3.1 Cross-sectional study of NHANES
3.1.1 Baseline characteristics of the included participants
Between 2005 and 2020, a total of 76,496 US residents participated in the NHANES survey. Among the 76,496 participants, 30,516 participants under 18 years, 5388 participants with missing dietary vitamin A intake data, and 2435 participants with missing depression assessment data were excluded from the study. Finally, 38,157 participants were eligible for analysis. The exclusion process was presented in Figure 2. Of the included participants, 3448 (9.04%) reported experiencing depression. Individuals with higher vitamin A intake were more likely to be men, Non-Hispanic White, better educated, married or living with a partner, insured, have a higher income, never smokers, have cancer or malignancy, have hypercholesterolemia, and free from hypertension, diabetes, coronary heart disease, or stroke. The baseline characteristics of the included participants were presented in Table 2.
Table 2 Baseline characteristics of the included participants.
|
vitamin A intake (μg)
|
|
Variables
|
Total
|
Q1 (<298.50)
|
Q2 (298.50-491.50)
|
Q3 (492.00-762.50)
|
Q4 (>762.50)
|
P-value
|
No.
|
38157
|
9524
|
9532
|
9551
|
9550
|
|
Depression, n (%)
|
|
|
|
|
|
< 0.001
|
No
|
34709 (90.96)
|
8424 (88.45)
|
8655 (90.8)
|
8767 (91.79)
|
8863 (92.81)
|
|
Yes
|
3448 ( 9.04)
|
1100 (11.55)
|
877 (9.2)
|
784 (8.21)
|
687 (7.19)
|
|
Gender, n (%)
|
|
|
|
|
|
< 0.001
|
Male
|
18796 (49.26)
|
4402 (46.22)
|
4438 (46.56)
|
4677 (48.97)
|
5279 (55.28)
|
|
Female
|
19361 (50.74)
|
5122 (53.78)
|
5094 (53.44)
|
4874 (51.03)
|
4271 (44.72)
|
|
Age (year), Mean ± SD
|
47.76 ± 18.70
|
45.27 ± 18.17
|
47.30 ± 18.59
|
48.84 ± 18.75
|
49.61 ± 18.97
|
< 0.001
|
Race/ethnicity, n (%)
|
|
|
|
|
|
< 0.001
|
Mexican American
|
5955 (15.61)
|
1702 (17.87)
|
1625 (17.05)
|
1474 (15.43)
|
1154 (12.08)
|
|
Other Hispanic
|
3651 ( 9.57)
|
1030 (10.81)
|
1007 (10.56)
|
860 (9)
|
754 (7.9)
|
|
Non-Hispanic White
|
16017 (41.98)
|
2952 (31)
|
3825 (40.13)
|
4341 (45.45)
|
4899 (51.3)
|
|
Non-Hispanic Black
|
8615 (22.58)
|
2861 (30.04)
|
2121 (22.25)
|
1895 (19.84)
|
1738 (18.2)
|
|
Other Race
|
3919 (10.27)
|
979 (10.28)
|
954 (10.01)
|
981 (10.27)
|
1005 (10.52)
|
|
Education level, n (%)
|
|
|
|
|
|
< 0.001
|
Less than high school
|
3504 ( 9.18)
|
1202 (12.62)
|
946 (9.92)
|
753 (7.88)
|
603 (6.31)
|
|
High school diploma
|
14129 (37.03)
|
4149 (43.56)
|
3672 (38.52)
|
3344 (35.01)
|
2964 (31.04)
|
|
More than high school
|
20524 (53.79)
|
4173 (43.82)
|
4914 (51.55)
|
5454 (57.1)
|
5983 (62.65)
|
|
Marital status, n (%)
|
|
|
|
|
|
< 0.001
|
Married /living with partner
|
22112 (57.95)
|
5064 (53.17)
|
5530 (58.02)
|
5767 (60.38)
|
5751 (60.22)
|
|
Divorced/separated/widowed
|
9556 (25.04)
|
2530 (26.56)
|
2418 (25.37)
|
2302 (24.1)
|
2306 (24.15)
|
|
Never married
|
6489 (17.01)
|
1930 (20.26)
|
1584 (16.62)
|
1482 (15.52)
|
1493 (15.63)
|
|
Family PIR, Median (IQR)
|
2.09 (1.09, 4.02)
|
1.62 (0.91, 3.18)
