Perinatal dietary patterns and symptomatic depression: A prospective cohort study

Abstract

To promote maternal and infant health, there is a need to optimise the dietary pattern of pregnant women to reduce perinatal depression. This prospective cohort study was conducted from June 2020 to February 2022, 300 women from a medical center were interviewed during late pregnancy and at 4–6 weeks postpartum. Dietary patterns were derived by factor analysis using a semiquantitative food frequency questionnaire. Symptomatic depression was defined using the Edinburgh Postpartum Depression Scale (EPDS, ranged 0–30). Their dairy, vegetable and fruit intakes were below the Taiwanese recommendations for pregnant women. Symptomatic depression (EPDS ≥10) affected 31.3% in the third trimester and 35.7% postpartum. Pre‐ and post‐EPDS scores were positively correlated (r = 0.386, p < 0.001). Approximately 55% of those depressed before delivery were also depressed postpartum. For late pregnancy, four dietary patterns were identified (‘Good oil’, ‘Vegetables and fruits’, ‘Omnivorous’ and ‘Refined‐grain and organ meats’). Dietary patterns were classified according to quartiles (Q). Higher omnivorous pattern scores reduced the risk of depression. For prenatal depression, with Q1 as a reference, the risk was reduced by 38% for Q2, 43% for Q3 and 59% for Q4 (p for trend = 0.068). These findings became evident postpartum (reduced risk by 68% for Q2, 69% for Q3 and 70% for Q4 (p = 0.031; p for trend = 0.0032). The association between dietary patterns and depression encourages the routine nutritional management of pregnant women.

Key messages

• Perinatal women have a higher prevalence of symptomatic depression, and more than half of women with antenatal depression also have postpartum depression.

• Pregnant women often face challenges in trying to achieve a healthy diet, as reflected in our findings on low vegetable and fruit intake.

• An omnivorous pattern was found to protect against symptomatic depression in late pregnancy, which persisted into the early postpartum period.

• Through effective nutritional counselling, paying attention to the risk of depression in perinatal women is expected to reduce women's pregnancy anxiety.

1. INTRODUCTION

Maternal stress, anxiety and depression may affect maternal health and increase the risk of premature delivery and low birth weight, ultimately affecting child health (Grote et al., 2010). The prevalence of postpartum depression precoronavirus pandemic worldwide and in Taiwan is reported to be 10%–20% (Iyengar et al., 2021) and 8.4% (Lin et al., 2019), respectively. That worldwide increased from 20% to 64% (Iyengar et al., 2021).

Dietary patterns broadly capture the overall picture of food and nutrient intake and may be more predictive of disease risk than these alone (Hu, 2002). Traditionally, women of Chinese ancestry have focused on postpartum dietary practices (Mao et al., 2016; Qin et al., 2021). Moreover, there is increasing interest among Asian women in their postpartum diets and mental health (Poh et al., 2005). Dietary patterns have been associated with depression in pregnant women (Baskin et al., 2017; Cao et al., 2020). Diet interventions can also reduce symptoms of depression and anxiety (Firth et al., 2019). In addition, the rapid development of digital network functions has dramatically changed people's lives and eating behaviours (Chenarides et al., 2021). Therefore, it is necessary to understand the dietary patterns and their health consequences in vulnerable groups.

In Taiwan, with the changes in the socioeconomic environment, the high employment rate and advanced maternal age lead to the lowest birth rate in the world. Considering the local food culture, there is an urgent need to understand the relationship between diet and maternal depression. The present study endeavours to identify dietary patterns among Taiwanese women during the third trimester of pregnancy and whether they are associated with perinatal symptomatic depression by way of a cross‐sectional and prospective follow‐up design in a clinical setting.

2. METHODS

2.1. Design and participants (Figure 1)

Figure 1.

Diagram of the research framework.

This prospective cohort study was conducted at the Obstetrics and Gynecology Clinic of the Tri‐Services General Hospital (TSGH) from June 2020 to February 2022, approved by the Human Research Ethics Committee (HREC) of the TSGH, Taiwan (C202005096). In Taiwan, pregnant women have access to freely available antenatal care from recognition of pregnancy through the National Health Insurance scheme for 14 visits. The study participants were healthy pregnant women who underwent routine prenatal care at a medical center at 29–40 weeks of gestation (third trimester). Given that α = 0.05 and power = 0.8, based on a previous study, the required sample size was estimated to be 200 (Miyake et al., 2018). The inclusion criteria were (1) age 20 years or older, (2) singlet pregnancy, (3) no recognisable high‐risk for major medical, surgical, or mental illness, (4) being mobile and (5) able to communicate and complete questionnaires.

