Daily Activity Energy Expenditure and Mortality Among Older Adults FREE

Observational studies have shown that older adults who report low physical activity levels are at elevated risk of mortality compared with those who report moderate or high levels of activity.^1- 6 These findings were based on questionnaire assessments of physical activity, which are subject to recall bias,^7- 8 are unable to account for free-living activity,⁹ and typically overestimate actual amounts of physical activity.^10- 11 Furthermore, self-reported physical activity does not provide accurate estimates of absolute amounts of activity (kilocalories per day) and thus cannot be evaluated to determine whether higher levels of activity-induced energy expenditure confer survival advantages.¹⁰

The most accurate and precise method of determining free-living energy expenditure uses water labeled with the stable isotopes of ²H and ¹⁸O (doubly labeled water).¹² When ingested, ²H is eliminated as water and ¹⁸O is eliminated as water and carbon dioxide. The excess disappearance rate of ¹⁸O relative to ²H is a measure of the carbon dioxide production rate, a direct measure of total energy expenditure. Therefore, over a standard amount of time individuals who have less ¹⁸O relative to ²H expend more energy. After accounting for resting metabolic rate and energy from the thermic effect of meals, an objective estimate of energy expended through free-living activity is derived. This technique allows for estimation of activity in an individual's normal environment over approximately 2 weeks. This method is valid when compared with energy expenditure in respiratory chambers¹³ and has excellent repeatability in younger and older adults¹⁴; therefore it is considered the gold standard measure of free-living activity energy expenditure.^9,12

The doubly labeled water method captures any form of physical activity ranging from purposeful exercise to simple fidgeting, whereas physical activity questionnaires generally address basic volitional activities (eg, household chores, walking, and vigorous exercise). Although this method can address whether higher levels of free-living activity energy expenditure are associated with mortality risk, to our knowledge no studies have been conducted to determine whether free-living activity energy expenditure assessed in this way is related to longevity among older adults. The purpose of this study was to determine the association of free-living activity energy expenditure, measured using doubly labeled water coupled with resting metabolic rate, with all-cause mortality in a group of high-functioning, community-dwelling older adults.

In 1997-1998, investigators from the University of Pittsburgh (Pittsburgh, Pa) and the University of Tennessee (Memphis) recruited 3075 participants aged 70 to 79 years from a random sample of white Medicare beneficiaries and all age-eligible black community residents to participate in the Health, Aging, and Body Composition (Health ABC) study. Eligibility criteria included self-reporting no difficulty in walking a distance of 0.4 km or climbing at least 10 stairs, independently performing activities of daily living, plans to live in the area for the next 3 years, and no evidence of life-threatening illnesses. The sample was approximately balanced for sex (51% women) and 42% of participants were black. Participants self-designated race/ethnicity from a fixed set of options (Asian/Pacific Islander, black/African American, white/Caucasian, Latino/Hispanic, do not know, other). Blacks were oversampled to ensure adequate numbers to examine whether results varied by race because of differences such as body composition between blacks and whites.

An energy expenditure substudy carried out in 1998-1999 enrolled 323 individuals and has been described in detail elsewhere.^15- 16 A total of 323 individuals were contacted based on random selection within sex and race and were asked to participate in a study on energy expenditure. Individuals who volunteered were paid. Twenty-one individuals were excluded from this analysis because of failure to complete the protocol, lack of appropriate urine volume specimens, or lack of isotope or resting metabolic rate data to meet a priori criteria, leaving an analytic sample of 302 individuals (150 men and 152 women). Sixty percent (n = 179) and 40% (n = 123) of individuals participated in the energy expenditure substudy in 1998 and 1999, respectively. Compared with the full Health ABC cohort, the energy expenditure substudy included 8% more blacks but there were no differences in age, sex, mobility function (measured as gait speed), self-reported walking ability, or self-reported physical activity (eg, walking, stair climbing, working, volunteering, and caregiving). Written, informed consent was obtained from each participant and was approved by the institutional review boards at the University of Pittsburgh and the University of Tennessee.

