Association between daily dietary intake trajectory and depressive symptom onset and transition among young adults: a longitudinal study

Abstract

Background

Little information is available on long-term dietary monitoring among young adults. We examined the associations between 28-month dietary intake trajectories and depressive symptom onset and transition.

Methods

Using complete daily dining records from Shanghai’s Intelligent Ordering System (IOS) (September 2021–December 2023; n = 6,447), we prospectively assessed dietary exposures prior to measuring depressive symptoms through two Beck Depression Inventory-II (BDI-II) administrations at 24-month intervals (2022 and 2024) in young adults aged 18–40 years. Group-based multi-trajectory models (GBMTs) identified monthly dietary trajectories (classified as Chronic Recommended, Chronic High/Low, or Fluctuating) for 28 nutrients and 15 food components. Multivariable logistic regression assessed associations with depressive symptom onset/transition.

Results

Among 6447 young adults, depressive symptom incidence was 9.7%, whereas depressive symptom improvement and progression incidences were 59.1% and 12.52%, respectively. Certain nutrient and food component trajectories were associated with lower depressive symptom onset risk, including chronic higher intake of carbohydrates, protein, sodium, oil, and sauce, and chronic lower intake of saturated fatty acids (SFAs). In contrast, non-recommended trajectories of zinc, refined grains, and light-colored vegetables were linked to higher onset risk. Notably, some trajectories (e.g., high fat, potassium) showed dual associations: they correlated with higher chances of depressive symptom improvement but also elevated progression risk.

Conclusions

Certain nutrients and dietary patterns showed protective effects against depressive symptom onset, while non-recommended trajectories increased the risk. Some patterns improved symptoms but increased progression risk. The nutritional medicine approach may be beneficial for preventing and promoting depression in young adults.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12916-025-04401-7.

Background

Depression is a major global mental health problem and the driver of mental health-related disability worldwide, with the burden of disease continuing to rise worldwide during the coming decades [1–3]. The prevalence of depression in China increased significantly from 1990 to 2017. In 2017, there were approximately 56 million individuals with depression in China, accounting for 21.3% of the global cases [4]. The one-year prevalence of major depressive disorder is approximately 6%, whereas the lifetime risk of depression is three times higher, ranging from 15% ~ 18% [5, 6]. However, young adults, such as university students, have a much higher prevalence of major depressive disorder (~ 30%) [7, 8], and almost 40% of these individuals experience their first episode of depression before the age of 20, with an average age of onset in the mid-20 s [6]. Among young adults, depression is a pernicious disruptor of developmental processes, conferring increased risk of academic decline, social disengagement, and substance abuse and the potential for increased suicidality. From 1990 to 2019, depression increased in rank among the leading causes of adolescent mortality from eighth to fourth, with its accompanying disability-adjusted life years (DALYs) increasing from 2.8% to 3.7% [9].

Although the determinants of mental disorders, including depression, are multifaceted, emerging and compelling research highlights nutrition as a crucial factor in the high prevalence and incidence of mental disorders, indicating that the importance of diet in psychiatric practice is on par with its importance in cardiology, endocrinology, and gastroenterology [2]. Several systematic reviews of randomized controlled trials and cohort studies have confirmed that dietary intervention has an impact on depression and anxiety, while healthy dietary indices, such as adherence to the Mediterranean diet and a low Dietary Inflammatory Index, are associated with a reduced risk of depressive outcomes [10, 11], and these results suggest that diet is a key modifiable intervention target for preventing the onset of depression. The literature indicates that dietary patterns and the intake of specific nutrients are correlated with the onset of depressive disorders [12, 13]. Compared with a “western” diet of processed or fried foods, refined grains, sugary products, and beer, a “traditional” dietary pattern characterized by vegetables, fruit, meat, fish, and whole grains was associated with lower odds of developing depression [14]. Some cohort studies have suggested a potential protective role of the Mediterranean dietary pattern (MDP), characterized by high consumption of vegetables, fruits, legumes, nuts, whole grains, fish, and olive oil; moderate wine intake; and a limited intake of red meat and dairy products, in the prevention of depressive disorders [15, 16]. Deficiencies in nutrients such as protein, B vitamins, vitamin D, magnesium, zinc, selenium, iron, calcium, and omega-3 fatty acids have significant negative effects on depressive symptoms, and these nutrients could offer a range of neurochemical regulatory functions in the improvement of depressive symptoms [12, 17]. Previous studies have examined the associations between dietary behaviour and depression through dietary patterns and average intake levels. However, individual dietary behaviour is subject to dynamic changes. Therefore, investigating the long-term characteristics of dietary behaviour and its association with the onset of and change in depression status is important in addressing this gap.

The traditional food consumption collection methods for young adults, including 3-day and 24-h diet recalls, food frequency questionnaire surveys, and weighed food records, can only be used to assess cross-sectional dietary intake, fail to capture dietary intake changes over time, and are subject to considerable recall bias [18]. Therefore, in the present study, we used digital daily food consumption records to evaluate dietary patterns and nutrient intake trajectories among young adults and to explore the associations between food consumption trajectories and incidence and changes in depressive symptoms.

Methods

Study design and participants

In this longitudinal study, participants were recruited through a multi-stage process: 1) Initial screening of 103,412 undergraduate students from all academic years and disciplines based on Intelligent Ordering System (IOS), which comprehensively records dining categories, consumption quantities, and meal timestamps for each student’s on-campus dining activities; 2) To guarantee the representativeness of the dietary evaluation for each student, we selected 43,290 participants who met all the inclusion and exclusion criteria. The inclusion criteria required participants have dietary intake records in the school cafeterias on more than 86 days (excluding the 42-day winter vacation and 65-day summer vacation, students had 258 school days. Participants who ate at school for more than one-third of these days (86 days) were included). Additionally, the participants were excluded if they dined in the school cafeteria with fewer than 1 meal of breakfast, lunch, or dinner during a month [18]. The exclusion criteria also included abnormal daily caloric intake (males: < 800 kcal/day or > 4000 kcal/day; females: < 500 kcal/day or > 3500 kcal/day), desire to gain weight, present use of a therapeutic diet or weight loss diet, and history of eating disorders. 3) In terms of mental health screening data acquisition, from 24,124 students who participated in mental health l screening examinations in 2022, we excluded those who had graduated, held non-Chinese nationality, and were unwilling to participate in follow-up surveys. Ultimately, 8,056 eligible participants were included for mental health follow-up investigations, and access authorization to their cafeteria dining data was obtained by formal invitation through campus email system with three follow-up reminders. 4) From the initial 8,056 students matched with their dining data, 6,447 participants possessing both psychological survey records and dining data were ultimately included as study participants through the data linkage process. Participants without monetary reward but received personalized nutrition assessment reports according to their dining records in the school cafeterias which summarizing their dietary patterns and nutrient intake as scientific feedback. The flow chart of the data collection and screening process is shown in Fig. 1 and the questionnaire is shown in Additional file 5. Although we collected long-term, accurate daily records of each meal of students from IOS, considering the possibility of students eating off campus or consuming take out, we conducted a 7-day food diary (7DFD) survey to evaluate on-campus diet constructs and patterns using the IOS to determine whether those data represented whole food consumption. Among 221 samples (this sample was independent of the main study and was used solely for validating the accuracy and representativeness of the IOS), for all nutrients and food components, the correlation coefficients between the IOS and 7DFD records ranged from 0.59 to 0.88 (all P value < 0.05). In the quartile classification of nutrient and food components intake, 70% to 97% of the results from the IOS and 7DFD were categorized into the same quartile. Bland‒Altman plots further indicated that 95% of the differences between the two methods for nutrients and food groups fell within ± 1 standard deviation of the mean difference. In conclusion, compared with the traditional 7DFD, the IOS used in present study can effectively represent the dietary intake of participants [19]. Our study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Medical Research, School of Public Health, Fudan University (protocol code: IRB#2019-01-0726S; data: 7 January 2019).