2.2. Data collection and baseline information

Participants completed questionnaires with a nutrition‐trained interviewer at the time of enrolment (third trimester) and early postpartum (4–6 weeks postpartum). Questionnaires comprised sociodemographic information (age, education level, marital status, economic status, occupation and family support), anthropometric and health‐related behaviours (weight, height, smoking, drinking, physical activity and sleep quality), reproductive status (parity, newborn sex, abortion/miscarriage experience, mode of delivery and infant feeding method) and medical history. In addition, during the second interview (4–6 weeks postpartum), the diet and locality for confinement were collected.

2.3. Dietary pattern

Dietary data were collected by a 33‐food group semi‐quantitative food frequency questionnaire (FFQ) with additional food preparation method questions to allow estimation of cooking oil. This FFQ has acceptable reproducibility and validity (Chien et al., 2013; Chou et al., 2011; Lee et al., 2011; Yu‐Ru Hu et al., 2022). Two‐dimensional food pictures were provided to assist in the estimation of portion size. Food and nutrient intakes were calculated accordingly. Dietary patterns were derived and named by factor analysis (principal component analysis, PCA) using daily food consumption frequency information. Factors were made independent of each other using a varimax orthogonal rotation with Kaiser normalisation. The variables with the absolute value of factor loading less than 0.3 were excluded, and four dietary patterns were generated after selecting the eigenvalue >1.4 (Hu et al., 2000). Using the factor loadings as weights and the frequency of various food group intakes, we constructed the respective factor scores for each participant.

2.4. Outcome ascertainment

The Edinburgh Postpartum Depression Scale (EPDS) was used to determine the risk of depression and has been verified for pregnant, maternal and nonpregnant women. The EPDS scale (range 0–30) was generated based on 10 questions, where questions 3 and 5–10 were reversed questions. The higher the score, the more severe the risk is. In the present study, an EPDS ≥score of 10 was considered symptomatic depression (Cox et al., 1987).

2.5. Statistical analysis

Categorical variables were expressed as numbers and percentages, and continuous variables were expressed as mean ± standard deviation. Differences among groups were tested using analysis of variance or χ ² tests, as appropriate. Logistic regression analysis was performed to assess the association between dietary patterns and perinatal depression. The potential covariates were those with p < 0.1 on univariable analysis or found to be important in the literature (Baskin et al., 2017; Cao et al., 2020). Factor scores for the dietary patterns were calculated using factor loading and consumption frequency. Pattern scores were classified according to the quartiles in the analysis.

3. RESULTS

A total of 300 pregnant women completed the third‐trimester questionnaire, and 258 completed the questionnaire after delivery. The reasons for the noncompletion of the postpartum questionnaire were employment obligations and newborn care (27 women), nonresponders (10 women) and study scheduling (five women) (Figure 1).

3.1. The characteristics of women in the third trimester and the risk of depression (Table 1)

Table 1.

Participant enrolment characteristics based on the Edinburgh Postnatal Depression Score (EPDS).

Characteristic	EPDS
	Total	<10a	≥10b	p
Number	300	206	94
Age (year)	32.9 ± 4.90	33.2 ± 4.94	32.1 ± 4.73	0.059
Gestation age (week)	34.3 ± 2.83	34.1 ± 2.80	34.7 ± 2.86	0.051
29–32	29.7	31.6	25.5	0.122
33–36	45.0	46.6	41.5
≥37	25.3	21.8	33.0
Education (%)				0.003
High school or less	9.30	5.30	18.1
College/University	72.0	74.8	66.0
Graduated School	18.7	19.9	16.0
Married (%)	97.7	99.5	93.6	0.004
Unemployed (%)	16.7	13.6	23.4	0.034
Perceived financial status (%)				0.081
More than enough	23.7	26.7	17.2
Other	76.3	73.3	82.8
Sedentary (%)	54.3	50.5	62.8	0.048
No medical history (%)	84.7	85.0	84.0	0.839
Primiparas (%)	64.7	61.7	71.3	0.106
Unplanned pregnancy (%)	31.7	30.1	35.1	0.387
Miscarriage experience (%)	24.7	22.3	29.8	0.165
Male fetus (%)	64.7	61.7	71.3	0.106
BMI (kg/m2)
Prepregnancy	22.4 ± 3.70	22.4 ± 3.62	22.3 ± 3.87	0.752
Late gestation	26.0 ± 3.69	26.0 ± 3.54	26.2 ± 4.00	0.623
BW change (kg)	9.42 ± 3.89	9.14 ± 3.67	10.0 ± 4.28	0.064
BW change (%)	17.0 ± 7.57	16.4 ± 7.14	18.2 ± 8.33	0.052
Macronutrients
Total energy (kcal)	1754 ± 423	1771 ± 430	1717 ± 407	0.307
Carbohydrates (%)	51.3	52.2	52.9	0.504
Protein (%)	15.2	15.8	15.5	0.498
Fats (%)	33.5	33.8	33.4	0.537
Fibre (g)	13.5 ± 5.3	13.6 ± 5.2	13.3 ± 5.4	0.585
Food groups (serving)
Grains	2.9 ± 0.9	3.0 ± 0.9	2.9 ± 1.0	0.621
Soy/fish/meat/egg	5.5 ± 2.3	5.6 ± 2.4	5.2 ± 2.1	0.218
Milk and dairy products	0.8 ± 0.6	0.8 ± 0.6	0.7 ± 0.5	0.044
Vegetables	1.2 ± 1.1	1.2 ± 1.1	1.2 ± 1.2	0.791
Fruits	1.1 ± 0.8	1.2 ± 0.7	1.0 ± 0.8	0.163

