Maternal diet during early childhood, but not pregnancy, predicts diet quality and fruit and vegetable acceptance in offspring

Abstract

Studies have identified prenatal flavour exposure as a determinant of taste preferences in infants; however, these studies have focused on relatively small samples and limited flavours. As many parents struggle with getting children to accept a variety of nutritious foods, a study of the factors influencing food acceptance is warranted. The objective of this study was to determine whether exposure to a wider variety of fruit and vegetables and overall higher diet quality in utero results in acceptance of a greater variety of these foods and better diet quality for offspring during childhood. This study is a secondary data analysis of pregnant women (n = 52) and their resulting offspring recruited for the Women and Their Children's Health study in NSW, Australia. Dietary intake of mothers and children was measured using food frequency questionnaires. Diet quality and vegetable and fruit variety were calculated using the Australian Recommended Food Score and the Australian Child and Adolescent Recommended Food Score. Associations between maternal and child diet quality and variety were assessed using Pearson's correlations and the total effect of in utero maternal pregnancy diet on childhood diet was decomposed into direct and indirect effect using mediation analysis. Maternal pregnancy and post‐natal diet were both correlated with child diet for overall diet quality and fruit and vegetable variety (P < 0.001). Mediation analyses showed that the indirect effect of maternal pregnancy diet on child diet was mediated through maternal post‐natal diet, particularly for fruit (P = 0.045) and vegetables (P = 0.055). Nutrition intervention should therefore be aimed at improving diet quality and variety in mothers with young children, in order to subsequently improve eating habits of offspring.

Introduction

Vegetables and fruit are cornerstones of a healthful diet, and their consumption has long been linked with chronic disease prevention and positive health outcomes (National Health and Medical Research Council, Department of Health and Ageing 2013). However, the 2007 Children's Nutrition and Physical Activity Survey showed that three quarters of Australian children aged 2 to 8 years did not meet vegetable recommendations for a healthy, nutrient‐dense varied diet, with inadequate fruit and vegetable intake, and high saturated fat and sugar intake highlighted (Department of Health and Ageing, Department of Agriculture, Fisheries and Forestry 2008). Poor eating habits in childhood have been shown to track into adulthood and can contribute to the development of chronic disease later in life (Berenson et al. 1992; Mikkilä et al. 2005; Northstone & Emmett 2008). It is therefore imperative to establish healthy eating habits in early childhood.

In westernized countries, most of the disease burden from poor nutrition is due to excess intake of energy‐dense, nutrient‐poor foods, which are foods high in energy, saturated fat and added or refined sugars and/or salt (National Health and Medical Research Council, Department of Health and Ageing 2013). These energy‐dense, nutrient‐poor foods are readily available and affordable (Beauchamp & Mennella 2009). Many parents report difficulties with feeding children a wide variety of nutrient‐dense foods from the core food groups (vegetables and legumes, fruit, whole grains, lean meat and vegetarian alternatives and dairy foods), particularly for vegetables (National Health and Medical Research Council, Department of Health and Ageing 2013). Encouraging vegetable intake from an early age remains a challenging area for both parents and dietetic professionals (Carruth et al. 2004; Maier et al. 2007). If children enjoy the flavour of a food, they are more likely to consume it (Benton 2004; Beauchamp & Mennella 2009). Young children are also more likely to accept a food if its flavour is familiar to them (Beauchamp & Mennella 2009). An exploration into food acceptance in young children can therefore lead to valuable insight upon which to base nutrition intervention.

The majority of observational studies of fruit and vegetable intake in children have focused on post‐natal influences on food choices (Benton 2004; Savage et al. 2007; Fisk et al. 2011). Experimental studies aiming to increase fruit and vegetable intake in young children have also focused primarily on post‐natal exposure (Birch et al. 1998; Forestell & Mennella 2007; Maier et al. 2007; Mennella et al. 2008). These studies have shown that exposure to a food through breast milk or formula or exposure to a solid food during infancy promotes long‐lasting effects, namely familiarity of and preference for this food (Schaal et al. 2000; Forestell & Mennella 2007; Mennella et al. 2008).

