Selective v. universal inquiry into the World Cancer Research Fund/American Institute for Cancer Research recommendations

There are at least fifty aspects of the WCRF/AICR recommendations which can be identified for evaluation that have no clear guidance on assessment methodology (see Appendix 1). In the initial parsing of the WCRF/AICR recommendations and planning of instrument development, the investigators of the four-nation Concordance Study⁽Reference Vossenaar⁹⁾ (Table 3) set out all of the fifteen major and fifty-two minor elements as criteria from which operative criteria could potentially be made (shown in the first two columns of the matrix in Appendix 1). Once the data of the two European sites were analysed⁽Reference Vossenaar⁹⁾, the operationalisation criteria developed were adapted to the four geographical sites (i.e. the Netherlands, Scotland, Mexico and Guatemala).

Table 3 Summary of the 1997 World Cancer Research Fund/American Institute of Cancer Research recommendations and the ability to assess them by means of the field-study and case–control study questionnaires used by Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Solomons and Anderson⁸⁾*

* Reproduced with permission from Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Solomons and Anderson⁸⁾.

† Variables are ranked as follows: +, evaluated; +/ − , could be evaluated; − , not evaluated.

‡ Assessed via anthropometry and total energy expenditure estimation.

The matrix developed by Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Solomons and Anderson⁸⁾, examining the data available from a survey and a case–control study, determined that between ten and eleven elements could be evaluated (Table 3). These were generally based on the criteria boundaries of the population goals of the WCRF/AICR⁽⁴⁾. Omitted from consideration among the original WCRF/AICR recommendation levels were those related to salt, fungal contamination, spoiling, chemical contamination, charring of fish and meat, and dietary supplements.

Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾ set forth the components of the recommendations to be assessed in the women from the Iowa cohort in tabular form (Table 4). These were developed from both the individual guidelines and population goals of the WCRF/AICR⁽⁴⁾. They included nine of a possible fifteen levels, with conscious omission of recommendation ‘Food supply and eating’ and recommendations 10 to 14 (i.e. ‘storage’, ‘preservation’, ‘additives and residues’, ‘preparation’ and ‘dietary supplements’). Recommendation 1 partly overlaps with recommendation 4 ‘vegetables and fruits’ and 5 ‘other plant foods’, so that it is partly covered.

Table 4 Summary of the 1997 World Cancer Research Fund/American Institute of Cancer Research (WCRF/AICR) recommendations and the operationalisation criteria and evaluation categories used by Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾*

* Adapted from Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾.

† 5 kg.

Evaluation criteria for population goals

The population-level recommendations of the WCRF/AICR⁽⁴⁾ are expressed as goals for populations. Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾ with archival data from a Guatemala field survey and Vossenaar⁽Reference Vossenaar⁹⁾ with the prospective inquiry in the two European sites of the Concordance Study⁽Reference Vossenaar⁹⁾ used as operative criteria the numerical target goals of the WCRF/AICR⁽⁴⁾. They compared sample means with the population goals, an approach recommended by Nishida et al. ⁽Reference Nishida, Uauy and Kumanyika¹⁰⁾ and described in detail below. Because of their specific interest in individual classification, Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾ did not use this level of classification.

In each of the two-nation Concordance Study (Vossenaar⁽Reference Vossenaar⁹⁾) and the Santa Rosa Study (Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾) a total of seven population goals (or their subcomponents) were examined, six of these being common to both studies. For five of these goals, evaluation methods and cut-off criteria used were similar. These included five of the recommendations with quantitative goals for food items as energy percentage (i.e. ‘to consume 45–60 % of energy from starchy or protein-rich foods of plant origin’; ‘refined sugar to provide < 10 % of energy’; ‘to restrict intake of alcoholic drinks to less than 5 % total energy for men and less than 2·5 % total energy for women’; ‘red meat (beef, lamb, pork) to provide < 10 % of energy’; ‘total fats and oils to provide 15 % to no more than 30 % total energy’).

The recommendation ‘to consume nutritionally adequate diets’ was examined in both studies, but twelve micronutrients were examined in the two-nation Concordance v. eight in the Santa Rosa study⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾. In addition, different nutrient adequacy criteria were used.

