The research presented in this review shows that the scientific effort to identify determinants of fruit and vegetable consumption among children and adolescents has increased substantially over the past four decades growing from a total of 11 published papers from 1958 through 1986 to 54 papers published only since 2000. This increase follows the growing evidence supporting the health benefits from eating ample amounts of fruit and vegetables [1
This review of determinants of fruit and vegetable consumption among children and adolescents reveals that the determinants supported by the greatest amount of evidence are gender, age, SEP, preferences, parental intake, and home availability/accessibility. Girls tend to have a higher or more frequent intake of fruit and vegetables than boys, and a corresponding pattern is seen for the younger age groups compared to the older age groups. As mentioned, solely descriptive prevalence papers were not included in this review. Such papers may contain information on age and gender differences. Therefore more documentation in this area might have been gained by including prevalence papers. SEP, preferences, parental intake, and home availability/accessibility are all positively associated with children and adolescents' fruit and vegetable consumption. In addition, also for nutritional knowledge, self-efficacy and shared family meals the evidence for positive associations is rather convincing.
A large number of potential determinants have been included in the quantitative scientific literature on fruit and vegetable consumption of children and adolescents. However, this review clearly shows that for many variables evidence is lacking. For the majority of the examined variables this lack of evidence is mainly due to lack of studies. For a minority of the identified variables (e.g. gender, age and SEP) a significant number of studies exist. Here, the literature is generally characterised by conclusive and similar findings within a large part of the papers and non-significant associations identified in the remaining papers. Only rarely conflicting results were seen.
Establishing epidemiological evidence for a given association implies the existence of only few cases of contradictory findings. Though being low in numbers, it is still important to analyse reasons for contradictory findings. Naturally, such findings may be due to methodological bias while others may reflect true differences between countries, geographical regions, time periods etc. Thus, contradictory findings should be seriously considered, as they may provide important information needed for creating new scientific hypotheses.
For variables studied in at least three papers included in the present review contradictory findings were observed for gender, age, SEP and nutritional knowledge.
In contrast to most of the papers analysing the influence of gender, four studies observed the highest or most frequent intake of fruit and/or vegetables among boys. Of these four papers, one US study by Burdine et al. (1984) [60
] and one Tasmanian study by Woodward (1985a) [110
] were both rather old publications compared to the majority of the included papers. The two remaining papers by Musaiger and Gregory (1992) [17
] and Rojas (2001) [114
] analysed study populations from Bahrain and Costa Rica (small sample). No other papers from South America and only five Asian papers were identified. The contradictory findings identified for gender may be due to methodological bias or they may reflect true associations existing in US and Tasmanian societies in the mid nineteen eighties. Finally, the two last contradictory findings may reflect true situations in geographical areas less investigated.
Three papers observed a positive association between age and intake of fruit and vegetables. As earlier described these three papers are based on the same Tasmanian data set. Data was collected in the early nineteen eighties, and it is therefore uncertain if these differences would still be observed at present.