Fig. 1.

The flow chart of the data collection and screening process

Assessment of depressive symptoms

Depressive symptoms were assessed using the Beck Depression Inventory (BDI-II) at baseline (2022) with one follow-up assessment conducted in 2024 [20, 21]. All enrolled participants (n = 8,806) completed the baseline assessment, and 92.0% (n= 8,100) completed the follow-up survey. The BDI-II is a 21-item self-report scale that assesses the severity of depressive symptoms in the past 2 weeks and includes somatic-affective (13 items) and cognitive (8 items) dimensions [22]. Each item is rated on a 4-point Likert scale ranging from 0 to 3, and total scores range from 0–63, with higher scores indicating more severe symptomatology. For the present study, depression was dichotomized into no depressive symptoms (0–9) or prevalent depressive symptoms. The higher the total score is, the more severe the subject’s depression (a total score of 0–14 is considered nondepressed, 15–27 is considered mild depression, and 28–63 is considered moderate–severe depression) for Chinese young adults [23]. The Chinese version showed better construct and concurrent validity and internal consistency among young adults [24, 25]. In the present study, we defined depressive symptom onset as the absence of depressive symptoms at baseline and onset at the follow-up. We also categorized depressive symptom transition as improvement and progression, where improvement refers to the status of depressive symptoms from being to not being or from severe to moderate or mild, and progression refers to the status of depressive symptoms from mild to moderate or severe as well as from moderate to severe.

Assessment of food composition, nutrient intake and diet quality

We assessed daily food composition and nutrient intake on the basis of daily food purchase records from the IOS from September 2020 to November 2023, while January to July 2020 represented the COVID-19 outbreak in Shanghai. Each food item ordered by participants through the IOS system was counted as one individual dietary record. The IOS contains 5714 dishes, and we first established a food database by weighing the materials of 1840 representative dishes before and after cooking. For the remaining 3874 dishes, we utilized a standardized recipe approach and calculated their weights on the basis of the composition of the representative dishes. The database includes information on food composition, such as the weight of raw food materials and condiments, as well as the nutrient composition of each meal. Details of the weighing and calculation methods have been described in a previous article [18]. Nutrient composition was determined via the Chinese Food Composition database (version 2023) [26]. Ultimately, we assessed the composition of 15 foods, including whole grains, refined grains, potato, light-coloured vegetables, dark-coloured vegetables, poultry meat, red meat, fish and shrimps, egg, bean, mushrooms, oil, salt, sugar, and sauce, and 28 nutrients, including energy, protein, fat, carbohydrate, fibre, cholesterol, vitamin A, carotene, retinol, thiamine (vitamin B1), riboflavin (vitamin B2), nicotinic acid (vitamin B3), vitamin C, vitamin E, calcium (Ca), phosphorus (P), potassium (K), sodium (Na), magnesium (Mg), iron (Fe), zinc (Zn), selenium (Se), copper (Cu), manganese (Mn), total fatty acid (TFA), saturated fatty acid (SFA), monounsaturated fatty acid (MUFA), and polyunsaturated fatty acid (PFA) contents. Given the lack of consistency in on-campus dining patterns (i.e., not every meal was consumed in the cafeteria daily), the daily calorie intake was derived as follows. First, the ​average calorie intake per breakfast, ​lunch, and ​dinner consumed within the campus cafeteria was ​calculated separately for each month. Second, these monthly averages for breakfast, lunch, and dinner were then ​summed. Third, this ​total monthly sum represents the estimated ​average daily calorie intake for that month. Nutrients and food groups intake over 28-month as shown in Additional file 1: Fig. S1. The recommended intakes for all nutrients and food groups as per the Chinese Dietary Guidelines are presented in Additional file 2: Table S2.1.

We used the Diet Balance Index-2022 (DBI-22), which was designed by the Chinese Central for Disease Control and Prevention for Chinese Residents [27], to assess comprehensive diet quality. The DBI-22 consists of 12 subgroups, each containing eight components that include cereals, vegetables and fruits, dairy products and soy products, various meats (both red meat and processed, poultry and game), seafood (fish and shrimp), eggs, cooking oils, alcoholic beverages, condiments (such as sugar and salt), dietary varieties, and drinking water. To assess diet quality via the DBI-22, three indicators are employed: the high bound score (HBS), the low bound score (LBS), and the diet quality distance (DQD). The HBS is used to evaluate the extent of excessive dietary intake, the LBS is used to determine the level of insufficient dietary intake, and the DQD is used to assess dietary imbalance. Each of these indicators is categorized into four levels: (1) minimal dietary issues; (2) low level of concern; (3) moderate concern; and (4) high concern. The evaluation criteria for excessive and insufficient intake are based on the Dietary Reference Intakes (DRIs) for The Chinese Dietary Guidelines. Intake exceeding the recommended levels for a specific age group is classified as excessive dietary intake, while intake below the recommended levels is termed insufficient dietary intake. The sum of the absolute values of excessive and insufficient intake reflects the degree of dietary imbalance. To assess the effect of dietary trajectory (exposure) on depressive symptom status (outcome), we established the exposure period one year prior to the measurement of depressive symptoms to ensure a clear temporal sequence between them.

Assessment of covariates

We assessed covariates in 2021. The following measures were considered pertinent covariates because they are associated with depression and are independent risk factors for depression among young adults based on previous studies [28–30]: age; sex; major; educational level; location; poverty-stricken college student; living-school experience; only child; education level of father and mother; relationship between father and mother; relationship with family members; and relationship with friends. The definition and category of all covariates were performed in Additional file 3: Table S3.1.

