The first year of life is a critical window for the development of future dietary behaviours and long-term health outcomes(Reference Gilley, Krebs, Marriott, Birt, Stallings and Yates1). Infant feeding practices play an important role in child health(2); breast-feeding, for instance, is the foundation of infant nutrition. Longer duration of breast-feeding has many health benefits for the child, such as a lower risk of type 1 diabetes, asthma(Reference Stoody, Spahn and Casavale3), sudden infant death syndrome, obesity and neurological conditions(Reference Dieterich, Felice and O’Sullivan4), and for the mother, such as a lower risk of breast and ovarian cancer(Reference Chowdhury, Sinha and Sankar5). Conversely, formula feeding has been linked to an increased risk of infections, asthma and atopic allergies, metabolic disease and obesity(Reference McNiel, Labbok and Abrahams6). Some evidence also supports an association between early introduction of solids before age 4 months and adverse health consequences, including higher obesity risk(Reference Stoody, Spahn and Casavale3,Reference Padhani, Das and Siddiqui7) . The WHO and Australian infant feeding guidelines recommend exclusive breast-feeding for the first 6 months of an infant’s life(8,9) . Solid foods are recommended to be introduced around 6 months of age to meet the infant’s additional energy and nutrient needs(10). Despite these recommendations, suboptimal infant feeding practices are highly prevalent globally(Reference Neves, Vaz and Maia11) and in Australia(12). A 2010 Australian national survey of infant feeding practices reported that most women (96 %) initiated breast-feeding, but rates of exclusive breast-feeding declined sharply to 2·1 % when infants were 6 months of age(12). Similarly, 35·3 % of infants were introduced to solids before 4 months of age(12). In addition, 34 % of infants were introduced to formula before 1 month of age, increasing to 69 % at 6 months of age(12). Therefore, it is vital to improve infant feeding practices to ensure long-term child health outcomes.
Prior research has highlighted maternal influences on child health outcomes and dietary behaviours(Reference Marshall, Abrams and Barbour13,Reference Ventura, Phelan and Silva Garcia14) . Previous studies in various countries, including Australia(Reference Amir and Donath15), the USA(Reference Zimmerman, Gachigi and Rodgers16) and Finland(Reference Vaarno, Niinikoski and Kaljonen17), have shown that a better pregnancy or postpartum maternal diet quality was associated with better infant feeding practices, including longer breast-feeding duration and introducing solids after age 6 months. However, no studies have investigated the associations between maternal diet quality in pregnancy and infant feeding practices in Australian women.
Evidence has also revealed that maternal diet quality or infant feeding practices are influenced by maternal factors such as age, educational attainment and country of birth(Reference Scott, Ahwong and Devenish18,Reference Doyle, Borrmann and Grosser19) . Therefore, the association between maternal diet quality and infant feeding practices could differ by maternal factors, but no studies have examined this to date. Thus, this study aimed to (1) investigate the associations between maternal diet quality in pregnancy and infant feeding practices (i.e. breast-feeding duration, timing of solids introduction and feeding mode) and (2) explore potential moderators. Knowledge of how maternal diet quality in pregnancy influences infant feeding practices and the potential moderators underlying this association will inform the design of dietary interventions in pregnancy and identify women at risk of poor diet quality and suboptimal infant feeding practices.
Methods
Study design and participants
The present study used longitudinal data from the Healthy Beginnings Trial, a randomised controlled trial to prevent early childhood obesity through a home-based intervention involving first-time mothers and their children(Reference Wen, Baur and Rissel20–Reference Wen, Baur and Simpson22). Details of the trial and results have been published elsewhere(Reference Wen, Baur and Rissel20–Reference Wen, Baur and Simpson22). The trial comprised a 2-year intervention stage conducted between 2007 and 2010 and a 3-year follow-up stage between 2011 and 2014. The intervention consisted of eight home visits by a trained early childhood nurse – one antenatal visit at 30–36 weeks gestation and seven postnatal visits starting from birth to 24 months(Reference Wen, Baur and Simpson21,Reference Wen, Baur and Simpson23) . During these visits, women were informed about healthy infant feeding, child and family physical activity and nutrition and social support. Nurses also provided telephone support. The control group received the usual care.
