点击显示 收起
1 From the Department of Food and Nutrition, Umeå University, Umeå, Sweden (CLL and GKJ), and the Department of Human Biology, Maastricht University, Maastricht, Netherlands (KRW).
2 Supported by grants from the Faculty of Social Sciences, Umeå University; The Swedish Institute; The Swedish Foundation for Healthcare Science and Allergy Research; The Swedish Nutrition Foundation; and the JC Kempes Memorial Fund. 3 Address reprint requests to CL Larsson, Department of Food and Nutrition, Umeå University, SE-901 87 Umeå, Sweden. E-mail: christel.larsson{at}kost.umu.se.
ABSTRACT
Background: It is difficult to obtain accurate reports of dietary intake; therefore, reported dietary intakes must be validated. Researchers need low-cost methods of estimating energy expenditure to validate reports of energy intake in groups with different lifestyles and eating habits.
Objective: We sought to validate the reported energy expenditure and energy and protein intakes of Swedish adolescent vegans and omnivores.
Design: We compared 16 vegans (7 females and 9 males; mean age: 17.4 ± 0.8 y) with 16 omnivores matched for sex, age, and height. Energy expenditure as reported in a physical activity interview and energy and protein intakes as reported by diet history were validated by using the doubly labeled water method and by measuring urinary nitrogen excretion.
Results: The validity of reported energy expenditure and energy and protein intakes was not significantly different between vegans and omnivores. The physical activity interview had a bias toward underestimating energy expenditure by 1.4 ± 2.6 MJ/d (95% CI: 2.4, 0.5 MJ/d). The diet-history interview had a bias toward underestimating energy intake by 1.9 ± 2.7 MJ/d (95% CI: 2.9, 1.0 MJ/d) but showed good agreement with the validation method for nitrogen (protein) intake (underestimate of 0.40 ± 1.90 g N/d; 95% CI: 1.10, 0.29 g N/d).
Conclusions: The physical activity and diet-history interviews underestimated energy expenditure and energy intake, respectively. Energy intake and expenditure were underestimated to the same extent, and the degree of underestimation was not significantly different between vegans and omnivores. Valid protein intakes were obtained with the diet-history method for both vegans and omnivores.
Key Words: Vegetarian • vegan • omnivore • adolescent • energy expenditure • energy intake • protein intake • physical activity • dietary assessment • doubly labeled water • biological markers
INTRODUCTION
Few validated dietary surveys have been done in adolescents. Published reports of these surveys showed that the diet-history (DH) method produced valid estimates of energy intakes (EIs), whereas the diet-record method underestimated intakes by 46% (1–3). Unvalidated dietary surveys of adolescents are also scarce, and often report EIs and energy expenditures (EEs) that are not consistent with long-term survival (4,5). Such reports often give the EE and EI data in relation to basal metabolic rate (BMR), reporting the food intake level (FIL = EI/BMR) and physical activity level (PAL = EE/BMR), which are too low for long-term survival.
Assessment of dietary intakes in populations with various lifestyles and dietary habits may give rise to methodologic problems, producing different biases in different populations. In addition, food databases used for calculating nutrient intakes are usually designed for the most prevalent eating habits in the population, which are typically omnivorous diets. Consequently, many vegetarian food items are missing from the databases and, thus, information about them must be collected and entered manually if correct nutrient intake values are to be obtained.
Doubly labeled water (DLW) can be used as a biological marker to validate reported energy intake (EIrep) and reported energy expenditure (EErep) (3,5). The method is expensive to use in large populations, and only a limited number of dietary surveys have been validated by the DLW method. Therefore, there is a need for simplified, less expensive, but still valid methods of estimating EErep. With a simplified validation method, EIrep could then be validated by comparison with EErep in large population studies; the ratio of EIrep to EErep should equal 1.0 if the estimates of EIrep and EErep are accurate (5).
Protein (nitrogen) intake can be validated by comparing it with nitrogen excretion in urine samples collected over 24 h (6,7). Because the proportion of total energy derived from protein is often relatively constant, protein intake is also an indicator of total food (energy) intake. The Swedish food database contains information on the protein, but not the nitrogen, contents of foods. To convert nitrogen excretion to protein intake, an average nitrogen intake of 16% in dietary protein is usually assumed (7). However, for veganism and other unusual dietary habits, this assumption may give rise to biased estimates of protein intake.
