Relationship between Prefecture-Level Yield of Not-for-Sale Fruits and Vegetables and Individual-Level Fruit and Vegetable Intake in Japan: A Cross-Sectional Study

This study investigated the relationship between prefecture-level yield of not-for-sale fruits and vegetables and individual-level fruit and vegetable intake in Japan. Data were drawn from the Japanese National Health and Nutrition Survey and National Crop Survey of 2016. Random intercept models were used for the analyses. Individual-level fruit and vegetable intake was used for the dependent variable, and prefecture-level yield of not-for-sale fruits and vegetables was used for the independent variable as a fixed effect. In addition, participants’ characteristics and health-related factors at the individual level were also put into independent variables as fixed effects. The prefectures were used as random intercepts. It was found that prefecture-level yield of not-for-sale fruits and vegetables was significantly related to individual-level fruit and vegetable intake (vegetable: B = 0.390, p < 0.001; fruit: B = 0.268, p = 0.003; fruits and vegetables: B = 0.357, p < 0.001). These relationships were also significant in the gender-specific analysis. Thus, the yield of not-for-sale fruits and vegetables might contribute to the intake of fruits and vegetables in Japan.


Introduction
The World Health Organization (WHO) and the UN Food and Agriculture Organization (FAO) promote increased consumption of fruits and vegetables because it can reduce the risk of certain noncommunicable diseases and the mortality these diseases entail [1]. Several meta-analyses have shown inverse associations between fruit and vegetable intake and the risk of developing cardiovascular disease or cancer and total mortality [2][3][4]. Additionally, it has been suggested that there is a positive relationship between fruit and vegetable intake and mental health [5,6]. Therefore, strategies to increase fruit and vegetable intake should be developed to promote public health.
Certain environmental factors have been known to influence fruit and vegetable intake [7]. A systematic review of environmental factors associated with fruits and vegetables intake among adults revealed an association with high household income, marriage, and good local availability of fruits and vegetables [7]. Local availability of fruits and vegetables, having a vegetable garden, and homegrown produce were positively associated with fruit and vegetable intake [7].
In Japan, it has been found that obtaining nonmarket food both from within and outside of the household could affect dietary intake [17][18][19][20][21][22][23][24]. Studies conducted in various locations throughout the country and in Satoyama (a traditional Japanese rural setting) showed that more nonmarket food is available in Satoyama than in other parts of Japan, and if people have more food-sharing partners, a greater diversity and quantity of food could be obtained from nonmarket food sources [17][18][19][20]. In addition, a significant relationship was found between types of land use and the ability to obtain greater amounts and varieties of nonmarket food [20,21]. Several studies have focused on vegetables in particular. One study found a positive relationship between the intake of vegetables received from neighbors and total vegetable intake [22]. In addition, it has been confirmed that vegetable cultivation and vegetable reception are related to vegetable intake [23,24]. Furthermore, in areas with flourishing fruit and vegetable cultivation, it is easy to obtain fruits and vegetables through social networks, even for individuals who do not cultivate fruits and vegetables themselves [25].
Thus, it was hypothesized that residents in areas with high yields of not-for-sale fruits and vegetables would have high intakes of fruits and vegetables. The mechanism would run as follows: the harvested not-for-sale fruits and vegetables are first consumed in the grower's home. Then, surplus fruits and vegetables are provided to their neighbors. In addition, some neighbors who have received more than they can use may provide their excess to their other neighbors. In an ecosystem of this type, residents in areas with high yields of not-for-sale fruits and vegetables are likely to enjoy high intakes of fruits and vegetables.
This study investigated the relationship between prefecture-level yield of not-for-sale fruits and vegetables and individual-level fruit and vegetable intake in Japan. The yields of the fruits and vegetables used in this study were taken to be the amount recorded in the Japanese national statistics. This figure does not include fruits and vegetables produced in a home garden or a community garden. However, previous studies in Japan have found that 96% of farmers grow vegetables for their own consumption and 84% give away the vegetables they grow to their neighbors [24]. In addition, prefectural-level ecological studies have already confirmed a positive correlation between farmers' yield of not-for-sale vegetables and vegetable intake [26]; however, ecological study has not clarified whether prefecture-level yield is associated with individual-level intake or not.

Study Design and Survey Outline
This was a cross-sectional study that used data from two Japanese national surveys. One of these was the National Health and Nutrition Survey (NHNS) 2016 [27]. That survey investigated Japanese citizens' physical condition, nutrient intake, and lifestyle and obtained basic data to comprehensively promote people's health, based on the health promotion law [27]. For NHNS 2016, the subjects of the survey lived in 475 census districts, including 10 census districts per prefecture that were randomly selected (15 districts in Tokyo, due to its large population) [27]. From among these districts, Kumamoto and a part of Tottori (in total 13 districts) could not be investigated due to the effects of natural disasters and were excluded [27]. Those who lived in the subject areas and were over 1 year old were surveyed from October to November 2016 [27]. There was a total of 24,187 subject households, of which 10,745 (26,354 people) could be surveyed [27]. The NHNS 2016 was the most recent large-scale survey that examined differences in prefectures. Relative to the standard survey year, the number of subjects in that study was large, and it was considered appropriate to use its data for the examination by prefecture conducted in this study.
The other major source of data was the National Crop Survey (NCS) 2016 [28,29]. This survey clarified the actual conditions of crop production and shipment and prepared materials to formulate production targets and promote various measures in the food, agriculture, and rural basic plans [28,29]. For this study, the fruit and vegetable data were extracted. NCS surveyed the crop acreage, yield, and shipment amount of all agricultural product shipping organizations in the target prefectures [28,29]. In addition, additional surveys were conducted on the entities with crops that were not shipped through the above organizations [28,29]. The NCS was conducted every 3 years for vegetables and every 6 years for fruits in all prefectures [28,29]. In other years, the top prefectures, which accounted for approximately 80% of the cultivated area, were targeted for each surveyed item [28,29]. In 2016, the vegetable survey was conducted for all prefectures [28], but data were not collected for fruits in 2016 for all prefectures [29], so for the prefectures where the fruit survey was not conducted, data from the latest survey of all prefectures (2014) were used.

Fruit and Vegetable Intake
This study used individual-level data on fruit and vegetable intake of NHNS as the dependent variables. Fruit and vegetable intake was provided and used for analyses. In addition, the total fruit and vegetable intake was calculated and used for the analyses. The unit used for all analyses was daily intakes (g) per participant.
The survey method for food and nutrition intake in NHNS was as follows: food and nutrition intake was determined by the dietary record, using a weighted method [27]. In addition, the standard amount was used for items that were used in small amounts and are difficult to weigh [27]. The survey day was set to a day of the week other than Sundays and public holidays, avoiding days when there was a special change in food intake [27]. The investigators of NHNS, who were registered dietitians, visited all of the surveyed households directly, checked the contents of the surveys, and corrected any deficiencies [27].

Yield of Not-for-Sale Fruits and Vegetables
This study used prefecture-level yield of not-for-sale fruits and vegetables of the NCS as independent variables. The yield of not-for-sale fruits and vegetables for each prefecture was calculated by subtracting the shipment amount from the yield using data for the yield and shipment amount for each prefecture from the NCS. The shipment and yield of the NCS were aggregated by annual tonnage for each prefecture [28,29]. Therefore, in this study, dividing by the participant's number for each prefecture [30] and the number of days per year (365) and multiplying by 10 6 , it was converted to the grams per person per day for the analysis.
NCS 2016 (vegetables and fruits) provides statistics on 55 major items (vegetables, 41; fruits, 14) produced in Japan [28,29]. However, the classification of vegetables and fruits in NCS is different from the classification used in NHNS. Therefore, in this study, the NCS data were rearranged according to the NHNS classification, following the procedure below. First, three items (potato, taro, and yam) that are not counted as fruits or vegetables in the NHNS were excluded. Then, the three items (strawberry, melon, and watermelon) counted among the vegetables in the NCS were converted to fruits, following the NHNS classification. In addition, Asian plum was classified as fruit in the NCS; however, it was counted as both of a fruit and a vegetable in NHNS. When eaten as a traditional pickle, it was classified as pickle, and it was a fruit otherwise in the NHNS. In this study, Asian plum was not included in either vegetable or fruit yield. However, the total vegetable and fruit yield included Asian plum. Finally, the total yields of 35 vegetables (cabbage, cucumber, Japanese radish, tomato, eggplant, carrot, leek, Chinese cabbage, green pepper, lettuce, onion, spinach, asparagus, green soybeans, turnip, pumpkin, cauliflower, burdock, komatsuna, sayaingen, sayaendou, crown daisy, ginger, sweet corn, celery, broad beans, bok choy, nira, garlic, Japanese butterbur, broccoli, mizuna, mitsuba, lotus root, and green peas), 16 fruits (strawberry, melon, watermelon, mandarin, apple, Japanese pear, pear, persimmon, loquat, peach, plum, yellow peach, grape, chestnut, pineapple, and kiwifruit), and 52 fruits and vegetables (all fruits and vegetables listed above and Asian plum) were used for the analyses.

Other Items
Prefecture, gender, age, lifestyle, drinking habits, smoking habits, body mass index (BMI), and energy intake were used for the analyses. These data were individual-level data provided by NHNS [27]. Gender, age, drinking habits, and smoking habits were determined with a self-administered questionnaire [27]. Prefecture and living style were ascertained from the national census before the survey [27]. BMI was calculated from measured height and weight data. For this, the body weight (kg) was divided by the square of the height (m) to obtain the BMI (kg/m 2 ). Energy intake was calculated from a survey of food and nutrition intake already written [27]. These variables were used as categories in the analyses.

Analyses
From the data on 26,354 individuals obtained from the National Health and Nutrition Survey, excluding those from participants aged younger than 20 and more than 79, those pregnant and lactating, and those with missing data in the items used for analysis, data from 15,046 participants aged 20-79 years (men, 6800; women, 8246) were used in this analyses. Table 1 shows the total number of participants and the number of participants by gender for each of the 46 prefectures (there are 47 prefectures in Japan; however, one was excluded because the survey was not conducted there due to a natural disaster [27]). First, in the preliminary analyses, the yields of not-for-sale fruits and vegetables and the intake of fruits and vegetables were calculated for each prefecture, and Spearman's correlation coefficient between these variables was calculated.
Then, random intercept models were used for the main analyses [31]. The prefectures were used as random intercepts. Fruit intake, vegetable intake, and fruit and vegetable intake were the dependent variables. These were used as interval scales (g). The yield of not-for-sale fruits, vegetables, and fruits and vegetables were used for independent variables as fixed effects. These were used as both of interval scale (g) and category (quartile categories), respectively. The trends using categorical order were also tested. Individuallevel gender, age, living style, drinking habits, smoking habits, BMI, and energy intake were adjusted as a fixed effect for the analyses. The number of participants by category for each variable used in the analyses was shown in Table 2. For the model construction, first, null models wherein nothing was input for the independent variable was created, and it was confirmed whether the variance of prefectures was significant. Then, individual-level variables were placed into the models. Finally, the prefecture-level yield of not-for-sale fruits and vegetables were input into the models. The goodness of fit for each model was determined from Akaike's information criterion (AIC) of each model [32]. For AIC, smaller values indicate better models [32]. All analyses were conducted using IBM SPSS Statistics for Windows, version 23.0; IBM Japan, Ltd.: Tokyo, Japan. Statistical significance was set at p < 0.05.

