade) .

DISCUSSION

This is the first study to describe TBK from 20 to 90 y of age in a multi-racial-ethnic cohort of African Americans, Asians,

by guest on February 5, 2016

ajcn.nutrition.orgD

ownloaded from

TBK CHANGES IN ADULTS DIFFER BY SEX AND RACE 75

A a a co N

a a ~

0 E a E- o

:L ~ ID f-

a a ~

§ N

20

B a a l{) ..-

a a a ..-

0 E a E- o

l{)

:L '" ID f-

a a g

a a l{) N

20

40 60

Age (y)

40 60

Age (y)

- African American .. ... Asian

._._._. Hispanic --- White

80

- African American ........ Asian ._._._. Hispanic --- White

80

FIGURE 2. Expected mean total body potassium (TBK) values and slopes estimated from the random coefficient regression model for women (A) and men (B) in 4 racial-ethnic groups. For women , the vertical line at age 30 y represents the cutoff of TBK decline. For men, the vertical line at age 31 y represents the cutoff of TBK decline.

whites, and Hispanics. In this cohort, the ages of 30 and 31 y were identified as the points at which TBK begins to decline in women and men, respectively. After these age cutoffs, African American women and Hispanic men experienced the most rapid rates of decline. Asians showed the lowest rate of decline for both women and men.

In the present study, a nearly identical age cutoff (30 and 31 y) was found for women and men. A previous longitudinal study of white subjects found that the cutoffs differed between women and men (17). In that study, minimal change in TBK was observed in women younger than 50 y, whereas TBK began to decline in men beginning at the age of 41 y (17). Results from a cross-sectional investigation, in which age- and sex-specific curves for TBK were adjusted for body weight (Klwt), showed a decrease in TBK after the age of 20 y in men and after puberty in women (19). In another cross-sectional study of 20-89-y-old women and men, the investigators fitted nonlinear models to describe the changes in TBK with age (25). Thus, differences in population character­istics, study aims, and analytic approaches make direct compar-ison across studies difficult. .

Sex differences in the rate of change in TBK with age in adults were previously described in both cross-sectional and lon­gitudinal studies (17, 19) . In agreement with those previous findings , the current study showed that men experience a more rapid loss of TBK with age than do women. However, this sex difference varied by race . The rate of decline in Hispanic men was 3 times that of Hispanic women, whereas the rate of decline in African American, Asian, and white men was twice that of their female counterparts .

To our knowledge, thi s is the first investigation of racial dif­ferences in the rate of decline in TBK in adults. Racial differences in TBK loss were previously estimated in black and white women in cross-sectional studies (11, 26) . In a study of African American and white women, Gallagher et al (11) reported a greater loss in TBK (weight and height adjusted) in African Americans than in whites, which agrees with the current study findings. In contrast, a cross-sectional study of 20-69-y-old women reported that the lifetime decline in TBK was 8% for black women compared with 22% for white women (26) . Differences in the age range studied and in the statistical adjustments made may account for these inconsistent findings.

There have been no previous reports of changes in TBK with age in Asian and Hispanic adults. Asians and Hispanics are among the most rapidly growing racial-ethnic groups in the United States (27), with increases of 20% for non-Hispanic Asians and Pacific Islanders and 21 % for persons of Hispanic origin between 1995 and 2000 (compared with 2% for non-Hispanic whites and 6% for non­Hispanic blacks). Baumgartner et al (15) reported a greater preva­lence of sarcopenia in elderly Hispanics than in non-Hispanic whites (15), which agrees with the current study finding that Hispanic men experienced a greater loss of skeletal muscle mass with aging.