The aim of this study was to validate a physical activity (PA) interview and a DH interview for estimating EE, EI, and protein intake in Swedish adolescent vegans and omnivores by comparison with biological markers of EE and protein intake. This validation study is part of a larger study comparing young vegans and omnivores in terms of their dietary intakes and nutritional status.
SUBJECTS AND METHODS
Subjects
Thirty-two adolescents living in Umeå, Sweden, were recruited by using advertisements and visits to schools: the subjects were 16 vegans and 16 omnivores matched to the vegans by sex, age (±6 mo), and height (±3.5 cm). The subjects had to be in good health, have no chronic diseases, and be aged 16–20 y. Vegans were defined as people eating foods of plant origin only and were included if they had been consuming such a diet for 6 mo and were planning to continue this diet. Omnivores were defined as people eating foods of both plant and animal origin and were included if they had been consuming such a diet for their entire life and were planning to continue this diet. In total, 65 vegetarians and 272 omnivores were interviewed before the 32 subjects were enrolled. All participants gave their written, informed consent. The study was approved by the Research Ethics Committee of the Medical Faculty, Umeå University.
Dietary assessment
DH interviews were conducted (by CLL) with all the subjects between January and June 1998. Each subject was questioned about his or her usual food intake during two 1–2-h interviews separated by 1–2 wk. The aim was to describe the subjects' typical eating patterns over the preceding 3 mo. During the first interview, meal patterns, food choices, and food frequencies were discussed. Subjects were asked to bring recipes and information about dietary supplements to the second interview. Subjects described average portion sizes of food items in terms of household measures, standard weights of food items, 3-dimensional food models, or validated food-portion photographs representing known weights (8). The food, beverage, and supplement intakes from the DH interview were entered into the dietary analysis program STOR MATS 4_03e, which uses the Swedish food composition database (version 2_97; Rudans Lattdata, Vasteras, Sweden). When composite foods not present in the database were reported, the subjects' own recipes were entered. Months of vegetarianism reported in the interview were defined as months of eating either lactoovovegetarian or lactovegetarian diets, which include dairy products and eggs or just dairy products, respectively, but no meat, poultry, or fish. Months of veganism were defined as months of eating foods of plant origin only. Body weight in light underwear was measured to the nearest 0.1 kg and height was measured to the last completed 0.5 cm.
Urine collections
After the first DH interview, each subject started the first of four 24-h urine collections. Because many factors can affect nitrogen balance, both before and during the urine collection, subjects were asked to provide information about any problems during the collection. To this end, subjects were asked if they had lost any urine, used any medication, had any problems with diarrhea, had atypical dietary intake or PA, or perspired a lot. However, it is difficult to estimate the influence of these factors on the results.
The completeness of the urine collections was verified by giving the subjects para-aminobenzoic acid (PABA) and measuring the amount excreted in the urine (9). Each subject took 3 tablets containing 80 mg PABA each: one tablet on awakening (after first emptying the bladder), one within 6 h (at the midday meal), and one within 12 h (at the evening meal). The 24-h urine collections began after the subjects awoke in the morning and emptied their bladders and continued until the next morning. Each 24-h urine collection was weighed and 2 samples from each collection were analyzed for PABA and nitrogen contents.
PABA in the urine was measured colorimetrically (9). Urine collections containing 85–110% of the PABA consumed were classified as complete. Urine collections containing <85% of the PABA consumed were classified as incomplete. Those urine collections containing <50% or >110% of the PABA consumed were rejected. For the collections containing 50–85% of the PABA consumed, the nitrogen content was adjusted by using a linear regression method (10) and the following equation.
RESULTS
Subjects
Some characteristics of the subjects are shown in Table 1. The mean age of the 32 subjects was 17.4 ± 0.8 y. The vegans had been practicing their dietary regimen for a mean of 1.9 y (range: 0.5–3.5 y). Female vegans had a higher body weight and body mass index (BMI; in kg/m2) than did female omnivores, but male vegans had a lower BMI and body water content than did male omnivores. Group averages for weight and BMI were not significantly different between vegans and omnivores.
