Background
A sufficient fruits and vegetables (FV) consumption contributes to reducing the risk of non-communicable diseases, including obesity, type 2 diabetes, cardiovascular diseases (CVD) and several cancers [
1‐
7]. Conversely, an insufficient consumption of FV has been estimated to be responsible for 19% of gastrointestinal cancers, 30% of ischemic heart diseases and 11% of strokes worldwide [
3]. It is therefore considered as one of the top 10 factors leading to mortality worldwide, including in developed countries in which inadequate nutrition has become one of the leading mortality risk factors [
3]. The world health organization (WHO) and the food and agriculture organization of the united nations (FAO) therefore recommend the consumption of at least 400 g of FV per day [
8,
9]. In France, this recommendation has been translated as follows: people are encouraged to eat at least 5 servings of FV (i.e. of 80 g each) per day [
10]. Low consumers of FV are defined as such when consuming less than 3.5 servings of FV per day [
10].
Studies over the past two decades have provided important information on the association between socioeconomic status and dietary habits [
11‐
18]. Overall, lower socioeconomic populations use to have a less healthy diet, and have higher proportions of small FV consumers in Northern Europe [
15,
19,
20], United States of America (USA) [
21], United Kingdom [
22] and Australia [
23]. In France, 35% of adults from the general population eat less than 3.5 servings of FV per day, this rate rises to 82.4% among underprivileged people using food aid [
24‐
26]. This trend is also reflected in children, as those from lower socioeconomic groups consume an average of 2.6 to 3.0 servings of FV per day, whereas in general population, FV consumption is ranging from 3.4 to 3.6 servings per day in children [
27]. Promoting healthy dietary habits among the youngest populations is an important issue since concrete food choices and dietary behaviors seems to originate in childhood and adolescence [
28‐
31].
A study performed in the USA among subjects from the Supplemental Nutrition Assistance Program (SNAP) (
N = 10,000), showed that a 30% increase in the FV budget decreased the cost of chronic diseases such as type 2 diabetes, CVD, strokes and obesity [
32]. The financial cost of such assistance at the national level has therefore been estimated as a major public health benefit [
32]. Several studies have documented the efficacy of FV vouchers on daily consumption among pregnant women and young children up to 5 from “nutritional risk” populations, like in the USA [
33,
34] or in Great-Britain [
35]. The findings showed that vouchers were associated with an increased consumption of FV, with a higher consumption when combining nutritional education [
34,
36‐
40].
But so far, no such intervention was specifically conducted among children, most of studies being performed at school. [
41,
42]. Yet, there is a real interest in promoting FV consumption simultaneously at school and at home, since it has been shown that dietary behaviours in these two environments are strongly associated [
34,
43,
44]. Additionally, results from such studies performed among disadvantaged populations are often limited due to recruitment and follow-up difficulties [
41]. The French National Nutrition and Health Program (
Programme National Nutrition Santé) (PNNS) is a national public health program aiming to improve the health of the general population through nutrition. Two major aims have been included in this program: 1) decreasing the number of individuals considered as low consumers of FV (less than 3.5 servings per day) [
45] by at least 25%, and 2) improving the nutritional status of disadvantaged population [
2].
In this overall context, the “Fruits and vegetables at home” study (Fruits et légumes à la maison) (FLAM study) aimed at assessing the effect of FV vouchers on the daily consumption of FV in children from low-income families. The present study relied on a mixed research method aiming to 1) determine whether children from disadvantaged households receiving FV vouchers during one year modified their FV consumption (quantitative method), and 2) understand how the intervention impacted dietary practices and identify barriers and levers for participation through interviews of families (qualitative method).
Discussion
This study showed a significant increase in FV consumption in children from disadvantaged families with a one-year allowance of FV vouchers compared to a control situation. No significant difference was observed in adults. Consumption of other food groups was not modified according to the study group.
