Measuring nutrition intake plays a pivotal role in obesity and nutrition studies. However, objectively measuring nutrition intake is not easy. Under natural settings with less restriction, two approaches are often used: self-report and biomarkers. Biomarker approaches are considered more accurate compared with self-report; however, they are typically expensive and require special equipment for each nutrient. Self-report approaches are less expensive and do not require special laboratory equipment or techniques, and can be applied to the measurement of any nutrition components from macronutrients (i.e. carbohydrates, fat, and protein) to micronutrients (i.e. minerals and vitamins); however, self-reported data are known to be fraught with bias (Schoeller et al., 1990; Dhurandhar et al., 2015).

Consider the example of biomarker versus self-report energy intake. Energy intake (EI) is defined as a sum of the total metabolizable energy content of consumed foods and beverages (Astrup, 2005). EI can be measured using biomarkers, especially by energy balance approaches in which respiratory gas analysis or doubly labeled water (DLW) are used for the calculation of energy expenditure (EE) (Ferrannini, 1988). EI can be calculated using EE accounting for change in energy stores (i.e. body composition) during the period of measurement. The respiratory gas analysis typically requires the participants to wear a gas mask or stay in a metabolic chamber (Leonard, 2012), which restricts participants’ behavior; thus, the measured EI using the respiratory gas analysis may not reflect EI from habitual eating behavior. DLW requires the participants to visit the laboratory to drink labeled water enriched with stable hydrogen and oxygen isotopes and collect urine for measurement of isotope levels. It is considered to reflect habitual eating behavior and widely accepted as a standard and reliable approach in measuring EI in free-living individuals (Schoeller et al., 1986; Speakman, 1997; Schoeller and van Santen, 1982). However, given the high cost of isotopes and laboratories to analyze collected urine, DLW is not practical in large-scale epidemiological surveillance (Banna et al., 2017).

EI measured using self-report approaches (EI_SR) is computed by summing up the energy intake from foods and beverages reportedly consumed during the period of measurement, which can commonly range from a day to a year depending on approaches (e.g. 24 hr recall, food diary, and food frequency questionnaire; Leung et al., 2017; Sugiyama et al., 2014; Oftedal et al., 2017). Given that self-report approaches are low cost and scale-up relatively easily, EI_SR is still widely used in epidemiological studies. However, EI_SR is biased compared with the EI measured by biomarker such as DLW (EI_BIO) (Schoeller et al., 1990) and is discouraged from use to estimate actual EI (Dhurandhar et al., 2015). The mechanism behind the reporting bias in EI_SR has been argued elsewhere. For example, EI_SR is computed from the information of reported foods and beverages using tables of food and beverage nutrients, composed of weights, calories, and the amount of macro- and micronutrients. Compositions of foods and beverages are different among brands, stores, and seasons (Martelli et al., 2018), and thus nutrient databases cannot be exhaustive. Consequently, EI_SR using such databases may not be accurate (McHenry et al., 1945). Further factors, including social desirability (i.e. people do not want to be seen as a ‘big eater’ in general) (Heymsfield et al., 1995) and weight/social status (Heymsfield et al., 1995; Johansson et al., 1998; Macdiarmid and Blundell, 1998; Johansson et al., 2001; Klesges et al., 1995), are known to be associated with reporting bias. Thus, the difference between EI_BIO and EI_SR is due to systemic biases rather than a random error.

Despite concerns about the accuracy of self-report approaches, they have been predominantly used in large-scale nutrition epidemiology due to their high feasibility. Approaches to mitigate the bias in EI_SR have been argued, including commonly used approaches to exclude the data of EI_SR that are considered unreliable. For example, one approach is to exclude those reporting EI_SR out of the range considered by the researchers to be plausible, such as from 500 to 3500 kcal/day (Banna et al., 2017). However, this rule does not consider individual variability and is not appropriate for those whose intake is truly above 3500 kcal/day, such as elite athletes or those whose body size is large and requires more than 3500 kcal/day to maintain their weight (Santos et al., 2014; DeLany et al., 2013).

Other approaches are to use EE prediction equations to account for individual variability in EI (McCrory et al., 2002; Huang et al., 2005; Vinken et al., 1999; Livingstone and Robson, 2000; Xie et al., 2003; Briefel et al., 1995; Huang et al., 2004). If the EE predicted considering various factors (e.g. age, weight, sex) is far from EI_SR, those individuals are considered under/overreporting their actual EI. Indeed, such an approach has evidently reduced misclassification (Tooze et al., 2012). Among proposed approaches, the Goldberg cutoffs Goldberg et al., 1991 have been used in nutrition epidemiology to assess nutritional status in specific populations and investigate the association between nutritional status and health and socioeconomic outcomes (Livingstone and Robson, 2000; Xie et al., 2003; Briefel et al., 1995; Huang et al., 2004; Martin et al., 2018). Note that nutritional status in this context means dietary intakes in general; not only EI but also nutritional constituents (vitamin, fish oil, etc.) and dietary styles (Mediterranean, vegan, etc.; Grimes et al., 2013; Molina-Montes et al., 2017; New et al., 2000; Shaheen et al., 2001; O’Sullivan et al., 2011). Therefore, the rationale of using the Goldberg cutoffs (and other similar approaches) to exclude data from individuals considered to have unreliable EI_SR is based on an assumption, ‘if total EI is underestimated, it is probable that the intakes of other nutrients are also underestimated’ (Livingstone and Black, 2003). The corollary for overestimation may also be true.

Bias-correcting approaches like the Goldberg cutoffs are frequently used in nutrition studies. By bias, we are referring to the difference between the estimates and their true values. Here, we focus on two typical estimands in nutrition epidemiology: the mean of nutrition intake, and the associations between nutrition intake and health outcomes in a population. While reducing bias in the mean nutrition intake is useful for determining the true value of population nutrition intake from self-reported data, it is also important to reduce bias in the associations between nutrition intake and health outcomes. However, whether the Goldberg cutoffs reduce bias has not been proved theoretically or empirically. In our previous study, we demonstrated that the Goldberg cutoffs do not necessarily eliminate the bias in estimating the associations between EI and various health outcomes (Ejima et al., 2019). Extending our previous study, we consider nutritional intake beyond EI to include biomarkers of sodium, potassium, and protein with the primary goal to examine if the Goldberg cutoffs eliminate or reduce the bias using a single dataset from the Interactive Diet and Activity Tracking in the American Association of Retired Persons (IDATA) study.

Evaluating the performance of the Goldberg cutoffs solely depending on a single empirical dataset is limited because the dataset is a single realization from an unobserved data-generation process. Further, if the dataset is too small, conclusions may be limited because of issues related to power to detect bias in associations or the power to detect reductions in bias. Therefore, in addition to the empirical data analyses above, we also generated data through simulation and analyzed those generated data that preserve the characteristics of the empirical data. The simulation further enables us to assess the impact of sample size on the performance of the Goldberg cutoffs.

Dietary self-reporting is a basic component of nutrition epidemiology, particularly for studies of nutrition-health relationships. However, reporting bias in nutrition intake is perhaps one of the greatest impediments to understanding the true effect of nutrition on disease. Failure to account for this bias in self-reports can affect the analysis and interpretation of studies designed to assess the influences of nutrition on health. Correlation or regression coefficients estimating associations between self-reported nutrition intakes and health outcomes can be subject to substantial error (Freedman et al., 2011; Gibson et al., 2017). To address such reporting bias issues, there has been increased interest in the use of statistical models in conjunction with biomarker data (Freedman et al., 2010; Brown et al., 2021).

In this study, reporting bias in self-reported nutrition intake was demonstrated using both empirical data analysis and a simulation study. We first examined whether the bias in mean nutrition intake and the bias in the associations between nutrition intake and health outcomes are reduced by applying the Goldberg cutoffs to an empirical dataset, IDATA. We confirmed the bias in mean self-reported nutrition intake was eliminated by applying Goldberg cutoffs to exclude extreme reporters (mostly under-reporters). This bias can be mostly explained by differences in reporting between the accepted and rejected cases rather than a difference in actual nutrition intake between the accepted and rejected cases. We further tested whether the associations between health outcomes and nutrition intake are biased when self-reported data are used, and whether the bias is reduced by applying the Goldberg cutoffs. The significant bias for the associations between EI and BW and the association between EI and WC were reduced but remained after applying Goldberg cutoffs. We did not observe significant bias in associations for other nutrition-outcome pairs, and therefore cannot make conclusions regarding the utility of applying the Goldberg cutoffs to reduce bias for these associations in the empirical data.

To overcome the limitations of the data (i.e. sample size and weak associations among many nutrition-outcome pairs), we generated and analyzed data preserving the characteristics of the IDATA for the assessment of the Goldberg cutoffs. We confirmed the existence of bias in estimated associations when the self-reported data were used, and these biases were reduced, but not eliminated, by applying the Goldberg cutoffs in general. We also confirmed large MSE and low coverage probability when self-reported data were used; MSE and low coverage probability were improved by using Goldberg cutoffs, but were not recovered to the levels of biomarker data. Overall, the empirical data analyses and the simulation study suggest that the Goldberg cutoffs improve the estimates in nutrition-health outcome associations; however, the biases were not completely removed.

This study has several strengths. First, although energy intake has traditionally been studied more with respect to Goldberg cutoffs, herein we also investigated three other nutrition intakes: sodium, potassium, and protein. In our previous study, we used only energy intake as a predictor (Ejima et al., 2019). However, given that the Goldberg cutoffs are believed to reduce the bias in nutrition intake broadly, we expanded the approach to these other nutrition intakes, which can be measured by biomarkers (Livingstone and Black, 2003). Second, we conducted the simulation in addition to empirical data analyses. There are two advantages of the simulation design. (1) The data generation process was based on the IDATA dataset. Thus, we were able to simulate datasets reflective of those that would be seen in the real world. Both the sample size of the generated dataset and the variables that we considered are reflective of what is often observed in nutrition studies. (2) Simulation can consider different models for the data-generating process. Although our objective was not to disentangle the underlying data generation mechanisms, we were able to generate data that preserve the characteristics of the IDATA.

The limitations of this study are worth noting. In the analysis of IDATA, we used only bivariate associations for nutrition-health outcomes. It is possible that including various covariates or different models may influence the existence or lack of bias in nutrition-health associations. In the data-generation process for the simulations, assumptions were made on parameter settings. The reporting bias was set to be normally distributed. Modeling the bias using other distributions (such as a heavier tailed distribution that may be more consistent with a zero-bounded intake) may improve model fit, and thus generated more realistic data. We simulated the bias with constant variance, consistent with the apparent stability of variance across the intake quantiles in Figure 4—figure supplement 2 and Supplementary file 3; however, modeling the variance as a function of intake quantile may alter the data-generating process. We used the self-reported nutrition intake reported by ASA24, as it reflects recent dietary intake. The bias we observed in this study may be different for other self-report methods. Finally, the simulation approach, although based on the IDATA, was nonetheless generated from a limited sample size. We are not aware of thresholds for sufficient sample size for empirical-data-based simulation, and we have had success with smaller plasmode based sampling in the past (Ejima et al., 2020; Alfaras et al., 2021); however, it is possible a larger, more representative, or more robust empirical data set may influence both the empirical and simulated results. Further as the data was from a single population in which white and relatively older people are predominant, generalizability of the study finding should be carefully considered.

In conclusion, our study demonstrated that the use of Goldberg cutoffs can effectively reduce or eliminate bias in mean nutrition intake measurements in the IDATA dataset. However, our analysis also revealed that these cutoffs have limited efficacy in addressing bias in the associations between nutrition intake and health outcomes, as both empirical and simulated results showed. Investigators considering excluding extreme reporters in epidemiological studies should consider whether such approaches are likely to be helpful in answering their specific research questions, as our study has shown that such cutoffs are more useful in estimating mean nutrition intakes than in studies of associations between nutrition intake and health outcomes.

We conducted a set of empirical analyses to assess the reporting bias due to self-report and examine whether the Goldberg cutoffs reduce reporting bias using the data from The Interactive Diet and Activity Tracking in the American Association of Retired Persons (IDATA) study (National Cancer Institute, 2018a). The reporting bias in the mean of nutrition intakes and the associations (regression coefficients) between nutrition intakes and health outcomes were assessed. We also conducted a simulation study to investigate the performance of Goldberg cutoffs using the results from the empirical analyses as ground truth. Note that in data analyses, we assumed the estimates from biomarkers are the true values, since the true values (i.e. population parameters) are unobservable.

Among various nutrition intakes (NI), we considered sodium intake (SI), potassium intake (PoI), and protein intake (PrI) in addition to energy intake (EI). These NI were selected because they can be measured by urine samples as well as self-reports. We denote self-reported NI by NI_SR: EI_SR, SI_SR, PoI_SR, and PrI_SR. We denote biomarker-based NI by NI_BIO: EI_BIO, SI_BIO, PoI_BIO, and PrI_BIO. The NI_SR data remaining after applying the Goldberg cutoffs (hereafter, we call those ‘accepted cases’) are denoted by NI_G: EI_G, SI_G, PoI_G, and PrI_G. Health outcomes (HO) that may be associated with NI were selected from the data dictionaries; body weight (BW), waist circumference (WC), heart rate after fitness test (HR), resting systolic blood pressure (SBP), resting diastolic blood pressure (DBP), and VO₂ max (VO2). The estimated regression coefficients are denoted by the subscript of the explanatory variables and the health outcomes (e.g., ${\beta }_{{EI}_{SR,},BW}$ represents the association between self-reported EI and body weight).

Data generation

Request a detailed protocol

We generated data composed of four variables: NI_BIO (biomarker), NI_SR (self-reported), HO (health outcomes: BW, WC, HR, SBP, DBP, and VO2), and FFM (fat-free mass). NI_BIO and FFM were resampled from the IDATA. HO is computed using a linear model using NI_BIO as a predictive variable:

$HO=a_0+a_1{NI}_{BIO}+ϵ,ϵ~N\left(0,{\eta }^2\right),$

where $a_0,a_1,$ and ${\eta }^2$ are the model parameters estimated from the IDATA. NI_SR is generated by adding an error term (i.e., reporting error), $e$, to NI_BIO, adapting the approach proposed by Ward et al., 2019:

${NI}_{SR}={NI}_{BIO}+e,$

where $e$ is determined by the percentile of NI_BIO in the empirical distribution, $p\left({NI}_{BIO}\right)$. The reporting error $e$ is assumed to follow a normal distribution parametrized by percentile-specific mean, $\mu \left(p\right({NI}_{BIO}\left)\right)$, and a constant standard deviation, $\sigma$: $e~N(\mu ,{\sigma }^2)$. µ and $\sigma$ were estimated by fitting polynomial models to the IDATA: $\mu \left(p\left({NI}_{BIO}\right)\right)=\sum _{k=0}^K{m}_{k}p{\left(N{I}_{BIO}\right)}^k$ , where ${\displaystyle K}$ is the order of the polynomial. The order of the polynomial function was varied from 1 to 5 and we selected the value of ${\displaystyle K}$ that yields the smallest MSE. We used k-fold cross-validation with 10 folds to calculate the test MSE for each model. This resulted in $K=1, 3, 3, 5$ for EI, SI, PoI, and PrI, respectively (Figure 4—figure supplement 2 and Supplementary file 4). For example, assuming EI_BIO is 3000 kcal, which corresponds to 90.4 percentile, the mean of the reporting error, μ, was –749.6 and the standard deviation of the reporting error, $\sigma$, was 759.1. Therefore his/her $e$ is randomly sampled from the normal distribution: $N\left(-\mathrm{749.6,759.1}\right)$ . We further confirmed our assumption of homoskedastic $\sigma$ is reasonable by the Goldfeld Quandt test (Goldfeld and Quandt, 1965; Supplementary file 3). The Goldfeld Quandt test compares variance in high and low values of a variable; we discarded the middle 20% of the total observations (eliminated the center 61 observations among 303 in total) to define the high and low groups. The simulated data were truncated so that NI_SR >0. The variable NI_G was generated by applying the Goldberg cutoffs to EI_SR using FFM (thus, NI_G is a subset of NI_SR). The above process was repeated for each NI. We generated $n=100$ individuals, and the data generation was repeated 1000 times, leading to 1000 replicates. The data generating process is summarized in Figure 4.

Figure 4 with 2 supplements see all

Download asset Open asset

The data can be obtained from NCI after permission, and we are not allowed to redistribute the data. Researchers interested in accessing the original data should submit a project proposal. Specifically, "To gain access to available IDATA data and/or biospecimens, you must submit a project proposal. These are reviewed by NCI. If your project is approved, you will be required to complete a Data Transfer Agreement (and a Material Transfer Agreement, if applicable) before you will be granted access". (https://cdas.cancer.gov/idata/). All simulations and analyses were performed using the statistical computing software R (version 4.0.1). Code book and analytic code will be made publicly and freely available without restriction at http://doi.org/10.5281/zenodo.7783842.

1. 1. Ejima A
   2. Yamamoto N

   (2023) Zenodo

   Bias in nutrition-health associations is not eliminated by excluding extreme reporters in empirical or simulation studies.

   https://doi.org/10.5281/zenodo.7783842

1. 1. Alfaras I
   2. Ejima K
   3. Vieira Ligo Teixeira C
   4. Di Germanio C
   5. Mitchell SJ
   6. Hamilton S
   7. Ferrucci L
   8. Price NL
   9. Allison DB
   10. Bernier M
   11. de Cabo R

   (2021) Empirical versus theoretical power and type I error (false-positive) rates estimated from real murine aging research data

   Cell Reports 36:109560.

   https://doi.org/10.1016/j.celrep.2021.109560

   • PubMed
   • Google Scholar
2. Book

   1. Astrup A

   (2005)

   3 - energy metabolism and obesity

   In: Mela DJ, editors. Food, diet and obesity. Woodhead Publishing. pp. 58–75.

   • Google Scholar
3. 1. Banna JC
   2. McCrory MA
   3. Fialkowski MK
   4. Boushey C

   (2017) Examining plausibility of self-reported energy intake data: Considerations for method selection

   Frontiers in Nutrition 4:45.

   https://doi.org/10.3389/fnut.2017.00045

   • PubMed
   • Google Scholar
4. 1. Bhutani S
   2. Kahn E
   3. Tasali E
   4. Schoeller DA

   (2017) Composition of two-week change in body weight under unrestricted free-living conditions

   Physiological Reports 5:13.

   https://doi.org/10.14814/phy2.13336

   • PubMed
   • Google Scholar
5. 1. Black AE

   (2000) Critical evaluation of energy intake using the goldberg cut-off for energy intake: Basal metabolic rate A practical guide to its calculation, use and limitations

   International Journal of Obesity and Related Metabolic Disorders 24:1119–1130.

   https://doi.org/10.1038/sj.ijo.0801376

   • PubMed
   • Google Scholar
6. 1. Briefel RR
   2. McDowell MA
   3. Alaimo K
   4. Caughman CR
   5. Bischof AL
   6. Carroll MD
   7. Johnson CL

   (1995) Total energy intake of the US population: The third national health and nutrition examination survey, 1988-1991

   The American Journal of Clinical Nutrition 62:1072S–1080S.

   https://doi.org/10.1093/ajcn/62.5.1072S

   • PubMed
   • Google Scholar
7. 1. Brown AW
   2. Aslibekyan S
   3. Bier D
   4. Ferreira da Silva R
   5. Hoover A
   6. Klurfeld DM
   7. Loken E
   8. Mayo-Wilson E
   9. Menachemi N
   10. Pavela G
   11. Schoeller D
   12. Vorland CJ
   13. Whigham LD
   14. Allison DB

   (2021) Toward more rigorous and informative nutritional epidemiology: The rational space between dismissal and defense of the status quo

   Critical Reviews in Food Science and Nutrition 2021:1–18.

   https://doi.org/10.1080/10408398.2021.1985427

   • PubMed
   • Google Scholar
8. 1. Cunningham JJ

   (1991) Body composition as a determinant of energy expenditure: A synthetic review and a proposed General prediction equation

   The American Journal of Clinical Nutrition 54:963–969.

   https://doi.org/10.1093/ajcn/54.6.963

   • PubMed
   • Google Scholar
9. 1. DeLany JP
   2. Kelley DE
   3. Hames KC
   4. Jakicic JM
   5. Goodpaster BH

   (2013) High energy expenditure masks low physical activity in obesity

   International Journal of Obesity 37:1006–1011.

   https://doi.org/10.1038/ijo.2012.172

   • PubMed
   • Google Scholar
10. 1. Dhurandhar NV
    2. Schoeller D
    3. Brown AW
    4. Heymsfield SB
    5. Thomas D
    6. Sørensen TIA
    7. Speakman JR
    8. Jeansonne M
    9. Allison DB
    10. Energy Balance Measurement Working Group

    (2015) Energy balance measurement: When something is not better than nothing

    International Journal of Obesity 39:1109–1113.

    https://doi.org/10.1038/ijo.2014.199

    • PubMed
    • Google Scholar
11. Website

    1. Diet History Questionnaire

    (2022) Diet history questionnaire II (DHQ II) for U.S. & Canada

    Accessed February 6, 2022.

    https://epi.grants.cancer.gov/dhq2/
12. 1. Dugas LR
    2. Harders R
    3. Merrill S
    4. Ebersole K
    5. Shoham DA
    6. Rush EC
    7. Assah FK
    8. Forrester T
    9. Durazo-Arvizu RA
    10. Luke A

    (2011) Energy expenditure in adults living in developing compared with industrialized countries: A meta-analysis of doubly labeled water studies

    The American Journal of Clinical Nutrition 93:427–441.

    https://doi.org/10.3945/ajcn.110.007278

    • PubMed
    • Google Scholar
13. 1. Efron B
    2. Stein C

    (1981) The jackknife estimate of variance

    The Annals of Statistics 9:1176345462.

    https://doi.org/10.1214/aos/1176345462

    • Google Scholar
14. 1. Ejima K
    2. Brown AW
    3. Schoeller DA
    4. Heymsfield SB
    5. Nelson EJ
    6. Allison DB

    (2019) Does exclusion of extreme reporters of energy intake (the goldberg cutoffs) reliably reduce or eliminate bias in nutrition studies? analysis with illustrative associations of energy intake with health outcomes

    The American Journal of Clinical Nutrition 110:1231–1239.

    https://doi.org/10.1093/ajcn/nqz198

    • PubMed
    • Google Scholar
15. 1. Ejima K
    2. Brown AW
    3. Smith DL
    4. Beyaztas U
    5. Allison DB

    (2020) Murine genetic models of obesity: Type I error rates and the power of commonly used analyses as assessed by plasmode-based simulation

    International Journal of Obesity 44:1440–1449.

    https://doi.org/10.1038/s41366-020-0554-2

    • PubMed
    • Google Scholar
16. 1. Ferrannini E

    (1988) The theoretical bases of indirect calorimetry: A review

    Metabolism 37:287–301.

    https://doi.org/10.1016/0026-0495(88)90110-2

    • PubMed
    • Google Scholar
17. 1. Freedman LS
    2. Kipnis V
    3. Schatzkin A
    4. Tasevska N
    5. Potischman N

    (2010) Can we use biomarkers in combination with self-reports to strengthen the analysis of nutritional epidemiologic studies?

    Epidemiologic Perspectives & Innovations 7:2.

    https://doi.org/10.1186/1742-5573-7-2

    • PubMed
    • Google Scholar
18. 1. Freedman LS
    2. Schatzkin A
    3. Midthune D
    4. Kipnis V

    (2011) Dealing with dietary measurement error in nutritional cohort studies

    Journal of the National Cancer Institute 103:1086–1092.

    https://doi.org/10.1093/jnci/djr189

    • PubMed
    • Google Scholar
19. 1. Gibson RS
    2. Charrondiere UR
    3. Bell W

    (2017) Measurement errors in dietary assessment using self-reported 24-hour recalls in low-income countries and strategies for their prevention

    Advances in Nutrition 8:980–991.

    https://doi.org/10.3945/an.117.016980

    • PubMed
    • Google Scholar
20. 1. Goldberg GR
    2. Black AE
    3. Jebb SA
    4. Cole TJ
    5. Murgatroyd PR
    6. Coward WA
    7. Prentice AM

    (1991)

    Critical evaluation of energy intake data using fundamental principles of energy physiology: 1. derivation of cut-off limits to identify under-recording

    European Journal of Clinical Nutrition 45:569–581.

    • PubMed
    • Google Scholar
21. 1. Goldfeld SM
    2. Quandt RE

    (1965) Some tests for homoscedasticity

    Journal of the American Statistical Association 60:539–547.

    https://doi.org/10.1080/01621459.1965.10480811

    • Google Scholar
22. 1. Grimes CA
    2. Riddell LJ
    3. Campbell KJ
    4. Nowson CA

    (2013) Dietary salt intake, sugar-sweetened beverage consumption, and obesity risk

    Pediatrics 131:14–21.

    https://doi.org/10.1542/peds.2012-1628

    • PubMed
    • Google Scholar
23. 1. Heymsfield SB
    2. Darby PC
    3. Muhlheim LS
    4. Gallagher D
    5. Wolper C
    6. Allison DB

    (1995) The calorie: Myth, measurement, and reality

    The American Journal of Clinical Nutrition 62:1034S–1041S.

    https://doi.org/10.1093/ajcn/62.5.1034S

    • PubMed
    • Google Scholar
24. 1. Huang TT-K
    2. Howarth NC
    3. Lin B-H
    4. Roberts SB
    5. McCrory MA

    (2004) Energy intake and meal portions: Associations with BMI percentile in U.S. children

    Obesity Research 12:1875–1885.

    https://doi.org/10.1038/oby.2004.233

    • PubMed
    • Google Scholar
25. 1. Huang TTK
    2. Roberts SB
    3. Howarth NC
    4. McCrory MA

    (2005) Effect of screening out implausible energy intake reports on relationships between diet and BMI

    Obesity Research 13:1205–1217.

    https://doi.org/10.1038/oby.2005.143

    • PubMed
    • Google Scholar
26. 1. Johansson L
    2. Solvoll K
    3. Bjørneboe GE
    4. Drevon CA

    (1998) Under- and overreporting of energy intake related to weight status and lifestyle in a nationwide sample

    The American Journal of Clinical Nutrition 68:266–274.

    https://doi.org/10.1093/ajcn/68.2.266

    • PubMed
    • Google Scholar
27. 1. Johansson G
    2. Wikman A
    3. Ahrén AM
    4. Hallmans G
    5. Johansson I

    (2001) Underreporting of energy intake in repeated 24-hour recalls related to gender, age, weight status, day of interview, educational level, reported food intake, smoking habits and area of living

    Public Health Nutrition 4:919–927.

    https://doi.org/10.1079/phn2001124

    • PubMed
    • Google Scholar
28. 1. Kipnis V
    2. Subar AF
    3. Midthune D
    4. Freedman LS
    5. Ballard-Barbash R
    6. Troiano RP
    7. Bingham S
    8. Schoeller DA
    9. Schatzkin A
    10. Carroll RJ

    (2003) Structure of dietary measurement error: Results of the open biomarker study

    American Journal of Epidemiology 158:14–21.

    https://doi.org/10.1093/aje/kwg091

    • PubMed
    • Google Scholar
29. 1. Klesges RC
    2. Eck LH
    3. Ray JW

    (1995) Who underreports dietary intake in a dietary recall? Evidence from the second National health and nutrition examination survey

    Journal of Consulting and Clinical Psychology 63:438–444.

    https://doi.org/10.1037//0022-006x.63.3.438

    • PubMed
    • Google Scholar
30. 1. Leonard WR

    (2012) Laboratory and field methods for measuring human energy expenditure

    American Journal of Human Biology 24:372–384.

    https://doi.org/10.1002/ajhb.22260

    • PubMed
    • Google Scholar
31. 1. Leung DH
    2. Heltshe SL
    3. Borowitz D
    4. Gelfond D
    5. Kloster M
    6. Heubi JE
    7. Stalvey M
    8. Ramsey BW
    9. Baby O

    (2017) Nutrition study investigators of the cystic fibrosis foundation therapeutics development N. effects of diagnosis by newborn screening for cystic fibrosis on weight and length in the first year of life

    JAMA Pediatrics 171:546–554.

    https://doi.org/10.1001/jamapediatrics.2017.0206

    • Google Scholar
32. 1. Livingstone MB
    2. Robson PJ

    (2000) Measurement of dietary intake in children

    The Proceedings of the Nutrition Society 59:279–293.

    https://doi.org/10.1017/s0029665100000318

    • PubMed
    • Google Scholar
33. 1. Livingstone MBE
    2. Black AE

    (2003) Markers of the validity of reported energy intake

    The Journal of Nutrition 133 Suppl 3:895S–920S.

    https://doi.org/10.1093/jn/133.3.895S

    • PubMed
    • Google Scholar
34. 1. Macdiarmid J
    2. Blundell J

    (1998) Assessing dietary intake: Who, what and why of under-reporting

    Nutrition Research Reviews 11:231–253.

    https://doi.org/10.1079/NRR19980017

    • PubMed
    • Google Scholar
35. 1. Martelli F
    2. Giacomozzi C
    3. Fadda A
    4. Frazzoli C

    (2018) Understanding seasonal changes to improve good practices in livestock management

    Frontiers in Public Health 6:175.

    https://doi.org/10.3389/fpubh.2018.00175

    • PubMed
    • Google Scholar
36. 1. Martin JC
    2. Moran LJ
    3. Teede HJ
    4. Ranasinha S
    5. Lombard CB
    6. Harrison CL

    (2018) Diet quality in a weight gain prevention trial of reproductive aged women: A secondary analysis of a cluster randomized controlled trial

    Nutrients 11:49.

    https://doi.org/10.3390/nu11010049

    • PubMed
    • Google Scholar
37. 1. Matthews CE
    2. Kozey Keadle S
    3. Moore SC
    4. Schoeller DS
    5. Carroll RJ
    6. Troiano RP
    7. Sampson JN

    (2018) Measurement of active and sedentary behavior in context of large epidemiologic studies

    Medicine and Science in Sports and Exercise 50:266–276.

    https://doi.org/10.1249/MSS.0000000000001428

    • PubMed
    • Google Scholar
38. 1. McCrory MA
    2. McCrory MA
    3. Hajduk CL
    4. Roberts SB

    (2002) Procedures for screening out inaccurate reports of dietary energy intake

    Public Health Nutrition 5:873–882.

    https://doi.org/10.1079/PHN2002387

    • PubMed
    • Google Scholar
39. 1. McHenry EW
    2. Ferguson HP
    3. Gurland J

    (1945)

    Sources of error in dietary surveys

    Canadian Journal of Public Health / Revue Canadienne de Sante’e Publique 36:355–361.

    • Google Scholar
40. 1. Molina-Montes E
    2. Sánchez M-J
    3. Buckland G
    4. Bueno-de-Mesquita HBA
    5. Weiderpass E
    6. Amiano P
    7. Wark PA
    8. Kühn T
    9. Katzke V
    10. Huerta JM
    11. Ardanaz E
    12. Quirós JR
    13. Affret A
    14. His M
    15. Boutron-Ruault M-C
    16. Peeters PH
    17. Ye W
    18. Sund M
    19. Boeing H
    20. Iqbal K
    21. Ohlsson B
    22. Sonestedt E
    23. Tjønneland A
    24. Petersen KE
    25. Travis RC
    26. Skeie G
    27. Agnoli C
    28. Panico S
    29. Palli D
    30. Tumino R
    31. Sacerdote C
    32. Freisling H
    33. Huybrechts I
    34. Overvad K
    35. Trichopoulou A
    36. Bamia C
    37. Vasilopoulou E
    38. Wareham N
    39. Khaw K-T
    40. Cross AJ
    41. Ward HA
    42. Riboli E
    43. Duell EJ

    (2017) Mediterranean diet and risk of pancreatic cancer in the European prospective investigation into cancer and nutrition cohort

    British Journal of Cancer 116:811–820.

    https://doi.org/10.1038/bjc.2017.14

    • PubMed
    • Google Scholar
41. Website

    1. National Cancer Institute

    (2018a) Cancer data access system

    Accessed December 11, 2018.

    https://biometry.nci.nih.gov/cdas/idata/
42. Website

    1. National Cancer Institute

    (2018b) Automated self-administered 24-hour (ASA24) dietary assessment tool

    Accessed June 20, 2018.

    https://epi.grants.cancer.gov/asa24/
43. 1. New SA
    2. Robins SP
    3. Campbell MK
    4. Martin JC
    5. Garton MJ
    6. Bolton-Smith C
    7. Grubb DA
    8. Lee SJ
    9. Reid DM

    (2000) Dietary influences on bone mass and bone metabolism: Further evidence of a positive link between fruit and vegetable consumption and bone health?

    The American Journal of Clinical Nutrition 71:142–151.

    https://doi.org/10.1093/ajcn/71.1.142

    • PubMed
    • Google Scholar
44. 1. Oftedal S
    2. Davies PS
    3. Boyd RN
    4. Stevenson RD
    5. Ware RS
    6. Keawutan P
    7. Benfer KA
    8. Bell KL

    (2017) Body composition, diet, and physical activity: A longitudinal cohort study in preschoolers with cerebral palsy

    The American Journal of Clinical Nutrition 105:369–378.

    https://doi.org/10.3945/ajcn.116.137810

    • PubMed
    • Google Scholar
45. 1. O’Sullivan A
    2. Gibney MJ
    3. Brennan L

    (2011) Dietary intake patterns are reflected in metabolomic profiles: Potential role in dietary assessment studies

    The American Journal of Clinical Nutrition 93:314–321.

    https://doi.org/10.3945/ajcn.110.000950

    • PubMed
    • Google Scholar
46. 1. Park Y
    2. Dodd KW
    3. Kipnis V
    4. Thompson FE
    5. Potischman N
    6. Schoeller DA
    7. Baer DJ
    8. Midthune D
    9. Troiano RP
    10. Bowles H
    11. Subar AF

    (2018) Comparison of self-reported dietary intakes from the automated self-administered 24-h recall, 4-d food records, and food-frequency questionnaires against recovery biomarkers

    The American Journal of Clinical Nutrition 107:80–93.

    https://doi.org/10.1093/ajcn/nqx002

    • PubMed
    • Google Scholar
47. 1. Santos DA
    2. Silva AM
    3. Matias CN
    4. Magalhães JP
    5. Fields DA
    6. Minderico CS
    7. Ekelund U
    8. Sardinha LB

    (2014) Validity of a combined heart rate and motion sensor for the measurement of free-living energy expenditure in very active individuals

    Journal of Science and Medicine in Sport 17:387–393.

    https://doi.org/10.1016/j.jsams.2013.09.006

    • PubMed
    • Google Scholar
48. 1. Schoeller DA
    2. van Santen E

    (1982) Measurement of energy expenditure in humans by doubly labeled water method

    Journal of Applied Physiology 53:955–959.

    https://doi.org/10.1152/jappl.1982.53.4.955

    • PubMed
    • Google Scholar
49. 1. Schoeller DA
    2. Ravussin E
    3. Schutz Y
    4. Acheson KJ
    5. Baertschi P
    6. Jéquier E

    (1986) Energy expenditure by doubly labeled water: Validation in humans and proposed calculation

    The American Journal of Physiology 250:R823–R830.

    https://doi.org/10.1152/ajpregu.1986.250.5.R823

    • PubMed
    • Google Scholar
50. 1. Schoeller DA
    2. Bandini LG
    3. Dietz WH