View this table:
TABLE 1 . Background data for the Swedish vegans and omnivores in 1997–19981
Validity of reported energy expenditure
Female vegans had significantly lower PALrep and PALmeas than did female omnivores, but the validity of EErep (PALrep/ PALmeas) was not significantly different between the 2 dietary groups (Table 2). However, the PA interview method underestimated EE by 1.44 ± 2.55 MJ/d (95% CI: 2.37, 0.52 MJ/d) (Figure 1).
View this table:
TABLE 2 . Validation of reported energy expenditure and energy and protein intakes in Swedish vegans and omnivores by using the doubly labeled water method and 24-h urine collections in 1997–19981
FIGURE 1. . Difference between reported (during physical activity interview) energy expenditure (EE) and measured (by the doubly labeled water method) EE plotted against the mean of the reported EE and measured EE in 16 omnivores () and 16 vegans (). The solid horizontal line represents all subjects (r2 = 0.5808). Negative values for the difference indicate that reported EE was lower than measured EE. r = -0.762, P < 0.05. Linear regression equation: y = 8.661 - 0.833x.
Validity of reported energy intake
Female vegans had a significantly lower FIL than did female omnivores, but the validity of EIrep (FIL/PALmeas) was not significantly different between the 2 dietary groups (Table 2); vegans underreported EI by 14%, whereas omnivores underreported EI by 12%. However, the DH interview method underestimated EI by 1.93 ± 2.65 MJ/d (95% CI: 2.89, 0.97 MJ/d) (Figure 2); the bias of the method was consistent in the 2 dietary groups.
FIGURE 2. . Difference between reported (during diet-history interview) energy intake (EI) and measured (by the doubly labeled water method) energy expenditure (EE) plotted against the mean of reported EI and measured EE in 16 omnivores () and 16 vegans (). The solid horizontal line represents all subjects (r2 = 0.0450). Negative values for the difference indicate that reported EI was lower than measured EE. r = -0.212 (NS). Linear regression equation: y = 0.396 - 0.196x.
Validity of reported protein intake
The mean recovery of PABA from the 24-h urine collections was 74% for vegans and 72% for omnivores (n = 15). A total of 23 urine collections were excluded. However, 3 or 4 collections were acceptable for 26 of the subjects: 24 urine collections were classified as complete and 81 were incomplete but the nitrogen content was adjusted as described in Methods.
Four collections were usable for each of 17 subjects, yielding 68 collections; 18 of these 68 collections were classified as complete and the nitrogen content of the remaining 50 collections was adjusted. Three collections were usable for each of 9 subjects, yielding 27 collections; 5 of these 27 collections were classified as complete and the nitrogen content of the remaining 22 collections was adjusted. Two collections were usable for each of 5 subjects, yielding 10 collections; 1 of these 10 collections was classified as complete and the nitrogen content of the remaining 9 collections was adjusted. One subject had a PABA recovery >130% in all 4 urine collections and was therefore excluded.
The calculated protein factor was 6.00 for vegans and 6.22 for omnivores. Female vegans reported a significantly lower (Nrep x 0.81)/Nmeas than did their omnivorous counterparts, but the validity of Nrep was not significantly different between the 2 dietary groups (Table 2). The reported nitrogen intake from the DH interview showed good agreement with the Nmeas in that the mean difference between (Nrep x 0.81) and Nmeas was -0.40 ± 1.90 g/d (95% CI: -1.10, 0.29 g/d) (Figure 3).
FIGURE 3. . Difference between 81% of reported (during diet-history interview) nitrogen intake and mean nitrogen excretion in 2, 3, or 4 different 24-h urine collections per subject (5 subjects with 2 collections, 9 subjects with 3 collections, and 17 subjects with 4 collections) plotted against the mean of 81% of nitrogen intake and nitrogen excretion in 15 omnivores () and 16 vegans (). The solid horizontal line represents all subjects (r2 = 0.0453). Negative values for the difference indicate that 81% of reported nitrogen intake was lower than nitrogen excretion. r = 0.213 (NS). Linear regression equation: y = -1.523 + 0.105x.