While extensive studies examining post‐natal influences on child taste preferences are available, limited data are available on the prenatal influence on taste development. It has been suggested that infants are born with innate preferences for sweet, salty and umami foods over bitter or sour foods (Beauchamp & Mennella 2009). However, humans can override these innate preferences and develop preferences for bitter or sour foods, including certain vegetables and fruit (Maier et al. 2007; Mennella et al. 2008; Beauchamp & Mennella 2009). Classic studies by Mennella and colleagues have increased our understanding of the influence of maternal diet and in utero flavour exposure on the future taste preferences of offspring (Mennella et al. 2001). By 13–15 weeks gestation, the fetus can perceive tastes and smells while still in the womb via amniotic fluid (Mennella et al. 1995, 2001), suggesting the earliest taste and smell experiences begin during gestation (Mennella et al. 2001). Indeed, Mennella has shown that prenatal exposure to certain flavours is associated with a greater acceptance of these foods in infancy (Mennella et al. 2001). In a randomised controlled trial, mothers in the experimental group drank 300 mL carrot juice 4 days per week for three consecutive weeks during the last trimester of pregnancy. At 5–6 months old, the infants who had been exposed to the carrot flavour prenatally exhibited less negative facial responses while being fed carrot‐flavoured cereal, relative to plain cereal (P = 0.01). The control group exhibited the opposite tendency, although this was not significant. (Mennella et al. 2001). Multiple animal studies on rat, pig, sheep and rabbit young support the finding that prenatal flavour learning influences feeding preferences (Bilkó et al. 1994; Bayol et al. 2007; Simitzis et al. 2008; Oostindjer et al. 2009). Although studies in humans have been of relatively small sample sizes and used only one or two foods, these early studies and animal models support the hypothesis for in utero flavour learning (Schaal et al. 2000; Mennella et al. 2001).

Whole foods contain not only macro‐ and micronutrients but a range of other non‐nutrient components, including phytochemicals that offer protective effects against disease (National Health and Medical Research Council, Department of Health and Ageing 2013). Therefore, exposure to a broader variety of flavours is associated with greater variety of food and therefore a broader range of macronutrients, micronutrients and non‐nutrient compounds consumed (National Health and Medical Research Council, Department of Health and Ageing 2013). Diet quality refers to both diet variety and nutritional adequacy, and high‐quality diet is associated with improved health outcomes and reduced risk of chronic disease (National Health and Medical Research Council, Department of Health and Ageing 2013). Therefore, willingness to accept a wide range of flavours will likely increase both diet variety and diet quality (Mennella et al. 2008). While food preference refers to foods that are found to be enjoyable and pleasant, food acceptance is also used in this study to refer to those foods which are willingly consumed.

There is a need to explore the relationship between maternal diet quality during pregnancy and the diet quality of their offspring. The primary aim of this study was to test whether maternal diet during pregnancy was associated with childhood diet quality at age 2–3 years. The secondary aim was to test whether maternal fruit and vegetable intake during pregnancy was associated with child fruit and vegetable consumption at 2–3 years. Maternal post‐natal diet was also considered as a potential predictor of the child's diet.

Key messages.

• Maternal post‐natal diet, not pregnancy diet, is associated with child diet.

• Maternal post‐natal diet, not maternal pregnancy diet, predicted child diet quality, and vegetable and fruit variety.

• Nutrition interventions should aim to improve diet quality and variety in mothers with young children.