Total energy expenditure was measured using doubly labeled water. This procedure was previously described in detail.¹⁵ Measurements were obtained between 2 visits separated by 2 weeks. On the first visit, participants ingested an estimated 2-g/kg total body water dose of doubly labeled water. This dose was composed of an estimated 1.9-g/kg total body water of 10% H₂¹⁸O and an estimated 0.12-g/kg total body water of 99.9% ²H₂O. After dosing, 3 urine samples were obtained at approximately 2, 3, and 4 hours. Two consecutive urine voids were taken during a second visit to the laboratory 15 days after the first visit. Plasma from a 5-mL blood sample was obtained from everyone but only used for those who had evidence of delayed isotopic equilibration likely caused from urine retention in the bladder (n = 28).¹⁵ Urine and plasma samples were stored at –20°C until analysis by isotope ratio mass spectrometry.

Dilution spaces for ²H and ¹⁸O were calculated according to the method by Coward.¹⁷ Total body water was calculated as the average of the dilution spaces of ²H and ¹⁸O after correction for isotopic exchange (1.041 for ²H and 1.007 for ¹⁸O). Carbon dioxide production was calculated using the 2-point doubly labeled water method outlined by Schoeller et al^13,18 and total energy expenditure was derived using the equation by Weir¹⁹ with a respiratory quotient of 0.86. All values of energy expenditure were converted to kilocalories per day and the thermic effect of meals was assumed to be 10% of total energy expenditure.²⁰ The within-subject repeatability of total energy expenditure was based on blinded, repeated, urine isotopic analysis and was excellent (mean [SD], 1.2% [5.4%]; n = 16) and compared well with rates given in a review article.¹⁴

Resting metabolic rate was measured via indirect calorimetry using a Deltatrac II respiratory gas analyzer (Datex Ohmeda Inc, Helsinki, Finland) and has been described in detail elsewhere.¹⁶ While in a fasting state and after 30 minutes of rest, a respiratory gas exchange hood was placed over the individual's head and the resting metabolic rate was measured for 40 minutes. To avoid a gas exchange created by the initial placement of the hood, only the final 30 minutes were used in subsequent calculations. Movement or sleeping during the test was noted and those values were excluded from the resting metabolic rate calculation.

Free-living activity energy expenditure was expressed in 2 ways.²¹ Activity energy expenditure was calculated as (total energy expenditure × 0.90) − resting metabolic rate; removing energy expenditure from the thermic effect of meals and subtracting energy devoted to basal metabolism. Activity energy expenditure was defined as the amount of kilocalories an individual expends in any activity per day. Physical activity level was calculated as total energy expenditure/resting metabolic rate. The division of total energy expenditure by resting metabolic rate, a major determinate of which is lean mass, adjusts for differences in body composition (in part reflecting weight and sex).²² This formula was adopted by the Food and Agriculture Organization, the World Health Organization, and the United Nations University.²³ These agencies have developed physical activity level categories (sedentary: 1.40-1.69; active, 1.70-1.99; vigorous activity, 2.00-2.40) but they were not used in this study because of the paucity of data to support their validity in older adults. Activity energy expenditure and physical activity level are highly correlated (r = 0.91) but offer different advantages (eg, simplicity of expression and inherently controlling for differences in body composition, respectively).

Physical activity over the past 7 days was assessed by an interviewer-administered questionnaire at the time of the doubly labeled water dosing. Questions about walking for exercise, other walking, climbing stairs, working for pay, and volunteering were assessed for both duration and intensity. Another question asked about caregiving but only the duration of the activity was assessed. The duration and intensity level of these activities were used to estimate energy expenditure with established metabolic equivalent values for each activity.²⁴ An additional question asked whether participants performed high-intensity exercise such as bicycling, swimming, jogging, racquet sports, climbing stairs, rowing, or cross-country skiing but information on duration and intensity were not collected.

Vital status was ascertained by telephone contact every 6 months over an 8-year period (1998-2006). Date of death was verified with death certificates and survival time was defined as the time of the second energy expenditure visit to the date of death or date of last contact. There were too few deaths to assess cause-specific mortality.

Self-reported health status (category score of 1-5 corresponding with categories of excellent to poor), body fat, body weight, and height were measured at the first energy expenditure visit. Body fat was assessed using dual-energy x-ray absorptiometry (QDR-4500, version 8.21, Hologic Inc, Bedford, Mass) and calculated as the ratio of body fat mass to total mass (percentage of body fat). Body weight was measured while the individual was wearing a hospital gown (no shoes) with a calibrated balance beam scale. Height was measured with a stadiometer. The following self-reported medical conditions were confirmed by treatment and/or medication: cardiovascular disease (hypertension, coronary heart disease, myocardial infarction, and stroke), lung disease, diabetes, hip or knee osteoarthritis, osteoporosis, cancer, and depression and were updated at the doubly labeled water visit in 1998-1999. Education (<high school, high school graduate, and >high school), smoking behavior (never, former, or current), and sleep duration were assessed during the first Health ABC annual clinic visit in 1997-1998.