Statistical analysis

The participants’ characteristics are presented as frequencies with percentages (Table 1). The participants’ volume (unit varies by specific food group or nutrient) of dietary intake are presented as means with standard deviations (Table 2). A chi-square test was performed to explore the distribution of depressive symptoms at baseline according to demographic characteristics (Table 1), as well as the onset and transition of depressive symptoms (Additional file 4: Table S4.1 and Additional file 5: Table S5.1). We used Student’s t test to compare the intake volume of nutrients and food components between nondepressive and depressive symptoms over 3 years (Table 2). Dietary intake trajectories are defined as distinct patterns of change in nutrient and food component consumption over time. We applied group-based multi-trajectory models (GBMTs) to identify food composition and nutrient intake groups with similar trajectories [31]. We fitted each model, allowing the formation of 1 to 7 groups adjusted by month. We identified the optimal number of groups on the basis of the following criteria: Akaike information criterion; Bayesian information criterion; entropy greater than 0·7; group size greater than 5% of the total sample for the smallest group; odds of correct classification greater than 5 for all groups; and model interpretability. Following selection, if quadratic or linear terms had large SEs and nonsignificant p values, we tested a lower-order model [32]. Trajectories were named based on two key dimensions: (1) absolute intake level relative to recommendations according to Chinese Food Composition database (e.g., recommended, slightly/moderately/vigorously low/high, lowest/highest) and (2) temporal patterns of intake over time (e.g., chronic, increasing/decreasing, rapidly increasing, or combinations like “initially decreasing then increasing”). We also used a chi-square test to screen dietary intake trajectories significantly associated with the onset and transition of depressive symptoms, as shown in Additional file 4: Tables S4.2 & S4.4, Additional file 5: S5.2 & S5.4. The significant trajectories are shown in Additional file 6: Fig. S6.1 & S6.2. Ultimately, we adopted logistic regression to examine multivariate associations of nutrient intake, food group intake and diet quality trajectories with the onset and transition of depressive symptoms, as shown in Figs. 2, 3 and Additional file 4: Tables S4.3.1 & S4.5.1, Additional file 5: S5.3.1 & S5.5.1. Each diet intake trajectory examined by three models: (1) crude model, only include diet intake trajectory; (2) first adjusted model, include significant covariates based on crude model; (3) final adjusted model, include total energy intake based on first adjusted model. Furthermore, we performed subgroup analyses stratified by educational attainment levels as shown in Additional file 4: Table S4.3.2 & S4.5.2 and Additional file 5: S5.3.2 & S5.5.2. The estimates of nutrient intake and food components associated with depression were summarized via odds ratios (ORs) and 95% confidence intervals (CIs). Statistical significance was determined with false discovery rate (FDR) -adjusted p-values using the Benjamini–Hochberg procedure at α = 0.05 to account for multiple testing. Statistical analyses were performed via R software (version 4.3.1).

Table 1.

Demographic among all overall participants

	Overall (N = 6447)	Non-depressive (N = 5760)	Depressive (N = 687)	P-value
Age				0.003
~ 20	3,267 (50.7)	2,880 (88.2)	387 (11.8)
21 ~ 24	2,016 (31.3)	1,814 (90.0)	202 (10.0)
25 ~	1,164 (18.1)	1,066 (91.6)	98 (8.4)
Sex				0.812
Male	3,407 (52.8)	3,041 (52.8)	366 (53.3)
Female	3,040 (47.2)	2,719 (47.2)	321 (46.7)
Major				0.852
Natural science	2,883 (44.7)	2,573 (44.7)	310 (45.1)
Social science	1,491 (23.1)	1,338 (23.2)	153 (22.3)
Biomedicine	2,073 (32.2)	1,849 (32.1)	224 (32.6)
Educational level				< 0.001
Bachelor	3,404 (52.8)	2,989 (51.9)	415 (60.4)
Master	1,120 (17.4)	1,013 (17.6)	107 (15.6)
Doctor	1,923 (29.8)	1,758 (30.5)	165 (24.0)
Poverty-stricken college student				0.002
No	5,588 (86.7)	5,019 (87.1)	569 (82.8)
Yes	859 (13.3)	741 (12.9)	118 (17.2)
Hukoua				0.517
Urban	4,520 (70.1)	4,031 (70.0)	489 (71.2)
Rural	1,927 (29.9)	1,729 (30.0)	198 (28.8)
Living-school experience				0.004
No	2,526 (39.2)	2,292 (39.8)	234 (34.1)
Yes	3,921 (60.8)	3,468 (60.2)	453 (65.9)
Only child				0.002
No	2,524 (39.1)	2,217 (38.5)	307 (44.7)
Yes	3,923 (60.9)	3,543 (61.5)	380 (55.3)
Education level of father				0.553
Secondary school and below	2,214 (34.3)	1,982 (34.4)	232 (33.8)
Associate Degree	2,292 (35.6)	2,056 (35.7)	236 (34.4)
Bachelor and above	1,941 (30.1)	1,722 (29.9)	219 (31.9)
Education level of mother				0.520
Secondary school and below	2,517 (39.0)	2,235 (38.8)	282 (41.0)
Associate Degree	2,279 (35.3)	2,045 (35.5)	234 (34.1)
Bachelor and above	1,651 (25.6)	1,480 (25.7)	171 (24.9)
Relationship between father and mother				< 0.001
Good	5,904 (91.6)	5,376 (93.3)	528 (76.9)
Poor	543 (8.4)	384 (6.7)	159 (23.1)
Relationship with family members				< 0.001
Good	6,286 (97.5)	5,666 (98.4)	620 (90.2)
Poor	161 (2.5)	94 (1.6)	67 (9.8)
Relationship with friends				< 0.001
Good	6,150 (95.4)	5,582 (96.9)	568 (82.7)
Poor	297 (4.6)	178 (3.1)	119 (17.3)

^aHukou refers to China’s household registration system, which classifies individuals based on their registered place of residence. It is commonly used in social and demographic analyses in China

Table 2.

The average daily nutrients and food components according to depressive symptom at baseline in 2021–2023 (3 months in 2021; full years in 2022 and 2023)