|
2.02 (1.04, 3.79)
|
2.28 (1.19, 4.45)
|
2.56 (1.26, 4.72)
|
< 0.001
|
Energy (kcal), Mean ± SD
|
2053.56 ± 866.82
|
1624.53 ± 695.16
|
1928.04 ± 703.30
|
2171.20 ± 785.27
|
2489.03 ± 1003.61
|
< 0.001
|
Protein (g), Mean ± SD
|
80.04 ± 36.67
|
61.10 ± 29.05
|
74.84 ± 29.82
|
84.92 ± 33.19
|
99.25 ± 42.02
|
< 0.001
|
Carbohydrate (g), Mean ± SD
|
248.11 ± 110.62
|
200.38 ± 92.41
|
231.61 ± 92.93
|
259.05 ± 100.59
|
301.23 ± 127.11
|
< 0.001
|
Total fat (g), Median (IQR)
|
72.00 (50.94, 98.68)
|
54.66 (37.88, 74.67)
|
69.54 (50.92, 91.38)
|
79.58 (58.15, 105.79)
|
88.59 (63.56, 120.27)
|
< 0.001
|
BMI (kg/m2), Mean ± SD
|
29.21 ± 7.10
|
29.48 ± 7.35
|
29.58 ± 7.13
|
29.21 ± 7.10
|
28.56 ± 6.77
|
< 0.001
|
Alcoholic drinks (drink/day), Median (IQR)
|
2.00 (1.00, 3.00)
|
2.00 (1.00, 3.00)
|
2.00 (1.00, 3.00)
|
2.00 (1.00, 3.00)
|
2.00 (1.00, 3.00)
|
< 0.001
|
Hypertension, n (%)
|
|
|
|
|
|
0.069
|
No
|
24959 (65.41)
|
6292 (66.06)
|
6201 (65.05)
|
6163 (64.53)
|
6303 (66)
|
|
Yes
|
13198 (34.59)
|
3232 (33.94)
|
3331 (34.95)
|
3388 (35.47)
|
3247 (34)
|
|
Hypercholesterolemia, n (%)
|
|
|
|
|
|
< 0.001
|
No
|
24863 (65.16)
|
6462 (67.85)
|
6214 (65.19)
|
6045 (63.29)
|
6142 (64.31)
|
|
Yes
|
13294 (34.84)
|
3062 (32.15)
|
3318 (34.81)
|
3506 (36.71)
|
3408 (35.69)
|
|
Diabetes, n (%)
|
|
|
|
|
|
0.016
|
No
|
33411 (87.56)
|
8337 (87.54)
|
8285 (86.92)
|
8345 (87.37)
|
8444 (88.42)
|
|
Yes
|
4746 (12.44)
|
1187 (12.46)
|
1247 (13.08)
|
1206 (12.63)
|
1106 (11.58)
|
|
Health insurance, n (%)
|
|
|
|
|
|
< 0.001
|
No
|
8081 (21.18)
|
2577 (27.06)
|
2127 (22.31)
|
1780 (18.64)
|
1597 (16.72)
|
|
Yes
|
30076 (78.82)
|
6947 (72.94)
|
7405 (77.69)
|
7771 (81.36)
|
7953 (83.28)
|
|
Coronary heart disease, n (%)
|
|
|
|
|
|
< 0.001
|
No
|
36666 (96.09)
|
9217 (96.78)
|
9164 (96.14)
|
9133 (95.62)
|
9152 (95.83)
|
|
Yes
|
1491 ( 3.91)
|
307 (3.22)
|
368 (3.86)
|
418 (4.38)
|
398 (4.17)
|
|
Stroke, n (%)
|
|
|
|
|
|
0.015
|
No
|
36741 (96.29)
|
9125 (95.81)
|
9172 (96.22)
|
9215 (96.48)
|
9229 (96.64)
|
|
Yes
|
1416 ( 3.71)
|
399 (4.19)
|
360 (3.78)
|
336 (3.52)
|
321 (3.36)
|
|
Cancer or malignancy, n (%)
|
|
|
|
|
|
< 0.001
|
No
|
34617 (90.72)
|
8867 (93.1)
|
8712 (91.4)
|
8562 (89.65)
|
8476 (88.75)
|
|
Yes
|
3540 ( 9.28)
|
657 (6.9)
|
820 (8.6)
|
989 (10.35)
|
1074 (11.25)
|
|
Work activity, n (%)
|
|
|
|
|
|
< 0.001
|
Moderate activity
|
8705 (22.81)
|
1938 (20.35)
|
2249 (23.59)
|
2181 (22.84)
|
2337 (24.47)
|
|
Vigorous activity
|
8571 (22.46)
|
2203 (23.13)
|
2096 (21.99)
|
2121 (22.21)
|
2151 (22.52)
|
|
Other
|
20881 (54.72)
|
5383 (56.52)
|
5187 (54.42)
|
5249 (54.96)
|
5062 (53.01)
|
|
Smoking status, n (%)
|
|
|
|
|
|
< 0.001
|
Never smokers
|
21468 (56.26)
|
5123 (53.79)
|
5342 (56.04)
|
5481 (57.39)
|
5522 (57.82)
|
|
Former smokers
|
8923 (23.38)
|
1849 (19.41)
|
2169 (22.75)
|
2382 (24.94)
|
2523 (26.42)
|
|
Current smokers
|
7766 (20.35)
|
2552 (26.8)
|
2021 (21.2)
|
1688 (17.67)
|
1505 (15.76)
|
|
Note: PIR, poverty income ratio.