Participant characteristics were assessed using analysis of variance for continuous variables and the χ² statistic for categorical variables. Because self-reported physical activity was not normally distributed, these data were expressed with medians and interquartile ranges and analyzed using the Kruskal-Wallis equality-of-populations rank test. Cox proportional hazard models were used to test the association between activity energy expenditure (activity energy expenditure and physical activity level) and all-cause mortality. We first tested for a continuous association by entering activity energy expenditure and physical activity level into the models as standardized units (per SD). To further evaluate linearity, we added a quadratic term (activity energy expenditure² or physical activity level²) to each model to test for curvilinear associations but these were removed because they were not statistically significant. Activity energy expenditure and physical activity level were also coded into tertiles with the lowest tertile serving as the reference group. Model 1 was estimated after adjusting for age, sex, race, and study site. Model 2 was adjusted for the factors in model 1 plus weight, height, percentage of body fat, and sleep duration. Model 3 was adjusted for the factors in model 2 plus self-reported health, education, smoking status and history, cardiovascular disease, lung disease, diabetes, hip or knee osteoarthritis, osteoporosis, cancer, and depression. The proportional hazards assumption was confirmed in all independent variables using Schoenfeld residuals.²⁵ Kaplan-Meier survival plots were analyzed using the log-rank test for trend, using the expected and observed survivor functions for each group over the entire follow-up time to calculate a test χ² statistic. Stata statistical software version 9.0 (StataCorp, College Station, Tex) was used for all analyses. Results were considered statistically significant at P≤.05.

Table 1 lists descriptive baseline characteristics for all participants and by tertile of activity energy expenditure. There were no differences in the distribution of age, race, educational level achieved, smoking status, self-reported fair or poor health, and total number of diseases between the activity energy expenditure tertiles. Prevalence of individual health conditions was similar across the activity energy expenditure tertiles. Women had lower levels of activity energy expenditure than men (mean [SD], 576 [251] kcal/d vs 769 [289] kcal/d; P<.01) but similar levels of physical activity (mean [SD], 1.68 [0.25] vs 1.72 [0.23]; P = .13). Persons in the highest tertile of activity energy expenditure were more likely to be from Pittsburgh and had a higher body weight, body mass index (calculated as weight in kilograms divided by height in meters squared), and lower percentage of body fat. Except for percentage of body fat (P = .56), all differences remained unchanged when adjusting for the sex imbalance across activity energy expenditure tertiles.

Over an average of 6.15 years (range, 0.21-7.53 years) of follow-up, 32 men (35.6 per 1000 person-years) and 23 women (23.9 per 1000 person-years) died with an overall cumulative mortality of 18.2%, which was similar to mortality patterns in the Health ABC cohort overall (43.2 per 1000 person-years for men and 27.3 per 1000 person-years for women).

The effects of activity energy expenditure and physical activity level as continuous variables with an added quadratic term were evaluated first. There was no evidence of a curvilinear association and therefore the quadratic term was removed from the model ([activity energy expenditure²], P = .95; [physical activity level²], P = .91 using model 1 covariates). When expressed as SD units in model 2 (Table 2), higher levels of activity energy expenditure and physical activity were associated with lower mortality risk (every 287 kcal/d for activity energy expenditure: hazard ratio [HR], 0.68 [95% confidence interval {CI}, 0.48-0.96]; every 0.24 for physical activity level: HR, 0.66 [95% CI, 0.47-0.93]). The estimates changed little when factors related to mortality risk were added to model 3.

Unadjusted and adjusted Kaplan-Meier survival estimates for activity energy expenditure and physical activity level stratified by tertiles are plotted in the Figure. Cox proportional hazard modeling was used to adjust for potential confounders (Table 2).There was no activity energy expenditure or physical activity level tertile interaction by sex (P = .58 and P = .78, respectively) or race (P = .61 and P = .48, respectively). Compared with the lowest activity energy expenditure tertile, those in the highest tertile had a lower risk of mortality (model 2: HR, 0.31; 95% CI, 0.14-0.69). Similar results were observed with physical activity level (HR, 0.40; 95% CI, 0.19-0.81). The absolute risk of mortality was 12.1% in the highest tertile of activity energy expenditure, 17.6% in the middle, and 24.7% in the lowest tertile. For physical activity level, risk of mortality was 12.0% in the highest tertile, 17.8% in the middle, and 24.7% in the lowest. The effects changed little after adjusting for smoking status, educational level, self-rated health, and prevalent health conditions.

