点击显示 收起
1 From the US Department of Agriculture, Agricultural Research Service, Beltsville Human Nutrition Research Center, Diet and Human Performance Laboratory, Beltsville, MD (JMC and JLS); the University of Minnesota, Division of Epidemiology, School of Public Health, Minneapolis (DRJ Jr); the Fred Hutchinson Cancer Research Center, Cancer Prevention Research Program, Seattle (MLI); and the University of South Carolina, School of Public Health, Departments of Epidemiology and Biostatistics and Exercise Science, Columbia (BEA).
2 Mention of a trademark or proprietary product or service does not indicate guarantee or warranty of the product or service by the US Department of Agriculture and does not imply its approval to the exclusion of other products or services that may also be suitable. 3 Address reprint requests to JM Conway, US Department of Agriculture/Agricultural Research Service/Beltsville Human Nutrition Research Center, Diet and Human Performance Laboratory Building 308, Room 122, Beltsville, MD 20705. E-mail: conway{at}bhnrc.arsusda.gov.
ABSTRACT
Background: Various methods are used by epidemiologists to estimate the energy cost of physical activity; these include physical activity records and recalls. However, there is limited validation of these methods against the doubly labeled water technique for determining energy expenditure (EE).
Objective: We compared EE as estimated by indirect methods (physical activity records and recall questionnaires) used in epidemiologic studies with EE obtained from doubly labeled water (EEDLW) in free-living men.
Design: We determined EEDLW, energy intake at weight maintenance, and EE from 7-d physical activity records (EERecord) and a 7-d physical activity recall questionnaire (EERecall) in 24 men aged 41 ± 2.0 y (
Key Words: Exercise energy intake Stanford 7-d physical activity questionnaire physical activity records physical activity recall doubly labeled water basal metabolic rate men
INTRODUCTION
Disease prevention strategies currently include recommendations for both eating a healthy diet and engaging in physical activity (1). Although many methods for assessing physical activity in free-living individuals have been proposed and directly or indirectly validated (24), energy expended during physical activity has proven difficult to measure. According to the FAO/WHO/UNU (5), physical activity is considered as including occupational activities, discretionary activities, optional household tasks, socially desirable activities, and activity for physical fitness and the promotion of health. This great variety of activities and the ability of humans to change their activity at will contribute to the difficulty in measuring physical activity in free-living humans.
Before designing and implementing nutrition education and disease prevention strategies, it is important to test the validity of epidemiologic methods that are currently used for estimating energy expenditure (EE), ie, physical activity questionnaires and records (69), against a criterion method. Although epidemiologic surveys can provide information about populations, they are conducted with less precision than what is required for metabolic studies of individuals.
Many data show that free-living EE can be measured under laboratory (3, 1012) and field (13) conditions by use of deuterium- and 18O-labeled water (2H218O). However, few epidemiologic methods for estimating EE from physical activity records and recalls have been cross-validated simultaneously against the doubly labeled water method under well-controlled conditions (1416). In the present study, we sought to provide independent estimates of EE from 7-d physical activity records (EERecord) and a 7-d physical activity recall (EERecall; with use of the Stanford 7-d physical activity questionnaire), measures of energy intake (EI) at weight maintenance, and measures of doubly labeled water (EEDLW). Our null hypothesis was that there would be no significant difference between EE estimated by the 2 previously validated physical activity methods, ie, the 7-d physical activity record (17) and recall (7), and EEDLW measurements.
SUBJECTS AND METHODS
Subjects
Twenty-seven men were recruited for participation in this study from a larger ongoing feeding study. One subject dropped out before completing the study and 2 subjects were eliminated because of questionable compliance with the dietary portion of the study. As a result, a total of 24 subjects aged 2765 y participated in the study. Subjects were originally recruited by advertisement at the Beltsville Agricultural Research Center in Beltsville, MD; at the Goddard Space Flight Center in Greenbelt, MD; and from the laboratory's computerized database of persons known to be interested in participating in human studies.
