AJCN Cancer Health Disparities Conference
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Horner, N. K
Right arrow Articles by Prentice, R. L
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Horner, N. K
Right arrow Articles by Prentice, R. L
Agricola
Right arrow Articles by Horner, N. K
Right arrow Articles by Prentice, R. L
American Journal of Clinical Nutrition, Vol. 76, No. 4, 766-773, October 2002
© 2002 American Society for Clinical Nutrition

Original Research Communication

Participant characteristics associated with errors in self-reported energy intake from the Women’s Health Initiative food-frequency questionnaire1,2,3

Neilann K Horner, Ruth E Patterson, Marian L Neuhouser, Johanna W Lampe, Shirley A Beresford and Ross L Prentice

1 From the Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center (NKH, REP, MLN, JWL, SAB, and RLP), and the Department of Epidemiology, University of Washington (REP, JWL, and SAB), Seattle.

2 Supported by the Fannie E Rippel Foundation, the Fred Hutchinson Cancer Research Center, the National Institutes of Health (T32 CA09661), and the National Cancer Institute (R03 CA80648 and R01 CA53996).

3 Address reprint requests to RE Patterson, Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue North, MP-702, Seattle, WA 98109-1024. E-mail: rpatters{at}fhcrc.org.


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Background: Errors in self-reported dietary intake threaten inferences from studies relying on instruments such as food-frequency questionnaires (FFQs), food records, and food recalls.

Objective: The objective was to quantify the magnitude, direction, and predictors of errors associated with energy intakes estimated from the Women’s Health Initiative FFQ.

Design: Postmenopausal women (n = 102) provided data on sociodemographic and psychosocial characteristics that relate to errors in self-reported energy intake. Energy intake was objectively estimated as total energy expenditure, physical activity expenditure, and the thermic effect of food (10% addition to other components of total energy expenditure).

Results: Participants underreported energy intake on the FFQ by 20.8%; this error trended upward with younger age (P = 0.07) and social desirability (P = 0.09) but was not associated with body mass index (P = 0.95). The correlation coefficient between reported energy intake and total energy expenditure was 0.24; correlations were higher among women with less education, higher body mass index, and greater fat-free mass, social desirability, and dissatisfaction with perceived body size (all P < 0.10).

Conclusions: Energy intake is generally underreported, and both the magnitude of the error and the association of the self-reporting with objectively estimated intake appear to vary by participant characteristics. Studies relying on self-reported intake should include objective measures of energy expenditure in a subset of participants to identify person-specific bias within the study population for the dietary self-reporting tool; these data should be used to calibrate the self-reported data as an integral aspect of diet and disease association studies.

Key Words: Dietary records • systematic bias • dietary measurement error • energy expenditure • postmenopausal women • Women’s Health Initiative


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Nutrient intake estimates from self-reporting tools such as food-frequency questionnaires (FFQs), records, and recalls are known to contain considerable random error. Of greater concern is the accumulating evidence that these nutrient estimates also incorporate systematic error. In particular, dietary self-reporting appears to substantially underestimate energy intake, and there is some evidence that this underreporting varies by participant characteristics. Researchers have reported that subjects who are female (1, 2), older (2), or overweight (3–12) can be predicted to underreport energy intake. Other traits hypothesized to influence the accuracy of reporting include income (5, 6, 10), education (2, 10, 13–15), social desirability (14, 16), body image (14), history of dieting or restrained eating (5), and depression (9).

To identify factors associated with dietary energy underreporting, it is necessary to have an objective measure of energy intake. In weight-stable participants, total energy expenditure (TEE) is used as a proxy for energy intake (17, 18), and the doubly labeled water method is generally considered the gold standard for assessment of TEE (19). However, that method requires the costly oxygen-18 isotope and isotope ratio mass spectrometry, which limits its use in large-scale epidemiologic studies. Predictive equations for basal metabolic rate (based on sex, age, weight, or height or on all of these factors) can be used to identify persons whose self-reported energy intakes fall below some physiologically plausible cutoff (2, 6, 10, 17). This method is easy and inexpensive, but predictive equations may not represent some population subgroups (20), include no measure of activity-related energy expenditure (AREE), and identify only extremes of reporting error (18, 21).

Here we use the factorial approach to measure TEE for comparison to self-reported energy intake, as estimated with the use of an FFQ. Factorial determination of objective energy intake involves measuring or estimating the components of TEE: resting metabolic rate (RMR), thermic effect of food, and AREE. Specifically, our objective was to examine the magnitude, direction, and predictors of energy intake underreporting on the FFQ used in the Women’s Health Initiative (WHI) Clinical Trial and Observational Study, a nationwide study of health and disease risk in > 160 000 postmenopausal women (21). An understanding of the participant characteristics associated with errors in self-reported dietary intake is critical for the interpretation of diet-disease relations in analyses that rely on self-reported dietary intake, and the development of a comprehensive measurement-error model has the potential to increase the validity and reliability of diet-disease association analysis.