Note: one serving size of grains is equal to two slices of bread; one cup cooked rice, pasta or cereal. Soy/fish/meat/egg: one oz cooked meat; half of tofu; two pieces of dried bean curd. Dairy products: eight oz milk; one cup yogurt; 1 1/2 oz cheese. Vegetables: one cup raw leafy vegetables; 1/2 cup of cooked vegetables. Fruits: one median piece of fruit; 1/4 cup dried fruit; 1/2 fresh, frozen or canned fruit; six oz fruit juice. All categorical and continuous variables were presented as either proportions or means ± standard deviation (SD), respectively. p < 0.05 was considered statistically significant. Each subscript letter denotes whose column proportions are significantly different from each other at the p = 0.05 level.

Abbreviation: BMI, body mass index.

^a EPDS <10, low risk of postpartum depression.

^b EPDS ≥10, high risk of postpartum depression.

Participants were aged 21–47 years (average: 32.9 years) and were at 34.3 weeks of pregnancy. Most (72.0%) had a college/university education, 83.3% had a career, about a quarter considered their financial situation to be more than enough, and more than half were sedentary. In addition, 15.3% of participants had a relevant medical history. Some 64.7% were primiparas, 31.7% had an unplanned pregnancy, 24.7% had a history of miscarriage and 64.7% of fetuses were male. BMIs before pregnancy and in the third trimester were 22.4 and 26.0 kg/m², respectively, a 17.0% increase. Those at risk of depression were less educated, unmarried, unemployed, had more experiences of miscarriage and were more sedentary than those with less risk (p < 0.05).

3.2. The prevalence of depression in the third trimester and early postpartum

The prevalence of prenatal depression (EPDS ≥10) was 31.3%, and postpartum depression was 35.7% (data not shown). There was a positive correlation between the before and after EPDS measures (r = 0.386, p < 0.001). Those who were at risk before delivery (51 women, 55.4%) were also at risk postpartum, with a 5.64‐fold risk (p < 0.001, data not shown). If EPDS ≥13 were the cut‐off point, the risk of depression would have been 12.3% in the third trimester and 19.4% in the early postpartum period (data not shown).

3.3. Diet of pregnant women

Compared with the acceptable macronutrient distribution ranges (AMDR) in Taiwan (Health Promotion Administration, Ministry of Health and Welfare, 2021), the average energy intake for all pregnant women was lower (<1800 kcal), fat intake was higher (>30%) and dietary fibre intake was only 13 g per day. Those at risk of depression ate less, except for carbohydrates. According to the food category, irrespective of depression status, the intake of milk/dairy, vegetables and fruits did not reach the recommendation. However, milk/dairy intake differed significantly between the two groups (p = 0.044). Vegetable (1.2 out of 4–5 servings, 24%–30%) and fruit (1–1.2 out of 3–4 servings, 33%–40%) intakes were far less than the recommendations (Health Promotion Administration, Ministry of Health and Welfare, 2018) (Table 1).

3.4. Factor analysis of dietary patterns

Four dietary patterns were derived by factor analysis from the FFQ for the third‐trimester diet, which explained 37.6% of the variance. The four dietary patterns were named (1) ‘Good oil pattern’ (MUFA vegetable oils, fatty fish), (2) ‘Vegetables and fruits pattern’, (3) ‘Omnivorous pattern’ (Dairy products, soy products, meats, poultry, eggs, fish and seafood) and (4) ‘Refined grains and organ meats pattern’ (white rice, sweetened beverages, white bread and organ meats). (Table 2) Daily energy intake and nutrient densities by quartiles of dietary patterns in the third trimester supported the pattern nomenclature used (Supporting Information: Table S1).