3.1.2 Association between dietary vitamin A intake and depression
In weighted multivariable logistic regression analyses, dietary vitamin A intake was evaluated as both continuous and categorical variables (Table 3). After controlling for potential confounding factors, the vitamin A intake (as a continuous variable) was negatively associated with depression (OR=0.92, 95%CI: 0.88-0.97, P=0.001). Similar inverse associations were found when vitamin A intake was converted into categorical variables. Compared with individuals with lower vitamin A intake in Q1, the adjusted OR values for vitamin A intake and depression in Q2, Q3, and Q4 were 0.89 (95%CI: 0.80-0.98, P=0.020), 0.89 (95%CI: 0.80-0.99, P=0.031), and 0.88 (95%CI: 0.78-0.99, P=0.032), respectively. Furthermore, the inverse associations remained strong across all models that adjusted for various covariates.
The RCS analysis found an L-shaped nonlinear relationship between dietary vitamin A intake and depression (P for non-linearity=0.010) after adjusting for all covariates (Figure 3). The results of the two-stage linear regression model revealed that the threshold effect inflection point between vitamin A intake and depression was 492.00 μg (Table 4). The results showed a different trend before and after the threshold point. Vitamin A intake was inversely associated with depression (OR=0.999, 95%CI: 0.999-1.000 P=0.002) on the left side of the inflection point. While on the right side, no significant association was found between vitamin A intake and depression (P=0.656).
The stratified analyses revealed that the trends of ORs remained steady across all stratifications (Figure 4), implying that the relationship of dietary vitamin A intake with depression was robust. In sensitivity analysis, the results remained robust despite the exclusion of individuals with missing covariate data. Dietary vitamin A intake was inversely associated with depression (OR =0.89, 95%CI: 0.83-0.95, P=0.001) after adjusting for all covariates. When dietary vitamin A intake was categorized, the adjusted ORs for Q2, Q3, and Q4 compared to Q1 were 0.83 (95% CI: 0.71-0.96, P=0.012), 0.85 (95% CI: 0.73-1.00, P=0.049), and 0.79 (95%CI: 0.66-0.94, P= 0.009), respectively.
Table 3 Association between dietary vitamin A intake and depression.
Dietary vitamin A intake (μg)
|
Event (%)
|
Non-adjusted model
|
Model 1
|
Model 2
|
Model 3
|
OR (95%CI)
|
P-value
|
OR (95%CI)
|
P-value
|
OR (95%CI)
|
P-value
|
OR (95%CI)
|
P-value
|
ln (vitamin A)
|
3448 (9.04)
|
0.80 (0.77-0.83)
|
<0.001
|
0.91 (0.87-0.94)
|
<0.001
|
0.92 (0.88-0.96)
|
<0.001
|
0.92 (0.88-0.97)
|
0.001
|
Quartiles
|
|
|
|
|
|
|
|
|
|
Q1 (<298.50)
|
1100 (11.55)
|
1 (Ref)
|
|
1 (Ref)
|
|
1 (Ref)
|
|
1 (Ref)
|
|
Q2 (298.50-491.50)
|
877 (9.20)
|
0.78 (0.71-0.85)
|
<0.001
|
0.87 (0.79-0.96)
|
0.005
|
0.89 (0.81-0.98)
|
0.020
|
0.89 (0.80-0.98)
|
0.020
|
Q3 (492.00-762.50)
|
784 (8.21)
|
0.68 (0.62-0.75)
|
<0.001
|
0.86 (0.78-0.95)
|
0.003
|
0.89 (0.80-0.99)
|
0.034
|
0.89 (0.80-0.99)
|
0.031
|
Q4 (>762.50)
|
687 (7.19)
|
0.59 (0.54-0.66)
|
<0.001
|
0.83 (0.74-0.92)
|
<0.001
|
0.87 (0.77-0.98)
|
0.022
|
0.88 (0.78-0.99)
|
0.032
|
Trend test
|
|
|
<0.001
|
|
<0.001
|
|
0.024
|
|
0.032
|
Note: CI, confidence interval; OR, odds ratio; ln (vitamin A), vitamin A intake values were natural logarithmised. Model 1 was adjusted for age, gender, race/ethnicity, marital status, education level, family income, health insurance, BMI, work activity, alcoholic drinks, and smoking status. Model 2 was adjusted for age, gender, race/ethnicity, marital status, education level, family income, health insurance, BMI, work activity, alcoholic drinks, smoking status, energy, protein, carbohydrate, and total fat. Model 3 was was fully adjusted, including age, gender, race/ethnicity, marital status, education level, family income, health insurance, BMI, work activity, alcoholic drinks, smoking status, energy, protein, carbohydrate, total fat, diabetes, hypertension, hypercholesterolemia, stroke, coronary heart disease, and cancer or malignancy.