Materials and methods

Study design

This study is a secondary data analysis of the Women and Their Children's Health (WATCH) prospective cohort study, which followed pregnant women and their offspring up to 4 years of age (Hure et al. 2012). All pregnant women less than 18 weeks gestation were eligible to participate in the WATCH study on the provision that they lived in the local or neighbouring areas and were able to commute to John Hunter Hospital (JHH) in Newcastle, New South Wales, Australia, for data collection. Participants were recruited through midwives at the JHH antenatal clinic, local media coverage and word of mouth. Pregnant women attended JHH for data collection at 19, 24, 30 and 36 weeks gestation. Additionally, data collection for both mothers and their offspring occurred at post‐natal quarterly intervals for the first 12 months, and then annually until 2–3 years of age (Hure et al. 2012). Between June 2006 and December 2007, 180 women were deemed eligible to participate. Of this sample 74% remained 2 years after study commencement (Hure et al. 2012). Detailed methods of the WATCH study are published elsewhere (Hure et al. 2012). This current study uses the dietary intake data from mothers and their offspring up to 3 years of age.

Ethics approval

The WATCH study was approved by the Hunter New England Human Research Ethics Committee in 2006 and was also registered with the University of Newcastle Human Research Ethics Committee (Hure et al. 2012). All participants in the WATCH study gave informed, written consent to participate. Participants did not receive any financial incentives to participate. Data were de‐identified before commencement of this study (Hure et al. 2012).

Data collection

Data were obtained on a number of health, anthropometric, socio‐economic and lifestyle variables (Hure et al. 2012). Anthropometric data were collected by an Accredited Practicing Dietitian and Level 1 Anthropometrist accredited by the International Society for the Advancement of Kinanthropometry. Data collection included height/length, weight, circumferences and skinfold thicknesses of both mothers and their children. Standardized methods of data collection were used, and are described in detail elsewhere (Hure et al. 2012). The following variables were selected for use in the current analysis.

Data on education, income and marital status were self‐reported and the questions were modelled on Women's Health Australia surveys (Brown & Dobson 2000). Further information relating to medical and socio‐economic data were obtained from the ‘Obstetrix’ database, the major record of antenatal information, birth outcomes and patient and family history in New South Wales, Australia (LeMay 2005).

Dietary intake

Maternal dietary intake was assessed using the validated Dietary Questionnaire for Epidemiology Studies (DQES) food frequency questionnaire (FFQ; Cancer Council Victoria 2005) developed by the Cancer Council of Victoria (Hodge et al. 2002). The DQES asks about intake of 74 foods; participants were asked to report intake frequency over the last 3–12 months, on a 10‐point scale of ‘never’ to ‘three or more times per day’ (Hure et al. 2009). Questions regarding total intake of fruit and vegetables were used to adjust for potential over‐reporting of intakes of individual fruit and vegetables (Hure et al. 2009). Nutrient intakes were computed from NUTTAB 1995 (Cancer Council Victoria 2005).

Child dietary intake was assessed from 2 years of age using the toddler version of the Australian Child and Adolescent Eating Survey (ACAES; Watson et al. 2009; Collins et al. 2013). Mothers completed the ACAES for their child, reporting the consumption frequency (ranging from ‘never’ to ‘four times per day’) for a range of foods over the preceding 6 months. The ACAES is a validated FFQ, which demonstrates acceptable reliability for ranking nutrient intakes in Australian toddlers 2–4 years old (Collins et al. 2013). Nineteen questions relate to the intake of vegetables and 11 items to fruit, with a separate section to adjust for seasonal fruit intake (Watson et al. 2009).

Australian Recommended Food Score

The Australian Recommended Food Score (ARFS) and the Australian Child and Adolescent Recommended Food Score (ACARFS) were used to calculate a diet variety score for mothers and their offspring (Collins et al. 2008; Marshall et al. 2012). The ARFS is a numerical value of diet quality and variety (Collins et al. 2008). It is calculated based on the level of alignment between reported consumption frequency of foods featured in the DQES/ACAES and Dietary Guidelines for Australian Adults and Australian Guide to Healthy Eating recommendations (Kant & Thompson 1992; The Children's Health Development Foundation, Deakin University 1998; National Health and Medical Research Council 2003). A higher score reflects greater adherence to national guidelines and therefore greater diet quality and variety. Alcohol was not included for this cohort. Respective maximum scores for the ARFS and the ACARFS were 72 and 73 (Hure et al. 2009). The scoring method for the ARFS and ACARFS are described elsewhere (Collins et al. 2008, 2013). Diet variety scores (sub‐scales) were calculated for consumption frequency of foods within the core food groups, including fruit and vegetables (Collins et al. 2008, 2013).