Women attending one of two antenatal clinics in metropolitan South Western Sydney were invited to participate in the trial(Reference Wen, Baur and Rissel20–Reference Wen, Baur and Simpson22). Women were eligible to participate if they were aged 16 years or over, 24–34 weeks pregnant, were local residents and could give informed consent. Additionally, women or their guardians were required to communicate in English. Following childbirth, women were excluded from the trial if their infant was diagnosed with a medical condition influencing physical activity, eating behaviours, weight or height/length. This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects were approved by the Sydney South West Area Health Service ethics review committee (RPAH Zone, Protocol No X10-0312 and HREC/10/RPAH/546) and the Deakin University Human Ethics Advisory Group—Health (HEAG-H 194_2021). Written informed consent was obtained from all subjects. This study is reported per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE-nut) guidelines (online Supplementary Table 1)(Reference Lacaht24).
Dietary intake
Dietary data were collected at baseline (24–34 weeks gestation) using a 15-item semi-quantitative FFQ(Reference Wen, Flood and Simpson25). The FFQ comprised ten dietary questions from the New South Wales Population Health Survey Australia(Reference Rutishauser, Webb and Abraham26,27) along with five additional dietary questions on bread, milk and beverage intake. The FFQ captured the frequency of intake (i.e. times/serves per day, week or month or rarely/never consumed) of vegetables, fruit, bread, breakfast cereal, cooked grains/cereals, milk, processed meat, takeaway meals/snacks, potato products, sugar-sweetened beverages, fruit juice and water(27). Participants also had the option of selecting ‘don’t know’ or ‘refused’. Online Supplementary Table 2 details the FFQ’s dietary questions, response options and adaptations.
Diet quality
The 2013 Dietary Guideline Index (DGI) was used to assess maternal diet quality at baseline (online Supplementary Table 3)(Reference Thorpe, Milte and Crawford28). The DGI is a food-based score that assesses adherence to age- and sex-specific recommendations of the 2013 Australian Dietary Guidelines(29). The DGI comprises thirteen components: seven encouraged (i.e. food variety, vegetables, fruit, grains/cereals (i.e. total cereal intake and type of bread consumed), lean meats and alternatives (i.e. total lean meats and alternatives and proportion of lean meats and alternatives to total meats and alternatives), dairy and alternatives, fluid intake (i.e. total fluid intake and proportion of water to total fluids)) and six discouraged (i.e. discretionary foods, saturated fat (i.e. trim fat from meat and type of milk consumed), unsaturated spreads and oils, salt intake (i.e. salt added during cooking and salt added during meal), added sugar and alcohol intake)(Reference Thorpe, Milte and Crawford28). Each component is scored 10, giving a total possible score of 130. For the present study, a modified DGI was used due to the availability of dietary data. The primary modifications were that diet variety, type of bread consumed, lean meat and alternatives intake, whether fat was trimmed from meat, salt intake and alcohol intake were excluded. Consequently, the modified DGI comprised eight components: vegetables, fruit, dairy and alternatives, fluid intake, discretionary foods, added sugar, grains/cereals (i.e. total cereal intake) and saturated fat (i.e. type of milk consumed). The sub-components grains/cereals and saturated fat were scored out of five. Scoring was categorical for saturated fat intake. Discretionary foods were reverse-scored. ‘Don’t know’ or ‘refused’ responses were assigned as missing and excluded from the analysis. Scoring was proportional for the remaining sub-components and components, with a maximum score suggesting that national recommendations had been met. The total score for the modified DGI ranged from 0 to 70, with a higher score suggesting better diet quality. For outcomes that showed a significant association with the continuous DGI score, further analyses were conducted to assess linear dose–response relationships by using DGI tertiles (i.e. tertile 1 = low, tertile 2 = moderate and tertile 3 = high), with DGI tertile 3 indicating better diet quality compared with DGI tertile 1.
Infant feeding practices
Information on infant feeding practices was collected using a questionnaire at 6 months postpartum via telephone interview and at 12 and 24 months postpartum via face-to-face interview in participants’ homes(Reference Wen, Baur and Rissel20,Reference Wen, Baur and Simpson22) . Detailed information on the questionnaire has been previously published(Reference Hector, Webb and Lymer30). At 6 and 12 months postpartum, women were asked whether their child had ever been breastfed, the total time their infant had been breastfed in weeks and months and at what age (in weeks and months) their child had been given infant formula or solid food regularly. At 24 months postpartum, women reported whether they were still breast-feeding their child, had never breastfed or the age in weeks and months that their child had stopped breast-feeding. Women responded to a similar question on infant formula.