   SUBJECTS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Study overview
We recruited 102 postmenopausal women aged 50–79 y from the Seattle area through direct mailings by using the Washington State Department of Licensure list, flyers, and newspaper advertisements. Interested women contacted us by telephone, and we provided an overview of the study. Potential participants were screened for conditions that might interfere with body-composition calculations or biomarker measurements: bowel disease, diabetes or hypoglycemia, kidney disease, chronic lung disease, liver disease, claustrophobia, incontinence, systemic steroid medication use, weight change > 4.5 kg in the 2 mo before enrollment, and alcohol intake > 2 servings/d (24 g). None of the women enrolled in this study were participating in the WHI study.

Participants completed questionnaires and collected their urine for 24 h before each of 2 clinic visits scheduled 1 wk apart. At the visits, we received the 24-h urine collections, reviewed completed questionnaires, took body measurements, conducted indirect calorimetry, and drew blood. To measure activity, participants wore an accelerometer for 3 consecutive days during the week between clinic visits. The Institutional Review Board at the Fred Hutchinson Cancer Research Center approved all procedures.

Energy intake reporting on the FFQ
Before the clinic visits, participants received the WHI FFQ in the mail (visit 1) and in person (visit 2). Both FFQs were completed at home by the participants without assistance. A registered dietitian checked the FFQs for completeness at each visit. The WHI FFQ asks about usual dietary intake during the previous 3 mo and contains 3 sections: 1) 19 adjustment questions that are used in the analysis software to calculate the nutrient content of specific food items, 2) a listing of 122 foods or food groups with questions about the usual frequency of intake and portion size for each entry, and 3) 4 summary questions about the usual intake of fruit, vegetables, and fat added to foods and used in cooking. We analyzed the FFQ dietary data by using the University of Minnesota Nutrition Coordinating Center Nutrient Database (22), and the algorithms for analysis are described elsewhere (23). Compared with food records and recalls in a study of 113 WHI study participants, the completed WHI FFQs underestimated energy intakes by {approx}8% and correlated with energy estimates from records or recalls at 0.4 (24). However, the WHI FFQ has not been validated against the doubly labeled water method.

Total energy expenditure
We objectively estimated energy intake by summing the components of TEE (RMR and AREE) with a 10% addition to this summation for the thermic effect of food (25). Below we provide details of our protocol and measures.

Resting metabolic rate
We measured RMR with a VMAX 2900 indirect calorimeter (SensorMedics, Loma Linda, CA). After 30 min for machine warm-up, volume calibration was conducted daily to <= 3% of 3 L according to the manufacturer’s recommendations. Gas calibrations were done before each subject was measured, with the use of 2 mixtures: 26% O2 and 0% CO2 and 20% O2 and 0.75% CO2.

We instructed participants to abstain from food and beverages, except water, for >= 8 h and to avoid strenuous activity for 48 h before each indirect calorimetry measurement. To begin the visit, participants rested quietly on a recliner for 30 min in the thermally neutral testing room. We explained the procedure and oriented participants to the equipment. The mixing chamber pump was turned on, and the plastic canopy was placed over the head and neck of the recumbent participant, with the vinyl skirt covering the torso. For the participants’ acclimation to the apparatus and our adjustment of pump speed, we allowed 2 min of data to expire before initiating formal data collection. We collected data points every 30 s and defined steady state as 10 min during which the volume of oxygen consumed, the minute ventilation, and the respiratory quotient did not vary by > 10%. If 10 min of steady state was achieved by 30 min of data collection, the test was concluded. If not, the test was continued until 10 min of steady state was achieved or to 45 min of data collection, whichever occurred first. Analyses indicated that RMR estimates from the first 5 min of calorimetry were significantly higher than the remainder of the measures. However, RMR estimates from indirect calorimetry segments meeting steady state criteria or from calorimetry extended beyond 30 min did not significantly differ from those at 30 min (26). Therefore, we used the 5–30-min segment of calorimetry measurements for our calculation of RMR and the mean of the 2 RMR measures in our estimation of TEE.

Activity assessment
We measured AREE with a uniaxial accelerometer (Caltrac; Muscle Dynamics Fitness Network, Torrance, CA), a 7-cm2 unit that is worn at the waist and that measures vertical accelerations of the body’s center of gravity. When movement occurs, a cantilevered beam in the monitor bends and emits a current proportional to the force acting on it. A computer in the monitor plots an acceleration curve and uses the area under the curve for the estimation of activity (27).