In addition, Vossenaar⁽Reference Vossenaar⁹⁾ assessed the BMI goal (i.e. ‘population average body mass indices throughout adult life BMI 21–23 (individual 18·5–25))’ and Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾ assessed vegetable and fruit consumption (i.e. ‘to consume ≥  7 % of energy from vegetables and fruits’).

Evaluation criteria for individual guidelines

The individual WCRF/AICR guidelines (see Table 1) were partly evaluated, along with population goals, in the Iowa study⁽Reference Cerhan, Potter and Gilmore⁶⁾ and the two-nation Concordance Project analysis⁽Reference Vossenaar⁹⁾. There was both homology and discrepancy in the application of criteria.

The individual guideline components evaluated by both authors were six in total. Of these, four were assessed using the same evaluation method (i.e. ‘consume five or more portions of vegetables and fruits a day’; ‘consume 600–800 grams of cereals, pulses, roots, tubers and plantains’; ‘limit alcoholic drinks to less than two drinks a day for men and one for women’; ‘limit intake of red meat to less than 80 grams daily’). However, of these four, only two were assessed using the same cut-off criteria. For example, for the recommendation to ‘consume five or more portions of vegetables and fruits a day’ the Iowa study used five servings per d as the cut-off value without specifying portion sizes, whereas the two-nation Concordance Study⁽Reference Vossenaar⁹⁾ used portions per d and set the portion size at 80 g. For the recommendation to ‘consume 600–800 grams of cereals, pulses, roots, tubers and plantains’ the Iowa study set the cut-off value at 400 g whereas the two-nation Concordance Study⁽Reference Vossenaar⁹⁾ had a lower limit of 600 g and an upper limit of 800 g.

The recommendation to ‘limit weight gain during adulthood to less than 5 kg’ was assessed as weight gain since age 18 years by both Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾ and Vossenaar⁽Reference Vossenaar⁹⁾. In addition, Vossenaar⁽Reference Vossenaar⁹⁾ used weight gain in the previous 10 years as a proxy for weight gain during adulthood. Further details are provided in Appendix 1.

Because the Concordance Study⁽Reference Vossenaar⁹⁾ was prospectively designed to evaluate the WCRF/AICR recommendations⁽⁴⁾, the study was able to incorporate the evaluation of a wider range of detail and nuance than could be derived from standard food intake investigations. An additional eleven recommendation components at the individual level were evaluated. A total of four recommendations (‘food supply and eating’ (recommendation 1); ‘maintaining physical activity’ (recommendation 3); ‘preparation’ (recommendation 13); ‘dietary supplements’ (recommendation 14)) were only assessed by Vossenaar⁽Reference Vossenaar⁹⁾ and most of the subcomponents were included.

Experience using the population-goal criteria as cut-off criteria for individual concordance

A hybrid format for evaluation was used by all three investigations⁽Reference Cerhan, Potter and Gilmore^6, Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo^7, Reference Vossenaar⁹⁾ and was created by applying the WCRF/AICR population-goal criterion (Table 1), to individual subjects in the study, in order to classify each individual as being inside or outside of the recommended range. In fact, these criteria are population goals to be applied as target averages in relation to the mean of variables assessed at a whole-population level. However, the authors generated prevalences or frequencies of individuals ‘non-concordant’ with the population cut-off criteria. We have named this pseudo-individualisation of population goals. In Appendix 1 this form of analysis has been indicated below the population-goal recommendation where appropriate as ‘Pseudo-individualised population goal’.

Nishida et al. ⁽Reference Nishida, Uauy and Kumanyika¹⁰⁾ came to grips with population goals (Fig. 1) in the context of the WHO/FAO consultancy that produced Technical Report 916 Diet, Nutrition and the Prevention of Chronic Disease ⁽¹¹⁾. Their concept is related to the issue of nutrient-intake goals for a population, which in turn are ‘based on the assumption that the first priority is to ensure national food security and equity of distribution of available food in accordance with individual needs’. This approach is essentially analogous to the concept of estimated average requirements (EAR) of the US dietary reference intake⁽Reference Murphy and Poos¹²⁾. Nishida et al. ⁽Reference Nishida, Uauy and Kumanyika¹⁰⁾ addressed the potential consequences and distortion if the population-goal values are applied as individual criteria as described above. They state ‘the Joint WHO/FAO Expert Consultation stressed that because population nutrient intake goals refer to population averages, the aggregate change would be substantially greater than intended if they were to be applied to the diets of individuals’.