Whether greater risk is associated with a more rapid loss of TBK (an index of the body's fat-free and skeletal muscle mass) in healthy adults is unknown. In patients with AIDS, those with the lowest potassium values were found to be closer to death at the time of study (28). Because skeletal muscle is attached to bone, and therefore pro­vides support for bone, a loss in skeletal muscle may increase the risk of bone fracture. We recently reported that appendicular skeletal mus­cle mass (as measured by dual-energy X-ray absorptiometry) was lower in prepubertal Asians than in prepubertal African Americans and whites (29). In the current study, Asians had the lowest TBK val­ues compared with the other racial-ethnic groups at the age of 30-31 y. However, because of the lower rate of decline in TBK in both Asian women and men after the age cutoff, other racial-ethnic groups sur­passed Asian men in the level of TBK reached in later years.

The recognition of racial differences in TBK loss may be of clin­ical importance because body composition varies by race. It was previously reported that Asian or Hispanic heritage is one variable associated with a significantly increased likelihood of osteoporo­sis in postmenopausal women (30) . The identification of racial dif­ferences in the rate of TBK loss needs to be followed up in meta­bolic studies to identify or clarify associations with health risk.

The observed racial differences in TBK may reflect in part dif- " ferences in dietary intake, accult4ration, body size, or physical activity, detailed information on which is not available for our study population. In one study, white women who had higher lev­els of habitual physical activity had 6.5 % more potassium per unit of fat-free mass than did their less-active counterparts (31). In that study, an active 70-y-old had the potassium content value of a sedentary 55-y-old woman (31). Although subjects participat­ing in vigorous or intense physical activity were excluded from

by guest on February 5, 2016

ajcn.nutrition.orgD

ownloaded from

76 HEETAL

the current study, differences in habitual or daily physical activity likely existed among the study participants, which may have con­tributed to the observed differences in TBK loss by race and age. Different races or ethnicities may have different diets and accul­turation issues that could influence body composition. Moreover, racial differences in body composition (7, 11, 29) are likely reflected in part in the TBK differences observed. Alternatively, the observed differences may be determined to an unknown degree by genetic differences. Another consideration is that the rate of decline in TBK appeared to be related to peak values for some of the racial-ethnic groups; however, this hypothesis could not be tested with the current data.

A limitation of this study is that TBK is not the ideal index of skeletal muscle mass throughout the life span. TBK may represent a different weighting of fat-free mass components at different stages of life as the skeletal muscle declines from its dominant pro­portion of the cell mass. Little information is available on how age, sex, and race influence TBK concentrations in skeletal muscle mass. It has been reported that the assumed steady state value for the ratio ofTBK to fat-free mass (TBKlfat-free mass = 64.2 mmollkg) is significantly lower in weight- and height-matched pairs of older (aged> 65 y) healthy white women than in younger (aged 19-35 y) women (32). The same question has not been explored in other racial groups. Therefore, TBK loss with age should be interpreted cau­tiously in relation to health risks or clinical outcomes.

In this study, racial-ethnic group was determined by subject self-report. Although other large national nutritional databases have categorized groups by race or ethnicity, a question arises regarding the appropriateness of studying discrete "race" groups, identified m'ai~I'Y by phenotypes such as skin color (33). In par­ticular, categorization of the Hispanic ethnic group requires acknowledgment of a cultural and genetic background of admixed populations with different combinations of Amerindian, European, and African ancestry.

In conclusion, the results of our study show that divergent pat­terns of TBK accumulation and loss are present for Asians and Hispanics compared with African Americans and whites through­out the adult years. Although similar age cutoffs for TBK decline were found in women and men, men attained higher peak values and experienced a much more rapid loss of TBK than did women throughout the age range studied. Because muscle mass is the major depot for protein storage against the catabolic stresses of starvation or illness, a decline in muscle mass, or sarcopenia, guar­antees a loss of reserves with aging. These findings emphasize the importance of sex- and race-specific interpretation of body-com­position results to define body-composition phenotypes and the need for further studies to explore and make explicit the associa­tions of body composition with health risks. n

QH was responsible for data pooling, screening, analysis, and manuscript writing. MH was responsible for data analysis and manuscript writing. SH was responsible for data analysis and manuscript writing and provided advice and consultation. JW, RNP, JA, ZW, and SBH were responsible for data collection . DG was responsible for study design, data collection, and manuscript writing and provided administrative support, supervision, and advice. No author had a conflict of interest in any company or organization sponsoring this study.