Statistical analysis

Participants were included if dietary data were available for at least one time point. Exploratory analysis found early and late pregnancy diets were strongly correlated. Therefore, the average of both time points was taken. The average ACARFS score was calculated for the child diet between ages 2 and 3 years. Pearson's correlation coefficients were used to assess the relationship between offspring ACARFS scores (total and subscales) and both maternal pregnancy and maternal post‐natal ARFS scores (total and subscales). The total effect of in utero maternal diet on childhood diet was decomposed into a direct effect and an indirect effect through maternal post‐natal diet using mediation analysis with standard errors of parameter estimates estimated using the bootstrap with 1000 bootstrap replications (MacKinnon & Dwyer 1993). Mediation analyses were performed unadjusted for confounders, and also adjusted for education, parity (first child or not), maternal age (years) and breastfeeding duration (weeks). All analyses were programmed using Stata v13 (StataCorp, College Station, TX, USA). We note that multi‐collinearity is an inherent problem in mediation analysis, particularly when strong correlation exists between the independent and mediating variables resulting in inflated standard error of mediating effects and therefore diminishing the power to detect them.

Results

Of the 180 women recruited, dietary intake data during pregnancy and offspring dietary data at age 2–3 years was available for 52 mother–child dyads. Only seven mother–child dyads completed FFQs at all time points: maternal intake in early pregnancy (19 weeks gestation), late pregnancy (36 weeks gestation) and at 2 and 3 years post‐partum; child intake at ages 2 and 3 years. Characteristics of the 52 mothers included are summarized in Table 1. Mothers were aged between 19 and 41 years (mean age 30.2 ± 5.4), and for the majority of mothers this pregnancy was their first (n = 26) or second child (n = 15). Around half of the women had a pre‐pregnancy body mass index (BMI) in the normal weight range of 18.5–24.9 kg m⁻² (52.9%), 25.5% were overweight (BMI 25–29.9 kg m⁻²) and 19.6% were obese (BMI > 30 kg m⁻²). One participant was underweight (BMI < 18.5 kg m⁻²), two women had a BMI of 35–39.9 kg m⁻² (obese class 2) and one had a BMI of greater than 40 kg m⁻² (obese class 3). Four participants had smoked at some point during their pregnancy. Of the 52 mothers in this analysis, 75% had a year 12 high school certificate or higher educational attainment, compared with 71.3% (n = 160) from the WATCH population cohort (Hure et al. 2012).

Table 1.

Demographic characteristics of women (n = 52) participating in the WATCH cohort study

Characteristic
	Min‐max values (n = 52)	Sample (n = 52)	WATCH cohort (n = 179)
Age (years), mean ± SD	19.4–41.2	30.2 ± 5.4	28.7 ± 5.7
Pre‐pregnancy weight (kg), median [IQR]	48–140	67.5 [18.5]	65.0 [21.0]
Height (cm), median [IQR]	156.3–182.6	165.7 [6.0]	164.2 [9.1]
Pre‐pregnancy BMI (kg m−2), median [IQR]	17.4–48.1	24.4 [6.1]	24.4 [7.8]
Length of gestation (weeks), median [IQR]	32–42	39.4 [2.0]	39.4 [2.0]
Nulliparous, n (%)		26 (50%)	77 (43%)
Educational attainment	N	%
Less than year 12 or equivalent	13	25%
Year 12, certificate, trade or apprenticeship	23	44%
University degree	16	31%

BMI, body mass index; IQR, interquartile range; SD, standard deviation.