The study protocol was approved by the Institutional Review Board at the Johns Hopkins University School of Medicine and by the Human Studies Committee of the US Department of Agriculture's Agricultural Research Service. Subjects were invited to attend an informational meeting, and those interested in participating in the study provided written, informed consent. At the beginning of the original aforementioned feeding study, each potential subject received a medical evaluation by a physician, which included the measurement of blood pressure, height, and weight and analysis of fasting blood and urine samples to screen for the absence of metabolic diseases. The present study was conducted at the US Department of Agriculture/Agricultural Research Service/Beltsville Human Nutrition Research Center in Beltsville, MD.
Experimental design
Each subject was studied over a 2-wk period. On day 1, the doubly labeled water (2H218O) was administered to subjects and urine was collected over the next 14 d. Physical activity records were kept during either the first or second week of the 2H218O protocol and whole-body calorimetry was performed during the opposite week. The 7-d physical activity recalls were administered on day 7 and again on day 14 of the 2H218O protocol.
Body composition
Weight was measured to the nearest 0.01 kg by use of an electronic balance (August Sauter, Ebingen, Germany) and height was measured to the nearest 0.1 cm by use of a stadiometer (Holtain Limited, Crymych, United Kingdom). Body mass index (BMI) was defined as weight divided by height squared (kg/m2). Percentage of body fat and lean body mass were determined by dual-energy X-ray absorptiometry (version 1.3, DPXL; Lunar Corporation, Madison, WI). The subjects were asked to not consume anything 3 h before dual-energy X-ray absorptiometry, to dress in metal-free clothing, and to remove all jewelry. The components of total body mass, ie, fat, soft tissue (lean body mass), and bone mineral, were used to calculate percentage body fat. Percentage of body fat was determined as follows:
RESULTS
There was a broad range in body weight, BMI, and percentage of body fat in the study population (Table 1). Although some of the subjects were overweight, most were within 130% of ideal body weight on the basis of the 1959 tables of the Metropolitan Life Insurance Company (24).
View this table:
TABLE 1 . Characteristics of the study population1
Measured BMR and comparisons among EEDLW, EI, EERecord, and EERecall are shown in Table 2. EEBMR ranged from 5.96 to 8.85 MJ/d. EI was in agreement with EEDLW with <0.5% underestimation. This difference was not significantly different from zero.
View this table:
TABLE 2 . Basal metabolic rate (EEBMR) and energy expenditure (EE) as determined by the doubly labeled water method (EEDLW), by physical activity records (EERecord), and by a physical activity recall (EERecall) and energy intake (EI)
Seven-day physical activity records overestimated free-living EEDLW by a mean of 7.9%, which was significantly different from zero. The difference between EEDLW and EERecord was
Graphic comparisons of EE, as determined by the different methods (EEDLW, EI, EERecord, or EERecall), are shown in Figures 13. Because the 7-d recall data from one of the subjects was >42 MJ/d, it was difficult to include these data on the y axis without distorting the graph in Figure 3; therefore, these data were omitted from the Bland-Altman plots but not from the remaining data analysis. Note that the y scale in Figure 3 is 10 MJ/d, so that differences between EEDLW and EERecall are visually reduced compared with the same level of difference in Figure 2, which compares EEDLW and EERecord. The Bland-Altman plot in Figure 3 shows a negative trend, reflecting that a major overestimation by use of EERecall occurred in several individuals.
FIGURE 1. . Difference between energy expenditure (EE) measured by the doubly labeled water method (EEDLW) and energy intake (EI) plotted against the mean of the 2 measurements according to Bland and Altman (22, 23). A negative sign indicates an overestimation and a positive sign indicates an underestimation by EI.
FIGURE 2. . Difference between energy expenditure (EE) measured by the doubly labeled water method (EEDLW) and by physical activity records (EERecord) plotted against the mean of the 2 measurements according to Bland and Altman (22, 23). A negative sign indicates an overestimation and a positive sign indicates an underestimation by EERecord.
FIGURE 3. . Difference between energy expenditure (EE) measured by the doubly labeled water method (EEDLW) and by physical activity recall (EERecall) plotted against the mean of the 2 measurements according to Bland and Altman (22, 23). A negative sign indicates an overestimation and a positive sign indicates an underestimation by EERecall.
