Issue 1: What counts towards fruit and vegetable intake?

The first issue relates to what is defined as being included in fruit and vegetable intake when intake is assessed, specifically what items that are consumed by people will be categorized as contributing to fruit and vegetable intake. In order to estimate fruit and vegetable intake a definition must specify what plant foods count towards such intake and what quantity corresponds to a serving if servings are being calculated⁽Reference Smith-Warner, Elmer and Fosdick²⁹⁾. The former seems pretty straightforward as it would seem clear what foods are in the category ‘fruits and vegetables’. However, that term covers a highly heterogeneous group of foods, especially across cultures and geographic locales⁽Reference Pomerlau, Lock and McKee³⁰⁾. Also, researchers often will include foods that are not technically fruits and vegetables or exclude certain foods from their definitions even though those foods are technically a fruit or a vegetable, the latter through either inadvertent omission or purposeful omission.

Legumes (dried beans and peas), which are not by definition fruits or vegetables, are often included in definitions of what is included in calculating fruit and vegetable intake, but not always in the same manner. Before MyPyramid, generally all servings of legumes were calculated in vegetable consumption, but with MyPyramid, legumes were first counted towards the meat and beans group and any servings of legumes remaining once that category's requirements were met were then allocated to the vegetable category⁽Reference Krebs-Smith, Guenther and Subar⁵⁾. As previously mentioned, Healthy People 2020 specifically includes legumes along with dark green and orange vegetables as a target category for increased intake⁽²⁷⁾. However, recent major studies in Europe have not included legumes (or potatoes) when calculating vegetable consumption⁽Reference Crispim, Geelen and Souverein³¹⁾. Legumes also potentially present a definition problem when assessment methodologies involve FFQ or screeners that do not specifically indicate in some way to respondents that legumes are to be considered in the fruit and vegetable category, as respondents might not consider them to be vegetables and thus would not count them unless prompted to do so.

An example of inadvertent omission is the definition that was initially used to calculate fruit and vegetable intake for Healthy People 2000. The Healthy People 2000 baseline data on fruit and vegetable intake were revised in part because in the original baseline data analyses, fruits and vegetables that individuals had consumed that were part of a mixed food (e.g. a stew containing meat and vegetables) were not counted in calculating fruit and vegetable intake because mixed foods had not been included in the original definition of fruits and vegetables⁽Reference Krebs-Smith, Cook and Subar²⁴⁾. Many food items that people prepare and consume contain a combination of vegetables, grains and meat or seafood, and thus not counting the fruits and vegetables contained in such foods could lead to significant underestimation of fruit and vegetable consumption.

Researchers often specifically omit from their definitions certain foods that are technically part of the fruit and vegetable category. Even when methodologies for measuring intake are used where all can be accurately counted in the results, such as 24 h recalls, the variable definitions for coding may disallow certain food items that contain fruits and vegetables or that are fruits or vegetables. Exclusion is often based on the idea that some sources of fruits and vegetables (such as French fries) have such a high energy content in relation to nutritive value that they should not be counted when calculating fruit and vegetable intake⁽Reference Cullen, Baranowski and Baranowski³²⁾. How much the behavioural approach affects results depends on what food items are not counted. Cullen et al.⁽Reference Cullen, Baranowski and Baranowski³²⁾ compared these two coding approaches using the same data sets and found that the intakes with the behavioural approach were 5 % to 15 % lower.

What foods that are technically fruits and vegetables are excluded from a definition and the rationale for such exclusions can be quite variable between studies. A few studies have excluded salads such as coleslaw because they are considered to be served in too small a serving size and the nature of the assessment tool cannot account for such small portions, or because of their high fat content they are deemed too nutritionally poor in relation to energy to count⁽Reference Ashfield-Watt, Welch and Day³³⁾. Condiments such as ketchup, and even sandwich toppings, such as lettuce and tomatoes, are also often excluded in the definition of fruit and vegetable intake when servings are directly assessed. In this case the rationale usually given is that the amounts are too small to constitute a serving or even half a serving and that the assessment method is based on number of servings rather than actual weights of everything consumed that is then converted into servings as is the case with 24 h recall methods⁽Reference Ashfield-Watt, Welch and Day³³⁾. Depending on specifically what is being excluded, this appears to be more an issue related to assessment methodology rather than pure definition as certainly lettuce and tomatoes are counted in all studies if they are consumed in larger portions such as in a salad. The logic in excluding condiments is that they may be frequently eaten yet their portion size is so small that to exclude them would have little effect on estimated intake whereas including them could lead to large overestimates of intake.

