The well-known association between physical activity and human health (24) motivates attempts to measure the amount of physical activity. We and others in nutrition and exercise physiology are particularly interested in calculating total energy expenditure (TEE) from activity records; however, two subjects with differing body weight performing the same physical activities will expend different amounts of energy. Therefore, it is necessary to adjust for body size. This is usually done by using the MET system (1,2), which classifies physical activity by intensity, where intensity is expressed as the ratio of the energy expenditure when performing specific activities to the resting metabolic rate (RMR) of the subject. This ratio is referred to as the MET value, and this system makes it possible to calculate the energy expenditure of a subject by multiplying his or her RMR by the MET values for the specific activities performed. Moreover, when using the MET system, it is often assumed that the RMR of human subjects is approximately 4.184 kJ·kg−1 body weight·h−1 (1). The MET system was developed using young, lean, primarily male subjects in whom RMR is actually close to 4.184 kJ·kg−1 body weight·h−1. However, because RMR is primarily a function of fat-free mass (16), RMR expressed per kilogram of body weight will decrease as the percent total body fat (% TBF) of subjects increases. This creates a dilemma: does one use the measured RMR, calculated RMR, or some combination of the two when calculating the energy expended in physical activity? In this regard, several authors (5,8,10,13,18,22) have noted that the MET system underestimates energy expenditure in heavy or obese subjects when measured RMR is used in the calculation. Racette et al. (18) went further and proposed a modified way of calculating energy expenditure of obese individuals using measured RMR plus (MET-1) times 4.184 kJ·kg−1·h−1. This issue is becoming increasingly important considering that the frequency of overweight and obesity is increasing worldwide, indicating a need for studies of the role of TBF content regarding the accuracy of the MET system, a topic that has so far been insufficiently investigated.
We investigated the following hypothesis: When % TBF increases, TEE will become increasingly inaccurate when measured basal metabolic rate (BMR) is used along with the MET system to calculate TEE from activity data. The hypothesis that the proposed calculation of Racette et al. (18) is more accurate than the standard calculation was also investigated. These hypotheses were tested using data obtained in two populations of women before and after weight gain using heart rate recording to determine the time spent in activities with varying intensities. In one of these populations (American women), weight gain occurred during a 1-yr period following voluntary weight loss. In the other (Swedish women), weight gain was a consequence of pregnancy. In both populations, before as well as after weight gain, heart rate was recorded for at least 3 d. These recordings took place during time periods when reference estimates of the TEE were obtained using the doubly labeled water method. These two populations were chosen because the weight gain in the American women was primarily fat mass, whereas that in the Swedish women contained relatively more fat-free mass. Thus, weight gain in these two populations would be expected to have different effects on BMR.
MATERIALS AND METHODS
Data for women participating in a study of pattern and cost of weight gain previously described by Votruba et al. (23) were used. These authors recruited women with a recent weight loss of at least 12 kg. Entry criteria were weight stability within 1 kg for at least 1 month, but not for more than 3 months, and a body mass index (BMI) between 20 and 30 kg·m−2. The protocol included assessments of body weight, BMR, body composition, TEE, and physical activity measured by heart rate recording at entry and 12 months later. Subjects were encouraged to use a weight maintenance strategy of their own preference between the two measurements. The present paper is based on data for 11 women (age 38 ± 7 yr at entry) who were selected from a larger cohort because they had gained body weight between the two measurements (3.8-40 kg) and had satisfactory heart rate recordings before as well as after weight gain. Details about these recordings are given by Schoeller et al. (20). In short, monitors (Polar Vantage, Stamford, CT) were worn during most waking hours for 3 d. Individual threshold values for activity levels were determined the week before the experiment by having the subjects walk at various controlled speeds. The threshold between light and moderate activity was the subject`s heart rate while walking at a rate of 3.7 km·h−1 (2.2 mph). The threshold between moderate and hard activity was the subject`s heart rate while walking at a rate of 6.3 km·h−1 (3.7 mph). The threshold between hard and very hard activity was calculated as 70% of maximum heart rate adjusted for age. A questionnaire was used to assess amount of sleep. Unrecorded time was considered to be spent in light activity. Thus, all time during the heart recording periods was divided into activity categories with the following MET values: sleep, 0.9; light activity, 1.3; moderate activity, 2.8; hard activity, 4.5; and very hard activity, 8.0. The study was reviewed and approved by the internal review board of the University of Chicago, and informed written consent was obtained from each volunteer (20).
