*Article* **Effects of Climate, Sun Exposure, and Dietary Intake on Vitamin D Concentrations in Pregnant Women: A Population-Based Study**

**Ya-Li Huang 1,2, Thu T. M. Pham 2,3, Yi-Chun Chen 4, Jung-Su Chang 4,5,6,7, Jane C.-J. Chao 4,5,8 and Chyi-Huey Bai 1,2,5,\***


**Abstract:** Background: Vitamin D deficiency (VDD) is a global micronutrient issue that commonly occurs in pregnant women, leading to adverse health outcomes. We examined the role of sunlight-related factors and dietary vitamin D intake on vitamin D concentrations among pregnant women in different climate zones. Methods: We conducted a nationwide cross-sectional survey in Taiwan between June 2017 and February 2019. The data of 1502 pregnant women were collected, including sociodemographic information and characteristics related to pregnancy, diet, and sun exposure. Serum 25-hydroxyvitamin D concentrations were measured, and VDD was assessed as a concentration of less than 20 ng/mL. Logistic regression analyses were used to explore the factors associated with VDD. Furthermore, the area under the receiver operating characteristic (AUROC) curve was used to analyze the contribution of sunlight-related factors and dietary vitamin D intake to vitamin D status stratified by climate zones. Results: The prevalence of VDD was 30.1% and was the highest in the north. Sufficient intake of red meat (odds ratio (OR): 0.50, 95% confidence interval (CI): 0.32–0.75; *p* = 0.002), vitamin D and/or calcium supplements (OR: 0.51, 95% CI: 0.39–0.66; *p* < 0.001), sun exposure (OR: 0.75, 95% CI: 0.57–0.98; *p* = 0.034), and blood draw during sunny months (OR: 0.59, 95% CI: 0.46–0.77; *p* < 0.001) were associated with a lower likelihood of VDD. Additionally, in northern Taiwan, which is characterized by a subtropical climate, dietary vitamin D intake (AUROC: 0.580, 95% CI: 0.528–0.633) had a greater influence on vitamin D status than did sunlight-related factors (AUROC: 0.536, 95% CI: 0.508–0.589) with a *z* value = 51.98, *p* < 0.001. By contrast, sunlight-related factors (AUROC: 0.659, 95% CI: 0.618–0.700) were more important than dietary vitamin D intake (AUROC: 0.617, 95% CI, 0.575–0.660) among women living in tropical areas of Taiwan (*z* value = 54.02, *p* < 0.001). Conclusions: Dietary vitamin D intake was essential to alleviate VDD in the tropical region, whereas sunlight-related factors played a greater role in subtropical areas. Safe sunlight exposure and adequate dietary vitamin D intake should be promoted appropriately as a strategic healthcare program.

**Keywords:** 25-hydroxyvitamin D [25(OH)D] concentration; diet; pregnant women; sunlight; Taiwan; vitamin D deficiency

**Citation:** Huang, Y.-L.; Pham, T.T.M.; Chen, Y.-C.; Chang, J.-S.; Chao, J.C.-J.; Bai, C.-H. Effects of Climate, Sun Exposure, and Dietary Intake on Vitamin D Concentrations in Pregnant Women: A Population-Based Study. *Nutrients* **2023**, *15*, 1182. https://doi.org/ 10.3390/nu15051182

Academic Editors: Louise Brough and Gail Rees

Received: 1 February 2023 Revised: 23 February 2023 Accepted: 24 February 2023 Published: 27 February 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Introduction**

Vitamin D deficiency (VDD) has become an urgent micronutrient issue globally [1] because of its high prevalence [2], and it has become a potential cause of non-communicable [3,4] and infectious [5,6] diseases. Although VDD has been addressed as a global public health problem in all age groups, the population-representative data regarding vitamin D were limited to several risky groups [7]. Pregnant women are a vulnerable population affected by VDD [1], which can lead to adverse pregnancy outcomes [8,9]. Moreover, VDD may result in health disparities [10], which leads to the increment of stillbirths and pregnancy-related deaths [11]. Hence, improving vitamin D status is necessary to upgrade the reproductive health and well-being of mothers and their infants.

The major factors for VDD are sun exposure and dietary vitamin D intake [12]. However, obtaining vitamin D through sun exposure can be inefficient or unsafe because of the skin cancer risk from ultraviolet radiation [13]. Additionally, the dermal synthesis of vitamin D was suggested to be influenced in different climate zones using an in vitro model [14]. The adequate achievement of vitamin D intake from diet alone is hard [15]. Therefore, vitamin D supplementation is a crucial nutritional priority recommended by many physicians to achieve optimal serum concentration [16] that could prevent short and long-term maternal and infant health complications [17].

Vitamin D status has been explored in the literature. However, population-based research on pregnant women in East Asia is still limited. To our best knowledge, relevant information regarding the potential effect of the climatic zone has not been explored. Taiwan is an East Asian island characterized by two climatic zones [18]. Based on this unique advantage, Taiwan has the opportunity to assess whether sunlight-related factors and dietary vitamin intake contribute differently to vitamin D levels among people living in different parts of the country. Exploring the prevalence of VDD and its potential risk factors among pregnant women in Taiwan is an important task to address the research gap and for future policy planning. This study aimed to assess the determinants of VDD and to examine the contribution of sunlight-related factors and dietary vitamin D intake to vitamin D status in different regions of Taiwan using a nationally representative survey.

#### **2. Materials and Methods**

#### *Study Population*

A national cross-sectional nutritional survey of pregnant women was conducted from June 2017 to February 2019 across Taiwan. A multiple-stage cluster sampling approach was used, including (1) the selection of eight layers according to geographical location (northern, central, southern, and eastern Taiwan) and (2) the random selection of hospitals (large and small sizes) from the list based on the number of women availing pregnancy-related services per year and the probability proportional to size in each layer and (3) the whole selection of participants arriving in the selected hospitals for antenatal examination with the expectation of 150–300 women from one or two hospitals in each layer enrolled based on the potential number of annual outpatients in each hospital [19]. The distribution of eleven selected hospitals across Taiwan was in Figure 1.

We calculated a sample size of 1062 based on 200,000 deliveries by pregnant women during the study period, with a 3% margin of error and a 95% confidence interval (CI). We recruited participants aged ≥15 years who were legal residents of Taiwan and who underwent antenatal examinations at the selected hospitals. A satisfactory sample of 1502 pregnant women was included in the final analysis after the exclusion of nonsingleton pregnancies, participants unable to understand and speak Mandarin, and incomplete questionnaires. All participants provided written informed consent before taking the survey.

**Figure 1.** The distribution of eleven selected hospitals across Taiwan.

#### **3. Data Collection**

During study periods, all pregnant women making an antenatal visit were enrolled consecutively. At recruitment, collection of questionnaires, physical examination and blood sample were performed. Information was obtained from standardized face-to-face interviews by trained interviewers using the structured questionnaires. Variables regarding participants' sociodemographic status, histories of diseases before and during pregnancy, pregnancy-related factors, and intake histories of prenatal and natal dietary supplements were collected by the self-reported baseline questionnaire. The dosage of supplements during pregnancy was asked and recorded in brand, exact dosage and frequency per week. Food frequency questionnaires was also used to record the intake frequency during past 3 months in 66 items of foods including egg, milk, meat, fish and vegetables. After interview of questionnaires, a 24 h dietary recall was recorded by trained dietitians. Food models were used to assist participants in recalling the food portion sizes and details of the dietary information. Then, we estimated participants' energy intake and nutrient intake from foods. The intakes of several nutrients (e.g., vitamin D) were labeled the sources of foods or supplements respectively. We used the online software Cofit Pro (Cofit Health Care, Taipei, Taiwan) to analyze participants' nutrient intake using the 2015 version of the Taiwan Food and Nutrient Database.

At the time of recruitment, pre-pregnancy body weight was self-reported by pregnant women, and their current body height and weight were measured. Blood samples were drawn, centrifuged, then froze (−80 ◦C) and analyzed in batches.

#### *3.1. Sociodemographic and Pregnancy-Related Characteristics*

Pregnant women were queried regarding their age (years); residential area; education level; household monthly income; religion; gravidity; parity; number of fetuses in the current pregnancy; gestational age; and body height (cm) and weight (kg) before pregnancy, which were used to calculate pre-pregnancy body mass index (BMI, kg/m2). Additional information related to pregnancy was extracted from the prenatal visit records of participants. The residence was categorized as living in Taiwan's northern, central, southern, or eastern regions.

#### *3.2. Dietary Characteristics*

Pregnant women were asked whether they consumed sufficient amounts of the four groups of the following food items: (1) dairy products (e.g., fresh milk, yogurt, cheese, cream cheese, and powdered milk); (2) eggs; (3) red meat (e.g., pork, beef, and mutton); and (4) nut fruits (e.g., stone fruit, nuts, pistachios, and almonds). Women also reported their frequency of using vitamin D and/or calcium supplements during pregnancy as "never", "less than 1 day per week", "2–5 days/week", and "almost daily". Then, this factor was recoded into two categories of usage, "yes" or "no", due to the small sample

size. The 24 h dietary intake was recorded to assess the intake of total energy (kcal), raw protein (g), raw fat (g), total carbohydrates (g), and vitamin D content (mg) and the use of vitamin supplements. The percentages of calories from protein, fat, and carbohydrates were also calculated [19].

The dosages of supplements were calculated if participants provided the exact dosage. However, these parameters were frequently missing, as were the brands and models of vitamins. Therefore, in the present study, we only analyzed the usage frequency of vitamin D-only or D-based supplements.

#### *3.3. Sunshine-Related Factors*

Sun exposure was estimated using the question, "Were you exposed to outdoor sunlight last month?" and the answers were categorized as "no" if exposed to sunlight for less than 10 min per day and "yes" if exposed to sunlight for more than 10 min per day. The seasons of blood draw were categorized according to the month of blood sample collection, as follows: sunny months (June to November) and rainy months (December to May) established according to the rainfall report of the Central Weather Bureau, Taiwan. Participants also reported whether they had to stay indoors (e.g., bedridden) for any reason during their pregnancy ("yes" or "no" response) and the number of methods used for sun protection (e.g., sunscreen, parasols, hats and outerwear with UV-block) and how often they are used.

#### *3.4. Vitamin D Deficiency Assessment*

As 25-hydroxyvitamin D [25(OH)D] has the long half-life (15 days) and relative stability of concentration in the blood [20], the circulating 25(OH)D is the useful biomarker of vitamin D in the human body [21]. The plasma 25-hydroxyvitamin D [25(OH)D] concentration was measured using an electrochemiluminescence immunoassay, as described previously [19]. Although there is no consensus in the definition of the suboptimal vitamin D level, VDD was defined as a 25(OH)D level of <20 ng/mL, which is a common threshold for people in at-risk groups, including pregnant women [22–24]. The cutoff point of less than 20 ng/mL was also recommended for use for VDD by Institution of Medicine, Academy of Medicine and American Academy of Pediatrics.

#### **4. Ethical Consideration**

This study was funded by the Health Promotion Administration, Ministry of Health and Welfare in Taiwan (C1050912) and was approved by the institutional review board of the government and selected hospitals (IRB number: N201707039).

#### **5. Statistical Analysis**

First, descriptive analysis was performed to explore the distribution of independent variables. We performed chi-square tests (for categorical variables) and *t* tests or Mann– Whitney tests (for continuous variables) to compare the distribution of independent variables between pregnant women with and without VDD. Second, logistic regression analysis was used to determine the factors associated with VDD. Two models were constructed. Model 1 comprised variables associated with VDD that had *p* < 0.1 in bivariate analysis, including age, residential area, parity, gestational age, pre-pregnancy BMI, egg intake, red meat intake, fat, vitamin D content, vitamin supplements, sun exposure, remaining indoors during pregnancy, and the season of blood draw. Gravidity and carbohydrate intake were removed from model 1 because they were highly correlated with parity (*rho* = 0.82) and fat intake (*rho* = −0.89), respectively (Table S1). Model 2 comprised factors associated with VDD that had *p* < 0.1 in model 1, including age, residential area, gestational age, red meat intake, vitamin D content, vitamin supplements, sun exposure, remaining indoors during pregnancy, and the season of blood draw. Odds ratios (ORs) and 95% CIs were reported, and *p* < 0.05 was considered statistically significant.

Further sensitivity analysis was performed and stratified by residential area (north vs. south and other regions) to examine the contribution of modifiable factors to vitamin D

status. Two models were constructed for each layer, including one model adjusted for sunlight-related factors (season of blood draw and sun exposure) and one model adjusted for dietary vitamin D intake (red meat and supplements). The area under the receiver operating characteristic (AUROC) curve was computed to compare the models. It is favored due to the characteristics of invariant and independent from the prevalence of the condition. All analyses were performed using R software (version 4.1.3; R Foundation for Statistical Computing, Vienna, Austria).

#### **6. Results**

#### *6.1. Characteristics of Study Participants*

The data contained several missing values, but the distribution of variables before and after removing the missing information was the same. Therefore, the entire data of the 1502 pregnant women were used for analysis. Overall, the mean 25(OH)D concentration was 25.5 ± 8.9 ng/mL, and the prevalence of VDD was 30.1% (weighted). Compared with women without VDD, those with VDD were younger (*p* = 0.017); lived in the north (*p* < 0.001); had uniparity (*p* = 0.01); were in the first trimester of gestation (*p* < 0.001); consumed high quantities of carbohydrates (*p* = 0.013) but insufficient eggs (*p* = 0.034), red meat (*p* < 0.001), fat (*p* = 0.023), and vitamin D and/or calcium supplements (*p* < 0.001); had little sun exposure (*p* = 0.001); remained indoors during pregnancy (*p* = 0.018); and had blood drawn during the rainy months (*p* = 0.004). These data are displayed in (Table 1).




#### **Table 1.** *Cont.*

Abbreviations: BMI, body mass index; IQR, interquartile range; NT\$, New Taiwan dollar; SD, standard deviation; VDD, vitamin D deficiency. \* Variables containing missingness of ≤0.6%, with the exception of remaining indoors during pregnancy, number of fetuses in this pregnancy, household monthly income, and vitamin supplements, which have 0.9%, 1.1%, 1.9%, and 2.5% missingness, respectively.

#### *6.2. Associated Factors of Vitamin D Deficiency*

As displayed in Table 2, the likelihood of VDD was significantly lower in pregnant women who were older (OR: 0.95, *p* < 0.001); lived in central (OR: 0.66, *p* = 0.010), southern, or eastern Taiwan (OR: 0.20, *p* < 0.001) or in the eastern and outlying islands (OR: 0.33, *p* < 0.001); were in the second trimester (OR: 0.72, *p* = 0.046) or the third trimester (OR: 0.60, *p* = 0.002); consumed sufficient red meat (OR: 0.50, *p* = 0.002); took vitamin D and/or calcium supplements (OR: 0.51, *p* < 0.001); received sun exposure (OR: 0.75, *p* = 0.034); and had blood drawn during the sunny months (OR: 0.59, *p* < 0.001).

In the sensitivity analysis, among participants living in northern Taiwan, dietary vitamin D intake (AUROC: 0.580, 95% CI: 0.528–0.633) had a greater influence on vitamin D status than did sunlight-related factors (AUROC: 0.536, 95% CI: 0.508–0.589). By contrast, among participants living in the south and other parts of Taiwan, sunlight-related factors (AUROC: 0.659, 95% CI: 0.618–0.700) were more influential than dietary vitamin D intake (AUROC: 0.617, 95% CI: 0.575–0.660). The differences in regional models were significant, with *z* value = 51.98, *p* < 0.001 for northern Taiwan and *z* value = 54.02, *p* < 0.001 for the remaining regions. These results are visualized in Figure 2.


**Table 2.** Factors associated with vitamin D deficiency via multiple logistic regression analysis models (*n*= 1502).

Abbreviations: BMI, body mass index; CI, confidence interval; NT\$, New Taiwan dollar; OR, odds ratio.

**Figure 2.** Contribution of sunlight-related factors and dietary vitamin D intake to vitamin D status in different regions of Taiwan.

#### **7. Discussion**

In the present study, the prevalence of 25(OH)D level < 20 ng/mL among pregnant women in Taiwan was 30.1% (weighted). The determinants of VDD included age, gestational age, red meat intake, vitamin D and/or calcium supplements, residential area, sun exposure, and the season of blood draw.

The occurrence of VDD [25(OH)D < 20 ng/mL] is common in pregnant women, although the rates vary in different Asian countries, ranging from 7% to 40.7% [25,26]. The present study found that VDD occurred more frequently in pregnant women living in northern Taiwan than in those living in southern Taiwan. A nationwide report on VDD among older adults (a risk group of VDD) had similar findings, reporting that VDD occurrence was higher in the north than in the south [27]. This phenomenon has several possible explanations. First, northern Taiwan has a higher latitude than other regions [28], and vitamin D status decrease with increasing latitudes [29]. Second, northern Taiwan has a humid subtropical climate, and sunlight may be of lower intensity than that in southern Taiwan and other regions characterized by a tropical monsoon climate. The association between age and VDD was found in the previous studies with the controversial findings. The former authors showed that age over thirty was the risk factor for VDD among pregnant women [26]. However, the current study indicated that younger age was a contributing factor for VDD, which was in line with other studies [30,31]. Our findings could be due to the habits of avoiding sunlight among almost youngers that they were likely to apply sun protection (e.g., using sunscreen, wearing long-sleeved clothes, preferring indoor activities). Thus, our findings indicate that it is worth planning VDD prevention, such as educating health literacy related to VDD and lifestyle changes in younger women, and such methods should be promoted integrating with efficient intervention strategies.

Regarding the impact of gestational age on maternal VDD, the findings are inconsistent across studies. Although several studies have reported that vitamin D status decreased during advanced gestation [32], our results are in line with those of studies reporting that the likelihood of VDD was reduced during the second and third trimesters. For example, Perreault et al. indicated that serum 25(OH)D concentrations were significantly greater in the last trimester compared to the first trimester [33]. Similarly, Savard et al. found that serum 25(OH)D levels significantly increased across trimesters [34]. In addition, Shen et al. noted a positive relationship between the increased vitamin D concentration and later gestational week [35].