We instructed participants to maintain their usual activities while wearing the monitor during waking hours for 3 consecutive days, including one weekend day. We input female sex and each woman’s height, weight, and age into her unit at visit 1. The participants cleared the screen to zero and attached the unit to their waistbands or belts under the right arm on arising each day. As needed, they used weightlifting and pedal modes for stationary, isotonic activities and for bicycling or rowing, which switched the accelerometer settings to prediction equations that more closely estimated the energy expended for these activities. Before retiring, participants recorded the readings in the "CALS USED ACTM" screen reflecting AREE only. Water activities precluded the wearing of the unit. The energy expenditure for these activities, though rare (n = 9 of 278 d), was added to the day’s monitor total (28). We included data if the accelerometer was worn for >= 22 waking h/d collectively and if the participant followed protocol instructions. We used the mean of the accelerometer measurements for 2–3 d for our estimate of TEE. Each participant completed the Physical Activity Scale for the Elderly (PASE) questionnaire before visit 1 (29), and scores were calculated for comparison with accelerometer data. The PASE questionnaire assesses the frequency and duration of low-, moderate-, and high-intensity activities over the previous week, and it has been used in studies of physical activity levels in older adults (30). The scale gives a total score that can range from 0 to > 400, with higher scores representing higher activity levels.

Subject characteristics
Anthropometry
We measured height, weight, and waist and hip circumferences at each clinic visit. We estimated body composition by using urinary creatinine concentrations from two 24-h urine collections 1 wk apart. Participants documented collection times and completeness, stored their urine at <= 4.4 °C in a refrigerator or cooler at home, and submitted collections at each clinic visit. Urinary creatinine concentrations were determined with a kinetic modification of the Jaffé alkaline picrate-reaction procedure using a Cobas Mira Plus Analyzer (Roche Diagnostics, Brandburg, NJ) according to the manufacturer’s instructions. The interassay CVs for low, medium, and high urine quality-control pool levels were 1.2%, 1.6%, and 1.6%, respectively. Daily creatinine excretions were excluded if they were < 780 mg/d or if urine was collected for < 23 h or > 25 h (31). We were unable to estimate body composition for 7 women because of urine collections that were incomplete, as calculated by Welles’s formula for older adults (32). The percentage of body fat (%BF) was derived from the fat-free mass (FFM) (weight – FFM = kg fat; kg fat/weight x 100% = %BF).

Psychosocial measures
We administered the Crowne-Marlowe Social Desirability Scale during the second visit (33). This is a 33-item, true-false questionnaire that measures a person’s tendency to provide the most socially desirable answer regardless of the truth. The tool discriminates between high total scores, representing a strong tendency to choose the socially desirable answer, and low total scores, representing a weaker tendency to do so, at a level of P = 0.05 or better. The scale has an internal consistency coefficient of 0.88 (Cronbach’s alpha) (14, 33).

We used the Stunkard-Sorensen silhouettes to assess the participants’ perceptions of body size (34). This instrument shows 9 silhouettes from thin to large, which are given a score of 1 to 9, respectively. Participants were asked to identify the silhouettes that best represented their perceived body size, the desired body size, and the healthiest body size. Dissatisfaction with body size was calculated as perceived body size minus desired body size. We also calculated the difference between current body size and the body size perceived to be the healthiest to determine its influence on energy reporting.

Other measures
We collected data on age, race, education, marital status, and household income from a self-administered questionnaire.

Statistical analyses
We present descriptive data ( ± SD) for our measures of energy expenditure, the energy intake reported on the FFQ (FFQ energy), and energy underreporting (FFQ energy - TEE), as well as Pearson correlation coefficients for multiple measures. To provide a general assessment of our factorial approach to estimating TEE, we present associations with participant characteristics that should influence TEE [age, body mass index (BMI; in kg/m2), FFM, %BF] and with characteristics that should not influence TEE [waist-to-hip ratio (WHR) and social desirability].

We used multiple linear regression to model the effects of participant characteristics on energy underreporting. The dependent variables were ratio measures of energy reporting. Specifically, we calculated a ratio of FFQ energy divided by TEE for our primary measure of energy reporting error. For analyses, these ratio variables were naturally log transformed to yield approximate normality, and the back-transformed means for FFQ energy divided by TEE are given in tables in the book by Fleiss (35). Independent variables were sociodemographic characteristics (age, income, and education), adiposity-related measures (BMI, FFM, %BF, and WHR), and psychosocial factors (social desirability, perceived body size, dissatisfaction with body size, and disparity from healthy size). To ensure the stability of our estimates, we divided our independent variables into tertiles, except in cases where meaningful categories existed, as follows. Age was divided into decades (50–59, 60–69, and 70–79 y), and education was grouped by level of attainment (high school or some college, bachelor’s degree, and advanced degree). BMI was categorized by using cutoffs from a 1998 consensus conference of the Expert Panel on the Identification, Evaluation, and Treatment of Overweight in Adults: < 25, 25–29.9, and >= 30 (36). A WHR > 0.80 is associated with increased health risks in women, and 0.80 was used as an upper cutoff (37).

