Association between plant-based diet indices and depressive symptoms among South Korean adults: a cross-sectional study using the 2014 and 2016 Korea National Health and Nutrition Examination Surveys Authors: Soohyun Park, Yoon Jung Park Published: 2026 ABSTRACT AbstractObjectiveThis study investigated the association between plant-based diet indices—overall plant-based diet index (PDI), healthful PDI (hPDI), and unhealthful PDI (uPDI)—and depressive symptoms in South Korean adults.MethodsThis cross-sectional study analyzed 5,846 participants (aged 19–64 years) using data from the 2014 and 2016 South Korea National Health and Nutrition Examination Survey. Dietary intake was assessed with a semiquantitative food frequency questionnaire, from which PDIs were derived. Depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9). Survey-weighted linear and logistic regression models were applied to assess associations, adjusting for sociodemographic, lifestyle, and clinical factors.ResultsIn fully adjusted models, higher overall PDI and hPDI were associated with lower PHQ-9 scores (β=−0.23; 95% confidence interval [CI], −0.41 to −0.04 and β=−0.16; 95% CI, −0.30 to −0.02 per 10-unit increment, respectively), whereas higher uPDI scores were associated with higher PHQ-9 scores (β=0.21; 95% CI, 0.07 to 0.35 per 10-unit increment). For clinical depressive symptoms (PHQ-9 ≥10), each 10-unit increase in overall PDI was associated with a 33% reduction in odds (odds ratio, 0.67; 95% CI, 0.50 to 0.89). Associations for hPDI and uPDI were attenuated and not statistically significant. Subgroup analyses revealed that these associations varied by sex, age, and obesity status.ConclusionGreater adherence to healthy plant-based foods and lower intake of less healthy plant-based foods were associated with fewer depressive symptoms among South Korean adults. These findings highlight the importance of plant-based food quality, rather than quantity alone, in supporting mental health.Keywords: Plant-based diet, Depression, Patient Health Questionnaire, Cross-sectional studies, National Health and Nutrition Examination Survey INTRODUCTION Background Depressive disorders are among the most common mental health conditions worldwide, affecting an estimated 332 million people. They remain a leading cause of disability, with the age-standardized disease burden increasing by 16.4% between 2010 and 2021 [1]. Beyond their psychological impact, depressive symptoms are associated with a higher risk of physical comorbidities, including cardiovascular disease (CVD) and obesity, and, in severe cases, can lead to suicide, substantially reducing quality of life [2]. In South Korea, the burden is particularly pronounced, with suicide rates nearly twice the OECD (Organisation for Economic Co-operation and Development) average, highlighting the urgent need for effective prevention strategies [3]. Identifying modifiable risk factors for depression is therefore a key public health priority. Recently, health policies and clinical guidelines have increasingly emphasized lifestyle interventions, particularly dietary modification, for both the prevention and management of mental disorders [4]. Dietary patterns may influence the development and course of depression through mechanisms such as inflammation and alterations in gut microbiota [5]. Evidence suggests that adherence to healthful dietary patterns, such as the Mediterranean diet, or avoidance of Western-style diets, may reduce the risk of depression [6]. Vegetarian and plant-based diets, in particular, have been linked to a range of health benefits, including reduced risks of chronic diseases and mortality [7]. However, findings on their association with depression have been inconsistent. This inconsistency may reflect differences in how plant-based diets are defined and the limited consideration of the quality of plant-based food within these dietary patterns [8,9]. Notably, not all plant-derived foods are beneficial; for example, higher consumption of refined grains has been associated with adverse health outcomes [10]. To address these limitations, plant-based diet indices have been developed to better capture diet quality. These include the overall plant-based diet index (PDI), healthful PDI (hPDI), and unhealthful PDI (uPDI), which distinguish between beneficial and less beneficial plant-based foods [11]. Unlike traditional dietary classifications, these indices allow for a more nuanced assessment of dietary patterns without excluding animal products. Previous studies have demonstrated an association between plant-based diet indices and mental health [12,13]. For example, a large prospective study from the UK Biobank found that higher hPDI scores were associated with a lower risk of depression, whereas higher uPDI scores were associated with increased risk [12]. Similar findings have been reported in US populations [13]. However, most of this evidence is derived from Western populations, and data from non-Western settings remain limited. Given that dietary habits vary substantially across cultures, these findings may not be directly generalizable. Traditional East Asian diets, for instance, are typically centered on rice, include higher consumption of soy products and fermented vegetables, and feature lower dairy intake compared with Western diets [14]. Examining these associations in East Asian populations is important to determine whether the observed relationships hold across different dietary contexts. Objectives This study aimed to investigate the association between plant-based diet indices and depressive symptoms using nationally representative data from the Korea National Health and Nutrition Examination Survey (KNHANES). We hypothesized that both overall PDI and hPDI would be inversely associated with depressive symptoms, whereas uPDI would be positively associated. METHODS Ethics statement This study was approved by the Institutional Review Board of Ewha Womans University (No. ewha-202510-0011-01). The requirement for informed consent was waived due to the retrospective nature of the study and the use of de-identified data. All procedures were conducted in accordance