Globally, over fifteen million infants are born preterm every year. Preterm birth (live birth < 37 weeks of gestation) and small for gestational age (SGA), which are linked with low birth weight (LBW) (live birth weight < 2500 g), are also significant causes of neonatal morbidity and mortality( 1 , Reference Hughes, Black and Katz2 ). Over 80 % of the world’s 2·5 million infants who die each year are of LBW, and the majority of these have been reported from low- and middle-income countries( 3 ). Preterm birth has been found to be associated with the diminishing child motor development and academic performance( Reference Moreira, Magalhães and Alves4 ) and neurodevelopmental impairment( Reference Schieve, Tian and Rankin5 ). It also has long-term effects on the risk of cardiovascular, pulmonary and metabolic diseases( Reference Marret, Ancel and Marpeau6–Reference Bonamy, Bendito and Martin8 ). Moreover, a number of studies have documented that LBW is associated with obesity, diabetes, hypertension and kidney diseases later in life( Reference Kensara, Wootton, Phillips, Patel, Jackson and Elia9–Reference Luyckx and Brenner12 ).
Maternal nutrition has a significant role in ensuring successful birth outcomes( Reference Ramakrishnan, Grant and Goldenberg13 ). Many studies have examined the association of maternal intake of single or a few nutrients, and/or foods, with adverse birth outcomes. In recent decades, there has been a growing interest in using a dietary pattern approach to assess overall dietary intakes because people do not habitually consume isolated single nutrients or single foods. It is also difficult to examine the separate effects of some nutrients due to the high level of inter-correlation among nutrients, for example, Mg and K. Analyses of single nutrient might potentially be confounded by the effect of dietary patterns since nutrients are commonly associated with certain dietary patterns. Therefore, the synergistic or antagonistic effects of overall dietary patterns could be significantly large enough to be measurable( 14–Reference Sacks, Obarzanek and Windhauser16 ).
A few prospective cohort studies have shown a beneficial effect of dietary patterns such as ‘prudent diets’, ‘Mediterranean diets’ and ‘Dietary Approaches to Stop Hypertension (DASH)’ during pregnancy, on lowering the risk of preterm birth( Reference Englund-Ogge, Brantsaeter and Sengpiel17–Reference Martin, Sotres-Alvarez and Siega-Riz20 ). These dietary patterns were mostly characterised by high consumption of vegetables, fruits and whole grains. Some prospective studies have shown a detrimental effect of ‘Western diets’ on increasing the risk of preterm birth, LBW and SGA( Reference Rasmussen, Maslova and Halldorsson21–Reference Knudsen, Orozova-Bekkevold and Mikkelsen23 ). This dietary pattern is commonly characterised by processed meat, potatoes, sweetened snacks and saturated fats, which contain pro-inflammatory nutrients, which act as a stressor on the hypothalamic-pituitary-adrenal system and subsequently increase the risk of adverse birth outcomes( Reference Shin, MohanKumar and Sirivelu24 , Reference Jeffery and O’Toole25 ).
Current prospective studies show an association between maternal dietary patterns during pregnancy and adverse birth outcomes; however, the findings remain inconsistent. In 2016, a traditional review was published by Chen et al. ( Reference Chen, Zhao and Mao26 ) and a systematic review by Kjollesdal et al. ( Reference Kjøllesdal and Holmboe-Ottesen27 ). These reviews focused specifically on the association between dietary patterns during pregnancy and birth weight or pregnancy outcomes, including gestational diabetes mellitus and hypertension disorder in pregnancy. They also used a single database, PubMed to search potential articles and considered publications up to 2015. A large number of studies have been published since Chen et al. ( Reference Chen, Zhao and Mao26 ). To our knowledge, there has been no current review conducted on the effects of maternal diets before and during pregnancy on preterm birth, LBW and SGA. The present study aimed to systematically review current evidence on the relationships between maternal diets, including single or selective food items and overall diets before and during pregnancy, and preterm birth, LBW and SGA.
Methods
Searching strategy
The Preferred Reporting Items for Systematic Reviews and Meta-Analyses guideline was followed to develop the systematic review( Reference Liberati, Altman and Tetzlaff28 ) and registered at PROSPERO 2018 CRD42018098714. A systematic search of PubMed, Embase, Scopus and the Cochrane library was carried out using keywords and subject search terms, including Mesh (PubMed) and Emtree (Embase). Additional relevant articles were also identified by Snowballing( Reference Greenhalgh and Peacock29 ) and Pearl growing methods( Reference Schlosser, Wendt and Bhavnani30 ). Publications from February 2002 to August 2018 were included, and the search was limited to studies published in the English language. The following search terms were used in different combinations of keywords, subject and Boolean searching: Maternal AND diet* OR food* OR nutrient* OR ‘dietary pattern’ OR ‘meal pattern’ OR ‘dietary habit’ OR ‘dietary intake’ OR ‘dietary consumption’ OR ‘eating consumption’ OR ‘eating pattern’ OR ‘eating behaviour’ OR ‘eating habit’ OR ‘food consumption’ OR ‘food intake’ OR ‘food habit’ OR ‘nutritional consumption’ OR ‘nutritional habit’ OR ‘nutritional pattern’ AND ‘preterm birth’ OR ‘premature birth’ OR ‘preterm delivery’ OR ‘premature delivery’ OR ‘preterm labour’ OR ‘premature labour’ OR ‘birth outcome’ OR ‘Low Birth Weight’ OR ‘Infant, Small for Gestational Age’ OR ‘Infant, Very Low Birth Weight’ OR ‘Infant, Extremely Low Birth Weight’ OR ‘birth weight’.
