Energy requirements in the eighth decade of life

¹ From the University of Wisconsin–Madison (SB and DS); the University of California, San Francisco (DB); the University of Pittsburgh (MED); the National Institute on Aging, Bethesda, MD (EMS and TBH); the University of Tennessee Health Science Center, Memphis (SBK); and the National Institute on Diabetes and Digestive and Kidney Diseases, Bethesda, MD (JEE).

² Supported by NIH contracts N01-AG-6-2106, N01-AG-6-2102, and N01-AG-6-2103.

³ Reprints not available. Address correspondence to DA Schoeller, Department of Nutritional Sciences, University of Wisconsin–Madison, 1415 Linden Drive, Madison, WI 53706-1571. E-mail: dschoell{at}nutrisci.wisc.edu.

ABSTRACT  
Background: Knowledge of energy requirements among relatively healthy elderly is limited.

Objectives: The objectives of the study were to measure total energy expenditure (TEE)–derived energy requirements in a biracial population of older adults without limitations to daily life and to test these empirical measures against national and international recommendations.

Design: TEE (measured by the doubly labeled water method), resting metabolic rate (RMR), activity-related energy expenditure (AEE), and body composition were measured in 288 persons aged 70–79 y selected from the Health, Aging, and Body Composition Study.

Results: TEE was lower in women (530 kcal/d; P < 0.0001) than in men because of the women's lower RMR and AEE. Fat-free mass explained the sex difference in RMR, but body weight failed to account for the women's lower AEE (1 kcal · kg^-1 · d^-1; P = 0.007). Blacks had lower TEE than did whites (100 kcal/d, P = 0.03), and that was explained by blacks' lower RMR. Physical activity level (TEE/RMR) did not differ significantly between sexes and races (1.70 ± 0.23). The World Health Organization (WHO) recommendations overestimated TEE by 10 ± 15% (P < 0.0001) in women but not in men, and the dietary reference intakes (DRIs) were accurate to 0 ± 14% (P = 0.1). Both WHO and DRI recommendations are based on an underestimated physical activity level, and WHO recommendations are based on overestimated RMR.

Conclusions: This study of well-functioning older adults confirms the racial difference in energy metabolism and supports the use of the 2002 DRIs. Because the DRIs and WHO recommendations underestimated PAL, new predictive equations of energy requirements are proposed.

Key Words: Aging • obesity • malnutrition • African Americans • sex • energy expenditure

INTRODUCTION  
It is projected that, by 2040, 20% of the US population will be =" BORDER="0">65 y old; by comparison, in 1997, 12% of the US population was aged =" BORDER="0">65 y, and at the beginning of the last century, that proportion was only 4% (1, 2). Recent data showed that one-half of those >65 y old have a chronic disease or some other form of disability, and, consequently, the total societal burden of physical impairment will increase with the rise in the elderly population (3, 4). To mitigate any age-related decline in autonomy, particularly that engendered by nutrition-related chronic diseases, it is essential to develop strategies to promote adequate nutrition.

Among these strategies, the maintenance of energy balance is critical but complex in the elderly (5–7). Total energy expenditure (TEE) decreases with age because of an equivalent reduction in resting metabolic rate (RMR) and activity-related energy expenditure (AEE; 7, 8). However, the ability to control food intake after overeating or undereating becomes impaired with age (9, 10). These problems, in conjunction with chronic diseases, chewing problems, polypharmacy, living alone, and low income, likely explain the high vulnerability of the elderly to energy imbalances. Ultimately, an inappropriate energy intake leads to weight gain or undernutrition, which, together with physical inactivity, further contributes to the decline of bodily functions and the development of age-associated chronic degenerative diseases (5, 6).

Dietary interventions are particularly challenging in the elderly because of the paucity of data on which accurate recommendations for age-related changes in energy requirements can be based (7, 11). This scantiness of data reflects 3 main factors. First, the elderly are a highly heterogeneous population ranging from the highly active newly retired to the institutionalized. Second, accurate energy requirements must be derived from free-living TEE measured by using the state-of-art, costly, expertise-requiring doubly labeled water (DLW) method (12). Third, low minority inclusion has impaired the ability to determine the general applicability of TEE estimates (13).

The Health, Aging, and Body Composition Study (Health ABC Study) provided a unique opportunity to evaluate energy requirements in the eighth decade of life in a cohort of people free of functional limitations and disabilities of daily living. This cohort constitutes a particularly interesting population compared with those of previous reports of studies using DLW because of its large size and generally well functioning status. In a group of 288 elderly persons, we specifically investigated whether TEE-derived energy requirements differ by sex and race and the extent to which the current World Health Organization (WHO; 14) and the new 2002 Dietary References Intake (DRI; 15) recommendations are accurate in such a population; then we generated new predictive equations of energy requirements applicable to the growing elderly population.