with the principles of the Declaration of Helsinki. The study was reported following the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines (Supplementary Material S1). Study design and setting This cross-sectional study used population-based data from the 2014 and 2016 KNHANES, a nationally representative survey conducted annually by the Korea Disease Control and Prevention Agency. Participants were selected using a stratified, multistage, clustered probability sampling design to represent the noninstitutionalized South Korean population. These survey years were selected because they were the only cycles that included both Patient Health Questionnaire-9 (PHQ-9) data (collected biennially since 2014) and semiquantitative food frequency questionnaire (FFQ) data (available from 2012–2016), both required for this analysis. Participants Among 15,700 survey participants, individuals were excluded sequentially according to the following criteria: age <19 or ≥65 years (n=6,560); missing or incomplete PHQ-9 (n=1,148) or FFQ data (n=1,894); implausible total energy intake (n=47); pregnancy or lactation (n=106); and current treatment for depression or nonresponse (n=99). After these exclusions, 5,846 participants were included in the final analysis (Fig. 1). Study participants selection flowchart. KNHANES, Korea National Health and Nutrition Examination Survey; PHQ-9, Patient Health Questionnaire-9; FFQ, food frequency questionnaire. Variables The primary outcome was depressive symptoms assessed using the PHQ-9, and the primary exposures were PDI, hPDI, and uPDI. Covariates included sociodemographic characteristics (sex, age, marital status, household size, education level, employment status, and household income), lifestyle factors (physical activity, smoking status, alcohol consumption, and body mass index [BMI]), clinical conditions (diabetes, hypertension, and CVD), and total daily energy intake. Sociodemographic data were collected through interviews conducted by trained personnel. Physical activity was classified as sufficient (≥150 min/wk of moderate activity, ≥75 min/wk of vigorous activity, or an equivalent combination) or insufficient. Alcohol consumption was categorized as a nondrinker or drinker based on lifetime drinking experience and drinking frequency over the past year. BMI (kg/m²) was categorized as underweight (<18.5 kg/m²), normal weight (18.5 to <23.0 kg/m²), overweight (23.0 to <25.0 kg/m²), or obese (≥25.0 kg/m²). For descriptive analyses, household income and BMI were presented as categorical variables (income quartiles and BMI categories), but both were treated as continuous variables in regression models. Diabetes was defined as fasting plasma glucose ≥126 mg/dL, use of antihyperglycemic medication or insulin, or a physician diagnosis. Hypertension was defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or use of antihypertensive medication. CVD included physician-diagnosed stroke, myocardial infarction, or angina. Total daily energy intake (kcal/day) was calculated from FFQ data and analyzed as a continuous variable. Data sources Assessment of depressive symptoms (outcome) Depressive symptoms were assessed using the PHQ-9, administered as part of the KNHANES health questionnaire. The PHQ-9 is a 9-item tool based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria. Participants rated their symptoms over the past two weeks on a 4-point scale ranging from 0 (“not at all”) to 3 (“nearly every day”). Total scores range from 0 to 27, with higher scores indicating greater severity of depressive symptoms. In this study, the PHQ-9 total score was analyzed as a continuous variable. Additionally, depressive symptoms were defined as a PHQ-9 score ≥10, a clinically validated cutoff with 88% sensitivity and 88% specificity [15], and were analyzed as a binary outcome. Dietary assessment and plant-based diet indices (exposure) Dietary intake was assessed using a validated 109-item semiquantitative FFQ administered in KNHANES [16]. Trained interviewers collected information on usual dietary intake over the previous year. For each item, participants reported consumption frequency using nine response categories and usual portion size per occasion based on standard portion options. Reported frequencies were converted to daily intakes (times/day) using the midpoint of each category, and daily intake was calculated by multiplying frequency by portion size (servings per occasion). The primary exposure—overall PDI, hPDI, and uPDI—was calculated based on established methods [11], with minor modifications to reflect South Korean dietary patterns and the available FFQ items. Briefly, individual food items were grouped into 18 categories based on nutritional and culinary similarity, and further classified into healthy plant-based foods, less healthy plant-based foods, and animal foods. Healthy plant-based foods included whole grains, fruits, vegetables, nuts, legumes, tea and coffee, and fermented foods. Less healthy plant-based foods included fruit juices, refined grains, potatoes, sugar-sweetened beverages, and sweets and desserts. Animal-derived foods included animal fat, dairy, eggs, fish and seafood, meat, and others (Table S1). Since sugar and cream added to coffee were assessed separately in the FFQ, plain tea and coffee were classified as healthy plant-based foods, added sugar as a less healthy plant-based food, and cream as an animal-derived food. A fermented foods group was created to reflect the high intake of kimchi and fermented soybean products in South Korean adults and was classified as a healthy plant-based food group based on emerging evidence suggesting potential benefits for depressive symptoms [17]. Vegetable oils were not included, as oil intake was not assessed in the FFQ. Daily intakes for each food group were summed and energy-adjusted using the nutrient density method (per 1,000 kcal) [18]. Participants were assigned quintile-based scores (1–5) for each food group. For overall PDI, all plant-based foods were scored positively and animal-derived foods inversely. For hPDI, healthy plant-based foods were scored positively and less healthy plant-based foods