Data for 15 out of 23 women participating in a study of energy metabolism before (13) and during (15) pregnancy were used in the present investigation. The other eight women were excluded because their heart rate recordings were incomplete. Healthy women planning pregnancy were enrolled in a study where body weight, BMR, body composition, TEE, and physical activity (measured by heart rate recording) were assessed before conception and in gestational week 32. The women were 30 ± 4 yr at the prepregnancy measurement, gained 3.4-17.7 kg of body weight between this measurement and gestational week 32, remained healthy throughout pregnancy, carried singleton fetuses, and delivered healthy infants. Details about the heart rate recording are given by Löf et al. (13). In short, monitors (Polar vantage NV, Polar Sverige AB, Stockholm, Sweden) were worn for 6 or 7 d during most waking hours. An average of 6.8 and 6.6 d were recorded before and during gestation, respectively. The recorded heart rates were converted to MET values using individual relationships, established shortly before the experiment, by measuring heart rate when performing activities with defined MET values as previously described (13). During the heart rate recording periods, the subject used a notebook to record activities carried out when the monitor was removed. Amount of sleep was estimated using this book. Unrecorded time in the waking state was assigned a MET value of 1.4 (13). The study was approved by the ethics committee at the University of Linköping.
BMR was measured in all subjects by indirect calorimetry as previously described (13,23). Briefly, after an overnight fast the respiratory gas exchange was measured using a Deltatrac Metabolic Monitor (Datex Instrumentarium Corp, Helsinki, Finland) while the women lay comfortably in a quiet, thermoneutral environment. The American women had slept in the general clinical research unit overnight while the Swedish women arrived at the hospital by car in the morning and rested for 45 min before the measurement.
Body composition and reference estimates of TEE.
As previously described (13,23), the doubly labeled water method (19) was used to assess TEE and total body water. The women were dosed with doubly labeled water using approximately 0.05 g of 2H2O (99.9%) and 1.5 g of H2 18O (10%) per kilogram of body weight (13,23). Urine samples were collected before and for 2 wk after dosing. Stable isotope enrichments were measured by isotope ratio mass spectrometry, and TEE was calculated from CO2 production assuming the food quotient to be 0.85 (3). Estimates of total body water were used to calculate TBF assuming the following water content of fat-free mass: 73.2% (American women), 71.8% (Swedish women before pregnancy) (14), and 74.7% (Swedish women in gestational week 32) (14).
Standard calculation of TEE.
Each MET value was multiplied by the measured BMR for the period of time when activities with this particular MET value were performed [BMR (kJ·h−1) · time (h) · MET]. Energy expenditure values calculated in this way, and representing time periods with all the different MET values used during each heart rate recording period, were then added together.
Proposed calculation of TEE.
For MET values ≤ 1, energy expenditure was calculated by multiplication of each MET value by the measured BMR for the period of time when activities with this particular MET value were performed [BMR (kJ·h−1) · time (h) · MET]. For MET values >1, BMR for each time period during which activities with such a MET value were performed was calculated and the result of the following calculation was added to this value: 4.184 kJ times the body weight of the subject times the time period during which activities with the particular MET value were performed times this MET value minus 1 [BMR (kJ·h−1). time (h) + 4.184 (kJ·kg−1·h−1) · time (h) · body weight (kg) · (MET-1)]. Energy expenditure values thus calculated, and representing time periods with all the different MET values used during each heart rate recording period, were then added together (18).
Values given are means ± standard deviations. Significant differences between mean values were identified using t-test for paired observations or by repeated ANOVA with subsequent post hoc analysis using Tukey`s multiple comparison test (9). All statistical analyses were done using STATISTICA software (version 6.0; StatSoft, Scandinavia AB, Uppsala, Sweden). Agreement between results obtained using different methods and the appropriate reference results was examined according to Bland and Altman (4).