Table 2.

Factor loadings for four major dietary patterns derived from principal component analysis with orthogonal rotation for the third trimester of pregnancy.

	Good oil pattern	Vegetables and fruits pattern	Omnivorous pattern	Refined‐grain and organ meats pattern
Vegetable oils, MUFA	0.889
Vegetable oils, PUFA	−0.883
Fatty fish	0.425
Vegetables		0.862
Fruits		0.850
Dairy products			0.599
Soy products			0.514
Meats			0.506
Poultry			0.497
Eggs			0.459
Soy drinks			0.424
Fish and seafood			0.418
Whole grains				−0.701
White rice				0.653
Sweetened beverages				0.416
White breads				0.377
Organ meats				0.375

Note: Factor loadings with absolute values ≥0.30 were listed in the table.

In general, the nutritional quality of the omnivorous pattern diet was the most favourable. For the ‘Omnivorous pattern’, more protein‐rich foods were consumed, with a linear trend. (Supporting Information: Table S2) This pattern reflects a wider range of food categories and essential nutrients, including both plant‐ and animal‐derived foods.

3.5. Effect of dietary patterns on the risk of perinatal depression (Tables 3 and 4)

Table 3.

Odds ratios for the effect of dietary patterns on depression symptoms in the third trimester (n = 300).

Dietary patterns	N	Depression symptoms
		Crude OR (95% CI)	p	p for trend	Model 1a OR (95% CI)	p	p for trend	Model 2b OR (95% CI)	p	p for trend
Good oil pattern				0.906			0.856			0.817
Q1 (Ref)	75	1.00			1.00			1.00
Q2	75	0.83 (0.42–1.65)	0.602		0.64 (0.29–1.40)	0.264		0.62 (0.28–1.37)	0.239
Q3	75	0.78 (0.39–1.56)	0.484		0.78 (0.36–1.71)	0.538		0.76 (0.35–1.67)	0.500
Q4	75	0.83 (0.42–1.65)	0.602		0.84 (0.39–1.83)	0.659		0.80 (0.37–1.77)	0.588
Vegetables and fruits pattern				0.789			0.958			0.766
Q1 (Ref)	75	1.00.			1.00			1.00
Q2	75	1.36 (0.68–2.73)	0.379		1.17 (0.53–2.58)	0.702		1.21 (0.54–2.68)	0.643
Q3	75	1.07 (0.53–2.17)	0.857		0.92 (0.41–2.04)	0.831		1.00 (0.44–2.27)	0.991
Q4	75	1.29 (0.64–2.58)	0.479		0.87 (0.38–2.01)	0.750		1.03 (0.41–2.60)	0.953
Omnivorous pattern				0.011			0.014			0.068
Q1 (Ref)	75	1.00			1.00			1.00
Q2	75	0.53 (0.27–1.04)	0.066		0.57 (0.27–1.23)	0.153		0.62 (0.28–1.36)	0.229
Q3	75	0.50 (0.26–0.98)	0.044		0.52 (0.24–1.10)	0.088		0.57 (0.26–1.28)	0.176
Q4	75	0.30 (0.15–0.62)	0.001		0.34 (0.15–0.77)	0.009		0.41 (0.16–1.05)	0.063
Refined grain and organ meats pattern				0.167			0.521			0.427
Q1 (Ref)	75	1.00.			1.00			1.00
Q2	75	0.77 (0.38–1.57)	0.468		0.58 (0.26–1.36)	0.197		0.58 (0.26–1.33)	0.198
Q3	75	0.88 (0.44–1.78)	0.720		0.71 (0.31–1.63)	0.419		0.75 (0.32–1.74)	0.497
Q4	75	1.59 (0.81–3.12)	0.175		1.13 (0.51–2.47)	0.768		1.28 (0.55–2.99)	0.571

Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio.

^a Adjusted for age (year), prepregnancy BMI (kg/m²), weight change, education, marital status, perceived financial status, occupation, miscarriage experience, medical history, gestation age (week), unplanned pregnancy, birth gender and physical activity.

^b Adjusted for the variables listed above and for total energy intake.

Table 4.

Odds ratios for the effect of the third‐trimester dietary pattern on early postpartum depression symptoms (n = 258).