Study selection
Any studies that observed maternal diets before and during pregnancy as the exposure variable and used preterm birth, LBW and SGA as outcome variables were included. Both observational (cross-sectional, case–control, cohort studies) and interventional studies (randomised controlled trials (RCT) and non-RCT) were included. Appropriateness of the statistical analyses was assessed, and studies with a high risk of bias, review articles and commentaries were excluded. Animal studies, conference papers and studies without full text were also excluded. Moreover, studies that focused on nutrients, sugar-sweetened or artificially sweetened beverages, alcohol intake, caffeine consumption and nutritional supplements alone or not in combination with dietary intake were excluded. Furthermore, studies that examined fish consumption contaminated with harmful substances (acrylamide, mercury and dioxin) or evaluated a biomarker of nutritional intake were excluded.
Dietary patterns are defined as the quantities, proportions, variety or combinations of different foods and beverages in diets, and the frequency with which they are habitually consumed( 31 ).
Selective diets are the intake of specific foods (not overall diets), created based on previous knowledge without using statistical or extraction method.
Data extraction
Data extraction and quality assessment were conducted by two independent reviewers (D. G. G. and M. W.). The following information was extracted in the observational studies: author, year, study area, study design, sample size, population characteristics, dietary assessments (method, period, course, extraction method and food groups), outcomes assessments (outcomes and percentage of cases), main findings and potential confounders (Table 1). If the studies identified were intervention studies, the following data were added: dietary assessments (follow-up period, control and interventional diets, intervention and control food groups and nutritional supplementation) and outcomes assessments (outcomes, percentage of cases in intervention and control groups) (Table 2).
Characteristics of observational studies with maternal diets and adverse birth outcomes (preterm birth, low birth weight (LBW) and small for gestational age (SGA))
PCA, principal component analysis; INMA, Infancia y Medio Ambiente; VFR, fruits–vegetables–rice; SFN, seafood and noodle; PCP, pasta–cheese–processed meat; AHEI, Alternative Healthy Eating Index; GI, glycaemic index; GL, glycaemic load; NND, new Nordic diets; DASH, Dietary Approaches to Stop Hypertension; DHQ, Diet History Questionnaire; HEI, healthy eating index.
Characteristics of interventional studies with maternal diets and adverse birth outcomes (preterm birth, low birth weight (LBW) and small for gestational age (SGA))
RCT, randomised controlled trial; MMN, multi micronutrients; FFS, fortified food supplements; CSB, maize–soya blend diets; CLD, cholesterol-lowering diets; HGI, high glycaemic index; LGI, low glycaemic index.
Quality assessment
The Newcastle–Ottawa Scale for observational studies( Reference Wells, Shea, O’Connell and Peterson64 ) and the Cochrane handbook for interventional studies( Reference Higgins, Altman, Sterne, Higgins and Green65 ) were used to assess the risk of bias for the eligible studies. Selection, comparability and outcome assessment were rated separately for observational studies, and the rating scores ranged from 0 (highest degree of bias) to 9 (lowest degree of bias) (online Supplementary Tables S3–S5). The risk of bias during selection, performance, detection, attrition, reporting and other potential sources was assessed for interventional studies, and the quality assessments were recorded as high risk of bias (X), low risk of bias (√) or unclear (?) (online Supplementary Table S1).
Data synthesis and analysis
Narrative analysis and qualitative summarisations were conducted to synthesise the included articles. We did not undertake meta-analysis due to wide diversity of maternal dietary intake. There were large variations in the methods, periods, extraction methods and food items across the studies and also for cut-offs for outcomes.
Results
A total of 23 363 records were identified by the electronic database searches (PubMed, Embase and Scopus). After removing duplicates, 10 429 records were screened by title reading and 183 records were screened by abstract reading. Of these, seventy-eight studies met the criteria for full document review. A further nineteen potentially eligible articles were found by manual search (snowballing and pearl growing) and from the Cochrane library. Following a thorough review of the articles based on the inclusion and exclusion criterion, forty eligible articles were included overall in the systematic review (Fig. 1).
Flow diagram showing the number of articles sourced at each stage of the systematic review.
Study characteristics
The thirty-four observational studies were mostly prospective cohort studies (thirty), three studies were case–control and one study was retrospective cross-sectional. Also, of six interventional studies, five were RCT and one was not. The majority of observational studies were from the Norway (Norwegian Mother and child cohort study (MoBa)) and Denmark (Danish national birth cohort). The study samples ranged from 309 to 72 072 in prospective cohort studies with the largest samples derived from the MoBa study. The sample sizes in case–control and RCT studies ranged from 1714 to 2521 and 290 to 1296, respectively. One non-RCT study was conducted on small samples (thirty-five intervention, thirty-five control group) (Tables 1 and 2).
Almost all observational studies used validated FFQ to assess maternal dietary intake. Two studies used a diet history questionnaire, and one study used an automated self-administered 24 h recall tool (ASA24). The FFQ ranged from fifteen food items in Italy to 360 items in Denmark (Table 1).