Our results though being modest, remain encouraging and are in line with previous studies performed using similar interventions. An English study compared three adult groups, a “control” group (
N = 64), a “nutritional advice on fruit and fruit juice” group (
N = 63) and a “coupons for fruit or fruit juice” group (
N = 63). Results showed that only the group receiving vouchers significantly increased fruit and fruit juice consumption compared to other groups [
35]. In Dunedin, New Zealand, a randomized study of 151 volunteers, (
N = 81 in the intervention group), showed a significant increase in overall food expenditure when coupons were distributed. However, the vouchers were not specifically targeting FV, but healthy foods items [
69]. In the USA, studies in the national WIC program showed significant positive impacts of vouchers on FV consumption, especially when strengthened by nutritional education [
34,
36,
38,
39]. In France, a randomized study on the effect of FV vouchers on consumption among adults was performed in the same area as FLAM study in 2008 [
56]. Vouchers did not show significant positive effects on the mean consumption of FV after 6 months; but they significantly decreased the proportion of very small FV consumers (i.e. < 1 servings a day) [
56]. Due to important attrition rates, results at 9 and 12 months attrition rates (respectively at 55.3 and 84.8%) were not available.
Non-randomized intervention studies were conducted in England, in the context of the Healthy Start program, in which vouchers were distributed to pregnant women and up to 4 years of age of the child. A total of 113 volunteers reported during the focus group that they increased quantitatively and qualitatively their FV consumption through coupons [
70]. A second study, conducted over a 5-month period (
N = 621), and using FV prescription by medical professionals did not show any significant effect on consumption, though the knowledge of the slogan “Five-a-day” was increased in the population [
71]. More generally, a review performed in 2012 on the effectiveness of subsidies in promoting healthy food purchases and consumption concluded that these types of interventions tended to be effective in modifying dietary behavior [
41].
Overall, these results tend to support the fact that vouchers alone may not be sufficient to increase consumption, and that the addition of nutrition education is suitable [
72]. On the other hand, Darmon and colleagues have shown that a nutrition education appears inefficient, if not supported by a financially affordable supply of healthy food [
73]. These findings are in line with our qualitative survey, showing that women with a baseline poor knowledge about FV who participated in the workshops durably modified their dietary habits. Furthermore, a financial support for the entire household is relevant (instead of children only), since parents are primary responsible for the food choices of the entire household, they also act as a model for their children [
71]. Our results were consistent with this assumption since the assessment of the adherence to the PNNS dietary guidelines showed that children’s dietary profiles were overall similar to their parents’ (Table
2) [
71,
74]. Furthermore, several works have highlighted the close interrelationship between FV consumption in parents and children, in both directions [
71]. An evaluation study of the “5 a day” program in Los Angeles showed that when a mother increased her daily FV consumption, it had a positive impact on the consumption of entire household [
37].
Surprisingly, no significant effect of the intervention on FV consumption has been shown in parents. However, the considerable use of vouchers suggested that the FV affordability in the households was increased. We assume that parents prioritized their children consumption, giving them the most of purchased FV, before raising their own consumption. A possible explanation for this might be that the number of vouchers sent did not always match with the size of the household (maximum 8 vouchers per month for households with 4 people or more).
A large majority of children (88%) in our study population had lunch at school canteens, in which the nutritional quality of foods is regulated in France [
75‐
77]. The school catering of Saint-Denis showed a good compliance to these recommendations [
78]. This may partly explain the significantly higher FV consumption (Exact Fischer test
p = 0.001, data not shown) in children compared to adults at baseline. However, though school canteens appeared to partly increase children’s FV consumption, the amount provided did not appear to compensate for an overall low consumption (compared to the general population).
Given the results on our primary outcome after 1 year follow-up (29.4 and 66.7% of low FV consumers in the intervention and control group respectively), and the number of children within each group (
n = 30 in the intervention group and
n = 34 in the control group), the effect size was therefore estimated at 37.3% and the power of the study at 31.9% [
79]. According to Cohens’ thresholds regarding effect sizes (0.20 is small and 0.50 is medium), the effect sizes of the intervention can be considered moderate in children (0.37), and small (0.15) in adults [
79].This could partly explain why no significant result was seen in adults, despite a lower proportion of small FV consumers in the intervention group (61.8%) compared to the control group (76.7%) after one year follow-up (Table
3). On the other hand, the results observed in children are quite encouraging for this kind of intervention. Compared to the study of Bihan and colleagues, our effect size in adults was lower (respectively 20.3% vs 14.9%). Regarding the daily consumption of FV, Herman and colleagues found an effect size of + 3.0 servings per day between households from voucher and control groups, while this difference was at estimated + 1.8 servings in children and + 1.1 servings in adults in our study. Burr et al. showed that the purchase of healthier food increased by 6% during the intervention.