inversely. For uPDI, less healthy plant-based foods were scored positively and healthy plant-based foods inversely. Scores across all 18 food groups were summed to generate each index, with a theoretical range of 18 to 90. Diet indices were categorized into tertiles to ensure adequate sample size in each category while maintaining variability for trend analysis. Continuous analyses (per 10-point increment) were also conducted to preserve statistical power and assess dose-response relationships. Statistical analysis All analyses used survey-weighted procedures to account for the complex, multistage probability sampling design. Sampling weights, stratification, and clustering were incorporated to produce nationally representative estimates and valid variance estimates for the South Korean population. Participant characteristics across overall tertiles of overall PDI are presented as numbers and percentages for categorical variables and weighted means with standard errors for continuous variables. Group differences were assessed using the Rao-Scott χ² test for categorical variables and weighted linear regression for continuous variables. Survey-weighted linear regression was used to examine associations with PHQ-9 scores, and survey-weighted logistic regression was used for depressive symptoms defined by the PHQ-9 cutoff. Diet indices were analyzed both as tertiles and as a continuous variable (per 10-point increment). Results were presented as β-coefficients for linear regression and odds ratios for logistic regression, each with 95% confidence intervals (95% CIs). Linear trends across tertiles were assessed by modeling the median value of each tertile as a continuous variable. Covariates were selected a priori to control for potential confounding, and sequentially adjusted models (Models 1–4) were fitted. Model 1 was unadjusted. Model 2 adjusted for age and sex. Model 3 additionally adjusted for household income, education level, marital status, household size, and employment status. Model 4 further adjusted for physical activity, smoking status, alcohol consumption, BMI, diabetes, hypertension, CVD, and total energy intake. To improve model stability, some variables were dichotomized before analysis: household size (living alone vs. with family) and education level (college or higher vs. high school or less). For descriptive purposes, household income and BMI were categorized, but both were treated as continuous variables in regression models. Subgroup analyses were conducted by sex (male, female), age group (19–29, 30–49, 50–64 years), and obesity status (presence, absence) using the fully adjusted model (Model 4). Within each sex, differences in food group intake by depressive symptom status were assessed using survey-weighted linear regression. Statistical analyses were conducted using SAS ver. 9.4 (SAS Institute Inc.). All analyses were two-sided, and P <0.05 was considered statistically significant. Participants with missing covariate data were excluded (listwise deletion), and no imputation was performed. RESULTS Participants Of the 15,700 individuals who participated in the KNHANES, 5,846 were eligible and included in the final analysis after sequential exclusions; the participant selection process is presented in Fig. 1. Descriptive data Participant characteristics by overall PDI tertiles are presented in Table 1. Compared with those in the lowest tertile, participants in the highest tertile were older and included higher proportions of women, married individuals, and employed individuals, whereas the proportions of college graduates, current smokers, and drinkers were lower. The highest tertile also had lower total energy intake and lower PHQ-9 scores, but a higher prevalence of diabetes, hypertension, and CVD. Household income, physical activity, and obesity status were similar across tertiles. Table 1.General characteristics of the participants across tertiles of overall PDICharacteristicOverall PDITertile 1 (n=1,722)Tertile 2 (n=1,850)Tertile 3 (n=2,274)P-valuea)Score range30–4950–5455–76-Age (yr)36.60±0.3740.76±0.4144.93±0.32<0.001Sex Male738 (54.6)705 (49.8)832 (46.6)<0.001 Female984 (45.4)1,145 (50.2)1,442 (53.4)Marital status Married1,214 (60.7)1,476 (72.1)1,954 (80.6)<0.001 Unmarried508 (39.3)374 (27.9)320 (19.4)Household size Alone131 (8.3)133 (6.9)137 (5.8)0.099 2–3 People828 (47.6)916 (48.5)1,215 (50.1) ≥4 People763 (44.0)801 (44.6)922 (44.1)Education level ≤Elementary83 (3.5)172 (6.7)290 (9.4)<0.001 Middle104 (5.3)151 (6.5)252 (9.8) High684 (43.6)718 (40.7)803 (36.9) ≥College851 (47.7)809 (46.1)927 (43.9)Employment status Employed1,119 (66.3)1,240 (68.8)1,554 (70.6)0.036 Unemployed603 (33.7)609 (31.2)720 (29.4)Household income (million KRW)d) <1110 (7.2)102 (4.8)142 (5.6)0.299 1–2198 (11.1)217 (11.2)292 (11.5) 2–4549 (32.3)588 (31.7)727 (32.3) >4863 (49.5)942 (52.4)1,107 (50.5)Physical activityb) Insufficiently active818 (43.7)862 (43.7)1,071 (44.4)0.920 Sufficiently active902 (56.3)988 (56.3)1,202 (55.6)Smoking status Ex, nonsmoker1,310 (71.4)1,500 (76.4)1,918 (81.0)<0.001 Current smoker409 (28.6)344 (23.6)351 (19.0)Alcohol drinking statusc) Nondrinker564 (28.3)745 (35.7)1,109 (45.1)<0.001 Drinker1,155 (71.7)1,099 (64.3)1,161 (54.9)Body mass index (kg/m²)d) Classification of obesitye) Underweight81 (4.7)93 (5.0)88 (3.9)0.543 Normal761 (42.9)744 (39.9)939 (41.3) Overweight361 (21.7)423 (22.3)519 (22.2) Obesity519 (30.7)588 (32.8)727 (32.5)Energy intake (kcal/day)2,168.87±23.152,054.48±22.101,950.09±19.34<0.001Diabetes No1,541 (95.6)1,630 (93.7)1,979 (91.5)<0.001 Yes81 (4.4)128 (6.3)197 (8.5)Hypertension No1,455 (85.4)1,505 (83.0)1,731 (77.4)<0.001 Yes264 (14.6)343 (17.0)541 (22.6)CVD No1,708 (99.5)1,820 (99.0)2,211 (97.6)<0.001 Yes14 (0.5)30 (1.0)63 (2.4)PHQ-9 score2.73±0.102.56±0.092.27±0.08<0.001Values are presented as mean±standard error or unweighted number (weighted percentage). Values are presented as mean±standard error or number (%). Missing data were <5% for all covariates. Specifically, missing values were observed for education level (n=2), employment status (n=1), household income (n=9), physical activity (n=3), smoking status (n=14), alcohol consumption (n=13), body mass index (n=3), diabetes (n=290), and hypertension (n=7); all other variables had complete data.PDI, plant-based diet index; KRW, South Korean Won; CVD, cardiovascular disease; PHQ-9, Patient Health Questionnaire-9.a)P-values were calculated using the PROC SURVEYFREQ (Rao-Scott χ² test) and PROC SURVEYREG.b)Physical activity was categorized as sufficient (≥150 min/week of moderate activity, ≥75 min/wk of vigorous activity, or an equivalent combination) or insufficient (not meeting these criteria).c)Alcohol consumption was defined as ≥1 drink per month during the past year.d)Variables were categorized for descriptive purposes in this table but treated as continuous or binary variables in the regression analyses.e)Body mass index categories were defined as follows: underweight (<18.5 kg/m²), normal (18.5 to <23.0 kg/m²), overweight (23.0 to <25.0 kg/m²), and obese (≥25.0 kg/m²). General characteristics of the participants across tertiles of overall PDI Values are presented as mean±standard error or unweighted number (weighted percentage). Values are presented as mean±standard error or number (%). Missing data were <5% for all covariates. Specifically, missing values were observed for education level (n=2), employment status (n=1), household income (n=9), physical activity (n=3), smoking status (n=14), alcohol consumption (n=13), body mass index (n=3), diabetes (n=290), and hypertension (n=7); all other variables had complete data. PDI, plant-based diet index; KRW, South Korean Won; CVD, cardiovascular disease; PHQ-9, Patient Health Questionnaire-9. P-values were calculated using the PROC SURVEYFREQ (Rao-Scott χ² test) and PROC SURVEYREG. Physical activity was categorized as sufficient (≥150 min/week of moderate activity, ≥75 min/wk of vigorous activity, or an equivalent combination) or insufficient (not meeting these criteria). Alcohol consumption was defined as ≥1 drink per month during the past year. Variables were categorized for descriptive purposes in this table but treated as continuous or binary variables in the regression analyses. Body mass index categories were defined as follows: underweight (<18.5 kg/m²), normal (18.5 to <23.0 kg/m²), overweight (23.0 to <25.0 kg/m²), and obese (≥25.0 kg/m²). Outcome data Of the 5,846 participants, 280 had depressive symptoms, defined as a PHQ-9 score ≥10. Main results Plant-based diet indices and depressive symptoms Associations between plant-based diet indices and PHQ-9 scores are presented in Table 2. In fully adjusted models (Model 4), higher overall PDI and hPDI were associated with lower PHQ-9 scores, whereas higher uPDI was associated with higher PHQ-9 scores. Specifically, compared with the lowest tertile, participants in the highest tertile had lower PHQ-9 scores for overall PDI (β=−0.26; 95% CI, −0.50 to −0.02) and hPDI (β=−0.32; 95% CI, −0.63 to −0.01), while uPDI showed a positive association (β=0.29; 95% CI, 0.03 to 0.55). Trend tests across tertiles were statistically significant for all three indices in Model 4 (P-trend=0.031 for overall PDI; 0.042 for hPDI; 0.029 for uPDI). Similar results were observed when PDIs were analyzed as continuous variables: per 10-unit increases in overall PDI and hPDI were inversely associated with PHQ-9 scores, whereas uPDI was positively associated (Table 2). Table 2.Association between plant-based diet indices and PHQ-9 scoresModelPer 10-unit incrementP-trenda)β-coefficient (95% CI)P-valuea)Tertile 1Tertile 2Tertile 3β-coefficient (95% CI)Overall PDI Model 10 (Reference)−0.16 (−0.41 to 0.08)−0.46 (−0.70 to −0.22)<0.001−0.38 (−0.57 to −0.20)<0.001 Model 20 (Reference)−0.12 (−0.36 to 0.13)−0.35 (−0.60 to −0.11)0.004−0.31 (−0.50 to −0.12)0.001 Model 30 (Reference)−0.08 (−0.33 to 0.16)−0.33 (−0.57 to −0.08)0.007−0.29 (−0.48 to −0.10)0.002 Model 40 (Reference)−0.11 (−0.36 to 0.13)−0.26 (−0.50 to −0.02)0.031−0.23 (−0.41 to −0.04)0.017hPDI Model 10 (Reference)−0.34 (−0.59 to −0.10)−0.49 (−0.73 to −0.25)<0.001−0.21 (−0.32 to −0.10)<0.001 Model 20 (Reference)−0.29 (−0.55 to −0.03)−0.40 (−0.69 to −0.11)0.008−0.18 (−0.31 to −0.04)0.010 Model 30 (Reference)−0.25 (−0.51 to 0.01)−0.41 (−0.70 to −0.12)0.006−0.19 (−0.33 to −0.06)0.005 Model 40 (Reference)−0.22 (−0.49 to 0.05)−0.32 (−0.63 to −0.01)0.042−0.16 (−0.30 to −0.02)0.025uPDI Model 10 (Reference)0.15 (−0.07 to 0.37)0.37 (0.12 to 0.61)0.0030.26 (0.13 to 0.39)<0.001 Model 20 (Reference)0.24 (0.02 to 0.47)0.47 (0.20 to 0.73)<0.0010.33 (0.19 to 0.48)<0.001 Model 30 (Reference)0.21 (−0.01 to 0.43)0.35 (0.08 to 0.61)0.0110.25 (0.11 to 0.40)<0.001 Model 40 (Reference)0.22 (0.00 to 0.44)0.29 (0.03 to 0.55)0.0290.21 (0.07 to 0.35)0.003Model 1, crude model; Model 2, adjusted for age (continuous) and sex; Model 3, further adjusted for average monthly household income (continuous), education level, marital status, household size, and employment status; Model 4, additionally adjusted for physical activity, smoking status, alcohol consumption, body mass index (continuous), diabetes, hypertension, cardiovascular disease, and total energy intake (continuous).PHQ-9, Patient Health Questionnaire-9; CI, confidence interval; PDI, plant-based diet index; hPDI, healthful plant-based diet index; uPDI, unhealthful plant-based diet index.a)P-values and P-trend were obtained using survey-weighted linear regression (PROC SURVEYREG). P for trend was calculated by assigning the median value of each tertile and modeling it as a continuous variable. Association between plant-based diet indices and PHQ-9 scores Model 1, crude model; Model 2, adjusted for age (continuous) and sex; Model 3, further adjusted for average monthly household income (continuous), education level, marital status, household size, and employment status; Model 4, additionally adjusted for physical activity, smoking status, alcohol consumption, body mass index (continuous), diabetes, hypertension, cardiovascular disease, and total energy intake (continuous). PHQ-9, Patient Health Questionnaire-9; CI, confidence interval; PDI, plant-based diet index; hPDI, healthful plant-based diet index; uPDI, unhealthful plant-based diet index. P-values and P-trend were obtained using survey-weighted linear regression (PROC SURVEYREG). P for trend was calculated by assigning the median value of each tertile and modeling it as a continuous variable. Associations between PDIs and the odds of depressive symptoms are shown in Table 3. In Model 4, the inverse association for overall PDI was attenuated in the tertile comparison (T3 vs. T1: OR, 0.72; 95% CI, 0.49 to 1.07; P-trend=0.107), but remained significant when modeled continuously (per 10-unit increment: OR, 0.67; 95% CI, 0.50 to 0.89). For hPDI, inverse associations observed in less adjusted models were no longer statistically significant after full adjustment. In contrast, uPDI showed a positive association in Models 1–3, but this was attenuated and became nonsignificant in Model 4 (T3 vs. T1: OR, 1.49; 95% CI, 0.97 to 2.28; P-trend=0.059; per 10-unit increment: OR, 1.23; 95% CI, 0.99 to 1.54) (Table 3). Collectively, higher overall PDI and hPDI were associated with lower PHQ-9 scores, whereas higher uPDI was associated with higher scores; corresponding associations with depressive symptoms were in the same direction but weaker after full adjustment. Table 3.Associations of plant-based diet indices with depressive symptomsModelPer 10-unit incrementP-trenda)OR (95% CI)P-valuea)Tertile 1Tertile 2Tertile 3OR (95% CI)Overall PDI Model 11 (Reference)0.87 (0.62 to 1.23)0.65 (0.45 to 0.93)0.0190.63 (0.48 to 0.81)<0.001 Model 21 (Reference)0.88 (0.62 to 1.25)0.66 (0.46 to 0.97)0.0320.62 (0.47 to 0.82)<0.001 Model 31 (Reference)0.91 (0.64 to 1.30)0.69 (0.47 to 1.01)0.0530.64 (0.48 to 0.84)0.001 Model 41 (Reference)0.87 (0.59 to 1.29)0.72 (0.49 to 1.07)0.1070.67 (0.50 to 0.89)0.006hPDI Model 11 (Reference)0.79 (0.58 to 1.07)0.71 (0.51 to 0.98)0.0410.89 (0.76 to 1.04)0.129 Model 21 (Reference)0.78 (0.56 to 1.07)0.68 (0.46 to 1.01)0.0570.88 (0.73 to 1.06)0.179 Model 31 (Reference)0.80 (0.58 to 1.11)0.63 (0.41 to 0.95)0.0260.85 (0.70 to 1.02)0.081 Model 41 (Reference)0.84 (0.59 to 1.19)0.71 (0.45 to 1.12)0.1410.87 (0.71 to 1.07)0.194uPDI Model 11 (Reference)1.01 (0.69 to 1.47)1.52 (1.08 to 2.14)0.0131.32 (1.10 to 1.58)0.002 Model 21 (Reference)1.11 (0.74 to 1.65)1.76 (1.17 to 2.64)0.0051.44 (1.17 to 1.79)0.001 Model 31 (Reference)1.10 (0.74 to 1.63)1.61 (1.07 to 2.44)0.0191.35 (1.09 to 1.67)0.006 Model 41 (Reference)1.14 (0.76 to 1.71)1.49 (0.97 to 2.28)0.0591.23 (0.99 to 1.54)0.058Model 1, crude model; Model 2, adjusted for age (continuous) and sex; Model 3, further adjusted for average monthly household income (continuous), education level, marital status, household size, and employment status; Model 4, additionally adjusted for physical activity, smoking status, alcohol consumption, body mass index (continuous), diabetes and hypertension, cardiovascular disease, and total energy intake (continuous).OR, odds ratio; CI, confidence interval; PDI, plant-based diet index; hPDI, healthful plant-based diet index; uPDI, unhealthful plant-based diet index.a)P-values and P-trend were obtained using survey-weighted logistic regression (PROC SURVEYLOGISTIC). P-trend was calculated by assigning the median value of each tertile and modeling it as a continuous variable. Associations of plant-based diet indices with depressive symptoms Model 1, crude model; Model 2, adjusted for age (continuous) and sex; Model 3, further adjusted for average monthly household income (continuous), education level, marital status, household size, and employment status; Model 4, additionally adjusted for physical activity, smoking status, alcohol consumption, body mass index (continuous), diabetes and hypertension, cardiovascular disease, and total energy intake (continuous). OR, odds ratio; CI, confidence interval; PDI, plant-based diet index; hPDI, healthful plant-based diet index; uPDI, unhealthful plant-based diet index. P-values and P-trend were obtained using survey-weighted logistic regression (PROC SURVEYLOGISTIC). P-trend was calculated by assigning the median value of each tertile and modeling it as a continuous variable. Subgroup analyses In subgroup analyses, by sex, overall PDI showed an inverse trend across tertiles in males, whereas hPDI showed an inverse trend in females; uPDI showed no clear trend in either sex. When stratified by age group, an inverse trend for overall PDI was observed among participants aged 30–49 years, but not in those aged 19–29 or 50–64 years; no consistent trends were observed for hPDI or uPDI across age groups. By obesity status, hPDI showed an inverse trend in participants with obesity, whereas uPDI showed a positive trend in those without obesity (Table 4). Overall, these findings indicated that associations between plant-based diet indices and depressive symptoms vary across subgroups and are not uniform. Table 4.Subgroup analyses of the associations between plant-based diet indices and depressive symptoms by sex, age, and obesity status (n=5,846)CharacteristicNo. of patientsTertiles of overall PDITertiles of hPDITertiles of uPDITertile 2Tertile 3P-trenda)Tertile 2Tertile 3P-trenda)Tertile 2Tertile 3P-trenda)Sex Male2,2750.59 (0.29 to 1.21)0.44 (0.22 to 0.89)0.0220.87 (0.45 to 1.68)0.99 (0.41 to 2.37)0.9271.01 (0.39 to 2.65)1.51 (0.57 to 3.97)0.277 Female3,5711.12 (0.72 to 1.73)0.94 (0.59 to 1.49)0.7440.79 (0.52 to 1.20)0.58 (0.35 to 0.99)0.0431.16 (0.75 to 1.79)1.43 (0.89 to 2.32)0.143Age (yr) 19–299171.08 (0.54 to 2.16)0.85 (0.35 to 2.03)0.7650.92 (0.46 to 1.84)1.31 (0.48 to 3.58)0.7700.42 (0.15 to 1.14)0.76 (0.29 to 1.99)0.906 30–492,8240.77 (0.45 to 1.33)0.38 (0.21 to 0.66)<0.0010.88 (0.54 to 1.44)0.56 (0.28 to 1.10)0.0931.36 (0.76 to 2.44)1.60 (0.90 to 2.85)0.109 50–642,1051.05 (0.52 to 2.14)1.36 (0.68 to 2.72)0.3400.43 (0.17 to 1.06)0.40 (0.17 to 0.96)0.1881.28 (0.66 to 2.48)1.92 (0.96 to 3.85)0.074Obesity statusb) Absence4,0120.94 (0.61 to 1.45)0.80 (0.50 to 1.28)0.3470.84 (0.54 to 1.31)0.88 (0.50 to 1.52)0.6341.25 (0.76 to 2.04)1.75 (1.03 to 2.99)0.035 Presence1,8340.68 (0.33 to 1.40)0.49 (0.24 to 1.00)0.0510.80 (0.41 to 1.57)0.40 (0.17 to 0.94)0.0290.97 (0.46 to 2.06)1.09 (0.50 to 2.40)0.830Values are presented as OR (95% CI). The lowest tertile (Tertile 1) served as the reference category (OR, 1.00). All subgroup analyses used the fully