It has been well established that sunlight is the main source of vitamin D. Hence, sun exposure and the summer season are the most important contributing factors to the vitamin D concentration. Nevertheless, if sun exposure cannot provide sufficient vitamin D because of factors such as sunlight intensity, time of exposure, and application of sun protection, the vitamin D status in the human body can be adjusted through nutrition and dietary intake. In the literature, the natural vitamin D content in foodstuffs is usually limited to vitamin D3 from animal products [36]. Our findings indicated that the consumption of red meat was associated with lower VDD rates. Moreover, the present study demonstrated that vitamin D and/or calcium supplements could reduce the likelihood of VDD.

In our sensitivity analysis, the effects of sunlight-related factors and dietary vitamin D intake on 25(OH)D levels varied by region. In northern Taiwan, dietary vitamin D intake was more important than sunlight-related factors for improving maternal vitamin D status; however, sunlight-related factors were the main sources of vitamin D for pregnant women living in the south and other parts of Taiwan, and vitamin D intake played a minor role. These variations in effectiveness corresponded to the variations in climate across Taiwan. These findings can assist health policymakers in designing regional strategies for the prevention of prenatal VDD.

To date, suboptimal vitamin D levels is mostly indicated for bone health but remain controversial across populations and countries. For some investigators, deficiency was defined as specific to bone; however, insufficiency was defined relating to other health outcomes. For others, deficiency covered diseased population and insufficiency covered at-risk population. One of the most commonly used definitions comes from the Endocrine Society Clinical Practice Guidelines [24]; vitamin D deficiency was defined as 25(OH)D values below 20 ng/mL (50 nmol/L), and vitamin D insufficiency was defined as 25(OH)D of 21–29 ng/mL (52.5–72.5 nmol/L). This guideline was accepted and used widely by the International Osteoporosis Foundation, American Association for Clinical Endocrinologists, Institute of Medicine, American Academy of Pediatrics, and government of Australia, New Zealand, Germany, Austria and Switzerland as well as in Taiwan. In any case, cut point is very important when looking at the results in 25(OH)D level.

Particularly in older adults, having a higher BMI or body fat percentage are significant subject-specific characteristics that negatively affect vitamin D metabolism [37]. Normalweight women reached the higher 25(OH)d level after vitamin D supplementation faster than women with obesity [38]. However, in pregnant women, the association between BMI and VDD was not consistent across the studies. While several studies showed that high BMI was associated with VDD, others showed that BMI was not statistically significantly associated with VDD [39,40]. Obesity is strongly associated with insufficient dietary vitamin D intake and low sun exposure. Pre-pregnancy obesity predicts poor vitamin D status in mothers [41]. In our study, pre-pregnancy BMI (as a continuous variable) was significantly different in two groups of VDD and non-VDD, but in logistic regression, after adjusting for confounders, pre-pregnancy BMI was not significantly associated with VDD. The findings for BMI (as a categorical variable) were also insignificant in multiple logistic regression. Obesity is not associated with 25(OH)D levels in our study.

The present study is the first national report on vitamin D status among pregnant women in Taiwan. Our findings demonstrated specific differences in the effects of sunlightrelated factors and vitamin D intake on vitamin D concentrations in distinct regions of Taiwan. However, several limitations should be considered. First, because this was a cross-sectional study, we can only note associations; we cannot determine the causal relationship. Second, several factors influencing vitamin D status were not assessed in our study, such as occupation and the brand and dose of supplements. Third, we used a selfreport questionnaire, which may introduce assessment bias because of subjective responses. Fourth, although the present study highlights the critical role of dietary vitamin D intake, the data on nutrient quantitation per serving are unavailable.

#### **8. Conclusions**

VDD was prevalent in pregnant women in Taiwan. On the basis of our findings, we recommend the promotion of a robust health policy regarding safe sunlight exposure and effective dietary vitamin D intake, with adjustments according to the characteristics of various climate zones. In doing so, clinicians can enhance maternal vitamin D status, reduce the VDD-induced burden, and improve health and well-being.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/nu15051182/s1, Table S1: Spearman's correlations among the studied variables (*n* = 1502).

**Author Contributions:** Y.-L.H., T.T.M.P., Y.-C.C., J.-S.C., J.C.-J.C. and C.-H.B.: conceptualization, methodology, validation, investigation, data curation, and writing review and editing draft. Y.-L.H., T.T.M.P. and C.-H.B.: formal analysis and writing—original draft. Y.-L.H., J.C.-J.C. and C.-H.B.: project administration. C.-H.B.: supervision and funding acquisition. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Health Promotion Administration, Ministry of Health and Welfare in Taiwan (grant number MOHW107-HPA-H-124-133204) and was partially supported by a grant from the Collaborative Research Project of College of Medicine, Taipei Medical University, Taiwan. The content of this research may not represent the opinions of the Health Promotion Administration, Ministry of Health and Welfare, Taiwan.

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of the government and selected hospitals (IRB number: N201707039).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors acknowledge all the participants and institutions for their contributions to the Nutritional Survey of Pregnant women in Taiwan.

**Conflicts of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

## *Article* **Self-Reported Intake and Circulating EPA and DHA Concentrations in US Pregnant Women**

**Keri Lanier 1, Breanna Wisseman 2, Cody Strom 3, Carol A. Johnston 4, Christy Isler 5, James DeVente 5, Edward Newton 5, Roman Pawlak 6, Brittany R. Allman-Tucker 7, Samantha McDonald <sup>8</sup> and Linda E. May 2,5,9,\***


**Abstract:** In the United States, pregnant women have low concentrations of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA), which are essential for fetal development. Although maternal blood provides accurate polyunsaturated fatty acid (PUFA) concentrations, venipuncture is expensive and not always accessible. PUFA-containing foods consumption, both omega-3 ad omega-6 is supposed to reflect in the status (plasma, RBC, adipose tissue) of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA). De novo synthesis of DHA and EPA during pregnancy is supposed to be higher compared to pre and/or post-pregnancy periods. Thus, this study aimed to determine the association between maternal self-reported dietary intake of foods high in DHA and EPA, along with vegetable oils as a source of omega-6 fatty acids, with maternal blood DHA and EPA concentrations. Pregnant women (13–16 weeks gestation) were recruited and asked to complete a food-frequency questionnaire (FFQ) and blood draw at enrollment and 36 weeks. Circulating concentrations of DHA and EPA were quantified and change scores were calculated. Correlations were done to determine associations between FFQ results and EPA/DHA maternal blood concentrations. Regression analyses were run to examine significant predictors of the main outcomes. Overall, PUFA-food consumption and RBC's DHA levels decreased from early to late pregnancy; self-reported PUFA-rich food consumption positively correlated with DHA and EPA levels. DHA concentration was predicted by self-reported PUFA-rich oils (sunflower/soy/corn/olive) consumption, but EPA concentration was predicted by maternal BMI. These findings suggest that EPA and DHA consumption decreased across pregnancy and the FFQ can be utilized as an effective method for estimating PUFA blood concentration during pregnancy.

**Keywords:** dietary assessment; food intake; polyunsaturated fatty acid; DHA; EPA; pregnancy

#### **1. Introduction**

In the United States (US), pregnant women usually have low ratios of omega-3 fatty acids to omega-6 fatty acids, due to a Western diet that prioritizes red meats, chicken, and corn oil, which exceeds the suggested omega-3s to omega-6s ratio of 1:4 up to 1:15 [1–3]. A diet high in cold-water fish, algae, and low intake of omega-6 fatty acids can help maintain the minimum suggested ratio, i.e., 1:4, of polyunsaturated fatty acids (PUFAs). This type of diet, high in omega-3 fatty acids, such as docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA), is important for the nervous system and health [4,5]. DHA and EPA

**Citation:** Lanier, K.; Wisseman, B.; Strom, C.; Johnston, C.A.; Isler, C.; DeVente, J.; Newton, E.; Pawlak, R.; Allman-Tucker, B.R.; McDonald, S.; et al. Self-Reported Intake and Circulating EPA and DHA Concentrations in US Pregnant Women. *Nutrients* **2023**, *15*, 1753. https://doi.org/10.3390/ nu15071753

Academic Editors: Leanne M. Redman, Louise Brough and Gail Rees

Received: 28 January 2023 Revised: 30 March 2023 Accepted: 3 April 2023 Published: 4 April 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

play a critical role in fetal development, especially the fetal nervous system [4]. These PUFAs influence fetal brain development as well as inflammatory properties throughout the body [4]. DHA, in particular, is important for developing neuronal connections, neurogenesis, and protection from oxidative stress in utero [4]. For this reason, it is important to be able to accurately measure whole-body PUFA levels in pregnant women.

Previously, food-frequency questionnaires (FFQs) have been validated in Chinese men and women as well as in Australian women in late pregnancy [6,7]. Similarly, in Japan, there has been some utility established for assessing self-reported DHA and EPA via questionnaire, in early and late pregnancy [8]. However, these studies utilized only red blood cells (RBCs) in their analyses, which can only provide an assessment of long-term PUFA consumption habits [9,10]; whereas plasma levels provide a short-term assessment of PUFA levels and may more closely mimic PUFA food intake. However, this assessment has not been done previously in US women. Furthermore, studies in other countries varied in timepoints of assessments during pregnancy; and/or utilized a FFQ that was not inclusive of PUFA-containing foods. Foods such as fish, sunflower/soy/corn/olive oils, and almond/cashew milk contribute to PUFA consumption [11,12]. DHA and EPA are important fatty acids that play an integral role in fetal neurological development; therefore, it is imperative that healthcare providers are aware of maternal PUFA intake. While venipuncture sampling is a practical method for assessing maternal DHA and EPA blood concentrations, this process is invasive, costly, and time-consuming. With fetal brain development beginning in early pregnancy, it is important to have a rapid method for estimating maternal PUFAs, allowing for early intervention if levels are too low [13]. FFQs are non-invasive, easy to distribute and understand, and provide a rapid assessment of maternal food intake, thus, making them a low-cost, clinic- and patientfriendly alternative to venipuncture blood sampling. While previous research has been done in other populations, there is no literature validating a FFQ with DHA and EPA levels in the United States during early and late pregnancy. Therefore, the purpose of the present study was twofold: (1) to measure DHA and EPA levels in RBC and plasma in early and late pregnancy, and (2) to determine the association, and possible predictors, between self-reported consumption of PUFA-containing foods with DHA and EPA concentrations in maternal RBC and plasma in early and late pregnancy. We hypothesize that: (1) PUFArich foods, DHA, and EPA levels in plasma and RBCs will be similar in early and late pregnancy, and (2) there will be a positive correlation, and possible predictors, between selfreported PUFA-rich food consumption and circulating PUFA plasma, but not necessarily RBC, concentrations.

#### **2. Methods**

#### *2.1. Study Design and Participants*

The present study was a post hoc secondary analysis of a larger prospective randomized controlled trial designed to examine the influence of maternal exercise during pregnancy on fetal and infant health outcomes [14]. Participants were enrolled if they were 18–40 years old, able to communicate in English, ≤16 weeks pregnant, had a pre-pregnancy BMI of 18.5–39.9 kg/m2, and had a singleton pregnancy. All women were required to receive written clearance from an obstetric provider to participate in the study. Participants were excluded from the study if they had pre-existing diabetes mellitus, hypertension, cardiovascular disease, co-morbidities known to affect fetal growth and well-being (e.g., systemic lupus erythematosus), or used tobacco, alcohol, and illicit drugs. All protocols were approved by the East Carolina University Institutional Review Board. Clinical Trial Registry is #NCT03517293. Written informed consent was obtained from each participant.

Pre-screening eligibility questionnaires and neonatal electronic health records were used to determine maternal age, gravida, parity, pre-pregnancy weight and height, education level, gestational weight gain (GWG), and gestational age (weeks). Height was measured using a stadiometer and weight was collected using a standard scale at 16 and 36 weeks gestation. Pregnancy weight was assessed at the same time points

using a calibrated digital scale. Pre-pregnancy weight was self-reported at enrollment (≤16 weeks). A standardized equation was used to calculate BMI at each time point [15]: BMI = ((weight (kg)) ÷ ([height (m2)])); BMI classification used standard cutoffs: normal weight: 18.5–24.9 kg/m2; overweight 25–29.99 kg/m2; obese ≥ 30 kg/m2.

#### *2.2. Maternal Food-Frequency Questionnaire*

Participants (N = 47) were asked to complete a food-frequency questionnaire (FFQ) at enrollment (13–16 weeks) and 36 weeks gestation to obtain self-reported PUFA levels. The FFQ asked women to specifically report foods, such as PUFA-rich foods, consumed during pregnancy [16]. The women were asked to report the frequency of consumption of foods based on the scale: 1—rarely or never eat the food, 2—eat the food once every 2 weeks, 3—eat the food 1–3 times/week, 4—eat the food 4–7 times/week, or 5—eat the food more than once per day. The individual PUFA-rich foods (white-flesh fish, other fish (e.g., salmon), almonds) were rated on the 5-point Likert scale [11,12]. Both the polyunsaturated margarines and sunflower/soy/corn/olive oils were rated on "Yes, you consume" or "No"; these two dichotomous measures were converted to No = 0 and Yes = 1 for analysis. All individual PUFA-rich food column numerical values were then summed for a PUFA summary score for each participant during pregnancy.

#### *2.3. Maternal Plasma and RBC Collection and Analysis*

A fasting venous blood sample was collected from women at enrollment (13–16 weeks) and 36 weeks gestation. All samples were completed following a ≥8 h fast and collected between 6–9 a.m. Blood was centrifuged and stored using standard procedures as described previously [14]. Both blood plasma and RBCs were utilized for analysis as plasma provides a representation of recent concentrations and RBCs provide longer term (~120 days) concentrations [9,10].

#### *2.4. Maternal DHA and EPA*

Chemicals and reagents: Optima grade acetonitrile, water, formic acid, methanol, and isopropanol were purchased from Fisher Scientific (Hampton, NH, USA).

Preparation of calibration and quality control standards: Working stock solutions were prepared for calibrators. Samples were screened for quality control (QC). Calibration curves were generated from 0.01–7.5 mg/mL. A positive cutoff limit was established at 10 mg/mL. Low and high QC samples were prepared by the addition of 10 or 500 ng/mL and were fortified as a QC solution.

Targeted LC/MS: An Agilent Poroshell (Agilent Technologies, Santa Clara, CA, USA) 120 EC-C8, 3 × 100 mm 2.7 μm column was used for separation of the analytes on an Exion HPLC. The column temperature was maintained at 32 ◦C. A gradient was used to separate the compounds using mobile phase A: 95:5 water with 0.1% formic acid:acetonitrile and mobile phase B: acetonitrile. A linear gradient was performed as follows: 0% B for 2 min, 90% B for 9 min, 90% B for 1 min, 0% B for 1 min, hold at 0% B for 5 min for a total run time of 18 min. The flow rate was 0.3 mL/min and 5 μL of sample was injected. MS-MS analysis was conducted using an AB SCIEX 3200 (Danaher Corporation, Toronto, ON, Canada) triple quadrupole mass spectrometer. The mass spectrometer was in negative ionization mode and analysis was conducted using multiple reaction monitoring (MRM). The source parameters were set to a curtain gas 50 psi, heater gas 50 psi, ion spray voltage 5500 V, and source temperature 500 ◦C. The instrument parameters were optimized using direct infusion of each analyte using a split tee injection with the LC flow. SCIEX Analyst software (v.1.6.2—Sciex Applied Biosystems, Framingham, MA, USA) was used for instrument control. Confirmation analysis was performed using MultiQuant where the calibrators and quality controls were carried through the same processes as the specimens being tested. Least squared regression with 1/x weighing was used to evaluate the linearity with adequate compensation for heteroscedasticity during all experiments.

#### 2.4.1. Solid Phase Extraction (SPE)

DHA and EPA were extracted from RBCs [17,18]. Plasma samples were prepared following a similar method. Plasma samples were prepared utilizing a 3.9:1 Optima grade H2O (Fisher Scientific, Hampton, NH, USA) to plasma solution, were vortexed, and homogenized. Aliquots of 490 mL of plasma solution were diluted to a 1 mL solution with 500 mL of methanol (MeOH) and 10 mL deuterated DHA (DHA-d5) and EPA (EPA-d5) internal standard solution. Immediately following solution preparation, both the RBC and plasma solutions were centrifuged at 13.2 rpm for 20 min. Strata-X reversed-phase SPE columns (Phenomenex, Torrance, CA, USA) and positive pressures (1 to 25 psi) were used to extract the supernatants on a Biotage Pressure+ manifold (Biotage, Charlotte, NC, USA). Columns were conditioned with 1 mL of MeOH and equilibrated with 2 mL of H2O. Supernatants were loaded onto the conditioned columns and were washed with 1 mL of 10:90 MeOH:H2O. The organic fraction of metabolites was collected by loading 1 mL of MeOH and 1 mL of 60:20:20 Acetonitrile(ACN):MeOH:IPA in duplicate, then evaporated using a steady flow of nitrogen gas and heat of 40 ◦C. Samples were reconstituted in 100 mL of 50:50:0.01 H2O:MeOH:formic acid and the solution was transferred into 100 μL autosampler vials for analysis on an AB SCIEX 3200 triple quadrupole mass spectrometer. A processed blank was extracted using the same method. All samples were stored and run in batches.

#### 2.4.2. Calibration Curve

A calibration curve was used to quantify the analytes. Stock solutions were prepared in ethanol with DHA and EPA standards, each at a concentration of 25 mg/mL. DHA and EPA standard solutions were prepared by serial dilution of the stock solutions in ACN to create primary standards at 0.01, 0.05, 0.1, 0.5, 1, 2.5, 5.0, and 7.5 mg/mL. Deuterated DHA and EPA were used as internal standards prepared at 0.5 mg/mL in ethanol. The deuterated DHA and EPA solution controlled for extraction recovery, injection of the mass spectrometer, and ionization variability. The stock solutions were processed and extracted using the same method of the plasma solution.

