Dietary patterns and diet quality during pregnancy and low birthweight: The PRINCESA cohort

Abstract Although the isolated effects of several specific nutrients have been examined, little is known about the relationship between overall maternal diet during pregnancy and fetal development and growth. This study evaluates the association between maternal diet and low birthweight (LBW) in 660 pregnant women from the Pregnancy Research on Inflammation, Nutrition,& City Environment: Systematic Analyses (PRINCESA) cohort in Mexico City. Using prior day dietary intake reported at multiple prenatal visits, diet was assessed prospectively using a priori (Maternal Diet Quality Score [MDQS]) and a posteriori (dietary patterns extracted by factor analysis) approaches. The association between maternal diet and LBW was investigated by logistic regression, controlling for confounders. Adherence to recommended guidelines (higher MDQS) was associated with a reduced risk of LBW (OR, 0.22; 95% confidence interval [0.06, 0.75], P < .05, N = 49) compared with the lowest adherence category (reference group), controlling for maternal age, education, height, marital status, pre‐pregnancy body mass index, parity, energy intake, gestational weight gain, and preterm versus term birth; a posteriori dietary patterns were not associated with LBW risk. Higher adherence to MDQS was associated with a lower risk of having an LBW baby in this sample. Our results support the role of advocating a healthy overall diet, versus individual foods or nutrients, in preventing LBW.

Human epidemiology and animal studies support that fetal growth is greatly influenced by maternal nutrition (Abu-Saad & Fraser, 2010;Dimasuay, Boeuf, Powell, & Jansson, 2016;Morrison & Regnault, 2016) and is supposed to be partially mediated by changes in maternal metabolism and hormone levels (Dimasuay et al., 2016).
Maternal diet composition plays a fundamental role early in pregnancy on organ development and differentiation, while in late pregnancy diet can be a major determinant of fetal growth rate and brain development (Jansson, 2016).
Individual nutrient effects on fetal growth have been studied (Brett, Ferraro, Yockell-Lelievre, Gruslin, & Adamo, 2014;Grieger & Clifton, 2014;Kubota et al., 2013;Lager & Powell, 2012;Pannia et al., 2016), those such as iron, zinc, calcium, folate, and n-3 polyunsaturated fatty acids, which have been associated with improved fetal health, healthier birthweight (>2,500 and <4,000 g), and increased rates of maternal and infant survival (Lowensohn, Stadler, & Naze, 2016). In addition, overall diet quality, which refers to the nutritional adequacy and food variety of an individual's dietary intake and its alignment with dietary recommendations, is also relevant. As opposed to the study of single nutrients or foods, indicators of diet quality and dietary patterns offer a broader assessment of overall adequacy of dietary intake (Borge, Aase, Brantsaeter, & Biele, 2017;Okubo et al., 2012).
Maternal diet as measured by dietary patterns as well as by diet quality scores during pregnancy has demonstrated inconsistent associations with birth outcomes. Studies carried out in New Zealand, Japan, and Denmark found that dietary patterns characterized by foods high in saturated trans fats, processed meat, sodium and added sugar, and low in vegetables, fruits, and fibre are negatively associated with birthweight (Knudsen, Orozova-Bekkevold, Mikkelsen, Wolff, & Olsen, 2008;Okubo et al., 2012;Thompson et al., 2010). In contrast, two studies found that no particular dietary pattern was significantly associated with birthweight (Bouwland-Both et al., 2013;Colón-Ramos et al., 2015). Findings related with diet quality are also heterogeneous; four studies suggested that increased diet quality during pregnancy was related to a reduced risk of LBW (Chatzi et al., 2012;Emond, Karagas, Baker, & Gilbert-Diamond, 2018;Rodriguez-Bernal et al., 2010;Timmermans et al., 2012). On the other hand, one study found no association between Mediterranean diet (MD) or Alternative Healthy Eating Index for Pregnancy with fetal growth outcomes (Poon, Yeung, Boghossian, Albert, & Zhang, 2013).
Inconsistent results between studies may be due to the particular food context of each country and also be related to the time point on which the maternal diet is evaluated during pregnancy. Most of the previous studies describing the relationship between dietary patterns in pregnancy and fetal growth have been done in high income countries; thus, there was limited evidence demonstrating this relationship in middle-and low-income populations in which diet can be a major contributor. In addition, every population adds complexity, diversity, and particular eating habits.
The aim of this study was to characterize maternal dietary patterns and diet quality during pregnancy and to evaluate the association with birthweight in a cohort of pregnant women who were clinically monitored on a monthly basis. We tested two different methods to define dietary patterns and evaluate associations with LBW, assuming diet recommendations supported by the scientific literature and adjusted to the Mexican context (Bonvecchio Arenas et al., 2015).