Nutrients and food components	2021.09–2021.12			2022.01–12			2023.01–12
	Non-depressive	Depressive	FDR-adjusted-P	Non-depressive	Depressive	FDR-adjusted-P	Non-depressive	Depressive	FDR-adjusted-P
Energy (kcal)	2037.01 (775.46)	2057.75 (832.72)	0.442	2057.44 (809.69)	2068.77 (831.41)	0.727	2078.31 (854.46)	2078.32 (870.36)	0.999
Protein (g)	80.54 (31.26)	80.97 (33.44)	0.611	82.94 (33.14)	82.84 (33.81)	0.902	85.60 (35.34)	85.00 (35.85)	0.416
Fat (g)	81.10 (35.63)	80.91 (37.51)	0.785	84.04 (38.35)	83.67 (37.90)	0.770	84.53 (39.78)	83.88 (39.27)	0.416
Carbohydrates (g)	251.93 (105.51)	257.12 (113.94)	0.069	248.06 (109.96)	251.85 (116.02)	0.158	249.87 (113.61)	251.90 (118.78)	0.416
Total Fatty Acids (TFA, g)	69.49 (31.74)	69.12 (33.42)	0.611	72.02 (34.12)	71.56 (33.40)	0.727	72.82 (35.26)	72.30 (34.45)	0.426
Saturated fatty acid (SFA, g)	20.10 (9.41)	19.98 (9.92)	0.611	20.85 (10.31)	20.77 (10.15)	0.780	21.06 (10.52)	20.89 (10.25)	0.418
Monounsaturated fatty acids (MUFA, g)	26.27 (12.94)	26.11 (13.71)	0.611	27.33 (14.19)	27.16 (13.95)	0.727	27.35 (14.33)	27.13 (13.89)	0.426
Polyunsaturated fatty acids (PUFA, g)	21.97 (9.95)	21.91 (10.35)	0.752	22.67 (10.45)	22.46 (10.21)	0.517	23.22 (11.05)	23.09 (10.98)	0.512
Fiber (g)	5.87 (2.72)	5.91 (2.83)	0.611	5.91 (2.95)	5.94 (3.06)	0.780	6.15 (3.12)	6.09 (3.24)	0.416
Cholesterol (mg)	638.90 (375.35)	630.60 (375.85)	0.486	698.51 (429.31)	694.49 (427.32)	0.770	736.12 (459.68)	730.36 (463.15)	0.512
Vitamin A ($\mu$g)	400.51 (229.72)	394.57 (229.13)	0.442	421.36 (246.48)	418.48 (241.92)	0.727	474.09 (329.97)	461.74 (289.76)	0.023
Carotene ($\mu$g)	1604.95 (1083.28)	1111.91 (212.36)	0.811	1657.71 (1219.76)	1655.54 (1226.54)	0.920	1915.41 (1456.48)	1883.12 (1426.09)	0.307
Retinol ($\mu$g)	216.30 (164.52)	212.36 (162.21)	0.442	233.21 (171.24)	230.96 (166.38)	0.727	260.28 (261.01)	252.12 (213.34)	0.061
Vitamin B1 (mg)	0.92 (0.39)	0.93 (0.41)	0.611	0.93 (0.40)	0.93 (0.41)	0.780	0.93 (0.41)	0.93 (0.43)	0.937
Vitamin B2 (mg)	0.71 (0.28)	0.71 (0.29)	0.607	0.74 (0.31)	0.73 (0.31)	0.780	0.78 (0.34)	0.77 (0.34)	0.212
Vitamin B3 (mg)	17.29 (7.80)	17.41 (8.34)	0.611	18.06 (8.50)	17.95 (8.68)	0.727	18.71 (8.85)	18.54 (9.08)	0.396
Vitamin C (mg)	71.38 (43.02)	71.85 (44.51)	0.611	77.97 (49.30)	78.63 (49.66)	0.727	84.58 (52.83)	84.43 (53.23)	0.883
Vitamin E (mg)	31.09 (13.99)	31.24 (14.54)	0.611	32.02 (14.60)	31.71 (14.38)	0.483	33.07 (15.64)	32.93 (15.74)	0.612
Calcium (Ca, mg)	319.13 (150.31)	315.58 (151.89)	0.442	320.58 (155.57)	321.39 (157.76)	0.825	346.91 (175.86)	343.68 (170.90)	0.400
Phosphorus (P, mg)	922.24 (358.43)	928.69 (380.49)	0.607	934.11 (370.68)	935.16 (379.05)	0.902	963.29 (397.15)	959.84 (401.68)	0.628
Potassium (K, mg)	1590.60 (627.31)	1600.65 (660.57)	0.611	1623.93 (661.79)	1627.99 (665.91)	0.803	1716.83 (726.95)	1710.75 (733.53)	0.631
Sodium (Na, mg)	4095.82 (1727.55)	4067.78 (1795.79)	0.611	4069.77 (1786.35)	4075.34 (1805.55)	0.902	4443.34 (2031.30)	4413.31 (2022.53)	0.426
Magnesium (Mg, mg)	253.14 (98.78)	255.20 (105.10)	0.598	253.42 (101.24)	253.94 (103.00)	0.825	265.77 (111.46)	264.50 (111.85)	0.541
Iron (Fe, mg)	15.66 (6.91)	15.64 (7.25)	0.861	15.52 (6.86)	15.45 (6.82)	0.770	16.38 (7.21)	16.18 (7.21)	0.207
Zinc (Zn, mg)	9.05 (3.80)	9.11 (4.06)	0.611	9.12 (3.91)	9.11 (3.98)	0.902	9.52 (4.20)	9.45 (4.62)	0.416
Selenium (Se,$\mu$g)	39.44 (16.41)	39.36 (17.15)	0.791	41.39 (17.87)	41.21 (18.14)	0.770	43.22 (19.33)	42.81 (19.55)	0.342
Copper (Cu, mg)	1.53 (0.64)	1.54 (0.67)	0.611	1.58 (0.71)	1.57 (0.71)	0.780	1.62 (0.81)	1.61 (0.79)	0.512
Manganese (Mn, mg)	3.17 (1.43)	3.21 (1.50)	0.442	3.19 (1.51)	3.25 (1.63)	0.158	3.21 (1.54)	3.25 (1.66)	0.340
Whole grains (g)	95.92 (66.22)	99.01 (73.64)	0.069	96.17 (76.64)	98.48 (80.49)	0.230	102.80 (85.27)	103.81 (90.91)	0.541
Refined grain (g)	143.65 (92.28)	148.71 (100.21)	< 0.001	149.11 (105.00)	157.90 (122.46)	< 0.001	147.06 (104.38)	123.40 (126.33)	< 0.001
Potato (g)	40.13 (41.74)	40.62 (40.63)	0.611	39.99 (47.04)	40.31 (46.53)	0.780	45.01 (49.17)	45.27 (50.84)	0.814
Light vegetable (g)	89.95 (63.04)	90.68 (65.03)	0.611	95.38 (74.50)	98.01 (82.23)	0.158	106.06 (84.01)	109.12 (95.96)	0.138
Dark vegetable (g)	94.80 (62.75)	96.71 (66.06)	0.368	100.84 (76.46)	104.26 (81.74)	0.046	123.38 (91.12)	126.33 (96.43)	0.146
Poultry meat (g)	99.12 (82.72)	101.75 (88.72)	0.368	120.27 (105.01)	121.79 (106.85)	0.727	120.44 (105.30)	123.19 (112.42)	0.234
Livestock meat (g)	138.75 (22.32)	137.94 (85.59)	0.752	143.93 (93.91)	145.52 (97.75)	0.727	151.01 (98.83)	150.65 (101.98)	0.871
Fishes and shrimps (g)	14.75 (22.32)	14.09 (21.01)	0.368	14.43 (25.25)	14.75 (27.53)	0.727	19.46 (30.31)	19.50 (31.12)	0.937
Egg (g)	72.49 (53.63)	72.14 (70.72)	0.752	84.06 (70.26)	85.57 (70.72)	0.483	91.70 (77.64)	92.84 (78.58)	0.430
Bean (g)	18.21 (16.40)	18.39 (17.35)	0.611	17.63 (19.41)	17.95 (20.39)	0.727	17.76 (19.74)	17.96 (20.44)	0.587
Mushrooms (g)	13.67 (34.23)	19.10 (34.47)	0.607	21.33 (38.56)	22.31 (43.74)	0.483	23.62 (42.98)	24.31 (47.21)	0.426
Oil (g)	36.97 (18.11)	37.31 (19.43)	0.607	39.34 (21.81)	39.95 (23.23)	0.286	41.80 (23.99)	42.58 (25.89)	0.146
Salt (g)	5.06 (2.59)	5.11 (2.76)	0.607	5.30 (3.10)	5.41 (3.25)	0.158	6.05 (3.57)	6.15 (3.76)	0.212
Sugar (g)	0.78 (1.90)	0.75 (1.82)	0.607	0.84 (2.14)	0.87 (2.29)	0.727	0.65 (1.80)	0.68 (2.02)	0.400
Sauce (g)	5.68 (4.16)	5.87 (4.57)	0.069	5.84 (4.82)	6.15 (5.19)	< 0.001	6.11 (5.02)	6.26 (5.22)	0.151
HBS	9.59 (2.31)	9.53 (2.18)	0.611	10.20 (2.06)	10.13 (2.02)	0.727	10.55 (2.11)	10.62 (2.20)	0.576
LBS	25.12 (2.57)	25.28 (2.40)	0.442	26.40 (2.73)	26.40 (2.63)	0.995	25.84 (2.88)	25.81 (2.76)	0.871
DQD	34.71 (3.40)	34.81 (3.26)	0.611	36.60 (3.16)	36.53 (3.17)	0.780	36.40 (3.37)	36.44 (3.42)	0.871