Three methods were widely used to evaluate the relationships between maternal dietary patterns and adverse birth outcomes: Factor analysis, cluster analysis and different dietary indices, such as the healthy eating index, Mediterranean diet, DASH, new Nordic diets and glycaemic index. The majority of studies (nine) used maternal dietary patterns derived from principal component analysis to examine the risk of adverse birth outcomes (Tables 1 and 2).
The majority of studies (40 %) assessed maternal dietary intake at the second trimester. Neonatal data were collected after delivery, from medical records and hospital registries. The percentage of preterm births ranged from 3·0 to 14·7 %, the percentage of LBW from 0·7 to 20·7 % and of SGA from 3·4 to 49·2 % (Tables 1 and 2).
The majority of included studies did not control for possible covariates, such as pregnancy complications (gestational diabetes mellitus and hypertension disorder in pregnancy), nutritional supplementations (Fe, Zn and folic acid) and caffeinated beverages (coffee, tea and cola). Over one-quarter of studies used self-administered postal questionnaires to collect maternal dietary intake, which might have selection bias (online Supplementary Table S2).
The quality assessment scores ranged from 3 to 8 for prospective cohort studies, 7 to 8 for case–control studies and 3 to 6 for cross-sectional studies. However, one prospective cohort( Reference Akbari, Mansourian and Kelishadi66 ) and one cross-sectional study( Reference Naghavi, Mahdavi and Moradi67 ) were excluded due to having a high risk of bias. The major concerns of included articles were selection bias (did not ascertain exposure) and comparability (did not control for any additional pertinent factors) (online Supplementary Tables S3–S5). The common concern for intervention studies was performance bias (the study participants and investigators were not blinded for the intervention allocation) (online Supplementary Table S1).
In the present review, the definition of SGA varied from birth weight < 2·5th percentile( Reference Knudsen, Orozova-Bekkevold and Mikkelsen23 ) to < 5th percentile( Reference Bouwland-Both, Steegers-Theunissen and Vujkovic32 , Reference Heppe, van Dam and Willemsen40 , Reference Heppe, Steegers and Timmermans41 ) to <10th percentile of gestational age( Reference Chia, de Seymour and Colega34 , Reference Emond, Karagas and Baker35 , Reference Halldorsson Th, Meltzer and Thorsdottir38 ). The majority of studies (seventeen articles) defined SGA as birth weight below the 10th percentile for the gestational age. However, three articles defined SGA as <5th percentile for the gestational age – these studies showed no association between SGA and maternal diet( Reference Bouwland-Both, Steegers-Theunissen and Vujkovic32 , Reference Heppe, van Dam and Willemsen40 , Reference Heppe, Steegers and Timmermans41 ).
There is well established evidence that preterm birth is the primary cause of LBW. Only one study excluded extremely preterm (<28 weeks) and post-term (>42 weeks) birth in the final analysis between maternal diet and LBW( Reference Brantsaeter, Birgisdottir and Meltzer33 ). For the analyses of associations between maternal diet and SGA, five studies( Reference Lu, Chen and He44 , Reference Mitchell, Robinson and Clark47 , Reference Poon, Yeung and Boghossian55 , Reference Ricci, Chiaffarino and Cipriani56 , Reference Thompson, Wall and Becroft58 ) excluded only preterm birth and two studies( Reference Mendez, Plana and Guxens46 , Reference Okubo, Miyake and Sasaki51 ) excluded both pre- and post-term birth from the final samples.
Maternal diets during pregnancy and preterm birth
Sixteen eligible articles reported on the association between maternal dietary patterns during pregnancy and preterm birth – from these articles, seven studies have shown beneficial effects of maternal diet on the risk of having preterm infants (Table 3). Greater adherence to maternal diet quality, including prudent, traditional, DASH, and vegetable–fruits–rice (VFR) diets were associated with lowering the risk of preterm birth( Reference Englund-Ogge, Brantsaeter and Sengpiel17 , Reference Martin, Sotres-Alvarez and Siega-Riz20 , Reference Chia, de Seymour and Colega34 ). A large prospective cohort study conducted in Norway, MoBa observed that prudent diets (characterised by high intake of vegetables, fruits, whole grains, cereals, oils, water as a beverage and fibre-rich bread) and traditional diets (high intake of potatoes and fish) lowered the risk of preterm birth( Reference Englund-Ogge, Brantsaeter and Sengpiel17 ). Martin et al. also reported that the DASH dietary pattern (high intakes of fruits, vegetables, nuts and legumes, low-fat dairy products and whole grains; low intake of Na, red and processed meats and sweetened beverages) was significantly associated with a lower risk of preterm birth( Reference Martin, Sotres-Alvarez and Siega-Riz20 ). Furthermore, Chia et al. found that those with high adherence to a VFR pattern had a lower risk of preterm birth( Reference Chia, de Seymour and Colega34 ).
Associations between maternal dietary patterns and preterm birth
VFR, vegetables–fruits–rice; AOR, adjusted OR; SFN, seafood and noodle; PCP, pasta–cheese-processed meat; HR, hazard ratio; GI, glycaemic index; GL, glycaemic load; NND, new Nordic diets; FFS, fortified food supplements; MMN, multi micronutrients; CSB, maize–soya blend diets; CLD, cholesterol-lowering diets; RR, risk ratio; DASH, Dietary Approaches to Stop Hypertension.