Table 4 Threshold effect analysis of the relationship between vitamin A intake and depression.
Inflection point
|
OR (95%CI)
|
P-value
|
Vitamin A intake<492.00 μg
|
0.999 (0.999-1.000)
|
0.002
|
Vitamin A intake≥492.00 μg
|
1.000 (1.000-1.000)
|
0.656
|
Likelihood ratio test
|
|
0.001
|
Note: CI, confidence interval; OR, odds ratio. Adjustment covariates included age, gender, race/ethnicity, marital status, education level, family income, health insurance, BMI, work activity, alcoholic drinks, smoking status, energy, protein, carbohydrate, total fat, diabetes, hypertension, hypercholesterolemia, stroke, coronary heart disease, and cancer or malignancy.
3.2 Causal association in MR analysis
3.2.1 Genetic IVs
We selected SNPs using a P threshold of 5×10-6, resulting in 101 SNPs that met the criteria. Out of the 101 SNPs, 84 were excluded during the clumping process. The F statistics for the identified SNPs surpassed the empirical threshold of 10, with values between 20.90 and 31.35. Two SNPs were not included in the outcome data. The harmonized process revealed no SNPs classified as palindromic SNPs. No SNPs were excluded using PhenoScanner V2. Finally, 15 eligible SNPs were selected as genetic IVs to evaluate the causal relationship between vitamin A and depression risk. The selection process for genetic IVs was illustrated in Figure 5.
3.2.2 MR Analyses
The IVW method found no significant causal relationship between vitamin A and the risk of depression (OR=0.39, 95%CI: 0.10-1.58, P=0.188). Similar estimates were found in MR-Egger method (OR=0.26, 95%CI: 0.02-4.18, P=0.357), weighted median method (OR=0.75, 95%CI: 0.11-4.93, P=0.763), simple mode method (OR=1.76, 95% CI: 0.06-49.39, P=0.745), and weighted mode method (OR=1.76, 95%CI: 0.05-56.69, P=0.755). The results of the MR analyses were listed in Table 5. The scatter plot of the MR analyses was presented in Figure 6. In sensitivity analyses, the Cochran's Q test revealed no significant heterogeneity (P=0.570). Both the MR-Egger intercept (P for intercept=0.739) and MR-PRESSO analysis (global test P=0.550) did not find any horizontal pleiotropy. Furthermore, the results of leave-one-out analysis demonstrated that no single SNP drove the MR estimate (Figure 7).
Table 5 Results of MR estimates.
Method
|
Number of SNP
|
OR (95%CI)
|
P-value
|
Heterogeneity
(P-value)
|
Pleiotropy
(P-value)
|
MR-PRESSO
(P-value)
|
IVW
|
15
|
0.39 (0.10-1.58)
|
0.188
|
0.570
|
0.739
|
0.550
|
MR-Egger
|
15
|
0.26 (0.02-4.18)
|
0.357
|
Weighted median
|
15
|
0.75 (0.11-4.93)
|
0.763
|
Simple mode
|
15
|
1.76 (0.06-49.39)
|
0.745
|
Weighted mode
|
15
|
1.76 (0.05-56.69)
|
0.755
|
Abbreviations: IVW, inverse-variance weighted; CI, confidence interval; OR, odds ratio; SNP, single-nucleotide polymorphisms.