Table 2 reports mean ARFS and ACARFS scores of mothers and children at each time point. Total ARFS and fruit and vegetable sub‐scores were very similar at both pregnancy time points. Mean pregnancy ARFS (average early and late) was 31 out of a maximum 72. Twelve participants obtained scores ≥40, and six obtained scores that were less than 20. Mean fruit score was 6.2 (maximum 14), and mean vegetable score was 12.2 (maximum 22). For children, mean ACARFS (total and sub‐scales) were very similar at ages 2 and 3. The mean total ACARFS in toddlerhood was 31.2 (maximum 73). Nine participants obtained scores of ≥40, and four participants obtained scores of ≤20. Mean fruit score was 6.0 (maximum 12), and mean vegetable score was 9.6 (maximum 21). Maternal diet between early and late pregnancy was found to be significantly correlated for ARFS total and fruit and vegetable sub‐scores (P < 0.0001). Therefore, the average of early and late pregnancy was used to show the association between maternal pregnancy and post‐natal diets and the offspring's diet at age 2 to 3 years (Table 3). Maternal pregnancy diet was found to be strongly correlated with maternal diet for both total ARFS (r = 0.85, P < 0.001) and vegetable sub‐score (r = 0.82, P < 0.001) and moderately correlated for the fruit sub‐score (r = 0.58, P < 0.001). Moderate correlations were found between maternal pregnancy diet and child diet (average of 2–3 years) for ARFS (r = 0.66, P < 0.001), fruit (r = 0.46, P < 0.001) and vegetables (r = 0.52, P < 0.001). Moderate correlations were also found between maternal post‐natal diet and child diet (average of 2–3 years) for total ARFS (r = 0.65, P < 0.001), fruit (r = 0.59, P < 0.001) and vegetable (r = 0.61, P < 0.001) sub‐scores (Table 3).

Table 2.

Mean (SD) of ARFS and ACARFS (total and sub‐scores) for women and their children participating in the WATCH cohort study, over time

Time point	Diet Quality Score (maximum score)	Maternal mean (SD)	Child mean (SD)
Pregnancy (average of early and late pregnancy)	ARFS total (72)	31.0 (9.7)	NA
	ARFS fruit score (14)	6.2 (2.6)	NA
	ARFS vegetable score (22)	12.2 (5.0)	NA
Post‐natal: maternal and child (child age 2 years)	ARFS/ACARFS total (72/73)	31.7 (8.7)	30.7 (9.5)
	ARFS/ACARFS fruit score (14/12)	5.2 (2.9)	5.6 (2.6)
	ARFS/ACARFS vegetable score (22/21)	13.7 (4.3)	9.3 (4.8)
Post‐natal: maternal and child (child age 3 years)	ARFS/ACARFS total (72/73)	28.6 (11.2)	31.1 (10.3)
	ARFS/ACARFS fruit score (14/12)	5.6 (3.7)	6.5 (3.0)
	ARFS/ACARFS vegetable score (22/21)	11.7 (5.3)	9.8 (4.7)
Average of post‐natal maternal and child (child age 2 and 3 years)	ARFS/ACARFS total (72/73)	30.3 (9.5)	31.2 (9.7)
	ARFS/ACARFS fruit score (14/12)	5.04 (3.2)	6.00 (2.8)
	ARFS/ACARFS vegetable score (22/21)	12.9 (4.7)	9.6 (4.7)

ACARFS, Australian Child and Adolescent Recommended Food Score; ARFS, Australian Recommended Food Score; NA, not applicable; SD, standard deviation.

Table 3.

Pearson's correlations between maternal pregnancy diet, maternal post‐natal diet and child diet at 2–3 years (n = 52 pairs)

	Maternal pregnancy and post‐natal diets		Maternal post‐natal and child diets		Maternal pregnancy and child diets
	Correlation	P‐value	Correlation	P‐value	Correlation	P‐value
Diet quality*	0.85	<0.001	0.65	<0.001	0.66	<0.001
Fruits	0.58	<0.001	0.59	<0.001	0.46	<0.001
Vegetables	0.82	<0.001	0.61	<0.001	0.52	<0.001

*Maternal diet quality measured using the Australian Recommended Food Score (ARFS) and child diet quality measured using the Australian Child and Adolescent Recommended Food Score (ACARFS), with fruits and vegetables sub‐scales from each tool.

