Dietary assessments : how to validate primary data before conclusions can be drawn By Leif

Dietary assessment methods result almost without exception in an underestimation of energy intake. Energy turnover should consequently be evaluated on the basis of studies of energy expenditure and only with exceptions on determinations of energy intake. An objective and reliable reference to validate whether obtained data from dietary assessment may be realistic or represent an underevaluation can be based on a comparison between the observed or registered energy intake and the theoretical calculations of energy turnover based on calculated BMR with the addition of a relevant PAL factor, which is based on the lifestyle including physical activity. It is not scientifically relevant to draw any conclusions regarding the relationship between the dietary intake of various nutrients and health in a long-term perspective if energy needs are not met by dietary intake according to this assessment, nota hene if body weight is maintained. "Although this may seem like a paradox, assessments for the assessment of energy from studies of dietary intake, it is conall exact science is dominated by the idea turnover, while energy expenditure (EE) sequently first of relevance to analyse the of approximation7' (Bertrand Russell) can be registered by direct and indirect primary goals of the study performed, and calorimetry. to what extent the method used is also


Introduction
In the latest edition of the FAO/WHO/ UNU recommendations for energy and protein intake (I), it was stated that energy turnover should preferably be evaluated on the basis of studies in energy expenditure and only with exceptions on determinations of energy intake.This is essentially due to the well-known fact that dietary assessment methods almost without exceptions result in an underestimation of energy intake (EI), no matter which method for dietary assessment has been used.Furthermore, energy turnover (ET) can be covered from endogenous release from energy stores, i.e. glycogen in liver and muscles and triglycerides in the adipose tissue, which calls for studies of body composition in addition to dietary Leif Hambraeus, Prof., Dept of Medical Sciences, Nutrition Unit, Faculty of Medicine, Uppsala University, Dag Harnmarskjolds vag 21, SE-752 37 Uppsala, Sweden.E-mail: Leif.Hambraeus@nutrition.uu.seThis article is based on a lecture held at the conference "Dietary assessments -how can we interpret the results?"The conference was arranged by Uppsala Food and Nutrition Centre (Uppsala Livsmedelscentrum) in Uppsala on Nov 1 1, 1997.

Various methods
As discussed in another paper by Lennernas (2), there are several forms of methods used for studies of dietary intake, based on interviews, records, frequency questionnaire or even double portion technique.Each of them has its pros and cons.Studies of the dietary intake may furthermore be aimed at determining the nutritional status of an individual, or be used for dietary counselling or to follow up dietary treatment.However, they can also be used for research purposes in metabolic studies or for epidemiological analysis of the diethealth relationship.Finally, they can be used in order to train personnel in the evaluation of nutritional status.The goal of the study is, however, of the utmost importance when selecting the optimal method for dietary assessment.Interestingly, Goldherg and Black (1998) recently reviewed how best to analyse and interpret reported energy intake data (3).
As Beaton has so rightly stated (4) there is not and will probably never be one single golden method for estimation of the dietary intake without error.Furthermore, different types of errors have different effects on analysis and interpretation.Consequently, the selection of method for dietary assessment must firstly depend on an analysis of the aim of the study, and secondly, data collected by means of one dietary assessment aimed to study the intake of one nutrient may not necessarily be as valid for analysing the effect of another nutrient.When discussing the relevance of conclusions drawn from obtained results optimal for any new, additional approach.

Validation
It is common that authors use various methods for dietary assessment in order to validate their findings.Thus data from frequency questionnaires in studies of larger populations have been validated by 24 hour recall interviews performed in a limited number of cases, or in some cases by means of 24 hour record methods.However, this capnot be considered an accurate form of validation as both methods may be characterised by a high risk of underestimation.

Energy turnover an objective reference
As it can be stated that the energy equation is a first-order equation, i.e. energyin = energyo,, and as the body gives priority to covering its energy needs, the dietary energy intake must cover energy needs when body weight and composition are maintained, i.e. available energy (endogenous + exogenous) must balance energy turnover and changes in energy stores.
If a person is in energy balance, which must be the case if bodyweight is stable and body composition unchanged during a certain length of time, then energy intake must balance energy turnover.

Calculation of BMR
Based on large population studies, equations have been established in order to calculate basal metabolic rate (BMR) with reasonable accuracy based on anthropo-How to validate dietary assessments metric data (weight, height, age and sex) (1).The energy turnover comprises BMR, as well as the energy needed for daily life, including characteristics of lifestyle and physical activity.
The daily activities can then be compensated for by expressing the lifestyle in relation to BMR.If various activities in daily life, as well as during various forms of physical activities, are expressed as multiples of BMR, the total energy turnover (ET) per 24 hour can be calculated based on the physical activity level (PAL) of the individual.
It has been postulated that for survival, 24 hour energy turnover represents about 1.27 times BMR, and for a sedentary lifestyle, total energy turnover represents about 1.55 times BMR.An objective and reliable evaluation whether obtained data of dietary energy intake from dietary assessment are realistic must be performed.A comparison between observed or registered energy intake (EI) and the theoretical calculations of ET based on calculated BMR with or without the addition for the lifestyle, using a relevant PAL factor, may indicate whether there might be an underevaluation (5).
Table 1 illustrates the flow sheet for estimation of energy turnover in an individual, including an example.

Forms of compensation
As studies of dietary intake result in an underevaluation, the use of various forms of compensation for energy adjustments has been proposed.However, such adjustments may involve potential serious errors per se.Firstly, differences in error structure between various nutrient vectors make energy adjustments hazardous.To what extent can we assume that the underestimation of dietary intake is the same for all nutrients?If the fat energy percentage is important, can we assume that this is the same for the missing energy units consumed but not reported?In epidemiological studies, food frequency instruments are usually used.In these studies it is quite obvious that because of differences in error structure between nutrients estimated by the same instrument (food frequency forms) the use of energy adjustments might lead to serious errors.
The underreport of energy intake was confirmed by Schoeller et al.  however, showed that it is possible to obtain congruence in estimating energy turnover from dietary assessments using dietary records and doubly labelled water technique at high energy turnover.He compared data from the doubly labelled water technique with those obtained in skilled dieticians using dietary records over a 7-day period in top athletes.However, he was also able to show that each individual had almost the same dietary intake per day, while the physical activity, expressed as training time, varied c o n .siderably from day to day.Thus there was no congruence between the dietary intake and physical exercise for shorter periods, i.e. 24 h.Branth and his collaborators (9) studied energy turnover in crew members participating in an offshore sailing race.Using food inventories before and after a 25-day-sailing period, in combination with measurements of changes in body composition data to evaluate endogenous energy release, they showed that energy turnover could be calculated with accuracy, when related to data by means of doubly labelled water technique (Table 2).

The risk of computerised statistical analyses
There is a risk of forgetting the validity of primary data and the physiological significance of the findings when computerised statistical analyses are performed.The present interest in the evaluation of diethealth relationship, based on epidemiological studies using modem and advanced statistical methods in computers, has resulted in a number of statements, the significance of wfiich can be questioned.
Furthermore, it is unfortunately often forgotten that epidemiological analysis can only be used to formulate hypotheses and not for qualitative or quantitative analysis of diet-health relationship.It should also be remembered that before any evaluation of dietary data can be performed to find any causal explanation for diet-health relationship using any statistical or epidemiological method, the basic physiological relevance of the primary datamust first be assessed.Thus data from any dietary assessment, which does not cover the minimal energy requirement  (12) were based on food frequency in which the reported energy intake was only equivalent to or lower than the expected BMR in the actual age group.This must mean 30 to 50% underestimation.If individuals do not lose body weight for a substantial number of months, it may be questioned whether it is possible to draw any conclusions regarding the relation between the intake of certain nutrients and the health panorama without knowing the composition of the "missing" energy units.In addition, it could be questioned to what extent food a based on food inventories frequencystudies during the last year can (total consumption divided by number of crew members) be considered to be representative of food based on total energy turnover minus endogenous energy released habits during the last 2-3 decades, which is " based on estimations of changes in body composition a must if long-term relationship between based on determinations by the doubly labelled water technique dietary habits and health panorama is to be considered.

Conclusions
Nutrition science is hampered by the fact based on BMR and a suitable PAL-factor, units in the dietary intake are coming.that there is a questionable precision in cannot be considered to be accurate for Interestingly, this has not made the authors most methods for dietary assessments.epidemiological analysis of relationship hesitate todraw conclusions in such classi-The theoretical calculations on BMR in between dietary intake and health.It is cal studies as those from the Framingham combination with the use of PAL-factor as thus recommended that validation mea-study (10).
indicator of energy turnover, however, sures, consisting of calculated EI/BMR 1n-a study of the dietary energy intake in seem to represent surprisingly well reliratios, should be included in dietary sur-2,500 individuals (mean age around 55 able data, although there is still some risk veys (5).years) they reported 2,150 kcal in men and of underevaluation of energy turnover 1f the data are considered as repre-1,580 kcal in females.If these data were to even using these calculations.This, in sentative in a long-term perspective, energy be related to a calculated BMR based on addition to the fact that the body gives balance must have been obtained, i.e. if the 70 kg bodyweight in male and 60 kg priority to cover energy needs and that data do not cover minimal energy need for sedentary life (BMR x 1.55) they must represent an underestimation.It is then necessary to find out whether the missing energy units are covered from a qualitatively identical diet or whether other sources of energy have been consumed and not recorded.It is then self-explanatory that no conclusions can be drawn regarding the relation between fat intake and health parameters, as long as it is not known from where the missing energy bodyweight in females, the BMR factor would be 1.2 to 1.3, which indicates that there must have been an underestimation, as the BMR factor for a sedentary normal population is given as 1.55, but usually is found to be between 1.6 and 1.8 in a normal population (5).
Recently published epidemiological studies on the relation between fat intake and breast cancer (1 1) have been based on food frequency studies where the mean energy intake also corresponded to a BMR and post-obese subjects: comparisons with validation against presumed energy requirements.Eur J Clin Nutr 1997;51:405-13.6. Schoeller DA: How accurate is self-reported energy intake?Nutr Rev 1990;48:373-9.7. Bratteby L-E, Sandhagen B, Fan H, Samuelson G: Total energy expenditure and physical activity as assessed by the doubly labeled water method in Swedish adolescents in whom energy intake was underestimated by 7-d diet records.Am J Clin Nutr 1998;67:905-11.8. Sjodin A, Andersson A, Hogberg J, Westerterp K: Energy balance in cross-country skiers: a study using doubly labeled water.Med Sci Sports Exerc 1994;24:720-4.9. Branth S, Hambraeus L, Westerterp K, Andersson A, Edsgren R, Mustelin M, Nilsson R: there must be a balance between energy intake, energy turnover and energy stores, makes prediction of the EI/BMR ratio a suitable way to validate food consumption data.In case energy needs are not met by dietary intake according to this assessment, and body weight is maintained, it is not scientifically relevant to draw any conclusions regarding the relationship between the dietary intake of various nutrients and health in a long-term perspective.
( 6 )  in 1990 when they analysed energy turnover by means of the doubly labelled water technique and related their findings to results obtained using dietary recall and record methods.Similar observations have been reported by Bratteby et al. (1998) in their studies on Swedish adolescents using 7 d dietary records (7).Sjodin et al. (8),

Table 1 . How to validate dietary assessment by comparing energy intake (EI) and energy turnover (ET) with BMR, including one practical example.
EI / (BMR x PAL) = 0.76 Comments: This example illustrates that if the dietary assessment is right, there is no room for physical training, as the EIIBMR-ratio is 1.6, e.g.only covers the normal energy turnover in an individual with a sedentary lifestyle.This is quite diferent,from his PAL, according to lifestyle and training record(2.5).