#### 2.4.3. Liquid Chromatography/Mass Spectrometry (LC/MS/MS)

Extracted samples were run on an AB SCIEX 3200 triple quadrupole mass spectrometer in negative ionization mode using previously published methods [17–19]. Change values for DHA and EPA were calculated between timepoints by subtracting 16-week concentrations from 36-week concentrations.

#### *2.5. Statistical Analyses*

Summary statistics were run for maternal descriptors, PUFA-rich foods from the FFQ, as well as levels from maternal blood. For FFQ, data was converted into the average number of times consumed per week as such: rarely or never eat the food = 0, eat the food once every 2 weeks = 0.5 times per week, eat the food 1–3 times/week = 2 times per week, eat the food 4–7 times/week = 5.5 times per week, or eat the food more than once per day = 7 times per week. Both the polyunsaturated margarines and sunflower/soy/corn/olive oils were rated on for whether they were used with foods (i.e., breads, vegetables) and with cooking. For both margarine and oil responses: Yes = 7 per week, No consumption = 0 per week; thus, the potential score for margarine and oils ranged from 0 to 14 considering use with food and with cooking to both questions. The summation of these columns provided the PUFA summary score for each participant during pregnancy. Data are reported as mean ± standard deviation (SD) unless data was not non-normally distributed then median (minimum, maximum) were reported. Difference values were determined by subtracting the 16-week value from the 36-week value for maternal lipid levels. Based on difference values (16 to 36-week change scores), all participants were coded as improved (increased score), no change, or decreased score for all DHA, EPA, and FFQ summary data. Participants that had a decreased DHA or EPA blood value and decreased FFQ value were coded as nonresponders, while those with increased DHA or EPA blood values with increased FFQ values were coded as responders. Thus t-tests were completed to compare non-responders with responders. Spearman's rank correlation tests were performed to determine relationships between maternal self-reported consumption of PUFA-rich foods with measured values from blood. Linear regressions were done to determine if self-reported values were predictors of maternal blood levels. Significance level was set a priori at 0.05 and SPSS was used for all analyses (SPSS 25.0 Chicago, IL, USA).

#### **3. Results**

Study Population. For this analysis, 156 pregnant women expressed interest; of these women, 145 were qualified and consented. Throughout the study, 38 participants were lost-to-follow-up with participant refusal (*n* = 6), moved, no time or lost interest (*n* = 29), discontinued due to drug use (*n* = 1), discontinued due to bed rest (*n* = 1), or miscarried (*n* = 1). Of the remaining 107 participants, 60 were excluded due to missing data for plasma, RBCs, and/or incomplete questionnaire data. Thus, a final sample of 47 pregnant women completed 16-week and 36-week FFQs and venipunctures for this post hoc analysis. On average, participants were 31 years old, had a mean BMI in a healthy range, with appropriate GWG, and delivered full-term healthy babies free from congenital issues (Table 1). The median response from the FFQ was that most women did not consume white fish, other fish, almond/cashew milk, or use polyunsaturated margarine on a regular weekly basis at 16 and 36 weeks gestation; at 16 and 36 weeks, participants reported using oil on foods and for cooking (Table 2). Overall, the PUFA summary decreased from 16 to 36 weeks gestation (Table 2).

**Table 1.** Maternal descriptors.


All values reported as mean ± SD. <sup>a</sup> Values reported as median (minimum, maximum) and used a Mann–Whitney U test due to non-normal distribution. BMI: body mass index.



All values reported as median (minimum, maximum) and used a Mann–Whitney U test due to non-normal distribution. <sup>a</sup> Values reported as mean ± SD. PUFA: polyunsaturated fatty acid.

#### *3.1. EPA and DHA Status*

Both maternal RBC DHA concentration and plasma DHA concentration decreased from 16 to 36 weeks. In contrast, maternal RBC EPA concentration increased from 16 to 36 weeks (Table 3). When comparing participants with overall decreased DHA or EPA blood values and decreased FFQ values (non-responders) to those participants with overall increased DHA or EPA blood values and increased FFQ values (responders), participants that have overall decreased DHA and EPA blood values have significantly increased GWG (*p* = 0.02) with no differences in maternal age, gravida, and pre-pregnancy BMI.

**Table 3.** Maternal blood EPA and DHA concentrations at 16 and 36 weeks gestation and the difference from early to late pregnancy.


All values reported as mean ± SD. <sup>a</sup> Values reported as median (minimum, maximum) and used a Mann– Whitney U test due to non-normal distribution. RBC: Red Blood Cell, DHA: docosahexaenoic acid, EPA: eicosapentaenoic acid.

#### *3.2. Correlation Analysis*

There were no correlations between maternal DHA or EPA in blood compared to self-reported fish consumption. There were moderate positive correlations between selfreported almond/cashew milk consumption at 36 weeks gestation with 36-week plasma concentration of DHA (*p* = 0.01, r = 0.582) as well as with the change value of plasma DHA (*p* = 0.041, r = 0.473) from 16 to 36 weeks (Table 4). Similarly, self-reported 36-week DHAand EPA-rich oil consumption (sunflower/soy/corn/olive) moderately correlates with 16-week EPA on RBCs (*p* = 0.04, r = −0.306), 36-week plasma DHA (*p* = 0.046, r = 0.464), and the change in plasma DHA (*p* = 0.02, r = 0.519) from 16 to 36 weeks (Table 4). Lastly, selfreported 36-week PUFA-rich summary score moderately correlates with 36-week plasma DHA (*p* = 0.01, r = 0.566), and the change in plasma DHA (*p* = 0.01, r = 0.567) from 16 to 36 weeks gestation (Table 4).

**Table 4.** Correlation between self-reported PUFA-rich food consumption and maternal blood and plasma DHA or EPA concentrations.


#### *3.3. Regression Analysis*

We found predictive models for plasma and RBC concentrations of EPA and DHA (Table 5). Controlling for gravida and 16-week self-reported PUFA-rich food summary, 16-week maternal BMI (*p* = 0.01) predicted 16-week maternal EPA on RBCs (Table 5); this suggests that changes in 1 unit of BMI correlates with about 2 ng/dL of maternal EPA, which may not be clinically meaningful. When controlling for gravida and GWG, 36-week self-reported PUFA oil consumption (*p* = 0.02) significantly predicted 36-week plasma DHA (Table 5); this suggests that if women added 2–3 servings/day of PUFA oil, this could

increase plasma DHA by 10 ng/dL, which could result in clinically meaningful differences. Other models that approached significance or did not have any significant individual predictors are not shown.


BMI: body mass index; PUFA: polyunsaturated fatty acid. Bolded headers indicate significant models. \* *p* < 0.05. Non-significant measures in each regression model includes: Model 1: gravida; Model 2: gravida, 16-week BMI; Model 3: gravida.

#### **4. Discussion**

The purpose of the study was: (1) to measure DHA and EPA levels from maternal blood in early and late pregnancy, and (2) to determine the association, and possible predictors, between self-reported consumption of PUFA-containing foods with DHA and EPA concentrations in maternal blood in early and late pregnancy. We hypothesized that: (1) PUFA-rich foods, DHA, and EPA levels will be similar in early and late pregnancy, and (2) there will be a positive correlation, and possible predictors, between self-reported PUFArich food consumption and circulating PUFA concentrations which was consistent with our findings. Our main findings were as follows: (1) DHA levels in maternal blood, and selfreported PUFA-food average weekly consumption, decreased from 16 to 36 weeks gestation; (2) self-reported PUFA-rich food average weekly consumption positively correlates with measured DHA and EPA levels in blood; and (3) DHA, but not EPA, concentration in blood was predicted by self-reported PUFA oils consumption.

We found both self-reported PUFA-food weekly consumption and measured DHA levels decreased from early to late pregnancy, which was different than expected. This suggests that there is decreased maternal consumption of PUFA-rich oils as pregnancy progresses. Previous literature on this topic reports conflicting results. One study notes that serum fatty acid concentrations of DHA, EPA, and total omega-3 PUFA's increase from the first to second trimester, with a slight, but continued, increase from the second to third trimester, which was attributed to an increase in lipids transported across the placenta after 20 weeks gestation [20]; however, our study was of US women with MS technology and the previous findings were in Brazilian women and LC technology, possibly explaining the difference in findings. Similar to our findings, another study reported decreasing levels of maternal plasma levels of DHA from 27 weeks gestation until delivery [21]. One further longitudinal study from Spain found an increase of total omega-6 PUFA's, a decrease of EPA, and no significant change of DHA, from the first to the third trimester of maternal plasma [22]. In the present study, decreasing concentrations of DHA on RBCs and in plasma were found from early and late pregnancy. The decreasing levels of DHA may be explained by a decrease in consumption of DHA and EPA-containing oils throughout the pregnancy. Previous studies have focused on the intake of fish throughout pregnancy [8]; whereas

the present study suggests a potential recommendation of increasing maternal PUFA-rich oil intake during pregnancy. Since those with generally decreased DHA or EPA in blood and on the questionnaire have significantly increased GWG, further research is needed to determine why these women may have an overall decrement in nutrition quality as pregnancy progresses.

As we hypothesized, we found self-reported 36-week PUFA-rich average weekly food consumption positively correlates with DHA and EPA levels at 36 weeks. Similar to the present study, previous research by Kobayashi et al. found a correlation between food intake and serum levels of EPA (R = 0.37) and DHA (R = 0.27) during pregnancy [8]. This study utilized a FFQ which had users rate foods such as fish, shellfish, and other fish products based on consumption and portion size during early (8–14 weeks) and late (26–35 weeks) gestation in Japan [8]. This differed from the current study in which correlations were analyzed for specific foods, allowing further assessment of which food items correlate best with the blood sample findings. Furthermore, the FFQ utilized in the Kobayashi et al. study has limited accuracy in measuring cooking oil as a possible source of PUFA intake; this is a limitation that our study was able to address [8]. Other research evaluated the validation of a FFQ measuring PUFA status in non-pregnant adults [7]. This study found a positive correlation between self-reported dietary DHA intake, specifically fish, with plasma DHA, but no correlation between plasma EPA [7]. The discrepancy between these findings and our findings of a positive correlation with both plasma DHA and EPA could be due to the type of FFQ used. The FFQ utilized in the previously mentioned research emphasized PUFA status from fish consumption, whereas the questionnaire utilized in the present study emphasized PUFA-rich foods, such as fish, oils, and margarine. These details provide validation regarding the participant's diet relative to a sensitive measure of EPA and DHA intake, thus accounting for the difference in results.

Finally, we found that 36-week DHA concentration was predicted by 36-week selfreported PUFA oil consumption, but EPA concentrations were predicted by maternal BMI. These findings suggest that maternal DHA levels could potentially be estimated by FFQ self-reported PUFA oil levels during pregnancy. This would provide a non-invasive method of assessing late pregnancy PUFA status, to ensure recommended levels are met, as an essential part of proper fetal development. Interestingly, maternal EPA concentrations were predicted with a negative association with maternal BMI. Similar to our findings, a study by Young et al. reported a negative association between BMI of non-pregnant women and omega-3 index [22]. The negative association was proposed to be due to altered metabolic pathways in the absorption and utilization of omega-3 fatty acids in women with obesity compared to healthy BMI [22]. The similar findings between our study and that of Young et al. suggests that the utilization of EPA is similar in non-gravid and gravid women. Further research is needed to accurately define the role of maternal BMI, most likely adiposity status, in pregnant women and EPA concentrations.

The strengths of our study include the unique comparison of early and late pregnant women with blood samples and questionnaire data. While our study provides valuable insights, this research is not without its limitations. First, the small sample and nonnormally distributed variables, may influence linear regression analysis; however, the uniqueness of the data argue for further investigation with a larger sample to enable the accuracy and efficiency of linear regression estimates. With more data, the type 2 error in the study would be reduced, leading to a higher sensitivity and greater generalizability of the outcomes. Furthermore, as with any self-report method for assessment, there is a potential for self-reported bias when responding. Women may feel compelled to alter their answers based on what they think they should be consuming, not what they consumed. It is important to note that some of the foods contained in the FFQ contained omega-3 and omega-6 fats; therefore, further research needs to explore how women's self-reported response relates to blood levels of both types of unsaturated fats. However, given the correlation between the self-reported data and the direct measurement of blood variables

and the low-cost, patient- and clinic-friendly use of FFQ relative to the use of venipuncture, this warrants further investigation.

Future research should expand upon the current study by assessing neonatal outcomes according to self-reported PUFA status. Furthermore, more research can be done to optimize the use of the FFQ and address a larger, more diverse pregnant population. The FFQ from the present study can be implemented in OB/GYN clinics with the goal of providing patient knowledge on PUFA-containing foods and creating obtainable goals for patients on the amount of those foods to consume. Further research should focus on overall nutrition quality in those women who have trends of increased GWG. This may create more education for patients and create better outcomes for neonates.

#### **5. Conclusions**

In conclusion, the present study found that average weekly PUFA consumption and blood levels seem to decrease throughout pregnancy. Importantly, the Sheffield FFQ seems to be an effective method for estimating late pregnancy DHA and EPA blood levels. This research allows for a compelling and non-invasive method for estimating DHA and EPA concentration in pregnant women, especially the third trimester. Furthermore, by utilizing a FFQ, women can be aware of their DHA and EPA status, thus, allowing a simpler approach for patients and clinical professionals to track PUFA intake throughout pregnancy. The present study provides insight into an easy, cost-effective method for estimating DHA and EPA status in pregnant women but warrants further nutrition analysis. Overall, the Sheffield FFQ might provide a noninvasive, low-cost method to estimate DHA and EPA status, especially in late pregnancy; however, further investigation with a larger sample is warranted.

**Author Contributions:** K.L. assisted writing the manuscript, formatted data in the tables, and performed literature review; B.W. assisted writing the manuscript, sample collection, formatted data; C.S. assisted writing the manuscript, preparing and cleaning data, sample collection, and running the DHA and EPA serum analysis; C.A.J. assisted with the data analysis; C.I., J.D. and E.N. assisted with the recruitment, retention of participants as well as review of the manuscript; R.P., B.R.A.-T. and S.M. assisted with the data interpretation, and review of the manuscript; L.E.M. completed study design, recruitment, oversaw all aspects of project, obtained funding, oversaw sample collection and processing, oversaw data cleaning and analysis, led/assisted/the writing process. All authors have read and agreed to the published version of the manuscript.

**Funding:** American Heart Association (AHA grant #15GRNT24470029; #18IPA34150006) and by ECU Internal Funds.

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of East Carolina University (protocol code IRB # 12-002524 and approved on 30 September 2021.

**Informed Consent Statement:** Written informed consent was obtained from the participants prior to publishing results.

**Data Availability Statement:** De-identified data can be shared upon request to the corresponding author.

**Conflicts of Interest:** The authors report no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

## *Article* **Associations of Dietary Patterns and Vitamin D Levels with Iron Status in Pregnant Women: A Cross-Sectional Study in Taiwan**

**Arpita Das 1, Chyi-Huey Bai 2,3,4, Jung-Su Chang 1,4,5, Ya-Li Huang 2,3, Fan-Fen Wang 6, Yi-Chun Chen 1,\* and Jane C.-J. Chao 1,4,7,8,\***


**Abstract:** Vitamin D is involved in the pathophysiology of anemia. This cross-sectional study was conducted using the Nationwide Nutrition and Health Survey in Pregnant Women in Taiwan database. We investigated associations among dietary patterns (DPs), vitamin D, and iron-related biomarkers in pregnant women. The principal component analysis revealed four DPs. Linear and logistic regression analyses were performed to investigate the association of DPs with anemia-related biomarkers. Plant-based, carnivore, and dairy and nondairy alternatives DPs were positively associated with serum vitamin D levels. After adjusting covariates, the pregnant women consuming plant-based DPs at the mid-tertile (T2) were associated with reduced risks of low serum folate and vitamin D levels, and those consuming carnivore DPs at higher tertiles (T2 and/or T3) were correlated with an increased risk of low serum iron levels but decreased risks of low serum transferrin saturation, vitamin B12, and vitamin D levels. The pregnant women consuming dairy and nondairy alternatives DPs at the highest tertile (T3) were associated with reduced risks of low serum folate and vitamin B12 levels. However, the processed food DP was not correlated with anemia-related biomarkers. Thus, plant-based, carnivore, and dairy and nondairy alternatives DPs were associated with the risk of low-serum-anemia-related variables.

**Keywords:** vitamin D; iron; dietary pattern; principal component analysis; gestational anemia

#### **1. Introduction**

Anemia during pregnancy or gestational anemia is a major health concern affecting approximately 38% of the global population (approximately 32 million individuals); this proportion ranges from 24% in the Western Pacific Region to 49% in Southeast Asia [1,2]. The World Health Organization (WHO) has defined anemia as a hemoglobin (Hb) level of <6.83 mmol/L (<11 g/dL) and severe anemia as an Hb level of <4.34 mmol/L (<7 g/dL) [3]. For pregnant women, the trimester-wise classification proposed by the Center for Disease

**Citation:** Das, A.; Bai, C.-H.; Chang, J.-S.; Huang, Y.-L.; Wang, F.-F.; Chen, Y.-C.; Chao, J.C.-J. Associations of Dietary Patterns and Vitamin D Levels with Iron Status in Pregnant Women: A Cross-Sectional Study in Taiwan. *Nutrients* **2023**, *15*, 1805. https://doi.org/10.3390/ nu15081805

Academic Editors: Louise Brough, Gail Rees and Roberto Iacone

Received: 21 February 2023 Revised: 17 March 2023 Accepted: 4 April 2023 Published: 7 April 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Control and Prevention (CDC) suggests that gestational anemia can be indicated by an Hb level of <6.83 mmol/L (<11 g/dL) in the first and third trimesters and that of <6.52 mmol/L (<10.5 g/dL) in the second trimester [4].

