Download Accuracy of Measuring Energy Intake & Expenditure in Epidemiological Studies and more Study notes Epidemiology in PDF only on Docsity!
Chapter 3. Can energy intake and expenditure (energy balance) be measured accurately in epidemiological studies? Is this important? 17
CHAPTER 3
chapter 3.
Can energy intake and
expenditure (energy balance)
be measured accurately
in epidemiological studies?
Is this important?
Walter C. Willett and Changzheng Yuan
have argued that self-reported ener- gy intake has no value and should be abandoned, and have extended this argument to all self-reported information. Others [4–8] have sug- gested that it is not realistically pos- sible to measure energy intake and expenditure with sufficient precision in epidemiological studies to assess energy balance, but that this is not a serious problem because other means can be used to evaluate the effects (e.g. on disease incidence or mortality) of energy balance as an exposure and to study the determi- nants (e.g. dietary factors and phys- ical activity) of energy balance as an outcome. In this chapter, the factors con- tributing to energy balance and the measurement of these factors are reviewed. Notably, energy intake and expenditure have important roles in human health, and in epidemiologi-
The roles of energy intake and expenditure are extremely important in human health and disease, for many reasons. Thus, the assess- ment and interpretation of energy intake and expenditure are major issues in epidemiological studies. Overweight and obesity have been recognized to be major risk factors for cancer, cardiovascular disease, diabetes, and many other health conditions. Therefore, the difference between energy intake and expendi- ture, frequently referred to as energy balance, has become of great inter- est, because of its direct relationship to long-term gain or loss of adipose tissue. For this reason, questions have arisen about whether energy intake and expenditure can be measured adequately in epidemiological stud- ies, to enable energy balance to be assessed adequately. Some [1–3]
cal studies, independent of their con- tribution to energy balance; although these other applications are not the focus of this chapter, they are also mentioned.
Components of total energy
expenditure
Total energy expenditure has tradi- tionally been partitioned into several components: resting metabolic rate (RMR), physical activity, thermo- genic effect of food, and adaptive thermogenesis (Fig. 3.1) [9]. RMR is quantitatively the most impor- tant, making up approximately 60% of total energy expenditure in an individual with moderate physical activity. In a moderately active indi- vidual, physical activity accounts for approximately 30% of total energy expenditure. The thermogenic effect of food (i.e. the metabolic cost of
18
diction models based on age, weight, and sex have been developed [4]. In principle, height should also be add- ed to the prediction models because, for the same weight, a taller person would be leaner, but height appears to add minimal variability. Because age, weight (or body mass index [BMI] plus height), and sex are routinely covari- ates in epidemiological studies, RMR is reasonably controlled for in most epidemiological analyses. After age, weight, and sex have been controlled for, physical activity assumes a relatively large role in de- termining the variation in energy ex- penditure among free-living individu- als. The true proportion of variation in energy expenditure accounted for by physical activity differs substan- tially among populations and is likely to be underestimated in most studies because of imperfect measurements of physical activity. Ravussin et al. [10] have demonstrated that even motor activity within the confines of a respiratory chamber (“fidgeting”) varies dramatically between individ- uals and can account for hundreds of kilocalories per day. Such differ- ences in activity would not be detect- ed by typical questionnaires. Thus, physical activity, which includes both fine motor and major muscle move- ment, is a major determinant of be- tween-person variation in energy
absorbing and processing macronu- trients) accounts for only about 10% of total energy expenditure. Adaptive thermogenesis (i.e. the compensa- tory capacity of an individual to con- serve or expend energy in response to variable intake of food or temper- ature extremes) has been estimated to be less than ± 10% of total energy expenditure [9]. In epidemiological studies, the thermogenic effect of food is not like- ly to vary appreciably, because this becomes important only on extreme diets, and this can generally be as- sumed to be constant. Adaptive ther- mogenesis is practically important, because it can account for resistance to weight loss in the face of moder- ate restriction of energy intake by downregulating metabolic process- es to become more energy-efficient. These differences in metabolic effi- ciency are difficult to measure even under highly controlled conditions as well as in epidemiological studies, so this needs to be recognized as a source of modest unmeasured vari- ation in energy expenditure. RMR is determined mainly by body weight, although this is primarily a function of lean body mass. Because mea- surement of RMR requires metabolic facilities and is therefore not feasible in epidemiological studies and most clinical investigations, a series of pre-
expenditure in many populations. In- deed, in most instances total energy intake can be interpreted as a crude measure of physical activity, espe- cially after controlling for body size, age, and sex.