adjusted Model 4, with the stratification variable excluded from the adjustment set.PDI, plant-based diet index; hPDI, healthful plant-based diet index; uPDI, unhealthful plant-based diet index; OR, odds ratio; CI, confidence interval.a)P-trend was obtained using survey-weighted logistic regression (PROC SURVEYLOGISTIC); P-trend was calculated by assigning the median value of each tertile and modeling it as a continuous variable.b)Obesity status was defined as body mass index <25.0 kg/m² (absence) and ≥25.0 kg/m² (presence). Subgroup analyses of the associations between plant-based diet indices and depressive symptoms by sex, age, and obesity status (n=5,846) Values are presented as OR (95% CI). The lowest tertile (Tertile 1) served as the reference category (OR, 1.00). All subgroup analyses used the fully adjusted Model 4, with the stratification variable excluded from the adjustment set. PDI, plant-based diet index; hPDI, healthful plant-based diet index; uPDI, unhealthful plant-based diet index; OR, odds ratio; CI, confidence interval. P-trend was obtained using survey-weighted logistic regression (PROC SURVEYLOGISTIC); P-trend was calculated by assigning the median value of each tertile and modeling it as a continuous variable. Obesity status was defined as body mass index <25.0 kg/m² (absence) and ≥25.0 kg/m² (presence). Food group intake differences by depressive symptoms Sex-stratified comparisons of food group intake by depressive symptom status are shown in Table 5. Among males, those with depressive symptoms consumed lower amounts of healthy plant-based foods—particularly whole grains, fruits, vegetables, and nuts—compared with those without depressive symptoms; lower intakes were also observed for fruit juices and potatoes. Intakes of refined grains, sweets and desserts, and most animal-derived foods were similar between groups. Table 5.Sex-stratified comparison of food categories by depressive symptom statusVariableMaleFemaleNondepressedDepressedP-valuea)NondepressedDepressedP-valuea)Healthy plant-based foods Whole grains0.466±0.0100.366±0.0500.0480.623±0.0100.523±0.0420.021 Fruits0.393±0.0110.225±0.031<0.0010.777±0.0140.643±0.0460.006 Vegetables1.007±0.0160.733±0.067<0.0011.456±0.0211.244±0.0770.008 Nuts0.035±0.0020.015±0.003<0.0010.047±0.0020.042±0.0110.606 Legumes0.135±0.0040.108±0.0180.1470.163±0.0040.157±0.0180.725 Tea and coffee0.589±0.0130.598±0.0800.9090.653±0.0140.783±0.0890.140 Fermented foods1.609±0.0271.466±0.1150.2261.799±0.0291.610±0.1300.154Less healthy plant-based foods Fruit juices0.046±0.0020.021±0.006<0.0010.034±0.0020.022±0.0040.006 Refined grains0.961±0.0110.971±0.0600.8700.798±0.0100.899±0.0430.021 Potatoes0.053±0.0020.028±0.004<0.0010.074±0.0020.068±0.0110.564 Sugar-sweetened beverages0.092±0.0050.136±0.0250.0820.051±0.0030.100±0.0180.006 Sweets and desserts0.402±0.0090.545±0.0870.1070.391±0.0080.501±0.0450.018Animal-derived foods Animal fat0.273±0.0080.424±0.0900.0950.235±0.0070.319±0.0400.039 Dairy0.318±0.0080.253±0.0370.0890.427±0.0090.418±0.0540.867 Eggs0.205±0.0050.265±0.0370.1130.276±0.0050.268±0.0240.730 Fish or seafood0.333±0.0080.319±0.0410.7410.444±0.0130.387±0.0390.172 Meat0.384±0.0060.364±0.0310.5220.335±0.0050.350±0.0220.490 Miscellaneous animal foodsb)0.042±0.0010.048±0.0070.3850.047±0.0010.053±0.0050.256Values are presented as mean±standard error of energy-adjusted intake (servings/1,000 kcal). Depressive status was defined as Patient Health Questionnaire-9 (PHQ-9) scores ≥10 (depressed) and PHQ-9 scores <10 (nondepressed).a)P-values were obtained using survey-weighted linear regression (PROC SURVEYREG).b)Miscellaneous animal foods refer to mixed dishes containing animal-based ingredients that are difficult to assign to a single animal food group (e.g., dumplings, pizza, South Korean blood sausage). Sex-stratified comparison of food categories by depressive symptom status Values are presented as mean±standard error of energy-adjusted intake (servings/1,000 kcal). Depressive status was defined as Patient Health Questionnaire-9 (PHQ-9) scores ≥10 (depressed) and PHQ-9 scores <10 (nondepressed). P-values were obtained using survey-weighted linear regression (PROC SURVEYREG). Miscellaneous animal foods refer to mixed dishes containing animal-based ingredients that are difficult to assign to a single animal food group (e.g., dumplings, pizza, South Korean blood sausage). Among females, those with depressive symptoms also had lower intakes of whole grains, fruits, and vegetables, along with lower fruit juice consumption. In addition, female participants reported higher intake of sugar-sweetened beverages, sweets and desserts, refined grains, and animal fat. Most other food groups—including nuts, legumes, tea/coffee, fermented foods, and most animal foods—did not differ meaningfully by depressive symptom status. Overall, depressive symptoms were consistently associated with lower intakes of key healthy plant-based foods in both sexes, whereas differences in less healthy plant-based foods and some animal-derived foods were more evident among females. DISCUSSION This study used KNHANES data to examine the association between plant-based diet indices and depressive symptoms among South Korean adults. Higher overall PDI and hPDI scores were associated with lower PHQ-9 scores, whereas higher uPDI scores were associated with higher PHQ-9 scores. These associations remained after adjusting for sociodemographic factors, lifestyle behaviors, clinical characteristics, and total energy intake. Similarly, higher overall PDI and hPDI scores were associated with lower odds of depressive symptoms, while higher uPDI scores were associated with higher odds; however, some associations were attenuated and no longer statistically significant after full adjustment. To date, relatively few studies have examined plant-based diet indices in relation to depression, and the available evidence is largely limited to specific populations [12,13]. In a large prospective cohort from the UK Biobank, overall PDI was not associated with depression risk, whereas hPDI and uPDI showed inverse and positive associations, respectively [12]. Similarly, an analysis of US National Health