↓, Negative association; ↔, null association; ↑, positive association.
In contrast, Rasmussen et al. and Martin et al. reported that Western diets characterised by high intake of salty and sweet snacks, processed meat, white bread and desserts, and a dietary pattern characterised by collard greens, coleslaw or cabbage, maize bread, red and processed meats, whole milk and vitamin C-rich drinks increased the risk of preterm birth( Reference Martin, Sotres-Alvarez and Siega-Riz20 , Reference Rasmussen, Maslova and Halldorsson21 ). Two prospective cohort studies documented that high glycaemic index diets, characterised by high intake of carbohydrates, added sugar and dietary fibre( Reference Englund-Ogge, Birgisdottir and Sengpiel36 ), and Western diets (high intake of salty and sweet snacks, white bread, desserts and processed meat products) had no significant association with risk of preterm birth( Reference Englund-Ogge, Brantsaeter and Sengpiel17 ).
Inconsistent findings were found in association between Mediterranean diets and preterm birth (Table 3). Mikkelsen et al. observed that Mediterranean diets (characterised by high intake of olive or rapeseed oil, intake of fish ≥2/week, >5/d fruits and vegetables, meat twice a week and two cups of coffee per d) were associated with lowering risk of early preterm birth( Reference Mikkelsen, Osterdal and Knudsen18 ). Saunder et al. reported that there was no association between adherence to Mediterranean diets (characterised by intake of vegetables, legumes, fruits and nuts, cereals, fish, meat and poultry, dairy products, alcohol and fat) and the risk of preterm birth( Reference Saunders, Guldner and Costet57 ). Moreover, in a study conducted in Norway, maternal adherence to the Mediterranean diet was not associated with preterm birth risk( Reference Haugen, Meltzer and Brantsaeter39 ).
Limited evidence was found on the relationship between maternal milk consumption and preterm birth (Table 4). A prospective cohort study conducted in Norway showed that high consumption of milk-based probiotic products was significantly associated with reduced risk of spontaneous preterm birth( Reference Myhre, Brantsaeter and Myking48 ). However, Heppe et al. reported that there was no association between milk consumption and the risk of preterm birth( Reference Heppe, van Dam and Willemsen40 ).
Associations between maternal milk consumption and adverse birth outcomes (preterm birth, low birth weight (LBW) and small for gestational age (SGA))
We found two prospective cohort studies with discrepant findings on the association between maternal fish consumption and preterm birth (Table 5). A study conducted in Denmark showed that low seafood consumption was a strong risk factor for preterm birth( Reference Olsen and Secher53 ). However, in the Netherlands, there were inconsistent findings observed in associations between consumption of total fish or different types of fish and preterm birth( Reference Heppe, Steegers and Timmermans41 ).
Associations between maternal fish intake and adverse birth outcomes (preterm birth, low birth weight (LBW) and small for gestational age (SGA))
Maternal diets during pregnancy and low birth weight
We found four eligible studies reporting on the association between LBW and maternal dietary patterns during pregnancy. The findings were mixed (Table 6). Hajianfar et al. showed a positive association between high intake of the Western dietary pattern and risk of having a LBW infant( Reference Hajianfar, Esmaillzadeh and Feizi22 ). Emond et al. found no significant association between maternal diet quality and LBW. Maternal exposures to high-quality diets (characterised by high intake of fruits, vegetables, legumes, whole grains, nuts, moderate alcohol consumption, long-chain n-3 fatty acids and polyunsaturated fats, and low intake of sugar-sweetened beverages and fruit juice, red and processed meats, trans-fatty acids and Na) had no association with LBW( Reference Emond, Karagas and Baker35 ). In addition, two RCT conducted in Cambodia and Burkina Faso examined the effect of providing maize–soya blended and fortified food supplements during pregnancy and reported no significant effects on LBW( Reference Huybregts, Roberfroid and Lanou60 , Reference Janmohamed, Karakochuk and Boungnasiri61 ). These RCT studies might not be comparable with the above observational studies since the effects of dietary interventions may influence diet only, and last for a limited time, so this result should be interpreted with caution.
Associations between maternal dietary patterns and low birth weight
AHEI, Alternative Healthy Eating Index; FA, fatty acids; FFS, fortified food supplements; MMN, multi micronutrients; CSB, maize–soya blend diets.
Three prospective cohort studies reported on the association between maternal fish consumption and LBW (Table 5). Olsen et al. reported that low fish intake was a strong risk factor for LBW( Reference Olsen and Secher53 ). Muthaya et al. documented that pregnant women who did not eat fish during the third trimester had a significantly higher risk of having LBW infants( Reference Muthayya, Dwarkanath and Thomas50 ). Brantsaeter et al. also reported that women with seafood consumption >60 g/d had a lower risk of having LBW infants( Reference Brantsaeter, Birgisdottir and Meltzer33 ).