Results for body weight and composition as well as for energy expenditure are shown in Table 1. During the 12-month period, the American women gained 13.9 ± 10.4 kg body weight (P < 0.01), 1.9 ± 1.8 kg fat-free mass, and 12.1 ± 9.0 kg TBF (P < 0.01). During this time their BMI increased (P < 0.001) from 24.2 ± 3.0 to 28.8 ± 4.7 kg·m−2, BMR by 0.35 ± 0.69 MJ·24 h−1, and reference estimates of their TEE by 0.84 ± 2.28 MJ·24 h−1. When expressed per kilogram of body weight, a significant decrease in BMR was found. TEE did not differ significantly from the reference estimates of TEE either before or after weight gain using either the proposed or the standard calculations. For the Swedish women, BMI before pregnancy was 24.4 ± 5.4 kg·m−2. Between the measurements before and during pregnancy, their body weight increased by 11.0 ± 4.8 kg (P < 0.001), TBF by 4.1 ± 3.8 kg (P < 0.001), and fat-free mass by 6.8 ± 1.9 kg (P < 0.001). Body fatness was very similar before pregnancy and in gestational week 32, at 32.1 and 32.8% TBF, respectively. Between these two measurements, BMR increased significantly by 1.24 ± 0.57 MJ·24 h−1, and this increase was also significant when expressed per kilogram of fat-free mass and per kilogram of body weight. The reference estimates of TEE increased significantly by 1.14 ± 1.35 MJ·24 h−1 between the prepregnant measurement and gestational week 32. On both these occasions, TEE obtained using the standard calculation was significantly lower than reference estimates of TEE. No such significant difference was obtained for TEE when the proposed calculation was used. As expected, the proposed calculation resulted in higher energy expenditure than the standard calculation both before and after weight gain in American as well as in Swedish women.
As shown in Figure 1, a significant (r = −0.65, P = 0.0004) linear relationship (y = 1.04 − 0.0039x) was found between % TBF (x) and the ratio TEE by the standard calculation/TEE by the proposed calculation (y) when including Swedish women before pregnancy and American women after weight gain (N = 26, BMI = 18-39) in the analysis. The point where this ratio equals 1 corresponds to 10.7% TBF. For women in gestational week 32, the standard/the proposed calculation was 0.94 ± 0.04 (N = 15). This is slightly higher than the ratio value of 0.91 that was observed for nonpregnant women with the corresponding body fat content (32.8% TBF), obtained using the regression equation given above. For these same 26 women (Swedish women before pregnancy and American women after weight gain), a corresponding analysis between BMI (x) and the ratio for TEE calculated by the standard calculation/the proposed calculation (y) yields the following regression equation: y = 1.06 − 0.0058x, r = −0.70, P = 0.00008.
When results obtained using the proposed calculation were compared with reference estimates of TEE according to Bland and Altman (4), no significant correlations for the difference between the two methods versus their average were obtained for either the American or for the Swedish women at any of the measurements. Likewise, no such significant correlations were obtained for results when using the standard calculation. Table 2 shows how the results obtained using these two calculations differ from reference estimates of TEE before and after weight gain in the American women, as well as before pregnancy and in gestational week 32 in the Swedish women. The table shows the average deviations from the reference estimates with their 95% confidence limits and the limits of agreement (± 2 SD) as suggested by Bland and Altman (4). Obviously, the average deviation from the reference value was larger for the standard calculation than for the proposed calculation in all cases except in American women before weight gain. It should also be noted that the 95% confidence limits were wider for the American than for the Swedish women. Finally, Table 2 shows that the limits of agreement for the proposed calculation were as wide as, or wider than, those obtained for the standard calculation.
Our results show that for women having a TBF content comparable with that of contemporary female populations, standard calculations of TEE will be lower than those obtained by means of the proposed calculation. This difference will increase as the TBF content of the subjects increases. Furthermore, our results indicate that for contemporary Western women, the proposed calculation is more accurate than the standard calculation, but as discussed below, this finding needs to be confirmed in future studies.
To obtain the results presented in Figure 1, we combined data obtained in American women after weight gain with data obtained in Swedish women before pregnancy. In this way, we were able to cover a broad range of %TBF, 19-51%, corresponding to a BMI between 18 and 39, a range that is likely to be quite typical for women in many Western populations (International Obesity Task Force, http://www.iotf.org). Using these data we identified a significant linear relationship between % TBF and TEE by the standard calculation/TEE by the proposed calculation, which indicates that these two calculations produce identical results at a TBF content of about 10%, a likely approximate average TBF content in the subjects used to develop the MET system. The conventional MET system is apparently associated with a bias in the great majority of Western women. Furthermore, as indicated by recent results (11) presenting the TBF content of a randomly selected population sample, this conclusion is likely to be true for men as well. In this context it is relevant to note that TBF was calculated using slightly different figures in the Swedish and American women regarding the water content of fat-free mass. These differences in hydration values were, however, too small to influence the results and conclusions presented above.