For purposes of comparison, we also examined the ratio of FFQ energy to RMR. Linear regression models using demographic characteristics and adiposity-related measures were controlled for age, education, and household income. Models using psychosocial factors were controlled for age, education, household income, and BMI. We used a linear contrast to test for trends across ordered categories. Pearson correlation coefficients were calculated for FFQ energy reported and TEE measured, and partial correlations were calculated as for the regression models. The z scores were used to test for differences between coefficients and for trends (38). PASE score tertiles were compared with average accelerometer measures. Statistical analyses used SAS version 6.12 (SAS Institute, Inc, Cary, NC). Trends were assigned significance if P < 0.05.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
This sample was 94% white, and 60% of the subjects were married or living as married. Sixty-two percent held college degrees, and 69% of those providing income information (n = 95) had household incomes of >= $35 000. Participants had a mean (± SD) BMI of 26.4 ± 4.8, %BF estimates of 30 ± 10%, FFM of 47.8 ± 4.7, and WHR of 0.78 ± 0.07.

The mean values and within-subject Pearson correlation coefficients for our measures of energy expenditure, FFQ energy, and energy underreporting are shown in Table 1Go. Duplicate indirect calorimetry measures of RMR were highly correlated (r = 0.92), as were the first and last days of AREE estimates measured on the accelerometer (r = 0.70). Analyses of the accelerometer data using paired t tests found no significant differences in AREE estimates between weekends and weekdays or between consecutive and nonconsecutive days of data collection (data not shown). FFQ energy at visit 1 and at visit 2 was also highly correlated (r = 0.79), and a two-tailed t test found no significant differences between the 2 measures.


View this table:
[in this window]
[in a new window]
  		TABLE 1 Measures of energy expenditure and self-reported energy intake in healthy postmenopausal women aged 50–79 y1
 
The distribution of estimates of energy-reporting errors identified in this group of 102 healthy postmenopausal women is shown in Figure 1Go. Thirty-six percent of our sample (n = 37) had underreported energy estimates of >= 2400 kJ (574 kcal) according to a comparison of FFQ and TEE estimates. The median for energy underreporting estimates in this sample was 20.8%, ranging from 80% underreporting to 140% overreporting (data not shown).



View larger version (20K):
[in this window]
[in a new window]
  		FIGURE 1. Distribution of energy-reporting error in 102 healthy postmenopausal women aged 50–79 y, expressed as energy intake (in kJ and kcal) from a food-frequency questionnaire minus total energy expenditure (resting metabolic rate + activity-related energy expenditure, with 10% adjustment for thermic effect of food). Negative values represent underreporting.

 
Correlation coefficients among various measures of energy expenditure, FFQ energy, and energy underreporting are shown in Table 2Go. TEE was highly correlated with RMR (r = 0.73) and AREE (r = 0.85), as expected, given that TEE is derived from these measures. However, AREE correlated with RMR only at r = 0.26. FFQ energy was only modestly correlated with the objective measures of energy expenditure, but strongly correlated with energy underreporting. That is, as FFQ energy intake increased, energy underreporting decreased (r = 0.81).


View this table:
[in this window]
[in a new window]
  		TABLE 2 Pearson correlation coefficients between measures of energy expenditure and self-reported energy intake in 102 healthy postmenopausal women1
 
To informally assess the validity of our factorial approach to estimating TEE, we present associations of TEE with participant characteristics that should influence TEE (age, BMI, FFM, and %BF) and those that should not influence TEE (WHR and social desirability) (Table 3Go). As expected, TEE significantly decreased with age (P for trend = 0.0005) and increased with BMI (P for trend = 0.0001), FFM (P for trend = 0.0001), and %BF (P for trend = 0.054). No significant differences were evident by WHR or social desirability. Other characteristics having no significant effect on TEE were education, annual income, and all measures of perceived body size (data not shown).


View this table:
[in this window]
[in a new window]
  		TABLE 3 Total energy expenditure by subject characteristics in 102 healthy postmenopausal women aged 50–791
 
Data on energy underreporting (FFQ energy divided by TEE) by age, education, household income, and adiposity are given in Table 4Go. A ratio of 1 would indicate a lack of systematic under- or overreporting for the group of women, whereas a ratio < 1 would indicate underreporting. There was a suggestion of greater underreporting (eg, smaller FFQ energy-to-TEE ratios) with younger age (P for trend = 0.07). Among the adiposity-related factors, BMI was not associated with underreporting of energy intake. The use of other BMI categories such as tertiles or those given in the Surgeon General’s Report on Nutrition and Health (39) also showed no associations with underreporting. Underreporting may increase with increasing FFM (P for trend = 0.11), although this trend was not significant. There were no significant differences in reporting error between participants who were married or living as married and all others (data not shown).


View this table:
[in this window]
[in a new window]
  		TABLE 4 Ratios of self-reported energy intake on a food-frequency questionnaire (FFQ energy) to total energy expenditure (TEE) and correlation coefficients by demographic characteristics and adiposity-related factors in 102 healthy postmenopausal women aged 50–79 y1
 
Partial correlation coefficients further illustrating how FFQ energy compares with measured TEE according to participant characteristics are also given in Table 4Go. There were significantly higher correlation coefficients between FFQ energy and TEE for women with less education (P for trend = 0.04), higher BMI (P for trend = 0.06), and greater FFM (P for trend = 0.03).