Dietary patterns	N	Depression symptoms
		Crude OR (95% CI)	p	p for trend	Model 1a OR (95% CI)	p	p for trend	Model 2b OR (95% CI)	p	p for trend
Good oil pattern				0.948			0.898			0.965
Q1 (Ref)	75	1.00.			1.00			1.00
Q2	75	1.01 (0.49–2.06)	0.980		0.87 (0.39–1.94)	0.735		1.04 (0.44–2.48)	0.929
Q3	75	0.83 (0.40–1.72)	0.614		0.70 (0.31–1.60)	0.399		0.70 (0.29–1.71)	0.433
Q4	75	0.93 (0.45–1.92)	0.841		0.96 (0.43–2.13)	0.918		1.08 (0.46–2.56)	0.857
Vegetables and fruits pattern				0.242			0.179			0.130
Q1 (Ref)	75	1.00.			1.00			1.00
Q2	75	0.79 (0.39–1.61)	0.514		0.75 (0.34–1.68)	0.487		0.66 (0.27–1.59)	0.349
Q3	75	0.48 (0.23–1.00)	0.048		0.39 (0.17–0.89)	0.025		0.33 (0.13–0.83)	0.019
Q4	75	0.83 (0.40–1.70)	0.608		0.61 (0.26–1.42)	0.250		0.50 (0.18–1.40)	0.189
Omnivorous pattern				0.018			0.008			0.032
Q1 (Ref)	75	1.00			1.00			1.00
Q2	75	0.48 (0.24–1.98)	0.042		0.37 (0.17–0.82)	0.014		0.32 (0.13–0.78)	0.012
Q3	75	0.39 (0.19–0.80)	0.010		0.35 (0.16–0.78)	0.010		0.31 (0.12–0.80)	0.016
Q4	75	0.36 (0.17–0.74)	0.006		0.32 (0.14–0.74)	0.008		0.30 (0.10–0.90)	0.031
Refined grain and organ meats pattern				0.980			0.421			0.171
Q1 (Ref)	75	1.00			1.00			1.00
Q2	75	1.08 (0.52–2.22)	0.842		0.77 (0.34–1.74)	0.525		0.83 (0.35–2.00)	0.684
Q3	75	1.17 (0.78–2.36)	0.672		0.72 (0.32–1.65)	0.437		0.68 (0.27–1.68)	0.397
Q4	75	1.10 (0.54–2.28)	0.790		0.64 (0.28–1.48)	0.296		0.42 (0.15–1.18)	0.099

Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio.

^a Adjusted for age (year), prepregnancy BMI (kg/m²), weight change, education, marital status, perceived financial status, occupation, medical history, gestation age at delivery (week), miscarriage experience, intentional pregnancy plan, delivery method, infant feeding method, infant health, physical activity, family support.

^b Adjusted for the variables listed above and for total energy intake

In the third trimester, the crude model, the higher the ‘Omnivorous pattern’ or the ‘Good oil pattern’ scores were the lower the risk of depression, but only the ‘Omnivorous pattern’ had significance. Compared with Q1, the odds ratios (ORs) and 95% confidence intervals (CIs) were 0.53 (0.27–1.04) for T2, 0.50 (0.26–0.98) for Q3 and 0.30 (0.15–0.62) for Q4, with a linear trend (p for trend 0.011). For the ‘Omnivorous pattern’, after controlling for potential covariates (Model 2), compared with Q1 had a 38% lower risk of depression for Q2, 43% for Q3 and 59% for Q4 (p for trend = 0.068).

Women in the omnivorous pattern remained at a significantly lower risk of early postpartum depression for Q2 (OR = 0.32, 95% CI (0.13–0.78), p = 0.012), Q3 (OR = 0.31, 95% CI (0.12–0.80), p = 0.016) and Q4 (OR = 0.30, 95% CI (0.10–0. 90), p = 0.031), with a linear trend (p for trend 0.032, Model 2).

There were no interactions between diet and education, unmarried, unemployed, miscarriage experience, physical activity and perinatal depression (data not shown). As far as possible effects of confinement diet on postpartum depression were observed, we could not detect any significant associations (data not shown).

4. DISCUSSION

Pregnant women at a medical center had a higher prevalence of postpartum depression than those in previous studies in Taiwan (Chen et al., 2007; Lin et al., 2019). The prevalence of depression (EPDS ≥ 10) in the third trimester was 31.3%, and 35.7% in the early postpartum. Almost 55% of the participants with prenatal depression continued with it postpartum. Lower education, unmarried, unemployed, experience of miscarriage, and less physical activity were associated with depression risk, as in other studies (Dennis & Dowswell, 2013; Liu et al., 2021).