Energy balance and deviations
from energy balance
Energy balance exists when weight is constant because energy expenditure equals energy intake. This can hap- pen when no components change or when a change in one component is compensated for by changes in other components. In adults, deviations from en- ergy balance are a critical concern because these underlie weight gain and ultimately obesity. Of particular concern to public health and epide- miologists are small increments in weight, such as 0.5–1 kg per year, which are typical of many high-in- come populations [11] and which over a period of 20–30 years lead to large changes in weight and major morbidity and mortality [12]. The de- viations from energy balance need- ed to produce this change in weight are very modest. For example, sim- ply on the basis of the energy con- tent of adipose tissue [4], if an adult man who consumes 2500 kcal/day (10 460 kJ/day) increases his ener- gy intake by only 1% while other fac- tors remain constant, over a 10-year period a theoretical weight gain of 10 kg would result. In reality, the in- crease in weight will be considerably less, because the additional energy cost of maintaining and moving the added body mass eventually equals the increment in energy intake and a new steady state in weight, i.e. bal- ance, is reached. By combining data on energy intake and weight gain and on the compensatory effects of added body mass on energy expenditure, Hall et al. [13] estimated that for each 10 kcal/day (42 kJ/day) increase in
Fig. 3.1. Components of energy expenditure during weight maintenance for a 70 kg man consuming 2500 kcal/day, and the potential modifying effect of adaptive thermogenesis. Adapted with permission from Horton (1983) [9].
RMR
TEF
TEE
AT
kcal/day
Resting metabolic rate
Thermogenic effect of exercise
Thermogenic effect of food
Adaptive thermogenesis
− 300 − 100 100 300 500 700 900 1100 1300 1500
RMR
TEF
TEE
AT
kcal/day
Resting metabolic rate
Thermogenic effect of exercise
Thermogenic effect of food
Adaptive thermogenesis
− 300 − 100 100 300 500 700 900 1100 1300 1500
20
of the fact that the data do not cap- ture many activities of daily living and fine motor movements. Physical activity records and 24-hour physical activity recalls, analogous to their corresponding dietary assessment methods, have been minimally used in epidemiological studies thus far. Motion sensors – small devices for monitoring physical activity – are be- coming sufficiently inexpensive to be used in epidemiological studies, but the best way to convert movement counts to energy expenditure is still being evaluated. The DLW measure, after subtracting energy expenditure due to RMR, is now often considered to be the gold standard for evalua- tion of other methods to assess energy expenditure due to physi- cal activity. The variation of these methods over 1 year is also shown in Table 3.1. The most consistent measure appears to be by acceler- ometer, expressed as counts over 1 day (CV%, 19%; ICC, 0.79), and
the DLW measure for physical activ- ity was considerably more variable than that for total energy expenditure (CV%, 23%; ICC, 0.51). As can be appreciated, the with- in-person CV% values both for total energy intake and for physical activ- ity assessed by all methods are all far greater than the approximately 1% deviation from energy balance that would be needed to evaluate small but important long-term de- viations from energy balance in individuals. Because the deviation would be calculated as the differ- ence between the variables, its within-person error would be even greater because it would include variability from both measures of energy intake and physical activity. Thus, as has been noted earlier [4], it is clear that available methods for measuring energy intake and phys- ical activity in epidemiological stud- ies, as well as methods considered to be the gold standard, are far from
adequate for assessing long-term deviations from energy balance in individuals. The relative absence of this approach in the epidemiological literature reflects this understand- ing. Further, it is unlikely that such methods will become available, because of inherent challenges in obtaining highly precise measure- ments of long-term behaviours of free-living individuals.
Alternative methods to
assess energy balance in
epidemiological studies
Fortunately, the study of energy bal- ance does not require measurements of energy intake and expenditure, because attained weight and chang- es in weight are readily measured with high precision, even by self-re- port [4]. These measurements of weight provide a simple but precise time-integrated measure of changes in energy balance. Also, weight and
Table 3.1. Distribution, within-person coefficient of variation (CV%), and intraclass correlation coefficient (ICC) for different measures of energy intake and expenditure in the Women’s Lifestyle Validation Study [25] (data provided by 622 female nurses in the USA aged 45–80 years) a
Method ( n = 622)
Time interval Mean (SD) Within-person SD
Within-person CV%
ICC
FFQ (kcal/day) 1 year 1901 (480) 286 15 0. Dietary records (kcal/day) ~6 months 1745 (334) 226 13 0. ASA24 (kcal/day) Every 3 months 1825 (475) 507 28 0. DLW (kcal/day) 6–12 months 2195 (360) 190 9 0. Weight (lb) Every 3 months 157.6 (33.4) 5.1 3 0. PAQ (MET-h/day) 1 year 16.5 (6.8) 5.3 32 0. Accelerometer (min/day) b^ ~6 months 19.5 (16.6) 8.9 46 0. Accelerometer (counts/day) ~6 months 243 056 (94 356) 45 827 19 0. PAEE (kcal/day) 6–12 months 708 (239) 166 23 0. PAEE (kcal/day) c^ 6–12 months 708 (237) 165 23 0. PAEE (kcal/day) d^ 6–12 months 708 (230) 165 23 0.