Infant feeding variables were derived by combining the infant feeding questions across three time points. They comprised of feeding mode at age 12 months, any breast-feeding duration and timing of solids introduction, which were derived by combining the infant feeding questions across three time points. Feeding mode at age 12 months was categorised as breast-feeding only or mixed feeding/formula feeding only, which was combined due to the small proportion of formula-feeding-only infants (3·8 %). Women who had never given infant formula and had breastfed for 12 months or more were grouped into the breast-feeding category. Women who had reported only ever formula feeding or who had reported both breast-feeding and formula feeding were grouped into the mixed feeding/formula feeding category. The age in weeks and months that solid food was first given was used to derive the timing of solids introduction variable. Timing of solids introduction was categorised as < 6 v. ≥ 6 months. Similarly, age in weeks and months that breast-feeding continued was used to calculate the breast-feeding duration variable. Two variables on breast-feeding duration were used: a continuous variable and a binary variable categorised as < 6 v. ≥ 6 months. Consistent with prior analyses from the Healthy Beginnings Trial(Reference Zheng, Campbell and Baur31), the timing of solids introduction and breast-feeding duration variables were analysed using 6 months as the cut-off to facilitate results interpretation, according to the recommendations of the WHO(2) and the National Health and Medical Research Council(8) regarding exclusive breast-feeding and introduction of solids recommendations at 6 months.
Covariates and potential moderators
Covariates and potential moderators were chosen based on a directed acyclic graph (online Supplementary Figure 1), which was constructed based on theory and the findings of prior literature(Reference Scott, Ahwong and Devenish18,Reference Scott, Binns and Graham32) . Socio-demographic data of women were collected at baseline, including age (in years), pre-pregnancy BMI (kg/m2), country of birth (Australia v. other), educational attainment (high school or less v. trade certificate/diploma v. university degree or higher), employment status (employed v. unemployed), marital status (married/ v. not married) and smoking status (non-smoker v. past/current smoker). Child sex was categorised as boy or girl.
Statistical analysis
Participant characteristics were summarised using descriptive analyses. Categorical and continuous variables are presented as percentages and means (SD) or medians (IQR), respectively. Analyses were conducted to assess differences in maternal diet quality scores and infant feeding variables by intervention allocation (intervention v. control group). No difference was found for all variables (DGI score: mean difference –0·01, 95 % CI –1·74, 1·72, P = 0·99; breast-feeding duration (months): mean difference –0·88, 95 % CI –2·07, 0·31, P = 0·15; breast-feeding duration (≥ 6 v. < 6 months): P = 0·18; timing of solids introduction: P = 0·18; feeding mode: P = 0·1). Consequently, the intervention and control groups were pooled for the present study, with intervention allocation included as a covariate to account for potential differences between groups. Differences between women included or excluded from the analysis were investigated using the t test for continuous variables and the χ2 test for categorical variables. Differences between DGI tertiles were explored using one-way analysis of variance for continuous variables and the χ² test for categorical variables.
Multivariable linear or logistic regressions were performed to investigate associations between maternal diet quality in pregnancy (exposure) and continuous (breast-feeding duration in months) or categorical (breast-feeding duration: < 6 v. ≥ 6 months; timing of solids introduction: before or at 6 months; feeding mode: breast-feeding or mixed feeding/formula feeding) infant feeding outcomes, respectively. Models were run first with maternal diet quality score as a continuous variable and then with diet quality categorised into tertiles. Crude models were adjusted for intervention allocation, and adjusted models included all covariates (i.e. maternal age, country of birth, educational attainment, employment status, marital status, smoking status, prepregnancy BMI and child sex) as well as intervention allocation. Variance inflation factors were checked for potential multicollinearity, and no significant correlations were found. P-trends for the association between categorical DGI tertiles and infant feeding variables were calculated using linear regression models with tertiles of diet quality analysed as a continuous variable.
Variables (i.e. maternal age, country of birth, educational attainment, employment status, marital status, smoking status, child sex, prepregnancy BMI and intervention allocation) that moderate the association between maternal diet quality in pregnancy and infant feeding practices were explored, informed by previous research(Reference Chimoriya, Scott and John33). Potential interactions between maternal diet quality and the covariates were investigated by including interaction terms in the adjusted model. Stratified analyses were conducted when significant interactions (P < 0·05) were observed. For models with breast-feeding duration (in months and < 6 v. ≥ 6 months) as the outcome, a significant interaction between maternal diet quality and country of birth was observed. Consequently, stratified analyses were performed to further test whether the associations between maternal diet quality and breast-feeding duration differed by country of birth. The likelihood ratio test revealed that the addition of an interaction term between maternal DGI and maternal country of birth significantly improved the model fit.
Sensitivity analyses
Multiple imputations by chained equations with ten datasets were performed, imputing missing covariates (online Supplementary Figure 1), to determine their effect on the analysis. Logistic regressions were conducted to predict the missing covariates. Stata’s ‘mi estimate’ command was used to pool estimates from the ten datasets. All analyses were performed using Stata version 18·0 (Stata Corp). Results were considered significant at P < 0·05.
Results
Of the 667 women recruited into the Healthy Beginnings Trial, 164 (24·6 %) were excluded for missing data on infant feeding practices, and a further 34 (5·1 %) for missing data on covariates (online Supplementary Figure 1). Thus, a total of 469 women were included in the final analysis. Compared with the included sample, more women from the excluded sample were younger, unemployed, unmarried did not go to university or had a lower DGI score in pregnancy (online Supplementary Table 4).
Participant characteristics
Table 1 summarises the characteristics of the 469 women by DGI score and DGI tertile. Overall, women had a mean DGI score of 47·5 (sd 9·5) and a mean prepregnancy BMI of 25·0 (sd 5·6). Most women were born in Australia (n 305, 65·0 %), employed (n 280, 59·7 %), married (n 423, 90·2 %), a non-smoker (n 309, 65·9 %), had attained a trade certificate or diploma (n 177, 37·7 %) or had given birth to an equal number of boys and girls. Women, on average, breastfed their infants for 6·1 (IQR 0·6–9·8) months. Most women breastfed their infants for < 6 months, introduced solids at or after 6 months or used formula or mixed infant feeding modes. There were significant differences across all three DGI tertiles for maternal age, country of birth, educational attainment, employment status, marital status and smoking status (P < 0·05). Across all three tertiles, women were predominantly born in Australia, but the proportion of Australian-born women decreased in DGI tertiles 2 and 3. Most women in DGI tertile 1 had finished high school or less, while women in DGI tertile 3 had completed university. From DGI tertiles 1 to 3, there were increasing trends in the proportion of women who were employed or married. Most women were non-smokers across all three tertiles; however, DGI tertile 1 had the smallest proportion of non-smokers compared with the other two tertiles. There was a non-significant difference (P = 0·4) in child sex across the DGI tertiles. There were significant differences across all DGI tertiles for breast-feeding duration (in months and > 6 v. ≥ 6 months) and feeding mode (P < 0·05). Compared with women in DGI tertiles 2 and 3, women in DGI tertile 1 breastfed their infants for a shorter duration and predominantly used formula or mixed feeding practices. Between tertile 2 and tertile 3, breast-feeding duration decreased slightly. There was a non-significant difference (P = 0·07) in the timing of solids introduction across tertiles.
Table 1. Demographic characteristics of women from the healthy beginnings trial (n 469) (Numbers and percentages; mean values and standard deviations; median values and interquartile ranges)

DGI score: 0 to 70, higher score indicates better diet quality. DGI tertiles scores: DGI tertile 1, 13·8–44·6; DGI tertile 2, 44·7–52·9; DGI tertile 3, 52·9–66·6. DGI, 2013 Dietary Guideline Index; TAFE, technical and further education.
Maternal diet quality in pregnancy and associations with infant feeding practices
Table 2 presents the results of the linear and logistic regressions for associations between maternal DGI score and infant feeding practices. For DGI score, the crude analysis revealed that DGI score was positively associated with longer breast-feeding duration (months: β 0·14, 95 % CI 0·08, 0·20; < 6 v. ≥ 6 months: OR 1·06, 95 % CI 1·03, 1·08). For every one-unit increase in DGI score, women were more likely to continue breast-feeding for a further 0·14 months, which equates to approximately 4 days, or women had 6 % higher odds of breast-feeding for ≥ 6 months. Significant associations remained in the adjusted analyses; however, the effect sizes were slightly smaller. Significant positive associations were also observed for DGI score and feeding mode at age 12 months (OR 1·04, 95 % CI 1·01, 1·07). For every one-unit increase in DGI score, women had 4 % higher odds of breast-feeding their child at age 12 months than formula feeding or mixed feeding. No evidence of association was found for the DGI score and timing of solids introduction.
Table 2. Linear and logistic regression analyses investigating associations between maternal diet quality in pregnancy and infant feeding practices among women from the healthy beginnings trial (n 469) (Regression coefficient and 95 % CI)

Crude models adjusted for intervention allocation. Adjusted models adjusted for maternal age, country of birth, educational attainment, employment status, marital status, smoking status, child sex, prepregnancy BMI and intervention allocation. β, regression coefficient; DGI tertile scores, median (IQR): DGI tertile 1, 37·4 (33·8–41·9); DGI tertile 2, 48·6 (46·7–50·4); DGI tertile 3, 56·9 (54·5–59·2). *Indicates P-trend. Significant at P < 0·05. DGI, 2013 Dietary Guideline Index.
When DGI score was analysed as tertiles, the crude analysis found that relative to women in DGI tertile 1, women in DGI tertiles 2 or 3 had 3·54 (95 % CI 2·13, 4·96) or 3·26 (95 % CI 1·84, 4·67) higher odds of having a longer duration of breast-feeding in months, respectively (Table 2). After adjusting for covariates, significant positive associations remained but with diminished effect sizes. Trend analysis showed a positive linear dose–response relationship between DGI tertiles and breast-feeding duration (crude: P < 0·001; adjusted: P = 0·01). Similar findings were found when breast-feeding duration was analysed as a binary variable (< 6 months v. ≥ 6 months). For feeding mode at age 12 months, compared with DGI tertile 1, women in DGI tertiles 2 or 3 were 3·20 (95 % CI 1·62, 6·34) or 2·24 (95 % CI 1·11, 4·56) times more likely to breastfeed their child for 12 months than to formula feed or mixed feed. Associations and effect sizes remained similar in the adjusted analysis. Trend analysis found a positive linear dose–response relationship between DGI tertiles and feeding mode in the crude model (P = 0·04) but not in the adjusted model (P = 0·1).
Stratified analyses
Stratified analyses showed the associations between DGI score (continuous score or tertiles) and breast-feeding duration (in months or < 6 v. ≥ 6 months) were significant among women born in Australia but not in women born overseas (DGI continuous score: breast-feeding in months: all participants—β 0·09, 95 % CI 0·03, 0·15; Australia—β 0·13, 95 % CI 0·06, 0·19; Other—β –0·03, 95 % CI –0·17, 0·11; < 6 v. ≥ 6 months: all participants—OR 1·04, 95 % CI 1·02, 1·07; Australia—OR 1·06, 95 % CI 1·03, 1·10; Other—OR 1·0, 95 % CI 0·96, 1·04; see Figure 1 for results for DGI tertiles). The likelihood ratio test revealed significant differences in stratum-specific effect sizes between women born in Australia and overseas (P < 0·05). This suggests that the significant positive associations between DGI score and longer breast-feeding duration were moderated or differed by maternal country of birth.

Figure 1. Forest plot of the associations between DGI tertile and breast-feeding duration in months (a) and ≥ 6 months v. < 6 months (b) stratified by country of birth with adjustment for covariates.
Sensitivity analyses
Imputing the missing covariates (n 1 to n 29 missing, 6·8 % overall missingness) increased the effect sizes slightly for the association between DGI score and most infant feeding variables, and associations were in the same direction (online Supplementary Table 5).
Discussion
This is the first study to investigate associations between maternal diet quality in pregnancy and infant feeding practices in Australian women. We found that a better maternal diet quality in pregnancy was associated with a longer breast-feeding duration. However, this association was observed in Australian-born women and not overseas-born women. We also found that a better diet quality during pregnancy was associated with a higher odds of breast-feeding than formula or mixed feeding when infants were 12 months old, but it was not associated with the timing of solids introduction. When maternal diet quality was categorised into tertiles, we found a positive linear dose–response relationship between maternal diet quality and breast-feeding duration.
We found that a better maternal diet quality in pregnancy was associated with longer breast-feeding duration, which is consistent with the few studies that have assessed postpartum maternal diet quality(Reference Zimmerman, Gachigi and Rodgers16,Reference Vaarno, Niinikoski and Kaljonen17) . For example, a study reported that 751 US women with a high Food Source Quality score were 24 % less likely to stop breast-feeding before 6 months than women with a moderate Food Source Quality score(Reference Zimmerman, Gachigi and Rodgers16). Notably, that study used a similarly brief FFQ to collect maternal dietary data, adding weight to the comparability of our results. Another study of 1797 Finnish families found that women with a high Index of Diet Quality score at 4 months postpartum breastfed for longer than women with a low score(Reference Vaarno, Niinikoski and Kaljonen17). Despite our findings showing a positive linear dose–response relationship between maternal diet quality in pregnancy and breast-feeding duration in the P-trend analysis, it is worth noting that the effect sizes for DGI tertile 3 were smaller than those of tertile 2. This is attributable to the shorter breast-feeding duration found in the highest DGI tertile compared with the mid DGI tertile. Further research is required to understand the dose–response relationship between maternal diet quality in pregnancy and breast-feeding duration. We also found that a better maternal diet quality in pregnancy was associated with higher odds of breast-feeding than formula or mixed feeding when infants were 12 months old. This finding aligns with a study of 360 women from the USA that found that women who consumed ≥ 3 servings of fruits and vegetables daily were more likely to breastfeed for 1 year compared with women who consumed fewer servings of fruits and vegetables(Reference Olson34). Likewise, another study of 149 US women reported that women with lower fat and higher fruit intakes were more likely to breastfeed to 6 months postpartum(Reference George, Hanss-Nuss and Milani35).
Several hypotheses may explain the beneficial role of a better maternal diet quality during pregnancy in promoting breast-feeding. It is possible that women who follow national dietary recommendations are more likely to follow recommendations for breast-feeding duration. Another potential reason is that women with a better diet quality during pregnancy may have better mental health or less stress(Reference Fowles, Bryant and Kim36), which has been shown to be linked with a longer breast-feeding duration(Reference Dias and Figueiredo37). Alternatively, women with a healthier diet during pregnancy may be less likely to consume processed foods exposed to endocrine-disrupting chemicals, such as per- and polyfluoroalkyl substances and bisphenols, through plastic food packaging(Reference Veiga-Lopez, Pu and Gingrich38). These chemicals have been found in breast milk, and prior research has shown that exposure to these chemicals could reduce breast-feeding duration through endocrine system disruption (e.g. impaired mammary gland development, lactogenesis and endocrine signalling)(Reference Criswell, Crawford and Bucinca39).
To the authors’ knowledge, this is the first study to report that the association between better maternal diet quality and longer breast-feeding duration is moderated by maternal country of birth, with significant associations found in women born in Australia but not overseas. Our finding reflects the ethnic and cultural differences in maternal diet quality or breast-feeding duration that have been extensively reported(Reference Deierlein, Ghassabian and Kahn40,Reference Dennis, Gagnon and Van Hulst41) . Our finding is consistent with an Australian(Reference Scott, Ahwong and Devenish18) study and a Portuguese(Reference Kana, Rodrigues and Fonseca42) study that identified that women born overseas were more likely to breastfeed for longer compared with women born in Australia and Portugal, respectively. The longitudinal Portuguese study also reported that migrant women breastfed for longer regardless of the time they had resided in Portugal(Reference Kana, Rodrigues and Fonseca42), which emphasises the importance of ethnic and cultural differences.
Other maternal socio-demographic factors could support our finding concerning maternal diet quality in pregnancy and breast-feeding duration. Research has shown that women who are older and tertiary educated have a better diet quality than women who are younger and with no tertiary education(Reference Deierlein, Ghassabian and Kahn40,Reference Bodnar and Siega-Riz43) . In our study, women with a high diet quality were more likely to be older and tertiary educated. Previous research has shown that women who are older are more likely to be tertiary educated, have greater levels of nutrition knowledge(Reference McLeod, Campbell and Hesketh44) and could, therefore, have greater awareness or knowledge of optimal infant feeding practices.
Maternal diet quality in pregnancy was not associated with the timing of solids introduction. Most prior research has used one variable to investigate the timing of solids introduction(Reference Padhani, Das and Siddiqui7). Few studies have explored maternal diet quality and timing of solids introduction. However, a Finnish study involving 1797 women, which assessed the timing of solids introduction as a continuous variable, found that women with a better postpartum diet quality introduced solids almost 1 month later than women with a poor diet quality(Reference Vaarno, Niinikoski and Kaljonen17), which contrasts with the present study’s finding. The lack of a significant association in our study could be because we used a categorical (v. a continuous) variable to assess the timing of solids introduction or because we assessed maternal diet quality at a different time point (i.e. pregnancy v. postpartum).
Our study has several strengths. Our study is the first to explore the relationship between maternal diet quality in pregnancy and infant feeding practices. We also found the relationship between maternal diet quality in pregnancy and breast-feeding duration was moderated by maternal country of birth. In contrast to previous research involving the over-representation of highly educated women, we used data from the Healthy Beginnings Trial, which comprises women from diverse educational and cultural backgrounds, supporting the generalisability of our study’s findings to the wider Australian population. We also have some limitations to acknowledge. The dietary assessment tool used to assess maternal diet was relatively brief. Consequently, the full DGI score could not be calculated. Modifications to diet quality indices are common because of the different dietary assessment tools used(Reference Hlaing-Hlaing, Pezdirc and Tavener45). However, we captured most key dietary intakes representative of encouraged and discouraged foods and relevant to health outcomes such fruits and vegetables and discretionary foods. We were also able to detect significant associations between maternal DGI score and breast-feeding, supporting the utility of using a short FFQ to derive a DGI score, similar to other studies(Reference Cleghorn, Harrison and Ransley46). Future research could replicate this analysis by using repeated measures of 24-hour recall data. In addition, only some of the dietary questions in the dietary assessment tool had been previously validated(Reference Rutishauser, Webb and Abraham26), which could have implications for its reliability. The DGI was modified based on the availability of dietary data to enable the calculation of the diet quality scores. These omissions could make comparison of the results with other diet quality research challenging(Reference Wingrove, Lawrence and McNaughton47). Parental reporting was used to determine the infant feeding variables, so parental reporting bias cannot be dismissed. However, infant feeding data were collected at several time points to help minimise this bias. As data on infant feeding practices were reported retrospectively by parents, there is a potential for recall bias. The current study did not assess exclusive breast-feeding; it would be desirable for future research to examine how maternal diet quality influences exclusive breast-feeding. Finally, there were differences in the excluded and included samples, with the excluded sample having a higher proportion of women who were younger, unmarried and unemployed. Inclusion of the excluded sample may have possibly weakened observed associations given that evidence has shown that women with these characteristics are less likely to practice infant feeding in line with recommendations(Reference Arora, Manohar and Hector48).
The findings from this study provide new evidence to support the initiation of dietary interventions in pregnancy, and such interventions should be tailored to cultural and ethnic groups. Nevertheless, additional research is needed to confirm our findings in wider population groups. Given adverse diet quality is also evident in preconception(Reference Pervin, Emmett and Northstone49), further studies could explore the relationship between maternal diet quality in preconception and infant feeding practices. Our understanding of mechanisms linking maternal diet quality and breast-feeding duration is limited; future research could explore the potential mechanisms.
In summary, in an Australian cohort, a better maternal diet quality in pregnancy was associated with longer breast-feeding duration in Australian women and not women born overseas. In addition, a better maternal diet quality was associated with a higher likelihood of breast-feeding v. formula or mixed feeding. Maternal diet quality was not associated with the timing of solids introduction. Our results add to the evidence supporting the implementation of dietary interventions to improve maternal diet quality in pregnancy, particularly in Australian-born women. Such interventions could potentially result in beneficial infant feeding outcomes. Moreover, interventions aimed at improving infant feeding practices should be initiated from pregnancy and target women born in Australia with poor diet quality.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S000711452500025X
Acknowledgements
The authors would like to acknowledge the participants of the Healthy Beginnings Trial.
This work was supported by a Deakin University Postgraduate Award scholarship (MJS-D). KML is supported by a National Health and Medical Research Council Emerging Leadership Fellowship (APP1173803). MZ is supported by the Australian National Health Medical Research Council (APP1124283).
M. J. S-D. and M. Z. conceptualised the idea and designed the research. L. M. W. collected data. M. J. S-D. performed the statistical analysis, interpreted the results and drafted the manuscript. M. Z. assisted with manuscript preparation. M. Z., K. M. L., E. A. S-G. and L. M. W. assisted with results interpretation. All authors critically reviewed the manuscript and approved the final version for submission.
The authors declare no conflicts of interest.