In contrast to most of the identified papers analysing the association between SEP and fruit and vegetable consumption among children and adolescents, three papers observed the highest or most frequent intake of fruit and vegetables among low SEP groups. One US study by Melnik et al. (1998) [84
] of 2nd
grade students in the city of New York found that 5th
grade students from low SEP households more frequently consumed fruit and vegetables than students from high/medium SEP households. Categorisation of SEP groups was based on a measure combining information on number of parents working, eligibility for free or reduced price school lunch, and use of federal assistance programs. The contradictory finding is not discussed within the paper. However, the study is characterised by an individual level response rate of 51%, which may have introduced a selection bias by which the low SEP participants are not representative for low SEP households in the city of New York. Another US study by Burdine et al. (1984) [60
] of 7th
grade students from Texas investigated determinants for fruit and vegetable consumption at home and at school respectively. Based on father's occupation, this study found that high SEP students ate fruit more frequently at home, while the most frequent intake of fruit at school was seen among low SEP students. As argued in the paper, these findings suggest that school nutrition subsidies may provide more opportunities for low SEP students to eat healthful foods. The division of total intake measures into measurements of what is consumed at home versus in school may be a new and important aspect to consider for future studies. Using a more general measure for intake that is not calculated by time of the day or different settings may hide important information. Finally, a longitudinal Chinese study by Wang et al. (2002) [20
] found that children whose mothers had higher educational levels were less likely to maintain a high fruit and vegetable diet. Within the paper it is argued that although these mothers were likely to have better access to the media and to health- and nutrition-related knowledge, their behaviour indicates that they are not aware of the health concerns related to higher fat foods. At the same time these families may also have more family resources, by which they can afford more expensive foods, such as meats and cooking oil. The authors also argue, that another possibility is that these women understand the need for energy-dense diets linked with high meat and fat intake and promoted concern for growth and development over the concerns for obesity and diet-related non-communicable disease. Conclusively it is stated that the findings from China suggest that mothers' nutritional knowledge, health consciousness and exposure to the media may influence their children's diet beyond the determining role of family resources and access to foods available to the community in developing countries undergoing a rapid social and economic transition. The present review only identified very few studies conducted in developing countries, and the results from the Chinese study exemplify the importance of initiating future studies involving study populations from countries with varying levels of developmental stages and with different political systems.
For nutritional knowledge, one US study by Lytle et al. (2003) [82
] found that students in the 25th
percentile for the knowledge score reported significantly greater intake of fruit and vegetables compared to students in all other quantiles. As stated in the paper, this finding may be due to a ceiling effect as most students scored very high on the knowledge questions.
In conclusion, several of the few contradictory findings identified within the present review may be attributable to methodological bias, while others may reflect realities.
No previous review comparable to the present review exists. Recently a systematic review was conducted by Blanchette & Brug (2005) [116
]. This review is restricted to the years 1990 to March 2005 and it focuses on determinants of fruit and vegetables consumption among 6–12 year-old children with emphasis on factors easily influenced through intervention. Therefore socio-demographic factors as ethnicity, gender, and SEP are not included. Past reviews have generally been characterised by either having a broader focus in terms of food categories (e.g. general eating behaviours and food patterns) [e.g. [13
]] or by being focused on a specific group of determinants (e.g. family and television watching) [118
]. The majority of the previous reviews did not aim at presenting a total overview of the literature on determinants. Rather, the aim has been to conduct an exploratory and non-exhaustive search to develop, understand and/or present a conceptual framework for understanding adolescent eating patterns and/or to inspire future development of interventions, policies and research [e.g. [13
]]. The aim of the present review was to conduct a comprehensive and exhaustive search from which the evidence of specific potential determinants of a specified food category, namely fruit and vegetables was systematically evaluated based on standardised procedures. Due to these marked differences in purpose and methodology it is therefore difficult to make detailed comparisons between the present review and related previous reviews.
This review has several strengths and limitations. Literature relevant for the present review was identified by searching Medline and PsycINFO. Inclusion of other databases may have led to identification of additional relevant papers that matched the inclusion criteria. For instance, it is rather surprising that no papers on industry promotions (e.g. advertising and marketing of fruit and vegetables or competitive food choices as chips, chocolate bars etc.) as a determinant of fruit and vegetable intake in children and adolescents were identified by the search and selection criteria. However, bibliographies of relevant papers including reviews and methodological papers were searched thoroughly until no more relevant papers emerged. Due to the language criteria, relevant information published in other languages than English may have been missed. The review is strengthened by the systematic approach by which procedures for evaluation and categorisation of all included papers were standardised. Within this review only significant associations are considered. In the field of epidemiology, the criteria for evaluation of estimated associations is a topic of discussion, and one recommendation is not to base evaluations on statistical significance alone [123
]. Preferably, both the significance and the magnitude of association of all included potential determinants should be evaluated. However, this has not been feasible as several papers only report estimates for significant associations or simply levels of significance without estimates of the associations.