ASA24, automated self-administered 24-hour dietary recall; CV, coefficient of variation; DLW, doubly labelled water; FFQ, food frequency questionnaire; ICC, intraclass correlation coefficient; METs, metabolic equivalents; PAEE, physical activity energy expenditure; PAQ, physical activity questionnaire; SD, standard deviation. a (^) ICC and CV% were calculated based on the original value. b (^) Moderate and vigorous activity (min/day), 1-minute bouts. c (^) Measure of activity assessed by DLW with weight regressed out. d (^) Measure of activity assessed by DLW with weight, age, and height regressed out.
Chapter 3. Can energy intake and expenditure (energy balance) be measured accurately in epidemiological studies? Is this important? 21
CHAPTER 3
changes in weight directly repre- sent the primary health concern due to deviations from energy balance, which is adiposity. Thus, the inabil- ity to evaluate long-term deviations from energy balance in individuals by measuring energy intake and ex- penditure is not important. The use of weight in epidemi- ological analyses, both as an ex- posure and as an outcome, is an important topic that has been dis- cussed widely [22]. When adjusted for height, often expressed as BMI, weight is widely used as a surrogate for adiposity. Although conceptually imperfect because it does not sep- arate lean mass and fat mass, BMI works remarkably well compared with gold-standard methods [23]. When it is used as an exposure, it is important for the analysis to address confounding by smoking, reverse causation due to underlying disease, and loss of lean mass due to frailty at older ages. When it is used as an outcome to study the effects of diet and activity, the study design needs careful consideration. In cross-sectional studies, re- verse causation can readily occur. In prospective studies with only a baseline measurement, the re- sults can be misleading, because a change in diet or activity will of- ten result in a change in weight for some period of time, and then a new steady-state weight will be reached. For example, if physical activity is increased, weight may decrease initially but does not continue to de- crease to zero. If most study par- ticipants have already reached a steady-state weight at baseline, an effect of physical activity on weight could be missed. A better design will usually be to examine change in diet or activity in relation to change in weight [24], which more closely approximates the design of a clinical trial. Unlike most studies with dis- ease outcomes, which can require many thousands of participants and
many years of follow-up, the effects of changes in diet and activity on change in weight can be investigat- ed with a few hundred subjects and 1 or 2 years of follow-up. Randomized trials should play a large role in ad- dressing the effects of diet on weight, because they better control for con- founding by variables that are hard to measure.
Other applications of data
on total energy intake and
expenditure
Although measurements of ener- gy intake and expenditure will not be useful for assessing energy bal- ance in epidemiological studies, they do play other important roles. For example, population trends in mean energy intake over time using 24-hour dietary recalls can provide useful information, because the ef- fects of within-person variation over time can be dampened with large sample sizes. If the method remains standardized over time, temporal trends can still be valid even if there is some systematic underestimation or overestimation. Unfortunately, standardized methods for physical activity assessment over time do not seem to have been used, so there is less certainty about temporal trends in energy expenditure. In nutritional epidemiological studies, assessment of energy in- take is also important as an adjust- ment variable for nutrient intakes, because the focus is primarily on the composition of diet rather than on absolute intakes. This is because the composition of diet is what can most realistically be changed by in- dividuals or a population [4]. Multiple aspects of dietary composition have been associated with weight chang- es [11], probably due to differences in satiety and possible hormonal ef- fects that favour or inhibit accumu- lation of lean mass versus fat mass. Adjustment for total energy intake
also has the benefit of cancelling correlated errors in nutrients, thus reducing measurement errors [4]. Assessment of physical activ- ity, primarily by structured ques- tionnaires, has documented the importance of moderate to vigorous activity in prevention of many dis- eases. Although these measures of physical activity have error, they have been validated by comparisons with more detailed assessments [21] and can thus provide useful informa- tion in prospective studies. The fact that these are based on self-reports rather than an objective measure is not important, because objective measures also have error and are subject to confounding. Even if a good measure of total energy ex- penditure from physical activity were available, this would not provide the important information on specific types of activity that can be obtained by questionnaires. Small motion sensors are now being incorporat- ed into epidemiological studies; the structure of their measurement er- rors is now being investigated.
Conclusions
Deviations from energy balance are important in human health and dis- ease. However, these cannot be assessed adequately in epidemi- ological studies by differences be- tween energy intake and expendi- ture, because very small long-term deviations in energy intake or ex- penditure can have major effects on body weight. Neither the available methods nor the foreseeable future methods will be sufficiently precise and accurate to measure these small differences. However, body weight and change in weight provide precise indicators of long-term deviations from energy balance and are widely available for epidemiological stud- ies. These simple and inexpensive measures of energy balance can be used as both exposure and outcome
Chapter 3. Can energy intake and expenditure (energy balance) be measured accurately in epidemiological studies? Is this important? 23
CHAPTER 3
- Hall KD, Sacks G, Chandramohan D, Chow CC, Wang YC, Gortmaker SL, et al. (2011). Quantification of the effect of en- ergy imbalance on bodyweight. Lancet. 378(9793):826–37. http://dx.doi.org/10.1016/ S0140-6736(11)60812-X PMID:
- Sanghvi A, Redman LM, Martin CK, Ravuss- in E, Hall KD (2015). Validation of an inexpensive and accurate mathematical method to measure long-term changes in free-living energy intake. Am J Clin Nutr. 102(2):353–8. http://dx.doi. org/10.3945/ajcn.115.111070 PMID:
- Wang YC, Gortmaker SL, Sobol AM, Kuntz KM (2006). Estimating the energy gap among US children: a counterfactual approach. Pediatrics. 118(6):e1721–33. http://dx.doi.org/10.1542/ peds.2006-0682 PMID:
- Willett W, Lenart E (2013). Reproducibility and validity of food-frequency questionnaires. In: Willett W, editor. Nutritional epidemiology. 3rd ed. New York (NY), USA: Oxford University Press; pp. 96–141.
- Roberts SB, Fuss P, Evans WJ, Heyman MB, Young VR (1993). Energy expenditure, aging and body composition. J Nutr. 123 (2 Suppl): 474–80. PMID:
- Freedman LS, Commins JM, Moler JE, Wil- lett W, Tinker LF, Subar AF, et al. (2015). Pooled results from 5 validation studies of dietary self- report instruments using recovery biomarkers for potassium and sodium intake. Am J Epidemiol. 181(7):473–87. http://dx.doi.org/10.1093/aje/ kwu325 PMID:
- Neuhouser ML, Tinker L, Shaw PA, Schoeller D, Bingham SA, Horn LV, et al. (2008). Use of recovery biomarkers to calibrate nutrient con- sumption self-reports in the Women’s Health Ini- tiative. Am J Epidemiol. 167(10):1247–59. http:// dx.doi.org/10.1093/aje/kwn026 PMID:
- Wong WW, Roberts SB, Racette SB, Das SK, Redman LM, Rochon J, et al. (2014). The doubly labeled water method produces highly repro- ducible longitudinal results in nutrition studies. J Nutr. 144(5):777–83. http://dx.doi.org/10.3945/ jn.113.187823 PMID:
- Hu FB (2013). Assessment of physical ac- tivity in nutritional epidemiology. In: Willett W, editor. Nutritional epidemiology. 3rd ed. New York (NY), USA: Oxford University Press; pp. 241–59.
- Manson JE, Stampfer MJ, Hennekens CH, Willett WC (1987). Body weight and longevity. A reassessment. JAMA. 257(3):353–8. http://dx.doi.org/10.1001/ jama.1987.03390030083026 PMID:
- Sun Q, van Dam RM, Spiegelman D, Heymsfield SB, Willett WC, Hu FB (2010). Comparison of dual-energy x-ray absorp- tiometric and anthropometric measures of adiposity in relation to adiposity-related bio- logic factors. Am J Epidemiol. 172(12):1442–
- http://dx.doi.org/10.1093/aje/kwq PMID:
- Smith JD, Hou T, Hu FB, Rimm EB, Spiege- lman D, Willett WC, et al. (2015). A compari- son of different methods for evaluating diet, physical activity, and long-term weight gain in 3 prospective cohort studies. J Nutr. 145(11):2527–34. http://dx.doi.org/10.3945/ jn.115.214171 PMID:
- Yuan C, Spiegelman D, Rimm EB, Ros- ner BA, Stampfer MJ, Barnett JB, et al. (2017). Validity of a dietary questionnaire as- sessed by comparison with multiple weighed dietary records or 24-hour recalls. Am J Epidemiol. 185(7):570–84. PMID:
