Epidemiology

Nutritional epidemiology: New perspectives for understanding the diet-disease relationship?

Abstract

Nutritional epidemiology is a subdiscipline of epidemiology and provides specific knowledge to nutritional science. It provides data about the diet-disease relationships that is transformed by Public Health Nutrition into the practise of prevention. The specific contributions of nutritional epidemiology include dietary assessment, description of nutritional exposure and statistical modelling of the diet-disease relationship. In all these areas, substantial progress has been made over the last years and is described in this article. Dietary assessment is moving away from the food frequency questionnaire (FFQ) as main dietary assessment instrument in large-scale epidemiological studies towards the use of short-term quantitative instruments due to the potential of gross measurement errors. Web-based instruments for self-administration are therefore evaluated of being able to replace the costly interviewer conducted 24-h-recalls. Much interest is also directed towards the technique of taking and analysing photographs of all meals ingested, which might improve the dietary assessment in terms of precision. The description of nutritional exposure could greatly benefit from standardisation of the coding of foods across studies in order to improve comparability. For the investigations of bioactive substances as reflecting nutritional intake and status, the investigation of concentration measurements in body fluids as potential biomarkers will benefit from the new high-throughput technologies of mass spectrometry. Statistical modelling of the dietary data and the diet-disease relationships can refer to complex programmes that convert quantitative short-term measurements into habitual intakes of individuals and correct for the errors in the estimates of the diet-disease relationships by taking data from validation studies with biomarkers into account. For dietary data, substitution modelling should be preferred over simple adding modelling. More attention should also be put on the investigation of non-linear relationships. The increasing complexity of the conduct and analysis of nutritional epidemiological studies is calling for a distinct and advanced training programme for the young scientists moving into this area. This will also guarantee that in the future an increasing number of high-level manuscripts will show up in this and other journals in respect of nutritional epidemiological topics.

Introduction

Nutritional epidemiology could be regarded as a subdiscipline of epidemiology, which provides a specific expertise that is also an integral part of nutritional sciences. As a subdiscipline of epidemiology, the overall definition of epidemiology will also hold for nutritional epidemiology. The scientific discipline of epidemiology is defined according to one of the grandfathers of this discipline in the 1970s, Abraham Lilienfeld, in his book ‘Foundations of Epidemiology’ as follows:1 ‘The study of the distribution of a disease or a physiological condition in human populations and of the factors which influence this distribution’. Regarding to this definition, nutritional epidemiology is dealing with nutritional exposures and their roles for the occurrence of diseases and impaired health conditions. The assessment of these exposures and the proper investigation of the link between exposure and end points form therefore the core activities of nutritional epidemiology.

The establishment of the relations between dietary exposures and fully established diseases, non-clinical intermediate end points and impaired health conditions are important information that also constitute the scientific knowledge of nutritional sciences. Many disciplines contribute with their approaches and specific knowledge to this scientific field. The field of nutritional epidemiology has evolved since the 1980s from a small speciality towards a major contributor to science. Nowadays, study results from the many epidemiological studies with dietary data including cross-sectional nutritional surveys form a large body of publications of this journal. Likewise, in other nutritional journals original publications in the area of nutritional epidemiology also form a large part of the articles. Together with molecular nutrition, physiology, toxicology and nutritional medicine, nutritional epidemiology can be considered as a major subdiscipline of nutritional sciences. Public Health Nutrition further represents the discipline that converts this knowledge into practise.

For the regular readers of scientific journals with interest in diet and nutrition, it is not easy to follow all the articles dealing with the relation between nutritional exposures and health risks. Many nutritional epidemiological articles address questions that are specific for end points ordered according to medical subjects such as endocrinology, cardiology and oncology. However, all of this more subject-related original research uses the research concepts that form the basis of nutritional epidemiology. This includes the areas of dietary assessment, description of nutritional exposure and statistical modelling of the diet-disease relationship. Some of them can be considered as unique contributions to nutritional sciences and medical research.

In nutritional epidemiology substantial progress has been achieved over the last years. Compared with the practise in nutritional epidemiology during the 1990s and the first 10 years of the new century, new approaches are gaining grounds. It seems as if a change of paradigm in how to conduct nutritional epidemiological studies, will take place in the near future.

Therefore, these new developments are described regarding the basis of nutritional epidemiology with the aim to improve the manuscripts in this area for the future. These new developments can often only be understood in the context of the historical background.

Dietary assessment

Dietary assessment in the early prospective epidemiological studies in the 1950s and 1960s focused mostly on cardiovascular risk factors and often included several thousands of study participants. It used either the method of food recording or the so-called Burke method that added to the recording method interviewer-guided meal based questions of habitual intake.2 With the advancement of the computer capabilities in the1980s, the application of retrospective self-administered dietary assessment instruments, the so-called semi-quantitative food frequency questionnaire (FFQ) method, became popular. This instrument asked for a limited number of habitual food intakes and was considered as superior to actual recording due to the retrospective nature covering a longer time period.3

The semi-quantitative FFQ became the prime dietary assessment instrument in large-scale epidemiological studies due to its easy application and reduced time burden for the study staff as well as for the participant. With the introduction of this dietary assessment instrument, also the concept of validation of the dietary instruments has been implemented as the new instrument was not aiming at the total diet but only for certain important aspects of dietary intake. The initial idea beyond the concept of validation was to show that the FFQ has sufficient accuracy and is able to estimate the dietary variable of interest to a certain extent compared with a reference method capturing total diet.4 In the early validation studies of FFQs data from food, records were used as the reference following the tradition since the 1950s.5 However, the coding of a food record turned out to be time consuming. Therefore, some groups proposed the 24-h recall method as alternative to a food record for validation by taking up some ideas proposed by Burke.2 This interviewer conducted instrument requests for the consumption of foods of the previous day in detail. The conduct of the interview through a trained interviewer can be further standardised through computer programmes that also include probing questions regarding potentially missed eating occasions and foods and a detailed food list.6 Compared with a food record, an unannounced conducted 24-recall is non-reactive regarding study specific food choices. Further, it turned out that a 24-h-recall did not need to be done face to face, but could also be done by telephone with the same quality.7, 8 Thus, compared with food records, the effort is reduced to apply measurements of the complete diet in larger populations.

The statistical approach of validation of the FFQ was further refined with the aim to calibrate the dietary data and to correct the estimate of the relative risk for the measurement error.9 The proposed design was the conduct of a validation study in a subgroup of the cohort population with multiple days of recording respective interviews.10, 11 In the multi-centric EPIC-study, only one 24-h-recall was conducted in a subgroup of the study sample (about 8% of the total sample).12 Later, it was revealed that the calibration formula in a situation of one FFQ and one 24-h-recall has limitations, compared with the situation of two and more 24-h-recalls per person.10

The statistical framework to identify and quantify measurement error of the FFQ had been further extended to the inclusion of biomarkers.13, 14 The use of biomarker information has the advantage that the random errors between self-reports of diet and objective measurements can be assumed to be zero. Otherwise, the random error between self-reporting instruments such as FFQ and 24-h-recalls is usually correlated and needs to be estimated. This might lead to unidentifiable measurement error models. The optimal design to identify the extent of measurement error in connection with FFQs is a subgroup study within the study population, including repeated measurements of all dietary assessment instruments and biological material as well.

In light of the FFQ or similar instruments such as dietary histories as prime assessment instrument in nearly all long-term studies established between 1980 and 2010, the question of measurement error when using such instruments is highly significant. Measurement error is distorting the estimate of relative risk and thus would lead to improper conclusion regarding the diet-disease relationship per unit of intake or across the whole range of intake if non-linear models are taken. Particularly important is the loss of power to detect existing true relationships, which would result in an increased chance of false negative results in studies with small power.15 But not only the estimate of relative risk is distorted by measurement error but also descriptive data about dietary intake and their relation to life-style and environmental data contain a bias. In this regard there is a lot of interest on the quantitative amount of measurement error with the use of this type of instrument. Usually, the correlation between the FFQ information and the reference data are in the order of moderate to good.16 The correlation coefficients are usually better if the variance of the energy assessment has been removed by energy adjustment (4; see below). However, the most recent validation studies of the FFQ with a reference instrument and objective biomarker information revealed a less optimistic picture taking energy and protein as example. Particularly, the NCI funded OPEN study attracted a lot of interest for their advanced statistical methodology and less promising results.17, 18

In light of the results of the validation studies with FFQs, a debate started whether measurement error and insufficient performance of the FFQ would have masked important diet-disease relationships in the large-scale cohort studies conducted so far.19 The new generation of large-scale epidemiological studies would therefore need to improve the quality of the dietary data substantially. There are several options to be discussed. One option is the use of the reference methods in addition to the FFQ. The reference methods such as 24-h-recalls or food records reflect total dietary intake and are short-term instruments. For a proper estimate of the diet of an individual several days need to be covered. The key statistical feature is the intra-individual variation of the dietary exposure between days. Statistical calculations suggest between 2–4 days for frequently eaten foods and up to 6 days for less often eaten foods.20

Another option is the use of innovative application of traditional instruments via the internet. These instruments have to be tailored for self-administration, which might result in less precision and less validity compared with the interviewer conducted methods. Web-based tools for self-administering 24-h-recalls are already available and under further investigation.21 Other groups already organised the complete cohort study with dietary assessment by web-based techniques.22 The critical issue is the validity of the web-based approach for self-administration compared with the interviewer approach forming currently the reference method. The overall philosophy of the application of the instruments developed so far is based on the insight, that each type of dietary assessment instrument provides specific information. This information can further be used by complex statistical algorithms to calculate the best estimate of habitual dietary exposures for each study subject from FFQ and reference data.20 As a consequence, the best estimates for each study subject will also form the intake distribution of that study. Precisely estimated dietary intake values favour the calculation of the relative risk on the continuous scale instead of using ranking procedures such as quantiles (see below).

A further and not yet fully developed option to improve dietary assessment is the application of a new technique that is based on the use of mobile phones.23 Mobile phone pictures of all foods consumed can be taken and transmitted to a server, which contains recognition programmes allowing a computerised analysis regarding type of food and amount. This technique is still under development but might be the tool of the future due to the low cost of conduct compared with interviews. In addition, with this technique the amounts of foods being eaten can be estimated with a still unreached precision.

Estimate of nutritional exposure

The dietary assessment instruments provide data on food intake and, if wanted, data on supplement use. The type of information about a food could be obtained in a written manner by food records or requested by interviewers in 24-h-recalls. Depending on the instrument, the requested details regarding a food could be variable, such as fat content, type of packaging, cooking procedures, and often the brand name. Thus, the coding of such foods will pose a challenge. It is estimated that at a particular point of time about 200 000 commercial foods are available in the supermarkets in western societies. Bar code readings of the labels of the commercial foods might help in some situations to handle the detailed information. In some areas such as the European Union, each commercial food has a specific label and some details of the food are available in a database. In general, all foods estimated in a study must be assigned a food code and have to be integrated into a food code system. Food code systems do not only provide a hierarchical ordering of the foods but also the connection to nutrient databases. The number of main food groups is usually in the order of 15–20 in such food code systems. About 40–50 food subgroups usually form the next level within the hierarchical food code. Main food and 1st order subgroup foods are often used for investigating risk relations.24 The use of a more detailed food code beyond the 1st order subgroups in risk analyses is usually restricted to specific hypotheses, for example, investigations of a food having a high concentration of a specific plant component.

Unfortunately, there is no world-wide common food coding system available for scientific use. Thus, study results on food group level and particular food subgroup level might not be directly comparable across studies partly because of a non-standardised food grouping system. Particular questions that is, regard the assignment of potatoes to the vegetable or staple group or nuts to the fruit or as an own group. The detailed composition of a food group regarding specific foods across studies can thus widely differ and would hamper systematic reviews and meta-analyses in the interpretation.

The need for standardisation of food tables25 with the prime aim to link this food table with a highly standardised and comparable nutrient database is well acknowledged in Europe. The European Union has invested into this issue over the last years by launching specific calls and by supporting collaborations. Particularly, the European Food Information Resource Network pushed for the standardisation of the national food tables to create an uniform Pan-European food coding system, such as nutrient databases that are already available or being made available in the future.26 This development is important in view of the development of a standardised 24-h-recall dietary assessment instrument for all European countries for surveys and epidemiological studies.27

Efforts had also been undertaken during the last years to extend existing nutrient databases towards bioactive compounds in addition to nutrients and to apply this knowledge in studies assessing the intakes. Thus, intake of i.a. carotenoids, flavonoids, phenols and phytoestrogens can nowadays be calculated by combining food code systems and specific nutrient databases.28

However, it needs to be discussed whether the calculation of dietary intakes of nutrients and other bioactive compounds by the assessment of food intake will provide the information that is required to judge the impact of a bioactive compound on disease risk. One disadvantage of the approach is the variation of concentrations of food compounds within the foods. Usually, a nutrient table makes use of the mean concentration in a food and not of the variation. Another disadvantage of the approach is the fact that intake does not directly reflect the internal dose as it does not consider bioavailability of the nutrients. The internal dose of a bioactive substance seems to be the most biologically important information, and the most relevant for disease risk. Inter-individual variation of resorption and transportation of the compounds due to genetic predisposition and the influence of other dietary compounds is considered by this approach which is not been considered by the dietary approach.

Direct measurements of these bioactive compounds in body fluids consist therefore of the alternative approach compared with the calculations of intake from food use and nutrient tables. The biomarker approach is also independent from the information on foods and their role in providing nutrients. Some foods or food groups are important sources not only for one bioactive compound, but often for many of them. Thus, it seems nearly impossible in many instances to distinguish statistically between the impact of a food and its major compounds on disease risk.

The use of human biomaterials to characterise and differentiate a study population regarding a specific nutritional exposure has therefore attracted a lot of attention. Research on nutritional biomarkers has been conducted for a long time, but a break through was not obtained so far. Still, the traditional nutritional biomarkers are dominating the discussion and only a few biomarkers could be added over the last years.29 Nonetheless, the measurement of concentrations of particularly bioactive compounds in body fluids had been conducted repeatedly in epidemiological studies and has added to our understanding of the role of diet on disease risk.30, 31

Research on biomarkers for nutritional intake, however, is ongoing and will hopefully generate substantial progress in the near future. In particular, there is a great demand for recovery biomarkers that directly help to validate dietary assessment instruments. A novel approach of detecting and using biomarkers for estimating specific dietary intakes and/or the relation to end points has been proposed in connection with the recently developed high-throughput technology of metabolomics.32 This technology based on mass spectrometry or nuclear magnetic resonance spectroscopy provides quantitative information on the presence of hundreds to thousands of metabolites in fluids in one run. It is speculated that each food, particular plant foods, will have its own finger print of metabolites. First studies are underway that link those fingerprints with the metabolites assessed in body fluids, either urine or blood components.33 Metabolomics offers the very interesting perspective that by analysing body fluid information on dietary intake can be derived. Further, these data can be used to investigate the risk relation to end points. However, it is still not definable in as much the biomarker approach, in general, will improve the estimate of the diet-disease relationship.

Statistical modelling of the diet-disease relationship

The FFQ as a semi-quantitative and usually biased instrument favoured the expression of nutritional exposure as quantile after ranking the study participants according to the estimated dietary exposure. The study participants were allocated to classes of intake, such as tertile, quartile of quintile and thereafter related to disease risk. It would be a mistake to label these quantiles with the dietary estimates derived from the semi-quantitative instrument due to the measurement error. The better estimate of the nutritional exposure within the FFQ-quantiles would be the average intake observed in the validation study for these subjects estimated in the reference instrument.34

This remark should direct the general interest to a common quality assurance system for published dietary intake values across all the disciplines of nutritional science. This means that given quantities such as a gram of fat should reflect the same true amount whether being used in animal experiments, human intervention studies or estimated in observational studies. Observational studies have—given the challenge of quality assurance—the problem that self-reports of diet might underestimate true intake.7, 35, 36 Therefore, even well-conducted epidemiological studies with several 24-HDRs might not provide the correct true intake values and their variation within the population. However, this type of measurement error in well-conducted studies seems to be small, compared with the systematic bias usually observed for FFQs and other types of instruments aiming at measuring habitual diet directly. It should be therefore the aim of nutritional epidemiology to present proper and comparable dietary data that can be immediately linked to nutritional knowledge generated in other fields. An important aspect of dietary intake assessment is to estimate true intake distributions within the study population and to reduce bias of the population estimates. Due to its feasibility in large population studies, FFQs will continue to be selected as the main dietary assessment instrument, coupled with validation sub-studies integrated into the data collection of the whole population. Therefore, it will still be demand to integrate the information from the validation studies into the overall assessment of diet in a study population.37

An important question is the adjustment for energy when the diet-disease relationship is modelled. Adjustment for energy means to model an iso-caloric situation. Energy intake itself will not be considered as important variable as energy intake is only one of the components comprising the energy balance. At constant body weight, energy intake of a subject reflects resting energy expenditure, physical activity and the individual metabolic capacity to utilise energy and macronutrients, but is not a variable predicting energy balance respective weight change. When energy adjusted models are used to describe the diet-disease relationship, the energy intake is assumed to be fixed and the increase or decrease of intake of a food or energy containing nutrients can only be accomplished by the exchange of other dietary variables. The most adequate modelling procedure with energy adjustment would be to run well-defined substitution models.38, 39, 40 Modelling the substitution of a dietary component with another dietary component means to formulate the modelling part in such a way that one dietary variable is replaced by another dietary variable, taking energy as the common unit across the dietary factors. Substitution models that hold energy constant are only useful to be applied if the dietary variable contains energy, such as macronutrients or foods. The application of substitution models, however, is in principle not restricted to holding energy constant. For example., in case of foods, also the total amount of food can be hold constant and 1 g of a specific food exchanged with 1 g of another food. Substitution models will give a much better insight into the health implication of changing a diet compared with models that assume a change also in energy intake or are unspecific regarding the replacement of the dietary exposure. It could be shown, for example, that the increase of poultry consumption will not have an effect on risk of colorectal caner, but the replacement of red meat by poultry.40

Modelling strategies of the diet-disease relationship should further investigate non-linear relationships such as thresholds and levelling off of effects. Regression models with one linear term per exposure variable assume a linear relationship across the range of the invested variables if the continuous variable is modelled. In principle, the non-linear risk modelling of dietary variables on a continuous scale is preferred against the modelling of categories.41 Modelling the diet-disease relationship on a continuous scale assumes that the range and distribution of dietary exposure within the study population is properly assessed (see above). There are several suggestions of modelling non-linear relationships. Most of them propose to add power terms to the modelling equation in addition to the linear term. For modelling non-linear relationships on a continuous scale the most oftenly used method is restricted cubic spline regression.42, 43 In this approach, knots are defined and the regression lines between the knots are modelled with simple, quadratic and cubic terms. The use of this technique requires the definition of a reference which could be different across studies. Other approaches include the search for the best fit by using the fractional polynomial technique.44 This technique has been developed to provide the most proper adjustment formula for a variable. By using this technique non-linear relations between dietary variables can be analysed and presented.

Discussion

Progress over the last years in the different fields of activities forming the core of nutritional epidemiology will change the practise regarding dietary assessment, formation of dietary variables and statistical modelling of the diet-disease relationship in the future. Subsequently, also the current standard of study presentations will change.

The most likely scenario of a well-conducted study in the future will include the use of multiple short-term instruments for food intake preferentially applied via the web in a kind of method mix coupled with complex statistical methods that calculate the best estimate of habitual food intake of an individual. The food data will be further corrected for measurement error using validation studies with biomarkers in subgroups integrated into the study. The thus estimated food intake will be linked to extended food tables allowing calculating the intake of nutrients and many bioactive compounds. The diet-disease relationship will be investigated with non-linear modelling using the quantitative values and substitution models with defined exchange relationships.

It is the question how we can train the young scientists that decide to be trained in the subject of nutritional epidemiology as a PhD student or young post-doc. It is obvious that the transformation of the newly developed approaches into a standard practise will require a lot of training. Therefore, new students need an updated curriculum for advanced training that allow the new generation of students to go beyond the current practise. The curriculum needs to have at least a master’s level and have to include basics in epidemiology and natural and medical sciences and advanced courses in the core fields forming nutritional epidemiological sciences. Only the establishment of specific training facilities for young sciences will allow to practice the increasing complexity of the subject and to push for further innovations. It is obvious that we need to establish new large scale epidemiological studies that adopt the new ideas and provide the data that are necessary to address the still unsolved question of the role of diet for disease risk.

It is the hope that efforts in training and adoption of the new methods will also result in manuscripts that adopt the new methods and reflect up-to-date techniques. This will also be a challenge for editors and reviewers for establishing a proper procedure to filter the best papers out of the incoming manuscripts and to present meaningful insights to the readers of this and other journals.

Nutrition is in addition to obesity, tobacco smoking and physical activity still considered as one of the major driving forces influencing life expectancy and the occurrence of chronic diseases in a population.45 By understanding major aspects of the diet-disease relationship, the potential of prevention of chronic disease can be better assessed and proper prevention measures developed by the discipline of public health nutrition. In addition, nutrition is deeply involved in all types of mechanisms that constitute life. By shedding light on the diet-disease relationship nutritional epidemiology helps to understand the enormous complexity of life a little bit better.

References

  1. 1

    Lilienfeld AM . Foundations of Epidemiology 1st edn Oxford University press: New York, NY, USA, 1976.

    Google Scholar 

  2. 2

    Burke BS . The dietary history as a tool in research. J Am Diet Assoc 1947; 23: 1041–1046.

    Google Scholar 

  3. 3

    Rohan TE, Potter ID . Retrospective assessment of dietary intake. Am J Epidemiol 1984; 120: 876–887.

    CAS  Article  Google Scholar 

  4. 4

    Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J et al. Reproducibility and validity of a semiquantitative food frequency questionnaire 1985 Am J Epidemiol 1985; 122: 51–56.

    CAS  Article  Google Scholar 

  5. 5

    Feskanich D, Rimm EB, Giovannucci EL et al. Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. J Am Diet Assoc 1993; 93: 790–796.

    CAS  Article  Google Scholar 

  6. 6

    Slimani N, Kaaks R, Ferrari P, Casagrande C, Clavel-Chapelon F, Lotze G et al. European Prospective Investigation into Cancer and Nutrition (EPIC) calibration study: rationale, design and population characteristics. Public Health Nutr 2002; 5: 1125–1145.

    CAS  Article  Google Scholar 

  7. 7

    Tran KM, Johnson RK, Soultanakis RP, Matthews DE . In- person vs telephone-administered multiple-pass 24-hour recalls in women: validation with doubly labeled water. J Am Diet Assoc 2000; 100: 777–783.

    CAS  Article  Google Scholar 

  8. 8

    Brustad M, Skeie G, Braaten T, Slimani N, Lund E . Comparison of telephone vs face-to-face interviews in the assessment of dietary intake by the 24 h recall EPIC SOFT program—the Norwegian calibration study. Eur J Clin Nutr 2003; 57: 107–113.

    CAS  Article  Google Scholar 

  9. 9

    Kaaks R, Ferrari P, Ciampi A, Plummer M, Riboli E . Uses and limitations of statistical accounting for random error correlations, in the validation of dietary questionnaire assessments. Public Health Nutr 2002; 5: 969–976.

    Article  Google Scholar 

  10. 10

    Hoffmann K, Kroke A, Klipstein-Grobusch K, Boeing H . Standardization of dietary intake measurements by nonlinear calibration using short-term reference data. Am J Epidemiol 2002; 156: 862–870.

    Article  Google Scholar 

  11. 11

    Jaceldo-Siegl K, Knutsen SF, Sabaté J, Beeson WL, Chan J, Herring RP et al. Validation of nutrient intake using an FFQ and repeated 24 h recalls in black and white subjects of the Adventist Health Study-2 (AHS-2). Public Health Nutr 2010; 13: 812–819.

    Article  Google Scholar 

  12. 12

    Kaaks R, Riboli E, van Staveren W . Calibration of dietary intake measurements in prospective cohort studies. Am J Epidemiol 1995; 142: 548–556.

    CAS  Article  Google Scholar 

  13. 13

    Rosner B, Michels KB, Chen YH, Day NE . Measurement error correction for nutritional exposures with correlated measurement error: use of the method of triads in a longitudinal setting. Stat Med 2008; 27: 3466–3489.

    Article  Google Scholar 

  14. 14

    Freedman LS, Tasevska N, Kipnis V, Schatzkin A, Mares J, Tinker L et al. Gains in statistical power from using a dietary biomarker in combination with self-reported intake to strengthen the analysis of a diet-disease association: an example from CAREDS. Am J Epidemiol 2010; 172: 836–842.

    Article  Google Scholar 

  15. 15

    Freedman LS, Schatzkin A, Midthune D, Kipnis V . Dealing with dietary measurement error in nutritional cohort studies. J Natl Cancer Inst 2011; 103: 1086–1092.

    Article  Google Scholar 

  16. 16

    Cade JE, Burley VJ, Warm DL, Thompson RL, Margetts BM . Food-freqency questionnaires: a review of their design, validation and utilisation. Nutr Res Rev 2004; 17: 5–22.

    CAS  Article  Google Scholar 

  17. 17

    Kipnis V, Subar AF, Midthune D, Freedman LS, Ballard-Barbash R et al. Structure dietary measurement error: results of the OPEN biomarker study. Am J Epidemiol 2003; 158: 14–21.

    Article  Google Scholar 

  18. 18

    Schatzkin A, Subar AF, Moore S, Park Y, Potischman N, Thompson FE et al. Observational epidemiologic studies of nutrition and cancer: the next generation (with better observation. Cancer Epidemiol Biomarkers Prev 2009; 18: 1026–1032.

    Article  Google Scholar 

  19. 19

    Schatzkin A, Kipnis V . Could exposure assessment problems give us wrong answers to nutrition and cancer questions? J Natl Cancer Inst 2004; 96: 1564–1565.

    Article  Google Scholar 

  20. 20

    Carroll RJ, Midthune D, Subar AF, Shumakovich M, Freedman LS, Thompson FE et al. Taking advantage of the strengths of 2 different dietary assessment instruments to improve intake estimates for nutritional epidemiology. Am J Epidemiol 2012; 175: 340–347.

    Article  Google Scholar 

  21. 21

    Subar AF, Kirkpatrick SI, Mittl B, Zimmerman TP, Thompson FE, Bingley C et al. The Automated Self-Administered 24-hour dietary recall (ASA24): a resource for researchers, clinicians, and educators from the National Cancer Institute. J Acad Nutr Diet 2012; 112: 1134–1137.

    Article  Google Scholar 

  22. 22

    Hercberg S . Web-based studies: The future in nutritional epidemiology (and overarching epidemiology) for the benefit of public health? Prev Med 2012; 55: 544–545.

    Article  Google Scholar 

  23. 23

    Daugherty BL, Schap TE, Ettienne-Gittens R, Zhu FM, Bosch M, Delp EJ et al. Novel technologies for assessing dietary intake: evaluating the usability of a mobile telephone food record among adults and adolescents. J Med Internet Res 2012; 14: 58.

    Article  Google Scholar 

  24. 24

    von Ruesten A, Feller S, Bergmann MM, Boeing H . Diet and risk of chronic diseases: Results from the first 8 years of follow-up in the EPIC-Potsdam study. Eur J Clin Nutr; e-pub ahead of print 6 February 2013 doi:10.1038/ejcn.2013.7.

  25. 25

    Ireland JD, Møller A . Lanngual food description: a learning process. Eur J Clin Nutr 2010; 64: S44–S48.

    Article  Google Scholar 

  26. 26

    Bouckaert KP, Slimani N, Nicolas G, Vignat J, Wright AJ, Roe M et al. Critical evaluation of folate data in European and international databases: recommendations for standardization in international nutritional studies. Mol Nutr Food Res 2011; 55: 166–180.

    CAS  Article  Google Scholar 

  27. 27

    Slimani N, Deharveng G, Unwin I, Southgate DA, Vignat J, Skeie G et al. The EPIC nutrient database project (ENDB): a first attempt to standardize nutrient databases across the 10 European countries participating in the EPIC study. Eur J Clin Nutr 2007; 61: 1037–1056.

    CAS  Article  Google Scholar 

  28. 28

    Zamora-Ros R, Knaze V, Luján-Barroso L, Kuhnle GG, Mulligan AA, Touillaud M et al. Dietary intakes and food sources of phytoestrogens in the European Prospective Investigation into Cancer and Nutrition (EPIC) 24-hour dietary recall cohort. Eur J Clin Nutr 2012; 66: 932–941.

    CAS  Article  Google Scholar 

  29. 29

    Tasevska N, Runswick SA, McTaggart A, Bingham SA . Urinary sucrose and fructose as biomarkers for sugar consumption. Cancer Epidemiol Biomarkers Prev 2005; 14: 1287–1294.

    CAS  Article  Google Scholar 

  30. 30

    Kristal AR, Till C, Platz EA, Song X, King IB, Neuhouser ML et al. Serum lycopene concentration and prostate cancer risk: results from the prostate cancer prevention trial. Cancer Epidemiol Biomarkers Prev 2011; 20: 638–648.

    CAS  Article  Google Scholar 

  31. 31

    Aune D, Chan DS, Vieira AR, Navarro Rosenblatt DA, Vieira R, Greenwood DC et al. Dietary compared with blood concentrations of carotenoids and breast cancer risk: a systematic review and meta-analysis of prospective studies. Am J Clin Nutr 2012; 96: 356–373.

    CAS  Article  Google Scholar 

  32. 32

    Llorach R, Garcia-Aloy M, Tulipani S, Vazquez-Fresno R, Andres-Lacueva C . Use of mass spectrometry fingerprinting to identify urinary metabolites after consumption of specific foods. J Agric Food Chem 2012; 60: 8797–8008.

