The Korean Elderly Environmental Panel (KEEP) study is a community-based, repeated-measures survey that was established to study the effects of diverse environmental exposures on the health of elderly Koreans [19]. The study methods and details are described in the literature [19]. Briefly, the KEEP study is composed of the KEEP-I and the KEEP-II. In KEEP-I, 560 elderly subjects were recruited in Seoul and followed up five times (six surveys) between 2008 and 2010, and in KEEP-II, 800 elderly subjects were recruited (400 from Seoul and 400 from Asan) and followed up twice (three surveys). Our study used data from the KEEP-II, which was conducted annually from 2012 to 2014. Most of the questionnaires, including weekly fruit/vegetable intake and the Korean version of the Short form Geriatric Depression Scale (SGDS-K), were investigated every year. The study participants were elderly adults aged ≥ 60 years with an appropriate cognitive level to understand the instructions of the survey, who regularly visited one of the two community welfare centers in the Seongbuk-gu area of Seoul and Asan, Korea. Of the 800 KEEP-II participants recruited in 2012, 416 and 327 participants were followed up in 2013 and 2014, respectively. New enrollment included 384 and 71 participants in 2013 and 2014, respectively. More details regarding follow-up status in Seoul and Asan are presented in Supplementary Material 1. From the baseline data of 1,255 individuals who participated in the study between 2012 and 2014, 5 participants whose response to the query regarding weekly fruit or vegetable consumption was “do not know,” 1 participant without SGDS-K information, and 23 participants who were diagnosed with depression and undergoing treatment at the baseline were excluded. However, subjects with depression who were not receiving therapy were included. Among the study subjects, 305 individuals who participated in the KEEP-II for three consecutive years were selected for the repeated-measures analysis (Figure 1).

The current study used the SGDS-K as a screening tool to assess depression. The SGDS-K comprises 15 questions that evaluate the participant’s mood in the preceding week. The participants choose the response “yes” or “no,” with “yes” counted as 1 and “no” counted as 0 to yield a range of total score from 0 to 15. As the SGDS-K score increases, people experience more depressive symptoms. A score of 8 was used as the cut-off point for the diagnosis of depression. Previous studies have established that the SGDS-K has high validity for the diagnosis of depression [20].

The dietary assessment was performed using a self-reporting method of responses to simple questions regarding fruit and vegetable consumption: “How often do you usually eat fruit? Fruits include fresh, dried, canned, and frozen fruit.” and “How often do you usually eat vegetables? Vegetables include fresh, dried, canned, and frozen vegetables.” The following choices were provided: (a) never, (b) less than 1 time/wk, (c) 1-3 times/wk, (d) 4-6 times/wk, (e) daily, and (f) do not know. The data signified the frequency, not the quantity, of consumption.

More detailed dietary information including the intake of each food item in grams, total energy intake, and daily nutrient intake were investigated for the year 2012 by using a semiquantitative food frequency questionnaire (FFQ) to assess the frequency of consumption (9 categories from rarely eaten to eaten more than 3 times/d) and portion size (categorized as small, average, or large) of 118 food items [21,22]. The amount of each food item was transformed into grams using the Computer Aided Nutritional Analysis Program for Professionals (CAN-pro 4.0, Korean Society of Nutrition, Seoul, Korea). However, because only 385 participants participated in the FFQ and the FFQ was conducted only in 2012, we decided to use the simple questions mentioned above.

Age was categorized into three groups: 60-69 years, 70-79 years, and ≥ 80 years. Education level was categorized into three levels (detailed information provided in Supplementary Material 2). Allowance was categorized as < 30.00, 30.00-69.99, and ≥ 70.00 units of 10,000 Korean won (KRW) per month (approximately US$10). Only current behavior was considered to assess smoking status and alcohol consumption. Household type was categorized as living alone or living with family, which refers to living with spouse, sons or daughters, or others. The presence or absence of underlying diseases was assessed. Disease status refers to a history of admission/surgery/emergency room visit within 1 year or a visit to a doctor without admission within 2 weeks. We used the short form of the International Physical Activity Questionnaire to assess physical activity levels. Physical activity was coded as < 600 or ≥ 600 metabolic equivalents of task (MET) min/wk, with MET measured as energy expenditure in calories per kilogram of body weight for 1 minute (kcal/min/kg) (Supplementary Material 2). Body mass index (BMI) was calculated as weight divided by height squared and categorized as <23.0 kg/m², 23.0-24.9 kg/m², and ≥25.0 kg/m², with ≥ 23.0 kg/m² classified as overweight according to the Korean Society for the Study of Obesity [23]. According to the Korean Society for the Study of Obesity, ≥ 25.0 kg/m² is classified as obese, although this value is classified as overweight by the World Health Organization [23,24].

The t-test was used to compare continuous variables such as age and total energy intake in participants with depression (SGDS-K ≥ 8) and without depression (SGDS-K< 8), whereas chi-square tests were used to compare categorical variables such as sex, age category, education, allowance, smoking status, alcohol consumption status, household type, disease status, physical activity, BMI category, weekly fruit intake, and weekly vegetable intake.

The SGDS-K was evaluated as both a continuous and categorical variable. The SGDS-K score as a continuous variable refers to the level of depression, and as a categorical variable, it refers to depression status (SGDS-K ≥ 8 or < 8). The association between the level of depression and the frequency of fruit or vegetable consumption was also estimated. Multivariate linear regression was used to estimate β, 95% confidence interval (CI), and p-value for SGDS-K scores (continuous variable). Multivariate logistic regression was used to estimate odds ratios (ORs), 95% CIs, and p-values for depression status (SGDS≥ 8 or < 8). The analyses were performed using three models: a crude model, an age- and sex-adjusted model, and a multiple covariate-adjusted model. The confounding factors used in the multiple covariate-adjusted models were sex, age category (the age- and sex-adjusted model used age as a continuous variable), education, allowance, smoking status, alcohol consumption status, household type, disease status, physical activity, and BMI.

For sensitivity analysis, reciprocal adjustment for weekly vegetable intake and weekly fruit intake was performed in multiple covariate-adjusted models. Also, we performed sensitivity analysis by combining “never” and “less than 1 time” together as the sample sizes of these categories were small. Furthermore, generalized linear mixed-effects models were used for repeated measures in the subset of subjects who repeatedly participated in the KEEP survey 3 times. Linear mixed-effects models were also constructed as crude, age- and sex-adjusted, and multiple covariate-adjusted models.

To assess dose-response relationships, we regarded the frequency of weekly fruit and vegetable intake as a continuous variable; “never,” “less than 1 time/wk,” “1-3 times/wk,” “4-6 times/wk,” and “daily” were converted into 0 times/wk, 0.5 times/wk, 2 times/wk, 5 times/wk, and 7 times/wk, respectively, as these values are the average value of each category. We then assessed the dose-dependent relationship of the level of depression, that is, SGDS-K score (continuous variable), and depression status (SGDS-K < 8 or ≥ 8) with fruit and vegetable consumption using linear regression and logistic regression analyses, respectively. R version 4.0.3. (https://cran.r-project.org/) and SPSS version 25.0 (IBM Corp., Armonk, NY, USA) were used for the statistical analysis. In the present study, a p-value < 0.05 was considered statistically significant.

All participants provided written informed consent, and the study protocol was approved by the Institutional Review Board (IRB) of the Seoul National University College of Medicine (IRB No. 0804-045-241).

Reciprocal adjustment for fruit or vegetable consumption frequency did not affect the overall results (Supplementary Materials 3 and 4). Furthermore, as we grouped “never” and “less than once/ week” together (Supplementary Materials 5 and 6), we found that the trends of results were not different from the main analyses. The present study performed additional analyses using generalized linear mixed-effects models to analyze the repeated measures pertaining to the subjects who participated in the survey three consecutive times (n= 305). The risk of depression decreased with an increase in the weekly frequency of fruit consumption in a dosedependent manner. In the multiple covariate-adjusted model, the ORs for depression were 1.34 (95% CI, 0.48 to 3.71), 0.66 (95% CI, 0.27 to 1.60), 0.53 (95% CI, 0.18 to 1.56), and 0.40 (95% CI, 0.16 to 1.02) at weekly frequencies of less than 1 time/wk, 1-3 times/wk, 4-6 times/wk, and daily. However, significant results were observed only in relation to weekly fruit consumption of 4-6 times/wk in the crude model and the daily fruit consumption in the crude and age- and sex-adjusted models (Table 3). There was no significant association between vegetable consumption and the level of depression or depression status with regard to repeated measures (Table 3).

Regarding the dose-dependent relationship, the SGDS-K scores had a negative linear relationship with the frequency of weekly fruit consumption (β, -0.04; 95% CI, -0.06 to -0.03). This pattern was consistent in the linear mixed-effects model (β, -0.04; 95% CI, -0.06 to -0.03). The greater the frequency of fruit consumption, the lower the risk of depression in both the cross-sectional model (OR, 0.90; 95% CI, 0.84 to 0.97) and repeated measures (OR, 0.87; 95% CI, 0.78 to 0.96), when the frequency of fruit consumption was considered as a continuous variable. The linearity was not significant in the case of vegetable consumption (Table 4).