| Study design
We analysed data from a prospective cohort of pregnant women conducted in Mexico City (O'Neill et al., 2013)

Key messages
• Our findings provide evidence that higher adherence to a good quality diet during pregnancy is associated with lower risk of having a low birthweight baby.
• Characterization of dietary patterns during pregnancy using a priori and a posteriori approaches may lead to identify food groups with major contribution to diet quality and significative impact on birthweight and other perinatal outcomes.
• The present work provides valuable new knowledge to be used for improvement of pregnancy care guidelines and complementary approaches to public health.
Inclusion criteria were (a) reliable recall of last menstruation; (b) agreement to prenatal visits every 4 weeks throughout their current pregnancy; and (c) written consent for their inclusion in the study.
Exclusion criteria were (a) previous presence of any medical or obstetric complication in the current pregnancy and (b) presence of multiple fetuses. Women who developed pregnancy complications such as gestational diabetes and preeclampsia were referred to a specialty hospital for follow-up. Eligibility was determined at screening and confirmed at the first visit. For the present study, an additional inclusion criterion was to have at least one complete dietary recall in both the second and third trimesters of pregnancy.
After screening for eligibility and informed consent, given at the first visit or at health clinics during recruitment, women were seen monthly over the course of their pregnancies. Information on clinical, anthropometric, and biochemical parameters and maternal diet was collected at each visit by a dedicated team composed of certified medical personnel and nutritionists with standardized training.

| Dietary variables
Data on maternal diet were collected through a multiple-step 24-hr dietary recall format (24H-DR) in the second and third trimesters of pregnancy by a nutritionist with standardized training. The multiple pass method is a five-step approach developed by the U.S. Department of Agriculture and designed to enhance the quality of the information from the 24H-DR (Blanton, Moshfegh, Baer, & Kretsch, 2006;Conway, Ingwersen, & Moshfegh, 2004).
Dietary information in the cohort was collected as individual foods; for the present analysis, foods made of more than one ingredient (e.g., sandwich) were disaggregated into their ingredients, except for beverages, fast food, and fried snacks, which were kept as a single unit. Food portion size was calculated according to the national reference system (Perez Lizaur, Palacios Gonzalez, & Flores Galicia, 2014).
We estimated daily intake for total energy, fat, protein, carbohydrates, and fibre. In addition, daily intakes of added sugars, calcium, iron, folate, and polyunsaturated, monounsaturated and saturated fats were analysed. The estimation of daily intake of energy and nutrients was calculated by using a food composition table compiled by the Mexico's National Institute of Public Health (Instituto Nacional de Salud Pública, 2012). To estimate added sugars, we used the method proposed by Louie et al. (2015) and also used by Sanchez-Pimienta, Batis, Lutter, and Rivera (2016) in the dietary analysis of the 2012 Mexican National Health and Nutrition Survey (Encuesta Nacional de Salud y Nutrición).
Before identification of dietary patterns, 693 individual foods were collapsed into 123 food groups to reduce complexity; these food groups were created on the basis of expected similar nutrient content.
To simplify some analysis and results, we further aggregated these 123 food groups into 12 major food groups, using a grouping similar to one proposed for Mexican food items (Aburto, Pedraza, Sanchez-Pimienta, , also shown in Table S1. To evaluate diet quality, we built a Maternal Diet Quality Score . The MDG discourage the intake of foods high in sugar, fat, and energy density and highly processed foods, but the MDG do not give guidelines for specific amounts of these foods. We used an upper limit of 10% of energy intake as the recommendation for HSFAS products proposed by Batis, Aburto, Sanchez-Pimienta, Pedraza, and Rivera (2016) for the Mexican healthy diet.