Human energy expenditure is mainly a function of body size and physical activity. However, the thermic effect of food also influences energy expenditure, although to a rather small extent. This effect is generally disregarded when the MET system is applied. This is probably an acceptable simplification because the original experiments carried out when establishing the MET values were conducted in fed subjects (17). Furthermore, it is also recommended that the MET system be based on estimates of RMR rather than BMR. However, estimates of RMR are sometimes, but not necessarily (16), obtained in subjects who are in the fed state. In addition, the current guidelines from the Food and Agricultural Organization (6) for calculating TEE of human adults are based on estimates of BMR. In the present study our calculations were performed using BMR rather than RMR. Because BMR often is slightly lower than RMR (16), our calculated TEE may have been slightly low. However, this effect is certainly too small to influence the general conclusions of our study. In addition, considering that BMR is more strictly defined than RMR, using BMR may even be preferable.
The heart rate recording technique for monitoring physical activity has limitations (21) because nonexercise factors, such as emotions, may influence the results obtained. A second limitation in the present study is that heart rate was measured only during parts of the time periods when TEE was assessed using doubly labeled water. Therefore, our results need to be confirmed in future studies. This limitation with respect to the length of the heart rate-recording period was especially pronounced in the American women, where only 3 d were monitored, whereas Swedish women were studied for 6 or 7 d. Furthermore, the procedure used to record heart rate in the American women was less detailed than that used in the Swedish women. Therefore, it is not surprising that, as indicated by the limits of agreement shown in Table 2, the deviation from reference estimates of TEE varied more in the American than in the Swedish women. Our results obtained in Swedish women before pregnancy suggest that the proposed calculation produces more accurate results than the standard calculation in a group of women with 19-45% TBF. The results obtained in American women neither confirm nor refute this conclusion. In addition, as shown in Table 2, the Bland-Altman (4) comparison indicated that compared with the standard calculation, the limits of agreement for the proposed calculation were as wide, or even wider. This observation indicates that the body fat content may affect the energy cost of performing different activities in a way not taken into account by the proposed calculation. Therefore, more studies are needed on the potential of this calculation to accurately assess energy expenditure in subjects with different levels of body fat. This point is illustrated by the study by Byrne et al. (5), who recently demonstrated that the standard estimate for RMR of 4.184 kJ·kg−1·h−1 was too high in a group of overweight and obese women. Furthermore, when walking at 5.6 km·h−1 (3.8 METs), the average energy expenditure of these women was 19.0 kJ·kg−1·h−1 (5), whereas our proposed calculation would have estimated a value of 15.1 kJ·kg−1·h−1, and the standard calculation would have estimated a value of 12.7 kJ·kg−1·h−1.
Our findings demonstrate the effect of weight gain on the results when TEE is calculated by means of the MET system. Using the relationship in Figure 1, it can be demonstrated that the bias in the results increased as the American women gained 13.9 kg of body weight and increased their average TBF from 30.8 to 39.3% because the ratio of TEE by the standard calculation/TEE by the proposed calculation decreased from 0.92 to 0.89, indicating that this increase in TBF is associated with estimates of TEE that are 3% lower with the standard than with the proposed calculation. It is interesting to compare these results with those obtained when the weight gain was due to pregnancy. In this case, % TBF remained almost the same. Nevertheless, in gestational week 32, TEE by the standard calculation/TEE by the proposed calculation was higher than in nonpregnant women with the same % TBF. This finding is explained by the increased BMR·kg−1 body weight observed during pregnancy. This illustrates the importance of considering the RMR or BMR of subjects when calculating their energy expenditure by means of the MET system.
The results reported in this paper are relevant in all situations when an estimate of the TEE of human subjects is requested. For example, recommendations regarding human energy requirements are based on information about daily activity pattern and MET values for various activities (6). Such information is also used to evaluate the accuracy of dietary surveys (12) in a variety of subjects including pregnant women (7). Thus, the accuracy of the MET system is of fundamental interest in the science of nutrition and of practical interest in many situations.
In conclusion, the decrease in RMR or BMR per kilogram of body weight, which is the consequence of an increased % TBF, confounds the standard MET system when using measured RMR as it is applied in most women and probably also in many men. Failure to recognize this decrease generally leads to estimates of energy expenditure that will be increasingly more underestimated as the body fat content increases. It is also important to keep in mind that the RMR of subjects may be altered due to physiological changes such as pregnancy, which may influence the results when the MET system is applied. Furthermore, our findings suggest that adopting a modification of the MET system, which takes the effect of the body fat content on the RMR and the energy cost of locomotion into account, is likely to improve the accuracy when calculating the TEE of groups. However, more studies are needed to confirm this conclusion and to further develop the MET system to make it applicable to individuals with different TBF content.
This paper was compiled during a visit by E. Forsum to the University of Wisconsin in Madison, made possible by grants from the Swedish Society of Medicine and Knut and Alice Wallenberg's Foundation.
1. Ainsworth, B., W. Haskell, A. Leon, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med. Sci. Sports Excerc.
2. Ainsworth, B., W. Haskell, A. Leon, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med. Sci. Sports Excerc.
3. Black, A., A. Prentice, and W. Coward. Use of food quotients to predict respiratory quotients for the doubly-labelled water method of measuring energy expenditure. Hum. Nutr. Clin. Nutr.
4. Bland, J. M., and D. G. Altman. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet
5. Byrne, N. M., A. P. Hills, G. R. Hunter, R. L. Weinsier, and Y. Schutz. Metabolic equivalent: one size does not fit all. J. Appl. Physiol.
6. Food and Agriculture Organization of the United Nations. Human energy requirements.
Report of a Joint FAO/WHO/UNU Expert Consultation. Food and Nutrition Technical Report Series No. 1, Rome, pp. 35-52, 2004.
7. Godfrey, K., S. Robinson, D. J. P. Barker, C. Osmond, and V. Cox. Maternal nutrition in early and late pregnancy in relation to placental and fetal growth. Br. Med. J.
8. Haggarty, P., M. Valencia, G. McNeill, et al. Energy expenditure during heavy work and its interaction with body weight. Brit. J. Nutr.
9. Hassard, T. Understanding biostatistics
, St Louis, MO: Mosby-Year Book, 1991, pp. 66-67, 75-97.
10. Howell, W., C. Earthman, P. Reid, K. Greaves, J. Delany, and L. Houtkooper. Doubly labeled water
validation of the compendium of physical activities in lean and obese college women (Abstract). Med. Sci. Sports Exerc.
11. Larsson, I., H. Bertéus Forslund, A. K. Lindroos, et al. Body composition in the SOS (Swedish Obese Subjects) reference study. Int. J. Obes.
12. Livingstone, M. B. E., and A. E. Black. Markers of the validity of reported energy intake. J. Nutr.
13. Lof, M., U. Hannestad, and E. Forsum. Comparison of commonly used procedures, including the doubly-labelled water technique, in the estimation of total energy expenditure of women with special reference to the significance of body fatness. Br. J. Nutr.
14. Lof, M., and E. Forsum. Hydration of fat-free mass in healthy women with special reference to the effects of pregnancy. Am. J. Clin. Nutr.
15. Lof, M., and E. Forsum. Activity pattern and energy expenditure due to physical activity
before and during pregnancy in healthy Swedish women. Br. J. Nutr.
16. McNeill, G. Energy intake and expenditure. In: Human Nutrition and Dietetics
, 10th edition, J. S. Garrow, W. P. T. James, and A. Ralph (Eds.). London: Churchill Livingstone, pp. 25-36, 2000.
17. Passmore, R., and J. V. G. A. Durnin. Human energy expenditure. Physiol. Rev.
18. Racette, S. B., D. A. Schoeller, and R. F. Kushner. Comparison of heart rate and physical activity
recall with doubly labeled water
in obese women. Med. Sci. Sports Exerc.
19. Schoeller, D. A., and E. van Santen. Measurement of energy expenditure in humans by doubly labeled water
method. J. Appl. Physiol.
20. Schoeller, D. A., K. Shay, and R. F. Kushner. How much physical activity
is needed to minimize weight gain in previously obese women? Am. J. Clin. Nutr.
21. Sjöström, M., U. Ekelund, and A. Yngve. Assessment of physical activity
. In: Public Health Nutrition.
, M. J. Gibney, B. M. Margetts, J. M. Kearny, and L., Arab (Eds.). Oxford: Blackwell Publishing, pp. 83-105, 2004.
22. Staten, L. K., D. L. Taren, W. H. Howell, et al. Validation of the Arizona activity frequency questionnaire using doubly labeled water
. Med. Sci. Sports Exerc.
23. Votruba, S. B., S. Blanc, and D. A. Schoeller. Pattern and cost of weight gain in previously obese women. Am. J. Physiol. Endocrinol. Metab.
24. World Health Organization. Diet, Nutrition and the Prevention of Chronic Diseases.
Report of a Joint Expert Consultation. Technical Report Series No 916, Geneva, 2003, paragraph 1.