We observed that pregnant women ate fewer servings of vegetables (1.2 servings/day) or fruit (1.0–1.2 servings/day), well below recommendations. An omnivorous pattern was found to protect against symptomatic depression in the third trimester, which lasted into the early postpartum period. The extent to which this effect may have lasted beyond this time is unclear. Since the confinement diet had no recognisable association with postpartum depression, it suggests that nutritional action needs to be taken earlier in its prevention.

4.1. Depression risk for perinatal women

Mental health disorders are common during the perinatal period. The depression risk has been found to be even higher during the specific pandemic (prevalence 20%–64%) (Iyengar et al., 2021), as supported by our study in Taiwan. Compared with previous data in Taiwan, the prevalence of depression 6–8 weeks postpartum was lower than in the present study (8.4%–10.3% vs. 35.7%) (Chen et al., 2007; Lin et al., 2019). Unfortunately, there are no local data available for prenatal depression comparisons.

Factors that add to the risk of depression during and after pregnancy include the pregnancy itself, term date, primipara and young maternal age. During the COVID pandemic, the uncertainty of the new virus for pregnant women and their newborns increased because of the physical and emotional burden. That also possibly raised maternal anxiety through greater isolation, loneliness, threats to income and access to healthcare in response to epidemic control measures. In addition, more time spent online with overwhelming and conflicting information became problematic. These various factors were likely to endanger the well‐being and mental health of pregnant women (Holmes et al., 2020; O'connor et al., 2021).

4.2. Diet pattern of the participants

4.2.1. Dietary intake during the third trimester

Optimising the nutritional status of pregnant women and their offspring is essential. Not only do physiological demands need to be met with appropriate weight gain, avoidance of obstetric complications, and preparation for lactation, but also provision for fetal growth and neonatal health. However, pregnant women are often challenged in their efforts to achieve healthy eating (Jardí et al., 2019; Rodríguez‐Bernal et al., 2013), as reflected in our findings of low vegetable and fruit intake. Similar results were found in a Hong Kong study with high sodium intake and insufficient fibre intake (Tsoi et al., 2022). It may be caused by traditional food culture or discomfort during pregnancy. Therefore, timely dietary advice and adjustments are necessary.

Compared to pregnant women and women in the same age group (19–44 years), reported in the Taiwanese nutrition survey (Health Promotion Administration, Ministry of Health and Welfare, 2018b) and as recommended in the Food Guides for the 3rd Trimester (Health Promotion Administration, Ministry of Health and Welfare, 2018a, 2021), total energy, percentage of carbohydrate, vegetable and fruit intakes were low. This problem may be more solvable than apparent, given that women are aware of the importance of nutrition during pregnancy for reasons that include physical discomfort (abdominal bloating) and altered taste, which hinders intake. However, constraints remain for these women, such as impaired mobility, along with time and inclination for shopping, food preparation and mealtimes. Regrettably, the increase in online food delivery with its cost and limited food options might do little to solve pregnant women's overall food system and nutrition needs (Chenarides et al., 2021).

4.2.2. Energy intake and dietary patterns

Depression and energy intake have a bidirectional relationship (Mayo Clinic, n.d.; Paykel, 1977; Simmons et al., 2016). Therefore, there is a possibility of mutual influence and correction. However, in our study, postpartum depression is associated with an omnivorous dietary pattern in the third trimester, which unquestionably is a sequence which allows the possibility that efforts to change diet towards the end of pregnancy may be of value to mental health postpartum.

In our data, there were no significant differences in energy and three macronutrient intakes between depressed and nondepressed participants (Table 1). Therefore, we controlled variables such as pre‐pregnant BMI, weight change and physical activity during pregnancy in Model 1 to explore the influence of dietary patterns. To make our findings more insightful, we further adjusted the total energy intake in Model 2, and the direction and magnitude of the findings remained the same.

Specifically, for postpartum depression after adjusting for energy, the magnitude of the relationship between omnivorous patterns and depression becomes stronger, suggesting that energy is a negative confounding factor; whereas the situation is not seen in the third trimester of pregnancy. These findings reflect that, in the cross‐sectional study (third trimester), we were unable to establish a causal relationship between dietary pattern and depression; however, in the follow‐up study (early postpartum), the effect of diet was greater after adjusting for energy, which suggests that the effect of diet is not through energy intake but through dietary pattern.