| Birthweight
Offspring's birthweight was obtained from medical records. Birthweight (grams) was divided into three categories: <2,500 for low, 2,500-3,999 for normal, and >4,000 for high according to the WHO classification (WHO, 2004).

| Potential confounders and intermediate variables
Maternal age, education, and number of pregnancies (parity) were obtained using questionnaires that collected data on sociodemographic variables, obstetric history, and detailed information about the pregnancy, including gestational age at birth. Maternal education was grouped by completed or no completed basic school (≥9 years and <9 years). Parity was divided into three groups (nulliparous, 1-2, and ≥3). Marital status was divided into two groups (married/partnered and divorced/single). Maternal height was measured at the first visit by trained staff using standardized methods (Lohman technique). Pre-pregnancy weight was self-reported by participants. The pre-pregnancy body mass index (pBMI) was calculated as pBMI = kg/m 2 and was categorized into five groups: underweight, ≤18.5; normal, ≥18.5-24.9; overweight, 25-29.9; obesity 1, 30-34.9; and obesity 2, ≥35.
Maternal weight was measured at the first and consecutive visits by trained staff using standardized methods (Lohman technique). Rate of gestational weight gain (RGWG; kg per week) was calculated in second and third trimesters and over the whole pregnancy. We categorized RGWG according to whether Institute of Medicine (Rasmussen & Yaktine, 2009) recommendations were met (insufficient, adequate, and excessive) based on ranges of the mother's pBMI.
Recommended weight gain in the second and third trimesters was based on the assumption that underweight, normal weight, overweight, and obese women should gain weight within the normal range of 0.44-0.58, 0.35-0.50, 0.23-0.33, and 0.17-0.27 kg per week, respectively.

| Statistical analysis
Descriptive statistics were computed for socio-demographic variables and maternal characteristics.