HBS High Bound Score, LBS Low Bound Score, and DQD Diet Quality Distance represent the three core dimensions of the Dietary Balance Index-22 (DBI-22), where HBS quantifies excessive dietary intake, LBS characterizes insufficient dietary intake, and DQD assesses dietary imbalance

Fig. 2.

The forest plots of associations between nutrients and food groups intake trajectories and depressive symptom onset

Fig. 3.

The forest plots of associations between nutrients and food groups intake trajectories and depressive symptom transition

Results

Baseline characteristics

As shown in Table 1, we included 6447 young adults, of whom 3040 (47.2%) were female; 3407 (52.8%) were male; 3267 (50.7%) were aged between 16 and 20 years; 2016 (31.3%) were aged between 21 and 24 years; 2883 (44.7%) were majoring in science, technology, engineering, or mathematics; 3404 (52.8%) were bachelor’s degree students; 1923 (17.4%) were doctoral students; 859 (13.3%) were poverty-stricken college student; 4520 (70.1%) originated from urban areas; 3921 (60.8%) had residential experience; 3923 (60.9%) were an only child; 1941 (30.1%) had a fathers’ education attainment of bachelor’s degree and above; 1651 (25.6%) had a mothers’ education attainment of bachelor’s degree and above; 543 (8.4%) had poor relationships between their parents; 161 (2.5%) had poor relationships with their family; and 297 (4.6%) had poor relationships with their friends.

Depressive symptom status at baseline and follow-up

The median depression score of young adults at the baseline survey was 3.0 (IQR 0–8.0), and the number of participants with depressive symptoms was 687 (406 with mild depressive symptoms and 281 with moderate and severe depressive symptoms). Among 5760 participants without depressive symptoms at baseline, 559 had depressive symptoms at follow-up, with an incidence rate of 9.7% (95%CI = 8.94%−10.47%). Among the 687 participants with depressive symptoms at baseline, 406 had improved depressive symptom status at follow-up, with an incidence rate of 59.1% (95%CI = 55.42%−62.77%), whereas 86 experienced depressive symptom progression by follow-up, with an incidence rate of 12.52% (95%CI = 10.04%−14.99%).

Differences in dietary intake by depressive symptom status

The volume of nutrient intake, food components and DBI score from September 2021 to December 2023 stratified by the presence of depressive symptoms at baseline are shown in Table 2. There were statistically significant between-group differences in the intake volumes of carbohydrates, Vit A, retinol, Vit B2, Fe, Mn, grains, light-coloured vegetables, dark vegetables, fish and shrimp, whole grains, oils, salts and sauces from September 2021 to December 2023 (all P values < 0.05). Notably, the differences were more pronounced for carbohydrates, vitamin A, retinol, grains, light-coloured vegetables, dark vegetables, and whole grains (absolute differences > 1; P values < 0.05).

Association of nutrients intake trajectory and food components intake trajectory with onset of depressive symptom

As shown in Additional file 3: Table S3.1, the prevalence of the onset of depressive symptoms differed across professions, household registrations, status of relationship with parents, and status of relationships with family and friends (P < 0.05). The results of the nutrient intake trajectory analysis by the GBMT and univariate analysis with depressive onset are shown in Additional file 3: Table S3.2. The chi-square test suggested that among the 28 types of nutrients, the intake trajectories of carbohydrates, proteins, SFAs, MUFAs, VitB3, Na, Zn, and Se were possibly associated with the onset of depressive symptom, with P values all < 0.10. The multivariate associations between nutrient intake trajectory and the onset of depressive symptoms are shown in Fig. 2 and Additional file 4: Table S4.3.1. According to the fully adjusted logistic regression model with energy, participants with chronic mildly high carbohydrate intake (7.68%) and chronic highly carbohydrate intake (7.25%) had a lower risk of depressive symptom onset than those who consumed the recommended amount of carbohydrates, with a prevalence of 10.8% (OR = 0.72, P < 0.001; OR = 0.69, P < 0.001).

Compared with participants who chronically consumed the recommended amount of protein, participants with chronic mildly high protein consumption, chronic moderately high protein consumption and chronic highly increasing protein consumption had a lower risk of depressive symptom onset (8.47%, OR = 0.75, P < 0.001; 8.76%, OR = 0.79, P < 0.001; 8.12%, OR = 0.68, P < 0.001). Compared with chronic rapidly increasing most high SFAs intake (11.8%), participants with chronic moderately increasing lower SFAs intake, initially increasing then decreasing lower SFAs intake, chronic lower SFAs intake and chronic lowest SFAs intake had lower had a lower risk of depressive symptom onset (8.0%, OR = 0.59, P < 0.001; 8.54%, OR = 0.68, P < 0.001; 11.0%, OR = 0.89, P = 0.024; 10.1%, OR = 0.78, P < 0.001). Participants with low MUFAs intake (10.3%) had a higher risk of depressive symptom onset than those with increasing high MUFAs intake (8.62%) (OR = 1.26, P < 0.001). Compared with chronic recommended VitB3 intake (10.6%), participants who are chronic slightly low VitB3 intake, chronic moderately low VitB3 intake, chronic moderately high increasing VitB3 intake suggested lower risk of depressive symptom onset (8.63%, OR = 0.80, P < 0.001; 9.29%, OR = 0.85, P < 0.001; 7.43%, OR = 0.65, P < 0.001), while chronic slightly low increasing VitB3 intake suggested higher risk of depressive symptom onset (11.6%, OR = 1.09, P = 0.008). Compared with chronic increasing Na intake (10.4%), chronic mildly high increasing Na intake, chronic moderately high Na intake and chronic highest increasing Na intake were associated with a lower risk of depressive symptom onset (9.97%, OR = 0.91, P = 0.024;6.54%, OR = 0.59, P < 0.001; 8.08%, OR = 0.75, P < 0.001). Compared with chronic recommended Zn intake (7.43%), chronic mildly lower but increasing Zn intake, chronic moderately lower Zn intake and the lowest but increasing Zn intake (11.0%, OR = 1.40, P < 0.001; 9.48%, OR = 1.18, P < 0.001; 10.9%, OR = 1.50, P < 0.001) were associated with higher risk of depressive symptom onset. Compared with chronic recommended Se intake (9.07%), chronic moderately lower increasing Se intake was associated with a lower risk of depressive symptom onset (7.13%, OR = 0.68, P < 0.001), while chronic vigorously lower increasing Se intake was associated with a higher risk of depressive symptom onset (11.2%, OR = 1.11, P = 0.018).