Maternal diets during pregnancy and small for gestational age
Ten studies used a dietary pattern approach to evaluate maternal dietary intakes. Of these, ‘alternative healthy eating index’, ‘Mediterranean diets’, ‘health-conscious’, ‘traditional’ and ‘varied’ patterns were significantly associated with reducing the risk of giving birth to SGA infants (Table 7). Emond et al. found that healthy diets characterised by high intake of fruits, vegetables, legumes, whole grains, nuts, and moderate alcohol consumption, intake of long-chain n-3 fatty acids from foods and supplements, and of polyunsaturated fats, and low intake of sugar-sweetened beverages and fruit juice, red and processed meats, trans-fatty acids and Na, decreased rates of SGA( Reference Emond, Karagas and Baker35 ). In a prospective cohort study conducted in the Infancia y Medio Ambiente cohort in Spain, Chatzi et al. showed that high adherence to ‘Mediterranean diets’ characterised by vegetables, fruits, legumes and nuts, cereals, seafood and milk products had a lower risk of delivering SGA( Reference Chatzi, Mendez and Garcia19 ). In addition, a Danish national birth cohort study has also found that the ‘health-conscious’ diets characterised by a high intake of vegetables, fruits, fish and poultry led to a significantly lower risk of having SGA infants( Reference Knudsen, Orozova-Bekkevold and Mikkelsen23 ). In a case–control study from New Zealand, Thompson et al. found that the traditional dietary pattern characterised by apples/pears, citrus fruit, kiwifruit/feijoas, bananas, green vegetables, root vegetables, peas/maize, dairy food/yogurt and water was significantly associated with a lower risk of having a SGA infant( Reference Thompson, Wall and Becroft58 ). In a large prospective cohort study conducted in China, Lu et al. reported that maternal consumption of a ‘varied or mixed’ dietary pattern, with high intakes of ‘noodles, bread, root vegetables, mushrooms, melon vegetables, sea vegetables, bean vegetables, poultry, seafood, animal organ meat, bean products, yogurt, sweet beverages, puffed food, confectionery and snacks’, was associated with a lower risk of having an SGA infant.
Associations between maternal dietary patterns and small for gestational age
VFR, vegetables–fruits–rice; SFN, seafood and noodle; PCP, pasta–cheese–processed meat; RR, risk ratio; AHEI, Alternative Healthy Eating Index; AOR, adjusted OR; GI, glycaemic index; GL, glycaemic load.
However, a ‘Western diet’ (characterised by high-fat dairy products, red and processed meat) was positively associated with the risk of SGA infants( Reference Lu, Chen and He44 ). In addition, in Japan, Okubo et al. found that consumption of a ‘wheat products’ pattern (characterised by a high intake of bread, fruit and vegetable juice, confectionery and soft drinks) was associated with increased risk of having a SGA infant( Reference Okubo, Miyake and Sasaki51 ).
In contrast, Knudsen et al. reported that adherence to the glycaemic index diet was not significantly associated with the risk of delivering SGA infants( Reference Knudsen, Heitmann and Halldorsson43 ). Studies conducted in Singapore and the USA by Chia et al. and Poon et al. also documented no dietary patterns (VFR, seafood and noodle, pasta, cheese, processed meat, healthy eating index, Mediterranean diets) as significantly associated with the risk of having a SGA infant( Reference Chia, de Seymour and Colega34 , Reference Poon, Yeung and Boghossian55 ).
Six studies reported results of fish consumption in relation to SGA deliveries, but the findings were inconsistent (Table 5). Halldorsson et al. found that high intake of fatty fish (salmon, mackerel, trout, herring and Greenland halibut) increased the risk of having a SGA infant( Reference Halldorsson Th, Meltzer and Thorsdottir38 ). In another prospective cohort study conducted in Spain, Mendez et al. reported that maternal intake of crustaceans and canned tuna (more than once/week) was associated with an increased risk of SGA( Reference Mendez, Plana and Guxens46 ). Contrary to this, Olson et al. and Heppe et al. found no association between maternal fish consumption and SGA infants( Reference Heppe, Steegers and Timmermans41 , Reference Olsen and Secher53 ). In a case–control study conducted in New Zealand, people with a low intake of fish had an increased risk of having SGA infants( Reference Mitchell, Robinson and Clark47 ). In another case–control study, Ricci et al. also reported that fish consumption was inversely associated with the risk of SGA( Reference Ricci, Chiaffarino and Cipriani56 ).
Two prospective cohort studies reported the beneficial effects of milk intake on having SGA infants, while one prospective cohort study reported no association (Table 4). In Denmark, maternal milk consumption during pregnancy was inversely associated with the risk of having SGA infants( Reference Olsen, Halldorsson and Willett54 ). In addition, Olmedo et al. found that high maternal dairy consumption was significantly associated with reduced risk of SGA( Reference Olmedo-Requena, Amezcua-Prieto and Luna-Del-Castillo52 ). However, Heppe et al. reported that there was no significant association between milk consumption and the risk of delivering a SGA baby( Reference Heppe, van Dam and Willemsen40 ).
Maternal diets before pregnancy and adverse birth outcomes
We found only four eligible studies that examined the relationship between preconception dietary patterns and adverse birth outcomes (preterm birth, LBW and SGA)( Reference Bouwland-Both, Steegers-Theunissen and Vujkovic32 , Reference Grieger, Grzeskowiak and Clifton37 , Reference Potdar, Sahariah and Gandhi63 ). Grieger et al. observed the association between pre-pregnancy dietary patterns in the 12 months before conception and preterm birth among Australian women. Those with a higher intake of protein/fruit had a lower risk of preterm birth, whereas the risk of preterm birth was higher in women with a high intake of fat/sugar/takeaways( Reference Grieger, Grzeskowiak and Clifton37 ). Recently, Hillesund et al. also showed an inverse association between a high healthy dietary score (characterised by high intake of fruits, vegetables, plenty of water, and low intake of added sugar, salt, sweet snacks and fast foods) and risk of having a preterm infant( Reference Hillesund, Bere and Sagedal59 ).