SUBJECTS AND METHODS  
Subjects
Initiated in 1997, the Health ABC Study consisted of a cohort of 3075 participants recruited at 2 field centers: the University of Pittsburgh and the University of Tennessee (Memphis). Participants were aged 70–79 y ( The energy expenditure (EE) substudy was conducted between July 1998 and June 2000, and it included a race- and sex-stratified random selection of participants from the larger Health ABC Study. Substudy exclusion criteria included recent blood transfusion or intravenous fluid administration, use of supplemental oxygen or insulin, and overnight travel immediately before or during the EE measurement period. Participants were recruited from a list of 500 persons randomly selected from the original 3075 subjects and were balanced by race and sex. A replacement list of 200 subjects, also stratified for race and sex, was generated. Whenever a participant from the primary list was ineligible or refused to participate, a participant from the same race and sex subgroup was contacted from the replacement list to determine his or her eligibility and willingness to participate. The field centers had lower yields in several cells, and, at the end of 1999, the EE substudy was unbalanced with respect to race. The study was extended to the year 2000 and a new primary list was generated, which oversampled participants in the race and sex categories most lacking at each field center. A total of 323 subjects were enrolled in the EE substudy. No differences were noted between the EE substudy and the entire Health ABC cohort with respect to body mass index (BMI; in kg/m²), age, sex, race, knee extension strength, self-reported physical abilities, or self-reported energy intake. The substudy cohort, however, had a 5% faster time on a standardized walk. We assessed the influence of this difference by adding 400-m walking speed to the multiple regression for physical activity level (PAL). The 5% difference in speed translated to a 1.5% difference in PAL, and the substudy can therefore be considered as representative of the overall Health ABC Study cohort.

Protocol
The testing protocols were completed in the field centers during 2 visits over a 2-wk period. The participants arrived for both visits in the fasted state. During visit 1, participants received a dose of DLW for the measurement of TEE according to a protocol previously described in detail (16), and body composition was determined by using dual-energy X-ray absorptiometry (DXA). Approximately 14 d (14 ± 1 d) after the DLW dosing, the participants returned to the field centers for visit 2. Body weight was recorded, and the participants were asked to relax in a prone position for 30 min in a quiet room with a light cover for thermal comfort. After this, RMR was measured for 40 min by using respiratory gas exchange. During this visit, 2 urine samples were collected for the endpoint DLW analysis. Between visit 1 and visit 2, the subjects were asked to maintain their normal activity pattern.

The protocol was approved by the Internal Review Boards at the University of Tennessee–Memphis and the University of Pittsburgh. The participants were fully informed of the purpose and potential risks of the experimental protocol, and individual informed written consent was obtained before the study.

Total energy expenditure measurements
The TEE was determined by using the 2-point DLW method according to Schoeller et al (17, 18). The specific DLW protocol applied during the Health ABC EE substudy (ie, dose, sample preparations, mass spectrometry analyses of deuterium and 18-oxygen isotopic enrichments in biological specimens, and EE calculations) was detailed elsewhere (16). Briefly, DLW was orally dosed at 0.2 g ¹⁸O and 0.14 g deuterium/kg estimated total body water. Results were calculated according to Racette et al (19) with a food quotient estimated at 0.86 from the third National Health and Nutritition Examination Survey (20) and Black's formula (21). Ten percent of the population displayed delayed urinary isotopic equilibration, and thus plasma enrichment was used for calculation (16).

Blinded repeat isotopic analyses were completed in 16 participants by using duplicate urine aliquots. For the total body water, the within-subject repeatability, calculated as the average percentage difference between the 2 analyses, was -0.1 ± 1.2%. The within-subject analytic repeatability for TEE was 1.2 ± 5.4%. Results were within the limits of expectation predicted for the typical analytic variation in our laboratory.

Resting metabolic rate and activity-related energy expenditure
The RMR was measured on one of 2 Deltatrac II respiratory gas analyzers (Datex Ohmeda Inc, Helsinki). After the subjects rested for 30 min, gas exchanges were measured for 40 min. Data from the first 10 min were excluded, as were the data for 2 min after any movement or loss of wakefulness. The remaining minute-by-minute data were averaged. Methanol burn tests were performed in duplicate once or twice a month. Carbon dioxide recovery averaged 100.1 ± 1.4% at the Pittsburgh site and 100.5 ± 1.5% at the Memphis site. Gas exchange ratios for methanol differed by 2.5% (P < 0.001) between sites; those at Memphis (0.666 ± 0.014; n = 71) were accurate, and those at Pittsburgh (0.683 ± 0.015; n = 48) were a little greater than theoretical. The respiratory ratios for participants enrolled at Pittsburgh were corrected accordingly.

The diet-induced thermogenesis was assumed to be 8% of TEE (22), and the AEE was calculated as 0.92 x TEE - RMR. PAL was calculated as the ratio of TEE to RMR.