DISCUSSION
Because of the importance of routine physical activity and leisure-time exercise in the prevention of disease and the maintenance of health, epidemiologists have developed methods to quantitate physical activity at population levels (2). The present study is one of few (1416) to test simultaneously the ability of different epidemiologic methods to estimate EE in free-living individuals and the first to test such methods under controlled feeding conditions. The 14-d excretion rates of 2H and 18O after an oral dose of 2H218O were used to calculate the free-living EE and served as the criterion method for this investigation. The small difference between mean EEDLW and EERecord (Table 2) in this study supports the advantage of physical activity records for estimating EE in a population. Although this study was conducted under carefully controlled conditions, the small size of the study population may have been a limiting factor and may explain the lack of significance when EERecord was regressed against EEDLW. However, in 10 of the subjects the difference between EEDLW and EERecord was <10%, indicating that physical activity records may be useful in some individuals for estimating free-living EE (Figure 2).
Physical activity level of population
The predicted physical activity level factors for light, moderate, and heavy activity were 1.56, 1.78, and 2.10 x BMR respectively (21). We observed a range in EEDLW/BMR almost identical to the predicted values of 1.582.05, indicating that our population consisted of men who ranged in activity from sedentary to heavy activity.
Energy expenditure from energy intake at weight maintenance
Human feeding studies have been conducted for >20 y at the Beltsville Human Nutrition Research Center (25). Long-term feeding studies are frequently conducted with subjects who are free to go about their typical daily activities while consuming a diet provided for them under a specified protocol. These studies differ from metabolic studies that are conducted while the participants are confined to a metabolic ward and from studies in which the participants consume food they procure and maintain records of food intake. During this study, the technique for determining EI at weight maintenance included weighing the subjects once per week. When a subject's weight changed by >2 kg, EI was adjusted by 837 kJ. This weight range was chosen because we had previously observed a variation in body weight of 2 kg when scientific personnel, who had excellent compliance, were fed a constant diet for 14 d (unpublished observations, JM Conway, 1994). Metabolizable energy intake was calculated from food tables (18) and the amount of energy consumed to maintain body weight was then hypothesized to be equal to the person's free-living EE or energy requirement.
Previous studies used EI as a prediction of both the energy cost of physical activity (26) and the energy requirements (27). In the present study, EI at weight maintenance was determined from the EI fed during the 2-wk study period, although the subjects had been fed a controlled dietary intake for >12 wk. As seen in Figure 1, the EI in only one subject was greater than 1 SD of the mean difference between EI and EEDLW. Because this subject had slowly gained weight, his EI was adjusted after the 14-d study period. Had we averaged EI and EEDLW over a longer period of time, the agreement between EI and EEDLW for this subject would have been <0.5 MJ/d. Nevertheless, it is remarkable that the maximum difference between EI and EEDLW was 1.5 MJ/d.
Energy expenditure from physical activity
Physical activity record
The uniformity in treatment of the physical activity records would tend to minimize the role of instruction and analysis in the error. In the present study, the observation that one-half of the 24 subjects had EERecord values within 11.2% of the mean EE determined by doubly labeled water strongly suggests that the MET values obtained from the Compendium of Physical Activities (7) for activities performed by the subjects were reasonable estimates of true energy cost. The error in determining the energy cost of physical activity from physical activity records in the other 12 subjects may be attributed to factors such as misreporting, body weight, degree of overweight, and environment (28). Earlier reports (14, 29) found that the MET intensities listed in the Compendium of Physical Activities (7) for overweight individuals may be inaccurate and that the inaccuracies may be different for weight- and non-weight-bearing activities. Estimation of EE from the frequency, duration, and type of activities recorded in the physical activity record may contribute errors, which could explain the observed R2 of 0.10 between EEDLW and EERecord. The high level of agreement between the 7-d physical activity records and the 14-d assessment made by the doubly labeled water methods partly reflects the overlapping time intervals; the present study did not assess how closely 7-d physical activity records reflect usual activity.
Both published reports on the Compendium of Physical Activities (7, 30) provide discussions on the recommendations for use and the limitations of the Compendium. However, it is worth restating here that the Compendium of Physical Activities was not developed to estimate the precise energy cost of physical activity for individuals, but rather to provide an activity classification system that standardizes the MET intensities for use in epidemiologic research. Furthermore, "this limits the use of the Compendium in estimating the energy cost of physical activity among individuals in ways that account for differences in body mass, adiposity, age, sex, efficiency of movement, geographic and environmental conditions in which the activities are performed. Thus, individual differences in EE for the same activity can be large and the true energy cost for a person may or may not be close to the stated mean MET level as presented in the Compendium" (30).