Fig. 1 Distinction between a population goal and an individual goal. The distribution of free sugars and SFA for three hypothetical populations, corresponding to: (1) an existing population whose intake of free sugars and SFA is higher than recommended by the joint WHO/FAO Expert Consultation⁽¹¹⁾ (right), (2) a population that has achieved the population cut-off of 10 % of energy as free sugars and SFA (centre), and (3) a population in which nearly all individuals have an intake less than 10 %, a situation that describes the result of misrepresentation of the population goal as an individual goal (left) (adapted from the report of the WHO Study Group⁽¹¹⁾). Reproduced with permission from Nishida et al. ⁽Reference Nishida, Uauy and Kumanyika¹⁰⁾.

The concept of population goals as they applied to ‘healthful eating’ to reduce the risk of cancer, however, are much more nuanced. For fruits and vegetables (recommendation 4), there is a strong analogy to an essential nutrient's intake goal, in which the target is something to be achieved for a positive end. Conversely, the goals with respect to sugar (recommendation 5d), alcohol (recommendation 6) and red meat (recommendation 7) are clearly not about nutrient or food security, but in the domain of tolerance of potentially harmful dietary substances. It is not to ‘achieve’ the target goal, but rather not to exceed it. In fact, this is closely aligned to free sugar and SFA, which are nutrients more of concern for their undesirable effects than their essentiality.

The WCRF/AICR population goals expressed as a numerical range (as for other plant foods (recommendation 5) and total fats and oils (recommendation 8)) are a species of hybrid, which recognise an essentiality (obligating a certain intake level) but also potential harm when consumed in excess.

All three studies⁽Reference Cerhan, Potter and Gilmore^6, Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo^7, Reference Vossenaar⁹⁾ engaged in this pseudo-individualisation of population goals for recommendations. Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾ used this approach to assess individual concordance with four population-goal components. These were: ‘population average body mass indices throughout adult life BMI 21–23 (individual 18·5–25)’ (recommendation 2); ‘active lifestyle equivalent to a physical activity level (PAL) of at least 1·75’ (recommendation 3); ‘total fats and oils to provide 15 % to no more than 30 % total energy’ (recommendation 8); ‘salt from all sources should amount to less than 6 grams/day for adults’ (recommendation 9).

Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾ used the pseudo-individualisation approach for six components. These were: ‘to consume nutritionally adequate diets’ (recommendation 1); ‘to consume 45–60 % of energy from starchy or protein-rich foods of plant origin’ (recommendation 5); ‘refined sugar to provide < 10 % of energy’ (recommendation 5); ‘to restrict intake of alcoholic drinks to less than 5 % total energy for men and less than 2·5 % total energy for women’ (recommendation 6); ‘red meat (beef, lamb, pork) to provide < 10 % of energy’ (recommendation 7); ‘total fats and oils to provide 15 % to no more than 30 % total energy’ (recommendation 8).

The above approach was used in the two-nation Concordance Study⁽Reference Vossenaar⁹⁾ where it was used to assess eleven population-goal components. These included the components evaluated by Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾ and Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾ described above (with the exclusion of 1a and salt) and the following population goals: ‘to consume varied diets’ (recommendation 1); ‘to consume ≥  7 % of energy from vegetables and fruits’ (recommendation 4); ‘to consume a variety of starchy or protein-rich foods of plant origin’ (recommendation 5); ‘when meat and fish are eaten, encourage relatively low temperature cooking’ (recommendation 13). The only component evaluated in this manner by all three authors was the population goal for ‘total fats and oils to provide 15 % to no more than 30 % total energy’ (recommendation 8). The reason for this approach is that the individual guidelines for fat do not include quantitative goals.