Because the accelerometer measure of AREE is novel and vulnerable to bias (participants were aware that we were monitoring their activity levels), we also examined the ratio of FFQ energy to RMR for purposes of comparison (data not shown). Overall, the associations of participant characteristics with ratios of FFQ energy to RMR were similar to those observed for ratios of FFQ energy to TEE, with P values differing only slightly. Specifically, for this outcome, the negative association of educational attainment and underreporting was suggestive (P for trend = 0.06), whereas trends in age (P for trend = 0.17) and FFM (P for trend = 0.20) were not. PASE scores were positively correlated with average accelerometer measures (r = 0.25, P < 0.01). PASE score tertiles (ie, low, medium, and high activity levels) corresponded to accelerometer means of 1723 kJ (412 kcal), 2029 kJ (485 kcal), and 2142 kJ (512 kcal), respectively, with a significant test for trend (P < 0.01).

Data on energy underreporting by psychosocial factors are given in Table 5Go. Women with high social desirability scores tended to be more likely to underreport energy intakes (ie, have lower FFQ energy-to-TEE ratios) than women with lower scores (P for trend = 0.09) (33). Women perceiving they were thin according to the Stunkard-Sorensen silhouettes (34) tended to underreport their energy intake more than women perceiving themselves as heavy (P for trend = 0.11). Again, similar trends resulted when ratios of FFQ energy to RMR were used, with a significant association between underreporting energy and social desirability (P for trend = 0.04) (data not shown). A suggestive trend in energy underreporting was seen for dissatisfaction with perceived body size (perceived size - desired size) by use of the FFQ energy:RMR outcome variable (P for trend = 0.09); that is, women closer to their desired size underreported energy more.


View this table:
[in this window]
[in a new window]
  		TABLE 5 Comparison of self-reported energy intake on a food-frequency questionnaire (FFQ energy) to total energy expenditure (TEE) and resting metabolic rate by psychosocial factors in 102 healthy postmenopausal women aged 50–79 y1
 
Partial correlation coefficients further illustrating how reported FFQ energy compares with measured TEE according to psychosocial characteristics are given in Table 5Go. There was a higher correlation coefficient between FFQ energy and TEE for women with higher social desirability (P for trend = 0.03), as well as a suggestive association for higher dissatisfaction with perceived body size (P for trend = 0.10).


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
These data indicate that healthy postmenopausal women underreported their energy intake by {approx}21% on the WHI FFQ. This degree of underreporting is consistent with that seen in doubly labeled water method studies (1, 40–43). Overall, the most important finding is that energy underreporting appeared to vary with participant characteristics, and evidence suggested more underreporting among women who were younger or had high social desirability scores. Johansson et al (2) reported an increase in energy underreporting with increasing age in a younger group (age 44 ± 9 y). However, studies of reporting errors in older women such as those in our sample found no significant age differences (1, 11, 15, 44–46). Our results may vary from those of others because of cultural and age differences among populations studied (47, 48), limited sample sizes, or differences in the measures of energy expenditure and energy intake (ie, the FFQ) (49).

We found no association between BMI and energy underreporting such as was found in numerous other studies (3, 4, 6–12), An interesting point regarding the association of adiposity with underreporting is that there is no plausible biologic reason that excess body fat would, in and of itself, cause women to underreport energy intake. Therefore, the various measures of body size and adiposity must be serving as surrogates for psychosocial characteristics that result in underreporting energy, such as poor awareness of intake or portion sizes, deliberate underreporting, and subconscious biasing toward intake that is perceived to be appropriate. We did observe several suggestive associations of psychosocial factors with underreporting. Postmenopausal women with high social desirability scores, which theoretically reflect a desire to answer questions in a socially desirable manner, underreported energy intake more than women with lower scores (P for trend = 0.09); this finding is consistent with findings in other studies (14, 16). We found that postmenopausal women perceiving themselves to be thin underreported their energy intake more than those perceiving themselves to be heavy, even when we controlled for actual BMI (P for trend = 0.11). These findings support the idea that body image, independent of measured BMI, is an important predictor of energy-reporting errors.