Body composition
Total body fat (fat mass, FM) and fat-free mass (FFM) were measured by DXA by using a Hologic 4500A Scanner (Hologic Inc, Watham, MA) with HOLOGIC software (version 8.21; Hologic Inc) for analysis. Health ABC validation studies, however, had shown that DXA FFM systematically overestimated FFM, and DXA FFM and FM were corrected accordingly (23).

Statistical analyses
Of the 323 subjects included in the EE substudy, 35 were excluded from the statistical analysis. Nine of the 323 subjects did not return to the field centers to complete the protocol, 2 failed to collect appropriate urine specimens, 8 had inconsistencies in their stable isotope data, 2 had highly variable RMR measurements, and 14 did not undergo DXA measurements within 15 d of the EE measurements.

Before the statistical analysis, the normality of the data was ascertained by using the Kolmogorov-Smirnov test. Differences between groups were tested by a factorial analysis of variance (ANOVA) with race, sex, and site as main effects. RMR was analyzed by an analysis of covariance (ANCOVA) with FFM as the covariate because regression analysis identified FFM as the main predictor of RMR (R² = 0.66). The model, however, rejected the hypothesis of slope homogeneity for race (FFM-by-race interaction: P = 0.02), but not for sex (FFM-by-sex interaction: P = 0.08). Thus, the interaction term for race was kept in the ANCOVA model for use in calculating the adjusted RMR values. To characterize the effects of the different slopes, RMR values adjusted for FFM were calculated at different percentiles of the population FFM. RMR was also adjusted for both FFM and FM; FFM was measured from DXA and FM was measured from isotope dilution to avoid circularities in analysis. Predictive equations of RMR were derived by using multiple regression analysis. Statistical analyses were performed by using STATVIEW software (version 5.01; SAS Institute Inc, Cary, NC) considering a P value of <0.05 to be significant. Values are expressed as means ± SDs, unless otherwise stated.

RESULTS  
Anthropometric data and body composition
The physical characteristics and body composition of the participants are presented in Table 1. To be consistent with the statistical analysis, the results are presented according to the main effects of the ANOVA. None of the 3-factor interactions were significant.

View this table:
TABLE 1. . Anthropometric and body-composition data¹

 
Sex effects
No sex differences by age or BMI were noted. As expected, the mean body weight of the men was 16% higher than that of the women. A significant sex-by-race interaction was observed for FM, which indicated that the general sex effect was dependent on race: FM weight was 25% higher in the black women than in the black men, whereas this sex difference was only 6% in the white subjects. FM expressed as percentage of body weight was one-third higher in the women than in the men.

Race effects
Age, weight, and BMI did not differ between races. The observation of a sex-by-race interaction for weight and BMI showed that the difference in weight and BMI between the blacks and the whites was dependent on sex. The weight of the black women was 10% greater than that of the white women, whereas no race difference was noted between the men; the same trend was noted for BMI. FM did not differ significantly between the black and white participants but was 5% higher in the white subjects than in the black subjects when expressed as a percentage of body weight. The sex-by-race interaction in FM showed, however, that most of the race difference was among the men rather than among the women.

Site effects
The age of the participants did not differ significantly between the 2 sites. Significant sex-by-site interactions were noted for weight (P = 0.05), BMI (P = 0.04), and FM (P = 0.02), and, as with the race effects, they were mainly attributed to differences among the women. Weight was 9% higher in the women in Pittsburgh than in the women in Memphis (73.7 ± 16.9 and 67.3 ± 13.9 kg, respectively), but it did not differ significantly between the men (Memphis: 82.8 ± 13.9 kg; Pittsburgh: 82.16 ± 13.2 kg). As expected, the same observation was made for BMI: the BMIs of the men from Memphis and Pittsburgh did not differ significantly (27.2 ± 4.5 and 27.5 ± 4.3, respectively), but the women from Pittsburgh had a 10% greater BMI than did the women from Memphis (29.0 ± 6.0 and 26.1 ± 5.1, respectively). Because FFM did not differ significantly between cities, these disparities were mainly attributed to differences in FM. Effectively, FM was shown to be 16% greater in the Pittsburgh women than in the Memphis women (30.5 ± 10.4 versus 26.0 ± 9.3 kg), whereas FM did not differ significantly between the men at the 2 sites (Memphis: 23.7 ± 8.0 kg; Pittsburgh: 23.3 ± 7.6 kg).

Energy metabolism
The energy components of the participants are presented in Table 2 stratified by the main effects of the ANOVA. None of the 3-factor interactions were significant.

View this table:
TABLE 2. . Energy expenditure components¹

 
Sex effects
As expected, the TEE was 25% lower in the women than in the men. This difference was due to a 21% lower RMR and a 31% lower AEE in the women. After statistical adjustment for FFM, RMR did not differ significantly between the men and the women. When AEE was normalized per unit of body weight, the difference observed in nonadjusted values was maintained. PAL did not differ significantly between the sexes, although a trend toward significance was observed.