Adolescence is a time period of rapid physical maturation and growth combined with a psychological and social development which is often accompanied by changes in social influences. As children move into adolescence family influences often decrease due to competing influences from other social settings [124
]. It could therefore be hypothesised that differences in determinants of fruit and vegetable consumption exist between children and adolescents. The present review includes children and adolescents in the age range of 6 to 18 years. It is therefore rather surprising that for the variables investigated most extensively, no differences in age exist between papers observing significant associations and papers that do not. The rest of the variables (the majority) included in this review are characterised by both being sparsely investigated and by showing few contradictory findings. Differentiating the results for these variables in relation to age is therefore hardly possible.
Methodological and design issues
There has been a clear quality improvement of the literature on determinants of fruit and vegetable consumption among children and adolescents over the past 15 years. In recent years a number of papers with high internal and external validity have been published [e.g. [25
]]. However, our analyses of all papers included in this review revealed some general issues on design and methodology that affect the validity of the generated results as well as the possibilities for comparing results from different papers.
A considerable part of the papers include analyses based on small study samples and many papers include samples that are non-representative or only representative of a restricted geographical area. Often the validity of the applied instruments are only considered very superficially or not mentioned at all. Insufficient confounder control also seems to be a problem in a number of papers. For example, in some papers the analyses are not adjusted for relevant socio-demographic factors such as gender, age and/or SEP. Other bias may be evident in the types of instruments used (recalls versus food frequencies) by age of participant. Parts of the literature on determinants of fruit and vegetable consumption among children and adolescents are therefore characterised by several methodological problems that may affect internal and external validity of the generated results.
Comparisons of results from the different papers may be problematic for several reasons. A large variety of approaches for conceptualising, operationalising, measuring, and coding the outcome variable(s) exists among the identified papers. Some papers consider frequency of intake, while others consider amount of intake, and while these may be inter-correlated and both lead to the desired enhancement of intake, they may have different determinants. For instance, increasing diversity of fruit in the child's home may increase the amount of intake, while it may have no influence on frequency of intake, if measured as daily versus less than daily. In contrast, it may also be hypothesised that an increase in frequency of intake may be accompanied by decreasing portions sizes. In terms of defining the outcome variables, the possibility for comparing results is also compromised by two other aspects: 1) In some papers fruit consumption and vegetable consumption is analysed as separate outcomes, whereas others conduct the analyses on one combined measure. Comparing results from papers applying different analytical approaches may therefore be difficult. Additionally, combined analyses of fruit and vegetable intake may hide the fact that eating fruit and vegetables respectively may be linked to different sets of determinants; 2) Comparing results from different papers may also be problematic due to the fact that some papers include potatoes in the vegetable measurement, while others do not. In almost half of the included papers that analyse vegetable consumption, insufficient information is provided to assess whether potatoes are included in the measurement of vegetable intake or not. It is therefore not possible to evaluate the bias introduced by different definitions of vegetables, when comparing results. Correspondingly, some papers include fruit juice in the fruit category, while others do not. In more than half of the included papers it is not possible to assess, whether fruit juice is included in the measurement of fruit consumption.
Conclusion and recommendations for research
The conceptual framework applied within the Pro Children project is one of the most comprehensive models applied within the research on fruit and vegetable consumption among children and adolescents. The present review shows that several areas in this model can be identified for which research is very limited or lacking:
• Psychosocial behavioural theories have been applied most often. Still, relatively few personal factors have been analysed extensively.
• Family-related factors have been investigated most extensively. However, this part of the literature is characterised by a large number of conceptually different factors that are each investigated only sparsely. Generally, and also in terms of establishing healthy food habits, parents function as important role models for their children. Parents are responsible for making healthy food items available and accessible to the child within the home and for supporting and encouraging the child to make healthy food choices overall. However, the present review shows that only for a very limited number of family-related factors, good evidence exists. To enable health promoters to make evidence based decisions, more studies on the influence of the family setting for influencing fruit and vegetable intake among children and adolescents are therefore needed.