Excluding fried potatoes, especially French fries, can have a substantial impact on results as one recent study showed that French fries account for a significant percentage of total vegetables consumed, accounting for 14 % of fruit and vegetable intake for adult women, nearly 18 % for adult men, and nearly 30 % for adolescents⁽Reference Guenther, Dodd and Reedy³⁾. Sometimes even non-fried potatoes are not counted towards vegetable intake. The Japanese system for classifying foods puts potatoes in the grain category along with rice and other grains and not in the vegetable category⁽Reference Takata, Maskarinec and Franke³⁴⁾. In the UK's ‘5-a-day’ programme all servings of potatoes are excluded when calculating fruit and vegetable intake⁽Reference Ashfield-Watt, Welch and Day³³⁾. Recent studies in Europe also have excluded potatoes when calculating vegetable intake⁽Reference Crispim, Geelen and Souverein³¹⁾. Since the end goal of increasing fruit and vegetable intake is to improve human health it might be logical to exclude certain foods that are technically fruits and vegetables if their consumption is actually counterproductive to that goal. Conversely, the reason the broad category of fruits and vegetables has been the target for assessment and intervention rather than specific fruits and vegetables or specific nutrients found in fruits and vegetables is that science has yet to clearly identify which components found in fruits and vegetables are crucial to human health⁽Reference Pomerlau, Lock and McKee³⁰⁾. In the context of variety of intake of fruits and vegetables being the goal and therefore the focus of assessment and intervention, excluding fried potatoes, or even all potatoes, may make theoretical sense if the focus is total or overall vegetable intake as opposed to categories of vegetable intake.

A possibly important issue with respect to including potatoes in fruit and vegetable intake is its potential effect on differences in fruit and vegetable intake by ethnicity or cultural background. While potatoes are a traditional staple for many Caucasian groups, they are not a staple for many Asian/Pacific Islander groups. Conversely, rice is a staple for the latter and not the former. A number of studies in the USA report that rice consumption is much higher for Asian Americans than for Caucasian Americans⁽Reference Takata, Maskarinec and Franke^34–Reference Batres-Marquez, Jensen and Upton³⁶⁾. The implication is that in populations where rice consumption is high, such as most Asian and Pacific Islander groups, differences in estimated fruit and vegetable intakes may occur by ethnicity in comparison to Caucasians because potatoes are counted towards fruit and vegetable intake and rice, of course, is not. Again, this would be more of a concern where the assessment evaluates overall or total vegetable intake as opposed to categories of vegetables.

Issue 2: How do you accurately measure fruit and vegetable intake?

There are a number of methods used for obtaining fruit and vegetable intake data from individuals, including 24 h dietary recalls, food diaries, measurement of biomarkers, FFQ and short screeners. The 24 h dietary recall method is often considered the most accurate method for obtaining fruit and vegetable intake data⁽Reference Greene, Resnicow and Thompson^37, Reference Subar, Thompson and Potischman³⁸⁾. However, in the past there have been a number of issues with its use. First and foremost was the fact that both the data collection process and the data analysis process were highly resource-intensive⁽Reference Subar, Thompson and Potischman³⁸⁾. Another issue was and is that dietary intake for a given individual tends to vary greatly from day to day so that a single 24 h dietary recall is usually not representative of an individual's typical intake and therefore multiple recalls are required for studies where an individual's typical intake is important⁽Reference Subar, Thompson and Potischman³⁸⁾. Computer-based, self-administered 24 h dietary recall methodologies have been developed in recent years that may address many of the traditional limitations of this method⁽Reference Zimmerman, Hull and McNutt³⁹⁾. However, more research is needed on these new methods.