The attrition rate of 30.4% is much lower than those found in a similar study previously performed by Bihan and colleagues in the same area [
56]. Moreover, it is in line with previous interventional researches performed among similar populations. Katz and colleagues performed in2001 an interventional study among low-income mothers with an attrition rate of 41% after one year duration [
80]. Nicholson and colleagues described in 2011 several retention strategies they implemented to improve retention rates in their interventional research in a low-income urban population. Attrition rates were 25% et 36% at 6 and 12 months respectively [
81]. Recruitment barriers have been explored through a qualitative analysis [
82]. Briefly, the main reasons reported by the individuals who refused to participate in the study were a lack of time, mistrust towards researchers and people coming from outside the neighborhood, and trouble with communicating with the interviewers.
Our study had some limitations. First, we used a food frequency questionnaire to assess food consumptions instead of a 24-h dietary recall which is usually the gold-standard for dietary assessment. This could have led respondents to misestimating some food consumptions (in particular FV), mostly due to a memorization bias. Plus, this type of assessment does not allow assessing the portion sizes [
83]. Moreover, due the young age of their child or its inability to respond to the questionnaire directly, some parents had to answer instead of their child (
N = 14) which may have led to less reliable information. Besides, the frequency dietary questionnaire we relied on has been specifically designed to be administered to disadvantaged groups, and was previously used in the French ABENA study performed among food aid users [
84,
85].
Another limitation pertains to the limited number of vouchers per household. Indeed, 34.8% of households included had at least three children. Thus, vouchers were not fulfilling the equivalent of one portion of FV a day for each person in these households. Sensitivity analyses in this group (
N = 14) showed that there was no significant increase in FV consumption in these families. However, caution is needed when interpreting this result given the very low number of participants in this sub-group. We could not totally exclude a contamination, i.e. a transfer of a part of FV vouchers towards the families of the control group. Nevertheless, families were included and followed separately and few have indicated knowing each other. A contamination seems therefore unlikely. Finally, results showed that families lost to follow-up were more precarious than the others (Table
4). This finding is in line with the literature showing that daily life difficulties in such populations make them harder to reach and follow in trials [
86]. We therefore explored the intervention effect by controlling the precariousness level of the families. In children from high level precarious families, difference in proportion of low FV consumers between the intervention and control group was lower than in other families, and was no longer significant (
p = 0.16, data not tabulated). This should be kept in mind before extending the measure, in order not to worsen social inequalities, trying to reduce them.
The major strengths of this study were the randomized design of the intervention and the use of face to face interviews to collect the data. Moreover, the percentage of participants lost to follow - up at one year was less (30.4%) than expected, given the targeted population [
56]. This could partly be due to regular solicitations of participants through follow-up questionnaire and workshops. The one-year duration of the FLAM study was another strength, a short duration being an usual limitation of such studies. It overcame the novelty effect of the vouchers and allowed them to be incorporated in usual purchase habits. Finally, the intention-to-treat analysis showed similar results, supporting the hypothesis of an efficiency of the intervention.
Finally, given the results of this interventional research, it is likely to consider its replication, or even its implementation at the national level (in the line with previous programs that were developed in the US or in the United Kingdom). However, maintaining this program in its current form in the long-term state seems difficult. Indeed, it would require significant human resources to handle nutritional education on such a large scale. Though, an adaptation of the financial incentives through a FV allowance for instance would be interesting to explore.
Acknowledgements
The authors thank Claudia Chahine and Cynthia Perlin dietitians for their important help on this work and for conducting the workshops throughout the study. We thank all FLAM investigators: Laëtitia Defoi, Anita Houeto, Anouchka Kponou, Lysa Tagherset and Marion Genest. We thank Paul Flanzy, Younès Essedik (computer scientists), Julien Allègre and Frédéric Coffinières (data-managers).
The authors also thank the Maison de la Santé of Saint-Denis and its former director Marjorie Painsecq, the municipal health centres of the city of Saint-Denis (CMS du Cygne, CMS Henri Barbusse, CMS des Moulins) for their hospitality and help, Saint-Denis neighborhood houses (Pierre Sémard, Romain Rolland, Floréal), and all the associations that were involved in the study: APIJ association, La maison des parents, the ludothèque Allende and the ASAFI association. Finally, the authors thank the members of the steering committee: Dr. Michel Chauliac (Direction Générale de la Santé), Samira Guedichi-Beaudouin (Mairie de Saint-Denis), Benjamin Cavalli (Programme MALIN).