| Dietary patterns
Factor analysis (FA) was applied to the women's daily intake (percentage of energy contribution) of each of the 12 food groups in order to reduce the large amount of diet information obtained into a smaller set of independent (non-correlated) factors (Yong & Pearce, 2013).
These factors reduced the 12 food groups to a smaller number based on their similar variability and allowed identification of specific dietary patterns (Thompson et al., 2010;Tucker, 2010) to facilitate interpretation of results. Percentage of energy contributed by each food group to daily total energy intake for each individual was obtained using the following formula: (Energy intake by food group * 100)/(Total energy intake). We used the percentage of variance explained by each factor and a screen plot to determine the number of factors (Yong & Pearce, 2013). Distinct dietary patterns were defined for the second and third trimesters and for both periods together, characterized by high loadings of specific food groups during these intervals. Food groups were considered to be descriptive of the "dietary pattern" if their factor loadings had a magnitude of 0.3 or greater (Knudsen et al., 2008;Varraso et al., 2012). The signs of the loadings show the direction of the correlation between each factor and food group (Yong & Pearce, 2013).
For each pregnant woman, a factor score in the respective dietary pattern was estimated, and then individual factor scores were divided into tertiles. FA allowed us to obtain dietary patterns by computing coefficients for each food group in the analysis; individual dietary pattern scores were calculated by multiplying these coefficients by the individual's consumption of the groups to provide a natural score for every participant (Crozier, Robinson, Godfrey, Cooper, & Inskip, 2009;Elstgeest, Mishra, & Dobson, 2012).
To evaluate the similarity of the factor loadings across trimesters in each dietary pattern for each woman, we estimated a coefficient of congruence (Lorenzo-Seva & Ten Berge, 2006). This coefficient is an index of factor similarity and is estimated as s were met and 0 if no recommendations were met. We defined the three following categories of adherence: low (0-2 points), medium (3-4 points), and high (≥5 points).
Differences of nutrient intakes and socio-demographic and maternal characteristics across the intake patterns and categories of MDQS were compared using chi-square and analysis of variance for categorical and continuous variables, respectively.
Multivariable linear and logistic regressions were used to assess the association between maternal diet (dietary patterns and MDQS) and the risk of having an LBW infant; models were adjusted for potential confounders (parity, baby's sex, mother's height and age, education, gestational age at the end of pregnancy, education level, and pBMI). We included the potential effect for RGWG in our models because it has been proposed as a potential intermediate variable in the association between maternal diet and fetal growth. Using FA on a total of 12 food groups, we identified two distinct dietary patterns in the second and third trimesters. The third and subsequent factors explained less variation than the first two and were less interpretable, so they were not considered further. Dietary pattern for whole pregnancy included the total number of 24H-DR obtained during second and third trimesters of pregnancy. This pattern was used in the association models because the sum of report intake represents more accurately the usual food intake.
The first factor explained 12.89% and 13.03% of the variation in the dietary data for the second and third trimesters, respectively. In both trimesters, this factor was characterized by high intake of white meat and eggs, low fat dairy products, cereals, tubers, fruits, and vegetables and low intake of HSFAS, sugary drinks, juices, and sodas; we termed this the "healthier dietary pattern." T A B L E 1 Socio-demographic and maternal characteristics in a sample of 660 women from the PRINCESA cohort Factor 2 explained between 11.83% and 11.22% of the variation in the second and third trimesters, respectively. It was characterized in the second trimester by high intakes of sugary drinks, juices and sodas, red and processed meat, cereals, and tubers. In the third trimester, it was characterized by high intakes of HSFAS, red and processed meat, and dairy products (both low and high fat); unlike in the second trimester, this factor was not characterized by a high intake of sugary drinks, juices, and sodas, so we termed it a "mixed dietary pattern." We used the denomination of "mixed dietary patterns" in second and third trimesters because factors for both dietary patterns were generated sequentially with the same 12 food groups (Table S1). The low congruence between trimesters can be explained by increase in adherence to the "healthier dietary pattern." It is important to mention that the sequential nature of the determination of factors implies the possibility that some women changed scores between trimesters.
Younger maternal age (24.25 ± 5.2 vs. 26.20 ± 6.2 years, P < .001) and nulliparity (35.64 vs. 29.33%, P < .08) were associated with lower adherence to healthier dietary pattern (tertile 1 vs. tertile 3). We did not observe significant differences with respect to other characteristics or across the tertiles of the mixed diet pattern.
The coefficient of congruence between trimesters 2 and 3 was very high (0.98) for the healthier dietary pattern and low (0.31) for the mixed dietary pattern. The factor loadings of the food groups from FA in the second and third trimesters and during the complete pregnancy are shown in Table 2.
On average, women with scores in the highest tertile of the two patterns reported less energy intake, higher intake of cereals and tubers, and lower intake of HSFAS (P < .05). Food groups with high positive factor loadings were highest in tertile 3, whereas food groups with high negative factor loadings were highest in tertile 1 for both dietary patterns. Food group intakes according to tertiles of dietary patterns are presented in Table S2. In accordance with the previous data, we also identified that intake of energy and some nutrients such as carbohydrates, total fats, SF, PUFAs, added sugars, fibre, and iron were higher (P < .005) in the second trimester compared with the third trimester. In contrast, intakes of folate and calcium were higher in the third trimester compared with the second trimester (P < .05). were among the groups with the lowest %EC in both trimesters. Figure S1 shows the percentage contributions of each food group to total energy intake in the second and third trimesters. These observed differences in MDQS, nutrient intakes, and energy contribution of food groups between trimesters indicate that the diet composition of women is modified for the better as pregnancy progresses. Abbreviations: CI, confidence interval; OR, odds ratio. a Logistic models adjusted for energy intake (continuous), dietary patterns (healthier and mixed) were mutually adjusted, pre-pregnancy BMI (normal, overweight, obesity 1, and obesity 2), parity (nulliparous, 1-2, and ≥3 pregnancies), gestational weight gain (insufficient, adequate, and excessive), maternal age (tertiles), maternal height (continuous), marital status (nonpartnered and married/partnered), maternal education (basic ≤ 9, superior > 9 years), term of gestation (preterm, term), and baby's sex (female, male).  (Knudsen et al., 2008;Okubo et al., 2012;Thompson et al., 2010).
In relation to consumption of discretionary food groups (sugary drinks, juices, sodas, HSFAS products, sugar, and candies), we observed that contribution of the total energy intake was lower in the third trimester than in the second trimester. We also identified that intakes of energy, carbohydrates, total fats, SF, and added sugars were higher in the second trimester compared with the third trimester.
With respect to diet quality, the proportion of women with high adherence to MDQS was greater in the third trimester than in the second trimester. However, we identified that the healthier dietary pattern was very similar between trimesters 2 and 3, but the mixed dietary pattern was not.
Only one study has explored dietary intake changes during pregnancy among 12,572 women in the United Kingdom; the results support that dietary patterns are similar throughout pregnancy, but diet composition quality is modified for the better as pregnancy progresses (Crozier et al., 2009), as observed in our study.
The principal strengths of the present study include the prospective design that provided a valuable opportunity to assess dietary differences in the second and third trimesters for the first time in low-income urban women. Our cohort included women with uncomplicated pregnancies, and thus, our findings could be relevant to a wide population of women worldwide. The results on diet composition and characterization of dietary patterns also provide insights into which foods are more accessible in this population context.
The high consumption of unhealthy food groups (HSFAS and SSBs) reported by these women could be explained by the lower prices and high market availability for this type of food, whereas the lower consumption of fruits, vegetables, legumes, and other traditional foods may be due to higher prices and low salaries (CONEVAL, 2015) that make it difficult for pregnant women to access healthy foods and high quality diets.
To our knowledge, this is the first study to report the characterization of maternal diet during pregnancy using two approaches: a posteriori (FA) and a priori (MDQS). In Mexico, no published studies to date have evaluated dietary patterns or diet quality in pregnant women. Considering that there is currently not one best approach to study overall diet, the use of these complementary approaches that include classification of diet patterns by selected nutriments and foods (a priori) and dietary patterns derived from specific populations (a posteriori) may be useful for identification of food groups that may contribute to pregnant women health.
The present work provides valuable new knowledge on areas of opportunity to improve the quality, dietary patterns, and composition of the diet in pregnant women who live in similar contexts. In summary, our findings provide evidence that higher adherence to MDQS is associated with lower risk of having a LBW baby. Our results are important for design and implementation of policy-based health prevention programs because they support the role of a healthy overall diet in preventing negative pregnancy outcomes such as LBW rather than promoting individual nutrient supplements or avoidance of individual foods or nutrients. This represents a more comprehensive and complementary approach to public health.
Eating a healthy diet during pregnancy is crucial for the future health of the unborn child and future generations; thus, all pregnant women should be encouraged to eat a healthier dietary pattern and high quality diet, using dietary recommendations that are simple, accessible, and well suited to the population context.
Further investigation of these findings, by replicating this methodology in different regions to identify social and environmental aspects related to accessibility of a better quality diet, is warranted.