Mediation analysis was performed to determine the direct effect of maternal pregnancy diet on child diet at age 2–3 years and the effect indirectly mediated through maternal post‐natal diet (Fig. 1). Mediation results are presented unadjusted for confounders and adjusted for education, parity, maternal age and breastfeeding duration in Table 4. After adjusting for confounders, a statistically significant total effect of in utero diet on child diet (coefficient = 0.64, P < 0.0001) was found for total diet quality (ARFS). However, neither direct nor indirect effects were statistically significant on their own (direct effect = 0.45, P = 0.125; indirect effect = 0.18, P = 0.419). For fruit variety, a statistically significant total effect of in utero diet on child diet (coefficient = 0.46, P = 0.014) was found, and this was comprised mainly (∼70%) of an indirect effect mediated through maternal post‐natal diet (coefficient = 0.32, P = 0.045). For vegetable variety, the statistically significant total effect of in utero diet on child diet (coefficient = 0.39, P = 0.03) was almost entirely comprised of an indirect effect mediated through maternal post‐natal diet (coefficient = 0.38, P = 0.055, borderline significant).

Figure 1.

Mediation model of maternal diet with potential direct (coefficient C) and indirect effects (product of path coefficients A and B) on child diet.

Table 4.

Mediation analysis quantifying direct and indirect effects of maternal diet during pregnancy on child diet at 2–3 years (n = 52 pairs)*

	Unadjusted			Adjusted †
	Coefficient	95% CI	P‐value	Coefficient	95% CI	P‐value
Diet quality ‡
Direct	0.55	0.02, 1.07	0.040	0.45	−0.13, 1.03	0.125
Indirect	0.20	−0.24, 0.63	0.379	0.18	−0.26, 0.62	0.419
Total	0.75	0.47, 1.02	<0.001	0.64	0.28, 0.99	<0.001
Fruits
Direct	0.23	−0.10, 0.57	0.170	0.14	−0.28, 0.56	0.515
Indirect	0.31	0.06, 0.55	0.013	0.32	0.01, 0.64	0.045
Total	0.54	0.26, 0.82	<0.001	0.46	0.10, 0.83	0.014
Vegetables
Direct	0.09	−0.36, 0.53	0.704	0.01	−0.50, 0.52	0.974
Indirect	0.46	0.08, 0.85	0.018	0.38	−0.01, 0.77	0.055
Total	0.55	0.28, 0.83	<0.001	0.39	0.04, 0.74	0.030

*The direct effect is the coefficient of maternal pregnancy diet on child diet and the indirect effect is the product of the path coefficients for maternal pregnancy diet on maternal post‐natal diet and maternal post‐natal diet on child diet. ^†Adjusted for education (ordinal), parity (first child or not), maternal age (years) and duration of any breastfeeding (weeks). ^‡Maternal diet quality measured using the Australian Recommended Food Score (ARFS) and child diet quality measured using the Australian Child and Adolescent Recommended Food Score (ACARFS), with fruits and vegetables sub‐scales from each tool.

Discussion

The purpose of this study was to establish whether maternal diet during pregnancy and after birth is a predictor of toddler diet quality or variety of fruit and vegetable intakes. This study found that much of the effects of maternal pregnancy diet on child diet are mediated indirectly through maternal post‐natal diet, particularly for fruit and vegetable variety. This supports the role of the mother's current diet quality in influencing the diet quality of their child at age 2 and 3 years. Although it has been suggested that exposure in utero to a wide variety of foods may develop a broader taste acceptance in offspring and therefore support greater acceptance of a variety of foods in early childhood (Mennella et al. 2001), the current study did not support a direct effect between maternal diet quality during pregnancy and diet quality of the offspring. Although initial correlations showed strong associations between pregnancy and child diet, these relationships were found to be mediated by the indirect effect of maternal post‐natal diet. For young women in the WATCH study, there was no significant difference in maternal diet quality between pregnancy and post‐partum, and this was previously shown in a cross‐sectional analysis of young Australian women (Hure et al. 2009).

The results of the current study also highlight the poor diet quality in this group of women during and after pregnancy. The mean ARFS scores for total diet quality and fruit and vegetable diet quality indicate that the consumption patterns of this group of women do not meet the national dietary guidelines for health maintenance and chronic disease prevention (National Health and Medical Research Council 2003). Indeed, many young Australian women are not meeting national recommendations for healthy, varied, nutrient‐dense diets as laid out in the dietary guidelines and nutrient reference values, regardless of pregnancy status (Hure et al. 2009). This is of particular concern during pregnancy, given that maternal nutrition during pregnancy is linked to long‐term metabolic and endocrine outcomes for the offspring, and poor maternal diet quality is associated with an increased risk of neural tube defects and lower birth weight (Kuzawa 2005; Hure et al. 2009; Wadhwa et al. 2009). Although it is suggested that the diet quality scores are arbitrary in terms of defining scores as ‘high’ or ‘low’, the ARFS is useful for ranking diet quality and variety at the population level (Hure et al. 2009). Hure et al. found that in a cohort of young women, those in the highest ARFS quintile (mean ARFS 42.6) still did not meet national recommendations for nutrient intake, regardless of pregnancy status (Hure et al. 2009). A similar finding was reported for a cohort of mid‐aged women, with participants in the highest ARFS quintile (mean ARFS: 45.9) still failing to meet recommended intakes for many nutrients (Collins et al. 2008). Therefore, it is suggested that a mean ARFS diet quality score greater than 40 needs to be targeted in order to meet national recommendations for nutrient intake. In a cohort of pre‐schoolers, the median score for the Australian Recommended Food Score for Pre‐Schoolers (ARFS‐P) was 36 (maximum score 55, minimum score 12). The ARFS‐P was closely modelled on the ARFS and it is suggested that a score of ≥42 indicated good diet quality (Collins et al. 2014).

Maternal post‐natal diet was found to mediate a strong effect on child fruit and vegetable acceptance. These findings mirror an increasing body of research that suggests familial environment, particularly maternal diet, is associated with child dietary intake (van der Horst et al. 2007; Pearson et al. 2009; Fisk et al. 2011). Given these findings, it is therefore not surprising that child ACARFS (total and sub‐scores) were suboptimal, reflecting those of mothers. This has significant implications for both further research and health promotion. The literature shows children are not consuming enough fruit and vegetables in their diets and getting children to eat vegetables is a particular challenge for many parents (Carruth et al. 2004; Booth et al. 2006; Maier et al. 2007; Department of Health and Ageing, Department of Agriculture, Fisheries and Forestry 2008). Multiple studies have shown that healthy habits in childhood track into later life, so the development of healthy eating habits early in life is vital (Berenson et al. 1992; Kelder et al. 1994; Mikkilä et al. 2005; Northstone & Emmett 2008). Many factors beyond the scope of the current study influence child acceptance of foods. In addition to innate taste preferences and research by Mennella supporting exposure to taste in utero and via breast milk, there is evidence for the role of genetics in taste preference (Fildes et al. 2014). In a study of twins (mean age 3.5 ± 0.27 years), Fildes et al. postulated that, while both genetics and environment had significant impact on food preferences, genetic effects were greater for vegetables and fruit, and environmental effects were greater for energy‐dense snacks and starchy foods (Fildes et al. 2014). This result poses interesting implications for nutrition interventions. Wardle & Cooke (2010) argue that an individual's experience with food can override this innate or genetic predisposition. Animal and human studies suggest that food preferences can be socially transmitted through food behaviours modelled in the home or by peers, and that feeding behaviours such as rewards and positive attention for healthy eating tend to increase acceptance (Wardle & Cooke 2010). Repeated exposure through ingestion of unfamiliar foods also leads to increased acceptance of that food (Cooke 2007). Conversely, coercive feeding practices or parents showing negative emotions at meal times can impair child enjoyment of eating (Wardle & Cooke 2010). Repeated exposure including social learning through parental, familial and peer food modelling and positive reinforcement of healthy eating are therefore all valuable strategies that assist with development of healthy food preferences in children.

It is an important finding of the current study that children appear to be influenced by their mothers intake at such an early age, as it supports the literature on social learning that mothers can act as role models for their children by consuming a wide variety of nutritious foods, including fruit and vegetables, in order to increase the likelihood that their children will accept the same (Wardle & Cooke 2010). This highlights opportunities for randomised controlled trials directed at improving mothers' diet quality in order to positively impact child diet quality.

This study had a number of limitations. Given this was a secondary data analysis, and diet quality was not the primary outcome, the study may have been underpowered to detect some associations between mother and child diet quality. As data collection took place over several years, not all participants attended every data collection session, and therefore, it is a limitation of this study that only 52 women of the recruited 180 women could be included in this analysis. We attribute the low retention rate for the FFQ measure to mothers not completing these questionnaires as they were allowed to take the FFQs home to complete to reduce the time burden associated with data collection. FFQ questionnaires have known limitations including recall bias, and there is the potential that the women wanted to report healthy intakes for their children, which would mean that the diet quality was overestimated. However, this seems unlikely on the basis of the low observed values for the ARFS and ACARFS. Although the data showed maternal diet to be a strong predictor, this study did not account for a number of factors that might influence acceptance of foods by the child. These factors include mothers' nutrition knowledge, maternal attitudes towards weight and food and behaviours involved in child feeding practices, such as repeated exposure to foods. Therefore, further research in this area would inform the design of targeted interventions. The aim of this study was the association between mother and child diet quality, and as such, father diet quality was not assessed. However, the importance of father diet quality of the child should not be overlooked and future studies should examine father–child diet quality associations. This study is strengthened by the longitudinal dietary intake data for women and their children collected prospectively throughout pregnancy and post‐partum for both mother and child. In addition, the AES and the ACAES are validated tools to assess dietary intake in these participant groups (Hodge et al. 2002; Watson et al. 2009).

The current study demonstrated that maternal diet quality post‐partum, not during pregnancy, is predictive of the diet quality of their child at age 2–3 years. This suggests that the quality of the mother's current diet influences the respective quality of the child's diet. Further research is warranted to investigate whether dietary intervention in the mother will have beneficial effects on the diet quality for both mother and child.

Conclusion

Testing the current hypothesis in a larger population sample is important, given the small sample size in the current study. However despite this, statistically significant relationships were found that suggest that future research testing interventions to optimize toddler dietary quality should simultaneously target the dietary intake of their mothers. Focus should be on encouraging healthy dietary habits for mothers through the inclusion of a wide variety of nutritious food. By offering these foods to offspring and modelling healthy habits, both mothers and their children have the potential to benefit.

Sources of funding

The WATCH study received funding from the University of Newcastle, the Newcastle Permanent Charitable Foundation and the John Hunter Hospital Charitable Trust. The study sponsors were not involved in the research design, implementation or publication.

Conflicts of interest

The authors declare that they have no conflicts of interest.

Contributions

AA wrote the majority of this manuscript and contributed to data analysis and interpretation. AH and CC designed the WATCH study, the methods for this secondary data analysis and contributed to the interpretation of results and writing of this manuscript. AH conducted data collection for the WATCH study. CO designed the statistical analysis for this data and CO, AH and AA conducted the statistical analyses. MJ contributed to the writing and revision of the manuscript. All authors contributed to the final revision of the manuscript.

Supporting information

Table S1. Scoring method for items in the Australian Child and Adolescent Recommended Food Score.

Table S2. Pearson's rank correlations between mothers' ARFS (total and sub‐scales) at various time points during pregnancy and their children's ACARFS (total and sub‐scales) at ages two and three.

Ashman, A. M. , Collins, C. E. , Hure, A. J. , Jensen, M. , and Oldmeadow, C. (2016) Maternal diet during early childhood, but not pregnancy, predicts diet quality and fruit and vegetable acceptance in offspring. Maternal & Child Nutrition, 12: 579–590. doi: 10.1111/mcn.12151.