To further understand an omnivorous pattern for the third trimester might contribute to the prevention of postpartum depression, we conducted a subgroup analysis with the EPDS cut‐off point of 10. Those with depression free (EPDS <10), although not significant, exhibited less risk for postpartum depression (Q1: ref, Q2: 0.25, Q3: 0.33, Q4: 0.22), this may be simply due to the smaller sample size. And even among those with depressive symptoms (EPDS ≥10) adopted the pattern had similar outcomes (Supporting Information: Table S4). This also strengthens our finding that diet is cause rather than effect.

4.3. Prenatal dietary pattern affects the risk of perinatal depression

The omnivorous pattern is positively correlated with almost every food item. The interpretation would be that this pattern represents a nutritious diet rather than a high energy intake (good appetite). Because the ‘Vegetables and fruits’ pattern also had a similar situation, no matter in terms of the energy value of various food items, the intake increased gradually from Q1 to Q4, but it was still not related to perinatal depression.

Our findings indicate that a dietary pattern that includes both plant‐ and animal‐derived protein‐rich foods among pregnant Taiwanese women decreases the risk of depressive symptoms. Although the average protein intake of the participants was within the recommended range and could be increased further by adherence to the high end of the omnivorous diet pattern, we cannot conclude that a high‐protein diet is a depressive symptom‐modifying diet. The diet in question is one of high nutrient density, a feature of more likely nutritional policy relevance, where there is the concomitant availability of an array of complementary nutrients, such as divalent cations (such as Fe, Ca, Zn, Mg and Cu), along with other micronutrients aside from amino acids and essential fatty acids in red/white meats and eggs, and phytonutrients in soy.

Rectifying gut microbiota dysbiosis could improve dietary optimisation for mental health through increased fruit and vegetable consumption (Zalar et al., 2018). A healthy diet contains various bioactive compounds that benefit from these pathways. Somewhat puzzling is the low, according to recommendations, fruit and vegetable intake in the omnivorous diet pattern category, itself relatively high in Q4 (Supporting Information: Table S3). It may be permissive to the remainder of the diet and favourable to mental health. Even so, the findings for omnivorous diet patterns should not be considered the best that might be achieved with diet, especially in more socioeconomically vulnerable populations or other cultural settings (Merino et al., 2021).

Omnivorous patterns have higher intakes of calcium, vitamin B complex and protein. Calcium and vitamin B complex can stabilise the nervous system, eliminate daily anxiety and improve insomnia caused by depression. The quality of protein foods is as important as their quantity. Red meat (including organ meat) was the most consumed protein food source in our study as a contributor to the omnivorous pattern, supporting previous findings suggesting that iron status is relevant to depression. Given that iron is essential for hematopoiesis, pregnant women whose iron intake is insufficient are at risk of fatigue, lethargy, lack of physical strength and depressive symptoms in any case and almost invariably take supplements, as in this study (Lee et al., 2020). Therefore, the association between dietary patterns and perinatal depression should reflect characteristics other than iron bioavailability. Moreover, Taiwanese traditionally believe in and eat pork liver to ‘nourish blood’. During pregnancy and postpartum confinement, there is evidence of the intake of organ meat (such as liver and kidney).

In a Taiwanese setting, we could not recognise an association of perinatal depression with the ‘Mediterranean‐like pattern’, unlike previous reports (Boutté et al., 2021; Huang et al., 2021). The lack of apparent protection of our ‘Mediterranean‐like pattern’ may be attributed to a low intake and limited variety of vegetables and fruits during the epidemic, making it difficult to verify the role of these two food categories in perinatal depression. However, it is increasingly recognised that the so‐called Mediterranean diet's associations with favourable outcomes are not only about the foods but also the sociocultural settings in which they are consumed (Bonaccio et al., 2022).

4.4. Collateral nutritional considerations

Although not studied directly in this investigation, there are several important areas of potential nutritional relevance to maternal and child health. These include the preconception diet with which women have entered pregnancy, the intrapartum diet in the first and second trimesters and the indirect effect of the overall food system. The observed low intake of fruits and vegetables in the third trimester may be attributable to any of these three nominated possibilities.

An illustration of the collateral nutritional considerations in our study is iodine nutrition. Taiwan's 1966 salt iodisation policy effectively reduced the prevalence of endemic goitre. However, in 2004, imported salt products were unrestricted, and most products were not iodised (Tang et al., 2014). Median urinary iodine data have showed that, for women aged 19–44, excretion was 146 μg/L during 2016–2017 (Chao et al., 2018), and more than half of women (51.5%) are lower than recommended as iodine adequate during pregnancy by WHO (150–249 μg/L). Thus, not only is iodine intake unlikely to be acceptable anyhow, perturbation in the supply chain of products that are iodine sources will put the unborn and breastfed infant at an even greater risk of iodine deficiency disorder. It is reasonable to speculate that iodine intake in pregnant women may not have been ideal. Given this situation, timely correction and guidance from dietitians may have been additionally necessary.