and Nutrition Examination Survey data reported no clear association for overall PDI, while higher hPDI was associated with lower odds of depressive symptoms and lower PHQ-9 scores, and higher uPDI was associated with higher odds and higher scores [13]. In contrast, the present study demonstrated significant associations for all three plant-based diet indices and PHQ-9 scores, with generally consistent directions for depressive symptoms. One possible explanation for these differences is variation in how plant-based diet indices are constructed. Most previous studies have applied indices developed in Western dietary contexts [11-13,19], whereas Asian dietary patterns—particularly in South Korea—are typically centered on rice with diverse side dishes. In South Korea, traditional fermented foods, such as kimchi and fermented soybean products, are consumed in substantial amounts (155.4 g/day in men and 105.5 g/day in women), warranting separate consideration in food classification [20,21]. Previous South Korean studies have often grouped these foods as “salty foods” and classified them as less healthy due to their sodium content and associations with chronic diseases [19,20]. In contrast, we classified them as healthy plant-based foods, considering their dietary importance and emerging evidence suggesting that fermentation-derived components may modulate gut microbiota and immune-inflammatory pathways relevant to mental health [17]. While sodium reduction remains an important public health priority [19,20], our approach provides a complementary perspective by incorporating potential benefits of fermented foods. Nevertheless, these findings should be interpreted with caution, as differences in index construction may partly explain variations in results. A clear understanding of how individual food groups contribute to disease risk, along with further evaluation of alternative scoring approaches, may help refine dietary recommendations [22]. In this study, hPDI was inversely associated with PHQ-9 scores and depressive symptoms, whereas uPDI showed positive associations. Notably, overall PDI also demonstrated an inverse association, suggesting that the benefits of healthy plant-based foods may substantially offset the potential harms of less healthy plant-based foods within the overall diet. These findings indicate that while increasing plant-based food intake may be beneficial, the quality of plant-based foods is likely more important for the prevention of depressive symptoms and the promotion of mental health. Accordingly, dietary strategies for mental health promotion should emphasize the selection of high-quality plant-based foods. Beyond mental health, adherence to a high-quality plant-based diet has also been associated with favorable long-term health outcomes. For example, a large prospective cohort study reported that higher hPDI was associated with lower risks of cancer, CVD, and type 2 diabetes, as well as a reduced likelihood of multimorbidity [22]. Several plausible mechanisms may explain the observed associations. First, the abundance of antioxidants and bioactive compounds in healthy plant-based foods may protect against depressive symptoms by reducing oxidative stress and modulating inflammation-related pathways. Plant-derived phytochemicals, particularly polyphenols found in fruits and vegetables, can inhibit the nuclear factor kappa B signaling pathway, thereby decreasing the production of pro-inflammatory cytokines and alleviating neuroinflammation, which may help improve depressive symptoms [5]. Second, a healthy plant-based diet is typically high in dietary fiber, which can influence the diversity and composition of the gut microbiota. Nondigestible carbohydrates are fermented by gut microbes to produce short-chain fatty acids, which play a role in regulating immune function and reducing systemic inflammation [23]. In this study, we classified fermented foods as healthy plant-based foods, given their potential to promote beneficial gut bacteria, including butyrate-producing species, and to contribute to anti-inflammatory effects. However, clinical evidence supporting the effects of fermented foods on depressive symptoms remains limited, with existing studies suggesting only modest benefits [17]. These mechanisms are broadly consistent with evidence from plant-based dietary patterns, including the Mediterranean diet, for which meta-analysis of randomized controlled trials has shown reductions in depressive symptoms, potentially mediated by improvements in gut microbiota and reductions in inflammation and oxidative stress [24]. Differences between male and female participants may reflect underlying biological, hormonal, and social factors that influence both psychological and physiological responses to diet [25]. As shown in Table 5, both sexes with depressive symptoms consumed lower intakes of healthy plant-based foods, particularly whole grains, fruits, and vegetables, indicating poorer overall diet quality. However, female participants with depressive symptoms also reported higher intakes of refined grains, sugar-sweetened beverages, and desserts, suggesting a greater contrast in diet quality. This pattern may help explain why the association between hPDI and depressive symptoms was more common in females. In contrast, among males, the pattern seemed to be driven mainly by lower intakes of healthy plant-based foods rather than higher intakes of less healthy foods, which may partly explain the stronger association observed for overall PDI, an index reflecting overall plant-based food consumption. Participants with depressive symptoms also shown distinct food intake patterns (Table 5). Female participants with depressive symptoms reported higher tea or coffee consumption, which may reflect greater reliance on caffeinated beverages to cope with fatigue or low energy; however, causal direction cannot be determined in this cross-sectional analysis [26]. Male participants with depressive symptoms consumed fewer potatoes than those without depressive symptoms. Although this may relate to reduced motivation affecting the selection of foods that require more preparation [27], the