Fruits and vegetables contain several nutrients that are known to prevent depressive symptoms. Defects in antioxidant defense systems are associated with a high incidence of depression. This is because depression increases the production of proinflammatory cytokines, which induce reactive oxygen species, which then trigger the lipid peroxidation process, compromising the brain [27,28]. Hence, antioxidants such as vitamins C, E, and β-carotene play an important role [3,27]. For example, vitamin E prevents peroxidation of ω3 polyunsaturated fatty acid [25]. Nutrients such as folate are essential for the biosynthesis of neurotransmitters such as serotonin, which stabilizes mood, as folate acts as a methyl donor, which is essential for the conversion of homocysteine to methionine by methylation [29]. Thus, folate deficiency is associated with a high risk of depression as neurotransmitters such as serotonin decrease [30]. Vitamin B6 is also known to mitigate the symptoms of hormone-related depression, as vitamin B6 is also involved in the synthesis of neurotransmitters that regulate emotions [31].

Conversely, a plausible explanation for the inverse association between depression and fruit consumption, in contrast to vegetable consumption, is that fruits can substitute for sugar-rich desserts. As sugar consumption has been associated with a higher incidence of depression in six countries [32], eating fruits instead of a sugarrich dessert may have a more desirable effect in preventing depressive symptoms. Additionally, several antioxidative nutrients, such as tangeretin, nobiletin, rutin, and resveratrol, are present only in fruits and not in vegetables [31].

First, the data were analyzed for both the level and risk of depression. Most previous studies focused on the OR of depression, whereas the present study also presented the level of depressive symptoms using the SGDS-K scores. Second, unlike most crosssectional studies, the present study conducted a longitudinal analysis using a generalized linear mixed-effects model. Lastly, to the best of our knowledge, this is the first study to analyze the associations between the level of depressive symptoms, which is explained by the SGDS-K score, and fruit and vegetable consumption in elderly Korean people exclusively, although a few studies have involved Korean adults [12,13] and adolescents [14,15]. Ji et al. [12] included female subjects below 70 years of age, which may not represent the elderly Korean population. A study by Ju & Park [13] covered both the sex and age of elderly subjects; however, vegetable and fruit consumption were not analyzed separately.

First, only the frequency of dietary intake was included in the analyses. Hence, data regarding the quantity, quality, cooking methods, nutrient density, and the proportion of fruits and vegetables in the total dietary intake were unavailable. In addition, it was not clear if fruit juice or kimchi was included in the fruit or vegetable consumption, considering the character of the question in the questionnaire. However, as the current study involved elderly subjects and considering that the KEEP-II study included numerous questions that were answered repeatedly every year, enquiries regarding the usual dietary behavior, such as the weekly frequency of consumption, may be more realistic and intuitive for the participants than enquiries about the exact quantity of each food item, which results in higher data reliability.

Second, depression was not assessed using clinical diagnoses; instead, the SGDS-K was used as a screening tool. However, considering that the subjects were elderly, feeling depressed may not warrant a visit to the doctor, as they are reluctant to burden their families and do not perceive depression as a serious condition. Accordingly, SGDS-K may be more inclusive.

Third, this study used a cross-sectional analysis. Hence, the causality of fruit and vegetable consumption with respect to the level of depression could not be proven. As depression can alter dietary style, reverse causation cannot be excluded in the correlation between the variables. Further longitudinal or cohort studies are needed to confirm the causality.

Fourth, more frequent fruit or vegetable intake, such as twice or thrice a day, was not considered. For example, elderly Korean individuals often eat kimchi at every meal. Thus, for future studies, a clearer representation of the food category is required.

Lastly, because the survey used a self-reporting method, recall bias cannot be excluded, especially in older adults. Furthermore, there may be residual confounding factors, such as total energy intake, which was obtained in only a small number of people who participated in the FFQ in 2012 (n= 385); thus, BMI was used instead. Cognitive status was not considered, as the participants of the KEEP survey were required to have an adequate cognitive level. Therefore, the current study is only applicable to fully cognitive elderly individuals with typical Korean dietary behavior.

In conclusion, the present study implies that the level of depression, which is explained by the SGDS-K score, is inversely associated with the frequency of fruit intake in a dose-dependent manner. Vegetable consumption did not display a dose-dependent relationship with depression level. However, SGDS-K scores were significantly negatively associated with vegetable intake compared to no vegetable consumption. Further studies are required to elucidate the underlying mechanisms and to identify the causality between the frequency of fruit and vegetable intake and the level of depression.