Ethical Approval
The NHNS data were obtained with permission from the Ministry of Health, Labor and Welfare. The NCS data were obtained from the Japanese official statistics portal site e-stat. The data did not contain any personally identifiable information. This study was approved by Takasaki University of Health and Welfare Research Ethics Committee (No.: 1907; approval date: 31 May 2019).

Preliminary Analyses
In Table 3, yields of not-for-sale fruits and vegetables and fruit and vegetable intake are shown according to prefecture. For the yield of not-for-sale amounts, vegetables ranged from 1.4 to 117.0 g, fruit ranged from 0.1 to 97.2 g, and fruits and vegetables ranged from 1.6 to 183.2 g. For mean intake by prefecture, vegetables ranged from 230.2 to 345.5 g, fruit ranged from 77.9 to 143.6 g, and fruits and vegetables ranged from 309.8 to 458.3 g. The Spearman's correlation coefficients (p-Value) between intake and yield among prefectures were 0.365 (0.010), 0.442 (0.002), and 0.401 (0.006) for vegetables, fruits, and fruits and vegetables, respectively.

Main Analyses
In Table 4, the relationships between prefecture-level yield of not-for-sale fruits and vegetables and individual-level fruit and vegetable intake, tested using the random intercept models, are shown. In addition, the AIC of each model and the p values of null models are shown in Table A1.  Random intercept model (random intercept: prefecture). Dependent variables: vegetable intake, fruit intake, and fruit and vegetable intake. Independent variables: prefecture-level yield of not-for-sale vegetables, prefecture-level yield of not-for-sale fruits, prefecture-level yield of not-for-sale fruits and vegetables. B: unstandardized regression coefficient, CI: confidence interval. Adjusted for gender (only all participants), age, family structure, body mass index, energy intake, smoking status, and drinking status as fixed effect. N = 15,046 (men: n = 6800, women: n = 8246). The p-value for trend of categories was less than 0.001. Similar significant associations were found when analyzed by gender. The null model's intercepts for all participants (p < 0.001), men (p = 0.010), and women (p = 0.001) were significant. For all participants, men, and women, the AICs were smaller when the individual-level variables were input from null models. The AICs were even smaller in the models that input the yield at the prefecture level.  17.468). The p-value for trend of categories was 0.013. Similar significant associations were found for analysis by gender instead of quartile categories for women. The null model's intercepts were significant in all participants (p = 0.001), men (p = 0.042), and women (p = 0.007). The AICs were smaller when the individual-level variables were input than null models for all participants, men, and women. For the models that input the yield at the prefecture level, the AICs were even smaller compared with the other models.
For fruits and vegetables, B was 0.357 (95% CI: 0.167 to 0.548) when the yield of not-for-sale fruits and vegetables was used as an interval scale. Similarly, there was a significant association in the quartile categories (Q1: Ref; Q2: B = 8.932, 95% CI = −12.960 to 30.824; Q3: B = 21.077, 95% CI = −1.517 to 43.671; Q4: B = 32.973, 95% CI = 11.870 to 54.076). The p-value for the trend of categories was 0.001. Similar significant associations were found when analyzed by gender. The intercepts of null models of all participants (p < 0.001), men (p = 0.005), and women (p = 0.001) were significant. For all participants, men, and women, the AICs decreased in the order of null models, individual-level models, and prefecture-level models.

