Investigating Associations Between Depressive Symptoms and Anti-/Pro-Inflammatory Nutrients in an Elderly Population in Northern China: A Bayesian Kernel Machine Regression Approach

Figures & data

Figure 1 Selection process of subjects.

Table 1 Baseline Characteristics of the 1865 Subjects [Median(IQR) or Frequency(%)]

Characteristic	Frequency (%) or Median(IQR)
	Non-Depression (n=1418)	Depression (n=447)	p-value
Age (years)	68(63–72)	65(61–70)	0.036
BMI(kg/m2)	24.70(22.22–27.01)	24.98(22.55–27.54)	0.078
Daily energy intake (kcal/day)	1398.97(1100.57–1582.35)	1304.99(1122.20–1610.83)	0.952
Sex			<0.001
Male	576(40.6%)	140(31.3%)
Female	842(59.4%)	307(68.7%)
Residence			0.033
Urban	380(26.8%)	143(32.0%)
Rural	1038(73.2%)	304(68.0%)
Employment			0.051
No	1211(85.4%)	398(89.0%)
Yes	207(14.6%)	49(11.0%)
Education level			0.422
Illiteracy	368(26.0%)	103(23.0%)
Primary school	460(32.4%)	146(32.7%)
Junior high school/above	590(41.6%)	198(44.3%)
Tobacco Smoking			0.941
No	1198(84.5%)	377(84.3%)
Yes	220(15.5%)	70(15.7%)
Alcohol Drinking			0.555
No	1281(90.3%)	408(91.3%)
Yes	137(9.7%)	39(8.7%)
Diabetes			0.757
No	1192(84.1%)	373(83.4%)
Yes	226(15.9%)	74(16.6%)
Hypertension			0.153
No	514(40.5%)	164(36.7%)
Yes	844(59.5%)	283(63.3%)

Table 2 Nutrient Content of Study Participants

Characteristic	Median(IQR)
	Non-Depression (n=1418)	Depression (n=447)	p-value
Carbohydrates (g)	195.25(160.29–235.58)	198.02(169.06–271.63)	0.002
Protein (g)	53.66(42.76–66.40)	58.36(47.25–71.69)	<0.001
Total fat (g)	35.00(26.82–44.47)	36.84(29.07–47.22)	0.002
β-Carotene (μg)	1516.91(781.85–2745.53)	1489.11(703.49–2665.51)	0.933
Fiber (g)	8.83(6.49–11.62)	9.01(6.81–12.08)	0.027
Cholesterol (mg)	381.02(230.35–435.12)	400.13(247.34–439.07)	0.070
Saturated fat (g)	4.00(2.51–6.01)	4.21(2.82–6.27)	0.025
Monounsaturated fat (g)	5.85(4.12–7.70)	6.13(4.62–7.71)	0.014
Polyunsaturated fats (g)	4.61(2.80–6.93)	5.37(3.15–7.58)	0.001
Niacin (mg)	8.87(6.86–11.00)	9.25(7.43–11.58)	0.004
Thiamine (mg)	0.72(0.57–0.89)	0.74(0.60–0.97)	0.013
Riboflavin (mg)	0.66(0.51–0.83)	0.70(0.54–0.87)	0.012
Vitamin B12 (μg)	0.22(0.14–0.33)	0.24(0.15–0.35)	0.124
Vitamin B6 (mg)	0.07(0.05–0.10)	0.07(0.05–0.10)	0.144
Fe (mg)	15.68(12.20–19.11)	16.48(13.31–20.14)	<0.001
Magnesium (mg)	217.65(167.20–266.38)	227.80(174.39–283.04)	0.002
Zn (mg)	7.19(5.66–8.70)	7.54(6.26–9.22)	<0.001
Selenium (μg)	40.64(32.48–48.53)	41.94(35.65–50.99)	0.001
Vitamin A (RE)	432.60(283.42–660.53)	453.87(277.51–683.43)	0.386
Vitamin C (mg)	51.88(27.52–86.17)	47.68(29.36–85.71)	0.923
Vitamin E (mg)	11.73(8.22–14.96)	12.43(9.18–16.52)	0.001
Folic acid (μg)	116.79(83.59–156.07)	121.04(88.64–165.75)	0.127

Figure 2 Factor selection using the LASSO logistic regression model. (A) LASSO coefficients of 22 candidate variables. (B) Identification of the optimal penalization coefficient (λ) in the LASSO model was achieved by 10-fold cross-validation and the minimum criterion. The left vertical line represents the minimum error, and the right vertical line represents the cross-validated error within 1 standard error of the minimum.

Abbreviation: LASSO, least absolute shrinkage and selection operator.

Table 3 Food Parameter Specific DII Score

Food Parameter Specific DII Score	Median(IQR)
	Overall(n=1865)
Anti-inflammatory
Carbohydrates (g)	–0.0289(–0.0457 – –0.0075)
Total fat (g)	–0.0919(–0.1484 – –0.0170)
Cholesterol (mg)	–0.0040(–0.0418 – 0.0189)
Saturated fat (g)	–0.0708(–0.1836 – 0.0890)
Vitamin B12 (μg)	–0.0518(–0.0702 – –0.0239)
Pro-inflammatory
Thiamine (mg)	0.0145(–0.0159 – 0.0399)
Riboflavin (mg)	0.0135(–0.0068 – 0.0311)
Polyunsaturated fats (g)	0.1880(0.1085 – 0.2562)
Fe (mg)	0.0031(–0.0080 – 0.0145)
Zn (mg)	0.0872(0.0164 – 0.1569)
Selenium (μg)	0.0757(0.0424 – 0.1067)

Figure 3 Associations between dietary pro-inflammatory nutrients and depression among the study population by BKMR model. The model is based on age, gender, education level, body mass index, employment status, smoking status, drinking status, daily energy intake, hypertension and diabetes, and other nutrients. (A) Cumulative effects of dietary pro-inflammatory nutrients (estimated value and 95% confidence interval). Compared with other nutrients in the 50th percentile, the nutrients are in a specific percentile (X-axis). (B) The single-nutrient effect (estimates and 95% credible intervals). (C) The univariate nutrient response function of each dietary pro-inflammatory nutrient has a 95% confidence band, and the other nutrients are fixed at the median. (D) Bivariate exposure response function of each two pro-inflammatory nutrients in depression.

Figure 4 Associations between dietary anti-inflammatory nutrients and depression among the study population by BKMR model. The model is based on age, gender, education level, body mass index, employment status, smoking status, drinking status, daily energy intake, hypertension and diabetes, and other nutrients. (A) Cumulative effects of dietary anti-inflammatory nutrients (estimated value and 95% confidence interval). Compared with other nutrients in the 50th percentile, the nutrients are in a specific percentile (X-axis). (B) The single-nutrient effect (estimates and 95% credible intervals). (C) The univariate nutrient response function of each dietary anti-inflammatory nutrient has a 95% confidence band, and the other nutrients are fixed at the median. (D) Bivariate exposure response function of each two anti-inflammatory nutrients in depression.