The multivariate associations between food component intake trajectories and depressive symptom onset are shown in Additional File 4: Table S4.5.1. Compared with chronic recommended refined grained consumption (6.96%), chronic mildly lower increasing refined grain, chronic moderately lower increasing refined grain consumption and chronic lowest increasing refined grain consumption were associated with higher depressive symptom onset risk (9.2%, OR = 1.37, P < 0.001; 11.0%, OR = 1.73; 12.0%, OR = 1.81), meanwhile chronic mildly high refined grain intake and chronic moderately high increasing refined grain intake also were associated with higher depressive symptom onset risk (9.17%, OR = 1.34; 10.3%, OR = 1.38). Compared with increasing chronic recommended light-coloured vegetable consumption, chronic mildly lower (9.20% vs 6.96%, OR = 1.44, P < 0.001), lowest increasing (11.0% vs 6.96%, OR = 1.37, P < 0.001) and chronic highest rapidly increasing (12.0% vs 6.96%, OR = 1.28, P < 0.001) light-colored vegetable intake were associated with a higher risk of depressive symptom onset. Compared with chronic recommended poultry (9.9%), chronic slightly higher poultry meat intake and chronic mildly higher increasing poultry meat intake were associated with a higher risk of depressive symptom onset (10.7%, OR = 1.07, P = 0.025; 11.3%, OR = 1.10, P = 0.009), while chronic mildly higher poultry meat intake and chronic moderately higher increasing light vegetable intake were associated a lower risk of depressive symptom onset (7.05%, OR = 0.72, P < 0.001; 9.26%, OR = 0.89, P = 0.006). Compared with chronic recommended oil consumption (10.2%), chronic moderately higher increasing oil consumption, chronic moderately higher increasing oil intake and chronic highest and rapidly oil increasing intake were associated with a lower risk of depressive symptom onset (7.48%, OR = 0.94, P = 0.020; 7.48%, OR = 0.62; 8.24%, OR = 0.63). Compared with chronic recommended sauce consumption (11.5%), chronic slightly higher sauce intake, chronic mildly higher sauce intake, chronic moderately higher increasing sauce intake, chronic vigorously higher and rapidly sauce increasing intake and highest rapidly increasing sauce consumption were associated with a lower risk of depressive symptom onset (8.87%, OR = 0.72, P < 0.001; 11.0%, OR = 0.90, P < 0.001; 8.77%, OR = 0.71, P < 0.001; 8.05%, OR = 0.60, P < 0.001; 9.02%, OR = 0.68, P < 0.001). For diet quality, a chronic slightly higher HBS (6.70%) was associated with a lower risk of depressive symptom onset than a chronic recommended HBS (10.17%) (OR = 0.67, P < 0.001). The majority of associations remained statistically significant following FDR adjustment.

Association of nutrients intake trajectory and food components intake trajectory with transition of depressive symptom

The multivariate associations between nutrient intake trajectories and depressive symptom transitions are shown in Fig. 3 and Additional file 5: Table S5.3.1. Compared with chronic recommendation fat consumption, chronic slightly higher, chronic mildly higher increasing and chronic moderately higher and highest initially decreasing then increasing fat consumption were associated with higher chance of depressive symptom improvement (65.26%, OR = 2.01, P < 0.001; 58.16%, OR = 1.26, P < 0.001; 59.0%, OR = 1.42, P < 0.001; 72.0%, OR = 3.14, P < 0.001) and higher risk of depressive symptom progression (12.68%, OR = 3.57, P < 0.001; 13.48%, OR = 2.88, P < 0.001; 16.0%, OR = 3.8, P < 0.001; 13.33%, OR = 2.21, P < 0.001; 12.0%, OR = 6.2, P < 0.001). Compared with chronic most high and decreasing MUFAs consumption, chronic moderately lower decreasing, chronic vigorously lower increasing, chronic vigorously lower and chronic lowest MUFAs consumption were associated with lower risk of depressive symptom progression (10.64%, OR = 0.96, P = 0.020; 6.38%, OR = 0.32, P < 0.001; 9.93%, OR = 0.61, P < 0.001), however, chronic some lower increasing and chronic vigorous lower MUFAs consumption were associated with lower chances of depressive symptom improvement (54.26%, OR = 0.77, P < 0.001; 51.79%, OR = 0.82, P < 0.001).

Participants who are chronic higher and increasing K consumption suggested a higher improvement chance (63.04% vs 55.06%, OR = 1.32, P < 0.001) and a lower progression risk (8.70% vs 13.29%, OR = 0.73, P < 0.001) of depressive symptom than who are chronic adequately increasing K consumption. Compared with chronic adequately increasing K consumption, chronic slightly low K consumption suggested a higher chance of depressive symptom improvement (56.12% vs 55.06%, OR = 1.12, P < 0.001) meanwhile a higher risk of depressive symptom progression (18.71% vs 13.29%, OR = 1.45, P < 0.001); chronic vigorously low K consumption suggested a lower chance of depressive symptom improvement (54.20% vs 55.06%, OR = 0.81, P < 0.001) meanwhile a lower risk of depressive symptom progression (9.92% vs 13.29%, OR = 0.63, P < 0.001). Compared with chronic recommended Zn consumption chronic low (66.67%), chronic moderately low increasing and chronic lowest decreasing Zn consumption (64.53%, OR = 0.93, P < 0.001; 56.44%, OR = 0.52, P < 0.001) as well as slightly high and initially decreasing then increasing and chronic highest decreasing Zn consumption (54.43%, OR = 0.47, P < 0.001; 58.06%, OR = 0.56, P < 0.001) were associated with lower chances of depressive symptom improvement.