The present study also supports the commonly held belief that subject compliance is a key limiting factor in the use of physical activity records. For example, the subject in Figure 2, who had the highest EEDLW of 17.22MJ/d and one of the higher differences (19.1%) between EEDLW and EERecord commented that he was so active it was hard for him to stop and record his activity.
Seven-day physical activity recall surveys
The mean difference between EEDLW and EERecall was much higher and had greater individual variation (Table 2) than that between EEDLW and EERecord or EI. When using the 7-d recall, physical activity was overestimated by >20% by 10 subjects and by >10% and <20% by 7 subjects, whereas the recall estimations of EE in 7 additional subjects were within 10% of EEDLW. The largest overestimations of the time spent in hard- and very-hard-intensity physical activity were made by one subject (204%) whose occupation was to detail cars with an 18-kg (40-lb) machine to buff the vehicles and one subject (116%) who was a carpenter. These data indicate that most subjects overestimated their physical activity when left to their own perceptions of level of exertion, ie, moderate, hard, and very hard. The level of overestimation could be lessened by scoring the 7-d recall differently, namely, by awarding a lower intensity than 6 and 10 METs/min to hard- and very-hard-intensity activity. Taken together, the general overestimation of EE by the 7-d recalls (Figure 3) and the mean ratio of EERecall/BMR of 2.36 indicate that the recall method has limited usefulness in the estimation of individual EE and is also limited in small groups. These results are mirror images to those of dietary recalls, which significantly underestimate EI (25).
By using basic principles of physiology and energetics, Black et al (31, 32) calculated EI cutoffs "below which a person of a given sex, age and body weight could not live a normal lifestyle." The minimal plausible level of habitual EI was reported as 1.35 x BMR. Because the physiologic value of EI and EE are identical for an individual in energy balance, this value of 1.35 x BMR can serve as a validity cutoff for any estimate of EE obtained from physical activity records or recalls. There were a few individuals who underestimated their physical activity when using the 7-d physical activity records and 7-d recall, as evidenced by EERecord/BMR ratios as low as 1.42 and EERecall /BMR ratios as low as 1.35 (Table 3).
View this table:
TABLE 3 . Ratios of energy expenditure (EE) as determined by the doubly labeled water method (EEDLW), by energy intake (EI), by physical activity records (EERecord), and by a physical activity recall (EERecall) to basal metabolic rate (BMR)1
More subjects underreported EE when using the 7-d physical activity record than when using the 7-d recall. Seven subjects underreported EE when using the physical activity records, whereas only 3 subjects underreported EE when using the 7-d recalls. These 3 subjects underreported their physical activity when using both methods, suggesting an overall difficulty in reporting physical activity.
In this study, the 7-d recall was a less reliable instrument for estimating EE than were the physical activity records. The large intraindividual variation may limit the usefulness of the 7-d recall in small populations. This serves as additional evidence for the difficulty in recalling physical activity from questionnaires and using those data to estimate the EE of human movement (28).
Conclusions
Estimating the true energy cost of physical activity remains one of the unsolved problems of nutritionists, exercise physiologists, and epidemiologists. Although doubly labeled water is considered a criterion method for estimating EE in free-living persons, the cost of isotopes and analyses and the requirement for an isotope ratio mass spectrometer prohibits 2H218O from being widely used in studies of large populations (26). Therefore, it is of considerable importance that physical activity records can be used in individuals to predict free-living EE. Physical activity records have the unique advantage of providing additional information on the types of activity and time devoted by individuals to specific activities.
Our null hypothesis was that there would be no significant difference between EE estimated by 2 previously validated physical activity methods, ie, the 7-d physical activity recall and the 7-d physical activity record, and the doubly labeled water method. The present study indicates that 7-d physical activity records may estimate the mean EE of population, but that the 7-d recall method has limited application because it was both biased and imprecise. The use of these methods to predict individual EE depends largely on the compliance of the population being studied and the ability of the subjects to correctly estimate time spent in activities of differing intensities.
ACKNOWLEDGMENTS
We thank Demetria Fletcher and Robert Staples for their technical assistance and the participants for their cooperation.
REFERENCES