This study has many limitations. In particular, the modest sample size limited our ability to detect many associations at conventional significance levels. The generalizability of our results is limited because we included only postmenopausal women from the United States. Although the correlation coefficient for reported FFQ energy and measured TEE (r = 0.24) was similar to the findings of others who used different tools in different populations (all: r = 0.22, P = 0.20) (13, 50), our measure of TEE was likely limited by >= 3 features of the AREE measurement. First, although we specifically instructed women to refrain from strenuous activity for 48 h before the indirect calorimetry measurements, 20 women (19.6%) recorded accelerometer data between 36 and 48 h beforehand. Second, the direction of the measurement error inherent in the accelerometer is unclear. AREE may be underestimated because the accelerometer measures movement in only one plane, does not account for fidgeting or anxiety-related energy expenditures, and is unable to discern variations in grade (51). Third, these participants did not wear the monitor during sleeping hours. One study that compared physical activity estimates determined with this accelerometer to those determined with indirect calorimetry and the doubly labeled water method found that the accelerometer underestimated by 55% in 45–84-y-old women (n = 35) (52). Conversely, another study found that this accelerometer overestimated activity > 2 mph in women aged 71.2 ± 3.5 y (51). Others described this accelerometer as unable to adequately discriminate between running speeds of 5–8 mph (53). Finally, although physical activity is theoretically an objective measure, participants may have altered their usual physical activity levels in response to wearing an accelerometer (54). RMR alone accounted for 60–70% of TEE (quartiles 1 and 3), and, when we examined the associations of participant characteristics with the ratio of FFQ energy to RMR (which excludes the accelerometer component), trends were almost identical to those seen for the ratio of FFQ energy to TEE, which indicates that errors in the accelerometer measures cannot entirely explain our findings. Nonetheless, one could postulate that errors in this measure of physical activity were associated with participant characteristics (eg, age and BMI) and could influence our results.

The ability to establish relative ranking of dietary energy intake from low to high was affected by participant characteristics, as evidenced by the correlation coefficients between FFQ energy and TEE. For example, dietary energy intake was ranked with more precision in relation to the objective TEE marker among women with lower levels of education, higher income, and greater BMI, social desirability, or dissatisfaction with perceived body size. These observations are limited by small sample sizes, and we know of no similar data in the literature with which to compare our findings.

This study supports the hypothesis that types of errors in dietary energy self-reporting vary by participant characteristics. These errors, when associated with dietary exposures or confounding factors of interest, could result in incorrect inferences regarding diet-disease relations. Studies relying on self-reported dietary intake to assess exposure to dietary components should include objective measures of energy intake (energy expenditure) in at least a subset of participants to identify sources of systematic error within the particular population studied and in the specific self-reporting tool used. The objective measure of energy expenditure in a subset of the study cohort should follow the self-reported intake. These data will allow an assessment of systematic bias for the self-reporting tool in the context of the study population and can be used to calibrate the self-reported data as an integral aspect of diet and disease association analyses. Future objective-measures substudies are needed that have the following characteristics: 1) larger sample sizes, 2) more-diverse populations, 3) additional biomarkers that could assist in determining whether underreporting varies by macronutrient source, and 4) a wider array of psychosocial measures to elucidate the observation that adiposity is associated with underreporting. Identification of characteristics associated with energy-reporting errors can lay the groundwork for the development of statistical methods equipped to adjust for person-specific systematic error in dietary self-reporting.