The multivariate associations between food groups intake trajectories and depressive symptom transitions are shown in Fig. 3 and Additional file 5: Table S5.5.1. Compared with chronic recommended egg consumption, chronic slightly low, chronic lowest, chronic moderately high, chronic moderately high increasing, chronic vigorously high increasing, chronic highest and rapidly increasing volume of egg consumption were associated with lower chance of depressive symptom improvement (60.71%, OR = 0.73, P < 0.001; 50.0%, OR = 0.47, P < 0.001; 60.98%, OR = 0.80, P < 0.001; 47.62%, OR = 0.64, P < 0.001; 58.33%, OR = 0.6, P < 0.001; 59.38%, OR = 0.52, P < 0.001), meanwhile chronic slightly low, chronic lowest, chronic moderately high, chronic moderately high increasing, chronic vigorously high increasing, chronic highest and rapidly increasing were also associated with lower risks of depressive symptom progression (11.43%, OR = 0.44, P < 0.001; 5.0%, OR = 0.14, P < 0.001; 12.2%, OR = 0.48, P < 0.001; 13.33%, OR = 0.3, P < 0.001; 9.72%, OR = 0.26, P < 0.001; 6.25%, OR = 0.16, P < 0.001). Compared with chronic recommended bean consumption, chronic moderately high increasing bean consumption was associated with a higher chance of depressive symptom improvement (65.67% vs 60.71%, OR = 1.15, P = 0.001) and a lower risk of depressive symptom progression (8.96% vs 11.31%, OR = 0.89, P < 0.001); chronic lowest bean consumption was associated a lower chance of depressive symptom improvement (51.02% vs 60.71%, OR = 0.56, P < 0.001) but a lower risk of depressive symptom progression (10.88% vs 11.31%, OR = 0.68, P < 0.001). Compared with chronic recommended salt consumption, chronic recommended increasing salt consumption was associated with a higher chance of depressive symptom improvement (69.85% vs 57.14%, OR = 1.52, P < 0.001) and a lower risk of depressive symptom progression (8.82% vs 14.91%, OR = 0.81, P < 0.001); while chronic moderately high increasing salt consumption was associated with a lower chance of depressive symptom improvement (51.82% vs 57.14%, OR = 0.86, P < 0.001) and a higher risk of depressive symptom progression (18.18% vs 14.91%, OR = 1.27, P < 0.001); chronic slightly lower and chronic highest and rapidly increasing salt consumption was associated with lower chances of depressive symptom improvement (56.1% vs 57.14%, OR = 0.88, P < 0.001; 62.67% vs 57.14%, OR = 0.84, P < 0.001) but lower risks of depressive symptom progression (12.68% vs 14.91%, OR = 0.84, P < 0.001; 5.33% vs 14.91%, OR = 0.27, P < 0.001). Compared with chronic moderately insufficient intake, chronic mildly insufficient and increasing and chronic mildly insufficient intake were associated with higher chances of depressive symptom improvement (58.73% vs 56.94%, OR = 1.25, P < 0.001; 63.01% vs 56.94%, OR = 1.42, P < 0.001), but higher risks of depressive symptom progression (15.66% vs 9.09%, OR = 2.27, P < 0.001; 10.27% vs 9.09%, OR = 1.4, P < 0.001). Most of the associations remained significant after FDR adjustment.

Discussion

Using 28-month daily dietary records in young adults, we identified several nutrient and food group intake trajectories associated with depressive symptom onset. Thirteen nutrient trajectories—including high CHO, high protein, low SFA, high Na, and moderate Se intake—were linked to lower risk, while non-recommended Zn and extremely low intake patterns were associated with higher risk. High fat and low K trajectories were linked to both greater chances of symptom improvement and higher progression risk. For food groups, high intake of oils, sauces, and HBS showed protective associations, while refined grains and light-colored vegetables were linked to increased risk. Poultry and egg intake showed mixed effects, and similar patterns were found for beans, salt, and LBS.

For macronutrient composition, our longitudinal daily nutrition records study suggested that young adults with chronic mildly high carbohydrate consumption and mildly high levels of increasing protein consumption during 28-month presented a lower risk of depressive symptoms than did those in the chronic trajectory group; however, a significant relationship between carbohydrate trajectories and depressive symptom transition among depressed individuals at baseline was not detected. Although few studies have assessed daily nutrient intake, observational studies have suggested that increased carbohydrate intake, including sugar consumption, is associated with an increased risk of depression [33, 34], and even a low-carbohydrate diet is regarded as a nutritional intervention for depression disorders [35]. However, a recent two-sample bidirectional Mendelian randomization study from the largest available genome-wide association studies indicated that the protective effect of relative carbohydrate intake on depression persisted when genetic variants associated with relative carbohydrate intake and major depressive disorder were used [36]. Carbohydrate-abundant diets can induce a lower hypothalamic–pituitary–adrenal (HPA) axis stress response, indicating a protective effect of carbohydrate consumption against stress and depression [36, 37]. Our study revealed that participants who consumed mildly high levels of increasing protein had a lower risk of depressive symptom onset than did those who did not, which is consistent with the findings of most previous studies [35, 38, 39]. The protective effect of protein intake may be explained by its role in neurotransmitter synthesis, particularly via amino acids like tryptophan and tyrosine [35, 40–43]. Moreover, dietary protein and fibre activate gluconeogenesis, affecting hormone precursors and glucocorticoid receptors and positively influencing anxiety and depression [44]. Compared with chronic recommended fat consumption, chronic slightly, mildly, or moderately greater fat consumption was associated with greater risk of depressive symptom progression, whereas slightly greater fat consumption was associated with greater odds of depressive symptom improvement; these results are consistent with those of a previous study [45]. SFA is a crucial component of total fat, and our study revealed that chronic moderately increasing lower SFA intake was associated with a lower risk of depression onset than chronic highest SFA intake. High-fat diets may negatively affect depression through inflammatory pathways and reduced brain-derived neurotrophic factor (BDNF) levels [46–50].

For micronutrients, our study revealed that a trajectory of moderately high increasing VitB3 intake was associated with a lower risk of depressive symptoms, which was consistent with previous studies showing that daily intake of vitamin B3, including supplementation, can reduce the risk of depression [51, 52]. Vitamin B3 is converted to nicotinamide, which has benzodiazepine-like properties [52]. It may treat depression by modulating adenosine triphosphate. without Sirtuin 1 activity [53], and it is related to the kynurenine pathway and affects brain receptors linked to depression [54]. Sodium intake in the human body is predominantly derived from the consumption of dietary salt, and similar results were observed for sodium and salt intake, and compared with the highest and fastest increasing and recommended sodium and salt intake groups, prolonged moderate and high Na consumption are associated with a lower risk of depression onset. A similar result was found only in a subgroup analysis of women utilizing the National Health and Nutrition Examination Survey, which indicated a negative correlation between high salt intake and depression levels [55]. A mouse experiment revealed a possible mechanism by which high salt intake induces active coping behaviour after experiencing fear stress by enhancing stress resilience, which may subsequently regulate depressive symptoms [56]. A previous study indicated that dietary salt is associated with depressive symptoms in Chinese adults in a U-shaped relationship, with both excessive and insufficient intake of dietary salt potentially increasing the risk of depressive symptoms [57]. In the present study, compared with chronic recommended zinc intake, a lower dietary zinc intake trajectory was not only associated with a greater risk of depressive symptom onset (chronic mildly lower increasing zinc intake and chronic lowest increasing zinc intake) but also a lower chance of depressive symptom improvement (chronic somewhat low zinc intake, chronic lowest decreasing zinc intake, and slightly high decreasing (1.5 years) and increasing (1.5 years) zinc intake). These results support previous findings: a meta-analysis revealed a risk ratio (RR) of 0.67 for depression based on dietary zinc intake [58], and depressed individuals had blood zinc levels 0.12 µg/mL lower than those of control subjects [59]. Zinc deficiency may lead to lower levels of synaptic zinc [60], consequently increasing glutamatergic activity via N-methyl-D-aspartate (NMDA) receptors [61], which are implicated in depression and stress-induced neurotoxicity [62], and impair BDNF activity [63].