REFERENCES 1. Moore FD, Olesen KH, McMurrey JD, Parker IHY. The body cell

mass and its supporting environment: body composition in health and disease. Philadelphia: WB Saunders Company, 1963.

2. Snyder WS , Cook Ml , Nasset ES , Karhausen LR, Howells GP, Tipton IH. Report of the group on Reference Man. Oxford, United Kingdom: Pergamon Press, 1975 :312-3.

3. Oberhausen E, Onstad CO. Relationship of potassium content of man with age and sex. In: Meneely GR, Linde SM, eds. Radioactivity in man. Springfield, IL: Thomas , 1965: l79-202.

4. Novak LP. Aging, total body potassium, fat-free mass , and cell mass in males and females between ages 18 and 85 years. 1 Gerontol 1972; 27:438-43 .

5. Cohn SH, Vaswani A, Zanzi I, Aloia JF, Roginsky MS, Ellis Kl. Changes in body chemical composition with age measured by total­body neutron activation. Metabolism 1976;25:85-95.

6. Bruce A, Andersson M, Arvidsson B, Isaksson B. Body composition. Prediction of normal body potassium, body water and body fat in adults on the basis of body height, body weight and age. Scand 1 Clin Lab Invest 1980;40:461-73.

7. Ellis Kl. Reference man and woman more fully characterized. Vari­ations on the basis of body size, age, sex , and race. Bioi Trace Elem Res 1990;26-27:385-400.

8. Wang ZM, Zhu S, Wang 1, Pierson RN lr, Heymsfield SB. Whole­body skeletal muscle mass: development and validation of total-body potassium prediction models . Am 1 Clin Nutr 2003 ;77:76-82.

9. Talso PI, Miller CE, Carballo AI, Vasquez 1. Exchangeable potas­sium as a parameter of body composition . Metabolism 1960;9: 456-7l.

10. Forbes GB , Gallup 1, Hursh lB. Estimation of total body fat from potassium-40 content. Science 1961 ; 133: 10 1-2.

11. Gallagher 0 , Visser M, De Meersman RE, et al. Appendicular skele­tal muscle mass: effects of age, gender, and ethnicity. 1 Appl Physiol 1997;83:229-39.

12. Baumgartner RN. Body composition in healthy aging. Ann NY Acad Sci 2000;904:437- 48.

13. Rosenberg IH, Roubenoff R. Stalking sarcopenia. Ann Intern Med 1995; 123:727-8.

14. Visser M, Kritchevsky SB , Goodpaster BH, et al. Leg muscle mass and composition in relation to lower extremity performance in men and women aged 70 to 79: the health , aging and body composition study. 1 Am Geriatr Soc 2002;50:897-904.

15. Baumgartner RN, Koehler KM, Gallagher 0, et al. Epidemiology of sarcopenia among the elderly in New Mexico. Am 1 Epidemiol 1998; 147:755-63.

16. Forbes GB, Reina lC. Adult lean body mass declines with age: some longitudinal observations. Metabolism 1970; 19:653-63 .

17. Flynn MA, Nolph GB, Baker AS"Martin WM, Krause G. Total body potassium in aging humans: a longitudinal study. Am 1 Clin Nutr 1989;50:713-7.

18. Flynn MA, Nolph GB , Baker AS , Krause G. Aging in humans: a con­tinuous 20-year study of physiologic and dietary parameters . J Am Coli Nutr 1992;11:660-72.

19. Pierson RN Jr, Lin DHY, Phillips RA. Total body potassium in health: effects of age, sex, height, and fat. Am J Physiol 1974;226: 206-12.