    CAS  Article  Google Scholar 

  33. 33

    Lloyd AJ, Favé G, Beckmann M, Lin W, Tailliart K, Xie L et al. Use of mass spectrometry fingerprinting to identify urinary metabolites after consumption of specific foods. Am J Clin Nutr 2011; 94: 981–991.

    CAS  Article  Google Scholar 

  34. 34

    Boeing H, Wahrendorf J, Heinemann L, Kulesza W, Rywik SL, Sznajd J et al. Results from a comparative dietary assessment in Europe: I. Comparison of dietary information derived from concurrently applied frequency questionnaires and quantitative measurement instruments. Eur J Clin Nutr 1989; 43: 367–377.

    CAS  PubMed  Google Scholar 

  35. 35

    Ferrari P, Slimani N, Ciampi A, Trichopoulou A, Naska A, Lauria C et al. Evaluation of under- and overreporting of energy intake in the 24-hour diet recalls in the European Prospective Investigation into Cancer and Nutrition (EPIC). Public Health Nutr 2002; 5: 1329–1345.

    CAS  Article  Google Scholar 

  36. 36

    Scagliusi FB, Ferriolli E, Pfrimer K, Laureano C, Cunha CS, Gualano B et al. Underreporting of energy intake in Brazilian women varies according to dietary assessment: a cross-sectional study using doubly labeled water. J Am Diet Assoc 2008; 108: 2031–2040.

    Article  Google Scholar 

  37. 37

    Zhang S, Midthune D, Guenther PM, Krebs-Smith SM, Kipnis V, Dodd KW et al. A new multivariate measurement error model with zero-inflated dietary data, and its application to dietary asessment. Ann Appl Stat 2011; 5: 1456–1487.

    Article  Google Scholar 

  38. 38

    Faerch K, Lau C, Tetens I, Pedersen OB, Jørgensen T, Borch-Johnsen K et al. A statistical approach based on substitution of macronutrients provides additional information to models analyzing single dietary factors in relation to type 2 diabetes in danish adults: the Inter99 study. J Nutr 2005; 135: 1177–1182.

    CAS  Article  Google Scholar 

  39. 39

    Schulze MB, Schulz M, Heidemann C, Schienkiewitz A, Hoffmann K, Boeing H . Carbohydrate intake and incidence of type 2 diabetes in the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study. Br J Nutr 2008; 99: 1107–1116.

    CAS  Article  Google Scholar 

  40. 40

    Daniel CR, Cross AJ, Graubard BI, Hollenbeck AR, Park Y, Sinha R . Prospective investigation of poultry and fish intake in relation to cancer risk. Cancer Prev Res (Phila) 2011; 11: 1903–1911.

    Article  Google Scholar 

  41. 41

    van Walraven C, Hart RG . Leave 'em alone—Why continuous variables should be analyzed as such. Neuroepidemiology 2008; 30: 138–139.

    Article  Google Scholar 

  42. 42

    Boeing H, Dietrich T, Hoffmann K, Pischon T, Ferrari P, Lahmann PH et al. Intake of fruits and vegetables and risk of cancer of the upper aero-digestive tract: the prospective EPIC-study. Cancer Causes Control 2006; 17: 957–969.

    Article  Google Scholar 

  43. 43

    Wu H, Dai Q, Shrubsole MJ, Ness RM, Schlundt D, Smalley WE et al. Fruit and vegetable intakes are associated with lower risk of colorectal adenomas. J Nutr 2009; 139: 340–344.

    CAS  Article  Google Scholar 

  44. 44

    Royston P, Sauerbrei W . Special topics involving fractional polynomials. In: Shewhart AA, Wilks SS (eds) Multivariable Model-Building: A Pragmatic Approach To Regression Analysis Based On Fractional Polynomials For Modelling Continuous Variables. John Wiley and Sons, Ltd: Chichester, UK, 2008.

    Google Scholar 

  45. 45

    World Health Organization, Global status report on noncommunicable diseases 2010. Description of the global burden of NCDs, their risk factors and determinants. April 2011. ISBN: 978 92 4 156422 9.

Download references

Author information

Affiliations

Authors

Corresponding author

Correspondence to H Boeing.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Boeing, H. Nutritional epidemiology: New perspectives for understanding the diet-disease relationship?. Eur J Clin Nutr 67, 424–429 (2013). https://doi.org/10.1038/ejcn.2013.47

Download citation

Keywords

  • epidemiology
  • dietary assessment
  • statistical modelling
  • measurement error
  • nutrition

Further reading

Search