Three population-goal components were assessed by both Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾ and Vossenaar⁽Reference Vossenaar⁹⁾ using the pseudo-individualisation approach. For these three components in common, different evaluation methods and cut-off criteria were used. Vossenaar⁽Reference Vossenaar⁹⁾ and Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾ used this method for seven components in common and used the same evaluation methods and cut-off criteria for all. The only component assessed in this manner by both Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾ and Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾ was ‘total fats and oils to provide 15 % to no more than 30 % total energy’ (recommendation 8), but different cut-off criteria were applied. A minimum value of 15 % energy from fat was set in the two-nation Concordance⁽Reference Vossenaar⁹⁾ and Santa Rosa studies⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾, whereas in the Iowa study, subjects with a fat intake below 30 % energy were not penalised.

Quantitative challenges in interpreting the recommendations

The major confounders for the evaluator trying to determine whether an individual – or a population – is following the WCRF/AICR recommendations are a series of quantitative inconsistencies between population goals and individual guidelines, further complicated by footnotes accompanying the original publication of the WCRF/AICR report⁽⁴⁾. Quantitative discrepancies in the recommended intake limits can be found in at least five recommendations according to the units chosen (Table 5).

Table 5 Quantitative inconsistencies of non-equivalency in the expression of guidelines and goals in the World Cancer Research Fund/American Institute of Cancer Research (WCRF/AICR) 1997 report's recommendations⁽⁴⁾

* 0·72 g/1000 kJ.

† 8368 kJ/d.

The following examples illustrate some of the potential problems:

An adult woman who reports consuming 850 g starch or protein-rich foods of vegetable origin would not be concordant with the individual guideline to consume between 600 and 800 g cereals, pulses, roots, tubers and plantains, since she would be exceeding the recommended range. However, she would be classified as adhering if the seven-portion criterion of the same foods, which has no upper limit, were taken as the operative cut-off.

An adult man who reports consuming 820 g fruits and vegetables per d would be above the acceptable range, with its 800 g upper limit. However, if the operative criterion were based on the expression in ounces, his 820 g would classify his intake as appropriate. The latter classification would also apply with the more-than-seven portion criterion, which is open ended, and has no upper limit. In a practical, survey sense, if data are clustered close to the boundary criteria levels, prevalences of concordant and non-concordant individuals in a sample could vary greatly depending on which of the expressions have been used to create the operative criteria for evaluation.

Quantitative directions

The primary recommendations calling for dietary change range from widespread vagueness, lack of quantitative specificity (or boundaries) to internal numerical inconsistencies when intake levels are provided. The three investigative groups⁽Reference Cerhan, Potter and Gilmore^6, Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo^7, Reference Vossenaar⁹⁾ have grappled with setting operative criteria and have often arrived at diverse cut-offs for the same recommendation components.

Quantitative uncertainty might evoke some confusion on the part of a consumer attempting to follow the guidance of the WCRF/AICR guidance. They might ask themselves: ‘When is my diet varied? What is minimally processed food?’ For the consumer these descriptions are generally subjective in their interpretation. For the evaluator, trying to classify behaviour as concordant, specific quantification is a requisite and some established cut-off criterion must be applied. Similarly, although a consumer may understand ‘year round’ and balance consistent intake across the year, establishing the continual nature of an eating behaviour in an evaluation instrument proved to be exceptionally challenging, requiring staggering of the survey throughout a calendar year. The vagueness of language also applies to individual recommendations such as ‘limit consumption’ and ‘consume modest amounts’ as in recommendation 8.

For at least a single recommendation (on sugar consumption) there was a general default consensus on the part of all three investigative groups⁽Reference Cerhan, Potter and Gilmore^6, Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo^7, Reference Vossenaar⁹⁾ to use the boundaries of 10 % energy contribution for sugar. Cerhan et al. ⁽Reference Cerhan, Potter and Gilmore⁶⁾ also applied the 6 g of salt boundary from the population-goal level to individuals, while the other two groups⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo^7, Reference Vossenaar⁹⁾ felt that their instruments could not quantify salt or Na consumption.