Race effects
The TEE was 3% lower in the blacks than in the whites. Because AEE did not differ significantly between the races, the difference in TEE was mainly due to a 4% lower RMR in the blacks. The race-by-sex interaction observed for TEE shows the sex specificity of the race difference; ie, TEE did not differ significantly between the black women and the white women, but an 8% TEE difference was noted between the black men and the white men. The relation between FFM and the RMR differed significantly between the races (Figure 1). To account for this difference, race-specific RMR/FFM slopes were maintained in the ANCOVA model, and adjusted RMRs were estimated at different percentiles (25th, 50th, 75th) of the covariate. At all FFM percentiles, RMR adjusted for FFM was significantly lower in the blacks than in the whites (P < 0.001 for all), and it was 100 kcal/d at the 50th percentile. The race difference in RMR was maintained even after adjustment for both FFM and FM. The absence of a difference between races in AEE was maintained after normalization per kilogram of body weight. PAL did not differ significantly according to race. It is interesting that age was a significant determinant of RMR. However, by plotting RMR/FFM residuals against age, we observed that the age effect did not differ significantly between the blacks and the whites (Figure 1). Thus, the race difference in slopes of the relation RMR/FFM was not due to an artifact related to age.

View larger version (25K):
FIGURE 1.. Top: Regression plot of resting metabolic rate (RMR) against fat-free mass (FFM) in the black (n = 139; •) and the white (n = 149; ) participants. For the blacks, the relation displayed an intercept of 410 kcal/d (95% CI: 307, 513 kcal/d) and a slope of 16.3 kcal · kg^-1 · d^-1 (95% CI: 14.3, 18.2 kcal · kg^-1 · d^-1) with an R² of 0.66. For the whites, the relation had an intercept of 374 kcal/d (95% CI: 283, 473 kcal/d) and a slope of 18.9 kcal · kg^-1 · d^-1 (95% CI: 17.0, 20.8 kcal · kg^-1 · d^-1) with an R² of 0.74. The slopes differed significantly between the blacks and the whites (P = 0.02). Bottom: Relation between the residuals of the top figure and age. The overall relation was significant (R² = 0.07, P < 0.0001) with y = 819 –11 x age in the blacks (n = 139; •) and 733 –10 x age in the whites (n = 149; ). The slopes did not differ significantly between races (race-by-age interaction: P = 0.8).

 
Site effects
The TEE was 6% higher in the Pittsburgh subjects than in the Memphis subjects (2229 ± 428 and 2102 ± 459 kcal/d, respectively; P = 0.01). This was attributable to a difference in AEE but not in RMR. Effectively, AEE expressed in kcal/d was 12% lower in the Memphis subjects (672 ± 459 kcal/d) than in the Pittsburgh subjedts (760 ± 672 kcal/d; P = 0.01), and the difference (10%) was still significant after normalization for body mass (Memphis subjects: 9.0 ± 3.5; Pittsburgh subjects: 10.0 ± 9.0 kcal · d^-1 · kg^-1; P = 0.04). As expected, these differences were also observed in PAL, which was 4% higher in the Pittsburgh subjects (1.73 ± 0.22) than in the Memphis subjects (1.67 ± 0.24; P = 0.02).

Accuracy of the WHO and 2002 DRI energy requirement predictions
In Figure 2, we compared the estimated and measured energy requirements for each sex and race subgroup. Each step of calculation—ie, RMR, PAL, and then TEE—is compared with the WHO-derived estimates. Individual age, height, and weight and an estimated PAL were used to calculate energy requirements by using the new nonlinear equation of the 2002 DRIs. Because obesity has been shown to increase EE, the cohort was divided according to BMI: normal-weight and overweight persons combined (20 > BMI < 29.9) and obese persons (BMI =" BORDER="0">30). Ten participants, with a PAL of 1.60 ± 0.20, had a BMI <20, and they were excluded from this analysis because the DRI predictive equations were calculated for a person at an ideal BMI of 25. By splitting our data according to BMIs <30 and =" BORDER="0">30, we allowed a meaningful comparison of the Health ABC Study data with the DRIs at an average BMI of 25. Several points emerged from a repeated-measures ANOVA. The WHO calculations overestimated RMR by 15% (P < 0.0001). This overestimation was greater for the blacks than for the whites (19% and 12%, respectively; P < 0.0001) and greater for the obese persons than for the lean or overweight persons (18% and 14%, respectively; P < 0.0001). We also found that the WHO-proposed PAL was low by 10% (P < 0.0001), and the difference was greater in the men than in the women (7% and 15%, respectively; P = 0.0003). The WHO-derived energy requirements overestimated TEE by 5 ± 15% (P < 0.0001). More detailed analysis showed that this overestimation was observed in the women, but not in the men (10% and 0.2%, respectively; P < 0.0001), and it was more important in the obese persons than in the normal-weight persons (8% and 2%, respectively; P < 0.0001) and more important in the blacks than in the whites (7% and 2%, respectively; P = 0.002). Conversely, the 2002 DRI energy requirement estimates were accurate with an overall difference of 0 ± 14% for TEE (P = 1.0). No effects of sex (P = 0.2) or BMI range (P = 0.4) appeared in the statistical analysis. A race effect (P = 0.002) was, however, observed, but the presence of a sex-by-race-by-BMI range interaction (P < 0.02) showed that the race effect was due only to a significant overestimation of energy requirements in obese black men (12%). Nonetheless, as was seen with the WHO recommendations, the range of PAL used in the DRI predictive equations (1.4 >PAL <1.6) was underestimated.

View larger version (66K):
FIGURE 2.. Mean (±SEM) measured and predicted resting metabolic rate (RMR), physical activity level (PAL), and total energy expenditure (TEE). Results are grouped by BMI (kg/m²) by using a cutoff of 30 to separate the high-energy cost of obesity. Comparisons were made with the recommendations of the World Health Organization (WHO) and with the 2002 dietary reference intakes (DRIs). Note that WHO energy requirements are calculated by using the product of a fixed PAL and sex-based estimation of RMR from body weight, whereas DRIs are calculated from a nonlinear equation by using age, height, weight, and an estimated PAL range of 1.4–1.6. The mean value of 1.5 is indicated on the figure. The WHO calculations overestimated RMR by 15% (P < 0.0001). The overestimation was greater for the blacks than for the whites (19% and 12%, respectively; P < 0.0001) and greater for the obese persons than for the lean or overweight persons (18% and 14%, respectively; P < 0.0001). WHO-proposed PAL was low by 10% (P < 0.0001), and the difference was greater in the men than in the women (7% and 15%, respectively; P = 0.0003). The WHO-derived energy requirements overestimated TEE by 5 ± 15% (P < 0.0001), but this overestimation was observed in the women, but not in the men (10% and 0.2%, respectively; P < 0.0001), and it was more important in the obese persons than in the normal-weight persons (8% and 2%, respectively; P < 0.0001) and more important in the blacks than in the whites (7% and 2%, respectively; P = 0.002). The 2002 DRI energy requirement estimates were accurate with an overall difference of 0 ± 14% with TEE (P = 1.0). No significant effects of sex (P = 0.2) or BMI range (P = 0.4) were noted. A race effect (P = 0.002), explained by an overestimation of energy requirements in black men, was observed (12%). The WHO recommendations underestimated the DRI range for PAL. The group with a BMI =" BORDER="0">30 included 23 black women, 14 white women, 18 black men, and 18 white men. The group with a BMI of 20–29.9 included 42 black women, 59 white women, 51 black men, and 53 white men.

 
Health ABC Study–derived predictive equations of energy requirements
Predictive equations of RMR based on the Health ABC Study cohort are proposed in Table 3. Because the RMR:FFM relation differed by race, the RMR predictive equations were independently generated for the blacks and the whites. The frequency distribution of PAL among subjects with BMIs above and below 30 are shown in Figure 3. Because PAL did not differ by BMI, sex, or race, TEE can be predicted by multiplying the derived RMR by 1.52, 1.67, and 1.85 for persons of below-average (taken at the 25th PAL population percentile), average (50th percentile) and above-average (75th percentile) physical activity, respectively.

View this table:
TABLE 3. . Predictive equations of resting metabolic rate based on weight or fat-free mass in whites (n = 149) and blacks (n = 139)¹

 

View larger version (30K):
FIGURE 3.. Frequency distribution of the physical activity levels (PALs) grouped by BMI (kg/m²). n = 206 persons with a BMI of 20–29.9; n = 72 persons with a BMI =" BORDER="0">30.

 

DISCUSSION  
In the Health ABC Study, the women had lower TEEs than did the men (539 kcal/d), as expected from previous studies (24–26). Among the previous studies, the largest one reported an overall sex difference (741 kcal/d) but no effect due to race on unadjusted values of EE (24). Because FFM is a main determinant of TEE, those authors adjusted TEE to account for the difference in FFM. In doing so, they unmasked a lower TEE in blacks than in whites (43 kcal/d), and the sex difference was maintained (383 kcal/d). In the Health ABC Study cohort, unadjusted TEE was lower in the blacks than in the whites (72 kcal/d), and the presence of a sex-by-race interaction showed that the race effect was attributable to a difference in the men, but not in the women. We propose that it is an oversimplification to linearly adjust TEE for FFM only, because TEE differences are explained by variations in RMR, AEE, or diet-induced thermogenesis, each of which should be adjusted for their determinant or determinants. Nevertheless, to compare the Health ABC Study data with data from previous reports (24), we also adjusted TEE by FFM and confirmed the race effect (146 kcal/d; P < 0.001), but not the sex effect (P = 0.23); however, these studies differ in sample size and the age range of the subjects (the largest study has 65 participants with a mean age of 65 y; 24). Six other studies, as reviewed elsewhere (13), investigated the race differences in TEE in prepubertal or early pubertal and middle-aged subjects. One of these studies showed a tendency for lower TEE in black subjects. In this current study, FFM differences explained the sex effect in TEE, but failed to account totally for the race difference.

On the other hand, because the relation between FFM and RMR has an intercept different from zero, Ravussin and Bogardus (27) concluded that RMR values should be adjusted for FFM by a multivariate-based approach. Theoretically, however, such an analysis can be performed only if the slopes for RMR versus FFM for blacks and whites or men and women are similar (28). In previous, smaller studies, slopes were not found to be different, but the Health ABC Study cohort is the largest group thus studied; we observed nonhomogeneity of the slopes for race but not for sex (ie, RMR differences between blacks and whites increased with FFM). Gannon et al (13) also reported that 10 of the 15 studies they reviewed had found a lower RMR in blacks than whites. The differences ranged from 81 to 275 kcal/d and were not explained by age, FFM, or method. Given the difference in the slopes, we applied the unequal-slope model of ANCOVA and calculated adjusted RMR at different percentiles of FFM in the population. Indeed, at all percentiles, the RMR adjusted for FFM was lower in blacks than in whites and was 99 kcal/d at the 50th FFM percentile (50.2 kg). Regarding the sex effect, our results do not support the observations of others (29–31) that RMR adjusted for FFM is lower in women than men. Given the age of our cohort, we speculate that the sex difference decreases with advanced age, perhaps because of concomitant decreases in sex hormones (32).

ANCOVA implies that FFM is a homogeneous compartment, but Heymsfield et al (33) showed that FFM is not an energetically homogeneous compartment, but, rather, that tissues vary with respect to heat production per gram of FFM. Viewed from this perspective, the nonhomogeneity of the slope suggests that, with increasing body mass, blacks respond with greater fractional increases in low-metabolic-rate tissues (ie, skeletal muscle, connective tissue, bone) than do whites. However, the similar intercepts suggest that blacks and whites are comparable in terms of the high-metabolic-rate residual mass (ie, heart, kidney, brain). Aloia et al (34) showed that body cell mass is similar in white women and black women in early adulthood, but declines with age to a greater extent in white women than in black women. Therefore, the unequal slopes observed in the present study may be specific to our age group and not simply due to the greater power resulting from the sample size.

It is widely accepted that physical activity and EE during physical activity decline with age, but activity in the elderly varies greatly, depending on health and independence. The mean values therefore depend on the subjects included in the sample. The Health ABC Study cohort was selected to reflect a relatively healthy, well-functioning group of older persons who were free of severe disabilities. Even so, the CV for AEE was 40.9%. The sex difference observed in unadjusted AEE in the Health ABC Study cohort was expected (24). AEE adjusted for FFM, however, did not show an effect of sex (P = 0.55) or race (P = 0.52) in the Health ABC Study cohort. Nevertheless, on the basis of theoretical considerations (35), we consider the use of the ratio of AEE to kilogram of body weight a better approach to adjusting AEE. When we did so, the women had an AEE 1.2 kcal · kg^-1 · d^-1 lower than did the men, without any effect of race. Other studies conducted in younger subjects had mixed findings with regard to low AEEs in black women (36, 37). Because low physical activity is a risk factor for obesity with increasing age (38), special attention should be given to interventions that increase physical activity in the elderly, especially black women, given that their peak prevalence of obesity (60%) occurs at a later age (60–69 y; 39).

PAL is an indirect index of physical activity and is useful in recommending energy intakes based on measured or estimated RMR and a level of physical activity. No effect of race or sex was noted in PAL, despite a tendency toward significance for sex. Several investigators have reviewed PAL values in the elderly (7, 11, 40, 41). Black et al (41) suggested that PAL is 1.62 for women aged 65–74 y and 1.48 for those aged >75 y. For men in those same age ranges, PAL is 1.61 and 1.54, respectively. Similar values for women (1.61) were found by Starling et al (11) in a recent review, but they reported a higher value for men (1.75). Our values are close to the estimate of Starling et al (11).

The WHO-derived energy requirements are calculated in 2 steps (14). First, the RMR is estimated from body weight according to a modified Schofield equation (42). Second, the resulting RMR is multiplied by PALs of 1.51 and 1.56 for men and women aged >65 y, respectively. Sawaya et al (43) and Roberts et al (40) suggested, on the basis of studies of relatively few subjects, that the WHO-derived PAL is underestimated in elderly men but not in elderly women. More recently and in a larger group of subjects, Starling et al (11) also reported that the WHO equations underestimated PAL among elderly men, but not among elderly black women. In contrast, we observed that the WHO equations do not underestimate energy requirements as previously suggested but, rather, overestimate them. This overestimation was only 5 ± 15% for the overall Health ABC Study cohort, but it was greater in women, blacks, and obese persons. More important, the WHO equations are derived from overestimated RMR and underestimated PAL. This WHO-derived RMR overestimation, but not the underestimation of PAL, was previously reported (40). It is relevant that the WHO database for derived RMR equations (42) contained few persons >60 y old.

In 1985 WHO recommended (14) that energy requirements be derived from measurements of TEE, and the 2002 DRI (15) sex-based energy-requirement equations were so derived. The DRI nonlinear sex-based equations were derived for an spectrum of adults aged >19 y old, but the contribution of elderly and African American persons with normal BMIs was small. The accuracy of these equations is, nevertheless, very good—a difference of 0 ± 14% from the actual TEE was noted—but the elderly in our study had PALs above the range suggested for people with low activity (1.4 > PAL < 1.6). No relevant race effects were noted, either because the error in the energy requirement estimates overlapped the race difference in TEE or because the DRI database was somehow weighted for the race effect by inclusion of data from some black persons. To overcome the limitations of the current WHO recommendations and to account for the DRIs' underestimation of PAL, we propose RMR prediction equations that take into account both race and sex and that can be easily used by clinicians.

ACKNOWLEDGMENTS  
We thank the staff at the Memphis and Pittsburgh Health, Aging, and Body Composition Study field centers for subject recruitment and specimen collection.

SB performed data analysis and drafted the manuscript. DAS supervised the data analysis and edited the manuscript. DB consulted on study design and edited the manuscript. MED supervised the data collection and edited the manuscript. FT supervised the data collection and edited the manuscript. EMS designed the study and edited the manuscript. TBH was the principal investigator, designed the study, and edited the manuscript. SBK consulted on study design, recruited subjects, and edited the manuscript. JEE designed the study and edited the manuscript. None of the authors had personal or financial conflicts of interest.