• An overview of the effectiveness of interventions conducted at schools to promote fruit and vegetable intake among school children reveals that multi-component approaches including active provision of fruit and vegetables at lunch are those approaches that have been most successful [15
]. Still, considering the fact that most interventions aiming at promoting fruit and vegetable intake among children and adolescents are conducted at school the number of observational (non-experimental) studies of potential school-related determinants are surprisingly low. Non-experimental studies on the association between school food environments and policies and adolescent eating patterns have been conducted [e.g. [126
]] but studies specifically analysing fruit and vegetable consumption are still lacking.
• Papers on potential influences of national level factors are almost absent. There is an obvious lack of papers looking at international differences in predictors. Future international comparative surveys should enable investigations of national level factors of importance e.g. price levels, policy, guidelines, supply, and exposure to mass media and commercials.
• Likewise, little research has been reported on the potential influence of community or neighborhood level factors. Future research should therefore study the influence of e.g. local access to fruit and vegetables through grocery stores, local food policies, exposure to mass media and commercials, and fruit and vegetable availability in leisure time facilities for children and adolescents, like for instance local sport clubs.
Across the identified areas for which research is lacking, future research would benefit from improvements in design and methodology. Almost every paper identified within this review was based on cross sectional data, and the need for future longitudinal analyses of children and adolescents' fruit and vegetable consumption is evident.
Half of the papers identified by this review are based on US study populations. There is therefore an obvious lack of knowledge concerning predictors of fruit and vegetable consumption among children and adolescents from other parts of the world.
Only few papers apply a theoretical approach. Introducing more theory based research may lead to more systematic research designs that ensure sound analytical models with sufficient confounder control. In addition, the theoretical frameworks (if any) of the papers included in the present review are mostly psychosocial and do not consider more structural environmental influences like nutritional policies and availability of fruit and vegetable in the different settings that children and adolescents take part in.
We strongly recommend that future studies keep a very broad and comprehensive theoretical scope, in order not to exclude important etiological components of importance for child and adolescent fruit and vegetable intake.
Introducing new comprehensive theoretical models should be accompanied by multilevel analytical approaches from which contextual effects can be estimated [127
]. To date, hierarchical models have been applied in a number of papers to adjust for the error introduced by cluster sampling [e.g. [55
]], but so far multilevel modelling of the effect of contextual factors has only been conducted within a few papers [24
Conclusion and recommendations for practice
Despite the lack of consistent evidence for many potential determinants of fruit and vegetable intake in children and adolescents, a few recommendations for practice can be provided based on the present review.
First of all, since fruit and vegetable intake appears to decline with age among children and adolescents, the present review confirms that intervention efforts are indeed needed to promote fruit and vegetable intake across childhood and adolescence. Furthermore, interventions to promote fruit and vegetable intake should especially be aiming at reaching youth from lower SEP groups, and specific efforts should be made to also reach boys. Such interventions should aim at improving preferences for fruits and vegetables, for example by more frequent exposure [128
] to fruits and vegetables by means of taste-testing games or school fruit and vegetable schemes [129
]. Such interventions can also help improve availability of fruit and vegetables in schools [42
Despite presenting solid documentation of the influence of several factors on fruit and vegetable consumption among children and adolescents this review also reveals that a long list of factors have only been sparsely investigated, and for several areas research is totally absent. Although the quality improvement of the research on children and adolescents' fruit and vegetable intake has been pronounced during recent years, a number of methodology problems have been identified. There is a need for further internationally comparative studies. At best, these should be theory-based multi-level studies in which both personal and environmental factors (family, school, local community, and national factors) are considered within a longitudinal design although we do realise that exploring such a broad range of potential determinants comes with measurement problems (e.g. long questionnaires or single item assessment of various constructs) [130
]. Such future research will generate more information on determinants and mediators of fruit and vegetable consumption among children and adolescents on which coming interventions should be tailored.