Gestational anemia increases the incidence rates of perinatal mortality, stillbirth, abnormal or retarded brain growth, and fetal morbidity [5,6]. Iron deficiency anemia is the most common type of gestational anemia and indicated by a serum ferritin level of <0.034 nmol/L (<15 μg/L) [7]. Other common causes of gestational anemia include folate (megaloblastic anemia) or vitamin B12 (pernicious anemia) deficiency, which contributes to maternal morbidities [8,9]. Fetal nutrient deficiencies may result from congenital abnormality, low birth weight, and preterm delivery [10,11]. Iron is a key micronutrient essential for tissue oxygenation and erythropoiesis. Blood loss, decreased iron intake, and impaired iron absorption could contribute to iron deficiency [12]. Gestational iron storage and the absorption of dietary iron are important for the maintenance of iron homeostasis. Ferritin is a protein form which stores iron and serves as a preliminary predictor of lower hemoglobin and anemia [13,14]. Hence, in the present study, the major variables related to anemia were ferritin followed by hemoglobin and serum iron levels. In a study that took place in the UK and Australia, a serum ferritin test in the first trimester was suggested to verify whether pregnant women needed to be referred for iron therapy, and serum ferritin levels were considered to be assessed in the first antenatal visit for women from areas with a high prevalence of iron-deficiency anemia, along with a full blood count test in early pregnancy for women at high risk of iron-deficiency anemia [15]. Additionally, a prospective cohort study of maternal and infant health and nutrition surveillance in Bangladesh determined maternal plasma ferritin levels at gestational weeks 14 and 30 and found that plasma ferritin levels in the late gestation of pregnancy were negatively correlated with infant birth weight [16], indicating the crucial role of ferritin as a form of iron storage in fetal growth outcome.

Several dietary nutrients affect iron balance, and the antioxidant vitamin C, as an acidic substance, promotes iron absorption [17]. Most earlier studies have focused on the role of vitamin C in dietary iron absorption [18,19]. However, few studies have explored the association between dietary patterns (DPs) and vitamin D levels in women with gestational anemia. Iron absorption was reportedly enhanced by vitamin D through reducing the levels of hepcidin and proinflammatory cytokines [20,21]. However, the role of vitamin D in anemia prevention and iron absorption remains controversial [22]. In animal- and population-based pregnancy studies, Qiu et al. [23] and Si et al. [24] both reported a positive association between blood vitamin D levels and iron status. A cross-sectional study conducted by Mayasari et al. revealed an association between dietary intake and serum hepcidin levels during pregnancy [25]. Furthermore, an evidence-based study conducted by Michalski et al. among Vietnamese women of reproductive age reported a positive association between serum, instead of dietary, vitamin D and Hb levels [26]. Additionally, Wong et al. found that serum vitamin D levels were positively associated with serum ferritin levels in Chinese pregnant women [27]. However, the aforementioned studies did not explore any other iron-related biomarkers. Our knowledge regarding DPs, vitamin D levels, and iron status remains limited. In the present study, DP was used as a supportive approach to investigate the association between overall dietary factors and disease outcomes [28]. Thus, we investigated the association of DPs with vitamin D levels and other iron-related biomarkers in pregnant women.

#### **2. Materials and Methods**

#### *2.1. Study Design and Population Demographics*

This population-based cross-sectional study was conducted using a database associated with the Nationwide Nutrition and Health Survey in Pregnant Women in Taiwan (2017–2019; Pregnant NAHSIT 2017–2019). Relevant data were collected from a total of 11 recognized hospitals in Taiwan. The inclusion criteria were as follows: being aged >15 years; receiving a maternal handbook; using an obstetric inspection service more than

once; being able to communicate in Mandarin, Taiwanese, and other languages; and being willing to participate in our study and provide consent. The exclusion criteria included having multiparity (>3) and being nonresponsive.

The data of 1502 pregnant women were used in the present study. After the participants signed the consent form, the researchers assigned the date for collecting data during their prenatal visits. The collected data were classified into the following four categories: sociodemographic, anthropometric, biochemical, and dietary (including supplements, such as milk powder, multivitamin/multimineral, iron, vitamin B complex, folate, vitamin D, and calcium, and dietary assessment) data. Sociodemographic and anthropometric data were obtained using a self-reported questionnaire, whereas dietary data were collected by well-trained registered dieticians during face-to-face interviews with the women. The data collection from all the questionnaires took 60–90 min. Biochemical analyses were performed using blood samples collected during prenatal visits. This study was approved by the Joint Institutional Review Board of Taipei Medical University, Taiwan (approval number: TMU-JIRB N201707039) and conducted in accordance with the ethical principles of the Declaration of Helsinki.

#### *2.2. Dietary Assessment*

Dietary assessment was performed using a 24 h dietary recall method and a validated semiquantitative food frequency questionnaire (FFQ) modified from the NAHSIT FFQ [25]. Food pictures and measurement cups or spoons were used when 24 h dietary recall was conducted by registered dieticians. The FFQ is the most commonly used, reliable, and cost-effective tool for nutrition surveys and has high reproducibility [29]. A total of 59 food items were identified using the FFQ. For the present study, a total of 32 food groups were developed based on the categories and nutrient contents of the aforementioned food items [25]. Food items having similar nutrient characteristics were categorized under the same group (Supplementary Table S1).

The daily, weekly, or monthly frequencies of food intake were recorded in the FFQ. The total monthly frequency of a particular food group was calculated. According to the 24 h dietary recall data, nutrient intake was calculated using Cofit Pro (Cofit Healthcare, Taipei, Taiwan), an online software available on the Taiwan Food Nutrient Database.

DPs can be determined using two approaches: a priori (a hypothesis-derived prospective study) and a posteriori (a data-driven, frequency-based retrospective study) methods [30]. Principal component analysis (PCA) was performed in the present study to determine the DPs of the women, because PCA (an a posteriori method) offered the highest interpretability with minimal information loss and reduced dataset dimensionality [31].

#### *2.3. Anthropometric and Biochemical Data Collection*

Pre-pregnancy body mass index (BMI) was calculated using body weight (kg) divided by height (m2). Both body weight and height before pregnancy were self-reported and collected in the questionnaire. Blood was drawn from the median cubital and cephalic veins. Serum hemoglobin (Hb) levels were measured using a hematology analyzer (Sysmex Corp., Kobe, Japan). Serum iron levels (μmol/L) were determined spectrophotometrically using a Beckman Coulter Unicel DxC 800 (Beckman Coulter, Brea, CA, USA) after iron was released by acetic acid from transferrin and reduced to ferrous iron by hydroxylamine and thioglycolate [25]. Serum ferritin levels were assessed by an enzyme-linked immunosorbent assay using the Beckman Coulter Unicel DxC 800 (Beckman Coulter, Brea, CA, USA) [25]. The total iron-binding capacity (TIBC, μmol/L) was evaluated by the colorimetric immunoassay method using the Beckman Coulter Unicel DxC 800 (Beckman Coulter, Brea, CA, USA) [32]. Transferrin saturation (%) was calculated by the percentage of serum iron levels divided by the TIBC value [33]. The serum levels of folate [34] and vitamin B12 [35] were measured using SimulTRAC-SNB radioimmunoassay kits (MP Biomedicals, Santa Ana, CA, USA) with 125I or 57Co as the tracer, respectively. Serum 25(OH) vitamin D levels were determined by an electrochemiluminescence immunoassay using an Elecsys vitamin

D total reagent kit with ruthenium-labeled vitamin-D-binding protein (Roche Diagnostics Ltd., Taipei, Taiwan) [36].

#### *2.4. Anthropometric and Biochemical Parameters in Gestational Anemia*

The Ministry of Health and Welfare, Taiwan, has recommended the following BMI-based classification of adults: underweight (<18.5 kg/m2), normal weight (18.5 to <24 kg/m2), overweight (24 to <27 kg/m2), and obesity (>27 kg/m2) [37]. Gestational anemia was defined according to the criteria outlined by the WHO and CDC. The normal cutoff values of serum iron and TIBC in women without anemia are 10.7 μmol/L (60 μg/dL) [38] and 42.96–80.55 μmol/L (240–450 μg/dL) [39]. The WHO recommends the following cutoff values for gestational anemia: serum ferritin level <0.034 nmol/L (<15 μg/L) [40] and transferrin saturation <16% [41]. The reference levels of serum folate for all age populations are 13.6–45.3 nmol/L (6–20 ng/mL) [42]. The Endocrine Society has defined vitamin D insufficiency as a vitamin D level of <75 nmol/L (<30 ng/mL) [43].

#### *2.5. Statistical Analysis*

Statistical analysis was performed using SPSS (version 22.0, IBM Corp., Armonk, NY, USA) and SAS (version 9.4, SAS Institute Inc., Chicago, IL, USA). A one-way analysis of variance was used for continuous variables expressed as mean ± standard deviation, whereas the chi-square test was used for categorical variables expressed as number and percentage. Tukey's post hoc multiple comparisons were performed to determine significant within-group differences among continuous variables. We identified DPs by PCA using SAS. A total of four DPs were identified through orthogonal varimax rotation with a mean eigenvalue of 1.0 and a factor loading of >0.30 [44]. Factor loadings of <0.30 were omitted for simplification. A high factor loading indicates a strong association between food groups and disease. For each DP, DP scores were calculated by total food intake (frequency/month) times factor loading. We used the following three models to verify the association between DPs and blood biomarker levels: model 1 (crude model), model 2 (adjusted for age, region of residence, parity, and trimester), and model 3 (adjusting factors in model 2 plus daily dietary intake). A simple linear regression analysis was conducted using the independent (DP) and dependent (biochemical biomarkers) variables to identify the trend (positive or negative) of association. Data are presented in terms of the regression of coefficient (β) and 95% confidence intervals (CIs). For further confirmation, each DP was categorized into tertiles. Tertiles 1 (T1), 2 (T2), and 3 (T3) represent the lowest, mid, and highest DP scores, respectively. Furthermore, a binomial logistic regression analysis was performed to identify the disease trend across the tertiles of each DP and biochemical biomarkers, and the odds ratios (ORs) of T2 and T3 were compared with the reference group (T1). Data are presented in terms of odds ratios and 95% CIs. The OR value of >1 or <1 with statistical significance indicates increased or decreased disease risk, whereas OR = 1 represents nonsignificant effects [45]. Statistical significance was set at *p* ≤ 0.05.

#### **3. Results**

#### *3.1. Sociodemographic and Anthropometric Characteristics of the Women*

Pregnant women in T3 of serum vitamin D were older (32.9 ± 4.9 vs. 32.0 ± 4.7 years) than those in T1 of serum vitamin D (Table 1). Most pregnant women in T3 of serum vitamin D lived in the southern part of Taiwan (32%), were primiparous (49.3%), and were in the third trimester of pregnancy (53%). The women across the vitamin D tertiles did not differ significantly in terms of education level, family monthly income, duration of sun exposure, or BMI.


**Table 1.** Sociodemographic and anthropometric characteristics of the women across the tertiles of serum vitamin D (*n* = 1502) 1.

<sup>1</sup> Continuous data are presented as the mean ± standard deviation, whereas categorical data are presented as the number and percentage in the parentheses. Different superscript letters for continuous variables indicate significantly different (*<sup>p</sup>* ≤ 0.05) using Turkey's post hoc test. <sup>2</sup> Tertiles of serum vitamin D levels: T1: 20 to >53 nmol/L, T2: 54 to >71 nmol/L, and T3: 72 to 154 nmol/L. <sup>3</sup> The *p*-value was determined using oneway analysis of variance test for continuous variables and chi-square test for categorical variables. BMI, body mass index.

#### *3.2. Biochemical Characteristics of the Women*

Pregnant women in T3 of serum vitamin D had higher levels of serum Hb (7.4 ± 1.3 mmol/L), iron (13.9 ± 7.8 μmol/L), TIBC (85.6 ± 17.1 μmol/L), folate (32.3 ± 17.0 nmol/L), and vitamin B12 (249.0 ± 169.8 pmol/L), but lower serum ferritin levels (0.05 ± 0.05 nmol/L) than those in T1 of serum vitamin D did (Table 2). Categorical classification revealed that the levels of serum Hb and folate were >6.76 mmol/L and ≥13.5 nmol/L, respectively, in most women in T3 of serum vitamin D. The number of individuals with anemia defined as Hb <6.83 mmol/L (<11 g/dL) in trimesters 1 and 3 or Hb <6.52 mmol/L (10.5 g/dL)

in trimester 2 was 322 (21.4%), and we did not further analyze the data based on the pregnant women with or without anemia due to there being much fewer women with anemia compared with those without anemia.


**Table 2.** Biochemical characteristics of the women across the tertiles of serum vitamin D (*n* = 1502) 1.

<sup>1</sup> Continuous data are presented as the mean ± standard deviation, whereas categorical data are presented as the number and percentage in the parentheses. Different superscript letters for continuous variables indicate significant difference (*<sup>p</sup>* ≤ 0.05) using Turkey's post hoc test. <sup>2</sup> Tertiles of serum vitamin D levels: T1: 20 to >53 nmol/L, T2: 54 to >71 nmol/L, and T3: 72 to 154 nmol/L. <sup>3</sup> The *p*-value was determined using one-way analysis of variance test for continuous variables and chi-square test for categorical variables. TIBC, total ironbinding capacity.

#### *3.3. Dietary Intake of the Women*

Daily dietary intakes of energy, macronutrients, iron, folate, vitamin B12, and vitamin D were determined using 24 h dietary recall data. Pregnant women in T3 of serum vitamin D had higher intakes of protein (g), fat (g and % of energy), iron, folate, and vitamin D, but lower intakes of carbohydrates (% of energy) than those in T1 of serum vitamin D did (Supplementary Table S2). No significant differences were found in pregnant women across the tertiles of serum vitamin D in terms of energy or vitamin B12 intake.

Pregnant women in T3 of serum vitamin D had higher monthly intake frequencies for supplements of multivitamin/multimineral, folate, and calcium than those in T1 of serum vitamin D did (Supplementary Table S2). Other dietary supplements such as milk powder (17.6%), iron (11.2%), vitamin B complex (18.0%), and vitamin D (11.1%) were not assessed for the monthly intake frequency because a lower proportion (<20%) of the women took these supplements.

#### *3.4. Dietary Patterns*

The PCA revealed a total of four DPs (Figure 1). All four DPs explained total variance of 9.35% (4.37%, 1.93%, 1.61%, and 1.44%). DPs were categorized and ranked on the basis of a threshold factor loading value (>0.30). Each DP was named according to their corresponding factor loading values and dietary component structures. The first pattern comprising a total of ten food groups was named the plant-based DP (DP-1) because the highest factor loadings were exhibited by bamboo shoots and melons; mushroom and related products; carrots, roots, and tubers; dark-colored vegetables; and legumes and various beans. Other food groups in DP-1 included marine plants and kelp; nuts and nut products; animal organ meat and blood; general soy products and gluten pasta; and aquatic fish, shell, shellfish, and seafood. The second pattern was named the carnivore DP (DP-2), which comprised the following six food groups from the highest to the lowest factor loadings: livestock lean meat; poultry meat; livestock lean meat; processed meat and meat products; herbs and spices; and salt. The third pattern was named the processed food DP (DP-3), which comprised the following six food groups: cake, pastry, and dumplings; salty buns and sweet buns; glutinous rice desserts and rhizome starch; pickled vegetables; deep water fish; and seafood products. Finally, the fourth pattern was named the dairy and non-dairy alternatives DP (DP-4), which comprised the following six food groups: milk and milk products; nondairy products, such as soy and rice milk; eggs; breakfast cereals and bread and related products; noodles and related products; and 100% pure juice and commercially available vegetable juice.

**Figure 1.** Factor loading of four dietary patterns identified by principal component analysis. The factor loadings of <0.30 were eliminated for simplification.

#### *3.5. Association of DPs with Serum-Anemia-Related Biochemical Variables*

Table 3 presents the association between plant-based DP (DP-1) and anemia-related biochemical variables. Serum ferritin levels in the crude model (model 1) were negatively (β: −0.06, 95% CI: −0.29, −0.01, *p* ≤ 0.05) associated with DP-1, but after covariate adjustment, there was no significant association between serum ferritin levels and DP-1. In contrast, serum TIBC in model 1 (β: 0.09, 95% CI: 0.02, 0.10, *p* ≤ 0.001) and serum vitamin D levels in all three models (model 1: β: 0.08, 95% CI: 0.02, 0.08, *p* ≤ 0.01; model 2: β: 0.06, 95% CI: 0.00, 0.06, *p* ≤ 0.05; model 3: β: 0.04, 95% CI: −0.00, 0.05, *p* ≤ 0.05) were positively associated with DP-1.

As shown in Table 4, in all the three models, carnivore DP (DP-2) was correlated with the reduction in serum iron levels by 0.07–0.08 μmol/L (model 1: β: −0.08, 95% CI: −0.49, −0.10, *p* ≤ 0.01; model 2: β: −0.07, 95% CI: −0.47, −0.07, *p* ≤ 0.01; model 3: β: −0.08, 95% CI: −0.50, −0.11, *p* ≤ 0.01). In addition, DP-2 was associated with the decrease in serum ferritin levels by 0.06 nmol/L (95% CI: −0.46, −0.04, *p* ≤ 0.05) but the increase in serum TIBC levels by 0.08 μmol/L (95% CI: 0.02, 0.10, *p* ≤ 0.01) in model 1. Changes in serum ferritin and TIBC levels were not significant after covariate adjustment. In all three models, serum vitamin D levels were positively associated with DP-2, and the increase in serum vitamin D ranged from 0.04 to 0.08 nmol/L (model 1: β: 0.08, 95% CI: 0.02, 0.10, *p* ≤ 0.01; model 2: β: 0.06, 95% CI: 0.00, 0.08, *p* ≤ 0.05; model 3: β: 0.04, 95% CI: −0.00, 0.07, *p* ≤ 0.05).


**Table 3.** The association of plant-based dietary pattern with anemia-related biochemical variables in serum evaluated by the generalized linear regression analysis 1.