reasons for lower potato intake in such participants remain unclear. Additionally, differences in certain food categories may be difficult to interpret, as these foods can vary widely in form, ranging from raw to highly processed products. Age-stratified analyses showed a significant inverse association between overall PDI and depressive symptoms only among participants aged 30 to 49 years. Conversely, among those aged 50 to 64 years, both overall PDI and uPDI were positively associated with depressive symptoms. Among young adults aged 19 to 29 years, overall PDI and uPDI showed inverse associated with depressive symptoms—a pattern that differed from other age groups—whereas hPDI showed a positive association. In the context of negative emotional states, such as stress, the consumption of highly palatable, energy-dense foods may serve as a coping mechanism that temporarily reduces unpleasant feelings [28]. Prior research suggests that comfort eating may weaken the link between adverse life events and perceived stress, particularly among individuals with relatively low levels of depressive symptoms [28]. In young adults, irregular eating patterns and meal skipping [29] may contribute to higher uPDI scores through greater reliance on convenience foods and snacks as meal substitutes. Given that meal skipping has been associated with depression [29], this may partly explain the observed inverse association. In addition, differences in food choice patterns across age groups may have contributed to the heterogeneity in the observed associations. In analyses stratified by obesity status, higher hPDI was significantly associated with lower odds of depressive symptoms in individuals with obesity, whereas higher uPDI was significantly associated with increased odds in those without obesity. The inverse association between hPDI and depressive symptoms in the obese group may be partly explained by shared oxidative stress pathways involved in both obesity and depression [30]. Antioxidant-rich plant-based foods may help reduce obesity-related low-grade inflammation, thereby contributing to improved mental health. However, subgroup differences in sample size and estimate precision may have influenced statistical significance, and the cross-sectional design limits causal inference. Limitations This study has several limitations. First, its cross-sectional design precludes causal inference, as the temporal relationship between plant-based diet indices and depressive symptoms cannot be established. Bidirectional associations and reverse causality cannot be excluded, and incident risk could not be assessed without longitudinal follow-up. Second, dietary intake was assessed using an interviewer-administered FFQ and may be subject to recall bias, potentially leading to nondifferential misclassification and attenuation of associations. In addition, dietary assessment was conducted at a single time point, limiting the ability to capture long-term dietary patterns. Depressive symptoms were assessed using the self-administered PHQ-9 and may be influenced by reporting bias or differences in interpretation. Third, the classification of fermented foods as healthy plant-based foods differs from the original PDI framework and may have influenced the observed associations. Given the high consumption of fermented foods in South Korea, this classification may have contributed to the overall pattern of associations in this study. Sensitivity analyses using alternative classifications (e.g., treating fermented foods as less healthy plant-based foods) were not performed in this study and should be considered in future studies. Fourth, selection and analytical biases are possible. Participants with missing covariate data were excluded (listwise deletion), and no imputation was performed. However, as missing data were minimal (<5%), substantial bias is unlikely. Although we adjusted for a wide range of covariates, residual confounding by unmeasured factors cannot be excluded. Finally, subgroup analyses may have been underpowered due to smaller sample sizes within strata, and nonsignificant findings should be interpreted with caution. While the findings are generalizable to South Korean adults aged 19 to 64 years, it may not extend to other age groups or populations with different dietary patterns. Despite these limitations, this study uses nationally representative data and provides, to our knowledge, some of the first evidence in South Korea linking the quality of plant-based dietary patterns with depressive symptoms. By focusing on overall dietary quality rather than individual nutrients, these findings offer a more comprehensive and practice perspective on the potential role of plant-based diets in mental health among South Korean adults. Conclusion In conclusion, higher overall PDI and hPDI, along with lower uPDI, were associated with lower depressive symptom severity as measured by PHQ-9 scores. Although these associations remained directionally consistent, they were attenuated after full adjustment. Overall, these findings highlight the importance of plant-based diet quality—emphasizing healthy rather than less healthy plant-based foods—rather than simply increasing total plant-food intake among South Korean adults. Footnotes Authors’ contributions Conceptualization: all authors. Data curation: SP. Formal analysis: SP. Supervision: YJP. Visualization: SP. Funding acquisition: YJP. Writing–original draft: SP. Writing–review & editing: all authors. All authors read and approved the final manuscript. Conflicts of interest None. Funding This work was supported by the National Research Foundation of Korea (RS-2025-00573031). Data availability The datasets analyzed in this study are available in the Korea National Health and Nutrition Examination Survey (KNHANES) repository, https://knhanes.kdca.go.kr/knhanes/main.do Supplementary materials Supplementary materials are available from https://doi.org/10.7762/cnr.2026.0007. Supplementary Material S1. Food groups included in the plant-based diet indices REFERENCES Associated Data Supplementary Materials Supplementary Material S1. Food groups included in the plant-based diet indices