4.5. Strengths and limitations

Although this study obtained cross‐sectional and longitudinal observations to link late pregnancy and postpartum depression in Taiwanese context, it has some limitations. First, due to the partly cross‐sectional design, we cannot assert a cause‐and‐effect relationship between diet and depression in the third trimester. Nevertheless, this relationship might be bidirectional. Second, we did not collect dietary data preconception or in either the first or second trimesters. The FFQ used has only 33 questions, and while each question is a group rather than a single food, it may not cover all the foods respondents ate, resulting in an underestimation of food intake. Although interviewers were trained in nutrition, participants may have differing perceptions of food content and portion sizes, leading to under‐ or over‐estimation of dietary intake. However, the shorter version of the FFQ is more feasible in the clinical setting than the longer version during the data collection phase. Moreover, because it is a food group, there is no need to group the foods before performing the factor analysis (Hu et al., 2000), just as successfully derived dietary patterns relate to diet quality in a Taiwanese shift worker study (Yu‐Ru Hu et al., 2022). Third, the participants were from only one medical center in Taiwan so that the findings may not be extrapolatable to other populations. Finally, instead of a cut‐off of 13, we used the EPDS ≥10, a looser definition, to provide adequate participant sample size for those with depressive symptoms for analysis.

4.6. Clinical implications

Taiwan has one of the lowest birth rates globally; therefore, adverse pregnancy outcomes are particularly distressing. A greater emphasis on diet in the reproductive age group, during pregnancy, and postpartum has become evident, recognising that it may play a role in perinatal mental health. The question is how the diet should be optimised so that the health of the mother and child might be advantageous. Our finding that a more omnivorous diet may be beneficial for maternal mental health should be a consequence of the health outcomes of the maternal–child dyad in general.

As far as the health system is concerned, consideration must be given to the current frequency of 14 free prenatal visits offered in Taiwan and the nutritional content and extent of nutrition counselling at such visits. Where there are compounding medical complications of pregnancy, such as gestational diabetes mellitus (GDM), this need will become more evident. Regrettably, in Taiwan, where the current study was conducted, access to nutritional counselling in pregnancy is limited to where GDM is found. Nevertheless, some guidance on contact time and counselling frequency is provided from related experience at the same medical center where at least three nutritional consultations are standard practice and where it would be evident that free prenatal check‐ups would be a logical part of the programme (Chan et al., 2021; Yang et al., 2019).

5. CONCLUSIONS

A varied omnivorous diet in the third trimester may reduce perinatal depression in Taiwan as indicated by there being a prevalence of more than 30% in the third trimester and early postpartum. Therefore, early and active nutritional counselling is necessary, even though a definitive mechanism is pending.

A deeper understanding of the diet patterns contributing to perinatal depression and how they might be managed during pregnancy is required. Amelioration of the problem of depression in perinatal women through effective nutrition counselling is likely to improve the overall health and well‐being of mothers and infants, as well as the risk of complications and survival prospects.

AUTHOR CONTRIBUTIONS

All authors contributed to the conception of the paper. Lin‐Chien Chan and Meei‐Shyuan Lee contributed to the conception and design of the study. Hsiu‐Hui Wang and Cheng‐Chieh Liu contributed to the acquisition and analysis of the data. Mark L. Wahlqvist and Jah‐Yao Liu contributed to the interpretation of the data. Lin‐Chien Chan, Meei‐Shyuan Lee and Mark L. Wahlqvist draughted the manuscript. All authors critically revised the manuscript, agree to be fully accountable for ensuring the integrity and accuracy of the work and read and approved the final manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

ETHICS STATEMENT

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Human Research Ethics Committee (HREC) of Tri‐Service General Hospital at the National Defense Medical Center, Taiwan (C202005096). Informed consent was obtained from all subjects involved in the study.

Supporting information

Supporting information.

Chan, L.‐C. , Wang, H.‐H. , Wahlqvist, M. L. , Liu, C.‐C. , Liu, J.‐Y. , & Lee, M.‐S. (2024). Perinatal dietary patterns and symptomatic depression: A prospective cohort study. Maternal & Child Nutrition, 20, e13561. 10.1111/mcn.13561

DATA AVAILABILITY STATEMENT

Data described in the manuscript will not be made available because the participants of this study did not agree for their data to be shared publicly. Codebook and analytic code will be made available upon request.