<sup>1</sup> The values of β and data in the parentheses indicate regression coefficient and 95% confidence interval (95% CI), respectively, after covariate adjustment in different models: model 1, crude model; model 2, adjusted for age, region of residence, parity, and trimester; and model 3, adjusted for age, region of residence, parity, trimester, and daily dietary intake, such as energy (kcal), carbohydrate (% of energy), protein (g and % of energy), fat (g and % of energy), iron (mg), folate (μg), and vitamin D (μg). \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, and \*\*\* *p* ≤ 0.001. TIBC, total iron-binding capacity.

**Table 4.** The association of carnivore dietary pattern with anemia-related biochemical variables in serum evaluated by the generalized linear regression analysis 1.


<sup>1</sup> The values of β and data in the parentheses indicate regression coefficient and 95% confidence interval (95% CI), respectively, after covariate adjustment in different models: model 1, crude model; model 2, adjusted for age, region of residence, parity, and trimester; and model 3, adjusted for age, region of residence, parity, trimester, and daily dietary intake, such as energy (kcal), carbohydrate (% of energy), protein (g and % of energy), fat (g and % of energy), iron (mg), folate (μg), and vitamin D (μg). \* *p* ≤ 0.05 and \*\* *p* ≤ 0.01. TIBC, total iron-binding capacity.

The processed food DP (DP-3) did not exhibit any strong association with anemiarelated biochemical biomarkers except vitamin B12 (Supplementary Table S3). Serum vitamin B12 levels were negatively associated with DP-3 in models 1 and 2 (model 1: β: −0.04, 95% CI: −1.44, 0.09, *p* ≤ 0.05; model 2: β: −0.05, 95% CI: −1.48, 0.02, *p* ≤ 0.05).

Table 5 presents the association between the dairy and nondairy alternatives DP (DP-4) and anemia-related biochemical variables. DP-4 was positively associated with serum TIBC in model 1 (β: 0.08, 95% CI: 0.02, 0.10, *p* ≤ 0.01). Furthermore, the serum vitamin D level was only positively associated with DP-4 in models 1 and 2 (model 1: β: 0.05, 95% CI: 0.02, 0.09, *p* ≤ 0.05; model 2: β: 0.04, 95% CI: −0.00, 0.08, *p* ≤ 0.05).

#### *3.6. Association of DPs with the Risk of Low-Anemia-Related Biomarkers*

As shown in Table 6, the binomial logistic regression analysis revealed that the pregnant women with the highest consumption levels (T3) of plant-based DPs (DP-1) were associated with a reduced risk of low ferritin levels (OR: 0.73, 95% CI: 0.57, 0.94, *p* ≤ 0.05) in model 1 compared with those with lower consumption levels (T1) of DP-1. However, there were no significant correlations between DP-1 and the risk of low serum ferritin levels after covariate adjustment. Additionally, the pregnant women with higher consumption levels (T2 and/or T3) of DP-1 were likely to have reduced risks of low folate and vitamin D levels compared with those with lower consumption levels (T1) of DP-1 in all the models.


**Table 5.** The association of dairy and nondairy alternatives dietary pattern with anemia-related biochemical variables in serum evaluated by the generalized linear regression analysis 1.

<sup>1</sup> The values of β and data in the parentheses indicate regression coefficient and 95% confidence interval (95% CI), respectively, after covariate adjustment in different models: model 1, crude model; model 2, adjusted for age, region of residence, parity, and trimester; and model 3, adjusted for age, region of residence, parity, trimester, and daily dietary intake, such as energy (kcal), carbohydrate (% of energy), protein (g and % of energy), fat (g and % of energy), iron (mg), folate (μg), and vitamin D (μg). \* *p* ≤ 0.05 and \*\* *p* ≤ 0.01. TIBC, total iron-binding capacity.

**Table 6.** Odds ratios (ORs) of low-anemia-related biochemical variables in serum across the tertiles of plant-based dietary pattern assessed by binomial logistic regression analysis 1.


<sup>1</sup> Three different models were performed in binomial logistic regression analysis: model 1, crude model; model 2, adjusted for age, region of residence, parity, and trimester; and model 3, adjusted for age, region of residence, parity, trimester, and daily dietary intake, such as energy (kcal), carbohydrate (% of energy), protein (g and % of energy), fat (g and % of energy), iron (mg), folate (μg), and vitamin D (μg). <sup>2</sup> Variables were divided into two levels on the basis of cutoff values in serum: hemoglobin, 6.52 mmol/L (10.5 g/dL); iron, 10.7 μmol/L (60 μg/dL); ferritin, 0.034 nmol/L (15 ng/mL); TIBC, 42.96 μmol/L (240 μg/dL); transferrin saturation, 16%; folate, 13.6 nmol/L (6 ng/mL); vitamin B12, 149.8 pmol/L (203 pg/mL); and vitamin D, 75 nmol/L (30 ng/mL). <sup>3</sup> Dietary pattern scores were divided into tertiles: T1 (reference), 0.56–38.85; T2, >38.87–65.61; and T3 >65.85–436.82. \* *p* ≤ 0.05 and \*\* *p* ≤ 0.01. TIBC, total iron-binding capacity.

As found in Table 7, the pregnant women with higher consumption levels (T3 and/or T2) of the carnivore DP (DP-2) were likely to have an increased risk of low iron levels in all the models. The pregnant women with higher consumption levels (T2) of DP-2 were associated with a decreased risk of low transferrin saturation (OR: 0.70, 95% CI: 0.54, 0.91, *p* ≤ 0.01) in model 2. T2 and T3 of DP-2 were correlated with reduced risks of low serum vitamin B12 and vitamin D levels in the adjusted models.

The processed food DP (DP-3) did not exhibit any prominent associations with anemiarelated biochemical biomarkers except serum vitamin D levels (Supplementary Table S4). The pregnant women with higher consumption levels (T2) of DP-3 were likely to have a reduced risk of low vitamin D levels in model 1 (OR: 0.71, 95% CI: 0.53, 0.95, *p* ≤ 0.05) and model 2 (OR: 0.68, 95% CI: 0.51, 0.92, *p* ≤ 0.05).

Table 8 demonstrates the associations between the dairy and nondairy alternatives DP (DP-4) and anemia-related biochemical variables. In Model 1, the pregnant women with the highest consumption levels (T3) of DP-4 were correlated with reduced risks of low serum TIBC (OR: 0.71, 95% CI: 0.54, 0.93, *p* ≤ 0.05), low vitamin B12 (OR: 0.73, 95% CI: 0.54, 0.97, *p* ≤ 0.05), and low vitamin D levels (OR: 0.72, 95% CI: 0.54, 0.96, *p* ≤ 0.05). After covariate adjustment, the pregnant women with the highest consumption levels (T3) of DP-4 were associated with decreased risks of low serum folate (models 2 and 3), low vitamin B12 (models 2 and 3), and low vitamin D (model 2).

**Table 7.** Odds ratios (ORs) of low-anemia-related biochemical variables in serum across the tertiles of carnivore dietary pattern assessed by binomial logistic regression analysis 1.


<sup>1</sup> Three different models were performed in binomial logistic regression analysis: model 1, crude model; model 2, adjusted for age, region of residence, parity, and trimester; and model 3, adjusted for age, region of residence, parity, trimester, and daily dietary intake, such as energy (kcal), carbohydrate (% of energy), protein (g and % of energy), fat (g and % of energy), iron (mg), folate (μg), and vitamin D (μg). <sup>2</sup> Variables were divided into two levels on the basis of cutoff values in serum: hemoglobin, 6.52 mmol/L (10.5 g/dL); iron, 10.7 μmol/L (60 μg/dL), ferritin, 0.034 nmol/L (15 ng/mL); TIBC, 42.96 μmol/L (240 μg/dL); transferrin saturation, 16%; folate, 13.6 nmol/L (6 ng/mL); vitamin B12, 149.8 pmol/L (203 pg/mL); and vitamin D, 75 nmol/L (30 ng/mL). <sup>3</sup> Dietary pattern scores were divided into tertiles: T1 (reference), 0.56–38.85; T2, >38.87–65.61; and T3 >65.85–436.82. \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, and \*\*\* *p* ≤ 0.001. TIBC, total iron-binding capacity.

**Table 8.** Odds ratios (ORs) of low-anemia-related biochemical variables in serum across the tertiles of dairy and nondairy alternatives dietary pattern assessed by binomial logistic regression analysis 1.


<sup>1</sup> Three different models were performed in binomial logistic regression analysis: model 1, crude model; model 2, adjusted for age, region of residence, parity, and trimester; and model 3, adjusted for age, region of residence, parity, trimester, and daily dietary intake, such as energy (kcal), carbohydrate (% of energy), protein (g and % of energy), fat (g and % of energy), iron (mg), folate (μg), and vitamin D (μg). <sup>2</sup> Variables were divided into two levels on the basis of cutoff values in serum: hemoglobin, 6.52 mmol/L (10.5 g/dL); iron, 10.7 μmol/L (60 μg/dL), ferritin, 0.034 nmol/L (15 ng/mL); TIBC, 42.96 μmol/L (240 μg/dL); transferrin saturation, 16%; folate, 13.6 nmol/L (6 ng/mL); vitamin B12, 149.8 pmol/L (203 pg/mL); and vitamin D, 75 nmol/L (30 ng/mL). <sup>3</sup> Dietary pattern scores were divided into tertiles: T1 (reference), 0.56–38.85; T2, >38.87–65.61; and T3 >65.85–436.82. \* *p* ≤ 0.05 and \*\* *p* ≤ 0.01. TIBC, total iron-binding capacity.

#### **4. Discussion**

#### *4.1. Association of Serum Vitamin D with Other Serum-Anemia-Related Biomarkers*

We showed that all anemia-related biochemical variables were significantly different across the tertiles of serum vitamin D levels in the pregnant women, except for transferrin saturation. Hence, pregnant women with higher serum vitamin D levels had higher serum Hb, iron, TIBC, folate, and vitamin B12 levels, which indicates better iron status. Similarly, Si et al. [24] found that plasma 25(OH) vitamin D levels were positively correlated with plasma Hb levels in each trimester of Chinese pregnant women. Additionally, Chinese pregnant women with vitamin D deficiencies (<50 nmol/L) in trimesters 1 and 2 were associated with an elevated risk of anemia compared with those without vitamin D deficiencies [24]. A cross-sectional study conducted in Vietnamese non-pregnant young women revealed that serum vitamin D levels, not dietary vitamin D intake, were positively associated with Hb levels, but not significantly correlated with anemia [26]. We also found that the pregnant women with higher serum vitamin D levels had lower serum ferritin levels, but the average ferritin levels were still within the normal range. A previous study demonstrated that serum 25(OH) vitamin D levels were not correlated with serum ferritin levels in Indonesian pregnant women in the first trimester; however, the pregnant women with insufficient (<75 nmol/L) or deficient (<50 nmol/L) 25(OH) vitamin D levels in the first trimester had a higher risk of developing anemia in the third trimester [46].

#### *4.2. Association of DPs with Serum-Anemia-Related Biomarkers*

Our findings from the linear regression analysis revealed that both the plant-based (DP-1) and carnivore (DP-2) DPs were negatively associated with serum ferritin levels in the crude mode, but positively correlated with serum vitamin D levels in all the models. In contrast, the processed food DP (DP-3) was negatively associated with serum vitamin B12 levels. The dairy and nondairy alternatives DP (DP-4) was positively correlated with serum TIBC and vitamin D levels. Consistently, our findings from the binomial regression analysis showed that both DP-1 and DP-2 were associated with a reduced risk of low serum vitamin D levels. DP-4 was correlated with decreased risks of low serum TIBC, folate, vitamin B12, and vitamin D levels.

Plant-based foods (non-heme iron source) are rich in fiber, phytate, oxalate, and/or polyphenols which could chelate with iron as an inhibitor of iron bioavailability, and they have less iron absorption compared with heme iron food sources [47–49]; thus, the plant-based DP (DP-1) could be correlated with a reduction in serum ferritin levels. Our study demonstrated that DP-1 was correlated with reduced odds of low serum folate and vitamin D levels in pregnant women. Similarly, a previous study reported that pregnant women consuming an ovo-lacto vegetarian or a low-meat diet were likely to have a lower risk of folate deficiency compared with those consuming a Western diet [50]. Additionally, pregnant women consuming a vegetarian diet had significantly higher serum 1,25-(OH)2 vitamin D levels compared with those consuming a nonvegetarian diet [51]. However, adults consuming a vegetarian diet or a plant-based diet were correlated with lower iron stores (lower serum ferritin levels) and a higher prevalence of anemia, probably due to the poor absorption of non-heme iron compared with those consuming a nonvegetarian diet [52,53].

Notably, the carnivore DP (DP-2) was associated with an increased risk of low serum iron levels in our study. However, a systematic review reported that the adults consuming a high intake of a carnivore/animal-based diet were positively correlated with iron status [54]. The possible reason for the association between DP-2 and low serum iron levels could be attributed to herbs and spices (such as chili paper, garlic, Thai leafy vegetables, shallot, tamarind, and turmeric) in DP-2, which are enriched in polyphenolic compounds and can hinder iron absorption by forming insoluble iron complexes [55]. We also found that DP-2 was correlated with reduced risks of low transferrin saturation, vitamin B12, and vitamin D levels. Norwegian women (36–39 years old) consuming a reindeer meat DP were likely to have slightly higher transferrin saturation (mean: 28%) compared with those consuming a fish (mean: 26%), average (mean: 27%), fruit/vegetables (mean: 24%), or Western/marine DP (mean: 26%) [56]. Dutch pregnant women who consumed higher vitamin B12 intake from animal foods such as meat, fish, or dairy food which were rich in vitamin B12 were correlated with higher plasma vitamin B12 levels [57]. A previous study showed that Caucasian pregnant women in Ireland consumed dietary vitamin D primarily from meat, eggs, and breakfast cereals [58]. Meat was the predominant food group in DP-2, and the pregnant women with higher intakes of DP-2 presumably had better serum vitamin D statuses.

The processed food DP (DP-3) was negatively associated with serum vitamin B12 levels. The excessive thermal treatment of foods during food processing may be attributed to reduced vitamin B12 content in foods [59]. Additionally, high intakes of ultra-processed foods were correlated with decreased intakes of certain vitamins such as vitamin A, B12, C, D, E, and niacin in adults [60].

After covariate adjustment, T3 of the dairy and nondairy alternatives DP (DP-4) was associated with reduced odds of low serum folate, vitamin B12, and vitamin D levels. Consistent with our findings, Cifelli et al. [61] demonstrated that dairy and individual dairy foods were correlated with increased serum folate and vitamin B12 levels in a US population. Dairy food also provided rich sources of vitamins B12 [62] and D [63], which could significantly contribute to serum vitamin B12 and vitamin D levels.

Overall, we identified that plant-based, carnivore, and dairy and nondairy alternatives DPs were positively correlated with serum vitamin D levels and a reduced risk of low serum vitamin D. Serum vitamin D status could be affected not only by dietary patterns but also by exposure to sunlight or the use of sun protection [64]. Our previous study showed that among 1502 pregnant women in Taiwan, 69.6% women had sun exposure ≥10 min/d in the previous month, and 61.7% women had blood drawn in sunny months between June and November [65]. Additionally, dietary vitamin D intake had a greater impact on serum vitamin D levels in the women who lived in the northern part of Taiwan, whereas serum vitamin D levels were more greatly influenced by sunlight-related factors in those who lived in the southern or other parts of Taiwan [65]. These vitamin-D-associated DPs may reduce the risk of anemia in pregnant women, because these DPs were also negatively correlated with other anemia-related biochemical variables such as serum folate and vitamin B12. A possible mechanism for the effect of vitamin D on anemia was its modulation in iron metabolism via the down-regulation of hepcidin [66,67]. Higher serum vitamin D levels could be beneficial for better iron statuses through reducing hepcidin at the transcriptional level and suppressing the expression of proinflammatory cytokines involved in iron imbalance [67]. Active vitamin D could down-regulate the production of endogenous hormone hepcidin, thereby improving iron release, iron recycling, and iron absorption [67], and further maintain iron status during pregnancy. A recent cross-sectional study reported that serum hepcidin levels were negatively associated with the consumption frequency of plant-based foods such as legumes, breakfast cereals, light-colored vegetables, and gourds and root vegetables in Taiwanese pregnant women [25]. In the present study, we did not analyze serum hepcidin, and further studies are necessary to identify whether vitamin-D-rich DP is correlated with serum hepcidin levels.

#### *4.3. Strengths and Limitations*

To the best of our knowledge, the present study pioneered the PCA-mediated identification method for the association of DPs with serum levels of vitamin D and iron biomarkers in Taiwanese pregnant women. PCA is commonly used in pragmatic analyses performed using correlation matrices of intake units to identify common DPs [68]. We used data from the Pregnant NAHSIT 2017–2019 and included pregnant women from different areas of Taiwan (northern, central, southern, and eastern). In addition, sociodemographic data (education and income levels) were also collected to complement our findings.

The present study has some limitations. First, because of the unavailability of data regarding serum vitamin C, hepcidin, and parathyroid hormone levels which could affect iron status, we could not assess the association of DPs with these biomarkers. Second, the use of the FFQ and self-reported data for body weight and height might have introduced biases, such as errors in overestimation or underestimation. To overcome or minimize the biases of the FFQ, we additionally obtained 24 h dietary recall data and used food pictures and measurement cups or spoons during data collection [69]. Third, we did not consider certain pathological conditions of pregnant women, such as morning sickness during the first trimester of pregnancy. Fourth, the data regarding dietary supplements and seasonality were limited. Finally, because of the cross-sectional study design, we could not establish any causal relationship among DPs, serum vitamin D levels, and iron status. Nevertheless, a correlation relationship was identified between DPs and serum levels of anemia-related biomarkers. Future cohort studies and randomized control trials are needed to overcome the aforementioned limitations.