In a prospective cohort study conducted on 847 Dutch mothers, Bowland et al. reported no significant association between pre-pregnancy dietary patterns in the 3 months before 10–13 weeks of gestation and risk of having SGA infants( Reference Bouwland-Both, Steegers-Theunissen and Vujkovic32 ). One RCT study conducted in India examined the effect of providing treatment snacks (fresh and dried green leafy vegetables, milk and dried fruit) and control snacks (low-micronutrient vegetables such as potato, tapioca and onion) from preconception to giving birth. They reported that there was no significant effect of the treatment snack on adverse birth outcomes( Reference Potdar, Sahariah and Gandhi63 ).
Discussion
The present study has reviewed forty articles comprising mostly of prospective cohort studies with few RCT, which focused on the association between maternal diets and adverse birth outcomes. The review suggests that better maternal diet quality during pregnancy, including ‘prudent’, ‘traditional’, ‘DASH’ and ‘VFR’, may reduce the risk of preterm birth( Reference Englund-Ogge, Brantsaeter and Sengpiel17 , Reference Martin, Sotres-Alvarez and Siega-Riz20 , Reference Chia, de Seymour and Colega34 ). It has been challenging to summarise robust evidence from the findings because of the variation in quantity and quality of maternal dietary intake and of the assessment techniques used. Dietary patterns were also named differently across the studies. However, the dietary patterns did share somewhat similar characteristics among the studies. For example, the ‘prudent’, ‘DASH’ and ‘VFR’ patterns all included similar food items, such as a high intake of fruits and vegetables. Only traditional diets did not include fruits and vegetables. Whole grains were also listed in the ‘prudent’ and ‘DASH’ diets. Fish and dairy products were included in Mediterranean diet, DASH and traditional diets. Therefore, a high adherence to dietary patterns characterised by fruits, vegetables, whole grains, fish and dairy products might have a protective synergetic effect on preterm birth. These diets are rich in anti-inflammatory nutrients or antioxidants, which might reduce both local and systemic inflammation. The dietary patterns characterised by collard greens, coleslaw or cabbage, red and processed meats and ‘Western’ diets commonly characterised by higher intakes of processed meat were associated with a higher risk of having preterm infants( Reference Martin, Sotres-Alvarez and Siega-Riz20 , Reference Rasmussen, Maslova and Halldorsson21 ). These diets contain pro-inflammatory nutrients, which act as a stressor on the hypothalamic-pituitary-adrenal system, and might also curb the transfer of nutrients through the placenta( Reference Shin, MohanKumar and Sirivelu24 , Reference Jeffery and O’Toole25 ).
The findings regarding adherence to a ‘Mediterranean diet’ during pregnancy and preterm birth have been inconsistent. A prospective cohort study by Mikkelsen et al. showed that higher adherence was associated with decreased risk of preterm birth( Reference Mikkelsen, Osterdal and Knudsen18 ). However, studies conducted in Norway( Reference Saunders, Guldner and Costet57 ) and France( Reference Haugen, Meltzer and Brantsaeter39 ) reported no association between Mediterranean diet and preterm birth. Khoury et al. also showed the beneficial effects of cholesterol-lowering diets in reducing the risk of preterm birth( Reference Khoury, Henriksen and Christophersen62 ). The discrepancy of these findings might be due to the variation of dietary assessments and sample size across the studies. There are challenges to establishing a specific definition of ‘Mediterranean diet’ due to socio-cultural and geographic differences, and therefore, the studies used different measurement techniques to assess ‘Mediterranean diet’ intake. The sample sizes also varied from 728( Reference Saunders, Guldner and Costet57 ) to 40 817( Reference Haugen, Meltzer and Brantsaeter39 ) in the prospective studies, while the RCT study included only 290 study participants( Reference Khoury, Henriksen and Christophersen62 ). The studies also used a variety of names for the ‘Mediterranean diet’, such as ‘Mediterranean-type diet’, ‘Mediterranean diet adherence’, ‘Mediterranean diet index’ and ‘cholesterol-lowering diet’. However, the food items were commonly characterised by a high intake of fish, dairy products, oils, whole grains, fruits, vegetables and legumes.