For food components, a mildly lower increasing refined grains intake trajectory was associated with a 65% higher risk of depressive symptom onset in the present study, whereas a chronic increasing moderately high grain intake trajectory presented a similar relationship before adjusting for confounders. However, the relationship between the whole grain intake trajectory and depression status was not significant. In contrast, previous cross-sectional studies reported decreasing odds of anxiety through higher intake of whole grains [64, 65]; however, while such relationships were not observed for refined grain intake in men, women in the highest quartile of refined grain consumption had greater risk of developing depression and anxiety [65]. In the present study, chronic mildly lower and lowest increasing light-coloured vegetable intake trajectories were associated with a greater risk of depression onset, whereas dark-coloured vegetable intake was not shown to have such a relationship. Light-coloured vegetables such as lettuce and onions contain quercetin, which may modulate signalling pathways responsible for suppressing neural apoptosis and BDNF [66, 67]. A chronic moderately higher increasing oil intake trajectory was negatively associated with the risk of depressive symptom onset, which is consistent with the SFA intake trajectory. In the present study, the associations between egg intake trajectory and depressive symptoms were complex among young adults with depressive symptoms. Chronic slightly low, lowest moderately high increasing, vigorously high increasing and highest and rapidly increasing egg intake trajectories were associated with a lower risk of depressive symptom progression, which seems to indicate a U-shaped relationship between egg intake and depressive symptoms. A previous cross-sectional study revealed no significant associations between egg intake and depression, anxiety, or psychological distress [68]. The present study revealed that the chronic lowest bean intake trajectory was associated with lower odds of improving depressive symptoms. An observational study revealed that the intake of high-quality protein from beans as well as meat, eggs and milk was significantly lower in depressed patients than in healthy individuals [69]. Our study revealed that chronic slightly higher and highest and rapidly escalating soy sauce intake trajectories were associated with lower risk of depressive symptom onset than the chronic recommended intake trajectory. Soy sauce commonly contains nutrients such as sodium, potassium, magnesium, calcium, iron, and small amounts of amino acids, which are beneficial to mental health, as mentioned above. An experimental study suggested that the addition of sauce to an older person’s meal can result in increased food intake, which further increases beneficial nutrient intake [70].

For all significant dietary intake trajectories, we specifically found that high intakes of certain dietary components, such as carbohydrates, fat, sodium, oil, and sauce, were beneficial for depressive symptoms but are risk factors for cardiovascular, endocrine and other physical diseases [71]. Inappropriate dietary quality, including excessive HBS and insufficient LBS, also yielded similar results. Our findings highlight the need for nuanced dietary strategies that balance mental and physical health in young adults, suggesting that targeted modifications to specific nutrient trajectories—such as optimizing complex carbohydrate and lean protein intake while maintaining micronutrient adequacy—could offer a promising avenue for depression prevention without substantially compromising metabolic health. To translate these insights into practice, we propose integrating dynamic monitoring tools and personalized nutrition approaches that account for individual risk profiles and lifestyle factors, while advocating for updated dietary guidelines that explicitly address the mental-physical health trade-offs observed in our study. Future intervention studies should rigorously evaluate whether such tailored dietary adjustments can effectively mitigate depression risk while preserving cardiometabolic health, ultimately paving the way for more holistic and evidence-based nutritional recommendations for young adult populations. Therefore, the previously established dietary recommendations focused on physical health may be inconsistent in the field of mental health; thus, dietary guidelines for mental health need to take this into account.

The main strength of our study was the achievement of precise measurement of daily dietary intake over three consecutive years through the IOS. First, we utilized the weighing method to assess the nutrient volume of each food; thus, we were able to evaluate the trajectories of nutrients and food components for young adults. By identifying beneficial or harmful dietary intake behaviour trajectories for depression onset and transition, this approach might reveal trends in behavioural pattern changes from a long-term monitoring perspective, providing a more comprehensive assessment of the impact of diet on depression. The limitations of our study are as follows. We only included data from a single university in Shanghai, which imposes limitations on the generalizability of the findings. However, considering that the IOS was trialled for the first time at this institution and has not been implemented in other organizations or the general community, we anticipate that its subsequent broader application will provide better population representativeness. Additionally, the university cafeteria is not the sole source of food for students, particularly for food types such as fruits, milk, and desserts, which are primarily sourced externally, therefore, this may lead to a certain degree of underestimation of dietary intake. Nonetheless, our 7-day dietary recall assessment revealed that the distribution of nutrient intake sources from the campus cafeteria is highly consistent with overall dietary intake. Therefore, except a few specific food items, the IOS is able to adequately represent both the absolute and relative levels of food consumption. A third limitation pertains to the potential confounding effects of the COVID-19 pandemic. The study timeframe, encompassing dietary assessments (September 2021–December 2023) and depressive symptom follow-ups (2022–2024), overlapped with the pandemic and its post-recovery phase in Shanghai. Although dietary data from January to July 2020—covering the initial outbreak in Shanghai—were excluded, no pandemic-specific variables (e.g., infection status, exposure to mobility restrictions, changes in mental health service access) were collected. Such variables might have independently influenced transitions in depressive symptoms, thereby precluding the quantification of the pandemic’s contribution to the 59.1% improvement rate and hindering the elimination of residual confounding with dietary trajectories.

Conclusions

In conclusion, the different trajectories of daily dietary intake identified were associated with different risks or chances of onset and transition of depressive symptoms among young adults, whereas some trajectories seemed to be disadvantage effect with physical health but beneficial to depression symptom. By identifying patterns of nutrient and food component intake associated with the onset and transition of depressive symptoms, this study offers new insights into how long-term dietary behaviors may influence mental health. These results highlight the potential for individualized dietary interventions and dietary monitoring tools to support early identification and tailored prevention strategies, particularly among young adult populations.

Abbreviations

IOS

Intelligent Ordering System

Vitamin B1

Thiamine

Vitamin B2

Riboflavin

Vitamin B3

Nicotinic acid

Ca

Calcium

P

Phosphorus

K

Potassium

Na

Sodium

Mg

Magnesium

Fe

Iron

Zn

Zinc

Se

Selenium

Cu

Copper

Mn

Manganese

TFA

Total fatty acid

SFA

Saturated fatty acid

MUFA

Monounsaturated fatty acid

PFA

Polyunsaturated fatty acid

HBS

High bound score

LBS

Low bound score

DQD

Diet quality distance

GBMT

Group-based multi-trajectory model

SE

Standard error

OR

Odds ratio

CIs

Confidence interval

FDR

False discovery rate

HPA

Hypothalamic–pituitary–adrenal

NMDA

N-methyl-D-aspartate

BDNF

Brain-derived neurotrophic factor

Authors’ contributions

CH, ZY and QJ were responsible for the conceptualization of the study, methodology design, statistical analyses, interpretation of the results, drafting of the manuscript, and critical revision for intellectual content. ZWQ, DJY and PH were responsible for the data collection, sorting, and cleaning. JYN, QHH, AZ were responsible for the conceptualization of the study, organization, methodology design, and critical revisions to the manuscript. All authors read and approved the final manuscript.

Funding

The 6th Round Three-Year Initiative Plan for Strengthening Public Health System Construction in Shanghai (Project No: GWVI-6).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Our study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Medical Research, School of Public Health, Fudan University (protocol code: IRB#2019-01-0726S; data: 7 January 2019). All enrolled participants provided written informed consent. Guardians provided written informed consent for participants who were under 18 years old.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.