#### **5. Conclusions**

This study is a novel attempt to identify the associations among DPs, serum vitamin D levels, and iron status in pregnant women. Plant-based (DP-1), carnivore (DP-2), and dairy and nondairy alternatives DPs (DP-4) are positively correlated with serum vitamin D levels. The medium intake of a plant-based DP (DP-1) is associated with higher levels of serum folate and vitamin D. The medium and high consumption of carnivore DP (DP-2) is correlated with higher levels of serum vitamin B12 and vitamin D. The high intake of dairy and nondairy alternatives DP (DP-4) is associated with higher levels of serum folate and vitamin B12. However, we found no strong association between DPs and serum levels of Hb and iron status, except the negative correlation between the carnivore DP (DP-2) and serum iron levels. Thus, the medium intake of a vitamin D-rich diet such as a plantbased, carnivore, or dairy and nondairy alternatives DP is suggested to be beneficial for preventing anemia in pregnant women due to better statuses of serum folate, vitamin B12, and vitamin D.

**Supplementary Materials:** The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/nu15081805/s1. Table S1: Food groups and subgroups for dietary assessment; Table S2: Daily dietary intake of women across the tertiles of serum vitamin D levels (*n* = 1502); Table S3: The association of processed food dietary pattern with anemia-related biochemical variables in serum evaluated by the generalized linear regression analysis; Table S4: Odds ratios (ORs) of low-anemia-related biochemical variables in serum across the tertiles of processed food dietary pattern assessed by binomial logistic regression analysis.

**Author Contributions:** Conceptualization, A.D., Y.-C.C. and J.C.-J.C.; data curation, C.-H.B., J.-S.C., Y.-L.H. and F.-F.W.; formal analysis, A.D.; writing—original draft preparation, A.D., Y.-C.C. and J.C.-J.C.; writing—review and editing, J.C.-J.C.; supervision, Y.-C.C. and J.C.-J.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study using the secondary data received no external funding.

**Institutional Review Board Statement:** The study was approved by the Joint Institutional Review Board of Taipei Medical University, Taiwan (approval number: N201707039; approval date: 7 November 2018) and conducted in accordance with the ethical principles of the Declaration of Helsinki.

**Informed Consent Statement:** All participants signed a consent form authorized by the team that conducted the Nationwide Nutrition and Health Survey in Pregnant Women in Taiwan.

**Data Availability Statement:** Data supporting the study findings are available from the database of Nationwide Nutrition and Health Survey in Pregnant Women in Taiwan. The data should be used for research purposes only. The study data are not publicly available.

**Acknowledgments:** We thank the Nationwide Nutrition and Health Survey in Pregnant Women (Taiwan) team for making their database available for our study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

## *Article* **Effect of Circadian Distribution of Energy and Macronutrients on Gestational Weight Gain in Chinese Pregnant Women**

**Wenjuan Xiong 1,†, Shanshan Cui 2,†, Jia Dong 3, Yuanyuan Su 1, Yu Han 1, Zhiyi Qu 1, Shihao Jin 1, Zhi Li 1, Lei Gao 1, Tingkai Cui <sup>1</sup> and Xin Zhang 1,\***


**Abstract:** Gestational weight gain (GWG) may be affected by the timing of dietary intake. Previous studies have reported contradictory findings, possibly due to inconsistent characterizations of meal timing. We conducted a birth cohort study in Tianjin to determine the effect of daily energy and macronutrient distribution in mid and late pregnancy on GWG. Dietary intake information in the second and third trimesters used three 24-h dietary recalls, and meal timing was defined in relation to sleep/wake timing. The adequacy of GWG was assessed using recommendations from the Institute of Medicine guidelines. Pregnant women who had a relatively high average energy and macronutrient distribution in the late afternoon–early evening time window exhibited a greater GWG rate and a greater total GWG than that in morning time window during the third trimester (β = 0.707; β = 0.316). Carbohydrate intake in the morning of the second and third trimesters (β = 0.005; β = 0.008) was positively associated with GWG rates. Morning carbohydrate intake in the second trimester was also positively associated with total GWG (β = 0.004). Fat intake in the morning of the third trimester (β = 0.051; β = 0.020) was positively associated with the GWG rates and total GWG. Excessive GWG of Chinese pregnant women was related closely to eating behavior focused on the late afternoon–early evening and carbohydrate and fat intake in the morning during the second and third trimesters.

**Keywords:** energy; macronutrients; circadian distribution; gestational weight gain

#### **1. Introduction**

Optimal Gestational weight gain (GWG) is essential to ensure the health of both the mother and the baby. However, GWG above or below the recommended guidelines of the Institute of Medicine is related to adverse perinatal outcomes, including gestational hypertension, gestational diabetes, cesarean delivery, premature birth, macrosomia, and infant mortality, as well as long-term negative outcomes in the offspring, including childhood obesity and adiposity [1,2]. Abnormal GWG is currently a serious obstetric issue. For example, in the United States, weight gained is higher than the Institute of Medicinerecommended range in 48% of women giving birth to full-term singleton infants, with 21% gaining insufficient weight [3]. In China, inadequate and excessive weight gain account for 27.2% and 36.6%, respectively [4]. On average, maternal weight increases as pregnancy progresses. The fastest weight gain occurs in the second trimester, and the weight gain rate in the third trimester slightly decreases [5]. In the second and third trimesters, weight gained includes maternal fat accumulation, extravascular fluid, placenta, uterus, and fetus growth [6]. Therefore, it is necessary to simultaneously pay attention to the GWG during the second and third trimesters and explore the associated factors.

**Citation:** Xiong, W.; Cui, S.; Dong, J.; Su, Y.; Han, Y.; Qu, Z.; Jin, S.; Li, Z.; Gao, L.; Cui, T.; et al. Effect of Circadian Distribution of Energy and Macronutrients on Gestational Weight Gain in Chinese Pregnant Women. *Nutrients* **2023**, *15*, 2106. https://doi.org/10.3390/ nu15092106

Academic Editors: Louise Brough and Gail Rees

Received: 26 March 2023 Revised: 24 April 2023 Accepted: 26 April 2023 Published: 27 April 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Emerging evidence suggests that the timing of food intake may affect weight gain. For instance, skipping breakfast, eating lunch late, and eating a large dinner have been associated with various indices of obesity [7]. Nevertheless, this subject has not been sufficiently studied in Chinese pregnant women, and one main methodological limitation is defining meal timing; conventional meal categories (i.e., breakfast, lunch, and dinner) and clock timing (external timing) to characterize meal timing may fail to accurately relate metabolic alterations in the context of the internal circadian rhythm [8]. Dim light melatonin onset is the recommended method for assessing the biological timing (internal circadian timing), which demands participants to stay in dim light conditions for a whole evening or more and undergo repeated blood or saliva collections to measure melatonin concentrations [9]. However, this method is unpractical for most epidemiological or clinical studies. A more practical approach to estimating the circadian time of food intake is to consider the timing of food intake relative to the sleep/wake cycle [10]. Moreover, changing an individual's meal timing in a real-world setting may be difficult, but changing the daily distribution of energy or macronutrients may be achievable. Therefore, in the present study, we investigated dietary intake and sleep/wake timing in the second and third trimesters to define mealtimes relative to sleep/wake timing. We examined the associations between individual daily energy and macronutrient distribution, macronutrient intake in different time windows, and GWG.

#### **2. Materials and Methods**

#### *2.1. Study Design*

Tianjin Maternal and Child Health Education and Service Cohort was a prospective cohort conducted at the Women and Children's Medical Care Center in Hebei and Heping districts of Tianjin, China, beginning in January 2018. The inclusion criteria for the cohort were: (1) age ≥ 18 years; (2) singleton pregnancy; (3) in the first trimester (8–13 weeks) at enrollment; and (4) no plan to move from Tianjin during the subsequent 4 years. The participants in this study were a subsample of this ongoing cohort. The exclusion criteria were (1) having no Chinese speaking or reading abilities; (2) an individual history of diabetes, hypertension, liver failure, renal failure, congestive heart failure, abnormal thyroid function, psychosis, or cancer; and (3) a positive test result for women of COVID-19, syphilis, human immunodeficiency virus, rubella, toxoplasmosis, varicella, or cytomegalovirus [11]. Accordingly, the present analysis included 149 pregnant women who had complete data on at least two of the three visits for the study and had not been locked down during pregnancy. Six of these women were excluded because their daily average energy intake was <500 kcal/d or >3500 kcal/d (first-trimester visit and second-trimester visit: *n* = 5; firsttrimester visit, second-trimester visit, and third-trimester visit: *n* = 1). Finally, 143 pregnant women were included (first-trimester visit and second-trimester visit: *n* = 34; first-trimester visit and third-trimester visit: *n* = 16; first-trimester visit, second-trimester visit, and thirdtrimester visit: *n* = 92) and 234 complete pieces of data for the participants were collected between 2018 January to 2021 December.

#### *2.2. Demographic Data and Covariates*

In the first trimester, through face-to-face interviews, a self-administered questionnaire was used to collect the general demographic data of pregnant women, including age, educational level, employment status, and economic circumstances [12]. The height and weight within 1 month before pregnancy were self-reported. The pre-pregnancy body mass index (BMI) (weight (kg)/height(m)2) was calculated using pre-pregnancy weight and height [11]. Physical activity was measured once per trimester: first trimester, 8–14 gestational weeks; second trimester, 16–27 gestational weeks; and third trimester, 28–37 gestational weeks. Physical activity evaluation was conducted by asking the participants whether they had performed any physical exertion and the duration of daily physical activity (0 = "0 h/d;" 1="≤0.5 h/d;" 2 = ">0.5 h/d and ≤1.0 h/d;" 3 = ">1.0 h/d and ≤2.0 h/d;" and 4 = ">2.0 h/d"). Metabolic equivalents of the task were analyzed as

reference thresholds of absolute intensities of the physical activities [13]. The pregnancy history, clinical history, gestational weight in each trimester, pre-delivery weight, and delivery condition were obtained from the women's medical documentation in the Women and Children's Medical Care Center.

#### *2.3. Estimation of GWG*

The evaluation indicators of GWG include total GWG across full pregnancy and the GWG rate in the second or third trimester.

First, the total GWG and the GWG rates were calculated as follows:

The total GWG = pre-delivery weight (kg) − pre-pregnancy weight (kg);

The GWG rate <sup>=</sup> Weight at the last obstetrician visit(kg) <sup>−</sup> Weight at the first obstetrician visit(kg) Gestational age at the last obstetrician visit(w) − Gestational age at the first obstetrician visit(w)

Second, to evaluate the adequacy of GWG according to the Institute of Medicine recommendation [14], the value of GWG in participants with different pre-pregnancy BMIs was reassigned by the recommended value. When GWG was within the range of recommended value (adequate): Values = 1; when GWG was below or above the recommended value:

$$\text{Values} = (\frac{\text{GWG}}{\text{the recommended lower limit}} + \frac{\text{GWG}}{\text{the recommended upper limit}}) / 2$$

Values > 1 represent excessive, values < 1 represent insufficient [15].

#### *2.4. Three 24-h Dietary Recalls*

Through a five-stage multiple-pass interviewing technique, three 24-h dietary recalls were conducted by trained researchers to assess the dietary intake in the second and third trimesters [16]. To further reduce recall bias and improve accuracy, the trained researchers explained the recording requirements of dietary recalls to pregnant women a few days before the survey. They suggested taking notes or photos of the food they consume [17]. Three 24-h dietary recalls were performed over consecutive days, including one on the weekend. The evaluation of dietary intake composition did not consider nutrient supplementation. The number of eating episodes was ascertained by the number of caloric events ≥50 kcal, with time intervals between food consumption ≥15 min. Additionally, meal clock timing for each eating episode was recorded. The intake of energy and macronutrients was calculated using the average of the three 24-h dietary recalls by the software Yingyangjisuanqi v2.7.6.10, with the Chinese database as a reference. Energy intakes <500 kcal/d or >3500 kcal/d were excluded from the analysis. Daily energy and macronutrient (carbohydrate, protein, and fat consumption) distribution was calculated as a percentage of total energy and macronutrient intake and divided into four time windows, as mentioned previously.

#### *2.5. Sleep/Wake Time and Daily Time Windows*

At each visit, pregnant women were required to report their usual wake time, bedtime, and sleep onset latency on weekdays. Daily food intake for the participants did not occur during the habitual sleep period on weekdays; therefore, the analysis of time windows was concentrated on the waking period. We divided the waking period into four time windows based on the relationship between the internal circadian time and the sleep/wake cycle [8,18]. The "morning" time window was defined as within 2 h after getting up. The "late morning–early afternoon" time window was defined as from 2 h after getting up to the middle of the waking period. The "late afternoon–early evening" time window was defined as from the middle of the waking period until 2 h before bedtime, and the "night" time window was defined as within 2 h before bedtime.

#### **3. Statistical Analysis**

Pregnant women were classified into mutually exclusive dietary patterns by latent profile analysis. Latent profile analysis could identify unobserved heterogeneity in multiple continuous response variables. The Akaike information criterion (AIC), Bayesian information criterion (BIC), and sample-size-adjusted BIC (aBIC) were used to determine the best-fitting latent profile model. Additionally, the Vuong, Lo, Mendell, and Rubin likelihood ratio test was used to determine whether adding an additional profile contributed to a significantly better-fitting model [19].

For continuous variables, the Shapiro–Wilk test was used to assess the distribution of variables. Data with parametric distribution are described as mean and standard deviation and were compared using the one-way analysis of variance or *t*-test. Data with nonparametric distribution are described as median and interquartile range and were compared using the Kruskal–Wallis or Mann–Whitney U tests. For categorical variables, data are described as numbers (percentages) and were compared using the chi-squared test or Fisher's exact test. Multiple comparisons were conducted using the Bonferroni post hoc test when necessary.

The Sample Wilcoxon Signed Rank Test was used to evaluate the adequacy of GWG in pregnant women with different dietary patterns. Spearman's correlation was used to explore correlations among macronutrient intake in different time windows and the GWG. Multiple Linear Regression Models (method = backward) were used to determine the effects of dietary patterns (independent variables) and macronutrient intake in different time windows (independent variables) on the adequacy of weight gain (dependent variable). Models were adjusted for age, educational level, employment status, household income, pregestational BMI, parity, the condition of gestational diabetes mellitus, gender of offspring, physical activity, daily sleep duration, time node of the time window, number of eating episodes, total energy intake, and gestational week of delivery. *p* < 0.05 was considered statistically significant. Statistical analyses were performed using SPSS version 24 and Mplus version 6.0.

#### **4. Results**

#### *4.1. Dietary Patterns Based on Energy and Macronutrient Distribution*

The model fit information for latent profile analysis model estimation based on one to five latent profiles is shown in Table S1. The Vuong, Lo, Mendell, and Rubin likelihood ratio test did not indicate that the data in the four-class model fit were significantly better than that in the three-class model (*p* = 0.227). However, as the number of latent profiles was raised, the values of AIC, BIC, and aBIC were reduced, and the entropy remained above 0.80. Based on the results of model fit tests, our research objective, and the goal of simplicity, the four-class model was identified as the best description of latent dietary profiles.

The four latent dietary profiles were characterized by average energy and macronutrient distribution in different time windows. Complete data of dietary recalls and sleep/wake time, 6.8% (*n* = 16, N = 234) were classified as having pattern 1, "high night distribution". This group had a relatively high average energy and macronutrient distribution in the night time window. A total of 40.6% (*n* = 95, N = 234) were classified as having pattern 2, "high late afternoon–early evening distribution". This group had relatively high average energy and macronutrient distribution in the late afternoon–early evening time window. Further, 31.2% (*n* = 73, N = 234) were classified as having pattern 3, "high late morning–early afternoon distribution," who had relatively high average energy and macronutrient distribution in the late morning–early afternoon time window, and 21.4% (*n* = 50, N = 234) were classified as having pattern 4, "high morning distribution," who had relatively high average energy and macronutrient distribution in the morning time window (shown in Figure 1).

**Figure 1.** Average energy and macronutrient distribution of time windows in different dietary patterns. Latent profile analysis. E, energy; C, carbohydrate; P, protein; F, fat.

#### *4.2. Participant Characteristics*

The sociodemographic and anthropometric characteristics and the dietary pattern composition did not appear to differ meaningfully across subsequent analyses (shown in Table 1).

#### **Table 1.** The characteristics of participants.


BMI, Body mass index; SD, Standard deviation.

Age, household income, eating episodes, and time node of the "early evening/night" time window of the pregnant women differed meaningfully by dietary profile in the second trimester. Age and time node of the "early evening/night" time window of the pregnant women also differed meaningfully by dietary profile in the third trimester (shown in Table 2).



*Nutrients* **2023**, *15*, 2106

**Table 2.** *Cont*.

significance

Bonferroni correction). Significant tests shown in bold. BMI, Body mass index; MET, Metabolic equivalents of task; SD, Standard deviation.

 identified with Bonferroni correction). d Values were significantly

 different from pattern 4 (α level for statistical significance

 identified with

Daily energy and macronutrient intake did not differ significantly by dietary profile in the second and third trimesters. The energy and macronutrient intake in different time windows differed significantly by dietary profile in the second and third trimesters (shown in Tables 3 and 4).

**Table 3.** The differences in energy and macronutrient intake during the second trimester between each dietary pattern.


<sup>a</sup> Values were significantly different from pattern 1 (α level for statistical significance identified with Bonferroni correction). <sup>b</sup> Values were significantly different from pattern 2 (α level for statistical significance identified with Bonferroni correction). <sup>c</sup> Values were significantly different from pattern 3 (α level for statistical significance identified with Bonferroni correction). <sup>d</sup> Values were significantly different from pattern 4 (α level for statistical significance identified with Bonferroni correction). Significant tests shown in bold. IQR, interquartile range.

In the second trimester, there were 14.3% (*n* =18) of pregnant women with insufficient GWG rates, 33.3% (*n* = 42) with adequate GWG rates, and 52.4% (*n* = 66) with excessive GWG rates. In the third trimester, there were 22.2% (*n* = 24) of pregnant women with insufficient GWG rates, 24.1% (*n* = 26) with adequate GWG rates, and 53.7% (*n* = 58) with excessive GWG rates. There were 23.9% (*n* = 22) of pregnant women with insufficient total GWG, 45.7% (*n* = 42) with adequate total GWG, and 30.4% (*n* = 28) with excessive total GWG.

#### *4.3. The Adequacy of GWG in Pregnant Women with Different Dietary Patterns*

Pregnant women with a high late afternoon–early evening distribution in the second (Median (IQR) = 1.31 (0.70), Z = 3.391, *p =* 0.001) and third trimesters (Median (IQR) = 1.34 (0.72), Z = 3.065, *p* = 0.002), pregnant women with high late morning–early afternoon distribution (Median(IQR) = 1.00(0.68), Z = 3.296, *p =* 0.001) in the second trimester, and pregnant women with high morning distribution in the second (Median(IQR) = 1.35 (0.81), Z = 2.838, *p* = 0.005) and third trimesters (Median (IQR) = 1.68 (1.14), Z = 2.374, *p* = 0.018) appeared to have excessive GWG rates. Pregnant women with a high late morning–early afternoon distribution in the second trimester (Median (IQR) = 1.00 (0.31), Z = 2.374, *p* = 0.018) appeared to have excessive total GWG (shown in Figure 2).

**Figure 2.** The effect of dietary patterns on GWG One Sample Wilcoxon Signed Rank Tests; Multiple Linear Regression Models (method = backward): (**A**) the effect of dietary patterns in the second and third trimesters on the adequacy of GWG rate; (**B**) the effect of dietary patterns in the second and third trimesters on the adequacy of total GWG \* Values were significantly different from 1.

**Table 4.** The differences in energy and macronutrient intake during the third trimester between each dietary pattern.



**Table 4.** *Cont*.

<sup>a</sup> Values were significantly different from pattern 1 (α level for statistical significance identified with Bonferroni correction). <sup>b</sup> Values were significantly different from pattern 2 (α level for statistical significance identified with Bonferroni correction). <sup>c</sup> Values were significantly different from pattern 3 (α level for statistical significance identified with Bonferroni correction). <sup>d</sup> Values were significantly different from pattern 4 (α level for statistical significance identified with Bonferroni correction). Significant tests shown in bold. IQR, interquartile range.

#### *4.4. Correlations between Macronutrient Intake in Different Time Windows and the GWG*

Fat consumption in the late afternoon–early evening of the second trimester was significantly positively correlated to the GWG rate of the second trimester (Spearman γ = 0.192, *p* = 0.031), fat consumption in the morning of the third trimester was significantly positively correlated to total GWG (Spearman γ = 0.220, *p* = 0.022) (shown in Figure 3).

**Figure 3.** Correlations between macronutrient intake in different time windows and the GWG. Spearman's correlation.

#### *4.5. Effect of Dietary Patterns and Macronutrient Intake in Different Time Windows on GWG*

In the second trimester, carbohydrate (β (95% CI): 0.004 (0.000, 0.008); *p* = 0.043), fat (β (95% CI): 0.023 (0.010, 0.036); *p* = 0.001), and protein intake (β (95% CI): 0.015 (0.005, 0.026); *p* = 0.005) in the late afternoon–early evening time window, protein intake in the late morning–early afternoon time window (β (95% CI): 0.016 (0.006, 0.027); *p* = 0.003), and carbohydrate intake in the morning time window (β (95% CI): 0.005 (0.001, 0.010); *p* = 0.018) were positively associated with the GWG rates. Carbohydrate intake in the morning time window (β (95% CI): 0.004 (0.001, 0.007); *p* = 0.005) was positively associated with total GWG, and protein intake in the morning time window (β (95% CI): −0.014 (−0.026, −0.002); *p* = 0.022) were negatively associated with total GWG.

In the third trimester, fat intake in the late morning–early afternoon time window (β (95% CI): −0.023 (−0.044, −0.001); *p* = 0.041) and protein intake in the late afternoon–early evening (β (95% CI): −0.034 (−0.059, −0.010); *p* = 0.007) and morning time window (β (95% CI): −0.042(−0.073, −0.012); *p* = 0.007) were negatively associated with the GWG rates. Carbohydrate (β (95% CI): 0.008 (0.000, 0.016); *p* = 0.037) and fat (β (95% CI): 0.051 (0.017, 0.085); *p* = 0.004) intake in the morning time window was positively associated with the GWG rates. Pregnant women who had a relatively high average energy and macronutrient distribution in late afternoon–early evening time window exhibited a greater GWG rate than in morning time window (β (95% CI): 0.707(0.038, 1.377); *p* = 0.039) (shown in Figure 2). Protein intake in the late afternoon–early evening time window (β (95% CI): −0.013 (−0.024, −0.001); *p* = 0.028), carbohydrate intake in the late morning–early afternoon time window (β (95% CI): 0.004 (0.001, 0.007); *p* = 0.014), and fat intake in the morning time window (β (95% CI): 0.023 (0.011, 0.035); *p* < 0.001) were significantly associated with total GWG. Pregnant women who had a relatively high average energy and macronutrient distribution in late afternoon–early evening time window exhibited a greater total GWG than that in morning time window (β (95% CI): 0.316(0.024, 0.607); *p* = 0.034) (shown in Figure 2).

#### **5. Discussion**

To the best of our knowledge, the present study is the first to be conducted on Chinese pregnant women to investigate the effects of daily energy and macronutrient distribution on GWG during the second and third trimesters.

Our study findings showed that the proportion of pregnant women with an inadequate, adequate, or excessive trimester-specific mean rate of GWG (14.3%, 33.3%, 52.4% in the second trimester; 22.2%, 24.1%, 53.7% in the third trimester) and total GWG (23.9%, 45.7%, 30.4%) was approximately similar to some studies [4,20–22]. However, it differed from a large retrospective cohort study conducted with Chinese singleton pregnant women with gestational diabetes mellitus [23]. This large population-based study was conducted with Chinese singleton pregnant women who delivered between January 2011 and December 2017 in Beijing [24]. However, this could be due to the differences between populations, such as physical conditions and regional dietetic culture.

Though neither daily energy intake nor physical activity differed significantly across all dietary patterns in the present study, we found that pregnant women with the high late afternoon–early evening distribution in the second and third trimesters appeared to have excessive GWG rates; macronutrient (carbohydrate, fat, and protein) intake in the late afternoon–early evening time window of the second trimester was associated with greater GWG rates. Moreover, pregnant women who had a relatively high average energy and macronutrient distribution in late afternoon–early evening time window exhibited a greater GWG rate and a greater total GWG than in the morning time window during the third trimester. These findings are consistent with other studies conducted in pregnant women [15,25] and non-pregnant adults [8,26,27], which supported that higher intake in the evening was associated with a higher risk of weight gain. A potential mechanism may be associated with circadian changes in total energy expenditure, including resting metabolic rate and the thermic effect of food [28]. Randomized crossover trials reported that the endogenous circadian rhythm in the total energy expenditure of healthy adults peaked in the biological morning or early afternoon and was lower in the biological evening [29,30]. If the total energy expenditure was reduced, coupled with high energy-dense food intake, it might cause a positive energy balance in pregnant women as a contributing factor for excessive weight gained.

Additionally, we found that pregnant women with high late morning–early afternoon distribution in the second trimester with high morning distribution in the second and third trimesters experienced excessive GWG rates or excessive total GWG. Protein intake in the late morning–early afternoon time window of the second trimester, carbohydrate intake in the morning time window of the second trimester, carbohydrate intake in the late morning–early afternoon time window of the third trimester, and carbohydrate and fat intake in the morning time window of the third trimester was associated with greater GWG rates or greater total GWG. This finding is inconsistent with other studies conducted on pregnant women [15,25] and non-pregnant adults [8,26,27,31], which supported that higher morning or lunch intake was associated with a lower risk of weight gain. One possible reason is the difference between Chinese and Western food cultures. Taking breakfast as an example, nearly 90% of Chinese ingested cereals and tubers products (rich in carbohydrates), approximately 50% ingested vegetables, fruits, meat, fish, eggs, and milk, and only approximately 30% ingested beans and nuts [32]. In our study, participants with high morning distribution or high late morning–early afternoon distribution did not have a better diet quality for fruit components, milk, and nuts [33,34]. There is no significant difference in micronutrient intake across four dietary patterns (shown in Table S2). Much deep-fried food (rich in carbohydrates and fat) belongs to traditional Chinese breakfasts, such as dough sticks [35]. A previous study showed 4–8 weeks of overfeeding healthy adults with a high-fat breakfast resulted in 2–4 kg of weight gain [36], and a high-fat breakfast did not change satiety a few hours after breakfast [37]. Additionally, an increase of 1 g of carbohydrates was related to an increment of 17 g in weight during pregnancy. In comparison, 1 g of sugar was associated with an increase of 26 g of weight during pregnancy [38].

Interestingly, we found protein intake in the morning time window of the second trimester, protein intake in the late afternoon–early evening and morning time window of the third trimester, and fat intake in the late morning–early afternoon time window of the third trimester were negatively associated with the GWG rates or total GWG. This could be because foods high in protein are typically less energy dense. In healthy women, high-protein intake has a greater effect on satiety and appetite control and less subsequent food intake [39]. Though there is no significant difference in micronutrient intake across four dietary patterns, participants in our study were accustomed to eating food rich in monounsaturated and polyunsaturated fatty acids as snacks in the late morning–early afternoon time window. For instance, in nuts and yogurt, monounsaturated and polyunsaturated fatty acids have been found to contribute to weight loss and obesity prevention [40]. The total GWG was related more closely to eating behavior during the third trimester. This could be due to nonmonotonic fetal growth. However, the biparietal diameter and head circumference show an accelerated increase in the second trimester, while the abdominal circumference and estimated fetal weight velocity peak in the third trimester [41]. Another possible cause was the dietary counseling [42] after BMI monitoring in the second trimester; some pregnant women (65.2%, not shown in the result) in our study changed their dietary patterns after the second trimester.

This study has many strengths. First, we concurrently collected data about sleep timing and meal timing, which enabled us to establish an index indicating the circadian time of food intake. Second, we used a prospective design to assess multiple time points of sleep timing and dietary recalls, which reduced the potential effect of seasonal fluctuations in sleep timing and meal timing. Third, we considered the relationship between relative and absolute energy and macronutrient intake and GWG, since this is achievable in weight management during pregnancy. This study also has several limitations. The study was conducted on relatively healthy women in early pregnancy; therefore, the results of this study cannot be generalized to all pregnant women, especially those with high-risk pregnancies. A small sample size of pregnant women with higher night distribution hinders the observation of the association between the night distribution of energy and macronutrients and GWG. Moreover, a more detailed classification of nutrient consumption was not considered, such as saturated fatty acids, monounsaturated fatty acids, and polyunsaturated fatty acids. Future studies with more comprehensive investigations of sleep status and diet conditions in a larger population of pregnant women are needed.

#### **6. Conclusions**

Excessive GWG of Chinese pregnant women was related closely to eating behavior focused on the late afternoon–early evening time window and carbohydrate and fat intake in the morning during the second and third trimesters. Our findings emphasize that it is necessary to pay attention to Chinese pregnant women with high energy and macronutrient distribution in the late afternoon–early evening and adjust the macronutrient intake based on internal circadian timing for GWG management. Additionally, clinicians should provide more well-directed nutritional advice for pregnant women in different trimesters.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/nu15092106/s1, Table S1: Model fit information for latent profile analysis by number of estimated profiles; Table S2: The differences in micronutrient intake during the second and third trimester between each dietary pattern.

**Author Contributions:** Conceptualization, W.X.; Data curation, X.Z.; Formal analysis, W.X., S.C., J.D., Y.S., Y.H. and Z.Q.; Funding acquisition, S.C. and X.Z.; Investigation, W.X., J.D., Y.S., Y.H., Z.Q., S.J. and Z.L.; Methodology, W.X., L.G. and T.C.; Project administration, X.Z.; Supervision, X.Z.; Writing—original draft, W.X. and S.C.; Writing—review and editing, S.C. and X.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by grants from the National Natural Science Foundation of China (nos. 81973070, 82003449, and 82273655).

**Institutional Review Board Statement:** The present study was approved by the Institutional Review Board of Tianjin Medical University (TMUhMEC2017020) and was conducted in agreement with the ethical norms of the Declaration of Helsinki.

**Informed Consent Statement:** Written consent was obtained from the pregnant women involved in the study.

**Data Availability Statement:** The raw data analyzed in the present study will be available from the corresponding author on reasonable request.

**Acknowledgments:** We acknowledge the Women and Children's Medical Care Center in Hebei and Heping districts of Tianjin China. Special thanks to the pregnant women who agreed to participate in this study.

**Conflicts of Interest:** The authors have no conflict of interest to declare.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

**Jiaomei Yang 1,\*, Qianqian Chang 1, Qiancheng Du 1, Shaonong Dang 1, Lingxia Zeng <sup>1</sup> and Hong Yan 1,2,3**


**Abstract:** The relationship between diet-related inflammation during pregnancy and congenital heart defects (CHD) is unclear. This study attempted to investigate the association between the dietary inflammation index (DII) during pregnancy, reflecting the overall inflammatory potential of the maternal diet, and CHD in Northwest China. A case-control study with 474 cases and 948 controls was performed in Xi'an City, China. Eligible women awaiting delivery were recruited, and their dietary and other information during pregnancy was collected. Logistic regression models were applied to estimate the risk of CHD in association with DII. The maternal DII ranged from −1.36 to 5.73 in cases, and 0.43 to 5.63 in controls. Pregnant women with per 1 higher DII score were at 31% higher risk of fetal CHD (OR = 1.31, 95%CI = 1.14–1.51), and the adjusted OR (95%CI) comparing the pro-inflammatory diet group with the anti-inflammatory diet group was 2.04 (1.42–2.92). The inverse association of maternal DII score with CHD risk was consistent across various subgroups of maternal characteristics. Maternal DII in pregnancy had good predictive value for CHD in offspring, with the areas under the receiver operating characteristic curve higher than 0.7. These findings suggested that avoiding a pro-inflammatory diet in pregnancy should be emphasized in the prevention of CHD.

**Keywords:** dietary inflammatory index; congenital heart defects; pregnancy; Chinese

#### **1. Introduction**

Congenital heart defects (CHD) are the most common congenital disorders globally, with the birth prevalence being 9.41‰ worldwide [1] and 9.00‰ in China [2]. It is estimated that millions of neonates are diagnosed with CHD every year worldwide [1], including 0.15 million in China [2]. CHD is the leading cause of infant morbidity and mortality from birth defects, and responsible for more than 0.26 million deaths globally [3], imposing great burdens on the family and society. The etiology for CHD is largely unknown, but previous research has shown that both genetic and environmental factors may contribute to CHD [4]. The major modifiable risk factors for CHD are generally accepted as maternal smoking, alcohol intake, dietary habits, and environmental exposures [4].

Previous studies have reported that maternal intakes of some nutrients, including folic acid, iron, selenium, zinc, and niacin, are associated with fetal CHD [5–8]. Maternal obesity, diabetes mellitus, and infection during pregnancy are reported to be associated with fetal cardiovascular development [9,10]. These maternal risk factors for CHD are associated with localized and systemic inflammatory cytokine milieu in the placenta and plasma [11]. One study has shown that whole blood cultures derived from mothers with CHD fetuses had higher levels of pro-inflammatory cytokines when activated with mitogen [11], emphasizing the importance of maternal inflammatory conditions in fetal cardiovascular development.

**Citation:** Yang, J.; Chang, Q.; Du, Q.; Dang, S.; Zeng, L.; Yan, H. Dietary Inflammatory Index during Pregnancy and Congenital Heart Defects. *Nutrients* **2023**, *15*, 2262. https://doi.org/10.3390/ nu15102262

Academic Editors: Louise Brough, Gail Rees and Jose Lara

Received: 11 April 2023 Revised: 8 May 2023 Accepted: 9 May 2023 Published: 10 May 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Pregnant women are usually in a low-grade systemic inflammation state due to physiological responses [12]. Diet plays a central role in the regulation of systemic inflammation through pro-inflammatory or anti-inflammatory components of foods and nutrients [13], and is also an important modifiable factor for the prevention of CHD [7,8,14–16]. Thus, it is important to investigate the association between pro-inflammatory diet in pregnancy and CHD to provide optimal recommendations for pregnant women to prevent fetal CHD. The Dietary Inflammatory Index (DII) is a literature-derived score for evaluating the overall inflammatory potential of a person's diet [13]. The DII was determined by peer-reviewed articles about the effect of diet on inflammatory biomarkers [13]. A higher DII score indicates that the diet is pro-inflammatory, while a lower DII score indicates that the diet is anti-inflammatory. The DII has been proven to be of value for the associations with health status in the general population [13], and has also been increasingly used as a predictor of pregnancy outcomes among pregnant women [17,18]. However, to our knowledge there has been no study assessing the association between DII during pregnancy and CHD risk. Previous studies have evaluated some maternal predictors in pregnancy for CHD [14,19,20], giving references for the early prediction of CHD. However, the predictive value of DII for CHD has not been assessed.

The present case-control study in Northwest China attempted to investigate the relationship between DII in pregnancy and CHD and assess the prediction value for DII on CHD.

#### **2. Materials and Methods**

#### *2.1. Study Design and Participants*

Between August 2014 and August 2016, we undertook a case-control study in six comprehensive hospitals in Xi'an City, Northwest China. These six hospitals have incorporated fetal echocardiography at 20th–24th gestational weeks into the routine prenatal ultrasound program to screen for CHD. The detailed study design has been reported previously [8,15,16]. Briefly, among pregnant women awaiting delivery in hospitals, those having fetuses with isolated CHD and no genetic malformation were included in the case group, and those having normal fetuses without any birth defects were included in the control group. Pregnant women with multiple pregnancies or diabetes were excluded because of potentially distinct etiologies. Qualified specialists in each hospital strictly enforced the standard criteria to diagnose birth outcomes. We also undertook a follow-up by telephone within one year after birth to confirm the diagnoses. We randomly selected controls in each hospital each month, and the ratio of the number of controls to cases included in the same hospital in the same month was 2:1. To detect a significant (*p* < 0.05) OR of 1.50 between high and low DII score groups with a statistical power of 80%, 305 cases and 610 controls would be required. A total of 474 cases and 948 controls with completed questionnaires were finally included in the analysis, meeting the sample size requirements.

The study was approved by the Xi'an Jiaotong University Health Science Center (No. 2012008). All participants provided informed consent before the survey.

#### *2.2. Dietary Assessment and DII Score*

We collected maternal diet information throughout pregnancy by face-to-face interviews while awaiting delivery using a semi-quantitative food frequency questionnaire (FFQ). The FFQ consists of 111 food items on the basis of a validated FFQ for pregnant women in Northwest China [21]. Women reported consumption frequency according to eight predefined categories and also recalled the portion sizes with the assistance of food portion images [22,23]. Maternal dietary habits tend to be stable throughout pregnancy [24]; thus, maternal diets throughout pregnancy are comparable with those in the 3rd–8th gestational week, the critical period of fetal cardiovascular development [7,8,15,16]. We applied the Chinese Food Composition Tables to derive maternal nutrient intakes during pregnancy [25,26].

We calculated the DII score using the methods described by Shivappa et al. [13]. We included 30 food parameters to calculate the DII score: 8 pro-inflammatory food parameters (energy, carbohydrate, total fat, protein, cholesterol, saturated fatty acid, vitamin B12, and iron) and 22 anti-inflammatory parameters (fiber, monounsaturated fatty acid, polyunsaturated fatty acid, *n*-3 fatty acid, thiamin, riboflavin, vitamin B6, folic acid, niacin, β-carotene, vitamin A, vitamin C, vitamin E, zinc, selenium, magnesium, caffeine, alcohol, garlic, onion, green/black tea, and pepper) that were available in the current study. We obtained the z-score by subtracting the "standard global mean" from the consumption amount recalled by each pregnant woman and dividing this value by the standard deviation. To minimize the "right skewness", this z-score was converted to a centered proportion. We then multiplied this proportion by the respective food parameter effect score according to the study by Shivappa et al. [13]. We finally summed all of the food-parameter-specific DII scores to create the overall DII score for each pregnant woman. In addition, we constructed a Mediterranean Diet Score (MDS) and a Global Diet Quality Score (GDQS) using the FFQ data according to the methods previously reported [14,27,28].

#### *2.3. Covariates*

Using a structured questionnaire, trained investigators collected the following covariates: (1) sociodemographic characteristics: maternal age, residence, education, work, and parity; (2) maternal health-related factors in early pregnancy: passive smoking, anemia, medication use, and iron/folate supplements use. Maternal age was grouped as two categories (<30 years/≥30 years). Residence included rural and urban areas. Maternal education was divided into two categories (junior high school or below/senior high school or above). Women with no paid employment outside their homes were classified as without employment, otherwise they were classified as in employment. Parity was categorized as two groups (0/≥1). The other covariates were treated as dichotomized factors (no/yes). Women with hemoglobin concentration <110 g/L in pregnancy were diagnosed with anemia.

#### *2.4. Statistical Analysis*

In univariate comparisons, the χ<sup>2</sup> test was adopted for categorical variable, and for continuous variables the Kruskal–Wallis test or Mann–Whitney U test was applied because of the non-normal distributions observed by the Shapiro–Wilk test. Considering the clustering in the design through hospitals, mixed logistic regression models were applied to evaluate ORs (95%CIs) for total CHD and CHD subtypes in association with maternal DII during pregnancy. The DII score was divided into three groups according to the 25th percentile and 75th percentile of the control distribution. The anti-inflammatory diet group was defined if the DII score was lower than the 25th percentile, the pro-inflammatory diet group was defined if the DII score was higher than the 75th percentile, and the intermediate group was defined if the DII score was in the range of the 25th percentile and 75th percentile. Potential confounders were controlled in the models if they were important priori confounders [4,8,29] and changed the estimates by more than 10% [30]. P for trend was calculated by including group specific median value in the model. Subgroup analyses were conducted according to maternal characteristics (maternal age, residence, education, occupation, parity, and maternal passive smoking, anemia, medication use, and iron/folate supplement use in early pregnancy). The interaction between maternal DII and each subgroup factors was tested by the likelihood ratio test comparing regression models with and without an interaction term. Sensitivity analyses were also conducted by dividing participants as three groups according to the tertiles of DII score in the control.

The receiver operating characteristic (ROC) curves were established to estimate the optimal cut-off values of DII during pregnancy for total CHD and CHD subtypes with the maximum Youden index. The areas under the ROC curves (AUCs) showed the accuracy of DII as a predictor for CHD. The AUC values indicated the predictive power as follows: >0.9, very good; >0.8, good; and >0.7, useful [31].

All analyses were conducted using the Stata software (version 15.0; StataCorp, College Station, TX, USA). Two-sided statistical significance was set at 0.05.

#### **3. Results**

#### *3.1. Characteristics of the Study Participants*

The distribution of DII scores in pregnancy among cases and controls is shown in Figure 1. The maternal DII ranged from −1.36 to 5.73 in cases, and 0.43 to 5.63 in controls. Pregnant women in the cases had a higher DII score than the controls (*p* < 0.001), with the medians (25th percentile, 75th percentile) being 4.83 (4.34, 5.23) and 4.63 (4.04, 5.08), respectively. The baseline characteristics of the three groups of maternal DII scores are displayed in Table 1. Among the cases, no difference in maternal characteristics existed among the three DII groups. Among the controls, participants in the intermediate group were more likely to be multipara, and mothers with higher DII score were more likely to take iron/folate supplements in early pregnancy. Maternal residence, education, occupation, parity, and maternal passive smoking, anemia, medication use, and iron/folate supplements use in early pregnancy were significantly different between cases and controls (all *p* < 0.05) (Table S1).


**Table 1.** Characteristics of the study participants according to three groups of maternal DII scores during pregnancy.

DII, Dietary Inflammatory Index. <sup>1</sup> The anti-inflammatory diet group indicates the DII score lower than the 25th percentile of the control distribution, the pro-inflammatory diet group indicates the DII score higher than the 75th percentile of the control distribution, and the intermediate group indicates the DII score in the range of the 25th percentile and 75th percentile of the control distribution. <sup>2</sup> *p* values are from *χ<sup>2</sup>* test for categorical variables and from Kruskal–Wallis test for continuous variables.

#### *3.2. Dietary Intakes and Dietary Quality Scores during Pregnancy among the DII Groups*

Pregnant women with higher DII score in pregnancy showed lower intakes of main food groups, including grains and tubers, vegetables, fruits, dairy, legumes, meats, fish, eggs, and nuts, both in cases and controls (all *p* < 0.001) (Table 2). Pregnant women with higher DII scores also showed lower MDS and GDQS scores in the case and control groups (all *p* < 0.001) (Table 2). Compared with the controls, case mothers had higher intakes of grains and tubers but lower intakes of other main food groups (all *p* < 0.001), and also had lower MDS and GDQS scores (both *p* < 0.001) (Table S2). Participants with higher DII score during pregnancy reported lower intakes of energy, carbohydrate, total fat, protein, cholesterol, fiber, saturated fatty acid, monounsaturated fatty acid, polyunsaturated fatty acid, *n*-3 fatty acid, vitamins (thiamin (vitamin B1), riboflavin (vitamin B2), niacin (vitamin B3), vitamin B6, folic acid (vitamin B9), vitamin B12, β-carotene, vitamin A, vitamin C, and vitamin E), minerals (iron, zinc, selenium, and magnesium), garlic, onion, and pepper both in cases and controls (Table S3). Participants in the cases consumed lower intakes than the controls of all dietary components included in the DII calculation except carbohydrate, caffeine, alcohol, green/black tea, and pepper (Table S4).

**Table 2.** Food groups intake and dietary quality scores during pregnancy according to three groups of maternal DII scores during pregnancy.



**Table 2.** *Cont.*

DII, Dietary Inflammatory Index; MDS, Mediterranean Diet Score; GDQS, Global Diet Quality Score. <sup>1</sup> The anti-inflammatory diet group indicates the DII score lower than the 25th percentile of the control distribution, the pro-inflammatory diet group indicates the DII score higher than the 75th percentile of the control distribution, and the intermediate group indicates the DII score in the range of the 25th percentile and 75th percentile of the control distribution. <sup>2</sup> *p* values are from Kruskal–Wallis test for continuous variables.

#### *3.3. Association between Maternal DII during Pregnancy and CHD*

The associations of maternal DII in pregnancy with the risks of total CHD, ventricular septal defects (VSD), and atrial septal defects (ASD) are displayed in Table 3. Compared with those in the anti-inflammatory diet group, mothers in the pro-inflammatory diet group had a higher risk of delivering fetuses with total CHD (OR = 2.04, 95%CI = 1.42–2.92), VSD (OR = 2.00, 95%CI = 1.25–3.19), and ASD (OR = 1.92, 95%CI = 1.22–3.03), with the tests for trend statistically significant (all *p* < 0.05). The risks of total CHD, VSD, and ASD were increased by 31% (OR = 1.31, 95%CI = 1.14–1.51), 29% (OR = 1.29, 95%CI = 1.07–1.55), and 25% (OR = 1.25, 95%CI = 1.04–1.50) for per 1 higher score of maternal DII in pregnancy, respectively.

**Table 3.** Associations between DII score during pregnancy and congenital heart defects.


DII, Dietary Inflammatory Index. <sup>1</sup> The anti-inflammatory diet group indicates the DII score lower than the 25th percentile of the control distribution, the pro-inflammatory diet group indicates the DII score higher than the 75th percentile of the control distribution, and the intermediate group indicates the DII score in the range of the 25th percentile and 75th percentile of the control distribution. <sup>2</sup> Adjusted for total energy intake, sociodemographic characteristics (maternal age, residence, education, occupation, and parity), and maternal health-related factors in early pregnancy (passive smoking, anemia, medication use, and iron/folate supplements use).

Subgroup analyses showed that the risks of total CHD, VSD, and ASD in association with maternal DII during pregnancy did not alter by maternal characteristics including maternal age, residence, education, occupation, parity, and maternal passive smoking, anemia, medication use, and iron/folate supplement use in early pregnancy (Figures S1–S3). When dividing participants as three groups according to the tertiles of DII score in the control, compared with the lowest tertile group, the highest tertile group showed higher risks of total CHD (OR = 1.66, 95%CI = 1.22–2.28), VSD (OR = 1.55, 95%CI = 1.03–2.33), and ASD (OR = 1.48, 95%CI = 1.08–2.02), with the tests for trend significant (all *p* < 0.05) (Table S5).

#### *3.4. The Prediction Value for Maternal DII during Pregnancy on CHD*

The ROC for maternal DII in pregnancy in the prediction of total CHD, VSD, and ASD is shown in Figure 2. The ROC indicated that maternal DII in pregnancy were useful in predicting total CHD, VSD, and ASD, with the AUC to be 0.79 (0.76, 0.81), 0.78 (0.74, 0.82),

and 0.77 (0.73, 0.80), respectively. The optimal DII cut-off values were 5.41 for total CHD (sensitivity: 67.3%, specificity: 77.3%), 5.31 for VSD (sensitivity: 66.7%, specificity: 79.0%),

**Figure 2.** The ROC for Dietary Inflammatory Index in pregnancy in the prediction of (**A**) total congenital heart defects, (**B**) ventricular septal defects, and (**C**) atrial septal defects. ASD, atrial septal defects; CHD, congenital heart defects; ROC, receiver operating characteristic curves; VSD, ventricular septal defects. The dotted line refers to the reference line that results from random selection.

#### **4. Discussion**

In the current case-control study, we found that higher maternal DII scores, indicating a more pro-inflammatory diet, were associated with higher risks of total CHD and its subtypes in fetuses. These inverse associations of DII score in pregnancy with CHD were consistent across various subgroups of maternal characteristics. We also observed that maternal DII in pregnancy had good predictive value for total CHD and its subtypes. To our knowledge, this is the first study to report data on maternal DII in pregnancy and CHD.

Although there has been no study exploring the relationship between maternal DII in pregnancy and CHD, previous research has shown that maternal pro-inflammatory diet in pregnancy is associated with adverse birth outcomes, such as premature birth, low birth weight, and small for gestational age [17,32,33], which are closely related with birth defects. Moreover, several previous studies have reported CHD risk in association with dietary patterns and dietary quality indices during pregnancy [15,34,35], which share some similar dietary components as the DII. For example, the one-carbon-rich dietary pattern during pregnancy, which was high in fish and seafood, was observed to be associated with a lower risk of CHD [35], and the Mediterranean diet during pregnancy, which was high in whole grains, fruits, vegetables, legumes, nuts, and fish, and, high in olive oil but low in saturated lipids, low to moderate in dairy, and limited in red meat, was reported to reduce CHD risk [14,34]. These similar dietary components may explain why those dietary patterns and scores all showed potential health benefits for fetal cardiovascular development. Compared with other dietary scoring systems such as MDS and GDQS that were also reported to show good predictive value for CHD [14], the maternal DII score reflects the inflammation potential of one diet as a whole and has been shown in high relation with maternal cytokine levels such as TNF-α, IL-1β, IL-8, IL-6, IL-10, MCP-1, and C-reactive protein [36,37]. The DII is based on an extensive literature search on the effect of diet on inflammation and is independent on specific means or recommendations of food/nutrient intake [13], which is different from the MDS and GDQS. Considering the importance of maternal inflammatory conditions on fetal cardiovascular development, the DII provides an easy and noninvasive way to assess the dietary inflammatory potential as a predictor for CHD. Findings from the present study imply that it is important to incorporate the suggestion of avoiding a pro-inflammatory diet in routine pregnancy management practices to prevent fetal CHD.

Several mechanisms may explain the higher risk of fetal CHD associated with higher maternal DII during pregnancy. First, the deleterious effect of a pro-inflammatory diet in pregnancy on fetal CHD may come from the increased pro-inflammatory cytokines. One recent study reported that placental inflammatory monocytes of maternal origin could change the cardiac tissue structure by migrating the embryonic heart [38]. Second, the higher systemic inflammation due to higher DII may cause a stress response, further influencing the normal development of the fetal cardiovascular system [39]. Third, it is possible that dietary inflammatory potential during pregnancy participates in the regulation of gut microbiota [40], which was reported to influence fetal CHD [41]. Fourth, the observed relationship between DII and CHD may be partly due to the low dietary quality of a proinflammatory diet. Previous research has reported that a higher maternal MDS, indicating a higher dietary quality, was associated with a lower DII score [32] and lower risk of CHD [14,34]. In fact, the present study also showed lower MDS and GDQS scores in the pro-inflammatory diet group and in the case group.

Our study provides valuable evidence on the risk of CHD in association with maternal DII score during pregnancy. However, some limitations merit discussion. First, we cannot exclude recall bias because data in pregnancy was recalled by participants awaiting delivery, although previous research indicated that mothers could recall information in pregnancy well after years [42,43]. Second, we cannot exclude exposure misclassification because we gathered dietary data in the entire pregnancy rather than in the 3rd–8th gestational week, the critical period of fetal cardiovascular development. However, previous research has shown maternal dietary habits are usually stable throughout pregnancy [24]. Third, we cannot exclude selection bias because we did not include CHD fetuses who had died before delivery at term. Fourth, we cannot separately assess the relationships between DII and other CHD subtypes because of the limited sample size. Finally, we cannot rule out the possibility of residual confounders, and cannot uncover a real causal relationship because of the case-control design.

#### **5. Conclusions**

The present study suggested that a higher DII score during pregnancy, indicating a more pro-inflammatory diet, was associated with higher CHD risk. Furthermore, the maternal DII score in pregnancy had good predictive value for fetal CHD. Our results implied that avoiding a pro-inflammatory diet could be an interesting target for prevention strategies to reduce the incidence of CHD in Northwest China. Routine pregnancy management should emphasize the importance of reducing dietary inflammation to prevent fetal CHD. Further studies are warranted to investigate the validity of the DII as a predicator for CHD in other populations, and further understand the mechanisms associating dietary inflammation in pregnancy with fetal CHD.

**Supplementary Materials:** The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/nu15102262/s1, Table S1: Characteristics of the study population among cases and controls; Table S2: Food groups intake and dietary quality scores during pregnancy among cases and controls; Table S3: Daily dietary components intake according to three groups of maternal DII during pregnancy; Table S4: Daily dietary components intake during pregnancy among cases and controls; Table S5: Associations between tertiles of maternal DII score during pregnancy and congenital heart defects; Figure S1: Subgroup analyses for the association between per 1 higher score of Dietary Inflammatory Index in pregnancy and the risk of total congenital heart defects; Figure S2: Subgroup analyses for the association between per 1 higher score of Dietary Inflammatory Index in pregnancy and the risk of ventricular heart defects; Figure S3: Subgroup analyses for the association between per 1 higher score of Dietary Inflammatory Index in pregnancy and the risk of atrial heart defects.

**Author Contributions:** J.Y. and H.Y. contributed to study concept and design; J.Y. and Q.C. drafted the initial manuscript; J.Y., Q.C., Q.D., S.D. and L.Z. conducted statistical analyses; J.Y., Q.C., Q.D., S.D. and L.Z. collected the data; J.Y., Q.C. and H.Y. revised the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Natural Science Foundation of China (82103852, 81230016), Shaanxi Health and Family Planning Commission (Sxwsjswzfcght2016-013), and National Key Research and Development of China (2017YFC0907200, 2017YFC0907201).

**Institutional Review Board Statement:** The study was in accordance with the guidelines of the Declaration of Helsinki and approved by the ethics committee of Xi'an Jiaotong University Health Science Center (No. 2012008) on 3 March 2012.

**Informed Consent Statement:** Informed consent was obtained from all participants in the present study.

**Data Availability Statement:** The datasets in the current study are available from the corresponding author on reasonable request.

**Acknowledgments:** The authors are grateful to all mothers who participated in this study, all staff who coordinated field work, and all investigators who contributed to data collection.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