Maternal dietary patterns, including higher adherence to ‘alternative healthy eating index’, ‘Mediterranean diet’, ‘health-conscious’, ‘traditional’ and ‘varied’ diets, were inversely associated with the risk of having SGA infants( Reference Chatzi, Mendez and Garcia19 , Reference Knudsen, Orozova-Bekkevold and Mikkelsen23 , Reference Emond, Karagas and Baker35 , Reference Lu, Chen and He44 , Reference Thompson, Wall and Becroft58 ). These findings were contradicted by a study conducted in the USA by Poon et al. which did not observe a significant association between dietary patterns (‘healthy eating index’ and ‘Mediterranean diet’) and SGA( Reference Poon, Yeung and Boghossian55 ). ‘Western diets’ and consumption of wheat products were significantly associated with an increased risk of SGA infants( Reference Knudsen, Orozova-Bekkevold and Mikkelsen23 , Reference Okubo, Miyake and Sasaki51 ). One possible explanation for this disagreement is that maternal dietary intake was only measured once during the third trimester in the infant feeding practices study II cohort( Reference Poon, Yeung and Boghossian55 ), while dietary intake in both the Infancia y Medio Ambiente( Reference Chatzi, Mendez and Garcia19 ) and NHBCS( Reference Emond, Karagas and Baker35 ) cohorts was measured in the first trimester and mid-pregnancy period, respectively. The dietary patterns inversely associated with the risk of SGA were termed differently, but were characterised by somewhat similar food items, such as high intake of vegetables, fruits, legumes, seafood/fish and milk, across the studies. Such diets are rich in antioxidants, fibre and unsaturated fats, which reduce both systemic and local inflammation( Reference Akbaraly, Shipley and Ferrie68–Reference Sen, Rifas-Shiman and Shivappa70 ).
Inconsistent findings have been observed in association between maternal fish consumption and preterm birth with reports of either positive( Reference Olsen and Secher53 ) or null association( Reference Heppe, Steegers and Timmermans41 ). The discrepancy of these findings might be due to the variety of measurement techniques used to assess fish intake and sample size. The nineteen large population-based European birth cohort studies showed that moderate fish intake during pregnancy had a beneficial effect on lowering the risk of preterm birth( Reference Leventakou, Roumeliotaki and Martinez71 ).
The association between maternal fish consumption and SGA infants has also been discrepant, with reports of either positive( Reference Halldorsson Th, Meltzer and Thorsdottir38 , Reference Mendez, Plana and Guxens46 ) or negative( Reference Mitchell, Robinson and Clark47 , Reference Ricci, Chiaffarino and Cipriani56 ) or null associations( Reference Heppe, Steegers and Timmermans41 , Reference Olsen and Secher53 ). This might be due to fish intake being assessed differently across the studies. Halldorsson et al. excluded women who were receiving fish oil supplements( Reference Halldorsson Th, Meltzer and Thorsdottir38 ), while Olsen et al. quantified long-chain n-3 fatty acids( Reference Olsen and Secher53 ). There was large variability in the sample sizes, from 657( Reference Mendez, Plana and Guxens46 ) to 44 824( Reference Halldorsson Th, Meltzer and Thorsdottir38 ).
Based on the findings from all prospective cohort studies, fish consumption has been inversely associated with the risk of having LBW infants( Reference Brantsaeter, Birgisdottir and Meltzer33 , Reference Olsen and Secher53 , Reference Muthayya, Dwarkanath and Thomas50 ), except in one study, which reported a null association( Reference Heppe, Steegers and Timmermans41 ), and which could be due to difference in quantity and quality of maternal fish intake( Reference Heppe, Steegers and Timmermans41 ). The findings might also be influenced by the intake of nutritional supplements, such as n-3 PUFA( Reference Muthayya, Dwarkanath and Thomas50 ) or n-3 fatty acid( Reference Brantsaeter, Birgisdottir and Meltzer33 ). The inverse association between fish intake and LBW might be due to fish being a major source of essential nutrients, such as Se, iodine, n-3 PUFA and protein, which are considered to have beneficial effects on reducing adverse birth outcomes( Reference Olsen, Sørensen and Secher72–Reference Philibert, Vanier and Abdelouahab74 ).
The discrepant findings between milk intake and preterm birth might be due to the use of different outcome assessments and sample sizes( Reference Heppe, van Dam and Willemsen40 , Reference Myhre, Brantsaeter and Myking48 ). For example, Myhre et al. assessed only spontaneous preterm birth (between 22 and 37 weeks of gestation)( Reference Myhre, Brantsaeter and Myking48 ), while Heppe et al. included both spontaneous and induced preterm birth (live birth < 37 weeks of gestation)( Reference Heppe, van Dam and Willemsen40 ). Maternal milk consumption that contains probiotics might decrease spontaneous preterm birth, particularly if caused by the premature rupture of membrane, possibly through an effect of probiotics on vaginal tract infections and lowering overall inflammatory response( Reference Yeganegi, Watson and Martins75 ). In addition, many studies showed that bacterial vaginosis was positively associated with preterm birth, which might be one explanation for the association between milk intake and reduced risk of spontaneous preterm birth( Reference Hillier, Nugent and Eschenbach76 , Reference Donders, Van Calsteren and Bellen77 ).
Beneficial effects of milk intake on reducing having an SGA infant are evident in two prospective studies( Reference Olmedo-Requena, Amezcua-Prieto and Luna-Del-Castillo52 , Reference Olsen, Halldorsson and Willett54 ), which is also confirmed by this review. However, one prospective cohort study did not show a significant association( Reference Heppe, van Dam and Willemsen40 ). This discrepancy might be due to maternal milk intake being measured differently across the studies. The gestational period in which maternal milk intake was assessed differed among studies, from being assessed at first trimester( Reference Heppe, van Dam and Willemsen40 ), mid-pregnancy( Reference Olsen, Halldorsson and Willett54 ) and between 20 and 22 weeks of gestation( Reference Olmedo-Requena, Amezcua-Prieto and Luna-Del-Castillo52 ). The findings were scarce, but suggestive of a negative association between milk consumption and SGA. However, it should be noted that excessive maternal milk intake (>2 glasses/d) was positively associated with increased risk of having large for gestational age infants( Reference Olsen, Halldorsson and Willett54 ). In addition, some studies showed positive associations between maternal milk intake and birth weight used as a continuous variable( Reference Heppe, van Dam and Willemsen40 , Reference Olsen, Halldorsson and Willett54 , Reference Mannion, Gray-Donald and Koski78 ).