   ACKNOWLEDGMENTS
 
We express our sincerest thanks to Maggie T Grate and Tamsen Johnston for their assistance with participant visits and materials.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. Johnson RK, Goran MI, Poehlman ET. Correlates of over- and underreporting of energy intake in healthy older men and women. Am J Clin Nutr 1994;59:1286–90.[Abstract/Free Full Text]
  2. Johansson L, Solvoll K, Bjorneboe GE, Drevon CA. Under- and overreporting of energy intake related to weight status and lifestyle in a nationwide sample. Am J Clin Nutr 1998;68:266–74.[Abstract]
  3. Heitmann BL, Lissner L. Dietary underreporting by obese individuals—is it specific or non-specific? BMJ 1995;311:986–9.[Abstract/Free Full Text]
  4. Ballard-Barbash R, Graubard I, Krebs-Smith SM, Schatzkin A, Thompson FE. Contribution of dieting to the inverse association between energy intake and body mass index. Eur J Clin Nutr 1996;50:98–106.[Medline]
  5. Lafay L, Basdevant A, Charles MA, et al. Determinants and nature of dietary underreporting in a free-living population: the Fleurbaix Laventie Ville Sante (FLVS) Study. Int J Obes Relat Metab Disord 1997;21:567–73.[Medline]
  6. Stallone DD, Brunner EJ, Bingham SA, Marmot MG. Dietary assessment in Whitehall II: The influence of reporting bias on apparent socioeconomic variation in nutrient intakes. Eur J Clin Nutr 1997;51:815–25.[Medline]
  7. Heerstrass DW, Ocke MC, Bueno-de-Mesquita HB, Peeters PH, Seidell JC. Underreporting of energy, protein and potassium intake in relation to body mass index. Int J Epidemiol 1998;27:186–93.[Abstract/Free Full Text]
  8. Braam LA, Ocke MC, Bueno-de-Mesquita HB, Seidell JC. Determinants of obesity-related underreporting of energy intake. Am J Epidemiol 1998;147:1081–6.[Abstract/Free Full Text]
  9. Kretsch MJ, Fong AKH, Green MW. Behavioral and body size correlates of energy intake underreporting by obese and normal-weight women. J Am Diet Assoc 1999;99:300–6.[Medline]
  10. Pomerleau J, Ostbye T, Bright-See E. Potential underreporting of energy intake in the Ontario Health Survey and its relationship with nutrient and food intakes. Eur J Epidemiol 1999;15:553–7.[Medline]
  11. Samaras K, Kelly PJ, Campbell LV. Dietary underreporting is prevalent in middle-aged British women and is not related to adiposity (percentage body fat). Int J Obes Relat Metab Disord 1999;23:881–8.[Medline]
  12. Tomoyasu NJ, Toth MJ, Poehlman ET. Misreporting of total energy intake in older men and women. J Am Geriatr Soc 1999;47:710–5.[Medline]
  13. Johnson RK, Soultanakis RP, Matthews DE. Literacy and body fatness are associated with underreporting of energy intake in US low-income women using the multiple-pass 24-hour recall: a doubly labeled water study. J Am Diet Assoc 1998;98:1136–40.[Medline]
  14. Taren DL, Tobar M, Hill A, et al. The association of energy intake bias with psychological scores of women. Eur J Clin Nutr 1999;53:570–8.[Medline]
  15. Cook A, Pryer J, Shetty P. The problem of accuracy in dietary surveys. Analysis of the over 65 UK National Diet and Nutrition Survey. J Epidemiol Community Health 2000;54:611–6.[Abstract/Free Full Text]
  16. Hebert JR, Clemow L, Pbert L, Ockene IS, Ockene JK. Social desirability bias in dietary self-report may compromise the validity of dietary intake measures. Int J Epidemiol 1995;24:389–98.[Abstract/Free Full Text]
  17. World Health Organization, Food and Agriculture Organization, United Nations University. Energy and protein requirements. Geneva: WHO, 1985.
  18. Starling RD, Toth MJ, Carpenter WH, Matthews DE, Poehlman ET. Energy requirements and physical activity in free-living older women and men: a doubly labeled water study. J Appl Physiol 1998;85:1063–9.[Abstract/Free Full Text]
  19. Schoeller DA, Shay K, Kushner RF. How much physical activity is needed to minimize weight gain in previously obese women? Am J Clin Nutr 1997;66:551–6.[Abstract/Free Full Text]
  20. Arciero PJ, Goran MI, Gardner AM, Ades PA, Tyzbir RS, Poehlman ET. A practical equation to predict resting metabolic rate in older females. J Am Geriatr Soc 1993;41:389–95.[Medline]
  21. Anderson G, Manson J, Wallace R, et al. Implementation of the WHI study design. Ann Epidemiol (in press).
  22. Schakel S, Buzzard I, Gebhardt S. Procedures for estimating nutrient values for food composition databases. Food Comp Anal 1997;10:102–14.
  23. Kristal A, Shattuck A, Williams AE. Food frequency questionnaires for diet intervention research. 17th National Nutrient Databank Conference Proceedings. Baltimore: International Life Sciences Institute, 1992:110–25.
  24. Patterson RE, Kristal AR, Tinker LF, Carter RA, Bolton MP, Agurs-Collins T. Measurement characteristics of the WHI food-frequency questionnaire. Ann Epidemiol 1999;9:178–87.[Medline]
  25. Poehlman ET, Melby CL, Badylak SF. Relation of age and physical exercise status on metabolic rate in younger and older healthy men. J Gerontol 1991;46:B54–8.[Medline]
  26. Horner NK, Lampe JW, Patterson RE, Neuhouser ML, Beresford SA, Prentice RL. Indirect calorimetry protocol development for measuring resting metabolic rate as a component of total energy expenditure in free-living postmenopausal women. J Nutr 2001;131:2215–8.[Abstract/Free Full Text]
  27. Vinken AG, Bathalon GP, Sawaya AL, Dallal GE, Tucker KL, Roberts SB. Equations for predicting the energy requirements of healthy adults aged 18–81 y. Am J Clin Nutr 1999;69:920–6.[Abstract/Free Full Text]