Comparison of the validity of the energy and protein intakes
A comparison of FIL/PALmeas with (Nrep x 0.81)/Nmeas by Wilcoxon's signed-rank test showed no significant difference between EIrep and reported protein intakes among the vegans, but a significant difference was found among the omnivores (P < 0.05; data not shown). There was no significant difference between males and females within either dietary group with regard to the validity of energy and protein intakes.
Urinary isotope concentrations
The mean (±SD) enrichments in the 96 background samples were 0.0146 ± 0.0001 AP (146 ± 1 ppm) for 2H and 0.1994 ± 0.0002 AP (1994 ± 2 ppm) for 18O. The SD between subjects was 0.0001088 AP (1.088 ppm) for 2H and 0.0001655 AP (1.655 ppm) for 18O, and the SD within subjects ranged from 0.0000026 AP (0.026 ppm) to 0.0000275 AP (0.275 ppm) for 2H and from 0.0000024 AP (0.024 ppm) to 0.0000853 AP (0.853 ppm) for 18O. The ratio of 2H to 18O in terms of space occupied by the isotopes was 1.0125 ± 0.0177.
Respiratory quotient and food quotient
The estimated respiratory quotient was 0.85. The observed FQ averaged 0.90 for all subjects.
DISCUSSION
Subjects
The subjects in the present study, except for the female vegans, were similar in height, weight, and BMI to a sample of 251 females and 342 males living in Umeå (mean age: 16.8 y; range: 15.9–18.2 y) (20). The female vegans weighed more and had a higher mean BMI than both the larger sample and the matched female omnivores in the present study, which is the opposite of what was observed in female vegans in England aged 20–89 y (21). Because we enrolled a small, nonrandom sample of female vegans, it is difficult to draw any firm conclusions about these differences between female vegans and omnivores. However, the findings might indicate that some of the female vegans chose to be vegans to reduce their weight and BMI, instead of indicating that their weight and BMI were consequences of their dietary habits.
Validity of reported energy expenditure
The DLW method measures EE accurately under normal conditions with a precision of 2–8% (13). When the DLW method was compared with room calorimetry, the difference between the methods was not significant (1.6 ± 2.6%) (22). In another study, the CV was 9% for repeated DLW measurements (23). In the present study, we used the multipoint method of Schoeller (13) to calculate EEmeas from DLW and we also used the method of Coward (24) to calculate EEmeas (data not shown). No significant difference between the 2 calculation methods was found.
The PA interviews showed a bias toward underestimation of the EErep of the subject groups. One explanation may be that the values used for the PA ratio were too low for the reported activities. Also, some of the subjects' activities might not have been reported in the interviews. However, the degree of underestimation was only 3% for vegans but was 12% for omnivores (Table 2). Subjects with a high EEmeas tended to report amounts of PA that led to underestimates of PA and subjects with a low EEmeas tended to report amounts of PA that led to overestimates of PA with the method used in this study (Figure 1). Because the omnivores had a higher PAL, they also had a greater degree of underestimation with this method than did the vegans. This suggests that the PA interview had a bias in measuring the extreme values of EE.
Validity of reported energy intake
The use of DLW as an independent validation method for EI involves the assumption that the subjects are in energy balance, which was confirmed by comparing body weights at the start and end of the DLW interval. Because the amount of energy stored as new tissue is <5% during the pubertal growth spurt, it is reasonable to assume that EI and EE are equivalent in adolescents during a limited period of time (2). The finding that EIrep was underestimated in the present study is in agreement with the findings of other studies. Dietary surveys in adolescents showed that EI was underestimated by 20–46% with 2-wk diet records, by 20–28% with 7-d diet records, and by 3% with the DH method (1–3).