To date, studies on preconception dietary patterns and adverse birth outcomes are scarce and the findings are mixed. Only two studies (one cross-sectional and one RCT) showed an association between preconception dietary patterns and preterm birth( Reference Grieger, Grzeskowiak and Clifton37 , Reference Hillesund, Bere and Sagedal59 ), while one prospective cohort( Reference Bouwland-Both, Steegers-Theunissen and Vujkovic32 ) and one RCT( Reference Potdar, Sahariah and Gandhi63 ) study reported no association. These studies were conducted using different dietary patterns, study designs and sample sizes. In addition, the period of preconception diet was assessed differently among these studies. For example, Grieger et al. observed the association between pre-pregnancy dietary patterns in the 12 months before conception and adverse birth outcomes( Reference Grieger, Grzeskowiak and Clifton37 ), while Bouland et al. evaluated diet in the 3 months before 10–13 weeks of gestation. Timing and duration of exposure (preconception diet) and its effects on maternal and fetal outcomes can be considered as a critical period, a sensitive period and cumulative effects, respectively. The period around conception (8–12 weeks before conception) is a critical period for gamete maturation and early fetal and placental development( Reference Stephenson, Heslehurst and Hall79 ). For example, folic acid supplementations during peri-conception could decrease the risk of neural tube defects( Reference De‐Regil, Peña‐Rosas and Fernández‐Gaxiola80 , Reference Mastroiacovo and Leoncini81 ) as well as LBW and SGA( Reference Hodgetts, Morris and Francis82 , Reference He, Pan and Hu83 ). Adolescence might be a sensitive period for poor lifestyle behaviours, such as harmful dietary intake, smoking and alcohol intake. Such pre-pregnancy risk factors could also have a cumulative effect on lifelong adulthood health as well as for future generations. For example, prevention of obesity in women of reproductive age has a crucial impact on pregnancy and birth outcomes( Reference Hanson, Barker and Dodd84 ).
The present review is limited by the wide diversity of maternal dietary intake measures, including the quality and quantity of foods. The review also included mostly prospective cohort studies, and so it is not possible to draw definite causal inferences. The findings are based on studies in mostly high-income countries, which limit the ability to generalise the evidence to a wide range of populations around the globe. Furthermore, for the studies of the relationship between maternal diets and SGA, results may have been influenced by variations in the prevalence of SGA infants and the use of different cut-offs for SGA across the studies. The operational definitions of SGA were different from birth weight <2·5th percentile( Reference Knudsen, Orozova-Bekkevold and Mikkelsen23 ) to <5th percentile( Reference Heppe, van Dam and Willemsen40 , Reference Heppe, Steegers and Timmermans41 ) to <10th percentile of gestational age( Reference Chia, de Seymour and Colega34 , Reference Emond, Karagas and Baker35 , Reference Halldorsson Th, Meltzer and Thorsdottir38 ). Because of these limitations, we have used a narrative approach for this systematic review rather than a meta-analysis.
In conclusion, we found several prospective cohort studies that examined the relationship between maternal dietary patterns during pregnancy, preterm birth and SGA. Despite some disagreements and variation of dietary patterns across the studies, some common characteristics were found in the relationships between maternal dietary patterns and the risk of preterm birth and SGA. Better maternal diet quality during pregnancy, characterised by a high intake of vegetables, fruits, whole grains, fish and dairy products, may have a beneficial effect on lowering the risk of preterm birth. Furthermore, maternal dietary patterns during pregnancy characterised by a high intake of vegetables, fruits, legumes, fish and milk products may reduce the risk of having SGA infants. No consistent associations were observed between maternal dietary patterns during pregnancy and LBW. Evidence on associations between pre-pregnancy diets and adverse birth outcomes is limited and inconsistent. Therefore, well-powered population-based prospective studies are needed to establish the association between pre-pregnancy diets and adverse birth outcomes. More specific definitions and homogenous measurements for dietary patterns are also needed for potential meta-analysis, to estimate a combined effect size.
Acknowledgements
The authors would like to acknowledge Megan Ferguson for her assistance with the editing of this publication, and Scott Macintyre (Herston librarian) for his assistance with the search strategy.
D. G. G. is supported by the UQ Research Training Scholarship. G. D. M. holds an Australian Health and Medical Research Council Principal Research Fellowship (APP1121844).
D. G. G. and G. D. M. designed the study; D. G. G. and M. W. conducted data extraction and quality assessment; D. G. G. analysed the data and wrote the paper; G. D. M. and M. W. revised the paper. All authors read and approved the final manuscript.
The authors declare no conflicts of interest.
Supplementary material
For supplementary material referred to in this article, please visit https://doi.org/10.1017/S0007114519002897