Self-reported physical activity was assessed to determine whether individuals with higher levels of free-living activity energy expenditure reported more physical activity (Table 3). The proportion of participants reporting high-intensity exercise, walking for exercise, other walking, volunteering, or caregiving did not differ across the activity energy expenditure tertiles. However, participants with higher free-living activity energy expenditure were significantly more likely to report climbing stairs and working for pay. Additionally, individuals with higher free-living activity energy expenditure self-reported longer duration and expended a greater number of kilocalories in total activity.

Several sensitivity analyses were performed using standardized units of activity energy expenditure (per SD of activity energy expenditure using model 3 covariates). After excluding individuals who died within the first year of follow-up, the strength and direction of the effect of activity energy expenditure on mortality was comparable with the original associations shown in Table 2 (after subtracting 4 deaths: HR, 0.75; 95% CI, 0.52-1.08). To test whether the effects were influenced by extreme activity levels, individuals with low-activity energy expenditure (20th percentile, after subtracting 15 deaths: HR, 0.79; 95% CI, 0.51-1.22) and high-activity energy expenditure (80th percentile, after subtracting 10 deaths: HR, 0.57; 95% CI, 0.31-0.97) were removed the analysis. Finally, data from individuals with 2 or more prevalent health conditions were analyzed to determine whether the effects remained in older adults with comorbidities (after subtracting 22 deaths: HR, 0.73; 95% CI, 0.47-1.13). Despite losing statistical power, results from these sensitivity analyses were consistent with the results in Table 2.

Higher free-living activity energy expenditure demonstrated a strong association with lower risk of mortality among older adults. The major strength of this study is the direct measurement of both total energy expenditure and resting metabolic rate resulting in an objective evaluation of free-living activity energy expenditure in a biracial sample of older adults.

The association of physical activity and mortality has been studied more often in younger adults than in older adults (>70 years). Investigation of modifiable risk factors for disease or disability in older adults is important for this growing segment of the population that contributes disproportionately to health care costs.²⁶ The few longitudinal studies of older adults that use self-reported physical activity questionnaires suggest a substantial protective effect against premature mortality (HR range, 0.50-0.80) when comparing the highest with the lowest activity group.^1- 3,27 For example, Gregg et al²⁷ showed that women aged 70 years or older experienced a 32% lower risk of mortality (HR, 0.68; 95% CI, 0.59-0.78) by being in the highest quintile of total activity. Although these studies have shaped our current understanding of whether physical activity is associated with lower mortality risks, they typically quantified general activities such as walking,² vigorous exercise,²⁸ or regularity of activity,¹ which ignore large components of energy expenditure such as usual daily activity.²⁹ The protective effects shown in Table 2, using the gold standard method for measuring activity energy expenditure, appear stronger than those derived from self-reported physical activity but are comparable with other objective measures of physical fitness.^30- 31 This likely reflects a more accurate activity assessment with the doubly labeled water method, which avoids known misclassification biases that occur with self-reported physical activity questionnaires.⁹ Therefore, this study suggests that earlier reports may have underestimated the benefits of physical activity among older adults, although this implication should be tested in future studies.

Our study suggests that any activity energy expenditure in older adults can help lower mortality risks, seemingly contradicting reports that exercise needs to be performed at a specific intensity.³² Unfortunately, the method of ascertaining free-living activity energy expenditure in this study does not provide guidance on the intensity or type of activity that may be important for public health recommendations. Results from the physical activity questionnaire suggest that the proportion of individuals who reported high-intensity exercise and walking for exercise (both in terms of duration and intensity) was similar across tertiles of free-living activity energy expenditure. Interestingly, individuals who worked for pay and climbed stairs were more likely to be in the high-activity energy expenditure tertiles. Another way to apply this information is to use the well-established metabolic equivalent values for physical activity. We found that for every 287 kcal/d in free-living activity energy expenditure, there is approximately a 30% lower risk of mortality. Using the average body weight in our sample of 76 kg and a metabolic equivalent value of 3.0, we estimated that individuals who perform 1 1/4 hours of activity per day will expend 287 kcal/d. Activities that meet a metabolic equivalent value of 3.0 include household chores (vacuuming the carpet, mopping the floor, washing windows, etc), child/adult care, lawn work, walking at a pace of 2.5 mph, and nonsitting work or volunteering. In support of this calculation, the total self-reported activity duration was approximately 30 and 60 minutes longer on average in the second and third tertiles of activity energy expenditure, respectively, compared with the lowest tertile of activity energy expenditure. Most importantly, this accumulation is from usual daily activities that expend energy and not necessarily from volitional exercise.