Discussion
This study investigated the relationship between prefecture-level yield of not-forsale fruits and vegetables and individual-level fruit and vegetable intake in Japan. As a result, the positive relationship between prefecture-level yield of not-for-sale fruits and vegetables and individual-level fruit and vegetable intake was found. To date, no studies have demonstrated such an association. This is the first study to identify the yield of not-for-sale fruits and vegetables as an environmental factor related to the intake of fruits and vegetables, and it may contribute to the development of a health-promoting food environment.
In preliminary analyses, it was found that there is a prefectural-level positive correlation between the yield of not-for-sale fruits and vegetables and fruit and vegetable intake in Japan. A previous study has reported a correlation of this type for vegetables, but it did not examine fruits or the total amount of fruits and vegetables [26]. This study found a prefectural-level positive correlation between the not-for-sale yield and intake for fruit and fruit and vegetables.
We found a positive relationship between prefecture-level yield of not-for-sale fruits and vegetables and individual-level fruit and vegetable intake. Many positive relationships between fruit and vegetable cultivation/receiving and fruit and vegetable intake have been reported [8][9][10][11][12][13][14][15][16][22][23][24]. In areas where there are many not-for-sale crops, much home consumption can be presumed, as well as obtaining fruits and vegetables from one's neighbors, and, as a result, the intake of fruits and vegetables is high. According to the results of this study, fruit and vegetable intake is about 33 g/day higher in prefectures with the highest yield (Q4) than in prefectures with the lowest yield (Q1) of not-for-sale fruits and vegetables. WHO and the FAO recommend a fruit and vegetable intake of 400 g/day or higher [1]. In addition, recent meta-analyses have reported that increasing fruit and vegetable intake to 800 g/day may reduce the risk of certain diseases and mortality [2]. In Japan, nutritionists recommend 350 g/day for vegetable intake and 100 g/day for fruit intake [33]. The contribution here of not-for-sale fruit and vegetable yields may not be significant. However, even if it is a small contribution, it may raise the base of fruit and vegetable intake.
Food access is positively associated with the consumption of fresh foods, such as vegetables and fruits [7]. However, in rural areas of Japan, food access has deteriorated, due to the disappearance of local grocery stores, which are thought to have a negative impact on the intake of fresh foods, such as vegetables and fruits [34]. Even under these circumstances, it is possible that the negative impacts were mitigated by cultivating not-forsale crops and distributing them within the region. In fact, although these data regard one island in Japan, nonmarket foods account for approximately 25% of the production price basis and around 17% of the caloric basis of food consumption as a whole [19]. In addition, approximately half of all food consumed is from nonmarket sources between the spring and autumn seasons [20]. In other words, the distribution of nonmarket foods may contribute to the healthy eating habits of Japanese people living in certain areas, where maintaining such an ecosystem may have important implications for the promotion of the health of residents. This necessitates maintaining agricultural cultivation in such areas. However, the number of farmers and cultivated land areas are declining in Japan [35,36]. The policies to revive agriculture may sustain the distribution of nonmarket foods. Furthermore, the higher the social cohesion, the higher the frequency of receiving fruits and vegetables from relatives or neighbors [25]. Therefore, programs to deepen ties among local residents may support the health of local residents through not-for-sale fruits and vegetables.

Limitation
This study had several limitations. First, since it was conducted in Japan, its results may only pertain to that country and not to other countries or regions. In Japan, it is common to share one's produce surplus to neighbors, so the results of this study were obtained. Therefore, results of similar studies will be different in countries and regions where there is no accepted culture of passing on one's surplus to others. In this study, only the yield of not-for-sale fruits and vegetables produced by farmers were used as independent variables. It should be noted that the crops produced by nonfarmers in home gardens and community gardens were not included. This study also examined data at the prefecture level. However, the actual transfer of surplus often appears within smaller regional units. Finally, this was a cross-sectional study. Therefore, the causal relationship of the association confirmed in this study cannot be confirmed.

Conclusions
This study revealed the positive relationship between the prefectural-level yield of not-for-sale fruits and vegetables and individual-level fruit and vegetable intake. An environment where not-for-sale fruits and vegetables could be harvested abundantly might be contributing to the resident's intake of abundant fruits and vegetables in Japan.

Informed Consent Statement: Not applicable.
Data Availability Statement: The raw data are not publicly available due to restrictions ethical.

Conflicts of Interest:
The author declares no conflict of interest.