To evaluate under- and overreporting and their determinants in the EPIC 24-hour diet recall (24-HDR) measurements collected in the European Prospective Investigation into Cancer and Nutrition (EPIC).

Cross-sectional analysis. 24-HDR measurements were obtained by means of a standardised computerised interview program (EPIC-SOFT). The ratio of reported energy intake (EI) to estimated basal metabolic rate (BMR) was used to ascertain the magnitude, impact and determinants of misreporting. Goldberg's cut-off points were used to identify participants with physiologically extreme low or high energy intake. At the aggregate level the value of 1.55 for physical activity level (PAL) was chosen as reference. At the individual level we used multivariate statistical techniques to identify factors that could explain EI/BMR variability. Analyses were performed by adjusting for weight, height, age at recall, special diet, smoking status, day of recall (weekday vs. weekend day) and physical activity.

Twenty-seven redefined centres in the 10 countries participating in the EPIC project.

In total, 35955 men and women, aged 35–74 years, participating in the nested EPIC calibration sub-studies.

While overreporting has only a minor impact, the percentage of subjects identified as extreme underreporters was 13.8% and 10.3% in women and men, respectively. Mean EI/BMR values in men and women were 1.44 and 1.36 including all subjects, and 1.50 and 1.44 after exclusion of misreporters. After exclusion of misreporters, adjusted EI/BMR means were consistently less than 10% different from the expected value of 1.55 for PAL (except for women in Greece and in the UK), with overall differences equal to 4.0% and 7.4% for men and women, respectively. We modelled the probability of being an underreporter in association with several individual characteristics. After adjustment for age, height, special diet, smoking status, day of recall and physical activity at work, logistic regression analyses resulted in an odds ratio (OR) of being an underreporter for the highest vs. the lowest quartile of body mass index (BMI) of 3-52 (95% confidence interval (CD 2.91–4.26) in men and 4.80 (95% CI 4.11–5.6l) in women, indicating that overweight subjects are significantly more likely to underestimate energy intake than subjects in the bottom BMI category. Older people were less likely to underestimate energy intake: ORs were 0.58 (95% CI 0.45–0.77) and 0.74 (95% CI 0.63–0.88) for age (≥ 65 years vs. < 50 years). Special diet and day of the week showed strong effects.

EI tends to be underestimated in the vast majority of the EPIC centres, although to varying degrees; at the aggregate level most centres were below the expected reference value of 1.55. Underreporting seems to be more prevalent among women than men in the EPIC calibration sample. The hypothesis that BMI (or weight) and age are causally related to underreporting seems to be confirmed in the present work. This introduces further complexity in the within-group (centre or country) and between-group calibration of dietary questionnaire measurements to deattenuate the diet—disease relationship.

To examine relationships between the ratio of energy intake to basal metabolic rate (EI/BMR) and age and body mass index (BMI) among Japanese adults.

Energy intake was assessed by 4-day semi-weighed diet records in each of four seasons (16 days in total). The EI/BMR ratio was calculated from reported energy intake and estimated basal metabolic rate as an indicator of reporting accuracy.

Residents in three areas in Japan, namely Osaka (urban), Nagano (rural inland) and Tottori (rural coastal).

One hundred and eighty-three healthy Japanese men and women aged ≥30 years.

The oldest age group (≥60 years) had higher EI/BMR values than the youngest age group (30–39 years) in both sexes (1.74 vs. 1.37 for men; 1.65 vs. 1.43 for women). In multiple regression analyses, age correlated positively (partial correlation coefficient, β = 0.012, P < 0.001 for men; β = 0.011, P < 0.001 for women) and BMI correlated negatively (β = −0.031, P < 0.001 for men; β = −0.025, P < 0.01 for women) with EI/BMR.

Age and BMI may influence the relative accuracy of energy intake among Japanese adults.

The aims of the present study were (1) to evaluate the degree to which underreporting of energy intake by repeated 24-hour recalls was related to gender, age, weight status, day of interview, educational level, smoking habits and area of living, and (2) to compare the dietary characteristics of underreporters with those of others.

Cross-sectional study. Ten 24-hour recalls were performed during a one-year period.

The Västerbotten intervention programme of cardiovascular disease and diabetes in Northern Sweden.

Ninety-four men and 99 women in four age groups: 30, 40, 50 and 60 years.

The prevalence of men and women with a food intake level (FIL; reported energy intake divided by estimated basal metabolic rate) below 1.2 was 44% and 47%, respectively. The youngest age group had higher FIL values than the oldest age group for both men (1.5 versus 1.1) and women (1.4 versus 1.1). The prevalence and magnitude of underreporting were directly related to body mass index (BMI; correlation coefficient: -0.47 (men) and -0.55 (women)). Smokers had a lower FIL value (1.1) than non-smokers (1.3). The nutrient density was lower for the group with high FIL values for protein and calcium and higher for fat and sucrose. The upper FIL group often had higher intake frequencies and larger portion sizes than the lower FIL group.

Underreporting of energy intake is prevalent when 24-hour recalls are used, but the prevalence differs between sub-groups in the population. BMI was the main predictor of underreporting but also old age and smoking seem to contribute in this aspect. Socially desirable food items were not underreported to the same extent as socially undesirable food items. The intake frequencies and portion sizes partly explained the differences in FIL.

To evaluate the ratio of energy intake to basal metabolic rate (EI/BMR) among young female Japanese adults, and to compare the lifestyle and dietary characteristics between relatively low and high reporters.

Dietary intakes were assessed over a 1-month period with a validated, self-administered, diet history questionnaire, and lifestyle variables were assessed by a second questionnaire designed for this survey. The ratio of EI/BMR was calculated from reported energy intake and estimated basal metabolic rate.

In total, 1889 female Japanese university students aged 18–20 years who were enrolled in dietetics courses.

Ninety-five per cent of the subjects were classified into a non-obese group (body mass index (BMI) <25 kg m−2; mean±standard deviation (SD): 20.8±2.6 kg m−2). EI/BMR was 1.43±0.40 (mean±SD). Sixty-eight per cent of the subjects showed an EI/BMR level below the possibly balanced value of 1.56, 37% showed EI/BMR below the minimum survival value of 1.27 and 2% of the subjects showed EI/BMR exceeding the maximum value for a sustainable lifestyle of 2.4. BMI, body weight and BMR decreased significantly with the increase in EI/BMR (P<0.001). The percentage of energy from carbohydrate was significantly higher, whereas those from fat and protein were significantly lower, among the lower EI/BMR groups. As for food groups, a significantly declining trend from the lowest to the highest EI/BMR groups was observed for cereals.

Underreporting, rather than overreporting, of energy intake was predominant in this relatively lean Japanese female population. BMI was the most important factor affecting the reporting accuracy of energy intake.

Energy intake determined from self-reported dietary assessment methods may be underreported. Therefore, it is important that such methods be validated against another with known validity for energy intake or energy expenditure.

We investigated potential underestimation of energy intake obtained from our semi-quantitative food-frequency questionnaire (FFQ) administered between 2000 and 2001 in the metropolitan area of Montreal, Canada. The study population included 246 adults aged 18 to 82 years. The ratio of energy intake to estimated basal metabolic rate (EI/BMR) was used to assess underreporting and physical activity was determined from self-administered questions. Comparison of the EI/BMR ratio with the Goldberg statistical cut-off allowed us to detect individuals who were low energy reporters (LERs). LERs and non-LERs were compared to determine if they differed on sociodemographic, anthropometric and lifestyle variables.

The EI/BMR ratio was 1.26 for men and 1.32 for women. LERs represented 43% of the sample of individuals. Male LERs accounted for 54% compared with 35% among females. Underreporting of energy intake was highest in men and individuals who were older, heavier, with higher body mass index and lower education level. A higher proportion of male LERs perceived their financial situation as adequate while a greater proportion of female LERs considered themselves poor.

Our data suggest that underreporting of energy intake from the FFQ was considerable and may bias dietary interpretation. As this was uneven across the sample, it is crucial to recognise the characteristics of LERs in order to increase the validity of reported energy intake.

To illustrate the effect of common mistakes when using 24-hour national dietary survey data to estimate the prevalence of inadequate nutrient intakes.

Raw data on nutrient intake from the Australian 1995 National Nutrition Survey were adjusted for within-person variance using standard techniques and corrected for underreporting using the criteria of Goldberg et al. The distributions for six nutrients were compared with current dietary reference values from the UK, USA and Australia.

A national sample of the Australian population with a 61.4% response rate.

Adjusting for within-person variance reduced the range of nutrient intakes to 66–80% of the raw data range and the proportion with intakes below the estimated average requirement (EAR) by up to 20%. Excluding underreporters further reduced the proportion below the EAR by up to 10%. Using the dietary reference values from different countries also resulted in some markedly different estimates. For example, the prevalence of low folate intakes ranged from <1 to 92% for adult women depending on the reference used. Except for vitamin A and protein, the prevalence of low intakes was invariably higher for women than for men.