Recommendation 1 illustrates an interesting conflating of the population goal and the individual guideline in the semantic sense. For example, the phrase ‘choose predominantly plant-based diets’ for individuals could be interpreted as anything greater than 50 % of intake from foods and beverages of plant origin or it could be calculated on a food-weight basis or a food-energy basis; the framers provide no guidance. Vossenaar⁽Reference Vossenaar⁹⁾ operationalised this cut-off on a weight (g food) basis given that individual guidelines are expressed in g amounts (Table 1). If we look at the corresponding population goal, ‘based primarily on foods of plant origin’, we would have to take the same semantic interpretation for a target group average, as we have for an individual goal. Hence, a population with the vast majority of its members being individually concordant would have a distribution for the plant:animal ratio shifted far to the left of a normative curve for which 51 % was the population mean.

At least two phrases in the individual recommendations leave doubt as to whether they represent health-promoting obligations or merely alternative, helpful suggestions. For recommendation 8, on meat consumption the individual guideline states ‘it is preferable to choose fish, poultry or meat from non-domesticated animals in place of red meat’. For recommendation 9, on salt and salting, the individual recommendation text reads ‘use herbs and spices to season foods’. A common-sense interpretation of this language might be that they merely constitute ‘menu suggestions’, with no imperative force. Vossenaar⁽Reference Vossenaar⁹⁾, at least, actually applied an operative criterion for the recommendation to prefer alternative meats (see Appendix 1), determining whether 51 % of flesh intake was in the alternate classes of fish and seafood or fowl. Since the salt recommendation was not evaluated by Vossenaar⁽Reference Vossenaar⁹⁾, the issue of spices substitution remains moot.

Table 5 illustrates that the varying ways of rephrasing some of the explicitly numerical goals do not provide strict quantitative equivalency. An individual who had all of their drinks of alcoholic beverages as wine would consistently have 40 % greater ethanol intake than one who consumed spirits. Equally troubling for the choice of evaluation criteria are the open-end v. the bounded criteria for daily intakes of healthful food groups such as fruits and vegetables (recommendation 4) and other plant foods (recommendation 5). Of course, there are reciprocal interactions within an individual consumer based on anatomical and physiological considerations. With a conventional three or four food intake eating pattern, satiety and gastric capacity would constrain consumptions of upwards of 1600 g of high-bulk and watery food items.

Staying with recommendation 4, the WCRF/AICR report⁽⁴⁾ specified that only a single 80 g serving of a natural fruit juice could be counted toward a day's fruit and vegetable allotment. The remainder had to be in a solid form. The underlying epidemiological or physiological basis for the non-equivalence of juiced or solid fruits and vegetables is not clear. Recently, Ruxton et al. ⁽Reference Ruxton, Gardner and Walker¹³⁾ reviewed the literature with a tentative conclusion that natural juices have the same preventive potency against chronic disease as their solid counterparts.

Reforming approaches to set operative criteria for population goals

Nishida et al. ⁽Reference Nishida, Uauy and Kumanyika¹⁰⁾ provide a beacon of clarity for what could be considered an orthodox approach to assessing populations for their concordance or compliance with population goals, when they are expressed in quantitative terms as a target population mean. A corollary of the Nishida logic, moreover, is that a representative sample of a population or subpopulation would be ideal for the application of population-goal criteria. Hence, in this regard, all of the publications considered are convenience samples and would not strictly apply as subjects of an orthodox population-goal analysis. This caveat notwithstanding, Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾ generally pursued the population-average assessment. Vossenaar⁽Reference Vossenaar⁹⁾, on the other hand, applied the population goal both as a target average at the whole-sample level and as a cut-off criterion at the individual-subject level.

At the level of assessing recommendation 1, related to dietary adequacy, both Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾ and Vossenaar⁽Reference Vossenaar⁹⁾ assessed individuals against the respective national standards for individual intake (for example, intake levels that would meet the requirements of 95–97·5 % of the healthy population). The population was assessed, however, as the percentage of individuals achieving national standards, which is the concept utilised within the EAR for a reference population⁽Reference Murphy and Poos¹²⁾, which has been adopted by the Food and Nutrition Board of the Institute of Medicine of the United States in its Dietary Reference Intakes ⁽¹⁴⁾. These EAR represent target mean nutrient intakes for normal, free-living North Americans in the USA and Canada. More recently, an adaptation of EAR for the global population has appeared within the UN system⁽¹⁵⁾. The EAR would be applied as population-goal intakes to evaluate the adequacy of the diet, in compliance or concordance with the first element of recommendation 1.