    (1990) Inaccuracies in self-reported intake identified by comparison with the doubly labelled water method

    Canadian Journal of Physiology and Pharmacology 68:941–949.

    https://doi.org/10.1139/y90-143

    • PubMed
    • Google Scholar
51. 1. Shaheen SO
    2. Sterne JA
    3. Thompson RL
    4. Songhurst CE
    5. Margetts BM
    6. Burney PG

    (2001) Dietary antioxidants and asthma in adults: Population-based case-control study

    American Journal of Respiratory and Critical Care Medicine 164:1823–1828.

    https://doi.org/10.1164/ajrccm.164.10.2104061

    • PubMed
    • Google Scholar
52. Book

    1. Speakman J

    (1997)

    Doubly labelled water: Theory and practice

    Springer Science & Business Media.

    • Google Scholar
53. 1. Sugiyama T
    2. Tsugawa Y
    3. Tseng C-H
    4. Kobayashi Y
    5. Shapiro MF

    (2014) Different time trends of caloric and fat intake between statin users and nonusers among US adults: Gluttony in the time of statins?

    JAMA Internal Medicine 174:1038–1045.

    https://doi.org/10.1001/jamainternmed.2014.1927

    • PubMed
    • Google Scholar
54. 1. Tooze JA
    2. Krebs-Smith SM
    3. Troiano RP
    4. Subar AF

    (2012) The accuracy of the goldberg method for classifying misreporters of energy intake on a food frequency questionnaire and 24-h recalls: Comparison with doubly labeled water

    European Journal of Clinical Nutrition 66:569–576.

    https://doi.org/10.1038/ejcn.2011.198

    • PubMed
    • Google Scholar
55. 1. Vinken AG
    2. Bathalon GP
    3. Sawaya AL
    4. Dallal GE
    5. Tucker KL
    6. Roberts SB

    (1999) Equations for predicting the energy requirements of healthy adults aged 18-81 Y

    The American Journal of Clinical Nutrition 69:920–926.

    https://doi.org/10.1093/ajcn/69.5.920

    • PubMed
    • Google Scholar
56. 1. Ward ZJ
    2. Bleich SN
    3. Cradock AL
    4. Barrett JL
    5. Giles CM
    6. Flax C
    7. Long MW
    8. Gortmaker SL

    (2019) Projected U.S. state-level prevalence of adult obesity and severe obesity

    The New England Journal of Medicine 381:2440–2450.

    https://doi.org/10.1056/NEJMsa1909301

    • PubMed
    • Google Scholar
57. 1. Xie B
    2. Gilliland FD
    3. Li Y-F
    4. Rockett HRH

    (2003) Effects of ethnicity, family income, and education on dietary intake among adolescents

    Preventive Medicine 36:30–40.

    https://doi.org/10.1006/pmed.2002.1131

    • PubMed
    • Google Scholar

Author details

1. Nao Yamamoto

   School of Human Evolution and Social Change, Arizona State University, Tempe, United States

   Contribution

   Software, Formal analysis, Investigation, Visualization, Methodology, Writing – original draft

   Contributed equally with

   Keisuke Ejima

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-3131-2368

2. Keisuke Ejima

   1. Department of Epidemiology and Biostatistics, Indiana University School of Public Health-Bloomington, Bloomington, United States
   2. Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore, Singapore

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration

   Contributed equally with

   Nao Yamamoto

   For correspondence

   keisuke.ejima@ntu.edu.sg

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1185-3987

3. Roger S Zoh

   Department of Epidemiology and Biostatistics, Indiana University School of Public Health-Bloomington, Bloomington, United States

   Contribution

   Formal analysis, Supervision, Methodology, Writing – review and editing

   Competing interests

   Reviewing editor, eLife

   "This ORCID iD identifies the author of this article:" 0000-0002-8066-1153

4. Andrew W Brown

   1. Department of Biostatistics, University of Arkansas for Medical Sciences, Little Rock, United States
   2. Arkansas Children’s Research Institute, Little Rock, United States

   Contribution

   Resources, Formal analysis, Supervision, Funding acquisition, Investigation, Methodology, Writing – review and editing

   Competing interests

   AWB declares: in the past three years (since 2019-09-19), Dr. Brown has received travel expenses from International Food Information Council; speaking fees from Eastern North American Region of the International Biometric Society, Purchaser Business Group on Health, Purdue University, and University of Arkansas for the Medical Sciences; monetary awards from American Society for Nutrition; consulting fees from LA NORC, Pennington Biomedical Research Center, and Soy Nutrition Institute Global; and grants through his institution from Alliance for Potato Research & Education, American Egg Board, National Cattlemen's Beef Association, NIH/NHLBI, NIH/NIDDK, and NIH/NIGMS. In addition, he has been involved in research for which his institution or colleagues have received grants or contracts from Center for Open Science, Gordon and Betty Moore Foundation, Hass Avocado Board, Indiana CTSI, NIH/NCATS, NIH/NCI, NIH/NHLBI, NIH/NIA, NIH/NIGMS, NIH/NLM, and Sloan Foundation. His wife is employed by Reckitt Benckiser

• 686

  views

• 62

  downloads

• 2

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.