Food diaries, in which the individual keeps a record of everything he/she eats over a given period of time, represent another method for assessing fruit and vegetable intake. Like 24 h dietary recall, food diaries have the potential to accurately capture true intake for a given period of time and thus might be a good method for evaluating the association between diet and chronic disease⁽Reference Bingham, Luben and Welch⁴⁰⁾. However, this method can be quite resource-intensive and it may be less adept than FFQ, for example, in assessing usual intake as opposed to intake for a specific day or time period which may not be representative of usual intake⁽Reference Smith-Warner, Elmer and Fosdick^29, Reference Crozier, Inskip and Godfrey⁴¹⁾. Another major concern with food diaries is that accuracy in reporting intake may vary greatly between respondents depending on factors such as education, socio-economic status and other respondent characteristics, or that certain respondent characteristics may be associated with either inaccurate reporting in diaries or with changing eating patterns during recording periods such that reported fruit and vegetable intake is higher than actual usual intake⁽Reference Crozier, Inskip and Godfrey^41–Reference Pietilainen, Korkeila and Bogl⁴³⁾.

Measurement of certain biomarkers such as plasma carotenoids is also used to indirectly assess fruit and vegetable intake in individuals although such measurements tend to be quite specific to certain types or categories of fruits and vegetables and are not necessarily good for evaluating overall fruit and vegetable intake⁽Reference Al-Delaimy, Ferrari and Slimani⁴⁴⁾. Fruit and vegetable intake patterns may also lead to differential results in biomarker levels⁽Reference Garcia, Mohan and Koebnick⁴⁵⁾. Such markers are also affected by numerous physiological factors beyond dietary intake⁽Reference Pollard, Wild and White⁴⁶⁾. Biomarkers might play a more valuable role in validating other assessment methods⁽Reference Bingham, Luben and Welch⁴⁰⁾ and in reflecting physiological nutritional status⁽Reference Pollard, Wild and White⁴⁶⁾ rather than in assessing overall fruit and vegetable intake.

Another methodological approach for gathering data on fruit and vegetable intake is the FFQ. An FFQ typically consists of a pre-coded form containing a large list (sixty to 120) of specific food items that assesses frequency with which each item is consumed and often the usual portion size when consumed. Any given food item may consist of a single food or a group of highly similar foods⁽Reference Teufel⁴⁷⁾. This approach does not require multiple administrations to assess an individual's usual intake⁽Reference Cade, Thompson and Burley⁴⁸⁾. Short screeners, which are really a type of FFQ, are also used for assessing fruit and vegetable intake. These are typically FFQ with fewer than twenty questions that ask about frequency of intake of various categories of fruits and vegetables over a given period of time⁽Reference Bogers, van Assema and Kester⁴⁹⁾. From a research standpoint, a short screener that is highly accurate in both ranking fruit and vegetable intake and assessing actual intake is highly desirable since short screeners can be easily administered to large populations at a relatively low cost⁽Reference Peterson, Hebert and Hurley⁵⁰⁾. However, there is concern that both FFQ and short screeners, while being somewhat accurate in ranking individuals by intake, are not very accurate in assessing actual intake⁽Reference Kim and Holowaty⁵¹⁾, with short screeners tending to underestimate actual intake and longer FFQ tending to overestimate actual intake⁽Reference Smith-Warner, Elmer and Fosdick^29, Reference Peterson, Hebert and Hurley⁵⁰⁾.

A related issue with regard to accurately assessing actual intake of fruits and vegetables is that assessment tools that require respondents to self-determine to some degree what foods they eat are fruits or vegetables might be understood differently by different individuals. One study evaluated respondents’ understanding of assessment tools specifically in relation to what the tools meant to include and not include in defining what constitutes fruit and vegetable intake⁽Reference Thompson, Subar and Smith⁵²⁾. Results indicated that many respondents misunderstood what to include or not include when determining their answers. For example, when queried about a question asking about ‘100 % fruit juice’ intake, many respondents did not understand the question and mistakenly included fruit-flavoured drinks when reporting intake. In a similar type of study respondents underwent cognitive testing regarding understanding of questions using the Behavioral Risk Factor Surveillance System (BRFSS) module and a newly created screener as the source of the questions⁽Reference Wolfe, Frongillo and Cassano⁵³⁾. A number of questions were found to be frequently misinterpreted and there were often ethnic/cultural differences in how a question was interpreted. For example, many whites when asked what they considered to be a ‘bean’ only included green beans, while many Hispanics and African Americans only included legumes. In a more recent study of American adults from various racial/ethnic backgrounds, there were significant differences between racial/ethnic groups on how to classify a number of foods (fruit, vegetable or other) including beans, potatoes and rice. In that same study, nearly a fifth of respondents considered rice to be a vegetable⁽Reference Thompson, Willis and Thompson⁵⁴⁾.