  28. Ainsworth BE, Haskell WL, Leon AS, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;25:71–80.[Medline]
  29. Washburn RA, Smith KW, Jette AM, Janney CA. The physical activity scale for the elderly (PASE): development and evaluation. J Clin Epidemiol 1993;46:153–62.[Medline]
  30. Schuit AJ, Schouten EG, Westerterp KR, Saris WH. Validity of the Physical Activity Scale for the Elderly (PASE): according to energy expenditure assessed by the doubly labeled water method. J Clin Epidemiol 1997;50:541–6.[Medline]
  31. Forbes GB. Human body composition. New York: Springer-Verlag, 1987.
  32. Welle S, Thornton C, Totterman S, Forbes G. Utility of creatinine excretion in body-composition studies of healthy men and women older than 60 y. Am J Clin Nutr 1996;63:151–6.[Abstract/Free Full Text]
  33. Crowne DP, Marlowe D. A new scale of social desirability independent of psychopathology. J Consult Psychol 1960;24:349–54.[Medline]
  34. Stunkard AJ, Sorensen T, Schulsinger F. Use of the Danish adoption register for the study of obesity and thinness. Res Publ Assoc Res Nerv Ment Dis 1983;60:115–20.[Medline]
  35. Fleiss JL. Statistical methods for rates and proportions. 2nd ed. New York: John Wiley & Sons, 1981.
  36. The Expert Panel on the Identification, Evaluation, and Treatment of Overweight in Adults. Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults: executive summary. Am J Clin Nutr 1998;68:899–917.[Medline]
  37. Groff JL, Gropper SS, Hunt SM. Body composition. In: Groff JL, Gropper SS, Hunt SM, eds. Advanced nutrition and human metabolism. 2nd ed. St. Paul: West, 1995:441–64.
  38. Zar JH. Biostatistical analysis. 2nd ed. Englewood Cliffs, NJ: Prentice-Hall, Inc, 1984:318.
  39. US Surgeon General. The Surgeon General’s report on nutrition and health. Washington, DC: US Department of Health and Human Services, 1988.
  40. Schoeller DA, Bandini LG, Dietz WH. Inaccuracies in self-reported intake identified by comparison with the doubly labelled water method. Can J Physiol Pharmacol 1990;68:941–9.[Medline]
  41. Sawaya A, Tucker K, Tsay R, et al. Evaluation of four methods for determining energy intake in young and older women: comparison with doubly labeled water measurements of total energy expenditure. Am J Clin Nutr 1996;63:491–9.[Abstract/Free Full Text]
  42. Martin L, Su W, Jones PJ, Lockwood GA, Tritchler DL, Boyd NF. Comparison of energy intakes determined by food records and doubly labeled water in women participating in a dietary-intervention trial. Am J Clin Nutr 1996;63:483–90.[Abstract/Free Full Text]
  43. Kroke A, Klipstein-Grobusch K, Voss S, et al. Validation of a self-administered food-frequency questionnaire administered in the European Prospective Investigation into Cancer and Nutrition (EPIC) Study: comparison of energy, protein, and macronutrient intakes estimated with the doubly labeled water, urinary nitrogen, and repeated 24-h dietary recall methods. Am J Clin Nutr 1999;70:439–47.[Abstract/Free Full Text]
  44. Goran MI, Poehlman RT. Total energy expenditure and energy requirements in healthy elderly persons. Metabolism 1992;41:744–53.[Medline]
  45. Kaczkowski CH, Jones PJH, Feng J, Bayley HS. Four-day multimedia diet records underestimate energy needs in middle-aged and elderly women as determined by doubly labeled water. J Nutr 2000;130:802–5.[Abstract/Free Full Text]
  46. Tomoyasu NJ, Toth MJ, Poehlman ET. Misreporting of total energy intake in older African Americans. Int J Obes Relat Metab Disord 2000;24:20–6.[Medline]
  47. Mennen LI, Jackson M, Cade J, et al. Underreporting of energy intake in four populations of African origin. Int J Obes Relat Metab Disord 2000;24:882–7.[Medline]
  48. Harrison GG, Galal OM, Ibrahim N, et al. Underreporting of food intake by dietary recall is not universal: a comparison of data from Egyptian and American women. J Nutr 2000;130:2049–54.[Abstract/Free Full Text]
  49. Kortzinger I, Bierwag A, Mast M, Muller MJ. Dietary underreporting: validity of dietary measurements of energy intake using a 7-day dietary record and a diet history in non-obese subjects. Ann Nutr Metab 1997;41:37–44.[Medline]
  50. Goris AH, Westerterp KR. Underreporting of habitual food intake is explained by undereating in highly motivated lean women. J Nutr 1999;129:878–82.[Abstract/Free Full Text]
  51. Fehling P, Smith D, Warner S, Dalsky G. Comparison of accelerometers with oxygen consumption in older adults during exercise. Med Sci Sports Exerc 1999;31:171–5.[Medline]
  52. Starling RD, Matthews DE, Ades PA, Poehlman ET. Assessment of physical activity in older individuals: a doubly labeled water study. J Appl Physiol 1999;86:2090–6.[Abstract/Free Full Text]
  53. Haymes EM, Byrnes WC. Walking and running energy expenditure estimated by Caltrac and indirect calorimetry. Med Sci Sports Exerc 1993;25:1365–9.[Medline]
  54. Montoya HJ, Kemper HCG, Saris WHM, Washburn RA. Measuring physical activity and energy expenditure. Champaign, IL: Human Kinetics, 1996.
Received for publication May 9, 2001. Accepted for publication November 7, 2001.





This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Horner, N. K
Right arrow Articles by Prentice, R. L
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Horner, N. K
Right arrow Articles by Prentice, R. L
Agricola
Right arrow Articles by Horner, N. K
Right arrow Articles by Prentice, R. L

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS