Prior to this study, only one study had been published on energy requirements in individuals in very late life . Further, most studies assessing energy requirements take a cross sectional approach, so no information regarding changes in energy requirements in late life were available. Since decreases in TEE increase obesity risk, understanding the energy requirements in late life and how they change over time is critical . Based on DLW measures of TEE, we found that the energy requirements in very late life (>80 years of age) were 2208 ± 376 kcal/d for men and 1814 ± 337 kcal/d for women. As expected, these TEE values are slightly greater than those reported by Rothenberg et al.  who reported TEE to be around 1936 kcal/d and 1506 kcal/d for men and women over 90 years of age, respectively. Further, these values were slightly lower than those reported in studies with subjects ranging from 55–79 years of age. Blanc et al.  reported TEEs of 2521 kcal/d (men) and 1885 kcal/d (women) for subjects between the ages of 70–79. Similarly, Carpenter et al.  reported TEEs of 2584 kcal/d and 1946 kcal/d in men and women over 55 years of age, respectively. Although the energy requirements decreased in both men and women from the 8th to 9th decade of life in our study (baseline to follow-up), the decrease in TEE was significant in men, but not in women. The average decrease in the energy requirement during the 7-year time span was −274 ± 338 kcal/d in men (p < 0.05) and −77 ± 322 kcal/d in women (ns). Other notable decreases were found in RMR, AEE and PAL in men, but not women.
The repeated measure design captured the average decrease in energy requirements due to aging, but it was also possible to gain some insight into the between individual differences in the rate of the aging effect. Trabulsi et al.  examined the precision of DLW measurements of TEE in 24 subjects measured twice within a 2-week period which is a short period that should be free of the influences of aging. They found that the coefficient of variation (CV) in TEE was 5.1% which included a 2.9% analytical variation and a 4.2% physiologic variation. This 5.1% CV is for a single measure thus contributed a 7.1% (5.1x1.4) CV to the change score, which was smaller than that observed in the men (11%) in our study. This suggests that our data represents actual changes in these subjects from baseline to follow-up. Further, we have previously reported the variation in RMR measurements on 2 consecutive days to be 3.0% . This 3.0% CV and 4.2% CV to the change score was also smaller than that which was observed in the men (6.75%) and women (6.89%) from baseline to follow-up.
We wanted to determine whether the changes in TEE from baseline to follow-up could be attributed to changes in RMR, changes in AEE (or PAL), or a combination of the two. Only the men showed a significant decrease in TEE. They also had significant decreases in RMR, AEE, and PAL. The average decrease in RMR was 70 ± 125 kcal/day which represents about 25% of the decrease in TEE. AEE decreased an average of 177 ± 301 kcal/day which accounts for the majority of the decrease in TEE. Therefore, we can conclude that the decreases in TEE in male subjects were mostly due to decreases in AEE rather than RMR. This suggests a priority for promoting physical activity in older persons. Since the female subjects did not show decreases in TEE, it is not surprising that neither RMR nor AEE (or PAL) decreased significantly from baseline to follow-up. It should be noted that women did start at a lower AEE and PAL at baseline, so any possible decline in activity would not be expected to be as great as that observed in men. However, their PAL level was not as low as a very sedentary individual, so it would have been plausible for the women to show a decrease in activity. A decline was simply not observed in the women in this study.
Since we observed decreases in both RMR and AEE in men, we then wanted to determine if specific variables could account for those changes. We set out to establish whether or not changes in FFM could explain the observed changes in RMR, and whether or not changes in body weight could explain the observed changes in AEE. Both RMR and FFM significantly decreased in male subjects from baseline to follow-up. Further, there was a positive correlation between the change in RMR vs. FFM for men (r = 0.49, p < 0.001). Therefore, approximately 49% of the decrease in RMR could be explained by decreases in FFM for men. This highlights the importance of not just increasing physical activity, but also performing more resistance exercise to prevent or slow down loss of muscle mass in aging adults, especially men. However, we acknowledge that this recommendation is only effective if elderly adults are willing and able to perform resistance exercise. The application of this recommendation is questionable since a very small percentage of the adult U.S. population actually meets physical activity guidelines  and less than 20% of men and women report strength training two or more times per week . Both AEE and body weight also decreased significantly in men from baseline to follow-up. However, a correlation analysis between changes in AEE vs. changes in body weight did not reach significance (r = 0.24, p = 0.10). Therefore, we cannot conclude that changes in body weight accounted for the changes in AEE that were observed in the male subjects.
In female subjects, both FFM and body weight decreased significantly, but there were no changes in AEE or RMR. This could be due to the fact that the decrease in FFM was relatively small (0.5 kg) and was possibly not a great enough change to alter RMR. Therefore, there was no significant correlation between changes in RMR vs. changes in FFM (r = −0.08, ns) or for changes in AEE vs. changes in body weight in women (r = −0.07, ns). It should be noted that the correlation analyses for both men and women had relatively small sample sizes (n = 47 for men and n = 40 for women). Therefore, it is possible the changes in these variables or total subject number were simply too small to yield significant changes or correlations. However, even with that small sample size, some significant changes and correlations were observed in men. Therefore, based on this data, we believe that elderly women may just not show the same pattern of change in TEE, RMR, and AEE that men do. Finally, it should also be noted that the women in our study did show decreases in AEE and RMR; they just were not statistically significant. It appears that there was a large amount of variation in this data, especially the AEE data in women at follow-up. This large amount of variation could explain the lack of statistically significant declines in women and/or also shows that variation in AEE may be greater in women in the 9th decade of life compared to men. Additionally, the women in this study had lower PALs to begin with compared to the men. This could have impacted the magnitude of decreases in PAL, or lack thereof, in women compared to men. This still indicates that changes in these variables in women are different than men in late life, but it may be partially due to women starting at lower PALs.
As shown above, sex differences of longitudinal change of TEE, RMR, and AEE were found, and the decreasing rate with aging was larger in men than in women. Although its biological reasons are unknown, the results of several previous longitudinal studies may have relevance to this phenomenon. Nakamura and Miyao  reported that the rate of biological aging calculated by using 7-year longitudinal data of forced expiratory volume, systolic blood pressure, red blood cells, albumin, and blood urea nitrogen was faster in men than in women. Kimura et al.  reported that the rate of physical fitness aging calculated by using 7-year longitudinal data of walking speed, functional reach, one leg stand, vertical jump and grip strength was also faster in men than in women. It would be interesting for future studies to explore the relationship between aging of metabolic aspects and biological or physical functional aspects.
Several prediction equations exist to predict or estimate TEE. As expected, there is some error associated with each of these measures as an estimate of physical activity is required. The DRI and WHO prediction equations have been widely used to estimate energy requirements. However, the WHO equation does not provide an equation for individuals over the age of 60 years. It was unknown how accurately either the DRI or WHO equations could predict TEE in an elderly population in their 8th or 9th decades of life. Only the DRI equation predicted a similar TEE as the measured TEE. In both men and women, the WHO equation significantly over-predicted TEE. This over-prediction occurred with using the lowest level of activity factor associated with the WHO equation (activity factors range from 1.6-2.6; 1.6 was used in this analysis). The Bland-Altman plots indicated that the over-prediction of TEE by the WHO equation occurred at all ranges of energy expenditure while the DRI was accurate at all ranges of energy expenditure. Additionally, the limits of agreement were much wider or greater for the WHO plot indicating worse agreement than that of the DRI plot with DLW measured TEE. Based on these results, it is apparent that the development of an age appropriate WHO equation is necessary for individuals in very late life. Our multiple linear regression analysis indicated that age, RMR, and FFM were the best predictors of TEE in this population. However, obtaining RMR and FFM measurements is difficult to do for the general public, so developing an equation using these variables is likely not clinically feasible.
Very little is known about the energy requirements in very late life. This study provides some initial insight; however, some limitations do exist. It may not be appropriate to extrapolate this data to other populations in the U.S. or worldwide. While this study represents one of the largest longitudinal studies on changes in EE components among older adults, only 27.48% of the initial sample completed the follow-up measurements. Compared to those who did not participate, the follow-up sample had a lower prevalence of cancer, had a greater physical performance score, and spent more time performing physical activity at baseline. We acknowledge that the sample is biased toward those individuals healthy enough to participate in the follow-up study. Consequently, we caution that the results may be skewed toward the healthier subset of the aged and thus are only modestly representative of the age-related changes in EE experienced during aging in late-life. However, it is also possible that these subjects are similar, and therefore, representative to those that are surviving into their 9th decade of life. Importantly, even if these subjects are “healthier” than others of the same age, disease prevalence and the number of diseases affecting at least some of the participants in this study was quite high compared to what one might find in a younger population. Therefore, due to the relatively small number of subjects and the many potential confounding variables (many diseases), we did not adjust variables such as TEE, RMR, AEE, or FFM for any of these potential confounders.