An unavoidable limitation in observational studies of physical activity is that unhealthy individuals who are at high risk of mortality are also less likely to be active. This bias is less likely in the current study because investigators from the Health ABC study specifically recruited high-functioning older adults who did not have mobility disabilities and thus would likely be capable of physical activity. Furthermore, adjustments for prevalent health conditions and self-reported health helped reduce this potential bias. Individual diseases as well as the total number of health conditions were not significantly different across the energy expenditure tertiles and the proportion of those self-reporting fair or poor health were slightly but not significantly worse in the lowest tertile of energy expenditure (Table 1). Finally, we also performed several sensitivity analyses to determine whether the results were influenced by extreme activity levels and comorbidity. The reduction in sample size caused some of these strategies to yield nonsignificant results but overall patterns suggest that higher levels of activity energy expenditure are beneficial for individuals with comorbidity and who have low to moderate activity levels.

Higher levels of physical activity are associated with reductions in coronary heart disease,³³ cancer incidence,²⁷ falls,³⁴ and physical disability.³⁵ The exact underlying mechanisms conferring protection are as yet unknown but these are likely to differ between younger and older adults. For example, biological aging is thought to be associated with increased oxidative stress, which contributes to higher levels of inflammation,³⁶ both of which are reduced with exercise.^37- 38 More research is needed to elucidate underlying mechanisms of how activity energy expenditure can protect older adults from premature mortality.

Free-living activity energy expenditure is influenced by body weight, age, sex, and sleep duration.¹² Those individuals who weigh more and are performing the same activity as their counterparts who weigh less expend more kilocalories.³⁹ Physical activity level is total energy expenditure as a function of resting metabolic rate, the latter being solely influenced by lean mass. This expression accounts for differences in body weight, which also corrects for sex differences in average energy costs (men vs women in our sample, 1.72 vs 1.68; P = .13).²² Nonetheless, we continued to statistically adjust for both weight and sex to account for their influence on mortality risk. In terms of other confounders, activity energy expenditure and physical activity level are equally affected by sleep duration because those who sleep less have a greater opportunity to expend more kilocalories through activity. Adjustment for sex, age, body weight, and sleep duration revealed a stronger effect of higher levels of activity energy expenditure, suggesting that controlling for these confounders refined the association.

Both activity energy expenditure and physical activity level decrease with age,^14,22 which prevented us from using the categories created by the Food and Agriculture Organization, the World Health Organization, and the United Nations University that do not recognize age-related changes in physical activity level. For example, using these established categories, 57% of our sample were sedentary, 33% were active, and only 10% were vigorously active. Additionally, the established categories revealed no benefits of energy expenditure in the moderate range. Therefore, further work to establish appropriate physical activity level categories suitable for older adults is needed.

Because the doubly labeled water method directly measures carbon dioxide production over an extended period of normal activity, it is considered the most accurate estimate of free-living activity energy expenditure. It is, however, expensive on an individual basis and requires special expertise limiting investigation to smaller sample sizes. Nonetheless, the strong protective effect observed with 302 participants demonstrates that a more accurate measurement substantially reduced the sample size needed to detect meaningful effects compared with what would have been required with a simply performed but less precise questionnaire assessment. This relatively small sample size limited our ability to assess cause-specific mortality but additional events will eventually allow examination of specific diseases and investigation into a potential threshold of activity energy expenditure that mediates these mortality associations.

In conclusion, we evaluated mortality risk with an accurate and objective measure of free-living activity energy expenditure. Free-living activity energy expenditure revealed a strong association with mortality risk suggesting that previous self-reported measurements may have underestimated the benefits of higher levels of physical activity in older adults. Efforts to increase or maintain free-living activity energy expenditure will likely improve the health of older adults.