20. Pierson RN, Wang J, Thornton lC, Van Itallie TB , Colt EWD. Body potassium by 4-pi counting: an anthropometric correction. Am J Physiol 1984;245:E234-9.

21. Pierson RN Jr, Wang 1, Thornton JC, et al. Biological homogeneity and precision of measurement: the boundary conditions for normal body com- .' position. In: Ellis Kl, Eastman JD, eds. Human body composition: in vivo methods, models , and assessment. New York: Plenum Press, 1993: 15-22.

22. Cleveland WS . Robust locally weighted regression and smoothing scatterplots. 1 Am Stat Assoc 1979;74:829- 36.

23. Naumova EN, Must A, Laird NM. Tutorial in biostatistics: evaluating the impact of 'critical periods' in longitudinal studies of growth using piecewise mixed effects models. Int 1 Epidemiol 2001 ;30: 1332-41.

24. SAS Institute Inc. SAS user's guide: statistics, version 8 edition. Cary, NC: SAS Institute Inc, 1999.

by guest on February 5, 2016

ajcn.nutrition.orgD

ownloaded from

TBK CHANGES IN ADULTS DIFFER BY SEX AND RACE 77

25. Kehayias JJ, Fiatarone MA, Zhuang H, Roubenoff R. Total body potassium and body fat: relevance to aging. Am J Clin NutI' 1997;66: 904-10.

26. Aloia JF, Vaswani A, Feuerman M, Mikhail M, Ma R. Differences in skeletal and muscle mass with aging in black and white women. Am J Physiol Endocrinol Metab 2000;278:E1153-7.

27. US Census Bureau. Population estimates program. December 2000. Internet: http://eire .cens us. gov /popes t/archi ves/national/nation3/ intfile3-l.txt (accessed 16 August 2002).

28. Kotler DP, Wang J, Pierson RN. Body composition studies in patients with the acquired immunodeficiency syndrome. Am J Clin NutI' 1985; 42: 1255-65.

29 . Song MY, Kim J, Horlick M, et al. Prepubertal Asians have less limb skeletal muscle. J Appl Physiol 2002;92:2285-91.

APPENDIX A

30. Siris ES, Miller PD, Barrett-Connor E, et al. Identification and frac­ture outcomes of undiagnosed low bone mineral density in post­menopausal women: results from the National Osteoporosis Risk Assessment. JAMA 2001;286:2815-22.

31. Hansen RD, Allen BJ. Habitual physical activity, anabolic hormones, and potassium content of fat-free mass in postmenopausal women. Am J Clin NutI' 2002;75:314-20.

32. Mazariegos M, Wang ZM, Gallagher D, et al. Differences between young and old females in the five levels of body composition and their relevance to the two-compartment chemical model. J Gerontol 1994; 49:M201-8.

33. Sauer NJ. Forensic anthropology and the concept of race: if races don't exist, why are forensic anthropologists so good at identifying them? Soc Sci Med 1992;34: 107-11.

SAS PROC MIXED code for fitting the random coefficient piecewise linear regression with a common interceptl

%macro connect (cutage); data temp;

run;

set tbkms; t age = age - &cutage; cucage = 1 * (cage> = 0); Cbefore = Cage* (l-cucage); cafter = cage*cucage;

proc sort data = temp; by sex; run;

proce mixed covtest noclprint data = temp; by sex; class unit race4 ; model tbk = race4 basewtkg basehtcm Cbefore cafter

cafter*race4/solution ; random intercept cafter/subject = unit type = un; %mend;

connect (25) connect (26)

connect (27) connect (28)

connect (29)

connect (30)

connect (31) connect (32)

connect (33)

connect (34) connect (35)

connect (36)

connect (37) connect (38) connect (39) connect (40)

I SAS Institute Inc, Cary, NC.

by guest on February 5, 2016

ajcn.nutrition.orgD

ownloaded from