REFERENCES  

1. Fowles D. A profile of older Americans. Washington, DC: Program Resource Department, Administration on Aging, US Department of Health and Human Services, 1999.
2. Balcombe NR, Sinclair A. Ageing: definitions, mechanisms and the magnitude of the problem. Best Pract Res Clin Gastroenterol 2001;15:835–49.
3. Baggio G. Occurrence of chronic diseases and their impact on physical disability over the whole spectrum of aging: from 65 to over 100 years of age. Z Gerontol Geriatr 1999;32:420–4.
4. Serow WJ. Economic and fiscal implications of aging for subnational American governments. J Aging Soc Policy 2001;12:47–63.
5. Meydani M. Nutrition interventions in aging and age-associated disease. Ann N Y Acad Sci 2001;928:226–35.
6. Morley J. The role of nutrition in the prevention of age associated diseases. In: Morley J, Glick Z, Rubenstein L, eds. Geriatric nutrition: a comprehensive review. New York: Raven Press, 1990:89–103.
7. Ritz P. Factors affecting energy and macronutrient requirements in elderly people. Public Health Nutr 2001;4:561–8.
8. Elia M, Ritz P, Stubbs RJ. Total energy expenditure in the elderly. Eur J Clin Nutr 2000;54(suppl 3):S92–103.
9. Roberts SB, Fuss P, Heyman MB, et al. Control of food intake in older men. JAMA 1994;272:1601–6.
10. Das SK, Moriguti JC, McCrory MA, et al. An underfeeding study in healthy men and women provides further evidence of impaired regulation of energy expenditure in old age. J Nutr 2001;131:1833–8.
11. Starling RD, Poehlman ET. Assessment of energy requirements in elderly populations. Eur J Clin Nutr 2000;54(suppl 3):S104–11.
12. Trabulsi J, Schoeller DA. Evaluation of dietary assessment instruments against doubly labeled water, a biomarker of habitual energy intake. Am J Physiol Endocrinol Metab 2001;281:E891–9.
13. Gannon B, DiPietro L, Poehlman ET. Do African Americans have lower energy expenditure than Caucasians? Int J Obes Relat Metab Disord 2000;24:4–13.
14. James W, Schofield WN. Human energy requirements. FAO publication. Oxford, United Kingdom: Oxford University Press, 1990:25.
15. Institute of Medicine. Dietary reference intakes: energy, carbohydrate, fiber, fat, fatty acids, cholesterol, protein and amino acids. A report of the Panel on Micronutrients, Subcommittees on Upper Reference Levels of Nutrients and Interpretation and Uses of Dietary Reference Intakes, and the Standing Committee on the Scientific Evaluation of Dietary Reference Intakes. Washington, DC: The National Academies Press, 2002 (Prepublication Copy, uncorrected proofs).
16. Blanc S, Colligan AS, Trabulsi J, et al. Influence of delayed isotopic equilibration in urine on the accuracy of the (2)H(2)(18)O method in the elderly. J Appl Physiol 2002;92:1036–44.
17. Schoeller DA, van Santen E. Measurement of energy expenditure in humans by doubly labeled water method. J Appl Physiol 1982;53:955–9.
18. Schoeller DA, Ravussin E, Schutz Y, Acheson KJ, Baertschi P, Jequier E. Energy expenditure by doubly labeled water: validation in humans and proposed calculation. Am J Physiol 1986;250:R823–30.
19. Racette SB, Schoeller DA, Luke AH, Shay K, Hnilicka J, Kushner RF. Relative dilution spaces of 2H- and 18O-labeled water in humans. Am J Physiol 1994;267:E585–90.
20. Kant AK. Nature of dietary reporting by adults in the third National Health and Nutrition Examination Survey, 1988–1994. J Am Coll Nutr 2002;21:315–27.
21. Black AE, Prentice AM, Coward WA. Use of food quotients to predict respiratory quotients for the doubly-labelled water method of measuring energy expenditure. Hum Nutr Clin Nutr 1986;40:381–91.
22. Visser M, Deurenberg P, van Staveren WA, Hautvast JG. Resting metabolic rate and diet-induced thermogenesis in young and elderly subjects: relationship with body composition, fat distribution, and physical activity level. Am J Clin Nutr 1995;61:772–8.
23. Tylavsky F, Lohman T, Blunt BA, et al. QDR 4500A DXA overestimates fat-free mass compared with criterion methods. J Appl Physiol 2003;94:959–65.
24. Carpenter WH, Fonong T, Toth MJ, et al. Total daily energy expenditure in free-living older African-Americans and Caucasians. Am J Physiol 1998;274:E96–101.
25. Morio B, Ritz P, Verdier E, Montaurier C, Beaufrere B, Vermorel M. Critical evaluation of the factorial and heart-rate recording methods for the determination of energy expenditure of free-living elderly people. Br J Nutr 1997;78:709–22.
26. Rothenberg E, Bosaeus I, Lernfelt B, Landahl S, Steen B. Energy intake and expenditure: validation of a diet history by heart rate monitoring, activity diary and doubly labeled water. Eur J Clin Nutr 1998;52:832–8.
27. Ravussin E, Bogardus C. Relationship of genetics, age, and physical fitness to daily energy expenditure and fuel utilization. Am J Clin Nutr 1989;49:968–75.
28. Lee J. Analysis of covariance by the SAS GLM procedure. Comput Biol Med 1987;17:221–5.
29. Ferraro R, Lillioja S, Fontvieille AM, Rising R, Bogardus C, Ravussin E. Lower sedentary metabolic rate in women compared with men. J Clin Invest 1992;90:780–4.
30. Arciero PJ, Goran MI, Poehlman ET. Resting metabolic rate is lower in women than in men. J Appl Physiol 1993;75:2514–20.
31. Poehlman ET, Toth MJ, Ades PA, Calles-Escandon J. Gender differences in resting metabolic rate and noradrenaline kinetics in older individuals. Eur J Clin Invest 1997;27:23–8.
32. Roshan S, Nader S, Orlander P. Review: ageing and hormones. Eur J Clin Invest 1999;29:210–3.
33. Heymsfield SB, Gallagher D, Kotler DP, Wang Z, Allison DB, Heshka S. Body-size dependence of resting energy expenditure can be attributed to nonenergetic homogeneity of fat-free mass. Am J Physiol Endocrinol Metab 2002;282:E132–8.
34. 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.
35. Schoeller DA, Jefford G. Determinants of the energy costs of light activities: inferences for interpreting doubly labeled water data. Int J Obes Relat Metab Disord 2002;26:97–101.
36. Sun M, Gower BA, Nagy TR, Trowbridge CA, Dezenberg C, Goran MI. Total, resting, and activity-related energy expenditures are similar in Caucasian and African-American children. Am J Physiol 1998;274:E232–7.
37. Wong WW, Butte NF, Ellis KJ, et al. Pubertal African-American girls expend less energy at rest and during physical activity than Caucasian girls. J Clin Endocrinol Metab 1999;84:906–11.
38. Poehlman ET, Toth MJ, Bunyard LB, et al. Physiological predictors of increasing total and central adiposity in aging men and women. Arch Intern Med 1995;155:2443–8.
39. Kuczmarski RJ, Flegal KM, Campbell SM, Johnson CL. Increasing prevalence of overweight among US adults. The National Health and Nutrition Examination Surveys, 1960 to 1991. JAMA 1994;272: 205–11.
40. Roberts SB. Energy requirements of older individuals. Eur J Clin Nutr 1996;50:S117–8.
41. Black AE, Coward WA, Cole TJ, Prentice AM. Human energy expenditure in affluent societies: an analysis of 574 doubly-labelled water measurements. Eur J Clin Nutr 1996;50:72–92.
42. Schofield WN. Predicting basal metabolic rate, new standards and review of previous work. Hum Nutr Clin Nutr 1985;39(suppl 1):5–41.
43. Sawaya AL, Saltzman E, Fuss P, Young VR, Roberts SB. Dietary energy requirements of young and older women determined by using the doubly labeled water method. Am J Clin Nutr 1995;62:338–44.