Recommendation 3 calls for ‘an active lifestyle equivalent to a physical activity level (PAL) of at least 1·75, with opportunities for vigorous physical activity’. Consistent with the Nishida-derived approach, the population-goal evaluation would seek to determine if the population sample had a physical activity level of 1·75. Based on the age, sex and body size of individuals in the population, one could estimate the underlying basal energy expenditure. The prescription of the WCRF/AICR⁽⁴⁾ would be that at least half the population expends energy in physical activity of three-quarters or more of their basal energy expenditure. Although diagnostic techniques to estimate total energy expenditure and expenditure due to physical activity exist⁽¹⁴⁾, none of them would be practically applied to a free-living population. Until a valid and reliable method for assessing individual activity-related energy expenditure is developed, WCRF/AICR recommendation 3 may elude evaluation at the population level. The second component criterion of ‘opportunities for vigorous physical activity’, interpreted literally, is not about behaviour; rather, it is about the social–physical environment. If, for instance, it is a social need for a rural adult population to carry water or chop firewood, then opportunities would be in place. For the rural middle class, the availability of exercise gymnasiums or running paths in the park would set the basis for meeting this ‘opportunities’ component criterion.

Recommendation 9 relates to salt and salting. Parsing the structure of the phrasing: ‘Salt from all sources should amount to less than 6 grams/day (0·25 ounces) for adults’, this clearly seems to be about an upper limit boundary criterion. The language of the footnote for children, which specifies less than 3 g/1000 kcal (0·72 g/1000 kJ), is consistent as a boundary prescription. For a population to be compliant, therefore, it would not be half of the population – but rather all of the population – to have a salt intake below the criterion. There is some confusion, however, between the reference to ‘salt’ (sodium chloride) from all sources, as referred to by the WCRF/AICR⁽⁴⁾, as all sources of ‘sodium’; Asian cuisine, of course, uses other Na salts, notably sodium glutamate, which would not be covered in any way by the present phrasing of recommendation 9's population goal. In terms of an assessment of compliance or concordance with the salt and salting recommendation, surveying a representative population sample for 24 h Na excretion could provide the classification. It is known from basic physiology that individuals in normal Na balance excrete over a 24 h period the amount of Na that they take in through their diet⁽Reference Melse-Boonstra, Rozendaal and Rexwinkel¹⁶⁾. It would be an expensive and cumbersome task to collect complete 24 h urine collections in a free-living setting. Moreover, the population falls out of concordance if a single individual in the sample is found to have more than the equivalence of 6 g of salt's worth of Na in his or her urine specimen. For children, the logistics would be far more severe, with the need to assess dietary energy intake and make accurate urine collections in the more challenging setting of a juvenile population.

It seems possible that recommendation 10 which states, ‘store perishable food in ways that minimize fungal contamination’ can be evaluated at the population level, but not in any survey of individual behaviour. The guidance is aimed at avoiding exposures to carcinogenic mycotoxins. For a given population, exposure would be conditional on the appropriate ambient climatic conditions for fungal growth and for reliance of the population on large intakes of susceptible grains or legumes. In the case of this recommendation, the locus of evaluation would not be in a surveyed population, but rather through systematic evaluation of the operative storage practices.

With respect to recommendation 11, the mandate to keep perishable foods chilled or frozen is inherently dependent on the infrastructure for refrigeration available to a society, and its appropriate use to preserve foods and prevent generation of carcinogenic toxins. In the evaluation, a proxy measure would be the reach and efficiency of refrigeration. On the one hand, if there are large gaps in the cooling and freezing systems, then the criterion could not be met. So, a systematic distribution of refrigeration could be the point of entry for population screening. If the population infrastructure is in place, proper use of refrigeration, such as temperature and storage of potentially problematic food items (for example, leftovers) are difficult to assess.

The status of exposure of a population to the carcinogenic effects of additives, pesticides, pesticide and other chemical contaminants is the basis of recommendation 12. However, such exposures are usually beyond the scope of the routine dietary survey investigations. Hence, this recommendation was omitted from consideration in the design of the Concordance Study⁽Reference Vossenaar⁹⁾ for Europe as well as that of Valdés-Ramos et al. ⁽Reference Valdés-Ramos, Cervantes and Mendoza-Perdomo⁷⁾ in Guatemala. In a broader consideration, however, there are sources for inquiry for the characterisation of a population as in or out of compliance with established international and local standards. This would begin at the various ministries in a country ranging from health and agriculture to environmental safety, and the specific regulatory bodies within these governmental entities. In the era of international food trade, the caveat of determining practices in the places of food production and processing, which might be overseas, arises as a final challenge in addressing this point.

Summary and reflections

Investigators need to continue to measure compliance and concordance with the WCRF/AICR tenets in free-living individuals. Such analysis will assist in the monitoring of dietary and health trends across populations, help to identify population-specific health education targets and also evaluate health interventions. Since the submission of the present paper for publication, the WCRF has published a new set of eight recommendations at two levels: public health goals and personal recommendations⁽⁵⁾. In this new report, population goals are set within a time range of 10 years, for example, ‘population average consumption of sugary drinks to be halved every 10 years’. This makes the monitoring over a 10-year time frame essential for the assessment of concordance and provides a further challenge for nutritional epidemiologists.

Agencies who make recommendations and guidelines also need to think about monitoring procedures and to identify practical ways of working with the research community. The process of operationalisation of evaluation criteria for dietary recommendations is not new. Assessing concordance with healthy eating guidelines is problematic, and similar difficulties are encountered in, for example, the ‘Healthy Eating Index’⁽Reference Guenther, Krebs-Smith and Reedy¹⁷⁾ which assesses concordance with the American dietary guidelines ‘My Pyramid’⁽¹⁸⁾. Similar questions, such as ‘what does < 10 % energy actual mean?’ arise. The WCRF/AICR recommendations⁽⁴⁾ are global, making assessment more complex because they include more than just food-based dietary guidelines, i.e. physical activity, body weight, food preparation methods, etc. Then again, assessment is simplified by the unity of recommendations across sexes (with the exception of alcohol), age and body weight. Most assessment instruments have a scoring protocol, which raises issues such as which components are relevant within a given population, the weightings given to each behaviour and possible modifying effects of different behaviours (for example, can high levels of activity reduce the risk posed by inappropriate dietary choices).

The WCRF/AICR recommendations are aimed at guiding both individuals and populations (public health); assessment procedures need to take account of this. Specific applications dictate the locus of analysis. For regression analyses, such as in the Iowa study⁽Reference Cerhan, Potter and Gilmore⁶⁾, in which an individual outcome (cancer diagnosis) is the variable, evaluation must take place at the individual level. When it comes to populations, both options are theoretically possible. For instance, until the advent of the dietary reference intakes⁽¹⁴⁾, it was standard practice in dietetic research to report the prevalence of individuals exceeding – or failing to meet – the RDA criterion⁽¹⁹⁾. This manner of assessing populations has given way, however, to the comparison of the measured group average with the EAR⁽Reference Murphy and Poos¹²⁾. The analogous advocacy for evaluating health guideline behaviour in populations must make reference to the population goals (such as is found in the work of Nishida et al. ⁽Reference Nishida, Uauy and Kumanyika¹⁰⁾ (Fig. 1)). So, it is not a matter of ‘either, or’ but rather how each level of recommendations can best be probed by appropriate survey instruments.

There are no simple answers for the design of food-frequency instruments aimed at assessing concordance of actual behaviours with recommended behaviours. However, sharing difficulties and building on the tools that have been developed from the four-nation Concordance Project should help the process to evolve. The jury is still out as to whether adapting the outputs of population surveys for evaluations of concordance to cancer-prevention recommendations or the design of specific questionnaires to test the recommendations will be the most appropriate approach. However, further judgements on the evaluation process will be possible following publication in the literature of the results of the Concordance Project.