Estimates of the prevalence of low nutrient intakes based on raw 24-hour survey data are invariably misleading. However, even after adjustment for within-person variance and underreporting, estimates of the prevalence of low nutrient intakes may still be misleading unless interpreted in the light of the reference criteria used and supported by relevant biochemical and physiological measures of nutritional status.

Research in industrialised countries has documented a high prevalence of underreported energy intakes associated with characteristics such as obesity. This paper examines the prevalence, patterns and impact of energy under- and overreporting on diet–obesity relationships in a middle-income developing country.

A 70-item food-frequency questionnaire was used. Underreporters had reported energy intakes < 1.35 × basal metabolic rate (BMR), overreporters > 2.4 × BMR. Multinomial models were used to identify characteristics associated with implausible reporting. Intakes were compared across reporting groups to assess evidence of bias. Associations between diet and obesity were compared with and without adjustment for implausible reporting.

Spanish Town, neighbouring the capital city of Kingston, Jamaica.

Eight hundred and ninety-one Jamaican adults, aged 25–75 years, were randomly recruited.

More women than men (38.6% vs. 22.5%) underreported, but more men overreported energy (23.7% vs. 16.0%). Underreporting was positively associated with obesity, special diets, smoking and age; age was inversely associated with overreporting. Underreporters estimated lower energy from potentially socially undesirable food groups (e.g. snacks) and higher intakes of ‘healthy’ foods (e.g. fruit) than did plausible reporters. For some of these food groups, significant differences in intakes among normal-weight versus obese subjects observed among plausible reporters were absent when implausible reporters were included. In models of food group–obesity associations, adjusting for implausible energy yielded more credible results that more closely resembled findings in plausible reporters.

Energy under- and overreporting are highly prevalent in Jamaica. Adjusting for implausible reporting may help to reduce bias in diet–health outcome associations.

To evaluate the ratio of reported energy intake to basal metabolic rate (EI/BMR) among pregnant Indonesian women, as well as identifying risk factors for being an underreporter.

Longitudinal study of dietary intake, using six repeated 24-hour diet recalls each trimester. Basal metabolic rate was estimated from body weight and physical activity from occupation. The lower 95% confidence interval for plausible EI/BMR was calculated and the proportion of underreporters estimated. Risk factors for being an underreporter were assessed in multivariate logistic regression analyses.

Purworejo District, central Java, Indonesia.

Pregnant women (n = 490).

For the three trimesters, EI/BMR ratio was 1.33±0.48, 1.53±0.43 and 1.52±0.40 (mean±standard deviation), respectively. The proportion of underreporters was 29.7%, 16.2% and 17.6%. Characteristics significantly associated with underreporting in at least one trimester included high body mass index and low education.

Levels of underreporting were low among the pregnant Indonesian women during the second and third trimesters. The low EI/BMR ratio during the first trimester likely reflects a true low intake due to nausea, rather than underreporting. Risk factors for being an underreporter included those known from developed countries, i.e. obesity and low education.

To assess the influence of different standards and restrained eating on underreporting in healthy, non-obese, weight-stable young subjects.

Eighty-three young adults (20–38 years, 55 women, 28 men) were assessed under weight-stable conditions with a 7-day dietary record and the three-factor eating questionnaire by Stunkard and Messick. Resting energy expenditure (REE; indirect calorimetry) plus data derived from physical activity records (PA) (Standard 1) or REE times an activity factor (AF) (Standard 2) was used as standard for total energy expenditure (TEE). For comparison, doubly labelled water (DLW) was used to measure TEE in a subgroup of subjects.

There was an association between self-reported energy intake and Standard 2 (r = 0.72) but not with Standard 1. When compared with DLW both calculated standards were inaccurate, but Standard 2 avoided high levels of overreporting. Using Standard 2 to identify ‘severe’ underreporting (SU; as defined by a deviation of energy intake (EI) and TEE of >20%), SU was seen in 37% of all subjects. It was more frequently found in women than in men (49% of women, 14.3% of men, P < 0.05). Underreporting subjects had a reduced EI (P < 0.01) but there were no significant differences in nutritional status (body weight and height, body mass index, fat mass and fat-free mass), energy expenditure and the proportion of energy from macronutrients between normal and underreporting subjects. However, high restraint was associated with a higher degree of underreporting in the total group, whereas disinhibition had an influence only in men.

A high prevalence of SU is seen in non-obese subjects. Characteristics of eating behaviour (restraint and disinhibition) were associated with underreporting but seemed to have a different influence in men and women.