The observed FQ of vegans in the United Kingdom was reported to be 0.86–0.88 (15). The higher FQ of vegans compared with that of omnivores results from the higher proportion of carbohydrates in the vegans' diet. A higher FQ would yield a lower calculated EE and better agreement with the reported EI in this study. However, we did not use the observed FQ because it might not be the true FQ as a result of underreporting or overreporting of carbohydrate, fat, or protein intakes. Further, because one of the aims of this study was to validate the DH method, it would be incorrect to introduce an error in the validation method.
Validity of reported protein intake
The findings that the mean recovery of PABA from all urine collections was 73% and that 75% of the samples had a PABA recovery <85% indicate that it is difficult to obtain complete 24-h urine collections. Unfortunately, this is a fairly typical result (10,25). Incomplete urine collections cannot be used to validate dietary data. To limit the consequences of this problem, a linear regression method was developed. This method allows for inclusion of urine collections with low PABA recovery (10). The fundamental principle of the linear regression method is described by Bingham and Cummings (9), who validated it in subjects confined to calorimeters. They concluded that an oral dose of PABA and urinary excretion of PABA are directly related, and that PABA is quantitatively excreted in the urine. The correlation coefficient was 0.99 for the relation between nitrogen and PABA recovery with the linear regression method (10), which indicates that any error in the adjustment method would probably be minor. There is no reason to believe that the relation between PABA excretion and nitrogen excretion would be physiologically different in the subjects in the present study than in other subjects. Also, there is no reason to believe that various amounts of protein intake would affect PABA excretion, and in turn affect the validity of the equation. In the present study, the nitrogen content of urine collections with PABA recovery between 50% and 85% was adjusted by using the linear regression method (10); this approach salvaged 81 of the 128 collections. For subject groups, the DH interview method showed good agreement with Nmeas.
The DH interview method seemed to overestimate nitrogen intake at high intakes and underestimate it at low intakes, especially in omnivores (Figure 3). One possible explanation for this finding is that if the diet has a high proportion of protein-rich foods, such as meat, fish, and milk products, then these food items may be part of the main meals and may be easier to remember than foods eaten between meals (26). These foods are considered to be socially desirable and tend to be overestimated in contrast with socially undesirable foods, such as sugary snacks. The latter tend to be underestimated, thereby lowering the total reported EI (27).
Comparison of the validity of energy and protein intakes
Selective underreporting of food and nutrient intakes by high-FIL compared with low-FIL reporters has been described (12,25). For example, in a small observational study, 101% of the protein intake but only 88% of the EI was reported (26). Similar selective underreporting was found in omnivores in the present study: 4 subjects (27%) had lower (Nrep x 0.81)/Nmeas values and 11 subjects (73%) had higher (Nrep x 0.81)/Nmeas values relative to their FIL/PALmeas values. However, no significant difference was seen among vegans or by sex in either dietary group. A possible explanation for the difference observed in omnivores is that nitrogen intake was overestimated at higher intakes and underestimated at lower intakes with the DH interview method, as discussed above and as shown in Figure 3. Another possible explanation for the difference in omnivores is that the dietary intake data, urine data, and EE measurements represented different periods of time. However, because the DH method covered 3 mo, the DLW method spanned 2 wk, and the urine samples were collected on 4 random days between the DH and the DLW period, the effect of day-to-day variation should have been small.
One of the aims of the present study was to validate methods for estimating EI and EE that would avoid the high cost of the DLW method. Therefore, a cost comparison of the methods is of interest. The cost of DLW and related laboratory materials and analyses in this study was $500 per subject, compared with $1 per subject for the PA interview form. However, the cost of the actual DLW can vary substantially over time, depending on supply and demand and the dollar exchange rate. The labor cost of collecting the DLW samples compared with the labor cost of performing the interviews is approximately the same. However, the labor cost of conducting the DH interviews plus performing the nutritional calculations is much higher because 7 h of labor is needed per subject.
Conclusion
This study showed that there were no significant differences between vegans and omnivores in the validity of EErep as estimated by a PA interview or the validity of reported energy and protein intakes as estimated by a DH interview. However, EI and EE were underestimated to the same extent, whereas valid estimates of protein intake were obtained.
ACKNOWLEDGMENTS
We thank all the adolescents who participated in this study.
REFERENCES